Secondary Predication and Adverbial Modification: The Typology of Depictives

Secondary Predication and Adverbial Modification This page intentionally left blank Secondary Predication and Adver...

Author: Nikolaus P. Himmelmann | Eva F. Schultze-Berndt

97 downloads 1468 Views 3MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Secondary Predication and Adverbial Modification

This page intentionally left blank

Secondary Predication and Adverbial Modification The Typology of Depictives

Edited by NIKOLAUS P. HIMMELMANN AND EVA SCHULTZE-BERNDT

AC

AC

Great Clarendon Street, Oxford ox2 6dp Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide in Oxford New York Auckland Cape Town Dar es Salaam Hong Kong Karachi Kuala Lumpur Madrid Melbourne Mexico City Nairobi New Delhi Shanghai Taipei Toronto With offices in Argentina Austria Brazil Chile Czech Republic France Greece Guatemala Hungary Italy Japan Poland Portugal Singapore South Korea Switzerland Thailand Turkey Ukraine Vietnam Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries Published in the United States by Oxford University Press Inc., New York #

editorial matter and organization Nikolaus P. Himmelmann and Eva Schultze-Berndt 2005 # the chapters their authors 2005 The moral rights of the authors have been asserted Database right Oxford University Press (maker) First published 2005 First published in paperback 2006 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this book in any other binding or cover and you must impose the same condition on any acquirer British Library Cataloguing in Publication Data Data available Library of Congress Cataloging in Publication Data Secondary predication and adverbial modification : the typology of depictives / edited by Nikolaus P. Himmelmann and Eva Schultze-Berndt. p. cm. Summary: ‘‘This book approaches depictive secondary prediction from a crosslinguistic perspective. It maps out all the phenomena and brings together critical surveys and new contributions on their morphosyntactic and semantic properties’’---Provided by publisher. Includes bibliographical references and index. 1. Grammar, Comparative and general---Syntax. 2. Grammar, Comparative and general--Verb phrase. 3. Grammar, Comparative and general---Adverbials. 4. Semantics. 5. Typology (Linguistics) I. Himmelmann, Nikolaus. II. Schultze-Berndt, Eva. P295.S413 2005 415---dc22 2004024152 Typeset by Newgen Imaging Systems (P) Ltd., Chennai, India Printed in Great Britain on acid-free paper by Biddles Ltd, King’s Lynn ISBN 0--19--927226--3 978--0--19--927226--6 ISBN 0--19--920434--9 (Pbk.) 978--0--19--920434--2 (Pbk.) 10 9 8 7 6 5 4 3 2 1

Contents Preface and chapter summaries Acknowledgements The contributors Abbreviations and glossing conventions 1 Issues in the syntax and semantics of participant-oriented adjuncts: an introduction Nikolaus P. Himmelmann and Eva Schultze-Berndt 2 Depictives in English and Warlpiri Jane Simpson 3 Adverbials and depictives as restrictors Thomas Mu¨ller-Bardey 4 Depictive agreement and the development of a depictive marker in Swiss German dialects Claudia Bucheli Berger 5 Quantifying depictive secondary predicates in Australian languages William B. McGregor 6 Depictives in Kartvelian Winfried Boeder 7 On depictive secondary predicates in Laz Silvia Kutscher and N. Sevim Genc¸ 8 Participant agreement in Panoan Pilar M. Valenzuela 9 Secondary predicates and adverbials in Nilotic and Omotic: a typological comparison Azeb Amha and Gerrit J. Dimmendaal 10 Asyndetic subordination and deverbal depictive expressions in Shona Tom Gu¨ldemann 11 Forms of secondary predication in serializing languages: on depictives in Ewe Felix K. Ameka

vii xv xvi xx

1 69 107

141

173 201 237 259

299

323

355

vi

Contents

12 Depictive and other secondary predication in Lao Nicholas J. Enfield

379

13 A semantic map for depictive adjectivals Johan van der Auwera and Andrej Malchukov

393

References Index of languages Index of terms

423 441 443

Preface and Chapter Summaries This book is concerned with depictive secondary predicates such as fresh in George bought the carrots fresh and related constructions in crosslinguistic perspective. These constructions have special semantic and syntactic properties and therefore have played a major role in modern theories of the syntax and semantics of natural languages. This book provides for new perspectives on these constructions in two ways. First, it considerably expands the database for them by showing that, in addition to the well-studied adjectival depictives such as fresh, a wide range of other expression types warrant an analysis as depictives, including, for example, some types of subordinate clause. This becomes apparent when considering data from a much wider range of languages than those which usually feature in the literature on this topic, especially since many of these languages have speciﬁc morphosyntactic means of marking depictives. Second, the book brings together different research strands which have tended largely to ignore each other because they diverge considerably both in terms of topic area (syntax vs. semantics) and in terms of their basic theoretical persuasions (cognitive linguistics, typology, lexical-functional grammar, neo-Davidsonian semantics, etc.). The contributions thus point to the need for a new, considerably richer and more complex systematics for the grammatical domain of participant-oriented adjunct constructions. Participant-oriented adjunct constructions differ from other types of adjunct construction (most importantly event-oriented adjunct constructions) in that they primarily convey additional information regarding one of the participants in the eventuality expressed by the main predicate (fresh in the example above tells us primarily something about the carrots and not about George or the manner of buying carrots). Although secondary predicates are perhaps prototypical examples of participant-oriented adjuncts, it has been noted that other types of adjuncts may also show participant orientation, and therefore exhibit a strong semantic afﬁnity to depictives. These include, for example, certain subject-oriented adverbials such as enthusiastically in John opened the door enthusiastically. Moreover, across languages we frequently ﬁnd that the same construction may render both participant- and event-oriented content, even beyond those adverbials that are typically regarded as subject-oriented. For this reason, the present volume also explores commonalities and differences between adverbials and depictives, based on the hypothesis that depictive and adverbial constructions cover a common semantic space. With regard to the more ambitious goal of presenting a comprehensive typology of the participant-oriented adjunct constructions attested in the world’s

viii

Preface and chapter summaries

languages, this book is clearly premature, in that very little is in fact known about depictive secondary predicates and related constructions in most of the world’s languages. That is, this is not a topic which is regularly studied in work on lesserknown languages, and there are only few grammars of such languages where the topic is mentioned at all (and none where it is extensively dealt with). This deplorable state of affairs is not going to change as long as adjuncts in general, and participant-oriented adjuncts in particular, do not make their way into the regular work package of ﬁeldworkers and descriptive linguists (along with standard topics such as grammatical relations and relative clauses). Thus, this book also has the purpose of contributing to this expansion of the standard set of descriptive topics by offering ideas and research strategies for tackling the language-speciﬁc intricacies of depictive secondary predicates and other participant-oriented adjuncts. Regarding the genesis of the volume, it may be useful to take note of the following. With the exception of the chapter by Valenzuela, the chapters appearing in this collection were ﬁrst prepared for the Workshop on Depictives in Crosslinguistic Perspective at the Ruhr-Universita¨t Bochum (7–9 June 2001), which was organized by the editors. All of these contributions were based on a position paper by the editors. The contributors considerably revised their papers after the workshop, taking into account feedback during the workshop as well as further comments by the editors. The introductory chapter was drafted after the workshop, when editing the individual chapters. Although it includes some aspects of the original position paper, it offers a different approach to the whole subject matter, reﬂecting a number of changes in our thinking which evolved during and in the aftermath of the workshop. It also contextualizes the other contributions with regard to a set of general themes. (Major parts of the original position paper have since—with considerable delay not attributable to the authors—appeared in Linguistic Typology, as Schultze-Berndt and Himmelmann 2004.) The current ordering of the chapters is motivated by the breadth of their approach, their theoretical orientation and the languages they deal with. Each chapter exhibits particularly close ties with the one preceding and the one following it. The chapters by Simpson and Mu¨ller-Bardey complement and expand major themes of the introductory chapter, which immediately precedes them. Simpson provides further details regarding syntactic issues in the analysis of depictives, and contrasts a language with clearly identiﬁable depictives (English) with a language where depictives appear to convey a broader range of meanings and functions (Warlpiri). Mu¨ller-Bardey offers a more detailed discussion of some of the core semantic issues, focusing on the distinction between depictives and manner adverbs, based on English and German data. The following three chapters share a focus on depictive constructions marked by agreement and other unambiguous markers of depictive status.

Preface and chapter summaries

ix

Bucheli Berger complements the discussion of English and German in Simpson and Mu¨ller-Bardey by providing data from non-standard varieties of German which exhibit interesting similarities to the Australian data discussed in the editors’ introduction, as well as in Simpson’s and McGregor’s contributions. McGregor elaborates on some of the intricacies involved in the interpretation of the Australian data, focusing on a subtype of depictives which to date has been largely ignored in the literature (quantifying depictives). While still focusing on agreement marking of depictives, Boeder’s chapter on Georgian is the ﬁrst in a series of chapters which take into view a much broader range of expression types than those usually considered in the literature on depictives, and which also discuss problems of delimitation between depictives and adverbials. All but the last of the remaining chapters share this common focus, which in turn is based on the semantic space for participant-oriented adjuncts established in the editors’ introduction. Apart from this common focus these chapters are interconnected by various areal, typological and topical links (see the short chapter summaries below). The ﬁnal chapter, by van der Auwera and Malchukov, puts the semantic space for participant-oriented adjuncts into a larger perspective by showing various links to neighbouring semantic spaces such as the ones for predicative complements, appositive attributes or pure manner adverbials. This chapter, which presents data from a worldwide sample of languages, heavily draws on the data presented in the preceding chapters.

Chapter summaries Nikolaus P. Himmelmann and Eva Schultze-Berndt, ‘Issues in the syntax and semantics of participant-oriented adjuncts: an introduction’ The editors’ introduction begins with a critical survey of previous classiﬁcations of adverbials and secondary predicates. It is argued that there exist signiﬁcant overlaps between participant-oriented manner adverbials (John angrily read the review), depictive secondary predicates, and so-called weak free adjuncts (Standing on a chair, John can touch the ceiling). Thus, from a crosslinguistic point of view, it is necessary to make use of an overarching category, participantoriented adjunct, which subsumes all of these construction types. Having established this category, the introduction then presents the range of semantic expressions which may show formal correlates of participant-orientation, and the morphosyntactic features typically associated with participant-oriented adjuncts crosslinguistically. Jane Simpson, ‘Depictives in English and Warlpiri’ Simpson compares depictive secondary predicates in English and Warlpiri. In English, there are heavy constraints on depictives in terms of their

x

Preface and chapter summaries

phrase structure position, their co-occurrence with other adjunct types and resultatives, their controllers, and their semantics, as well as the semantics of the main predicate. Simpson shows that most of these constraints are absent in Warlpiri, and proposes that these differences stem from basic syntactic differences between the two languages. These include the use of word order (English) vs. case marking (Warlpiri) to mark grammatical relations, and the possibility, in Warlpiri, of making use of temporal-aspectual clitics on depictives, which allow for an independent speciﬁcation of the relation of the state denoted by the secondary predicate to the event denoted by the main predicate. Simpson concludes that in Warlpiri, depictives are true adjuncts, while in English they are comparable to complements. Thomas Mu¨ller-Bardey, ‘Adverbials and depictives as restrictors’ Mu¨ller-Bardey applies analytical concepts originally developed for the analysis of indeﬁnite NPs to the analysis of adverbials and depictives, and as a result proposes a new subclassiﬁcation of circumstantial adverbials and depictives. Of fundamental importance is the partition of the clause, i.e. the idea that various kinds of operators including frequency adverbs such as often and always divide clauses into two parts, one functioning as the restrictor for the operator, the other as its nuclear scope. Mu¨ller-Bardey claims that various kinds of circumstantial adverbials (including manner, time, and location) do not behave uniformly with regard to restrictor formation, since only some of these adverbials allow readings in which they are part of the restrictor. ‘Pure’ manner adverbials, in particular, are always part of the nuclear scope, and in this regard they differ from depictives, which behave like those adverbials which can be part of the restrictor. Claudia Bucheli Berger, ‘Depictive agreement and the development of a depictive marker in Swiss German dialects’ Bucheli Berger discusses the different morphological markers found on (adjectival) predicative complements and depictive secondary predicates in Swiss German dialects. Three main dialect areas can be distinguished: one where secondary predicates and predicative complements are unmarked and are therefore not formally distinguished from adverbials (thus comparable to Standard German), one where both depictives and predicative complements show obligatory agreement with their controller (thus comparable to Romance and many Australian languages), and a third area where depictives (but not predicative complements) take a marker which is restricted to this construction, a crosslinguistically rare phenomenon. The emergence of this depictive marker is explained in terms of reanalysis of a former agreement marker, and the syntactic and semantic properties of depictive constructions showing the marker are discussed in some detail.

Preface and chapter summaries

xi

William McGregor, ‘Quantifying depictive secondary predicates in Australian languages’ McGregor investigates participant-oriented quantiﬁers in Australian languages, which comprise two main expression types: those that indicate the cardinality of a set of participants (e.g. ‘alone’, ‘as a pair’, ‘as a group of three’), and those that indicate iterated involvement (e.g. ‘one by one’, ‘few by few’). The formal manifestation of these expression types is subject to variation: constructions found in Australian languages may involve agreement of a numeral with its controller (‘identical marking’), use of a numeral ‘one’ in the sense of ‘together’, locative marking of the quantifying expression, and other, more speciﬁc, lexical and grammatical means. For each construction, the question is addressed of whether or not it is an instance of a dedicated depictive construction, which can be distinguished from NP-internal attributes on the one hand, and from adverbials on the other. Winfried Boeder, ‘Depictives in Kartvelian’ In his survey of depictives in Georgian and Svan, Boeder gives special attention to the overlapping semantic range of depictive and adverbial expressions. Depictives in Georgian show case agreement with their controller, and may convey a condition or state, an intention, a phase or life stage, and quantity. There exist, however, a number of non-agreeing, adverbial constructions involving the socalled adverbialis case or the instrumental, which may also receive a participantoriented interpretation, e.g. in the domains of posture, manner, comparison, function/role, and concomitance. The close semantic relationship between depictive and adverbial adjuncts in Georgian manifests itself in considerable variation in the use of the two construction types, both within and between dialects. It is also shown by the fact that adjuncts of the two types can be coordinated, and that the same interrogative marker can be used with both. Boeder provides an explanation of this overlap between adverbial and depictive constructions in terms of a regular metonymic relationship between the property predicated of a participant and characteristics of the event itself. Silvia Kutscher and N. Sevim Genc¸, ‘On depictive secondary predicates in Laz’ Kutscher and Genc¸ discuss participant-oriented adjuncts in Ardes¸en Laz, a Kartvelian language spoken in Turkey. Laz, unlike the related language Georgian discussed in the contribution by Boeder, does not exhibit case agreement of depictives with their controllers. Consequently, it is argued that Laz only has general adjunct constructions which may have participant-oriented or event-oriented interpretations, including unmarked adjectives and adverbials, and instrumental or locational adjuncts. One important difference between participant-oriented adjuncts and manner expressions is possibly found in the domain of prosodic realisation, though: the former may constitute an intonation unit of their own, while manner expressions are always integrated into the verb phrase.

xii

Preface and chapter summaries

Pilar M. Valenzuela, ‘Participant agreement in Panoan’ The Panoan languages of the Amazon region exhibit a very extensive system of participant agreement in that participant orientation is overtly marked on both clausal and phrasal adjuncts, and generally S, A, and O controllers are all formally distinguished. In her contribution, Valenzuela explores the participant orientation system in the Panoan language Shipibo-Konibo. Here, marking of participant orientation occurs on expressions of manner, quantity, location, and life stage, among others, as well as on emphatic pronouns and same-subject clauses. Valenzuela discusses the semantic distinctions conveyed by different types of agreement marking, and accounts for differences in marking patterns on adjuncts of various types by establishing a semantic hierarchy. Azeb Amha and Gerrit J. Dimmendaal, ‘Depictive secondary predicates in Nilotic and Omotic: a typological comparison’ Amha and Dimmendaal discuss various types of participant-oriented adjuncts in languages belonging to two distinct language groups spoken in adjacent areas of (north-)eastern Africa, the Nilotic branch within Nilo-Saharan and the Omotic branch of Afroasiatic, and address the correlation between the formal properties of these adjuncts and other typological properties of the languages in question. In Turkana as well as in other Nilotic languages, participant-oriented adjuncts are predominantly verbal, and formally distinct from manner adverbials, but the authors show that the distribution of expression types across the two constructions may vary, thus providing further evidence for the view that, in a crosslinguistic perspective, adverbial and depictive constructions ‘compete’. In contrast to the Nilotic languages, the Omotic languages Wolaitta and Maale systematically distinguish between same-subject converbs showing agreement with the main clause subject, and free-subject converbs which allow the independent expression of a subject. The authors argue that the former should be analysed as depictives and the latter as adverbial clauses, and that the switchreference function of these converb clauses is parasitic upon the depictive/ adverbial distinction. Tom Gu¨ldemann, ‘Asyndetic subordination and deverbal depictive expressions in Shona’ In Shona as well as other Bantu languages, participant-oriented expressions mainly take the form of converb clauses (or free-subject dependent predicates), which are prosodically marked and which may also function as part of a complex predicate, or as a relative clause. Gu¨ldemann reviews depictive-like uses of clauses of this type and discusses the arguments for and against assigning the status of a (morphosyntactically distinct) depictive construction to them. A major concern in this exploration is the investigation of the role of focus structure

Preface and chapter summaries

xiii

for the analysis of depictives. In the concluding sections of his chapter, Gu¨ldemann goes well beyond the Shona data and assembles crosslinguistic support for the view that depictives in general represent the assertive focus of a clause. Felix K. Ameka, ‘Forms of secondary predication in serializing languages: on depictives in Ewe’ Ameka shows in his contribution that meanings conveyed by depictives e.g. in European languages are often expressed by means of multiverb constructions (serial verb constructions) in languages like Ewe, a Kwa language spoken in Ghana and Togo. However, different types of participant-oriented adjuncts, including ideophones, prepositional phrases, and noun phrases, also occur in this function, contra Larson’s (1991) hypothesis that serializing languages do not have nominal secondary predicates. Ameka, moreover, presents a new analysis of a ubiquitous enclitic ¼i in Ewe which serves as a marker of secondary predicates and predicative complements, but also as an adverbializing sufﬁx. It also appears on some verb phrases in a series, which raises the intriguing question as to how serial verbs and verbal adjuncts can be distinguished, in particular in an isolating language such as Ewe. Nicholas J. Enﬁeld, ‘Depictive and other secondary predication in Lao’ In Lao, depictive expressions are predominantly verbal. Since Lao, like Ewe, is a typical isolating language, lacking inﬂectional morphosyntactic categories such as case-marking and agreement, verbs are deﬁned distributionally, and there is no distinction between ﬁnite and non-ﬁnite verb forms. Enﬁeld suggests, therefore, that the distinction between a primary and a secondary predicate can only be made in the following way: the secondary predicate is an adjunct in the sense that it can be omitted without changing the basic semantics of the primary predication. While Lao clearly has secondary predicates in this sense, it does not have a speciﬁc depictive construction involving verbs, since not only depictive but also resultative and adverbial meanings can be conveyed by unmarked verbs functioning as secondary predicates. Enﬁeld also addresses pragmatic restrictions on speciﬁc combinations of primary and secondary predicate, related to the relevance and/or conventionality of the combination. Johan van der Auwera and Andrej Malchukov, ‘A semantic map for depictive adjectivals’ Van der Auwera and Malchukov explore the formal relationships that are found crosslinguistically between adjectives in depictive function, and adjectives in other, arguably related functions (where ‘adjective’ is taken in a broad sense to include stative verbs, property nominals, participles, and relative clauses). The results are presented in the form of semantic maps, which are based on the

xiv

Preface and chapter summaries

assumption that crosslinguistic similarities in terms of formal properties can be taken to reveal proximity in semantic space. The ﬁndings show that depictives occupy an intermediate position between predicatives (i.e. adjectives used as a main predicate), predicative complements, appositive structures, and adverbials, and therefore can be said to have semantic afﬁnities with all four of these categories. London April 2004

NPH ESB

Acknowledgements The editors are very grateful to the contributors, who have been not only exceptionally cooperative in sticking to deadlines and tailoring their contributions to the design of the book but also extremely patient, staying on board despite the fact that the review and editing process took much longer than originally expected. The editors also gratefully acknowledge the generous support of the German Research Foundation (Deutsche Forschungsgemeinschaft), which included a research fellowship for Schultze-Berndt from 1999 to 2001 as well as the complete funding of the workshop in June 2001. Many thanks to Judith Ko¨hne and Jan Strunk for help with the proofs and indices. Finally, we thank John Davey at OUP for his encouragement and for seeing this work through.

The Contributors Felix K. Ameka teaches in the Department of African Languages and Cultures, Leiden University, and is a Research Associate in the Language and Cognition Group, Max Planck Institute for Psycholinguistics, Nijmegen, The Netherlands. His main interests are in language description, typology, semantics, pragmatics, anthropological linguistics, and ethnography of communication with special reference to West African languages. Currently he is working on the documentation of the language and culture of Likpe (Ghana) and (with James Essegbey) preparing a modern comprehensive grammar of Ewe. He is Editor-in-Chief of the Journal of African Languages and Linguistics. Azeb Amha is a post-doctoral researcher at the Leiden University Centre for Linguistics and teaches in the Department of African Languages and Cultures at Leiden University, The Netherlands. She works on Omotic languages, and is interested in typological/areal comparative studies. She has published The Maale Language (2001) as well as articles in scholarly journals and collections. She is currently writing a grammar of Wolaitta (with Gerrit Dimmendaal) and is documenting Zargulla, an endangered Omotic language. Winfried Boeder is Professor of Linguistics at the University of Oldenburg, Germany. His main interests are language typology and historical linguistics. He works on Kartvelian (South Caucasian) comparative and historical morphosyntax, and on the historical sociolinguistics of Georgian. Claudia Bucheli Berger is a research associate in an empirical project investigating the syntax of Swiss German dialects at the University of Zurich, Switzerland. She just ﬁnished her doctoral thesis on serial verbs in French Creoles, focussing on their implications for typology and theories of creolization. Gerrit J. Dimmendaal is Professor of African Linguistics at the University of Cologne, Germany. His main interests are language typology and Nilo-Saharan languages. His more recent publications include Surmic Languages and Cultures (1998, with Marco Last), ‘Morphology’ (2000, in African Linguistics: An Introduction, edited by Heine and Nurse), ‘Number marking and noun

The contributors

xvii

categorization in Nilo-Saharan languages’ (2001, Anthropological Linguistics 42), and ‘Places and people: ﬁeld sites and informants’ (2001, in Linguistic Fieldwork, edited by Newman and Ratliff). He is currently editing a book on the Central Sudanic family. Nicholas J. Enfield is a scientiﬁc staff member in the Language and Cognition Group, Max Planck Institute for Psycholinguistics, Nijmegen, The Netherlands. Recent publications include Linguistic Epidemiology: Semantics and Grammar of Language Contact in Mainland Southeast Asia (2003), Ethnosyntax: Explorations in Grammar and Culture (edited volume, 2002), and numerous journal articles on issues in grammar, semantics, semiotics, and interaction. Continuing research is based on regular ﬁeldwork in mainland Southeast Asia, especially Laos. N. Sevim Genc¸ graduated in linguistics at the University of Cologne, Germany, with a thesis on Laz word formation. She is currently committed to helping preserve Laz as a native language. Tom Gu¨ldemann is a Research Associate at the Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany, and teaches African linguistics at Leipzig University. He specializes in Bantu and Khoisan languages and has a major interest in typology, areal and historical linguistics. He is currently the director of two major projects: ‘Focus in South African Languages’ at the ZAS Berlin, and ‘Documentation of Western !Xo˜o’, funded by the Volkswagen Foundation. His more recent works include ‘Khoisan’ (2000, in African Linguistics: An Introduction, edited by Heine and Nurse), Quotative Constructions in African Languages: A Synchronic and Diachronic Survey (2001), and ‘Present progressive vis-a`-vis predication focus in Bantu: a verbal category between semantics and pragmatics’ (2003, Studies in Language 27). Nikolaus P. Himmelmann is Professor of Linguistics at the University of Bochum, Germany. His major research interests include typology, grammaticization theory, prosody and grammar, as well as language documentation and description. He is the author of an in-depth study of the typology and grammaticization of demonstratives and articles (Deiktikon, Artikel, Nominalphrase, 1997) and co-editor of The Austronesian Languages of Asia and Madagascar (2005) and Essentials of Language Documentation (2006). He has done ﬁeldwork in the Philippines, Sulawesi and East Timor, and published widely on a number of core issues in Austronesian grammar, including the nature of lexical and syntactic categories and voice.

xviii

The contributors

Silvia Kutscher has written a doctoral thesis on the noun phrase structure of Ardes˛en-Laz, and currently holds a Lise Meitner scholarship at the University of Cologne. Her research interests include lexical semantics, argument structure, and linguistic typology with a focus on Germanic, South-Caucasian, and Finno-Ugric languages. Together with N. Sevim Genc¸ she is involved in documenting the endangered language Laz. William B. McGregor is Professor of Linguistics at Aarhus University in Denmark. His main interests are Aboriginal languages of the Kimberley region of the far north-west of Australia, in particular Gooniyandi, Nyulnyul, Warrwa, and Unggumi. Among other things, he is also interested in language typology, numerals, the origins and evolution of language, and linguistic theory (speciﬁcally, the development of an adequate formal-functional theory). He has published grammars of four languages, and is working on descriptions of other. His most recent publications include Verb Classiﬁcation in Australian Languages (2002, Berlin: Mouton de Gruyter) and The Languages of the Kimberley, Western Australia (2004, London: Routledge Curzon). He has just ﬁnished revising Frs. Nekes and Worms’ 1953 monograph Australian languages, which appeared in February 2006, published by Mouton de Gruyter, and has edited a collection of papers on the history of Australian Aboriginal linguistics (submitted to Paciﬁc Linguistics). Andrej Malchukov is a member of the St Petersburg Typology Group (Institute of Linguistic Research, Russian Academy of Sciences). He specializes in Tungusic languages, and his habilitation thesis deals with aspects of the morphosyntax of the North Tungusic language Even. His main interests are in morphosyntactic and semantic typology. Recently ﬁnished projects include the typology of attributive relations (Dependency Reversal in Noun-attribute Constructions: Towards a Typology, 2000), the typology of transcategorial operations (Nominalization/verbalization: Constraining a typology of transcategorial operations. LINCOM, 2004), and the typology of adversative coordination. Currently he is working on the typology of case-marking systems within the PIONIER project ‘Case cross-linguistically’ at the University of Nijmegen, The Netherlands. Thomas Mu¨ller-Bardey was a researcher in the department of General and Comparative Linguistics at the University of Mainz, Germany, from 1992 to 2001. His main research interests include switch-reference, converbs and related phenomena both from a typological and a formal semantics point of view. He now lives in Cologne and works in public relations. Eva Schultze-Berndt is Professor of Linguistics at the University of Graz, Austria. Her research interests include typology, grammar of spoken

The contributors

xix

language, lexical semantics, language contact, and language documentation and description. She has published on complex and secondary predication, verb semantics, word classes, and construction-based approaches to grammar, both from a typological perspective and with a focus on the Northern Australian language Jaminjung, based on her own ﬁeldwork. Jane Simpson teaches linguistics at the University of Sydney, Australia. She works on central Australian languages (Warumungu and Warlpiri) and southern Australian languages (Kaurna and Thura Yura languages). She is interested in syntax, lexical semantics, comparative linguistics, child language acquisition, language maintenance, and lexicography. She has co-edited books on Indigenous Australian place names, Australian languages, and Aboriginal history. Her most recent book is A Learners’ Guide to Warumungu (2002). She is presently working with a team studying children’s acquisition of mixed languages and code-switching in Central Australian languages. Pilar M. Valenzuela teaches Spanish linguistics at Chapman University in Southern California, USA. She is interested in functional-typological syntax, language contact and change, indigenous South American languages, and anthropological linguistics. She has published a book and various articles on Panoan languages. Topics of in-depth study include synchronic and historical syntax, ethnobiological taxonomy, and borrowing of verbs. Her doctoral dissertation, Transitivity in Shipibo-Konibo Grammar, was awarded the Mary R. Haas Book Award by SSILA. Her professional experience includes work as researcher, instructor, and consultant in language development and intercultural-bilingual education among the indigenous peoples of Peru and Bolivia, as well as teaching linguistics to speakers of native languages. Together with a group of collaborators, she has compiled and edited the bilingual autobiography of a Shipibo woman, which will be published in Peru. Johan van der Auwera teaches English and general linguistics at the University of Antwerp, Belgium. His research, conducted within Antwerp’s Centre for Grammar, Cognition and Typology, concerns the typology and areal typology of grammatical semantics. Currently the focus is on modality, mood and sentence type, and negation. Earlier work concerned connectives, relative clauses, Functional Grammar, and phasal adverbials.

Abbreviations and Glossing Conventions In glossing example sentences, the following conventions obtain: Preﬁxes and sufﬁxes are separated by hyphens, clitics by an equal sign (¼), and inﬁxes are enclosed by <>. If root and afﬁxes are not separated in the vernacular, a colon is used between grammatical category labels and root meaning in the gloss (e.g. rls.ug:getting). If one formative represents two or more morphosyntactic categories, a period is used to separate the categories in the gloss (e.g. rls.ug). In the case of pronouns/pronominal clitics and afﬁxes, an exception is made with regard to the period convention in order to save space, hence: 1sg, 2pl, etc. (but: 1sg.acc). The following abbreviations are used in this volume: 1 2 3 a abl abs acc act adess adj adjz adv aff agr all anaph andat ant antip aor

ﬁrst person second person third person agent-like argument of canonical transitive verb; A-orientation ablative absolutive accusative actor adessive adjective adjectivizer adverb(ial(izer)) afﬁrmative agreement allative anaphoric (article or deictic) andative (case) anterior (event) antipassive aorist

Abbreviations and glossing conventions appl art asp assoc aux ben caus clf coi com comp compl cond cont cop cvb dat decl def dem der des det detr dim dir dist distr do.along dpct ds du dur eff emph erg ess evid excl exp

applicative article aspectual marker associative auxiliary benefactive causative classiﬁer coincidence comitative complementizer completive (aspect) conditional continuative (aspect) copula converb dative declarative deﬁnite demonstrative derivational marker desiderative determiner detransitivizer diminutive directional distal distributive do continuously while in motion depictive (marker) different subject dual durative effector (agent marker) emphatic ergative essive evidential exclusive experiential

xxi

xxii

Abbreviations and glossing conventions

f fam foc fqy fs fut g gen hab hon hsy hts ideo ill imp imprs inan inch incl incomp ind indf iness inf ingr ins intens intent intr intrss io ipfv iprf irr lig lk loc m mal man

feminine familiar (referent) focus frequency (‘times’) free-subject dependent predicate future general (gender) genitive habitual honoriﬁc hearsay evidential high tone sufﬁx ideophone illative imperative impersonal (third person) inanimate inchoative inclusive incompletive (aspect) indicative indeﬁnite inessive inﬁnitive ingressive instrumental intensiﬁer intentive (personal purposive) intransitive interessive (complement of interest) indirect object imperfective imperfect irrealis ligature linker locative; location masculine malefactive manner

Abbreviations and glossing conventions med mid min mod mot n nneg nmlz nom o obj obl onom opt ord orig p part pass pau perl pfv pl pn poss pot pred pref prev prf priv prog proh prol pron prop prox

xxiii

medial (distance) middle minimal modal element motative case (‘motion to or from’) neuter non- (e.g. nsg nonsingular, nf non-feminine, npst nonpast) negation, negative nominalizer/nominalization nominative second argument of transitive verb; O-orientation object oblique onomatopoeia optative (Georgian examples: optative of aorist series) ordinal (form of numeral) origin(ative) (case) patient-like argument of canonical transitive verb partitive passive paucal perlative perfective plural personal name (marker) possessive potential predicative (marker) preﬁx preverb perfect privative (‘without’) progressive prohibitive prolative pronoun proprietive (‘having’) proximal/proximate

xxiv

Abbreviations and glossing conventions

prs pst ptcl ptcp purp pv q quot rdp rec recp ref refl rel rem rep res restr rls rs s sbj sbjv seq sg sim simil spec ss stat sub subs subsec suff sup temp top tr transl

present past particle participle purposive patient voice question particle/marker; interrogative quotative reduplication recent past reciprocal referential reﬂexive relative remote past repetitive aktionsart resultative restrictive marker realis restricted-subject dependent predicate single argument of canonical intransitive verb; S-orientation subject subjunctive sequentive/sequential (relator) singular simultaneous converb/participle similative speciﬁc (article) same subject stative subordinator subsequent (event) subsecutive sufﬁx superlative temporal (marker) topic transitive/transitivizer translative

Abbreviations and glossing conventions ua ug uq util vblz vis vn voc vsf

unit augmented undergoer uniﬁed quantiﬁer utilitive verbalizer visible verbal noun vocative verb stem formative

xxv

This page intentionally left blank

1 Issues in the syntax and semantics of participant-oriented adjuncts: an introduction NIKOLAUS P. HIMMELMANN AND EVA SCHULTZE-BERNDT

1.1 Depictive secondary predicates in crosslinguistic perspective Constructions such as (1), where fresh functions as a depictive secondary predicate, provide a number of challenges for syntactic and semantic theory, and are frequently invoked as crucial evidence in syntactic analyses. (1) George bought the carrots fresh. Recent accounts have focused in particular on the following issues: Since depictive constructions contain two different predicates—the main predicate and the secondary predicate—they are a crucial test for theories of predication, in particular predication theory within generative grammar (Williams 1980; 1983; see Rothstein 2001 for a summary of the relevant literature). Depictive secondary predicates, being simultaneously predicates and adjuncts, take part in two relations within the clause, and thus pose problems for an analysis either in terms of constituency or in terms of dependency. Nevertheless, they are also invoked as evidence for controversial proposals in phrase structure theory such as the subjectinternal hypothesis (e.g. Roberts 1988). The role of the subject (or controller) of a depictive secondary predicate (the carrots in (1)) is often restricted to core argument functions (subject, object). For this reason they are also used as evidence in discussions of grammatical relations (e.g. Hale 1983; Kroeger 1993: 30).

N. P. Himmelmann and E. Schultze-Berndt

2

Depictive secondary predicates provide a major diagnostic for the stage-level vs. individual-level distinction (Carlson 1977; Kratzer 1995; Ogawa 2001) and currently play an important role in discussions of verbal aspect/aktionsart (e.g. Rapoport 1999). Depictive secondary predicates also provide evidence for focus theories (e.g. Winkler 1997). Discussions of these issues tend to make use of a very limited number of example types drawn from a few European languages (most importantly English, of course) and Japanese. This book, in contrast, attempts to present the full range of depictive constructions in crosslinguistic perspective. It is thus concerned with questions such as: Do all languages have (the same kinds of) depictive constructions? How are depictive constructions distinguished from related constructions in a given language? What kind of semantic content is typically rendered by a depictive construction? The contributions in this volume address these questions drawing on data from languages belonging to more than a dozen different families. This book thus builds on the short pioneering explorations of the typology of depictives by Nichols (1978a; 1978b) and Mu¨ller-Bardey (1990), and complements and expands the recent survey by Schultze-Berndt and Himmelmann (2004). The crosslinguistic perspective immediately brings to the fore an issue which is almost totally ignored in the literature on secondary predicates, but regularly surfaces in the literature on the classiﬁcation of adverbs: the problem of distinguishing between depictives and (certain kinds of) adverbials. These are widely considered two distinct types of adjunct constructions, but the boundary between them is far from clear. As a very preliminary illustration, compare the two English examples in (2) with the German example in (3): (2) a. Claire left the room angry. b. Claire left the room angrily. (3)

Claire hat wu¨tend das Zimmer verlassen.

Both (2a) and (2b) have to be translated into German as (3). That is, in English there (often) is a fairly straightforward formal contrast between a depictive construction and an adverbial construction, but no such clearcut contrast exists in German. This difference between English and German is not restricted to morphosyntactic form. German native speakers would not normally consider (3) as semantically ambiguous, but rather as vague with regard to the meaning difference between the two English examples.

Issues in the syntax and semantics of participant-oriented adjuncts

3

German is not unique in this respect: time and again when looking for depictive constructions in the languages of the world we ﬁnd that the same morphosyntactic construction may render depictive and adverbial content. For this reason, this chapter begins with a brief survey of some of the proposed classiﬁcations of adverbials and secondary predicates and then explores commonalities and differences between adverbials and depictives (section 1.2). It is argued that a number of adjunct types, including those referred to as depictives, circumstantials, oriented manner adverbs, and weak free adjuncts in studies of English syntax, share so many characteristics that at times they become virtually indistinguishable and form a single domain for crosslinguistic comparison. The term participant-oriented adjuncts is used here to refer to this domain. Participant orientation is found not only with standard examples of depictive adjuncts, i.e. adjuncts expressing a physical or mental state or condition (e.g. ‘angry’, ‘alive’, ‘hungry’, ‘drunk’, ‘raw’, ‘hot’), but also with a much larger semantic range of adjuncts, including quantiﬁcational, comitative, and locative adjuncts. Section 1.3 provides more details on the semantic range of participant-oriented adjuncts from a crosslinguistic perspective. Section 1.4 is a programmatic sketch of a morphosyntactic typology of participant-oriented adjuncts, taking into account the syntactic properties, the word class and internal structure, and the morphological marking displayed by the adjuncts in question.

1.2 The classiﬁcation of adverbials and secondary predicates In this section, after introducing some basic distinctions and terminology regarding adverbials and secondary predicates (section 1.2.1), we ﬁrst focus on participant-oriented adverbials and argue that in crosslinguistic perspective these are not necessarily distinguished as syntactic and semantic categories from depictive secondary predicates (section 1.2.2). In Section 1.2.3 we compare depictives with another type of secondary predicate, i.e. circumstantials, and again argue that these two types of secondary predicate may be clearly distinguishable in some but not necessarily in all languages. In section 1.2.4, we turn to so-called free adjuncts which have been a major concern in the semantics literature. We argue that weak free adjuncts are in fact identical to what is called circumstantial in the literature on secondary predicates. strong free adjuncts are considered independent clausal constructions only loosely connected to the matrix clause they modify. Finally, we brieﬂy consider predicative complements, another construction type which is closely related to, and sometimes indistinguishable from,

4

N. P. Himmelmann and E. Schultze-Berndt

depictives (section 1.2.5). Section 1.2.6 summarizes the major distinctions established in this section. 1.2.1 Some basic distinctions At ﬁrst sight, the classiﬁcation of secondary predicates, which are also known as predicative attributes (e.g. Paul 1919; Halliday 1967), predicative adjuncts (Hengeveld 1992a), and copredicates or copredicatives (e.g. Nichols 1978b; Plank 1985), appears to be straightforward. They are generally subdivided into depictives and resultatives. Depictives express a state that holds during the reference time of the event encoded by the main predicate. Example (1) above states that the carrots were fresh at the time that George bought them. Resultatives, on the other hand, express a state which is interpreted as a result of the state of affairs encoded by the main predicate. In (4) the carrots are soft as a result of having been boiled. (4) George boiled the carrots soft. Depictives and resultatives have in common the predicative nature of the adjunct (hence the term secondary predicate). That is, the state encoded by the secondary predicate is interpreted as holding for one of the participants of the main predicate (henceforth termed the controller), which in (1) and in (4) happens to be the direct object (for other possibilities see section 1.4.2). This participant orientation is often seen as the main feature distinguishing secondary predicates from adverbials, which are considered to be exclusively event-oriented. As we will see shortly, however, participant orientation and event orientation are not clearly aligned with depictive and adverbial constructions, respectively. Before turning to the classiﬁcation of adverbials, we may note two problems with the simple classiﬁcation of secondary predicates into depictives and resultatives. First, other more complex types of classiﬁcation have been proposed (Halliday 1967; Nichols 1978a; 1978b; 1981; Plank 1985; Mu¨llerBardey 1990), and we will have a closer look at a putative third type of secondary predicates in section 1.2.3. Second, it is not clear whether resultatives are in fact secondary predicates. Crosslinguistically, it is quite clear that resultative notions are often expressed by complex predicates, not by adjunct constructions. This is true for Warlpiri (Simpson, Ch. 2, this volume), but also for English and German—at least this has been argued by Dowty (1979), and is the central claim in Neeleman (1994) and Winkler (1997). Consequently, resultatives only play a minor role in this volume. The classiﬁcation and analysis of adverbials remains a highly controversial issue. Numerous subdivisions have been proposed (for a recent overview see

Issues in the syntax and semantics of participant-oriented adjuncts

5

Tenny 2000), and the classes resulting from the application of different criteria hardly overlap. Furthermore—and this is perhaps the most confusing phenomenon in this grammatical domain—it is the rule rather than the exception that a given adverb or adverbial allows for a number of different readings and thus often also belongs to a number of different classes. Compare the following three examples: (5) a. Elaine spoke naturally. b. Naturally Elaine spoke. c. Elaine spoke, naturally. The difference between (5a) and (5b) is obvious: In (5a) naturally says something about the way Elaine spoke, i.e. it in some sense modiﬁes the verb and is thus called a predicate-level adverb. In (5b) naturally conveys an evaluative stance of the speaker towards the proposition as a whole (for the speaker, Elaine speaking was a natural thing to happen). Here, naturally functions as a sentence-level adverb. In these two examples, the two quite different functions correlate with a difference in position (clause-initial vs. clause-ﬁnal). However, it is not the case that differences in meaning and syntactic scope are always clearly correlated with positional differences, as shown by (5c). Here naturally is prosodically detached from the rest of the clause and forms an intonation unit of its own, and both interpretations are possible.1 Variable position, sometimes correlating with a meaning difference and sometimes not, is a highly conspicuous feature common to both adverbials and depictive secondary predicates, as will become apparent throughout this chapter as well as the whole book. There are four basic kinds of parameter employed in the classiﬁcation of adverbials. Adverbials may be classiﬁed by their internal constituency (e.g. simple (lexical) adverbs, adjective-derived adverbs, prepositional phrases, adverbial clauses) or by their morphological marking (e.g. English adverbs in -ly, adverbials marked with an instrumental adposition or case). While there is little controversy about these two ways of classifying adverbials, these morphosyntactic features rarely correlate to a signiﬁcant degree with the two remaining parameters, semantics and syntactic distribution and scope (such that, for example, -ly adverbs in English would always be manner adverbs semantically). Hence, internal constituency and morphological marking are of rather limited interest for the classiﬁcation of adverbials. 1 In spoken language, the two interpretations may often be distinguishable by prosodic cues in rendering naturally, which is of no direct import to the current argument.

6

N. P. Himmelmann and E. Schultze-Berndt

The classiﬁcation of adverbials according to semantics or syntactic distribution and scope is much more problematic. This is because there are different subparameters in each domain, and many taxonomies in fact mix semantic with syntactic criteria. Syntactic sub-parameters include positional variability (e.g. for English: adverbials with ﬁxed sentence-ﬁnal position vs. adverbials with variable position, see Filipenko 2000) and syntactic scope, which in part at least correlates with position. The best-known scope distinction is the one between sentence-level and predicate-level adverbials, which is as much a syntactic as a semantic distinction. The most common and widespread semantic classiﬁcation of adverbials is the one by semantic function (or logical role), distinguishing adverbials of manner, location, time, degree, etc. (see also section 1.3). Here some confusion easily arises because terms such as manner, location, time, reason/ cause, condition, which are widely used in descriptive and pedagogical grammars, are used in a different, much more speciﬁc sense in the more recent theoretical literature, where a much larger number of semantic (or syntactic-semantic) adverbial classes are distinguished. For example, ‘manner adverb(ial)’ may be used either in its broad descriptive sense of ‘an item which usually/potentially conveys something about the manner in which an action is performed’, or in the narrower sense of ‘an item which actually conveys the manner in which an action is performed, and nothing else’. Geuder (2000), for example, distinguishes the following four subtypes of manner adverb functions: pure manner, transparent, agentive, and resultative (not to be confused with resultative secondary predicates). These are illustrated in (6)–(9) (taken from Geuder 2000: 29–35). Note in particular that the same lexical item, e.g. angrily or stupidly, can appear in more than one function. (6) a. John shouted at them angrily. (pure manner) b. John answered the question stupidly. (pure manner) (7)

He angrily broke the door open. (transparent)

(8)

John stupidly answered the question. (agentive)

(9)

They loaded the cart heavily. (resultative)

In (7)–(9) the ‘manner’ adverbs do not exclusively convey the manner in which an action is performed in the same sense as in (6). Instead, at least (7) and (8) also convey statements about one of the participants of the main event. In (7) John was angry while breaking the door open; in (8) it was stupid of John to answer the question (for (9) see Geuder 2000: Ch. 3). In this

Issues in the syntax and semantics of participant-oriented adjuncts

7

sense they are participant-oriented adverbs rather than pure manner adverbs, which are exclusively event-oriented. Before taking a closer look at participant-oriented adverbials, a note on terminology is in order. We will follow Geuder in using the term pure manner when referring to adverbials (or more precisely: adverbial uses) which convey manner and nothing else. Adverbials in the uses illustrated in (7) to (9) will be termed oriented (manner) adverb(ial)s. We will use the unmodiﬁed term manner adverb(ial) in the broad sense, i.e. with reference to an adverb(ial) which conveys manner (and thus is event-oriented) but may also have an additional orientation. In fact, as will be shown in sections 1.2.2 and 1.3.3 below, and in several of the contributions to this volume, for manner adverbials in this broad sense it can often be said that participant orientation and event orientation are simultaneously present, or can be derived from one another by metonymic shift. Note also that participant orientation is not the only kind of additional orientation. There is, for example, also speakerorientation, which characterizes evaluative uses such as (5b) above. 1.2.2 Oriented adverbials vs. depictives The fact that adverbials are not necessarily and exclusively event-oriented but instead may exhibit semantic orientation towards a participant has been widely noted in the (semantics) literature; see Platt and Platt (1972), Jackendoff (1972: 47–107), Bartsch (1976: 144), Dik et al. (1990: 31), McConnell-Ginet (1982), Frey and Pittner (1998), and Ernst (2000b), among others. Most of these authors only discuss manner adverbials (in the broad sense), although occasionally other types of adverbial are also taken into consideration. Jackendoff (1972: 57–8), for example, also regards in order to clauses (as in Bill has ruined his chances of an inheritance in order to kill his mother) as subject-oriented expressions. This issue is the main concern of section 1.3 where we make an attempt to determine the full semantic range of participant-oriented adverbials. In the current section, the main concern is as follows. Orientation towards a participant is a key feature shared by oriented adverbials and secondary predicates. This immediately raises the question of how the two construction types differ. One of the few authors who devotes extensive discussion to this question is Geuder (2000), on which the ﬁrst part of this section therefore heavily draws. We limit the discussion to two of the three types of oriented manner adverbs distinguished by Geuder, i.e. agentive and transparent manner adverbs. Agentive adverbs such as stupidly in (8) above have also been called subject-oriented adverbs and syntactically belong to the sentence-level adverbials. They ascribe a certain characteristic to the agent on the basis of the

8

N. P. Himmelmann and E. Schultze-Berndt

event which it performs. The agentive orientation of the adverb in (8) is clear from paraphrases such as (10). (10)

It was stupid of John to answer the question.

In the ‘pure manner’ use of the same adverb in (6b), in contrast, stupidity is ascribed to the way John answers the question. Ernst (2000a; 2000b) makes the intuitions behind these paraphrases more explicit by stating that agent-oriented adverbials evaluate the agent with respect to the event that is performed, where the comparison class consists of other possible events. In the case of (8), these alternatives would include silence, or an explicit refusal to answer the question. In the case of pure manner adverbials, the comparison is with other events of the same speciﬁc type. In the case of (6b) these would be other events of answering the same question, which could take place in more or less stupid ways.2 These differences in interpretation here clearly correlate with a different position of the adverbial (preverbal vs. clause-ﬁnal). Although agentive adverbs are participant-oriented, they differ from depictives in a number of ways. Semantically, depictives do not convey any evaluation of the agent’s action and thus, among other things, may be oriented towards participants other than the agent (as in John drank his coffee hot). Depictives are also never to be interpreted as predicating over the whole state of affairs: John answered the questions drunk cannot be paraphrased as ??It was drunk of John to answer the question (compare (10) above). These differences in scope often correlate with differences in syntactic position between agentives and depictives. In English, agentives are only found sentence-initially and in pre-auxiliary position, while depictives generally occur in post-predicate position. In other languages, other formal characteristics may set agentives apart from depictives. In German, for example, agentives have a special sufﬁx (-weise). Without having investigated this in detail, we think it very likely that agentive adverbials are relatively easily distinguishable from depictives on formal as well as semantic grounds in all languages. Therefore, they will not be considered any further in the remainder of this chapter. The second type of participant-oriented adverbs—transparent adverbs3— are very similar to depictives, however. Compare the following examples: (11) a. John read the review slowly. (pure manner) b. John angrily read the review. (transparent) c. John left the party angry. (depictive) 2 Geuder (2000: 137) proposes to re-analyse Ernst’s ‘comparison classes’ as alternatives in the sense of focus semantics. 3 Regarding the rationale for this less than transparent term, see Geuder (2000: ch. 2.1.1).

Issues in the syntax and semantics of participant-oriented adjuncts

9

As shown in detail by Geuder (2000: ch. 5), angrily in (11b) is not a pure manner adverb like slowly in (11a). It does not exclusively indicate the manner of reading, which is why (11b) cannot be paraphrased as John read the review in an angry manner. Rather, it is participant-oriented in that it describes the emotional condition of the subject at the time of reading. In this respect it bears a close semantic relationship to the depictive in (11c), which also encodes that John was angry at the time of leaving the party. There is a rather subtle semantic difference between the transparent and the depictive construction. The depictive construction in (11c) merely entails that the two states of affairs of leaving the party and being angry overlap in time. Note in particular that the departure here is not necessarily related to the state of being angry. The adverb in the construction in (11b), on the other hand, conveys a closer factual link between the activity denoted by the main predicate and the bodily or psychological condition of the subject. This link pertains to the fact that the way in which the subject is engaged in the activity denoted by the main predicate suggests something about its bodily or psychological condition which in turn suggests that activity and bodily or psychological condition are inherently linked in some way. The precise nature of the factual link varies. Two possibilities mentioned by Geuder are causal consequence and motive. In (11b), the link appears to consist in a causal consequence: John becomes angry because of reading the review. An example showing a motivational link between the state conveyed by the transparent adverb and the activity denoted by the main predicate is (12), where the state of being hungry is understood as the motivation for the boy’s return to his parents. (12) The boy hungrily returned to his parents. (Geuder 2000: 204) The major semantic difference between depictives and transparent adverbs thus pertains to the relation established between the state of affairs denoted by the main predicate and the condition denoted by the participant-oriented adjunct. In the case of depictives, the relation is one of mere temporal overlap, while in the case of transparent adverbs it goes beyond mere temporal coincidence and reﬂects a ‘deeper’, factual link (often causal or motivational). An important correlate of this semantic difference would appear to be the fact that transparent adverbs, unlike depictives, always seem to exhibit agent orientation (on the range of controllers for depictives, see section 1.4.2 below). Geuder discusses the intricacies involved in providing a detailed formal representation of the meaning differences between depictives and transparent adverbials. There is one point in this discussion that we wish to

10

N. P. Himmelmann and E. Schultze-Berndt

emphasize here. Within the framework of Davidsonian semantics, where verbs carry an event variable and adverbials are analysed as predicates of events, there is an apparently straightforward way of representing the difference in meaning between (11a) (pure manner) and (11c) (depictive), as shown in (13). (13) a. read(e) (x, review) & slow(e) (pure manner) b. leave(e) (x, party) & angry(x) (depictive) The representations in (13a) and (13b) make a clear-cut distinction between event orientation and participant orientation: in (11a) John read the review and his reading was slow while in (11b) John left the party and was angry at the same time. As Geuder notes, there are two major problems with this approach. First, it does not offer an equally simple representation for transparent adverbs. Second, and more importantly, the representation of the depictive construction in (13b) only captures the participant orientation of angry. It does not represent the relation of temporal overlap between the two states of affairs of leaving and being angry, and thus, for example, does not explain why individual-level predicates such as clever cannot function as depictives (consider *John left the party clever). Geuder consequently proposes more complex representations for depictives and transparent adverbs, and states in conclusion: In fact, depictives have been found to carry an almost ‘adverbial’ reading: Their interpretation is linked to the event variable of the verb. The difference between depictives and transparent adverbs lies merely in the fact that depictives assert the independence of a concurrent state while the adverbial forms assert the existence of a closer factual connection to the event. (Geuder 2000: 213)

That is, at least in some of their uses, manner adverbs and depictives are very similar indeed. It will thus not come as a surprise that the different meanings conveyed by depictives and (in particular participant-oriented) manner adverbials are not distinguished on the formal, i.e. morphosyntactic, level in all languages, as seen in the following discussion. The semantic differences between depictives and participant-oriented uses of manner adverbs in English may be rather subtle, and native speakers not specializing in semantics usually ﬁnd it very hard to make them explicit. But there is little doubt that such differences exist. This may be due to the fact that in English most manner adverbs involve the sufﬁx -ly and thus differ formally from depictives. In German, on the other hand, for many bases expressing bodily or psychological conditions such as traurig ‘sad’, wu¨tend

Issues in the syntax and semantics of participant-oriented adjuncts

11

‘angry’, etc. there is no morphological difference between the use as a manner adverbial and as a (primary or secondary) adjectival predicate, making it questionable whether these uses can be distinguished at all.4 Geuder (2000: 192) mentions the possibility that they may be distinguished by position (and stress), such that (14a) only allows the depictive reading and (14b) only allows a manner reading (pure manner or transparent). German (14) a. daß Hans den Raum traurig verließ that John the room sad left [with the major accent on traurig] ‘ . . . that John left the room sad’ b. daß Hans traurig den Raum verließ that John sad the room left ‘ . . . that John left the room sadly’ (Geuder 2000: 192) However, to us—and, we believe, most native speakers—this difference is far from clear-cut, and we would argue that both (14a) and (14b) are vague rather than ambiguous with respect to a depictive or an adverbial (pure manner or transparent) interpretation. While the emotional condition traurig in (14b), but probably not in (14a), can be construed as a motivation for leaving the room (which may simply be due to the iconicity of ordering; cf. Geuder 2000: 212), in both examples an interpretation of pure temporal overlap as well as of pure manner can be construed. (See also our discussion of example (3) above.) In languages such as German where depictive and adverbial constructions are often not distinguished by morphological marking, it is common to diagnose the differences between the two by the paraphrases they allow (cf. e.g. Bartsch 1972: 140f.; Pittner 1999: 97). It is assumed that depictive constructions, but not adverbials, can be paraphrased by a biclausal construction where the element corresponding to the depictive functions as a primary predicate. Thus, (15) is a possible paraphrase of (1). (15)

The carrots were fresh (when George bought them).

Conversely, adverbial constructions, but not depictives, can be paraphrased with a clause where the element corresponding to the adverbial functions as the main predicate, or by an attributive construction where the element corresponding to the adverbial is a modiﬁer of a verbal noun. Thus, both (17a) and (17b) are possible paraphrases of (16), but (18a) and (18b) are 4

See also van der Auwera and Malchukov (Ch. 13, this volume) for Dutch.

12

N. P. Himmelmann and E. Schultze-Berndt

not possible paraphrases of (1). (16)

George bought the carrots fast.

(17) a. The buying of the carrots (by George) was fast. b. The fast purchase of the carrots (by George). (18) a. ??The buying of the carrots (by George) was fresh. b. ??The fresh buying of the carrots (by George). However, even these examples may be used to show that the supposedly clear semantic difference between participant-oriented (depictives) and event-oriented adjuncts (adverbials) is not quite so straightforward. That is, the paraphrase tests which are supposed to be diagnostic for this difference are ﬂawed in that it is possible to paraphrase the ‘adverbial’ construction in (16) as (19), in analogy to the paraphrase of the depictive construction in (15). (19) George was fast (when he bought the carrots). Thus, as has also been pointed out by Plank (1985), semantic intuition and paraphrase tests are often at a loss to distinguish between a participantoriented and an event-oriented adjunct in the absence of overt formal marking, since it appears that one and the same adjunct (such as fast in (16)) can be construed as simultaneously participant-oriented and event-oriented. From a wider crosslinguistic perspective, there is further evidence for a much more pervasive overlap between participant-oriented and eventoriented adjuncts than the traditional dichotomy between secondary predicates and manner adverbials suggests. One similarity between manner adverbials and depictives in many languages, including English, is that both can be substituted by the manner interrogative (‘how’) (see Boeder, Ch. 6, this volume, for Georgian). Moreover, constructions that can be identiﬁed as depictive constructions on formal grounds may also encompass expressions that we might regard as primarily event-oriented on semantic grounds—in other words, the morphosyntactic marking contradicts our semantic intuitions in these cases. For example, adverbs like slowly and fast are classiﬁed by almost all authors as pure manner adverbs, i.e. as exclusively event-oriented adverbs, on semantic grounds (although, as we have already seen in (19), speed can also be ascribed to a participant directly). However, there are languages in which the translation equivalent of slowly behaves exactly like a depictive. As (20) shows, in the Australian language Martuthunira, the manner adjunct jarruru ‘slow(ly), careful(ly)’ agrees in case with a controller in the same way as the typical depictive wanka ‘alive’. (In (20a), the controller is not overtly expressed.)

Issues in the syntax and semantics of participant-oriented adjuncts

13

Martuthunira (20) a. ngunhaa malumalu-npa-waa¼rru jarruru-u waruul dem.nom dark-inch-purp¼now slow-acc still ‘(It’s slowly getting cool now as the sun goes down). That will make everything go dark slowly.’ (Dench 1988: 121) b. nhulaa miyu mungka-rnuru wajupi-i wanka-a¼l dem.nom cat eat-prs grasshopper-acc alive-acc¼temp ‘this cat eats grasshoppers alive.’ (Dench 1995: 182) A number of authors have proposed that a process of metonymic shift may provide the explanation for the fact that event-oriented adjuncts may be interpreted as participant-oriented (and vice versa), a possibility already hinted at in our discussion of the transparent use of angrily illustrated in (11b) above. Consider (21). (21)

Fred ate the sausages ravenously.

In the discussion of this example, Platt and Platt (1972: 237) note: We do perhaps talk about ravenous eating but in saying this we imply that the eater is ravenous. It may be objected that this ravenous quality displays itself in the manner of the eating. This is not denied. The manner of eating is an outward and visible sign of an inner ravenous quality of the eater. To say that someone eats ravenously is a judgement about the eater.

The authors then conclude that ravenously is primarily a participant-oriented adverbial (where the participant in this case is the agent), but that the nature of the agent may be inferred from the manner in which the event is performed, and this is how the adverbial derives a secondary, event-oriented sense. A similar conclusion is drawn by Geuder (2000: 214–16) for the adverbial sadly in the translation of (14): the manner of leaving the room is not sad as such, but it allows the conclusion that the subject is sad at the time of leaving the room. In an analogous fashion, Vogel (1997: 412) interprets German leise ‘quiet’ as a primarily event-oriented adverb. When it is predicated of an animate being (as in John is quiet), this is interpreted as ‘John generally behaves in a quiet manner’ (see also Eisenberg 2002: 63). In the analyses just referred to, the adverbials in question are interpreted as polysemous, being basically event-oriented but also allowing a derived participant-oriented reading (or vice versa). However, this would not appear to be the only possible analysis. Instead, the adverbials in question could be considered to be semantically general, ‘vague’, or ‘underspeciﬁed’ (Do¨lling 2003). That is, they exhibit both event-oriented and participant-oriented

14

N. P. Himmelmann and E. Schultze-Berndt

components simultaneously, without one being primary and the other secondary. The metonymic transfer is inherent in their lexical semantics, because it is always possible to ascribe the manner of performing an event to the agent, and to conclude the agent’s nature from the way in which the event is performed. A more detailed argument along these lines is put forward by Boeder (Ch. 6, this volume). It is this dual nature of manner adverbials (in a broad sense) that allows us to paraphrase (16) as both (17) and (19). The dual nature of many types of adjunct with regard to event- and participant-orientation also points to an explanation for the crosslinguistic differences in coding hinted at above. In Schultze-Berndt and Himmelmann (2004) we have suggested that there exists a continuum between expression types (such as expressions of physical condition) where the participantoriented component prevails and those (such as expressions of time) where the event-oriented component prevails. Manner expressions are situated in the middle of this continuum and are therefore encoded as adverbials in some languages (as in the case of agent-oriented adverbials in -ly in English), and as depictives in others, as shown for Martuthunira in (20). Languages may also exhibit an overlap of both adverbial and depictive coding strategies for certain semantic types of adjunct, in particular manner adjuncts, with no or only very subtle differences in meaning (as shown by Geuder (2000) for English, and Boeder (Ch. 6, this volume) for Georgian), or as a stage in a diachronic development from one preferred strategy to another (see Pinkster 1988: 227 for Latin). As a third possibility, a language may choose a nonspeciﬁc coding strategy, as has just been argued for German manner adjuncts. Given that participant orientation and event orientation may thus overlap and cooccur in a broad variety of linguistic structures, the following terminological conventions will be useful for keeping track of crosslinguistically recurrent (non-)distinctions. The terms construction and adjunct are used in this chapter only in reference to morphosyntactically deﬁnable units while expression is used primarily in reference to semantically deﬁnable units. Participant orientation and event orientation are understood here to be primarily semantic relations. A depictive construction (or secondary predicate) is a construction primarily dedicated to conveying participant orientation, although, as we will see, certain expressions involving depictive constructions probably have to be regarded as event-oriented in some languages. Frequently, participant orientation is iconically reﬂected in the formal marking of a depictive construction, e.g. in the agreement of the depictive with its controller. The important point, though, is that the depictive construction differs in at least one morphosyntactic feature from other adjunct constructions. This feature may pertain to syntactic position or

Issues in the syntax and semantics of participant-oriented adjuncts

15

distribution or to the fact that the construction is morphologically unmarked while all other adjunct constructions are marked. Similarly, an adverbial construction is a distinctive construction (e.g. one marked by adverbial afﬁxation, or by a distinct position) which is primarily dedicated to conveying event orientation, although in some cases it may be interpreted as participant-oriented. The term general adjunct construction, on the other hand, is used for adjuncts which are neutral with regard to the event-oriented/participant-oriented distinction and thus allow for both a participant- and an event-oriented interpretation as an unmarked reading. With regard to (morphosyntactically deﬁned) adjuncts, we thus distinguish between depictive, adverbial, and general adjunct constructions (see also Schultze-Berndt and Himmelmann 2004). As these deﬁnitions make clear, semantic interpretation as participant- or event-oriented and morphosyntactic properties of a construction do not necessarily align. Consequently, in the current framework it is not a contradiction to speak of participant-oriented uses of an adverbial construction (e.g. the transparent uses of English manner adverbs discussed above) or event-oriented instances of a depictive constructions (e.g. the agreeing temporal expressions discussed in section 1.3.11 below.). 1.2.3 Depictives vs. circumstantials In this section, we will introduce a further subdivision in the domain of depictive constructions as deﬁned in the previous section, into depictives proper and participant-oriented adjuncts of another type, so-called circumstantial (or conditional) secondary predicates. The latter are usually morphologically identical to depictives; in English, for example, neither depictives nor circumstantials carry a special morphological marking. The main crosslinguistic ﬁnding is similar to the one in the preceding sections. In some languages, including English, depictives proper and circumstantials are two closely related but nevertheless distinct constructions, the main difference resulting from their interaction with focus. In other languages there may be just a single construction for both functions. As brieﬂy mentioned in section 1.2.1 above, it is common in the literature on secondary predicates to distinguish between depictives and resultatives, and especially in the more recent literature this already completes the top level of the taxonomy. Some authors, however, discuss a third, related construction. Halliday (1967: 78–81) terms this construction condition(al) (short for ‘conditional attribute’). Halliday himself notes that ‘the distinction between attribute [i.e. depictive—NPH and ESB] and condition may seem somewhat arbitrary; and certainly it may not be very clear to which type a

16

N. P. Himmelmann and E. Schultze-Berndt

given token should be assigned’ (1967: 80). Despite these problems of delimitation and the fact that they share a number of important features with depictives and resultatives (1967: 62), Halliday clearly sets conditionals apart from depictives/resultatives in his overall systematics. It is thus unclear whether or not conditionals should also be considered secondary predicates in his taxonomy. In some of the examples that Halliday provides for conditionals, e.g. those in (22a) and (22b), the conditional meaning seems to arise from the presence of a modal element in the main predicate. For example, the relation between the adjunct raw and the main predicate (and the undergoer) in (22b) seems to be identical to that in you eat them raw, which Halliday considers a depictive construction (cf. also Aarts 1995: 78f.). (22) a. I can carry it empty. b. You can’t eat them raw. c. She died young. (Halliday 1967: 78) In other examples, e.g. (22c) (which according to Halliday is ‘surely conditional’ as opposed to depictive she died happy), it appears to be the secondary predicate itself which contributes the ‘conditional meaning’. The term condition is used here in two quite different senses. In (22a) and (22b) it refers to a precondition which could also be rendered by a conditional clause (I can carry it if it is empty). In (22c), on the other hand, condition refers to a life stage (she died when she was young). We argue here that there may be some support for distinguishing the former usage (i.e. examples (22a) and (22b)) from depictives while the latter is in fact depictive. To date, conditionals have generally not been acknowledged as a distinct type of secondary predicate in the literature, with a few exceptions. The major exception is Nichols (1978a: 117; 1981; see also Plank (1985: 169–70), and Simpson (Ch. 2, this volume)), who regards conditionals as a subtype of what she calls circumstantial secondary predicates, thus making it clear that they are secondary predicates. Examples of conditional circumstantial secondary predicates in Nichols’s sense are found in (23). (23) a. This food is not supposed to be nice cold. [overheard utterance] b. I can’t work hungry. The other two subtypes of circumstantial secondary predicate recognized by Nichols are temporal (As a child he lived in Paris or I knew him young) and concessive (Even dead I won’t forget). Quite clearly, then, Nichols’s term

Issues in the syntax and semantics of participant-oriented adjuncts

17

circumstantial, which we will adopt here, covers the same ground as Halliday’s and Simpson’s conditional.5 At ﬁrst sight, the distinction between circumstantial and depictive secondary predicates appears to be primarily a semantic one, pertaining to the logical relation between main and secondary predicate. In the case of depictives (He bought the carrots fresh, They left outraged), the logical relation between main and secondary predicate is exclusively one of temporal overlap: the state referred to by the depictive holds true at the same time as the event expressed by the main predicate (and it may have held true before that point in time and keep on holding true after it). In the case of circumstantials, other semantic links are evoked in addition to the basic pure temporal overlap. Following Nichols, these ‘other semantic links’ could be classiﬁed as condition, concession, and temporal (in this regard circumstantials are similar to the transparent uses of manner adverbs discussed above).6 Since this semantic distinction at times ‘may seem somewhat arbitrary’, as Halliday rightly stresses, one may wonder whether there are any formal correlates which would allow us to make the distinction more operational. Two major candidates for such formal correlates which are occasionally invoked in the literature are negation and position. We take up negation ﬁrst. Circumstantial secondary predicates, such as cold and hungry in (23) and as a young girl in (24), appear to be outside the scope of negation, i.e. in these examples it is possible to negate the main predicate without negating the secondary predicate (see also Simpson, Ch. 2, this volume). (24) As a young girl Sarah did not travel to Paris alone. This clause can certainly not be understood as denying that Sarah was a girl in the same way that (25) can be read as denying that they were outraged. (25) They didn’t leave outraged (they left happy). That is, the negator in depictive constructions such as (25) can be interpreted as having either the depictive alone in its scope, or both the depictive and the main predicate (No! They didn’t leave outraged, they didn’t leave at all), but not just the main predicate as is usually the case with circumstantials.

5 Note that Halliday also uses the term circumstantial, but in a very different sense. His circumstantial is a cover term for the following four constructions: beneﬁciaries (roughly: indirect objects), range (e.g. wall in he jumped the wall), attribut(iv)es (depictives, resultatives), and his condition(al)s (1967: 52–3). 6 In the case of temporal circumstantials (As a child he lived in Paris) the additional semantic link consists in the fact that the secondary predicate (as a child) contributes to the determination of the reference time of the clause in a way similar to temporal adverbials (Twenty years ago he lived in Paris).

18

N. P. Himmelmann and E. Schultze-Berndt

With regard to the second formal parameter, position, it has been noted that in English depictives usually have to occur in post-predicate position, while circumstantials often occur in clause-initial position. This is completely unproblematic for As a child he lived in Paris and Even dead I won’t forget. In the other examples of circumstantials mentioned above, initial position is at least marginally possible, although for some of the presumed temporal circumstantials, e.g. (26e, f) the result is not too good (as indicated by the question marks): (26) a. Empty I can carry it. b. Raw you can’t eat them. c. Cold this tea isn’t good. d. Hungry I can’t work. e. ??Young I knew him. f. ??Young she died. The preceding observations point to the fact that what is actually at stake here is focus. Typical depictive secondary predicates always provide focal information, as has been argued at length by Winkler (1997) (see also Geuder 2000: 199–200; Gu¨ldemann, Ch. 10, this volume). They are part of the focus domain of the clause they appear in and, at least in standard examples such as They left angry and She returned drunk, they in fact carry the sentence accent and thus function as the focus exponent. Being part of the focus domain explains why depictives are usually not acceptable in initial position in English (except in contrastive contexts), and why they are within the scope of clause-level negators. Circumstantials, on the other hand, contribute presupposed information to an utterance. They are generally unstressed (unless they attract contrastive stress) and outside the scope of clause-level negators. Thus, they appear either in pre-predicate position (as in (24)), or they are clause-ﬁnal and unstressed (see also Pittner 1999: 115 for German). Thus (26c) is equivalent to (27a), while (27b) is only acceptable in contrastive contexts (such as this tea isn’t good cold, it’s good hot). (27) a. This tea isn’t good cold. b. ?? This tea isn’t good

COLD. COLD.

In this view, the fact that young in I knew him young and she died young is usually stressed and strongly disfavours initial position implies that it is a depictive and not a circumstantial in these examples (pace Halliday and Nichols). This is also supported by the negation test, since She didn’t die young allows a reading in which only the secondary predicate is in the scope of negation (She didn’t die young, she died in her seventies).

Issues in the syntax and semantics of participant-oriented adjuncts

19

To sum up this section: in English it is possible to distinguish two kinds of otherwise very similar secondary predicates, depictives proper and circumstantials. Both are participant-oriented adjuncts which convey a state of affairs which temporally overlaps with the state of affairs conveyed by the main predicate. They differ, however, in that depictives are part of the focus domain and convey focal information while circumstantials do not. This explains their positional and prosodic differences (or rather preferences) as well as the fact that depictives, but not circumstantials, can be in the exclusive scope of a negator. On the semantic level, circumstantials often evoke a speciﬁc semantic link (condition, concession) to the main predicate in addition to the basic pure temporal overlap characterizing depictives. This can be regarded as a pragmatic inference deriving from the non-focal quality of these adjuncts, not as a deﬁning characteristic of the construction: if the adjunct is not in focus, it is part of the presupposition and in restrictor position (cf. Mu¨ller-Bardey, Ch. 3, this volume), and its function is that of restricting the interpretation of the constituent(s) in focus. While it is very likely that participant-oriented adjuncts may appear in focus or in the presupposition in all languages where participant-oriented adjunct constructions occur, it is not necessarily the case that this difference with regard to information structure is also manifest on the morphosyntactic level. That is, it would appear to be possible that a language does not distinguish between depictive and circumstantial constructions, but has a single construction conveying participant orientation which is open to both a focal and a presuppositional interpretation, based primarily on pragmatic inferences. The fact that depictives and circumstantials thus are very closely related constructions gives rise to different readings of the term depictive: In one sense, it refers to participant-oriented adjuncts which not only are formally distinct from event-oriented adjuncts, but also belong to the (morphosyntactically delimited) focus domain. These may be called depictives proper. On the other hand, it may refer to adjunct constructions conveying participant orientation regardless of their information-structural status, hence encompassing depictives proper and circumstantials. In line with widespread usage in the current literature (including most contributions to this volume), these may be simply called depictives. We will be using depictives in the broad sense whenever it is important to emphasize that a given argument or statement pertains to both depictives proper and circumstantials. 1.2.4 Depictives vs. free adjuncts So-called free adjuncts (e.g. Having unusually long arms, John can touch the ceiling) are also participant-oriented and share essential similarities with

20

N. P. Himmelmann and E. Schultze-Berndt

depictives such as denoting a condition or state which temporally overlaps with the state of affairs denoted by the main predicate. Here we argue that so-called weak free adjuncts are actually identical to what has been called circumstantial secondary predicates in the preceding section, thus unifying insights from two different strands in the literature which appear to have been unaware of the fact that they deal with essentially the same phenomenon. Strong free adjuncts are only loosely attached to, and thus in fact outside, the clause headed by the primary predicate, and hence strictly speaking not secondary predicates. The distinction between strong and weak free adjuncts has been proposed by Stump (1985), who provides the most detailed semantic analysis of free adjuncts to date. The following two of his examples (1985: 41–2) may serve as an initial illustration: (28) a. Standing on a chair, John can touch the ceiling. (weak) b. Having unusually long arms, John can touch the ceiling. (strong) In (28)a, the initial adjunct is similar to a conditional clause in that it restricts the interpretation of the modal, as in If he stands on a chair, John can touch the ceiling. In (28b), on the other hand, the adjunct does not have such a restricting function and rather resembles a causal clause (Because he has unusually long arms, John can touch the ceiling). This difference correlates with a difference in entailments. The actual truth of the strong adjunct in (28b) (i.e. that John has unusually long arms) is always entailed by the truth of the whole sentence, while the actual truth of the weak adjunct in (28a) (i.e. that John is standing on a chair) is not necessarily entailed by the truth of the whole sentence (Stump 1985: 41–64). The difference between weak and strong free adjuncts is linked in part to the meaning of the adjunct. If the adjunct involves an individual-level predicate (roughly: the subset of stative predicates which refers to inherent, essential properties) then it is always strong. That is, individual-level predicates like having unusually long arms in (28b) may not function as weak free adjuncts. Only stage-level predicates such as standing on a chair or drunk allow weak uses. However, a weak use is always just a possibility, and very much depends on the larger context in which the adjunct occurs, as repeatedly emphasized by Stump (1985: 55 and passim).7 Thus given the right context, standing on a 7 Note that we deviate here in a minor way from the terminological conventions used by Stump. For Stump, there are two kinds of free adjunct, strong and weak, the latter allowing two kinds of use, which are referred to in various ways including strong vs. weak. In the present chapter, weak free adjunct refers to a stage-level predicate used in weak free adjunct function, while strong free adjunct refers to an individual-level or a stage-level predicate used in strong free adjunct function.

Issues in the syntax and semantics of participant-oriented adjuncts

21

chair in (28a) can be interpreted as a description of an actual occurrence of John standing on a chair and can thus function as a strong free adjunct. In this case the actual truth of the adjunct would be entailed, and a causal rather than a conditional link between adjunct and main predicate would be evoked. Furthermore, the difference between strong and weak adjuncts only becomes apparent in a limited number of morphosyntactic contexts. One such context are modally marked main predicates, as in the above examples. Other contexts discussed by Stump are frequency adverbs (e.g. Lying on the beach, John sometimes smokes a pipe; see also Mu¨ller-Bardey, Ch. 3, this volume) and generic or habitual sentences (e.g. Drunk, he drives very dangerously). In these contexts, weak free adjuncts function as arguments of a clause-level operator (such as a modal or a frequency adverb), while strong free adjuncts remain outside the scope of this operator. Importantly, then, weak free adjuncts are part of the same clause as the (ﬁnite) main predicate because both function as arguments of the same operator. Strong free adjuncts, on the other hand, express ‘a proposition logically distinct from that expressed by the main clause’ (Stump 1985: 276–7). The preceding short summary of the major differences between strong and weak free adjuncts will have made it clear that weak free adjuncts have much in common with the circumstantial secondary predicates discussed in the preceding section. Both are non-focal participant-oriented adjuncts which preferably occur in modal or habitual expressions and often involve a conditional link between adjunct and main proposition. Stump (1985: 87f., 272–7) in fact claims that some of those prepositions which mark circumstantial secondary predicates such as as in as a child serve to turn strong free adjuncts into weak ones (or individual-level predicates into stage-level predicates). Stump’s example is: A blonde, Mary might look something like Jane (only strong) vs. As a blonde, Mary might look something like Jane (weak or strong). A possible objection to collapsing weak free adjuncts and circumstantials in a single category may be differences with regard to lexical category, position, and prosody. Standard free adjunct examples usually consist of participial expressions in initial position, which is the most commonly attested position. Circumstantials are often illustrated with single adjectives, usually also in initial position but occasionally also in other positions. But this difference with regard to standard examples is purely superﬁcial. Weak free adjuncts may also simply consist of a single adjective; they may also occur after the superordinate clause and sometimes even immediately after the subject of the superordinate clause (Stump 1985: 6). As far as prosody is concerned, note that the fact that both strong and weak free adjuncts are generally separated with a comma from the superordinate

N. P. Himmelmann and E. Schultze-Berndt

22

clause is an orthographic convention in English and does not necessarily indicate a prosodic break. It is not unlikely that strong and weak adjuncts are actually distinguished by the presence vs. the absence of a prosodic break. But this is an empirical issue impossible to decide at this point, as we lack crucial evidence. However, we may note that the occurrence of a clear prosodic break appears to turn typical depictives into independent clause-like units or, in Stump’s terminology, strong free adjuncts, even when they occur in utteranceﬁnal position. Compare the following examples (the comma here represents a clear prosodic break).8 (29) They returned early in the morning, (totally) drunk. (30) They didn’t leave, outraged. In example (30), the scope of negation makes it clear that we no longer deal with a ‘true’ depictive (in contrast to example (25) above). The only possible interpretation of this sentence is that the subject didn’t leave at all, being outraged. Because of its prosodic detachment outraged here is no longer part of the same clause (and proposition) as leave and hence it is not a secondary predicate but rather constitutes a reduced clause of its own. To conclude—strong free adjuncts are another type of participantoriented expression which differs quite clearly from depictives and circumstantials in that it is not part of the same clause as the primary predicate and hence not, strictly speaking, an adjunct (nor a secondary predicate). Weak free adjunct, on the other hand, appears to be simply another term for circumstantial secondary predicates. In the preceding section we suggested that circumstantials/weak free adjuncts are non-focal participant-oriented adjuncts, differing from depictives proper in that they provide presupposed information rather than conveying focal information. In Stump’s discussion of circumstantials/weak free adjuncts, which is concerned with distinguishing these from strong free adjuncts rather than from depictives proper, the major characteristic of this type of adjunct is its interaction with an operator in the main clause (e.g. a modal element, a frequency adverb). Mu¨ller-Bardey (Ch. 3, this volume) adopts basic insights from Stump’s analysis to the analysis of depictives (in the broad sense) and recasts it in terms of the so-called partition of the clause. In this analysis, the operator divides the clause into two parts, with one part functioning as a restrictor for the operator and the other part functioning as its nuclear scope. Depictives (in the broad sense) can occur in either part. When they occur as part of the nuclear scope (as in George usually 8

For a spontaneous example, see ex. (18) in Schultze-Berndt and Himmelmann (2004: 68).

Issues in the syntax and semantics of participant-oriented adjuncts

23

drives home drunk), they are depictives proper in our terminology. When they occur as part of the restrictor (as in Drunk, George drives very dangerously where drunk restricts the occurrences of George’s driving dangerously), they are circumstantials in our terminology. Thus, there appear to be two different, not necessarily overlapping properties distinguishing between depictives proper and circumstantials/weak free adjuncts. On the one hand, the distinction between depictives proper and circumstantials appears to correlate with being in focus vs. being in the presupposition (as proposed in the previous section). On the other hand, it also appears to correlate with being in the restrictor vs. being in the nuclear scope, as suggested by Mu¨ller-Bardey’s analysis. A further exploration of these relationships is not a straightforward matter, as it involves a fair number of complex issues. To begin with, it is not clear whether the focus/ presupposition and the restrictor/nuclear-scope distinctions themselves differ substantially (regardless of how they relate to depictives) or whether they are just different ways of making essentially the same distinction. Mu¨ller-Bardey (Ch. 3, this volume, sections 3.2.3 and 3.2.4) argues that they are not, although examples where items are distributed across non-matching categories (i.e. an item in the restrictor is in focus or an item in the nuclear scope is presupposed) are hard to come by (but see Mu¨ller-Bardey’s example (47) and discussion). Assuming that focus/presupposition and restrictor/nuclear scope actually refer to substantially different distinctions, one relating to information structure, the other to logical form, it then becomes a rather intricate empirical problem to determine whether they make different predictions with regard to the distinction between depictives proper and circumstantials. One major difference here is that the restrictor/nuclear-scope distinction requires the presence of an operator inducing the partition of the clause, which in turn predicts that circumstantials only occur in clauses which include such operators (recall from section 1.2.3 above that circumstantials generally involve some additional semantic link between secondary and main predicate which we broadly characterized as condition, concession, or temporal).9 We are currently not in a position to pursue this issue any further. In fact, it is not clear to us whether there is much to be gained by further pursuing it, since for practically all uncontroversial examples of either depictives proper 9 So far, we have not come across any reasonably natural examples which would allow us to resolve this issue. A relevant example would have to look something like ??Fresh we drank orange juice neat that night, where fresh would have to be presupposed information (see also Simpson, Ch. 2, this volume, section 2.2.5).

24

N. P. Himmelmann and E. Schultze-Berndt

or circumstantials known to us, the two properties match perfectly: depictives proper are both in focus and in nuclear scope, circumstantials are presupposed and in the restrictor. To simplify the ensuing exposition, we simply assume that being non-focal is the major and deﬁning feature of circumstantials. 1.2.5 Depictives vs. predicative complements For some authors (including Halliday 1967; Nichols 1978a; Napoli 1989) the notion of (depictive) secondary predicate also includes constructions where the second predicative element is not an adjunct but a complement, i.e. where it is in some sense selected by the main predicate, and is therefore obligatory, as in (31) (from Nichols 1978a: 114). (31) a. They elected him president. b. Rocks serve them as support. It cannot be denied that there are certain functional and formal relationships between predicative complements and depictives. Functionally, both are obviously participant-oriented and convey partially independent predications about one participant of the main predicate. Moreover, a secondary predicate can rarely be selected completely independently of the main predicate (see also section 1.4.1 below). Simpson (Ch. 2, this volume) in fact argues that in a language like English, these restrictions are so strong that the depictive elements in standard depictive examples such as They returned drunk may have to be analysed as complements rather than adjuncts. Even if one does not want to follow Simpson’s argument for English, it is clear that once one acknowledges that there are at least weak selectional restrictions between main and secondary predicates, the question of distinguishing depictives and predicative complements becomes a complex issue. In fact, given the close surface similarities between predicative complements and depictive adjuncts, distinguishing them would appear to be just another instance of the well-known problem of distinguishing complements and adjuncts in general (see Dowty 2003 for a recent survey and discussion of complement/adjunct similarities and historical transitions between them). Formally, in many (though by no means all) languages, predicative complements and depictives may have the same internal constituency and the same morphological marking (if any). Examples of the latter include the predicative marker as in English, illustrated in (31b), the Russian instrumental (van der Auwera and Malchukov, Ch. 13, this volume) and the predicative marker ¼ i in Ewe (Ameka, Ch. 11, this volume). Plank (1985) in fact advances the hypothesis that depictive constructions are always derivative of

Issues in the syntax and semantics of participant-oriented adjuncts

25

predicative complement constructions, i.e. that they always exhibit the same structure and marking as predicative complement constructions. However, Bucheli Berger (Ch. 4, this volume) provides counter-evidence from Swiss German dialects with a generalized depictive marker which is not found on predicative complements. Nevertheless, in those instances where predicative complements and depictives have the same internal constituency and the same morphological marking, one may wonder whether we are dealing with distinct construction types. The answer depends very much on whether selectional restrictions between the main predicate and its complements are formally manifest in a given language. Of major importance in this regard is clear-cut evidence for the obligatoriness of some argument expressions but not others. It would appear that once again this is a grammatical domain where there is ample room for crosslinguistic variation. Simpson’s comparison of Warlpiri and English adjuncts and complements in Chapter 2 illustrates one fairly clear case of such variation. See also Ogawa (2001: 9), who claims that the multiple object construction in Korean is a depictive secondary predicate construction despite the fact that the second object is obligatory. On the basis of the pattern of formal distinction vs. indistinction, van der Auwera and Malchukov (Ch. 13, this volume) conclude that depictives and predicative complements are in principle distinct, but adjacent on their semantic map of construction types. 1.2.6 Summary of basic distinctions regarding participant-oriented adjuncts In this section, we have discussed a number of adjunct types in English which share the characteristics of allowing participant-oriented readings and of denoting a state or condition which temporally overlaps with the state of affairs designated by the main predicate, i.e. depictive secondary predicates, participant-oriented manner adverb(ial)s, circumstantial secondary predicates, weak free adjuncts, and strong free adjuncts. The discussion suggests a somewhat revised set of terminological distinctions, with their crosslinguistic applicability in view. In particular, we propose that the term depictive proper (or depictive in the narrow sense) be reserved for participantoriented adjuncts which are part of the focus domain of a sentence, i.e. which function as focus exponents. Such adjuncts encode a state which contributes a signiﬁcant characteristic to the main event—for example, leaving drunk or outraged is different from simply leaving (cf. also Geuder 2000: 197). In English, they typically occur in post-predicate position and carry the main sentence stress.

26

N. P. Himmelmann and E. Schultze-Berndt

Participant-oriented adjuncts outside the focus domain will be termed circumstantials, following Nichols (1978b). As we have argued in sections 1.2.3 and 1.2.4, this category also comprises Halliday’s conditionals and the adjunct type that has been referred to as weak free adjunct in the semantics literature. Whether the distinction between depictives proper and circumstantials is universally applicable is an empirical question; it is questioned for Warlpiri by Simpson (Ch. 2, this volume). For the time being, we propose the term depictives in the broad sense to comprise both depictives proper and circumstantials. In line with much of the recent literature, most contributors to this volume use the term depictive in this broad sense.10 The term depictive will, moreover, be restricted to those participantoriented adjuncts which can be distinguished from event-oriented adjuncts— adverbials—on morphosyntactic grounds. This terminology allows us to describe cases of mismatch between morphosyntactic and semantic properties: a construction type which generally serves to convey event orientation may still receive a participant-oriented interpretation in some instances. For English, a case in point are the transparent adverbs discussed by Geuder (2000), which are formally distinguished from depictives by the presence of the sufﬁx -ly. As brieﬂy mentioned in section 1.2.2 and further discussed in section 1.3, the reverse type of mismatch can also be found: for some languages we can speak of event-oriented depictives, i.e. constructions which usually convey participant orientation, but which may receive a primarily event-oriented interpretation under certain circumstances. The term general adjunct construction was introduced in section 1.2.2 for constructions which are not morphosyntactically marked for either event- or participantorientation. So far, we have only considered participant-oriented expressions which clearly form a constituent of their own within the matrix clause headed by the main predicate. However, there are further morphosyntactic options for participant-oriented expressions, which are not always easily distinguishable from depictives in the broad sense as just deﬁned. The main types here are loosely adjoined reduced clauses, and incorporated expressions, discussed in turn below. 10 In passing, it may be noted that the currently most widespread usage of depictive is actually somewhat narrower than just deﬁned, since the term is generally used only in reference to adjectival or participial adjuncts exhibiting participant orientation. However, this appears to be based on the fact that the most common and straightforward examples of depictive secondary predicates in English and other European languages involve adjectives and participles. As further discussed in section 1.4, there is no systematic basis for limiting depictive status to these two lexical categories. Some of the contributions to this volume restrict their discussion to depictives of this type, though.

Issues in the syntax and semantics of participant-oriented adjuncts

27

Loosely adjoined reduced clauses which are participant-oriented in that they share an argument with the matrix predicate have also been termed strong free adjuncts. Unlike depictives, they render a logically completely independent proposition. One indication of this independence is the fact that they are usually prosodically detached from the main clause. Furthermore, reduced clauses are always outside the scope of a negator of the main predicate (as in They didn’t leave, outraged), and impose fewer restrictions on the semantic nature of the participant-oriented expression. Thus, in English, participant-oriented reduced clauses, in contrast to depictives, may contain individual-level predicates like that in (28b). There is, however, some evidence that the restriction of depictives to stage-level predicates may not be universal (which reduces the number of criteria distinguishing depictives from reduced clauses/strong free adjuncts). For example, Simpson (Ch. 2, this volume) points out that individual-level predicates could be analysed as depictives in Warlpiri, e.g. when they function as predicates on the objects of perception verbs. In addition to the cases of participant orientation discussed so far, we may note in passing that in the literature, the term (depictive) secondary predicate has occasionally also been applied to participant-oriented expressions which do not have adjunct status, but are part of the main predicate (i.e., the two jointly form a complex predicate). This issue is taken up again brieﬂy in section 1.4.1; see also Schultze-Berndt and Himmelmann (2004: 69–72) for discussion and illustration. Having introduced the most important terminological distinctions, we are now in a position to review the range of semantic domains for which depictive constructions can be used (section 1.3), as well as the morphosyntactic features of depictive constructions found in different languages (section 1.4).

1.3 The semantic range of participant-oriented adjuncts While, as we have seen in section 1.2, the oriented nature of manner adverbials has received some attention in the literature, there is only occasional mention of the participant-oriented nature of e.g. adjuncts of quantity, concomitance, comparison, or location.11 In this section, we will brieﬂy summarize the evidence, from a crosslinguistic perspective, for the claim that 11 Discussions of the Latin praedicativum are a major exception, in that they tend to include a detailed listing of the different semantic domains for which secondary predicates can be used. In fact, in works such as Menge (2000: 354–9) roughly the same kind of semantic domains are mentioned as in the following sections (thanks to Utz Maas for reminding us of this very useful source).

28

N. P. Himmelmann and E. Schultze-Berndt

expressions that have traditionally been assigned to the realms of ‘adverbials’ and ‘depictives’, respectively, in fact form part of the same semantic domain, in that they all share the feature of participant orientation (as well as event orientation), albeit to different degrees. Such evidence, we will argue, comes from the fact that the encoding of these adjuncts, at least in some languages, differs in no way from that of prototypical depictive secondary predicates (such as fresh in example (1) above), which express a physical state or condition that holds of one of the participants of the main predicate at the time of the main event. Such ‘prototypical depictives’, in many languages, are rendered by general adjunct constructions which allow for both participantand event-oriented readings (see section 1.2.6). For languages where participant orientation and event orientation are distinguished by morphosyntactic means, the resulting two construction types—depictives in the broad sense and adverbials—are in competition within the semantic domain outlined here. Building on the hierarchy ranging from prototypical participantoriented expressions to prototypical event-oriented expressions in SchultzeBerndt and Himmelmann (2004: 119–23), but incorporating the more complex ﬁndings presented in the contributions to this volume, we propose in Figure 1.1 a (very tentative) semantic map for this semantic domain.12 The expression types included in the map will be brieﬂy discussed in sections 1.3.1–1.3.14 below. In section 1.3.14, the semantic map will be tested against four of the languages represented in this volume, the Appenzell dialects of Swiss German (Figure 1.2), Georgian (Figure 1.3), Warlpiri (Figure 1.4), and Shipibo-Konibo (Figure 1.5). The semantic map, as presented here, centres around the most prototypical participant-oriented adjuncts, i.e. those encoded as depictives in all languages that have depictive constructions at all. As mentioned above, these are adjuncts conveying a physical condition or state. Since this adjunct type has already been discussed extensively in the preceding sections, it is not included in the following review. For expository reasons, the order of presentation of the semantic categories in the following subsections also does not follow the semantic map in all respects. The arrangement of the other expression types on the semantic map is designed to reﬂect, ﬁrst and foremost, their likelihood of depictive encoding— a semantic domain that is encoded as a depictive adjunct in more of the languages surveyed by us is placed closer to the centre. The dotted lines connecting various expression types should be read as hypothesized paths of 12 For the methodology of semantic maps, see van der Auwera and Malchukov (Ch. 13, this volume), and references therein.

Concomitance (with a dog)

Intention (wanting to be beautiful)

Benefactive/ Malefactive (for/on you)

Resultant state / anterior event (drunk ; having burnt the field)

Association (with a hat) Distributivity (three by three)

Collective (alone ; as a group of three)

Emphatic pronoun (herself )

Mental/emotional condition (sad ; purposefully)

Physical condition (fresh ; green)

Posture/ configuration (flat ; in a circle)

Purpose (in order to train the dog) Function (purpose) (in order to be skinned) Simultaneous event (grinding grain)

Manner (fast ; carefully)

Comparison (like a horse) Location (of participant) (in the canoe ; from the balcony) Location (of event) (in the camp)

Function (Role) (as a teacher)

Life stage (as a child) Atmospheric condition (in the dark)

Order ([as the] first)

Time (in the morning ; yesterday)

Figure 1.1 A tentative semantic map for participant-oriented expressions

Frequency (twice)

30

N. P. Himmelmann and E. Schultze-Berndt

context expansion of depictive marking (alternatively, they can of course also be read as paths of context expansion of adverbial marking encroaching on the core domain of participant orientation, but this possibility will not be further discussed below). The expression types enclosed in a box connected to the centre by straight lines are envisaged as located on a separate plane, i.e. in a third dimension of the semantic map; these are the expression types which are most likely to be encoded by clausal adjuncts, with somewhat different properties from the other adjuncts (see further sections 1.3.12 and 1.3.13). Semantic domains found adjacent to each other in the semantic map are likely to receive the same type of morphosyntactic coding, whether in a depictive construction or in an adverbial or general adjunct construction. An example of the latter would be the construction involving the marker als ‘as’ in German, which covers expressions of life stage (‘as a child’), function and role (‘as a teacher’), and order (‘as the ﬁrst’). However, due to its twodimensional nature, the map cannot always adequately represent semantic closeness of this type, and no attempt has been made to visually represent potential identity of marking other than by spatial contiguity. The major evidence that we draw on in this section is the semantic range of genuine depictive constructions (in the broad sense), which, as the contributions in this volume clearly show, can vary considerably. Depictive constructions usually display some form of agreement between the depictive adjunct and its controller (see section 1.4.5.1). Alternatively, there is some other clear formal marker of the construction such as the dedicated depictive marker in a few Swiss German dialects. Constructions showing agreement, in our view, constitute the paradigm case of participant-oriented adjunct constructions, because the predicative relationship between controller and adjunct is overtly reﬂected in their morphosyntactic structure. Although evidence from agreement is not always unequivocal,13 in all of the examples of relevance here participant orientation provides the most straightforward explanation of the agreement facts. 1.3.1 Mental or emotional condition Expressions of a mental or emotional condition such as sad(ly), angry/angrily, or ravenous(ly), when appearing with adverbial marking, belong to those expressions typically subsumed under ‘oriented adverbials’. The use of this term already suggests that, at least when predicated of the subject, expressions of this type not only convey information about the internal state or condition 13

See Schultze-Berndt and Himmelmann (2004: 82–4) for further discussion.

Issues in the syntax and semantics of participant-oriented adjuncts

31

of a participant, but are also event-oriented in that they almost inevitably make a statement about the manner in which an event is performed, which after all is often the only clue to the internal state of a participant (see section 1.2.2 for a detailed discussion). This distinguishes expressions of a mental or emotional condition from expressions of a physical condition, which are found at the centre of the semantic map in Figure 1.1. The latter are more exclusively participant-oriented because the time-stable nature of the physical state or condition means that it is less likely to be perceived as inﬂuencing the manner of the event. Needless to say, the boundary between physical and mental condition is not always clear-cut: consider e.g. concepts like ‘hungry’. This also holds for the other semantic categories discussed. The prediction made here is that there will be languages which distinguish expressions of physical condition from expressions of mental or emotional condition by encoding the former as depictives and the latter as adverbials. While we have not yet found an example of a language where this distinction in encoding is consistently made, there are a number of languages which behave like English in that expressions of mental or emotional condition, unlike expressions of physical condition, allow for a variation between depictive and adverbial coding, focusing on either the participant-oriented aspect or the event-oriented aspect of the expression; other relevant examples are Georgian (Boeder, Ch. 6, this volume), and Latin (see e.g. Menge 2000: 354f.). Finally, there should be languages where mental or emotional condition is consistently encoded by a depictive construction. Shipibo Konibo (see Valenzuela, Ch. 8, this volume, ex. (64)) and many Australian languages including Warlpiri appear to be languages of this type (cf. Hale 1982; Simpson 1991; for an example, see Simpson, Ch. 2, this volume, ex. (29b)). 1.3.2 Posture and conﬁguration Like expressions of a mental or emotional condition, expressions of posture such as backwards and expressions of conﬁguration such as in a circle describe a property of a participant. At the same time, when the posture is that of the agentive participant, it may alter signiﬁcantly the way in which an event is conceived. Semantically, therefore, these expressions are situated between expressions of a physical condition (the prototypical depictives) and manner expressions, and have been represented in this way in the semantic map in Figure 1.1. The formal evidence available to us so far seems to support this intermediate position, in that expressions of posture and conﬁguration can oscillate between adverbial and depictive coding (in languages where the distinction

32

N. P. Himmelmann and E. Schultze-Berndt

between the two construction types exists).14 Examples are provided by Bucheli Berger (Ch. 4, this volume) and Boeder (Ch. 6, this volume, section 6.4.1); in both of the languages discussed by these authors—the north-eastern dialects of Swiss German, and Georgian, respectively—a subgroup of posture expressions are coded as depictives, while the remaining posture expressions and prototypical manner expressions (e.g. of speed) are coded as adverbials. 1.3.3 Manner Even the most prototypical manner adjuncts usually associated with event orientation, e.g. expressions of speed such as quickly and slowly, are not distinguished from depictives in their morphosyntactic coding in all languages. For example, manner expressions may be unambiguously marked as depictive by agreement with their controller. Relevant examples can be found in Martuthunira (see ex. (20)), Warlpiri (Simpson, Ch. 2, this volume, ex. (20)), Diyari (see ex. (48)), and Shipibo-Konibo (Valenzuela, Ch. 8, this volume). In some Omotic languages, obligatorily controlled converb constructions (see section 1.4.4.5), identiﬁed as genuine depictive constructions by Amha and Dimmendaal (Ch. 9, this volume), are used to encode both physical condition and manner (see in particular their ex. (36b)). In Ewe (Ameka, Ch. 11, this volume), the predicative marker ¼i found on nominal depictives and some expressions of physical condition occurs—albeit in a more strongly grammaticalized form—on expressions of manner as well. Van der Auwera and Malchukov (Ch. 13, this volume) discuss examples from several other languages where manner adverbials and depictives share major formal characteristics, e.g. Basque and the Tungusic language Even. In some languages, e.g. Italian (Napoli 1975), Turkana (Amha and Dimmendaal, Ch. 9, this volume), and Shona (Gu¨ldemann, Ch. 10, this volume), manner is a semantic domain where adverbial and depictive coding alternate with the same expression, or positive and negative versions of the same expression, with often only very subtle semantic differences. A similar variation can be observed between corresponding expressions in closely related languages (for examples see Amha and Dimmendaal, Ch. 9, this volume, and Boeder, Ch. 6, this volume). This further supports the hypothesis that expressions of manner are simultaneously event-oriented and participantoriented, and that one or the other component can be ‘highlighted’ by the morphosyntactic characteristics of the construction involved. 14 In serializing languages, posture is often encoded by a verb in a verb series (see Ameka, Ch. 11, this volume, and Enﬁeld, Ch. 12, this volume, for examples), and thus falls outside the domain of adjuncts as deﬁned here.

Issues in the syntax and semantics of participant-oriented adjuncts

33

1.3.4 Comparison Closely related semantically to expressions of manner are expressions of comparison (similatives) such as like a horse in He eats his food like a horse. From a semantic point of view, adjuncts of this type clearly have a participantoriented component: the similarity to another entity is ascribed to one of the participants (generally the agent; in the example above it is the subject and not the food which is compared to a horse). At the same time, the comparative expression also says something about the manner in which the event (in this case, the eating) is performed, and in this respect it is event-oriented. From a crosslinguistic perspective, we therefore expect a similar variation in coding to other manner adverbials. This expectation is indeed borne out. As Schultze-Berndt and Himmelmann (2004: 112–13) show, in languages where depictive secondary predicates denoting condition or state agree with their controller in case, similative expressions also tend to show agreement. This is illustrated in (32) for the Australian language Gooniyandi. A paraphrase that captures the relationship between the similative adjunct and the agent is ‘the woman ﬁghts being like a man (at the time of ﬁghting)’. Gooniyandi (32) goornboo15 thirri gardboowoona yoowooloo-jangi-ngga woman ﬁght she:belts:them man-simil-erg ‘the woman ﬁghts like a man’ (McGregor 1990: 346) The participant-oriented nature of similative expressions is not always reﬂected formally, though. In Georgian, for example, secondary predicates of condition or state, but not similative expressions, show case agreement with their controller, although the latter still show semantic agreement in number (Boeder, Ch. 6, this volume). As Haspelmath and Buchholz (1998: 322) point out, in many languages expressions of comparison are formally related to expressions of role or function (see section 1.3.5). This is true, for example, for Russian, where both expression types are marked with instrumental case, and for Ewe, where both may take the connector abe´ ‘like’ and/or the predicative marker ¼i (Ameka, Ch. 11, this volume). For this reason, expressions of comparison are placed between manner expressions and expressions of function/role in the semantic map.

15 The NP representing the controller, goornboo ‘woman’, here does not carry the ergative marker, since ergative marking of agents is optional in Gooniyandi (cf. McGregor 1992b; 1998b).

34

N. P. Himmelmann and E. Schultze-Berndt

1.3.5 Function and role A typical example of an adjunct of function or role is as a present in They gave him the book as a present, which semantically clearly relates only to the object (the theme of the transfer), and neither to the subject (the agent) nor to the indirect object (the recipient). Expressions of function/role have therefore been adduced as typical examples of participant-oriented adjuncts (see Nichols 1978a). This analysis ﬁts in with the view expressed by Stump (1985) that in English, the marker as (and presumably its translation equivalents) serves to convert individual-level predicates—which cannot function as depictives—into stage-level predicates, i.e. expressions of a temporary state (see also section 1.2.4 above). Arguably, however, expressions of function or role also have an event-oriented component (e.g. as a present above can also be interpreted as making a statement about the manner of presenting). Thus, in Georgian, where depictives and adverbials are distinguished by the presence vs. absence of agreement, expressions of function and role receive adverbial coding (Boeder, Ch. 6, this volume), and good examples of depictive coding of these expressions are surprisingly hard to come by. In some languages, expressions of function and role carry special predicative markers such as the so-called ‘essive’ case (see section 1.4.5.2). In other languages, they take a copula or auxiliary verb, just as they would as main predicates (cf. Enﬁeld, Ch. 12, this volume, for Lao; Gu¨ldemann, Ch. 10, this volume, for Shona; and Amha and Dimmendaal, Ch. 9, this volume, for Turkana). They may also be marked with an instrumental or locative case or adposition, and/or share their marking with expressions of comparison (see section 1.3.4) or expressions of life stage (see section 1.3.6); this is the case e.g. in English, German, Russian, Finnish, and also in Ewe (Ameka, Ch. 11, this volume). Most of these constructions probably have to be regarded as general adjunct constructions. Although participant-oriented adjuncts of role or function are fairly frequent in English and many other languages, there appear to be languages which cannot express these functions by an adjunct, but which have to resort to biclausal constructions. One of these is Laz (Kutscher and Genc¸, Ch. 7, this volume). 1.3.6 Life stage Adjuncts of life stage such as as a young girl in (24) belong to the standard examples of depictive or circumstantial secondary predicates in the literature (see section 1.2.3). Their participant-oriented nature may be reﬂected in their formal marking, as in Georgian (Boeder, Ch. 6, this volume).

Issues in the syntax and semantics of participant-oriented adjuncts

35

In other languages, however, either the predicative nature (see the remarks in section 1.3.5 above) or the temporal nature of life-stage expressions is coded in preference over their participant-oriented nature; expressions of life stage are therefore placed between prototypical participant-oriented expressions and temporal expressions in the semantic map. Schultze-Berndt and Himmelmann (2004: 121) have found that in many Australian languages with a strict requirement of case agreement of participant-oriented adjuncts with their controller, expressions of life stage receive temporal or locative marking and thus fall outside the domain of (formally marked) participant-oriented adjuncts. The same is true for a subset of life stage expressions in ShipiboKonibo (Valenzuela, Ch. 8, this volume), while others carry the formal markers of participant orientation. In this language, therefore, life stage expressions exhibit variation between depictive and adverbial coding. 1.3.7 Quantiﬁcation and order Numerals and other quantiﬁers in adjunct position also often exhibit a semantic relationship to one participant of the main predicate, i.e. participant orientation. The clearest case can be made for quantiﬁcational adjuncts in a collective reading, which indicate the size of the set of entities involved as a participant in an event. An example is alone in Mark ate the cakes alone again; the adverb here only indicates the quantity of the subject, not of the object. Formally, collective quantiﬁcational adjuncts, in particular numerals expressing the cardinality of a set of participants, are often rendered by a distinctive construction which only applies to these adjuncts and which can arguably be regarded as a depictive construction (see Schultze-Berndt and Himmelmann 2004: 107–9). McGregor (Ch. 5, this volume) provides a detailed survey of constructions involving participant-oriented quantiﬁcational adjuncts in several Australian languages. The characteristic formal property of some of the constructions he discusses is agreement, making these adjuncts clearly depictive. Further examples of agreeing quantiﬁcational adjuncts are provided by Boeder (Ch. 6, this volume) for Georgian. Distributive quantiﬁcational adjuncts, or, in McGregor’s terms, adjuncts expressing iterated co-participation (as in The schoolchildren walked two by two/in pairs) could also be regarded as participant-oriented. From a crosslinguistic perspective, however, there seems to be less evidence for a formal manifestation of participant orientation in this type of expressions, which very frequently consist in reduplicated forms of a numeral (for more discussion, see McGregor, Ch. 5, this volume, and Kutscher and Genc¸, Ch. 7, this volume).

36

N. P. Himmelmann and E. Schultze-Berndt

Ordinal numerals in adjunct function, as in She wanted to arrive home ﬁrst, are also arguably participant-oriented. In some languages, such as German, these expressions share their marking with expressions of function/role and life stage (see sections 1.3.5 and 1.3.6). For Latin, ordinal adjuncts, displaying agreement with their controllers, are presented as a clear case of praedicativa, i.e. depictives (e.g. Menge 2000: 356). The ﬁnal expression type involving quantiﬁers to be considered here consists of expressions of frequency such as twice. While modifying the event as a whole rather than being participant-oriented, and therefore found at the margin of the semantic map in Figure 1.1, in Shipibo-Konibo they receive the type of depictive encoding also found in expressions of manner (Valenzuela, Ch. 8, this volume). 1.3.8 Emphatic pronouns Emphatic pronouns in adjunct position can be illustrated with the example She drove the truck herself. Trivially, there exists a relationship of coreference between the emphatic pronoun herself and the subject of the sentence. Expressions of this type could thus also be regarded as participant-oriented adjuncts. One reason for adopting such an analysis is their close semantic relationship to participant-oriented quantiﬁers (since e.g. herself in the example above can be interpreted as alone); see McGregor (Ch. 5, this volume) for a discussion of this issue. The morphosyntactic facts of Panoan languages provide an additional argument. In Shipibo-Konibo, agreement between the emphatic pronominal and the participant of the main clause does not just manifest itself in shared values for person, number, or case; rather, emphatic pronouns display the type of agreement speciﬁcally restricted to participant-oriented adjuncts (i.e. depictives), and this is therefore the analysis that Valenzuela (Ch. 8, this volume) adopts. If we accept this evidence, emphatic pronouns may well be considered as very close to prototypical participant-oriented adjuncts, which is where they have been placed in the semantic map in Figure 1.1. 1.3.9 Concomitance and association Expressions of concomitance—marked with a comitative or a comparable case or adposition—are generally regarded as adverbials. However, some authors (see in particular Frey and Pittner 1998 and Pittner 1999: 101) have noted that they exhibit a special relationship to one of the participants of the main predicate, in that the latter is said to be accompanied by the referent of the comitative phrase. This can be illustrated with the example They brought

Issues in the syntax and semantics of participant-oriented adjuncts

37

in the prisoner with his accomplice, where the comitative phrase is interpreted with respect to only the object, not the subject. Still, the participant-oriented nature of the adpositional phrase in English is not obvious from its encoding. For this reason, the comitative phrase here is probably more appropriately characterized as a general adjunct construction. Evidence for an analysis of comitative expressions as participant-oriented adjuncts comes from languages where they show agreement with the noun phrase whose referent is the participant that is accompanied. In (33), from Warlpiri, the comitative expression is in addition ergative-marked in agreement with the agent (see Schultze-Berndt and Himmelmann 2004: 110f. for further examples). Warlpiri (33) kurdu ka ngurlu kipi-rni karnta-ngkajinta-rlu child prs mulga.seed winnow-npst woman-com-erg ‘The child is winnowing mulga seed with the woman.’ (Hale 1982: 272) Similarly, in Shona (Gu¨ldemann, Ch. 10, this volume), accompaniment is expressed by a converbal construction in the same way as other participantoriented adjuncts. However, this formal manifestation of participant orientation is not found in all of the languages with dedicated depictive constructions. An example is Georgian (Boeder, Ch. 6, this volume), which has participant-oriented expressions of concomitance which are marked with instrumental case rather than showing agreement. Interestingly, many languages appear to make a formal distinction between expressions of concomitance or accompaniment where the concomitant has a certain degree of independence (i.e. it is typically animate) and expressions where the concomitant is more closely associated with, and less independent of, the controller. A prototypical example of such an associative relation is adornment (e.g. with a hat), although any inanimate entity carried in close physical contact, and sometimes also a person’s children, may fall under this category. Compare the Warlpiri example in (33) with the one in (34); the associative relationship in the latter is indicated by a different marker glossed traditionally as ‘proprietive’. (34) kurdu ka-rla ngarrka-ku rdanpa-rni kuyu-kurlu-ku child prs-3sg.io man-dat accompany-npst meat-prop-dat ‘The child is accompanying [the man]i [with meat]i.’ (Hale 1982: 277) While in Warlpiri, as these examples show, both comitative- and proprietive-marked adjuncts exhibit case agreement with a controller, in

38

N. P. Himmelmann and E. Schultze-Berndt

Shipibo-Konibo participant agreement is restricted to proprietive expressions (which have a similar semantic range to the Warlpiri ones), but in some other Panoan languages it also extends to comitatives (Valenzuela 2003 and Ch. 8, this volume). From a semantic point of view, too, it makes sense to regard associative expressions, where the association between the accompanying and the accompanied entity is more direct, as more strongly participant-oriented than expressions of concomitance. Further evidence comes from the fact that if a distinction between the two construction types is made, it is the marker signalling an associative relationship which is also often involved in the marking of expressions of mental or emotional condition, manner or posture. ‘Association’ is therefore singled out as a separate category in the semantic map in Figure 1.1, and placed in between expressions of physical condition and expressions of concomitance. This constitutes a reﬁnement and, to some extent, revision, of the hierarchy proposed in Schultze-Berndt and Himmelmann (2004: 119–23), where expressions of concomitance are regarded as a single category which is ascribed a relatively high degree of participant orientation. 1.3.10 Location and direction It has occasionally been noted in the literature that locative expressions are not necessarily exclusively event-oriented, i.e. indicate the location of an event as a whole. Rather, they can also have a participant-oriented reading in that they indicate the location of only one of the participants in an event, as in (35) (see e.g. Mu¨ller-Bardey 1990; Maienborn 2000; 2001; Takezawa 1993: 55). (35) The cook prepared the chicken in a Marihuana sauce. (Maienborn 2000: 155) As demonstrated by Maienborn (2000; 2001), these two cases are distinguished by syntactic position in languages like German (see also the discussion in Mu¨ller-Bardey, Ch. 3, this volume), as well as by the fact that participant-oriented locative phrases are substituted by the manner interrogative, not the locative interrogative. It is not surprising, therefore, that participant-oriented locative phrases occur in depictive constructions. In Warlpiri, for example, participantoriented locative phrases show agreement with their controller (see examples (18) and (19b) in Simpson, Ch. 2, this volume). With verbs of ‘emission’, e.g. speaking or throwing, a source (or ‘ablative’) expression can also have a participant-oriented interpretation and indicate the location of the ‘emittor’ or ‘source’, as in The boy threw stones from the

Issues in the syntax and semantics of participant-oriented adjuncts

39

roof. Examples of this type are discussed for Laz by Kutscher and Genc¸ (Ch. 7, this volume) and for Panoan languages by Valenzuela (Ch. 8, this volume). Finally, directional phrases can also be participant-oriented. An example from Walmatjarri, with agreement of the directional phrase, is given in (36). Being directed towards the sandhill is conceptualized here as a temporary condition of the participant who is moving. Walmatjarri (36) jilji-karti-rlu ma-nya wanyjani yapa-warnti nganpayi-rlu sandhill-all-erg aux-3pl.o leave:pst child-pl man-erg ‘The man left the children on his way to the sandhill.’ (Hudson 1978: 35-6) While in the preceding examples the agreement facts appear to have a semantic basis, reﬂecting the participant-oriented nature of the adjunct, there are a number of languages where agreement of locative phrases is generalized in that it occurs in all instances, even if the location is that of the event as a whole (including all participants) rather than just that of a single participant. Typically, the participant triggering agreement in this case is the agentive, and presumably most prominent, participant, as in the Warlpiri example in (37); see also Valenzuela (Ch. 8, this volume) for relevant examples from ShipiboKonibo. Warlpiri (37) ngarrka-ngku man-erg ‘The man is and Bresnan

ka jarnti-rni karli ngurra-ngka-rlu prs trim-npst boomerang camp-loc-erg trimming the boomerang in the camp.’ (Simpson 1983: 57)

Similarly, in the East Georgian mountain dialect Tush, agreement seems to have been generalized to all expressions of a spatial source (see Boeder, Ch. 6, this volume). We consider these cases as instances of overgeneralization of an originally more limited agreement pattern. Consequently, location of a participant and location of an event are distinguished in the semantic map in Figure 1.1, with the former closer to the core than the latter. 1.3.11 Time and atmospheric condition Temporal expressions constitute the perhaps most puzzling case of expressions whose formal make-up, in at least some languages, suggests an analysis as participant-oriented adjuncts—even though temporal expressions are generally considered a paradigm case of event-oriented adverbials. Still, it is

40

N. P. Himmelmann and E. Schultze-Berndt

well known that they may agree with a participant of the main predicate in Old Greek and in Latin (Menge 2000: 356f.; Pinkster 1988). In several Australian languages, too, including Warlpiri and Yankunytjatjara, a subset of temporal nominals show optional case agreement with the subject (A or S argument) of the clause (Hale 1982: 281; Goddard 1985: 256–9; SchultzeBerndt 2006: 203; see also ex. (21) in Simpson, Ch. 2, this volume). As suggested in Schultze-Berndt and Himmelmann (2004: 119), we consider the depictive coding of time expressions as an overgeneralization of a depictive construction beyond its core semantic domain. Depictive encoding is always ‘exceptional’ in the sense that to our knowledge there is no language where all temporal adjuncts regularly agree with another constituent in the matrix clause. Rather, this type of agreement is usually limited to a subset of temporal adjuncts and often optional. Time expressions are therefore included in the semantic map shown in Figure 1.1, albeit at the very margin. Interestingly, Bucheli Berger (Ch. 4, this volume, ex. (47)) reports that the depictive marker in the north-eastern Swiss German dialects is marginally acceptable with expressions of atmospheric circumstance such as ‘dark’ in a sentence such as They went home (in) dark(ness), which can be interpreted as a metonymic reference to a time of the day, and hence as event-oriented. On the other hand, these expressions seem to be more strongly participantoriented than other temporal expressions, since the participant involved may be affected directly by an atmospheric circumstance such as darkness (consider also the discussion of a similar Latin expression in Boeder, Ch. 6, this volume, ex. (126)). We very tentatively propose here that such expressions may provide a link between more prototypical participant-oriented expressions and temporal expressions; they are therefore placed between the two in the semantic map in Figure 1.1. 1.3.12 Resultant state (anterior event) and simultaneous event In Schultze-Berndt and Himmelmann (2004: 98–106), we argue at length that event nominals, converbs or participles, and consequently also clauses headed by these, can have depictive status if they are obligatorily controlled by a participant of the main clause. This means that the subject of the subordinate clause is unexpressed and understood as coreferential with a participant of the main clause. This and the following subsection are devoted to semantic expression types most likely to be encoded as clausal constructions. A particularly clear case for depictive status can be made if these constructions either show agreement with a controller or involve what is generally termed ‘switch-reference marking’, which usually involves at least a distinct marker (‘same-subject marking’) for subject orientation of the clause (see further

Issues in the syntax and semantics of participant-oriented adjuncts

41

sections 1.4.4.5 and 1.4.5.1). Because the latter constitutes a speciﬁc coding property, expressions of this type are represented on a different plane in the semantic map; this issue is taken up again in section 1.3.14. In many languages, including several discussed in this volume, converbs with overt same-subject marking display a contrast in taxis or ‘relative tense’. Wolaitta and Maale (Amha and Dimmendaal, Ch. 9, this volume) distinguish anterior forms (expressing an event that has been completed prior to that of the main predicate) and simultaneous forms (expressing an event simultaneous to that of the main predicate). Shipibo-Konibo (Valenzuela, Ch. 8, this volume), in addition, also has subordinate clauses displaying subsequent marking. The taxis system in Shona is even more complex (Gu¨ldemann, Ch. 10, this volume). We will discuss anterior and simultaneous expressions before turning to subsequent expressions. Anterior converbs, ‘past participles’, or ‘resultative participles’, and clauses headed by these, encode a state resulting from a completed event, and are semantically very close to ‘prototypical’ depictives (compare the participial drunk with the adjective sober). In all the languages with dedicated depictive constructions surveyed by us, expressions of this type in fact receive depictive encoding. Examples can be found in several contributions to this volume, including Amha and Dimmendaal (Ch. 9), where the relevant examples are those for anterior same-subject converbs in Wolaitta and Maale. For further examples and discussion, see the semantic maps and references for the north-eastern Swiss German dialects, Georgian, Warlpiri, and Shipibo-Konibo in section 1.3.14, as well as Schultze-Berndt and Himmelmann (2004: 103–6). The following Warlpiri example illustrates agreement of the simultaneous converb with the controller in ergative case. Simultaneous event expressions straightforwardly fulﬁl the criterion of temporal overlap between a depictive and a main clause: the participant of the main clause (here, the man) is said to be (in the state of) running at the same time as being involved in the event encoded by the main clause (seeing a kangaroo). Warlpiri (38) wati-ngki marlu nya-ngu parnka-nja-karra-rlu man-erg kangaroo see-pst run-cvb-ss-erg ‘The man saw the kangaroo while [the man was] running.’ (Hale et al. 1995: 1442) Other examples of simultaneous same-subject clauses with depictive coding can be found in the contributions by Valenzuela (Ch. 8), Amha and Dimmendaal (Ch. 9), and Gu¨ldemann (Ch. 10).

42

N. P. Himmelmann and E. Schultze-Berndt

1.3.13 Subsequent event It appears that a subsequent event is less readily attributed to a participant of the main clause (in terms of formal correlates of participant orientation) than an anterior or simultaneous event, judging from the ﬁnding that dedicated marking of participant orientation is more frequently found on anterior and simultaneous expressions than on subsequent expressions. In a subsection entitled ‘Prospective deverbal depictives’, Schultze-Berndt and Himmelmann (2004: 104–6) demonstrate that subsequent expressions, at least in the Australian languages surveyed for that paper, are more likely to receive depictive coding if they speciﬁcally convey an intention, rather than subsequent taxis in general. In other words, it is the intention as a pre-state of the event, not the event itself, that is predicated of the controller. A specialized converbal form of this type, the preparatory purposive form in Warlpiri, is illustrated in example (10d) in Simpson (Ch. 2, this volume). In the semantic map in Figure 1.1, it is proposed (in the form of a link in the semantic map) that depictive coding may spread from expressions of intention to purposive or subsequent expressions in general. An example of a language with generalized agreement marking on subsequent expressions appears to be Shipibo-Konibo (Valenzuela, Ch. 8, this volume; see also Figure 1.5 below). A third type of subsequent event, conceptualized as a pre-state and attributed to a participant, has to be identiﬁed in order to account for the agreement facts of Latin and Georgian (see Boeder, Ch. 6, this volume, ex. (33)). These languages exhibit ‘gerundive’ forms conveying a purpose in the sense of an intended function, as in The bear lay there to be skinned. A comparable example is (39). Martuthunira (39) murna-ngka-nguru warra, ngayu thathu-lalha mulurru close-loc-abl emph 1sg.nom send-pst straight [thanturri-waa yanti-ngka waruul] go.down-purp.o side-loc still ‘From close up, I sent it [the spear] straight [for it] to go down into [the emu’s] side.’ (Dench 1995: 264, l. 7) It is hardly possible to ascribe an intention to the bear in the example above, or the spear in the Martuthunira example in (39). Rather, it is the intended function that is ascribed to the participant, and that explains the formal marking of participant orientation on subsequent clauses of this type.

Issues in the syntax and semantics of participant-oriented adjuncts

43

1.3.14 Summarizing the evidence for the semantic map Throughout this section (and in fact throughout this volume), it has become evident that many adjunct types that are widely analysed as adverbials and hence, at least implicitly, as event-oriented, have a participant-oriented semantic component which may or may not be formally reﬂected in a given language. It has been argued that at a semantic level, participant orientation can be regarded as a feature of adjuncts conveying e.g. a mental or emotional condition, an associative relationship, a quantity, or a life stage, but also of expressions of manner, function or role, and concomitance, and of a subset of locative expressions. As argued in section 1.3.8 above, emphatic pronouns are also good candidates for participant-oriented adjunct status. In the semantic map introduced in Figure 1.1, the expression types just listed are arranged according to their postulated degree of participant orientation. At the margin of the map are found, furthermore, semantic categories such as temporal and frequency expressions which are only rarely encoded as depictives (but rather as adverbials or general adjunct constructions), but which nevertheless should be considered as part of the same semantic domain because of the morphosyntactic facts in some languages.16 Finally, adjuncts representing an event as anterior, simultaneous or subsequent with respect to the event encoded by the main predicate are represented on a separate plane from the other expressions in Figure 1.1. This is because the type of depictive marking involved may have a different origin from that employed on other adjuncts, i.e. in switch-reference (same-subject) morphology (see further sections 1.4.4.5 and 1.4.5.1). As both Valenzuela (Ch. 8, this volume) and Gu¨ldemann (Ch. 10, this volume) argue, same-subject morphology may constitute the main source for depictive coding in a language and may in fact spread to other, intraclausal adjuncts. In other words, formal marking of participant orientation may arise on these constructions independently, without any implicational relationship to depictive marking even on the core semantic category (physical condition) in the lower plane of the map—the constructions in question are represented as a ‘second core type’ of depictive marking, as it were, on a different level. The links between these and other expression types that are nevertheless present in the map can, in this case, be read in at least three ways (and note that no exhaustive representation has been aimed at). First, depictive marking of the switch-reference type as typically found on the second plane may spread to constructions on the ﬁrst 16 Note also that only the positive version of the various expression types is represented in the map, even though the negative counterparts (e.g. ‘without’ vs. ‘with’ in the domain of concomitance and association) may exhibit interesting differences in behaviour (see section 1.4.1), and would therefore have to be included in a more ﬁne-grained version of the map.

44

N. P. Himmelmann and E. Schultze-Berndt

plane—at least this is what Valenzuela argues for Panoan languages. Second, clausal constructions involving switch-reference marking may in fact be the major strategy in a language to encode expression types on the lower plane such as location or concomitance (the Shona case, according to Gu¨ldemann). Finally, if in a language a depictive construction involves e.g. case agreement, this marker may be found both on expressions on the lower plane and on the more clausal expressions represented on the upper plane, regardless of whether the latter carry additional markers of restricted reference (e.g. same-subject marking), as in Warlpiri (cf. Hale 1994), or not, as in Georgian (see Boeder, Ch. 6, this volume). Figures 1.2–1.5 illustrate a few test cases for the semantic map outlined in this and the preceding subsections. They represent the ﬁndings for those four languages discussed in some detail in this volume which exhibit genuine depictive constructions. These are the Appenzell dialects of Swiss German (Figure 1.2), Georgian (Figure 1.3), Warlpiri (Figure 1.4), and Shipibo-Konibo (Figure 1.5). The prediction tested for these languages is that the semantic range corresponding to the depictive construction is a subset of the domain of participant-oriented adjuncts included in the semantic map, and that this semantic range is represented by a contiguous segment of the map, including the centre (expressions of physical condition). In the maps below, semantic expression types which, when occurring in adjunct position, are encoded as depictives in the language in question are shaded in grey. Partial shading indicates that only a subset of the relevant expression types receives depictive encoding. Hatching represents variation between depictive and adverbial or general adjunct coding. Absence of information on a particular expression type is indicated by a question mark next to that segment in the map. Figure 1.2 represents the expression types on which the dedicated depictive marker is found in the north-eastern (Appenzell) dialects of Swiss German (Bucheli Berger, Ch. 4, this volume). Depictive marking is basically restricted to the core area of the map—adjectives encoding a physical condition, and anterior-resultative participles (with adjectival morphology). The only extensions beyond that core domain are a subset of posture expressions, and, for some speakers, expressions of atmospheric condition such as ‘dark’ (see also section 1.3.11). In Georgian, depictive constructions can be clearly distinguished from adverbial or general adjunct constructions by the presence of case agreement (and sometimes number agreement). Agreement is found on depictive adjectives, participles, numerals, and nouns in the genitive and instrumental case. The range of expression types displaying agreement is shown in Figure 1.3,

Concomitance (with a dog)

Association (with a hat) Distributivity (three by three)

Collective (alone ; as a group of three)

Emphatic pronoun (herself )

Intention (wanting to be beautiful)

Benefactive/ Malefactive (for/on you)

?

Resultant state / anterior event (drunk ; having burnt the field)

Mental/emotional condition (sad ; purposefully)

Physical condition (fresh ; green)

Posture/ configuration (flat ; in a circle)

Purpose (in order to train the dog) Function (purpose) (in order to be skinned) Simultaneous event (grinding grain)

Manner (fast ; carefully)

Frequency (twice)

Comparison (like a horse) Location (of participant) (in the canoe ; from the balcony) Location (of event) (in the camp)

Life stage (as a child) Atmospheric condition (in the dark)

Function (Role) (as a teacher) Order ([as the] first)

Time (in the morning ; yesterday)

Figure 1.2 Semantic range of depictive marking in the north-eastern (Appenzell) dialects of Swiss German (Bucheli Berger, Ch. 4, this volume)

46

N. P. Himmelmann and E. Schultze-Berndt

following Boeder (Ch. 6, this volume). Not only expressions of physical condition (often encoded by means of anterior participles), but also expressions of life stage and collective quantiﬁcation, and a subset of posture expressions and expressions of mental or emotional condition, regularly involve depictive agreement. Variation between depictive and adverbial or general adjunct encoding is found in the domains of function/purpose (expressed by means of ‘future’ or ‘gerundive’ participles) and association. The hatched section in the ‘physical condition’ segment represents the encroachment of a general adjunct marker, the adverbialis case, into the core domain of participant orientation. Finally, agreement on locative constructions is only found in the Tush dialect of Georgian, and is restricted to source expressions. Figure 1.4 shows the semantic range of case agreement on adjuncts in Warlpiri, based on a number of sources including Hale (1982; 1994), Hale et al. (1995), Simpson (1991; Ch. 2, this volume; p.c.). In Walpiri, the semantic range covered by case agreement—interpreted by us as a depictive marker— goes far beyond that seen for Georgian, in that it also includes expressions of manner, association, and participant-oriented location. Furthermore, case agreement is also found in clausal expressions which take specialized ‘complementizer’ or converb markers, and encode simultaneous events, anterior events, and intention or preparation with respect to a subsequent event. Variation between the use of agreeing and non-agreeing expressions has been described for event-oriented location, time, and concomitance. The shaded section of the ‘frequency’ area tentatively represents agreement with the subject on the adverb tarnnga ‘always’ (see example (10e) in Simpson, Ch. 2, this volume). Collective meanings (‘alone’, ‘as a group’) are expressed by preverbs, i.e. as part of a complex predicate, rather than by adjuncts, so the absence of agreement on these expressions cannot be regarded as counterevidence to the semantic map proposed here. As Figure 1.5 shows, the Panoan language Shipibo-Konibo, as described by Valenzuela (2003; Ch. 8, this volume), is another case of a language where formal marking of participant orientation is found in a very large, and sometimes surprising, semantic range, extending to expression types at the margin of the map such as location of a participant, distributive quantiﬁcation, subsequent and simultaneous event, manner, comparison, and even frequency. Expressions of life stage exhibit variation between depictive and adverbial (temporal) coding (see section 1.3.6). In Shipibo-Konibo, participant agreement is also (marginally) attested on an expression type not discussed so far, termed ‘interessive expressions’ by Valenzuela, which comprises both benefactive and malefactive adjuncts. The motivation for depictive coding of expressions of this type is not

Concomitance (with a dog)

Benefactive/ Malefactive (for/on you)

Resultant state / anterior event (drunk ; having burnt the field)

Association (with a hat) Distributivity (three by three)

Collective (alone ; as a group of three)

?

Emphatic pronoun (herself )

Intention (wanting to be beautiful)

Mental/emotional condition (sad ; purposefully)

Physical condition (fresh ; green)

Posture/ configuration (flat ; in a circle)

Purpose (in order to train the dog) Function (purpose) (in order to be skinned) Simultaneous event (grinding grain)

Manner (fast ; carefully)

Comparison (like a horse) [Source expressions in Tush]

Location (of event) (in the camp)

Location (of participant) (in the canoe ; from the balcony)

Life stage (as a child) Atmospheric condition (in the dark)

Function (Role) (as a teacher) Order ([as the] first)

Time (in the morning ; yesterday)

Figure 1.3 Semantic range of depictive agreement in Georgian (Boeder, Ch. 6, this volume)

Frequency (twice)

Concomitance (with a dog)

Benefactive/ Malefactive (for/on you)

Resultant state / anterior event (drunk ; having burnt the field)

Association (with a hat) Distributivity (three by three)

Collective (alone ; as a group of three)

Emphatic pronoun (herself )

Intention (wanting to be beautiful)

Mental/emotional condition (sad ; purposefully)

Physical condition (fresh ; green)

Posture/ configuration (flat ; in a circle)

Purpose (in order to train the dog) Function (purpose) (in order to be skinned) Simultaneous event (grinding grain)

Manner (fast ; carefully)

Frequency (twice)

Comparison (like a horse) Location (of participant) (in the canoe ; from the balcony) Location (of event) (in the camp)

Life stage (as a child) Atmospheric condition (in the dark)

Function (Role) (as a teacher) Order ([as the] first)

Time (in the morning; yesterday)

Figure 1.4 Semantic range of case agreement on adjuncts in Warlpiri (Hale 1982; Simpson 1991 and Ch. 2, this volume)

Concomitance (with a dog)

Benefactive/ Malefactive (for/on you)

Collective (alone ; as a group of three)

Emphatic pronoun (herself )

Intention (wanting to be beautiful)

? Function (purpose) (in order to be skinned)

Resultant state / anterior event (drunk ; having burnt the field)

Association (with a hat) Distributivity (three by three)

Purpose (in order to train the dog)

Simultaneous event (grinding grain)

Mental/emotional condition (sad ; purposefully)

Physical condition (fresh ; green)

Manner (fast ; carefully)

Posture/ configuration (flat ; in a circle)

Comparison (like a horse) Location (of participant) (in the canoe ; from the balcony) Location (of event) (in the camp)

Life stage (as a child) Atmospheric condition (in the dark)

Function (Role) (as a teacher) Order ([as the] first)

? ?

Time (in the morning ; yesterday)

Figure 1.5 Semantic range of participant agreement in Shipibo-Konibo (Valenzuela, Ch. 8, this volume)

Frequency (twice)

50

N. P. Himmelmann and E. Schultze-Berndt

completely clear, but a possible link, indicated in the semantic map, connects ‘interessives’ to expressions of mental or emotional condition, since a benefactive expression can be interpreted as the intention of the agentive participant to beneﬁt the referent of the adjunct expression. Still, this potential motivation cannot be extended to the malefactive use of these expressions. The semantic map proposed here is compatible with the hierarchy that Valenzuela (Ch. 8, this volume) proposes, based on purely language-internal criteria, and according to which emphatic pronouns and life-stage expressions exhibit the greatest degree of participant orientation, and expressions of time and concomitance (which do not receive participant agreement marking) exhibit the greatest degree of event orientation, with expressions of location, manner, quantiﬁcation, and clausal expressions located in between the two poles. The semantic map is also supported by ﬁndings for other languages, such as Latin, or the Australian languages Martuthunira and Yankunytjatjara, which were included in our survey but are not discussed in detail in this volume. Needless to say, the map and the predictions associated with it are still tentative and preliminary in that they remain to be tested on many more genetically and areally diverse languages. In particular, the proposed paths of context expansion of depictive marking would have to be corroborated by historical and comparative research (of the type undertaken by Valenzuela 2003 for Panoan languages).

1.4 Parameters for a morphosyntactic typology of participant-oriented adjuncts In this section, we will discuss ﬁve parameters that play a role in a morphosyntactic typology of participant-oriented adjuncts. These are restrictions on the combination of oriented adjuncts and main predicates (section 1.4.1), restrictions on the syntactic function or semantic role of the participant that functions as the ‘controller’ (section 1.4.2), restrictions on the syntactic position of the adjunct (section 1.4.3), the word class and internal structure of the adjunct (section 1.4.4), and its morphological marking (section 1.4.5). As we have seen in section 1.3, languages may have many subtypes of participant-oriented adjuncts, and the parameters do not necessarily align for each subtype. Therefore, establishing a comprehensive typology of participantoriented adjuncts is a huge task which in a volume like this can hardly be begun. Furthermore, in a functionally oriented typology, participant-oriented elements other than secondary predicate and adverbial constructions should also be considered. For example, in many languages the participant-oriented element and the main predicate together form a single complex predicate (see

Issues in the syntax and semantics of participant-oriented adjuncts

51

further section 1.4.1). A further alternative is biclausal constructions (see Schultze-Berndt and Himmelmann 2004: 67–9). Thus, in Laz (Kutscher and Genc¸, Ch. 7, this volume) expressions of function/role are usually not adjuncts, but separate main clauses. In this programmatic survey, we will restrict ourselves to participant-oriented adjuncts, which, however, as already noted, include some types of subordinate clause. 1.4.1 Restrictions on the combination of oriented adjuncts and main predicates Hardly any crosslinguistic research has been undertaken so far with regard to restrictions on the classes of main predicates that occur in constructions with a participant-oriented adjunct. For English, Rapoport (1993: 178) suggests that depictive secondary predicate constructions ‘are restricted to verbs that cause a change in the state or location of their objects’. Rapoport (1999) claims that depictives with object controllers are not possible with activity verbs in English, but Warlpiri appears to be less restrictive (see Simpson, Ch. 2, this volume, for further discussion and examples). Bucheli Berger (Ch. 4, this volume) provides a good overview of possible predicate classes of depictive and main predicate in another rather ‘restrictive’ language, a north-eastern dialect of Swiss German. A number of authors, without stating absolute restrictions on the classes of predicates involved, note certain collocational restrictions between depictives and main predicates occurring with depictives (see e.g. Enﬁeld, Ch. 12, this volume). Nichols (1978a) treats such lexically ﬁxed combinations as a distinct subtype of depictive constructions (‘bound co-predicates’; cf. also Simpson, Ch. 2, this volume). Grammatical correlates of a conventionalized combination of depictive and main predicate have been noted by Kuno and Takami (1993: 130–4). They show that restrictions on Heavy NP Shift with depictives in English, which had previously been attributed to syntactic differences between subject and object depictives, depend on the predictability of the combination of verb and depictive, as illustrated in (40). Heavy NP Shift of the object NP leaves the verb and the depictive in immediate syntactic contiguity, thus making them more like complex predicates, which is only warranted if the combination encodes a conventional event (leaving a party sober as opposed to leaving a party angry). (40) a. ?John left angry [the reception for the ambassador from Ulan Bator]. (Kuno and Takami 1993: 131) b. Mike never leaves sober [parties that he goes to]. (Kuno and Takami 1993: 132)

52

N. P. Himmelmann and E. Schultze-Berndt

The tendency for the combination of depictive and main predicate to form ﬁxed collocations is also evident in the formal and diachronic relationship between depictive constructions and periphrastic predicates or copular constructions, where the former main predicate takes on an auxiliary or copular function and the former depictive becomes the main predicate (see e.g. Paul 1919: 52; Haspelmath 1995: 43; Nedjalkov 1995: 99–100; Johanson 1995; Hengeveld 1992a: 237–49). Obviously, when the ﬁrst verb is semantically general but not completely devoid of semantic content, as e.g. the positional verb in The man lay dead in his home for three days, it becomes difﬁcult to distinguish between copular and depictive construction. Some further examples of this type are discussed by Boeder (Ch. 6, this volume). The conventionalized nature of certain combinations may also account for the fact that in many languages, translation equivalents of English depictive constructions are complex predicates. For example, the participant-oriented element may be an incorporated noun or adjective, a construction type not attested in English or in the languages discussed in this book (see SchultzeBerndt and Himmelmann 2004: 69–72 for some examples and discussion). A second possibility is the occurrence of a participant-oriented expression as a preverbal element; an example from Warlpiri can be found in the contribution by Simpson (Ch. 2, this volume, ex. (13)). Alternatively, the translation equivalent of a depictive in English may appear as one of the verb phrases in a serial verb construction, as shown for Ewe by Ameka (Ch. 11, this volume), for Lao by Enﬁeld (Ch. 12, this volume), and for Mandarin Chinese by van der Auwera and Malchukov (Ch. 13, this volume, exx. (25) and (26)). Here, as with copular constructions, the distinction between complex predicates and adjunct constructions may become blurred in some instances, in particular if the language is of the isolating type with no distinction between ﬁnite and nonﬁnite verb forms. Arguments that can be adduced for a more adjunct-like status of one of the verb phrases in a series are its optionality (Enﬁeld, Ch. 12, this volume) and its formal marking (Ameka, Ch. 11, this volume). Notwithstanding the difﬁculties of distinguishing complex predicates and participant-oriented adjunct constructions in some cases, crosslinguistically speaking the prediction is that the more expected and conventionalized an event is, the more likely it is to be expressed by a complex predicate or an incorporating structure, where the formal expression iconically reﬂects conceptual closeness. In the extreme case, a highly conventionalized complex event may be expressed by a single predicate. Amha and Dimmendaal (Ch. 9, this volume) brieﬂy discuss the case of highly speciﬁc verbs derived from ideophones in both Nilotic and

Issues in the syntax and semantics of participant-oriented adjuncts

53

Omotic languages, which, they argue, ‘preempt the need for other strategies rendering a similar content, e.g. the need for a separate manner adverb or secondary predicate’. In Ilokano, the need for a separate participantoriented adjunct may also be circumvented, due to the highly ﬂexible and productive voice morphology which is characteristic of Philippine languages. In the following example, the main predicate (kinilaw) semantically corresponds to a depictive in other languages. Ilokano (41) kinilaw¼da ti sida rls.pv:raw¼3pl.poss art ﬁsh ‘They ate the ﬁsh raw.’ (lit.: ‘They ‘‘rawed’’ the ﬁsh.’) (C. Rubino, p.c.) Note that this strategy is available only for a rather small number of predicates, i.e. those where a conventionalized, culturally well-established practice is being referred to in an abbreviated way and recovered by pragmatics, as it were: the most likely interpretation of ‘raw’ as a main predicate in (41) is as ‘eat raw’. Constructions involving participant-oriented adjuncts, on the other hand, are expected to express more remarkable, unexpected events. This is explicitly commented on by Schroeder (2003) with reference to Turkish and by Gu¨ldemann (Ch. 10, this volume) in his discussion of participant-oriented adjuncts in Shona. Often, the unexpected state of affairs has a negative form. In German, for example, adjectives with the negative preﬁx un- are frequently found as depictives, and sometimes are actually restricted to this function, while their positive counterparts are not found in adjunct position. Bucheli Berger (Ch. 4, this volume) discusses the form ooggesse ‘uneaten’ in a northeastern dialect of Swiss German. This form, for pragmatic reasons, can take on an agent-controlled interpretation (i.e. ‘not having eaten’). A counterpart from standard German is ungefru¨hstu¨ckt ‘not having had breakfast’, which likewise only appears in depictive function, whereas its positive counterpart cannot be used as an adjunct at all. On the other hand, some examples also point to the tendency for negative states of affairs to be encoded as adverbials, even when their positive counterparts are encoded as depictives. This seems to hold for some manner expressions in Nunggubuyu (see van der Auwera and Malchukov, Ch. 13, this volume, ex. (46)). Boeder (Ch. 6, this volume, section 6.5) also discusses some minimal pairs of this type in Georgian; he in fact concludes that negative properties are less likely to be conceived of as qualities attributable to a participant.

54

N. P. Himmelmann and E. Schultze-Berndt

1.4.2 Restrictions on the controller A second parameter in the typology of participant-oriented adjuncts concerns restrictions on the syntactic role of the controller (i.e. the participant to which the adjunct relates). For depictive secondary predicates, it is generally assumed that they can be controlled by subjects and objects or, more precisely, by the single core argument of an intransitive predicate (S), the actorlike core argument (A) of a transitive predicate, and the undergoer-like argument (O) of a transitive predicate. It is widely agreed that controllers in English appear to be more or less restricted to subject and object functions (but see below). This ﬁnding has sometimes been generalized, and it has been claimed that there are general restrictions either on the syntactic function or on the thematic role of potential controllers, e.g. to theme, agent, and patient (see Williams 1980; McNulty 1988; for counter-arguments to these positions, see Bayer 1997: 210–24 and Mu¨ller 2002: 180–9). However, languages differ considerably with regard to which participants are easily accessible as controllers. Simpson (Ch. 2, this volume) contrasts English and Warlpiri in this respect: in Warlpiri, but not in English, datives and other oblique objects may function as controllers. Examples of depictives controlled by dative-marked indirect objects can also be found in Georgian (Boeder, Ch. 6, this volume). Oblique agents in passive clauses, too, are frequently found as controllers, e.g. in Australian languages like Martuthunira (Dench 1995; for examples see Schultze-Berndt and Himmelmann 2004: 73, and McGregor, Ch. 5, this volume). Finnish also regularly allows oblique controllers (Nichols 1978a: 120–1). Furthermore, one has to be careful in making generalizing statements even for a single language because there appear to be considerable differences with regard to acceptability and usage depending on speaker, context, genre, and medium (spoken vs. written usage). For example, Nichols (1982) reports that the acceptability of depictives with object controllers in Russian varies dramatically depending on factors including the case and topicality of the controller, word order, the presence of other potential controllers, and the involvement of the controller in more than one control relationship, but also on individual preferences. German, just like English, is generally rather restrictive in allowing mainly subject and object controllers (as is also demonstrated for Swiss German dialects by Bucheli Berger, Ch. 4, this volume). Still, in our corpus of spontaneous spoken German we also found a number of examples with oblique controllers which would not be judged acceptable in written German. An example is (42), where the depictive clearly predicates on Tasse ‘cup’, the

Issues in the syntax and semantics of participant-oriented adjuncts

55

complement of the preposition aus. Similar examples are cited by Paul (1919: 49–57), Plank (1985: 175), and Schultze-Berndt and Himmelmann (2004: 74). German (42) und sie haben dann am Nachmittag and they. nom aux then in:the:dat afternoon [aus [derselben Tassei]]PP unausgespu¨lti den from the.same:dat cup unrinsed the:acc Kaffee getrunken coffee drunk ‘and in the afternoon they then drank their coffee from the same cup unrinsed.’ [overheard utterance] Similarly, the following English utterance, where the controller of the depictive unlocked is embedded in a PP, is attested, although out of context it is judged unacceptable by most speakers. (43) she came in [through [the back doori]]PP unlockedi [overheard utterance] Thus, it may well be the case that medium and text types (communicative genres) within a given language vary with regard to restrictiveness in the choice of the controller, with less planned and more contextualized types such as everyday conversations being less restrictive in this regard. 1.4.3 Syntactic position/phrase structure properties It would appear obvious that participant-oriented adjuncts differ in their positional characteristics across languages, depending on the general linguistic type of the language in question (conﬁgurational vs. non-conﬁgurational, VO vs. OV, etc.). Amha and Dimmendaal (Ch. 9, this volume) provide an illustrative case study with data from Nilotic and Omotic languages. In terms of surface structure, crosslinguistic positional differences pertain not only to broad topological parameters (e.g. preverbal vs. postverbal position) but also to positional variability. As adjuncts, it is to be expected that participant-oriented adjuncts show at least some positional variability (which in some languages would set them apart from arguments which are positionally more restricted), but languages clearly differ as to how much positional variability they allow for. Of major import in this regard is the question of whether or not different positions correlate with different meanings and constructions, and in particular with focus assignment.

56

N. P. Himmelmann and E. Schultze-Berndt

While the preceding observations should be largely uncontroversial, the syntactic position of participant-oriented adjuncts is in fact one of the most problematic issues in a typology of depictives and adverbials simply because the precise position in the hierarchical syntactic structure is still a matter of debate even for better-analysed languages (see e.g Winkler 1997: 17–91; Mu¨ller 2002: 173–207 for recent surveys of some of the positions proposed for depictives proper in English and German). While it has been argued by some authors that depictives—in particular, subject-oriented depictives—are clause-level adjuncts (see e.g. Williams 1980; Rothstein 1985; Nakajima 1990), most authors follow the arguments of Andrews (1982) in assuming that depictives are adjoined somewhere within the VP. Attachment to different levels of the VP is used to account for the distinction between manner adverbs and depictives (see section 1.2.2), and also for the different subtypes of depictives such as depictives proper and circumstantials (see section 1.2.3), and subject- vs. object-oriented depictives (see section 1.4.2). For further discussion and references, see Simpson (Ch. 2, this volume), and Mu¨ller-Bardey (Ch. 3, this volume). Note that, in line with the argument in section 1.2.3, depictives proper (as distinct from circumstantials) by deﬁnition are restricted to focus position, which more often than not will also be a verb-phrase internal position (see also Kutscher and Genc¸, Ch. 7, this volume; Gu¨ldemann, Ch. 10, this volume). Thus, the possibility for making a principled distinction between depictives proper and circumstantials rests on the availability of a reasonably circumscribed focus position. In a language like Warlpiri, with a considerably greater freedom in word order and very little surface evidence for a VP, the distinctions relying on constituency become blurred (Simpson, Ch. 2, this volume). Crosslinguistically, therefore, the syntactic position of participantoriented adjuncts of different types appears to be much more variable than is usually assumed, and an investigation of this topic has only barely begun. The problems regarding the precise phrase structural position of participantoriented adjuncts are but one aspect in unravelling their complex syntax. Another major challenge is to model the ‘dual’ role of these constituents as participant-oriented ‘predicates’ and as verbal adjuncts. The predicative nature cannot be directly captured in models relying exclusively on constituency, except by positing a small clause analysis (see Winkler 1997: 18–50 for an overview of the debate on this controversial analysis). Other suggestions for capturing the particular nature of depictives, depending on the model assumed, include positing an irreducible syntactic relationship of predication between the depictive and its controller (see Rothstein 2001 for an overview and references), or a double relation of dependency (e.g. Nichols 1978a: 120; McGregor 1997c: 171–3, and Ch. 5, this volume). A further

Issues in the syntax and semantics of participant-oriented adjuncts

57

alternative is to limit the syntactic analysis of depictives to their syntactic role as adjuncts, as is standard practice for other participant-oriented adjuncts, and to relegate all aspects of participant orientation to semantics. In fact a number of authors, including Plank (1985: 183), Steube (1994), and Du¨rscheid (2002: 70–1), have argued for just this approach to depictives in German. But while German depictives are morphologically unmarked and thus favour such an account, in those instances where participant orientation has morphological repercussions such as agreement with the controller, a ‘semanticsonly’ approach obviously faces the challenge of providing a convincing account of the morphological facts. For further discussion and references, see also Schultze-Berndt and Himmelmann (2004: 74–7). 1.4.4 Word class and internal structure The fourth parameter in a typology of participant-oriented adjuncts concerns the word class and internal structure of the adjunct. Of course, the constructions to be included here depend on the semantic range of adjuncts that one considers. Work on depictive secondary predicates has largely concentrated on adjectival secondary predicates, but, as shown in section 1.3, many types of adverb, adpositional phrase, or case-marked noun phrase, and other complex ‘adverbials’, including subordinate clauses, may also be considered participant-oriented. 1.4.4.1 Adjectives Simple adjectives and adjective phrases are the most widely recognized instances of participant-oriented adjuncts; they are amply illustrated throughout this chapter and in the contributions to this volume. A number of contributors, including Bucheli Berger, Simpson, Mu¨llerBardey, McGregor, and van der Auwera and Malchukov, mainly restrict their discussion to adjuncts of this type. However, in not all languages do adjectival depictives constitute the most ‘basic’ type of depictives. Obviously, languages lacking adjectives, or possessing only a small number of adjectives, have to rely on alternative means of forming participant-oriented adjuncts, which may include adverbs, ideophones, deverbal forms, and subordinate clauses. In fact, the use of adjectives as participant-oriented adjuncts seems to presuppose the possibility of employing adjectives as main predicates—Ewe, for example, lacks both (Ameka, Ch. 11, this volume). 1.4.4.2. Nouns In languages where nouns generally are individual-level predicates, i.e. express time-stable characteristics,17 participant-oriented 17 But see Ogawa (2001), who proposes to apply the stage-level vs. individual-level distinction also to nominal predicates.

58

N. P. Himmelmann and E. Schultze-Berndt

adjuncts tend to be adjectival or verbal. If a noun appears in adjunct position in these languages, it is in a construction with a dedicated predicative marker or a copula (see section 1.4.5.2), or in a construction with a special case marker or adposition. In these cases, the construction as a whole serves to convert the nominal into a stage-level predicate (cf. also section 1.2.3). In languages like Warlpiri, on the other hand, with no clear distinction between nouns and adjectives, or stage-level and individual-level predicates, nominals are much more freely employed as participant-oriented adjuncts (see Simpson, Ch. 2, this volume). 1.4.4.3 Adverbs and related parts of speech A part of speech restricted to adjunct function is generally classiﬁed as ‘adverb’. Adverbs may be participantoriented, event-oriented, or both, as discussed in detail in section 1.2. Examples of participant-oriented adverbs in English are the oriented manner adverbs in -ly discussed in section 1.2.2, and unmarked quantifying adverbs such as alone (see section 1.3.7). If a class of ‘adverbs’ always conveys participant orientation, it may be justiﬁed to speak of a word class restricted to functioning as a depictive constituent; this special case will be discussed in 1.4.4.4. below). A class of words which may be used as adjuncts, but also as main predicates (with or without a copula), but not as attributes, are often subsumed under adverbs, although traditionally, the term ‘predicative adjectives’ has also been used. Typically, these exhibit participant orientation. An example from German is barfuß ‘barefoot’ and its dialectal variants, which can be used as a main predicate (e.g. sie war barfuß ‘she was barefoot’) or as an adjunct, as in (44), but not as a modiﬁer (*die barfuße Frau ‘the barefoot woman’). German (Cologne dialect) (44) un alle Kinder durfte fro¨her bla¨ckfo¨ß loufe, and all children may:pst.3pl formerly barefoot walk nur ich nit only 1sg not ‘And in those days all children were allowed to walk barefoot, except for me.’ (Bhatt and Lindlar 1998: 173) In a number of languages, a special word class of (often sound-symbolic) uninﬂecting words, variously termed ‘ideophones’, ‘expressives’, or ‘soundsymbolic adverbs’, may appear in adjunct position and allow for participantor event- orientation (cf. the contributions in Hinton et al. 1994 and Voeltz and Kilian-Hatz 2001). Participant-oriented ideophones are illustrated with

Issues in the syntax and semantics of participant-oriented adjuncts

59

the following two examples from Hausa18 (for similar examples from Ewe see Ameka, Ch. 11, this volume). Hausa (45) kana`r yaa tsaya`a k`ıika`m colonel 3sg.m.pst stay ideo:tall&motionless ‘The colonel stood there tall and motionless.’ (Newman 2000: 254) (46) naa gan shı` tik, haihu`war 1sg.pst see 3sg.m.o ideo:stark.naked birthgiving:poss uwarsa` mother.3sg.m.poss ‘I saw him naked as the day he was born.’ (Jaggar 1992: 92) Amha and Dimmendaal (Ch. 9, this volume) mention the case of several Nilotic and Omotic languages where ideophones may have participantoriented manner interpretations. Unlike in Hausa and Ewe, these have to be introduced by a defective verb ‘say’. This is also true for ideophones in a similar function in Shona (Gu¨ldemann, Ch. 10, this volume). Yet another type of adverbial-like part of speech allowing participantoriented uses is found in many northern Australian languages which have a closed class of verbs. Most ‘verbal’ notions, but also notions encoded by adverbs and verbal particles in Germanic languages, are expressed not by verbs, but by uninﬂecting, inherently predicative elements which form an open class. These are variously known as ‘preverbs’, ‘coverbs’, and ‘verbal particles’ in the literature (see also Simpson, Ch. 2, this volume, for preverbs in Warlpiri). They mainly function as constituents of complex predicates together with a verb (like mung ‘look at’ in (47)), but may also occur independently as participant-oriented adjuncts. In (47), the allative marker on the coverb gurdij ‘stand’ indicates that this is oriented towards the undergoer (O) of the main predicate—in this case, that the pig and not the horse is standing up (see further Schultze-Berndt 2000: 112–13, 2001). Jaminjung gurdij-bina (47) mung gani-ngayi-m¼biyang pigibigi look.at 3sg:3sg-see-prs¼now pig(abs) stand-all(o) ‘It [the horse] is looking at the pig [that is] standing up.’ (ﬁeldnotes, E. Schultze-Berndt)

18

We would like to thank Carmen Dawuda for drawing our attention to these Hausa examples.

60

N. P. Himmelmann and E. Schultze-Berndt

1.4.4.4 Word classes restricted to occurrence as a depictive adjunct Little attention has been paid in the literature to the fact that some languages appear to have word classes restricted to depictive function. Members of these word classes only occur in adjunct position and in this respect resemble adverbs. However, these lexemes show obligatory agreement with a participant of the clause. For example, Pinkster (1988: 224–6) discusses Latin forms such as invitus ‘reluctant’ which are traditionally classiﬁed as adjectives because of agreement in case, number, and gender, but which can function neither as attributes nor as main predicates, but only as depictive secondary predicates (praedicativa). Similar nonverbal word classes which show obligatory agreement and which are restricted to secondary predicate function have been described for some Australian languages, e.g. Kayardild (Evans 1995: 227–31), Martuthunira (Dench 1995: 53), Yankunytjatjara (Goddard 1985: 57), Diyari (Austin 1981a: 107–8), and Warlpiri (Hale 1982: 279–80; 1983; Simpson 1991: 123ff., and Ch. 2, this volume). Semantically, they comprise notions of posture, quantiﬁcation, and manner. An example is (48) from Diyari. Diyari (48) a. wata yini parraparra pithi-ya not 2sg.sbj energetic(abs) fart-imp ‘Don’t fart loudly!’ (Austin 1981a: 107) b. nhulu karna-li kirra parraparra-li warra-yi 3sg.nf person-erg boomerang(abs) energetic-erg throw-prs ‘The man throws the boomerang energetically.’ (Austin 1981a: 107) From a typological perspective, these forms are probably best regarded as a distinct part of speech, the only function of which is occurrence as depictive secondary predicates. 1.4.4.5 Verbal and deverbal depictives The use of deverbal predicates as depictives is well attested for European languages, where participles can be used just like adjectives, although, as shown by Bucheli Berger (Ch. 4, this volume) for some Swiss German dialects, there may in fact be restrictions on the verb classes that these participles may come from. Following Haspelmath (1995), we use the term converb for ‘participles’ which are used primarily as adjuncts. In Schultze-Berndt and Himmelmann (2004: 98–106) we argue that converbs with restricted reference and null subject—Haspelmath’s (1995: 9) ‘implicit subject’ converbs—are always participant-oriented because they are obligatorily controlled (see also Mu¨ller-Bardey 1990: 2–3; Haspelmath 1995: 17–20).

Issues in the syntax and semantics of participant-oriented adjuncts

61

Often, ‘implicit subject’ converbs with an S/A controller are in functional opposition with converbs which are not obligatorily controlled (or which even obligatorily express their own subject) and which typically have a differentsubject interpretation (Haspelmath’s ‘free subject’ and ‘explicit subject’ converbs). Functionally, in this case, the contrasting forms are part of a switchreference system (see e.g. Austin 1981b; Haiman 1983; Nichols 1983; SchultzeBerndt 2006: 186–9), and there has been a tendency to analyse both as adjuncts of the same type (i.e. calling them simply ‘adverbials’). Formally, however, in a given language, same-subject converbs may constitute genuine participant oriented adjuncts, while other converb types (which are neutral with regard to participant orientation) instantiate a general adjunct construction. Two languages which exhibit exactly this type of contrast, the Omotic languages Wolaitta and Maale, are discussed by Amha and Dimmendaal (Ch. 9, this volume). Another language where same-subject clauses, unlike different-subject clauses, formally mark participant orientation is Shipibo-Konibo, discussed by Valenzuela (Ch. 8, this volume). See sections 1.3.12–14 for further discussion. In many languages, the boundary between biclausal constructions, i.e. constructions involving two main predicates, and deverbal adjuncts showing participant orientation is difﬁcult to draw. This is either because the language lacks a ﬁnite/nonﬁnite distinction, as does Lao (Enﬁeld, Ch. 12, this volume) or because translation equivalents of ‘standard’ depictive examples involve a verb with the properties of ﬁnite verbs, often an auxiliary or copular verb ‘be’ linking a predicative nominal to the rest of the clause. Borderline cases of this type are presented by both Turkana (Amha and Dimmendaal, Ch. 9, this volume) and Shona (Gu¨ldemann, Ch. 10, this volume). In these languages, predicates of subordinate clauses in ‘depictive’ function take obligatory bound subject pronominals. They thus seem to express their own subject and not to meet the condition of obligatory control proposed for depictives by Schultze-Berndt and Himmelmann (2004). As the above-mentioned authors argue, though, clauses of this type do not allow NPs (including free pronouns) as subjects, and in this sense they can be said to be obligatorily controlled. 1.4.5 Morphological marking Morphological marking constitutes the ﬁfth and probably most straightforward parameter in a typology of participant-oriented adjuncts. In the simplest case, of course, participant-oriented adjuncts do not have any morphological marking. This is the case for adjectival depictives in English, which are distinguished from e.g. manner adverbials (including oriented adverbials) by the absence of formal marking (see section 1.2).

62

N. P. Himmelmann and E. Schultze-Berndt

In the following subsections, only the main types of overt morphological marking will be considered (Schultze-Berndt and Himmelmann (2004) discuss a few other types which are more marginal crosslinguistically, such as nonspeciﬁc linkers and genitive marking of adjuncts). The types of morphological marking—if any—of participant-oriented adjuncts that are found crosslinguistically formally reﬂect, on the one hand, the semantic subclasses of participant-oriented adjuncts (e.g. special markers for comitative and similative adjuncts: see section 1.3). On the other hand, they reﬂect functional relationships that the participant-oriented adjuncts bear to various other construction types, such as adverbials and predicative complements. These functional relationships are represented in the form of a semantic map in Mu¨ller-Bardey (1990) and, in a much more elaborate form, in van der Auwera and Malchukov (Ch. 13, this volume). Disregarding the specialized markers just mentioned, it seems to hold as a crosslinguistic generalization that participant-oriented adjuncts generally capitalize on marking strategies that are also found in other constructions. A speciﬁc marker of participantoriented adjuncts (more speciﬁcally in this case, of depictives) was only found in one language so far, the north-eastern (Appenzell) variety of Swiss German (Bucheli Berger, Ch. 4, this volume), where it can be shown to originate from agreement marking. 1.4.5.1 Agreement In many languages, agreement is used to establish a link between a participant-oriented adjunct and its controller. In these languages, agreement can be regarded as a major criterion for the status of an adjunct as a depictive, as reﬂected throughout section 1.3. Agreement, as usually understood, involves the categories of gender, number, and/or case. Languages exhibiting agreement of this type usually have NP-internal agreement as well, although, as the contribution by Boeder (Ch. 6, this volume) on Georgian shows, the latter does not necessarily display the same characteristics as the marking on depictives. Further examples, discussed in this volume, come from Swiss German dialects (Bucheli Berger, Ch. 4) and from Australian languages (Simpson, Ch. 2, and McGregor, Ch. 5). It is also well known that many Romance and Slavic languages exhibit depictive agreement. Agreement can also be used in a wider sense, for a marking strategy that does not involve any independently existing category (such as gender, number, case), but involves forms which have the sole purpose of unambiguously identifying the controller of an adjunct (such as S or A)—which implies, by deﬁnition, that there is no corresponding ‘agreement’ within NPs. This is discussed under ‘other strategies for indicating restricted reference’ in

Issues in the syntax and semantics of participant-oriented adjuncts

63

Schultze-Berndt and Himmelmann (2004: 84f.). If this marking strategy is employed on subordinate clauses and indicates co-reference of the (often null) subject of this clause and the controller, this is commonly referred to as switch-reference marking rather than agreement. The most common type is same-subject marking (i.e. orientation towards an S/A pivot), as illustrated for Wolaitta and Maale by Amha and Dimmendaal (Ch. 9, this volume). However, speciﬁc forms signalling O orientation are also attested, e.g. in the Australian languages Warlpiri (Simpson 1988; Hale 1994), Martuthunira (Dench 1988; see example (39) above), and Jaminjung (Schultze-Berndt 2001; see example (47) above), and in Shipibo-Konibo (Valenzuela, Ch. 8, this volume). The actual markers of participant orientation can be of different types, e.g. specialized converb forms (as in Wolaitta and Maale), or case markers in subordinating function which restrict the reference of a controller (as in Warlpiri and Jaminjung). For the Panoan language family, of which ShipiboKonibo is a member, Valenzuela (Ch. 8, this volume) argues that markers of participant-orientation, although synchronically unanalysable forms, arose diachronically through a combination of a specialized subordinating case marker and an additional case marker added in agreement (in the traditional sense). In Shipibo-Konibo, the same markers are found not only on subordinate clauses, but also on intraclausal adjuncts of lower complexity, which justiﬁes the use of the term ‘participant agreement’ rather than ‘switchreference marking’. In a language like Warlpiri, too, case markers in agreement with the controller appear both on clause-internal depictives, and in addition to the switch-reference marking in participant-oriented subordinate clauses.19 Yet another type of agreement between a subordinate clause and a participant of the main clause can be illustrated with Shona (Gu¨ldemann, Ch. 10, this volume). As already mentioned in section 1.4.4, subordinate predicates in 19 Corbett (1998: 195) argues that case-marking on a modiﬁer within a noun phrase is not triggered by the head noun but ‘imposed’ on both NP constituents by the governing verb, and that it should therefore not be subsumed under agreement. The same would presumably hold for identical casemarking of a controller and a depictive. Two arguments can be made in defence of the use of the term ‘agreement’ to include case-marking. The ﬁrst—also acknowledged by Corbett—is that in languages of the fusional type, the categories of number, gender, and case are often expressed by portmanteau morphemes and that it therefore makes little sense to separate them. The second is that, as indicated above, the involvement of case in depictive marking cannot be clearly separated from switch-reference marking. Clearly, one would want to analyse e.g. same-subject marking as involving the marking of a relationship between a target and its controller (i.e. agreement proper), and not maintain that it is assigned by the verb of the ﬁrst clause. This analysis of identical case marking as directly signalling a relationship between the two case-marked constituents could then be extended to non-clausal depictives.

64

N. P. Himmelmann and E. Schultze-Berndt

depictive function in Shona, although being obligatorily controlled in the sense that they do not allow an NP subject, take an obligatory bound pronominal. The orientation of the subordinate predicate can thus be indicated by person/noun class agreement of this bound pronominal with the controller. 1.4.5.2 Copulas and other predicative markers In many languages participantoriented adjuncts take predicative markers, i.e. markers which are also found on main predicates and on predicative complements, thus reﬂecting the function of the adjuncts in question as (secondary) predicates. A paradigm case of a predicative marker is the so-called essive case in Finnic languages. In Finnish, it is found not only on predicative complements and expressions of life stage and function or role (in these functions, its English translation equivalent is as), but also on participant-oriented manner expressions, as example (49) shows. Finnish (49) ha¨n la¨hti hiljaise-na huoneesta he left quiet-ess from:room ‘He went quiet out of the room.’ (Nichols 1978a: 123) The clitic ¼i in Ewe, which, as Ameka (Ch. 11, this volume) argues, is a predicative marker, has remarkably similar functions to the essive in Finnish. A predicative marker found with nominals both in main predicate and depictive function in the Eastern Nilotic language Bari is brieﬂy discussed by Amha and Dimmendaal (Ch. 9, this volume). There are also languages where not only main predicates but also secondary predicates appear with an obligatory copula, another type of predicative marker. The copula may be an invariable particle, as in Berber (Maas, p.c.), or belong to the verb class, as in Lao (Enﬁeld, Ch. 12, this volume). In the latter case, it may appear in a converbal form (see section 1.4.4.5) when marking secondary predicates. This is a crosslinguistically frequent strategy, found in languages as diverse as Turkish (Boeder and Schroeder 1998; Schroeder 2003; Johanson 1995), Shona (Gu¨ldemann, Ch. 10, this volume), and Wolaitta (Amha and Dimmendaal, Ch. 9, this volume). 1.4.5.3 Adpositional/case marking In many languages, certain semantic cases or adpositions such as locative or comitative are regularly widely employed on participant-oriented adjuncts. A well-known case of instrumental marking on participant-oriented adjuncts is Russian (see Jakobson 1936; Nichols 1978a; 1981; Janda 1993). Further examples of participant-oriented adjuncts taking an instrumental case or adposition can be found in Japanese

Issues in the syntax and semantics of participant-oriented adjuncts

65

(Takezawa 1993), Georgian (Boeder, Ch. 6, this volume) and the related language Laz (Kutscher and Genc¸, Ch. 7, this volume), and in Ewe (Ameka, Ch. 11, this volume). In section 1.3.10 we pointed out that certain locative adjuncts have to be regarded as primarily participant-oriented rather than event-oriented, in that they do not locate the event as a whole but only one of the participants. Locative cases and adpositions, moreover, may take on functions going beyond the purely spatial ones. In this case, too, the resulting constituents may have participant-oriented interpretations; examples can be found in Ewe (Ameka, Ch. 11, this volume, ex. (5b)), and Wolaitta (Amha and Dimmendaal, Ch. 9, this volume). Usually, locative-marked adjuncts of this type will also have to be regarded as instances of general adjunct constructions. There are other types of case marking or adpositional marking on participant-oriented adjuncts which cannot be assigned a prototypical function such as instrumental or locative. An example is the adverbialis case in Georgian, the functions of which are discussed in some detail by Boeder (Ch. 6, this volume). 1.4.5.4 Restrictive markers In section 1.2.3, depictive secondary predicates in the narrow sense were deﬁned as those participant-oriented adjuncts which are part of the focus domain in a given clause. This is consistent with the observation that depictives are often semantically more speciﬁc than the main predicate, and thus make the main contribution to the information conveyed by the clause. Depictives also often seem to be accompanied by certain particles or clitics which are associated with the focus of an utterance (see Ko¨nig and van der Auwera 1990: 344; Gu¨ldemann, Ch. 10, this volume). Schultze-Berndt (2002) discusses particles and clitics with primarily restrictive function (e.g. ‘just’, ‘only’) in several Australian languages, which are frequently found on secondary predicates. She argues that the restrictive marker indicates that only the speciﬁc event expressed by the depictive occurs, to the exclusion of possible alternatives. It seems that restrictive markers are obligatory or near-obligatory with depictives in some languages. In the Australian language Jaminjung, a restrictive clitic is obligatory with quantiﬁers in depictive function (ﬁeldwork Schultze-Berndt; see exx. (27) and (28) in McGregor, Ch. 5, this volume). Nichols (1982: 339) explicitly regards what she calls the ‘delimiter’ esˇˇce (lit. ‘still’) in Russian as a near-obligatory grammatical marker in the subtype of depictive constructions which convey life stage. Bucheli Berger (Ch. 4, this volume) reports that in the Swiss German dialect of Diepoldsau, the restrictive

66

N. P. Himmelmann and E. Schultze-Berndt

particle asa ‘so, so much’ appears to be obligatory on depictives, while in the surrounding dialects it is merely frequent in this construction. Finally, in Shipibo-Konibo, the restrictive or ‘emphatic’ marker -bi appears to be obligatory on expressions of life stage and emphatic pronouns, both considered as highly participant-oriented by Valenzuela (Ch. 8, this volume). Since the same marker also frequently occurs on adjunct types other than depictives in Shipibo-Konibo and the other languages mentioned here, it cannot be regarded as a depictive marker as such, but its high frequency and strong grammaticalization in this construction type constitutes further evidence for the close link between restrictive marking and depictive constructions.

1.5 Summary In this chapter, we have outlined a number of issues arising in a crosslinguistic approach to the syntax and semantics of depictive secondary predicates. Most importantly, we have seen that such an approach forces us to treat as a single domain a number of semantic adjunct types which to date have generally been treated as two distinct domains, that of depictives and adverbials. Here we have argued that the domain as a whole can be described in terms of the semantic features of participant orientation and event orientation. Thus, this chapter can also be read as a contribution to the further systematization of adverbials (see in particular section 1.2.1).20 The crosslinguistic approach shows that the morphosyntactic manifestation of participant- or event-orientation is an area of considerable variation (section 1.2). In some languages there exist speciﬁc depictive constructions, i.e. constructions which mainly or exclusively convey participant orientation (e.g. by means of agreement), which differ from adverbial constructions, i.e. constructions which mainly or exclusively convey event orientation. There are also adjunct constructions which remain unspeciﬁed for either event- or participant-orientation (these constructions are termed general adjunct constructions here), and all three construction types can be found in the same language. From the crosslinguistic variation in the semantic range of depictive and adverbial constructions, we have concluded that all of the semantic adjunct types discussed in section 1.3 are semantically both participant- and eventoriented (to very different degrees), and that there is competition as to which orientation is highlighted in morphosyntactic form through the use of a depictive or an adverbial construction. Perhaps the clearest example of this double nature is provided by manner expressions, which are discussed at 20

See also Mu¨ller-Bardey (Ch. 3, this volume) for a somewhat different approach to this topic.

Issues in the syntax and semantics of participant-oriented adjuncts

67

length in sections 1.2.2 and 1.3.3. Instances where morphosyntactic form appears to contradict semantic preference (as when temporal adjuncts agree in case with one of the arguments of the main predicate) are seen as cases of overgeneralization. The different degrees of participant- and event-orientation at the semantic level, as well as the variation in encoding, can be represented by way of a semantic map. In the map proposed in section 1.3, primarily or exclusively participant-oriented adjuncts, i.e. adjuncts conveying a condition or state of one of the arguments of the main predicate, are found at the centre, and primarily or exclusively event-oriented adjuncts, e.g. temporal adjuncts, at the periphery. The semantic map predicts, on the basis of the ﬁndings reported in this volume, that if participant orientation has a morphosyntactic manifestation in a given language, then the construction in question will cover contiguous segments of the map. Not only the semantic range but also the formal properties of adjunct constructions exhibiting participant orientation are subject to crosslinguistic variation (section 1.4). For example, participant-oriented adjuncts may pose more or fewer restrictions on the syntactic function (e.g. subject, object, indirect object) of possible controllers. And while the standard examples of depictives usually discussed in the literature are adjectives, participant-oriented adjuncts may have a more complex constituency and be built around members of various word classes. Noun phrases and adpositional phrases, ideophones, and subordinate clauses headed by ‘participial’ or converbal forms of verbs can all serve as participant-oriented adjuncts, and languages may exhibit a preference for adjuncts of one or the other type. In section 1.4.4 the possibility is raised that some languages even have a word class restricted to depictive function. In preparation of the crosslinguistic comparison of depictive secondary predicates, a number of concepts and distinctions discussed in the literature had to be critically reviewed and modiﬁed (section 1.2). Perhaps the most important point here is the hypothesis advanced in section 1.2.3 that depictives (in the broad sense) can be either in focus or part of the presupposition, and that at least in some languages this difference is manifest on the constructional level, such that a distinction has to be made between depictives proper and circumstantials. Much remains to be done to test this hypothesis, which among other things needs to be properly speciﬁed for each of the semantic adjunct types discussed in section 1.3. Further cross-linguistic research is thus likely to lead to further reﬁnements in the typology of participant-oriented adjuncts in terms of their semantics, their morphosyntax, and their status in terms of information structure.

This page intentionally left blank

2 Depictives in English and Warlpiri JANE SIMPSON

2.1 Introduction Typical and often-cited examples of ‘depictive attributes’ (Halliday 1967) in English include: (1) a. I ate the meat raw. b. She left drunk. c. I ate the meat drunk. The adjectives (in boldface) in (1) act as secondary predicates, attributing a quality to a participant (underlined) at the time of action denoted by the verb. The term ‘attribute’ was introduced by Halliday to label an English construction involving a ‘characteristic ascribed to one of the participants . . . that relates speciﬁcally to the process in question’ (Halliday 1967: 62). ‘Depictives’ are an ‘attribute’ subtype ‘which characterises the attribuant in relation to the process, but as a concomitant, not a result of the process’ (p. 63). Thus, attributes in Halliday’s sense are both participant-oriented and event-oriented, like manner adverbs; but unlike English manner adverbs, they focus more on the participants than on the process (Schultze-Berndt and Himmelmann 2004, and introduction to this volume). The sentences in (1) have translation equivalents in Warlpiri. (2) contains a nominal, wanka ‘raw’, predicated of the understood object, and a nominal yarnunjuku-rlu¼ju, ‘hungry’, predicated of the subject, expressed as a clitic -lu. The verb ngarni ‘eat’ always takes an Ergative subject and so yarnunjuku has Ergative case agreeing with that subject. Thus in both English and Warlpiri I thank the participants in the Workshop on Depictives, Ruhr-Universita¨t Bochum, 7 June 2001, and participants at the 2002 Blackwood Workshop on Australian Languages for stimulating papers and discussion. And I am very grateful to Nikolaus Himmelmann and Eva Schultze-Berndt for excellent editorial remarks which have made me rethink much of the paper.

70

J. Simpson

a state of affairs can be predicated of a participant by means of a nominal, without mediation by an adposition or semantic case-marking.1 (2) Nga-rni¼lki ka¼lu wanakiji¼ji. eat-npst¼then prs¼333.sbj bush.tomato¼top yarnunjuku-rlu¼ju Nga-rni ka¼lu eat-npst prs¼333.sbj hungry-erg¼top ngula¼ju ka¼lu wanka nga-rni. that¼top prs¼333.sbj raw eat-npst ‘Bush tomatoes, they eat them then; they eat them when hungry, and they eat them raw.’ [Hale 1966: 82]

1 Glossing conventions used only in this paper: 111: ﬁrst person plural exclusive, 222: second person plural, 33: third person dual, 333: third person plural, 122: ﬁrst person plural inclusive. The boundary between clitics and sufﬁxes is not clear-cut in Warlpiri; endings that are uncontroversially clitics are marked with an equals sign, ¼, while other endings are marked with a hyphen, -. In some Warlpiri examples prosodic breaks are indicated by one or two dashes (- or --), in line with conventions used in their sources. Warlpiri sentences are followed by reference to their source. Published sources are given in parentheses. Unpublished sources are given in square brackets. Much of this material is also contained in digital form in ASEDA (Australian Studies Electronic Data Archive). The following abbreviations are used:

Darby Jampijinpa 1984¼Darby Jampijinpa Warumungku Watikirli, Warlpiri Literature Production Centre Inc. Yuendumu, 1994. Dinny 1979¼Dinny Japaljarri France, Tape #402 (copy of #63). Recorded at Yuendumu 1979. Transcribed by Tess Ross Napaljarri. Hale 1959¼Hale, Kenneth (1959). Walbiri Field Notes. ms 3171, also digitally in the collection ASEDA 0050. Canberra: Australian Institute of Aboriginal and Torres Strait Islander Studies. Followed by page number or section reference. Hale 1966¼Hale, Kenneth (1966–7). Warlpiri Field Notes: Yuendumu, Hooker Creek, Lander River, Warrabri, N.T. ms 865, also digitally in the collection ASEDA 0050. Canberra: Australian Institute of Aboriginal and Torres Strait Islander Studies. Followed by page number or section reference. Ngarlinjiya¼Robertson, Ngarlinjiya Mary Nungarrayi, Transcript of tape no. 629a recorded at Yuendumu, 23 May 1990 by Peggy Rockman Napaljarri and Lee Cataldi, 5 June 1990. Later published (1994) as ‘Witikirli’, in Peggy Rockman Napaljarri and Lee Cataldi (ed.), Warlpiri Dreamings and Histories: Yimikirli. (The Sacred Literature Series.) San Francisco, London, Pymble: HarperCollins. Swartz¼Swartz 7 June 1996: Luke 17:26 Toby Japangardi¼Transcription of 60-minute tape recording made at Willowra by Toby Japangardi around 1984/5, transcribed by Mary Laughren, Dec. 1988. Warlpiri Dic.¼The Warlpiri Dictionary, digital version May 2000, by permission of Mary Laughren. Sentences are cited by reference to the lemma within which they occur (e.g. (Warlpiri Dic: ngakaparla)), and if there is a source given for the example sentence, that also is included (e.g. (Warlpiri Dic. wankaru: C:ML warrarna)). Where I have been able to ﬁnd the original source, I have checked that, because the example sentences in dictionaries sometimes have to be edited to make the illustration clear. The translations are those given in the Dictionary if possible, modiﬁed occasionally if they are misleading.

Depictives in English and Warlpiri

71

In these examples wanka and yarnunjuku-rlu superﬁcially ﬁt the criteria of Schultze-Berndt and Himmelmann (2004: section 2.8) for depictives. Both are nominals without tense or mood marking. Neither is obligatory. Wanka, yarnunjuku-rlu and the verb ngarni are separate predicates, and the state of affairs expressed by wanka/yarnunjuku-rlu holds at a time within the timeframe expressed by ngarni. The subject of wanka is the same as the object of ngarni, while the subject of yarnunjuku-rlu is the same as the subject of ngarni. Neither wanka nor yarnunjuku-rlu forms a complex predicate or a low-level constituent with ngarni, and equivalent secondary predicates appear in a variety of sentence positions. (Warlpiri does have many complex predicates consisting of preverbs and verbs (section 2.2.3 below).) However, while in (1) and (2) the Warlpiri and English secondary predicates appear to be doing much the same thing semantically, there are more syntactic and semantic constraints on the appearance of the English depictives in (1) than on secondary predicates in Warlpiri. I suggest that many of these differences stem from basic syntactic differences between Warlpiri and English. In each language, these secondary predicates act syntactically as part of different systems of grammar. In Warlpiri, as I shall show, these secondary predicates are part of a general system of secondary predication using nominals, which act syntactically as adjuncts, while in English, depictives have closer ties with the ways of expressing complements. In the ﬁrst part of the chapter, I describe and contrast the systems within which depictives operate in the two languages. In the second part I compare and contrast the constraints on depictives in English with the relative freedom of secondary predicates in Warlpiri.

2.2 Differences between Warlpiri and English Obvious differences in behaviour between depictives in Warlpiri and English stem from: the way grammatical functions are expressed using word order in English and using case agreement in Warlpiri; the wide range of denotation of Warlpiri nominals; the complementarity of depictives and resultatives in English, and related to this, the constraints on the appearance of more than one secondary predicate; the use of clitics in Warlpiri to express temporal information, information structure and connections between predicates. I discuss each in turn.

J. Simpson

72

2.2.1 Word order vs. case agreement In English, depictive predicates normally occur following an intransitive verb (He died young) or the object of a transitive verb (He ate the meat drunk), and, for some verbs, following certain prepositional complements (He served the meat to them raw). They differ from manner adverbs in that they cannot precede the verb (*He drunk ate the meat, he drunkenly ate the meat). In terms of phrase structure position, Andrews (1982) showed that English depictives, regardless of whether they are predicated of the subject or object, occur inside the Verb Phrase (VP) constituent, as (3) illustrates. His ﬁndings are repeated by Legendre (1997) for French. Koizumi (1994) argues that in Japanese depictives predicated of objects and some subjects are in the VP, while some depictives predicated of the subject are generated as sisters to VP. Koizumi extends the ﬁndings to English to account for the marginal acceptability of sentences like (3h). Eating ﬁsh raw upsets Mary. Eating ﬁsh naked upsets Mary. What Mary did was eat the ﬁsh naked. What Mary did was eat the ﬁsh raw. Eat the ﬁsh raw though Mary did, she was nonetheless blamed for her attitude. f. Eat the ﬁsh naked though Mary did, she was nonetheless blamed for her attitude. (Koizumi’s 108a) g. What John did was drive his car drunk. h. ?What John did drunk was drive his car. (Koizumi’s 108b)

(3) a. b. c. d. e.

Non-depictive secondary predicates are expressed in English by other phrase structure positions. Sentence margins (clause-initial and clause-ﬁnal) are places where adjuncts modifying the whole clause, or the subject, may appear (4a–d), often resembling depictives in meaning. This is especially true of longer adjuncts which may simultaneously say something about the event (4b–d). Clause-ﬁnal position can also be used for heavy adjuncts modifying arguments (and even adjuncts) other than the subject (Quirk et al. 1985: 425) (4e–h). They may be ‘strong’ detached adjuncts (Stump 1985; Himmelmann and Schultze-Berndt, Ch. 1, this volume) which may have causal or concessive readings, or they may be ‘weak’ and integrated, as English depictives are. (4) a. b. c. d.

Drunk, John went home. John locked himself out of his house, drunk and bewildered. Crazy as it seems/Crazy as she is, Mary loves John. Mary loves John, crazy as it seems/crazy as she is.

Depictives in English and Warlpiri

73

e. Mary loves John, crazy as he is/even drunk. f. Mary gave John her car, drunk as he was/brand-new as it was. g. Mary gave her car to John, drunk though he was/brand new though it was. h. ?For Mary he’d do anything, pregnant or not. Depictives and sentence adjuncts in English behave differently under negation, regardless of whether the depictive is predicated of the subject or the object. Depictives, being weak and integrated, are in the scope of negation. Negation of the verb can negate both the verb and the depictive, or just the depictive. (5) a. He didn’t arrive drunk. (either He arrived sober, or He didn’t come at all.) b. He didn’t eat the meat raw. (either He ate it cooked, or He didn’t eat it at all.) c. He didn’t eat the meat naked. (either He ate it fully clothed, or He didn’t eat it at all.) The verb can be negated on its own only if there is a heavy pause following the verb, in which case the most plausible interpretation is not as a depictive, but as a strong detached sentence-margin adjunct, and, as Quirk et al. (1985: 425) argue, the heavier the adjunct is the more acceptable the sentence (6b and c, compared with 6a; 6g compared with 6f). These adjuncts say something about the state of the participant they are predicated of, and also about the reason the main event did not take place (see Nichols 1981: 42 for similar ﬁndings in Russian), and acceptability often depends on this (6d compared with 6e). Getting a similar reason reading when a bare adjective or adjective phrase is predicated of an object is harder (6f, g), and is helped by the presence of an overt subject (6h). ??He didn’t arrive, drunk. ??He didn’t eat the meat, naked. He didn’t arrive, too drunk to remember the appointment. He didn’t arrive, drunk as usual. ?He didn’t die, naked as usual. ??He didn’t eat the meat, raw. ?He didn’t eat the meat, disgustingly raw and crawling with maggots. h. He didn’t eat the meat, raw as it was/it being raw.

(6) a. b. c. d. e. f. g.

Hence English depictives are best tested inside a VP or inside a sentential argument or adjunct, as in (3), since detached adjunct readings are much less plausible in these environments.

74

J. Simpson

Inside the VP a depictive predicated of the subject or object must follow NP arguments, but may follow or precede PP arguments with intransitive verbs, as in (7a–h), unless the preposition is acting more like a casemarker and less like a predicate in its own right (7i, j). (7a–d) also show that there need be no overt noun phrase of which the secondary predicate is predicated. The subject of the secondary predicate can have an unspeciﬁed referent (Laughren 1992; Himmelmann and Schultze-Berndt, Ch. 1, this volume). (7) a. b. c. d. e. f. g. h. i. j.

Jumping fully clothed into the water was a bad idea. Jumping into the water fully clothed was a bad idea. Prancing naked around the room for an hour was a bad idea. Prancing around the room naked for an hour was a bad idea. Running naked into the street was Jones’s ﬁrst solecism. Running into the street naked was Jones’s ﬁrst solecism. Arriving drunk at the party was Jones’s second solecism. Arriving at the party drunk was Jones’s second solecism. Staring at him drunk was Jones’s third solecism. *Staring drunk at him was Jones’s third solecism.

Depictives may also follow or precede PP arguments with transitive verbs, when predicated of the object (8a–d). However, when predicated of the subject (8e, f), sentence ﬁnal position seems preferable. (8) a. After handing the baby to its mother crying, Jones sighed with relief. b. After handing it crying to its mother, Jones sighed with relief. c. After having waiters serve food raw to the visitors, they brought in the elephants. d. After having waiters serve food to visitors raw, they brought in the elephants. e. ??After having waiters serve food naked to visitors, they brought in the elephants. f. After having waiters serve food to visitors naked, they brought in the elephants. Depictives predicated of objects must precede nonﬁnite VP adjuncts (9a, b). Depictives predicated of subjects of transitive verbs are difﬁcult to get with nonﬁnite VP adjuncts (9c, d). However, if the verb is intransitive, then depictives are more acceptable, as in (9e) (thanks to the editors for this example).

Depictives in English and Warlpiri

75

(9) a. Burning her alive to justify their cause must have seemed like a good idea. b. *Burning her to justify their cause alive must have seemed like a good idea. c. ??After eating the meat drunk to satisfy their hunger they went home. d. *After eating the meat to satisfy their hunger drunk they went home. e. After driving home drunk to avoid being stuck at the host’s place, they went to sleep Differences between depictives predicated of subjects and depictives predicated of objects have received much attention (Legendre 1997 for French; Koizumi 1994 and Yatsushiro 1999 for Japanese; Winkler 1997 for English). In the generative literature the usual solution has been to assume that depictives predicated of objects are attached close to the V (e.g. as sisters of V0 ), and depictives predicated of subjects are attached further away from the V (e.g. as sisters of a VP whether inside a VP or not). Thus phrase structure position in English determines in part whether or not a secondary predicate is interpreted as a depictive, what the scope of negation is, and whether a depictive can be interpreted as predicated of a subject or object. In Warlpiri, phrase structure position is used for information structure purposes (for example, to indicate what is questioned), but is not used for grammatical function or to indicate that a phrase is predicated of another word. For example, the secondary predicate wankaru in (10) appears in a variety of positions, sentence-ﬁnal (10a), sentence-initial (10b), before the verb (10c), after the object and before an adjunct (10d), before the object (10e)—and other places are possible too. It can be predicated of the subject of an intransitive verb (10a, c), of the object of a transitive verb (10b, d), of the subject of a transitive verb (10e, f). parnka-ja wankaru. (10) a. Jinta-kari kala one-other used.to run-pst alive ‘One would run off alive.’ [Hale 1966: 1138–9] b. Wankaru ka¼jana ngaja-rni kurdu warrarna-rlu. live prs¼333.sbj bear-npst child skink-erg ‘The warrarna skink gives birth to its young live.’ [Warlpiri Dic. wankaru: C:ML warrarna]

76

J. Simpson c. Marlu -- panti-rninja-warnu, ngula; kangaroo spear-nmlz-assoc that wankaru wuruly-parnka-ja. alive hide-run-pst ‘That kangaroo, after being speared, it would run off and hide alive.’ [Hale 1966: 349–50] d. Puuly-marda-rni ka wardapi wankaru¼wiyi hold-npst prs goanna live¼before paka-rninja-kungarnti-rli. hit-nmlz-purp2-erg ‘He holds onto the goanna still alive before killing it.’ wankaru: C]

[Warlpiri Dic.

e. Kirdanyanu-kirdanyanu-rlu kala¼lu¼nganpa father-poss-father-poss-erg used.to¼333.sbj¼111.obj wankaru-rlu¼wiyi yujuku ngurrju-ma-nu- tarnnga-ngku. alive-erg¼before hut good-cause-pst always-erg ‘Our fathers when they were still alive used always to build us huts.’ (Warlpiri Dic. wiyi: BWJ1979:20) f. Yungu¼rlupa¼nyanu warrawarra.ka-nyi wankaru-rlu¼(yi)jala. so.that¼122.sbj¼refl look.out.for-npst live-erg¼also Kula¼rlipa jarda-nguna-mi jalangu¼ju. not¼122.sbj sleep-lie-npst today¼top Yungu¼rlupa¼nyanu wankaru-rlu warrawarra.ka-nyi. so.that¼122.sbj¼refl live-erg look.out.for-npst ‘So we can look out for ourselves (when) lively also. We won’t sleep now, so we can look out for ourselves (while feeling) lively.’ [Hale 1959: 131ns; also Warlpiri Dic. wankaru] We have not found evidence to link phrase structure position with whether the secondary predicate is predicated of a subject or object. Instead, what the secondary predicate is predicated of is shown by case agreement. In (10a–d) above, absence of case is interpreted as Absolutive, the case of the subject of intransitive verbs. In (10e) the Ergative case on wankaru-rlu shows that it is predicated of the Ergative case-marked subject, kirdanyanu-kirdanyanu-rlu. (10f) has no overt Ergative case-marked subject, but the verb is transitive and a predicate agreeing with the understood subject must have Ergative case. As in English, the secondary predicate can be within the scope of negation, by being in the same prosodic unit as a clause starting with the sentential negator kula. The secondary predicate may precede or follow the verb—the examples in (11) are predicated of the object.

Depictives in English and Warlpiri

77

(11) a. Kula¼lpa¼lu wankaru ka-nja-nu, na:. not¼pst¼333.sbj alive carry-nmlz.go-pst no ‘They did not take them along alive, no.’ [context: they did carry them along, and they made them alive afterwards] [Hale 1966: 892] yinjalykinyi¼ji. b. [..] yirnmi ka¼rnalu nga-rni cooked prs¼111.sbj eat-npst tuber¼top Kula¼lpa¼rnalu nga-njarla rdarri, not¼pst¼111.sbj eat-irr raw purra-nja-warnu-mipa ka¼rnalu nga-rni yinjalykinyi¼ji. cook-nmlz-after-only prs¼111.sbj eat-npst tuber¼top ‘Cooked is how we eat the yinjalykinyi tuber. We cannot eat it raw. Only after it has been cooked do we eat the yinjalykinyi tuber.’ [Warlpiri Dic: yinjalykinyi] There are examples in which the verb is denied while what appears to be a secondary predicate remains taken as true (12). Yarnunjuku-ku in (12) is most probably a secondary predicate, predicated of the Dative ﬁrst person inclusive object. Its interpretation is vague between unrestrictive (as the translation suggests), concessive (‘even though we are hungry’), or restrictive (‘us hungry ones’). Such vagueness of interpretation is discussed in Section 2.2.2. (12)

yarnunjuku-ku. Kula¼ka¼ngalpa yi-nyi not¼prs¼122.obj give-npst hungry-dat ‘He doesn’t give any to us who are hungry.’ (context: speaker is starving, and complains about not being given anything) [Warlpiri Dic: ngurru: rno]

Warlpiri has little surface evidence for a VP. It is thus hard to ﬁnd parallels in Warlpiri for the constraints on ordering depictives depending on what they are predicated of. There are phrases consisting of nominalized verbs preceded by their arguments and to a limited extent other constituents, but the preference is for very little material to appear in such phrases. (13) is a possible example; however it is usually possible to interpret such apparent depictives as ‘preverbs’ (predicates that modify the verb) (Section 2.2.3). (13)

Lani parnka-nja-wangu¼(l)pa¼npa karri-yarla¼rni. scared run-nmlz-without¼pst¼2.sbj stand-irr¼hither ‘You would stay without running away scared.’ [Hale 1959: 7.238]

Example (14) shows other relevant syntactic features of Warlpiri. Nominals can act as the main predicator of a clause, since there is no obligatory

78

J. Simpson

copula (14i, iii, iv), but stance verbs can be used as copulas (ngunami in 14vii). There is no requirement for an overt subject (14ii). As will be discussed in section 2.2.2, there is no morphosyntactic distinction between nouns and adjectives and manner adverbs. Consequently (14iii) is syntactically ambiguous between a coordination of two primary predicates, or a primary predicate and a secondary predicate: it is still unripe, it is green ﬁrst or it is still unripe, ﬁrst green. These factors combine to make it difﬁcult in isolation to parse nominal predicates following a pause as primary or secondary, let alone as depictives or as detached adjuncts or even manner adverbs. (14)

i. Pirarrpirarrpa nyanungu¼ju yipintiri wita. orange that¼top berry small ii. Pirarrpirarrpa-jarri ka. [ . . . .] orange-become.npst prs iii. Ngula¼ju ka2 wanka¼juku; yarringkarringki¼wiyi. [..] that¼top prs raw¼still green¼before iv. Wanka¼ju yarringkarringki¼wiyi. raw¼top green¼before v. Ngula-jangka¼ju pirarrpirarrpa-jarri¼lki that-from¼top orange-become.npst¼then vi. nyanungu¼ju - yipintiri¼ji. [ . . . ] that¼top berry¼top vii. Pirarrpirarrpa¼lku¼lu nguna-mi yurnmi¼ji. orange¼then¼333.sbj lie-npst ripe¼top i. ‘It is orange, that little yipintiri berry. ii. It becomes orange [..] iii. It is still unripe, ﬁrst green. iv. The unripe one is ﬁrst green. v. After that it becomes orange then, vi. that berry.[..]. vii. The ripe ones are orange then.’ [Hale 1966: 834–5]

Example (14) above also shows another feature of Warlpiri secondary predicates—their appearance with temporal-aspectual clitics (Green 1987) (italicized in 14). These provide a precise indication of how to connect the time-reference of the clause with the time-reference of the state associated with the element to which they are attached.

2

It is unusual to have the auxiliary ka without a verb as the main predicate.

Depictives in English and Warlpiri

79

(15) a. ¼ lku, ¼lki:3 X is in state S at time T, and X wasn’t in state S before time T (14v, vii) b. ¼ wiyi: X was in state S before time T, and X is not in state S after time T (14iii, iv) c. ¼ juku, ¼jiki: X is in state S at time T, and X was in state S before time T (14iii) The use of these clitics allows great ﬂexibility for Warlpiri secondary predicates to appear with verbs of different aspectual classes (sections 2.2.2, 2.2.6, 2.3.3). 2.2.2 Types of secondary predicate As the previous section has suggested, there is no clear-cut distinction between secondary predicates translating as depictives and other nominals. In both instances, case is used to determine what participant the predicate is predicated of, that is, orientation towards a participant (Himmelmann and Schultze-Berndt, Ch. 1, this volume). The ﬁrst and most obvious instance are the so-called discontinuous noun phrases (Hale 1983; Nash 1986; Simpson 1991; Hale et al. 1996). Nominals with the same case can be interpreted as denoting the same entity, regardless of whether they are contiguous in the sentence. (16)

Maliki-rli¼ji yarlku-rnu wiri-ngki. dog-ERG¼1.obj bite-pst big-erg ‘The/a big dog bit me.’; ‘The/a dog bit me, a big one.’; etc.

In (16) both ‘dog’ and ‘big’ have Ergative case, and the interpretation of the nominals and the relation between them is vague as to whether the nominals are attributes or heads and, if attributes, restrictive or non-restrictive. Obviously the normal interpretation would be for malikirli to be the head and wiringki the attribute. But it seems that only context determines whether the attribute is restrictive, ‘The big dog bit me’ (the ‘merged’ reading, terms due to Nash 1986 and Hale 1983), or non-restrictive (‘unmerged’ reading), ‘The dog bit me, a big one’. I have found no morphosyntactic means for distinguishing these uses of nominals from the depictive use in (2). The question then arises about those nominals which, unlike wiringki in (16), can’t really be interpreted as restrictive attributes. Warlpiri nominals range from deﬁnite and fully referential expressions to ones which are almost 3 Warlpiri clitics and sufﬁxes with high vowels usually agree in backness with the last vowel of the stem to which they attach.

80

J. Simpson

exclusively predicative in use. One such class are ‘action nominals’ (Simpson 1991), secondary predicates which, unlike depictives, describe activities or emotions, rather than states, of some participant. These agree in case with the case assigned to that participant, manyu ‘play, fun’ in (17a, b). (17) a. Maliki-jarra-rlu ka¼pala¼nyanu manyu-ngku yarlki-rni. dog-two-erg prs¼33.sbj¼refl play-erg bite-npst ‘The two dogs are biting each other in play.’ (Warlpiri Dic.: yarlkirni) manyu ngari yangka b. Ngula-piya¼jala ka¼lu that-like¼contrast prs¼333.sbj play just anaph junpurr-ma-ni ﬂap.lips-do-npst ‘Like that they playfully blow (ﬂap their lips)’ [Hale 1966: 0102/Warlpiri Dic.: junpurr(pa)] [junpurr-mani’s subject has Absolutive case.] Another such class are secondary predicates indicating place. If these attribute location to an object only, they agree in case with the object, whether this is Absolutive (unmarked) (kuyu in 18a), or Dative (the understood Dative argument cross-referenced by the auxiliary clitic ¼rla in 18b). If the sentence is intransitive, then the location nominal is unmarked. If the sentence is transitive, then the location nominal agrees in case with the subject, but only if the place is more closely associated with the subject (18d, e) than with the object or the event (Hale 1982; Laughren in press). (18f) shows an example where the location nominal is unmarked, despite the subject having Ergative case. (18e) also shows that the nominal may receive overt case-marking in agreement with the subject whether the nominal is inherently locative (yantarli) or marked with Locative case (ngurrangka). kuyu. (18) a. Parlku-ngka ka¼rnalu purra cooking.trench-loc prs¼111.sbj cook.npst meat. ‘We cook meat in a pit.’ [Warlpiri Dic.: jarnngi] walya karrkurr-pu-ngu b. Kala¼lu¼jana¼rla used.to¼333.sbj¼333.obj¼dat earth crunch-pst rdaku-ngka-ku¼ju. hole-loc-dat¼top ‘They (children) would crunch the earth over them (tadpoles) in the hole.’ [Warlpiri Dic.: karrkurrpa: Nora Nungarrayi] [The understood object, the tadpoles, has Dative case.]

Depictives in English and Warlpiri

81

c. Ngurra-ngka¼lpa nguna-ja purlka¼pardu. camp-loc¼pst lie-pst old.man¼dear ‘The dear old man was lying in camp.’ [Dinny 1979] d. Nga-rnu¼lpa yangka ngurra-ngka-rlu eat-pst¼pst anaph camp-loc-erg ‘He was eating in camp.’ [Dinny 1979] (Overt nominal subjects of ‘eat’ always have Ergative case.) wapal.nya-nyi yantarli-rli e. [ . . . ] yapa-kari-rli ka¼rla person-other-erg prs¼dat look.for-npst at.home-erg ngurra-ngka-rlu. camp-loc-erg ‘[ . . . ] another person who is at home in camp looks out for it.’ [Warlpiri Dic.: riwi: jne] ngurra-ngka¼jala f. Kaji¼ka¼ngku¼lu if¼prs¼2.obj¼333.sbj camp-loc¼contrast Kulu-parnta-mipa-rlu¼ju. panti-ni. spear-npst anger-full-only-erg¼top ‘They might spear you in camp indeed. All the ﬁghters.’ 1959: 7.221]

[Hale

Another kind of secondary predication is found in the part–whole syntax of Warlpiri (Hale 1981), which permits nominals denoting body parts to be predicated of some participant in the clause. Furthermore, these body parts can themselves have modiﬁers, forming noun phrases which agree in case with the case assignable to any nominal denoting that participant, unmarked Absolutive in (19a), Ergative in (19b). There seem to be no intonational or morphosyntactic grounds for distinguishing these noun phrases from other nominals used as secondary predicates in Warlpiri; thus compare (19a) with (19c, d).

(19) a. Parrulka ka

kakarda kirrirdi wapa. turkey prs neck long walk.npst ‘The turkey walks with long neck.’ [Hale 1959: 799]

b. Jirri¼kila ka nga-rni, kalwa-ngku, tadpole¼foc prs eat-npst egret-erg ngapa-ngka-rlu, kakarda kirrirdi-rli water-loc-erg neck long-erg ‘It’s tadpoles that it eats in the water, the egret, that long necked (one).’ [Hale 1966: 590]

82

J. Simpson c. Yangka yi¼ka kurdu-wangu wapa. anaph cause¼prs child-without move.npst ‘One who walks around without children’ [in deﬁnition of ‘childless’] [Hale 1966 pt. 3: 0378] d. Nyurrurla ka¼nku¼lu pingka wapa. you.pl prs¼222.sbj slow move.npst ‘You walk slowly.’ [Hale 1966 pt. 3: 0239]

Thus case agreement on action nominals, locations and body parts indicates orientation towards a participant, as Himmelmann and SchultzeBerndt suggest (Ch. 1, this volume). But case agreement cannot always be easily interpreted in this way. For example, manner nominals such as those denoting speed of an action denote a property of the event, rather than of the participant expressed by the subject. They still receive case-marking in agreement with the subject, as in (20a, b); see Himmelmann and SchultzeBerndt (Ch. 1, this volume) for discussion and similar examples from other languages. (20) a. Turlka-ni kapi¼rna¼ngku wakurturdu-rlu. pinch-npst fut¼1.sbj¼2.obj hard-erg ‘I’ll pinch you hard.’ [Hale 1959: 933] (turlkani’s subject has Ergative case.) b. Yankirri¼ji ka wakurturdu-nyayirni emu¼top prs fast-very jukajuka-parnka. head.back.forth-run.npst ‘The emu runs fast with head going back and forth.’ 1959: 8.55]

[Hale

Finally, nominals denoting time also may agree in case with the subject. (21a) shows that if the subject is Absolutive, then the time nominal has no overt case-marking. In (21b, c) the transitive verbs require Ergative subjects and the time nominals also have Ergative case (see also (10e)).

(21) a. Pirrarni

nyurnu-jarri-ja kurdu. yesterday sick-become-pst child ‘The baby got sick yesterday.’ [Hale 1959: 7]

b. Nya-ngu¼rna¼ngku pirrarni-rli. see-pst¼1.sbj¼2.obj yesterday-erg ‘I saw you yesterday.’ [Hale 1959: 19]

Depictives in English and Warlpiri

83

c. Jalangu-rlu ka¼rna¼nyarra yirri-pura today-erg prs¼1.sbj¼222.obj tell.npst luwa-rnu kuyu yinya¼ji yangkurra-rlu¼nya¼lu yesterday-erg¼emph¼333.sbj shoot-pst meat that¼top yapa-patu-rlu¼ju puluku manu marlu. person-many-erg¼top cow and kangaroo ‘Today I tell you that yesterday the people shot a cow and a kangaroo over there.’ [Warlpiri Dic.: yangkurra: ppj 10/87] Not all time nominals agree in case all the time with the subject, but what determines the variation has not yet been clariﬁed. In sum, Warlpiri nominals have a wide range of uses. Many of these can be seen as participant-oriented adjuncts (Himmelmann and Schultze-Berndt, Ch. 1, this volume), where case agreement indicates the participant to which they are oriented. However, case agreement with the grammatical function Subject sometimes has no obvious semantic connection with the participant expressed by the subject, as with the time nominals, and perhaps manner nominals such as ‘fast’. For time nominals, in fact, their case-marking seems to have clausal scope (Laughren, in press; Simpson 1991). Such case-marking could result from the extension of a pattern that was originally semantically motivated by participant orientation (Himmelmann and Schultze-Berndt, Ch. 1, this volume). 2.2.3 Resultative/depictive contrast An important difference between English and Warlpiri is that in English syntactically similar structures are used for both ‘predicatives of being’, or depictive secondary predicates (22a), and ‘predicatives of becoming’, or resultative secondary predicates (22b) (Jespersen 1965: ch. 27.17; Halliday 1967). In Warlpiri, ‘predicatives of being’ and ‘predicatives of becoming’ are handled in different parts of the grammar. (22) a. They burned Joan of Arc alive. b. They burned the steak black. Burning Joan alive was an error. depictive Burning the steak black was an error. resultative (22) shows how similar the structures for expressing depictives and resultatives can be. In fact, out of context some English sentences can be ambiguous between a resultative reading and a depictive reading (23a). It is also hard to ﬁnd acceptable sentences containing both a resultative and a depictive (23b) (but see also section 2.2.4 below).

84

J. Simpson

(23) a. She cooked it dry. Her hair fell loose. The water ran cold. (Halliday 1967: 65) b. ??They burned Joan of Arc black alive. *They burned Joan of Arc alive black. Moreover, both depictives and resultatives are also in approximate complementary distribution with nonﬁnite VP complements of verbs (24a–d). ((24a, c) are acceptable in the reading ‘He was to become teetotal drunk’.) (24) a. b. c. d.

*Ordering him to become teetotal drunk was a bad idea. *Ordering him drunk to become teetotal was a bad idea. *Convincing him to become teetotal drunk wasn’t hard. *Convincing him drunk to become teetotal wasn’t hard.

The major difference between English resultatives and depictives has to do with what they can be predicated of. Resultatives must be predicated of the object of a transitive verb, can only be predicated of the subjects of certain intransitive verbs, and can appear in a fake reﬂexive construction, as in She cried/ate herself sick. Resultatives furthermore are unique—not more than one resultative can occur in a single clause (*He boiled the pot dry black) unless they are overtly conjoined. Depictives can be predicated of the subject or object of a transitive verb, and the subjects of many intransitive verbs, but do not appear in fake reﬂexive constructions. In Warlpiri, free nominals which agree in case with what they are predicated of predicate a state or location or an action or emotion of their subject. Change of state is handled by other means. These include: the use of compound verbs (-jarri(mi) ‘become’ in (14ii, iv) and (21a) indicates a change of state), and the use of special ‘translative’ resultative case sufﬁxes (-karda in 25aii) which normally attach to nominals, or complementizer sufﬁxes (-kijaku in 25aiii) which can attach to nominals or nominalized verbs. These resultative case sufﬁxes create secondary predicates that are not as restricted as their English counterparts; they can be predicated of the subject of a transitive verb (25c). Moreover, even the special resultative case sufﬁx is not restricted: two secondary predicates with the resultative sufﬁx can occur in the same clause (25b), and a secondary predicate with a resultative sufﬁx can co-occur with a secondary predicate indicating the state of a participant (25c). (25) a.

i Ngardaly-kiji-rni ngula¼ji over-turn-npst that¼top ii warlu-ngka kuyu wardapi ﬁre-loc meat goanna

yangka kuja¼ka yapa-ngku anaph when¼prs person-erg yurnmi-karda cooked-transl

Depictives in English and Warlpiri

85

iii ngardaly-ngardaly-kiji-rni kuyu wardapi linji-kijaku. over.over-turn-npst meat goanna dry-lest ‘Ngardaly-kijirni is when a person keeps turning a goanna over on the ﬁre from one side to the other to cook it and to prevent it from becoming dried out [lest it become dry].’ [Warlpiri Dic.: ngardaly-kijirni: PPJ <9/86] b. Yulpu-karda janka-ja pirrjirdi-wangu-karda, yilara. soft-transl burn-pst hard-without-transl ﬂesh ‘It has cooked the meat soft so that the ﬂesh is no longer (nice and) ﬁrm.’ [Warlpiri Dic.: jurlkuly-jurlkuly-janka: krn] c. Yurrpa-rninja-rla ka¼rlipa nga-rni manu ka¼rlipa grind-nmlz-after prs¼122.sbj eat-npst and prs¼122.sbj parnti.nya-nyi mulyu-ngku miirnta-rlu ngurrju-karda. sniff-npst nose-erg cold-erg good-transl ‘We grind it up and eat it and we also inhale it through the nose when we have a cold to get better.’ [Warlpiri Dic.: munyuparntiparnti: jne] As well as there being no resultative/depictive complementarity in Warlpiri, there is also no complementarity with nonﬁnite verb complements. This is because Warlpiri makes little use of sentential or VP arguments (Hale 1982). However, Warlpiri does show a complementarity between resultative (change of state) and depictive (simultaneous state) in a different part of the grammar, namely the preverb–verb relation (Nash 1982). Preverbs are an open part of speech which combine with the closed class of verb roots to create verb meanings. Some preverbs express mostly change of state; others such as rdilypirr ‘holey, having a hole’ occur with different verbs to express simultaneous state (26a) or change of state (26b): (26) a. Kiwinyi-rli¼ji pu-ngu -- rdilypirr-karri mosquito-erg¼1.obj bite-pst holey-stand.npst ka¼rna¼rla marlaja. prs¼1.sbj¼dat cause ‘A mosquito bit me and I have a bite from it.’ [Hale 1959: 812] b. Kajipa¼npa¼rla rdilypirr-paka-karla kapanku ngapa ka if¼2.sbj¼dat hole-chop-irr quickly water prs juurl.pi-nyi¼rni. jump-npst¼hither ‘If you hit a hole in it [tree] quickly the water spurts hither.’ [Hale 1966: 0321]

86

J. Simpson

2.2.4 More than one secondary predicate In the previous section I suggested that resultatives and depictives were in approximate complementary distribution in English, but that in Warlpiri this complementary distribution does not hold for free nominals. However, in English apparent depictives can co-occur with both resultatives and depictives. I show here that the situations in which this can happen are quite constrained. For example, (27a–d) show that, marginally, a depictive and a resultative can appear, provided the depictive is outside the resultative. (27e, f) show that, marginally, two depictives can appear, provided the subject-oriented one is outside the object-oriented one. This has been used to argue for the close relation between resultatives and object-oriented depictives and the verb (Winkler 1997). (27g) contains two secondary predicates, both predicated of the same participant. (27h, i) show examples with a resultative and a depictive in the correct order. The ﬁrst is predicated of the object of shoot, the second of either the subject or the object. Both seem unacceptable. (27) a. They soon cook tender young [resultative, conditional] (Halliday 1967: 80) b. *They soon cook young tender [conditional, resultative] c. ??Hammering metal ﬂat hot makes sense [resultative, conditional] d. *Hammering metal hot ﬂat makes sense [conditional, resultative] e. ?Fred’s upset with me—me eating the meat raw drunk was the last straw. f. *?Fred’s upset with me—me eating the meat drunk raw was the last straw. g. I drink lemon juice neat fresh. [depictive, conditional] (Halliday 1967: 80) h. *After shooting the man dead asleep, they completed their paperwork. i. *After shooting the man dead drunk, they completed their paperwork. The more acceptable sentences are those where the event is interpreted as habitual, or generic in some way. This was noted by Halliday (1967: 78), who argued that in (27a, g) the second adjective is not a depictive but a ‘conditional attribute’ (‘that’s the condition under which I will drink it neat’), although he admitted that the distinction is sometimes not always clear. Thus depictives and resultatives are in complementary distribution with each other, but either can appear with a nominal representing a conditional attribute. The importance of the habitual or generic context is discussed in the next section.

Depictives in English and Warlpiri

87

Another context in which there appear to be two depictives is given in (28a), in which two adjectives predicated of the subject can appear on either side of a PP complement, in a context that is not generic. (28) a. After prancing naked into the ofﬁce drunk, John’s career came to a sudden end. b. ??After prancing into the ofﬁce naked drunk, John’s career came to a sudden end. The kind of relations the two adjectives naked and drunk have with prance and John seem similar, and it would be hard to argue that they are not both depictives. However, the lesser acceptability of (28b) suggests an ordering restriction against having both adjectives together. This in turn suggests the correctness of approaches involving building up a complex VP structure in which depictives may be attached at different levels of the VP: [[[prance naked] into the room] drunk]. While the appearance of adjectives as depictives in English is tightly constrained, this is not the case in Warlpiri. More than one predicative nominal may appear in the same clause, predicated of subject and object (29a) or predicated of the same participant (29b, c). However, as (29b) shows, the interpretation is vague between depictive, manner nominal and head, and restrictive and non-restrictive attribute. (29) a. Purra-mi ka¼rnalu warlu-ngka manu ka¼rnalu cook-npst prs¼111.sbj ﬁre-loc and prs¼111.sbj wanka nga-rni yarnunjuku-rlu¼ju raw eat-npst hungry-erg¼top ‘We cook it [bush banana] in the ﬁre and we eat it raw when hungry.’ [Warlpiri Dic: ngakaparla] b. Mirrimirri-rli ka miyi yangka muku-nga-rni yarnunjuku-rlu, ravenous-erg prs food anaph all-eat-npst hungry-erg ‘Ravenous he eats all the food hungry.’ or ‘The hungry one eats all the food ravenously.’ [Hale 1966: 0180] c. Nga-rni ka¼lu yapa-ngku yurnmi¼ji eat-npst prs¼333.sbj person-erg ripe¼top manya-nyayirni ngurrju - purra-nja-wangu. soft-very good cook-nmlz-without ‘People eat the ripe ones/them ripe as they are very soft and good without being cooked.’ [Warlpiri Dictionary: yajukurlu: knr] This free appearance of predicative nominals is a characteristic of adjuncts.

88

J. Simpson

2.2.5 The inﬂuence of modals and negation In this section I consider the importance of modals and negation in providing contexts which allow the inference of connections between secondary predicates and participants, other than connections through temporal coincidence. Such connections are essential in English for most instances of more than one secondary predicate denoting state. Halliday’s ‘conditional attributes’ introduced in the last section predicate a state of some participant, and assert the temporal coincidence of that state with the state/event represented by the main clause. The acceptability of these depends in large part on whether a nontemporal connection between that state and the main clause can be inferred.4 The interpretation of the connection depends on the tense, mood, and aspect of the main clause. For example, in (27a) the conditional reading is assisted by the simple present and the possibility of a generic or habitual interpretation. If these are changed, the sentence becomes less acceptable (30a, b). These can be made more acceptable by expanding the secondary predicate and making it a sentence-margin adjunct with its own nonﬁnite verb, prosodically distinct from the main clause (30c). It is tempting to say that these ‘conditional attributes’ are sentence-margin adjuncts, and indeed they mostly are. However, in that case they should be unacceptable inside VP phrases acting as arguments. But, provided the form of the verb allows the habitual interpretation, two secondary predicates can marginally appear inside a VP argument, as in (30d). (30) a. b. c. d.

*That chicken cooked tender young. *After cooking tender young his turkey was placed on a dish. ?That chicken cooked tender, being young. ?Drinking lemon juice neat fresh will cause you stomach problems.

The improvement in generic contexts of the acceptability of two secondary predicates is part of more general ways in which modality can inﬂuence the acceptability of depictives in English. The importance of generic or nomic 4 This property is discussed by Nichols (1981) for Russian ‘circumstantial predicate nominals’, which correspond to Halliday’s conditionals, except that gender and number agreement (and sometimes case) indicate what the controller is, as well as the construction type.

(i) Sladkij on neprijatnyj sweet: m.sg it:m.sg unpleasant:m.sg Sweet, it (tea) isn’t good/It isn’t good sweet.

(Nichols 1981: 40)

A circumstantial predicate nominal as in (i) asserts ‘the temporal coincidence or logical consequence of the two predications’ (Nichols 1981: 39). Thus (i) is an assertion that ‘the tea’s being unpleasant is a [potential] consequence of its being sweetened’ (Nichols 1981: 40). The similarities between English and Russian as to construction types are quite strong, despite the fact that English lacks case, gender and number agreement.

Depictives in English and Warlpiri

89

readings for allowing depictive readings was noted by Stump (1985) and Hale and Keyser (2001 and earlier versions). One way of creating such generic readings is by adding a modal and negation. This very often makes secondary predicates more acceptable. (31) a. He won’t drink orange juice neat fresh. b. You can’t cook chickens tender young. Other examples of forms with depictives improved by modality and negation include passives. In the past tense, regardless of whether the passive agent is present, depictives predicated of the agent are generally bad (32a–c, e, f), ﬁtting the pattern that it is hard to predicate depictives of objects of prepositions (Williams 1980). But once the modal can is added and negated, the sentences become much more acceptable (32d, h). (32) a. b. c. d. e. f. g. h.

?The song was sung drunk. ??The song was sung by them drunk. *The song was sung drunk by them. This song can’t be sung drunk. *Cartons of beer are often carried drunk. ?Cartons of beer are often carried drunk by them. ?Cartons of beer are often carried by them drunk. Cartons of beer can’t easily be carried drunk.

This suggests embedding of the VP structure inside a modal phrase, with a modal/negative as a higher predicate. (33) gives a possible structure. (33)

=

ModalP Secondary Predicate [conditional]

=

ModalP VP

=

Modal

VP Secondary Predicate (depictive) We have now seen in English object-oriented depictives appearing both next to the verb and following a PP argument, subject-oriented depictives appearing at the edge of the VP, and conditional attributes outside both subject- and object-oriented depictives. Clearly, the interaction with modality and negation deserves much further study. I do not have the data to discuss this in Warlpiri, but for English I will add observations in passing throughout this chapter of where English examples can be improved by changing negation and modality.

90

J. Simpson

2.2.6 Temporal and other modiﬁcation of secondary predicates In Warlpiri, many of the kinds of interpretation given by conditional attributes and sentence-margin adjuncts in English are provided by clitics which both constrain and add to the interpretation of the nominal. These include the temporal-aspectual clitics already discussed, as well as information structure clitics, and clitics which indicate concession, condition, counterexpectation, etc. Most of these clitics can attach to almost any part of speech, and their correct use is not well understood. In (34) the Dative argument of the main predicator, the nominal lijija, is ‘the Nangalas’. Without the clitic¼jala, this sentence could have been interpreted as having an appositional noun phrase expressing the Dative argument, ‘his mothers-in-law, the Nangalas’, or ‘the Nangalas, his mothersin-law’. The clitic¼jala forces makurnta-warnu-ku to be interpreted as a secondary predicate, and contrasts this information with the information in the main clause that Jakamarra is lusting after these women whom he is forbidden to approach because they are his in-laws. Nangala-warnu-ku (34) Jakamarra yalumpu¼ju lijija subsection that=top covetous subsection-assoc-dat makurnta-warnu-ku¼jala. in-law.assoc-dat¼contrast ‘That Jakamarra is always after all the Nangalas even though they are his mothers-in-law in fact [whom he should avoid].’ [Warlpiri Dic.: lijija] Temporal aspectual clitics also set up contexts in which pragmatic inferences about nontemporal connections are readily made. Thus (35a, b) both contain the temporal clitic ¼lku indicating a change of state. One is best translated with a concession ‘even though’, and the other with a reason ‘because’. (35) a. Parnka-ya¼rni¼ji¼li¼rla yapa. Ngayi¼lki run-imp¼hither¼top¼pl.sbj¼dat person Just¼then ka paka-rni warrarda Japaljarri-rli prs hit-npst always Japaljarri-erg Nakamarra¼ji wata¼lku. Nakamarra¼top unconscious¼then ‘You people come here quickly to help her. Now Japaljarri is still hitting Nakamarra (even though she is) now unconscious.’ [Warlpiri Dic.: wardarrarra: PPJ 10/87]

Depictives in English and Warlpiri

91

b. Yurrpu-rnu¼rna¼ngku waja, pala-pala-rlu¼lku leave-pst¼1.sbj¼2.obj indeed tired-tired-erg¼then wardiji-rla kanunju waja mulga-loc under indeed ‘I have left you here under the mulga trees (because I am) now exhausted!’ [Ngarlinjiya] Another important determinant of pragmatic inferences is word-order position, since sentence margin and pre-Auxiliary positions are used in Warlpiri to indicate discourse prominence (Swartz 1991; Laughren 2002). Winkler (1997) and Himmelmann and Schultze-Berndt (Ch. 1, this volume) note the importance of information structure information, in particular focus, for determining interpretations of secondary predicates. In (36) the initial position and the question focus clitic¼nya ensure that the focus of the question is on the secondary predicate. nguna? (36) Yarnunjuku¼nya ka¼npa hungry¼qfoc prs¼2.sbj lie.npst maarr-nguna. Yuwayi, yarnunjuku ka¼rna yes hungry prs¼1.sbj listless-lie.npst ‘Are you lying down HUNGRY?’ ‘Yes. Hungry, I am lying down listless.’ [Hale 1966: tape 2.32, 0208] But even without the addition of clitics, in (37) the nominal predicate is readily susceptible to the inference of nontemporal connections, because it is in a discourse prominent position immediately preceding the Auxiliary kajipangkulu. The translation is Hale’s. (37) Ngaka¼lku¼ju yawarra-kurlu-rlu kajipa¼ngku¼lu later¼then¼top wound-having-erg if¼2.obj¼333.sbj pangi-karla¼wiyi? gouge-irr¼before ‘They will spear you ﬁrst, though they have spear wounds?’ [Hale 1959: 7.220] The interaction between context on the one hand and discourse-prominent positions, discourse clitics, and temporal-aspectual clitics on the other allows for a wide range of pragmatic inferences as to the connection between the secondary predicate, the participant, and the main predicator. 2.2.7 Summary English makes use of particular positions in the VP for depictive, resultative, and nonﬁnite VP complements. Adjuncts modifying the Subject, Object and

92

J. Simpson

clause appear at sentence margins, with strong (detached) interpretations. Depictive, resultative, and nonﬁnite VP complements can all be predicated of Subjects or Object, but for resultatives and nonﬁnite VP complements, the controller is determined by the lexical entry for the verb (thus persuade requires object control, while try requires subject control; transitive verbs require resultatives to be predicated of their objects). By contrast, depictives can be predicated of subjects or objects, and what an adjunct modiﬁes is more loosely determined. Whether the controller of depictives depends on the verb will be discussed in Section 2.3.3. Warlpiri makes much use of secondary predicates denoting state. States resulting from the action denoted by the verb (the equivalents of resultatives), however, are formed with case or complementizer sufﬁxes, and do not exist as free predicate nominals in opposition to depictives. Agreement determines what a secondary predicate is predicated of. However, agreement is also used on a wide range of other nominals, from restrictive and non-restrictive attributes to predicates denoting location in space and time, duration, manner, and body part. These range from the clearly participant-oriented to those like time locatives which apply over the whole event. Temporal-aspectual clitics can specify whether the state precedes, continues during, or starts close to the event denoted by the main predicators. Non-temporal interpretations are determined in part pragmatically, in part by word order and in part by the attachment of clitics, which can by and large attach to elements other than secondary predicates. There seems to be no obvious morphosyntactic evidence distinguishing the depictive use of secondary predicates in Warlpiri from uses like those of sentence-margin adjuncts in English, or indeed from attributive modiﬁers.

2.3 Constraints on depictives in English compared with Warlpiri Himmelmann and Schultze-Berndt (Ch. 1, this volume) list ﬁve parameters playing a role in classifying participant-oriented adjuncts, which include depictives. Two have already been discussed in the previous section, as part of showing how depictives ﬁt into the grammar of English and Warlpiri. These were the restrictions on the syntactic position of the depictive and its morphological marking. The discussion so far has indicated that depictives in Warlpiri have more adjunct-like properties than depictives in English. In this section I compare English and Warlpiri with respect to the other three parameters: what can act as a depictive, what a depictive can be predicated of, and what verbs the depictive can appear with.

Depictives in English and Warlpiri

93

2.3.1 What can act as a depictive In English, the kinds of adjective or nominal or prepositional phrase that can act as a depictive or resultative are semantically restricted. Adjectives that can appear as depictives or resultatives are those that denote a transitory, bounded, temporally restricted state, so-called stage-level predicates.5 Thus (38a, b, c) are not acceptable because the adjectives denote more permanent properties, ‘individual-level’ predicates. (38d) is acceptable because intelligently is an adverb. (39a, b) are acceptable, because they are detached, sentence-margin adjuncts, interpreted in this case in apposition to what they are predicated of. (38) a. *I ate the meat organic. b. He arrived *intelligent/*knowing/?knowledgeable c. He read the book *intelligent/*knowing/*knowledgeable (unacceptable with either resultative or depictive reading) d. He read the book intelligently/knowingly/knowledgeably. (39) a. ?I ate the meat, organic and nutritious b. The new secretary arrived, eager, intelligent and ﬁring on all cylinders. In Warlpiri, there is no obvious evidence for the syntactic relevance of the stage-level versus individual-level distinction,6 because nominals that denote individual-level properties can usually be interpreted as attributive to an understood head, ‘the lame one, the speckled one’, or as appositional, as in (40a). There are examples showing nominals that appear to denote individual-level predicates being used as secondary predicates in ways which seem foreign to an English speaker. For example, there is a nominal pina which means roughly ‘knowing, knowledgeable, knowledgeably’, and is often used as a secondary predicate. Sufﬁxed with the translative sufﬁx -karda it can denote a change of state, like a resultative, as in (40b). ngapa-ku (40) a. Pararri ngula¼ji yangka kuja¼ka¼rla rainbow that¼top anaph which¼prs¼dat water-dat jaa(r)l-karri nguru-ngka mawulpari-mawulpari block-stand.npst sky-loc banded ‘A rainbow is that which, banded with colour, blocks the rain in the sky.’ [Warlpiri Dic.: pararri: PPJ <9/86, translation JHS] 5 There are many labels given to this distinction: a temporary state of affairs (Bolinger 1971), a temporally restricted state (Dowty 1972), a ‘stage-level predicate’ (Carlson 1977; Stump 1985; Winkler 1997), a ‘relational predicate’ (Ogawa 2001), a ‘bounded predicate’ (Wyngaerd 2001). Nichols (1981: 122) describes similar constraints in Russian for ‘bound co-predicates’ (depictive equivalents). 6 However, see Laughren (in press) for an argument that the Object-control complementizer -kurra selects a dynamic stage-level predicate, or else forces a stage-level reading on the nominal to which it attaches.

94

J. Simpson b. Nya-nyi ka¼lu pina-karda see-npst prs¼333.sbj knowledgeable-transl ‘They look at it and become knowledgeable by so doing.’ [Hale 1966: 1567]

The textlet in (41) gives further examples of the use of pina, with the perception verb purdanyanyi in (iv). (41)

nyina-mi pina yangka i. Yi¼ka¼lu-cause¼prs¼333.sbj sit-npst knowing anaph ii. --yi¼ka¼lu¼jana ngarrka-patu¼wiyi cause¼prs¼333.sbj¼333.obj initiated.men-pl¼before purda.nya-nyi, listen.to-npst iii. kuja¼ka¼lu¼nyanu ngarri-ni ‘yajini’, thus¼prs¼333.sbj¼refl tell-npst ‘yajini’ ngarrka-patu-rlu¼wiyi, men-pl-erg¼before kurdu-ngku¼ju yangka purda.nya-nyi iv. ngula¼lu¼jana that¼333.sbj¼333.obj child-erg¼top anaph listen-npst pina-ngku¼jala. v. [ka¼lu¼jana]7 [prs¼333.sbj¼333.obj] knowing-erg¼contrast vi. Kaji ngarrka-jarri¼lki, If man-become.npst¼then ngarri-ni -- kuja-rlu¼yijala. vii. ngula ka pina-ngku¼ju that prs knowing-erg¼top tell-npst thus-erg¼also viii. Ka¼jana yangka kurdu-ngku¼wiyi purda.nya-nyi. prs¼333.obj anaph child-erg¼before hear-npst i. ‘They learn [are knowledgeable] ii. because they listen to initiated men iii. when they call each other ‘yajini’ [‘you cross-cousin’, term used by adult men], ﬁrst. iv. They listen to them as children, [or: The children listen to them] v. and they really know. vi. When such a one becomes an initiated man vii. then he uses it [that word]knowledgeably, like that also. [or: The knowledgeable one uses it like that also] viii. He hears them ﬁrst as a child.’ [Hale 1966: 0358; Warlpiri Dic.: yajini] 7

The material in square brackets is omitted in the Warlpiri Dictionary entry.

Depictives in English and Warlpiri

95

In (41i), pina appears with the stance verb nyina ‘sit’ functioning as a semicopula, acting as the complement of the copula: ‘they are knowledgeable’. In (41iv, v) pina appears with Ergative case, apparently agreeing with the subject of a perception verb ‘listen, hear’: ‘listen being/becoming knowledgeable’—a mixture of a depictive and resultative meaning. In (41v) pina appears with Ergative case agreeing with the subject of a verb of speaking, and can be translated as a manner adverb ‘knowledgeably’, or conceivably as an attributive modiﬁer of an understood head. The remaining two lines show two other nominals agreeing with the subject—one, kuja-rlu ‘thus’, is used to show how the process happens (linking back to the whole event of learning by listening), and the other ‘child-erg-before’ has a temporal interpretation ‘as a child’ used with a perception verb. As this discussion shows, in Warlpiri a nominal agreeing in case with an argument of the sentence can be translated in a variety of ways, depending on what inferences (other than temporal coincidence) are made about the connection between the secondary predicate, the process, and the participant. Much more work needs to be done to determine the role that the individual vs. stage-level nature of predicates plays in the formation of these inferences. 2.3.2 What a depictive can be predicated of In English a depictive is normally predicated of a subject, or an object, as we have seen. However, there are some differences in acceptability. We have already seen that a depictive predicated of a subject must appear outside a depictive predicated of the object (27e, f repeated here). (42) a. ??Fred’s upset with me—me eating the meat raw drunk was the last straw. b. *?Fred’s upset with me—me eating the meat drunk raw was the last straw. (43a, b) show that a subject-controlled depictive can occur inside the clefted part of a cleft construction, while (43c, d) show that an object-controlled depictive cannot occur inside the clefted part of a cleft construction, presumably because the object is not overtly present in the cleft. (43d) is blocked by the restriction against control by a prepositional object. (43) a. b. c. d.

What John did naked was run down the street. What John did drunk was eat the ﬁsh. *What John did raw was eat the ﬁsh. ??What John did to the ﬁsh raw was eat it.

96

J. Simpson

I turn now to a subclass of objects, Dative objects of ditransitive verbs. While these share many properties with regular objects (passivizability, for example), they do not allow depictives. Depictives in such sentences are predicated of the secondary object (i.e. the Theme in 44a, b), but not the Dative object (the Recipient/Goal in 44c) (examples from Hale and Keyser 2001). It is hard to argue for a semantic restriction, as Koizumi (1994: 48) notes, because the passive of such sentences is acceptable (44e, f). (44) a. b. c. d. e. f.

(Hale and Keyser 2001) I gave the baby its bottle full. (Hale and Keyser 2001) I handed the mother her baby crying. *I gave the baby the bottle crying. (Hale and Keyser 2001) (Koizumi’s 64a) *They gave the patients the drugs drunk. The patients were given the drugs drunk. (Koizumi’s 64b) After being given the drugs drunk, the patients complained.

However, if the verb and the secondary object form a complex predicate, then depictives predicated of the Recipient/Goal are marginally more acceptable. (45)

??Giving him the injection unconscious was sensible.

Depictives predicated of the Recipient/Goal are also made more acceptable by changing the polarity and modality to force a generic conditional interpretation. (46) a. ?You can’t give them bottles crying. b. You can’t give them injections unconscious. The situation in Warlpiri is quite different. Dative arguments of the verb can control secondary predicates, whether the verb is a two-place predicate (47a, b), or a three-place predicate (48). rdipi-ja nyurnu-ku. (47) a. Ngula-jangka¼ju¼lu¼rla that-from¼top¼333.sbj¼dat meet-pst dead-dat ‘Later on they encountered him dead.’ (Darby Jampijinpa 1984: 26) b. Yarnunjuku-ku mayi ka¼ngku ngurlkurr-ngurlkurr-wanti? hungry-dat question prs¼2.obj gulp-gulp-fall.npst ‘Are you hungry that it (food) is going down you in a gulp?’ [Hale 1966: 0377] (48)

yu-ngu, Pina-wangu-ku¼ju¼nganpa knowing-without-dat¼top¼111.obj give-pst wali¼lpa¼rnalu puta paja-rnu. well¼pst¼111.sbj half taste-pst ‘He gave some to us unknowingly [i.e. we didn’t know what it was]. And we tasted a little.’ [Toby Japangardi 1988]

Depictives in English and Warlpiri

97

In English, depictives predicated of arguments of prepositions generally are judged unacceptable (cf. Halliday 1967) as in (49a, b). Depictives predicated of Recipient/Goals expressed as to prepositional phrases are similarly unacceptable (49c, d, f). (Williams 1980: 2b) *John loaded the wagon with hay green. (Williams 1980: 3b) *I presented John with it dead. ??They gave the award to him naked/in his best clothes. ?Presenting awards to them in their shabbiest clothes seemed sensible. e. I gave the bottle to the baby full. f. *I gave the bottle to her crying.

(49) a. b. c. d.

Again, depictives predicated of arguments of prepositions are marginally more acceptable if the verb is complex (50a, b) or the prepositional argument can be reanalysed as the object of the verb (50c, d), or if modals and negation are added (50f): (50) a. ??Giving the injection to him unconscious was sensible. b. Giving birth to their young live is a characteristic of skinks. c. ?They looked at him unconscious. ?Looking at him unconscious aroused their pity. d. ?They operated on her asleep. ?Operating on her asleep was the only way. e. ?They did it to him asleep. ?Doing it to him asleep was unfair. f. ?They won’t give it to him naked/in rags. [he is naked/in rags] The more a preposition phrase acts as an adjunct, the harder it is to have a depictive predicated of the prepositional argument. (51) a. ??They did it for him asleep. b. ??They danced on him sedated. The situation in Warlpiri is less clear. It appears that Warlpiri lacks constraints on predicating states of participants in a clause. Laughren (1992: 239) writes: A stative ‘when’ predicate, as opposed to a Perception predicate, is not selected by the verb as part of its L(exical) C(onceptual) S(tructure), and its logical subject is not restricted to any syntactic relation with respect to the matrix verb. A ‘when’ predicate is associated with a temporal value which modiﬁes or restricts the temporal value associated with the matrix predicate.

However, the majority of secondary predicates resembling depictives that I have found have been predicated of arguments of the verb, not of arguments

98

J. Simpson

of adjuncts. That is, very few second predicates (other than restrictive or appositional attributes), are predicated of the arguments of the major semantic case-sufﬁxes Locative, Ablative, or Allative which are commonly found in adjuncts indicating place, source and goal. To illustrate this, of 359 instances of wankaru ‘alive’ found in a search of the Warlpiri corpus of texts, only one had the Locative ending on it (none had the Ablative or Allative endings), and it was probably a complementizer use in an absolute construction. (52)

Ngula-rra¼ju yapa yangka-rra-piya kuja¼lpa¼lu that-pl¼top person anaph-pl-like thus¼pst¼333.sbj nyurruwiyi nyina-ja Nuwa-ngka wankaru-rla¼wiyi. old.days sat-pst Noah-loc alive-loc¼before ‘People were like that in the old days when Noah was alive before.’ [‘And as it was in the days of Noah’, Luke 17: 26] [Swartz]

In (52) I construe Nuwa-ngka wankaru-rla as a phrase forming an absolute construction; a nominal predicate wankaru acts as an argument of a sufﬁx denoting a predicate, the Locative ‘when’, which itself has a temporalaspectual clitic added. That nominal predicate in turn has an overt subject, marked with the Locative in agreement: Nuwa-ngka. The whole phrase relates the time of the event ‘people were like that’ to a time before the time of speaking ‘before now when Noah was alive’. Schultze-Berndt and Himmelmann (2004) exclude such absolute constructions from their deﬁnition of depictives. The only structural difference between this phrase and the other secondary predicates we have been looking at is that the subject Nuwangka is overt, and is not construable on its own as an argument or adjunct of the main clause. Of course, there are many examples of two or more predicative nominals with the same semantic case-marking in the one clause, but these are often coordinated predicates, or else they may be restrictive or appositional modiﬁers, as in (53). (53) a. Nyarrpara-ngurlu ka nyampu¼ju jirrmily-karli-mi where-from prs this¼top ooze-ﬂow-npst rurrpa-ngurlu? hole-from ‘What opening is it oozing out from?’ [Warlpiri Dic: jirrmilypa] jirrmily-karli-mi b. Milpa-ngurlu ka eye-from prs ooze-ﬂow-npst ‘It is oozing out from the sore eyes’.

wijini-ngirli. sore-from [Warlpiri Dic: jirrmilypa]

Depictives in English and Warlpiri

99

(53a) is best treated as having a semantic case sufﬁx -ngurlu ‘from’ which takes as argument a single merged functional constituent ‘what opening’, but the boldfaced constituents in (53b) could be translated as restrictive attributes ‘from the sore eyes’, or as appositional predicates ‘from the eyes, the sore ones’. (54) shows the same ambiguity of interpretation between restrictive and apposition ‘from there, in the distance’, ‘from there in the distance’. (54) Nyurrurla¼rna¼nyarra yinya-ngurlu nya-ngu¼rnu you.pl¼1.sbj¼222.obj that.rem-from see-pst¼hither wurnturu-ngurlu. far-from ‘I saw you from back there in the distance as I came.’ [Hale 1966: 0329] The ﬂexibility of interpretation of secondary predicates seen in (53) and (54) reinforces the point made in the previous section, that it is very difﬁcult to ﬁnd morphosyntactic evidence distinguishing depictives from other uses of nominals in Warlpiri.

2.3.3 Semantic class of verb The last point to discuss is the relation of the depictive predicate to the verb class. Several authors (e.g. Bresnan 1982) have noted semantic restrictions in English on which verbs depictive predicates may appear with. Nichols (1981) noted that in Russian the lexicosemantic class of the verb is important for interpreting secondary predicates. Most of the lexicosemantic verb classes that Nichols identiﬁes for ‘bound co-predicates’ (which include depictives) are similar to those where depictives are most commonly found in English: verbs of motion or position (Nichols gives Russian examples translated as ‘Many students go barefoot’, ‘The train arrived empty’, ‘The plates lay in the cupboard wet’, ‘He sat happy’), verbs of existence, state, change of state (‘He was born blind’, ‘He lived alone’); verbs of ﬁnding and meeting (‘They found him dead’, ‘I met him on the road drunk’); verbs of perception (‘I didn’t see one of them drunk’); verbs of remembering (‘I remember you as tiny’). More recently, the aspectual classes of verbs appearing with depictive predicates have been discussed. Both Winkler (1997) and Rapoport (1999) make a distinction between the behaviour of secondary predicates with stative verbs and the behaviour of secondary predicates with the other three of Vendler’s four verb types: achievements, activities, and accomplishments (Vendler 1967). I will consider the evidence for depictives with each aspectual

100

J. Simpson

type in turn, looking at the constraints in English and comparing them with Warlpiri data. The Warlpiri comparison is quite limited, because I rely here on data in the Warlpiri corpus, and not on elicitation, and because we do not yet have an adequate account of the interaction of temporal-aspectual clitics, verb class, prosodic unit, and word order in Warlpiri. 2.3.3.1 States In (55) the verbs all denote enduring states and appear with secondary predicates which resemble depictives in meaning. Winkler (1997) and Rapoport (1999) give several reasons for saying that the sentences in (55) contain secondary predicates predicated of objects which are not true depictives. First, the secondary predicates need not be stage-level predicates (55b). Second, the secondary predicate is very often not an adjunct but a complement. Together with the object it indicates what the subject likes/ wants/prefers. This is shown by the fact that, as Rapoport asserts, the object and the secondary predicate can move together (55c), whereas depictives with other verb types cannot (55d). There are also differences in entailment. (55e) with a state verb, like, is acceptable, while (55f) with a non-state verb is not acceptable. (Rapoport 1999: 654) Jones prefers her coffee black. She likes/prefers/wants/needs her men intelligent. ?Her coffee black is what Jones prefers. (Rapoport 1999: 654) She cut the bread hot. *The bread hot is what she cut. (Rapoport 1999: 654) e. She liked her food hot, but she did not like her food. f. *She cut the bread hot but she did not cut the bread.

(55) a. b. c. d.

Finding secondary predicates predicated of the subjects of statives is not easy, because there is a conﬂict set up between the temporary state involved in a depictive and the enduring state involved in many statives. However, modal and generic contexts improve the readings, and allow a depictive to appear following a secondary predicate complement (56e). (56) a. ?I knew the answer sober, but now that I’m drunk I can’t remember it. b. ???I knew the answer young [better: I knew the answer when young] c. ? John owned a car quite young [better: John owned a car when quite young, or he was quite young to own a car] d. *Jake owns chickens young (Winkler 1997: 6) e. I’ve never seen him sober awake.

Depictives in English and Warlpiri

101

But if the state denoted by the verb is stage-level, and not an enduring individual-level state, then depictives are possible: (57) a. b. c. d.

Many linguists were available drunk. (temporary) Many linguists were intelligible drunk. (temporary) Many linguists sounded intelligent drunk. (temporary) ?*Many linguists were intelligent drunk. (enduring)

Examples with statives sound better when the presence of a comparative allows the state to be interpreted as temporary: (58) a. I look better naked. (slogan on T-shirt popular in the US in 2001) b. Babies are better heavier. The greater acceptability of stative verbs with secondary predicates that denote temporary events is generalized by Winkler (1997: 366–9) as a claim that both the verb and the secondary predicate must be of the same aspectual type—both stage-level (temporary) or both individual-level (enduring). Thus The letter arrived torn (Winkler’s 87b) has two stage-level predicates (arrived and torn), while He was born defective, He likes women big each have two individual-level predicates. Like Rapoport, Winkler considers the examples with individual-level predicates not to be depictive constructions at all, but to be a related construction. In Warlpiri it is hard to ﬁnd similar clashes of state verbs and stage- or individual-level secondary predicates. Warlpiri has many stative sentences, where states are predicated of a subject either by directly using a nominal or by having a stance verb as well. Several nominal predicates can occur in such sentences (59). The temporal-aspectual clitics allow for resolving conﬂicts over enduring states versus temporary states. (59)

Rdarri ka nyina wajirrki-wajirrki ¼wiyi, raw prs sit.npst green-green¼before parru¼ju, murntu¼ju ka nyina maru¼lku, unripe¼top, ripe¼top prs sit.npst black¼then ‘They are raw, still green, the unripe ones, the ripe ones are black then.’ [Hale 1966: 1644]

Because there is no obligatory copula it is often hard to determine whether the predicates are working in coordination or whether one is primary, as in (60a): (60) a. Ngawu-ngawu lirrjiri yurlpu-wangu-jangka, bad really.dry ochre-without-from kurdiji nyampu¼ju. shield this¼top ‘This shield is in bad shape, all dry and brittle from having no red ochre.’ [Warlpiri Dic.: lapa: PPJ 10/85]

102

J. Simpson b. Lirrjiri nyampu = ju karli, dry this¼top booomerang, mapa-rninja-wangu-jangka ngawu-ngawu Japaljarri. rub-nmlz-without-from bad subsection name ‘This boomerang is dry; it is bad as it hasn’t been oiled, Japaljarri.’ [Warlpiri Dic.: lirrjiri: PPJ 10/85]

As for stative verbs, it is difﬁcult to ﬁnd test examples, because in Warlpiri the obvious candidates, such as verbs of perception, often have non-stative readings (e.g. ‘obtain by seeing/smelling’). (41) showed examples of depictives predicated of the subjects of such verbs in what appear to be stative uses. (61) shows examples of depictives predicated of the objects of such verbs. (61) a. Yirnmi¼lki kaji¼ka¼npa parnti.nya-nyi nganayi = ji. ripe¼then if¼prs¼2.sbj smell-npst whatsname¼top Yipintiri = ji. berry.sp.¼top ‘You can smell it when it’s ripe that whatsname, the yipintiri berry.’ [Hale 1966: 0288] b. Kaji¼lpa¼npa karla-karla mulju¼rlangu, kaji¼ka¼npa if¼pst¼2.sbj dig-dig.npst soakage¼for.example if¼prs¼2.sbj nya-nyi liirlpari - ngapa-wangu, lawa. see-npst white water-without nothing ‘If you dig a soakage, say, you might see that it is white, without water, empty.’ [Warlpiri Dic.: liirlpari: c:darby] (61a) has a reading like a depictive, while (61b) has a reading more like the English complement reading (what the speaker sees is ‘the soakage being white, waterless’). See Laughren (1992; in press) for more discussion. Note that the English counterpart of (61a) with a depictive is unacceptable: You can smell the berry ripe now. It must have an explicit temporal conjunction when, as in the translation. It is likely that the Warlpiri depictive reading is also achieved through the temporal-aspectual clitic ¼lki, as Laughren (1992: 231) claims. In sum, secondary predicates denoting state can be predicated of subjects of stative verbs and nominals acting as the main predicate in Warlpiri. However, the restrictions on predicating states of the objects of perception verbs observed by Laughren resemble those in English, and probably derive from general principles of linking events with different inherent aspects. 2.3.3.2 Activities Rapoport (1999) observed that English activity verbs allow depictives to be predicated of their subjects but not their objects. Secondary

Depictives in English and Warlpiri

103

predicates following activity verbs and predicated of the object are interpreted as resultatives if possible, with a consequent change from activity to accomplishment or achievement. (62) a. b. c. d. e.

Jones phoned Smith sad Jones pushed Smith sick Jones chased Smith angry Jones slapped Smith sober I kicked John depressed

f. ?*John hit Tom naked

*Jones phoned Smith sad. *Jones pushed Smith sick. *Jones chased Smith angry. *Jones slapped Smith sober. *I kicked John depressed. (Rapoport’s ex. 3, 4a–f) ?*Mary praised the professor drunk (Koizumi’s 111a, b).

In Warlpiri, both subjects and objects of verbs of impact and concussion can have secondary predicates predicated of them as in (63). (63) a. Ngayi¼lki ka¼npa¼ji paka-rni yakayaka¼lku, just¼then prs¼2.sbj¼1.obj hit-npst hurting¼then mikinpa¼lku, murrumurru¼lku. injured¼then sore¼then ‘You just keep on hitting me (even though I’m) sore all over and injured and in pain.’ [Warlpiri Dic.: yakarra: PPJ 10/87] b. Wati-ngki¼nyanu paka-rnu mata-ngku¼lku. man-erg¼refl hit-pst tired-erg¼then ‘The man hit himself, tired by then.’ (Laughren 1992: ex.11c) Here Warlpiri is more ﬂexible than English in allowing depictives to modify both subjects and objects, but it remains to be studied whether this is due to the presence of temporal-aspect clitics, as is probably the case for acceptable depictives predicated of the objects of perception verbs. 2.3.3.3 Achievements and accomplishments Rapoport (1999) observed a difference in behaviour between transitive causative achievements, which allow depictives predicated of the object (64a, c), and intransitive achievements, which do not allow depictives predicated of their subject even though that subject has the same semantic role as the object in the transitive clauses (64b, d). The tendency in English is to interpret the depictive as a resultative predicated of the subject. (63) a. b. c. d.

I fried the potatoes raw. * The potatoes fried raw. I boiled the lobster alive. *The lobster boiled alive.

(Rapoport’s (Rapoport’s (Rapoport’s (Rapoport’s

2a) 9a) 2c) 9c)

104

J. Simpson

Once again, modals and negation improve the clauses, as conditional readings become easier to obtain: (65) a. This juice won’t freeze fresh. b. The potatoes won’t fry raw. c. They become dangerous hungry. example.)

(Thanks to the editors for this

Warlpiri has lots of intransitive achievements formed by compounding nominals denoting the state with the verbaliser -jarrimi ‘become’. The interpretation of secondary predicates with these varies. Some resemble depictive predicates, ‘as they are hungry’ in (66a), while others resemble resultative predicates (describing the state achieved, ‘ripe’, by means of a temporalaspect clitic¼lki in 66b). (66b) suggests that secondary predicates occurring with achievement verbs can be interpreted as modifying the end state. nganyinganyi-jarri wiyarrpa. (66) a. Yarnunjuku ka¼lu hungry prs¼333.sbj impatient-become.npst poor.thing ‘They are getting impatient as they are hungry, poor things.’ [Hale 1966: 0311] b. Marnikiji ka parru-jangka¼ju berry prs unripe-from¼top maru-rra-jarri-nja-ya-ni¼jiki¼jala yirnmi¼lki¼ji. black-off-become-nmlz-go-prs¼still¼contrast ripe¼then¼top ‘Conkerberries after being unripe turn black still by contrast, ripe then.’ [Hale 1966: 0205] To conclude, English places restrictions on the co-occurrence of depictives with the arguments of verbs of different aspectual classes. While the investigation of Warlpiri data in this light has barely begun, it seems likely that there are fewer restrictions, in part because the temporal-aspectual clitics allow independent speciﬁcation of the relation of the state denoted by the secondary predicate to that denoted by the main predicate.

2.4 Conclusion Comparing depictives in English with secondary predicates in Warlpiri has revealed that the Warlpiri secondary predicates have much more ﬂexible word order, and they also have much more ﬂexible interpretations, in part due to the independent speciﬁcation of time and event structure made possible by the temporal-aspectual clitics, as well as the discourse clitics, and the use of word order for information structure purposes. There are fewer

Depictives in English and Warlpiri

105

constraints on what can be a depictive predicate in Warlpiri, and also fewer constraints on what a depictive predicate can be predicated of (secondary predicates are freely predicated of Dative arguments). In fact, there seems little reason to distinguish a class of depictive predicates separate from other secondary predicates which assert the overlap in time between a state and the state or event denoted by the main predicate. This leads us to consider whether English depictives are truly part of the adjunct system of English, since they are so much more constrained than their Warlpiri counterparts with similar meanings. It has usually been assumed that depictives are adjuncts in English, although Halliday (1967) implies that they are participants. Debate has revolved around whether depictives and resultatives are both adjuncts (Jespersen 1965; Stowell 1981; Schein 1995). This has the advantage of capturing the optionality of both depictives and resultatives, and also the similarity between depictives and resultatives. Alternatively, it has been argued that resultatives are complements (or parts of complex verbs) and depictives are adjuncts (Chomsky 1985; Green 1973; Dowty 1979; Carrier Duncan and Randall 1992; Goldberg 1995; Winkler 1997). This approach loses the parallelism between resultatives and other nonﬁnite complements, and depictives. A third possibility is to say that both depictives and resultatives are arguments, optional subcategorizing predicate complements (Bresnan 1982 and, implicitly, Quirk et al. 1985), which, like other subcategorizing complements, have constraints on what can be their controller. This has the advantage of capturing the approximate complementary distribution with other complements, and goes part-way towards explaining the similarity between depictives and resultatives. However, it provides no explanation for the greater freedom of appearance of depictives. The case for calling English depictives adjuncts in contrast with resultatives rests on several claims: they are always optional; semantically they resemble manner adverbs; they can be predicated of either subjects or objects in transitive verbs; they are apparently less restricted than resultatives as to which verb classes they can appear with (e.g. Green 1970: 278–9); more than one depictive can appear. This case has been weakened by the evidence summarized in this paper. In English, stative secondary predicates predicated of subjects of transitive verbs and objects have different properties, and both in turn differ from secondary predicates predicated in modal and generic contexts. Moreover what depictives can be predicated of is limited by the aspectual class of the verb. As Rapoport (1999) and Winkler (1997) have shown, depictives, not only resultatives, show restrictions according to lexicosemantic class of verb.

106

J. Simpson

Once these are factored in, the counterexamples to the complementary distribution of depictives and resultatives appear less compelling. Depictives and resultatives are then seen as part of an intricate system of building predicate meaning in the verb phrase, which is distinct from the sentence margin adjunct system. Differentiating the one as adjunct and the other as argument appears then to be unmotivated.

3 Adverbials and depictives as restrictors ¨ LLER-BARDEY THOMAS MU

3.1 Introduction This chapter proposes a model to characterize depictives and different classes of adverbials in terms of their capability to be restrictive in quantiﬁcational relations. In essence, we apply to adverbials the framework which Diesing (1992) and others have developed for indeﬁnites, and make a case for analysing depictives as event predicates adjoined to a higher VP level. Consider the following example from Stump (1985: 99): (1) Drunk, John drives very dangerously. Here, driving dangerously is related to a multitude of situations, hence some sort of quantiﬁcation obtains. We can conceive of this as brought about by a covert habitual or generic operator residing in the Simple Present. However, this quantiﬁcation is overtly restricted. It is the depictive drunk which brings this restriction about. Thus, the sentence is interpreted as ‘When he is drunk, he drives very dangerously’. Note that word order plays a role here. A depictive or an adverbial in postverbal position is much less likely to function as a condition for habituality: (2) These days, John drives home drunk. Here we are not talking about what happens when John is drunk. Rather (2) states that a situation of John driving drunk holds habitually.

This chapter has beneﬁted from comments by the editors of this volume and an anonymous reviewer. I want to express my gratitude to them.

108

T. Mu¨ller-Bardey

The same variation can also be observed with adverbials: (3) a. On the balcony, we sweat. b. We sweat on the balcony. With this example, the preposed adverbial, just like the depictive in (1), can provide a condition for a recurrent action in (a), but not in (b). We take these observations to be crucial for the analysis of depictives and adverbials. In developing our analysis, we draw a parallel to the model which, among others, Diesing (1992) has developed with respect to indeﬁnite nominals. Our ﬁrst claim, laid out in section 3.2, is that this model can be extended to adverbials and depictives. The result of this move is that an adverbial such as on the balcony in (3) is appropriately construed as a predicate of events, disposing of a variable. By topicalizing the adverbial, it becomes associated with the restrictive range of a quantiﬁer, the latter binding the variable. Thus, the event variable associated with the verb and that of the adverbial are linked: for any relevant event (of being) on the balcony, there is an event of sweating. Applied to depictives, such an approach would aptly reﬂect the dual nature of the depictive: it has two variables, an individual variable, linked to the subject (John in (1)), as well as an event variable.1 A second claim emerging from our investigation relates to the observation that some types of adverbial, manner adverbs in particular, resist restrictor formation. (4) a. Quietly, the prisoner talks to himself. b. The prisoner talks to himself quietly. With (4a), an interpretation conveying ‘When he acts quietly, the prisoner talks to himself’ (analogous to (1) and (3a)) is much less readily available. Among adjective-based adverbials, this seems to distinguish manner adverbs from depictives. We will propose an explanation for this difference in section 3.3, cast in both semantic and syntactic terms. Semantically, manner adverbs will emerge as referring to processes, a notion subordinate to events. 1 This is not to say that both events—the one that the verb qualiﬁes and that of the depictive—are identiﬁed by necessity. In (i), for example, the event of being hoarse is not affected by the numerality of events of greeting:

(i) Severely hoarse, John greeted every guest only brieﬂy. Stump (1985: 43) calls this type of free adjunct ‘strong’, as opposed to the weak one of (1). Often, it comes along with a causal (or concessive) overtone, as with (i). In this chapter, we will not deal with this type of adjunct, but will focus instead on the interplay between what, in the introduction to this volume, Himmelmann and Schultze-Berndt call ‘circumstantials’ and ‘depictives in the narrow sense’.

Adverbials and depictives as restrictors

109

As a syntactic reﬂex of this, manner adverbs turn out to be more deeply embedded in the syntactic tree than depictives. The structure of the chapter is as follows: In section 3.2, we establish adverbials and depictives as predicates which partake in Re(strictor)formation. In 3.2.1, the partition of the clause via quantiﬁcational adverbs is presented as a target structure. The derivational means to arrive at this structure (i.e. Re-movement) is introduced in 3.2.2, with reference to indeﬁnites. We go on to discuss the relation between Re-formation and focus structure in 3.2.3, distinguishing two basic types of post-syntactic movement, Re-movement and Pre(supposition)-movement. We apply Re-movement to adverbials and to depictives in 3.2.4, subsequently addressing the issue of the relation between temporal and eventive variables. In section 3.3, we broaden the perspective by classifying various sorts of adverbials with respect to their ability to be restrictive, highlighting especially their relation to depictives. We compare depictives and manner adverbs in 3.3.1, and in 3.3.2 we go on to present a pragmatically informed explanation for the fact that a subclass of manner adverbs, i.e. mental-state adverbs, resist Re-movement. Subsection 3.3.3 lays out a basic distinction between event vs. process adverbials—only the former are possible restrictors—and a translation of this difference into a syntactic perspective: event-related adverbials are adjoined at a position higher than the one occupied by process adverbials in the extended VP. After developing the system for German, we undertake an adaptation to English in 3.3.4. In 3.3.5 we integrate depictives into the picture by submitting them to a series of syntactic tests, suggesting their base position to be in the upper region of the extended VP. Section 3.4 concludes.

3.2 Adverbials and depictives as eventive predicates An important requirement for an appropriate analysis of (some) adverbials and depictives is that they should be renderable as restrictors for quantiﬁcational elements accompanying the main verb. This makes them amenable to an analysis in the vein of Diesing (1992), devised for indeﬁnites and going back to Lewis (1975). In this section we will lay out the foundations of such an analysis. 3.2.1 Quantiﬁcational adverbs and event variables 3.2.1.1 In the introduction, we have observed the quantiﬁcational effect for fronted adverbials and depictives as it originates with the Simple Present of

110

T. Mu¨ller-Bardey

the verb, if we take this tense as comprising an operator for habituality. The analysis of Lewis—and the research it sparked off—capitalized on a different quantiﬁcational category, namely, frequency adverbs. These can also be brought to bear on our issue: (5) At the bus stop, Paul always reads the newspaper. Evidently, the fronted adverbial here restricts the interpretation of always. In what follows, adverbs like always will be called Quantiﬁcational Adverbs (QA). According to Lewis, Quantiﬁcational Adverbs in connection with an indeﬁnite should be cast in the mould of (non-symmetrical) determiners like every: (6) A good bassoonist always becomes a star. (7) Every good bassoonist becomes a star. The QA, just like the determiner, induces a tripartite structure by successive application to its two arguments, ﬁrst the NP, then the VP: (8) [8 x: good-bassoonist(x)] become-a-star(x) Put in set-theoretical terms, the structure says that the set of individuals that are good bassoonists is a subset of the set of individuals that become a star. This procedure will not be sufﬁcient for a rendering of QAs in the context of adverbials. The ontological issue here is to determine the (type of) individuals forming the sets that the adverb is to quantify over. As is often done in the literature, we can identify them as the Davidsonian event variable (extended to cover states as well; for a recent overview on this topic see Pianesi and Varzi 2000). According to Davidson (1967), verbs not only exhibit variables for roles like ‘agent’ or ‘patient’, but they also carry an event variable. Furthermore, verbs are not the only kind of predicates of events, but adverbials also qualify events. This conception went against the traditional one of regarding adverbials as modiﬁers of the verb (phrase), according to which an adverbial would be applied to a verb and the outcome of that would be a verb (phrase) again, but increased in speciﬁcity. By way of illustration, the predicate ‘to stab Caesar’ becomes, for example, the predicate ‘to stab Caesar at the dinner’, if the adverbial ‘at the dinner’ is applied to it. However, Davidson argued, such an analysis cannot satisfactorily account for the entailment patterns characterizing (9)–(11); (9) entails (10a), (10b), as well as (11), while (10a) and (10b) each entail (11): (9) Brutus stabbed Caesar with a knife at the dinner

Adverbials and depictives as restrictors

111

(10) a. Brutus stabbed Caesar with a knife b. Brutus stabbed Caesar at the dinner (11) Brutus stabbed Caesar Assuming that adverbials are predicates of the event argument of the verb, and hence conjoined with the verb, immediately allows for an explanation of these patterns. The fact that a conjunction of (10a) and (10b) does not entail (9) is taken as evidence that the event variable is existentially bound for each sentence (which, in linguistic adaptations of Davidson’s framework, is normally understood to be a function of tense). Hence, we obtain the following Davidsonian representations, with ‘e’ designating the event variable:2 (12)

(9e) stab(b,c,e) & with-a-knife(e)& at-the-dinner(e)

(13) a. (9e) stab(b,c,e) & with-a-knife(e) b. (9e) stab(b,c,e) & at-the-dinner(e) (14)

(9e) stab(b,c,e)

3.2.1.2 Now that we have the variable, we can implement the quantiﬁer-like rendering of frequency adverbs (which gives rise to the term ‘quantiﬁcational adverb’), applied to (5):3 (15)

[8 e: at-the-bus-stop(p,e)] read(p,the-newspaper,e)

Similar afﬁnities exist between, for example, often, mostly, and the determiners many and most, respectively. As for the habitual or generic operator which we assumed to be involved with the habitual interpretation of sentences with the Simple Present, it too would be appropriately reﬂected by truth conditions similar to (15). An overt analogue to this operator would be the QA usually.4 2 Cf. Landman (2001: 2–15) for a recent exposition, where it is neatly demonstrated that the absense of monotonicity with VPs, in particular, impedes any other way to capture droppability, that is, the entailment of (10b) from (9), for example. 3 For the semantically initiated reader, three caveats are in order: First, evidently, stative predicates like ‘ill’ or ‘drunk’ would require a maximalization operator in order to provide a base for any type of operator (i.e. make events countable; see Johnston 1995 for related effects with when-clauses). Second, it should be noted that the ‘Neo-Davidsonian’ theory has two pillars. One is extending the concept of the event variable to stative verbs, the other is the assumption that even thematic roles of arguments are predicates of the events. While I endorse the ﬁrst of these (but cf. Katz (2000) for a critical assessment), I do not subscribe to the second. It would render vacuous the current proposal. Third, it is clear that we would need more elaborate machinery to take care of the relation between the events. Actually, a more appropriate representation would be:

(i) [8 e: at-the-bus-stop(p,e)] (9e') e e' & read(p,the-newspaper,e') 4 Restrictiveness does also surface in modal contexts, as in At the bus stop, Paul would have read the newspaper. In this chapter, however, we will conﬁne ourselves to the temporal domain.

112

T. Mu¨ller-Bardey

Thus, our introductory example (1)—repeated here as (16)—would be analysed as (17): (16) Drunk, John drives very dangerously. (17) [GEN8 e: drunk(j,e)] drives-very-dangerously(j,e) What we observe with this format is that the QA divides the clause into two parts. The string within the brackets is the ‘restrictor’ (Re) of the operator, while the predicate following it is its ‘nuclear scope’ (NSc). What is remarkable about this division is that it shapes the global structure of the clause. This is called the ‘partition’ of the clause. It has been established by Diesing (1992) (in turn elaborating on Heim 1982 and Lewis 1975) on the basis of the interpretation of nominals. Motivation for the partition theory came from the famous donkey anaphora: (18) Usually, if a man rides on a donkey, he beats it with a stick. The problem of Donkey Anaphora resides, of course, with the lack of c-command of the pronoun (it) by the indeﬁnite NP, a donkey. The solution to this problem rested on the assumption that indeﬁnites are taken to be predicates instead of quantiﬁers (i.e. they are considered expressions of type <e,t> , rather than e,t> ,t>). On the basis of this assumption, the binding mechanism proceeds in the following way: A quantiﬁer unselectively binds any open variable within its Re, while the remaining variables—which are conﬁned to the Nsc—are bound by ‘existential closure’, a default existential operator. For (18), this yields the following logical form: (19) [8 e,x,y: man(x) & donkey(y) & ride-on(x,y,e)] (9z) stick(z) & with(z,e) & beats(x,y,e) Here, the QA is able to connect the indeﬁnite a man to it, while the indeﬁnite a stick is rendered by a narrow scope existential, as intended. 3.2.2 Movement and the partition of the clause 3.2.2.1 As for the partition induced by adverbs, the question arises which constituents belong to the Re, and which to the NSc. It is not the predicate– argument structure by which this is determined. In the literature concerning this issue, indeﬁnites have attracted a lot of attention (in the wake of Diesing 1992), since for them, being in the Re as opposed to being in the NSc implies

Adverbials and depictives as restrictors

113

different truth conditions. It will be useful to quote Diesing’s Mapping Hypothesis at this point: (20)

Mapping Hypothesis (Diesing 1992) Material from VP is mapped into the NSc. Material above VP is mapped into the Re.

That is, if a constituent moves out of the VP to a higher position, it thereby enters the restrictor. One way to perform this movement is Scrambling. Consider a widely cited example relating to bare plurals from German, with the particle sequence ja doch ‘after all’ functioning as a boundary marker for the VP: (21)

a. weil ja doch Linguisten Kammermusik spielen because after all linguists chamber music play b. weil Linguisteni ja doch ti Kammermusik spielen because linguists after all chamber music play

There are two possible types of reading here. In one, the bare plural is interpreted existentially, as a statement about some linguists. In another, it is taken to be generic. The latter interpretation emerges with (21b), where the indeﬁnite has left the VP, while the former is associated with (21a), where the indeﬁnite has remained within the conﬁnes of the VP. The two readings can be stated as follows: (22)

a. [GEN8 e: C(e)] (9x) linguist(x) & play-chamber-music(x,e) b. [GEN8 e,x: C(e) & linguist(x)] play-chamber-music(x,e)

We should note that we are encountering here two kinds of non-overt functions: one is the tacit generic operator GEN8, which we have already associated with usages of the Simple Present in English. As mentioned before, it is largely a zero-equivalent to usually.5 The other non-overt function is the predicate variable C, which encodes the contextual restriction for the event variable at the current stage of discourse. It is a standard ingredient of the Re, but we will only represent it if necessary. 3.2.2.2 This kind of variation between restrictive indeﬁnites and nuclear indeﬁnites is not only observable in German. Nor is it conﬁned to Scrambling. 5 gen and usually are intensional QAs (cf. Chierchia 1995b), in contrast to the other QAs occurring in this chapter, which are extensional. We will ignore this distinction in what follows, since nothing hinges on it.

114

T. Mu¨ller-Bardey

Such effects can also be detected in English, albeit indirectly, that is, laid over with feature-driven movement. Two types of movement are of relevance here. One is (short) Subject Raising (to Spec-T, technically), which is obligatory in English except for intransitives with an expletive where the expletive occupies the subject position. The following contrast obtains: (23) a. There is usually a surgeon in the operation tent. b. A surgeoni is usually ti in the operation tent (after the battle). With an expletive, an indeﬁnite receives an existential interpretation. In contrast, if, in the absence of an expletive, the subject raises, it can receive a generic interpretation (as in 23b). Another relevant type of feature-driven movement in English is topicalization, i.e. movement to Spec-C, which also makes available a generic reading: (24) a. I usually devour beans (when you see me in the kitchen). b. Beansi I usually devour ti (if I get hold of any). We should add two qualiﬁcations here. First, the genericity effect can be bypassed by reconstruction, that means interpreting the indeﬁnite in its original site, a possibility generally connected with movement to Spec-C (see Fox 2000). Second, genericity can equally obtain without displacement if focus is on devour. (For the issue of focus, see further 3.2.3.) It may appear to be not without problems to assume that some types of movement—here, Subject-Raising and Topicalization—can absorb the effects of another—here, Scrambling. In line with Chomsky (1999), we assume there to be an interface condition according to which, roughly, NPs and adverbials that are situated above the (extended) VP are interpreted either as presuppositional or as being restrictive for some type of operator (for example, a quantiﬁcational adverb). There are several kinds of feature driven movement in the ‘narrow syntax’, among them movement to the EPP-/subject-position (Raising), movement to the accusative position (Object Shift), and movement to Spec-C (Topicalization), which raise elements out of the VP for their own purposes. If an NP is thus raised out of the VP, it will be interpreted accordingly (except in the case of reconstruction for Spec-C). However, there are also variants of linearization being produced in the phonetic component. This is, essentially, what Scrambling is like. These variants are handed over to the interface as well. According to the interface condition, the moved versions, as opposed to the non-moved ones, will be singled out for restrictive interpretation (or, as demonstrated in the next section,

Adverbials and depictives as restrictors

115

presuppositional interpretation). In this sense, Raising and Topicalization can absorb the effects of Scrambling. 3.2.3 Re-movement vs. pre-movement 3.2.3.1 We have established that with indeﬁnites, there are restriction effects by movement. We will be looking at adverbs from this angle in section 3.2.4. But before doing that, we have to address another type of effect of movement, Pre(supposition)-movement, which is similar to restriction and sometimes hard to tell apart from it. As we will see, it is also of relevance to adverbs. Consider once more an example of Scrambling in German: (25)

a. weil Jan im Seminar immer [vp einen Sachsen because Jan in-the class always a Saxonian foppte] teased ‘because Jan, in the class, always teased a Saxonian’ b. weil Jan im Seminar einen Sachseni immer because Jan in-the class a Saxonian always [vp ti foppte] teased

(26)

a. [8 e: in-the-class(e)] (9x) Saxonian (x) & tease(j,x,e) b. [8 e,x: in-the-class(e) & Saxonian (x)] tease(j,x,e)

Movement in the presence of the QA immer ‘always’ yields the expected variation: In (a), we are talking about one Saxonian, in (b), about a multitude of Saxonians (‘Whenever there is a Saxonian in the class, Jan teases him’). At the same time, (25b) may have another interpretation where the indeﬁnite does not become a restrictor for the QA. In this case, the indeﬁnite is shifted because it has a presuppositionally prominent status, often translated as partitive: one of the Saxonians he teased, the others he left alone. Once again, traces of this effect of movement can be observed in English, for instance with subject-raising, as has already been pointed out by Milsark (1974) (cited in Diesing 1992: 59): (27) a. There are some ghosts in my house. b. Some ghosts are in the pantry (others are in the attic). We may safely assume that this kind of movement also subsumes the often-noted dislocation of deﬁnites (The thief, I did not see), which is equally driven by presupposition. Consequently, there seem to be two kinds

116

T. Mu¨ller-Bardey

of post-syntactic (non-feature-driven) movement: Re-movement and Pre(supposition)-movement: (28) Types of post-syntactic movement a. Re-movement b. Pre-movement6 The ﬁrst kind of movement is sensitive to QAs, but the latter is not. 3.2.3.2 Regarding the last examples, the question might be raised whether we are dealing with speciﬁc types of movement, or, alternatively, are touching on variations of focus structure. As concerns presupposed NPs, they are undoubtedly instances of the complement of focus, viz. the background. So maybe the proposed positional regularities are side-effects of the division between focus and background? As is well known, the focus–background division is also amenable to a movement account. It is commonly assumed that focus may project to the entire VP according to a rule that for German, for instance, says ‘stress the preverbal argument if there is one, and the verb otherwise’. Now, what will happen if according to communicative needs it is not the entire VP which is to be in focus but only one of its sub-constituents? In that case, narrow focus would apply: (29) [vp . . . X . . . [Y]f . . . ] But there is an alternative to this approach, which yields the same effect in terms of the assignment of constituents to focus and background. One can tailor the VP without invoking narrow focus by evacuating it (through some kind of anti-focus movement; see Reinhart 1995): (30) (X) . . . [vp . . . (X) . . . Y . . . ]f Here, focus projection is preserved—i.e. no narrow focus assignment occurs—but the content is alterable by moving some part of it into the 6 One may speculate whether Pre-movement is a subcase of a broader phenomenon. In German, Scrambling makes available alternatives of scope order (see Frey 2000): (i) a. Paul hat jedem Kind mindestens ein Bild gezeigt (8>9) Paul has every[dat] child at-least one picture shown

b. Paul hat mindestens ein Bildi jedem Kind ti gezeigt (9>8, 8>9) Paul has at-least one picture every[dat] child shown In fact, if we put together this kind of scopal movement with our Pre-movement, we obtain a natural class of quantiﬁer movement as opposed to predicate movement in the case of Re-movement.

Adverbials and depictives as restrictors

117

background space. Note that the focus domain is identical to the domain of existential closure. Thus any predicate remaining in the VP can be existentially bound without attracting stress, i.e. narrow focus. According to this approach, the focus domain is identical to the NSc, and the background, to the Re. Applying this approach to our example (21b), this would yield (marking stressed words by small caps): (31) weil Linguisteni ja doch [vpti Kammermusik spielen]f because linguists after all chamber-music play ‘because linguists play chamber music after all’ The indeﬁnite ‘chamber music’ is the exponent of the rule of focus projection (not of narrow focus). Hence, movement of Linguisten can be conceived of as movement out of the NSc as well as movement out of the focus domain. The question arises whether Re-movement—the type of movement that we are primarily concerned with—can be entirely reduced to this kind of focus shift. In the next section, we will argue that adverbials provide evidence that there is in fact more to Re-movement.7 3.2.4 Extending the indeﬁnites framework to adverbials and depictives 3.2.4.1 Having established the mechanisms of assigning indeﬁnites to the Re, as opposed to the NSc, we turn to adverbials. Our basic claim here is that adverbials should be treated analogously to indeﬁnite nominals. That is, their positional variation is in large part due to alternating assignments to either Re or NSc. This, in turn, presupposes the assumption that they are in fact predicates of events. Taken as a whole, such a theory would provide a signiﬁcant improvement in our understanding of adverbials in that it reduces some of their peculiarities to facts that are already well understood. The focus of the following discussion will be on Scrambling in German, with occasional consideration of Topicalization in English. As will become apparent, generalizing from indeﬁnites to adverbials will only be successful if one draws a basic distinction between the two ‘pragmatic’ dislocation types 7 The contextual variable (C) mentioned above is furthermore an appropriate tool to account for the formation of the Re with deﬁnites and names (cf. Rooth 1995). Consider:

(i)

Er he

hat Maria immer ein Bonbon has Maria always a sweet

geschenkt. given

For this sentence one will only ﬁnd an appropriate representation if additional events are supplied, notwithstanding the fact that adverbials are not involved: ‘For all relevant events involving Maria (i.e. meeting her), she was given a sweet at that event’.

118

T. Mu¨ller-Bardey

mentioned above, Re-movement vs. Pre-movement. Let’s start with the following example: (32) a. Wir haben oft auf dem Balkon gegrillt. we have often on the balcony barbecued ‘We often barbecued on the balcony.’ b. Wir haben auf dem Balkoni oft ti gegrillt. we have on the balcony often barbecued ‘On the balcony, we often barbecued.’ Here, we have a shift of the PP ‘on the balcony’ to the front of the quantiﬁcational adverb.8 This shift can come about either by Re-movement or by Pre-movement.9 The latter would occur in a context, for example, when somebody says, ‘You have a nice balcony!’ This establishes the balcony as a discourse object, which in turn may be a reason to let the PP undergo Premovement. This movement has no inﬂuence whatsoever on the interpretation of the quantiﬁcational adverb. Hence, from the perspective of the quantiﬁcational adverb, the PP could be regarded as reconstructing: (33)

[many e: C(e)] on-the-balcony(e) & barbecue(w,e)

The second way to reach (32b) is by Re-movement. This is plausible as a response to some neighbour saying, ‘From my window [across the street], I can look over to your balcony. You rarely seem to sit there reading. So what do you use it for?’ There are other types of example, where the possibility of Pre-movement seems to be much less likely. In these instances the possibility of Re-movement comes to the fore, as for example in (34) (cf. (3)): (34)

Wir haben auf dem Balkon oft geschwitzt. We have on the balcony often sweated ‘On the balcony, we often sweated.’

8 In contradistinction to NPs, scrambling PPs implies a source position that is already outside the core VP. We will return to the issue of the base position of adverbials below. 9 The claim is not unheard of that adverbials do not scramble at all, as maintained for instance by Fanselow (2001), on the grounds that Long Scrambling cannot be found for adverbials. However, it is unclear why this would have to carry over to Short Scrambling. Furthermore, this claim is not substantiated with PPs.

Adverbials and depictives as restrictors

119

This conveys very clearly the meaning: ‘Often, when there was an eventuality of us being on the balcony, we sweated in that eventuality.’ In cases like these, the PP is being scrambled as a predicate of events:10 (35) [many e: C(e) & on-the-balcony(e)] sweat(w,e) As suggested by the translations of the preceding examples, Topicalization in English produces similar effects, when adverbials (event predicates) are shifted to the Re by movement to the top (Spec-C): (36) On the balconyi, we usually [vp ti [vp barbecue]] The site of origin of the adverb assumed here might appear to be unusual, as the adverb could never surface there. We will return to this issue in 3.3.4. At this point, it should be noted that we are dealing here with a source site of movement that is an extended form of the VP. 3.2.4.2 Up to now, when discussing predicates—indeﬁnites and adverbials alike—that remain within the NSc, we have only considered wide focus. But both predicate types can also be the sole constituent in the NSc (taken as focus domain here): (37) Usually, Paul reads [a crime story]f (38) Usually, we barbecue [on the balcony]f We might account for this by assuming that the core VP is pushed into the Re. For adverbials, this touches on the well-known observation that they resist being integrated in the focus projection (see Gussenhoven 1992). Some authors (Delﬁtto and Bertinetto 2000; Koster and Zwart 2000) argue that sentences like (38) are the result of moving the VP upwards: (39) Usually we [vp barbecue]i [vp ti [on the balcony]] Without getting into the details of this proposal (with the obviously problematic movement of an inner segment), we note that this would be very much in accordance with what we think connects position and interpretation of adverbials: the restrictor results from movement out of the (extended) VP:11 (40) [usually e: C(e) & barbecue(w,e)] on-the-balcony(w,e) 10 We have to take ‘many’ to be asymmetrical here, i.e. the implicit standard applies to the (relevant) events taking place on the balcony. 11 Note that indeﬁnites and adverbials are not in complete agreement with respect to the partition of the clause. For, if an indeﬁnite remains in the conﬁnes of the VP, it is targeted by existential closure, instead of being bound by the QA, as can be observed with (25) and (26) above: The indeﬁnite ‘a Saxonian’ is bound by the QA only in (25b) and (26b). The reason for this is that there can only be one nominal per variable. In order for the nominal to remain in the VP and still have the variable be

T. Mu¨ller-Bardey

120

Hence, narrow focus can likewise be modelled along the lines of the partition approach. 3.2.4.3 Turning to depictives, it will be clear, by now, that the movement analysis carries over to them. An example like (41) would receive the analysis of (42): (41) Paul hat betrunken meistens die Hintertu¨r benutzt. Paul has drunk mostly the back door taken. ‘Drunk, Paul took the back door in most instances.’ (42) Paul hat betrunkeni meistens [ti [die Hintertu¨r benutzt]] Paul has drunk mostly the back door taken As mentioned above, depictives are predicates of individuals (of the noneventive type), and additionally, they have an eventive variable. Hence, ‘drunk’ would be ‘drunk(x)(e)’. The eventive variable also plays a role in a simple clause like Paul is drunk (here it would be bound by tense). Furthermore, we said that adverbials partake in Re-formation by virtue of an eventive variable, and pointed out that depictives are adverbials in this sense. Now the question might be raised whether the variable that depictives contribute to the Re (or the NSc, for that matter) is the same as the eventive variable they have as predicates, or whether, alternatively, they acquire it as adverbials. In this regard, we can refer to the so-called individual-level constraint: (43)

#

Intelligent, Paul takes the front door [# if restrictive]

It is impossible to have an individual-level predicate as restrictive depictive.12 One can give two explanations for this constraint. One consists in capitalizing on the line of reasoning by Rapoport (1991), who argues that individual-level predicates do not have an event role. The other one would be to follow Katz (1993), who takes this constraint as evidence that individual-level predicates come along with a generic operator which binds their eventive variable. bound by the QA, the variable would have to be identiﬁed with the subject, with the verbal predicate being fronted. This does occur, in fact, with idiom chunks: ti

fu¨hren lodge

(ii) a. weil Beschwerdeni immer Sachsen ti because complaints always Saxonians

fu¨hren lodge

(i) a. weil because

immer Beschwerden Sachseni Saxonians always complaints

b. [(8 x: Saxonian(x)] lodge-complaints(x)

b. [(8 x: lodge-complaints(x)] Saxonian(x) 12

However, such an adjective is possible as a strong adjunct (cf. n. 1): Intelligent, Paul did not even attempt to take the back door.

Adverbials and depictives as restrictors

121

Either way, we can draw two conclusions from this. First, our approach is generally corroborated since we can safely infer that the eventive role is in fact what is responsible for the role in quantiﬁcation. Second, even if adjectives are assumed to acquire an adverbial eventive role just when they are not predicates of clauses or attributes, it is their lexical eventive role that this must be based on. 3.2.4.4 A potentially problematic issue pertaining to the nature of event variables is their relation to temporal adverbials. Within our model of Re-movement, do we ascribe to event variables what would genuinely only hold of time variables? Delﬁtto and Bertinetto (2000), in a paper dealing with issues related to ours, only acknowledge one type of adverbials as really partaking in a partition of the clause in terms of quantiﬁcation over event variables, and that is the class of temporal adverbials. That means that of the following two sentences, only the ﬁrst one would be a case in point: (44) At 8 a.m., John always goes swimming. (45) In Paris, John always drinks red wine. However, Delﬁtto and Bertinetto (2000) do not make use of a contextual accommodation (C), which would render spatial descriptions appropriate for restrictive purposes. Nor do they acknowledge that the Re is not conﬁned to topicalized material. Regarding local PPs (such as on the balcony), it is worth pointing out that when they are in the Re, their preposition does not seem to be of relevance in a strictly spatial sense. Hence, they do not contrast with other local prepositions (as, for example, under the balcony). This points to equivalence with other types of restrictive elements, speciﬁcally, narrowly temporal ones. Depictives can also be brought to bear on this issue. Geuder (2000: 196) claims that they do not take part in the combination of temporal variables. However, it is quite easy to come up with examples exhibiting depictives that combine with the main predicate in a temporally restricted manner (i.e. not simply overlapping in time): (46) Die Blumen machten schon frisch einen schlechten the ﬂowers gave already fresh a bad Eindruck nicht erst welk. impression not not until wilted ‘The ﬂowers already looked bad fresh, not only when wilted.’ Through the inﬂuence of the phasal adverbs ‘already’ and ‘not until’, the depictive state is viewed as temporally bound, as a phase in a series of states

122

T. Mu¨ller-Bardey

(i.e. ‘fresh’ preceding ‘wilted’).13 Hence, there is ample evidence that the restrictor of QAs is not conﬁned to temporal adverbials. 3.2.4.5 We conclude this section by returning to the issue of focus, which we have already discussed in 3.2.3. There is reason to believe that Re-formation should not be reduced to the division between focus and background. Evidence comes from Re-movement of adverbials; cf. example (47): (47) Damals hat sie [auf dem Tennisplatz]f gewo¨hnlich Then has she on the tennis court usually geschwitzt. sweated ‘At that time, she usually sweated on the tennis court.’ This can be an appropriate statement in a context where the activity which usually made her perspire is at issue: (48) [most e: on-tennis-court(e)] sweat(s,e) This means, the PP in focus functions restrictively, but is raised above the QA in accordance with its restrictiveness.14 Thus, Re-formation might be a rather self-contained area of grammar.

3.3 Sorting out Re-suitable adverbials 3.3.1 Non-restrictive adverbs Taking stock, we have now established that depictives, like adverbials, contribute to the partitioning of the clause as predicates of events. In this section 13 Even the very restriction Geuder adduces as evidence of the fundamental non-temporality of depictives does not seem to be sustainable. He claims that depictives cannot anchor to the ‘reference point’—an undoubtedly temporal individual—of perfect tenses. Examples like (i), however, do not seem to be unreasonable, displaying exactly such a reading:

(i)

Already/Even half-full, the tank had come off the rear fastening.

14

It is not entirely clear whether this is different in any signiﬁcant way from what Herburger (2000: 69–73) discusses as ‘second occurrence focus’, von Fintel’s (1994) account of which—according to which the partition should generally not be read off from a non-enriched focus structure—she dismisses. This type of mismatch between focus structure and partition would obtain, for example, if the above example would be a rejoinder to the following utterance: (i) Damals hat sie in der Turnhalle gewo¨hnlich [geschwitzt]f Then has she at the gym usually sweated ‘At that time, she usually sweated at the gym.’ In this case, the focus structure of the ﬁrst sentence is being continued on a non-superﬁcial level in the second sentence. Note also that there is a group of adverbials which, when focused, not only can but must be interpreted restrictively in preposed position, namely, if/when-clauses, for which see Johnston (1995).

Adverbials and depictives as restrictors

123

we will develop a more ﬁne-grained typology of (de-)adjectival adverbs, encountering a subtle distinction between manner adverbs and depictives. Our claim about the appropriateness of the predicate approach to adverbials does not hold in an unqualiﬁed way. There seem to be two groups of adverbial, one Re-suitable, the other tied to the NSc and generally not in the same way a predicate of events as the ﬁrst kind. The central member of the NSc-group is the manner adverb. Within tree structure, we will argue, adverbials of this group range in the lower part. Depictives, despite appearances, belong to the ﬁrst, the Re-suitable group. Consider the following German example with the manner adverb ‘quietly’: (49) Paul hat leise meistens die Zementsa¨cke geschleppt. Paul has quietly mostly the cement bags hauled ‘Mostly, Paul quietly hauled the bags of cement.’ It is difﬁcult to associate a reading with this sentence which has the Re consisting in the manner adverb (i.e. of the type of ‘When acting quietly . . . ’). In fact, this becomes even more apparent with an adverb like ‘carefully’: (50)

#

Paul hat vorsichtig meistens die Jalousien runtergelassen. Paul has carefully mostly the blinds lowered ‘Mostly, Paul carefully lowered the blinds.’

Again, this sentence can hardly be taken to mean ‘Most events of executing something carefully were events of lowering the blinds’, as represented by (51): (51) [most e: C(e) & carefully(e)] lower(p,the-blinds,e) This means that the manner adverb is not able to form the Re. For, if it were, the sentence would be wrong in a context where all other activities were carried out carelessly by Paul (assuming that ‘mostly’ presupposes ‘not all’).15 Preferably, however, we understand (52): (52) [most e: C(e)] lower(p,the-blinds,e) & carefully(e) That is, in most of the relevant instances, he lowered the blinds cautiously. The adverb is unable to join the Re. Movement of the adverb is not without 15 Our mapping mechanism should not exclude shifting (or copying) material from the c-command domain of the QA into the Re if there is narrow focus. Exploiting this possibility, if we resort to the adverb ‘only’, which is otherwise not in the scope of this chapter, we can get the manner adverb to be at least a partial exponent of the Re:

(i) Paul hat vorsichtig nur Paul has carefully only

die kleinen Pakete geo¨ffnet the small packages opened

(ii) [(8 e,x: open(p,e,x) & package(x) & carefully(e)] open(p,e,x) & small(x)

124

T. Mu¨ller-Bardey

effects, however. An appropriate rendering would be: ‘What he mostly does cautiously, is lowering the blinds.’ That is, the notion of caution is already established as a discourse object, if only indirectly. Recall that we have subdivided Scrambling (as well as Topicalization) into Re-movement and Pre-movement, the former serving the purposes of Re-formation, while the latter is presupposition-driven. Hence, we can conclude that manner adverbs do not undergo Re-movement, but only Pre-movement.16 We will now give a pragmatics-informed explanation for the resistance of the mental state adverbs of the type of ‘carefully’, in particular, to occur in the Re. This leaves us with a number of further adverbial types the nonrestrictiveness of which is yet to be explained. A comprehensive syntactic account for the behaviour of this group follows the discussion of the mental state adverbs. 3.3.2 Mental-state adverbs 3.3.2.1 Since Jackendoff (1972), it has been known that the members of the class of adverbs to which ‘carefully’ belongs (along with, for example, ‘stupidly’ and ‘cleverly’) display an ambiguity between a manner variant and a variant known as ‘subject orientation’.17 This ambiguity is reﬂected by a distinction between postverbal and preverbal position, respectively: (53) a. John checked the ﬁgures carefully. b. John carefully checked the ﬁgures. The subject-oriented subtype ascribes the property of carefulness to the subject by virtue of his or her acting in the way encoded by the VP. A near-paraphrase is provided by the underlying adjective: (54)

It was careful of John to check the ﬁgures.

There have been several attempts to relate the subject-oriented variant of these adverbs to the manner reading, mostly by extension of the latter, proposing an abstract higher predicate as its host (see e.g. McConnell-Ginet 1982). But we should note that the subject-oriented adverb functions in a way that is entirely different from what we have seen with predicates of events. 16 Pre-movement is marked here by intonation. The moved adverb is likely to carry a fall–rise tone, rendering it a contrastive topic (Jacobs 1997). Supposedly, contrastiveness is how presuppositionality is realized here. What has been established is not necessarily the adverb as such, but a series of notions belonging to one dimension, in the case of ‘cautiously’, say, the dimension of conforming to standards of carefulness. 17 Subject-orientation contrasts with ‘speaker-orientation’ with adverbs like regrettably, cf. John regrettably missed the train.

Adverbials and depictives as restrictors

125

The adverb (as well as the predicative adjective) conveys a comment on the content of the VP. Accordingly, focus-sensitivity obtains for the adverb, as demonstrated by the following variation: (55) Stupidly, John kissed her on the [lips]f (56) Stupidly, John [kissed]f her on the lips Focus-sensitivity in turn has consequences for the patterns of entailment connected to modiﬁers (cf. Geuder 2000: 125f., drawing on Wyner 1994). For adverbials that are focused, the pattern we exempliﬁed with examples (9)–(11) above does not obtain. Hence, (55) does not entail (57): (57) Stupidly, John kissed her. 3.3.2.2 It can be concluded that subject-oriented adverbs should not be regarded as predicates of events, which implies that they cannot form the restrictor of a quantiﬁcation over events. For this reason, they are similar to the class of sentence adjuncts, with which they share several positional properties (cf. Frey 2000).18 This would raise the issue of whether the resistance of manner adverbs like ‘carefully’ to be restrictive may be causally related to their occurrence as a subject-oriented adverb, given that in this latter case restrictiveness is never an issue. Geuder (2000) further pursues this line of reasoning. He capitalizes on the fact that subject-oriented adverbs (which he calls ‘agentive’) are de-adjectival, stemming from the domains of capabilities (cleverly), social standards (rudely), and attentional standards (carefully). He notices that with these adverbs, an individual is ascribed the respective property; this individual is identical to the of-argument in the adjectival paraphrase in (54) above. However, with the corresponding manner adverb, the individual variable is not clearly discernible (see further below). Geuder accounts for the semantics of subject-oriented adverbs by the following schema. Generally, the adverb expresses an assessment of the agent in his or her decision to execute a certain action, described by the VP. This assessment is done, ﬁrst, in view of the goals the agent can be assumed to pursue—that is, how do the expected consequences match up with the goals? 18 This said, one has to point out that subject-oriented adverbs are not intensional operators on propositions either, primarily for the reason that there are no opacity effects to be detected with them: cf. Geuder (2000: 126), referring to Wyner (1994). On the other hand, Geuder shows that the subjectoriented adverb does involve reference to the event described by the VP. It can relate to a conjunction of events, but not a disjunction. In other words, it can operate on ‘big events’. Geuder, drawing on Asher (1993) and others, therefore assumes that the argument of a subject-oriented adverb is an intermediate category ‘fact’.

126

T. Mu¨ller-Bardey

This is the ‘causal’ component. Second, this decision is taken to be indicative of a disposition on the part of the agent. This is the ‘mental’ component. In order to illustrate the two components, the causal and the mental component of stupidly and carelessly are given here (Geuder 2000: 160): (58) stupidly causal: effects of the action do not match the (possibly long-term) goals of the agent mental: lack of intellectual capability (59) carelessly causal: effects of the action do not fully match the intended effects of the action or create unwelcome side effects mental: lack of attention 3.3.2.3 How do manner adverbs of the carefully class come into play? Quite surprisingly, it is not easy to give a compositional account for them. While we have no difﬁculty in understanding She opened the package carefully, it is hard to see how this relates to the underlying adjective in She is careful (at least much harder than in She carefully opened the package in her ofﬁce). Are manner variants therefore better candidates for being predicates of events than their subject-oriented analogues? This is questionable, since it is not clear how anything can be a careful event. That is why we found that the manner adverbs of the carefully class resist being assigned to the Re. Geuder’s (2000: 182f.) solution to this problem is the suggestion that the manner variant of these adverbs underlyingly refers to the subject-oriented variant. His claim is that for every occurrence of the former, there is an occurrence of the latter, such that the VP of the manner variant is instantiated or implied by the subject-oriented variant, in the sense that it is the encompassing term. The hearer, on getting the manner adverb, reconstructs (tentatively) the more speciﬁc subject-oriented adverb by abduction, taking recourse to the context. Consider the following example for a manner adverb of this class: (60) The aides worded the text carefully. Assume a context of formulating a communique´ after talks over a diplomatic quarrel, where both sides did not reach an accord, but want to leave the door open for further diplomatic efforts. Then the speciﬁc proposition—involving the subject-oriented adverb—might be: (61) The aides carefully mentioned those points that are shared by both sides.

Adverbials and depictives as restrictors

127

Hence, the postverbal position in (60) signals the abductive presence of the more central, subject-oriented variant of these adverbs.19 In this way, we can give a pragmatic explanation as to why mental-state adverbs like carefully do not undergo Re-formation. 3.3.2.4 The preceding discussion suggests the following hypothesis. If a manner adverb does not have a subject-oriented analogue, it should not resist forming the Re, at least not as strongly as carefully. This expectation is indeed borne out, as can be seen with our initial example (49), repeated here: (62) Paul hat leise meistens die Zementsa¨cke geschleppt. Paul has quietly mostly the cement-bags hauled ‘Mostly, Paul quietly hauled the bags of cement.’ This sentence should be excluded in a context where Paul carried out all other activities loudly (a restriction which did not hold analogously with carefully in (50) above). But this is in fact exactly what (62) would mean if it were an acceptable construction. Most speakers, however, show some reluctance to accept it in the ﬁrst place. Hence, we should continue our search for reasons preventing manner adverbs from occupying the Re. The explanation we develop in the next section will actually target an even more comprehensive class than manner adverbs. 3.3.3 Event-related vs. process-related adverbials In the current subsection and subsequent ones, we will outline two classes of adverbials in a primarily syntactic perspective, according to their disposition to take part in the partition of the clause. Depictives and manner adverbs will turn out to side in different ways. Let us start with a synopsis of adverbials in general. 3.3.3.1 In Frey and Pittner (1998) and Frey (2000), an attempt is made to detect the base positions of adverbials of the type that is not to be associated with functional categories (in contrast to the adverbial subsystem under scrutiny in Cinque (1999), although there is some overlap here). To this end, they apply a battery of criteria that have proved useful in sorting out the merging sites of subjects and objects in German, primarily comprising complex topicalization, existential closure, focus projection, and effects of Principle C and quantiﬁer scope (some of which are illustrated below). 19 Geuder (2000: 184) sees a close parallel between the subject-oriented adverb and its manner analogue in that both induce a set of alternatives—in the case of the former, this is to be identiﬁed with the focus value of its argument, while with the latter, this is the set of ways to execute the action described by the VP.

128

T. Mu¨ller-Bardey Sentence adjunct Finite Event-related (temporal, causal) Subject Event-internal (local, instrumental) Object Process-related (manner) V

Figure 3.1 Schematic order of base positions of adverbials in German (according to Frey and Pittner 1998)

The relative order of adverbials emerging from this—certainly not a surprise from the point of view of what is known from works on English—spans from sentence adverbs in the high position c-commanding the highest verb (the ﬁnite position) to manner adverbs in the low position c-commanding the lowest verb trace in the VP-shell. Compare the following English example: (63) Johni [fortunately [has [ti spokenj [to his mother [nicely [tj about the letter]]]]]] In the middle range, we ﬁnd, for example, temporal and local adverbials. Without going into details, we give here a broad conspectus of the adverbial classes in their merging positions according to Frey and Pittner (1998). Figure 3.1 gives a schematic overview for German. As for English, one has to keep in mind that, ﬁrst, the order of the middle-range adverbials is inverted, so that a left-branching structure obtains (see the next subsection), and, second, that the subject moves obligatorily. From the discussion of subject-oriented adverbs in 3.3.2, it should be clear that sentence adjuncts (like fortunately)—which subject-oriented adverbs are closely related to—are not plausible candidates for restrictorhood.20 As for the other classes, we have already seen representatives of the mid-range to be restrictive, to which we can add depictives. Towards the lower end of the schema, we have seen that manner adverbs disfavour the Re. On closer inspection, it will become obvious that these are not the only adverbials resisting attraction by the Re. 20

In fact, they never belong to the NSc either, being similar in that respect to strong adjuncts: see ex. (1).

Adverbials and depictives as restrictors

129

3.3.3.2 First, according to Maienborn (2000), there is a subtype of local adverbials, which she calls ‘internal locative modiﬁers’ and which have very special properties. Consider: (64) Paul hat die Forelle an einem langen Spiess gebraten. Paul has the trout on a long spit roasted ‘Paul roasted the trout on a long spit.’ Maienborn notes that interrogatives pertaining to inner locatives involve a manner and instrument ‘how’ instead of a locative ‘where’. Quite obviously, ‘on a long spit’ is not situating the event as a whole, but only a part of it. Maienborn proposes the following kind of semantics: (65) (9e,e") roast(p,the-trout,e) & part-of(e",e) & loc(e",on(a-long-spit)) That means: ‘There is an event e of Paul roasting the trout, such that the location of an e" which is part of e is on a long spit’ (where further details of the relation between these two events are construed contextually). Now, testing internal locative modiﬁers for restrictiveness, we observe that they are rather reluctant to move to the Re: (66) Ich habe an einem Spiess selten eine Forelle gebraten. I have on a spit rarely a trout roasted ‘I rarely roasted a trout on a spit.’ This sentence will hardly ever receive the interpretation of ‘Only few instances of my doing something on a spit were instances of roasting a trout’. Instead, only the reconstructive reading (quantifying over instances of either roasting a trout or roasting something on a spit) comes to mind. In search for an explanation for this restriction, the representation in (65) which Maienborn (2000) gives to sentences with internal locative modiﬁers might provide a hint. In (65), an event variable e" is introduced, in addition to the standard variable e which external locative modiﬁers would predicate over. Speciﬁcally, e is assumed to contain e". Hence, we might hypothesize that it is prohibited for partial individuals and events to partake in the partition of the clause. This hypothesis can be supported both syntactically and semantically. The ﬁrst possibility is based on the assumption of a two-tiered VP. Assume that the comprehensive event is associated with the higher shell, the partial event (or process) with the lower one. We would capitalize on the relation of inclusion, thus ending up with some type of locality constraint. This prevents the extraction of an event variable belonging to the lower-shell VP outside the higher-shell VP.

130

T. Mu¨ller-Bardey

The corresponding semantic explanation would be as follows. Any modiﬁer which is to be dislocated from the Re is to be bound by the QA. However, an adverbial like on a long spit corresponds to ‘part-of(e",e) & loc(e",on(a-long-spit))’. And this necessitates binding not only of e, but also of e". We could combine these two lines of explanation in the following Restrictor Constraint: (67) Restrictor Constraint Adverbials merged in the lower VP shell do not undergo Re-movement, because they refer to sub-events. 3.3.3.3 The notion of partiality might be extended to many, if not all, manner adverbs, as an account for their inability to join the Re. There are many subtypes of manner adverbs. But even those that Ernst (1984) identiﬁes as ‘pure manner adverbs’—for the reason that they are strictly event-related and do not lend themselves to multiple readings or usages—only focus on one particular aspect of the event that the verb introduces. For example, loudly in sing loudly, which at ﬁrst glance may seem to relate to the same event that sing does, describes but one particular perceptual dimension of it, which at the same time cannot be conceived as isolated from the event of singing as a whole. Similarly, the manner adverbs dealt with in Geuder (2000), which are selected according to the property (termed ‘orientedness’) of implying an assertion about the subject or object, all relate to an intrinsic part of the verbal event, as is evident from the various representations of their meaning which Geuder considers. This group includes resultative adverbs (load the cart heavily), manner variants of subject-oriented adverbs (open the package carefully), and psychological adverbs (read the article angrily). As yet another case in point, dimensional adverbs like ‘quickly’ invoke a subevent, as seen in heat up quickly: The speed relates to the subevent of measuring out a speciﬁc stretch of a scale, not the event of heating up as a whole (cf. Eckardt 1997: 40). The subevent variable cannot be severed from the variable of the encompassing event. Some corroborating evidence for this idea comes from coordination. An adverbial like in the park may hold of the event variables of more than one verb. In contrast to that, a manner adverb is tied to one speciﬁc verb: (68) Wir haben im Park [[Trinklieder gesungen] und we have in-the park drinking-songs sung and [Fussball gespielt]] soccer played ‘In the park, we sang drinking songs and played football.’

Adverbials and depictives as restrictors (69)

131

#

Wir haben laut [[Trinklieder gesungen] und we have loudly drinking-songs sung and [Fussball gespielt]] soccer played ‘We sang drinking songs and played football noisily.’

The manner adverb, in contrast to the locative adverbial, is not able to relate to the two predicates in a symmetrical way.21 3.3.3.4 We now turn again to the order of adverbials in Frey and Pittner (1998) (and Frey 2000). In the middle range of the order of circumstantial adverbs, there seems to be some sort of cut-off point, relating to the adjunction site of the adverbials, such that the items following it include an additional event variable and resist being restrictive. We will offer more evidence for this claim, which relates to instrumentals. Instrumentals are likely to be placed in positions in the region of the locative modiﬁers in Frey’s order, and we would expect there to exist a variation similar to the one with locatives. To this end, we need brieﬂy to review some of Frey’s criteria developed for German. One is the following scope test. In German, the base order can be determined by absence of scopal ambiguities. For example, the indirect object with most ditransitive verbs precedes the direct object. Scrambling makes available a second scopal order: (70) a. Paul hat jedem Kind mindestens ein Paul has every [dat] child at-least one Bild gezeigt (8>9) picture shown ‘Paul showed every child at least one picture.’ jedem Kind ti b. Paul hat mindestens ein Bildi Paul has at-least one picture every[dat] child gezeigt (9>8, 8>9) shown ‘Paul showed at least one picture to every child.’ A second criterion is focus projection. With the unmarked order, focus can extend to the entire VP: (71)

Dann hat Lotte [einem Besucher [ein buch gegeben]]f Then has Lotte a[dat] visitor a[acc] book given ‘Then Lotte gave a visitor a book.’

21 This restriction does not seem to hold with VPs that give rise to a uniﬁed (atomic) event, such as ‘to call the waiter and order a juice’. By contrast, singing drinking songs and playing soccer involves two independent events. This conﬁrms the hypothesis that the event level determines the use of manner adverbs.

132

T. Mu¨ller-Bardey

When, as a marked option, the direct object scrambles out of the VP, a wide focus reading is not possible, that is, focus cannot project to the VP with the scrambled direct object included: (72)

Dann hat Lotte [ein Buchi [einem besucher Then has Lotte a[acc] book a[dat] visitor [ti gegeben]]]f given #

As the third, and ﬁnal, criterion, consider Complex Topicalizaton. A VP in unmarked position can be fronted: (73)

[Einem Besucher [ein Buch gegeben]]i hat Lotte ti a[dat] visitor a[acc] book given has Lotte

So can the sequence of direct object plus adjacent verb: (74)

[Ein Buch gegeben]i hat Lotte einem Besucher ti a[acc] book given has Lotte a[dat] visitor

But, signiﬁcantly, moving the VP with the direct object being left behind (as an instance of what is known as Remnant Movement), is much more marked. It is licit only if the direct object left behind is focused. The reason for this is a violation of the Proper Binding Condition: (75)

?

[Einem Besucher ti gegeben]j hat Lotte ein Buchi tj a[dat] visitor given has Lotte a[acc] book

Thus, Complex Topicalization, along with focus projection, constitutes an indicator for the derivation of the clausal order. 3.3.3.5 On the basis of these tests, let us take a closer look at the adverbial class of instrumentals. The following example has both a restrictive and a reconstructive reading, (a) and (b) respectively, but the latter obtains much more easily: (76)

weil because am on-the

ich I Auto car

mit dem Hammer selten etwas with the hammer rarely something repariere. ﬁx

a. ?‘When I use the hammer, it is only in rare instances that this happens in connection with repairing the car.’ b. ‘Only rarely do I ﬁx my car by using a hammer.’ With (a), it may be the case that most instances of my ﬁxing the car actually involve the hammer.

Adverbials and depictives as restrictors

133

Now, let us compare this to an adverbial which is related to the instrumental, i.e. the comitative: (77) weil ich mit Karl selten Kuchen backe because I with Karl rarely cake bake ‘because with Karl, I rarely bake cakes’ Like (75), this sentence also has a reconstructive as well as a restrictive reading. But the latter is accessible much more readily than in the case of the instrumental. We believe that this is symptomatic of a subtle dividing line between instrumentals and comitatives that is also reﬂected in the tests of Frey and Pittner (1998). According to their criteria, both the instrumental and the (subject-related) comitative have their base position between the subject and the object. But their discussion also contains indications that set one adverbial type apart from the other. First, we may apply the focus criterion. In the most unrestricted case, the comitative seems to precede the instrumental: (78) Er hat [mit einem Freund mit einem he has with a friend with a Kleintransporter einen Schrank herbeigeschafft]f minivan a cupboard transported ‘With a friend, he transported a cupboard with a minivan.’ (79)

?

Er hat [mit he has with Freund einen friend a

einem Kleintransporter mit einem a minivan with a Schrank herbeigeschafft]f cupboard transported

Second, when we consider Complex Topicalization, we observe that it seems to be much less of a problem to topicalize the VP including an instrumental, with a comitative left behind, than a topicalization with the reverse distribution. Effectively, this should be taken to mean that the instrumental is closer to the verb than the comitative: (80) [Mit dem Hammer den Wagen repariert]i hat er mit with the hammer the car repaired has he with Karl ti Karl (81)

??

[Mit Karl ti den Wagen repariert]j hat er mit with Karl the car repaired has he with dem Hammeri tj the hammer

134

T. Mu¨ller-Bardey

In conclusion, comitatives are more appropriate candidates for being restrictors than instrumentals. And this correlates with the higher underlying position of the former in comparison to the latter. It is only comitatives that are unambiguously event-related. Taking stock, we have established that within the spectrum of adverbials (Figure 3.1), the mid-range members are suitable for the restrictor; that is, temporal and local modiﬁers, to which we have added the comitative. In contrast to that, the instrumental and anything below it is not admitted to move to the Re. Pivotal in this range are manner adverbs. We have also shown that internal locatives display similar propensities. This gives the following classiﬁcation (‘low, mid, high’ refer to height in tree structure):22 (82) Restrictive vs. non-restrictive adverbials a. Restrictive mid: temporal, local, comitative, depictive b. Non-restrictive 1. high: sentence-adjuncts, mental-state manner 2. low: internal-locative, instrumental, manner The higher and mid-range adverbials are presumably adjuncts to the higher VP shell, correlating with the event, while the lower adverbials revolve around the lower VP shell, corresponding to subevents or the category of process. 3.3.4 Application to English data Once we try to transfer these results to English, we seem to be confronted with a challenge for our approach. In this subsection we will discuss these problems, and will ultimately show how they can be converted into an argument in favour of the theory of restrictive adverbial types. 3.3.4.1 As has been observed by Larson (1988), several types of adverbial show signs of being c-commanded by the object of a verb. This holds for the temporal modiﬁer and the indirect object in the following example: (83)

John gave the books to themi on each others’i birthdays.

22 It is tempting to compare this hierarchy to the one of Schultze-Berndt and Himmelmann (2004), which, however is not a straightforward exercise because of substantial conceptual differences. The two hierarchies appear to be in large parts compatible with each other. However, if we equate ‘condition or state’ in Schultze-Berndt and Himmelmann’s hierarchy with ‘depictive’ in our (82), we realize that both accounts are at variance with each other regarding the relation between manner adverbs and depictives. For the two authors, it is manner adverbs over depictives that group with circumstantial adverbials; for us, it is depictives over manner adverbs.

Adverbials and depictives as restrictors

135

This is a Principle A-effect, but we also observe Principle C-effects: (84)

*John gave the books to heri on Maryi’s birthday.

This would appear to contradict our claim that temporal and local PPs are located above the core VP. But we can reconcile both observations. Under certain conditions, the c-command relation seems not to obtain, one of these being VP-Ellipsis, as noted by Lechner (2001) ( < > designating the ellipsis site): (85) a. John [gave the books to themi] on each others’i birthdays and Mary did < > on their ﬁrst day of school b. *John [gave the books to themi] on each others’i birthdays and Mary did < > on each others’i ﬁrst day of school It is acceptable to have a binding relationship in the antecedent of VP ellipsis, as was the case with non-conjoined clauses, but as soon as this relationship shows up in the ellipsis clause as a consequence of copying, the sentence becomes unacceptable. As for Principle C, this restriction can be replicated. Copying the VP into the reconstruction site obviously does not result in a c-command relation, since no Principle C effect is visible: (86)

John [gave the books to heri] at Christmas and Jill did < > on Maryi’s birthday

In addition to VP-ellipsis, as Lechner (2001) goes on to point out, VP-Fronting is another instance where the adverbial is not c-commanded by the object, at least not obligatorily: (87)

John promised to give the book to her next year and [give the book to her]i he did ti on Mary’s birthday

The important point to emphasize in our context is that one can conclude from the above observations that beside an adjunction site for adverbials that is in the c-command range of the object, there must exist an alternative adjunction site above the object. According to Lechner (2001), the object moves to the Spec-Position of the higher V for case reasons and thus can c-command any adverbial adjoined to the lower VP shell. In contrast, an adverbial adjoined to the outer shell will not be in the c-command domain:23 (88)

[(modiﬁer) [tsubject object-1 [V (modiﬁer) [tobject-1 [V object-2]]]]]

23 An obvious question is why the higher adjunction site must be activated in ellipsis clauses and can be activated in VP-fronting. We refer to Lechner (2001) for an account for this in terms of economy conditions.

136

T. Mu¨ller-Bardey

A modiﬁer of the temporal adverbial type can be adjoined to the lower VP or to the higher VP. Hence we can stick to our proposal that modiﬁers from the lower range of the layered VP are unable to be restrictive. If adverbials move to the Re, they originate as adjuncts to the higher shell.24 3.3.4.2 We can even strengthen our point by utilizing binding data. While it is possible for the subject to bind into an adverbial in the Re, this does not hold true of objects: (89)

a. Mostly, in hisi kitchen, a cooki is the absolute ruler b. ??Mostly, in hisi kitchen, Mary asks a cooki after a recipe

The reason for this distribution is this: The restrictive part of the clause is somewhere at the higher margin of the extended VP, as we have maintained. Assuming that an adverbial like in his kitchen originates above this higher shell, it may reconstruct into its source position, facilitating binding by the subject, and still being in a position to be assigned to the Re (cf. Chierchia 1995a: 138 for a similar approach): (90)

Mostly ti a cook [in his kitchen]i is the absolute ruler

For the object in (89b), however, no such reconstruction is possible if the restrictiveness of the adverbial is to be preserved. The (ordinary) locative cannot reconstruct to a position below the object. We should expect converse effects for the internal locative, an adverbial that we have assigned to the process-related, lower VP shell. We predict that internal locative modiﬁers are obligatorily c-commanded by objects. Taking VP Ellipsis, this seems indeed to be the case: (91)

*John served himi spaghetti on Jill’s plate and Jim did < > on Carl’si plate

Principle C here does not seem to be circumventable, revealing the robustness of the low position for the internal modiﬁer.25 Hence, the theory of restrictive adverbs accords with the binding data. 24 A number of authors (Koster and Zwart 2000; Haider 2001) point out that there are indications for a left-branching order in the English VP, of the kind of ‘[ [ [ [(manner) V object] (manner)] location] time]’. Evidence stems from VP-Fronting:

(i) He wanted to give the book to Mary at her party some time next week and [give the book to her in the garden]i he did ti on her birthday If one considers the binding evidence in (83), (84) to be less important, one would here have a straightforward way to endorse the hierarchy that we maintain to hold between adverbials, i.e. time > location > manner. 25 Prosodic evidence points to the same conclusion: Maienborn (2000) notes that internal locatives attract stress with intransitive verbs, while ordinary locatives do not. Attracting stress is an object-like property. It is understandable if we assume that the lower VP shell corresponds to the stress cycle.

Adverbials and depictives as restrictors

137

3.3.5 Integrating depictives 3.3.5.1 What is the impact of the theory of restrictive adverbials for depictives? Let us restate what we know about them, by way of the following examples: (92)

a. weil er betrunken oft seine Ex-Frau anrief because he drunk often his ex-wife called ‘because (when) drunk, he often called his ex-wife’ b. weil er oft betrunken seine Ex-Frau anrief because he often drunk his ex-wife called ‘because he often called his ex-wife drunk’

For (92a), there is a reading available—the reconstructive reading—that corresponds to the sole interpretation of (92b), the version without an overtly ﬁlled restrictor. Events of drunken calling are asserted to take place ‘often’ (i.e. exceeding the degree of expectation). But equally natural is a reading in which the depictive forms the Re: a signiﬁcant subset of events of being drunk is also a set of events of calling his ex-wife. Hence, depictives, according to the criteria outlined above, are in a higher position compared to manner adverbs and other types of adverbial that are tied to the NSc. In semantic terms, this means that the depictive introduces an eventive variable that the quantiﬁcational adverb identiﬁes with that of the VP. That is, it does not represent a constitutional sub-event of the verb event, as was said to be the case for lower adverbials. Support for this reasoning comes from an insight in Rothstein (2000). She has found that depictive events are never properly included in the verb phrase event in terms of their ‘runtime’. This would be in contrast to properties of more deeply embedded modiﬁers. At this point, one has to concede that our ﬁndings seem to be at odds with a rather well-received hypothesis, according to which depictives, even subjectoriented ones, have a position usually circumscribed as ‘inside the VP’, as opposed to adjoined to IP/S. This has been argued for on the basis of constituency tests like VP-Fronting and Clefting (cf. Andrews 1982), which obligatorily include the depictive in the VP, in contrast to other types of modiﬁer like local adjuncts: (93) {In his room, Relaxed} George wanted to watch the game a. and watch the game {in his room, relaxed} he did b. and watch the game he did {in his room, *relaxed} However, note that this barely warrants the conclusion that the position of depictives is to be sought below that of local adjuncts, or, in the system of Frey (2000), between event-internal and process-related adverbials. 3.3.5.2 While binding data targeting the VP are hard to come by for depictives—as a consequence of their being contained in the VP— there are

138

T. Mu¨ller-Bardey

some indications which suggest that depictives originate in a position in the upper part of the entire VP, higher than manner adverbs at least. First, subject-oriented depictives cannot be bound by the object (Winkler 1997: 72): (94)

*Mary met every boy angry at his father. [*subject-oriented reading of angry]

Second, Complex Topicalization in German shows that evacuating the object out of the VP before fronting it is ﬁne for manner adverbials, but considerably less so for depictives: (95)

[Laut angeschnauzt]i hat Paul seinen Freund ti loudly yelled at has Paul his Friend ‘Loudly, Paul yelled at his friend.’

(96)

?

[Traurig tj angerufen]i hat Paul seinen Freundj ti sad called has Paul his friend ‘Sad(ly), Paul called his friend.’

Third, the scope test from Frey (2000) mentioned in 3.3.2 shows that depictives (here represented by a quantiﬁcational phrase ‘in every condition’), but not manner adverbs, c-command the object in the base order: They show ambiguities when following it, but not when preceding it (taking ‘in an x-condition’ to be a subject-oriented PP-depictive): (97) a. Paul hat mindestens ein Auto in jeder Verfassung Paul has at least one car in every condition gefahren [9>8, 8>9] driven ‘Paul drove at least one car in every condition.’ b. Paul Paul Auto car

hat in fast jeder Verfassung mindestens ein has in almost every condition at least one gefahren [8>9] driven

(98) a. Paul hat mindestens eine Frau auf jede Art und Paul has at least one woman in every kind and Weise umworben [9>8] manner courted ‘Paul courted at least one woman in every manner.’ b. Paul hat Paul has mindestens at least

auf in eine one

fast jede Art und Weise almost every kind and manner Frau umworben [8>9, 9>8] woman courted

Adverbials and depictives as restrictors

139

3.3.5.3 It is not only with Re-movement, but also with Pre-movement that it becomes obvious that depictives are adjoined to the higher VP shell, in contrast to manner adverbs. Pre-movement, we recall, is the kind of scrambling that does not affect quantiﬁcation and is driven by the presuppositional status of the moved item. Let us once again compare a depictive and a manner adverb example: (99) Paul hat krank selten Post bekommen. Paul has ill rarely mail received (100) Paul hat laut selten gesungen. Paul has loudly rarely sung The depictive in (99) can be interpreted as an outcome of Re-movement: ‘Few of the instances of being sick were instances of receiving mail.’ At the same time, there exists the possibility of Pre-movement. An appropriate context would be the givenness of illness in general. But in another variant of Pre-movement, the depictive would designate a temporal relation (compare also 3.2.4 ): ‘During his illness, there were few instances of him receiving any mail.’ Hence, the predicate of illness is not divided by the QA. In this sense, ‘ill’ relates to the ‘external situation’ of Chierchia (1995b), as opposed to the ‘internal’ one, which serves as a restriction. We may take the external usage as an indication that the depictive may move into the position of temporal adverbials. This type of shift is not open to manner adverbs. Not only is the restrictive reading excluded for (100), as has been shown. A reading in terms of the external event cannot be associated with sentence (100) either. This is due to the lower, sub-event position of manner adverbs.

3.4 Summary In this chapter, we have tried to lend substance to the conventional wisdom that depictives qualify an argument of the clause and simultaneously relate to its predicate. Taking recourse to the pattern of indeﬁnites, we have shown depictives—and adverbials—to be predicates of events that can form the restrictor for quantiﬁcational adverbs. We introduced a basic typology of non-feature-driven movement, distinguishing between Re-movement (to the restrictor) and Pre-movement (triggered by presupposition). In many instances, a raising of an adverbial or depictive can alternatively come about by either movement type. Not all adverbials may undergo Re-movement, which led us to develop a typology of adverbials according to their readiness to lend themselves to

140

T. Mu¨ller-Bardey

restrictor formation. We observed parallels in the order in which the adverbials merged in the VP. The highest adverbials are sentence adjuncts and the lowest are manner adverbs, instrumentals, and internal locatives—both regions are excluded from Re-formation. Mental-state adverbs are excluded from restrictiveness because they group with sentence adjuncts by inference. Adverbials of the lower range are predicates of sub-events (or processes), which prevents them from restricting quantiﬁcation over the (whole) event denoted by the main predicate. Mid-range adverbials are potentially restrictive. English data have been shown to adhere to the same regularities, even though their order is the inverse of the German order. The restrictiveness of depictives correlates with the fact that they could be shown to be merged above the range of manner adverbs. Hence, these two categories seem to be fundamentally distinct.

4 Depictive agreement and the development of a depictive marker in Swiss German dialects CLAUDIA BUCHELI BERGER

4.1 Introduction In this chapter I will provide a description of predicative and depictive agreement and marking in Swiss German dialects, applying the general typological framework of Schultze-Berndt and Himmelmann (2004 and Ch. 1, this volume) to the varieties of a single language. Swiss German dialects can be subdivided into three groups on the basis of the pattern of adjectival inﬂection and depictive marking that they exhibit. Three patterns of adjectival inﬂection and two patterns of depictive marking can be distinguished (section 4.1). After discussing agreement in gender and number of depictive secondary predicates in one dialect area (section 4.2), I will present evidence of the emergence of a depictive marker in another dialect area (section 4.3) and establish the historical relation between these two phenomena. Syntactic function and semantic range of depictive expressions containing this depictive marker will be examined in section 4.4. In this chapter, I use the term ‘depictive’ for constructions that in the German literature have also been called halbpra¨dikativ (Behaghel 1923: I x 490, III x 1150), ‘predicative attribute’ (Paul 1919: 49), or ‘copredicative’ (Plank I am very grateful to the editors, Eva Schultze-Berndt and Nikolaus Himmelmann, for many useful suggestions and comments. I also thank Balthasar Bickel, Elvira Glaser, Michele Loporcaro, Elke Ronneberger-Sibold, Stefan Sonderegger, and Iwar Werlen for commenting on preliminary versions of this chapter. I thank Annette Brechbu¨hl and Theres Gru¨ter for improving my English and, most importantly, all the native speakers, especially Joe Manser, for answering my questions. This research was supported by the Swiss National Science Foundation, project no. 11-57121.99 (Dialektsyntax des Schweizerdeutschen / Syntaktischer Atlas der Deutschen Schweiz (SADS)), which provided me with the opportunity to collect data and to write this chapter. As the research for SADS was still in progress while this chapter was being written, Table 4.1 and Map 4.1 should be considered preliminary.

142

C. Bucheli Berger

1985). With regard to the distinctions made in the introduction to this volume, ‘depictive’ is used in its broad sense, covering depictives in the narrow sense and circumstantials. The term ‘predicative’ is reserved for instances where the adjective or past participle forms the predicate together with the auxiliary sii ‘to be’ or haa ‘to have’. As deverbal predicates, past participles can be used like adjectives, although there are some restrictions on the verb class (section 4.4). Present participles will not be considered in this chapter, as they cannot be formed productively (see further section 4.4.2). 4.1.1 Swiss German Swiss German dialects are spoken in the south-western part of the Germanspeaking area. Swiss German is classiﬁed as Alemannic and Upper German, sharing some characteristics, e.g. the loss of the preterite tense, with other southern German dialects such as Bavarian or Swabian. It is divided into three subgroups: Low Alemannic, spoken in Basel and its surroundings; High Alemannic, spoken in the Jura, in the relatively ﬂat area of the Mittelland, and in the lower hills of the Alps; and Highest Alemannic, spoken in the higher Alps (See Hotzenko¨cherle 1961). This geographical situation accounts in two ways for some of the peculiarities of the Swiss German dialects. First, the isolation in the mountains has led to the preservation of many features of Old and Middle High German. For instance, the agreement of predicative adjectives with the subject has been preserved in Highest Alemannic, whereas it has disappeared in other German dialects in Switzerland, Germany, and Austria, as well as in Standard German. Secondly, geography and history show that contact with other languages has been very intensive, from the beginnings of the Alemannic settlements (after ad 500) until today. Nowadays, the Swiss German dialects, especially those in the Highest Alemannic group, are in contact with French in the west, Italian in the south, and Rhaeto-Romance in the east. Regarding the agreement of predicative and depictive adjectives, it is evident that contact with Romance languages may also have contributed to its retention (cf. Moulton 1941: 46). Swiss German and Standard German form a diglossia (Ferguson 1959). While Swiss German is used exclusively in speech situations, Standard German is the language of all written communication. Dialect is pervasive in everyday life, used by everyone on a daily basis. There is no social class to which it is restricted. Dialect is not written, with the exception of linguistic studies and informal, personal communication between younger speakers (letters, e-mails, SMS, diaries, notes, shopping lists, etc.). Thus, in writing, Standard German is the variety used almost exclusively, and it is the language of instruction and communication at school. During breaks, however, it is perfectly normal to

Development of a depictive marker in Swiss German dialects

143

speak dialect, even at university. Exceptions to the general use of dialect in spoken communication are formal situations, such as religious ceremonies, news broadcasts on TV and radio, in parliament and most political events, and so on, where Standard German is spoken. To sum up, Swiss German dialects are widely used and not endangered, in contrast to the situation in France and Germany, where dialects have a very restricted social domain of use. In addition to the traditional local dialects, we now also ﬁnd ‘koines’ covering a whole region. They lack very specialized local features in order to allow for broader communication, particularly in the large urban areas of Zu¨rich, Bern, and Basel. Being a fully functional and vital form of communication, Swiss German has changed over the last 100 years as a result of changes in society and technology. On the one hand, specialized terminology has disappeared together with the disappearance of the related objects and processes. On the other hand, loanwords from English and other languages have made their way into the dialect recently, in the same way as French words were integrated a few centuries ago. However, despite these waves of borrowing, the crucial observation is that morphology and syntax tend to remain intact. 4.1.2 Adjectival agreement and marking Swiss German dialects show adjectival agreement to different degrees. All of them, as in Standard German, have attributive agreement. That is, the attributive adjective agrees in gender and number with the noun after the indeﬁnite article in all dialects, as in en alt-e Wii (a.m.sg old-m.nom.sg wine) ‘an old wine’ and e alt-i Fla¨sche (a.f.sg old-f.sg.nom bottle) ‘an old bottle’. After the deﬁnite article, agreement follows a reduced paradigm (compare de alt Wii ‘the old wine’ and di alt-i Fla¨sche ‘the old bottle’). In some dialects the adjective also agrees with the subject when functioning as part of the predicate as in a¨r ischt alt-e (he is old-m.sg) and schi ischt mied-i (she is tired-f.sg). Map 256 in Sprachatlas der Deutschen Schweiz1 (SDS iii) shows that this predicative agreement is found only in the Highest Alemannic dialects. In the same zone, depictive constructions show agreement with the subject or the object, e.g. Milch heiss-i triichu (milk:f.sg hot-f.sg drink) ‘drink milk hot’. Table 4.1 provides an approximate synopsis of the different types of Swiss German dialects with regard to predicative, depictive and attributive agreement of adjectives and past participles. The depictive marker mentioned in the ﬁfth column will be illustrated in detail in section 4.3. For geographical details, see Map 4.1. 1

Atlas of Swiss German Dialects.

Places where no depictive marking is attested gfrornä (>1) 18–50% 51–100% gfrorni (>1) 12–50% 51–100%

N

Map 4.1 ‘You must fry ﬁsh ﬁngers frozen’. Copyright: Projekt Dialektsyntax

Development of a depictive marker in Swiss German dialects

145

Table 4.1. Adjectival agreement in Swiss German dialects Type

Attributive agreement

Predicative agreement

Depictive agreement

Depictive marker

Geographical area

A B C D E

þ þ þ þ þ

þ þ

þ þ

þ

Mittelland, Jura Alps (Wallis) North-east Not attested Not attested

Dialects of type A, spoken in the Mittelland and Jura, resemble Standard German in that they exhibit only attributive agreement but no agreement in predicative or depictive constructions. Dialects of type B have both depictive and predicative agreement, in addition to attributive agreement; these are the dialects of the Alps (Wallis), discussed in section 4.2. The north-eastern dialects of type C have a peculiar feature in that they do not exhibit agreement outside the attributive construction, but have a specialized depictive marker (see section 4.3). I have not yet encountered a dialect which follows either of the following logically possible agreement types: ‘type D’, marking attributive and depictive but not predicative agreement, or ‘type E’, marking attributive and predicative but not depictive agreement. In places where nowadays the dialect is in the process of changing from type B to type A, a mixed pattern of type D or E has not yet been observed with native speakers who are well anchored in the local community and speak the local dialect. For instance, during direct interviews in Muotathal (a canton of Schwyz), two dialectal speakers consistently followed a type A pattern whereas another consistently followed a type B pattern. No native speaker produced a mixed pattern of type D or E. Further research in the transition zones is required to conﬁrm this observation (cf. Bucheli Berger and Glaser, 2004). The currently documented distribution leaves open all possibilities for implicational relationships between depictive, predicative, and attributive constructions. Two types of implicational hierarchy have been discussed in the typological literature: 1. Designing a ‘semantic map’ for this constructional domain, van der Auwera & Malchukov (Ch. 13, this volume) propose the following ordering: pred(icative)—dep(ictive)—attr(ibutive). This ordering excludes an expression strategy which applies only to predicative and attributive but not to depictive constructions. Their comparison of various Germanic languages, such as Swedish, Norwegian, Standard

146

C. Bucheli Berger

German, and English, further suggests that this ordering in fact reﬂects the following implicational hierarchy with regard to the realization of agreement marking: pred>dep>attr. 2. Plank (1985: 173) claims that adjectives which can be used depictively can always be used predicatively as well. Applying this claim to agreement, we get the following implicational relationship: dep>pred. If a language or a dialect shows depictive agreement, then predicative agreement must also be present. If there is no predicative agreement, there is no depictive agreement either. Put in more general terms, the claim would be that if there is a marking strategy for depictive adjectives, the same marking strategy has to be found with predicative adjectives as well. Obviously, the existence of a depictive marker in dialects of type C contradicts this generalization. With regard to the agreement facts, synchronic Swiss German data do not provide clear-cut evidence supporting either of the two implicational hierarchies. The Swiss German agreement types A, B, and C are perfectly consistent with both hierarchies, but decisive data for type D or E are missing. Type D, if attested, would constitute evidence for the ﬁrst implicational hierarchy; type E, on the other hand, would support the second implicational hierarchy. 4.1.3 Data sources The data used in this chapter come from native speakers who use the dialect every day, who are well anchored in the local community, and whose parents spoke the same dialect. Most of the depictive examples were in fact overheard by myself in ‘participant observation’ in the area of Appenzell, in restaurants, while travelling, and while visiting relatives. In addition, depictive examples were taken from grammars and the SDS (iii: 257), provided that they seemed representative of natural speech. In this way, a small corpus of depictive examples was created. In a second step, these examples were submitted to the judgement of native speakers in different procedures. I indicate the source of the judgement by means of the following abbreviations at the end of each example: SADS 1: In multiple-choice questions in written questionnaires, native speakers were confronted with several variants and asked to choose the one corresponding to their dialect. This procedure was repeated with ﬁve to ten native speakers (2,710 in total) in each of the 385 places from the SADS place grid. A short story before each question provided a context for the sentence under investigation. Among other phenomena,

Development of a depictive marker in Swiss German dialects

147

four depictive expressions were presented to the speakers in the ﬁrst and second questionnaire. Two contained a subject controller (‘These children came to the wedding uncombed’, ‘Being very ill, this child went on a trip’), two an object controller (‘You must drink the milk hot’, ‘You have to fry ﬁsh ﬁngers frozen’), alternating adjectives and past participles. SADS wanted to establish a reliable picture of the geographical distribution of different depictive constructions. Table 4.1 and Map 4.1 reﬂect the (still preliminary) results of this investigation. SADS 2: Additionally, in the Wallis, the SADS team talked to the same native speakers in direct interviews in order to compare these oral answers with the written answers of the questionnaires. The speakers were asked to provide the correct form and especially the correct pronunciation of the sentences under investigation. Fieldwork CBB: The depictive expressions collected during my ‘participant observations’ in Appenzell, found in grammars or invented by myself, were submitted to the judgements of a few native speakers, one in the village of Appenzell, one in Trogen and one in Herisau, in order to cover the entire area of Appenzellerland. In section 4.3 I will present the depictives in the dialect of the village of Appenzell, where one native speaker, Joe Manser, kindly provided me with the correct pronunciation and spelling for these examples. Furthermore, a special questionnaire testing differences between adverbs and depictives was sent to twentyfour native speakers in ten places in the area of the depictive marker. As for my ‘participant observation’, talking to women in the kitchen and gossiping with neighbours proved to be particularly fruitful situations for obtaining spontaneous depictive expressions. In the former case, advice regarding the way in which something has to be cooked, fried, etc. was provided; in the latter, it was the expression of being shocked by the fact that someone had done something while not being in the usual or correct state, which provided good depictive examples. In the former case, it is evident that modality2 is constantly present in depictive constructions and could be considered as an accompanying feature of depictive pragmatics. While ‘kitchen’ depictives describe an expected or usual state, ‘gossiping’ depictives rather refer to an unexpected or surprising state. In the latter case, separate negation of the depictive secondary predicate through the preﬁx un-/on-/oo- ‘un-’, e.g. in ‘unwashed’, ‘uncombed’, etc., is very frequent since this construction allows the speaker to suggest the required state at the same time (see e.g. (12), (37), 2 Modality may even have an inﬂuence on the grammaticality of depictive constructions (see e.g. Simpson, Ch. 2, this volume, for English).

148

C. Bucheli Berger

and (39)). Furthermore, I observed that depictives are often found in narratives, when the speaker wants to insert a short description (see example (17)). As there is no ofﬁcial orthography for Swiss German dialects, I adopt the recommendations by Dieth (1986), except that I do not indicate open vowels. All errors in orthographic representation are, of course, mine and are not the responsibility of the native speakers. In the following sections, I discuss two of the three adjectival agreement types shown in Table 4.1. Regarding type B, I present the different constructions that show agreement in gender and number in section 4.2. I limit the data to the dialects of the Wallis. This short description should allow for a better understanding of the historical relationship of the type B pattern to the emergence of the depictive marker in the type C pattern, which will be discussed in section 4.3. I will not examine type A, as it shows no agreement nor special depictive marker, and in that respect is similar to Standard German (Plank 1985; Schultze-Berndt and Dawuda 2001).

4.2 Adjectival agreement in the south-west of Switzerland (Wallis) Adjectival agreement is an old feature inherited from Old High German, which had agreement in case, gender, and number in attributive, predicative, and depictive function (Braune 1987: x 247; Behaghel 1923: i x 138).3 In Middle High German predicative and depictive agreement is very rarely attested (Paul 1919: 49) and has tended to disappear. Dialects of type A joined this development and lost adjectival agreement, apart from attributive agreement. Dialects of type B were very conservative, and preserved agreement with some modiﬁcation. The forms are well described in attributive or predicative use (Wipf 1910; Bohnenberger 1913; Stucki 1917; Brun 1918) and also in depictive function (Clauss 1929; Henzen 1927, Hotzenko¨cherle 1934; Szadrowsky 1936; Fischer 1960; Dauwalder 1992). Recently, the loss of agreement and thus dialect change has been investigated by a native speaker (Fuchs 1993) in a sociolinguistic framework for the village of Steg in the Wallis. In this section, I present the range of constructions exhibiting adjectival agreement in gender and number. I demonstrate that agreement in depictive constructions is not an isolated phenomenon but simply part of an adjectival agreement system in the Highest Alemannic dialects. The following data illustrate the dialects of the Wallis, a valley that still has very intense use of adjectival agreement. 3

Forms without agreement co-ocurred in all constructions, i.e. variation was already present.

Development of a depictive marker in Swiss German dialects

149

4.2.1 Agreement paradigms in the Wallis Following Wipf (1910), there are two agreement paradigms for attributive adjectives, a full paradigm and a reduced one, distinguishing number, gender, and, regarding case, the dative and the genitive from a single form representing the former nominative and accusative. Additionally, the bare form of the adjective, without inﬂection, can occur. The actual situation of attributive agreement has not been investigated, only the predicative and depictive forms. The predicative and depictive agreement paradigm resembles the full one, without dative or genitive forms, as there is no dative or genitive depictive agreement. It contains only the single form in which the former nominative and accusative fall together: (1) Predicative and depictive agreement paradigm in the Wallis m.sg -e/-ne or -a¨/-na¨ or -a/-na f.sg -i/-ni n.sg -s pl -i/-ni The form without the nasal consonant is used with adjectives and with ‘weak’ past participles ending in -t, while the form with the nasal consonant occurs with ‘strong’ past participles, with vowel change in the stem (Ablaut and Ru¨ckumlaut) and ending in a vowel. For the exact geographical distribution of the plural forms, see SDS iii: 256. In a secluded part of the Wallis, the Lo¨tschental, the agreement paradigm still distinguishes between nominative and accusative in the singular masculine and feminine: nom.m.sg -e/-ne or -a¨/-na¨; acc.m.sg -n/-ø; nom.f.sg -i/-ni; acc.f.sg -a/-na. 4.2.2 Predicative agreement In this section, I present two constructions showing predicative agreement. One consists of sii ‘to be’ plus predicative adjective, and the other is the resultative construction4 with sii ‘to be’ or haa ‘to have’ plus past participle of the verb. This in turn is closely related to the perfect construction, which has the same constituents but differs from the resultative construction in the absence of agreement. 4 The term ‘resultative’ refers in this chapter only to the aspecto-temporal value focusing on the result of an action that has just happened. It is also called Zustandspassiv ‘passive of state’ but is not identical to the passive (which is formed in some of the dialects of type B with the auxilary choo ‘to come’ þ past participle (þ agreement)).

150

C. Bucheli Berger

Most of the adjectives used predicatively with the auxiliary sii ‘to be’ show agreement with the subject: Wallis, Visperterminen ¨ r ischt alt-e (2) a. A he is old-m.sg ‘He is old.’ (SADS 2)

b. Schi ischt mied-i. she is tired-f.sg ‘She is tired.’ (SADS 2)

However, the presence or absence of agreement can be distinctive on a semantic-syntactic level (see Fuchs 1993) for certain adjectives such as schwer ‘heavy/difﬁcult’ or hert ‘hard/difﬁcult’. Depending on the lexical meaning of the subject, the semantics of the adjective adapts. With a concrete subject, the adjective schwer has the concrete meaning ‘heavy’ and the adjective agrees with the subject: Wallis, Visperterminen (3)

Das Jungi ischt schwer-s. this child:n.sg is heavy-n.sg ‘This child is heavy.’ (SADS 2, see Fuchs 1993)

If the subject is abstract, the meaning of the adjective is abstract, i.e. ‘difﬁcult’ for schwer, and the adjective does not agree with the subject: Wallis, Visperterminen ¨ fgab (4) D¼U ischt schwer (*-i). the¼task/job:f.sg is heavy/difficult (*-f.sg) ‘This job is difficult.’ (SADS 2, see Fuchs 1993) These two examples demonstrate that adjectival agreement can assume a specialized function at the syntax–semantics interface if the adjective allows for two interpretations. Closely related to this phenomenon is the case of telic verbs, which also show distinctive agreement in constructions with the auxiliaries sii ‘to be’ and haa ‘to have’ plus past participle. If a resultative meaning is intended, past participles usually agree with the respective controller. If a perfect or narrative past meaning is intended, there is no agreement. The following examples illustrate this distinction. SADS had asked native speakers to describe a picture of a basket which had fallen down. As both meanings, the resultative and the perfect/past, could potentially be activated5 in the native 5 For problems with the elicitation of the resultative construction in a written questionnaire, see Bucheli and Glaser (2002). In direct interviews, working with a picture showing broken, fallen, emptied, and sewn items in a child’s room was a very successful strategy (see also Fuchs 1993).

Development of a depictive marker in Swiss German dialects

151

speaker’s mind, one native speaker from the Lo¨tschental in the Wallis produced an inﬂected past participle, i.e. a resultative: Wallis, Lo¨tschental (5) Dr¼Chorb ischt uis-gg-lort-a¨. the¼basket:m.sg is out-pst.ptcp-empty-m.sg ‘The basket is in an emptied state.’ (SADS 1) Another speaker from the same place gave a sequence of events in the perfect, which serves the function of past narrative tense. In this case, the past participle cannot inﬂect: Wallis, Lo¨tschental (6) Dr¼Chorb ischt um-gg-falln(*-a¨) und the¼basket is over-pst.ptcp-tip (*-m.sg) and d¼Epfl sind usa-gghid (*-i). the¼apple:pl are out-pst.ptcp-fall (*-pl) ‘The basket has tipped over and the apples have fallen out.’ (SADS 1) The following sentences illustrate that the same aspectual distinction applies to transitive telic verbs. In order to probe the position of the predicate’s constituents in subordination, the SADS presented native speakers with sentence (7) without agreement, i.e. . . . gizalt het ‘paid has’ and positional variants of it such as . . . het gizalt ‘has paid’, also without agreement. Surprisingly, some native speakers from the Wallis corrected the writing of the past participle so that it would agree with the direct object ts Auto ‘the car:n.sg’. This means that some of the native speakers did not read the intended narrative past meaning (without agreement) into the sentence, but rather a descriptive resultative meaning, which requires agreement, as in (7): Wallis, Agarn (7) Ich ha¨ kei Ahnig, ob schi ts¼Auto I have no idea if she the¼car:n.sg scho gi-zalt-s het. already pst.ptcp-pay-n.sg has ‘I have no idea whether she has already paid for the car (i.e. she has the car in a paid state).’ (SADS 1) The following example illustrates that the resultative function was marked in the same way, i.e. by the presence of agreement, in Old High German.

152

C. Bucheli Berger

Old High German (8) Phigboum habeta sum gi-flanzot-an in fig-tree-m.sg has-past one pst.ptcp-plant-acc.m.sg in sinemo uuingarten. his wine garden ‘Someone had a fig tree planted, i.e. possessed a planted fig tree in his wine garden.’ (Tatian (AD 830), Sievers 1892: 102,2, s. 146, line 23) With the last examples, I have shown that predicative agreement in the dialect of the Wallis is an old feature, yet—more importantly—that it has now acquired its own regularities and semantic-syntactic constraints. 4.2.3 Depictive and verbal complement agreement Let us now consider the depictive construction in the Wallis. Depictive adjectives and past participles always show agreement with the controller in gender and number. That is, there are no semantic-syntactic constraints similar to the ones described above for the predicative constructions. The following four examples are all from Visperterminen. The ﬁrst two illustrate depictive agreement in reference to the object of the main clause: Wallis, Visperterminen (9) Du¨ mu¨oscht d¼Milch de heiss-i triich-u. you must the¼milk:f.sg (then) hot-f.sg drink-inf ‘You must drink the milk (while it is) hot.’ (SADS 1, based on an example in SDS iii. 257) (10) Fisch-sta¨b-jini mu¨oscht doch gfror-ni abraat-u. fish-finger-dim.pl must:2.sg but pst.ptcp-freeze-pl fry-inf ‘But you have to fry fish fingers (while they are still) frozen.’ (SADS 1) In the following two examples, the depictive agrees with the subject of the main clause: (11) Aso chrank-s ischt das Chind uf d¼Reis ggangu! so ill-n.sg is this child:n.sg on the¼trip:f.sg pst.ptcp:go ‘While [s]he was so ill, this child went on a trip [with his/her class].’ (SADS 1, based on an example in Sonderegger and Gadmer 1999) (12) Un-gg-schtroolt-i sind d¼Jung-ini zer not-pst.ptcp-comb-pl are the¼children-pl to.the Hochzit cho! wedding pst.ptcp:come ‘These children came to the wedding uncombed.’ (SADS 1)

Development of a depictive marker in Swiss German dialects

153

Controllers with syntactic functions other than subject or object were found neither in the written investigations of the SADS nor during the direct interviews and elicitation of SADS ﬁeldwork in the Wallis. This corresponds to the semantic restrictions observed for Standard German (Himmelmann and Schultze-Berndt (Ch. 1, this volume)) and the dialect of Cologne (Schultze-Berndt and Dawuda 2001), where oblique controllers are rarely observed but are considered to belong to the domain of unplanned speech and limited acceptability. The preceding examples have to be distinguished from the following example, where the adjective nass ‘wet’ agrees with the object controller das ‘that-n.sg’: (13) Du¨ mu¨oscht das nit nassu-s mach-u! you should that:n.sg not wet-n.sg make-inf ‘You shouldn’t make that wet!’ (SADS 2, after an example in Szadrowsky 1936) The adjective nass ‘wet’ in this example cannot be omitted, since it conveys the semantic core of the causative predicate nass machu ‘to make wet’. Therefore, it cannot be considered a depictive (which is always optional— see Schultze-Berndt and Himmelmann 2004: 65). Instead, this example instantiates a different construction, i.e. a verbal complement construction, which is a third type of construction where non-attributive adjectives agree with a controller. To sum up this and the preceding section, the examples from the Wallis illustrate that depictive agreement in these dialects is embedded in the adjectival agreement system, together with predicative and verbal complement agreement. Details of the adjectival agreement paradigm6 and of the range of its uses, especially for verbal complements, remain a matter for further investigation. Dialects of type B thus have preserved an old feature of Old High German—predicative and depictive agreement. However, nowadays they make use of this feature in novel ways, in particular in the predicative domain.

4.3 The depictive marker in the north-east of Switzerland (Appenzellerland) In this section, I provide a description of depictive constructions involving a genuine depictive marker in the north-east of Switzerland, more precisely in 6 In some places in the Wallis there is a strong tendency to use the singular neuter sufﬁx for all human singular referents, as the singular neuter pronoun is often used to refer to them.

154

C. Bucheli Berger

Appenzellerland (dialect type C). Following a brief review of the previous literature (section 4.3.1), the syntactic function of the depictive marker is discussed in section 4.3.2. In section 4.3.3, the diachronic relation between the depictive marker and the agreement facts discussed in section 4.2 is established. Section 4.3.4 contains a short note on the restrictive marker ase/asa. If not otherwise indicated, the examples in this and the remaining sections come from the village Appenzell. 4.3.1 Previous literature Regarding type C dialects, two dictionaries—Tobler (1837: 27) for the dialects of Appenzellerland, and Idiotikon (1881: 200–2) for Swiss German in general—list a few examples of depictives with a depictive marker, both under the entry for also/ase/asa ‘so, so much, etc.’. On its map for predicative agreement, SDS (1975: iii. 256) identiﬁes some villages as having ‘generalized’ depictive ‘agreement’, based on some spontaneous utterances by native speakers. In Appenzeller Sprachbuch (Sonderegger and Gadmer 1999: 35, 147), the authors emphasize the originality of the dialects of Appenzell in this regard with some examples of depictive adjectives and past participles. As the investigations of SADS have revealed, the area of Appenzellerland represents today’s heartland of the depictive marker; depictive marking is still well anchored in the everyday use of this dialect (see (14) and Map 4.1). (14) Du moscht d¼Mo¨lch abe waam-e trink-e! you must the¼milk:f.sg (but) hot-dpct drink-inf ‘You have to drink the milk hot.’ (SADS 1, based on an example in SDS iii. 257) Depending on the phonology of the particular dialect, the forms of the depictive marker are as in (15): (15) -e/-ne or -a¨/-na¨ or -a/-na The form without the nasal consonant is used with most of the adjectives and with ‘weak’ past participles ending in -t; the form with the nasal consonant occurs with some adjectives and with ‘strong’ past participles with vowel change in the stem (Ablaut) and ending in -e (see also Sonderegger and Gadmer 1999: 147). The distribution rules of the allomorphs of the depictive marker resemble those of the sufﬁxes of the agreement paradigm in the dialects of the Wallis (see (1)). Moreover, the allomorphs of the depictive marker in (15) are identical with the allomorphs of the masculine singular forms in the agreement paradigm in (1). This suggests a diachronic relationship which is further discussed in 4.3.3.

Development of a depictive marker in Swiss German dialects

155

4.3.2 Syntactic function of the depictive marker Considering example (14) in the previous section, we observe that the depictive secondary predicate waam ‘hot’ refers to Mo¨lch ‘milk’, which is feminine singular. Despite this gender, waam carries the sufﬁx -e, which would indicate masculine singular in the dialects of the Wallis. In those dialects the sufﬁx -i would render feminine singular. The next example shows that this phenomenon is not restricted to feminine singular. The same depictive sufﬁx appears on troche ‘dry’ and nass ‘wet’, both referring to Gra¨a¨s ‘grass’, which is neuter singular: (16) Chaa me s¼Gra¨a¨s troch-ne besse schniide as can one the¼grass:n.sg dry-dpct better cut than nass-ne? wet-dpct ‘Can one cut the grass better when it is dry than when it is wet?’ (fieldwork CBB) In the dialects of the Wallis, the correct sufﬁx would be -s. As multiple examples in the following sections will illustrate that all depictive secondary predicates carry the same sufﬁx, regardless of the gender and number of the controller, I limit the discussion to these two examples. Taking into account these observations, I think it is justiﬁed to call the sufﬁx a ‘depictive marker’. Note that in further contrast to the situation in the Wallis, where predicative, depictive, and verbal complements carry agreement in gender and number, no predicative agreement occurs in the dialect of Appenzell (cf. map in SDS iii. 257; Sonderegger and Gadmer 1999: 147). A predicative adjective shows no inﬂection or other sufﬁx. Neither do past participles which are used as resultatives with the auxiliaries sii ‘to be’ or haa ‘to have’ show any agreement. There is also no agreement or marking of verbal complements. Now, the function of the depictive marker in depictive constructions cannot be the identiﬁcation of the controller in the main clause, as it always has the same form. The explicit marking of a depictive predicate, rather, serves the purpose of distinguishing the depictive function from the predicative or the adverbial function of ‘adjectives’, because the latter two functions have no agreement or marker in dialects of type C (nor in dialects of type A (see Table 4.1), nor in Standard German). In (17), for instance, it is the depictive marker alone which conveys the distinction between an adverbial and a depictive reading. A farmer, standing with me at a place where there used to be a wooden bridge, wanted to tell me that the bridge had been swept away as a whole. He pointed to the place alongside the river where

156

C. Bucheli Berger

the bridge was lying in one piece and said the following: (17) Het d¼Brogg gaanz-ne devoo-gg-schwo¨mmt. has the¼bridge:f.sg whole-dpct away-pst.ptcp-sweep ‘[The water of the river] swept the bridge away as a whole.’ (fieldwork CBB) The use of the depictive marker -ne forces a depictive reading—that is, ganz ‘whole’ is interpreted as referring to d Brogg ‘the bridge’. If the farmer had focused on the totality of the action, he would have used ganz as an adverb—with no marking—which would then have meant ‘totally, completely’. And he probably would have pointed to the place where we were standing (i.e. the original location of the bridge). In the adverbial reading, the bridge could have been swept away in several stages, but without remainders at the original location. The following example shows that when an adverb modiﬁes a depictive secondary predicate, the position and presence of the depictive marker establish the distinction between them. In this case the adverb precedes the depictive: (18) Denn bin-i scho zfrede ggsee, wenn de¼Lismer then am-I just happy pst.ptcp:be when the¼sweater frisch gg-wa¨sch-ne ode guet freshly pst.ptcp-wash-dpct or well uus-gg-lo¨ffted-e wide paraad ggsee ischt. out-pst.ptcp-air-dpct again ready pst.ptcp:be is ‘Then I was happy enough when the sweater was ready again, freshly washed or well aired.’ (fieldwork CBB) 4.3.3 Diachronic origin As mentioned in 4.3.1, the forms of the depictive marker are identical with the forms of masculine singular in dialects with agreement. In dialects with a depictive marker, there is no predicative agreement or other type of marking on adjectives used predicatively. It is thus evident that there must be a diachronic relationship between the loss of adjectival agreement and the grammaticalization of the depictive marker in exactly the same form as the masculine singular in the agreement paradigm. It is usually assumed (e.g. in SDS iii. 257; Hotzenko¨cherle 1961) that agreement of predicative and depictive adjectives and past participles with the controller is a feature which used to be much more widespread in the Alemannic dialects than it is today or was at the beginning of the twentieth century, as represented on map SDS iii. 256. Its existence in Old and Middle High German—discussed in section 4.2—supports this basic assumption. As Appenzellerland is adjacent to the Highest Alemannic zone—where

Development of a depictive marker in Swiss German dialects

157

agreement is still productive—it is plausible that, in some way, agreement was present in the dialects of Appenzell or the Northeast of Switzerland at the onset of further grammaticalization. Concerning the process of further grammaticalization, there are two plausible scenarios. The ﬁrst scenario assumes, as a ﬁrst step, the loss of predicative and verbal complement agreement. This is precisely the analysis proposed by Behaghel (1923: i x140) and by Paul (1998: x393) for Middle High German. The second step involves the reanalysis of depictive agreement. Such a reanalysis would be a necessary consequence if one assumes that depictive agreement implies the existence of predicative agreement (see section 4.1.2 above). This would make it impossible for depictive agreement to exist on its own. A tendency towards a system-internal change would then have existed. In the process of restructuring, the masculine form won over the others, perhaps because it was the least marked one (Wolfgang Ulrich Wurzel, p.c., 2001). The result was a grammaticalized depictive marker looking exactly like the masculine singular in the agreement paradigm. In the second scenario, the ﬁrst step consists of a generalization of all agreement marking into one form (for whatever reason). As a result, the form of the masculine singular would be used for all numbers and genders in all constructions, i.e. predicative, depictive, and perhaps also the verbal complement. The second step would involve the loss of this generalized marking in predicative and verbal complement constructions. It is evident that this scenario is consistent with the traditional diachronic analysis of the depictive marker as ‘generalized agreement’ (See SDS iii. 257 or Hotzenko¨cherle 1961). The following data from older sources could support this hypothesis, since they provide evidence of the former existence of predicative constructions with ‘wrong’ agreement, i.e. a ‘generalized agreement’. The masculine singular form is used with non-masculine singular controllers: (19) Es ist asa us-g-macht-a. it:n.sg is totally upon-pst.ptcp-agree-suff (m.sg?) ‘It has been fully arranged.’ (original translation by Tobler: ‘Es ist vo¨llig ausgemacht’: 1837: 27 (original spelling)) Dialect of Thal, canton of St Gallen (20) [die Garben sind] tro¨sch-ne. the sheaf:pl are pst.ptcp:thresh-suff (m.sg?) ‘The sheaves are threshed.’ (SDS spontaneous material7 (Dieth spelling)) 7 I am very grateful to Rudolf Tru¨b, editor of the SDS, who helped me ﬁnd this example in unpublished spontaneous material from the SDS archives. The original spelling of this example was transferred into Dieth spelling.

158

C. Bucheli Berger

However, these two examples constitute the only relevant data supporting this scenario so far. In fact, (19) is taken from a dictionary published in 1837, where it is said to be representative of all dialects of Appenzellerland, though this may well be a misprint or misunderstanding. Example (20) is to be taken more seriously, as it comes from a spontaneous utterance noted during interviews of the SDS in Thal (canton of St Gallen) some sixty years ago. Thal is situated in the immediate surroundings of Appenzellerland. In its neighbouring village, Rheineck, the depictive marker is still productively used by two out of six native speakers today (SADS database). However, as interviewers and native speakers of the SDS concentrated on lexical material, (20) could also represent the beginning of an unﬁnished depictive construction. Therefore, as long as we have no more historical evidence, the existence of a stage involving generalized predicative use of the sufﬁx -e/-ne, -a¨/-na¨, or -a/-na remains a hypothesis. Regarding the place and time of the grammaticalization of the depictive marker, there is no clear evidence that it took place in Switzerland itself (i.e. after ad 500, the time of the ﬁrst Alemannic settlements in Switzerland), since the depictive marker can also be found in Bavarian (Merkle 1975: 171; Zehetner 1985: 133; Bucheli Berger and Glaser (2004); Birgit Ro¨der, p.c.; Robert Hinderling, p.c.) and Vorarlberg German (Eugen Gabriel, p.c.). There is no conclusive evidence as to whether monogenesis or polygenesis should be assumed. Furthermore, Bavarian depictives are still awaiting a detailed description which would allow for a comparison with the Swiss depictive marker on the level of syntactic function and semantic range. It is probable that the process of reanalysis took place at the end or just after the Middle High German period because, as mentioned above, late Middle High German still showed some cases of agreement (Behaghel 1923: i x140; Paul 1998: x393) but some ‘generalized agreement’ with the masculine singular sufﬁx appeared (Elvira Glaser, p.c.). And it is probable that it happened before the Reformation because all varieties of Appenzell have the depictive marker (see Map 4.1). The fact that there is a major religious separation between Innerrhoden (Catholic) and Ausserrhoden (Protestant) would have militated against one group accepting an innovation of the other. 4.3.4 The restrictive marker ase/asa The depictive sufﬁx often tends to co-occur with the preposed word ase/ asa ‘so, so much’, a ‘restrictive marker’ following the terminology of Schultze-Berndt and Himmelmann (2004: 91–3), and Himmelmann and Schultze-Berndt (Ch. 1, this volume). Ase/asa also has other functions, most

Development of a depictive marker in Swiss German dialects

159

of which are of demonstrative character and convey a similar meaning of ‘so, so much, in this way, same’ (cf. Idiotikon 1881: 200–02): (21)

Ase mosch daa mache! so must:2.sg this do ‘You have to do this in this way.’

(fieldwork CBB)

(22) E het wider ase blo¨o¨d tue. he is again so/same stupid pst.ptcp:do ‘He again behaved in the same stupid way.’ (fieldwork CBB) As a restrictive marker on depictive secondary predicates, it is not used in the same way in every village in the depictive marker zone. For example, in the dialect of the village of Appenzell itself, in the centre of depictive marking, ase seems to be optional. When it is present, it contributes to the meaning by intensifying the value of the depictive secondary predicate and may be translated as ‘so, so much’ throughout: (23)

Ase chrank-ne ischt da Go¨o¨fli of d¼Reis ggange! so ill-dpct is this child:n.sg on the¼trip pst.ptcp:go ‘While (s)he was so ill, this child went on a trip [with his/her class].’ (SADS 1, based on an example in Sonderegger and Gadmer 1999)

In the dialect of Diepoldsau (canton of St Gallen), on the border with Austria, the restrictive marker asa seems to be obligatory. Native speakers constantly added it when they were confronted with depictive examples not containing it. In some contexts, it can be translated as ‘so, so much’ (as in (24)), but it can also appear in contexts where this intensiﬁed meaning makes little sense and asa remains untranslated (as in (25)). Diepoldsau (24) Du moascht d¼Milch aber asa hoass-a trink-a! you must the¼milk:f.sg (but) so.much hot-dpct drink-inf ‘You have to drink the milk (so) hot.’ (SADS 1) (25) Asa restr as to.the ‘These

on-gg-schtra¨a¨lat-a sind di Kiand not-pst.ptcp-comb-dpct are these children:pl Hoarzig koo! wedding pst.ptcp:come children came to the wedding uncombed!’ (SADS 1)

From a diachronic perspective, the occurrence of ase/asa as a preposed restrictive marker in depictive constructions is not surprising. It derives from Old and Middle High German also ‘totally, so, so much’, which also served as

160

C. Bucheli Berger

a preposed restrictive marker. The Bavarian equivalent oisa seems to have the ¨ B 1963: 152; Merkle 1975: 171; Zehetner 1985: 133; BWB: 300, same function (WO 302; Bucheli Berger and Glaser (2004)).

4.4 Syntactic function and semantic range of depictive constructions containing the depictive marker The following observations hold for depictive constructions containing the depictive marker in the dialect of Appenzell, type C. Most of the observations are probably also correct for dialects of type A and B but, as they could differ in some details, my claims are restricted to Appenzell. The syntactic function which the controller in the main clause is allowed to assume is limited to that of the subject or the object. Reference to an indirect object, carrying dative morphology, is blocked since, in this case, the subject would still be interpreted as the controller. Nevertheless, depictive adjectives can refer to experiencer subjects coded morphologically as dative (see example (29) below), while past participles are not encountered in this context. The dialects of Appenzell thus only allow for a very limited syntactic range of depictives with the depictive marker. This range covers mainly the ﬁrst two positions of the accessibility hierarchy for grammatical relations (Keenan and Comrie 1977).8 The verb of the main clause can be either transitive9 or intransitive. The assumption by Schultze-Berndt and Himmelmann (2004: 63) that the main predicate is usually a verb of motion, ingestion, affect/manipulation, or change of state can be fully conﬁrmed. The depictive secondary predicate can occupy two positions in the main clause. Either it is placed in clause-initial position, where it is strongly stressed (see e.g. (37)); alternatively, it can occur in the so-called Mittelfeld10 (‘middle ﬁeld’), after the direct object, where it is also stressed but less so than in initial position (see e.g. (26)). Looking through the examples, the reader will notice the limited semantic range of depictives with the depictive marker. It covers only a part of what Schultze-Berndt and Himmelmann (2004: 63–5) consider as typical depictives. The depictive marker in Appenzell is found with depictive secondary 8 Without a depictive marker, depictives controlled by indirect objects and adjuncts are also rejected by native speakers. 9 I follow Helbig and Buscha (1994: 53) in considering a verb transitive if it has an accusative argument. 10 The Mittelfeld is a frame which is deﬁned on the left-hand side by the ﬁnite (inﬂected) verb form, which has to occur in second position, and on the right-hand side by the past participle or inﬁnitive, placed towards the end of the sentence.

Development of a depictive marker in Swiss German dialects

161

predicates encoding a physical state, condition, or colour (cold/hot, old/ young, raw, wet/dry, ill, frozen, peeled, rotten, green, etc.). With regard to bodily posture, it is difﬁcult to interpret the grammatical status of the multiple answers translating expressions such as Standard German rittlings ‘astride’ and ba¨uchlings ‘belly-down’. They seem to belong either to the adverbial domain (querawa¨g ‘crossways’, verkehrt ‘in the wrong way’, buuchvora¨ ‘belly in the front’, de langeweg ‘the long way’) or to the depictiveadjectival domain (platt ‘ﬂat’). In the latter case, the elicited translations also included present participles (e.g. bu¨u¨chlige ‘belly-ig-dpct’—see example (38)). It thus appears that expressions of bodily posture constitute a transition zone in that some expressions are constructed adverbially, others depictively. Role, function, and life stage (‘as a young woman’ or ‘as a page’) are expressed—in Swiss German in general as well as in the dialect of Appenzell—by means of the preposition as/als ‘as’ without the depictive marker, and are therefore identical to Standard German (cf. Schultze-Berndt and Himmelmann 2004: 64). In the following sections, I illustrate these general statements with various examples, discussing depictive adjectives (4.4.1) and depictive past participles (4.4.2) separately. 4.4.1 Adjectives As mentioned above, adjectives expressing a state or condition are often found as depictive secondary predicates. We saw in the previous sections that they occur with either an object controller ((16) and (17)) or a subject controller (23) in the main clause. It is common to make a pun on the two readings of a depictive expression when the reference of the depictive secondary predicate is ambiguous, as in (26). Note that this would be impossible in a dialect where the depictive has to agree with its controller. (26) I ha miini Muete no ledeg-e kennt. I [m.sg] have my mother:f.sg still single-dpct pst.ptcp:know ‘I knew my mother when I was unmarried.’ or ‘I knew my mother when she was unmarried (I am an illegitimate child).’ [play on words by Joe Manser: fieldwork CBB] Colour depictives with object controllers occur; subject controllers have not been encountered. Talking in the garden in autumn: (27) Vo¼s gfru¨u¨t, wo¨r¼i die Tomaate before¼it freeze:3.sg would¼I these tomato:pl

162

C. Bucheli Berger denn no gru¨en-e eweg nee. then still green-dpct away take:inf ‘Before it freezes [i.e. it gets too cold], I would take away these tomatoes [from the plant] while they are still green.’ [Otherwise they will get black and will rot. Inside the house, where it is warm, they will get red and will be edible.] (fieldwork CBB)

About a rare habit because one usually drinks coffee with milk in Appenzell: (28) Si sa¨a¨t, eren Vatte hei de¼Kafi all schwaz-e she says her father has:sbjv the¼coffee always black-dpct trunke. pst.ptcp:drink ‘She says that her father always drank his coffee black.’ (fieldwork CBB) Adjectives as depictive secondary predicates can refer to an experiencer (me ‘1.sg.dat’) even if these are morphologically marked as dative: (29) Ase mu¨ed-e ischt me daa denn nome en Sii So tired-dpct is i.dat that then no.more in.the mind choo. pst.ptcp:come ‘Being so tired, I just couldn’t think of that any more/that did not come to my mind.’ (fieldwork CBB) Because dative experiencer verbs are rare in Swiss German,11 it was difﬁcult to ﬁnd such an example. In this position, past participles do not seem to occur as depictive secondary predicates. 4.4.2 Past participles In the dialect of Appenzell, the past participle takes the preﬁx gg-/k-. ‘Weak’ past participles have a sufﬁx -t/-d, while the ‘strong’ forms have stem vowel change (Ablaut) and end in -e (see Sonderegger and Gadmer 1999: 165). If a separate negation of the depictive secondary predicate is required, the past participle carries the preﬁx oo-/un- ‘un-’. Nu¨d ‘not’ is used to negate either the main predicate or the main predicate together with the depictive in the Mittelfeld. In this section, I will ﬁrst discuss the semantic and syntactic restrictions on the formation of participles serving as depictive secondary predicates. I then 11 In Swiss German, most experiencer expressions involve haa ‘have’ þ noun or adjective, e.g. Ich ha Angscht ‘I am scared’ or Ich chum Angscht u¨ber ‘I get scared’, corresponding to Standard German experiencer verbs with dative subjects such as Mir graut or Mir wird Angst.

Development of a depictive marker in Swiss German dialects

163

proceed to present the various semantic and syntactic linkage possibilities between the main clause predicate and the depictive secondary predicate. Finally, I present two examples which represent borderline cases of the depictive marker construction in Appenzell. 4.4.2.1 Semantic and syntactic restrictions on participle formation Only past participles which can also be used attributively can be used as depictives. The verbs from which these participles are derived have to express a change of state, the resulting state being permanent. Moreover, the object must be affected to a high degree. Hence telic verbs are the best candidates.12 German verbs with the preﬁx ver- occur frequently, since they express the fact that an action has reached its end and affected its subject or object to the highest possible degree (e.g. Standard German verblu¨hen ‘to wither’ (for ﬂowers), verfaulen ‘to rot’, verkochen ‘to overcook’). Consequently, past participles of intransitive atelic verbs such as schlafen ‘to sleep’ cannot be used either for attributes or for depictives. The past participle geschlafen ‘slept’ cannot refer to its subject attributively, since there is no change of state. Thus: Standard German (30) a. schlafen ‘to sleep’ b. *der geschlafene Junge *‘the slept boy’ Secondly, there is a restriction on the type of argument which the depictive secondary predicate can refer to attributively. In its relation with the depictive predicate, it must have the semantic role of the undergoer. As described by Haspelmath (1994), past participles used as attributes in English and Standard German predominantly refer to an undergoer. This observation also holds for Swiss German dialects. Therefore, intransitive (telic) verbs with an actor subject cannot refer to this actor subject as an attribute. For instance, the verb aufstehen ‘to stand up’, as past participle aufgestanden ‘stood up’, can refer to Junge ‘boy’ neither as an attribute nor as a depictive: Standard German (31) a. aufstehen ‘to stand up’ b. *der aufgestandene Junge *‘the stood-up boy’

12 In the terminology of Breu (1994), the aktionsart of these verbs is gradually terminative or totally terminative.

164

C. Bucheli Berger

However, intransitive telic verbs with an undergoer as subject can refer attributively to the subject. The verb verfaulen ‘to rot’, for instance, can be used as an attribute of Apfel ‘apple’: Standard German (32) a. verfaulen ‘to rot entirely’ b. der verfaulte Apfel ‘the rotten apple’ Thus, the dialectal verb vefuule ‘rot’ can occur as a depictive secondary predicate and receives the depictive marker in the dialect of Appenzell: (33)

Ase vefuuled-e chaa me daa nu¨d bring-e! so pst.ptcp:rot-dpct can one that:n.sg not bring-inf ‘One can’t bring that so rotten.’ (fieldwork CBB)

Transitive verbs generally have an undergoer argument to which they can refer attributively. Thus, das gescha¨lte Ei ‘the shelled egg’ is grammatical, but die gescha¨lte Frau ‘the shelled woman’ in the sense of ‘who shelled the egg’ is not possible: (34)

scha¨len das gescha¨lte Ei *die gescha¨lte Frau ‘shell’ ‘the shelled egg’ *‘the shelled woman’ (who shelled sth.)

In the same way, a transitive verb used as a depictive secondary predicate must refer to its undergoer. Consider the past participle ggscho¨llt of the dialectal verb scho¨lle ‘to shell’ in the following example: (35)

Ond denn hend¼s¼ne d¼Eie gg-scho¨llt-e and then have¼they¼them the¼egg:pl pst.ptcp-shell-dpct procht! pst.ptcp:bring ‘And then they brought them the eggs shelled!’ (fieldwork CBB)

It is of no importance whether the shared argument is a subject or an object of the main clause; what is crucial is that it is the undergoer of the depictive predicate: (36)

Gg-scho¨llt-e so¨nd die Eie a¼de Sonn gg-lege! pst.ptcp-shell-dpct are the egg:pl on¼the sun pst.ptcp-lie ‘Shelled, the eggs were lying in the sun!’ (fieldwork CBB)

Participles derived from reﬂexive verbs are particularly well suited to depictive use, since they always have an undergoer which is both subject and object at the same time. Thus, the past participle of sich

Development of a depictive marker in Swiss German dialects

165

wa¨sche ‘to wash oneself’ can be used as a depictive, here with the negative preﬁx oo- ‘un-’: (37) Oo-gg-wa¨sch-ne isch¼[s]i zo¼nis choo. un-pst.ptcp-wash-dpct is¼she to¼us pst.ptcp:come ‘Unwashed, she came over to us.’ (fieldwork CBB, based on an example in Sonderegger and Gadmer 1999) As a short note, it has to be added that the Appenzell dialect has no productive present participles with the old sufﬁx -ende/-end, either in attributive and depictive use or as a simultaneous converb (gerund). This is in contrast to Standard German, which could use the present participle in -ende (der tanzende Junge ‘the dancing boy’) to describe an ongoing process at the moment of speech. In the dialect of Appenzell, there would be no other way to express ‘the dancing boy’ than through a relative clause. Behaghel (1928: 360, 435) notes that the old sufﬁx -ende/-end has become -ing/-ig in Alemannic. Today, in the dialect of Appenzell, we observe that a form in -ege marks the function of the gerund of simultaneity (38) with verbs of motion (‘limping’, ‘jumping’, etc.) and body activities (‘crying, i.e. producing tears’). (38) Hink-ege ischt e a ere vebei gg-lauffe. limp-gerund is he to her along pst.ptcp-pass.by ‘Limping, he passed her.’ (fieldwork CBB) This form -ege can be decomposed into -eg and -e. From a synchronic perspective, the latter could be the depictive marker. The former could represent, on the one hand, an allomorph of -ing/-ig, as proposed by Behaghel. On the other hand, there is also the sufﬁx -ig (<-ic). This sufﬁx usually serves to derive adjectives from nouns (Sumpf ‘marsh’, sumpﬁg ‘marshy’) or from verbs (haue ‘cut, hit’, hauig ‘sharp’) which would ﬁt perfectly with the needs of an attributive or depictive function. Further research is needed to clarify this issue. Note in this regard that, in other German dialects outside Switzerland, the sufﬁx -ig has also been reinterpreted. In the Upper German dialect of Lechrain, the sufﬁx -ig is used without any additional sufﬁxing for the marking of depictive past [sic!] participles (see Walch, 2005). 4.4.2.2 Semantic and syntactic linkage possibilities Table 4.2 lists all possible combinations of main verbs and deverbal depictive secondary predicates (past participles). The controller, i.e. the participant shared by the two predicates, can be either an actor or an undergoer of the main verb. However, as I stated above, the controller always has to have the role of the undergoer with respect to the depictive secondary predicate. Therefore, on the vertical axis of

Table 4.2 Possible combinations of main verbs and deverbal depictives Main-clause verb

Controller ¼ subject

Depictive secondary predicate

Controller ¼ actor; transitive verb

Controller ¼ actor; intransitive verb

Controller ¼ undergoer; intransitive verb

Controller ¼ undergoer; transitive verb

Derived from transitive verb; controller¼ undergoer¼ object

Type 1

Type 2

Type 5

Type 6

Si hets ooggha¨a¨ssne tue.

Ooggschtra¨a¨lede so¨nd d Goofe as Hochzig choo.

Vehaune tro¨chned s Brood gschwinde.

E het die Chue ooaagglueegede kaufft.

‘She has done this without being asked to.’

‘These children came to the wedding uncombed.’

‘Cut into pieces, bread dries more quickly.’

‘He bought this cow without having looked at it.’

*Type 3

*Type 4

Type 7

Type 8

?

?

Ase vefuulede so¨nds denn all no dei gglege!

Fischsta¨a¨bli mo me doch gfroorne abro¨o¨tle.

‘They were still lying there so rotten!’

‘But you have to fry ﬁsh ﬁngers frozen.’

From (telic) intransitive verb; controller¼ undergoer¼ subject

Controller ¼ object

Development of a depictive marker in Swiss German dialects

167

Table 4.2, we only have to distinguish whether the depictive predicate is transitive or intransitive. On the horizontal axis, however, we can distinguish between transitive and intransitive main clause verbs, and also between controllers which are actors and those which are undergoers with regard to the main verb. Table 4.2 shows the interaction of these syntactic and semantic criteria. The empirically attested possibilities show a strong tendency towards keeping the semantic role of the controller unchanged. Types 5, 6, 7, and 8 fulﬁl this condition: the undergoer is shared by the two verbs. Types 1, 2, 3, and 4, where the semantic roles differ between main and depictive predicate, on the other hand, seem to be dispreferred: types 3 and 4, where depictives based on intransitive verbs are predicated on main clause actors, could not be found, and type 1, where depictives derived from transitive verbs are predicated on main-clause transitive actors, seems to be very rare. As the secondary predication describes a state, an undergoer seems to be the preferred controller of this state, rather than an actor. However, type 2, where depictives derived from transitive verbs are predicated on main clause intransitive actors, is well attested. In the following, I give an example for each type of linkage, if it is attested. Type 1 is very rare, being attested only by example (39) in the current corpus. Here, the depictive past participle of the transitive verb ha¨a¨sse ‘ask/ order someone to do something’ is connected to the transitive main verb tue ‘to do’, which has an actor subject. The controller si ‘she’ is an undergoer in relation to the depictive secondary predicate ha¨a¨sse and an actor in relation to the main verb tue. (39) Si het¼s oo-gg-ha¨a¨ss-ne tue. she has¼this un-pst.ptcp-order-dpct pst.ptcp:do ‘She has done this without being asked to/without any order.’ (Manser 2001: entry ooggha¨a¨ssne) Type 2 is very common. For example in (40), the transitive depictive verb schtra¨a¨le ‘to comb [one’s hair]’ is linked to the intransitive main clause verb choo ‘to come’, which has an actor subject: (40) Oo-gg-schtra¨a¨led-e so¨nd d¼Goofe as Hochzig un-pst.ptcp-comb-dpct are the¼children:pl to.the wedding choo! pst.ptcp:come ‘These children came to the wedding uncombed.’ (SADS 1) Goofe ‘children’ is the undergoer of the depictive secondary predicate schtra¨a¨le ‘to comb’ and the actor of the main verb choo ‘to come’. As verbs

168

C. Bucheli Berger

of motion are preferred predicates in the main clause (cf. Schultze-Berndt and Himmelmann 2004: 63), this type is unproblematic and can be constructed easily. Let me now proceed to the remaining linkage possibilities, 5, 6, 7, and 8, which are all well attested. As a common trait, the controller here is an undergoer in relation to both the main and the depictive predicate. In the case of 6 and 8, when the controller is the direct object of the main clause, the verb of the main clause must be transitive. In the case of types 5 and 7, the controller is the subject of the main verb. Type 5, e.g. in (41), combines the transitive depictive verb vehaue ‘to cut something into pieces’ with the intransitive main clause verb tro¨chne ‘to dry’, which has an undergoer subject. Brood ‘bread’ represents the undergoer of both verhaue and tro¨chne; see also example (36) (‘lying in the sun shelled’). (41) Vehau-ne tro¨chned s¼Brood gschwinde. pst.ptcp:cut.to.pieces-dpct dries the¼bread quickly ‘Cut into pieces, the bread dries more quickly.’ (fieldwork CBB) In type 6, the depictive past participle of the transitive verb aalueege ‘to look at’ is connected to the transitive main verb chauffe ‘to buy something’ (42): (42) E het die Chue oo-aa-gg-lueeged-e kaufft. he has this cow(f.sg) un-at-pst.ptcp-look-dpct pst.ptcp:buy ‘He bought this cow without having looked at it.’ (Manser 2001: entry ooaagglueegede) Chue ‘cow’ is the undergoer of both predicates. Another instance of type 6 (‘to bring something shelled’) was presented in (35) above. Type 7, illustrated by (43), combines a depictive past participle of an intransitive verb—here vefuule ‘to rot’—with an intransitive main verb which has an undergoer as subject—here deiligge ‘to lie there, be there’. The controller, the clitic pronoun ¼s ‘they’, remains the undergoer of both predicates. (43)

Ase vefuuled-e so¨nd¼s denn all no dei gg-lege! so pst.ptcp:rot-dpct are¼they then ever still there pst.ptcp-lie ‘They were still lying there so rotten/ in such a rotten state!’ (fieldwork CBB)

Type 8 links the depictive past participle of an intransitive verb—in (44) gfru¨u¨re ‘to freeze’13 to a transitive main verb, here abro¨o¨tle ‘to fry [sth]’. In 13 The g- in this case is part of the verb stem and not the past participle marker, cf. De Bronne gfru¨u¨rt. ‘The fountain is freezing’.

Development of a depictive marker in Swiss German dialects

169

this example too, the controller Fischsta¨a¨bli ‘ﬁsh ﬁngers’ represents an undergoer with regard to both main and secondary predicate. (44) Fisch-sta¨a¨b-li mo me doch gfroor-ne abro¨o¨tl-e. fish-fingers-dim.pl must one but pst.ptcp:freeze-dpct fry-inf ‘But you have to fry fish fingers [while they are still] frozen.’ (SADS 1) In (33) above, another example of type 8 was presented (‘to bring something so rotten’). To sum up, the presentation of all attested types of linkage has revealed that in the preferred depictive expressions, the shared participant of both predicates has the semantic role of undergoer. Those combinations which require a switch from undergoer to actor tend to be rare or impossible, except for type 2. 4.4.2.3 Borderline cases I brieﬂy discuss two examples which are at the limits of depictive constructions. The ﬁrst example is still from the Appenzell dialect. Recall that we stated above that the controller has to be the undergoer of the depictive secondary predicate. However, the following sentence seems to represent a case where the controller is the actor of the depictive predicate. (45) Me so¨nd oo-gg-ess-ne as Fescht ggange. we are un-pst.ptcp-eat-dpct to.the party pst.ptcp:go 1. ‘Having not been eaten, we went to the party.’ 2. ‘Not having eaten/hungry, we went to the party.’ (fieldwork CBB) The transitive depictive secondary predicate esse ‘to eat’ is combined with the intransitive main clause verb gaa ‘to go’. Following the syntactic-semantic rules described in the previous sections, one automatically constructs an attributive relation between the past participle ooggessne ‘uneaten’ and an undergoer, which in this case typically is an edible entity. But there is no such participant—only the actor subject me ‘we’ is there. As a result, this subject is reinterpreted as the undergoer of the depictive secondary predicate ‘uneaten’, causing the ﬁrst reading. This reading resembles a type 2 linkage but semantically it is quite surprising, because there is no danger of cannibalism or wild animal attacks and hence it is again pragmatically reinterpreted. A second reading appears that is theoretically impossible according to the syntactic-semantic rules: the controller is interpreted as the actor of the depictive secondary predicate (‘we in the state of having not eaten’). This example, then, represents an exceptional instance of a controller functioning as actor vis-a`-vis main and secondary predicate. Since native speakers

170

C. Bucheli Berger

expect to have to explain at least the second reading, it is evident that we are dealing with an example beyond the limits of depictive linkage. Note that the past participle ooggesse ‘uneaten’ can function as a depictive without problems when used in other linkage types, e.g. type 6, as shown in (46). (46) Di(e) Chriesi hem-me(r) oo-gg-ess-ne mo¨se ligge loo. the cherries have-we un-pst.ptcp-eat-dpct must lie let ‘We had to leave these cherries there uneaten.’ (Sonderegger and Gadmer 1999: 147) This conﬁrms that it is not the form of ooggessne but the linkage type which is the cause of the problems in (45). The second example is even more problematic since semantically, the expression carrying the depictive marker does not predicate on one of the participants of the main predicate, but rather refers to the atmospheric circumstances (and, metonymically, possibly the time of day). It is constructed after an utterance originally overheard in the area of Westallga¨u, in Stiefenhofen (Manfred Renn, p.c.; cf. Bucheli Berger and Glaser, 2004). I suggested a similar example to seventeen native speakers in the Swiss depictive marker zone. As expected, most of them refused it, but four persons accepted it when a depictive marker was used. In the dialect of Diepoldsau, the native speaker again added the restrictive marker asa: (47) Mear seand asa hell-a fort und we are restr bright-dpct away and asa dunkl-a heicho. restr dark-dpct home.come:pst.ptcp ‘We left in broad daylight and came home in the dark.’ (fieldwork CBB)

4.5 Summary After showing that agreement is still found in the south-western area of the Wallis (section 4.2), I have described the development of a depictive marker in the dialects of Appenzell (section 4.3). This has been presented as an instance of grammaticalization, more precisely, a restructuring of the agreement paradigm. Constructions with a depictive marker are restricted by the semantics of the depictive predicate as well as by the syntactic function of the controller (section 4.4.1). Furthermore, as seen in section 4.4.2, the depictive past participle has undergoer orientation exclusively (never actor orientation).

Development of a depictive marker in Swiss German dialects Mittelland/Jura (type A) Pred

Dep

171

Wallis (type B) Pred

Attr

Dep

Attr

Adv

Adv Appenzell (type C) Pred

Dep

Attr

Adv Figure 4.1 German

Semantic maps for morphological marking on adjectives in Swiss

There is a tendency to avoid a switch from a depictive undergoer to a mainclause actor, with the shared controller preferably remaining the undergoer of both verbs. Swiss German dialects today thus present three completely different strategies for the marking of depictives. The northern dialects of type A mainly follow the (northern) strategy of German dialects and Standard German, marking neither depictives nor adverbials nor predicatives. The south-western dialects of type B, especially in the Wallis, show agreement except for adverbs, still preserving an Old High German strategy and thus demonstrating a strategy also similar to that of the neighbouring Romance languages in the south. We can, therefore, establish a clear typological division between north and south. In addition to this, on the east–west axis, we ﬁnd the north-eastern dialects of type C, which constitute (part of) an innovation zone, where a typologically unexpected strategy of explicit depictive marking has emerged. The differences in marking strategies between the three types of adjectival agreement are summarized in the semantic maps in Figure 4.1, designed following van der Auwera and Malchukov (Ch. 13, this volume).

This page intentionally left blank

5 Quantifying depictive secondary predicates in Australian languages WILLIAM B. McGREGOR

5.1 Introduction Secondary predicates are well known in Australian languages, and have been widely reported in the literature, in both descriptive grammars and in theoretically oriented articles. Best described are depictive secondary predicates or depictives, in expressions such as (1)–(3), in which the depictive agrees in case-marking with the controller (see Himmelmann and Schultze-Berndt, Ch. 1, this volume, section 1.3).1 Martuthunira (1) nhulaa miyu mungka-rnuru wajupi-i wanka-a¼l near:you cat eat-prs grasshopper-acc alive-acc¼then ‘That cat eats grasshoppers alive.’ (Dench 1995: 182) This is a revised version of a paper presented to the workshop ‘Depictive secondary predicates in crosslinguistic perspective’, Ruhr-Universita¨t Bochum, June 2001. I am grateful to Eva Schultze-Berndt and Nikolaus Himmelmann for encouraging me to rework the original paper, and for helpful comments on previous drafts. Preliminary research was conducted during tenure of a fellowship in the Catholic University of Leuven, and prompted by a request by Fritz Schweiger to present a seminar on numerals in Australian languages in Salzburg. Financial support for my ﬁeldwork has come from the Australian Institute of Aboriginal and Torres Strait Islander Studies, the Australian Research Council, and the Max-Planck-Gesellschaft. My greatest debt is, however, to speakers of various Kimberley languages who have imparted information about their languages to me over the past two decades, including: yJack Bohemia, yCarmel Charles, Reni Chestnut, yBigfoot Jagarra, yDave Lamey, Susie Lamey, Maudie Lennard, yFreddy Marker, yMorndi (Bill) Munro, David Street, and Mervin Street, among many others. 1 In most examples I have retained the orthography of the source; this is generally a standard practical orthography. (In some languages, e.g. Gooniyandi, oo is used to indicate the high back vowel u.) For Gooniyandi, rrr represents the apical tap (rr) followed by the apico-postalveolar glide (r). Morpheme divisions usually follow the source; however, in some languages complex morphophonemic rules obscure morpheme boundaries, especially in inﬂecting verbs. In these cases, when the morphemic composition is not relevant to the point being made, the word is given an approximate gloss as a whole, suitable to the context. Commas in the source line of example sentences indicate prosodic breaks, either the end of intonation units, or pauses.

174

W. B. McGregor

Warlpiri (2) wati-ngki paka-rnu karnta mata-ngku¼lku man-erg hit-pst woman tired-erg¼then ‘The man, tired by then, hit the woman.’ (Laughren 1992: 204) Gooniyandi (3) goornboo galintha nangbani woman young:woman she:died ‘The woman died young.’ Most examples of depictives one ﬁnds in the Australianist literature illustrate temporary qualities of entities—for instance, states of animates (examples (1) and (2)), or age (example (3)). However, many languages exhibit quantity expressions in what appear to be depictive uses. An illustration is provided by Jiwarli example (4), in which kayanu ‘one’ is used in the sense ‘alone’: here the numeral does not merely specify the numerosity of the referent set (the speaker), but does so in relation to the referent event. Jiwarli (4) kumpa-ja muntu-parnti-purra warrkamu-rri-ngu kayanu sit-pst morning-abl-time work-inch-ipfvSS one:nom ‘I’ve been working alone since morning.’ (Austin 1997: 9) Such quantifying expressions are the topic of this chapter. I attempt to identify the range of expressions in Australian languages that admit quantifying secondary predicate interpretations. I also raise the question of their status: for instance, do expressions such as (4) really represent constructions, in the sense of Construction Grammar (Fillmore 1988; Goldberg 1995; Kay and Fillmore 1999; Zwicky 1987) and Semiotic Grammar (McGregor 1997c)? If so, do they (or some proper subset of them) represent a separate secondary predicate construction in their own right, or do they fall within a general class of depictives? Alternatively, is the quantifying secondary predicate sense merely an interpretation of another, non-secondary predicate construction type—does, for instance, the quantiﬁer function as an adverbial, or as a modiﬁer in an NP that also contains a referring nominal or pronominal (presumably an ellipsed ﬁrst person singular pronominal in (4))? The exposition is organized as follows. In section 5.2 I begin by outlining the range of meanings of numerals in Australian languages, summarizing the results of a previous investigation (McGregor 1998a). Following this, in section 5.3 I identify two types of quantifying expression that would seem to be reasonable candidates for secondary predicates. These are discussed

Depictive secondary predicates in Australian languages

175

and evaluated in terms of grammatical status: are they secondary predicate constructions or do they instantiate another type of construction (e.g. adverbial, NP-modiﬁer)? Section 5.4 brieﬂy turns to mass quantiﬁers, and discusses their grammatical status. Section 5.5 presents a brief conclusion, remarking on the grammatical relations involved.

5.2 Senses of numerals in Australian languages As is well known, numerals constitute quite small sets in most Aboriginal languages—typically there are terms for ‘one’, ‘two’, ‘three, a few’, and ‘many’. Despite the small size of the sets, and the widely presumed lack of interest in quantiﬁcation amongst Aboriginal peoples traditionally, numerals show a rather extensive range of senses and uses. Figure 5.1 presents in summary form the range of senses for the word ‘one’ attested in a sample of over one hundred Australian languages. Similar representations are possible for the senses of the terms for ‘two’, ‘three’, and ‘many’. For each term it is possible to group the senses into ﬁve general categories: (a) Quantifying: in which the quantiﬁcational sense is in the fore, and the numeral is used principally to indicate quantity, generally cardinality; (b) Reference-related: in which the numeral, in addition to providing information about quantity concomitantly deploys this for referential purposes; (c) Qualifying: in which the numeral is used to designate a quality displayed by an entity instead of the cardinality of a set; (d) Involvement: which concerns the relation of the referent entity vis-a`-vis other like entities that might have been involved in the situation; and (e) Temporally modulated: in which the numeral interacts signiﬁcantly with that which is quantiﬁed over a span of time, invoking some element of sequence, rank order, or frequency. These senses are most comprehensively represented for ‘one’, and for the other quantiﬁers decreasingly as cardinality increases, doubtless a reﬂection of the corresponding decrease in textual frequency, if Gooniyandi and English can be presumed typical. Nevertheless, each major category (a)–(e) is represented for each quantiﬁer, in at least one language. Doubtless many gaps in the speciﬁc senses result from inadequacies in descriptions (few grammars provide detailed discussion of quantifying words) and of textual corpora included in grammars, which usually provide few relevant examples.

176

W. B. McGregor quantifying

temporally modulated

distributed: 1 each

1 by 1, 1 at a time

borderline ordinal: 1st in birth sequence

augmented: 1 more

little by little

frequency: 1 time, action, situation

mass individuation small size comparison and intensification

involvement unique participation

cardinality of set is 1

comparative co-participation

approximately 1

exclusive involvement

reference-related indefinite definite specific, indefinite: the particular a particular, certain one

unity

qualifying separateness

identical: sameness

become alone

nonspecific indefinite

unithood: unification as whole

become scattered or separated

selective: one of a group

come together, gather, meet

one side of

indefinite standard of reference

join together, unite, mix

arbritary, any single item from given set or universe

put in one place

generic

Figure 5.1 Range of senses of ‘one’ in Australian languages, after McGregor (1998a). (italics¼verbal expression; underlining¼morphological modiﬁcation required.)

Cross-cutting these ﬁve semantic dimensions are three different syntactic environments in which numerals are attested: NP-internal use as determiners (deﬁnite (‘a certain’), indeﬁnite (‘someone’, ‘something’); speciﬁc (‘a particular’), nonspeciﬁc (‘any’), etc.), attributing modiﬁers (attributing quantity of the entity N), and

Depictive secondary predicates in Australian languages

177

perhaps even entity-speciﬁers (‘heads’). An example of NP-internal use of jirrawu- ‘one’ in Kija as a speciﬁc determiner is: Kija (5) timana-ny ta-ny jirrawu-ny horse-m that-m one-m ‘that one stallion’ [Kofod n.d.] NP-external use, in which the numeral appears to indicate something about the event, not just one of the entities involved in it. Potentially included here are putative depictives (e.g. ‘alone’, ‘together’), as in (4) above, and frequency adverbials (e.g. ‘once’, ‘twice’) as in (6): Warrwa (6) jawu jina wila kujarra-ngal swim he:said/did water two-fqy ‘He swam across the river twice.’ Use as event-speciﬁers, as primary predicates, in VPs (e.g. ‘unite’, ‘divide’), where the quantiﬁer ‘functions as’ a verb. In the clearest cases derivational morphology marks the new status of the numeral as a verb; however, in various languages from the far north-west of the continent explicit marking of the new status does not occur. An example of the former type, employing derivational morphology, is: Pitjantjatjara (7) kunmanara-nya kuwari kutju-ri-ngu Kunmanara-abs now one-inch-pst ‘Kunmanara became alone today.’ (Goddard 1996: 55) The investigation reported in McGregor (1998a) is primarily concerned with delimiting the full range of senses of numerals. It does not address questions of grammar, or whether the meanings identiﬁed are systematically distinct. The present chapter is an attempt to address these two questions in a subset of the data, and determine whether it is possible to distinguish the second syntactic grouping, the NP-external group, systematically from the NP-internal group, and, within the former, whether and to what extent it is possible to distinguish between depictive secondary predicates and adverbials. That is to say: are we justiﬁed in making a constructional distinction corresponding to the meaning differences? Or is the apparent NP-internal and NP-external distinction merely motivated by contextual

178

W. B. McGregor

considerations, and does it represent different senses of a single semantically vague construction type? For example, does kayanu ‘one’ in (4) belong within an argument NP which also includes ngatha ‘I’ (which happens to be ellipsed, presumably because it presents given information), the ‘alone’ sense emerging as a pragmatic implicature of ‘I one have been working since morning’?

5.3 Cardinal quantifying secondary predicates On semantic grounds the only candidates for secondary predicates come from the ﬁnal two categories identiﬁed in the previous section—(d) involvement and (e) temporally modulated—these being the only types that invoke something beyond the entity itself, where the numeral is related semantically not just to the entity but to the event. In both cases we ﬁnd expression forms reminiscent of the typical Australian depictive secondary predicate—a nominal that is case-marked in agreement with the NP it attributes on. Granted that the expression involving the numeral does not function in determining reference (and hence as an NP-level modiﬁer), these constitute plausible candidates for depictives on formal grounds also (as per Himmelmann and Schultze-Berndt, this volume, section 1.3). We now discuss categories (d) and (e) in order. As we will see, determining whether we really do have depictive constructions can be a difﬁcult matter, and involves considerations that take us deeper than mere expression form. 5.3.1 Involvement: co-participation Category (d) relates to how the entity participated in the event, whether alone or together with some other entity or entities as a member of a larger group. In various Australian languages each numeral can be used in this way, meaning, respectively, ‘alone, singly’, ‘together as a pair’, ‘together as a triplet, together as a few’, and ‘together as a large group, together as a big mob’. Some illustrative examples are, for ‘alone’, (4) above and (8); for ‘together as a pair’, (9) and (10); for ‘together as a triplet’, (11) and (12); and for ‘together as many’, (13–15). Notice that, as the Gooniyandi examples illustrate, the quantiﬁer gets ergative marking in transitive clauses in agreement with the controller, and the construction thus resembles the widespread depictive construction of Australian languages remarked on in section 5.1. Secondary predicate expressions showing these formal features of agreement will be referred to as Identically Marked Expressions (IMEs).

Depictive secondary predicates in Australian languages

179

5.3.1.1 One Gooniyandi (8) igi marriyali ngirnda nyamani girli gand nganyi-ngga no wife’s:mother this big same can’t 1.sg-erg yoowarni-ngga ngabbila, one-erg I:will:eat:it ‘No, mother-in-law, it’s too big. I can’t eat it alone.’ 5.3.1.2 Two Gooniyandi (9) thoodgoowirrani roob-ja garndiwirri¼nyali they:began:sliding:down rope-loc two¼again ‘They started sliding down the rope together.’ Wagiman (10) gay-giwu wilhma ga-ba-ya-guju larima that-du walk 3-pl-go.prs-du two ‘Those two are walking along as a pair (i.e. together).’ 1999: 68)

(Wilson

5.3.1.3 Three Gooniyandi (11) thoodjirranirri ngarloodoo yoowooloo, ngarloodoo we:paucal:descended three man three thoodjirranirri, banda-yirra, girili-nhingi, we:paucal:descended ground-all tree-abl ‘We got down, the three of us; we went down together to the ground, from the tree.’2 Marthuthunira (12) ngunhaa nganarna-lu, yilu ngathu manku-yangu jarrkurti-lu that-nom 1.pl.excl-eff this 1.sg.eff grab-pass.ptcp three-eff ‘That fellow was grabbed by us, by this fellow and me, by all three of us [together].’ (Dench 1995: 201) 2 As Eva Schultze-Berndt has pointed out to me, the NP ngarloodoo yoowooloo could also be analysed as a depictive, conveying a sense like ‘we descended as three men’. However, this does not appear to be the appropriate interpretation in this particular example: immediately preceding these words the narrator has said that a group of people were laying out stones below a burial platform, at which point the speaker descended from the platform to the ground. He then corrects himself, indicating that in fact it was three men (including he himself) who descended. The fact that in the previous clause the speaker says nganyi thoodngani (I I:descended), rather than e.g. yoowarni thoodngani (one I:descended) strongly suggests against the depictive interpretation in this particular instance.

180

W. B. McGregor

5.3.1.4 Many Gooniyandi (13) garndiiwangoorroo¼nyali mirdjirrrimi mijimab, gininyjirrrarri, many¼again we:tied:them:up mixed we:mixed:them maroowa-ya, murderer-loc ‘We tied them all up together, and put them together with the murderer.’ Gooniyandi (14) ngidi-ngga milayirrra, garndiwangoorroo-ngga, we-erg we:saw:them many-erg ‘Together as a group we watched them.’ Nyikina (15) mogodo mukudo many ‘Let us go

yanar-ma yanarr-ma we:shall-go together.’

(Nekes and Worms 1953: 726)

This use of numerals seems to be fairly widespread, though it is best attested for ‘one’, diminishingly as cardinality increases. In Jaminjung it seems that the numeral must host the restrictive clitic ¼(C)ung in order to serve as a secondary predicate (Schultze-Berndt 2002)—we return to this below. A somewhat different pattern is attested in a small number of languages, where instead of indicating the cardinality of the set, the notion of togetherness is highlighted, and the word for ‘one’ is used for ‘together as a group’, regardless of numerosity. Expressions displaying this pattern will be referred to as unifying expressions (UEs). In languages with UEs the ‘one’ word often takes a marker specifying the person and/or number of the group. For instance, in Wangkangurru nguyu-bula (one-dual) means ‘just the two’ (Luise Hercus, p.c.)—example (16). And in (17) wumba ‘one’ takes person and (for categories above the minimal) number afﬁxes (Merlan 1982: 92). Wangkangurru (16) ari nguyu-bula thangkarda we.du.excl one-du sit ‘We two are (sitting) here by ourselves.’

(Luise Hercus p.c.)

Mangarrayi (17) malam-gala wumba-rla-(yi) wurlayaj three-pl one-pl-(prop) they:went ‘Three men went by themselves.’ (Merlan 1982: 92)

Depictive secondary predicates in Australian languages

181

Gaagudju shows a slight variation on this theme: the preﬁx-taking bases -gadu and -gaduwa are both used in expressing the quantity ‘one’. However, only the latter may also be used to express the ‘together as a group’ sense, in which case it takes preﬁxes and/or sufﬁxes indicating the person and number of the uniﬁed group (Harvey 2002: 164–6).3 This is illustrated by (18). Gaagudju (18) ngameneega nj-djaa-ni-mana nginjaa-gaduwa-mana why 2.a-prs-sit-m.ua 2-alone-m.ua ‘Why are the pair of you sitting alone?’ (Harvey 2002: 165) UEs seem to be an areal feature of the Arnhem Land region, being found in Gunwinjguan languages also, including Mayali and Ndje´bbana. In Ndje´bbana person–number sufﬁxes are attached to wara´bba ‘one’ as in wara´bba-njabba (one-me) ‘me alone’. (Unfortunately McKay 2000 does not give examples for cardinalities larger than one.) Ndje´bbana (19) ngaya´mala yi-be´na ngarra´ma njana 3.min.f.emph 3.min.f.s-go:away female and nake´mala la´rlana ka-be´na, wara´bba. 3.min.m.emph male 3.min.m.s-go:away one ‘The female went away and the male went away, alone.’ 2000: 205)

(McKay

No UEs in the available corpora display case-agreement marking, though this may be due to the fact that examples are not plentiful, and most have absolutive controllers. A third pattern—which will be referred to as locative expressions (LEs)—is exhibited by DjambarrpuyNu, where the locative form of ‘one’ waNgany-Nur is used to mean ‘together’, as illustrated by the following example:4 DjambarrpuyNu (20) Nilinyu Nuli ga warkthu-n manda, waNgany-Nur 1 þ 2.du hab ipfv.1 work-1 du one-loc ‘Us two are working on the same (work), together.’ (Wilkinson 1991: 591)

3 There are a number of complexities concerning third person referents that we need not go into here (Harvey 2002: 166). 4 The digits 1–4 in glosses for DjambarrpuyNu verbs indicate inﬂections expressing tense, mood, and aspect meanings.

182

W. B. McGregor

These are not the only means of expressing the meaning ‘together as an n-tuple’, though they seem to be the main means associated with numerals and/or quantiﬁers. Various languages employ speciﬁc words with this meaning, that bear no close etymological relation to any quantifying lexeme. Sometimes these words take pronominal or number afﬁxes cross-referencing the whole group, sometimes they are invariant. Many Nyulnyulan languages, for instance, have something cognate with Nyulnyul ngidirrngin ‘alone’ (as in (21)) and yambun ‘together’. Nyulnyul (along with other closely related languages of the Western Nyulnyulan subgroup) has in addition the poorly attested preﬁxing nominal -malk(an) ‘alone’, illustrated in (22). Other languages with separate words for ‘alone’ and/or ‘together’ include Kayardild (examples (23) and (24)), Yir Yoront, Warrgamay, Nyamal, and Anindilyakwa. Expressions involving such lexical items will be referred to as specific lexical expressions (SLEs). Nyulnyul (21) buu inam dubak ngidirrngin smoke he:put:it tobacco alone ‘He smoked the tobacco alone.’ Nyulnyul (22) djinji ni-malg Nan-djalen bob-og djen jinji ni-malk ngan-jalin bub-uk jin honey:eater 3.sg-alone 1.sg-see:it ﬂower-loc 3.min.obl in-galen maNer in-kalin mangirr it-goes always ‘I see that a honey eater is always alone resting on a ﬂower.’ (Nekes and Worms 1953: 485) Kayardild (23) nga-ku-l-da bakii-n-da mardala-a-ja marin-d 1-incl-pl-nom all-nmlz-nom paint-detr-actualized self-nom kantharrkuru rarrthararrth alone:nom individually:nom ‘We all paint ourselves.’ (Evans 1995: 228) Kayardild kantharrkuru ngudi-nangku banga-wu (24) ngada 1.sg.nom alone:nom throw-neg/pot turtle-obj ‘I can’t throw the turtle over by myself.’ (Evans 1995: 372) Languages may employ more than the one strategy—and use speciﬁc ‘alone’ or ‘together’ words along with the usual numerals, as is the case in

Depictive secondary predicates in Australian languages

183

Gooniyandi and Jaminjung. Quite likely this is a widespread feature, although it is remarked on in few grammars. Another strategy frequently employed—though apparently never the sole strategy—is the use of pronominals, in pronominal expressions (PEs). In Gooniyandi, oblique pronominals can be used in this way, as illustrated by the following example: Gooniyandi (25) igi igi girrangi¼nyala, yinigawanma, ngabbinggirra no no your:pl¼again what:will:we:do you:pl:will:eat:it girrangi¼nyala marriyali:, your:pl¼again wife’s:mother ‘No, no, it’s yours. What will we do? You eat it yourself, mother-in-law.’ The PE in Jaminjung involves the appropriately case-marked form of the cardinal pronominal, followed by the restrictive enclitic ¼(C)ung (Eva Schultze-Berndt, p.c.). We have now identiﬁed ﬁve modes of expression—which are not necessarily all mutually exclusive (this does not affect the discussion below)—as potential candidates for quantifying depictive secondary predicate constructions: IMEs (Identically Marked Expressions): involving a numeral showing the same case-marking as the controller; UEs (Unifying Expressions): involving the lexical item ‘one’, usually with an afﬁx specifying person and number features of the controller; LEs (Locative Expressions): involving a locative-marked quantiﬁer; SLEs (Speciﬁc Lexical Expressions): involving speciﬁc lexical items meaning ‘alone’,‘together’, and so forth; and PEs (Pronominal Expressions): involving a pronominal referring to the controller. Do these represent genuine secondary predicate constructions, or are they merely expressions that admit secondary predicate interpretations? We begin with the seven criteria that must be met, according to Schultze-Berndt and Himmelmann (2004), for a clause-level secondary predicate expression to count as a construction: i. It contains two separate predicative elements, the main predicate and the depictive, where the state of affairs expressed by the depictive holds within the time-frame of the eventuality expressed by the main predicate. ii. The depictive is obligatorily controlled, i.e. there exists a formal relation to one participant of the main predicate, the controller, which is usually interpreted as a

184

iii.

iv. v. vi.

vii.

W. B. McGregor predicative relationship (i.e. the depictive predicates an eventuality of the controller). The controller is not expressed separately as an argument of the depictive[ . . . ]. The depictive makes a predication about the controller which is at least in part independent of the predication conveyed by the main predicate, i.e. the depictive does not form a complex or periphrastic predicate with the main predicate[ . . . ]. The depictive is not an argument of the main predicate, i.e. it is not obligatory[ . . . ]. The depictive does not form a low-level constituent with the controller, i.e. it does not function as a modiﬁer of the controller[ . . . ]. The depictive is nonﬁnite (to be understood as: not marked for tense or mood categories), or the dependency of the depictive on the main predicate is indicated in other formal ways[ . . . ]. The depictive is part of the same prosodic unit as the main predicate [ . . . ]. (Schultze-Berndt and Himmelmann 2004: 77–8)

Condition 1 seems to be unproblematically satisﬁed: there are two predicative elements in each of the ﬁve expression types we have discussed, and the depictive ‘as an n-tuple’ holds at least throughout the event denoted by the main predicate. Condition 2 also appears to be met, in most instances at least: the depictive quantity is obligatorily controlled, in the above examples, by a participant (often ellipsed) in the clause. The main question concerns UEs and PEs: it is arguable that the ‘semantic’ controller is indeed expressed—by the agreeing or cross-referencing afﬁx in the case of those UEs that take one, and by the person-number features of the pronominal itself in the case of PEs—as an argument of the depictive. Whether or not this is actually so depends on precisely how the concepts of ‘controller’, ‘argument’, and ‘separate expression’ are construed. For instance, in (18), one possibility would be to take nginjaa- . . . -mana (you- . . . -m.ua) as an argument of -gaduwa- ‘alone’, following the lead of pronominal argument analyses for verbs (e.g. Bresnan and Mchombo 1986; Cootes 1989); simultaneously it might be taken as the controller and separate from the depictive (it is clearly morphologically separate). On the other hand, the pronominal argument analysis might be rejected (see e.g. McGregor 1997c : 104–5; Simpson 1991 for rejection on clause-level), and it might be maintained that the pronominal element crossreferences or agrees with—but is not—the controller. These alternative analyses give different answers to the question as to whether the Gaagudju expression is a depictive. Moreover, the controller is not phrasally separate

Depictive secondary predicates in Australian languages

185

from the depictive, and condition 2 might be taken as not met for this reason. The two analytical questions just raised are empirical questions, answerable only through careful examination of the facts of the language (in relation to a particular grammatical theory), and cannot be adequately dealt with in a typologically oriented paper such as this. We can only conclude that UEs and PEs remain potential depictives, provided that the ﬁnal condition under 2 is interpreted as ‘phrasally separate’. Condition 3 seems to be consistently met by the ﬁve expression types identiﬁed in this paper: the depictive does not form a complex or periphrastic predicate with the main verb. Nor is there any evidence that the depictive is an argument of the verb, so condition 4 is also satisﬁed. Condition 6 causes no problems either, since none of the modes of expression we have dealt with shows marking of ﬁniteness on the depictive element. The remaining two conditions are somewhat problematic. In relation to 7, it is unfortunately the case that few descriptions provide prosodic information, making it impossible to decide whether the condition is met. Such information is not provided for many of the above examples (e.g. (21–24)). At least for IMEs and PEs some examples are available indicating the possibility of integration on the same intonation contour as the main predicate—see (8) and (25), respectively. However, the LE (20) does not satisfy the condition, and would have to be rejected as a depictive. I ﬁnd this rather unsatisfactory, especially given that the locative phrase could conceivably be an afterthought: there is reason to believe that afterthoughts serve the same grammatical relation as they would if incorporated in the normal way in the utterance. (For instance, they are generally case-marked in the normal way, by the same case-marker as would be used if they had been incorporated in the same intonation contour as the main predicate.) Ultimately, it does not seem justiﬁable (given our present state of knowledge) to reject any of the ﬁve modes of expression as potential depictive constructions on the basis of the prosodic condition. Condition 5 is also difﬁcult to evaluate without detailed knowledge of a language. For SLEs in Nyulnyul the speciﬁc lexeme in an example such as (21) cannot be a part of an NP denoting the controller, since if it were it would be marked by the ergative postposition, being the only element of the subject NP actually present. Similar examples with -malk ‘alone’ in transitive clauses argue against this word forming part of an NP with the controller. In the same way, the impossibility of ergative marking on the second pronominal in (25) argues against the pronominal forming an NP with the ellipsed controller in Gooniyandi PEs. For LEs and UEs insufﬁcient evidence is available to permit a decision one way or another on the issue of co-constituency with the controller.

186

W. B. McGregor

As for IMEs, three major analytical possibilities present themselves:5 The numerals could be modiﬁers in NPs with nominal or pronominal heads. In example (8), for instance, yoowarni-ngga (one-erg) ‘by one’ and nganyi-ngga (1.sg-erg) ‘by me’ could form a single NP ‘by me as one’. Similarly, in (9) garndiwirri¼nyali (two¼again) ‘two again’ could be an elliptical NP with head ngiidi ‘we’. These NPs could serve as participants (arguments). Second, it could be that the quantiﬁer expressions are separate units from the NPs representing the controllers. Thus, in (8) nganyi-ngga (1.sg-erg) ‘by me’ and yoowarni-ngga could be separate, not being co-constituents of a single NP; the pronominal NP could then serve as the controller. In a similar way, in (9) garndiwirri¼nyali (two¼again) ‘two again’ and ellipsed ngiidi ‘we’ could be separate units, the latter serving as controller. In both cases the numeral could be an external unit linked to the controlling NP by some grammatical relation (see section 5.5 for further discussion). Third, the numeral and controller (whether present or ellipsed) might together constitute a split NP—either discontinuous (McGregor 1997a) or fractured (McGregor 1989), as perhaps in (8). I favour the second analysis, though acknowledge that the evidence is not completely compelling. One piece of evidence in favour of the separate phrasal status of the quantiﬁer expression and its controller comes from the fact that the latter can be ﬁlled out with other material, including another instance of the quantiﬁer. As a possible illustration, consider (26), from a personal narrative describing a lengthy hunting and ﬁshing trip. Gooniyandi (26) yilba warrgoomjirri, garndiwirri¼nyali diribjirri for:good we:worked two¼again we:entered ngidi garndiwirri, we two ‘For good we worked; together we went back to work, the two of us.’ (McGregor 1990: 584) The ﬁrst instance of the numeral cannot form an NP with the ﬁnal two words; the two items must represent distinct grammatical units. Granted the role uniqueness hypothesis (e.g. Harada 1975; McGregor 1997c : 118–19), 5 Analysis of the numeral as an adverbial cannot be maintained, since in Gooniyandi adverbials do not take case-marking postpositions in agreement with their semantic controllers.

Depictive secondary predicates in Australian languages

187

only one can serve in a participant role. Since the second unit clearly serves such a role, the ﬁrst cannot. This also suggests a semantically plausible interpretation of the utterance, in which the ﬁrst numeral speciﬁes the togetherness of the two individuals as they returned to work.6 Clear-cut examples such as this are rare in my Gooniyandi corpus, due to the widespread propensity for ellipsis of given information. Nevertheless, the fact that a few can be found attests to the existence of at least some IMEs that satisfy condition 5. The depictive in (26) hosts the enclitic ¼nyali again. In some languages it seems that similar enclitics attached to numerals specify the ‘joint participation’ interpretation. This is the case for ¼(C)ung in Jaminjung, as illustrated by (27) and (28). According to Eva Schultze-Berndt (p.c.) these admit only the ‘joint participation’ interpretation, and preclude a plain attributive NP-internal interpretation. Condition 5 is thus presumably satisﬁed. Furthermore, the enclitic could be interpreted—in the environment of the IME—as a marker of secondary predicate status. (Incidentally this is not the case for Gooniyandi ¼nyali ‘again’: its presence on a numeral does not signal the ‘joint participation’ sense; NP internal interpretations such as ‘the same n’ (for some numerical value of n) are also admitted.) Jaminjung (27) mindi-ngayi-m jirrama-ni¼wung mung 1.du.incl-see-prs two-erg¼restr look:at ‘You and me will watch it, both of us.’ (Schultze-Berndt 2002: 236) (28) luba¼wung gan-kuga-yirrag, Line Creek-nyunga, many¼restr 3.sg:1.sg-take-pst Line Creek-orig ‘He took us away (being) many, from Line Creek.’ (Eva SchultzeBerndt, p.c.) Various other languages (e.g. Martuthunira and Warlpiri—see examples (1) and (2) above) employ temporal enclitics at least optionally in depictives. Whether it is only instances involving these enclitics that should be treated as depictives is beyond the scope of the present investigation to determine.

6 The free translation differs from that provided in McGregor (1990: 584), ‘For good we worked, the two of us went back to work’, which misses part of the meaning expressed. In favour of the reinterpretation, observe that the previous clause stated ‘I went to the [police] station to work’. The present utterance then adds that it was actually two people who went back to work, as a pair. This situation is contrasted, in the following sentence, with the activity of the third member of the group, who went back alone to his place of work.

188

W. B. McGregor

In summary, it would seem that each of the ﬁve expression types identiﬁed above as potential candidates for co-participation depictives satisfy conditions 1–6 of Schultze-Berndt and Himmelmann (2004); condition 7 is satisﬁed by some IMEs and PEs. No instances of the other three expression types satisfy the latter condition, though this perhaps represents lacunae in the data. On this basis, the ﬁve expression types—and particularly IMEs and PEs—may be tentatively taken to be genuine instances of depictive constructions. One difﬁculty in evaluating our ﬁve types against criteria 1–7 has, however, been glossed over: what counts as a predicate or predicative element? All of the quantifying depictives discussed in this section count in context as predicative expressions. But if the expressions they occur in are depictive constructions specifying predicative relations between the controller and the depictive, I submit, depictives must also represent predicative elements grammatically: the predicative sense must be part of the constructional meaning, and not arise merely as a contextual interpretation. I have suggested elsewhere (McGregor 1997c : 171–4; see also section 5.5 below) that this happens when the dependency relation of attribution is involved, and the depictive attributes a property of the controller, on which it is dependent. Attribution is precisely the relation found NP-internally between the entityspecifying nominal and a quality-specifying nominal, for instance, between timana-ny ‘horse-m’ and jirrawu-ny ‘one-m’ in example (5) above, and externally between NPs in attributive clauses, which, like (29), attribute a property of an entity (McGregor 1997c : 143). Warrwa (29) kinya baalu yirdany this tree long ‘This tree is tall.’ Evidence that attribution is involved in the various secondary predicate expression types could come from paradigmatic alternations of the following two types. One is the existence of agnates in which the quantifying expression and the entity-specifying nominal form an NP together. This has already been noted for IMEs—agnate with (9) is thoodgoowirrani roob-ja bidi-yoorroo garndiwirri (they:began:sliding:down rope-loc they-du two) ‘The two of them began sliding down the rope’. This argues in favour of IMEs as secondary predicate constructions. The other piece of evidence is that there are complex sentence agnates in which one clause is an attributive relational clause linking the two expressions, controller and depictive (see also Nichols 1978a). It is not difﬁcult to imagine

Depictive secondary predicates in Australian languages

189

that UEs, LEs, and SLEs could have agnates of this type (NP-internal agnates seem less likely). For instance, for (16) one can imagine the existence of a biclausal agnate ?ari thangkardi, ari nguyu-bula (we.du. excl sit, we.du.excl one-du) ‘We two are sitting here, we are alone’. Similar agnates can be envisaged for examples such as the LE (20) (‘We two are working, we are together’), and the SLE (21) (‘He smoked the tobacco, he was alone’). Unfortunately, I have found no examples of such complex sentences. Nevertheless, some Nyulnyulan examples cited in Nekes and Worms (1953) are suggestive: for instance, in (30) we ﬁnd wud ‘alone’ in what appears to be an attributing clause with the copula verb -in ‘be’.7 This suggests that wud ‘alone’ can be used attributively and thus that (some) SLEs might be amenable to analysis as depictives. However, at least some SLEs (perhaps ngidirrngin ‘alone’ in (21)) would seem to be participant-oriented adverbials, rather than depictives. Jabirrjabirr (30) yai yaNge-djed bindan, ginjiNg wamb wo@d gaNk-en yay yangki-djid bindan kinyingk wamb wud kangk-in we:two we:will-go bush that man alone he:must-be ‘Let us go into the bush, but that man must stay [lit.: be] by himself.’ (Nekes and Worms 1953: 904) In the absence of such agnates, there is less evidence that attribution is involved as a grammatical relation—and more reason to believe that the predicative sense arises as a pragmatic interpretation. This seems to be the case for PEs of the type found in Gooniyandi, in which the pronominal appears in oblique form. Corresponding to (25) it is difﬁcult to imagine either of the following would be acceptable in the analyses indicated (where K in (32) represents ‘clause’)—and I have encountered no examples resembling either: Gooniyandi (31)

ngabbinggirra girrangi]np, marriyali np[gidi-ngga you:pl:will:eat:it np[you:pl-erg your:pl]np wife’s:mother ‘np[You yourself]np eat it, mother-in-law.’

7 This is not the verb for ‘sit’, but an inﬂecting verb that is used somewhat like an auxiliary along with preverbs in compound verb constructions, and as a general copula verb ‘be’ in attributing clauses. The latter may be verbless, as in many Australian languages; the verbal construction is typically used for temporary qualities. Note in particular that wud ‘alone’ is not a secondary predicate in this example—it is obligatory, and so fails by condition 4. Indeed, it may be considered to express the main predicate, since the verb serves as a mere copula. Granted this, an alternative analysis of this clause as an adverbial construction—e.g. in which wud ‘alone’ serves as a manner modiﬁer of the verb—is implausible. (What is crucial is not the part-of-speech membership of the special lexeme, but the grammatical construction it enters into.)

190

W. B. McGregor

Gooniyandi (32)

ngabbinggirra marriyali, girrangi]k k[gidi you:pl:will:eat:it wife’s:mother k[you:pl your:pl]k ‘Eat it mother-in-law, you are yourself.’

There is other evidence suggesting that PEs in Gooniyandi merely admit the quantifying depictive interpretation rather than actually encode it. Thus plural oblique pronominals in PEs admit not only the co-participation interpretation but also a non-co-participation interpretation: that each individual acted separately to bring the event about. Thus, in (33) there is no implication that the three individuals acted jointly as a group, rather than severally alone, as distinct individuals. Other Gooniyandi examples lend further support to the proposal that the plural oblique pronominal is vague with respect to the interpretations ‘act together as an n-tuple’ and ‘act alone or separately, being a member of an n-tuple’.8 Only the former interpretation counts as a quantifying sense: in the latter interpretation the quantity does not hold of the participant as it is engaged in the event. Gooniyandi (33) ngarloodoo-ngga birrangi¼nyali rirrwirrarnirri three-erg their¼again they:pulled:themselves thaanoonggoo up ‘The three of them pulled themselves up by themselves.’ For these two reasons I am inclined to the view that the oblique pronominal in these examples does not enter into a dependency relation with the NP denoting the participant, the controller. It perhaps serves in an adverbial relation, specifying how the event was performed, rather than attributes a quality of the performers as they enacted it. We conclude that the best candidates for secondary predicate constructions are Identically Marked Expressions. Unifying Expressions, Locative Expressions, Speciﬁc Lexical Expressions, and Pronominal Expressions generally make poorer candidates. Some SLEs probably involve participantoriented adverbials (Himmelmann and Schultze-Berndt, Ch. 1, this volume); PEs of the type found in Gooniyandi (with oblique pronominals) are poor candidates for secondary predicate constructions, although the Jaminjung 8 In (25), of course, the plural pronominal has singular reference, being used as a marker of respectfulness and distance between individuals in an avoidance relation. Thus the two interpretations do not arise, although they could if instead the quoted utterance had been addressed instead to a pair of individuals. (Actually, it is likely that more than two interpretations are available: it is unlikely that any possible partitioning of the referent set is excluded.)

Depictive secondary predicates in Australian languages

191

type (involving cardinal pronouns and the usual secondary predicate enclitic) would seem to be better candidates. 5.3.2 Iterated involvement The second category of quantifying senses potentially representing a secondary predicate was (e) temporally modulated. Three subtypes are attested, all usually involving morphological modiﬁcation of the quantiﬁer lexeme: frequency; restricted ordinal (usually limited to birth sequence); and iterated involvement (distributed). Only the iterated involvement subtype— which speciﬁes iterated co-participation of equally numerous proper subsets of a larger set throughout the event—is a good candidate for a depictive secondary predicate. As for frequency markers, these quantify events, and are not participant-oriented. They typically show up as adverbials—for example, employing special adverbial derivational morphemes, such as -ngal ‘times’ in Nyulnyulan languages. As for ordinals, these are rarely expressed as quantiﬁers in Australian languages; sequence is usually expressed by spatial metaphor—usually ‘ahead’, ‘behind’, and ‘middle’ for ‘ﬁrst’, ‘last’, and ‘intermediate’, respectively. When quantiﬁers are used, they are either verbal expressions (e.g. Anindilyakwa uses inchoative verbs to indicate birth sequence), or clearly belong within NPs as dependents on the entity nominal. What I have referred to as iterated co-participation resembles the co-participation expressions discussed in section 5.3.1, except that the co-participation does not apply over the entire time-frame of the event to all entities involved, but rather to equally numerous subsets in sequence, one after the other. Thus, the meanings expressed are ‘one by one’, ‘one at a time’, ‘one each’, ‘singly’, ‘two by two’, ‘two at a time’, ‘two each’, ‘pair-wise’, and so forth. As in many of these English expressions, the quantiﬁer is normally reduplicated, either totally or partially, sometimes with a linking element. Such reduplicated expressions are found in a fair number of languages— though in less than half of the languages in the sample—especially for ‘one’. Examples include Nunggubuyu anyjaa-nyjaa-nyjaabugij ‘one by one’, ‘one at a time’, ‘singly’ (Heath 1984: 495); Pitjantjatjara and Kukatja kutju kutju ‘oneby-one’ (Goddard 1996: 56; Valiquette 1993: 62); and Yir-Yoront yi-yirr-w yi-yirr-w ‘to each one separately’ is presumably an instance, despite the different gloss. Nekes and Worms (1953) report on this use of reduplicated words for ‘one’ in Nyulnyulan languages, and I have an example in my own Warrwa corpus. Partial reduplications are attested in Wik Mungkan thonthonam—which could be analysed as either ‘other-one’ or as a partial reduplication of thonam ‘one’—for ‘one after another’, ‘one-by-one’ (Kilham et al. 1986: 220); Ngalakgan wah-wangginy ‘one at a time’ from wangginy ‘one’

192

W. B. McGregor

(Brett Baker, p.c.); and probably DjambarrpuyNu waNga’waNgany ‘each one’ (Wilkinson 1991). Illustrative examples are:9 Ngalakgan (34) wah-wangginy burru-wulup nugu-jerrk one-one 3.pl.nom-bathe i.def-bird ‘The birds are bathing one at a time.’ (Brett Baker p.c.) Jiwarli (35) kayanu-rru kayanu-rru ngurntaira one-now one-now will:lie ‘They lie down one by one.’ (Austin 1997: 190) In addition to the cardinal interpretations, ‘one-one’ often admits a mass interpretation ‘little by little’, as in example (36): Gooniyandi (36) yoowarni yoowarni, dardigirrmiwirrra, one one they:understood:them ‘They understood [English words] little by little.’ or ‘They understood [learnt] English words one by one.’ In Nyulnyul the reduplication of war ‘other’, warawar, admits the sense ‘one-by-one’, as well as a variety of other senses including ‘severally’, ‘separately’, ‘one another’, and so forth. Although formed on the ‘other’ word, rather than the ‘one’ word, there is a close relation between the two, the word for ‘one’, warang, being built on the ‘other’ word. (McGregor 1997b discusses this point further.) Corresponding forms for ‘two by two’, ‘three by three’/‘few by few’, and ‘many by many’ are attested in just a few languages. The following Mparntwe Arrernte examples illustrate each of these: Arrernte (37) nwerne tnye-rlepe-ke therre-ke-therre-le 1.pl.a dig-do.along-pst.cont two-dat-two-adv crowbar nyente-kerte-le crowbar one-prop-ins ‘We dug two at a time with one crowbar each.’10 (Wilkins 1989: 346) 9

In the Ngalakgan example below I indicates a noun class. The exact interpretation of this example is uncertain. The scene described involved four persons and two crowbars. But whether each pair had one crowbar between them, or had two while digging (then giving to the other pair when their turn came) is not clear. In any event, it does not affect the point being made here, that the digging was done in a pair-wise fashion. 10

Depictive secondary predicates in Australian languages

193

Arrernte (38) kele ikwerenge arratye-rle tyape urrpetye-ke-urrpetye nwerne alright then true-foc grub three-dat-three we:all tnye-ke-le-rle, . . . dig-pst.cont-ss-foc/rel ‘After a while, sure enough, we were digging out grubs a few at a time.’ (Wilkins 1989: 499) Arrernte (39) itne nhenge atningke-m-atningke anteme irreke they anaph many-uq-many now became ‘Then the [numbers of people] grew, crowd upon crowd.’ Wilkins, p.c.)

(David

Iterated involvement is not, however, always expressed by a reduplicated numeral. In Kukatja we ﬁnd the irregular kutjunpa ‘one by one’, which is clearly constructed from kutju ‘one’ by addition of -npa. And in Jaminjung it seems that the continuous enclitic is employed:11 Jaminjung (40) ga-da-m jungulug-mayan 3.sg.nom-fall-prs one-cont ‘They [fruit] fall off one by one.’ (Schultze-Berndt 2000: 80) On purely semantic grounds these ‘number by number’ expressions could count as secondary predicates. Unfortunately, few sources discuss them in detail, making it difﬁcult to evaluate their grammatical status. Five of the seven criteria for secondary predicate constructions (SchultzeBerndt and Himmelmann 2004) would seem to be met: 1: there are two (semantically) predicative elements, with the depictive holding in the time frame of the main event; 3: the depictive does not form a complex or periphrastic predicate with the main verb; 4: the depictive is not an argument; 5: the depictive does not form an NP with its controller (there is no evidence that the quantifying expression can form an NP with its controller); and 6, the depictive is not ﬁnite. As for 7, no examples show the putative depictive in the same intonation contour as the primary predicate; this may well be a reﬂection of the inadequacies in the data. The few examples for which

11 Eva Schultze-Berndt informs me (p.c.) that reduplicated quantiﬁers in this function are not attested in her Jaminjung corpus.

194

W. B. McGregor

prosodic information is available involve the depictive on a different intonation contour. Condition 2, that the depictive is obligatorily controlled, is problematic: although it is apparently satisﬁed by examples (34–40) above, it is not in (41), where ‘one by one’ is not controlled by either the ‘subject’ or the ‘object’. Rather, it applies to an unexpressed entity, that does not even serve as a participant. This example raises the analytical question: is the ‘number by number’ expression a ‘ﬂoating’ term that can be predicated of any relevant entity in the referent situation, or is this so only in Panyjima? Given the paucity of examples, we cannot rule out the former possibility in Australian languages in general. Panyjima (41) nhupalu mankarrpa yata karnku-nma yikamarta yikamarta you:two ﬁrm shield hold-imp one one ‘You two hold that shield ﬁrmly, for one [spear] after another.’ (Dench 1991: 225) Ignoring this example and the problems it raises, we still need to decide whether a case can be made for recognizing an attributive dependency relation between the reduplicated quantiﬁer and the controlling NP. To begin with, it seems unlikely that agnates exist for examples such as those cited above that involve the reduplicated quantiﬁer in an NP or in a relational clause. It seems improbable, that is, that wah-wangginy (one-one) ‘one-by-one’ in (34) could form an NP with nugu-jerrk (i(def)-bird) ‘the bird(s)’, or could be attributed of it in an attributive clause. However, a case could be made if it could be argued that (34–40) are instances of IMEs, inasmuch as examples with reduplicated quantiﬁers show identical marking with the controller: generally none or zero. Analysis of the reduplicated quantiﬁers as adverbials is also possible. But there is evidence against such an analysis at least in Mparntwe Arrernte. In examples (37) and (42) the reduplicated quantiﬁer is marked by the sufﬁx -le, a sufﬁx with a wide range of functions, including ergative, instrumental, and locative case-marking, as well as an adverbial derivational use. Although Wilkins (1989) consistently glosses it adv in these examples (repeated in my citations), it is not clear that this is the best analysis. Comparison with (38) and (39) reveals that the -le marker is not necessarily present, as would be the case if this were really the derivational sufﬁx -le adv (Wilkins 1989: 176–7); rather, its presence apparently depends on the grammatical role of the controller.

Depictive secondary predicates in Australian languages

195

Arrernte (42) ngke-ke-rle are-me-le, nwerne crack.out-pst.cont-rel see-npst.prog-ss 1.pl.a tnye-rle.pe-ke therre-ke-therre-le crowbar dig-do.along-pst.cont two-dat-two-adv crowbar nyente-kerte-le one-prop-ins ‘When we saw the evidence of witchetty grubs ‘‘cracking out’’ we started digging around. Two people at a time dug with one crowbar each.’ (Wilkins 1989: 499) The Arrernte evidence favours a secondary predicate analysis over an adverbial analysis, consistent with the remarks of Schultze-Berndt and Himmelmann (2004: 122–3). However, languages may differ in this respect. For instance, in Panyjima attribution might not be involved, as suggested by example (41). And the only putative example I have found of a transitive ‘subject’ controller in Warrwa is (43), from a text on wurnan ‘exchange, trade’; however, the controller in this example could equally be the ‘object’, the recipient of the gift, in which case a better translation would be ‘They gave things to one another two by two (in pairs)’. Warrwa (43) kujarra-kujarra, yirra-na ngirranayirr, two-two they-erg they:gave:them ‘Two by two they gave things to one another.’ More detailed information is required before any conclusion can be conﬁdently drawn. At present it can only be said that IMEs involving reduplicated quantifying expressions are potential instances of depictive secondary predicate constructions conveying the ‘number by number’ sense. They presumably represent instantiations of a general construction. Corresponding to the four other expression types distinguished in the previous section for co-participation depictives, the only type attested with the ‘number by number’ sense is the LE, as in (44), for which a ‘one by one’ interpretation would seem not unreasonable. On the other hand, it could be that the numeral and the following nominal, which displays the same case-marking, together constitute a referential NP, in which case the depictive sense would be a contextual interpretation. Evaluation of the status of this expression as a depictive construction must await further grammatical evidence. It remains to be seen whether further investigation will reveal expression types corresponding to any of the other co-participation types identiﬁed in the previous section.

196

W. B. McGregor

DjambarrpuyNu (44) warkthu-rr Narra Na¨thilnydja ga-n girri’ ga-nha rurrwuyu-na work-3 1.sg prior ipfv-3 clothes ipfv-4 wash-4 Na¨paki-w walalaN, wa¨Na-kurr mala-Nu-wurr bunbu-kurr white:person-dat 3.sg.dat place-perl pl-?-perl house-perl bay, waNga’-waNgany-Nur bunbu-Nur OK one-rdp-loc /abl house-loc /abl ‘The ﬁrst job I did was washing clothes for the white people, in all the houses, in each house.’ (Wilkinson 1991: 601)

5.4 Other types of quantifying depictive secondary predicate construction As already illustrated with example (36) above, numerals in Australian languages often also admit mass interpretations. Other nominals that sometimes express mass (though never count) quantiﬁcation, ‘big’ and ‘little’, also participate in quantifying depictive secondary predicate expressions. Little can be said about these rather poorly represented expressions, they are mentioned mainly for the sake of completeness, and we do not enter into discussion of their status. Examples (45) and (46) are probable IMEs. The subject of loorrloorryi ‘it ﬂowed’ in (45) is presumably an NP such as gamba ‘water’; jiginyaginya ‘very little’ may serve as a secondary predicate on this ellipsed NP. Alternatively, jiginyaginya ‘very little’ could be part of the subject NP whose head has been ellipsed. In (46) lambalambardi-nhingi-ngga (little-little-from-by) ‘from little ones’ is unlikely to be a part of an NP including the NP yangbalangga ‘by the young people’. (The presence of the ablative postposition on the quantity expression speciﬁes an initial quantity (age) as state interpretation—that the young people began to have it as youngsters. Without the ablative, the interpretation would be that the young people were little ones at the time of the having: ‘the young people should have it now as little ones’.) Gooniyandi (45) mangarri wardji gamba lanngarri, ngarrja—lil bit, not it:went water top little little.bit jiginyaginya loorrloorryi, very.little it:ﬂowed ‘Water wasn’t going over the top [of the trough]; just a little bit. It was ﬂowing over a little.’

Depictive secondary predicates in Australian languages

197

Gooniyandi (46) yangbala-ngga¼rni goorrijjawoorra, young:people-erg¼seq they:might:hold:it lambalambardi-nhingi-ngga, little:little-abl-erg ‘The young people should have it now, from an early age.’ Parallel to the ﬁfth category of cardinal senses (section 5.2), i.e. (e) temporally modulated, we also ﬁnd ‘mass by mass’ expressions. Again this sense is poorly attested, (47) being one of the very few examples available (cf. (36) above). Arrernte (47) unte kwenhe, Arrernte-ke kalty-irre-tyeke 2.pl.s emph Arrernte-dat knowledge-inch-purp kweke-me-kweke! little-uq-little ‘You should learn Arrernte little by little!’ [Don’t try to do it all at once.] (Wilkins 1989: 345) This may be a secondary predicate construction, though it is not an IME since kweke-me-kweke ‘little by little’ is not marked by the dative in agreement with the apparent controller.

5.5 Conclusion We have examined two potential types of quantitative depictive secondary predicate in Australian languages, co-participation and iterated involvement. Five different expression types are attested for the former. Pronominal Expressions involving oblique pronominals are almost certainly to be discounted as genuine secondary predicate constructions; those involving cardinal pronouns make better candidates. Of the other four expression types, Identically Marked Expressions seem the most widespread in Australian languages, and the best candidates for secondary predicate constructions. Unifying Expressions, Locative Expressions, and some Speciﬁc Lexical Expressions are also good candidates; however, some of these may be adverbial constructions, and they are not always easily separated from quantifying depictives.12 As for the iterative involvement type, IMEs are the 12 Thus Wilkinson (1991) takes the DjambarrpuyNu LE to be an adverbial expression, perhaps because the locative marker generally speciﬁes an adverbial-type relation (see also Himmelmann and Schultze-Berndt, Ch. 1, this volume). It is impossible with the limited information available to me to decide between the adverbial and depictive analyses.

198

W. B. McGregor

most widespread expression types; they represent good candidates for secondary predicates, though insufﬁcient information is available to permit ﬁrm conclusions concerning their status. In section 5.3 we encountered a number of difﬁculties in deciding whether or not expressions with quantifying secondary predicate senses count as instances of secondary predicate constructions. It was suggested that the seven criteria of Schultze-Berndt and Himmelmann (2004) should be augmented by a semiotic (formal-functional) condition, namely that the depictive element enters into an attributive dependency relation with its NP controller, as per McGregor (1997c : 171–4). This excludes instances in which the secondary predicate interpretation is a contextual interpretation, not conveyed by the construction itself. Evidence for the grammatical relation of attributive dependency is that the depictive lexeme has the potential of occurring NP internally as a modiﬁer of the entity-denoting nominal, or as a predicate in an attributive relational clause. This is a sufﬁcient, though not necessary condition—otherwise, iterated involvement expressions would probably be immediately excluded. Instead, it was suggested that a necessary condition is that the items in question can occur in the same construction as words demonstrably occurring in secondary predicate constructions according to the above condition. Granted similarity with a co-participation depictive construction, this weaker criterion will admit at least some iterated involvement expressions as quantifying secondary predicate constructions. Essential to a secondary predicate construction is, I have suggested, a dependency relation of attribution between the secondary predicate and its controller. But this does not provide a complete grammatical description of secondary predicate constructions (McGregor 1997c : 171–4). Another dependency relation must be identiﬁed, between the attributive dependency relation and the core of the clause (the verb and the participants). This is an adverbial-type relation of enhancement—i.e. embellishment or circumstantiation (McGregor 1997c : 61, 137)—typically temporal. This captures in a grammatical relation the semantic relation Schultze-Berndt and Himmelmann (2004) described under condition 1, to the effect that the depictive must apply to the controller within the time-frame of the event expressed by the main predicate. Figure 5.2 illustrates the relevant features of the grammatical structure of the ﬁnal clause of example (8), a typical quantifying co-participation IME.13 13 In this diagram the headed arrows indicate hypotactic dependency relations, the head pointing to the dependent. The equals sign (¼) indicates the grammatical relation of elaboration (of which attribution is a subtype); and x indicates enhancement.

Depictive secondary predicates in Australian languages Actor/ Agent

SoA

NP

VP

kand

nganyi-ngga

yoowarni-ngga

ngabbila

can’t

I-ERG

one-ERG

I : will : eat : it

199

=attribution x temporal contiguity Figure 5.2 Grammatical analysis (partial) of example (8)

There is a need for richer descriptions before we can progress much further in understanding quantifying secondary predicate constructions in Australian languages. Perhaps the most signiﬁcant gap is prosodic: little information on prosody in depictive secondary predicate constructions is available. Prosodic differences might perhaps permit us to distinguish between secondary predicates and (split) NPs. On the other hand, I have reservations about the need for condition 7 of Schultze-Berndt and Himmelmann (2004). Granted that intonation units in various languages represent information units, separation into distinct prosodic units need not necessarily argue against constructional status as a secondary predicate. Perhaps, however, a weaker condition would prove adequate: the potential of re-expression of the quantiﬁer in the same prosodic unit as the main predicate. I have been concerned here with the question of the status of quantifying depictives as constructions expressing secondary predication as a part of their essential constructional meaning. Whether they represent a separate quantifying type construction in their own right is a different question. It seems that quantifying IMEs are typically found in languages with IME depictive secondary predicate constructions, and would presumably belong to the latter constructions, probably as a semantically (though not grammatically) distinguishable subtype. Whether quantifying depictive constructions always represent subsets of depictive constructions is an issue that warrants investigation. Could a language, for instance, have an IME as a depictive construction type, but no corresponding quantifying IME—the latter being expressed by, say, an LE or SLE construction?

200

W. B. McGregor

Among many questions that have not been raised in this paper concerning quantifying depictives, one is worth mentioning in winding up the discussion: are there restrictions on the grammatical roles that can be attributed on? The examples we have discussed show a restricted range of roles: the controllers always serve in participant roles, and these seem to be restricted to the major ones of transitive subject and object, and intransitive subject. A small set of other clause types have been illustrated, including reﬂexive/reciprocals, in which again it is the subject that is the controller. In no instance have we seen another participant role such as indirect object of a clause of speech as the controller (examples of such controllers for non-quantifying depictives are given in the introduction to this volume, section 1.4.3); nor have we seen controllers in other grammatical relations such as cognate objects (e.g. ‘He sang a song alone’ where ‘alone’ is controlled by ‘song’), or body-part ranges (e.g. ‘He shot him in the head alone’, where ‘alone’ attributes on ‘head’). Given that depictive IMEs in Warlpiri can have indirect object controllers (Simpson, Ch. 2, this volume), it seems likely that the same will hold true of quantifying depictive IMEs.

6 Depictives in Kartvelian WINFRIED BOEDER

6.1 Introduction The purpose of this chapter is to give a short survey of Kartvelian, speciﬁcally Georgian and Svan, depictives; to delineate the extent to which ‘adverbials’ have a restricted orientation in these languages; and to present some tentative thoughts on the semiotic and semantic relationship between ‘state’ depictives on the one hand, and manner and similitive expressions on the other, as conceptualized in traditional rhetorical theory. This survey does not claim to be exhaustive, nor does it claim to solve problems of grammatical theory. Rather, it draws attention to a speciﬁc form and distribution of depictives that has parallels in other languages and that presupposes the availability of unambiguous morphological means in the respective domains of participantoriented expressions. Many Georgian examples are taken from literary texts of novelists (who are indicated in brackets, the immediate source of the examples being introduced by apud).1 Some of these examples may be more elaborate than most of the non-literary ones, but the relevant phenomena are essentially the same in the dialects. To show this, many examples are taken from these variants (which I am indebted to the editors of this volume, Eva Schultze-Berndt and Nikolaus Himmelmann, for encouraging work on Georgian depictives in the ﬁrst place, and for their extremely valuable suggestions and corrections of earlier versions of this chapter; to Christoph Schroeder (University of Osnabru¨ck) for a stimulating discussion on the question of metonymy (section 6.5.2), and for giving me the opportunity to read his careful analysis of German and Turkish depictives (Schroeder 2003); to my main informant, Rezo Kiknadze (Lu¨beck), for his constant help as a native speaker of Georgian; to Giorgi Tsotsanidze (Tbilisi) who provided some of the Tush examples; to Tedo Uturgaidze (Tbilisi) for a discussion of Tush grammar; to Lata Shukvani (Tbilisi/Mu¨nster) for some Svan examples; and to David Brown (Braunschweig) for correcting my English. All mistakes are of course mine. 1 In quoting Georgian sources, the following abbreviations are used: Chr=S ˇ aniZe et al. 1978; Dial=Gigineisˇvili et al. (1961); KEGL=Cˇikobava et al. 1950–64; K.v 1=K . vacˇ. aZe 1999; . vacˇ. aZe 1996; K.v 2=K K.v 3=K . vacˇ. aZe 2000b; Pr 1=Svan prose texts from SˇaniZe et al. 1939 (Upper Bal . vacˇ. aZe 2000a; K.v 4=K dialect); Pr 2=Svan prose texts from Davitiani et al. (1957) (Lower Bal dialect).

202

W. Boeder

are also indicated in brackets). Dialect texts offer some features of the spoken language, but so far no specimens of authentic urban colloquial speech are available. If not otherwise stated, the examples are Georgian. Besides Georgian and Svan, the Kartvelian or South Caucasian language family comprises Mingrelian and Laz, which form a continuum of dialects that are very similar to each other. The Laz dialect described by Kutscher and Genc¸ (Ch. 7, this volume) represents a variant without case agreement. On the other hand, agreeing depictives can be found in Mingrelian, which is an immediate neighbour of Georgian.2 After a short outline of morphological and positional resources in section 6.2, I will give a survey of the different categories of depictives and related adjuncts on the basis of the typological predictions outlined by Himmelmann and Schultze-Berndt (Ch. 1, this volume). The ﬁrst group shows agreement (3.1) that occurs most often with nominative controllers and which in this case shows a speciﬁc formal overlap with predicative complements. The depictive-like agreement of Tush ‘ablatives’ is one of the unsolved problems dealt with in this section. Formal similarity raises some problems of delimitation from copular and other constructions (3.2). Participant-oriented adjuncts without agreement (section 6.4) typically show adverbial and instrumental case forms, but also include a similitive postpositional construction. Section 6.5 deals with the formal and semantic relationship between depictives and adverbials, and with the behaviour of similitive constructions. Finally, I will venture a tentative interpretation of the Kartvelian coding properties differentiating the hierarchy of participant-oriented adjuncts (6.6).

6.2 Essentials on case-marking and agreement To begin with, Georgian and Svan offer (a) case-marking, (b) agreement, and (c) word order as a solid basis for the description of participant orientation and for distinguishing depictives from attributive modiﬁers. 6.2.1 Agreement and word order of modiﬁers and depictives Case-marking is more or less agglutinative in Kartvelian, but Svan has many morphophonemic processes that make word forms rather intransparent. 2 Cf. gaxareb-ul-ep-k´ gakociis mus ˇusˇi /ude-sˇa (gladden-PTCP-PL-ERG their.respective they.parted . ˙ ipsˇiZe 1914: (texts) 60, 15) ‘They parted and went home into their homes happy’; arti house-dir; Q osuri ucˇat monc.q˙ili kigexe midgasˇi-ren sapules (one:nom woman:nom black:adv array:ptcp.nom she.is.sitting somebody:gen-it.is tomb:dat; Xubua 1937: 17, 7) ‘A woman was sitting at somebody’s tomb dressed in black’. NB: The case traditionally called ‘ergative’ is used with subjects of both transitive and intransitive aorist verbs in Mingrelian.

Depictives in Kartvelian

203

Table 6.1. Georgian case and number suffixes

Nominative Vocative Ergative Genitive Dative Instrumental Adverbial

Singular

Plural

k.arg-i k.ac-i k.arg-o k.ac-o k.arg-ma k.ac-ma k.arg-i k.ac-is k.arg-ø k.ac-s k.arg-i k.ac-it k.arg-ø k.ac-ad

k.arg-i k.ac-eb-i k.arg-o k.ac-eb-o etc. Archaic plural: Nominative Oblique

k.arg-n-i k.ac-n-i k.arg-ta k.ac-ta

To facilitate morphological understanding, I give the formally transparent paradigm of the Georgian syntagm k.arg- k.ac- ‘good man’ in Table 6.1. In Modern Georgian non-poetic language, adjectival (and participial) modiﬁers normally precede their head noun. They agree with their head in case, but not in number. Compare nominative singular k.arg-i k.ac-i and plural k.arg-i k.ac-eb-i, ergative singular k.arg-ma k.ac-ma and plural k.arg-ma k.ac-eb-ma, etc. As we can see from the paradigm in Table 6.1, agreement is not simply copying of inﬂectional morphemes. Old Georgian had a morphologically more transparent case agreement: genitive k.arg-isa k.ac-isa, dative k.arg-sa k.ac-sa, instrumental k.arg-ita k.ac-ita etc. Modern Georgian and Svan dialects show a variation between total lack of formal correspondence between nominal and adjectival inﬂectional morphemes (e.g. genitive k.arg-i k.ac-is, dative k.arg-i k.ac-s, instrumental k.arg-i k.ac-it, etc.) and different forms of ‘reduced’ correspondence, one of which is used by the literary norm (as in Table 6.1). However, regardless of their formal make-up, all adjectival forms are glossed in the same way as the nominal forms they agree with: k.arg-i k.ac-it (good-ins man-ins), etc. Case agreement occurs in depictive adjectives, participles, numerals (see 6.3.1.3) and nouns in genitive and instrumental case (resulting in double case-marking; see 6.3.1.2 and 6.3.1.4). Examples of case agreement in adjectival depictives are: (1) mama cocxal-i movidaq˙e father:nom alive-nom he.came(aor).to.them ‘Their father came home alive.’ (Fereidanian; Dial 259, K. v 3) . (2) tkvenistana pativsacem k. ac-s xel-cariel-s ver ˙ you.like:dat honourable:dat man-dat hand-empty-dat impossible gavistumreb ˙ I.will.send.him.off.as.a.guest ‘I cannot let go a respectable man like you empty-handed as my guest.’ (N. Lomouri apud K. v 1: 222)

204

W. Boeder

The depictive adjectives in these examples have the ‘full’ inﬂection of head nouns. In this regard they are similar to discontinuous modiﬁers placed apart from their head (ApridoniZe 1986: 50), as in: Svan (3)

asˇxv hilv-s laxvedne vokvr-sˇv l´-la¨¯b-s one:dat mule-dat he.will.give.him gold-ins ptcp-load-dat ‘He will give him a mule loaded with gold.’ (Pr 1: 332,31)

On the other hand, unreduced inﬂection distinguishes depictives and modiﬁers in discontinuous noun phrases from head-adjacent modiﬁers. Compare k.ac-s xel-cariel-s in (2) with xelcariel-ø k.ac-s ‘empty-handed man’. Another difference between modiﬁer and secondary predicate is numbermarking. In contrast to nominal modiﬁers, depictives sometimes show number agreement, which is the same as with predicative complements (as in (4)): with plural controllers they can occur in the singular (as in (5a)) or in the plural (as in (5b)), depending on factors that need not concern us here: gaoceb-ul-eb-i vacˇ. r-eb-i merchant-pl-nom amaze-ptcp-pl-nom ‘The merchants were amazed.’ gaoceb-ul-eb-i (5) a. vacˇ. r-eb-i merchant-pl-nom amaze-ptcp-pl-nom saxl-idan house-from (4)

iq˙vnen they.were(aor) gamovidnen they.came.out

b. gaoceb-ul-i gamovidnen saxl-idan amaze-ptcp-nom they.came(aor).out house-from vacˇ. r-eb-i merchant-pl-nom ‘The merchants came out of the house amazed.’ (A. Cereteli _ apud K. v 1: 227) However, literary Georgian prefers the Old Georgian nominative plural form (see Table 6.1): (6)

vacˇ. r-eb-i merchant-pl-nom

(7)

gaoceb-ul-n-i gamovidnen saxl-idan vacˇ. r-eb-i merchant-pl-nom amaze-ptcp-pl-nom they.came.out house-from ‘The merchants came out of the house amazed.’

gaoceb-ul-n-i amaze-ptcp-pl-nom

iq˙vnen they.were

Depictives in Kartvelian

205

Whereas the old nominative plural is felt as archaic in other contexts (e.g. vacˇ. ar-n-i gamovidnen ‘merchant-pl-nom they.came.out’), it is still the rule in predicative, depictive and discontinuous modiﬁer constructions (ApridoniZe 1986: 50; see also (38), (39), (55), (115), and section 6.3.2.1). In other words: number agreement is a feature where predicatives and depictives go together (Plank 1985; see also 6.3.1.3). As for constituent order, depictives and related constructions most often precede the verb (as in (1), (2), (5), (7)), which is also the almost exclusive position of manner adverbials (ApridoniZe 1986: 42). But depictives may also follow the verb, with or without intervening constituents (as in (8); K.vacˇ. aZe 1996: 227). They normally follow their controller; they very rarely precede it, as in (5b). In most cases, position thus also distinguishes depictives from noun phrase-internal premodiﬁers: ˇsier-i (8) dabrunda gel-i gverd-eb-gaxvret-il-i ˙ he.returned(aor) hungry-nom wolf-nom side-pl-pierce-ptcp-nom ‘The hungry wolf returned with his sides pierced.’ ?* ‘the wolf returned hungry, with his sides pierced’ (Kartlian; Dial 303, K.v 3) ‘With pierced sides’ in (8) has to be interpreted as a depictive, and ‘hungry’ as a noun phrase-internal premodiﬁer, rather than as a preposed depictive. 6.2.2 Controllers of agreement in depictives The Georgian verb is polypersonal. Subject, direct object, and indirect object are coded in the verb, but do not necessarily have a verb-external counterpart (‘pro-drop’). Therefore, ﬁrst-, second-, and third-person controllers of depictives need not be verb-external constituents (Boeder 2002): (9) sait mi-di-xar, sad mi-xval where prev-go(prs)-you.are where prev-you.go(fut) egret-i daVoneb-ul-i? such-nom depress-ptcp-nom? ‘Where are you going, where will you go, depressed as you are?’ (Kartlian; Dial 281, no. 228,53, K. v 3) ise tq˙e=sˇi (10) ˇsen ravac ro bZane, ˙ you as conjunction you.ordered(aor).it so forest=in davtie gak.ocˇ. v-il-i ˙ I.left(aor).him bind-ptcp-nom ‘I did as you ordered [me] and left him bound in the forest.’ (Imerkhevian; Dial 381, no. 306,28, K.v 3)

206

W. Boeder

In addition, subject and object markers in the verb are not always overt. This is a result of morphological slot competition, as for instance in: (11) me tkven msˇier-s ar daktovebt ˙ I you:pl hungry-dat not I.will.leave.you ‘I will not leave you [plural] hungry. ’ (Kartlian; Dial 313, K.v 3) (12) me c amevedi cariel-i _ I I.went(aor).away empty-nom ‘I went away empty-handed.’ (Gurian; Dial 420, K.v 3) The verb form da-k-tov-eb-t in (11) consists of a perfectivizing preverb (that ˙ gives the verb form a future meaning), a root (tov- ‘leave’), a thematic sufﬁx ˙ (which assigns the verb form to the present tense series), a second-person object marker k- ‘you’ (Standard Georgian g-), and a plural morpheme -t (which pluralizes the direct object ‘you’). The ﬁrst-person subject marker v- (as in da-v-tov-eb ‘I will leave him/her/it’) is suppressed, because the object ˙ marker k- ﬁlls the person marker slot. In (12), -v- is lost before the initial v- of the root -ved- ‘go’. Notice that personal pronouns (as in (11) and in (12) above) and verbinternal person markers (as in (9) and (11)) do not combine with modiﬁers in head-modiﬁer syntagms (ApridoniZe 1986: 50). Only depictives and appositions can be linked to them. Verb-external pronouns do occur, but explicit ﬁrst- and second-person pronouns are unmarked for case and it is doubtful whether they should be considered controllers of agreement. The pronoun me ‘I’ is a form used for ﬁrst-person subjects, direct objects, and indirect objects. In Georgian and Svan, subject case-marking is aligned according to tense series and transitivity. Roughly speaking, the ergative is used with transitive verb subjects of aorist series verbs, the dative with transitive verb subjects of the perfect series and ‘indirect’ verbs, the nominative elsewhere. Ergative-marked depictives do occur but they are rare: (13) sada xar=o? gaxarebul-ma damiZaxa where you.are=quot gladdened-erg he.called(aor).me ‘‘‘Where are you?’’, he called me happily.’ (Kakhian; Dial 222, K.v 3) (14) tan-sˇisˇvel-ma da pex-sˇisˇvel-ma daic q˙o ˙ _ body-naked-erg and foot-naked-erg he.began(aor).it siarul-i going-nom ‘Naked and barefoot he began to walk.’ (Iesaia 20,2 Biblia 1989)

Depictives in Kartvelian

207

By far the most frequent type is a depictive controlled by a nominativemarked subject (as in (15)), including passives (as in (16)): Svan (15) pisˇir me-cˇde @ li kor=te [ . . . ] txvim many:nom ptcp-go.away:nom he.is house=to head me-k.vsˇe ptcp-break:nom ‘Many have gone home with broken heads.’ (Chr 18, 36–7) Georgian unda dejtesos gu¯cexvel-i (16) brinZ#-i rice-nom it.is.necessary that.it.be.sown(opt) unhusked-nom ‘Rice must be sown unhusked.’ (Atcharian; Dial 409, K.v 3) For examples with dative subject marking see (46), (47), (123b). Examples (17–22) illustrate depictive agreement in Svan, where it seems to differ from Georgian: while depictives agree with dative objects (as in (17) and (18)), both nominative and ergative subjects seem to occur in the same ‘nominative’ form as in object-oriented nominative depictives (compare (22) with (19–21)): Svan luq˙cˇu¯ra (17) h´ngr-a¨l-s saddle-pl-dat ptcp:break:dat itxax a¨gi=te ˙ they.brought(iprf).them.back.for.themselves home=to ‘They brought the saddles back home broken.’ (Pr 1: 253, 8) (18) luva¨r-s asˇtxvix ptcp:live-dat they.bury(prs).him ‘They bury him alive.’ (Pr 1: 67, 1) esxvı¯d (19) c. ic. la¨r l´ja¨r chicken:pl.nom ptcp:alive:nom they.met.with.him ‘He found the chicken alive.’ (Pr 1: 251, 18) a¨xgva¨bs (20) T. ariel l´-dgar Tariel:nom ptcp-die:nom they.found.him ‘They found Tariel dead.’ (Pr 1: 61, 15)

W. Boeder

208

(21) amnem [ . . . ] tvib=isga ˇcvadk.va¨r ˙ he:erg [ . . . ] lake=in he.put(aor).him.down.into . ¨ ter-ar-l´-xp´re eye-pl-ptcp-dig.out:nom ‘He put [sc. the boy] down into the lake with his eyes scratched out.’ (Pr 2: 29, 22, K.v 4) (22) dı¯na-d a¨dbine lizela¯¨ l haga¨r girl-erg she.began.it going:nom barefoot.‘nom’ ‘The girl began to walk barefoot.’ (The ergative of ‘barefoot’ would be haga¨r-d.) All ‘nominative’ forms of Svan should perhaps be interpreted as basic and unspeciﬁed for case. Direct object depictives occur in the nominative or in the dative, in conformity with split-ergative case alignment; the dative case is used with present tense series verbs, otherwise the nominative: Svan (23) tvep-s sga de@sgi ser bid=te ka lusˇdbuna riﬂe-dat in he.puts.it already case=to prev ptcp:clean:dat ‘He put the riﬂe into the case cleaned.’ (Pr 1: 54, 24) Georgian (24) unda c. q˙al=sˇi gadaagdon isav it.is.necessary water=in that.they.throw(opt).him again picar=ze gak.rul-i board=on bound-nom ‘They must throw [sc. Sizmara] into the water, bound on a board.’ (Khevsurian; Dial 68, K. v 3) mi/q˙vandes gulsgaxetkil-s! (25) kal-s ra/sa-V woman-dat what-only they.took.her.there heart.broken-dat ‘How could they take the woman there terriﬁed as she was!’ (Tush; Dial 116, no. 101, 30) Indirect object depictives (in the dative) are rare: (26) ima-t am ambav-ma moumzadebl-eb-s that-pl.obl this:obl news-erg unprepared-pl-dat moasc. ro it.reached(aor).them3 ‘This news found them unprepared.’ (V. Barnovi apud K. v 1: 222) 3

The verb form moasc.ro is transitive (ergative subject), but has an indirect, and no direct, object.

Depictives in Kartvelian

209

(27) k.odala-s puVur=sˇi mZ#domare-s c. avac. q˙di woodpecker-dat tree.hole=in sitting-dat I.came(aor).upon ‘I came upon a woodpecker sitting in a tree-hole.’ (Vazˇa Psˇavela apud K.v 1: 222) I was unable to ﬁnd a Svan example of a depictive controlled by a dative object (nor does K.vacˇ. aZe 2000b mention one). We may say that at least in narrative texts, nominative subjects and objects are the most frequent controllers of depictives. Other subjects and objects do occur, but indirect objects seem to be extremely rare. It remains to be investigated if other controllers are marginally possible.

6.3 Participant-oriented constructions showing agreement This section reviews the semantic range of depictives in Georgian and Svan, using the semantic domains established in the introduction to this volume. 6.3.1 Semantic categories 6.3.1.1 Posture, clothing, condition, state This group is amply discussed and documented in the Georgian grammatical literature under the heading of ‘adverbial of modality’ (vitarebis garemoeba, I. Imnaisˇvili 1948), ‘predicate having an attribute’ (atributiviani ˇsemasmeneli, Glonti 1955), ‘predicative . ˙ ˙ ˙ complement’ (predik.atuli damateba, K.vacˇ. aZe 1957), ‘predicative speciﬁcation’ ˙. ˙ (lit.: ‘determination’) (predik.atuli gansazVvreba, Basilaia 1966, K. vacˇ. aZe 1996, ˙ Burcˇ. ulaZe 2002), ‘predicative expansion’ (gavrcobili ˇsemasmeneli, K.iziria 1977), or ‘momentaneous characteristics’ (momentobrivi maxasiatebeli, Enukasˇvili ˙ 1977).4 Most examples are (anterior-)resultative participles (Boeder 1999b), as in (28–31), or adjectives that denote some deviance from a normal state, as in (32): Svan laxtexa (28) jarv nagza=unVve [ . . . ] l´-psˇir ˙ two week=after [ . . . ] ptcp-increase:nom they.returned ‘after two weeks [sc. the goats (nominative)] came back multiplied.’ (Pr 2: 323, 36–7, K.v 4) 4 Contrary to Leo K.vacˇ. aZe’s (1999: 51) judicious assessment of ‘predicative speciﬁcations’ (that they are today what they have always been), some authors derive them from nominal modiﬁers. They argue that these modiﬁers changed their position, came to be connected with the verb, lost their head noun, and ‘secondarily’ became adverbials of modality or ‘predicative speciﬁcations’ (see Burcˇ. ulaZe 2002, with further references), or even that they are ‘semantic-stylistic variants’ of nominal modiﬁers, and that there is no speciﬁc structural property that corresponds to this ‘stylistic nuance’ (Davitiani 1973: 250–1).

210

W. Boeder

Georgian (29) deda Niblia mxr-eb-gasˇl-il-i mother Niblia:nom shoulder-pl-open-ptcp-nom miegeba Zvirpas meuVle-s she.met(aor).him dear:dat husband-dat ‘Mother Niblia welcomed her dear husband with open arms.’ (T. Razik.asˇvili apud K.v 1: 223) (30) c. evida axla e k.ac-i gaxareb-ul-i he.went(aor) now this:nom man-nom gladden-ptcp-nom sax=sˇi house=in ‘Now this man went home happy.’ (Rachian 515, K. v 3) (31) na-svam-i martavda mankana-s ptcp-drink-nom s/he.controlled(aor) car-dat ‘S/He drove the car drunk.’ (lit.: having drunk) ˇsier-i, (32) ˇsevida ai k.ac-i mc. q˙urval-i he.went(aor).in art man-nom hungry-nom, thirsty-nom ‘The man entered [sc. the mill] hungry and thirsty.’ (Gurian; Dial 434, K.v 3) There is at least one other type of ‘state’, which is not (anterior-)resultative: future participles (marked by a circumﬁx sa- . . . (-el)-) with a gerundive meaning belong in this group (see also 6.4.3). They occur both as modiﬁers (da-sa-k.lav-i ‘to be slaughtered’ in: da-sa-k.lav-i gocˇ. -i viq˙ide (prev-prefkill-nom piglet-nom I.bought.it) ‘I bought a piglet for slaughter’) and as depictives: ˇsua-dVe-mde ˇsua sopl-is (33) datv-i meore dVe-s bear-nom second day-dat middle-day-till village-gen middle moedan=ze ga-sa-tq˙aveb-el-i egdo ˙ square=on prev-pref-skin-suff-nom he.lay.there ‘On the next day, the bear lay in the middle of the square till noon, to be skinned.’ (Vazˇa Psˇavela apud K. v 1: 222) These future participles are a counterpart of (anterior-)resultative participles: with resultatives, a present (or, rather, simultaneous) state is interpreted in terms of its causal relationship to a past event; with gerundives, the concurrent state is one of ﬁnality: in (33), ‘to be skinned’ is predicated of the bear for the time when it lies on the square.

Depictives in Kartvelian

211

6.3.1.2 Phase Secondary predicates denoting a phase also typically agree with their controller even in those instances where the adjunct itself is already case-marked (as in (37)). (34) is Z#er k.idev bavsˇv-i maxsovs he:nom still again child-nom I.remember.him ‘I remember him when he was still a child.’ . (35) igi patara daoblda ˙ he:nom little:nom he.became(aor).an.orphan ‘He became an orphan as a little child’ / ‘he was a little child when he . . . ’ (Vazˇa Psˇavela apud K.v 1: 223) Svan (36) mu dı¯na-s adc. wile jesˇdsemi l´-za¨j-s father:nom girl-dat marries(prs).her thirteen poss-year-dat ‘The father gives his daughter in marriage at the age of thirteen.’ Georgian (37) tekvsmeti c. l-isa-m c. armateb-it daamtavra Tbilis-is ˙ ˙ 16 year-gen-erg success-ins he.completed(aor).it Tbilisi-gen gimnazia gymnasium:nom ‘He sucessfully ﬁnished secondary school in Tbilisi at the age of 16 years.’ [from a calendar] Notice that Georgian and Svan employ different forms. Georgian has a genitive of age that is inﬂected;5 in Svan (and in Mingrelian and Laz, see Kutscher and Genc¸, Ch. 7, this volume), a derivational afﬁx is used: Svan has a possessive preﬁx l´-; compare l´-gzel ‘having a child (gezal)’ with l´-za¨j ‘having year’ in Ilia semi l´-za¨j li (Ilia three:nom poss-year:nom is) ‘Ilia is three years old’ (Pr 1: 43,8). 6.3.1.3 Quantity The following examples illustrate quantity phrases restricting the number of the subject: ˇcwen sam-i (38) c. indac. in k.ac-i c. avedit beforehand we three-nom man-nom we.went(aor) marto-n-i ˙ alone-pl-nom ‘Beforehand, we three men went alone.’ (Mokhevian; Dial 35, K. v 3) 5

Cf. also igi tekvsmeti c.l-isa-a (s/he:nom sixteen year-gen-is) ‘S/He is sixteen years old’. ˙

212

W. Boeder

(39) movedit sam-n-i we.came(aor) three-pl-nom ‘The three of us came.’ Svan

Vvasˇ (40) vosˇtxv ma¯rem bVe=xa¨n-ka four men:erg ravine=from-away male.ibex:nom ka¨¯xvtu¨rned we.dragged.it.out ‘We were four men who dragged the ibex out of the ravine.’ (Pr 1: 40, 22) (41) esˇxu misˇgu mu arda Cimbil-s One:nom my father he.was Siberia-dat ‘Only my father was in Siberia.’ (Svan; Pr 1: 5, 13) As seen in (38) and (39), the Old Georgian nominative plural forms with -n-i are used in these depictives. But the oblique counterpart (-ta, see Table 6.1) expected for the ergative etc. does not seem to be acceptable, the singular being used instead:

Zlivs sam-ma (42) es did-i kva this:nom big-nom stone:nom scarcely three-erg (*sam-ta) avc. iet (*three-obl.pl) we.lifted(aor).it ‘The three of us scarcely lifted this big stone.’ Notice that the forms with -n-i are also used for many predicative complements, especially in numerals and pronouns; (39) corresponds to: sam-n-i vart (three-pl-nom we.are) ‘we are three’, ra-n-i vart (what-pl-nom we.are) ‘what are we?’, tagv-eb-i arian mimal-ul-n-i (mouse-pl-nom they.are hiddenptcp-pl-nom) ‘the mice are hidden’ (cp. the ‘modern’ plural marker -eb- in isi-n-i megobr-eb-i arian (s/he-pl-nom friend-pl-nom they.are) ‘They are friends’). The use of speciﬁc forms shared by depictives and predicative complements seems to point to a similar semantic and syntactic status— whatever the formal basis of this similarity might be. 6.3.1.4 Place In contrast to, for example, Old Greek (Lu¨bker 1837; Boeder 1999a), location and time are not normally expressed by agreeing expressions in Georgian or Svan. There is one exception, though. In one of the east Georgian mountain dialects, Tush, the point of departure, which is coded by the instrument case and which we call ‘ablative’, seems to agree with the

Depictives in Kartvelian

213

subject (which has ergative case-marking in (43–45) and dative case-marking in (46) and (47) according to Georgian alignment rules): Tush (43) abano-t momdinara/-m balk.un-ita-m bath-ins coming-erg balcony-ins-erg gadaxeda-d’ [ . . . ] he.looked(aor).down-and ‘[They took Guram quickly into the bathroom;] coming from the bath, he looked down from the balcony and [ . . . ]’ (UturgaiZe 1960: 123, no. 48, 165) . vasˇrap/`ı (44) Revazo-m cxen=z-ita-ma=v i/Vv Revaz-erg horse=on-ins-erg=just he.took.it pitcher:nom ‘Revaz, sitting astride on a horse, just picked up the pitcher.’6 (45) iZaxes Z#avax-ita-ma, gvisˇvelet, vin xart they.called Javakhe-ins-erg help.us who you.are qml-ian-i sword-poss-nom ‘[The Chechens have come on a raid.] People called from (the settlement) Javakhe: Help us, those of you with a sword.’ de /@ c. q˙o` top-is sreva (46) /¯ı/k cixe=sˇi-ita-s there fortress=in-ins-dat he.began.it riﬂe-gen shooting:nom Dido-eb-isa-d Dido-pl-gen-adv ‘[Dja’o went into the fortress and] he began to shoot at the Dido men there from the fortress’ (UturgaiZe 1960: 97, no. 29, 19) . tav enaxav (47) pirvel mosaxvev=sˇi-ita-s Kumala/urt-is ﬁrst bend=in-ins-dat Kumala’urta-gen head:nom he.saw.it ‘from the ﬁrst bend [sc. of the path] he could see the beginning of [sc. the settlement] Kumala’urta’ (UturgaiZe 1960: 97, no. 29, 12) Our hypothesis is that case agreement in Tush occurs only in cases like (43–47) where the ablative designates not only a point of departure, but also the position of the subject during the action denoted by the main predicate. 6 This and the following example are from an ethnographic book by Giorgi CocaniZe (Giorgobidan giorgobamde, Tbilisi, 1990, p. 225).

214

W. Boeder

The following examples, on the other hand, are different: Tush (Georgian) (48)

ert k.ac gamovard cix-ita-j one:nom man:nom he.rushed(aor).out fortress-ins-‘nom’ gaqureb-ul-ı`, c. q˙al-moc. q˙ureb-ul-ı` heated-ptcp-nom water-become.thirsty-ptcp-nom ‘One man rushed out of the fortress heated and thirsty.’ (Dial 106, No. 85, 13-14, K.v 3) / e /@ s da-i-d’ [ . . . ] (49) adg ak.avan=sˇig-ita-j she.stood.up cradle=in-ins-‘nom’ this:nom sister-nom-and ‘The sister stood up from the cradle and [ . . . ]’ (UturgaiZe 1960: 123, no. 110, 12) . ˇsor-ita-i=ve=v (50) eg sapon-i=v naxevar this:nom soap-nom=quot far-ins-‘nom’=just=quot half ert k.ar-eb-s miasxi=v, naxevar one door-pl-dat pour(aor).it=quot half meore k.ar-eb-sa=v second door-pl-dat=quot ‘Pour half of this soap from far at one door, and half of it at the second door.’ (UturgaiZe 1960: 114, no. 46, 170) /¯ı /m Vor-m potl-eb=sˇi/-ita-j (51) memr turı` then apparently that:obl pig-erg leaf-pl=in-ins-‘nom’ naxa-d’ [ . . . ] it.saw(aor).it-and ‘[A pig frightens the other animals; the cat is going to eat;] then the pig, they say, saw it from within the leaves [ . . . ]’ (Dial 122, no. 109, 34) (52)

xalx movidisa-d’ ˇsevidis cixe=sˇig q˙vela/ sopl-eb-it all village-pl-ins people:nom come-and enters.it fortress=in ‘People from all villages are used to coming and entering the fortress.’ (Dial 106, no. 85, 3)

In (48) and (49), the ablative does not denote the location of the subject during the action, but the starting point of a movement of the subject. In (50) and (51), the ending -ita-j in fact does mark a location, but -j cannot be the result of nominative agreement, since ‘to pour’ and ‘to see’ are transitive verbs that require an ergative subject marker (i.e. we would expect ˇsorita-m in (50) and potl-eb=sˇi/-ita-m in (51)). In these examples, -itaj seems to be an allomorphic variant of -ita- or -it (the latter is illustrated in (52)); synchronically, -j is perhaps a word juncture marker in these instances (it is

Depictives in Kartvelian

215

glossed as ‘nom’). In other words, -ita-j seems to have two interpretations: it may represent (a) an instrumental þ nominative sequence (based on agreement, i.e. nominative agreement in parallel with ergative and dative agreement as illustrated in (43)–(47))7, or (b) an allomorph of the instrumental case-marker. In some cases, the locative meaning component is made precise by postpositions followed by -ita-. This position of -ita- shows that it is used as a phrasal case sufﬁx. Sequences of this type, for instance =z(e)-ita ‘on’ in (44) and =sˇi-ita ‘in’ in (46), (47), and (51), occur in the the eastern Georgian mountain dialects that have been in contact with Chechen and other east Caucasian languages, where multiple case-markings are very common. In these languages we also ﬁnd cases of subject agreement in adverbs like ‘here’, ‘inside’, ‘around’, but also ‘quickly’. These forms of agreement remain to be explored. There are two residual problems. First, expressions with -ita- also occur as modiﬁers, for instance in: cxen=ze-it(a) k.ac-ma (horse=on-ins man-erg), cxen=ze-it(a) k.ac-s (horse=on-ins man-dat) etc. ‘the man on the horse’ (T. UturgaiZe, p.c.). Could examples (43–47) therefore be explained in terms of modiﬁer constructions? We believe that this is unlikely. Notice, ﬁrst, that these modﬁers have a locative, not an ablative meaning (‘on the horse’, not ‘from the horse’). Second, if cxen¼z-ita-ma were a modiﬁer in (44), it would have to precede its head noun. (An appositive interpretation, as in the English translation, cannot, however, be ruled out.) Third, (45) cannot be explained on the basis of a head-modiﬁer construction: the verb is in the plural, but Z#avax-ita-ma is in the singular, and cannot mean ‘those living in Javakha’ (with ellipsis of a head noun); if it is analysed as a depictive, the singular is expected (as with predicative complements; see 6.2.1; see also (5b), (11), (17), (19), (24), and (28)). Note, incidentally, that modiﬁers in Georgian noun phrases have to be either in the genitive or adjectival. Thus, -ita- seems to convert postpositional phrases into adjectivals to make them available for attribution. A second problem is the question of why agreement occurs with ablatives but not in other locative expressions. An answer may be given along the lines just hinted at. As the instrumental case sufﬁx -ita- converts postpositional locatives into adjectival forms that make them available as modiﬁers, it makes postpositional phrases also available as agreeing depictives: postpositional

7 Unfortunately, I have not been able to ﬁnd an unambiguous example with nominative subject agreement, but a present tense variant of (46) would probably be: cixe=sˇi-ita-j ic.q˙ebs top-is sreva-s ‘He begins to shoot from the fortress’, with a nominative subject (because the verb is a present-tense form).

216

W. Boeder

phrases in general cannot agree in case, but expressions inﬂected for genitive or instrumental can (see e.g. -isa-m in (37); see also Boeder 1995). These highly tentative suggestions regarding Tush ablatives obviously are in need of further investigation. 6.3.2 Problems of delimitation 6.3.2.1 Detached participles One problem of delimitation of depictives from similar constructions is the lack of prosodic data. To the best of my knowledge, investigations on Georgian intonation have not dealt with depictive and similar constructions so far. However, intonational detachment from the rest of the clause can be an indicator of non-depictive status (see the introduction to this volume and Schroeder 2003), and the participial clause in the following example, considered in the literature as ‘predicative speciﬁcation’, seems to be a sentence-margin adjunct whose detachment is marked by a comma (see, however, the detached construction abano-t momdinara/-m ‘coming from the bath’ in (43) without a comma): k.maq˙opileb-it gataceb-ul-t, (53) raVac ˙ some.speciﬁc satisfaction-ins ravish-ptcp-pl.obl Vor-eb-s ertmanet-isa=tvis bevrZ#er im that:obl pig-pl-dat each.other-gen=for many.times c. autavazebiat ding-i they.have.presented.it muzzle-nom ‘Ravished by some feeling of satisfaction, those pigs have pushed each other with their muzzles many times.’ (I. Cˇ. avcˇ. avaZe apud ApridoniZe 1986: 50) Notice that the participle agrees with the dative subject (the pigs): like predicative complements, it is marked by the old oblique plural marker -t (see 6.2.1). 6.3.2.2 Copular constructions Traditional Georgian grammar tends to confuse depictive and copular constructions. These are indeed difﬁcult to distinguish in some cases, where the analysis of these expressions as depictive or copular depends on whether the verb is interpreted as a copula or as a main verb. As far as I can see, the following examples are not depictive but copular, with a non-omissible predicative complement. The copular verbs mean ‘to become and be’: nac. q˙en-i darcˇen (54) Zm-eb-i brother-pl-nom annoyed-nom they.became.and.remained ‘The brothers became annoyed.’ (Imeretian; Dial 444, K.v 3)

Depictives in Kartvelian

217

(55) gamovedit or-n-i obol-n-i am we.came(aor).out two-pl-nom orphan-pl-nom this:obl . trial mic. a=ze u-pur-o-d, ˙ turning.round earth=on neg-bread-suff-adv u-pul-o-d, u-bina-o-d, neg-money-suff-adv neg-lodging-suff-adv u-nugesˇ-o-d, u-tvistom-o-d ˙ neg-comfort-suff-adv neg-kinsman-suff-adv ‘As a result, we were two orphans on this rotating earth, without bread, without money, without a roof, without consolation, without relatives.’ (I. Cˇ. avcˇ. avaZe apud K. v 1: 174) (56) bicˇ. -i martlac k.arg-i izrdeboda boy-nom truly good-nom he.grew(iprf).up ‘The boy indeed became [‘grew’] [a] good [one].’ (Gr. AbasˇiZe apud K. v 1: 222) The meaning of the last sentence is not: ‘The boy grew up, being really good’ or the like. Again, compare the following semantically similar sentences: . ˇcem col-s (57) esa=o vapatiev=o, ˙ this:nom=quot my wife-dat I.forgave(aor).her=quot ˇsvil-i=o, cocxal-s davtovem=o, mara ˙ offspring-nom=quot alive-dat I.will.leave.her=quot but unda movk.la bicˇ. -i boy-nom it.is.necessary that.I.kill(opt).him ‘I have forgiven my wife for having a child, I will let her live, but I must kill the boy.’ (Rachian; Dial 520, K. v 3) Svan (58) mesme na¨¯t=i desˇ acvirx lEja¨r third part:nom=also not.possible they.let.it ptcp:live:nom ‘They did not even let a third [sc. of the soldiers] survive.’ (Pr 1: 4, 16) (59) prinvel-s ar gausˇobs ca=ze mimaval-s mouk.lav-s bird-dat not he.will.let.it.go sky=on going-dat unkilled-dat ‘[sc. Adua] will not let any bird that ﬂies in the sky escape without having killed it.’ (Khevsur; Dial 19, no. 14, 12–13 apud K.v 3) (57) and (58) are copular, (59) is depictive. Semantically, ‘leave’ in (57) is a kind of causative of ‘remain’ in (54): in both cases, the complement (‘annoyed’, ‘alive’) cannot be omitted, but ‘unkilled’ in (59) can. The following example (60a) is perhaps ambiguous between a predicative complement and a depictive reading. But the depictive reading can be

218

W. Boeder

enforced by postposing the adjective phrase (as in (60b)), and the nondepictive reading by giving ˇsor-idan ‘from afar’ a focusing intonation (‘From a distance, it looked white and beautiful, but when you came nearer, . . . ’): (60) a. ˇsor-idan mta tetr-i da lamaz-i far-from mountain:nom white-nom and beautiful-nom ˇcanda it.appeared ‘The mountain appeared from afar(,) white and beautiful.’ or: ‘The mountain looked white and beautiful from afar.’ (L. Gotua apud K.v 1: 222) b. ˇsor-idan mta ˇcanda tetr-i da lamaz-i 6.3.2.3 Adverbial complements Optionality is generally seen as a major criterion for distinguishing depictives not only from the copular constructions illustrated in the previous section but also from verbal complements, which likewise are considered to be obligatory. This criterion is not always easy to apply in the case of verbs of stance and posture that normally require some specifying adverbial expression. Agreeing nominals can ﬁll this position, as shown in (61) and (62). However, these do not differ in any way from participles in depictive function, discussed in section 6.2.1, except that they appear to be ‘more obligatory’. Therefore, it might be wise to analyse them as depictives rather than introducing a new category for them: ˇcakanc-ul-i da daic. q˙o pikr-i (61) daZ#da exhaust-ptcp-nom he.sat.down and he.began.it thought-nom ‘He sat down exhausted and began to think.’ (Mokhevian; Dial 34, K. v 3) ˇsesˇineb-ul-i (62) didxan izˇda dedupal-i longtime she.sat queen-nom frighten-ptcp-nom ‘Thequeen sat there frightened foralong time.’ (Gurian;Dial 432, K.v 3) 6.3.2.4 Topicalization and quantiﬁer ﬂoating Since modiﬁers precede their head noun, the adjectives in the following examples cannot be simply postposed modiﬁers. But it is not clear if these adjectives should be considered depictives. Alternatively, their position could be the result of the topicalization of their head noun (comparable to cases of ‘split topicalization’ in English and German; H. van Riemsdijk’s term apud Kniffka 1996): (63)

[torola-m] ganabva ostatur-i icis ˙ ˙ ˙ lark-erg budging:nom masterful-nom it.knows.it ‘As for budging, [the lark] can do it in a masterly manner.’ (I. Gogebasˇvili apud K. v 1: 222) cp. torola-m ostatur-i ganabva icis ‘The lark knows how to ˙ ˙ ˙ do a masterful budging.’

Depictives in Kartvelian (64) simVera k.arg-i singing:nom good-nom ‘To sing well is what you dance?).’ cp. k.argi simVera gcodnia!

219

gcodnia you.know(prf).it obviously know (but let us see how you ‘You know a good song!’

Some quantiﬁers seem to have the same positional and semantic properties, and L. K. vacˇ. aZe considers an example such as the following one, a ‘predicative speciﬁcation’, i.e. as a depictive: (65) c. amsvlel-i bevr-i minaxam=o leaving-nom many-nom I.have.seen.them=quot da momsvlel-i k.i ara=o and coming-nom but not=quot ‘I have seen many people leave, but nobody who returns.’ (Kakhian; Dial 202, K.v 3) Svan (66) min-s eser sacˇkva¨r-s xahvdix xva¨j-s they-dat quot gift-dat they.give.it.to.them many-dat ‘As for gifts, they give many.’ (Pr 1: 5,2) Yet the following example seems to offer a real case of ‘quantiﬁer ﬂoating’, since ˇc¯ı ‘all’ does not show case-marking (expected: ˇc¯ı-s ‘all-dat’): Svan (67) zura¯l woman:nom ‘The woman

ˇc¯ı ka l´dga¨r-s isˇ´lda¨ni all prev she.enumerates dead-dat enumerates all the dead.’ (Pr 1: 8, 1) (68) daqa¨r-s eser ka xahvdi cˇ¯ı goat:pl-dat quot prev he.will.give all ‘He will give all the goats.’ (Pr 1: 390, 31)

6.4 Participant-oriented adjuncts without agreement In section 6.3 we discussed various semantic classes of participant-oriented adjuncts showing agreement, and argued that these can be regarded as depictives, and delimited them from copular or complement constructions also showing agreement. Lack of agreement, however, does not prevent restricted, if not unambiguous, orientation. Some adjuncts without agreement are oriented towards the object with transitive verbs and towards the subject elsewhere.

220

W. Boeder

6.4.1 Adverbial expressions without case-marking Svan has quite a few idiomatic postpositional phrases that function as secondary predicates: Svan (69) amecˇu xocˇa gvi la¨nı¯sdda here good heart:nom we.sat.down ‘We sat down here in a good mood.’ (Pr 1: 36, 6) (70) xek.va¨d da¨va¨ xola¯m gu=zˇi lipsˇvde he.wanted dev:gen bad:dat heart=on sending.away:nom ‘He wanted to send the dev away bad-tempered.’ [referring to the dev, a fabulous man-like being] (Chr 154, 5) Some Georgian counterparts are also non-agreeing expressions such as cud xasiat=ze (bad character=on) ‘in a bad mood’, but others are agreeing participles: moxarebul- ‘gladdened, happy’; cf. (30). Where nonagreeing idiomatic expressions exist, these are usually preferred, but there is no categorial exclusion of agreeing expressions for any given semantic category. Oriented adverbial expressions of state (‘posture’) are illustrated by expressions involving non-agreeing adverbs like q˙iramala ‘head over heels’, . pirkve ‘prone’, gul-aVma (heart upwards) ‘with the face upwards’, gul-daVma (heart downwards) ‘prone’, tav-daq˙ira ‘head ﬁrst’, etc. with subject or object orientation: (71) tavdaq˙ira gadaesˇva head.ﬁrst s/he.plunged.over ‘S/He plunged head ﬁrst into

c. q˙al=sˇi water=in the water.’

(72) tavdaq˙ira gadausˇva head.ﬁrst s/he.made.him/her.plunge.over ‘S/He made him/her plunge head ﬁrst into

c. q˙al=sˇi water=in the water.’

Svan ˇc. er=xa¨n-cˇu (73) usˇgul xoc. bina salda¨t-s reverse he.hung.him.up soldier-dat ceiling=from-down ‘[sc. The giant] hung the soldier upside down from the ceiling.’ (Chr 170, 8) The same is true for an ‘ornative’ adverb like amara ‘having only x(+gen)’, which has no adjectival counterpart. Semantically it belongs to the

Depictives in Kartvelian

221

instrumental ornative or concomitance expressions (as in (99)): (74) marto-d-marto modioda xanZ#l-is amara ˙ ˙ alone-adv-alone:nom he.used.to.come dagger-gen having.only ‘He used to come all alone, having only his dagger.’ (V. Barnovi, KEGL s.v.) ‘Ornamental’ depictives for body-part postures and clothing, on the other hand, are most often coded by participial possessive compounds (see (14), (29), (75)). All these uninﬂected expressions allow no agreement: as preferred lexicalized units they block the use of alternative agreeing expressions. Although most postures are coded by uninﬂected units in Georgian, there is no intrinsic necessity for such a form. In fact, there are some more speciﬁc expressions that can be inﬂected, as in: pex-mok.vec-il-i (75) taxt=ze iZ#da ˙ ˙ divan=on he.sat foot-fold-ptcp-nom ‘He sat on the divan with his legs folded.’ (KEGL s.v.) Svan also has, for instance, an agreeing adjective ‘upright’: Svan (76) er k.acˇ l´g some upright:nom stands ‘Some people stand upright.’ (Pr 1: 8, 11) 6.4.2 Adverbial case forms with essive functions One of the most important coding forms of adverbial relations in Kartvelian languages is the adverbial case, which may convey an ‘essive’ meaning. Examples (77) and (78) are from Georgian, and examples (79) to (81) from Svan. (77) ekim-ad musˇaobs doctor-adv s/he.works ‘S/He works as a doctor.’ (78) k.arg ekim-ad itvleba good doctor-adv s/he.is.counted ‘S/He is considered a good doctor.’ Svan xak.u (79) cˇ a¨ˇs-d ˙ husband-adv she.wants.him ‘She wants him as her husband.’ (Pr 1: 256, 37)

222

W. Boeder

. (80) c. el-d esvsiped mule-adv I.turned.her ‘I turned her into a mule.’ (Pr 1: 71, 26) (81) ucxo k.uma¨ˇs-d a¨nqdeni strange cattle-adv he.will.come ‘He will come as a strange head of cattle (says the giant sorcerer).’ (Pr 1: 62, 16) The adverbial case also occurs on possessive adjectives, where it is preceded by the possessive sufﬁx -ian (which is a phrasal sufﬁx, as seen in (83)). Expressions like these denote concomitance, and can also be interpreted as participant-oriented. (82) col-sˇvil-ian-ad movida wife-child-poss-adv he.came ‘He came with his wife and children.’ ˇcamoatara (83) c. q˙al-ma mtel-i xe tavis ˙ water-erg it.carried.down whole-nom tree:nom its:refl tot-eb-ian-Z#irk.v-eb-ian-ad ˙ ˙ branch-pl-poss-trunk-poss-adv ‘The water carried down the whole tree with its branches and trunk.’ (V. Barnovi apud Glonti 1978: 85) ˙ Svan k.e@sa¨rsˇa lu-zra¯l-l´-bopsˇv-d, (84) an´ga¨nx ma¨g they.rose all:nom king’s.family:nom poss-woman-poss-child-adv ma¨g cicv-a¨r i zˇeV-a¨r-i lu-gocˇ -d, ˙ poss-piglet-adv cat-pl.nom and dog-pl.nom=and all:nom i la¨jbinex lilxin acˇa¨dx l´-c. h-a-d they.went ptcp-invite-pl-adv and they.began banquet:nom ‘All the king’s family rose, wife and children and piglets included, the cats and dogs all went as guests and began the banquet.’ (Svan; Chr 165, 21–2) Georgian (85) [xorc-s] meat-dat c. q˙al=sˇi water=in

ˇsecˇ. amt nu not:imp you.eat.it moxarsˇul-s; mxolod boiled-dat only

naxevr-ad half-adv cecxl=ze ﬁre=on

ˇsemc. var-s roasted-dat ˇsemc. var-i roasted-nom

an or ˇc. amet eat:imp

Depictives in Kartvelian

223

tav-pex-ian-ad da ˇsigneul-ian-ad head-foot-poss-adv and inner.parts-poss-adv ‘You must not eat [the meat] half-roasted or boiled in water; but only roasted on a ﬁre, with head and feet and intestines.’ (Exodus 12, 8–9, Biblia 1989) The expression ‘with head and feet’ in the last example must refer to the object of eating. Notice that this expression parallels state/condition depictives in the preceding context: ‘half-roasted, boiled’, which are condition/state depictives with case agreement. Besides derivational and inﬂectional forms, there is a periphrastic expression of concomitance consisting of the postposition -tan plus ertad ‘together’. Once again, this expression has a participant orientation which allows for different controllers and thus may give rise to ambiguity. In the following example, the agent either ate it in company of the dog, or ate it and the dog: ˇsecˇ. ama (86) ZaVl=tan ertad dog=with together s/he.ate(aor).it.up ‘S/He ate it up with the dog.’ or ‘S/He ate it and the dog up.’ 6.4.3 Future participle with adverbial case-marking From the point of view of morphological form and participant orientation, Georgian purpose expressions can also be regarded as non-agreeing participantoriented expressions. They are future participles marked with the circumﬁx sa-__-(e)l- (see 6.3.1.1) and the adverbial case sufﬁx -ad. They seem to be oriented towards subjects of intransitive verbs and towards direct objects of transitive verbs: da-sa-sveneb-l-ad (87) k.unZul=ze miva island=on s/he.will.go prev-pref-relax-suff-adv ‘S/He is going to the island to relax.’ (88) k.unZul=ze gaagzavna da-sa-sveneb-l-ad island=on s/he.sent.him/her prev-pref-relax-suff-adv ‘S/He sent him/her to the island to relax.’ Svan (89) lic axa¨j nicˇvaresˇ i totre la-bra¯l-d ˙ ˙ water:nom bring face:pl.gen and hand:pl.gen ptcp-wash-adv ‘Go fetch some water to wash our hands and faces.’ (Pr 1: 376, 34–5) In addition to the future participle followed by the adverbial case illustrated in (89), Svan has a second supine formation with the directional sufﬁx

224

W. Boeder

-te instead of -d: Svan (90)

Cxvita¨gn la¨-txviar=te acˇad Tskhvitagan:nom ptcp-hunt=to he.went ‘Tskhvitagan went hunting.’ (Pr 1: 376, 4)

6.4.4 Other functions of the adverbial case The adverbial case also forms simple adverbs most of which seem to have a semantic subject orientation: (91)

prtxil-ad midiodnen ˇcven-i mgzavr-eb-i cautious-adv they.walked our-nom traveller-pl-nom ‘Our travellers walked cautiously.’

(92)

matarebel-i zant-ad daiZra ˙ ˙ train-nom lazy-adv moved ‘The train moved lazily.’ (R. GvetaZe apud K. v 1: 181) ˙ . . ˇsua mindor=ze lurZ#-ad molaplape tba mocˇanda ˙ middle ﬁeld=on blue-adv blazing:nom lake:nom showed ‘In the middle of the ﬁeld a shimmering blue lake appeared, [‘shimmering bluely’].’ (Vazˇa Psˇavela apud K. v 1: 181) Sˇisˇia gul-ian-ad atvalierebda Shishia:nom heart-poss-adv he.inspected(iprf).them am napexur-eb-s this footprint-pl-dat ‘Shishia was inspecting these footprints excitedly.’ (T. Razik.asˇvili apud K.v 1: 181)

(93)

(94)

There are a few ‘real’ manner adverbs with the sufﬁx -a instead of -ad: ˇckar-a ‘fast’, xmamaVl-a ‘loudly’, nel-a ‘slowly’, etc. These adverbs seem to be exclusively process-oriented. Example (95) illustrates a typical contrast between a real manner adverb (‘fast’) and a depictive, which is well known from other languages: (95) a. ˇckar-a mivida fast-‘adv’ s/he.went.there ‘S/He went there fast.’ (process-oriented) b. mocˇkarebul-i mivida accelerated-nom s/he.went.there ‘S/He went there in a hurry.’ (subject-oriented)

Depictives in Kartvelian

225

‘Fast’ in (a) is a simple manner adverb; the participle in (b) is a subjectoriented state depictive. It is an open empirical question as to which adverbs with the full adverbial case sufﬁx -ad are exclusively process-oriented and which also allow participant orientation. In many instances, it will in fact be quite difﬁcult to resolve this issue because of the possibility of metonymic shifts, as further discussed below in section 6.5.2. Quantifying adjectives with adverbial case-marking seem to behave like agreeing quantiﬁers in every other respect. Compare (65) and (66) above with the examples in (96–98). (96) pul-i blom-ad akvs money-nom abundant-adv s/he.has ‘S/He has plenty of money.’ (‘abundantly’) Svan (97) dia¨r-s masa¨r-d ama¯rex bread-dat abundant-adv they.prepare.it ‘They prepare plenty of bread.’ (‘abundantly’) (Sv 1: 8, 24) masa¨r-d (98) ma¯¨ Vvra-s ik.edx ﬁne-dat they.get abundant-adv ‘They get a high indemnifying ﬁne.’ (Sv 1: 6, 21) 6.4.5 Instrumental case forms The instrumental case occurs in different contexts: concomitance expressions (see (99)), quantity expressions (see (100)), manner expressions (see (37), (101); cf. (91)): (99) ˇsakr-it svam sugar-ins you.drink.it ‘Do you drink tea with

ˇcai-s tu rZ-it? tea-dat or milk-ins sugar or with milk?’

ˇsedis (100) ˇc ir-i urm-eb-ita da gamodis ˙ plague-nom it.goes.into cart-pl-ins and it.comes.out misxl-ob-ita=o ounce-collective-ins=quot ‘Calamity enters in [quantities carried by] carts and comes out in ounces.’ (101) siprtxil-it midiodnen ˇcven-i mgzavr-eb-i caution-ins they.walked our-nom traveller-pl-nom ˙ azbegi ‘Our travellers were going along there with caution.’ (A. Q apud K. v 1: 180)

226

W. Boeder

There is an interlingual variation between instrumental-marked and depictive participles. The following expression in Svan takes instrumental case-marking: Svan (102) anVrix ma¨g la¨xiada¯l-usˇ they.come all joy-ins ‘They all come cheerfully.’ (Pr 1: 66, 37) In Georgian, on the other hand, ‘cheerfully’ would be rendered by an agreeing participle, i.e. gaxareb-ul-eb-i (gladden-ptcp-pl-nom); see also (13), (30), and note 3. 6.4.6 Expressions with =vit ‘like’ Expressions of comparison are similar to adverbial and instrumental casemarked expressions, except that they have postpositional marking instead of case-marking. They seem to be oriented towards subjects or direct objects, but their orientation is not formally marked.8 Notice that the postposition =vit ‘like’ occurs with either the ‘nominative’ form (-i-) or the dative (Zera-sa=vit, k.ata-sa=vit in (110))—interchangeably according to ˙ V. Imnaisˇvili (1997: 84). The ‘nominative’ is not an agreement marker: it also occurs with oblique controllers, for instance the ergative in (107). (103) veravin cek.vavda lek.ur-sa da ˇcacˇnur-s nobody:nom danced Lekuri-dat and Chachnuri-dat ˙ azbeg-i=vit Q Qazbegi-‘nom’=like ‘Nobody could dance the Lekuri and the Chachnuri like Qazbegi.’ (V. Imnaisˇvili 1997: 87) Svan (104) pek=sˇa¯l esvgeni hok.er=isga ﬂour=like I.will.fall.down ground=in ‘I will fall down to the ground like ﬂour.’ (Pr 1: 69, 1) Vumir-s (105) topi pindix=sˇa¯l azzi riﬂe:gen bullet=like he.sends ﬁr-dat ‘He throws the ﬁr like a bullet.’ (Pr 1: 60, 20) However there is at least one feature that points to a participant orientation: the similitive construction ‘agrees’ in number with the relevant participant 8 The same is true for the periphrastic alternative with the word rogorc ‘like’, which has no particular orientation: xelebi rogorc prtebi gasˇala (hands like wings he.opened.them) ‘he opened his arms like wings’ (cp. (108)).

Depictives in Kartvelian

227

(K.vacˇ. aZe 1996: 177–8): Zegl-eb-i=vit idgnen (106) q˙vela-n-i kv-is all-pl-nom stone-gen monument-pl-‘nom’=like they.stood ‘They all stood there like monuments.’ (M. Z#avaxisˇvili apud K. v 1: 177) (107) Zura-m da Erek.le-m tagv-eb-i=vit dasunes Zura-erg and Erekle-erg mouse-pl-‘nom’=like they.sniffed.it akauroba hereabouts:nom ‘Zura and Erekle sniffed around here like mice.’ (M. Z#avaxisˇvili apud K. v 1: 178) (108) xel-eb-i prt-eb-i=vit gasˇala hand-pl-nom wing-pl-‘nom’=like he.spread.them ‘He opened his arms like wings.’ (M. Z#avaxisˇvili apud K.v 1: 178) The primarily semantic basis of this agreement, though, is evident from the fact that collectives count as ‘plurals’: ˇsemovidnen (109) glex-eb-i gac. vrtnil lask.ar-i=vit peasant-pl-nom trained army-‘nom’=like they.came.in ‘The peasants came in like a trained army.’ (M. Z#avaxisˇvili apud K.v 1: 178) Furthermore, generics allow both singular and plural forms: ˇcxavis, (110) [cˇxik.v-i] xan k.rux-i=vit xan jay-nom sometimes broody.hen-nom=like it.clucks sometimes Zera-sa=vit c. ivis, k.ata-sa=vit k.navis, an ˙ kite-dat=like it.shrieks cat-dat=like it.mews or ZaVl-eb-i=vit q˙eps dog-pl-‘nom’=like it.barks ‘[A jay] sometimes clucks like a broody hen, sometimes shrieks like a kite, mews like a cat, or barks like a dog [‘like dogs’].’ (T. Razik.asˇvili apud K. vacˇ.antiraZe 1978: 41–5) ˙

6.5 The relationship between depictives and non-agreeing adjuncts 6.5.1 Formal relationship As seen in the preceding section, the adverbs with the adverbial case sufﬁx -ad and the instrumental sufﬁx -it allow participant-oriented readings, and this makes them similar to the condition/state group of depictives discussed in section 6.3.1. And indeed, there are some properties that the condition/state depictives, on the one hand, and expressions of manner and concomitance, on the other, have in common.

228

W. Boeder

6.5.1.1 Interrogability First, both can be asked for by rogor ‘how?’ (see I. Imnaisˇvili 1957: 676): (111)

a. How did the merchants come out of the house?—Amazed. (see (5)): condition/state b. How did the train move?—Lazily. (see (92)): manner c. How did he come?—With his wife. (see (82)): concomitance d. How did you drink tea, with sugar or milk?—With sugar. (cˇai rogor dalie, ˇsakrit tu rZit?—sˇakrit) (see (99)): concomitance but: e. What age/*How did he become an orphan?—As a child. (see (35)): phase f. How many were in their party?/*How did they come?—Three [of them]. (see (39)): quantity

6.5.1.2 Coordination Coordination of depictives of the same category with or without ‘and’ is possible (see (14), (32), and (48)); depictives that belong to different categories are not normally coordinated (see Schroeder 2003 for some discussion). However, Leo K.vacˇ.aZe (1996: 225) in his standard syntax of Modern Georgian notes that what he calls ‘predicative modiﬁers’ (i.e. depictives) and manner adverbials can be coordinated (cf. I. Imnaisˇvili 1957: 676); however, he considers such coordinations ‘stylistically unjustiﬁed’ (stilistik.urad gaumartlebeli). Since coordination presupposes both syntactic ˙ ˙ and semantic identity of the conjuncts on some level, both depictives and oriented manner adverbials can be assigned to essentially the same category of adjuncts. Furthermore, the fact that they can be coordinated is evidence of a semantic overlap between depictives and adverbials on the basis of a metonymic relationship (see further below 6.5.2). Examples: (112) is midioda c. q˙nar-ad he:nom he.walked(iprf) quiet-adv da dapikrebul-i and absorbed.in.his.thoughts-nom ˙ azbegi apud ‘He walked quietly and absorbed in his thoughts.’ (A. Q K. v 1: 225) (113) Murtuza mad-ian-ad da gamgeleb-ul-i Murtuza:nom appetite-poss-adv and become.a.wolf-ptcp-nom scˇamda he.ate(iprf) ‘Murtuza had an appetite like a wolf.’ [‘M. was eating with appetite ˙ azbegi apud I. Imnaisˇvili 1957: 676) and having become a wolf.’] (A. Q

Depictives in Kartvelian

229

(114) tumca c. q˙nar-is oxvr-it magram gaxarebul-i although quiet-obl sigh-ins but gladdened-nom c. avidoda he.used.to.go ‘Although sighing, but still happy, he used to go [to fulﬁl the ˙ azbegi apud I. Imnaisˇvili 1957: 676) orders of his mistress].’ (A. Q (115) siprtxil-it da gacˇumebul-n-i midiodnen caution-ins and silenced-pl-nom they.went(iprf).there ˙ azbegi ‘They walked cautiously [‘with caution’] and silently.’ (A. Q apud K. v 1: 225; cf. (91)) 6.5.1.3 Cases of variation and overlap Condition/state/concomitance depictives and manner expressions are used in similar contexts, with variation occurring within the same language, across related languages, and in diachronic development. Condition/statemanner There is some variation between agreeing depictives and adverbial case forms within Georgian, as illustrated in (116) and (117). Subtle differences in meaning remain to be investigated. xorc-i um-i / um-ad miq˙vars meat-nom raw-nom / raw-adv I.like.it ‘I like meat raw.’ (117) a. ˇsoridan mta tetr-i da lamaz-i from.far mountain:nom white-nom and beautiful-nom ˇcanda (=(38)) it.appeared b. ˇsoridan mta tetr-ad da lamaz-ad from.far mountain:nom white-adv and beautiful-adv ˇcanda it.appeared ‘From afar the mountain appeared(,) white and beautiful.’

(116)

In some instances, Svan has the adverbial case form where Georgian seems to prefer agreeing depictives. Thus, the translation equivalents of the adverbials in the Svan examples in (118) to (120) would carry nominative case in agreement with the intransitive subject in Georgian. Compare (118) with (2) and (12), (119) with (11), and (120) with (13), (30). Svan (118)

ha¨ri-d a¨gi=t’ a¯nqa¨d empty-adv home=to he.came ‘He came home empty-handed.’ (Pr 1: 54, 29)

230 (119)

(120)

W. Boeder m´tma-d asa¨dx ˙ hungry-adv they.remained ‘They remained hungry.’ (Chr 169, 13) ocˇa¨dd amzˇi mugvri-d we.went.away thus sulky-adv ‘Thus we went away sulkily.’ (Pr 1: 39, 9–10)

Negative condition/state or manner With negative expressions of condition, state, or manner, likewise, there is some variation between adverbials (121a) and depictives (121b) in Georgian. Both types of expression are marked with the negative preﬁx u- and there is no clear difference in meaning. (121)

a. Epemia ga-u-xedel-ad ic. va Euphemia:nom pref-neg-remove-adv she.lay(aor)

b. Epemia ga-u-xdel-i ic. va Euphemia:nom pref-neg-remove-nom she.lay ‘Euphemia lay there in her clothes.’ (R. GvetaZe apud K.v 1: 222) ˙ In contrast, their positive counterparts do not seem to allow the adverbial variant, but only the agreeing expression: (122)

login=sˇi a. cˇacmul-i davc. eki dressed-nom I.lay(aor) bed=in b. *cˇacmul-ad davc. eki login=sˇi dressed-adv I.lay(aor) bed=in ‘I lay in bed with my clothes on.’

Negative concomitance or manner Similarly, concomitance expressions have negative counterparts, marked by a circumﬁx u-___-o, which exhibit the same variation. That is, unlike their positive counterpart, they can either have invariable adverbial case-marking (123a) or exhibit agreement. In (123b), the expression of negative concomitance agrees in dative case with its controller: (123)

a. Tek.lia-sa=c u-vaxsˇm-o-d daeZina Teklia-dat=too neg-supper-suff-adv she.went.to.sleep b. Tek.lia-sa=c u-vaxsˇm-o-s daeZina Teklia-dat=too neg-supper-suff-dat she.went.to.sleep ‘Teklia, too, went to sleep without supper.’ (T. Razik.asˇvili apud K.v 1: 222)

Variation also has a diachronic dimension. There are medieval Georgian examples with adverbial case-marking whose modern equivalents require an

Depictives in Kartvelian

231

agreeing depictive: Medieval Georgian (124) a. ˇseve, vnaxe igi turpa mtirl-ad da ˙ ˙ I.went.in I.saw.her that:nom lovely:nom weeping-adv and creml-dasxm-ul-ad tear-pour.over-ptcp-adv ‘I went in; I saw that lovely one weeping and ﬂooded in tears.’ (Rustaveli 1170(1174), 1; trans. M. Wardrop) Modern Georgian b. vnaxe I.saw.her

... ...

mtiral-i da creml-dasxm-ul-i ˙ weeping-nom and tear-pour.over-ptcp-nom

However, this last example possibly does not really belong here since the status of verba sentiendi constructions remains to be investigated (a verbal complement analysis may be more appropriate than a secondary predicate analysis). 6.5.2 Semantic relationships between depictives and non-agreeing adjuncts 6.5.2.1 Condition, state, concomitance, and manner One may speculate about the reason for the phenomena of variation and overlap between depictive and adverbially marked adjuncts illustrated in the preceding section. First, condition/state depictives characterize the predicate insofar as a metonymic relationship exists between the property predicated of a participant and the characteristics of the ‘action’ denoted by the matrix verb; this is the double relationship recognized by traditional Georgian grammarians and by general linguists (Nichols 1978b: ‘double dependency’): depictives ‘characterize the object and the action at the same time’ (Enukasˇvili 1977: 162). In He walked happily, the way of walking can express the interior state of the agent. At least some manner adverbs have the same property. Second, with negative adjectives and participles, the unmarked variant seems to be the manner/concomitance coding with adverbial case-marking. This seems plausible if we think of them as expressions that code a less intimate relationship between some property and a participant. The idea that whenever variation is possible, condition/state depictives code a higher degree of participant relatedness is conﬁrmed by the observation of my consultant (Rezo K.ik.naZe) that the participant-related (b) variants in (121) are more ‘ﬁgurative’ (xatovani). Negative properties like not being dressed or ˙ having had no supper are less typical qualities attributable to some referent, because they rely on reﬂection rather than on direct observation. Third, in some instances attribution of properties to a participant in the form of condition/state depictives is ‘ﬁgurative’ in the sense that properties of

232

W. Boeder

Properties of participants coded by participant-oriented agreement of condition/state depictives Metonymy with a totalizing effect (evidenced by coordination)

Synecdoche with a specifying effect (evidenced by variation)

Properties of events coded by concomitance and manner adverbials

Figure 6.1 Relationship between classes of depictive and adverbial expressions

the event are conferred to the participant involved. This might be called a synecdochic transfer, which, in the formal framework of classical rhetorics, can be subsumed unter the scheme of hypallage, as in (125), where ‘mischievously’ primarily refers to the subject, and only by metonymy to the event. (125) tovl-is c. q˙l-eb-ma daic. q˙es celk-ad dena snow-gen water-pl-erg they.began mischievous-adv ﬂowing:nom mindrvr-ad da xev-eb=sˇi ﬁeld-adv and gorge-pl=in ‘The snow-water began to ﬂow mischievously down the ﬁeld and into the gorges.’ (Vazˇa Psˇavela apud K. v 1: 181) In poetry, this type of transfer can even give rise to more extreme examples as in:9 Latin (126) ibant obscuri sola sub nocte they.walked dark:m.pl.nom alone:f.abl.sg under night:f.abl.sg per umbram through shadow:f.sg.acc ‘Dark [nominative plural!] they walked through the shadow under the lonely night.’ (Vergil, Æneid vi. 268), instead of: ‘Lonely they walked under the dark night.’ We have, then, two metonymic extensions across the categories of depictives: a transfer from participant-related qualities to events, and a transfer from concomitance and manner properties to participants (Figure 6.1). Following the typology of metaphoric relationships developed by David Sapir (1977), we may say the following: the coordination of condition/state depictives with concomitance and manner adjuncts is an extension from the participant to the event with which it is connected by metonymy, and 9 This is Servius’ commentary (fourth century ad): aut hypallage est: sub obscura nocte soli ibant, aut, sub sola nocte, id est, ubi nihil alius est praeter noctem ‘It is either a hypallage: under the dark night they walked alone, or: under the night alone, that is, where there was nothing but night.’ See Gerber (1871: 570–3) for more examples and some discussion.

Depictives in Kartvelian

233

the quality of the participant is ‘totalized’. The variation between negative concomitance and manner depictives and condition/state depictives is an extension from circumstantial event properties to the participants with which they are connected by synecdoche, a ‘reduction to a pertinent part’, namely the participant on which or with which the negative property does not occur. 6.5.2.2 Similitive constructions In Georgian grammatical tradition, similitives are considered adverbial expressions of manner (K. vacˇ. aZe 1996: 176–9), which are correspondingly asked for by rogor ‘how?’. Alternatively, however, ra-sa=vit ‘what-dat=like’ (‘like what?’) is sometimes used. V. Imnaisˇvili (1997: 88) feels that this last form is less appropriate for clauses where the similitive construction ‘is connected with the verb’. So ‘how?’ seems to be connected with an adverbial interpretation, whereas rasavit seems to favour a depictive interpretation. Be this as it may, the distinction does not seem to be clear-cut, and we may ask again for the reason for this partial overlap that manifests itself—however weakly—in a semantic number agreement. My impression is that this is a case of ‘interplay of external and internal metaphor’ (Sapir 1977: 25–8). Take for example ‘He opened his arms like wings’ (example (108)). The same comparison may be read as an internal metaphor (or metaphor proper), where the hands are like wings (an interpretation which is underlined by number agreement); and as an external metaphor (or analogy according to the Aristotelian theory of metaphor), where the hands are to the person and the action of opening his or her hands like the wings are to a bird and the action of opening its wings. Although an analogy where A is to A’s domain as X is to X’s domain does not presuppose a similarity between A and X, but between A and its domain on the one hand and X and its domain on the other (Sapir 1977: 23), A and X can be thought of as being similar on the basis of the analogy. In our example, the hands become similar to wings on the basis of their function in the act of opening them (in addition to some conceivable similarity between wing and hand that we might be inclined to detect in the ﬁrst place!). The indeterminateness of orientation in similative constructions could thus be the result of a transition from a primary ‘analogy’ between the compared propositions (‘He opens his arms’ ‘A bird opens its wings’) to a metaphorical interpretation of the NP governed by ‘like’.

6.6 Summary and conclusions Looking back at the different forms of Georgian depictives and related participant-oriented constructions, we get a picture as in Table 6.2. The

234

W. Boeder

Table 6.2 Georgian (and Svan) participant-oriented expressions Coding 1. 2. 3. 4.

Condition/state Phase Quantity Concomitance

5. Comparison

6. Manner

7. Location in Tush

Possible controllers

Question word how?

agreement

core argument

(a) periphrastic -tan ertad ‘together with’ 9 (b) possessive sufﬁx > > (-ian) þ adverbial = case (-ad) > (c) future part. þ > ; adv. (-ad) (a) periphrastic: rogorc ‘like’ (b) postposition =vit ‘like’, possible (semantic) number agreement (a) adjective þ adverbial case (-ad) (b) verbal noun þ instrumental case (-it) agreement

no

9 > > > > > > > > > > > intr. subject/tr. > > > direct object > > > > > > > = no

how?

> > > > > > > > > > > > > > > > intr. subject/tr. > > > direct object > > ;

no

subject

subjects only (?)

semantic categories 1–7 have the properties of an implicational hierarchy (Himmelmann and Schultze-Berndt 2004). Higher up in the hierarchy, in the ﬁrst three categories, we ﬁnd agreement as an unambiguous device of participant orientation. These three categories also share the feature of allowing a superordinate clause paraphrase (‘We went home happy’‘We were happy when we went home’). If we neglect comparison, which is not coded as depictive in Georgian, categories 4 and 6 share case-marking, mainly adverbial case-marking, which can be semantically participant-oriented; but since this is not overtly coded, the interpretation depends on contextual information, as well as on the transitivity of the verb and the lexical-semantic category of the adverbial form itself: some forms with adverbial case-marking are process-oriented, others are participant-oriented. This is, in a sense, a weakening of restricted orientation coding. Some of the categories that are lower in the hierarchy have alternative periphrastic expressions with postpositions, such as =vit ‘like’ in category 5,

Depictives in Kartvelian

235

and independent words, as in -tan ertad ‘together with’ in category 4. These expressions have no restricted orientation at all. We have, then, a scale of restricted orientation coding: agreement is the best, independent coding of participant orientation; coding by case depends on the context; and marking by postpositions and independent words involves no restrictions on participant orientation at all. This also explains the gap in category 5: comparison cannot be coded by agreement or case in Georgian, therefore it falls out of the continuum of the hierarchy. From this point of view, category 7, the place and time adverbials, should have case coding or postpositions, and indeed, it normally has in literary Georgian. The agreement occurring with Tush Georgian (locative) ablatives is exceptional in the lowest category of the hierarchy. Still, we normally expect that an implicational scale works on contiguous segments, and the ‘gap’ between categories that show participant orientation by agreement, and those that do not, disturbs this notion of contiguity. However, the scale should also be seen in the light of our discussion of morphological availability (6.3.1) and lexical blocking (6.4.1). Georgian and Svan have no dedicated morphological means for depictives (of the type described by Bucheli Berger, Ch. 4, this volume). Case and number agreement exist independently of depictives in adjectives, participles, etc. that have a modiﬁer or predicative complement function, but for the categories without participant orientation there are simply no such adjectival formations available. Similarly, lexical blocking by non-oriented lexemes or idioms and the like is not ‘grammatical’. It is only those categories that offer the same resources as the categories higher up in the scale that count. In this sense, lack of agreement by itself does not falsify the implicational hierarchy. Finally, we may ask whether the degree to which depictive (agreeing) constructions in a language cover the implicational hierarchy correlates with something else in that language. The following speculation may be relevant here. Agreement or any kind of cross-reference is a prerequisite of an unambiguous orientation towards a controller. But mere existence of agreement in a language does not imply its use in participant orientation: languages like German have morphological agreement in the noun phrase, but not in depictives. Rather, unambiguous coding of participant orientation correlates with a general tendency in some languages to avoid any constituent that is not formally related to a nominal or verbal head, and to prefer orientation towards nominal heads (Boeder and Schroeder 1998). The semantic map of depictives proposed by Himmelmann and Schultze-Berndt (Ch. 1, this volume), which predicts the highest incidence of formal depictives with ‘condition/state’ and their lowest incidence with ‘location in time

236

W. Boeder

and space’, could be related to the degree in which those languages allow discontinuity—‘structural distance’—between constituents that belong together. In other words, the extension of the depictive coding strategy into the less usual areas of the domain possibly correlates with ‘freedom of word order’.

7 On depictive secondary predicates in Laz SILVIA KUTSCHER AND N. SEVIM GENC¸

7.1 Introduction This chapter deals with morphosyntactic, semantic, and prosodic characteristics of depictive secondary predication in Laz. We show that Laz adjunct expressions generally cannot be divided into depictive and adverbial constructions on the basis of their morphosyntactic properties. We also deal with some prosodic characteristics of adjuncts expressing manner and state, and discuss to what extent depictive expressions may be delimited from manner adverbials on the grounds of intonational patterns. Concerning the semantic interpretation of adjuncts in Laz, we argue that they are vague with respect to participant- or event-oriented readings. Laz clause-level adjuncts thus are general adjunct constructions in the terminology proposed in the introduction to this volume. An interesting exception to this generalization are distributive numeral expressions, which can be argued to be a genuine depictive construction. A sister language of Georgian, spoken on the south-eastern coast of the Black Sea, Laz is the only member of the South Caucasian family which is spoken primarily outside Georgia. The vast majority of its speakers live in Turkey and are bilingual. An increasing number of young Laz, however, are ﬂuent only in Turkish. In addition, native speakers of Laz restrict using their mother tongue to private communications amongst friends and family members. The variety of Laz discussed here is the one spoken in the city of Ardes¸en and the villages of the Ardes¸en region. Although this dialect (Ardes¸en-Laz) differs from other Laz varieties with respect to the case-marking system (see Kutscher 2001: ch. 5), it is similar to the other dialects in that no variant of Laz exhibits The authors would like to thank Eva Schultze-Berndt, Nikolaus Himmelmann, and Katrin Lehmann for numerous comments on previous versions of this chapter, and Britt Temme for brushing up our English. We also are grateful to the patience and helpfulness of the Laz speakers in the region of Ardes¸en and in Germany, who have supported our work for several years now.

238

S. Kutscher and N. Sevim Genc¸

NP internal agreement or an adverbialis case, in contrast to the Georgian varieties discussed by Boeder (Ch. 6, this volume). The data presented in this chapter consist of utterances taken from a corpus of spoken texts recorded on location in Turkey. Some texts from this corpus have been published in Wodarg (1995) and Kutscher and Genc¸ (1998). Examples from these publications are marked with W (for Wodarg) and K/G (for Kutscher and Genc¸) followed by an abbreviated title of the source text and the reference number of the intonation unit (e.g. K/G murunZxi 003). Other examples are from our research on positional verbs elicited with stimuli of the Language and Cognition Group of the Max Planck Institute for Psycholinguistics in Nijmegen. These are marked ‘Posit’. Examples not marked for their source have been elicited for the purposes of this article. The structure of the chapter is as follows. Section 7.2 gives a short overview of Laz basic clause structure. Section 7.3 focuses on how NP-internal modiﬁers can be delimited from adjuncts. Section 7.4 concerns participantoriented and event-oriented manner expressions, and gives an overview on their segmental and distributional characteristics (7.4.1) followed by a few remarks on prosodic communalities and differences (7.4.2). Section 5 deals with participant-oriented and event-oriented uses of adjuncts in instrumental case (7.5.1), motative case (7.5.2), and locational nominals (7.5.3). Section 7.6 argues that distributive numerals are expressed by a genuine depictive construction in Laz. Section 7.7 deals with expressions of role and life stage, which in Laz are biclausal in nature. Section 7.8 concludes.

7.2 Morphosyntactic essentials Laz is basically an SOV language, exhibiting the categories case and number in nominal expressions and a rich inventory of verbal categories with up to ten different morphological slots to be ﬁlled in the predicate (see Kutscher 2001: ch. 1). Predicates in Laz are head-marking—i.e. depending on the valence of the verb, verbal inﬂection is mono- or polypersonal. With polypersonal verbs the ﬁnite verb inﬂects for both actor and undergoer as in (1). (1) cemc¸am1 hit:[2>1]sg.prs ‘You beat me.’ 1 Examples are written in the Lazog˘lu/Feurstein orthography introduced to the Laz community in Turkey in 1984. It deviates from the Caucasianists’ transcription in the following graphemes (): , , , , <s¸¼sˇ>, , , . ˙

On depictive secondary predicates in Laz

239

Laz is an active language (Klimov 1974), i.e. monopersonal verbs subdivide into two classes, depending on whether the verb takes a controlling or non-controlling single core argument. Controlling single core arguments are marked as actor on the predicate, cf. the ﬁrst person marker b- in (2a). Non-controlling single core arguments are marked as undergoer, cf. the ﬁrst person marker m- in (2b). This is indicated by an arrow (>) in the gloss. (2) a. bulur go:1sg.prs ‘I go.’ b. mac¸inden sneeze:>1sg.prs ‘I sneeze.’ Note that information on person and number in Laz predicates is not marked by a single preﬁx but rather results from the interaction of preﬁxes and sufﬁxes. These are portmanteau forms coding tense/aspect/mood simultaneously (see Mattissen 1995).2 Argument NPs are always non-obligatory and often are omitted in discourse. If present, all argument NPs in Ardes¸en-Laz are unmarked for case, as opposed to other Laz dialects. This holds for the actors of polypersonal predicates (3a, b), primary and secondary objects (3b), and for the single core argument of monopersonal inactive predicates (3c). (3) a. baba pencere kosuy father window clean:[3>3]sg.prs ‘Father wipes the window.’ b. nana baba bere mec¸ay mother father child give:[3>3]sg.prs ‘Mother gives the child to Father.’ c. bic¸ˇi as¸ kˇurinen boy be_afraid:>3sg.prs ‘The boy is afraid.’ 2 Person-marking in the glosses includes information on the macro-roles actor and undergoer. Undergoers are marked with a preceding angled bracket, >. In polypersonal predicates this bracket is to be read as ‘acts on’. In monopersonal–inactive predicates, it marks the single core argument as a non-controlling argument. The single core argument of monopersonal active predicates is an actor and has no special marking.

240

S. Kutscher and N. Sevim Genc¸

Adjuncts, e.g. goals (4a) or instruments (4b) take oblique case-marking. (4) a. Vecekˇules¸ a es¸ kˇaftaten ( . . . ) ˇ Vecekule-s¸a es¸kˇa-ftaten ‘place name’-mot up-go:1pl.fut.pfv ‘We went up to Vecekˇule.’ (K/G gecekule 005) goc¸xu; s¸ kˇomu \3 b. Z#arite Z#ari-te goc¸xu s¸kˇomu water-ins clean:3sg.pst.pfv eat:3sg.pst.pfv ‘He cleaned it with water and ate it.’ (K/G proVoni 031) Expressions of location (5a) or time (5b), however, are always unmarked for case. (5) a. Trabzoni ya (0.3) xolo ˇteyyare kocepxedi Trabzoni ya xolo tˇeyyare ko-ce-pxedi Trabzon quot again plane emph-down-sit:1sg.pst.pfv ‘He said: ‘‘In Trabzon I sat down in an airplane once again’’.’ (K/G Ferat dayi 1 002/3) b. a mapxa ndVa; (0.1) xvala a mapxa ndVa xvala one sunny day alone kes¸ kˇebuc¸vi \ k-es¸kˇebuc¸vi emph-up: lead:[1>3]sg.pst.pfv ‘One sunny day I brought the cows

pucepe puci-pe cow-pl

up alone.’

(W Kuhtext 004)

Having set out the major characteristics of Laz clause structure, in the following sections we will turn to the nature of participant-oriented adjunct constructions beginning with delimiting unmarked adjuncts from NP-internal modiﬁers.

7.3 Participant-oriented adjuncts vs. NP-internal modiﬁers In contrast to Georgian (see Boeder, Ch. 6, this volume), Laz NPs only inﬂect on the last element of the phrase and do not exhibit NP-internal agreement, 3 In transcripts of audio data intonational units are delimited as follows: \ ¼ ﬁnal intonation, i.e. pitch drops to the base line; ; ¼ medial intonation, i.e. pitch ends in mid-range either level or slightly falling; / ¼ progredient intonation, i.e. ﬁnal pitch is clearly rising. A number in brackets, e.g. (0.1), indicates the length of a pause in seconds.

On depictive secondary predicates in Laz

241

as (6) illustrates (see Kutscher 2001: ch. 2 for details). (6) (. . .) pasiari pasiari rusty ‘I sew it

lems¸ ite kelebuc¸ˇi / lems¸i-te ko-elebucˇ¸ i needle-ins emph-sew:[1>3]sg.pst.pfv together with a rusty needle.’ (K/G korme 011)

In (6) the instrument-NP pasiari lems¸ ite ‘with a rusty needle’ is case marked only on the semantic nucleus (lems¸ i ‘needle’) of the NP, whereas the attribute pasiari ‘rusty’ is a bare adjective stem. The last lexical element of an NP can either be the semantic nucleus (as in (6)) or a possessive pronoun (cf. (7)). (7) pasiari lems¸ i s¸ kˇimi-te kelebuc¸ˇi pasiari lems¸i s¸kˇimi-te ko-elebucˇ¸ i rusty needle poss.1sg-ins emph-sew:[1>3]sg.pst.pfv ‘I sew together it with my rusty needle.’ Hence, the right border of an NP can clearly be identiﬁed by inserting a possessive pronoun into the phrase. NP-internal modiﬁers can thus be formally distinguished from participant-oriented adjuncts following the NP. Compare (8a) with (8b). s¸ kˇimi bimxor (8) a. ˇtuZ#a cari hot bread/meal poss.1sg eat:1sg.prs ‘I eat my hot meal.’ ˇtuZ#a bimxor b. cari ¸skˇimi bread/meal poss.1sg hot eat:1sg.prs ‘I eat my meal hot.’ In (8a) the adjective ˇtuZ#a ‘hot’ is part of the NP and functions as an NP-internal modiﬁer to its head noun. The position of s¸kˇimi in (8b) shows that the modiﬁer ˇtuZ#a ‘hot’ is outside the NP. It serves as a clause-level adjunct with object-oriented reading. Although the semantic nucleus of the NP functions as the controller of this participant-oriented adjunct, ˇtuZ#a ‘hot’ exhibits no morphological means to show this semantic relation overtly. In this respect as well, Laz contrasts with Georgian. While participant-oriented adjuncts following their controller NP can be distinguished from NP-internal modiﬁers, the possessive pronoun insertion test obviously does not work for adjuncts that are placed to the left of their controller-NP. Hence, ˇtuZ#a ‘hot’ in (8a) may either be read as an NP-internal modiﬁer or as a participant-oriented adjunct of its own. In actual utterances,4 4 Note that Laz is mainly a spoken language. There are no widespread mass media in the language and only a few printed books and journals. Most Laz do not read and write in their language. There is no standard variety.

242

S. Kutscher and N. Sevim Genc¸

however, a participant-oriented adjunct construction differs from one with NP-internal modiﬁcation with respect to prosody. A participant-oriented adjunct may form an intonation unit of its own (see section 7.4.2), while this does not hold for NP-internal modiﬁers.

7.4 Depictive and manner expressions This section is concerned with participant-oriented expressions of state (i.e. the kind of expression widely used to exemplify depictives) and eventoriented manner expressions. We use the term depictive in this section as a convenient shorthand for ‘participant-oriented expression of a state’, i.e. for a purely semantically deﬁned expression type. As we will see, there is little evidence to support the distinction of two different (formal) constructions for these two kinds of expression. 7.4.1 Similarities in distribution In Laz, there are no segmental means to distinguish manner adverbials and depictives. For manner expressions, compare the adjectives vrosi ‘good, well’ and evedi in (9), for depictive compare the participle okˇokˇotˇeri in (10). (9) a. vrosi gamaZ#edi; gazirasen ya kˇoc¸epe \ vrosi gamaZ#edi gazirasen ya kˇoc¸i-pe good look_through:2sg.pst.pfv see:>2sg.fut.pfv quot man-pl ‘ ‘‘Look carefully. You’ll see (something),’’ the men said.’ (K/G murunZ#xi 032) b. evedi evedi quick ‘Come (10)

komoxti \ ko-moxti emph-come:2sg.imp here quickly!’ (W Bienentext 034) masa-s¸ i cindo;(1.9) okˇokˇotˇeri eo-zun \ table-gen surface folded_in_half on-lie:3sg.prs ‘It is on the table, folded in half.’ (Posit)

The position of an adjunct constituent of any kind is more or less free, i.e. it depends on discourse pragmatic rules. The preverbal position is the preferred position for both depictive and manner expressions with respect to text frequency and in terms of the order usually given in elicited utterances. Note that preverbal position is the preferred slot for focused constituents. For example, interrogative pronouns are obligatorily placed there and cannot be intonationally detached from the predicate: compare (11).

On depictive secondary predicates in Laz (11)

Voma yesterday ‘What did *Voma

243

cuma skˇani noVa mu iindru? brother poss.2sg market what buy:3sg.pst.pfv your brother buy on the market yesterday?’ cuma skˇani mu noVa iindru?

In this position, participant-oriented and manner expressions as well as other kinds of constituents tend to have a falling pitch accent with the fall continuing on the following predicate (cf. ﬁgures 7.1 and 7.2). Topics are placed utterance-initially. Depictive expressions may precede or follow their controller, as the examples in (11) illustrate. In (12a), the controller ˇtoc¸ˇi ‘rope’ precedes the depictive adjunct kˇirkˇoleri ‘wound up’, while in (12b) the controller ˇtoc¸ˇi ‘rope’ follows the adjunct kˇotˇeri ‘folded’. Note that the depictive expressions in both utterances are prosodically separated from the adjacent constituents.5 (12) a. ˇtocˇ¸i ; kˇirkˇoleri ; (0.1) buc¸ˇis¸ i cindo goozun \ ˇtocˇ¸ i kˇirkˇol-eri bucˇ¸ i-s¸i cindo goo-zun rope wind-ptcp stump-gen surface on-lie:3sg.prs ‘The rope is on the tree stump in a wound fashion.’ (Posit) ˇtikˇina / (0.5) kˇotˇeri / (1.1) ˇtoc¸ˇi goozun \ b. bos¸ i ˇtikˇina ˇtocˇ¸ i goo-zun bos¸i kˇotˇeri empty basket fold-ptcp rope on-lie:3sg.prs ‘On an empty basket there is a rope in a folded state.’ (Posit) Furthermore, depictive expressions may appear in clause-initial position; cf. kˇuc¸xe ¸ˇcunc¸ˇu ‘barefoot’ in (13). (13)

kˇuc¸xe ¸ˇcunc¸ˇu kˇuc¸xe cˇ¸ uncˇ¸ u foot-naked ‘She sent her

Z#ari moVapay Z#ari moV-ap-ay water bring-caus-[3>3]sg.prs barefoot for water.’ (W Fadume 009)

Depictive expressions may also be placed after the predicate as is the case with the depictive kˇuru ‘plain’ in (14). (14)

nezi bimxor / (0.1) kˇuru / nezi bimxor kˇuru walnut eat:1sg.prs plain ‘I eat the walnuts plain [i.e. without any other food].’ (K/G nezi 036)

5 At this point, it remains to be investigated what kinds of semantic effect correlate with these alternative positions and prosodic breaking.

244

S. Kutscher and N. Sevim Genc¸

These variations in word order are also found with manner expressions. They too may follow the predicate, as shown for the manner adverbial evedi ‘quickly’ in (15). (15) Kedir Kedir Kedir ‘Kedir,

Kedir Kedir Kedir Kedir,

a; (0.1) a one come on.

a komoxti \ evedi; a ko-moxti evedi one emph-come:2sg.pst.pfv quick Quickly!’ (W Bienentext 032)

In post-predicate position, both depictive and manner expressions exhibit the characteristics of afterthoughts, i.e. they constitute a separate intonation unit and follow a prosodic unit with ﬁnal intonation. Manner adverbials may also be non-adjacent to the predicate, as is the case in (16), where the interrogative pronoun mi ‘who’ directly precedes the predicate. (16) bitˇumi vrosi mi ompulu? all well who hide:[3>3]sg.pst.pfv ‘Who has hidden everything well?’ In conclusion, we can state that in Laz depictive (participant-oriented) and manner (event-oriented) expressions cannot be distinguished by segmental or distributional means. They would therefore appear to be formally expressed by a single construction type which allows both participantoriented and event-oriented readings, thus exemplifying a general adjunct construction as deﬁned in the introduction to this volume. However, there may be prosodic differences between the two expression types, as further discussed in the next section. 7.4.2 Prosodic differences? In this section, we provide preliminary evidence for the observation that depictive and manner expressions differ with regard to at least one aspect of prosodic marking. In preverbal position, depictives may be intonationally detached from the main predicate of the clause, i.e. they may form an intonation unit of their own. In contrast, manner adverbials cannot be intonationally separated from the predicate in this position. Note that these are very tentative observations, since Laz prosody is still very poorly understood. In Laz, elements in preverbal position show a strong tendency to form an intonational unit with the following predicate. This holds for adverbials as well as depictive expressions, as the following examples demonstrate. Figure 7.1

On depictive secondary predicates in Laz

245

Figure 7.1 Wave form and pitch extraction for example (9b)

shows the waveform and pitch extraction for the manner adverbial in (9b) above. The vertical cursor line crossing the graphs indicates the word boundary between the adverb evedi ‘quickly’ and the predicate komoxti ‘come’. Figure 7.1 clearly shows rising Fø on the ﬁrst syllable of the manner adverb evedi followed by a fall on the second syllable. The fall continues smoothly into the ﬁrst syllable of the predicate komoxti. A continuous contour like the one in Figure 7.1 is found on most of the preverbal depictive expressions in our corpus, as well. Figure 7.2 illustrates the Fø contour of the depictive expression okˇokˇotˇeri ‘folded’ given in example (10) above. The relevant section of the contour is marked with a vertical line in the ﬁgure. Once again the falling contour on the depictive participle okˇokˇotˇeri ‘folded’ is continued on the following predicate eozun ‘it lies on sth.’. In contrast to adverbial manner expressions, with depictive expressions we also ﬁnd examples in our corpus where the depictive expression is intonationally detached from the following predicate by a clear break in the

246

S. Kutscher and N. Sevim Genc¸

Figure 7.2 Wave form and pitch extraction for example (10)

Fø contour, as is illustrated in (17) and ﬁgure 7.3.6 (17) gzalineri / (0.6) mendaxtey \ by_foot go_to:3pl.pst.pfv ‘They went there by foot.’ (K/G askerepe 111) As Figure 7.3 illustrates, the Fø contour in this case is quite the opposite of those given in Figures 7.1 and 7.2. In ﬁgure 7.3 the Fø clearly rises on the last syllable of the depictive gzalineri ‘by foot’ (in all likelihood realizing a highrising boundary tone). The onset pitch of the following predicate, mendaxtey, is clearly much lower and in the mid range typical for the onset of a new intonation unit. As for manner adverbials, we did not ﬁnd an Fø contour in our corpus comparable to the one illustrated in Figure 7.3 for depictive expressions. Therefore, we tested this intonation pattern by elicitation. The test utterance 6

The semantic effects of the prosodic break need further investigation.

On depictive secondary predicates in Laz

247

Figure 7.3 Wave form and pitch extraction for example (17)

(18) was taken from a narrative in our corpus. In the original utterance, the manner adverbial vrosi ‘well’ and the predicate dvompuli ‘hide it’ are part of a single continuous pitch movement (see Figure 7.4). (18) bitˇumi oZ xone bitˇumi oZxone all attic ‘Hide everything

vrosi vrosi good well in

dvompuli / dv-ompuli emph-hide:[2<3]sg.imp the attic.’ (K/G askerepe 104)

For our test, the second author of this chapter, a native speaker of Laz, produced the utterance illustrated in (19), i.e. a high-rising, unit-ﬁnal pitch on the manner adverbial vrosi followed by a break and a mid range onset on the predicate dvompuli. (19) bitˇumi oZxone vrosi / dvompuli all attic well emph-hide:[2>3]sg.imp ‘Hide everything well in the attic.’

248

S. Kutscher and N. Sevim Genc¸

Figure 7.4 Wave form and pitch extraction for example (18)

This test utterance was played to two other native speakers. Both speakers clearly rejected it on the grounds that it was ‘not sounding right’. Both of them only accepted intonation patterns close to the one found in (18), i.e. with no intonational break. To conclude, manner adverbials and depictive expressions appear to be very similar with respect to (the lack of overt) marking and distribution. But depictive expressions may constitute intonation units of their own (typically with progredient ﬁnal intonation) when in pre-predicate position. In contrast, sequences of manner adverbial plus predicate strongly disfavour prosodic breaking. Preliminary analyses of other adverbial expressions support the impression that the restrictions concerning possible intonation patterns demonstrated for manner adverbials in this section also hold for other adverbials. But the intonational properties of adverbial and depictive expressions deﬁnitely need further investigation. For the time being, we consider them to instantiate a single construction type, i.e. a general adjunct construction.

On depictive secondary predicates in Laz

249

7.5 Other types of adjunct that allow participant-oriented readings This section provides data on other types of expressions which share semantic characteristics with depictive expressions in Laz. Section 7.5.1 investigates the semantic range of the instrumental case, section 7.5.2 discusses uses of expressions marked with motative case, and section 7.5.3 deals with locational nouns. For all three kinds of expressions, we will argue that no clear boundary can be drawn between participant-oriented and eventoriented constructions. 7.5.1 Instrumentals The sufﬁx -te mainly marks NPs referring to the instrument of an event. Its use is restricted to non-human referents (see Kutscher 2001: ch. 5 for details). In addition to encoding an instrument, the sufﬁx covers cause (20, 21), purpose (22), and accompaniment (23) readings. In the following, we discuss these uses in more detail. An instrumental case may mark adverbials of cause, as is the case with the verbal noun obiru ‘play’ in (20) which gives the reason why the speaker returned home late. (20)

leba domau \ himndVa / (0.1) obirute him-ndVa obiru-te leba do-mau dem-day play:vn-ins late emph-become:>1sg.pst.pfv ‘On this day I was late because of playing.’ (W Kuhtext 006)

Instrumentals of cause are not restricted to event-oriented uses. Compare the noun s¸ kˇurina ‘fear’ in instrumental case in (21). The example is from a narrative about a mother and a daughter harvesting tea up in the mountains away from their village. On their way home they reach a river and as dusk comes, the mother begins her evening prayer. While the mother is praying, the daughter hears some strange noises and thinks the river ghosts are coming to get her. The mother, knowing that her daughter is a timid person and would prefer to go home and not pray at the bank of the river, thinks that her daughter is hallucinating and says: (21) (. . .) s¸ kˇurinate ognam s¸kˇurina-te ognam fear-ins hear:[2>3]sg.prs ‘You only hear it because gecekulesa 028)

ya ya quot you are frightened.’

(K/G

250

S. Kutscher and N. Sevim Genc¸

In this context, the nominal s¸ kˇurinate denotes a mental state of the actor of the predicate ognam ‘you hear it’, i.e. the daughter, and hence functions as a participant-oriented expression. The same holds for the instrumental phrase parate ‘with money’ in (22) which denotes the purpose of the activity of the actor participant of the main predicate pZ#ilum ‘I pluck’: (22) ¸cayi ; (0.1) parate pZ#ilum c¸ayi para-te pZ#ilum tea money-ins pluck:1sg.prs ‘I am harvesting tea for money [in order to get money].’ The instrumental in (22) also relates to the mental state of the actor or rather, it encodes that the actor is accompanied by a certain mental state, namely the thought of money. Hence, more literally one may translate (22) as ‘I pluck tea with money on my mind’. An instance of the participant-oriented use of the accompaniment meaning of the instrumental is the instrumental phrase kˇakˇalite ‘with a stone’ in (23). (23) mbuli kˇakˇalite ces¸ kˇidu mbuli kˇakˇalite ces¸kˇidu cherry stone-ins swallow:3sg.pst.pfv ‘S/he swallowed the cherry with its stone.’ As can be shown by the insertion of a possessive pronoun in (24) (see section 7.2 above), kˇakˇali ‘stone’ in (23) is not internal to the object NP mbuli. It forms a constituent of its own and functions as a participant-oriented adjunct. (24) mbuli s¸ kˇimi kˇakˇalite ces¸ kˇidu ! *mbuli kˇakˇalite s¸ kˇimi ces¸ kˇidu ˇ mbuli s¸kimi kˇakˇali-te ces¸kˇidu cherry poss.1sg pip-ins swallow:3sg.pst.pfv ‘S/he swallowed my cherry with its stone.’ In some cases the instrumental expression is semantically vague as to a participant- or event-oriented reading, as seen in (25). kˇibrite cari imxoy (25) Z#uneri ˇ Z#un-eri kibri-te cari imxoy hurt-ptcp teeth-ins meal/bread eat:3sg.prs ‘S/he eats the meal with hurting teeth.’ In (25), the hurting teeth of the actor are the instruments of the eating process, i.e. on the one hand the expression is event-oriented. Simultaneously, the instrumental phrase in (25) refers to the physical state of the eater and therefore is participant-oriented.

On depictive secondary predicates in Laz

251

To conclude, the instrumental allows event-oriented as well as participantoriented uses. Moreover, in some instances the orientation of the instrumental is in fact vague. Hence, Laz instrumentals instantiate what Himmelmann and Schultze-Berndt (Ch. 1, this volume) call a general adjunct construction. In the following section, we will show that comparable ﬁndings hold for motative phrases. 7.5.2 Motatives The term motative originates in the grammar of Pazar-Laz written by Rosen (1844) and captures the particular semantics of this case, which only encodes that the referent of a nominal thus marked has moved, but it is vague with respect to whether the referent is moving towards a goal (as in (26a); see also (4a) above) or moves away from a source, as in (26b). (26) a. bere oxori-s¸ a amulun bere oxori-s¸a amo-ulun child house-mot into-go:3sg.prs ‘The child goes inside the house.’ b. bere oxoris¸ a bere oxori-s¸a child house-mot ‘The child goes out

gamulun gamo-ulun out-go:3sg.prs of the house.’

The direction of movement is usually speciﬁed by a spatial preﬁx to the predicate, e.g amo- ‘into’ in (26a) or gamo- ‘out’ in (26b). Motative phrases most frequently are event-oriented, as in the preceding examples. In some cases, however, we have found motative-marked phrases which clearly have participant-oriented semantics. In these cases, the predicate denotes an event of emission and the motative relates to the source of this event; compare (27) for illustration. (27)

pa rmakluVepes¸ a pa‹ rmakluVi-pe-s¸a railings-pl-mot ‘I called her from

buoxam \ naana buoxam nana call:[1>3]sg.prs mother the railings: ‘‘Mother.’’ ’ (K/G nezi 044)

In (27), the narrator of the story is accidentally caught in a storehouse and tries to call her mother so that she may be released. The actor of the predicate buoxam ‘I call her’ is behind the railings of the storehouse, while the undergoer, the person called, is outside the storehouse, coming down the street. Interestingly, the motative phrase in these participant-oriented uses expresses the

252

S. Kutscher and N. Sevim Genc¸

location of the controller, not its movement towards a goal or from a source. The calling, however, is directed away from the controller in (27), which makes the controller the source of the event expressed in the predicate. Hence, the motative phrase at the same time is also event-oriented. Compare (28), where the caller comes to the storehouse and calls a person standing behind the storehouse’s railings. In this case the motative-marked phrase is purely eventoriented, denoting the direction of the calling but not the location of the caller. (28) serende-s¸ a moxtu do storehouse-mot come:3sg.pst.pfv and pa rmakluVepe-s¸ a uoxu railings-mot call[3>3]sg.pst.pfv ‘S/he came to the storehouse and called to her towards the railings.’ To conclude, Laz has a general adjunct construction of location—the motative-marked phrase. These motative phrases may be used to express the location of a participant and at the same time are event-oriented in that they indicate the direction to which the event is oriented. 7.5.3 Locational nominals In addition to the general adjunct construction of location dealt with in the preceding section, Laz also has expressions in which the location of a participant is expressed by an unmarked locational nominal, e.g. cindo ‘top’ in (29). (29) cindo keizdaman / cindo ko-e-izdaman top emph-up-pull:3pl.prs ‘They [the boys] pull it (the food) up [from a shelf in the kitchen] being on the top [i.e. in the store room above the kitchen].’ (K/G lu 062) The landmark expression cindo ‘top’ in (29) denotes the location of only one of the participants in the event, namely the boys. Conversely, the preverb e- ‘up’ denotes the direction of the food which is being pulled up, i.e. the food is on its way up to the boys. With respect to participant orientation, Laz locational nominals, which consist of a locative adverb plus the sufﬁx -ndo, differ from the locative adverbs from which they are derived.7 The latter are compatible with 7 Landmark expressions with the sufﬁx -ndo like cindo ‘top’ are nominals. They can appear in core argument function and may be modiﬁed by a genitive NP (cf. (15a)).

(i)

cindo mus¸ i buzi diu cindo mus¸i buz do-iu top poss.3sg ice mod-become:3sg.pst.prf ‘Its surface became ice.’ (K/G ¸cxomepe 006)

On depictive secondary predicates in Laz

253

spatial preverbs of the same semantic domain. Thus, the locative adverb cin ‘up’ in (30) speciﬁes or emphasizes the spatial semantics of the preverb e- ‘up’. (30) cin e-izdaman up up-pull:3pl.prs ‘They pull it up [above].’ Locative adverbs are event-oriented only. The adverb cin ‘up’ denotes the upward movement of the food and, in contrast to the locational nominal in (29), does not relate to the location of the actor. The puller in (30) may as well transport the food away from herself with the help of a block and tackle. Since the nominal expressions may be participant-oriented, there is a functional difference between locational adjunct expressions containing a nominal with -ndo sufﬁx and those with an adverb. Nevertheless, locational nominal expressions cannot be analysed as a genuine depictive construction, since they may also be used as locative modiﬁers. In this function, the locational expression refers to the location of the event to which the predicate relates. Compare (31), where the locational expression oxoris¸ i cindo ‘on top of the house’ relates to the place where the event of the speaker’s spreading the hazelnuts is taking place. (31) oxori-s¸ i cindo ntxiri goo-bobVam house-gen top hazelnut on-lie_mass:[1>3]sg.prs ‘I spread hazelnuts on top of the house.’ Laz locational nominal expressions may also serve to relate only to a part of the event expressed by the predication (i.e. internal modiﬁer function, cf. Maienborn (2001)). Compare (32) for illustration. (32) kˇafri cindo me-buc¸ˇadi nail top at-hammer:[1>3]sg.pst.pfv ‘I hammered the nail on the upper part [of the door].’ The locational nominal cindo ‘top’ in (32) does not relate to the location of the whole event, since only the hammering and the object of the hammering is located on the upper part of the door. The actor of the event is not located there. Hence, the locational nominal speciﬁes an internal aspect of the event Conversely, bare forms, i.e. expressions without -ndo sufﬁx, are locative adverbs. They cannot appear in core argument function (see Kutscher 2001, chapter 3 for details) and cannot be modiﬁed by a genitive NP. (ii)

ntxiri cin goobobVam -> *oxori-s¸ i cin hazelnut above spread.on:[1>3]sg.prs ‘I spread hazelnuts above.’

254

S. Kutscher and N. Sevim Genc¸

but does not locate the event as a whole. In this function, the locational nominal is both participant-oriented (relating to the location of the nail, but not the actor) and event-oriented. Locational nominals can also be used as frame-setting expressions (Maienborn 2001). In this function, the locational nominal does not relate to the location of the event directly but sets a frame for it. Compare (33), where a young woman is instructed by her mother-in-law how to escape from the Russian soldiers who are about to invade the village. (33) oxori doloxendo /nekˇnape pencerepe vrosi; (0.3) oxori doloxendo nekˇna-pe pencere-pe vrosi house inside door-PL window-PL good doc¸ˇadi / (0.5) (. . .) do-ocˇ¸ adi emph-nail:[2>3]sg.pst.pfv ‘Inside the house, nail the doors and windows well . . . ’ askerepe 098–102)

(K/G

The locational expression oxori doloxendo ‘inside the house’ in (33) is intonationally detached and has a progredient intonation contour. It sets the frame of the hammering event denoted in the following intonation unit nekˇnape pencerepe vrosi docˇ¸adi ‘nail the doors and windows well’. In sum, locational nominal expressions in Laz serve the three different functions which Maienborn (2001) argues to be typical for locative adjuncts: internal and external modiﬁcation and frame-setting. With respect to these functions, locational nominal expressions in Laz are event-oriented. In addition, locational nominals in some uses exhibit purely participantoriented readings. Hence, locational nominal expressions in Laz are another instance of the general adjunct construction.

7.6 A genuine depictive construction: distributive numerals This section deals with quantiﬁer expressions. On constructional grounds, these have to be divided into two subclasses in Laz: collective quantiﬁcational and distributive quantiﬁcational expressions. Collective quantiﬁcational expressions are similar in construction to the manner expressions discussed in section 7.4 above. They are construed with unmarked quantiﬁers such as xvala ‘alone’ in (34) and may have participant-oriented readings (34a) or are vague with respect to participant and event orientation (34b).

On depictive secondary predicates in Laz

255

(34) a. xvala es¸ kˇa-ftare / alone up-go:1sg.fut.pfv ‘We will go up alone.’ (K/G xvala 004) ˇ b. kis¸ i ¸coyi xvala pskˇudur winter village alone live:1sg.prs ‘In the winter I live on my own in the village.’ Bare numerals, however, can only be interpreted as event-oriented. Compare the numerals ar ‘one’ and cur ‘two’ in (35), which denote the frequency of the soldiers’ comings. (35)

askerepe a moxtey ; askere-pe ar moxtey soldier-pl one come:3pl.pst.pfv ‘The soldiers came once, they came

cu moxtey / cur moxtey two come:3pl.pst.pfv twice.’ (K/G askerepe 019)

Reduplicated numerals, in contrast, are clearly participant-oriented, as in (36). (36)

askerepe curcur moxtey soldiers two:rdp come:3pl.pst.pfv ‘The soldiers came in pairs’

These participant-oriented numeral expressions can only have a distributive meaning (e.g. ‘two by two’ in the above example) and cannot be used to express the total number of referents of the phrase,8 like the German construction with zu ‘to’ þ numeral.9 Other participant-oriented quantiﬁers may also be reduplicated, compare (37). (37)

xvalaxvala / (2.3) bulur \ xvala:rdp bulur alone go:1sg.prs ‘I walk totally alone.’ (K/G inonu 058)

In these cases, however, the reduplication seems to be emphatic or, in the case of Turkish loans, may be induced by the Turkish source construction. Reduplication in these emphatic cases, however, is non-obligatory and hence not a genuine constructional means to encode participant orientation. In contrast, for numerals functioning as participant-oriented adjuncts 8 When used in argument position, numerals are sufﬁxed with -(i)ti (for details, see Kutscher 2001: ch. 3.2.2). 9 Compare German Sie kamen zu zweit (3.pl came at two) ‘The two of them came’, referring to exactly two persons coming, in contrast to the distributive construction Sie kamen in Paaren (3.pl came in pairs) ‘They came two by two’, referring to more than two persons.

256

S. Kutscher and N. Sevim Genc¸

reduplication is obligatory. Therefore, we would argue that reduplicated numeral expressions are an example of a genuine depictive construction.10

7.7 On expressing role and life stage in Laz Expressions of role such as I got the book as a present are often considered to be typical examples of depictive secondary predicate expressions. Interestingly, adjunct constructions of this type cannot be found in Laz. Role expressions in Laz always have to be biclausal constructions. Both the role predication and the main predication form independent main clauses, each containing a ﬁnite verb. An example is given in (38). (38) baba s¸ kˇimi doktori on \ hastahane ic¸alis¸ ay \ baba s¸kˇimi doktori on hastahane ic¸alis¸ay father poss.1sg doctor be:3sg.prs hospital ‘My father is a doctor. He works at the hospital.’

work:3sg.prs

Life-stage expressions, which often occur in the same type of construction as role expressions (e.g. she lived in Paris as a child), occur in two types of construction in Laz. The ﬁrst type of life stage construction is again biclausal in nature. The life stage is expressed in a ﬁnite copula construction such as bortˇi ‘I was’ in (39). In contrast to biclausal role expressions, however, the life-stage expression in biclausal life-stage constructions is the predicate of a subordinate clause marked by the motative sufﬁx -s¸ a. The motative in these constructions, together with imperfective aspect, expresses the simultaneity of the two predications. (39) bere bortˇis¸ a / lu zeri va bimxortˇi \ ˇ bere borti-s¸a lu zeri var-bimxortˇi child be:1sg.pst.ipfv-mot traditional meal neg-eat:1sg.pst.ipfv ‘When I was a child, I didn’t eat lu zeri.’ The subordinate clause construction is obligatory for this type of life-stage expression and not merely an alternative as is the case e.g. in English, where

10 Subject to the condition that distributive numeral expressions are analysed as depictive expressions. McGregor (Ch. 5, this volume) provides a detailed discussion of numerative expressions in some Australian languages. As he points out, distributive quantiﬁcational expressions (or, in his terms, iterative co-participation) do not as easily qualify as depictive expressions. Although semantically they are participant-oriented and predicative, at least in some of the Australian languages under consideration, they lack agreement with the designated controller. In the Panoan language Shipibo-Konibo (Valenzuela, Ch. 8, this volume), on the other hand, numerals in distributive function show agreement with the controller.

On depictive secondary predicates in Laz

257

the subordinate clause when I was a child could be turned alternatively into the depictive construction as a child. The fact that the subordinate clause is marked by a sufﬁx expressing simultaneity which is similar in form to the motative case marker -s¸ a (see section 7.3.2) gives rise to the question of whether this construction is a converb construction. Since the ﬁnite verb of the subordinate clause allows for an explicit subject NP and may have complements that are not coreferential with a participant of the matrix clause, we analyse the sufﬁx in this use as a temporal conjunction. Compare (40), where nana s¸ kˇimi ‘my mother’ is the subject of the predicate komoxtu ‘she came’. (40) nana s¸ kˇimi oxori ko-moxtu-s¸ a mother poss.1sg house emph-come:3sg.pst.pfv-mot kitabi golobioni book read:1sg.pst.pfv ‘When my mother came home, I was reading a book.’ The second type of life stage expressions has two variants. It may either contain a temporal adverb, as in (41b), or, for expressions denoting the exact age of the controller, a numeral is combined with the participle Z#aneri ‘of age’, as in (41a). (41) a. cur cur two ‘Two

Z#aneri Z#an-eri year-ptcp years old I

osinapu ceboc¸ˇi osinapu cebocˇ¸ i speak:vn begin:[1<3]sg.pst.pfv began to speak.’

b. ordo Vuru early die:3sg.pst.pfv ‘S/he died young.’ This kind of life stage construction is an instance of the general adjunct constructions discussed in sections 7.4 and 7.5.

7.8 Conclusion In this chapter we have presented an overview of expressions in Laz which allow participant-oriented readings. We have argued that although Laz has a wide range of expressions of this kind, there are no unique segmental or distributional means marking a given construction as a genuine depictive construction. The single major exception to this claim is reduplicated numerals functioning as distributive quantiﬁers (section 7.6). The fact that otherwise there appears to be no genuine depictive construction in Laz seems

258

S. Kutscher and N. Sevim Genc¸

to be based on two factors: (a) the lack of case agreement on various levels, including agreement between controllers and participant-oriented adjuncts; (b) the fact that participant- and event-oriented adjuncts share essentially the same distributional possibilities. They also appear to share essentially the same possibilities of prosodic marking, in particular the ability to appear in intonation units of their own. The only possible exception with regard to prosody pertains to the observation discussed in section 7.4 that manner adverbials in pre-predicate position cannot be separated prosodically from the following predicate. However, this observation as well as other features of prosodic marking and packaging are still in need of much deeper exploration.

8 Participant agreement in Panoan PILAR M. VALENZUELA

8.1 Introduction Panoan is a well-established linguistic family in South America; it comprises some thirty languages spoken in the Amazon regions of Peru, Brazil, and Bolivia. participant agreement (PA) can be considered the typologically most salient feature of Panoan grammar. It refers to the use of a distinct inﬂectional morphology on adjuncts, in correlation with the syntactic function of the participant they are predicated of.1 In this contribution, I examine PA in Shipibo-Konibo (SK), a Panoan language spoken by c.30,000 people settled along the Ucayali river and its main tributaries in the Peruvian Amazon. Although the discussion will be restricted to this single language, most of the points made also apply to the sister languages for which sufﬁcient description is available. Consider the following sentences where the locative The present account of the SK Participant Agreement system is based on Valenzuela (2003). I would like to thank Eva Schultze-Berndt and Nikolaus Himmelmann for their invitation to participate in this volume and for helpful comments on a previous version of this chapter. My analysis of the Panoan PA system has beneﬁted from input from audiences who attended my talks at the Michigan Historical Linguistics Group hosted by Sally Thomason, and the MPI for Evolutionary Anthropology in Leipzig, in 2001 and 2002; I am very grateful for this. Foremost, I am indebted to the various Shipibo speakers who shared with me data and valuable insights on this fascinating aspect of their language; especially, I express my gratitude to Kesin Beso, Sanken Bari, and Yoi Sani. Jakiribi ichabires ira´ke! 1 In addition to PA, Panoan languages share the following morphosyntactic features: dominantly agglutinative morphology with ‘polysynthetic tendencies’ in the verb; use of sufﬁxes and enclitics, except for a closed set of body-part preﬁxes which play an oblique function; basic AOV/SV order (following the labels in Dixon 1979 and 1994) and constituent order correlations conforming to this syntactic type; no (or very incipient) cross-referential pronominal marking in the verb or auxiliary; (split-)ergative alignments manifested through case-markers attached to the last NP word; syncretism/polyfunctionality involving a case enclitic -n which may indicate ergative, instrument/means, genitive, locative/allative or other oblique functions; a fairly complex switch-reference system; various types of transitivity-dependent concord; and innovative evidential systems with non-cognate morphology across languages from different subgroupings. Comparative analysis and reconstruction of the Panoan PA morphology can be found in Valenzuela (2003). See also Sparing-Cha´vez (1998) for Amawaka, Loos (1999) for Kapanawa, and Fleck (2003) for Matses.

260

P. M. Valenzuela

ßotSiki takes the marker -ß, -ßon or -ø when oriented towards the S, A, or O participant, respectively:2 (1) ßaki¯-ra ßotSiki-a-ß pa"ki¯-ki¯. child:abs-evid up-abl-s cause.to.fall:mid-compl ‘A/the child fell from high up.’

S

(2) ßaki¯-n-ra piko ßimi-n tsaka-ki¯ hiwi ßotSiki-ßon A child-erg-evid Piko:abs fruit-ins hit-compl tree up-a ‘The boy up in the tree hit Piko with a fruit [e.g. while Piko was passing by].’ (3) piko-n-ra jami ki¯ntı´ ßotSiki a-ki¯. O Piko-erg-evid metal pot:abs up do:tr-compl ‘Piko placed the metal pot high up.’ (only the pot is high up) An in-depth analysis of the Panoan PA system constitutes an important contribution not only to Amazonian linguistics but also to typology and other linguistic sub-ﬁelds; it might include peculiarities and intricacies that have not been found in other languages, and support or undermine previous hypotheses concerning participant-oriented adjuncts and similar constructions. In particular, in Panoan languages the very unusual situation arises that adjuncts may show PA, but there is no NP-internal agreement (see Schultze-Berndt and Himmelmann 2004: 83). The present chapter is organized as follows. Section 8.2 introduces the SK case-marking system. This information is necessary because, as shown in Valenzuela (2003), PA in Panoan is diachronically related to case-agreement. Sections 8.3 and 8.4 present the overt markers involved in the coding of PA and provide an overview of participant-oriented clausal and non-clausal adjuncts; adjuncts which cannot exhibit PA are also dealt with. Drawing on this analysis, section 8.5 proposes a classiﬁcation of adjuncts based on the PA patterns they allow. Also, a tentative correlation between adjunct types and varying degrees of participant orientation versus event orientation is offered. Section 8.6 summarizes the ﬁndings of the present study and raises other relevant issues.

2 The SK data throughout this article are given using IPA symbols, except for the following conventions: the retroﬂex approximant consonant //, which has several phonetic realizations, is represented by the symbol for simplicity and stands for a nasalized vowel. Primary stress falls on the ﬁrst syllable of the word, unless the second syllable is heavy, in which case this latter syllable attracts the stress. Deviation from this basic pattern is indicated through an acute accent. The following glosses are only used in the present chapter: ant.o ¼ s/a: previous event, dependent object is coreferential with matrix subject; hsy2: shorter hearsay; pst1 earlier today past; pst2: yesterday past; pst4: several years ago past.

Participant agreement in Panoan

261

8.2 The Shipibo-Konibo case-marking system SK exhibits a fairly consistent ergative-absolutive case-marking system for both pronouns and nouns. The morpheme -n is employed to indicate ergative, instrument/means, genitive (except for ﬁrst and third person singular), interessive (except for ﬁrst, second, and third person singular, as well as ﬁrst person plural), locative/allative, and temporal. While the marker -n is the only means to convey the ﬁrst four functions when nouns are involved, there are alternative ways to encode locative/allative and temporal meanings (see e.g. examples (4), (15), (28) and (38) for locative/allative, as well as examples (29) and (78) for temporal). The morpheme -n is realized through various allomorphs whose distribution is, to a large extent, morphophonologically conditioned (Valenzuela 1998; 2003: 118–26). -N-marked forms exhibit the following endings: -n; -an, -i¯n, -in; -kan, -ti¯n, -tan; -man; -nin; -ton, -tonin. Addition of -n to NPs ending in the plural/collective -ßo results in the sequences -ßaon, -ßoan, or -ßoon. Absolutive case is indicated through zero or -a; the latter allomorph is attested on the following pronouns: ﬁrst and second person singular, ﬁrst person plural, and the interrogative tso- ‘who’. Table 8.1 shows SK personal and human interrogative pronouns in the genitive, ergative and absolutive case. Let us now focus on the illustration of the SK case-marking system. Sentences (4–7) exemplify the case-marking on the ﬁrst and second person singular pronouns when playing the S, A, and O functions (for nouns see e.g. (1–3)). There is no contrast between active and inactive S: (4) ¯i-a-ra kako-nko ka-ißa-ki¯. 1sg-abs-evid Kako-all go-pst2-compl ‘I went to Kako yesterday.’

S

Table 8.1. Shipibo-Konibo Pronouns Pronoun

Genitive

Ergative

Absolutive

1 sg. 2 sg. 3 sg. 1 pl. 2 pl. 3 pl. ‘who’

nokon mi-n hawi¯n no-n mato-n hato-n haßaon tso-n

¯i-n mi-n ha-n no-n mato-n hato-n haßaon tso-n

¯i-a mi-a ha no-a mato hato haßo tso-a

262

P. M. Valenzuela

(5) ¯i-a-ra mi-n hama´-ki¯. 1sg-abs-evid 2sg-erg kick-compl ‘You kicked me.’

O/A

(6) ¯i-n-ra mi-a hama´-ki¯. 1sg-erg-evid 2sg-abs kick-compl ‘I kicked you.’

A/O

(7) mi-a-ra ransa-[a]i. 2sg-abs-evid dance-incomp ‘You are dancing.’

S

Examples (8–12) illustrate different endings and functions of -n-marked forms. Disyllabic nouns exhibit the marker -n when ending in an unstressed vowel (pia, inka, and also the compound manan-ßawi¯, whose second root conforms to this pattern), and -Vn (with certain vowel harmony) when ending in a sibilant (hisis). Disyllabic nouns whose ﬁnal segment is a stressed vowel or a nasal exhibit the endings -kan (pia´) and -man (sanki¯n, Sinan), respectively. The ﬁrst segments in -kan and -man have been analysed as neutralized realizations of a ﬁnal root consonant (Valenzuela 1998). After addition of -n, disyllabic nouns undergo stress shift according to the rule stated in footnote 3; e.g., /"inka/> /in"kan/, /hi"sis/> /"hisisin/. Trisyllabic nouns, most of which are loanwords, add the marker -nin (sisar, maria, matS´ıto, paßoro, and saßaro); stress remains unaffected. As for the functions played by -n in the examples below, these are ergative, genitive, instrument/means, temporal, and allative (see also ex. (2) for instrument): (8) sanki¯man-ra hawi¯n piakan pia-n kojaparo Sanken:erg-evid 3sg.poss nephew:gen arrow-ins tucunare´:abs tsaka-ki¯. hit-compl ‘Sanken hit a tucunare´ [type of ﬁsh] with his nephew’s arrow.’

Sinaman manan-ßawi¯-n pi¯i-ki (9) inka-n Inka-erg mind:means hill-turtle-gen wing-obl si¯ki¯-nan-a iki break-mal-compl.ptcp aux ‘The Inka, with his mental power, broke the motelo’s [type of tortoise] wing.’ (Valenzuela 1997) (10) hisis-in-ra sanki¯man ßaki¯ nati¯ß-ki¯ ischimi-erg-evid Sanken:gen child:abs bite-compl ‘A/the ischimi [type of ant] bit Sanken’s child.’

Participant agreement in Panoan

263

(11)

sisar-nin-ra maria-nin wai ri¯ra-ki¯ matS´ıto-nin Ce´sar-erg-evid Maria-gen chacra:abs felled-compl machete-ins ‘Ce´sar felled [trees in] Maria’s chacra with a machete.’ (Valenzuela 1997)

(12)

saßaro-nin-ra paßoro-nin ka-kas-ai ¯i-a Saturday-temp-evid 1sg-abs cove-all go-des-incomp ‘On Saturday I want to go to the cove.’ (Valenzuela 1997)

Instances of -n added to a plural-marked noun can be found in exx. (70) and (78). SK case-markers are enclitics operating at the NP level. They attach to the last word of the corresponding NP and may precede evidential/illocutionary force markers. In sentences (13–15), the ergative case-marker follows a coordinated NP, a compound noun modiﬁed by an adjective, and a relative construction, respectively: (13)

[tita ßi¯tan papa]-n-ra hato-n ßaki¯-ßo mother and father-erg-evid 3pl-gen child-pl.abs ¯isi¯-ai advise-incomp ‘Mother and father advise their children.’

(14)

[wiso-ino sina]kan-ronki honi ra"ßi¯ pi-ki¯. black jaguar ﬁerce:erg-hsy man two:abs eat-compl ‘I heard that the ﬁerce black jaguar ate [the] two men.’

(15)

[honi kako-nko-ni-a-ß noko-ißat-a]-tonin-ra man Kako-loc-lig-abl-s meet:mid-pst2-compl.ptcp-erg-evid rono ri¯ti¯-ki¯. snake:abs kill-compl ‘The man who arrived from Kako yesterday, killed a snake.’

The ergative-absolutive alignment is kept even when one or both arguments remain unexpressed; unless otherwise indicated, the absence of a required argument is interpreted as a third person singular pronoun: (16) a. ¯i-a-ra hama´-ki¯. 1sg-abs-evid kick-compl ‘He/she/it kicked me.’ b. ¯i-n-ra hama´-ki¯. 1sg-abs-evid kick-compl ‘I kicked him/her/it.’ Ergative-absolutive case-marking persists even if the arguments of a transitive clause lack important agent or patient characteristics (Valenzuela 1997).

264

P. M. Valenzuela

8.3 Interclausal and intraclausal participant agreement As mentioned in the introduction, the term participant agreement (PA) refers to the use of a distinct inﬂectional morphology on adjuncts, in correlation with the syntactic function of the participant they are predicated of. PA takes place at the intraclausal and interclausal levels, where ‘clausal’ expressions are those containing either a verbal predicate or a copula verb. Intraclausal PA involves adjunct words and phrases (including interrogatives), emphatic pronouns, and non-clausal conjunctions, whereas interclausal PA is found on dependent same-subject marked clauses and clausal conjunctions. However, as will be discussed below, the boundary between these two levels is not always clearcut. Table 8.2 introduces the overt morphology involved in the SK PA system and lists the syntactic environments in which it is attested. As can be observed in Table 8.2, the markers -aß, -i, and -noß are used to encode S-agreement; on the other hand, -ßon, -kin, and -noßon convey A-agreement. It must be stated that the markers involved in the overt coding of PA are not further analysable in SK, at least synchronically. Although -ßon, -kin, and -noßon (the latter may be segmented into -no þ -ßon) carry the ending /n/ which suggests the presence of the ergative case-marker, the sequences -ßo, -ki, and -no are not transparent (although -ki is formally

Table 8.2. Overt participant agreement markers in Shipibo-Konibo Morpheme S-agreement

A-agreement

-(a)ß

-ßon

-i

-kin

-noß

-no(n)ßon

Non-clausal adjuncts

Same-subject clauses

complement of interest life-stage proprietive privative (only A forms) emphatic pronoun conjunction manner quantity ablative (only S forms) locative/allative (only A forms) manner quantity conjunction –

anterior

simultaneous

subsequent

Participant agreement in Panoan

265

identical to an oblique marker: see ex. (9) and footnote 8). In addition, the markers encoding S-orientation do not resemble the absolutive case-marker.3 Sections 8.3.1 and 8.4 examine the coding of PA at the interclausal and intraclausal levels. PA is also attested on locative- or similative-marked demonstratives functioning as conjunctions at different syntagmatic levels inside and beyond the clause. However, PA on conjunctions has been omitted since its treatment is not directly relevant to the present discussion (see (102) for illustration). 8.3.1 Interclausal participant agreement The main reference-tracking mechanism of SK is its elaborate switchreference system. Dependent reference-marked clauses, which are obligatorily verb-ﬁnal, may (immediately) precede or follow their matrix clause, or occur embedded in it. In most instances, reference-marked clauses ground the situation described by the main clause; however, the exact relationship between the events in the two clauses is not speciﬁed but left to be inferred from the context. Alternatively, a reference-marked clause can encode an event that advances the main storyline. Hence, SK reference-marked clauses may correspond to English temporal, purposive, conditional, reason, concessive, and even coordinate clauses. Same-subject markers signal syntactic dependency (they occur in lieu of ﬁnite morphology marking aspect and illocutionary force), identity of subjects (where subject is the conﬂation of S and A), and the relative temporal or logical order of the events depicted by the dependent clause and its matrix clause.4 But most interestingly, same-subject markers alternate in correlation with the S/A function of the matrix verb subject. Consider examples (17a) and (17b): katSio (17) a. nokon papa-ra [wai mi¯no-aß] 1sg.poss father:abs-evid chacra:abs burn-ant.ss.s to.forest ka-ki¯ go-compl ‘After burning the chacra my father went to the forest.’

3 While most of the agreement markers have no other function in the language, -ßon and -kin are identical in form and compatible in meaning with the benefactive and associative applicatives, respectively (Valenzuela 1999: 68–9; see also examples (18b) and (45)). 4 Given its morphosyntactic and semantic characteristics, the marked verb would ﬁt the deﬁnition of converb, and thus is comparable to the phenomena discussed in Gu¨ldemann’s (Ch. 10) and Amha and Dimmendaal’s (Ch. 9) contributions to this volume.

266

P. M. Valenzuela b. nokon papa-n-ra [wai mi¯no-ßon] 1sg.poss father-erg-evid chacra:abs burn-ant.ss.a ßana-ki¯ atsa manioc:abs sow-compl ‘After burning the chacra my father sowed the manioc.’

In (17a) and (17b), the events in the dependent clause (given in square brackets) precede the events in the matrix clause, and subjects are the same; but in (17a) the matrix verb ka- is intransitive while ßana- in (17b) is transitive. Since PA is controlled by the matrix clause subject, the same-subject marker -aß is selected in (17a) to code S-agreement, while -ßon is selected in (17b) to code A-agreement. The next example illustrates the distribution of same-subject markers when the events in the dependent and matrix clauses are overlapping or simultaneous. Example (18) consists of two ﬁnite clauses, each one of them modiﬁed by a simultaneous same-subject marked clause; -kin is used when the matrix clause subject plays the A function, while -i is required when this controller works as S: (18) a. ¯i-a-ra [nokon tita-n ¯isi¯-kin] 1sg-abs-evid 1sg.poss mother-erg advise-sim.ss.a ani-a iki big-do:tr.compl.ptcp aux ‘My mother raised me giving me advice.’ b. ¯i-a-ra iki ¯i "si¯-ja 1sg-abs-evid advice-prop cop [nokon tita-n joij-ai 1sg.poss mother-erg tell-incomp.ptcp:abs ninka´-ßon-katit-i] hear-ben-pst4-sim.ss.s ‘I am wise following [lit. ‘‘hearing to her beneﬁt’’] what my mother used to tell me.’ Finally, examples (19) and (20) illustrate the use of the same-subject markers corresponding to subsequent or following events showing S- and A-orientation or control: (19)

ha-ki¯ [wi¯stı´ora mi¯tsa´ . . . no-a i-ti 1pl-abs do:intr-inf exist-compl one beautiful raotia ainßo i-noß] woman do:intr-subs.ss.s adorned ‘ . . . we must be adorned in order to be a beautiful woman.’

Participant agreement in Panoan

267

(20) wi¯stı´ora ni¯ti¯ wino´-ma-ßon, katSio ßo-ti one day:abs pass:mid-caus-ant.ss.a to.forest take-inf iki [ßi¯ti¯-ma-noßon] aux smell-caus-subs.ss.a ‘After one day, one takes [the treated dog] to the forest to have it smell [at animals].’ Note that the terms S-orientation and A-orientation, used to account for the distribution of same-subject markers, directly translate into agreement with intransitive and transitive matrix verbs respectively. That is, it could have simply been stated that -aß, -i, and -noß are required when the matrix verb is intransitive, whereas -ßon, -kin, and -noßon must be selected when the matrix verb is transitive. This is the main reason why PA has been previously accounted for in terms of ‘transitivity agreement’ (Valenzuela 1999) or ‘transitivity concord’ (Loos 1999), given that linguists devoted to the study of Panoan languages have concentrated almost exclusively on the switch-reference system, i.e. on interclausal PA (but cf. Valenzuela 1999). However, when examining this feature at the intraclausal level, especially in the instances where alternate inﬂections are possible, it becomes evident that transitivity is not the determining factor (see especially the discussion in 8.4.1 and 8.4.3). Before turning to intraclausal participant agreement, a few words on different-subject marking are in order. 8.3.2 Different-subject marking Different-subject-marked clauses in SK can be divided into (a) those which lack an object that is coreferential with the matrix clause subject (including intransitive ones) and (b) those whose object is coreferential with the matrix clause subject (i.e. instances of object-to-subject coreferentiality). For the sake of brevity, only clause combinations in which the event in the dependent clause is anterior to or simultaneous with the event in the matrix clause will be considered here. Different-subject-marked clauses lacking an object or whose object is not coreferential with the matrix clause subject may take the aspect morphemes -ai or -ki¯ and thus can be said to be (more) ﬁnite.5 In accordance with their incompletive and completive aspect functions, -ai is used when the two events are viewed as contemporary or overlapping, while -ki¯ is selected when 5

A ﬁnite verb in SK consists minimally of a root and an aspect marker.

268

P. M. Valenzuela

the event in the marked clause is presented as anterior. In the majority of cases, these clauses take additional subordinating morphology, in particular the temporal -tian, which can be said to have specialized as different-subject marker: (21) [piko´-ki¯-tian]-ßi no-a ani a-kan-ai . . . take.out:mid-ant-ds-emph 1pl-abs big do.tr-pl-incomp ‘From the moment we are born, they [our parents] take care of us . . . ’ (22) [ina´ tsaka-ki¯-tian] domesticated.animal:abs shoot.with.arrow-ant-ds wi¯stı´ora ainßo-nin ha-"ßi¯ ainßo-ßo ki¯na-ki¯ one woman-erg 3.sg-com woman-pl.abs call-compl ‘After [the men] shot at the domesticated animal, a woman called the other women accompanying her.’ (23) [hi¯ni¯-n ri¯ti¯-ai-tian]-ra ﬂowing.water-erg kill-sim-ds-evid ainßo sai ik-ai woman:abs onom:cry.out.for.help do:intr-incomp ‘Since (s)hei was drowning, the womanj cried out for help.’ (i.e. the one drowning is not the one crying out for help) *‘Since shei was drowning, the womani cried out for help.’ In the sentence immediately above, the subject of the dependent clause is hi¯ni¯ ‘ﬂowing water’. A literal translation would be ‘Since the ﬂowing water was killing heri, the womanj cried out for help’; there is no verb equivalent to ‘drown’ in SK. If -ai-tian strictly coded simultaneity of events and nonidentity of subjects, the second interpretation above would be perfectly possible. However, the drowning woman being the object of the dependent clause cannot be the one crying out for help, or else it would be an instance of object-to-subject coreferentiality which is treated below. Hence, in addition to coding non-identity of subjects, -ki¯ -tian and -ai-tian indicate that the object of a transitive marked clause is non-coreferential with the matrix clause subject. The more idiosyncratic reference-marker -a indicates that the object of a dependent clause conveying an anterior event is coreferential with the subject (S/A) argument of its matrix clause; i.e., differently from -ai-tian and -ki¯ -tian this marker restricts the reference of the dependent clause in the same way SS markers do. This marker is glossed as Anterior Object-to-Subject Coreferentiality (ant.o ¼ s/a). Consider examples (24)

Participant agreement in Panoan

269

and (25), involving an intransitive and a transitive matrix verb subject respectively: (24) [ha-n rao-n-a]-ra ¯i-a 3sg-erg medicine-vblz.tr-ant.o ¼ s/a-evid 1sg-abs ka-wan-ki¯ go-pst1-compl ‘S/he treated me with plant medicine, and I left.’ (25) [ha-n ¯i-a rao-n-a]-ra ¯i-n 3-erg 1-abs medicine-vblz.tr- ant.o ¼ s/a-evid 1sg-erg nonti ßi¯na-wan-ki¯ canoe:abs look.for-pst1-compl ‘S/he treated me with plant medicine, and I looked for a canoe.’ In a converse situation, i.e. when the dependent subject and the matrix object are coreferential, different-subject marking must be employed (see ex. (21)). There is no dedicated morpheme, equivalent to -a, coding object-tosubject coreferentiality with simultaneous events. However, the restriction that ‘different subject’ entails object-to-subject non-coreferentiality still holds, as has been shown by example (23).

8.4 Intraclausal participant agreement and participant agreement on adjuncts of intermediate syntactic status PA markers are also attested on certain word-level and phrasal adjuncts providing optional adverb-like information about a core argument of the clause; namely, on locative, manner, quantity, life-stage, affective, proprietive, and privative adjuncts, as well as emphatic pronouns. In 8.4.1–9 the coding of PA in each of these contexts is examined. It is also shown that certain manner and quantity expressions exhibit an intermediate syntactic status, between the phrasal and clausal levels. While three different sets of PA markers occur on same-subject clauses, only two of these sets are found on non-clausal adjuncts: -aß (S) versus -ßon (A) and -i (S) versus -kin (A). In some instances, e.g. on the locative/allative expressions discussed in 8.4.1, only one of the PA markers occurs, contrasting with zero. 8.4.1 Locative/allative: -ßon (A-orientation) vs. -Ø (S- / O-orientation) Locative/allative adjuncts receive the marker -ßon when semantically oriented towards an A participant, but remain unmarked when predicating of S/O participants; in terms of overt marking, therefore, PA on locatives/allatives

270

P. M. Valenzuela

is organized in the ergative-absolutive fashion. Consider the following examples where locatives are Sa- and So-oriented: (26) hawi¯rano-ki inka-bo ha-a? where-q Inka-pl.abs exist-compl.ptcp ‘Where do the Inkas live?’ (27) ßotSiki-ra ¯i-a oßa-[a]i / tsini-ai / ßi¯wa-[a]i up-evid 1sg-abs sleep-incomp play-incomp sing-incomp ‘I sleep / play / sing high up [e.g. in a higher area inside the house]’ hi¯ni¯-nko pa"ki¯-ki¯ (28) naSi-ra ¯i-a bathe:sim.ss.s-evid 1sg-abs ﬂowing.water-loc drop:mid-compl ‘While I was bathing, I fell in the water [I was in the water already].’ Whereas examples (26–28) involve different types of intransitive verb, examples (29–33) involve (main) transitive ones. The absence of -ßon in (29) indicates that the locative is oriented towards the O participant and the A participant is not at the same location; see also ex. (3):

ß¯ia-ti tSomo (29) ha-tian ani nonti-n wi¯stı´ora atsa that-temp big canoe-loc one manioc drink-inf jar:abs na-ja´sa-n-kan-ai interior-sitting.position:tr-tr-pl-incomp ‘Then, they put a jar of manioc beer inside the big canoe.’ (The agents are not in the canoe; Valenzuela 2002: 469) Examples (30–33) also involve transitive verbs or clauses; however, the presence of -ßon indicates semantic orientation of the locative adjunct towards the A participant. Whether the O participant is in the same location (examples (30) and (31)) or in a different location (examples (32) and (33), also example (2)) is irrelevant: pi-ai? (30) hawi¯rano-ßon-ki ¯ipa-n where-a-q paternal.uncle-erg eat-incomp.ptcp ‘Where does paternal uncle eat?’ (31) . . . ani nonti-n-ßon wi¯stı´ora atsa ß¯ia-ti tSomo big canoe-loc-a one manioc drink-inf jar:abs ki¯jo-kan-ki¯ ﬁnish-pl-compl ‘ . . . they ﬁnished a jar of manioc beer in the big canoe.’ (Valenzuela 2002: 469)

Participant agreement in Panoan

271

(32) ¯i-n-ra oin-ki¯ [honi wino-t-ai] 1sg-erg-evid see-compl person:abs pass-mid-incomp.ptcp:abs hi¯ni¯-nko-ßon ﬂowing.water-loc-a ‘Ii saw someonej passing (in a canoe) in the wateri.’ (I was bathing in the water, and I saw someone passing in a canoe.) (33) ha-tian nokon tita-n joij-a iki ¯i-a that-temp 1sg.poss mother-erg 1sg-abs tell-compl.ptcp aux mana-i ka-ti] [ha ti¯naman-ßon 3:abs on.the.shore-a await-sim.ss.s go-inf:abs ‘Then my mother told me to go wait for him on the shore.’ Whereas in example (32) the locative is oriented towards the A argument of the main clause, in (33) it is oriented towards the A argument of the subordinate clause. As shown in examples (29–33), the presence of -ßon does not depend on the transitivity of the verb or clause (since these are all transitive), but on the syntactic function of the participant towards which the locative is semantically oriented. In sum, overt A-agreement through -ßon is used to indicate the A participant is found at the location where the event takes place; the O participant may or may not be at that location too. Zero agreement indicates either S- or O-orientation. 8.4.2 Ablatives: -ß (S-orientation) vs. -Ø (O-orientation) Differently from locative/allative adjuncts, ablatives exhibit the additional marker -ß to signal S-orientation, but receive no overt morphology when the ablative is semantically oriented towards the O participant. A-orientation can only be coded through a locative construction. Examples (34) and (35) illustrate S-orientation, while example (36) shows O-orientation: ho-a? (34) hawi¯rano-a-ß-ki mi-a where-abl-s-q 2sg-abs come-compl.ptcp ‘Where do you come from?’ ma"pi¯-ki¯ (35) ¯i-a-ra hi¯ni¯ mi¯rano-a-ß 1sg-abs-evid ﬂowing.water inside:loc-abl-s go.up-compl ‘I went up from inside the water.’ (36) ¯i-n-ra jami ki¯ntı´ ßotSiki-a ßi¯-ki¯ 1sg-erg-evid metal pot:abs up-abl bring-compl ‘I brought the metal pot from high up.’

272

P. M. Valenzuela

As mentioned above, it is not possible for an ablative structure to receive A-orientation agreement morphology; however, a -ßon-marked locative structure may be used instead for this purpose. Therefore, although the English translations in (37a, b) and (38a, b) contain the ablative preposition ‘from’, the SK constructions in the (b) examples involve locative structures: ninka´-ki¯ (37) a. ¯i-n -ra katSio-ki¯-a 1sg-erg-evid in.the.forest-loc-abl hear-compl ‘I heard it from the forest.’ (The noise came from the forest.) ninka´-ki¯ *katSioki¯a-ßon b. ¯i-n-ra katSio-ßon 1sg-erg-evid in.the.forest-a hear-compl ‘I heard it from the forest.’ (I was in the forest; the noise may have come from the forest or not; lit. ‘I heard it in the forest.’) (38) a. ¯i-n-ra ßaki¯ paki¯-ki¯ 1sg-erg-evid child:abs cause.to.fall-compl wi¯jo´ti-aino-a hammock-loc:loc-abl ‘I [accidentally or purposefully] caused the child to fall from the hammock.’ (Only the child was in the hammock.) b. ¯i-n-ra ßaki¯ paki¯-ki¯ 1sg-erg-evid child:abs throw-compl wi¯jo´ti-ain-ßon *wi¯jo´tiainoa-ßon hammock-loc-a ‘I caused the child to fall from the hammock.’ (I was in the hammock too; lit. ‘I caused the child to fall in the hammock.’) Alternations involving locative versus ablative adjuncts like the ones illustrated above are not possible with the transitive movement verb ß¯i‘bring’ which only allows for an unmarked and hence O-oriented ablative adjunct: (39) a. ¯i-n-ra ian napo-ri-ki¯-a 1sg-erg-evid lake interior-about-loc-abl ßi¯-ki¯ onpaß contained.water:abs bring-compl ‘I brought water from approximately the middle of the lake.’ b. *¯i-n-ra ian napo-ri-ßon onpaß 1sg-erg-evid lake interior-about-a contained.water:abs

Participant agreement in Panoan

273

ßi¯-ki¯ *naporiki¯aßon bring-compl ‘I brought water from approximately the middle of the lake.’ One may speculate that in order to bring something from one place to another both agent and patient participants must share the source and destination locations. In this case, it is the adjunct form showing semantic orientation towards the patient-object that has become grammaticalized.6 Whereas a locative structure is employed to overtly mark A-orientation, an ablative structure may be used instead of a locative one to encode overt S-agreement. In fact, there are instances of alternate locative and ablative S-oriented adjuncts, but unlike (37) and (38) these alternations are not accompanied by an obvious shift in meaning. In the following examples, the location of the S argument is coded through a locative adjunct in the (a) sentences but through an ablative one in the (b) counterparts (examples (40a) and (41b) were spontaneously offered by native speakers when discussing a different topic, while (40b) and (41a) were elicited for comparison): (40) a. ¯i-manaon-ra popo ki¯o ki¯o-wan-ki¯ ni¯ti¯ 1sg-on.top.of-evid owl:abs cry cry-pst1-compl day:abs ßaßat-ai-tian clear-sim-ds ‘The owls cried above me [i.e. on top of my house] very early in the morning.’ popo ki¯o ki¯o-wan-ki¯ b. ¯i-manaon-ki¯-a-ß-ra 1sg-on.top.of-loc-abl-s-evid owl:abs cry cry-pst1-compl ßaßat-ai-tian ni¯ti¯ day:abs clear-sim-ds ‘The owls cried above me [i.e. on top of my house] very early in the morning.’ ki¯o ki¯o-wan-ki¯ ja"mi¯ (41) a. ¯i-naman-ra aSa´ 1sg-under-evid asha´:abs cry cry-pst1-compl at.night ‘The asha´ toads sang and sang under my house [lit. ‘under me’] last night.’ b. ¯i-namami¯-a-ß-ra aSa´ ki¯o ki¯o-wan-ki¯ 1sg-under:loc-abl-s-evid asha´:abs cry cry-pst1-compl 6 In contrast, the agent-oriented alternative has been selected in the sister languages Matses (Valenzuela 2003: 902) and Sharanawa (Tournon and Valenzuela, ﬁeldnotes).

274

P. M. Valenzuela ja"mi¯ at.night ‘The asha´ toads sang and sang under my house [lit. ‘under me’] last night.’

It was previously noted that PA on locatives/allatives follows an ergativeabsolutive distribution, since -ßon encodes A-orientation while S- and O-oriented adjuncts remain unmarked. Ablatives, on the other hand, can be said to exhibit an S-marked system, whereby only S-oriented adjuncts receive overt marking, i.e. -ß. But if locatives/allatives and ablatives are treated as a single place-adjunct category, it can be concluded that PA operates on a tripartite basis, with the markers -ßon for A, -ß for S, and -Ø for O. This classiﬁcation is supported by the fact that both locative/allative and ablative markers may have a relative locational function. 8.4.3 Participant agreement in causative and applicative constructions This section examines how PA operates in complex monoclausal sentences such as those involving causatives and applicatives. First, consider the following causative expressions involving an intransitive base verb. Examples (42a) and (42b) differ only in the agreement the locative adjunct takes, i.e. -Ø vs. -ßon; the change in meaning is clear. Sentence (42a) indicates that the speaker stays out of the forest but sends his dog to the forest to bark (at animals); by contrast, in (42b) both the speaker and the dog are in the forest and the speaker has the dog bark (at animals): hoho (42) a. ¯i-n-ra nokon otS´ıti 1sg-erg-evid 1sg.poss dog:abs onom:bark i-ma-[a]i nii mi¯ran do:intr-caus-incomp forest inside ‘I make my dog bark in the forest.’ (Speciﬁes location of causee.) b. ¯i-n-ra nokon otS´ıti hoho 1sg-erg-evid 1sg.poss dog:abs onom:bark i-ma-[a]i nii mi¯ran-ßon do:intr-caus-incomp forest inside-a ‘I make my dog bark in the forest.’ (Speciﬁes location of causer.) Examples (43a) and (43b) also differ in meaning. The former indicates that only the dog is on the palm-bark ﬂoor, while the latter expresses that the

Participant agreement in Panoan

275

speaker is (also) up there (Shipibo houses generally stand on poles, so that the ﬂoor and the ground are at different levels): (43) a. ¯i-n-ra nokon otS´ıti ransa-ma-[a]i 1sg-erg-evid 1sg.poss dog:abs dance-caus-incomp tapo-n palm.bark.ﬂoor-loc ‘I make my dog dance on the palm-bark ﬂoor.’ (Speciﬁes location of causee.) ransa-ma-[a]i b. ¯i-n-ra nokon otS´ıti 1sg-erg-evid 1sg.poss dog:abs dance-caus-incomp tapo-n-ßon palm.bark.ﬂoor-loc-a ‘I make my dog dance on the palm-bark ﬂoor.’ (Speciﬁes location of causer.) The data examined above are compatible with the analysis provided so far. That is, the locative bears the marker -ßon when predicating of the A-causer but zero when predicating of the O-causee. However, in the next example the ablative ‘from Kako’ bears S-orientation agreement even though the absolutive NP (Wesna) is either the causee-object (44) or the applicative-object (45) of the clause (recall that O-orientation is coded by -Ø): (44) ¯i-n-ra wi¯sna ho-ma-ki¯ 1sg-erg-evid Wesna:abs come-caus-compl kako-nko-ni-a-ß Kako-loc-lig-abl-s ‘I made Wesna come from Kako.’ (45) ¯i-n-ra wi¯sna noko´-kin-ki¯ 1sg-erg-evid Wesna:abs meet:mid-assoc-compl kako-nko-ni-a-ß Kako-loc-lig-abl-s ‘I came with Wesna from Kako.’ The examples above were provided by different Shipibo collaborators. Interestingly, when asked whether (44) and (45) could take the zero-marked ablative form kakonkonia, one of these collaborators, Mr Sanken Bari, explained that kakonkonia is actually the ‘‘correct’’ form although speakers normally produce kakonkoniaß in these contexts. This observation (provided by a very insightful speaker) suggests that the absolutive Wesna triggers semantic agreement on the ablative, since it is in fact the S argument of the base predicate.

276

P. M. Valenzuela

An alternative way to account for the presence of overt S-agreement in (44) and (45) would be to claim that in these types of complex predicate the adjunct may show agreement with a base verb argument. I will return to this point shortly. I have just shown that in causative or applicative constructions involving an intransitive base verb it is possible to obtain overt S-orientation agreement through -ß on an ablative structure, to indicate that the adjunct predicates of the base subject. However, an analogous situation does not hold when the base verb is transitive; for example, it is not possible to employ the A-orientation agreement marker -ßon on a locative expression to signal orientation towards the causee: pi-ma-[a]i (46) a. ¯i-n-ra nokon otS´ıti 1sg-erg-evid 1sg.poss dog:abs eat-caus-incomp tapo-n palm.bark.ﬂoor-loc ‘I make my dog eat on the palm-bark ﬂoor.’ (Speciﬁes location of causee.) pi-ma-[a]i b. ¯i-n-ra nokon otS´ıti 1sg-erg-evid 1sg.poss dog:abs eat-caus-incomp tapo-n-ßon palm.bark.ﬂoor-loc-a ‘I make my dog eat on the palm-bark ﬂoor.’ (Speciﬁes location of causer.) In the sentence immediately above, the marker -ßon necessarily signals that the locative adjunct is oriented towards the causer. One possible conclusion is that in causative and probably also applicative constructions an adjunct cannot be oriented towards a base verb argument—semantic agreement on ablatives controlled by an intransitive base verb subject being the exception, as in (44) and (45). An alternative analysis is to posit that it is possible for a base verb argument to control PA; however, when the base verb is transitive, both subjects compete for A-status and A-agreement on adjuncts is necessarily linked to the highest ranked A-participant. Although I have not been able to come up with speciﬁc tests that would help me decide between these two alternatives, I feel inclined towards the semantic agreement interpretation due to Mr Sanken Bari’s judgements cited above. 8.4.4 Complement of interest: -aß (S-orientation) vs. -ßon (A-orientation) In SK a ‘complement of interest’ encodes the participant in whose interest one undergoes a change of internal state or carries out an action. Complements of interest are thus generally the second argument or stimulus of extended

Participant agreement in Panoan

277

intransitive verbs, and carry the interessive case-marker -on -n, which may or may not be followed by PA morphology. Examples (47–9) illustrate interessive-marked adjuncts without PA. (47)

ha-on ki¯¯in-ai ¯i-a-ra 1sg-abs-evid 3sg-intrss want-incomp ‘I love him.’

(48)

Sinan-ßi¯not-i ha-tian honi mi-on that-temp man:abs 2sg-intrss mind-fail-sim.ss.s ka-[a]i go-incomp ‘Then the man is going to forget you.’

(49)

honi joßi¯n-wan-ki¯ ¯i-on-ra 1sg-intrss-evid man:abs ask.in.matrimony-pst1-compl ‘The/A man asked [my parents] for me in matrimony.’

PA morphology on interessives has been attested in elicitation, but in these instances they generally have an affective meaning, either benefactive or malefactive. It must be pointed out that only example (51) comes from a spontaneous text, and the meaning of this interessive-marked expression seems to be ablative rather than affective. Interessives agreeing with S and with A will now be discussed in turn. The overt S-orientation morpheme -ß has been attested on interessivemarked pronouns; however, these pronouns need to take the ablative case (which in turn is built upon the locative) before addition of -ß. That is, the interessive pronouns in (50a) and (51–4) are composed of the following morphemes: pronominal root þ interessive -on þ locative -mi¯ þ ablative -a þ participant agreement (-mi¯ is a locative allomorph that precedes the ablative on bases ending in a nasal): aß¯i-ai (50) a. ¯i-a-ra mi-on-mi¯-a-ß 1sg-abs-evid 2sg-intrss-loc-abl-s get.used.to/learn-incomp ‘I learn for you/I learn from you.’7 Sentences (50a) and (50b) are formally almost identical, but the interessive pronoun in the latter example is not in the ablative case and cannot take PA;

7 Speakers differ as to the preferred interpretation of this utterance. For example, while for Mr Yoi Sani the benefactive reading was the preferred one, Mr Sanken Bari allowed for the second reading only.

P. M. Valenzuela

278

consequently, (50b) does not allow for a benefactive reading: b. ¯i-a-ra mi-on aß¯i-ai 1sg-abs-evid 2sg-intrss get.used.to/learn-incomp ‘I am going to get used to you.’8 Compare also (50a) with the example below, where ¯ionmi¯a is in the ablative but lacks overt agreement morphology; as mentioned above, its meaning seems to be ablative rather than affective. In (51), the spirit of the Sun Medicine plant talks, in her dreams, to a woman who has drunk a preparation made from that plant: (51)

ß¯ia-a, hawi¯ kopı´-ki mi-n ¯i-a what reason-q 2sg-erg 1sg-abs drink-compl.ptcp i¯ -on-mi¯ -a onan-kas-kin . . . ? 1sg-intrss-loc-abl know-des-sim.ss.a ‘Why did you drink me, wanting to learn from me . . . ?

In the next examples the interessive-marked pronouns are in the ablative and, when in combination with intransitive verbs, exhibit overt S-agreement morphology. The addition of the emphatic -ßi contributes to the malefactive rather than benefactive reading: (52)

ha-ra ¯i -on-mi¯-a-ß-ßi 3:abs-evid 1sg-intrss-loc-abl-s-emph wano-ki¯. *¯ion(ßi) marry.a.woman-compl ‘He got married in spite of being with me.’

(53)

ßaki¯ mawat-a. mi-on-mi¯-a-ß-ßi *mion(ßi) 2sg-intrss-loc-abl-s-emph child:abs die-compl.ptcp ‘The child died being under your care.’ (It is an accusation; maybe you didn’t take good care of the child, or you didn’t take it to the onanya in order to be treated.)

The analysis of the examples provided here suggests that the ablative marking in (50a) and especially in (52) and (53) is grammatically motivated; i.e. ablative case-marking is simply used to allow the adjunct to take overt S-agreement morphology, rather than to code ablative function. The next 8

One speaker offered the alternative form: (i)

mi-ki aß¯i-ai ¯i-a-ra 1sg-abs-evid 2sg-obl learn/get.used.to-incomp ‘I am going to get used to you.’

Participant agreement in Panoan

279

example exhibits overt A-agreement, given that the verb kinan- is transitive and requires an ergative-marked subject: (54) mi-on-ßon ßaki¯-n kinan-a *mion 2sg-intrss-a child-erg vomit-compl.ptcp ‘The child vomited being under your care.’ In sum, overt PA on interessives is not common in SK. In fact, the only instances attested so far have been obtained through elicitation. Nevertheless, native speakers agree as to the grammaticality of these constructions, even though their interpretation of them may differ. 8.4.5 Life stage: -ß (S-orientation) vs. -ßon (A-orientation) Life-stage expressions may also show S- or A-agreement overtly; these agreement-marked forms are followed by the emphatic -ßi. Example (55) illustrates S-orientation: ßaki¯-ß-ßi (55) ha-rı´ßa-ki¯ ßaki¯ tS¯ia-a-ti, exist-rep-compl child:abs very.very.skinny-do.tr-inf child-s-emph too-ki¯-tian poko-mi¯-a-ßi ßaki¯ maßko become.pregnant-ant-ds womb-loc-abl-emph child:abs small a-ti do:tr-inf ‘There is also [a plant with special powers] to make the foetus very very skinny, to make the foetus small when a girl becomes pregnant at/from a very early age.’ Participant agreement involving the expression of a speciﬁc age is provided in (62) when dealing with proprietives. The temporal -tian, introduced as different-subject marker in 8.3.2, may be used on life-stage expressions referring to the A or O argument of a transitive clause; i.e., these expressions are not participant-oriented. Consider the next sentence involving two human core participants: (56) ¯i-n-ra onan-ki¯ sanki¯n ßari ßaki¯-tian-ßi 1sg-erg-evid know-compl Sanken Bari:abs child-temp-emph ‘I have known Sanken Bari since I was a child/since he was a child.’ In the example below, ßaki¯ -Soko receives the A-orientation marker -ßon: (57) mapo´ a-kin pi¯o-ti r-iki ßaki¯-Soko-ßon-ßi clay do.tr-sim.ss.a begin-inf evid-cop child-dim-a-emph ‘To make pottery one has to begin at a very young age.’

280

P. M. Valenzuela

Agreement-marked life-stage expressions may code the equivalent of English ‘since’, i.e. a temporal-ablative sense, despite the absence of ablative marking. Compare the following pairs of sentences involving intransitive verbs: (58) a. ¯i-a-ra ti¯¯it-i pi¯o-koot-a iki 1sg-abs-evid work-sim.ss.s begin-mid-compl.ptcp aux ßaki¯-tian-ßi child-temp-emph ‘I began to work when [being a] child.’ b. ¯i-a-ra ti¯¯it-i 1sg-abs-evid work-sim.ss.s ßaki¯-ß-ßi child-s-emph ‘I began to work as a child.’ ‘I have worked since [being] a

pi¯o-koot-a iki begin-mid-compl.ptcp aux

child.’

(59) a. ßiri-ra rima-n ka-a iki Biri:abs-evid Lima-all go-compl.ptcp aux ßaki¯-Soko-tian-ßi child-dim-temp-emph ‘Biri went to Lima when [being a] little girl.’ b. ßiri-ra rima-n ka-a iki ßaki¯-Soko-ß-ßi Biri:abs-evid Lima-all go-compl.ptcp aux child-dim-s-emph ‘Biri went to Lima since [being a] little girl.’ Some speakers prefer (58a) and (59a) when the situations expressed do not hold any more at the moment of utterance; their (b) counterparts, on the other hand, probably entail that the speaker is still working and Biri is still in Lima. But what is of relevance here is that it is justiﬁed to establish a distinction between true temporal non-participant-oriented expressions marked by -tian, as opposed to participant-oriented life-stage expressions marked by -ß and -ßon. This is compatible with the fact that -tian is used to encode different-subject (8.3.2), whereas PA morphology takes part in the coding of same-subject (8.3.1). An S-marked life-stage adjunct may also allow for a concessive interpretation. This is probably because life-stage adjuncts receive emphatic marking and the emphatic -ßi is often used on dependent clauses to convey

Participant agreement in Panoan

281

a concessive/adversative meaning.9 The expressions below were given a concessive reading: (60) ¯i-a-ra ti¯¯it-ai ßaki¯-ß-ßi 1sg-abs-evid work-incomp child-s-emph ‘I work in spite of [being a] child.’ (61) ha-ra ransa-[a]i otS´ıti-ni-ß-ßi 3:abs-evid dance-incomp dog-lig-s-emph ‘It dances in spite of [being a] dog.’ Note that on life-stage adjuncts the S-agreement morpheme -ß occurs without ablative marking. As will be shown in 8.4.7, this is also the case with regard to emphatic pronouns. I will address the relevance of this issue in 8.5. 8.4.6 Proprietive and privative: -ßon (A-orientation) vs. -Ø / -ß (S-orientation) Participant agreement also occurs on proprietive and privative modifying adjuncts; however, overt S-orientation has only been attested with age expressions like the one in ex. (62), but not with other types of intransitive clauses, as shown in (63):

Sinan-ai miSki-ti, (62) pitSika ßaritia-ja-ni-ß moa ﬁve year-prop-lig-s already think-incomp ﬁsh.w/hook-inf:abs pia a-ti, kano´ti a-ti arrow do:tr-inf:abs bow do:tr-inf:abs ‘At ﬁve years of age, [the boy] already thinks of ﬁshing with hook, making arrows, making bows.’ (63) tita-ra ka-[a]i rario-ja *rario-ja(-ni)-ß mother:abs-evid go-incomp radio-prop ‘Mother is leaving taking the radio with her.’ (lit. ‘is leaving with [a] radio’)

9

Thus consider the following example involving a concessive/adversative clause: (i)

nami-[j]a i-ßon-Bi-ra kikin itSa very much meat-prop do.intr-ant.ss.a-emph-evid a-ma-jama-kan-ki¯ ¯i-a 1sg-abs do.tr-caus-neg-pl-compl ‘Even though they had a lot of meat, they did not offer me anything.’ ‘They had a lot of meat, but they did not offer me anything.’

282

P. M. Valenzuela

It may be argued that while age is an inherent property of the S participant, a radio is not. However, consider examples (64) and (65), both involving A-agreement: (64) nato honi-n-ra oin-ki¯ sina´-ja-ßon ¯i-a this man-erg-evid 1sg-abs see-compl anger-prop-a ‘This man looked at me [and he was] angry.’ (lit. ‘looked at me with anger’) ainßaon-ronki matS´ıtoma-ßon-ßi wai (65) ßawi¯ Turtle Woman:erg-hsy machete:priv-a-emph chacra:abs a-ka´ti-ai do:tr-pst4-incomp ‘It is said that the Turtle Woman worked in the chacra without a machete.’ All expressed verbs in ex. (62), including Sinan-, are transitive and in principle one would expect the adjunct expression to carry A-orientation rather than S-orientation morphology (i.e. -ßon rather than -ß). It may be the case however that age expressions of this type assume an unexpressed copula verb and S-oriented PA morphology has become ﬁxed in this context. This point, as well as the lack of PA on adjuncts in other types of intransitive clauses, requires further study. 8.4.7 Emphatic pronouns: -ß (S-orientation) vs. -Ø (A- / O-orientation) Emphatic pronouns are used to highlight the fact that the subject participant carried out the action her/himself. The fact that some emphatic pronouns exhibit participant orientation morphology suggests that they play an adjunct function in SK. Emphatic pronouns are formed by a nominative (-n) marked pronoun, followed by the lexicalized emphatic clitic -ßi, and either the morpheme -ß to signal S-orientation or simply zero when the emphatic pronoun is A-oriented.10 Examples (66–8) illustrate S-oriented emphatic pronouns involving different persons: piko´-ki¯ (66) wi¯sna-ra ha-n-ßi-ß Wesna:abs-evid 3sg-nom-emph-s take.out:mid-compl ‘Wesna freed herself/came out herself.’

10 I consider O-orientation is not possible, since I have not been able to ﬁnd any example where an emphatic pronoun refers to an already expressed O argument (cf. exx. 66–70).

Participant agreement in Panoan

283

(67) mato-n-ßi-ni-ß mato rami-t-ai 2pl-nom-emph-lig-s 2pl.abs harm-mid-incomp ‘You yourselves get harmed/You harm yourselves.’11 (68) (i¯-a-ra) ßi¯no-ki¯ ¯i-n-ßi-ß-ßi-Saman 1sg-abs-evid 1sg-nom-emph-s-emph-intens marry.a.man-compl ‘I [woman speaking] got married by myself [i.e. without my parents’ intervention].’ Note that -ßi is attested twice on the emphatic pronoun in (68); as mentioned above, the ﬁrst occurrence is an instance of lexicalization whereas the second one following PA marking is a productive emphatic. In the next sentences, where the emphatic pronoun is A-oriented, addition of the general A-agreement marker -ßon would yield ungrammatical utterances: (69) nokon tita-n-ra ha-n-ßi ha honi ¯i-a 1sg-poss mother-erg-evid 3sg-nom-emph 1sg-abs that man:abs ßi-ma-ki¯ get-caus-compl ‘My mother herself gave me in matrimony to that man.’ no-n (70) no-n-ßi tari a-ßon-jama-kin 1pl-erg-emph tari do:tr-ben-neg-sim.ss.a 1pl-gen no-n ki¯jo-ai ßi¯nßo-ßaon a"ß¯i husband-pl.gen custom:abs 1sg-erg ﬁnish-incomp ‘Because we ourselves do not manufacture taris [i.e. traditional men’s clothes] for [them], we are ﬁnishing with our husbands’ traditional customs.’ In sum, emphatic pronouns refer to an S or A argument (although arguments may be left unexpressed) and combine pronouns marked in the nominative-accusative fashion with the morphemes -ß for S-orientation or -Ø for A-orientation. Whereas the adjunct expressions examined in 8.4.1–7 can be considered as clearly non-clausal, the boundary between intraclausal and interclausal PA becomes blurry when dealing with manner and quantity adjuncts since these may require the presence of an accompanying verb.

11 A ﬂuent bilingual offered the following Spanish translation (emphasis added): ‘Entre ustedes mismos se hacen dan˜o’. The use of the preposition entre in this context signals a kind of manner adjunct function. A similar interpretation can be obtained when translating the quantity adjunct forms hatikaß [all-S] hatı´ßon [all-A] as entre todos; i.e., ‘acting altogether’ rather than simply todos ‘all’.

284

P. M. Valenzuela

8.4.8 Manner: -aß/-i (S-orientation) vs. (-a)-ßon / (-a)-kin (A-orientation) Manner adjuncts are relatively more complex in that they may exhibit up to two sets of agreement markers: -ßon and -kin for A-orientation, as opposed to -aß and -i for S-orientation. The following examples involve the interrogative hawi¯ ki¯ska´ hawi¯ ki¯skat ‘how, why’ exhibiting S-agreement through -aß : (71) hawi¯ ki¯skat-aß-ki mi-a ho-a/ what simil-s-q 2sg-abs come-compl.ptcp piko-t-a? take.out-mid-compl.ptcp ‘How did you come/escape?’ (72) ha-ska-ra i-ki¯-tian-ki ßi¯-kati-kan-ai that-simil-evid be-ant-ds-hsy2 come.nsg-pst4-pl-incomp aSi-ißa, Ashi-loc.person hato-n papa, hato-n tita, hato-n ßaki¯ 3pl-gen father:abs 3pl-gen mother:abs 3pl-gen child:abs mawa´-ki¯-tian die-ant-ds i-ßon hawi¯ ki¯skat-aß-ki mawat-a what simil-s-hsy2 die-compl.ptcp do:intr-ant.ss.a onan-kas-i know-des-sim.ss.s ‘Like this [i.e. because of this], [many people] came to his place to know how their father, their mother, their child had died.’ (Soi Rawa 1995: 25) However, the manner interrogative does not carry S-agreement when the verb encodes a state of being, rather than an event happening or other type of intransitive event: (73) hawi¯ ki¯ska´-ki mato i-t-ai? what simil-q 2pl.abs be-prog-incomp ‘How are you [pl.]?’ Examples (74) and (75) illustrate the second S-agreement marker, -i (which is identical to the S-oriented simultaneous same-subject marker shown in (18b)). In (75), S-agreement occurs on a comparative construction: (74) hawi¯ ki¯ska´t-i mato i-t-ai? what simil-s 2pl.abs do:intr-incomp ‘What is going on with you [pl.]?’

Participant agreement in Panoan

285

(75) ¯i-a-ra itSa ni-ki¯ ßi¯noma ki¯skat-i, 1sg-abs-evid much walk-compl unmarried simil-s ßi¯no-jama[-a] iki ik-aß-ßi¯ ¯i-a do:intr-ant.ss.s-emph 1sg-abs marry.a.man-neg-compl.ptcp aux ‘I have walked a lot like single women [do]; however, I never had a lover.’ The next example contains the expression ni¯ "ti¯ -Saman-aß ‘quietly:’ (76) ha that:abs hawi¯n 3sg.poss

oin-taanan, see-ant.ss nonti canoe:abs

ni¯ "ti¯Saman-aß quiet-intens-s napon to.the.middle

ka-ßon, honi-n go-ant.ss.a man-erg ßoton-ßain-a push-andat-compl.ptcp

iki . . . aux ‘Seeing that, getting closer [acting] very quietly, the man pushed his canoe to the middle [of the river] (Ministerio de Educacio´n and ILV 198212: 3) But given that time-place-manner words may function as predicates taking verb inﬂection directly, under an alternative analysis one could interpret ni¯ "ti¯Saman-aß as ‘behaving very quietly’, -aß being the S-oriented same-subject marker for anterior events (introduced in (17a)).13 That is, as mentioned above, when discussing manner adjuncts we arrive at a domain where the intraclausal and interclausal levels are not neatly distinguished. Let us now turn to A-orientation. The ﬁrst two examples below illustrate the A-agreement marker -ßon. Example (77) belongs to a text where the Shipibo learn from the white monkey how to have sexual intercourse: hawi¯ ki¯ska´-ßon-ki a-ti iki no-n (77) tSaikonı´, brother.in.law:voc what simil-a-q do.tr-inf aux 1pl:gen awin tSota-kin? wife:abs have.sexual.intercourse-sim.ss.a ‘Brother-in-law, how do we have to do to have sexual intercourse with our wives?’ 12 Ministerio de Educacio´n, and Instituto Lingu¨ı´stico de Verano. 1982. Quirica n.8 Libro de lectura n.8, Shipibo-Conibo. Yarinacocha: Centro Amazo´nico de Lenguas Auto´ctonas Peruanas ‘Hugo Pesce’. 13 This alternative interpretation is not plausible for hawi¯ ki¯skat-aß in ex. (71), since in this case the verb a(k)- if present would adopt the form ak- before a sufﬁx beginning with a vowel, thus yielding *hawi¯ ki¯skat-ak-aß.

286

P. M. Valenzuela

(78) ikaßßi rama-tian ßontako-ßaon onan-jama-ki¯ but now-temp unmarried.girl-pl.erg know-neg-compl mapo´ a-ti, clay do:tr-inf.abs mi¯sko´ moa-tian ki¯"ni¯-ßo ha-ska´-Saman-ßon that-simil-intens-a different already-temp design-pl.abs a-ti do.tr-inf:abs ‘But nowadays young girls do not know how to make pottery; similarly, [they do not know how] to make traditional designs.’ Somewhat similarly to ni¯ "ti¯Samanaß in (76), ha-ska´-Saman-ßon may be a shortened form of ha-ska´-a-Saman-ßon (that-simil-do:tr-intens-ant.ss.a), and hence it may be clausal too. Sentences (79–81) illustrate the A-agreement marker -kin: (79) hawi¯ ki¯ska´-kin-ki mi-n pi-ka´Samai? what simil-a-q 2sg-erg eat-des.neg.incomp.ptcp ‘Why don’t you want to eat?’ (80)

ak-a´-ki hakon-Saman . . . hawi¯ ki¯ska´-kin S¯i¯i what simil-a onom do:tr-ant.o ¼ s/a-hsy2 good-nicely piko-t-ai. take.out-mid-incomp ‘ . . . how to varnish the hot pottery so that it turns out pretty’

Analogously to sentence (75), the next example involves a comparative construction. This utterance is a message from Mrs Ranin Ama to indigenous peoples outside Peru: mato joij-ai (81) itSa-ßi-ri¯s ¯i-n much-emph-purely 1sg-erg 2pl.abs tell-incomp mato-n ki¯jo´-ma-jama-kan-wi¯ mato-n a"ß¯i, ﬁnish:mid-caus-neg-pl-imp 2pl-gen custom:abs 2pl-gen hoi; language:abs no-n ak-ai ki¯ska´-kin-rißi a-kan-wi¯ 1pl-erg do:tr-incomp.ptcp simil-a-also do:tr-pl-imp ‘I beg you not to let your customs and your languages disappear; keep them just like we keep ours!’ Again, it is plausible that some instances of (hawi¯) ki¯ska´ (pronounced with a long ﬁnal /a/) are actually followed by the semantically generic transitive verb

Participant agreement in Panoan

287

a(k)- ‘do (tr.),’ and so these similative-marked expressions would turn out to be clausal. This cannot be conclusively determined, however, given that the longer duration of the vowel might (also) be due to the stress. Manner adjuncts may be derived from adjectival roots by addition of the morpheme -n; these words do not exhibit PA: ho-"wi¯! (82) iSto-n quick-adv come-imp ‘Come quickly!’ ßi¯-"wi¯! (83) iSto-n quick-adv bring-imp ‘Bring [it] quickly!’ However, as mentioned earlier, in SK adjunct elements may take verb morphology directly to function as intransitive predicates; hence iSto-n can also be used as the intransitive verb ‘to hurry’, In this case, iSton- takes PA through the same-subject markers described in 8.3.1, as any other verb: pa"ki¯-ki¯ (84) iSto-n-i-ra ¯i-a fast-adv-sim.ss.s-evid 1sg-abs cause.to.fall:mid-compl ‘Being in a hurry, I fell.’

Sinan-ßi¯no-ki¯ (85) iSto-n-aß-ra ¯i-a fast-adv-ant.ss.s-evid 1sg-abs mind-fail-compl ‘Being in a hurry, I forgot [it].’ (86) iSto-n-kin-ra ßi¯-jama-ki¯ ¯i-n fast-adv-sim.ss.a-evid 1sg-erg bring-neg-compl ‘Being in a hurry, I didn’t bring [it].’ Like non-clausal adjuncts in causative clauses involving a transitive base verb (see ex. (46)), iSto-n-ßon in the example below agrees with the causer necessarily: (87) iSto-n-ßon-ra ¯i-a fast-adv-ant.ss.a-evid 1sg-abs ‘Being in a hurry, he didn’t let me

ßi¯-ma-jama-ki¯ bring-caus-neg-compl bring [it].’

Let us now consider the behaviour of the element kikin. Kikin can function as an intensiﬁer modifying an adjective, in which case it occurs unmarked (adjectives may precede or follow their head in SK, but kikin must precede the whole (noun adjective) complex): (88) kikin sina´ [extremely brave] ‘extremely brave’

288

P. M. Valenzuela

(89) kikin ainßo mi¯tsa´ [extremely woman beautiful] ‘extremely beautiful woman’ Another instance in which kikin occurs unmarked is when functioning as an adjectival modiﬁer meaning ‘true, legitimate, elegant, luxurious’: (90) kikin ßoßo [elegant house] ‘good, luxurious house’ However, kikin may also function as an adjunct external to the noun phrase and hence show the agreement morphology under discussion; kikin is glossed here as ‘extremely’. In examples (91–4) kikin shows S-agreement through -i and -aß : (91) kikin-i paki¯-t-a extremely-s cause.to.fall-mid-compl.ptcp ‘He fell hurting himself hard.’ (92) kikin-i ki¯jo-t-a extremely-s ﬁnish-mid-compl.ptcp ‘It ﬁnished [disappeared] completely.’ tSankat-a (93) kikin-aß extremely-s stand-compl.ptcp ‘He is standing very elegantly dressed and/or posing.’ (94) kikin-aß jaka-t-a extremely-s sitting.position:intr-mid-compl.ptcp ‘He is sitting like posing [e.g. crossing his legs].’ A functional distinction becomes obvious when analysing the examples above: whereas kikini can be said to modify the event, kikinaß rather predicates about the S participant. Let us now turn to transitive clauses and A-orientation: (95) kikin-a-kin ki¯jo-a extremely-do:tr-sim.ss.a ﬁnish-compl.ptcp ‘He ﬁnished [it] completely.’ (96) kikin-a-kin ri¯ti¯-a extremely-do:tr-sim.ss.a kill-compl.ptcp ‘He really killed [him].’14 (97) kikin-ßon pi-ki¯ extremely-a eat-compl ‘He ate elegantly (like the mestizos, with tablecloth, silverware, taking out the bones from the ﬁsh . . . .), or ‘He was dressed elegantly and ate.’ 14

A Spanish version offered by a ﬂuent bilingual collaborator is: Lo mato´ bien matado.

Participant agreement in Panoan

289

Again, there seems to be a semantic distinction between the more eventoriented kikin-a-kin and the rather participant-oriented kikin-ßon. Note also in the above examples that kikin requires addition of the semantically generic transitive verb a(k)- before taking -kin but not before taking -ßon. This difference suggests that similative-marked expressions taking -kin might in fact be clausal too, and that at least some constructions taking -kin and -i are functionally distinct from those taking -ßon and -aß (compare also hawi¯ ki¯ska´-ßon ‘how’ and hawi¯ ki¯ska´-kin ‘why’ in sentences (77) and (79)). It is necessary at this point to distinguish between formal and functional participant orientation. Whereas all expressions marked by -(a)ß, -ßon, -i, and -kin are formally participant-oriented, only those marked by -(a)ß and -ßon are semantically participant-oriented; expressions marked by -i and -kin are best considered as semantically event-oriented. Let us also examine the SK equivalent of ‘well’, which is based on the adjectival root hakon ‘good’. When functioning as adjunct, hakon takes PA marking too. The following examples illustrate S-agreement (I have not been able to ﬁnd any semantic distinction between (98a) and (98b)): (98) a. hakon-aß ha-kan-wi¯! good-s exist-pl-imp ‘Live well!’

b. hakon-i ha-kan-wi¯! good-s exist-pl-imp ‘Live well!’

When turning to transitive clauses, it can be seen that the presence of the transitive a(k)- is required before any A-agreement inﬂection is added; thus, the utterances below are to be considered clausal: (99) a. hakon-a-ßon joa´ a-"wi¯ good-do:tr-ant.ss.a cook-imp ‘Cook well!’ b. hakon-a-kin joa´ a-"wi¯! good-do:tr-sim.ss.a cook-imp ‘Cook well!’ For one speaker, the sentences immediately above have slightly different meanings. While the (a) version is interpreted as ‘cook carefully’, the second expression is read as referring to the quality of the food (Mr Sanken Bari, personal communication, 2002). Under the same analysis applied to constructions based on kikin, it could be said that hakon-a-ßon is semantically oriented towards the A participant who is going to cook well or carefully, while hakon-a-kin modiﬁes the event which results in good food and hence is not semantically participant-oriented. But constructions based on kikin and

290

P. M. Valenzuela

hakon still differ in that only -kin-marked forms require the transitive verb a(k)- with the former, while both -kin- and -ßon-marked forms require a(k)with the latter. Let us close this section with the following text example which carries two instances of S-agreement in the same clause: -i on the manner/ degree modiﬁer kikin and -ß on the life-stage adjunct ßaki¯: no-a ß¯ita (100) ha-tian kikin-i ßaki¯-ß-ßi that-temp extremely-s child-s-emph 1pl-abs tooth:abs ki¯jo-t-ai ﬁnish-mid-incomp ‘Then, in spite of being extremely young we lose our teeth.’ 8.4.9 Quantity At least two different PA sub-patterns are found with quantiﬁers. One is illustrated by the quantiﬁer proper hatik hatı´ ‘all’ and the second by the numeral kimiSa ‘three’. 8.4.9.1 hatik hatı´ ‘all’: -aß (S-orientation) vs. -ßon (A-orientation) When in adjunct function, the quantiﬁer hatik hatı´ may take the PA markers -aß and -ßon as shown below: (101) ha-ska´-ki¯-tian, hatik-aß-i ßo-kan-a that-simil-ant-temp altogether-s-emph go.nsg-pl-compl.ptcp iki, ni¯¯i-t-i aux go.up-mid-sim.ss.s ‘Then, everybody started to go up.’ hato-n (102) ha-ska´a-ßon-ki, hatı´-ßon-ßi that-simil-ant.ss.a-hsy2 altogether-a-emph 3pl-gen ßaki¯-ßo-ki notsi-n-kan-a iki. child-pl.abs-hsy2 get.upset-tr-pl-compl.ptcp aux ‘Then, they scolded their children altogether.’ (Soi Rawa 1995: 16) 8.4.9.2 Numeral: -Ø / -i (S-orientation) vs. -ßon / -a-kin (A-orientation) Like manner adjuncts, numerals may also take two different sets of agreement markers. There is a slight difference with respect to manner adjuncts though, since the S-agreement markers on numerals are -Ø and -i (rather than -aß and -i); the A-agreement markers are the same (i.e., -ßon and -kin). As proposed for kikin, alternate S- and A-agreement markers encode an obvious semantic distinction. Consider examples (103–6) involving the numeral kimiSa ‘three’: (103) ha-ßo-ra kimiSa-ßo hiki-ki¯ 3sg-pl.abs-evid three-pl enter-compl ‘They entered in groups of three.’

Participant agreement in Panoan

291

(104) ha-ßaon-ra kimiSa-ßon-ßo naranSa ßi-ki¯ 3sg-pl.erg-evid three-a-pl orange:abs get-compl ‘They got oranges acting in groups of three.’ (105) ha-ra kimiSa-i paki¯-t-a iki 3:abs-evid three-s cause.to.fall-mid-compl.ptcp aux ‘He fell three times.’ pi-a iki (106) ha-n-ra kimiSa-a-kin 3sg-erg-evid three-do:tr-sim.ss.a eat-compl.ptcp aux ‘He ate three times.’ KimiSa-ßo and kimiSa-ßon-ßo are semantically participant-oriented adjuncts, they even take the plural -ßo; kimiSa-i and kimiSa-a-kin, on the other hand, are best analysed as semantically event-oriented. Note also that only kimiSaa-kin is clausal. In sum, manner and quantity adjuncts differ from those described in 8.4.1–7 in the following respects: they allow for the markers -i and -kin in addition to -aß and -ßon, they may require addition of the generic transitive verb a(k)before taking A-agreement thus making the border between clausal and nonclausal adjuncts fuzzy, and they may exhibit a mismatch between formal and semantic participant orientation. Also, it seems justiﬁed to distinguish the quantiﬁer proper hati hatik ‘all’ from numerals such as kimiSa ‘three’. 8.4.10 Comitative and temporal Comitative and temporal adjuncts do not bear PA, which suggests that these are not viewed as participant-oriented. Consider the following intransitive and transitive clauses involving comitatives, which are marked by the postposition ßi¯tan: (107) ¯i-a-ra wai-nko kai papa ßi¯tan. *ßi¯tanaß *ßi¯tani 1sg-abs-evid chacra-all go:incomp father com ‘I go to the chacra with father.’ (108) ¯i-n-ra wai si¯pai papa ßi¯tan. 1sg-erg-evid chacra:abs slash:incomp father com *ßi¯tan-ßon *ßi¯tankin ‘I slash the chacra with father.’ Nevertheless, as shown in Valenzuela (2003: 906–9), comitatives may carry PA in certain genetically related languages. Temporal adjuncts carry no participant agreement morphology in SK (or to my knowledge in any other Panoan language), probably due to the fact that

292

P. M. Valenzuela

these modify the event as a whole rather than any participant (the only known exception involves the equivalent of ‘the day after tomorrow’ in Matses—Valenzuela 2003: 909).15 The next examples involve the time expressions rama ‘now’, ja0 mi¯kiri ‘early in the morning’, and ßakiS ‘one day from today (i.e. ‘yesterday’/‘tomorrow’)’: / ka-[a]i /pi-ai. (109) rama-ra oß a-[a]i now-evid sleep-incomp go-incomp eat-incomp *ramaßon *ramaß *ramai *ramakin ‘Now I am going to sleep/to go/to eat.’ ja|mi¯kiri-ßi (110) ßakiSa ¯i-a one.day.away:evid 1sg-abs early.in.the.morning-emph wi¯ni-t-ai standing.position-mid-incomp ‘Tomorrow I am going to wake up early in the morning.’

ß¯inan wai-nko-ni-a (111) ßakiSa ¯i-n one.day.away:evid 1sg-erg guaba:abs chacra-loc-lig-abl ßi-i ka-[a]i get-sim.ss.s go-incomp ‘Tomorrow I am going to bring guabas from the chacra.’ After providing an overview of clausal, non-clausal, and intermediate syntactic status adjuncts in SK and examining the environments in which PA is (and is not) attested, the next section offers a classiﬁcation of adjunct types based on the PA morphology they may receive.

8.5 Agreement patterns and adjunct types The description of the SK participant agreement system provided above can be summarized and reorganized in the following way. First, three basic agreement patterns occur on adjuncts: Pattern 1 -(a)ß S -ßon A -Ø O Pattern 2 -i S -kin A unattested Pattern 3 -noß S -noßon A n.a. (only as SS markers) Pattern 1 occurs on all adjunct types examined here (except for comitative and temporal), although not all three agreement markers have been attested 15 The fact that temporals do not take part in PA is compatible with the observation that time adverbials in the Australian language Warlpiri optionally agree in case with the clausal subject, while locative or manner expressions receive case agreement necessarily (Simpson 1991: 208).

Participant agreement in Panoan

293

with every adjunct type. Pattern 2 is found on simultaneous same-subject marked clauses and on manner and numeral adjuncts. Certain manner and numeral expressions necessarily require an accompanying verb and hence can be said to be always clausal. In addition, when both Pattern 1 and Pattern 2 agreement markers are possible, it seems that the former conveys semantic orientation towards a core participant whereas the latter is event-oriented. Pattern 3 is a combination of the sequence -no (a former locative) þ the markers in Pattern 1, and occurs on subsequent same-subject marked clauses only. Second, adjuncts may be subdivided into four groups or adjunct types, depending on the agreement markers they receive (8.3.1 and 8.4.1–9); a ﬁfth group is characterized by the inability to take PA marking (8.4.10). I will deal with each adjunct type below. Third, a hierarchical organization is attested so that adjuncts taking the agreement markers of Pattern 2 also take those of Pattern 1, and adjuncts taking the agreement markers of Pattern 3 also take those of Patterns 2 and 1. For example, only same-subject marked clauses take the agreement markers -noß and -noßon of Pattern 3, and they also take those of Patterns 2 and 1. Manner adjuncts take the agreement markers -i and -kin of Pattern 2, and these adjunct expressions also take the agreement markers of Pattern 1. Finally, place, life-stage and other non-clausal adjuncts only take the agreement markers -(a)ß, -ßon, and -Ø of Pattern 1. Pattern 1 > Pattern 2 > Pattern 3 Fourth, it is possible to tentatively correlate the various adjunct types with certain semantic criteria. 8.5.1 Adjunct type 1 Emphatic pronouns and life-stage adjuncts (including proprietive constructions expressing a speciﬁc age) exhibit the peculiarity of being able to take the S-agreement marker -ß without requiring that it be preceded by the ablative -a (which itself is necessarily built upon the locative); I will refer to the relevance of this below, when discussing adjunct types and semantic criteria. However, emphatic pronouns and life-stage adjuncts differ in terms of the A-agreement markers they take; i.e., -Ø for the former but -ßon for the latter (as well as for proprietive) adjuncts. ha-n-ßi-ß ha-n-ßi ßaki¯-ß-ßi ßaki¯-Soko-ßon

[3sg-nom-emph-s] [3sg-nom-emph] [child-s-emph] [child-dim-a]

‘she herself (S-oriented)’ ‘she herself (A-oriented)’ ‘as a child (S-oriented)’ ‘as a little child (A-oriented)’

294

P. M. Valenzuela

8.5.2 Adjunct Type 2 The relevant feature of adjuncts in Type 2 is that these take -aß for S-agreement (i.e. ablative marking or -a must precede the addition of -ß) and -ßon for A-agreement. In this group are found place adjuncts (i.e. locative/ allative and ablative combined), affectives (i.e. PA-marked interessives), and hatı´- hatik- ‘altogether’.

ßoßo-nko-ni-a-ß ßoßo-n-ßon i¯-on-mi¯-a-ß ¯i-on-ßon hatik-aß hatı´-ßon

[house-loc-lig-abl-s] [house-loc-a] [1sg-intrss-loc-abl-s] [1sg-intrss-a] [all-s] [all-a]

‘from the house (S-oriented)’ ‘in the house (A-oriented)’ ‘on me (S-oriented)’ ‘on me (A-oriented)’ ‘altogether (S-oriented)’ ‘altogether (A-oriented)’

8.5.3 Adjunct Type 3 Manner and numeral adjuncts are grouped together here, since these take agreement markers from both Pattern 1 and Pattern 2; while addition of -kin seems to require the semantically generic transitive verb a(k)-, this is not always the case with -ßon: 8.5.3.1 KimiSa ‘three’ Pattern 1 Pattern 2

kimiSa-ø vs. kimiSa-i vs.

kimiSa-ßon kimiSa-a-kin

8.5.3.2 Hawi¯ ki¯ska´ hawi¯ ki¯skat ‘how/why’ Pattern 1 hawi¯ ki¯skat-aß vs. hawi¯ ki¯ska´-ßon Pattern 2 hawi¯ ki¯skat-i vs. hawi¯ ki¯ska´(-a)-kin 8.5.3.3 Kikin ‘extremely’ Pattern 1 kikin-aß vs. kikin-ßon Pattern 2 kikin-i vs. kikin-a-kin 8.5.3.4 Hakon ‘good, well’ Pattern 1 hakon-aß vs. hakon-a-ßon Pattern 2 hakon-i vs. hakon-a-kin 8.5.4 Adjunct Type 4 Under Adjunct Type 4 are grouped same-subject marked clauses depicting events that are anterior, simultaneous or subsequent with respect to the events in their matrix clause (section 8.3.1). All three agreement patterns are involved.

Participant agreement in Panoan

295

8.5.5 Adjunct Type 5 Comitative and temporal adjuncts do not receive any participant agreement morphology and hence are grouped together under Adjunct Type 5. 8.5.6 Adjunct types and semantic criteria As anticipated already, it is possible tentatively to correlate the formally based adjunct types established above with functional criteria such as degrees of participant-orientedness vs. event-orientedness (Schultze-Berndt and Himmelmann 2004); this proposal is based on the semantic distinction made for certain adjuncts that take agreement patterns 1 and 2 (i.e. those in Adjunct Type 3). Emphatic pronoun and life-stage expressions (i.e. those under Adjunct Type 1) can receive the marker -ß directly. -ß has been reconstructed as a former case-marker conveying S function (Valenzuela 2003: 918–23; see also section 8.6 below). Since Adjunct Type 1 members are/were able to receive the case-marking of their S controller directly without requiring locative marking, they may be considered as especially participant-oriented.16 Place (locative/allative and ablative) and affective expressions, as well as ‘altogether’ (i.e. Adjunct Type 2 members) are also interpreted as semantically participant-oriented, but these cannot take the marker -ß directly. Manner and numeral adjuncts can be said to be semantically participantoriented (when taking -aß and -ßon) or event-oriented (when taking -i and -kin). Finally, comitative and temporal adjuncts, as well as life-stage expressions marked by -tian, do not take any PA morphology.17 The proposal above can be summarized as the continuum in Figure 8.1 (inspired by Schultze-Berndt and Himmelmann 2004: 119–23). PO stands for Participant-Orientedness, EO for Event-Orientedness, and AT for Adjunct Type. Although the position occupied by certain adjunct types on this continuum might not seem obvious from a functional perspective, the fact 16 There is, however, an alternative analysis to account for the seeming privileged status of life-stage adjuncts, which do not require the presence of the ablative -a before taking the S-agreement marker -ß. In SK a noun may function as an inchoative predicate by taking verbal morphology directly. For example, oß¯i ‘moon’, oß¯i-ki¯ [moon-compl] ‘turned into moon’, oß¯i-aß. . . . [moon-ant.ss.s] ‘after turning into moon . . . ’. A sequence such as ßaki¯-aß [child-aß] is interpreted as ‘after becoming a child/after behaving as a child’. Therefore, it may be the case that life-stage adjuncts are assigned the S-agreement marking -ß instead of -aß in order to avoid conﬂation and ambiguity between the inchoative and the life-stage readings. 17 If this analysis is extended to dependent clauses, it could be proposed that anterior and subsequent same-subject marked clauses are to be interpreted as semantically participant-oriented, whereas simultaneous same-subject clauses are rather event-oriented. Different-subject clauses marked by -tian are not participant-oriented in either formal or semantic terms. Nevertheless, samesubject marked clauses are kept here as a single category.

296

P. M. Valenzuela

PO

EO

AT1

AT2

- vs. - on

-a vs.- on

(-a ) vs. (-a)- on -a vs. - on -i vs. -a-kin -i vs. -kin -no vs. -no on

no PA

Emphatic pronouns Life-stage

Place Affective ‘altogether’

Manner Reason Numeral

Comitative Temporal

AT3

AT4

SS clauses

AT5

Figure 8.1 Participant-orientedness continuum

that life-stage and temporal adjuncts are found at the most participantoriented and the most event-oriented ends, respectively, is certainly very revealing.

8.6. Conclusions and ﬁnal remarks In this chapter I have offered a detailed treatment of the PA system of SK. I have shown that certain adjuncts carry a distinct inﬂectional morphology in agreement with the syntactic function of the participant they are predicated of. There are three basic agreement patterns that occur on adjuncts, the overt markers being -aß, -i, and -noß for S-orientation versus -ßon, -kin, and -noßon for A-orientation; O-orientation, where possible, is always encoded by zero. PA takes place at the interclausal and intraclausal levels. Interclausal PA is found on dependent same-subject marked clauses, which may exhibit all three agreement patterns. Intraclausal PA occurs on word-level and phrasal place, manner, quantity, life-stage, affective, proprietive and privative adjuncts. PA is also found on emphatic pronouns coreferential with an S or A argument, which indicates that these may function as adjuncts. The emphatic clitic -ßi is systematically attested with three different kinds of non-clausal participant-oriented adjuncts: complement of interest, life-stage, and emphatic pronouns (see ‘restrictive’ markers in the introduction by Himmelmann and Schultze–Berndt, p. 65, this volume). PA in complex monoclausal sentences involving causatives and applicatives has also been examined. In these constructions, an adjunct cannot be oriented towards a base verb argument, semantic agreement on ablatives controlled by an intransitive base verb subject being the exception (8.4.3). The boundary between interclausal and intraclausal PA is not always evident. Manner and numeral expressions exhibit intermediate syntactic status, since they may require the presence of the semantically generic transitive verb a(k)- before receiving A-orientation marking. In addition,

Participant agreement in Panoan

297

given that manner words are capable of functioning as predicates by taking verb inﬂection directly, alternative analyses of these expressions as either non-clausal or clausal adjuncts are possible, as discussed with regard to ni¯ "ti¯-Saman-aß in 8.4.8. Unlike other non-clausal adjuncts which only allow for the overt PA markers -aß and -ßon, manner and numeral adjuncts also take -i and -kin. A functional distinction has been found in some of these instances: expressions marked by -i and -kin are best interpreted as semantically event-oriented whereas those taking -ßon and -aß are participant oriented. Therefore, it is necessary to distinguish between formal and functional participant orientation. Whereas all expressions marked by -(a)ß, -ßon, -i, and -kin are formally participantoriented, only those marked by -(a)ß and -ßon are semantically participantoriented. In other words, manner and numeral adjuncts may exhibit a mismatch between formal and semantic participant orientation. I have also proposed a classiﬁcation of adjunct types based on the PA morphology they allow. Of the three basic agreement patterns that occur on adjuncts, Pattern 1 is attested on all adjunct types except for comitative and temporal, Pattern 2 is found on same-subject marked clauses and on manner and numeral adjuncts, and Pattern 3 is restricted to same-subject marked clauses. These agreement patterns are hierarchically organized so that adjuncts taking the agreement markers of Pattern 2 also take those of Pattern 1, and adjuncts taking the agreement markers of Pattern 3 also take those of Pattern 2 and 1: 1 > 2 > 3. Adjuncts have been subdivided into four groups or adjunct types, depending on the agreement markers they receive; a ﬁfth group is composed of comitatives and temporals which cannot take PA marking. The ﬁve formally based adjunct types have been tentatively correlated with functional criteria, namely degrees of participant orientation versus event orientation. Emphatic pronoun and life-stage expressions are considered as the most participant-oriented given their ability to receive the marker -ß (reconstructed as a former case-marker conveying S function, see below) directly. Place and affective expressions, as well as ‘altogether’, are also semantically participant-oriented, even though they require the ablative marker -a before taking -ß. Manner and numeral adjuncts can be said to be semantically participant-oriented (when taking -aß and -ßon) or eventoriented (when taking -i and -kin). Finally, comitative and temporal adjuncts do not take any participant agreement morphology. As mentioned at the end of section 8.5, although the position occupied by certain adjuncts on the Participant-Orientedness–Event-Orientedness continuum in Figure 8.1 is not semantically transparent, the fact that life-stage and temporal adjuncts are found at opposite ends is compatible with the proposal outlined here.

298

P. M. Valenzuela

In the introduction to this chapter I mentioned that Panoan PA diachronically involves case agreement. Valenzuela (2003) carried out a comparative study examining languages from ﬁve different branches of Panoan and, among other ﬁndings, concluded that the PA morphology resulted from the combination and fusion of two different layers of case-marking: peripheral cases þ core cases. The ﬁrst layer involved the following peripheral markers: *-ßo sequential (possibly ablative), *-/a ablative, -ki locative/ oblique, *-no locative/allative. The second layer involved the core casemarkers *-n, *-ø, and *-ß for the A, O, and S functions, respectively. Reﬂexes of the peripheral markers are easily found in the current languages, whereas the tripartite case-marking system in question is attested in a couple of languages such as Amawaka and Kashibo-Kakataibo, which belong to two different branches of the family. Valenzuela (2003) also hypothesizes that the origin of the Panoan PA system is to be found at the interclausal level. Although the Panoan PA system can be viewed as an instance of case agreement on adjuncts, it must be highlighted that we are dealing with a crosslinguistically very unusual situation, where a nominal controller determines the shape of adjunct elements without affecting the shape of its modiﬁers inside the NP. In fact, PA constitutes the only domain of agreement in Panoan languages (except for what has been called ‘transitivity agreement’, a set of systematic covariance which mainly depend on the transitivity status of the verb and target the VP—Valenzuela 2003: 587–607). PA in Panoan constitutes a previously undescribed instance of split ergativity in which a tripartite case-marking system is retained/found on adjuncts, in combination with different types of split-ergative systems on the NP. In turn, the Panoan tripartite system is unusual since both A and S arguments receive overt marking. In this way, it can be considered a truly tripartite system in that it cannot be seen as the result of the combination of nominative-accusative and ergative-absolutive patterns; if this were the case the S argument would be zero-marked (as described for some Aboriginal Australian languages—Austin 1981b).

9 Secondary predicates and adverbials in Nilotic and Omotic: a typological comparison AZEB AMHA AND GERRIT J. DIMMENDAAL

9.1 Introduction From a typological point of view, verb-initial and verb-ﬁnal languages are known to constitute each other’s mirror image. Crosslinguistically, the position of the verb relative to the subject and object correlates, in statistically signiﬁcant ways, with a number of other syntactic properties. Amongst the set of widespread implicational universals in verb-initial as against verb-ﬁnal languages are such properties as the preponderance of prepositions as against postpositions and the tendency for auxiliaries to precede rather than follow the verb. Moreover, complement clauses tend to precede the matrix verb in verb-ﬁnal languages, whereas they follow the latter in verb-initial languages. There also appears to be a tendency towards distinct coding mechanisms for the expression of morphosyntactic relations in these two language types. As argued by Nichols (1986: 104), in her study on the morphological marking of grammatical relations, the head-marked clause pattern favours verb-initial order, whereas dependent-marking strategies favour verb-ﬁnal order. The authors would like to express their gratitude to Nikolaus Himmelmann and Eva Schultze-Berndt for extensive comments on an earlier draft of this chapter. We would also like to thank the participants of the Workshop on Depictive Secondary Predicates (Ruhr-Universita¨t Bochum, 7–9 June 2001) for their critical comments and questions. We also gratefully acknowledge the help of Rafed El-Sayed for data on modern Egyptian Arabic. Azeb Amha is a native speaker of the Semitic language Amharic and the Omotic language Wolaitta; data on these two languages are based on her native speaker intuitions. Data on other languages are based on ﬁeldwork carried by the authors at earlier points, as well as on sources quoted in the paper. The present contribution emerged from a research project on ‘Participant marking in African languages’, which was made possible through grant number HE 574/31-1 from the DFG (the German National Science Foundation). We would like to express our deep gratitude to the DFG for this generous support.

300

A. Amha and G. J. Dimmendaal

A barely studied domain in this respect concerns the expression of depictive secondary predication in languages belonging to these two polar types. This topic will be central to the present contribution. First, we intend to investigate the structure of such clauses in verb-initial Nilotic languages. Next, we will have a closer look at secondary predication in Omotic languages, which are typically verb-ﬁnal. More speciﬁcally, we will be addressing the following questions: 1. Which syntactic categories play a central role in the expression of depictive secondary predicates in these two language families? 2. What kind of intragenetic variation do these two language families manifest? 3. In what ways are adverbial constructions and secondary predicates in competition with each other? Below, we will ﬁrst discuss secondary predication in Turkana and a number of other Nilotic languages (section 9.2), followed by a survey of this syntactic phenomenon in Wolaitta and other Omotic languages (section 9.3). As shown in these two sections, secondary predication in these two language families involves different syntactic categories whose position relative to the main clause also varies. As further argued in these sections, a lexical dimension needs to be brought in as well, in order to explain differences between genetically related languages which are similar from a typological point of view. In spite of major intergenetic differences between Nilotic and Omotic, both genetic groupings also share a common strategy, as argued in section 9.4. The question to what extent these language families are typical for the syntactic type to which they belong can only be answered in full detail once a fully-ﬂedged study has been carried out on secondary predication in other verb-initial and verb-ﬁnal languages, an ultimate goal lying beyond the scope of the present contribution. The ﬁnal section (9.4), however, constitutes an initial attempt to compare our preliminary ﬁndings with secondary predication in a number of other language families, in order to be able to distinguish wider and more universal properties from language-speciﬁc ones.

9.2 Depictives and adverbs in Nilotic 9.2.1 Turkana Within Nilo-Saharan, the genetic unit to which Nilotic languages belong, a verbinitial syntax is restricted geographically to languages spoken in the eastern and south-eastern zones of this phylum. More speciﬁcally, this type of syntactic structure is found in the Kadu group (Sudan), Surmic (Sudan and Ethiopia),

Secondary predicates and adverbials in Nilotic and Omotic

SUDA N

ET HI OP I A

DEM . RE P . C ON GO

UGAND A K EN Y A

Nilotic

TANZANIA

Omotic

Map 9.1

Distribution of Nilotic and Omotic languages

301

302

A. Amha and G. J. Dimmendaal

Kuliak (Uganda), and Nilotic (Congo, Ethiopia, Kenya, Sudan, Tanzania, Uganda); see also Map 9.1. The present contribution focuses on the Nilotic group within the set of verb-initial Nilo-Saharan languages, because the syntactic phenomenon of secondary predication is understood best with respect to this subgroup. We begin our discussion with an investigation of the Eastern Nilotic language Turkana (spoken in Kenya). Turkana is a verb-initial language with a fairly extensive head-marking strategy at the clausal level, as shown in Dimmendaal (1983a). The verbal root in Turkana may be expanded with pronominal markers for subject and object (in the case of transitive verbs); moreover, the same language has a verbal dative marker, as well as markers for direction (ventive and itive), with instrumental marking on verbs being restricted to subordinate clauses. Turkana also has dependent-marking strategies at the clausal level, involving a case distinction between Nominative and Absolutive as core case distinctions, as well as Locative and Instrumental case as peripheral case distinctions (cf. Dimmendaal 1983a; 2006). Secondary predicates always follow the main clause in Turkana and other verb-initial Nilotic languages; this also appears to be the case in Nilotic languages with SVO or S AUX OV constituent order, the other common constituent order types in this family. One frequent type of secondary predication in Turkana involves a conjugated verb expressing a simultaneous state or condition (‘blind’ in (1b) ); such verbs are formally identical to main clause predications of this type (1a): (1) a. e-m Ud Uk-aan-a` 3-be.blind-hab-mid ‘It/(s)he is blind.’ e-mUdUk-aan-a` b. ka`iˆo` a-doun-ere ikO~kU why 3.pst-born-pass.sbjv child:abs 3-be.blind-hab-mid ‘Why was the child born blind?’ Whereas pronominal subject marking on the verb is obligatory for conjugated verbs in Turkana, the presence of syntactic subjects and objects is governed by pragmatic considerations (Dimmendaal 1983b). Thus, in example (1a) a syntactic (pronominal) subject may be added, but this is not required from a syntactic point of view. In secondary predicates, however, such independent pronouns are excluded (compare also Gu¨ldemann, Ch. 10, this volume, for a description of similar syntactic properties as found in the Bantu language Shona). Thus, in example (1b) above, involving passive voice marking on the main verb (or more properly, impersonal active voice marking, as there is no object-to-subject movement involved in Turkana), the secondary predicate

Secondary predicates and adverbials in Nilotic and Omotic

303

refers to a state the object in the main clause (‘child’) ﬁnds itself in; here, ‘child’ may not be referred to by way of an independent pronoun in the secondary predicate. An expression such as ‘being born in a blind way’, i.e. a construction about the manner in which an action or process takes place, would not make sense from a semantic point of view. Other verbs or adjectives, however, may appear either as secondary predicates or, alternatively, as derived manner adverbs, with distinct formal properties. Most manner adverbs in Turkana are derived in fact either from adjectives or from relative clauses, which in turn are derived from verbs. Thus, the stative verb root -ron ‘be bad’ occurs as an attributive clause form -aronon(ı`) when modifying a noun. This latter form constitutes the basis for the manner adverb, nı-aronon(ı`) ‘badly’. (Word-ﬁnal vowels between parentheses in the Turkana examples in the present contribution are de-voiced before pause, and extra short elsewhere.) The following examples illustrate the alternative use of this stative verb ‘be bad’, ﬁrst as a depictive secondary predicate involving object orientation (3a), and next as a manner adverb (3b). e`-rono` 3-be.bad:sg ‘It/(s)he is bad.’ (3) a. e`-pe`s-e-te` Ne`si e`-rono` 3-kick-asp-pl 3sg.abs 3-be.bad:sg ‘They kick him/her/it badly.’ b. e`-pe`s-e-te` Ne`si ni-aronon(ı`) 3-kick-asp-pl 3sg.abs rel-badly ‘They kick him/her/it in a bad way.’

(2)

In example (3b), the (derived) manner adverb expresses information on the way in which a particular act is carried out, rather than a result or state the object or goal ﬁnds himself in, as in (3a). There are only few non-derived adjectives in Turkana (i.e. modiﬁers occurring in their morphologically simplest form when modifying a noun), most property concepts being expressed by way of stative verbs, as in (1) to (3), and in the next example: (4)

a`-ˆa`m-e`-te kec(ı`) akiriN e`-jon 3.pst-eat-asp-pl 3pl.nom meat:abs 3-be.raw ka`iˆo` pe-a-ye`n-e`-te aki`m(i) because neg-3.pst-know-asp-pl ﬁre:abs ‘They ate the meat raw, because they did not know ﬁre.’

304

A. Amha and G. J. Dimmendaal

Note also that Turkana uses a distinct consecutive or narrative mood form for verbs in order to express a sequence of actions, e.g. in storytelling (Dimmendaal 1983a: 176); the latter verb type is accordingly formally distinct from secondary predicates such as ‘(be) raw’ in example (4). When non-derived adjectives are used in a predicative sense in Turkana, they require a habitual extension. The habitual marker could be analysed as a kind of verbalizer. It should be noted, however, that the same sufﬁx may also be used with basic verbs in order to express a habitual state. (5) a. emoN lo-mug(ı`) bull agr.m-brown ‘a/the brown bull’ b. e`-mu`gy-aan-a` 3-brown-hab-mid ‘(S)he/it is brown.’ When such adjectives are used as secondary predicates, however, no pronominal reference marker or middle voice marker is required: (6)

k-i`tUm mu`gy-aa`n imp-paint brown-hab ‘Paint it brown!’

Here, the secondary predicate does not refer to a particular style or manner in which the act of painting takes place; rather, it expresses a resultative meaning affecting the pronominal object (which is expressed by way of a zero anaphor). In other words, the habitual sufﬁx on the secondary predicate is compatible with depictives (as in 1b) as well as with resultatives (as in 6). Turkana distinguishes between verbal and non-verbal sentences. The latter typically involve predicative constructions expressing identity. They consist of a subject plus predicative element (a noun (phrase) or some nominal modiﬁer), linked by an auxiliary verb (glossed as ‘be’), when reference is made to a temporal state other than the present. A syntactic subject pronoun is required when a categorical statement is made; with thetic statements on the other hand, subject pronouns may be omitted. e-ra`i epeyo`non(ı`) (7) a. (Ne`si) 3sg.abs 3-be guest:abs ‘(S)he is a guest.’ b. (Ne`si) a`-ra`i epeyo`non(ı`) 3sg.abs 3.pst-be guest:abs ‘(S)he was a guest.’

Secondary predicates and adverbials in Nilotic and Omotic

305

This type of non-verbal construction is also found in secondary predicates, more speciﬁcally in clauses expressing a role, function or life stage. In these types of secondary predicates, there is no independent tense marking on the auxiliary verb; the latter obligatorily takes [þ past], rather than [past], tense marking. (8) a`-bU k-ı`bo`y-ı`kin-(ı`) kide a` awuy(e`) 3.pst-come 3.subsec-stay-dat-mid west on house:gen a`-ra`i epeyo`non(ı`) 3.pst-be guest:abs ‘(S)he sat down on the west side of the house as a (proper) guest.’ (9) etUmiar-i`t akitOi a`-ra`i Ebela` 1sg:use-asp wood:abs 3.pst-be club:abs ‘I am using this piece of wood as a club.’ Note also that Turkana does have a separate Instrumental case form; unlike languages such as Russian (Nichols 1978a), however, Turkana or other Nilotic languages never appear to use this peripheral case form in secondary predicates. As shown in section 9.3 below, Omotic languages do use case-marked nominals for depictive expressions. 9.2.2 Depictives and adverbs in Kipsikiis The verbal strategies discussed with respect to secondary predication in Turkana above are found elsewhere in Nilotic as well. Differences in the actual formal expression of these notions appear to be largely due to diverging strategies for person-marking and other inﬂectional properties of verbs; alternatively, they are the result of differences in the formal expression of manner adverbs. For example, in Southern Nilotic Kalenjin, a cluster of closely related dialects all of which have a verb-initial syntax, secondary predication also involves conjugated verbs following the main predicate. But whereas in Eastern Nilotic Turkana the third person marker is E-/e-, Kipsikiis and other Southern Nilotic languages use either a zero marker or a preﬁx kO-/ko-; the vowel quality of the preﬁx is determined by vowel harmony principles in these languages, and depends on the quality of the following root vowel. Rottland (1982: 228) reconstructs such a third person preﬁx *kO- for Proto-Kalenjin and Proto-Southern Nilotic (p. 243). As further pointed out by Rottland (1982: 122), this third person preﬁx alternates with zero synchronically in Southern Nilotic languages. The exact syntactic and pragmatic distribution of these alternative strategies is not clear, according to the same author (p. 228). One dependent clause type, however,

306

A. Amha and G. J. Dimmendaal

where kO-/ko- is obligatory involves secondary predicates expressing resultatives. The following examples illustrate the use of this person marker in the Kalenjin lect Kipsikiis (data from Toweett 1979): (10) a. nwak-iit 3:be.short-asp ‘(S)he/it is short.’ b. til kO-nwak-iit cut 3-be.short-asp ‘Cut it [to be] short.’ Objects as well as subjects in Kipsikiis may be expressed by way of zero anaphora (as in Turkana). The secondary predicate, however, requires a preﬁx kO-/ko-. (11)

ra´t ko`-k`i`im-´i´it tie 3-be.strong-asp ‘Tie it [to be] strong/tie it solidly.’

(12)

am ko-ya¨ eat 3-be.good ‘Eat [it] properly.’

A closer inspection of constructions with the preﬁx kO-/ko- plus verb stem in Kipsikiis shows that the function of such forms is not only to describe the condition or state that the agent or patient of the main clause ﬁnd themselves in; their semantic and functional range includes the description of processoriented verbs involving intensiﬁcation in contexts where languages like Turkana, for example, would use a manner adverb instead: Kipsikiis (13)

ru´ ko-rw-ak sleep 3-sleep-stat ‘Sleep well.’

Turkana (14)

to-per lO~kOjOkO~n(i) imp-sleep well ‘Sleep well.’

These examples show that the referential domain for either (basic or derived) manner adverbs or secondary predicates is not necessarily identical between genetically related languages which are typologically similar otherwise. As example (13) further shows, such expressions may involve a repetition of verbs.

Secondary predicates and adverbials in Nilotic and Omotic

307

This makes such constructions similar to cognate object constructions in other languages. Note, however, that in Kipsikiis the manner expression involves a conjugated verb, rather than a noun; this verb may also take verbal extensions, as the following example shows: (15) a-sir-chinii ko-sir-chiin-ak 1sg-write-dat.mediative.asp 3-write-dat.mediative-stat ‘I write well for him/her.’ Kipsikiis, and in fact most other Southern Nilotic languages, have relatively few non-derived manner adverbs (Toweett 1979: 370). There is one productive strategy of nominal derivation involving two gender preﬁxes, kip(feminine) and kaa- (masculine—restricted in application), which also plays a role in the formation of manner adverbs. The two preﬁxes (or proclitics) may be added to nouns, adjectives, or verbs in Kipsikiis and other Southern Nilotic languages, in order to express either a condition, a state, a comparison with the subject or object of the main clause, or a manner in which an event is carried out. In other words, the derived forms are used in general adjunct constructions involving both participant-oriented and process-oriented ⁄ Ok ‘hen, fowl’, interpretations. This is illustrated below with the noun I ng‘k the adjective kOOi ‘tall, long’, and the verb -chuulya ‘go past the destination’: (16) Ng‘ook kip-ing‘kOk 3:be.selﬁsh der-chicken ‘(S)he is as selﬁsh as a chicken.’ (17) tyen-i kip-kOOi 3:sing-asp der-long ‘(S)he sings always.’ (18) mnyan-i kip-chuulya 3:sicken-asp der-go.past.destination ‘(S)he sickens continually.’ A number of such derived forms have become lexicalized as nouns; thus, kip-sirich means both ‘rhino’ and ‘as a rhino’: (19) kiim kip-sirich 3:be.strong rhino ‘(S)he is strong as a rhino.’ Obviously, it is not possible in a paper of this length to survey secondary predication for all Nilotic (or Omotic) languages, also because in-depth studies on the structure of secondary predication are lacking for several of these languages. Nevertheless, a general pattern emerges from the existing

308

A. Amha and G. J. Dimmendaal

literature, suggesting that the competition in the semantic range of participantoriented adjuncts between secondary predication and the use of manner adverbs is widespread in the Nilotic family. It is further observed that systems are never isomorphic between genetically related languages for at least two reasons. First, there is an important lexical dimension, involving the number of manner adverbs available in individual languages. Consequently, there is an oscillation between adverbial constructions involving manner adverbs, on the one hand, and depictive constructions (which in these languages are verbal in nature), on the other. Second, the referential domain for manner adverbs may change (e.g. their compatibility with animate versus inanimate or concrete versus abstract subjects or objects, or stative versus dynamic verbs), thereby affecting their distribution. 9.2.3 Depictives and adverbs in Bari There appears to be an additional historical process causing divergence between genetically related languages. In a number of Nilotic languages adjectives and nouns may be used either in an adverbial sense or as secondary predicates. This may be illustrated for Bari, an Eastern Nilotic SVO language. As shown by Spagnolo (1933: 196–9), Bari has a few simple, non-derived adverbs (e.g. parik ‘much, very’, do¨ ‘a little’). Parallel to Southern Nilotic Kipsikiis, nominal constructions in Bari may commonly be used in an adverbial sense in order to express information on the manner in which a particular event takes place (data from Spagnolo 1960): (20)

-kOndya witiwiti turn:antip whirlwind ‘to pirouette, circle around others [like a whirlwind]’

However, such nominal forms in Bari may also be preceded by the (third person) predicative marker a, when assigning a speciﬁc property (state or condition) to a (subject or object) participant of the main predicate: a witiwiti ‘turning, like a whirlwind’ (Spagnolo 1960: 325). This marker is cognate historically with the third person past tense marker a- in Turkana, as found in examples (1b, 7b, 8). The marker a is attested independently as a predicative marker with nominal and adjectival complements synchronically in Bari (Spagnolo 1933: 237, 249 and passim). Compare also the following main predication and the secondary predication expressing a participant-related meaning: (21) a. nan a Nutu’ 1sg pred man ‘I am a man.’

Secondary predicates and adverbials in Nilotic and Omotic

309

b. kitani a Nutu’ work pred man ‘Work like a man!’ Accordingly, in Bari (and possibly other Nilotic languages) adjuncts may have a depictive reading or an adverbial function. Rather than changing the categorical status (as in Turkana above, where verbs are changed into manner adverbs in order to modify an event described by the verb in the main clause), the oscillation between a depictive and an adverbial function is expressed through the presence vs. absence, respectively, of a predicative marker a in Bari, where such constructions always seem to involve nouns and adjectives.

9.3 Depictives and adverbial clauses in the Ometo branch of Omotic The Omotic branch within Afroasiatic is usually divided into an Eastern and a Western branch. The Ometo languages constitute a major subgroup within Western Omotic (see Hayward 1990). In the present contribution we are concentrating on two Ometo languages, Wolaitta and Maale. Our main concerns are the so-called ‘simultaneous’ verb constructions and ‘converbs’ (see Adams 1983: 140–4), also known as ‘imperfect gerund’ and ‘perfective gerund’ constructions (Lamberti and Sottile 1997: 222–4). In languages like Wolaitta these two constructions are treated as (adverbial) subordinate forms along with the conditional, concessive, and complement verb forms. The two constructions in question are formally and functionally similar to some extent. For the purposes of this paper we will use the labels simultaneous converb and anterior converb (borrowed from Haspelmath and Ko¨nig 1995) to identify these related structures. We will show that for both simultaneous and anterior converb constructions, same-subject and differentsubject converb clauses can be distinguished, and that these differ considerably in their morphosyntactic make-up. Both converb constructions, and in fact all dependent clause constructions, precede the main clause in Omotic languages. Below, we argue that in Wolaitta same-subject converbs are genuine depictives, but that different-subject converbs are adverbials or general adjunct constructions. In some cases, however, same-subject anterior converbs have an adverbial interpretation. This phenomenon presents further evidence in favour of the claim that even in languages with a morphological distinction between depictive and adverbial constructions, these two constructions may freely encroach upon each other’s territory, as argued by

310

A. Amha and G. J. Dimmendaal

Schultze-Berndt and Himmelmann (2004). In our short survey below we will also show that, in spite of the fact that Maale and Wolaitta are relatively closely related Ometo languages, they are different in their expression of participant orientation. 9.3.1 Simultaneous converb clauses in Wolaitta In Wolaitta, there are two types of simultaneous construction involving ‘same-subject simultaneous clauses’ and ‘different-subject simultaneous clauses’. These dependent clause types strictly occur preceding the main clause. The subject, i.e. the participant taking the semantic role of ‘actor’ or ‘agent’, takes Nominative case in Wolaitta and other Ometo languages. In same-subject simultaneous clauses, the subjects of the dependent verb and the main verb are coreferential. In such clauses, the subject is expressed only once, either overtly as in (22), or covertly (if it is understood from the context). (22)

haatta-ı´ miicc-iı´ddı´ m-ee´si water-m.nom laugh:caus-sim.ss eat-3m.sg.ipfv ‘Tragedy occurs as one is enjoying oneself and thus is not alert.’ (lit. ‘Water drowns one [while] making [one/the victim] laugh.’)

The dependent verb in simultaneous clauses is inﬂected with a morpheme -iı´ddi when the subject is third person masculine singular (22), or when it is plural (23a); for other singular subjects, the dependent verb is marked by -aı´dda´ (23b). (23) a. /as-at-ı´ harg-iı´ddı´ /oott-o´sona person-pl-pl.nom be.sick-sim.ss work-3pl.ipfv ‘The people work while they are sick.’ b. miSir-ı´ya´ ka´tta gaac}c}-aı´dda´ ye´t}t}-ausu woman-f.nom grain:abs grind-sim.ss sing-3f.sg.ipfv ‘The woman sings while grinding grain.’ Thus, apart from marking simultaneity, the morphemes -iı´ddi and -aı´dda also indicate inﬂectional properties of the subject with respect to number and/or gender. Since the co-occurring main verb is distinctly marked for each person, no ambiguity arises from the reduced inﬂection in the dependent verb. (See Table 9.1 for a full inﬂectional paradigm.) The same-subject simultaneous clauses illustrated in (22) and (23) are genuine depictives, as they satisfy all the criteria suggested in the literature for distinguishing depictives from other constructions. These include, among

Secondary predicates and adverbials in Nilotic and Omotic

311

Table 9.1. Wolaitta person marking on verbs Secondary predicate Person

1sg 2sg 3f.sg 3m.sg 1pl 2pl 3pl

Main predicate

Same-subject

Different-subject

sim

ant

sim

ant

pfv

ipfv

fut

-aı´dda´ -aı´dda´ -aı´dda´ -ı´´ıddı´ -ı´´ıddı´ -ı´´ıddı´ -ı´´ıddı´

-a´da´ -a´da´ -a´da´ -ı´dı´ -ı´dı´ -ı´dı´ -ı´dı´

-ı´Sin -ı´Sin -ı´Sin -ı´Sin -ı´Sin -ı´Sin -ı´Sin

-ı´nı´ -ı´nı´ -ı´nı´ -ı´nı´ -ı´nı´ -ı´nı´ -ı´nı´

-aa´si -a´dasa -iı´si -aa´su -ı´da -ı´deta -ı´doso na

-aı´si -aa´sa -ee´si -au´su -oo´si -ee´ta -oo´so na

-ana´ -ana´ -ana´ -ana´ -ana´ -ana´ -ana´

Table 9.2. Simultaneous constructions and converbs in Ometo compared on the basis of the first six criteria for distinguishing depictives (Schultze-Berndt and Himmelmann 2004: 77f.)

Criterion 1 ¼ cotemporality 2 ¼ control 3 ¼ predicates on controller 4 ¼ not obligatory 5 ¼ not modiﬁer of controller 6 ¼ nonﬁnite

Simultaneous converbs

Anterior converbs

SS þ þ þ þ þ þ

DS /þ þ þ þ

DS þ þ þ þ

SS þ/ þ þ þ/ þ þ

others, the fact that the dependent verb is cotemporaneous with the main predicate, it is obligatorily controlled by the subject of the main predicate and shows agreement with it, and it makes a distinct predication about the subject from that expressed by the main verb. (See Table 9.2 for a comparison with other constructions.) In different-subject simultaneous clauses, the dependent verb is invariably marked by -ı´Sin(i), whereas the co-occurring main verb is distinctly inﬂected for each person. (24) miSir-ı´ya´ ka´tta gaac}c}-ı´Sin bita´nee mı´tta woman-f.nom grain:abs grind-sim.ds man:m.nom wood:abs k}e´r-eesi split-3m.sg.ipfv ‘The man splits wood, while the woman is grinding grain.’

312

A. Amha and G. J. Dimmendaal

The event expressed by the dependent verb in (24) occurs simultaneously to that expressed by the main verb. Unlike the dependent verb in same-subject simultaneous clauses illustrated in (22) and (23), however, it has its own overt (disjunct) subject, itself non-coreferential with the subject of the main verb in (24); moreover, the dependent verb does not show agreement nor does it predicate on any of the core arguments of the main verb. Thus, sentence (24) is not depictive. Rather, this sentence represents an adverbial clause which is in paradigmatic contrast, for example, with the conditional in (25). (25)

miSir-ı´ya´ ka´tta gaac}c}-ı´kko bita´nee mı´tta woman-f.nom grain:abs grind-cond man:m.nom wood:abs k}e´r-eesi split-3m.sg.ipfv ‘If the woman grinds grain, the man splits wood.’

However, the different-subject simultaneous clauses can have an objectoriented interpretation, as illustrated in (26a), which contrasts with the depictive in (26b). (26) a. polı´ce´ ta´na´ ma´tt-u´wa-n police:nom 1sg.abs drunkenness-abs-loc de/-ı´Sin /oı´k}k}-iisi exist-sim.ds catch-3m.sg.pfv ‘The police caught me [while I was] drunk.’ b. polı´ce´ ta´na´ ma´tt-u´wa-n de/-iı´ddi police:nom 1sg.abs drunkness-abs-loc exist-sim.ss /oı´k}k}-iisi catch-3m.sg.pfv ‘The police caught me [while he was] drunk.’ In (26a), the predicative relation holds between the object of the main verb (i.e., ta´na´ ‘me’) and the secondary predicate ma´ttu´wa-n de/-ı´Sin ‘be in state of drunkenness’, which expresses additional information on the object noun ta´na´ ‘me’. In (26b), the relation holds between the subject of the main verb and the secondary predicate ma´ttu´wa-n de/-iı´ddi ‘be in a state of drunkenness’. In (26a), the different-subject simultaneous clause is functionally, but not formally a depictive secondary predicate. In other words, clauses with the sufﬁx -ı´Sin(i) allow for a participant-oriented as well as a process-oriented interpretation, i.e. they are general adjunct constructions. The (disjunct) subject of this dependent clause may be spelled out phonetically, as in (27), although this is not common.

Secondary predicates and adverbials in Nilotic and Omotic (27)

313

polı´ce´ ta´na´ ta´a´nı´ ma´tt-u´wa-n de/-ı´Sin police:nom 1sg.abs 1sg.nom drunkenness-abs-loc exist-sim.ds /oı´k}k}-iisi catch-3m.sg.pfv ‘The police caught me [while I was] drunk.’

By contrast, (26b) would be unacceptable if the subject of the dependent verb were to be phonetically realized. As we will show in the next section, the contrast between the constructions illustrated in (26a) and (26b) also serves to denote the pragmatic function of switch-reference. 9.3.2 The expression of anteriority in Wolaitta Like the simultaneous converb construction, the anterior converb construction (also known as participle, gerundive, or short perfect) in Wolaitta has two different formal realizations, depending on whether or not the subject of the converb and the main verb are identical. Consider the following examples: (28) a. na/a-ı´ faatana´ /aa´ºº-ı´dı´ /ufaı´tt-iisi child-m.nom exam:abs pass-ant.ss be.happy-3m.sg.pfv ‘The boy is happy, having passed his exam.’ b. na/a-ı´ faatana´ /aa´ºº-ı´nı´ /aayy-ı´ya child-m.nom exam:abs pass-ant.ds mother-f.nom /ufaı´tt-aasu be happy-3f.sg.pfv ‘The mother is happy with her child having passed the exam.’ The different-subject anterior converb construction has an invariable sufﬁx -ı´n(i). In line with the analysis we proposed for the different-subject simultaneous clauses above, we propose that the different-subject anterior clause is adverbial in nature, since it does not show agreement or subject control. In the same-subject converb construction, the anterior converb is marked by -a´da´ or by its shortened form -a´ when the subject is third person feminine or ﬁrst and second person singular. When the subject is third person singular masculine or when it is plural, the sufﬁx -ı´dı´ or its shortened form -´i is attached to the converb. This agreement pattern is identical to that observed in same-subject simultaneous clauses in the previous section. Consider the following examples: (29) a. /ı´ na-at-a´ kaass-ı´dı´ zin/-iss-iı´si 3m.sg.nom child-pl-abs play:caus-ant.ss lie down-caus-3m.sg.pfv ‘He brought the children to bed after having played with them.’

314

A. Amha and G. J. Dimmendaal b. /a´ na-at-a´ kaass-a´da´ 3f.sg.nom child-pl-abs play:caus-ant.ss zin/-iss-au´su lie down-caus-3f.sg.ipfv ‘She brings the children to bed after having played with them.’ c. /etı´ na-at-a´ k}u´ma miz-ı´dı´ 3pl.nom child-pl-abs food:abs eat:caus-ant.ss zin/-iss-ana´ lie down-caus-fut ‘They will put the children to bed after having fed them.’

The same-subject simultaneous construction and the same-subject anterior construction differ from each other in their temporal interpretation. The former expresses a situation which occurs simultaneously with that expressed by the main verb, whereas the latter basically—though not exclusively— expresses a situation which is anterior to that expressed by the main verb. Accordingly, the actual tense–aspect interpretation of both the simultaneous converb and the anterior converb depends on the tense–aspect of the main verb, as can be seen from the examples in (29). Despite the invariable form of the converb in (29), (29a) refers to a past/completed event, (29b) indicates that the situation described is presently taking place or that it is a habitual action, whereas (29c) describes a situation that is yet to take place. Another feature distinguishing the simultaneous converb from the anterior converb construction is the fact that the anterior converb is not necessarily the semantically dependent or optional secondary verb. Anterior converbs may also be used in compound verb structures, for example, with the anterior converb and the main verb expressing a single event (30a, b). In this verbal compound, neither of the two verbs can be omitted and neither of the two is semantically ‘the main predicate’; compare Azeb Amha and Dimmendaal (2006) for further details on phonological, morphosyntactic, and semantic features of such compounds. miiSSa´a [/ekk-ı´ y-iisi] (30) a. zal/a´ncca-ı´ trader-m.nom goods:def.abs take-ant.ss come-3m.sg.pfv ‘The merchant brought the goods.’ (lit. ‘The merchant taking the goods came.’) b. harga´nc-ı´ya [/aa´ºº-a´ woºº-aasu] patient-f.nom pass-ant.ss descend-3f.sg.nom ‘The patient (f.) turned over.’

Secondary predicates and adverbials in Nilotic and Omotic

315

The examples in (30) indicate that in some of its uses it may be difﬁcult to treat the same-subject anterior converb as an optional peripheral level category, which constitutes a problem for a depictive analysis. However, in other uses it appears fairly clear that the same-subject anterior converb functions as a depictive secondary predicate. Some typically participant-oriented expressions, for example, expressions of a condition, physical state, or posture, can in fact only be expressed by the same-subject anterior converbs and no other converb construction in Wolaitta, as the following examples illustrate: (31) a. na/-ı´ya sa´k-ett-a´da´ zin/-aa´su child-f.nom be.sick-pass-ant.ss lie.down-3f.sg.pfv ‘The girl is lying down sick.’ b. /ı´ /ek}k}-ı´dı´ m-ee´si 3m.sg.nom stand.up-ant.ss eat-3m.sg.ipfv ‘He eats standing.’ c. /etı´ hiyyes-ı´dı´ simm-ı´dosona 3pl.nom be.poor-ant.ss return-3pl.pfv ‘They returned poor.’ There is, however, a non-converbal alternative to the anterior construction illustrated in the examples in (31). The information communicated in these examples may also be expressed by means of (peripheral) case-marked nominals without an apparent change in meaning. This is illustrated in (32a, b). (32) a. /ı´ /ek}-u´wa-n m-ee´si 3m.sg.nom standing-abs-loc eat-3m.sg.ipfv ‘He eats standing.’ hiyyesa´tetta-n simm-ı´dosona b. /etı´ 3pl.nom poorness:abs-loc return-3pl.pfv ‘They returned poor.’ Note that the construction involving a noun/adjective afﬁxed with the Locative case marker -n above may also be used to express some clearly adverbial notions, as can be seen from example (33b), in contrast to the depictive notion in (33a). (33) a. miiSSa´a gee´SSa-n zaar-ite goods:def.abs clean-loc return-3pl.imp ‘Return the container clean.’ b. miiSSa´a /akee´ka-n zaar-ite goods:def.abs attention-loc return-3pl.imp ‘Return the container carefully.’

316

A. Amha and G. J. Dimmendaal

In contrast to examples (28a) and (29), where the anterior converb expresses anteriority and appears to have a more adverbial function, the state described by the converb in (31a–c) is to be interpreted as being cotemporaneous with the situation expressed by the main verb. Expressions such as those in (31a–c) are parallel to same-subject simultaneous clauses, which are genuine depictives, as argued in the previous section. However, the converb marker and the simultaneous marker cannot be freely exchanged in these examples. Thus the converb marker in (31a) cannot be replaced by a simultaneous marker; consequently, (34) is anomalous: (34) ? na/-ı´ya sa´k-ett-aı´dda´ zin/-aa´su child-f.nom be.sick-pass-sim.ss lie.down-3f.sg.pfv ‘The girl is lying down sick.’ On the basis of a survey of similar situations in a variety of other languages, Schultze-Berndt and Himmelmann (2004: 103) argue that the cotemporal interpretation of anterior converbs is possible because these ‘encode a poststate of the eventuality which is denoted by the lexeme from which it derives . . . the form focuses not on the eventuality itself, but its resultant state.’ We take the same position with regard to the interpretation of the converb constructions in (31a–c), all of which express durational or temporary states or conditions. Despite the formal and functional similarity between them (see Table 9.2), the same-subject simultaneous converb construction and the same-subject anterior converb construction in Wolaitta have been analysed as belonging to two totally different construction types. Approaching these constructions from the point of view of participant vs. event orientation has the advantage of allowing a more precise statement of their morphosyntactic and semantic similarities and differences. Moreover, such an approach brings to light the fact that the same-subject/different-subject (i.e. the switch-reference) system, which has been central in the discussion of clauses expressing simultaneous or anterior action in Wolaitta above, is not the main function of these sufﬁxes. The switch-reference system in this case is parasitic upon, or exploits, a system which exists independently of it; this system allows the converb to be used extensively in clause chaining. That is, in the case of the ‘different-subject’ simultaneous construction, the main function of the morpheme -ı´Sin(i) is to show that the clause headed by the verb to which this sufﬁx is attached is a dependent clause expressing a situation which takes place alongside the event expressed by the main clause; here, the focus is on the temporal overlap between the situations expressed by the two clauses. In other words, -ı´Sin(i) makes no explicit reference about the presence or absence of identity between

Secondary predicates and adverbials in Nilotic and Omotic

317

the subjects of the dependent verb and the main verb. In dependent clauses marked by the morphemes -aı´dda´j-iı´ddı´, on the other hand, not only the temporal overlap is important but also the fact that the agents/actors in the situations described by the dependent and the main verb are identical. The fact that switch-reference systems may depend on independently existing constructions has already been demonstrated for other languages (see e.g. Simpson 1988; Wilkins 1988). A similar situation is observed for Maale, as shown below. We argue that same-subject simultaneous converbs marked by -aı´dda´ and -iı´ddi are genuine depictives, whereas same-subject anterior converb constructions marked by -a´da´ and -ı´dı´ (or their shortened allomorphs -a´ and -ı´), while being clearly depictive in form, may have either participant-oriented or event-oriented functions. The ‘different-subject’ morphemes -ı´Sin(i) and -ı´n(i), expressing primarily simultaneity and anteriority respectively, are genuine temporal adverbial clause markers attached to a verb heading a clause with an overt subject. Finally, we note that the depictive and adverbial categories cross each other’s boundaries, e.g. when the adverbial -iSin(i) receives an object-related interpretation. By contrast, the same-subject converb markers -a´da´ and -ı´di, which can be treated as depictives on formal grounds (through agreement) and functional grounds (as illustrated in example 31) may also be used in adverbial clauses expressing anteriority (see examples 28a and 29). Furthermore, shortened forms of -a´da´ and -ı´di are used in complex predicates (see example 30). These facts suggest that the same item is used at different junctures ranging from adverbial (peripheral-level) to depictive (core-level) as well as to lexical (nucleus-level) construction types in the sense of Van Valin and LaPolla (1997). A similar position has been argued for by Azeb Amha (2001) in her description of the Omotic language Maale, which is discussed next. 9.3.3 Simultaneous events and converb constructions in Maale Like Wolaitta, Maale draws a morphological distinction between samesubject and different-subject anterior converb forms: The former is marked by -´i, the latter by -e´m; in addition, there is another same-subject anterior converb marker in Maale, i.e. -a´//o. However, verbs with this latter sufﬁx are used exclusively in order to express anteriority, whereas -´i may also be used in order to express simultaneity (through reduplication), as shown below. None of the three converb markers in Maale is inﬂected for person, number, or gender; these latter properties are also absent from main verbs in Maale.

318

A. Amha and G. J. Dimmendaal

(35) a. /ı´zı´ mı´s’-o´ tı´k’-ı´ makiin-aa 3m.sg.nom wood-abs cut-cvb.ss car-loc c’aan-e´-ne load-pfv-aff.decl ‘After having cut the wood, he loaded it on a car.’ b. /ı´zı´ mı´s’-o´ tı´k’-e´m nu´u´nı´ makiin-aa 3m.sg.nom wood-abs cut-ant.ds 1pl.nom car-loc c’aan-e´-ne load-pfv-aff.decl ‘Him having cut the wood, we then loaded it on a car.’ c. /iya´ta´ mu´/-a´//o /a´a´º-a´-ne 3pl.nom eat-ant.ss go-ipfv-aff.decl ‘After having eaten, they leave.’ The same-subject converb marker -´ i in Maale thus may be used to express anteriority, as in (35a), or a simultaneous condition, as in (36a). Consequently, it appears to be a depictive marker. However, as (36b) demonstrates, some constructions with -ı´ can be interpreted either as process-oriented (expressing manner) or as participant-oriented (expressing state of the subject). (36) a. /iya´ta´ dik}a´tt-ı´ zag-a´za 3pl.nom be.afraid-cvb.ss see-when ‘frightened, when they looked . . . ’ (describing two friends who realized too late that they were close to a wild animal) b. /atsı´ tats-ı´ mu´/-a´-ne. person:m.nom be.slow-cvb.ss eat-ipfv-aff.decl ‘The man eats slowly.’ The simultaneous clause also distinguishes between same-subject and different-subject forms. Note that same-subject clauses always involve the same marker and are not speciﬁed for relative tense (simultaneity vs. anteriority), although simultaneity is expressed by way of reduplication of the converb (as in 37). Different-subject simultaneous clauses are marked with the sufﬁx -nte (as in 38b). (37)

laall-e´ll-a´ wo´ntsi woº-ı´ woº-ı´ woman-f-nom mill:abs kill-cvb.ss rdp /ayn-a´º-a´-ne sing-vsf-ipfv-aff.decl ‘The woman sings while grinding.’

Secondary predicates and adverbials in Nilotic and Omotic

319

(38) a. laall-e´ll-a´ wo´ntsi woº-ı´ /ayn-a´º-a´-ne woman-f-nom mill:abs kill-cvb.ss sing-vsf-ipfv-aff.decl ‘Having ground, the woman sings.’ b. laal-e´ll-a´ wo´ntsi woº-a´-nte nu´u´nı´ woman-f-nom mill:abs kill-ipfv-sim.ds 1pl.nom /ayn-a´º-a´-ne sing-vsf-ipfv-aff.decl ‘While the woman grinds, we sing.’ In summary, Maale and Wolaitta make use of parallel constructions in order to express participant-oriented (depictive) and event-oriented (adverbial) meanings, namely through the use of converbs. In Wolaitta, subject agreement in same-subject converbs is marked on both simultaneous and anterior converbs, but not on other dependent verb forms. Maale, on the other hand, does not mark agreement on verbs, thereby making it difﬁcult to distinguish depictives from adverbial clauses. Nevertheless, the Maale forms provide evidence in favour of the claim made earlier with respect to Wolaitta, namely that the same-subject simultaneous construction and the same-subject anterior converb construction represent similar construction types; in Maale, these two constructions are expressed by one and the same form, i.e. a converb marked by -ı´ (the former without reduplication and the latter with reduplication). Both same-subject converb types mainly express depictive content, although the anterior converb may occasionally be interpreted as an adverbial. In contrast, we have interpreted the different-subject converbs as instances of general adjunct constructions.

9.4 Some wider prospects 9.4.1 Explaining intragenetic variation The syntactic position of secondary predicates relative to the main clause appears to correlate with the verb-initial versus verb-ﬁnal parameter crosslinguistically (although this needs to be checked in detail). This may be further illustrated here with data from other Semitic languages as a tertium comparationis. Within Semitic, one ﬁnds verb-initial languages like (Classical) Arabic and verb-ﬁnal languages such as Amharic. The latter, as well as other Ethiopian Semitic languages, are known to have converged towards verb-ﬁnal languages under the inﬂuence of Ethiopian Cushitic (and Omotic) languages.

320

A. Amha and G. J. Dimmendaal

In modern Egyptian Arabic, which has a VO constituent order, predicative adjectives follow the main predication when used as secondary predicates: (39) idhin il-bab iswid paint the-door black ‘Paint the door black.’ (40) kul il-lahma nayya ˙ eat def-meat raw ‘Eat the meat raw.’ In Amharic on the other hand, these predications precede the main verb or clause: (41) t’_˝k’ur k’´bba-w black paint:imp-3m.obj ‘Paint it black.’ (42) s_˝gawn t’_˝rewun b_˝law meat:def.acc raw:def.acc eat:imp ‘Eat the meat raw.’ Parallel to the Omotic languages discussed above, the Ethiopian Semitic language Amharic uses anterior clause-chaining strategies involving converbs in order to express simultaneous or anterior action: (43) l_˝ju rot’-o gabba boy:def run-ant entered:3m.sg ‘The boy came in running.’ As the survey for Nilotic as well as for Omotic above should have shown, systems are never isomorphic even between closely related languages which are otherwise typologically similar, due to changes in morphosyntactic strategies (both derivationally and inﬂectionally) as well as to differences in the referential domain for adverbs. It may be expected, therefore, that interesting differences occur between verb-ﬁnal Semitic languages, as well as between Semitic languages using either a verb-initial or a SVO constituent order, in the way they encode depictive secondary predicates. 9.4.2 A common pattern Apart from opposite syntactic strategies as used in verb-initial Nilotic versus verb-ﬁnal Omotic languages, which constitute mirror images of each other typologically, a common depictive strategy is attested in both genetic groups, involving the use of ideophones. For reasons of space, this topic is discussed

Secondary predicates and adverbials in Nilotic and Omotic

321

only brieﬂy here. All Nilotic and Omotic languages make use of predicative ideophones, although in some languages the system is far more elaborate than in others. The Nilotic language Bari has ‘adverbs of character’ as Spagnolo (1933: 202) calls them, introduced by the defective verb ‘say’; these constructions either modify the event, i.e. they have an adverbial function (as in (44)), or they are participant-related, i.e. they have a depictive function (as in (45)): (44) walE a-’duNgo¨ adi liyuk knife pred-cut:antip pred-say ideo ‘The knife cut cleanly.’ (45) Nutu sisida kana, adi lyaa(lee) people sit only say ideo ‘The people are sitting down being vain . . . doing nothing.’ Such constructions with ‘say’ are also common in Omotic languages, where they always precede the main clause in line with general constituent order properties of these languages; compare Amha (2001: 254 ff.) for a description of this phenomenon in Maale. It is also common in both language families to derive main verbs from such ideophones. As pointed out by Spagnolo (1933: 206), ‘many verbs are descended from such adverbs’, e.g wOlEgga ‘to act quickly’ from wlEk ‘quickly’. Toweett (1979: 387) refers to similar cases whereby ideophones are turned into verbs in Southern Nilotic Kipsikiis. As shown by Amha (2001: 256ff.), such predicative ideophones can also be changed into transitive or intransitive verbs in the Omotic language Maale. Obviously, the operation of such derivational strategies has consequences for the lexical-semantic as well as the syntactic structure of individual languages; the incorporation of semantic roles such as manner, condition, or state in derived verbs of this type preempts the need for other strategies rendering a similar content, e.g. the need for a separate manner adverb or secondary predicate (whether ideophonic or non-ideophonic). These additional, derivational strategies thus illustrate another way in which the complex interaction between the syntactic and lexical structure may cause historical divergence even between closely related and typologically similar languages in the way they express either depictive or adverbial meanings.

This page intentionally left blank

10 Asyndetic subordination and deverbal depictive expressions in Shona ¨ LDEMANN TOM GU

10.1 Introduction Shona, the major Bantu language of Zimbabwe, has recourse to several strategies for expressing participant orientation in adjuncts. These include adpositional phrases of a type that is widely attested crosslinguistically in this function, and which involve, in Shona, such preﬁxes as similative sa- ‘like, as’ or ideophones in adjunct function. In this chapter, however, I will be concerned with a different construction. Here, the adjunct is instantiated by a special type of verb form, exempliﬁed by atsamwa in (1), which displays normal segmental verb inﬂection, but which is prosodically marked as subordinate.1 (1) ipapo . . . mambo waka-muruka a-tsamwa then 1:king 1.rem.pst-stand.up 1.fs.stat-show.displeasure ‘Thereupon the king rose angrily.’ (orig. ‘da . . . stand der Ko¨nig erzu¨rnt auf’) (Sicard 1965: 336)

My thanks go to Nikolaus Himmelmann and Eva Schultze-Berndt for inviting me to the workshop ‘Depictive secondary predicates in cross-linguistic perspective’ and for their extensive and fruitful comments on earlier drafts, to the workshop participants for additional input to this chapter, and last but not least to the Shona speakers, in particular Samson Huni, who introduced me to their interesting language. 1 The following conventions apply to the representation of examples in this chapter. Constituents relevant for the discussion are normally highlighted by boldface. The orthography of cited examples is retained. Since there are slight differences between the different sources (i.e. Fortune and Sicard vs. others), the same linguistic element can have more than one representation. I have added most glosses

324

T. Gu¨ldemann

Such dependent predicates would belong to the class of ‘deverbal depictives’ in terms of Schultze-Berndt and Himmelmann (2004: 98–105) provided that it can be shown that they are not only semantically, but also morphosyntactically participant-oriented and hence proper depictive constructions. Without doubt they play a central role in Shona because verbs in Shona quite generally compensate for the absence of a large class of adjectives, which frequently function as a major depictive expression type in other languages. One main concern of this chapter, to be addressed in Section 10.4, will be to answer the question to what extent the relevant Shona verb forms meet the crosslinguistic deﬁnition for depictive secondary predicate constructions proposed by Schultze-Berndt and Himmelmann (2004: 77f.). A second concern is an investigation of the relevance of focus structure for the analysis of depictives. Starting with the characteristics of the Shona construction and supplementing these with crosslinguistic data found in Schultze-Berndt and Himmelmann, I will argue in Section 10.5 that the default focus behaviour of depictives is another promising candidate for being a criterial feature of dedicated depictive constructions (i.e. depictive constructions which are formally participant-oriented and differ morphosyntactically from other adjunct constructions). However, before these two main concerns of the chapter can be addressed, it is necessary to outline the complex system of prosodic predicate marking in Shona (Section 10.2) and to single out the subtype that actually occurs regularly in depictive-like functions. This subtype consists of dependent predicates which are not used as adnominal modiﬁers, which have potentially free subject reference and which encode a temporal macro-category called ‘taxis’ (Section 10.3). The Shona data are drawn mostly from published narrative texts (Sicard 1965) and sources on grammar (Dale 1972; Fortune 1955). These are supplemented by spoken language data from my own research.

10.2 Asyndetic prosodic subordination and taxis predicates in Shona Many Bantu languages display an asyndetic clause-linkage strategy whereby the dependency of a clause is marked exclusively by formal features of and sometimes modiﬁed/added to the original translations. If no source is given in the translation line, it comes from my own research on Shona. In the glosses, arabic numerals indicate person categories of speech-act participants if directly followed by sg or pl; otherwise, they refer to nominal agreement classes of Bantu languages, which establish an elaborate gender system for third-person referents.

Asyndetic subordination and deverbal depictive expressions in Shona

325

its inﬂected verbal predicate. This is often achieved by a suprasegmental modulation of the verb. In the absence of any other marking at the clause or predicate levels (at least in most afﬁrmative forms), particular prosodic patterns are the only grammatical signal to distinguish main, relative, and adverbial clauses. This strategy is recognized as a potential feature of Proto-Bantu by Meeussen (1967: 113, 121; see ‘Conjunctive verb forms’), and is discussed from a wider cross-Bantu perspective by Gu¨ldemann (1996: ch. 3). A special instance of this strategy is also found in Shona. The traditional classiﬁcation of Shona verb forms was established by Fortune (1955). He sets up inter alia eight ‘moods’ which are ‘‘inﬁnitive’’, ‘‘imperative’’, ‘‘indicative’’, ‘‘potential’’, ‘‘participial’’, ‘‘relative’’, ‘‘subjunctive’’, and ‘‘hortative’’. Three of these moods, namely ‘‘indicative’’, ‘‘participial’’, and ‘‘relative’’, are distinguished primarily by different suprasegmental patterns operating over one and the same segmental form of the verb.2 Generally speaking, the prosodic form of a verb is determined by a complex of different factors. These are: (a) the lexical tone class and number of syllables of a verbal lexeme, (b) the presence and type of preﬁxes marking cross-reference to subject and object, (c) the presence and type of afﬁxes marking predication-operator functions like polarity, tense, aspect, etc., (d) the presence of consonants triggering phonetic tone depression, (e) ordinative accent, and last but not least (f) the grammatically assigned suprasegmental melody (see e.g. Fortune 1984 regarding so-called ‘‘tone conjugations’’). In this chapter, it is neither possible nor necessary to present all details of the complex interaction of these factors in determining the prosody of Shona verb forms. I will limit myself to presenting the most salient generalization on the three classes of tone patterns at issue. Of major import in this regard is the tone on the initial subject preﬁx. In verb forms of the ‘‘indicative mood’’, it is low for all ﬁrst and second persons (alias speech-act participants), but high for all third persons, i.e. referents marked for noun class. This tonal distinction is neutralized in all subordinate predicates. In the ‘‘relative mood’’, the subject preﬁx bears a low tone irrespective of the type of subject. In the ‘‘participial mood’’, the tone of the subject preﬁx is always high. Compare the following short paradigm of tengesa ‘sell’ in the recent past (an accent marks high tone).

2 The segmental differences which do exist but are restricted to certain environments, are explained from a functional perspective in Gu¨ldemann (1997a; 1997b). The same sources can be consulted in general for a more extensive discussion of asyndetic clause linkage in Shona.

326

T. Gu¨ldemann a´tengesa ‘(s)he has sold’ ‘‘relative’’ ate´ngesa ‘(s)he who has sold’ ‘‘participial’’ a´te´ngesa´ ‘(s)he having sold’

(2) ‘‘indicative’’

ndate´ngesa ‘I have sold’ ndate´ngesa ‘I who have sold’ nda´te´ngesa´ ‘I having sold’

The philological terms ‘‘indicative’’, ‘‘relative’’, and ‘‘participial mood’’ are misleading from a functional perspective: apart from the most unusual use of ‘‘mood’’, they suggest a clustering of linguistic functions which is different from the one actually found. The ‘‘indicative mood’’ is not particularly problematic in this regard as it covers predicates of non-modal main clauses as in (3). (3) nda-muka ma-Ngwanani a-no 1sg.rec.pst-get.up 6-morning 6-dem ‘I got up this morning.’ (Fortune 1955: 278) ‘‘Relative’’ and ‘‘participial mood’’ are functionally related in that they both mark subordinate clauses. However, the ‘‘relative mood’’ does not cover all relative clauses, but only so-called direct ones. These constitute a formally deﬁned category where the initial cross-referencing preﬁx on the relative predicate agrees in person, gender, and number with the head noun. It renders mostly subject relative clauses as in (4). But it can also convey nonsubject relatives displaying so-called ‘‘subject inversion’’ as in (5), provided the postverbal noun in the dependent clause is clearly higher on the animacy hierarchy than the head noun cross-referenced on the dependent predicate and thus qualiﬁes semantically for the agent role. (4) nyoka dzi-no-gara mu-muti dzi-no-bata shiri 10.snake 10.rs-prs-live iness-tree 10-prs-catch birds ‘Snakes which inhabit trees catch birds.’ (Dale 1972: 310) (5) sadza ri-no-dya m-Nana 5.porridge 5.rs-prs-eat 1-child ‘porridge which the child eats’ (lit. ‘porridge which eats the child’) (Fortune 1955: 185) The ‘‘participial mood’’ comprises several uses, two of which are relevant here. On the one hand, it occurs on predicates of adverbial-type clauses, as exempliﬁed in (6) and (7). The position and prosodic linkage of these clauses vis-a`-vis the main clause is variable (see (6) and (7) for different positions and Section 10.4 regarding prosody).

Asyndetic subordination and deverbal depictive expressions in Shona

327

(6) va-svika-po va-bereki vaka-ti . . . 2.fs.ant-arrive-adess.anaph 2-parent 2.rem.pst-quot ‘Having arrived there, the parents said . . . ’ (7) garwe raka-ona mu-komana a-gwira mu-rwizi 5.crocodile 5.rem.pst-see 1-boy 1.fs.ant-fall iness-river ‘The crocodile saw the boy having fallen into the river.’ On the other hand, this ‘‘mood’’ occurs in so-called indirect relatives, which are deﬁned by the lack of grammatical agreement between head noun and relative predicate as in (8). Most non-subject relative clauses are formed in this way.3 (8) iyi ndi-yo imba Sadza aka-vaka 9.dem cop-9.anaph 9.house 1.pn 1.fs.rem.pst-build ‘This is the house which Sadza built.’ (Fortune 1955: 187) In view of the actual functional range of ‘‘relatives’’ and ‘‘participials’’, I will replace the misleading philological terms by more transparent labels in the glosses of examples and, when convenient, also in the text. The new terms are motivated by a consistent formal characteristic, namely the status of subject reference in the dependent predicate. Verb forms of the ‘‘relative mood’’ have restricted subject reference, because their subject must be coreferential with the head noun of the relative construction. Accordingly, they are called restricted-subject dependent predicates (glossed as RS at the subject concord). As opposed to this, verb forms of the ‘‘participial mood’’ do not imply obligatory coreference with a main-clause participant (in fact, they never cross-reference a main-clause participant when occurring in ‘‘indirect’’ relatives). Therefore, they are called free-subject dependent predicates (glossed as fs at the subject concord). It is important to recognize, though, that the latter term does not exclude the possibility of actual control of the ‘‘participial’’ subject by a main-clause participant as shown by the adverbial participials given in (6) and (7). It is not sufﬁcient to clarify the distinction of verb forms merely in terms of prosodic features, because these interact with other morphosyntactic and semantic mechanisms to yield a yet more complex functional differentiation of predicate types. This concerns in particular the class of free-subject dependent predicates. As just mentioned, this predicate type displays a further opposition between an adnominal use in (‘‘indirect’’) relative clauses and a non-adnominal use in adverbial-type clauses. 3 See Gu¨ldemann (1997a, b) for a functional motivation of the formal afﬁnity between adverbial hypotaxis and non-subject relative clauses in Shona.

328

T. Gu¨ldemann

Predicates of the latter type exclusively share an important semantic feature: they encode a temporal macro-category which is called taxis here. Taxis denotes the time relation between the communicated state of affairs and another state of affairs which is encoded or implied in the discourse context and which serves as the temporal reference point.4 I prefer ‘‘taxis’’ to ‘‘relative tense’’, another common label for this function, as it stresses a semantic independence from other temporal macro-categories, in particular tense and aspect. The Shona forms at issue represent a speciﬁc variant of taxis which combines the indication of a time relation with the indication of the dependent status of the relevant clause so that this kind of time-marking is always oriented towards a syntactically related main clause. The temporal feature of taxis that is typical for free-subject dependent predicates is responsible for the fact that they contrast with all other prosodically marked predicates not only syntactically, but also in semantic terms. Table 10.1 summarizes the relevant distinctions between the predicates differentiated by prosodic patterns. In the leftmost column, it presents the segmental make-up of the verb forms affected by the variable prosodies and speciﬁes their meanings in the right-hand columns. The different meanings depend on prosody and syntactic context. The semantic labels for free-subject dependent predicates in the rightmost column refer to the concept of taxis and show that their common denominator can be characterized mostly as stativity, and cotemporality with the main predicate. Such functions as ‘‘anterior’’ and ‘‘proximative’’ (¼ ‘be about to do’) also refer to cotemporality, namely when a state results from a previous event or is the precondition of a following one. The semantic dissociation between free-subject dependent predicates marking relative time (alias taxis) and all the other verb forms in Table 10.1, viz. main-clause forms and direct and indirect ‘‘relatives’’, which mark absolute time or modality, is shown by the following fact. In some instances, a segmental grammatical marking is incompatible with a taxis reading and thus entirely excluded from this type of use5 (these forms are not listed in Table 10.1, see Gu¨ldemann (1997a; 1997b) for some discussion). Alternatively, a given segmental marking has a biased, restricted, or entirely different meaning when occuring on free-subject dependent predicates as opposed to its use in other clause types. 4 The term ‘‘taxis’’ appears to go back to Jakobson (1971) and is commonly used in this sense in Russian linguistics (see e.g. Maslov 1988: 64). 5 As mentioned later in section 3.1, there are a number of purely grammatical triggers for the use of free-subject dependent predicates. These triggers are responsible for the fact that many free-subject dependent verb forms are used more widely than the meanings given in Table 10.1 would lead one to expect.

Asyndetic subordination and deverbal depictive expressions in Shona

329

Table 10.1. Affirmative verb forms affected semantically and syntactically by prosodic ‘‘mood’’ marking in Shona (after Gu¨ldemann 1997a; 1997b) Philological terms

Indirect Direct Relative Adverbial [‘‘Indicative’’] [‘‘Relative’’] [‘‘Participial’’] [‘‘Participial’’]

Terms proposed in this chapter

Main clause predicate

RS dependent predicate

Form 1

sbj.pref-ri ku-stem-a

FS dependent predicate Meaning

present progressive

2 sbj.pref-stema 3 sbj.pref-stem-eb

present present stative

4 sbj.pref-a-stem-a

recent past/perfect/present stativec

5 sbj.pref-aka-stem-a

remote past/present stativec

6 sbj.pref-cha-stem-a

persistived/future

7 sbj.pref-o-stem-a

‘‘exclusive’’e

?

?

simultaneous progressive simultaneous simultaneous stative anterior/ simultaneous stativec simultaneous stativec persistived/ proximative coincidence/ sequential

Notes: a only with defective verbs ri ‘be (at)’, na ‘be with’; b only with a few intransitive inchoative verbs, stem assimilates to ﬁnal vowel; c relevant meaning only with inchoative verbs; d marginal in Zezuru and Standard Shona; e deontic modality and temporal notion ‘now’.

The second case is brieﬂy exempliﬁed here by the so-called ‘‘exclusive’’ form (the last row in Table 10.1). As an ‘‘indicative’’ form, it is only used in permissive and directive questions as in (9) and in assertions focusing on the immediate beginning of an action (often translated as ‘now’). As a ‘‘participial’’ form, it marks temporal coincidence when preposed to the main clause as in (10), and mostly sequentiality when postposed to it as in (11). (9) ndo-pinda here 1sg.excl-enter q ‘am I now to enter? [can/should I enter?]’

(Fortune 1955: 276)

(10) zuva ro-tsvuka aka-svika pa-rukova 5.sun 5.fs.coi-become.red 1.rem.pst-arrive adess-river ‘at sunset [i.e. as the sun turned red] he reached a river.’ (Dale 1972: 61)

330 (11)

T. Gu¨ldemann aka-simuka o-dzokera ku-musha 1.rem.pst-rise 1.fs.seq-return loc-home ‘He rose up and returned to his village.’ (Dale 1972: 89)

The different functioning of the three ‘‘moods’’ is perhaps best illustrated by contrasting their use in rendering the same proposition in essentially the same context. The triplet of complex sentences in (12) illustrates such a contrast in the expression of the proposition ‘[four small puppies] are outside’ occurring in a subordinate construction headed by the main clause predicate ‘see’. In (12a), the conjunction kuti ‘that’ heads a complement clause with the embedded predicate twakange turi in the ‘‘indicative mood’’. In (12b), the reference of the noun tuhanda ‘puppies’—now a main-clause object—is restricted by a direct relative clause with the embedded predicate twakange turi in the ‘‘relative mood’’. The same state of affairs is conveyed in (12c) by means of the ‘‘participial mood’’ (free-subject dependent predicate) form turi. The dissociation in time marking is reﬂected by the different forms of the embedded predicates. Like the main predicate yakaona ‘saw’, twakange turi in both (12a) and (12b) displays absolute temporal deixis with remote-past marking on the auxiliary nge (which in turn requires ‘‘participial’’ turi for grammatical reasons: see section 10.3.1). The dependent predicate in (12c) is the plain form turi which encodes simultaneous taxis. (12)

hanzvadzi-komana pa-yaka-ngo-buda 9.sibling-boy 16time.assoc-9.fs.rem.pst-just-come.out pa-nze adess-outside ‘When the brother arrived, . . . ’ a. yaka-ona kuti tu-handa tu-na twaka-nge 9.rem.pst-see that 12-puppy 12-four 12.rem.pst-be tu-ri pa-nze 12.fs-be adess-outside ‘ . . . he saw that there were four small puppies outside’ b. yaka-ona tu-handa tu-na twaka-nge 9.rem.pst-see 12-puppy 12-four 12.rs.rem.pst-be tu-ri pa-nze 12.fs-be adess-outside ‘ . . . he saw the four small puppies that were outside’

Asyndetic subordination and deverbal depictive expressions in Shona

331

c. yaka-ona tu-handa tu-na tu-ri-po 9.rem.pst-see 12-puppy 12-four 12.fs-be-16adess.anaph pa-nze 16-outside ‘ . . . he saw four small puppies [being] outside’ Having outlined the form and function of prosodically marked taxis verb forms, it is important to recognize that Shona also possesses two taxis forms which are marked segmentally and do not belong prosodically to the class of free-subject dependent predicates.6 The ﬁrst one with the structure sbj.prefka-stem-a marks sequentiality when it is postposed to the main clause as in (13) and an open conditional protasis when it is preposed, as in (14). (13)

a-ba mari mu-chitoro a-ka-tiza 1.rec.pst-steal money iness-shop 1-seq-run.away ‘He stole money in the shop and then made off.’ (Dale 1972: 260)

(14)

u-ka-ramba 2sg-cond-refuse wa-ko 1.assoc-2sg.poss ‘If you refuse this (Dale 1972: 261)

mu-komana uyu baba 1-boy 1.dem 1.father a-no-zo-tambudzika 1-prs-then-grieve young man, your father will grieve.’

More important for the present discussion is the second gram, sbj.prefchi-stem-a, illustrated in (15), which marks general simultaneity. Since it is one of the most frequent taxis forms and can be analysed as the semantic counterpart of the main-clause present (which has the form sbj.pref-nostem-a), it is regularly assigned to the class of free-subject dependent predicates in Shona grammars, despite its unrelated suprasegmental form. We will return to this form below in section 10.3.4. (15)

ndaka-uya nezuro mvura i-ci-naya kwa‰o 1sg.rem.pst-come yesterday 9.rain 9-sim-rain much ‘I came yesterday, the rain falling hard.’ (Fortune 1955: 295)

10.3 The functional range of free-subject dependent predicates, with special reference to the expression of depictive content As indicated in the previous discussion, the functional extension of the ‘‘participial mood’’ (including the chi-form) is far greater than suggested by the philological term. The major contexts of use of what I call here 6

Cf. Gu¨ldemann (1998) for a cross-Bantu perspective on such verb forms.

332

T. Gu¨ldemann

free-subject dependent predicates are: (a) (b) (c) (d)

grammatically controlled dependent predicates predicates of indirect relative clauses predicates of asyndetic adverbial background clauses ‘depictives’ (i.e. participant-oriented adjuncts)

The ﬁrst three contexts will only be exempliﬁed brieﬂy, while the last use is illustrated in more detail since it is the central topic of this chapter. 10.3.1 Grammatically controlled dependent predicates One of the most salient uses of free-subject dependent forms is their obligatory occurrence after conjunctions and auxiliary verbs. Both usage contexts are secondary grammaticalizations of the strategy of asyndetic subordination, but they go in different directions on what Lehmann (1988: 216–17) identiﬁes as a scalar continuum between two opposite poles: the ‘‘elaboration of a phrase into a more fully developed construction which contains its own predication with all the accessories’’ on the one hand, and the ‘‘compression of a fully ﬂedged clause to a nominal or adverbial constituent of a matrix clause’’, on the other. Free-subject dependent predicates preceded by conjunctions such as kana ‘if ’ in (16) reﬂect a clear tendency in Shona (and other Bantu languages) to elaborate an asyndetic subordination toward syndetic and semantically more explicit clause-linkage strategies. (16) kana pai-ne chi-manikidzo chi-kuru if adess.fs.pst.hab-be.with 7-obstacle 7-big ‘If there is a big obstacle . . . ’ (Dale 1972: 302) The use after auxiliaries, illustrated in (17), represents the downgrading of dependent clauses to sub-constituents of complex predicates. It has the effect that the free-subject feature of the ‘‘participial mood’’ is replaced by obligatory subject control triggered by the auxiliary. (17) ha-ndi-fe ndaka-ita i‰ i neg-1sg-die:prs 1sg.fs.stat-do 8inan.dem ‘I never do this.’ (lit. ‘I don’t die having done this.’)

(Fortune 1955: 358)

10.3.2 Predicates of indirect relative clauses 7 In the context of indirect relatives, the subject of the free-subject dependent predicate is by deﬁnition not controlled from outside the dependent 7

The chi-form is excluded from the relative use because it marks simultaneous taxis exclusively.

Asyndetic subordination and deverbal depictive expressions in Shona

333

clause, i.e. the free-subject dependent predicate does not agree with the head noun. Example (8) has illustrated its use in an asyndetic relative structure. However, it is more common that additional segmental markers are employed in this grammatical context. In Standard Shona and central dialects, the associative/genitive linker a (glossed as assoc) preceded by the respective class concord of the head noun is directly attached to the subordinate verb as in (18); the recruitment of a linking demonstrative agreeing with the head noun is typical for Manyika dialects, as seen in (19). (18) mu-kadzi wa-nda-uya ne-hari 1-woman 1.assoc-1sg.fs.rec.pst-come com-9.pot ya-ke 9.assoc-1poss ‘the woman with whose pot I came [woman ‘‘of’’ I came with her pot]’ (Fortune 1955: 190)

”a”a wa-Ngu aka-teNga (19) mombe idzo 10.cattle 10.dem 1.father 1.assoc-1sg.poss 1.fs.rem.pst-buy ‘the oxen which my father bought’ (Fortune 1955: 186) In addition to the fact that such free-subject dependent predicates must be adjacent to their head noun in the superordinate clause (which indicates their embedding in a complex nominal) and the fact that they do not encode taxis, these formal properties justify the synchronic distinction of the construction from all the other usage contexts for free-subject dependent predicates discussed here. 10.3.3 Predicates of asyndetic adverbial background clauses In the domain of adverbial hypotaxis, free-subject dependent predicates display the greatest freedom in terms of position and integration vis-a`-vis the main clause. Preposed, they mostly instantiate background clauses and serve to create textual cohesion (see (6) above and (20) below). Free-subject dependent predicates in adverbial use may also follow the main clause, as seen in (21) and (22). (20) a-va ma-ngwanani vaka-muka 6.fs.ant-become 6-morning 2.rem.pst-get.up ‘In the morning [lit. ‘it having become morning’] they [class 2] got up.’ (Dale 1972: 73)

334

T. Gu¨ldemann

(21) ndaka-uya mu-Jerumani ma-kore ma-viri 1sg.rem.pst-come iness-Germany 6-year 6-two aka-pfuura 6.fs.stat-pass ‘I came to Germany two years ago.’ (lit. ‘two years having passed’) (22) waka-uya kuno ini ndi-si-nga-de 1.rem.pst-come here 1sg 1sg.fs-neg-sim-want ‘He came here without my wish.’ (lit. ‘I not wanting’) 1955: 297)

(Fortune

In general, the subject of adverbially used free-subject dependent predicates can be coreferential with a main-clause participant as in (6), or it may have independent reference as in (20–22). 10.3.4 Depictive-like uses Finally, I demonstrate in more detail how free-subject dependent predicates can be oriented towards a participant of the main predicate, i.e. take over the function of typical (and not so typical) depictive secondary predicates. Here, the form always shows person-gender-number agreement with one of the main-clause participants, and in this sense it can be said to be controlled. The ﬁrst set of examples show the use of free-subject dependent predicates for the expression of condition and state—one of the semantic core domains of depictives. (23) aka-zo-regera zi-nyoka ra-fa 1.rem.pst-then-leave 5augmentative-snake 5.fs.ant-die ‘Eventually he left the big snake [being] dead [or: having died].’ (Dale 1972: 72) (24) asi kana mu-pfuta u-ri ku-kura zvakanaka but if 3-pfuta.plant 3-prog-grow well waka-svibirira 3.fs.stat-be.vivid.in.colour ‘But when the pfuta plant would grow well in vivid green . . . ’ (25) Cemedzai waka-kura a-ri mu-sikana 1.pn 1.rem.pst-grow.up 1.fs-be 1-girl waka-rurama kwa‰o 1.rs.stat-be.righteous very ‘Cemedzai grew up as [lit. being] a very righteous girl.’ (orig. ‘Cemedzai wuchs als ein sehr rechtschaffenes Ma¨ dchen heran’) (Sicard 1965: 107)

Asyndetic subordination and deverbal depictive expressions in Shona

335

(26) aka-wana shumba i-vete pa-zasi 1.rem.pst-ﬁnd 9.lion 9.fs-sleep:stat 16adess-down po-muti 16.assoc-tree ‘He found a lion sleeping at the foot of the tree.’ (Dale 1972: 72) ku-kuya (27) waka-kura a-si-Nga-zini 1.rem.pst-grow.up 1.fs-neg-sim-know inf-grind ‘She grew up without knowing how to grind’ (orig. ‘ . . . das wuchs auf, ohne das Getreidemahlen zu lernen’) (Sicard 1965: 388) In passing, it may be noted that free-subject dependent forms in some of these examples indicate that a state or condition resulted from a change of state; the particular marking depends on the meaning of the respective verb. For example, there is a regular distinction between states achieved in a more recent or a more remote inchoative event. Compare rafa in (23) and wakasvibirira in (24) in connection with the different tense meaning of the corresponding main clause forms, namely recent vs. remote past. In addition, there exists a very restricted, lexicalized pattern: the special stative stem form seen in ivete in (26) only occurs with less than ten inchoative verbs. It can also be seen that the constructions headed by free-subject dependent predicates can display quite different degrees of propositional and structural elaboration. They range from one-word predicative expressions as in (23) and (24), over clauses with complements and adjuncts as in (26) and (27), to a construction which is itself biclausal as in (25), containing a secondary copulative predicate and a predicate nominal modiﬁed by a relative clause. However, as Schultze-Berndt and Himmelmann (2004: 98–101) observe, there is no clear-cut distinction between simple and complex deverbal depictives, but rather a gradual transition from more phrase-like to more clause-like expressions. This also holds for the Shona construction. The chi-form (ﬁrst mentioned at the end of section 10.2 above) of the verb da ‘want’ followed by an inﬁnitive in (28) conveys intention, differing semantically and formally from a simple purpose clause. This expression of an internal cognitive condition/state represents the Shona counterpart of what Schultze-Berndt and Himmelmann (2004: 103–5) discuss under the label prospective deverbal depictive. (28) dzaka-nzwa izwi ra-ke ndoku-bva 10.rem.pst-hear 5.voice 5.assoc-1poss consequence-turn.out dza-buda pa-nze dzi-chi-da ku-mu-ruma 10.fs.ant-come.out adess-outside 10-sim-want inf-1obj-bite ‘ . . . they heard his voice and came out wanting to bite him/ with the intention to bite him’

336

T. Gu¨ldemann

Concomitance or lack of accompaniment is also expressed by means of free-subject dependent predicates. Here, the construction is based on the defective verb na ‘be with’ as in (29) and (30). (29) yaka-ona hanzvadzi-komana i-ne imbwa i-na 9.rem.pst-see 9.sibling-boy 9.fs-be.with 10.dog 10-four ‘She [ ¼ hanzvadzsikana ‘9.sister’] saw the brother with four dogs.’ (30) naka-gara 2.rem.pst-live ‘They lived for ‘sie lebten eine 1965: 13)

ˆa-si-no-mwana ci-nhambo 7-short.time 2.fs-neg-be.with-child some time childless.’ (lit. ‘not having a child’; orig. Zeitlang, ohne ein Kind zu bekommen’) (Sicard

A third use of participant-oriented free-subject dependent predicates is the expression of manner. Compare: (31) aka-taura zv-ose izvi aka-shatirwa 1.rem.pst-speak 8inan-all 8dem 1.fs.stat-become.angry ‘He said all these things in a state of anger.’ (Dale 1972: 72) (32) aka-gara no-mu-kadzi wa-ke a-chi-fara 1.rem.pst-live com-1-woman 1.assoc-1poss 1-sim-be.happy kwazvo very ‘He lived with his wife very happily.’ (Dale 1972: 72) (33) i-ka-ci-ti yaka-nyatso-pfaˆa kwa‰o 9sbj-seq-7inan.obj-quot 9.fs.stat-truly-be.gentle very ‘And it replied very gently indeed.’ (Fortune 1955: 299) The following example pair demonstrates the contrast between the use of the verbal manner deictic daro ‘like that’ as a subject-oriented adjunct in (34) and as an attributive nominal modiﬁer in a direct relative clause in (35). (34) zvaka-ngo-ramba zvi-chi-ngo-daro nzira y-ose 8inan.rem.pst-just-persist 8-sim-just-like.that 9.path 9-all ‘These things continued all along the way just like that.’ (35) vaka-ona tu-mi-rwi twe-zvi-yo chi-bage mu-punga 2.rem.pst-see 12dim-4-heap 12.assoc-8-grain 7-maize 3-rice ne-zvi-mwe-wo zvaka-daro com-8inan-other-foc 8.rs.stat-like.that ‘They found small heaps of grain, maize, rice, and also other things like that.’

Asyndetic subordination and deverbal depictive expressions in Shona

337

The segment from a dialogue given in (36) illustrates the use of two ideophones as well as the verbal proforms ºini and ºero (same as daro above) as subject-oriented adjuncts referring to the sound of a trap when falling. Since ideophones and other mimetic elements are preferably introduced in Shona by means of the dummy verb ti (glossed mimesis), the following clauses mostly involve free-subject dependent predicates based on this verb (see Amha and Dimmendaal, Ch. 9, this volume, for similar constructions in Nilotic and Omotic languages). (36) ri-va ra-wa 5-trap 5.prf-fall aka-ti ra-wa ri-ci-ti ku-ºini 1.rem.pst-quot 5.prf-fall 5-sim-mimesis inf-how ivo vaka-ti ra-wa ri-ci-ti bu 2.dem 2.rem.pst-quot 5.prf-fall 5-sim-mimesis ideo:fall.ﬂatly ha-ri-wi iye aka-ti ra-Ngu 1.pron 1.rem.pst-quot 5.assoc-1sg.poss neg-5-fall:prs ri-ci-ºero 5-sim-like.that ri-no-wa ri-ci-ti warakata 5-prs-fall 5-sim-mimesis ideo:fall.with.rustling.sound ‘‘‘ . . . the trap has fallen.’’ He said, ‘‘With what sound did it fall?’’ She said, ‘‘It fell like ‘bu’.’’ He said, ‘‘Mine does not fall with such a sound, it falls like ‘warakata’.’’ ’ (orig. ‘ ‘‘.. die Falle ist gefallen.’’ Er erwiderte: ‘‘Wie ho¨rte es sich an, als sie ﬁel?’’ Jene sagte: ‘‘Sie machte beim Fall bu.’’ Da sagte er: ‘‘Bei meiner ho¨rt es sich nicht so an, wenn sie fa¨llt, sondern nur wie ein leises Rascheln.’’ ’) (Sicard 1965: 343) A last usage domain of participant-oriented free-subject dependent predicates is the expression of location. This is apparently less typical for depictives from a crosslinguistic perspective, but frequent in Shona. The clause pattern is based on the defective copulative verb ri ‘be (at)’ and used regularly after main verbs like siya ‘leave’, wana ‘ﬁnd’, ona ‘see’, and nzwa ‘perceive’. (37) apa adess.dem va-siye 2-leave:sbjv ‘the purpose

zano raka-nge ri-ri re-kuti 5.plan 5.rem.pst-be 5.fs-be 5.assoc-comp v-ana va-ri-po ipapo 2-child 2.fs-be-16adess.anaph 16.dem.anaph here was that they would leave the children right there’

338

T. Gu¨ldemann

(38) ma-purisa aka-va-wana va-ri-po 6-police 6sbj.rem.pst-2obj-ﬁnd 2.fs-be-adess.anaph ‘the police found them [they being] there.’ (Dale 1972: 72) One might be tempted to view constructions as in (7) and (38), which involve perception verbs as matrix predicates, as complement clause constructions. However, example (12) shows that there is a structural contrast between a canonical factive complement clause involving the conjunction kuti ‘that’ given in (12a) and a clearly participant-oriented free-subject dependent predicate in (12c). Moreover, the free-subject dependent copulatives with ri in (7) and (38) are formally identical to those in (25) and (37), which are clearly unlike verbal complements. For these reasons, I do not analyse free-subject dependent predicates after perception verbs as predicative complements, but rather as participant-oriented adjuncts.

10.4 Participant-oriented free-subject dependent predicates: depictive or general adjunct construction? When trying to identify the grammatical constructions involved in the different uses of free-subject dependent predicates reviewed in the preceding section, some can be singled out fairly easily as pertaining to separate construction types on formal and semantic grounds. This is true of the ﬁrst two uses presented in section 10.3. Grammatically triggered dependent predicates (see 10.3.1) occur as parts of two construction types: either they constitute the nucleus of elaborate and syndetically marked adverbial clauses—their occurrence being triggered by the particular conjunctive element (example (16) above)—or they establish a semantically unitary predicate in conjunction with an auxiliary that represents its syntactic nucleus and determines subject reference (example (17) above).8 As predicates in adnominal indirect relatives (see 10.3.2), they belong to the modiﬁer of a complex nominal which is also indicated, in the majority of cases, by segmental relative markers. The occurrence of such relative markers and the temporal characteristics of the verbs (marking absolute rather than relative time) distinguish adnominal free-subject dependent predicates from their non-adnominal counterparts. This leaves two types of use of free-subject dependent predicates which minimally share the following negative features: the relevant verb form is not downgraded to a sub-constituent of a predicate, it has no purely grammatical 8 The use of dependent predicates as part of verbal periphrasis is crosslinguistically frequent and commonly distinguished from their use as clause nucleus. See e.g. Wilkins (1988) for a relevant case in the Australian language Mparntwe Arrernte and Amha and Dimmendal (Ch. 9, this volume) regarding the Omotic language Wolaitta.

Asyndetic subordination and deverbal depictive expressions in Shona

339

trigger, and it does not depend on a noun. These are the uses in asyndetic adverbial clauses (see 10.3.3) and depictive-like expressions (see 10.3.4)—a functional range that apparently motivated the philological label for this ‘‘mood’’. I will argue in this section that these two types cannot be clearly separated on formal grounds and hence together constitute what SchultzeBerndt and Himmelmann (2004: 78) call a general adjunct construction, deﬁned as follows: Depictive content is rendered by the same construction type as adverbial content. Differences between the two expression types exist, if at all, only on the semantic level. In this scenario, it only makes sense to speak of (general) adjunct constructions rather than of speciﬁcally depictive or adverbial constructions. (This does not preclude the existence of several adjunct constructions in a language, some of which may be speciﬁcally adverbial—the important point for our purposes is that there is no genuine depictive construction.)

‘Genuine depictive constructions’ are deﬁned by Schultze-Berndt and Himmelmann (2004: 77) by a list of features; these will be reviewed in the following for the free-subject dependent non-adnominal taxis predicates in Shona. 10.4.1 Separate predicate with interlacing time-frame vis-a`-vis main predicate The semantic-syntactic makeup of all free-subject dependent predicates ( ¼ no grammatical control and non-embedded) clearly qualiﬁes them as independent predicates separate from the main predicate. Sections 10.2 and 10.3 have also shown that, in their non-adnominal taxis use, they almost invariably express a state of affairs holding within the temporal frame of the main predicate. Only one taxis marker, the ka-sequential, never marks cotemporality and also is virtually limited to restricted subject reference (see section 10.2). It will therefore not be considered any further in this section. Otherwise, a non-simultaneous reading is only possible for two other taxis forms, namely the o-form, as in (11), and the chi-form, as in (39). (39) va-no-zo-sara va-chi-dyiwa ne-garwe 2-prs-then-remain 2-sim-be.eaten com-crocodile ‘They would remain behind and be eaten by the crocodile.’ Like the ka-form, these two forms must follow the main predicate, their subject is mostly controlled by a preceding nominal, and they encode a sequential meaning. This indicates a restricted and well-deﬁned tendency of some taxis predicates to extend into other uses, but does not contradict the general conclusion that they, as a set, meet Schultze-Berndt and Himmelmann’s ﬁrst criterion.

340

T. Gu¨ldemann

10.4.2 Obligatory control of argument I have discussed and exempliﬁed in previous sections the fact that free-subject dependent predicates, as opposed to forms of the other subordinate mood, the ‘‘relative mood’’, can in principle display their own subject in both adnominal and non-adnominal uses. From a typological perspective, the nonadnominal taxis predicates of Shona are thus comparable to ‘‘free-subject converbs’’. That they are not obligatorily controlled is a major difference to genuine depictives in the terms of Schultze-Berndt and Himmelmann. This statement, while generally adequate, needs some qualiﬁcation, though. First, there exists a construction that contrasts semantically with the relevant class of freesubject dependent predicates whereby the latter turn out to be clearly depictive. This is due to their overtly marked subject cross-reference, which is controlled by a main-clause participant. The contrasting construction is a true sentence adverbial which has the form of a headless direct relative clause with a stative predicate and the Class 8 subject concord preﬁx zvi-. The latter has an inherent connotation of manner—a feature widely attested for this class in Bantu—and is associated conceptually with the Class 8 noun zvinhu ‘matters, ways’, so that the construction in fact has an implicit nominal head. In the following, I give three examples, (40), (41), and (42), each to some extent providing a minimal pair with a participant-oriented freesubject dependent predicate presented earlier, namely (34), (25), and (31), respectively. The latter are given again in the (b) examples for a convenient comparison. (40) a. u-sa-ita zvaka-daro 2sg-neg.imp-do 8man.rs.stat-like.that ‘Don’t do it like that; don’t act that way.’

(Dale 1972: 176)

b. zvaka-ngo-ramba zvi-chi-ngo-daro nzira 8inan.rem.pst-just-persist 8-sim-just-like.that 9.path y-ose 9-all ‘These things continued all along the way just like that.’ (41) a. ndaka-ita basa ra-ngu 1sg.rem.pst-do 5.work 5.assoc-1sg.poss zvaka-rurama 8man.rs.stat-be(come).righteous ‘I did my work righteously.’

Asyndetic subordination and deverbal depictive expressions in Shona

341

b. Cemedzai waka-kura a-ri mu-sikana 1.pn 1.rem.pst-grow.up 1.fs-be 1-girl waka-rurama kwa‰ o 1.rs.stat-be(come).righteous very ‘Cemedzai grew up as a very righteous girl.’ (orig. ‘Cemedzai wuchs als ein sehr rechtschaffenes Ma¨dchen heran’) (Sicard 1965: 107) (42) a. ndaka-mu-tuka zvaka-shata 1sg.sbj.rem.pst-1obj-scold 8man.rs.stat-become.bad ‘I scolded him angrily.’ (Dale 1972: 41) b. aka-taura zv-ose izvi aka-shatirwa 1.rem.pst-speak 8inan-all 8dem 1.fs.stat-become.angry ‘He said all these things in a state of anger.’ (Dale 1972: 72) A second point regarding subject control concerns the general status of verbal subject cross-reference in Bantu. While the cross-reference markers could be analysed as expressing verbal arguments in some contexts, in other contexts they clearly function as mere agreement markers. To mention just one example, the agreement function is evident in direct relatives with subject inversion. This construction has been exempliﬁed in (5) for Shona as one among other relativization strategies, but it happens to be the only relative construction type in a number of other Bantu languages. In these relative constructions, the relevant preﬁxal concord does not index the verb’s subject but its object, which is the head noun of the complex nominal. Therefore, the verb preﬁx which normally cross-references the subject can as such not be taken as argument marking. Finally, there exists a strong tendency to express the subject of a freesubject dependent predicate separately as soon as it is not coreferential with a participant of the main clause. This is so even in those instances where the nominal subject does not serve referential disambiguation as with speech act participants. A relevant example is (22), where ini ‘I’ does not contribute any semantic information, owing to the fact that ﬁrst person singular reference is expressed also in the subject preﬁx of the free-subject dependent predicate ndisingade. It follows that dependent predicates without nominal subjects are ‘controlled’ by a main-clause participant. One might be tempted now to view non-adnominal free-subject dependent predicates with controlled subjects and with explicit subjects as each constituting a separate construction type, and to argue that the former represent true depictives. However, this hypothesis has little evidence in its favour.

342

T. Gu¨ldemann

On account of both formal and semantic features, there is a gradual cline rather than a clear-cut distinction between the different types of nonadnominal free-subject dependent predicate, so that it is more appropriate to subsume them under a general adjunct construction (see also below). I discuss additional data in section 10.5 showing that rendering a semantically depictive expression by means of a free-subject dependent predicate does not only depend on the criterion of control, but also on yet another important parameter, i.e. focus. Considering all the above remarks, the answer to the question of obligatory argument control in non-adnominal free-subject dependent predicates turns out to be generally negative, but not entirely so. 10.4.3 Semantic independence The type of taxis predicate at issue here is always semantically autonomous vis-a`-vis the main predicate. A ‘‘participial’’ verb form (otherwise functioning as free-subject dependent predicate) only occurs as a sub-part of a complex predicate expressing a unitary state of affairs in one of its grammaticalized uses brieﬂy outlined in section 10.3.1, namely as the content sign in a verbal periphrasis. However, this phenomenon, which is also relevant for converbs in other languages viewed as true depictives by Schultze-Berndt and Himmelmann, has been identiﬁed as a separate construction. 10.4.4 Non-argument status Taxis predicates are not obligatory complements/arguments of the main predicate. This is also valid for the instances occurring after perception verbs. This was discussed at the end of section 10.3.4, inter alia in connection with the opposition shown earlier between the true proposition-type clause in (12a) and its depictive-like counterpart in (12c), a free-subject dependent predicate in participant-oriented interpretation. 10.4.5 Non-modiﬁer status The delimitation of non-adnominal from adnominal free-subject dependent predicates has been treated in section 10.3.2 and the beginning of 10.4. To the extent that non-adnominal dependent predicates, when used depictively, are subject-controlled, they are similar to the second type of adnominal verb form in Shona: the group of restricted-subject dependent predicates used in direct relative clauses (see section 10.2). However, there is a prosodic distinction between the corresponding moods, ‘‘participial’’ vs. ‘‘relative’’ (in

Asyndetic subordination and deverbal depictive expressions in Shona

343

the traditional terms). Moreover, the arguments for a structural separation of non-adnominal and adnominal free-subject dependent predicates also apply here. The major semantic-functional difference was illustrated above by the contrast between the ‘‘relative’’ in (12b), with its absolute temporal deixis alias ‘‘tense’’ and its function of restricting the reference of an entity, and the participant-oriented free-subject dependent predicate in (12c). 10.4.6 Nonﬁnite status Dependent taxis predicates, most of which are assigned traditionally to the ‘‘participial mood’’, display subject cross-reference and create paradigms based on a fairly extensive range of verb forms. These are the suprasegmentally marked forms given in Table 10.1 and the segmentally marked chi-form, most of them having been represented in this chapter by relevant examples. At ﬁrst glance, these features do not make such taxis-marking freesubject dependent predicates look particularly ‘nonﬁnite’. Nevertheless, they qualify as nonﬁnite in terms of Schultze-Berndt and Himmelmann’s (2004: 77) deﬁnition, that is, ‘‘not marked for tense and mood’’. The variable temporality and modality features found in basic main-clause forms are all converted in their dependent-clause counterparts, which are suprasegmentally and/or segmentally distinct, into taxis categories as deﬁned in section 10.2. These are neutral with respect to modality and must also be distinguished from tense in the narrow deictic sense used here. Taxis marking is in fact also relevant for similar verb forms in other languages called ‘‘participles’’, ‘‘gerunds’’, ‘‘converbs’’, etc. which are used in depictive expressions and are unambiguously nonﬁnite. In line with the above description, as well as Schultze-Berndt and Himmelmann’s (2004: 102–5) discussion of ‘‘anterior-resultative deverbal depictives’’ marking a ‘‘post-state’’ and ‘‘prospective deverbal depictives’’ marking a ‘‘pre-state’’, I assume that most dependent taxis forms express a state of affairs holding ‘‘within the time-frame of the state of affairs expressed by the main predicate’’. The fact that Shona retains a high degree of inﬂectional variability (as shown in Table 10.1) must be seen as a function of the way in which Shona has recourse to a typologically fairly rare strategy of only employing suprasegmental features for the distinction of clause types, and thus leaving the segmental morphology largely intact. 10.4.7 Clause-level status The discussion of the above criteria of dedicated depictives has left out so far a very important feature. It is so basic that Schultze-Berndt and Himmelmann

344

T. Gu¨ldemann

(2004: 77) seem to have felt the need to integrate it in the initial phrase of their deﬁnition instead of listing it together with the other six features. It is the requirement that ‘‘a depictive secondary predicate construction is a clauselevel construction’’. In other words, the second predicative element pertains to the domain of a higher main clause. When applying this criterion to the whole range of non-adnominal freesubject dependent predicates or ‘‘participials’’ in Shona, it turns out that many tokens do not meet it and in this sense are not typical depictives. That is, these predicates display a variable position and prosodic integration vis-a`vis their main clause so that they cannot all be handled by the concept of a ‘‘clause-level construction’’. First, free-subject dependent predicates preposed to the main clause constitute adverbial subordinate clauses within a complex sentence and contain topical, backgrounded information outside the immediate assertion frame of the matrix. They are thus ruled out as candidates for depictives. This argument can also be applied to a group of postposed free-subject dependent predicates, namely those not integrated in the main clause prosodically. Since prosodic linkage is deﬁned ﬁrst of all by the pause behaviour of the respective predicate, it is difﬁcult to evaluate in written texts. A potential clue could be punctuation. In (43) and (44), for example, the comma could reﬂect that the dependent predicate constitutes its own clause unit so that the second English translation, which is less depictive-like, would be more adequate. (43) mu-rume aka-shanduka nyoka, a-ri mu-”undo 1-man 1.rem.pst-turn.into snake 1.fs-be iness-bush ‘The man turned into a snake in the bush / when/as soon as he was in the bush.’ (orig. ‘Der Mann verwandelte sich im Busch in eine Schlange.’) (Sicard 1965: 22) ku-mu-ˆuraya, ˆo-tya nyoka, (44) ˆa-ka-kona 2-seq-be.unable inf-1obj-kill 2.fs.coi-fear snake ‘and could not kill him in fear of the snake’ / ‘because he was afraid of the snake.’ (orig. ‘und konnte ihm aus Furcht vor der Schlange . . . nichts antun.’) (Sicard 1965: 368) While the evaluation of the integration of postposed dependent predicates according to punctuation must remain guesswork to a considerable extent, this is much easier with audio-recordings. The following is an example pair from my own ﬁeld data.

Asyndetic subordination and deverbal depictive expressions in Shona

345

(45) rai-zo-dya mu-komana uyu # 5.hab.pst-then-eat 1-boy 1.dem imbwa dza-ke dzi-si-po 10.dog 10.assoc-1poss 10.fs-neg-adess.anaph ‘ . . . it would then eat this boy while/because his dogs would not be there’ (# ¼ pause) (46) vai-zo-siya imbwa dzi-ri mu-musha 2.hab.pst-then-leave 10.dog 10.fs-be iness-home ‘ . . . they would leave the dogs in the village’ In (45), the clause with the negative locative ‘‘participial’’ phrase dzisipo is preceded by a pause so that it constitutes unambiguously a separate adverbial clause. This is in principle independent from the additional fact that it has its own subject, which points to the same analysis. In (46), the dependent verb form dziri and its locative adjunct belong to the same intonation unit as the main predicate, and the subject of the former is controlled by the object of the latter. The subtype of free-subject dependent predicate constituted by this and similar tokens is evidently akin to a genuine depictive secondary predicate construction. One can generalize that the alternation of non-adnominal free-subject dependent predicates with respect to position and prosodic linkage correlates with semantic-functional differences between the clauses thus established. When they are preposed to and outside the domain of the main clause, they instantiate adverbial background clauses.9 When postposed to and not contained in the main clause, they can be characterized as the nucleus of postposed adverbial clauses that are similar to ‘‘depictivelike reduced clauses’’ as discussed by Schultze-Berndt and Himmelmann (2004: 67–9; see also their introduction to this volume). The third and last subtype is constituted by postposed dependent predicates which are contained in the main clause; these represent the closest Shona gets to something in the way of dedicated deverbal depictives. However, since the features shared by these three subtypes of non-adnominal free-subject dependent predicates, in particular their marking properties, outweigh their differences, I do not grant them the status of separate constructions, 9 A semantic change in the clause-linkage type is a general effect of preposing a subordinate clause in Shona. For example, a dependent clause introduced by sezvi- encodes ‘how’ when occurring after the main clause, but ‘because’ when occurring before it. In a parallel fashion, a postposed clause with kuti conveys a ‘that’-type linkage, while a preposed one marks condition. This phenomenon is also found with the ‘‘participials’’, with ka and chi expressing temporal relations when following the main verb (see section 10.2) and condition when preceding it.

346

T. Gu¨ldemann

which I did with adnominal dependent predicates in indirect relative clauses and with grammatically triggered dependent predicates in syndetic clause linkage and auxiliary periphrases; I view them instead as constituting a general adjunct construction. Summing up the above comparison, it can be said that the class of nonadnominal free-subject dependent predicates in Shona consistently satisﬁes ﬁve of the seven criteria in Schultze-Berndt and Himmelmann’s deﬁnition of a dedicated depictive construction. The differences between the Shona strategy and the latter are twofold: (a) it has potentially free-subject reference so that it is not obligatorily controlled as a construction type; (b) it can be and in fact is frequently used as the nucleus of asyndetic adverbial subordinate clauses. Regarding the ﬁrst difference of free subject reference, however, comparative Bantu data could reveal in the future how close the relevant dependent predicates actually are to genuine depictives. It is conceivable that free-subject subordinate clauses, as soon as they come to show coreference regularly, could develop to obligatorily controlled clauses, ultimately yielding a dedicated depictive construction. Chewa, for example, possesses a converblike form ka´-stem-a with obligatory subject control ‘‘which may be translated as ‘while’ ’’ (Watkins 1937: 99), but possibly qualiﬁes as a dedicated depictive. This form is suspiciously close to a simultaneous-taxis verb form sbj.prefka-stem-a, which is widespread in Bantu (Gu¨ldemann 1996: 138–43; 1998); compare a likely cognate form in Shona exempliﬁed in (13) and (14) and even has a reﬂex in Chewa in a dependent predicate conveying ‘when, if ’. It should be investigated in the future whether these two verb forms are related historically in that the person-inﬂected predicate was grammaticalized to an obligatorily controlled predicate through the loss of subject cross-reference under coreference with a main-clause participant. If this idea turned out to be true, one would have a case on the diachronic dimension where a taxis predicate ended up as a dedicated deverbal depictive construction. If non-adnominal free-subject dependent predicates in Shona are analysed as a whole, they are best conceived of as covering a continuous scale of syntactically de-ranked predicate types which differ from each other in their kind and degree of formal and semantic interlacing with the main predicate. This is shown in Figure 10.1. The extreme poles of this scale are constituted by the two types of grammatically controlled dependent predicate referred to in Section 10.3.1: those used as the nucleus of fully elaborate and syndetically linked adverbial clauses are on the end of greatest syntactic and semantic freedom, while those controlled inter alia in their subject reference by an auxiliary

Asyndetic subordination and deverbal depictive expressions in Shona Construction syndetic adverbial type cause linkage

asyndetic general adjunct construction

auxiliary periphrasis

loose bond Subtypes

347

strong bond preposed adverbial

postposed adverbial

postposed depictive

Figure 10.1 Constructions with non-adnominal free-subject dependent predicates on a scale of semantic-syntactic bond with the main predicate

within a verbal periphrasis are on the other end of extreme syntactic and semantic downgrading. In between, one ﬁnds all asyndetic ‘‘participial’’ types discussed in this section, which can be assigned to the asyndetic general adjunct construction. The subtype displaying the tightest bond with the matrix clause, i.e. postposed, prosodically integrated free-subject dependent predicates, can be identiﬁed as the threshold to dedicated depictives. In Gu¨ldemann (1997a; 1997b), I have proposed that free-subject dependent predicates alias ‘‘participials’’ have started out in Shona as a functionally unmarked and syntactically fairly loose adjunct construction at the margin of a clause, and only acquired a closer semantic and syntactic relation to the main clause nucleus in certain contexts. This also throws light on the apparent anomaly that, in its depictive use, this construction type regularly covers parts of the less typical semantic range of depictive content (e.g. location), but has gaps in its central domain (e.g. quantity). This appears to be related to the history of the construction: its more adverbial-like uses like the expression of time, location, manner, etc. have diachronic precedence over the more participant-oriented functions of quantity, condition, etc. That is, dependent ‘‘participials’’ have progressed in their range of uses from the outer periphery to the inner core of a clause, whereby one of such variants has come to provide a major portion of depictive expressions in Shona.

10.5 Assertive focus as an additional feature of depictives The previous section established that a depictive interpretation of free-subject dependent predicates in Shona is tied to the clause position after the verb or verb phrase. It is important to recognize in this respect that the linguistic material in this position normally constitutes, or is at least part of, the asserted information of a sentence in Shona (and other Bantu languages).

348

T. Gu¨ldemann

Accordingly, it can be concluded that depictively used dependent predicates represent the assertive focus10 of a clause, like other items in this position. This functional analysis can also be used to clarify the feature of ‘‘being in the domain of the main clause’’, discussed in section 10.4.7 above. The characteristic of being postposed to the clause core and prosodically integrated is a formal reﬂex of the fact that depictive dependent predicates are parts—in fact, central parts—of the assertion made by the sentence containing them. This functional distinction, in conjunction with the criterion of subject control, is capable of excluding all non-adnominal dependent predicates of Shona that do not qualify as expressing depictive content. Dependent predicates which are preposed and those which are postposed, but prosodically detached, are outside the information structure of the main clause. Dependent predicates that are grammatically controlled by auxiliaries are inside the clause, but do not constitute independent focal information units. This observation on depictively used dependent predicates in Shona leads to the hypothesis that an additional feature could be valid for depictives from a more general perspective: they are in the domain of the assertive focus of a clause. As indicated above, this is in fact contained implicitly in SchultzeBerndt and Himmelmann’s ‘‘clause-level’’ criterion. While not mentioning it explicitly in their deﬁnition (but see Himmelmann and Schultze-Berndt, Ch. 1, this volume), these authors also recognize and discuss its explanatory potential, as well as its formal correlate of prosodic integration (SchultzeBerndt and Himmelmann 2004: 91). At least two points should be made regarding this proposed characteristic. First, it only helps to distingush a depictive construction from some related constructions, but does not single it out against all of them: while adnominal modiﬁers or many resultatives indeed differ from depictives in not being directly asserted (see also below), clause-internal adverbial expressions seem to have a focus behaviour similar to that of depictives. The limits of the feature’s heuristic potential is, of course, not particularly remarkable because this applies also to the other criterial features of depictives. Second, the assertive-focus status only represents the default case, but does not hold for all and every depictive token. This becomes clear from the observation that 10 The classiﬁcation of focus types and the terminology are parallel to Hyman and Watters (1984) and Dik (1997a). It is important to note in this respect that the communicative point of assertive focus is a simple information gap which the speaker expects to exist for the hearer. This is an unmarked pragmatic constellation which must not be confounded with the situation where constructions of contrastive focus occur. Linguists repeatedly fail to recognize the former as pertaining to the domain of focus so that focus, per se is equated inappropriately with contrastive focus.

Asyndetic subordination and deverbal depictive expressions in Shona

349

even genuine depictives can appear in clearly extrafocal environments as in (47) where barfuß constitutes the clause topic (cf. also Himmelmann and Schultze-Berndt, this volume, section 2.3). German (47) barfuß wirst du den nicht mal am Strand zu sehen barefoot will you him not even on the beach to see bekommen get ‘Barefoot, you wouldn’t even spot him on the beach.’ However, examples like (47) seem to represent marked cases vis-a`-vis the general proﬁle of depictives in terms of information structure. It should in fact be expected that more complex clause conﬁgurations (e.g. with an extraposed topic, with more than one focus, or with contrastive focus) can override phenomena found in the least-marked pattern of one assertive focus per clause. On account of the data available, I therefore maintain at present that the assertive-focus status of depictives is a robust generalization for central instances of this expression type. Having made these qualiﬁcations, I will try to show in the following that there are indeed indications that this functional feature is also relevant for depictives in other languages. (i) A ﬁrst characteristic likely to corroborate the focal nature of depictives is that they can be distinguished in some languages from other closely related expressions by displaying some sort of prosodic salience, as reported by Schultze-Berndt and Himmelmann (2004: 83–4). In some Australian languages (see Austin 1981a: 108; McGregor 1992a: 316; Hale 1994: 193), stress on a secondary predicate disambiguates a depictive modiﬁer constituting (part of) the asserted focus from a segmentally identical attributive modiﬁer conveying information which is backgrounded and not asserted directly. Another case is the following pair of sentences from German: the secondary predicate roh in (48a) is a depictive and accordingly receives word stress (indicated by an accent), while the secondary predicate blank in (48b) is a resultative which cannot bear stress. German (48) a. er ißt das La´mm ro´h he eats the lamb raw ‘He eats the lamb raw.’ (Winkler 1997: 291) b. dort fegt der Wind den Hı´mmel blank there sweeps the wind the sky clear ‘There the wind sweeps the sky clear.’ (Winkler 1997: 282)

350

T. Gu¨ldemann

(ii) A linguistic detail from Jacaltec (and other Mayan languages) points in the same direction. The construction type exempliﬁed in (49) is a major device to express typical depictive content in these languages. Jacaltec (49) tz’op xu spitzc’a naj blind did was.born clf ‘He was born blind.’ (lit. ‘It was blind that he was born.’) 1977: 329)

(Craig

It can be seen that the depictive tz’op ‘blind’ is the main asserted predicate of a cleft construction. Such a structural equivalent to more canonical depictives of other languages seems ‘‘rather surprising’’ at ﬁrst glance. However, assuming a general focality of depictive expressions, the highlighting of the relevant state of affairs within the focal main clause of a bisected construction appears in a very different light, turning out to be merely a variation on a universal theme. (iii) Another similar hint comes from the behaviour of a type of verb form marking assertive predication focus observed in Nguni varieties. This so-called ‘‘long form’’ stands in opposition to the morphologically unmarked ‘‘short form’’ occurring in contexts where the predicate is extra-focal. Doke (1992: x807) writes regarding the occurrence of this functional distinction with depictive taxis forms: ‘‘When the verb is followed by a participial tense, the short form is used . . . ’’ Compare the following example showing that the use of the secondary predicate eqhuga blocks in the main clause the use of the focal ‘‘long form’’ of the present marked by the preﬁx ya:11 Nguni (50) a. u-hamba e-qhuga 1-walk:prs 1.sim-limp ‘he walks limping’ (Doke 1992: x807)

b. *u-ya-hamba eqhuga 1-foc-walk:prs 1.sim-limp ‘he walks limping’

In other words, the depictive dependent predicate has the same effect for the main predicate as any other non-predicate constituent that is the utterance focus, such as asserted nominal or adverbial terms, content interrogatives, and focus markers. This phenomenon receives a straightforward explanation as soon as the assertive-focus nature of depictive secondary predicates is recognized. They are largely incompatible with a 11 It is not excluded that the b version is a possible utterance. The important point is that the ‘‘long form’’ followed by a participial does not express assertive focus.

Asyndetic subordination and deverbal depictive expressions in Shona

351

second assertive focus, here on the main predicate. It would be worth investigating in more detail whether, in the presence of a depictive, the occurrence of other focus constituents is marked or even impossible in other languages, too. (iv) Simpson (Ch. 2, this volume) reports from English a detail giving evidence to the same effect, namely that ‘‘apparent depictives can co-occur with both resultatives and depictives, but the situations in which they can do so are quite constrained’’. This approximate complementarity means that clauses do not normally have two depictives that are independently predicated. Again, the functional explanation would be that clauses with more than one asserted focus are pragmatically marked and therefore likely to be avoided. (v) Focality of depictives also seems to have a correlate in the semantic proﬁle of dedicated depictives as far as this can be discerned from the data in Schultze-Berndt and Himmelmann (2004; cf. also Himmelmann and Schultze-Berndt, Ch. 1, this volume, Section 1.4.1). If the present assumption is adequate, these can be expected to have a bias towards encoding states of affairs which are in some sense more salient pragmatically. This is indeed corroborated by the available data in two respects. On the one hand, ‘‘expected’’ or ‘‘habitual’’ depictive content—that is, states of affairs which are culturally unremarkable—are not expressed by dedicated depictives in several languages. Instead, there is a tendency to integrate such expressions into the main predicate to varying degrees. A possible motivation, in line with the present approach, is that they are not prone to be encoded by a structure which is inherently focused. SchultzeBerndt and Himmelmann (2004: 69–72) present several cases involving complex predicates or incorporation, e.g. in Mayali and Turkish. For the complete conﬂation of the depictive content with the main predicate consider (51) from Ilokano. Ilokano (51) kinilaw ¼ da ti sida rls.pv:raw ¼ 3pl.poss art ﬁsh ‘They ate the ﬁsh raw.’ (lit. ‘They ‘‘rawed’’ the ﬁsh.’) (C. Rubino, p.c.) On the other hand, expressions which are restricted to depictive function frequently seem to convey concepts which are in some sense remarkable as secondary predicates or at least represent the marked member of a conceptual opposition. A well-known example is German barfuß ‘barefoot’ and some of its counterparts in various other European languages

352

T. Gu¨ldemann

belonging to a small class of so-called ‘‘predicative adjectives’’. In so far as this concept is the marked value vis-a`-vis ‘with shoes’, at least today in many parts of Europe, it can be argued that its greater pragmatic signiﬁcance, i.e. counter to the general cultural expectation, has fostered its encoding by means of a structural class which is inherently focused. Another similar case is the uniquely depictive red-handed in English, insofar as the general cultural expectation is that offenders manage at least to get away from the place where they commit the deed. Another possibly parallel example given by Schultze-Berndt and Himmelmann (2004: 97) is Latin invitus ‘reluctant’. It is possible that there are some semantic universals as to which member of such an opposition pair of concepts is more apt to being expressed by a conventionalized depictive. However, several examples given above indicate that the formal behaviour of a great portion of such expressions is not steered by universal tendencies, but rather by culture-speciﬁc norms and the expectations resulting therefrom. To mention just one obvious case: while the concept of ‘eating (meat/ﬁsh) raw’ is culturally marked and expressed by a construction that typically conveys depictive content in Europe and other areas (cf. (48a) from German), example (51) from Ilokano shows that it is common in other cultures and can therefore be treated formally according to alternative strategies that are less typical for depictives. (vi) Schultze-Berndt (2002) and Himmelmann and Schultze-Berndt (Ch. 1, this volume) also report that depictives in some languages involve linguistic material that is related to restrictive focus morphemes encoding ‘just, only, still’ (see also Bucheli Berger, Ch. 4, this volume). In line with the functional explanation given by the above authors, such a recurrent formal association between focus and depictives can be motivated by the inherent focus nature of the latter. (vii) A ﬁnal observation relates to an emerging correlation between the unmarked position of depictive expressions and the general word order pattern in an individual language. Judging from the data in Schultze-Berndt and Himmelmann (2004) and my familiarity with several languages of different types, VO languages tend to have depictives occurring after the verb or verb phrase while OV languages have them before it (see also Amha and Dimmendaal, Ch. 9, this volume, for the same conclusion). Without being able to lay out here the evidence for this, several crosslinguistic indications lead me to assume that the position of a single nominal object with respect to the verb is the general position of the assertive clause focus in many languages. Provided this assumption is correct, the parallel between the order of objects and depictives would be a result of the similar pragmatic property of

Asyndetic subordination and deverbal depictive expressions in Shona

353

these constituent types: namely that they present, in the unmarked case, new asserted information. Summing up the above data, the evidence seems sufﬁcient for justifying a systematic investigation of the idea that the inherent focal status of depictives is a promising criterial feature with crosslinguistic relevance, and has heuristic potential for distinguishing such expressions from formally related ones.

10.6 Summary This chapter has dealt with one of the major expression types for participantoriented adjuncts in the Bantu language Shona. This construction is deverbal in that ﬁnite predicates which are syntactically dependent on the main predicate convey the depictive content. Since the subordinate status of the secondary predicate is rendered primarily by suprasegmental features, the segmental morphological variation of these verb forms approaches that of main-clause forms; their semantics, however, are reduced from tense, aspect, and modality meanings to several types of cotemporal taxis. From a typological perspective, I propose to analyse this type of dependent predicate in line with Schultze-Berndt and Himmelmann (2004), as a ‘general adjunct construction’, because it has not only depictive functions but also several types of adverbial functions. The depictive reading arises when the secondary predicate is controlled by a participant of the matrix predicate and is in the domain of the assertive focus of the entire clause. On the basis of Shona and crosslinguistic data, the inherent nature of assertive focus is proposed as being a criterial feature of depictives in general.

This page intentionally left blank

11 Forms of secondary predication in serializing languages: on depictives in Ewe FELIX K. AMEKA

11.1 Introduction This chapter has two main goals. First, I want to argue that Ewe, a verb-serializing Kwa language of West Africa, has nominal depictive secondary predicates, contra the claim by Larson (1991) that serializing languages predominantly have verbal secondaries. . . . assuming serial constructions to be a form of secondary predication, the chief difference between a serialising language like Yoruba [or Ewe FKA] and a nonserialising language like English reduces to a matter of what secondary predicate categories are allowed. . . . Serialising languages show non-nominal secondaries, whereas non-serial languages show non-verbal secondaries. (Larson 1991: 206)

A second aim is to demonstrate that the same form ¼i which is optionally used to mark nominal secondaries also occurs in subtypes of serial verb constructions (SVCs). This raises the question of whether the SVCs could be characterized as instantiations of depictive situations. This question will be explored in the light of the deﬁnition of depictives proposed by SchultzeBerndt and Himmelmann (2004), which excludes serial verb constructions. The plan of the chapter is as follows. First, a typological overview of the language is given (section 11.2). Some structures that function as equivalents of depictive constructions, including different types of multiverb constructions, are outlined in section 11.3. The predication marker ¼i is then introduced in For discussions of some of the issues discussed here, I am grateful to Evershed Amuzu, Chris Collins, James Essegbey, and Eva Schultze-Berndt. I am also greatly indebted to the editors, Nikolaus Himmelmann and Eva Schultze-Berndt, who read earlier versions and offered many valuable criticisms and suggestions for improvement.

356

F. K. Ameka

section 11.4. In section 11.5, I discuss the properties of some types of serial verb constructions marked by the predicative marker and relate them to the deﬁning characteristics of participant-oriented constructions.1 Section 11.6 summarizes the chapter and defends the view that serializing languages have nominal secondaries.

11.2 Typological overview of the language Ewe is a tone language spoken in West Africa. It is isolating with agglutinative features. It makes use of compounding, reduplication and triplication, and afﬁxation processes in the formation of new words. It has ideophones which are multi-categorial (see Ameka 2001). Ewe is an SVOX language. Grammatical relations are deﬁned by constituent order and distribution. The subject must be expressed at least once in each clause including imperative constructions. It is a hypertransitive language with obligatory expression of both Actor-like and Undergoer-like semantic arguments in a simple clause (see Essegbey 1999). Ewe has both prepositions and postpositions (Ameka 2003a). The complement of a prepositional phrase can be an NP, a postpositional phrase or a clause. Postpositional phrases can function in argument positions in a clause. Clausal negation is expressed by a bipartite morpheme: me´ . . . o; me´ is cliticized onto the ﬁrst element in the verb cluster while o occurs at the end of the clause but before any utterance-ﬁnal particles. Constituent negation is marked by the privative ma- involving nominalization and adjectivalization. The nominalized privative forms function like any other nominal and, as we shall see below, can be used as depictives. Consider the example below, 1 The Ewe data are drawn from various sources: oral performances of folk tales and other verbal arts and TV and radio drama, as well as written plays and ﬁctional novels. The written sources include the following works: Ayeke, Kodzo (1996) Hl~Obiabia [Revenge-taking]. Accra: Bureau of Ghana Languages. Ayeke, Kodzo (1998) Asitsu at~Oawo [The ﬁve male rivals]. Accra: Bureau of Ghana Languages. Bii Setsoaﬁa, H. K. (1982) Fia Tsatsala [The wandering chief]. Accra: Bureau of Ghana Languages. Bureau of Ghana Languages (1967) e modzaka [Entertain thyself]. Accra. HlOmatsi, Yao (1995) Agbe nye nusi newOe [Life is what you make it]. Accra: Bureau of Ghana Languages. Kwamuar, Sebastian (1997) EwO moya na TOgbi AgOkOli [It surprised Chief Agokoli]. Accra: Bureau of Ghana Languages. Obianim, Sam (1990) Agbezuge. Accra: Sedco.

The examples culled from these sources are referenced in two ways. If they are taken from the paper version, the author, year and page number are given. If the example is cited from my electronic database, it is referenced with the author and year and a concordance line number in square brackets. Utterances heard during conversations as well as examples in the linguistic literature are also used. As a native speaker, I have constructed some of the examples, all of which have been cross-checked with other native speakers for acceptability and interpretation.

Forms of secondary predication in serializing languages

357

in which the head nominal of the possessive phrase is a form derived from the verb nyo´ ‘become good’ by privative preﬁxation and reduplication: (1)

e´¼f e´ ma-nyo-ma-nyo´ 3sg¼poss priv-become.good-priv-become.good ‘its/his/her not being good’

Adjectives, derived or underived, are only used attributively. Adjectives are not used predicatively without derivation (e.g. with the adverbializing sufﬁx discussed in section 11.4.3). Properties can be predicated of entities in a number of ways: by the use of property verbs, the use of adverbials, the use of the ‘do’ verb wO plus a property nominal or ideophonic word (see example 2a), or the use of the locative verb le ‘be at’ plus an adverb or an ideophonic word (see 2b). (2) a. E´¼wO tsi 3sg¼do water Noun ‘It is watery.’ b. E´¼le legbee 3sg¼be.at:prs ideo:long ideophonic adverb ‘It is long.’ Property verbs are used in underived form in predicate position, but in attributive position a derived adjective has to be formed. In example (3a), the transitive property verb nya´ ‘know’ is used with its obligatory complement nu´ ‘thing’ to predicate a quality of the NP subject. In (2b) an adjective derived from the property VP in (3a) by compounding its component parts is used attributively. Similarly, in example (2c), the privative form of the derived adjective in (3b) is used attributively. (3)

a. ame¼a nya´ nu´ person¼def know thing ‘The person is clever.’ b. ame nya´-nu´´ person know-thing:hts ‘clever person’ c. ame ma-nya´-nu´´ person priv-know-thing:hts ‘unintelligent person’

Constituent focus is signalled by placing the focused constituent, followed by the particle/clitic (y)e´ ‘foc’, in initial position within the core clause

358

F. K. Ameka

(see (17) for an example). If the fronted constituent is linked to a non-subject function in the clause, it need not be marked by the particle (Ameka 1992). Ewe makes use of at least three types of multiverb constructions: a serial verb construction (SVC), an overlapping clause (Duthie 1996), and a consecutive clause. The SVC and the overlapping clause are of immediate relevance in the context of the present paper and are introduced in the next section.

11.3 Strategies for expressing participant orientation in Ewe Ewe does not seem to have a speciﬁc construction whose sole function is to mark a constituent as being depictive. A clitic ¼i ‘pred’ which is used to mark various adjunct structures is the closest to being a dedicated form for this function (see section 11.4). However, there are several constructions that can be interpreted as participant-oriented. These structures include various adjunct NPs and PrepPs (sections 11.3.1 and 11.3.2) and a similative construction marked by abe´ . . . ene´ ‘like . . . as’ (section 11.3.3). In addition, ideophones in predicative and adjunct positions in a clause can encode the condition or state of a participant in the rest of the clause (11.3.4). Moreover, some functional types of serial verb construction (and multiverb constructions in general) involve participant orientation, especially those dealing with associated posture or appearance. These are discussed in section 11.3.5. In each of these structures, the constituent with depictive content or the structure as such can be marked with a clitic/particle ¼i. However, the function of this form is wider than marking depictive status. The range of constructions in which this marker occurs and their functions are discussed in sections 11.4 and 11.5. The latter section is dedicated to the serial verb construction types in which the marker ¼i occurs, and relating it to the issues of serialization and secondary predication. 11.3.1 NPs in adjunct slot In the examples below there is no formal indicator of a secondary predicative function. However, the highlighted NPs provide information about the condition or state of a participant linked to a core clause argument position. As such they function as participant-oriented adjuncts. In (4a) the NPs are in focus position in the two clauses but are not overtly marked by the focus marker (y)e´. In the ﬁrst clause, the fronted NP pertains to the condition of the participant with the object function while in the second clause, which is intransitive, the fronted NP relates to the participant linked to the subject function. The relevant NP in (4b) consists of a noun and an adjective, and characterizes the state in which the participant that functions as subject in the

Forms of secondary predication in serializing languages

359

SVC carried out the motion. In (4c), the highlighted NP is a predication on the object NP aha la´ ‘the drink’, specifying quantity. (4) a. Ama´ma wo´¼dzi¼m; ama´ma ma¼ga¼yi . . . naked 3pl¼bear¼1sg naked 1sg.pot¼rep¼go ‘Naked I was born; naked I shall return . . . ’ (opening lines of a dirge) b. e´¼gbO va´ ası´ fu´f lu 3sg¼return come hand empty ‘S/he came back empty handed.’ c. ame sia´a ame no-na aha la´ vi a´e´ person all person drink-hab drink def little indef ‘Everybody drinks the drink a little.’ Quantity is expressed within the NP by quantiﬁers and adjectives with quantity semantics. Both classes of modiﬁers come after the head noun but before a determiner. Thus in (4c) above, if there were no deﬁniteness determiner before the quantity expression then we would have a simple NP functioning as object: aha vı´ a´ e´ (drink little indef) ‘a little drink’. If the adjunct NP in (4c) were oriented towards the event as a whole, a derived adverbial form would have to be used, yielding an expression like . . . no-na aha la´ vı´e´ (drink-hab drink def a.little) ‘drinks the drink a bit’.2 The NP vi a´ e´ ‘little (nominalized) indef’ in (4c) is an adjunct but it is predicated of the object ‘the drink’ in the clause. In terms of its meaning and function it is depictive-like. 11.3.2 Prepositional phrases Information pertaining to a companion, or the location, or condition of a participant during the manifestation of a situation can be expressed in a prepositional phrase in adjunct position in a clause. Such phrases may be headed by the comitative preposition kple´ ‘com’ (5a) or le ‘loc’ (5b). (5) a. e´¼trO! dzo´ le Agbeko gbO! kple´ dziku´ 3sg¼turn leave loc name place com anger ‘He turned away from Agbeko in anger.’ b. e´¼wO dO¼a le dOme-Vı´ me 3sg¼do work¼def loc stomach-white containing.region ‘S/he did the work in hunger (on empty stomach).’ 2 The form vı´-e ‘little’ is a de-adjectival adverb composed of the basic adjective vi ‘small’ and an adverbializing sufﬁx-i which undergoes dissimilation. Incidentally, the nominalized form of the adjective vı´ ‘small’ by low-tone preﬁxation yielding vi~ ‘small, little [noun]’ is what is used as the head of the NP in vi a´ e´ ‘a little’.

360

F. K. Ameka

In (5b), the complement of the locative preposition is a postpositional phrase headed by the postposition me ‘containing region’. The conﬁguration or position of a participant while carrying out an event may be expressed using similar locative prepositional phrases. For instance, if an event was performed by a participant standing, the phrase le tsi-tre nu (loc remain-upright mouth) ‘in standing position’ could be used. 11.3.3 Similative constructions One of the ways of expressing similarity in comparison in Ewe is by the use of a similative or semblative construction with the standard of the comparison circumscribed by the form abe´ X ene´ ‘like/as X as’. The standard of comparison (X) can be an NP or a clause. When the similative construction is used to predicate a role of a participant in the clause, it can be interpreted as having depictive-like function. Example (6) is a serial verb construction introduced by the temporal subordinator ha´fı´ ‘before’ and the similative construction pertains to the role of the referent of the ﬁrst object NP (Zanu) as a teacher. (6) ha´fı´ wo´¼e Zanu va´ mı´a de afı´ı` before 3pl¼take.out name come 1pl hometown here abe´ nu´fı´ala´ ene´ simil teacher simil ‘[I noticed this] before Zanu was transferred to our hometown here as a teacher.’ (Ayeke 1996 (2432–3)) 11.3.4 Ideophones Ideophones can be used to denote qualities, manners, or properties of entities or events. In Ewe, ideophones are multi-categorial and can belong to the grammatical word class of adjectives, nouns, quantiﬁers, verbs, or adverbs (Ameka 2001). When used in predicative adjunct position, ideophones can be interpreted as participant-oriented (see also Amha and Dimmendaal, Ch. 9, this volume). As noted above, properties can be predicated of entities by using a verbal construction headed by the performance verb wO ‘do’ or the locative verb le ‘be at’ and an ideophonic property word (see example 3b). The latter is illustrated in (7), which is the standard response to a riddle. (7) MO!¼a way¼def bo¼O lot¼def ‘The road

le lu´bu´i be.at:prs ideo:narrow fo le gbadzaa surface be.at:prs ideo:ﬂat.and.wide is narrow while the farm is huge and wide.’

Forms of secondary predication in serializing languages

361

Ideophones in adjunct position in sentences with semantically speciﬁc verbs can also be interpreted as expressing a participant’s condition or state during the event. This can be illustrated with respect to some ideophones that can be used in the description of motion events. Westermann (1930: 107–9) lists about forty ideophones that can be used to modify the general move verb zO ‘walk’ ‘‘according to the manner of going’’. In some of the uses, however, some of the ideophones pertain more to the participant than to the manner of walking. For instance, the ideophone hloyihloyi, which Westermann glosses in combination with the verb zO as ‘to walk with many objects, clothes etc. dangling round one’, describes the condition in which the ﬁgure is, rather than the manner of movement. An example of this participant-oriented usage is (8), which is a serial verb construction. (8) e´¼zO va´ yi hloyihloyi 3sg¼move come go ideo:objects.dangling.round ‘S/he walked past with objects and clothes dangling round him/her.’3 Similarly, an ideophone like kodzokodzo ‘body bent forward stooping’ predicates a posture or conﬁguration of a ﬁgure while in motion. Such ideophones contrast with others which either conﬂate event and participant orientation in a motion event description (e.g. e´ ¼zO bOhObOhO [3sg¼move ideo:heavy.walk.of.a fat.person] ‘A fat man moved with a heavy walk’) or focus on the manner of movement (e.g. zO kpo´o´ kpo´o´ ‘walk quietly’). Thus not all usages of ideophones are participant-oriented. Although in some usages they can be interpreted as having depictive function, they are not dedicated solely to this function. 11.3.5 Serial verb constructions and other multiverb structures Like the structures discussed in the preceding sections, some instantiations of multiverb constructions can express depictive-like meanings. I will ﬁrst introduce serial verb constructions (SVCs) and then overlapping clauses, illustrating the forms that express participant orientation. An SVC in Ewe is a sequence of two or more verb phrases (including any complements and adjuncts) without any marker of syntactic dependency linking them. The VPs in the sequence are construed as occurring within the same temporal frame, share the same mood (e.g. imperative), and can be formally marked for different aspect categories. All VPs in the series share the 3 Since the condition of having objects dangling on a person will inevitably affect his/her manner of movement, the interpretation of the orientation of this ideophone hloyihloyi could be rather vague between event or participant orientation—an illustration of a situation where it is sometimes impossible to distinguish the two.

362

F. K. Ameka

same syntactic subject, but it is expressed only once with the initial VP. The individual verbs can function as independent verbs in simple clauses (in the same form). The verbs can be individually focused or questioned, but the VPs cannot be formally independently negated, although the negative operator may be interpreted as scoping over individual component VPs. The VPs together function as a single clause. There is no limit on the number of VPs that can constitute an SVC, except for restrictions on production and comprehension (see Ameka 2002; 2006). Some tokens of this construction can express depictive content without being formally marked as such. Consider example (9). (9)

e´¼[do´ dzi]vp [do go]vp 3sg¼put heart exit outside ‘She came out in courage (and stood in front of the crowd)’ (Ayeke 1998:15)

In this example, the ﬁrst verb and its complement denote and attribute a condition (that of being courageous, lit. ‘having put on heart’) to the subject and in that state the subject carried out the state of affairs represented in the second VP (‘coming outside’). Another instantiation of SVCs which can be seen as an equivalent of depictive structures in other languages is one where the subject is in a posture or in certain clothing while carrying out the state of affairs described in the second VP. For example: [le (10) a. Mı´e´¼[le klo dzı´]VP1 1pl¼be.at:prs knee upper.surface be.at:prs ´ ]VP2 ku´ku´ e-m hat remove-prog ‘We are on our knees begging.’ b. E´¼[ta avO]VP1 [yi dOme]VP2 3sg¼wrap cloth go work ‘She went to work in cloth.’ SVCs of the West African type have been analysed as having multiple VP heads which are co-dependent on each other syntactically and semantically (e.g. Dechaine 1993; Bodomo 1997; Ameka 2006). Even though constituent VPs function as heads, each VP can have an independent interpretation with respect to features such as negation, focus, or question. It is this possibility that seems to allow for the interpretation of certain types of SVC as being equivalent to depictive expressions in other languages.

Forms of secondary predication in serializing languages

363

Similarly, predicate structures within a multiverb construction known as the overlapping clause (OVC) in the Ewe linguistic literature can be interpreted as being oriented to one of the participants. An OVC is different from the SVC in being a biclausal structure. It consists of two clauses juxtaposed to each other without any overt connector. The referents of the subjects of the two clauses are different and each must be obligatorily expressed. The subject argument of the second clause is coreferential either with a non-subject argument of the ﬁrst clause, as in (11a), or with the situation characterized by the ﬁrst clause (see (11b)), or it indexes the spatio-temporal features of the situation represented in the ﬁrst clause (example (11c)). (11) a. E´¼a nu´ na´ mı´ mı´e´¼u 3sg¼cook thing dat 1pl 1pl¼eat ‘She cooked for us and we ate.’ b. E´¼fo¼m wo`¼se˜´ 3sg¼hit¼1sg 3sg¼hard ‘S/he hit me it was hard’, i.e. ‘S/he hit me hard.’ c. [Context: Two people discussing their attendance at the Old Students Association meetings and one says]: Me¼de e´¼didi 1sg¼reach 3sg.imprs¼be.far ‘I have been [to these meetings] it is a long time now.’ (overheard conversation, August 2003). Each clause in an OVC can be independently negated and marked for its own aspect and modality values. However, the two clauses should share the same temporal frame. The individual verbs in the clauses can function as independent verbs in simple mono-verbal clauses. The second clause in an overlapping clause is used to express various relations such as manner, result, consequence, or simultaneity relations as well as an evaluative stance with respect to the ﬁrst clause (see Ameka 2003b). The second clause can also be oriented towards a participant in the ﬁrst clause as in (11a) or (12). (12)

wo`¼nya´-se e`¼gblO¼e 2sg¼say¼3sg 3sg¼mod-hear lit: ‘You said it and it was hearable.’ (i.e. ‘It was pleasant.’)

In (12), the second clause characterizes a property of an argument, the content of the speech, or arguably of the state of affairs in the ﬁrst clause. If it was a monoclausal structure it could have been considered a kind of depictive or manner adjunct. The OVC as such being biclausal is not a depictive

364

F. K. Ameka

structure. However, some structures that can be interpreted as having depictive content can be paraphrased as OVCs, as we shall see below, especially in section 11.5.

11.4 The predicate (predication) marker ¼i Many categories of expressions and constructions outlined in the previous subsection such as adjunct NPs, ideophones, and VP components of SVCs can be optionally marked by the form ¼i in various constructions. I suggest that this marker is a predication marker, and furthermore that some of the constituents that are so marked can be analysed as secondary predicates. The contexts in which the predication marker is found are outlined in the next two sections. This section focuses on its use to mark nominal and ideophonic constituents with adjunct or objective complement function, while section 11.5 concentrates on its use in serial verb constructions. Broadly speaking, there are four groups of functional contexts in which the marker ¼i is found. First, it is used to mark objective complements of verbs of creation, cognition and perception in three place constructions (section 11.4.1). Second, it is used to mark nominal (including privative nominalizations) and ideophonic expressions in adjunct function that provide information about one of the core participants in the clause during the evolution of the state of affairs characterized by the main predicate in the clause (section 11.4.2). Third, it is used in certain types of SVC to mark a component VP as pertaining also to a non-subject argument in the clause (section 11.5). These three uses contrast with a fourth usage in the morphological status of the marker: for the three uses mentioned so far the form is a clitic, but in the fourth usage the form is a derivational afﬁx. It is thus used as a denominal and deadjectival sufﬁx (section 11.4.3). As a clitic, the marker is optional in all the ﬁrst three contexts of use outlined above; the present section on the predication marker will therefore conclude with some observations on its optionality (11.4.4). Ameka and Schultze-Berndt (2000) demonstrated that the contexts in which the predication marker occurs (especially in its clitic form) are analogous to the structures that have been analysed as depictive secondary predication in languages across the world. Part of the argument is based on the functional relationship between predicative complements and depictive adjuncts (see. e.g. Plank 1985; Schultze-Berndt and Himmelmann 2004; Bucheli Berger, Ch. 4, this volume; van der Auwera and Malchukov, Ch. 13, this volume). This line of argument is further pursued below.

Forms of secondary predication in serializing languages

365

For ease of identifying the marker in the examples provided, it should be noted that the marker ¼i has several surface realizations depending on the preceding vowel. It is realized as /¼i/ after high close vowels and as /¼e/ after non-high vowels. (In some dialects the assimilation is regressive in the sense that the half-close vowels are raised to high and the marker stays high too.) It sometimes fuses with a preceding /a/ vowel to yield a front half-open vowel /E/. In some dialects, e.g. Kpele (Collins 1993), there is an epenthetic palatal glide preceding it rendering the form /yi/. 11.4.1 The occurrence of ¼i in three place constructions The predication marker functions as an optional marker of nominal ‘object complements’ of certain factive verbs (e.g. wO ‘do, make’, o´ ‘pile, install’, kpa ‘carve’, gbi ‘weave’, tia´ ‘elect’, 4lO ‘write’) as shown in the highlighted segments in the examples (cf. Amuzu 1993: 61–72). (13) du¼a o´ Kofı´ ﬁa¼e town¼def install name chief¼pred ‘The town installed Koﬁ as chief.’ (Amuzu 1993: 62) (14) e´yata wo`¼tsO! anyı´ sia na´¼m be´ ma´¼tsO! therefore 3sg¼take ground this give¼1sg quot 1sg.sbjv¼take na´ wo` give 2sg ne´ na¼gbi ka¼e purp 2sg.pot¼weave rope¼pred ‘ . . . therefore he gave me this clay to give to you so that you can weave it into a rope . . . ’ (Kwamuar 1997: 25) In example (15) the controller of the pred marker is the 1sg object which is shared by the two verbs xO ‘get’ and wO ‘do’ in a serial verb construction. The SVC itself functions as a complement clause. (15) me´¼ga¼dze be´ na¼xO¼m 3sg.neg¼rep¼ﬁt quot 2sg.pot¼get¼1 sg 0 a´¼wO srO¼wo`¼e o 2 sg.sbjv¼make spouse¼2sg.poss¼pred neg ‘It is no longer ﬁtting that you take me to make [me] your spouse.’ (Ayeke 1996 [103]) (16) wo´¼NlO wo´ nya eka´¼e 3pl¼write 3pl word one¼pred ‘They write them as one word.’ (Collins 1993: 24)

366

F. K. Ameka

As is evident from the examples so far, the clitic ¼i tends to occur towards the rightmost edge of the clause but typically before any sentence-level ﬁnal operators such as the o ‘neg’ in (15), or illocutionary particles. Interestingly, however, when the target constituent with which the marker occurs is fronted for focus, the marker remains in postverbal position and cliticizes onto the controller—in the examples given so far, the direct object. (17) below is related to (13) above in the sense that the object complement is focused and in initial position but the ¼i marker is not pied-piped with it, but rather attaches to the controller. (17) ﬁa¼e´ du¼a o´ Kofı´¼i chief¼foc town¼def install name¼pred ‘A CHIEF the town installed Koﬁ.’ Similarly, the marker is also used to optionally mark nominal ‘object complements’ of verbs of perception and cognition (e.g. kpO! ‘see’, bu ‘think’) (cf. Amuzu 1993: 61–72). (18) A´ma kpO! sr~O¼a ga-tO!¼e name see spouse¼def money-owner¼pred ‘Ama saw her husband as a rich person.’ (Amuzu 1993: 61) (19)

. . . gake´ mı´e´¼bu wo´ na´ne´ke´¼e o but 1pl.neg¼think 3pl nothing¼pred neg ‘ . . . but we did not think them anything’ (i.e. ‘we did not think anything of them’) (Ayeke 1996: 82)

In these examples, the object complement designates a judged state of the object referent and as such can be analysed as participant-oriented. 11.4.2 The occurrence of ¼i with depictive secondary predicates The clitic ¼i also functions as an optional marker of nominal depictive secondary predicates on the object (20, 21) or subject (22) of the clause. (20) wo´ dzila´ eve¼a¼wo´ ku´ 3pl parent two¼def¼pl die 0 he´¼gble˜´ wo´ ı´ tsyO´ e-vı´¼wo´¼e mod¼leave.behind 3pl down orphan-dim¼pl¼pred ‘Both of their parents died and left them [as] orphans.’ (Obianim 1990 [2165])

Forms of secondary predication in serializing languages (21)

wo´¼le´ Kofı´ ﬁaﬁ¼e 3pl¼catch name thief¼pred ‘They caught Koﬁ as a thief.’ (Amuzu 1993: 72)

(22)

Nu´tsu a´e´¼wo´ nya´ dze´-na´ ame¼e, . . . man indef¼pl mod appear-hab person¼pred ‘Some men do indeed appear as people, . . . (but they are just empty barrels).’ (Ayeke 1996 [22])

367

In examples (20–22) the highlighted NPs function as adjuncts; however, in each case they pertain to a property or state of an argument of the main predicate. These structures therefore satisfy the deﬁnition of depictives. Interestingly, the clause in (20) is an SVC in which such a nominal depictive occurs. The evidence here shows that in a serializing language like Ewe (and in serial structures) nominal secondaries do occur (contra the claim by Larson (1991) with which this chapter opened). In addition, ideophonic expressions that function as participant-oriented adjuncts in a clause may also be optionally marked by ¼i ‘pred’. In example (23), the ideophone mu´mu ‘raw’ is a predication, so to speak, on the referent of the object, ‘the meat’. Were this ideophone used attributively, it would occur within the NP and in that case before the determiner la´ ‘the’— la˜ mu´mu la´ ‘the raw meat’. (23)

wo´¼u la˜ la´ mu´mu¼i 3pl¼eat meat def ideo:raw¼pred ‘They ate the meat raw.’

Furthermore, the clitic ¼i can also occur as an optional marker on privative nominalizations in secondary predicate function. Examples (24a, b) are both taken from the same author. However, the marker is used in the former and left out in the latter, even though the contexts are comparable. This illustrates, in part, the optionality of the marker to which we return in 11.4.4. Ada˜ledzogbe xe´ fe¼a nE (24) a. e´ye ame-ga˜´ and person-big name ﬁll debt¼def dat.3sg nu-ma-ke-ma-ke¼e mouth-priv-open-priv-open¼pred ‘And the big man A. paid back [lit. ‘ﬁlled’] the debt to him without opening the mouth [i.e. without saying anything].’ (Ayeke 1998: 41)

368

F. K. Ameka b. e´¼nO anyı´ nu´-ma-u-ma-u . . . 3sg¼be.at:nprs ground thing-priv-eat-priv-eat ‘She stayed without eating . . . [until six in the evening].’ (Ayeke 1998:72)

(25)

e´¼ku´ vi-ma-dzi-ma-dzi¼i 3sg¼die child-priv-bear-priv-bear¼pred ‘She died childless.’

Amuzu (1993), inspired by Halliday (1967), suggested that the marker in the usages described so far—object complements and depictives including the adjunct privative nominalizations—can be viewed as an attribute marker (cf. Duthie 1996). The argument is that in each case the constituent it marks can be seen as representing an attribute of the controller (in the sense of Halliday, where the term extends to the relationship between a depictive and its controller). I have used a broader term, predicative marker, to be more inclusive of the other contexts of use of the marker ¼i. 11.4.3 The adverbializing function of -i The marker -i also appears to have a derivational function, serving as an adverbialiser (glossed ‘adv’ in the examples below). This function and the predication marking functions described above have been kept separate in traditional descriptions of Ewe. The perspective adopted here is that these functions are related in the following sense: all the contexts in which the ¼i marker is found relate to complements or adjunct-like constituents that are oriented to a participant and/or the event characterized in the clause (see Ameka and Schultze-Berndt 2000 for a plausible diachronic development). The adverbs derived using the marker tend to be event-oriented, as the examples below will make clear. There are three categories of terms that the adverbializing sufﬁx attaches to. 11.4.3.1 The marker -i occurs as a denominal adverbializing sufﬁx on manner adverbials: (26)

a´le´ wo´¼de´ ası´ fufu¼a kple´ tOme-la˜¼wo´ mi-mi thus 3pl¼put hand fufu¼def com river-meat-pl rdp-swallow me glugluglu alO-tsO-tsO-e containing.region ideo:gulp arm-rdp-be.near-adv ‘Thus they started swallowing the fufu and the ﬁsh gulping them down very quickly.’ (Bureau of Ghana Languages, e modza´ka I: 14)

Forms of secondary predication in serializing languages (27)

369

e´¼gble˜´ to´ na´ fo¼nye sr~O, . . . he´¼xO¼e 3sg¼spoil ear dat brother¼1sg.poss spouse mod¼get¼3sg le e´¼sı´ se-se˜´-e loc 3sg¼hand rdp-be.strong-adv ‘He coaxed my elder brother’s wife . . . , and took her away from him forcefully.’ (Bii Setsoaﬁa 1982: 8)

This adverbializing sufﬁx -i is distinct from the clitic ¼i in a number of respects. First, the sufﬁx -i changes the grammatical class of the form it is attached to, while the clitic ¼i does not. In examples (26) and (27) the sufﬁx is denominal. Second, and this follows from the ﬁrst, it is obligatory and, unlike the clitic, it cannot be detached and the nominal base to which it is attached moved away from it. Compare the acceptability of example (17) to the unacceptability of *alO-tsO-tsO ye´ wo`¼mi nu´¼a-e (arm-rdpbe.near foc 3sg¼swallow thing¼def-adv) ‘Quickly she swallowed the food’. Third, there is a formal difference between the de-adjectival adverbializing sufﬁx illustrated immediately below and the predicative ¼i clitic, in the sense that the former undergoes dissimilation when it is attached to a high front vowel ending stem, as explained in footnote 3, while the latter does not. Thus in (28b) vı´-i (small-adv) becomes vı´-e ‘slightly’, but in example (25), when the clitic attaches to vi-ma-dzi-ma-dzi ‘not bearing a child’, it stays an ¼i, namely vi-ma-dzi-ma-dzi¼i ‘without bearing a child’. 11.4.3.2 The marker -i also functions as a de-adjectival adverbializing sufﬁx on manner adverbials in the southern and standard dialects. In this context of use, it has a dialectal variant in the inland dialects, namely, - e´ ‘adv’ (Ameka 1991: 91). (28)

(29)

a. ne´ Ma´wu´ l~O ko la´, ye¼a¼to´ if God love just top logophoric¼pot¼pass dodo´kpO¼a´ me bObO-e exam¼def containing.region easy-adv ‘If God likes it, he will go through the exam easily.’ (Ayeke 1996: 27) b. e´ye dzitsı´tsı´ la´ ha˜ nu yi vı´-e and worry def also mouth go small-adv ‘And the worry too has reduced slightly.’ (Ayeke 1996 [29]) wo´a´¼wo´¼e´ nye´ yevu´ si¼wo´ 3pl¼pl¼foc cop European rel¼pl Vı´-e, eve-lı´a´¼wo´¼e´ nye´ white-adv two-ord¼pl¼foc cop

f e´ Nu´tı´-gbale˜ fu poss body-skin be.white IndiatO!¼wo´ Indian¼pl

370

F. K. Ameka ) ame-si¼wo´ f e´ Nu´tı´-la˜ le dzi -e person-rel¼pl poss body-ﬂesh be.at:prs red-adv ‘[America is made up of three groups of people.] They are the Europeans whose skin is white; the second are the Indians whose skin is red [and the descendants of the blacks].’ (HlOmatsi 1995 [1829])

The de-adjectival colour adverbial words in (29)—V´ı-e (or V´ı- ´e in the inland dialects) ‘white’ and dzi )-e (or dzi -) e´ in the inland dialects) ‘red’— occur in adjunct position in the clause and can be used to modify any verbs. In the example, V´ı-e is used to modify the semantically general verb fu ‘become macro-white’ to zoom in on the focal white nature of the colour of the Europeans. The form dzi )-e, on the other hand, is used as a complement of the locative verb le ‘be at’ and together they predicate the quality of redness of the skin of the Indians. 11.4.3.3 Furthermore, the marker -i occurs as a lexicalized component of an ‘adverbializing’ sufﬁx -tOe (tO ‘kind’ þ ¼i ‘pred’) (cf. Westermann 1930: 184; Ansre 1966: 229; Agbadja 1983; Collins 1993: 20–3). This can be interpreted as deriving from the adverbializing function involving a nominal base illustrated under 11.4.3.1 above. (30) Kofı´ zO eﬁe-tOe name walk king-kind.adv ‘Koﬁ walked king-like / in the manner of a king.’ (Collins 1993: 20) (31)

e´¼do´ dzi o´ Nyo´tugbe f e´ nya¼wo´ Nu´ nE 3sg¼put heart answer name poss word¼pl side dat.3sg Nu´tsu-tOe . . . man-kind.adv ‘He picked up courage and responded to Nyotugbe’s words in a manly way . . . ’ (Ayeke 1996 [82]).

There is variation among the dialects with respect to the degree of lexicalization or ﬁxedness of the marker -i to the nominalizing form -tO ‘kind’. In the standard dialects and many of the dialects spoken in Ghana both southern and inland, the form -tOe ‘adv’ does not have an internal structure, it has become a single morpheme in the function of adverbializer. For some dialects in Togo, however, such as Kpele (Collins 1993) and other coastal and interior dialects (see Agbadja 1983), the -tO and the -e form can be separated and treated as independent morphemes in a way similar to how the constituent to which the clitic ¼i ‘pred’ and the

Forms of secondary predication in serializing languages

371

constituent it is marking can be stranded. Thus the -tO nominal can be fronted and the ¼i marker left behind as shown in (32), based on the form in (30).4 (32) eﬁe-tO Kofı´ zO¼e king-kind name walk¼pred ‘King-like Koﬁ walked.’ Thus far we have identiﬁed two formal markers that occur on adjunct-like and complement-like structures in Ewe: (a) a denominal, de-adjectival adverbializing sufﬁx, -i and (b) the predicative clitic ¼i. The latter is used to optionally mark participant-oriented adjuncts that correspond to prototypical depictive secondary predicates (see section 11.4.2). A related context in which the marker occurs is when it marks object complements in three-place constructions. Such complements have been argued to be similar in form and function to depictives, and the identity of marking in Ewe supports this. Plank (1985), for instance, argues that the former are derivative of the latter. Thus far we can conclude that Ewe does have nominal secondary predicates, and depictive ones at that. As such, Larson’s claim that the difference between serializing languages like Ewe and non-serializing languages like English is in the lack of nominal secondaries in serializing languages is, from the point of view of Ewe, without empirical foundation. The challenge of the Ewe phenomena, however, is that the predication marker ¼i is not obligatory. In the next subsection I take up this issue of the optionality of ¼i. 11.4.4 The optionality of the marker ¼i As is evident throughout the discussion, the adverbializing derivational sufﬁx -i is obligatory, however, the clitic ¼i that marks complements and adjuncts is optional. By this I mean that its presence or absence does not affect the structure of the construction in which it appears. It appears that its presence is more for ease of processing. It signals that there is a constituent in the clause that pertains to a participant rather than to the event as a whole. Such information can also be derived from the structure and from other contextual information such as the semantics of the main predicate. For instance, the function of the object complement of verbs of creation, 4 The separability of the -tO base form and the -e is suggestive of the context of the development of the adverbializing sufﬁx -tOe. A scenario for the development is that -tO forms occurring as nominal adjuncts and having event or participant orientation are regularly marked by the clitic ¼i which is realized as an /e/. Some dialects continue to treat this as the clitic while others have lexicalized the two forms, treating it as a single adverbializing morpheme distinct from the nominalizing -tO form and the clitic ¼i as well as the denominalizing adverbial sufﬁx -i.

372

F. K. Ameka

perception, and cognition can be inferred from their occurrence in three place constructions and the frame of the verbs. Similarly, the adjunct status of an NP or ideophonic expression or a quantifying expression can be inferred from structural properties of utterances. Nevertheless, there are various factors that might coerce the use of the marker. An important feature is dialect preferences. Anlo (or southern dialect) speakers tend to regularly use the marker in the contexts where one expects it. Inland dialect speakers tend to leave it out. In the literary standard, being based on the Anlo dialect, there is a preference for the form to be used. However, since the local dialects of the producers of the literary forms affect the form of their language, some writers (especially those with inland dialect backgrounds) tend to leave the marker out. Frequency of use and structuration or lexicalization seem to be leading to the more or less obligatory use of the marker in certain contexts. Attention has already been drawn to the variation across dialects with respect to its ﬁxedness in the denominal adverbializing sufﬁx -tOe above. Another context where this is increasingly the case is in serial verb constructions involving guided conveyance or accompanied deictic motion, as we shall see below. Thus in collocations like kplO¼e yi¼i (accompany¼3sg go¼pred) ‘lead him/ her away’ and e¼e do¼e (take.out¼3sg exit¼pred) ‘take it out’ the marker appears to be obligatory. In fact, the marker has become more or less lexicalized in some forms: vE ‘come:pred’ as in tsO¼e vE (take¼3sg come:pred) ‘bring it here’ and do´ kpo¼e (put log¼pred) ‘failed’. In a few contexts, the absence of the marker could lead to ambiguity and affect the interpretation. For instance, in (33) there is a preference for the ¼i marker to be present. If it was left out, it could lead to a different interpretation. In this case its presence makes the structure of the utterance explicit and therefore facilitates processing as well. The source of the trouble in (33) is that the quantiﬁer eka´ ‘one’ can follow the head noun as a modiﬁer and constitute an NP with it. In that case the sentence would mean ‘The parents and the children ate the same thing’. However, the desired contextual interpretation is that they ate together, and in this case eka´ ‘one’ functions on its own as a participant-oriented adjunct, with the subject NP as controller. As such the ¼i marker disambiguates this structure. (33) evı´¼a´¼wo´ kple´ wo´ dzila´¼wo´ u-a nu´ child¼def¼pl com 3pl.poss parent¼pl eat-hab thing eka´¼e one¼pred ‘The children and their parents eat together.’

Forms of secondary predication in serializing languages

373

The generalization is that the clitic ¼i as a predicative and secondary predicate marker is optional. Some conditions might, however, lead to a preference for its occurrence on the participant-oriented adjunct and complement expressions in clauses. A tendency towards increasing obligatoriness is developing based on structural disambiguation, structuration and lexicalization, and frequency. Individual differences based on dialect preferences also affect the degree to which speakers might use or not use the marker. Serial verb constructions are one of the vehicles for the emergence of lexicalized forms involving the predication marker. In the next section, I discuss the types of SVC and conditions under which the marker ¼i is found in such structures.

11.5 Serial verb constructions and depictive secondary predication It was noted in section 11.3 that some serial verb structures can be interpreted in context as equivalents of depictive expressions. In addition, there are subconstructions of serial verb constructions in which a marker ¼i appears. I claim that the ¼i marker found in SVCs is the same as the predication marker described in the preceding section. That is, it is the same marker that marks object complements in three-place constructions and adjunct NPs as depictive. If that is so, a question that this raises is whether the VP components in SVCs which are marked by ¼i can be seen as depictive secondary predicates. My claim is that a VP component in an SVC is marked by ¼i to signal that it pertains to a non-subject argument of the SVC clause also. That is, the marker signals that the sub-event represented by the VP component on which it occurs is not a predication only of the subject but also of another core participant. Recall that SVCs in Ewe are same-subject structures. I will return to this question after describing the contexts in which the ¼i marker occurs in SVCs. Traditionally, the ¼i marker in SVCs, even though it has the same form as the predication marker described in the preceding section, has been treated as distinct from it. Ansre (1966) called it the redundant object marker, while Lewis (1985) called this form serial ¼i, to distinguish it from other ¼i forms. Collins (1993) argues that it is an EC (empty category) case assigner. He explicitly rejects a secondary predicate analysis of this form. The suggestion being made here is that, given the formal identity of marking and given the similar semantics of the structures involved, the ¼i marker is used in SVCs to indicate that a VP component is oriented to a non-subject participant. The constructions in which the marker occurs are bona ﬁde SVCs because the VP components are codependent in syntax

374

F. K. Ameka

and semantics on each other and share the same subject as well as the same temporal frame. The main difﬁculty for Schultze-Berndt and Himmelmann (2004: 69–72) in accepting SVCs as possible depictive constructions is that the VPs in SVCs are not adjuncts. In fact, in a way the VP components marked by ¼i in SVCs are like object complements of verbs of creation, perception and cognition which are not optional, and yet are participantoriented. Since object complements are related to depictives, the VP components in SVCs marked by ¼i could be related to depictives, too. The semantics involved are also comparable. One can distinguish three semantic subtypes of the serial verb constructions in Ewe in which the marker occurs. 11.5.1 ‘Cumulative participant’ SVCs There are expressions in which the subject and object of the ﬁrst verb are collectively involved in carrying out the event denoted by VP2 as in (34a). In this case it could be argued that the ¼i marker is used to signal that the referent of the object of VP1 is also an effecter of the sub-event in VP2. The default interpretation of an SVC is that the VP components share the same subject: they need not share any other arguments. As such, the ¼i marking could be seen as making explicit that the object of VP1, which is not shared by VP2 in this case, is involved in the realization of the sub-event as well. Signiﬁcantly, the SVC structure in (34a) can be contrasted with a semantically related overlapping clause construction as shown in (34b) which is explicit about the coreference between the referent of the object in the ﬁrst clause (or VP1) and the subject of the second clause (or VP2). (SVC) (34) a. E´¼[yO! evı´¼a´¼wo´]vp [fo fu´]vp¼i 3sg¼call child¼def¼pl hit bone¼pred ‘S/He called the children together.’ (overlapping clause) b. E´¼yO! evı´¼a´¼wo´k wo´k¼fo fu´5 3sg¼call child¼def¼pl 3pl¼hit bone ‘S/He called the children [and] they gathered together.’ The difference between (34a) and (34b) (i.e. between the SVC and the overlapping clause) is that in (34a), both the participant coded as subject 5 Ewe makes extensive use of verb–noun collocations in rendering predicate meanings (see e.g. Essegbey 1999). The VP fo fu´ ‘gather’, which literally means ‘hit bone’, is one such collocation where ‘gather together as a group’ is construed as bones hitting each other or coming in contact (where ‘bones’ are used metonymically for ‘body’).

Forms of secondary predication in serializing languages

375

and the participant coded as object of the ﬁrst verb are concomitantly participating in the realization of the state of affairs expressed in the VP2 and its complement. That is, both the caller and the children gathered together. In (34b) only the children called gathered together: the caller is not part of it. In other words, the subject pronoun of VP2 is coreferential only with the object NP in the ﬁrst clause. In a sense, in (34a) the VP2 (the verb plus its inherent complement) can be thought of as being predicated secondarily of the subject and of the object of the ﬁrst verb. In (34b) it is only predicated of its own subject, which happens to be coreferential to the object of the ﬁrst verb.6 11.5.2 ‘Accompaniment’ SVCs The second type of construction in which the marker ¼i is optionally found is where the object NP of the ﬁrst verb phrase denotes an entity which accompanies the subject argument in carrying out VP2. Such an accompaniment can be a true instrument as in (36), where the stick is used in effecting the action described by VP2, or it can be an emotion, a state, or a condition, as in (35). In example (35), VP1 expresses the emotional condition of the subject participant in carrying out VP2. The emotional condition is like a comitative argument which co-participates in VP2. The ¼i marking suggests that VP2 is oriented towards the non-subject argument as well. (35) e´¼da [tsO! nu´xaxa]vp [yi yOme]vp¼e 3sg¼mother take worry go grave¼pred ‘Her mother went to the grave in worry.’ (lit. ‘took worry [and] went to the grave [with it]’, ‘took worry to the grave’) (adapted from Obianim 1990 [4403]) 6 Just as adjunct-like phrases in monoclausal structures can be participant- or event-oriented, I believe that adjunct-like clauses, such as the second clause in an Ewe OVC, can also be interpreted as having an orientation to a participant or the event in the ﬁrst clause, as I have demonstrated above. The relationship between such tight-knit biclausal constructions and their orientation to those at the monoclausal level needs to be further explored. It is not only SVCs with a component VP marked by the predication marking clitic ¼i that can be paraphrased as an OVC. Example (a) below involving the object complement of the verb wO ‘do, make’ marked with the clitic ¼i can also be paraphrased as an OVC as shown in (b):

a.

afO ade´¼a´ wO wo´ ka´ta˜´ sO-sO¼e foot six¼def do 3pl all rdp-equal¼pred ‘The six feet [i.e. the grave] made them all equal.’

b.

afO ade´¼a´ wO wo´ ka´ta˜´ wo´¼sO foot six¼def do 3pl all 3pl¼equal ‘The six feet [i.e. the grave] made them all they were equal.’

376

F. K. Ameka

(36) Koﬁ tsO! atı´¼E fo Yao yi name take stick¼def hit name pred ‘Koﬁ took the stick [and] hit Yao.’ (Collins 1993: 36; glosses modiﬁed FKA) 11.5.3 ‘Guided conveyance’ SVCs A third type of serial verb construction in which the predication marker appears is one which could be characterized as a case of accompanied motion. That is, the subject of the SVC is accompanied by the object of VP1 and together they undertake a motion event represented in VP2. The examples typically involve cases of guided conveyance. In this context it seems that the ¼i marker signals that the VP2 component of the SVC pertains to the object of VP1 as well. (37) E´¼ku tsi v 3sg¼scoop water come:pred ‘S/He fetched water [and] brought it.’ (38) Koﬁ kplO af e´nO¼a dzo´¼e name. lead madam¼def leave¼pred ‘Koﬁ led the woman away (to a hide-out)’ (HlOmatsi 1995 [718]) There is increasing evidence accumulating from research in languages with serial verb constructions that a verb in the series may be restricted in function to providing information about a participant in the clause in a way similar to that of participant-oriented adjunct-like expressions in monoverbal clauses (see e.g. Zavala, 2006). Aissen (2000), for instance, shows that the agreement behaviour between the controller and one of the verbs in serial constructions in Tzotzil signals their being oriented to the particular participant. Some of the examples she gives include motion constructions of the kind involving guided conveyance that we discussed above for Ewe. She argues further that they are instances of depictive secondary predication and that they ﬁll the same slot in Tzotzil where other secondary predicate terms occur. The unifying feature of the three sub-types of serial verb constructions in Ewe in which the predication marker occurs is that the referent of the object of VP1, together with the shared same subject of the SVC construction, are the effectors of VP2. As such it can be argued that ¼i marking is to signal that the VP2 is a kind of secondary predication on the non-subject [¼argument]of the SVC. This function may be related to depictive secondary predication function in the same way that object complements may be linked to depictive-type structures.

Forms of secondary predication in serializing languages

377

11.6 Conclusion Does Ewe, a serializing language, have nominal secondaries, especially depictive ones? This is one of the questions I aimed to address in this chapter. In section 11.3 I showed that various structures in Ewe exhibit participant orientation, although they are not formally marked as such and thus may not be considered dedicated depictive constructions. Such structures include adjunct noun phrases that express a property and are oriented towards a participant rather than the event as a whole; ideophonic expressions; various adjunct prepositional phrases introduced by comitative, allative, or locative prepositions; and similative constructions. SVCs and biclausal overlapping clauses can also often be used as translation equivalents of depictive expressions in other languages, e.g. where one of the VPs in a serial construction encodes the psychological state, physical appearance, or posture of the subject in carrying out the other subevents. Apart from these structures expressing depictive content, Ewe also has a general marker of predicative adjuncts, which is the clitic ¼i (with various allomorphs). It is used to optionally mark participant-oriented nonverbal constituents such as object complements of creation, perception, and cognition verbs, adjunct NPs including quantiﬁer phrases, and also ideophones. It also marks component VPs of SVCs to indicate that the particular VP component, while in its primary predication role it relates to the subject of the SVC (SVCs in Ewe being same-subject constructions), also secondarily predicates on a non-subject core argument of the clause. It appears that a denominal and de-adjectival adverbializing sufﬁx -i derived from this marker. The bridging context for this grammaticalization seems to be the use of the marker with the structure consisting of a nominal headed by -tO ‘kind’ in adverbial function. Such adverbials are vague between participant and event orientation. Furthermore, there are degrees of lexicalization of the clitic/sufﬁx in this and other contexts. From all this, we can conclude that Ewe has nominal (depictive) secondary predicates as well as verbal secondaries. It seems that the difference between the serializing languages and the non-serializing languages a` la Larson is not in the absence of nominal secondaries in the former type of languages. Rather, it seems to lie in the lexical typological properties of the languages. Recall that Ewe does not have predicative adjectives. Furthermore, some property concepts, especially those of human propensity, are nominal. By contrast, one source of depictive secondary predicates in languages like English and German is the adjective. This difference in lexicalization could

378

F. K. Ameka

explain why Larson’s sentence John came back a perfect wreck is not easily translatable into the West African serializing languages I am familiar with. Given that Schultze-Berndt and Himmelmann (2004) as well as Himmelmann and Schultze-Berndt (Ch. 1, this volume) insist that participant-oriented expressions which should be considered depictive secondary predicates have adjunct status, the function of component VPs in SVCs as participant-oriented secondary predicates raises at least two questions for the deﬁnition of depictive secondary predicates. First, none of the verbs in an SVC is optional, and thus the candidates for depictive status in such structures are not adjuncts in the classical sense. Recall, however, that object complements are also not prototypical adjuncts, but that they can and have been related to depictives as noted above. Second, syntactically all verbs in an SVC are equal in rank—none is secondary. It seems, however, that in languages where the component VPs of SVCs can be independently focused or questioned, it is also possible for one of the components to act as a secondary predicate on a non-subject participant. This, it seems to me, is what is happening in Ewe. It is important, however, on all levels, to make a distinction between structures that can be interpreted as having depictive function or content and structures that are dedicated to the coding of this function. The form ¼i in Ewe is a dedicated predication marker which signals participant orientation of a constituent in a clause, including depictives. There are other structures that can be interpreted as having depictive content but which do not constitute depictive constructions in their own right.

12 Depictive and other secondary predication in Lao NICHOLAS J. ENFIELD

Lao (South-Western Tai, Laos/Thailand/Cambodia) is a typical isolating language, lacking inﬂectional morphosyntactic categories such as casemarking and agreement. It is of interest from a crosslinguistic perspective on depictive secondary predication, since most analyses of depictive expressions in other languages appeal largely to morphosyntactic patterns of agreement and ﬁniteness, which are not overtly marked in Lao.

12.1 Preliminaries on Lao grammar I begin by presenting some facts about the grammar of Lao which are relevant to a discussion of depictive and other secondary predication. These concern the nature of word classes, the nature of complex predication, and the general under-determination of distinctions in grammatical relationships, leading to heavy context-dependence of interpretation. One large formally deﬁned word class is the verb class. Words in this class express a range of concepts which in English are expressed in three distinct word classes: verb, adjective, and adverb. Grammatically, verbs take direct negation with the negator bo`o`1, may be marked by the preverbal progressive marker kamlang2 and the postverbal perfective marker le`e`w4 as well as other tense/aspect/modality markers, and may freely function as unmarked modiﬁers in noun phrases. Minor sub-distinctions within this class may be established (e.g. accessibility to certain types of reduplication), and these correspond to distinctions more markedly grammaticalized in other languages (e.g. between ‘property concept words’ and ‘action/event words’; Thanks to members of the audience at a presentation in Bochum, 7 June 2001, for comments on some of the ideas in this chapter, and especially to Eva Schultze-Berndt and Nikolaus Himmelmann for encouragement and advice.

380

N. J. Enfield

Enﬁeld, 2004). There is no distinct class of adverbs. Adverbial meanings are often expressed by verbs in certain constructional slots. The following examples show vaj2 ‘fast’ as a main predicate (1), as a modiﬁer in a noun phrase (2), and as an adverbial modiﬁer of a main predicate le`e`n1 ‘run/go’ (3):1 (1) lot1 khan2 nii4 vaj2 vehicle clf this fast ‘This vehicle is fast.’ (2) lot1 vaj2 khan2 nii4 ngaam2 vehicle fast clf this beautiful ‘This fast vehicle looks good.’ (3) lot1 khan2 nii4 le`e`n1 vaj2 vehicle clf this run fast ‘This vehicle goes fast.’ The basic structure of the Lao clause is as follows (‘AM’ stands for ‘aspectmodality marking’, which may occur on either side of the verb): (4) Left Position—Subj—AM-[V (Obj)]-AM—final-ptcls— right position Noun phrases in almost any position may be ellipsed if their referent is contextually retrievable. Left aspect-modality marking may occur no more than once per clause. There are many ways in which two verbs can appear together in a single clause, involving adjunction, complementation, or compounding (Enﬁeld 2003; to appear). An important point in the context of analysing secondary predication in this language is the complete lack of overt morphological marking of relationships between elements within complex structures. Many languages have various forms of the verb showing distinctions in ‘ﬁniteness’, indexed by presence or lack of tense/ aspect/modality marking, dedicated nonﬁnite forms, patterns of agreement or cross-referencing, which all can help to decide what role a certain element 1 Transliteration used here follows IPA convention except for velar nasal /ng/, glottal stop /q/, mid-high front vowel /eˆ/, low front vowel /e`/, central vowel (schwa) /e/, low central vowel /a/, high back unrounded vowel /u`/, low back vowel /o`/. Tones are represented by numbers following each syllable (ø¼unstressed/atonal, 1¼/mid level/, 2¼/high rising/, 3¼/low rising/, 4¼/high falling/, 5¼/low falling/). Examples are drawn from a corpus of spontaneous texts (including personal narratives, procedural descriptions, folk tales, conversations, and the like) collected in Vientiane in 1996 and 1997. Numbers in brackets following the translation of each example refer to paragraph number in text transcriptions. Examples with no such reference number are constructed in consultation with native speakers.

Depictive and other secondary predication in Lao

381

has. In analysing depictive constructions in a wide range of languages, Schultze-Berndt and Himmelmann (2004) make extensive reference to patterns of agreement and other overt morphological marking of the role of different predicates in complex expressions (see also Ch. 1, this volume). In Lao, however, there is no agreement, no case-marking, no explicit ﬁniteness distinction. There are, however, overt markings of aspect and modality (e.g. negation, perfective marking, illocutionary particles) which can be useful in analysing the structural status of certain elements in complex expressions. Also useful are tests involving movement and paraphrase. These are employed in the discussion below. The chapter is structured as follows. In section 12.2, I discuss participantoriented expressions2 involving unmarked verbs in V1 and V2 structures, ﬁrst where the depictive element is V2, then where the depictive element is V1. Some ambiguities in analysis are discussed. While these structures can and often do convey participant orientation, they are not dedicated depictive constructions, since the same structures are also used to express adverbial (manner) and resultative meanings. That is, they are general adjunct constructions in terms of the distinctions proposed in the introduction to this volume. In section 12.3, I look at two ways in which nominals can contribute to depictive expressions. They may appear as predicative nominal phrases, or in an adjunct structure headed by the verb peˆn3 (elsewhere a copula ‘be’). The nominal hosted by peˆn3 describes the physical (or other) form of one of the main clause arguments. Once again, these are not dedicated depictive but general adjunct constructions, since the same structures may also be used with manner adverbial and resultative meanings. Section 12.4 concludes.

12.2 Depictive secondary predication by unmarked verb Many unmarked combinations of verbs in Lao clauses can be analysed as involving secondary predication. A predicative element (typically a member of the verb class) makes a predication in addition to a main or primary predication. This secondary predicative element may be omitted without changing the basic semantics of the primary predication, and in this sense secondary predicates are always adjuncts. The term secondary predication refers to this phenomenon generally, and the various semantically deﬁnable subtypes may be termed depictive, adverbial, and resultative (perhaps among others in addition). I use the term expression to refer to structures 2

For the term participant-oriented expression, see the introduction to this volume.

382

N. J. Enfield

with a certain meaning, but where the structure is not dedicated to expressing just that meaning. Thus, a depictive expression is a complex structure in which a secondary predication has depictive meaning. Semantically, I make the following distinctions among secondary predications (although the distinctions are in practice not always neat): depictive Expresses the incidental and transient state of one of the participants in a primary predication. There is no connection of cause, result, or manner between the two predications. Stock examples are She ate the ﬁsh raw and He gave the lecture nude. adverbial (manner) Says something about the manner of the primary predication, as in He ate fast and She spoke hesitantly. resultative Expresses something that happens or is the case because the primary predication takes place. Typical examples are She licked the platter clean and He broke it in half. 12.2.1 V2 depictives Many clauses in Lao are of the form (NP1) V1 (NP2) V2, where V1 and V2 are in a relationship of either coordination or subordination (with only the former normally allowing insertion of clause-linking particles without altering the truth-conditional meaning of the complex expression). In the latter case, either V1 or V2 may be the secondary predicate. We ﬁrst consider structures in which V2 conveys a depictive meaning. A prototypical if not archetypal depictive secondary predicate is ‘raw’. It predicates a changeable/transient state of a (usually edible) primary predicate object argument, and has little if any bearing on how the action involving that object (typically ‘eating’) is carried out. While the word order of English distinguishes between He eats meat raw versus He eats raw meat, an equivalent construction in Lao makes no surface distinction: (5) man2 kin3 3sg eat i. ‘S/he eats ii. ‘S/he eats

siin4 dip2 meat raw meat raw.’ raw meat.’

In the ‘depictive’ reading (5i), the main predication is the idea that ‘S/he eats meat’, and predicated in addition to this is the idea that the meat is, at this time, in the state of being ‘raw’. A verb such as vaj2 ‘fast’ can appear in the same position, with adverbial meaning (see example (3), above). The difference here is that the secondary element vaj2 ‘fast’ does not make a predication about the meat, but about the manner of the event.

Depictive and other secondary predication in Lao

383

In the (5ii) reading, dip2 ‘raw’ is a nominal modiﬁer, forming a constituent with the nominal head siin4 ‘meat’ to give a noun phrase siin4 dip2 ‘raw meat’ which could, for example, function as a subject argument: (6)

siin4 dip2 bo`ø dii3 meat raw neg good ‘Raw meat is no good.’

The difference between the two interpretations of the role of dip2 ‘raw’ in (5) corresponds in many other languages to an overt morphosyntactic distinction (e.g. in patterns of agreement, word order, derivational morphology). The distinctions in Lao are covert in (5). They come to light when we look at possibilities for movement, paraphrase and insertion. The distinction between an analysis of a nominal followed by a stative predicate (as in (5)) as a complex noun phrase (with N þ modiﬁer structure) or as a simple noun phrase (N) followed by a secondary predicative element can be made explicit if the noun phrase is more complex (e.g. with a determiner included, as in (7, 8)), or ‘extracted’ and fronted (9, 10): (7)

man2 siø kin3 [siin4 dip2 nii4]np 3sg irr eat meat raw this ‘S/he’s going to eat this raw meat.’ [does not entail that she’ll eat it raw]

(8)

man2 siø kin3 [siin4 nii4]np dip2 3sg irr eat meat this raw ‘S/he’s going to eat this meat raw.’

(9)

[siin4 nii4]i man2 siø kin3 øi dip2 meat this 3sg irr eat raw ‘This meat, s/he’s going to eat raw.’

(10)

[siin4 dip2 nii4]i man2 siø kin3 øi meat raw this 3sg irr eat ‘This raw meat, s/he’s going to eat.’ [does not entail that she’ll eat it raw]

Such tests using constituent extraposition can reveal differences in grammatical behaviour between potentially depictive elements such as dip2 ‘raw’ in (11a) and modiﬁers such as n˜aj1 ‘big’ in (12a). The (b) examples show that the nominal head of the object argument can be fronted in the case of the depictive expression, but not in the case of the noun-modiﬁer expression: (11) a. man2 kin3 siin4 dip2 3sg eat meat raw ‘S/he eats meat raw.’ (also possible: ‘S/he eats raw meat.’)

384

(12)

N. J. Enfield b. siin4 man2 kin3 dip2 meat 3sg eat raw ‘Meat, s/he eats raw.’ a. man2 juu1 hu`an2 n˜aj1 3sg live house large ‘S/he lives in a large house.’ b. *hu`an2 man2 juu1 n˜aj1 house 3sg live large (A house, s/he lives in large.)

A ﬁnal point concerns the yes-answer properties of the two readings of (5). Ambiguity in a question between depictive and noun-modiﬁer readings (i.e. (5i) vs. (5ii), above) can be removed in an afﬁrmative answer (in favour of the depictive reading), by using both the primary and secondary verb together (without the object argument), as opposed to just the primary predicate (in which case the response remains ambiguous): (13) Q: man2 kin3 3sg eat i. ‘Does s/he ii. ‘Does s/he

siin4 dip2 bo`o`3 meat raw Q eat meat raw?’ eat raw meat?’

A1: kin3 dip2 eat raw ‘(Yes, s/he) eats (it) raw.’ [depictive interpretation] A2: kin3 eat i. ‘(Yes, s/he) eats (it).’ [non-depictive interpretation, preferred] ii. ‘(Yes, s/he) eats (it raw).’ [depictive interpretation, possible] It seems that particular combinations of main verb and depictive element in expressions such as (5), (8), (9), and (11b) are associated with familiar, conventionalized combinations of primary and secondary predications. This does not mean that the combination must be typical in itself, and in fact it may be that depictive expressions are more likely to be used when the combination of predications is pragmatically marked (as, for example, in the celebrated case of ‘eating meat raw’). But the combination must make sense with reference to some kind of conventionalized or sensible scenario (i.e. as ‘eating meat raw’ presupposes a contrasting norm of ‘eating meat cooked’). Out of context, a V2 depictive jeˆn3 ‘cool’ is perfectly natural with a primary predication kin3 ‘eat/consume’ but not with

Depictive and other secondary predication in Lao

385

tak2 ‘scoop’: (14)

nam4, man2 dajø kin3(/*tak2) jeˆn3 water 3sg attain eat(/scoop) cool ‘Water, he drank cool.’ (not: ‘Water, he scooped cool.’)

It is my impression that depictive expressions are subject to pragmatic restrictions on speciﬁc combinations of primary and secondary predicate, related to the relevance and/or conventionality of the combination (see Enﬁeld 2002a for a study of this phenomenon in associated postural constructions). The issue needs further exploration. Participant-oriented expressions such as (5), (8), (9), and (11b), involving dip2 ‘raw’, are not formally distinct enough from secondary predicates of other kinds (e.g. adverbials) to be regarded as distinct ‘depictive constructions’. The following examples, involving vaj2 ‘fast’ as a V2 adverbial expression, show the same grammatical properties as have been illustrated for unmarked V2 depictive expressions: (a) manner versus modiﬁer ambiguity (15), (b) disambiguation by different insertion of demonstrative (16a, b), (c) disambiguation by topicalization of the object nominal head alone (17a) vs. object nominal together with secondary predicate (17b): (15)

man2 3sg i. ‘He ii. ‘He

khap2 lot1 vaj2 drive car fast drives cars fast.’ drives fast cars.’

(16) a. man2 khap2 lot1 vaj2 nii4 3sg drive car fast this ‘He drives this fast car.’ b. man2 khap2 lot1 nii4 vaj2 3sg drive car this fast ‘He drives this car fast.’ (17) a. lot1 man2 khap2 vaj2 car 3sg drive fast ‘Cars, he drives fast.’ b. lot1 vaj2 man2 khap2 car fast 3sg drive ‘Fast cars, he drives.’ 12.2.2 V1 depictives Some verbs, such as verbs of posture and wearing, can have depictive function in V1 position. These verbs include in their internal semantics an event

386

N. J. Enfield

(e.g. ‘putting on a hat’ or ‘moving into a sitting position’) followed by a resultant state (e.g. ‘wearing a hat’ or ‘being in sitting position’). V1-V2 strings with posture verbs in V1 position are thus interpretable as either a series of actions (e.g. ‘sit and then read’) or an overlapping of state and action (e.g. ‘sitting [i.e. in the state of having sat] and reading at the same time’: see Enﬁeld 2002b). These can appear to be compounds (‘He drunk-drives’), subordinating adverbial constructions (‘He drunkenly drives’), or clause chains (‘He [gets] drunk [and then] drives’). The distinction is difﬁcult to make in the absence of simple morphosyntactic tests, and tests such as paraphrase and movement can upset the integration of such structures and make judgements difﬁcult. Here are two examples of V1 depictive expressions: (18) man2 nang1 qaan1 pu`m4 3sg sit read book ‘He sat reading a book.’ (or: ‘He read a book sitting.’) (19) man2 maw2 maa2 hu`an2 3sg drunk come house ‘He came home drunk.’ It is possible, but dispreferred, for the depictive element in these expressions to appear after the primary predicate: (20) ??man2 maa2 hu`an2 maw2 3sg come house drunk ‘He came home drunk.’ The preference for the constituent order in (18) and (19) is presumably due to the nature of the ‘depictive’ element being an expression of ‘processfollowed-by-result’ (e.g. ‘sitting-followed-by-being-seated’, ‘putting-hat-onfollowed-by-wearing-hat’, ‘getting-drunk-followed-by-being-drunk’). The depictive expression needs to occur in V1 position to preserve the iconicity of the process occurring before the state, which then holds over the primary predicate. This is not the case when the depictive element expresses a simple state (which does not inherently presuppose a preceding process) such as ‘raw’, ‘fresh’, or ‘cool’ (cf. section 12.2.1 above).

12.3 Use of nominals in depictive expressions Nominals are used in depictive expressions in Lao in at least two ways. First, certain constituents which appear formally to be noun phrases (but which turn out to be stative predicates) can have depictive function. Second, a type

Depictive and other secondary predication in Lao

387

of adjunct headed by a copula verb peˆn3 hosts nominals with a range of depictive, adverbial, and resultative meanings. 12.3.1 Body part noun þ stative verb as V2 depictive An apparent case of an NP in depictive function is as follows: (21) man2 mu`a2 hu`an2 mu`u`2 paw1 3sg return house hand empty ‘S/he returned home empty-handed.’ While the sequence mu`u`2 paw1 (hand empty) can indeed be analysed as an NP meaning ‘empty hand(s)’, in this context it is a predicative NP, not a referential NP—i.e. it is predicating a state of affairs (‘having empty hands’) rather than a thing (‘empty hands’). Nominals in Lao may occasionally be used predicatively—e.g. in equational expressions such as phuu5-nan4 qaaj4 kho`o`j5 (person-that brother I) ‘That person [is] my brother’. But they are quite restricted in this function. Most importantly, they can take no verbal trappings (negation, aspect marking, etc.) whatsoever. The sequence mu`u`2 paw1, however, is of a different type. As a general rule in Lao, sequences of the form ‘body part noun’ þ ‘stative verb’ can function as predicates, taking full verb trappings (see Clark 1996 on this phenomenon in South-East Asian languages more generally). The following examples show the sequences taa3 khiaw3 (eye green) (elsewhere a noun phrase ‘green eye(s)’) and phom3 de`e`ng3 (hair red) (elsewhere a noun phrase ‘red hair’) as main predicates, taking direct negation and irrealis marking, respectively:3 (22) khon2 laaw2 bo`ø taa3 khiaw3 person Lao neg eye green ‘Lao people [are] not green-eyed.’ (23) luuk4 caw4 caø phom3 de`e`ng3 child 2sg irr hair red ‘Your child will [be] red-haired.’ Thus, the depictive element mu`u`2 paw1 ‘empty hand(ed)’ in (21) can be analysed as a predicate in V2 position, and is therefore not a genuine case of an unmarked ‘nominal’ having depictive function. Instead, it belongs to the type of construction discussed in 12.2.1 above. 3 This is similar to a ‘topic-comment’ construction, but not the same. The difference here concerns the placement of verb marking (negation, irrealis marking). In a regular topic-comment construction, verb marking occurs immediately before the verb (e.g. ton4-sak2 baj3 bo`ø no`o`j4 (tree-teak leaf not small) ‘Teak trees, [the] leaves are not small’).

388

N. J. Enfield

12.3.2 Depictive/resultative/adverbial adjuncts marked by peˆn3 ‘be’ A construction which involves an adjunct headed by the copula verb peˆn3 often has depictive meaning, but also shows resultative and adverbial meanings. In these cases, the complement of peˆn3 ‘be’ is always a nominal.4 In the following example, the verb phrase ho`o`p5 phuu2 ‘carry [a] mountain’ takes an adjunct formed by peˆn3 and its nominal complement nuaj1 (a classiﬁer): (24) bak2 n˜ak1 kum3phan2 ho`o`p5 phuu2 peˆn3 nuaj1 fam.m ogre K. carry.in.arms mountain be clf ‘The Ogre Kumphan carried the mountain whole.’ (201) In (24), the word nuaj1 is the classiﬁer for mountains (among other things), and the adjunct peˆn3 nuaj1 in this context means ‘whole’ or ‘as a unit’. The expression is depictive in that it makes an extra predication about the state (as ‘whole’ rather than in parts) of one of the participants involved in the main predication (i.e. the object of ho`o`p5 ‘carry in arms’). The use of sortal classiﬁers in peˆn3-adjuncts with the meaning ‘whole, as a unit’ is fully productive in Lao. The marker of the peˆn3-adjunct construction is the verb peˆn3 ‘be’, which has a range of main verb functions in other contexts. For example, it may serve as a copula (25), and as a predicator of illness (26): (25) to`o`n3 nan4 kho`o`j5 peˆn3 nak1-hian2 time that 1sg be student ‘At that time, I was a student.’ (243) (26) peˆn3 khaj5 be fever ‘[They] had fever.’ (139) It may also appear as a postverbal modal expressing inherent, learnt, or acquired ability:5 (27) ho`q2 bo`ø peˆn3 deˆeˆ4 ﬂy neg be ptcl ‘[He] couldn’t ﬂy, you know.’ (169) 4 There are only a few exceptions. For example, the sequence peˆn3 pok2katiq2 (be normal/regular) (where pok2katiq2 is a verb), is a common adverbial adjunct meaning ‘in normal/regular fashion’. 5 In some uses, and in some dialects, peˆn3 may be used with this meaning as a preverbal complement-taking predicate—cf. Stung Treng Lao (NE Cambodia) bo`ø peˆn3 vaw4 laaw2 (neg ‘be’ speak Lao) ‘don’t know how to speak Lao’ versus Vientiane Lao vaw4 laaw2 bo`ø peˆn3 (speak Lao neg ‘be’) ‘don’t know how to speak Lao’.

Depictive and other secondary predication in Lao

389

(28) mu`ng2 kaø heˆt1 kin3 bo`ø peˆn3 do`o`k5 2sg foc make eat neg be ptcl ‘You don’t know how to cook [them]!’ (42) The peˆn3-adjunct makes a secondary predication which can express meanings ranging from depictive to resultative to adverbial. Some examples are difﬁcult to categorize uniquely, since they combine elements of more than one adjunct type. Like example (24) above, the following two examples have genuinely depictive semantics, in that the secondary expressions describe the form or state of the main clause object at the time of the main verb event taking place (with no necessary relation of cause or manner holding between the two predications): (29) khaw3 kin3 siin4 peˆn3 to`o`n1 3pl eat meat be chunk ‘They ate [the] meat in chunks.’ cia4 (30) man2 he`e`ng5 le`ø ke`q2 qo`o`k5 peˆn3 phe`e`n1 3sg dry ptcl scrape/peel exit be clf/sheet paper ‘[When] it’s dry, then peel it off in/as paper sheets.’ (113) Similarly, in the next example, the peˆn3-adjunct is depictive in that it describes the physical arrangement of the argument of the main clause ‘they’ during the time that the action of the main clause takes place: (31) khaw3 nang1 kin3 khaw5 peˆn3 the`e`w3 3pl sit eat rice be row ‘They sat and ate in rows.’ A ﬁnal example is perhaps more abstract, where ‘being in English’ is a predication about the ‘form’ of a story: (32) man2 law1 lu`ang1 nii4 peˆn3 phaa2saa3 qang3kit2 3sg tell story this be language English ‘He told this story in English.’ Peˆn3 adjuncts can express a state of the primary predicate object which is not incidental to the primary predication but rather results from it. The main semantic difference between these expressions and genuine depictive expressions is the temporal relationship between the two predications. In these cases, the secondary predication is true after the primary predication, while in the case of depictives, the two predications overlap temporally. There are a number of subtypes of these ‘state as result’ secondary predications.

390

N. J. Enfield

For example, the peˆn3 adjunct may express the form of the primary predicate object as resulting from the primary predication, in terms of physical transformation or modiﬁcation (33, 34), a transformation in status or social role (35), or coming into existence (36): (33) maa2 paat5 peˆn3 sii1 liam1 come slice be four sides ‘Bring [the wood and] cut [it] into four-sided [pieces].’ (114) (34) liaw3 beng1 su`ak4 khanaat5 nii4 pu`aj1 peˆn3 phong3 turn look rope size this dissolved be powder ‘[They] turned [and] looked [and saw] a rope of such size dissolved into powder.’ (133) (35) phen1 leej2 haj5 buat5 peˆn3 phaq1 3hon then give ordain be monk ‘Then he had [me] ordained [as] a monk.’ (321) (36) can3thaa2 me`e`1-khaw4 keet5 luuk4 peˆn3 sat2 C. queen born child be animal ‘Chantha the queen gave birth to children [as] animals.’ (153) There are also examples of adjuncts in peˆn3 which are primarily manneradverbial, such as the following: (37) man2 khap2 lot1 peˆn3 labiap5 3sg drive vehicle be regulation/orderliness ‘He drives in an orderly way.’ Finally, some peˆn3 adjuncts are comparable to ‘predicative complements’ such as I consider him a brother which in English are morphosyntactically and semantically quite distinct from depictive complements: (38) kuu3 thu`u`3 khon2 nii4 peˆn3 qaaj4 1sg regard person this be older.brother ‘I regard this person [as a] brother.’ Semantically, this is not depictive. ‘Being the speaker’s brother’ is not being asserted in the peˆn3 adjunct merely as an incidental state of the primary predicate object. Clearly, there is a semantic subordination of ‘being brother’ to ‘regarding’ that something be the case. Unlike the examples we have seen so far, in this case the peˆn3 adjunct cannot be omitted without changing the meaning of the main verb thu`u`3. (This is in explicit contrast to, say, example (35).) In (38), thu`u`3 means ‘regard, consider’. If the peˆn3 adjunct were removed, the meaning of the verb thu`u`3 would be ‘hold, carry’, and the

Depictive and other secondary predication in Lao

391

sentence as a whole would mean ‘I carried this person’ (cf. English I regarded this person as a brother versus I regarded this person).

12.4 Summary Depictive secondary predications are commonly expressed in Lao by the use of an unmarked predicative element (verb) in addition to the primary predicate. The status of secondary verbs as secondary is not overtly manifest in Lao (e.g. by distinct nonﬁnite forms). The nature of the grammar denies the analyst patterns of agreement or verb ﬁniteness, thus making Lao—like any heavily isolating language—a useful addition to crosslinguistic research on a topic which has hitherto centred heavily around analysis in terms of those morphosyntactic indices. Secondary predicative verbs appear in V1-V2 sequences, with sub-types in which either the V1 or V2 element makes the secondary predication. V2 depictives express simple states (such as ‘raw’ or ‘fresh’), while V1 depictives express post-states which are the inherent outcome of prior events (such as ‘seated/having-sat’ or ‘drunk/having-drunk’). Nominals may be hosted in adjuncts headed by the verb peˆn3 ‘be’, where the nominal describes the form of one of the primary predicate participants. A common use of this construction features a classiﬁer, giving the meaning ‘whole’ or ‘as a unit’. The peˆn3 adjunct can also express not just form but resultant form, thus expanding beyond simple depictive meaning. These various structures have broader secondary predicative function, being also able to express adverbial and resultative meanings. Since they are not restricted to the expression of depictive semantics only, I conclude that there is no dedicated depictive construction in Lao. Nevertheless, the subdistinctions among depictive and other functions of these secondary predicate structures are essentially semantic in nature, and I would be surprised if such sub-distinctions did not correspond to some kind of demonstrable distinction in grammatical behaviour, however subtle.

This page intentionally left blank

13 A semantic map for depictive adjectivals JOHAN VAN DER AUWERA AND ANDREJ MALCHUKOV

13.1 Introduction Schultze-Berndt and Himmelmann (2004) is an exploratory crosslinguistic study of what they call ‘depictives’. Examples are shown in (1) and (2). (1) George left the party angry. (2) As a young artist James lived in Dublin. Essential for depictive constructions is that they make a secondary predication about a main predication participant, that the secondary predication happens outside the referential expression identifying the main predication participant, and that the temporal interval of the state of affairs denoted by the secondary predication overlaps with the state of affairs denoted by the main predication. The secondary predication typically depicts a state that the main predication participant is in—as in (1)—or a phase or function—as in (2). The primary focus of Schultze-Berndt and Himmelmann (2004) is the link between depictives and adverbials, e.g., the relation between the depictive illustrated in (1) and the adverbial shown in (3). (3) George left the party angrily. They argue that the depictive in (1) and the adverbial in (3) are semantically very close, and that languages may have constructions that cover both and

This chapter was completed while van der Auwera was a Fellow at the Royal Flemish Academy for Arts and Sciences and Malchukov at the University of Antwerp’s Centre for Grammar, Cognition and Typology. Thanks are due to both institutions. We are grateful to the comments of the editors as well as of those of audiences in Bochum and in Vilnius, especially also to Thorstein Fretheim for making us reconsider Norwegian. Finally, we thank Uma Pappuswamy for help on Tamil.

394

J. van der Auwera and A. Malchukov

whose meaning may therefore be neutral between these uses. Dutch, for instance, renders both angry and angrily with woedend. Dutch (4)

Georges heeft het feestje woedend verlaten. George has the party angry/angrily left

In this chapter the exploration continues. There are three big differences. First, we only discuss adjectival constructions, not only plain adjectives, but also verby and nouny adjectives, adjectives in serial constructions, participles, and relative clauses. For this reason also, we focus on state depictives—as in (1)—rather than depictives that pertain to functions or phases—as in (2)—or further semantic domains. Second, we focus on the relation between depictive adjectivals and non-depictive adjectivals more than on the relation between depictives and adverbials. Third, we will report our ﬁndings in the form of a semantic map. It is important to stress the exploratory nature of the chapter. The hypotheses offered will be illustrated with data from some ﬁfteen languages only. For many hypotheses, we could offer illustrations from other languages, from diverse genetic stocks and regions; but because the information collected is often incomplete, it is not possible to offer our results as fully based on a sensible sample of the world’s languages. The study reported here therefore only counts as a pilot study. For this reason also, no in-depth analysis of any of the language-particular phenomena will be attempted. Section 13.2 will brieﬂy spell out the main idea of the semantic map approach. Section 13.3 builds up a semantic map that includes depictives, with data from a variety of languages. Section 13.4 is the conclusion.

13.2 Semantic maps The basic idea underlying the semantic map approach is simply that similar concepts are expressed in similar ways. Consider the sentences in (5), and more particularly the functions of the modal verb must. (5) a. Mary must go home now. b. Mary must be home now. (5a) expresses an obligation and (5b) a high probability. Obligation and high probability are by no means the same concepts, yet they are similar. An obligation is a situational necessity: there is something in the state of affairs described in (5a), maybe somebody’s wish or command, that necessitates Mary’s leaving. A high probability is also a kind of necessity, but it is crucially

A semantic map for depictive adjectivals

395

epistemic or inferential and refers to a judgement of the speaker and a degree of commitment. Epistemic necessity is the necessity of a judgement relative to other judgments. Perhaps the speaker believes that Mary always goes to work by bike, perhaps he notices that Mary’s bike is no longer there, and he then deduces that Mary must be home. In many languages obligation and high probability have to be expressed in different ways. In the Tungusic language Evenki, for instance, the sufﬁx -mAchin is used for situational necessity, and -nA for epistemic necessity. Evenki (Nedjalkov 1997: 269, 264, 265, 265) (6) a. Minggi girki-v ilan-duli chas-tuli suru-mechin-in. my friend-1sg.poss three-prol hour-prol go.away-obl-3sg ‘My friend must go/leave in three hours.’ b. Su tar asatkan-me sa:-na-s. you that girl-acc.def know-probability-2pl ‘You probably know that girl.’ In these constructions Evenki reﬂects the difference between obligation and high probability. With the modal auxiliary must, English reﬂects the similarity. These observations immediately yield a small map for necessity (Figure 13.1). Situational and epistemic necessity occupy two distinct areas in semantic space. This is symbolized by the fact that each is named explicitly. But the concepts are related: this is symbolized by the connecting line. In English must can be used for both, in Evenki -mAchin is used for situational necessity, but -nA for epistemic necessity (Figures 13.2 and 13.3). Of course, when there situational necessity

epistemic necessity

Figure 13.1 A semantic map for necessity

situational necessity

epistemic necessity must

Figure 13.2 English must

situational necessity

epistemic necessity

-mAchin Figure 13.3 Evenki -mAchin and -nA

-nA

396

J. van der Auwera and A. Malchukov

are just two uses, the value of drawing shapes and lines is minimal, but for most of the domains for which semantic maps have been proposed, there are many uses and far from trivial connections. For modality, for instance, an elaborate map has been proposed by van der Auwera and Plungian (1998), accounting for semantic relations between necessity and possibility types, subtypes of situational necessity and possibility, as well as for the relation between modal uses and related non-modal ones. A good discussion of recent uses of the semantic map approach is Haspelmath (2003). For what follows, we have to mention four properties of the semantic map approach. First, the approach has a synchronic as well as a diachronic purpose. Synchronically, the goal is to understand dimensions of semantic space and the ways they are covered by language. Diachronically, the goal is to describe how linguistic elements change their meanings. The lines now become paths, and often linguistic elements can move along these paths in only one direction. With respect to necessity, for instance, there is a bulk of research suggesting that a marker of situational necessity may extend its use to epistemic necessity, but never the other way round. Second, whenever a linguistic marker has more than one use, these uses normally have to be contiguous. Consider the abstract semantic maps in Figure 13.4. The conﬁgurations in A and B are allowed, for the spread of the markers aaa and bbb obeys the contiguity requirement. This point can be illustrated with the modals may and can and their employment in (a) capacity, a type of situational possibility that is internal to a participant in the situation, usually the subject, (b) permission, a situational possibility external to the participant, and (c) uncertainty or epistemic possibility. (7) a. *I may swim. [Participant-internal situational possibility] b. You may go now. [Participant-external situational possibility] c. John may be next door. [Epistemic possibility]

A.

B.

C.

use 1

use 2

use 1

use 2

use 1

ccc

use 3

bbb

use 2

Figure 13.4 The contiguity requirement

aaa

use 3

use 3

ccc

A semantic map for depictive adjectivals participant-internal situational possibility

participant-external situational possibility

397

epistemic possibility may

Figure 13.5 English may

participant-internal situational possibility

participant-external situational possibility can

epistemic possibility

Figure 13.6 English can

(8) a. I can swim. [Participant-internal situational possibility] b. You can go now. [Participant-external situational possibility] c. *John can be next door. [Epistemic possibility] The conﬁguration in C (Figure 13.4) ﬂouts the contiguity restriction, and indeed, English does not have any possibility modal that would be used to express only a participant-internal situational and an epistemic possibility. Nevertheless, the conﬁguration is not impossible. It is semantically acceptable, but only if use 3 and use 1 are developments out of a use 2 which is no longer in existence (see e.g. van der Auwera and Plungian 1998 for examples). Even when use 2 is not a common ancestor, we can imagine a constellation in C with a non-semantic motivation, such as language contact. Third, for any two contiguous uses there may be constructions that are at work for one use to the exclusion of the other. (5a), for instance, can only be situational and not epistemic. There may also be constructions that can be taken in both uses, without much or any appreciable semantic difference. The example in (9) is due to Coates (1983: 17) (9) A: Newcastle Brown is a jolly good beer. B: Is it? A: Well it ought to be at that price. (9) makes sense, and almost the same sense, in both a situational reading— for a high price the brewer should provide good beer—and an epistemic reading—it is expensive and therefore I presume it to be good. Finally, locations on the map have so far been called ‘functions’ and ‘uses’. The map as such is noncommittal as to whether different functions amount to different meanings. The mini-maps in Figures 13.2, 13.5, and 13.6,

398

J. van der Auwera and A. Malchukov

for instance, are compatible both with monosemic hypotheses of the English modals, postulating unique meanings and multiple uses, and with polysemic accounts, postulating multiple meanings.

13.3 A semantic map for depictives 13.3.1 Subject and object depictives Depictive adjectival predicates (abbreviated as dep) are secondary predicates; they are ‘copredicative’. The main predication in (1) does not concern George’s anger but his departure. (1) George left the party angry. One can make a rough subdivision between two subtypes of depictive. The subtype illustrated in (1) is the subject depictive. It predicates a property of the subject of the sentence. A depictive can also predicate a property of the object, and this can then be called an object depictive. (10) I eat the meat raw. There are other subtypes, and we follow Schultze-Berndt and Himmelmann (2004) in their claim that the S/A/O concepts are better than subject and object, but for the purposes of our chapter, subject and object will do. 13.3.2 Predicatives and attributives Two non-depictive uses of English adjectives are illustrated in (11) and (12). (11) The angry young men left the party. (12) George was angry. The use in (11) will be called attributive (attr) and the one in (12) predicative (pred), short for predicative of main predication. We claim that the depictive use is intermediate. On the one hand, in its predicative use the adjective (angry) is the main predicate; in its depictive and attributive uses it is not. On the other hand, the attributive use operates within the referring expression that identiﬁes the main predication participant. This is not the case for either the predicative or the depictive use. The intermediacy of depictives vis-a`-vis both predicatives and attributives is easy to represent (see Figure 13.7).1 Note that the morphology of the English adjective does not reﬂect these distinctions at all: the adjective is invariable in 1 In this and the remaining ﬁgures, the following abbreviations are used: adv—adverbial, app— appositive, attr—attributive, compl—complementative, dep—depictive, restr—restrictive.

A semantic map for depictive adjectivals pred

dep

399

attr

Figure 13.7 Depictive, intermediate between predicative and attribute

Pred

Dep

syntax (copula)

Attr

morphology (invariable form)

Figure 13.8 The morphology and syntax of the English depictive

each function. The syntax is more discriminative: at least it singles out the predicative use, in that the predicative use needs a copula, and the depictive and restrictive uses do not.2 The constellations needed for English are also needed for Russian. Russian adjectives have ‘long’ and ‘short’ forms. The long form is found in each of the three uses; the short form is essentially restricted to the predicative use. Russian (13) a. On byl vesel-yj/vesel. he was cheerful-m.sg.nom/cheerful:m.sg ‘He was cheerful.’ b. On vernulsja vesel-yj/*vesel. he returned cheerful-m.sg.nom/cheerful:m.sg ‘He returned cheerful.’ c. Vesel-yj/*vesel student otozvalsja. cheerful-m.sg.nom/cheerful:m.sg student replied ‘A/The cheerful student replied.’ Russian also makes special use of the instrumental case. It is used for predicatives and depictives, but not for attributives. So this is a special use that separates attributives, on the one hand and predicatives and depictives on the other.3 2

The additional discriminatory role of syntax may be a property of other languages as well, and it is probably a very general property with respect to the distinction between predicatives, depictives, and attributives. Our information on syntax is usually more restricted, however. For this reason this chapter concentrates on morphology. 3 Russian also shows that a language may have more than one depictive strategy. Finer semantic, syntactic, or sociolinguistic factors may then inﬂuence the choice. In the case at hand, Nichols (1981) identiﬁes the following factors as favouring the instrumental strategy: non-subject control, iterativity,

400

J. van der Auwera and A. Malchukov

Pred

Dep

Attr

Long form

Short form

Figure 13.9 The long and short forms of the Russian adjective

Pred

short form

Dep

long form

Attr

instrumental form

Figure 13.10 The long, short, and instrumental forms of the Russian adjective

Russian (14) a. On byl vesel-ym. he was cheerful-m.sg.ins ‘He was cheerful.’ b. On vernulsja vesel-ym. he returned cheerful-m.sg.ins ‘He returned cheerful.’ c. *Vesel-ym student otozvalsja. cheerful-m.sg.ins student replied ‘A/The cheerful student replied.’ This predicative–depictive alignment can also be found in Moroccan Arabic. In all three uses adjectives agree in gender and number, but only in the attributive use do they also agree in deﬁniteness. Moroccan Arabic (Maas, p.c.) (15) a. Lqi-t-ha ferhan-a. ﬁnd-pfv-i->her happy-f.sg ‘I found her happy.’ negation, and a high register. As factors that favour the long form she discusses feminine gender, aorist-type context, and a low register. Another point of interest is that the Russian instrumental has a role or function depictive use when applied to nouns. (i) On rabotaet uchitel-em He works teacher-ins.m.sg ‘He works as a teacher.’

A semantic map for depictive adjectivals

Pred

Dep

401

Attr

agreement in gender and number, but not in definiteness agreement in gender, number, and definiteness Figure 13.11 Agreement of the Moroccan Arabic adjective

b. Um-i ferhan-a. mother-my happy-f.sg ‘My mother is happy.’ c. Um-i l-ferhan-a. mother-my def-happy-f.sg ‘my happy mother’ It is also possible for a language to mark the depictive in a way that is different from both the predicative and the attributive. The contiguity requirement further entails that the predicative and the attributive use cannot have an identical marker. This situation is found in the Sudanic language Ngiti. Compare the sentences in (16). Ngiti (Kutsch Lojenga 1994: 348) (16) a. `Iza` rı` ka`zu`. meat be raw ‘Meat is raw.’ b. Ma m-u`ka sı`ta` ka`zu` r/. I 1sg-eat:pfv potato raw from ‘I have eaten potatoes raw.’ c. Ma m-u`ka ka`zu` sı`ta`. I 1sg-eat:pfv raw potato ‘I have eaten raw potatoes.’ By itself the adjective ka`zu` has an attributive use, as shown in (16c). In the predicative use there is a copula (16a) and in (16b), the depictive use, the adjective is accompanied by the locative postposition r/ ‘from, in’. If the attributive and the predicative share a strategy, it will be found in depictives, too. Note that this does not actually preclude depictives from having a dedicated strategy. This situation can be illustrated with the Australian language Mangarayi. It offers two options for encoding depictives. Like many other Australian languages, Mangarayi has a ‘generalized

402

J. van der Auwera and A. Malchukov Pred

Dep

Attr

adjective and r

adjective and copula

bare adjective

Figure 13.12 The Ngiti adjective

pred

dep

‘generalized subordinate clause’

attr

‘descriptive subordinate clause’

Figure 13.13 Depictives in Mangarayi

subordinate clause’, used for adverbial, relative, and complementative as well as depictive functions. More relevant for our discussion is the construction which Merlan (1982: 21) calls the ‘descriptive temporal clause’, a construction which could well be exclusively depictive. (17) illustrates a subject and an object depictive. Mangarayi (Merlan 1982: 21) (17) a. Wula-jiray na-jijga-wa. 3pl>3sg-eat:pst m.nom-small-det ‘They ate it (the ﬁsh) (when he was) small.’

Na-jijga-wa. b. Buy/ 4iyan-wu-ni teach 3sg>1pl.excl-aux-pst 1sg-small-det ‘He taught us [when I was] small.’ Mangarayi serves to make a further point. The depictive in these examples is based on a predicative adjective pattern. Adjectives in Mangarayi are a subclass of nominals but when used in predicative function they may show either N-like or V-like characteristics, depending on the person of the subject. When the subject is a third person, the adjective is nouny and it retains nominal characteristics such as gender, number, and case. When the subject is a ﬁrst or second person, however, the adjectives are verby in that they show person/number agreement with the subject, in the same way as verbal predicates. As is clear from the examples in (17), depictives follow the same pattern as predicatives: in (17a) the adjective is nouny and in (17b) it is verby. However, depictives still differ from predicative adjectives in that they additionally take an article-like sufﬁx, -wa, which marks both appositive and free adjectives; as in na-jijga-wa ‘a small one’. Thus what looks like

A semantic map for depictive adjectivals

403

a specialized depictive form in Mangarayi is actually a ‘mix’ of predicative and attributive strategies. A ﬁnal illustration comes from Germanic. We have already mentioned that in English the adjective is invariable in each use. The opposite is found in Swedish, which has the adjective agreeing with the controller in each use (see also the Wallis dialects of Swiss German—Bucheli Berger, Ch. 4, this volume). An intermediate one is found in German, with agreement only in the attributive use. Swedish (18) a. Bulle-n a¨r varm / bull-ar-na a¨r varm-a. bun-def.g is warm:g.sg bun-pl-def are warm-pl ‘The bun is warm / The buns are warm.’ b. Jag a¨ter ga¨rna bulle-n varm / bull-ar-na varm-a. I eat gladly bun-def.g warm:g.sg bun-pl-def warm-pl ‘I prefer to eat the bun / buns warm.’ c. en varm bulle / varm-a bull-ar a warm:g.sg bun:g.sg warm-pl bun-pl ‘a warm bun / warm buns’ German (19) a. Die Schwaben sind froh. the Swabians are glad ‘The Swabians are glad.’ b. Sie sind froh weggegangen. They are gladly gone.away ‘They gladly went away.’ c. Die froh-en Schwaben . . . The glad-pl Swabians ‘The happy Swabians . . . ’ And with Norwegian past participles we ﬁnd number agreement that is obligatory with attributives, possible in depictives and impossible in predicatives. In the latter case the neutralized form—a neuter singular form—is used, which is possible with depictives, too. Norwegian (A˚shild Næss, p.c.) (20) a. Bil-ene er stja˚l-et/*stja˚l-ne. car-def.pl is steal-pst.ptcp.n/steal-pst.ptcp.pl ‘The cars are stolen.’

404

J. van der Auwera and A. Malchukov

b. De kom skuff-ede/skuff-et tilbake. they came depress.pst.ptcp.pl/depress-pst.ptcp.n back ‘They returned depressed.’ c. stja˚l-ne/*stja˚l-et biler steal-pst.ptcp.pl/steal-pst.ptcp.n car ‘stolen cars’ (21) Agreement Swedish Adj Norwegian pst.ptcp German Adj English Adj

Pred — Dep — Attr yes yes yes no yes yes no no yes no no no

Once again, the intermediacy of the depictive is obvious. It also obvious from the agreement facts of the north-east dialects of Swiss German (Bucheli Berger, Ch. 4, this volume), which have no agreement for predicatives, full agreement for attributives, and a neutralized form for depictives. 13.3.3 Restrictive and appositive attributives In semantic terms, attributives can be subdivided into restrictives and appositives. In restrictive attributives, the attribute is marshalled to help identify the referent. In the context of (11) there might be two groups of young men, the ones that were angry and those that were not. In that setting, angry is meant to be necessary to identify the referent, to restrict the possible reference of the young men. (11) The angry young men left the party. This is not the case for appositives. The angry young men picks out the same referent, but the angriness of the young men does not serve to identify this group. There is no such role for depictives or predicatives either. The semantic map can therefore be extended to yield a new version, in which the attributive function is split into appositive (‘app’) and restrictive (‘restr’) functions (Figure 13.14). In English, there is little formal reﬂection of the distinction between appositive and restrictive adjectives. But consider (22). (22) My father, angry as always, left the party. pred

dep

app

Figure 13.14 Attributives split up into appositives and restrictives

restr

A semantic map for depictive adjectivals

405

angry as always could be taken as appositive to my father, and thus still belonging to the domain of reference, with left the party as the predication. However, it can just as well be taken to be outside the referential domain, just like the clear depictive, angry, in (1). Note that the pragmatics reﬂected by the intonational independence of angry as always can be found with depictives, too (see Ch. 1, section 1.2.4 for further discussion). (23) Angry, George left the party. So at least postposed appositives are more like depictives than restrictives. Not surprisingly, from this perspective, the postposed appositives of German do not have agreement, different from restrictives but like depictives. German (24) Die Schwaben, froh wie immer, sind langsam weggegangen. the Swabians happy as always are slowly gone.away. ‘The Swabians, happy as always, slowly went away.’ A language that clearly shows the differences between restrictives and appositives is Mandarin. In Mandarin, the most common4 depictive strategy is the serial verb construction, the juxtaposition of verbs, including verbs that correspond to adjectives in other languages. Mandarin (Li Renzhi, p.c.) (25) Taˆ bı`sha`ng yanjıˆng le`i le. (s)he close eyes tired pfv ‘[S]he closed her eyes [feeling] tired.’ Verb serialization is also used to encode appositive relative clauses, called ‘descriptive’ by Li and Thompson (1981: 614ff.) Mandarin (Li and Thompson 1981: 614) (26) Woˇ maˇi le yi jia`n yıˆfu ta`i da`. I buy pfv one clf outﬁt too big ‘I bought an outﬁt [which turned out to be] too big.’ Trivially, the deﬁnitionally monopredicative ‘predicative’ use is not accessible to the deﬁnitionally polypredicative serial verb construction. Less trivially, the restrictive relative clause cannot use it either: it uses the particle de instead. 4

Other options—the adverbial de marker and a compound-like structure—are discussed in Zhang (2000).

406

J. van der Auwera and A. Malchukov

pred

dep

app

restr

Figure 13.15 Verb serialization in Mandarin

Mandarin (Li and Thompson 1981: 614) (27) Woˇ maˇi le yi jia`n ta`i I buy pfv one clf too ‘I bought a/the outﬁt that was too

da` de yıˆfu. big rel outﬁt big.’

What we need for Mandarin is therefore the constellation in Figure 13.15. 13.3.4 Complementatives A further use of the English adjective can be called complementative (compl). (28) I consider John intelligent. Like the depictive and the two attributives, the complementative use establishes a kind of secondary predication. The main predication in (28) is that the speaker considers a proposition and it is within that proposition that we ﬁnd a second predication—intelligent is predicated of John. The complementative is different from the attributive and the depictive, however. The difference with the depictive is that the entity that the secondary predicate applies to is not an entity within the primary predication. Whereas George is an argument of the predicate leave in (1), John is not an argument of the predicate consider in (28). The difference between the complementative and the attributive is that they make different kinds of referring expression. Whereas the angry young men in (11) is of the same type as the young men, John intelligent is not of the same type as John. John intelligent refers to a state of affairs. The kind of reference or reiﬁcation is furthermore rather special. John intelligent is not a good enough entity to ﬁgure as the subject of a passive sentence. (29) *John intelligent is considered by me. In fact, passivization makes John resemble objects. (30) John is considered intelligent. We propose to consider both the complementative and the depictive as intermediary between the attributive and the predicative, and we preliminarily

A semantic map for depictive adjectivals

407

compl

pred

app

restr

dep Figure 13.16 The complementative

put them at an equal distance from the predicative and the appositive (Figure 13.16). Note that the complementative in (28) allows copula support, a fact that is compatible with the constellation proposed in Figure 13.16. (31)

I consider John to be intelligent.

Semantically, this means that at least some complementatives can be taken as primary predications, though subordinate ones, just like the ﬁnite and copula-supported subordinate that John is intelligent in (32). (32)

I believe that John is intelligent.

Figure 13.16 does not only say that both depictives and complementatives are intermediate between predicatives and attributes; it furthermore posits a direct connection between them. This is clear with perception verb complementatives, as in (33). (33)

I saw George angry.

In (33) the subject did not only see that George was angry, he also saw George. He saw him angry, of course, but this is rather like the way that the speaker of (10) ate meat and he ate meat raw. We interpret this to mean that certain complementatives, like those of perception verbs, are very close to object depictives. We could put forward the stronger claim that they are object depictives, and that the construction does not distinguish between both interpretations. Note also that English perception verb complements do not allow copula support, and neither do pure object depictives. (34) a. *I saw him to be worried. b. *I ate the meat to be raw. If we do not look at syntactic properties such as copula support, and restrict ourselves to morphology, English must be analysed as a language with a highly polyfunctional adjectival strategy (Figure 13.17).

408

J. van der Auwera and A. Malchukov compl

pred

app

restr

dep Figure 13.17 The polyfunctional English adjective

This constellation almost ﬁts the long adjective of Russian, too. We have already discussed its predicative, depictive, and attributive uses. As to the complementatives, it reaches the complementative of perception verbs, but not those of cognitive verbs. But then we have just argued that perception object complementatives are depictives or very close to them. Russian (35) a. Ja videl ego vozrashchajushch-ego-sja s raboty. I saw him return-acc.m.sg-refl from work ‘I saw him returning from work.’ b. *Ja scitaju ego umn-ogo. I regard him intelligent-acc.m.sg ‘I consider him intelligent.’ The typical complementative strategy is the instrumental strategy. Russian (36) a. Ja videl ego vozrashchajushch-ym-sja s raboty. I saw him return-ins.m.sg-refl from work ‘I saw him returning from work.’ b. Ja scitaju ego umn-ym. I regard him intelligent-ins.m.sg ‘I consider him intelligent.’ We propose Figure 13.18 for the long and instrumental adjectives of Russian. Note that the ﬁgure does not distinguish between subtypes of complementatives, to wit between perception and cognition complementatives. Another constellation seems appropriate for the depictives in the Kwa language Ewe (constituents marked with the predicate marker -i, Ameka, Ch. 11, this volume) and in the West Atlantic language Fula. In Fula verbal forms can be classiﬁed into three groups. First, there are the ‘independent tenses’,

A semantic map for depictive adjectivals

409

compl

pred

app

restr

dep

long form

instrumental form

Figure 13.18 The long and instrumental forms of the Russian adjective

compl

pred

app

restr

dep Figure 13.19 Depictives and complementatives in Fula

which are used in independent sentences as well as in adverbial and complement clauses introduced by conjunctions. Second, ‘relative tenses’ are the forms used in relative clauses. Then there is a third group, involving both a continuous and a stative tense, just as in independent clauses, but these are reserved for what Arnott (1970: 285) describes as either modal subordinate clauses describing concomitant processes or subordinate clauses after verbs of perceiving and ﬁnding. In (37) we get what we would call depictive and complementative uses. This gives us Figure 13.19. Fula (Arnott 1970: 285–7) (37) a. ‘O-warti ‘emo-hiim.oo. he-return:pst he-ponder:cont.prs ‘He came back pondering.’ b. Mi-tawii-e ‘ebe-kaba. _ I-found-them they-ﬁght:cont.pres ‘I found them ﬁghting.’

410

J. van der Auwera and A. Malchukov

compl

pred

app

restr

dep Figure 13.20 Depictives aligned with complementatives and appositives

The Papuan language Alamblak also shows the alignment of Figure 13.19, and the strategy employed is the incorporation of the adjective root into the verb. Alamblak (Bruce 1984: 175–6) (38) a. Miyukham fa-nfri-me¨-an-m. fruit eat-raw-rem.pst-1sg-3pl ‘I ate fruit raw.’ b. Ye¨nr fe¨hm hti-bro-me¨-r-m. child pigs see-big-rem.pst-3sg.m-3pl ‘A child saw pigs as (being) big.’ A third constellation that the map allows is that shown in Figure 13.20, a strategy for complementatives, depictives, and appositives, but neither for predicatives nor for restrictives, and the latter should not share any strategy excluded for the three other functions. Mandarin exempliﬁes this situation. The serial verb construction illustrated in depictive (25) and appositive (26) use, and of which we said that it is excluded for predicatives and restrictives, is allowed in complementatives. Mandarin (Li Renzhi, p.c.) (39) a. Woˇ ka`njia`n ta¯ (za`i) xia`o. I see him/her dur smile ‘I see him/her smiling.’ b. Woˇ re`nwe´i ta¯ co¯ngming. I think [s]he clever ‘I consider him/her smart.’ 13.3.5 Adverbials We take it that all the adjectival uses described so far can be called ‘participantoriented’. Put simply, the adjectives apply a predication to a participant,

A semantic map for depictive adjectivals

411

compl

pred

app

restr

dep

adv Figure 13.21 Adverbials

not to an event or a state of affairs. This is different with adverbials, such as the manner adverb angrily. The expression of the manner in which something happened predicates a property of an event. Of course, the event involves participants, so whatever is predicated of an event pertains to the participants as well. So the predication of properties of events is related to the predication of properties of participants. The type of participant property predication that is most similar to event property predication is the depictive one. In (1) and (3) we are dealing with a secondary predication; in neither does the predicate function in any referential domain, nor does it establish a complementative state of affairs. This yields the map in Figure 13.21. The semantic range of the adv node—or, perhaps better, a further mapping of various adverbial notions—includes manner, concomitance, and simultaneity. This issue will not be discussed here, though, and we refer to Schultze-Berndt and Himmelmann (2004) for the proposal that there is a hierarchy of relevant adverbial notions. Whereas English has a minimal pair with the participant-oriented depictive angry in (1) and the event-oriented manner adverb angrily in (3), Dutch does not. (1) George left the party angry. (3) George left the party angrily. Dutch (4) Georges heeft het feestje woedend verlaten. George has the party angry/angrily left We interpret this to mean that the form woedend is vague between the two perspectives, or, in terms of the theory of parts of speech, that to a certain extent Dutch simply does not distinguish between adverbs of manner and adjectives. In the terminology of Hengeveld (1992a), Dutch would have

J. van der Auwera and A. Malchukov

412

a ﬂexible parts of speech system with a ‘macro-category’, servicing what English distinguishes as adjectives vs. manner adverbs. Given that Dutch attributive adjectives are variable, however—see the -e in (40)—and the other uses are not, we would not really want to say that the macro-category is found in attributives. Dutch (40) De woedende man heeft het feestje verlaten. the angry man has the party left ‘The angry man left the party.’ It is better to say that to a large extent Dutch does not distinguish between predicative, complementative, and depictive adjectives on the one hand and adverbs on the other (see Figure 13.22). As in the previous sections, we can explore whether our materials manifest any of the constellations sanctioned by the semantic map and its contiguity hypothesis. Fula is a case in point. Like English, it has a strategy available for the depictives, complementatives, and adverbials, but unlike English, this strategy is not allowed in predicatives. This strategy is Arnott’s ‘third tense’, already illustrated in (37) and now shown to have adverbial uses too. Figure 13.23 reﬂects this alignment. Fula (Arnott 1970: 287) (41) ‘Ebe-njod’ii, sey Bello wari. they-sit:stat then Bello came ‘When they were sitting down, Bello came.’

compl

pred

app

dep

adv Figure 13.22 The invariable adjective and adverb in Dutch

restr

A semantic map for depictive adjectivals

413

compl

pred

app

restr

dep

adv Figure 13.23 The ‘third tense’ in Fula

Given the existence of the clearly adverbial use, and given the analysis of Dutch woedend, one could argue that (37a) above is not simply depictive, but instead neutral between depictive and adverbial. Another illustration of the constellation in Figure 13.23 is the use of the ‘adverbial’ participle in Tamil. This participle is invariable, and it contrasts with a variable participle used for attributives. An arguably adverbial use is illustrated in (42a): Kumar is not merely depicted as being an alcoholic; the drinking problem describes the manner in which Kumar wastes his money. The use in (42b) is either simply adverbial too or it is neutral between adverbial and depictive. Tamil (Lehmann 1993: 273–4) (42) a. Kumaar tan pana-tt-ait tinam kuti-ttu Kumar he:obl money-obl-acc daily drink-adv.ptcp paazaakku-kir-aan. waste-prs-3sg.m ‘Kumar wastes his money [by] drinking daily.’ b. Kumaar oot-i va-nt-aan. Kumar run-adv.ptcp come-pst-3sg.m ‘Kumar came running.’ The reason for allowing depictiveness as a component of the meaning of (42b) is that we can also use this to describe (43a), viz. as a construction neutral between an object depictive and a complementative. (43b) is a nondepictive complementative.

414

J. van der Auwera and A. Malchukov

Tamil (Lehmann 1993: 273–4) (43) a. Kumaar ivvaluvu cattam-aakap peec-i naan itu Kumar this.much sound-adv speak-adv.ptcp I this varaikum keetka-av-ill-ai. until hear-inf-be:neg-n ‘So far I have not heard Kumar speak so loud.’ b. Kumaar intap pustaka-tt-aip pati-ttu ippootu Kumar this book-obl-acc read-adv.ptcp now muti-tt-aan. ﬁnish-pst-3sg.m ‘Kumar has now ﬁnished reading this book.’ The Mangarayi ‘generalized subordinate clause’ is a case of a strategy available for all positions, except for the predicative one. (44a) is complementative or adverbial, and (44b) looks like being vague between relative, depictive, and adverbial readings (Figure 13.24). Mangarayi (Merlan 1982: 15–16) ˜ a-yiri.wa-ni (44) a. N (w)a-Nala-gala.wu-b 2sg-see-pst.cont sub-1pl-hang.up-pst.punctual na-dajam-gan. n.loc-capparis.umbonata-loc ‘You saw when/that we hung it up in a Capparis umbonata [tree species].’ jaN/ wa-ø-ma-n˜. b. Wurg ø-ga-ni ø-waNgij hide 3sg->3sg-aux.pst n.abs-child die sub-3sg-aux-pst ‘He hid the child who/as he died.’ compl

pred

app

dep

adv Figure 13.24 The ‘generalized subordinate clause’ in Mangarayi

restr

A semantic map for depictive adjectivals

415

The Australian language Nunggubuyu is useful for showing a strategy that links up only adverbials and depictives. In this language, adjectives (‘adjectival nouns’, in the terminology of Heath 1984) are a subclass of nouns: they take nominal class, number, and case sufﬁxes when used as atttributives. However, adjectives differ from nouns in that they are ‘verby’ as predicates: predicative adjectives take the intransitive verb preﬁxes for person, number, and gender— though not any other verbal morphology. A subclass of the ‘adjectives’, including lexemes such as ‘fast’, ‘slow’ and ‘alone’, can be used in the depictive function and they can then pattern as predicative adjectives. Nunggubuyu (Heath 1984: 517) (45)

Ni-bu-buri ni-wiriwiri-wugij. 3m.sg-sit-pst.cont 3m.sg-alone-clitic ‘He was sitting alone.’

What is of interest here is that there is a second pattern, in which the depictives may pattern as adverbials taking the ana- class preﬁx, characteristic of nominals in adverbial function. Nunggubuyu (Heath 1984: 153) (46) a. Ni-ya:-ri: ni-NuluNulug. 3m.sg-go-npst 3m.sg-fast ‘He goes fast.’ b. Wa:ri ani-ruma-N ana-NuluNulug. not 3m.sg.neg-go-npst ana-fast ‘He is not going fast.’ In positive sentences both constructions may be used. However, in negative sentences, as in (46b), only the adverbial pattern is available. If we assume that (46b) is no less depictive than (46a), then we must conclude that the ana strategy is used for both depictives and adverbials. This is shown in Figure 13.25. More evidence for a predicative–depictive–adverbial link comes from Basque. In Basque, izan ‘be’ is the only copula which may combine with a participle that agrees with the subject in number and case. All other copulas and semi-copulas require a participle in -ta, which Saltarelli (1988: 55) catalogues as functioning as an adverbial of manner. (47a) illustrates the predicative use of -ta, and (47b) the adverbial use. Basque (Saltarelli 1988: 63, 55) (47) a. Gu neka.tu-ta g-a-u-de. we:abs tired:pfv.ptcp-man 1pl.abs-prs-be-abs.pl ‘We are tired.’

J. van der Auwera and A. Malchukov

416

compl

pred

app

restr

dep

adv

verby adjectives ana constructions

Figure 13.25 Depictives in Nunggubuyu

compl

pred

app

restr

dep

adv Figure 13.26 The -ta participle in Basque

b. Esku-etan n-u-en-a utz-i-ta hand-pl.loc 1sg.erg-have:pst-comp-sg.abs leave-pfv-man korri-ka joa-n n-in-tz-en. run-adv go-pfv 1sg.abs-pst-aux-pst ‘Dropping what I had in my hands, I ran.’ Depictive participles, it turns out, also take the form in -ta. This gives us Figure 13.26. Basque (Miren L. On˜ederra, p.c.) (48)

Jon-ek kafe-a bero.tu-ta John-erg coffee-sg.det.abs heated:pfv.ptcp-man eda-ten du. drink-prs.ptcp aux.3sg ‘John drinks [his] coffee heated.’

A semantic map for depictive adjectivals

417

compl

pred

app

restr

dep

adv Figure 13.27 Case-agreeing participles in Even

In the Tungusic language Even there is evidence for a link between adverbials and attributives, via depictives (Figure 13.27). In (49a) we see a participle in restrictive use; it agrees with its head in case (and number). (49b) shows the same case agreeing participle in what might be called a detached appositive use, but perhaps better a depictive use. In (49c) we see a reanalysis of the participle marker -ri-/-di- augmented by a speciﬁer -kan ‘exactly’ and the ACC marker -v as a temporal (‘converbial’) marker -rikam ‘when’. Even (Malchukov 1996: 375) (49) a. Etiken girka-d-di-w bej-u nemke-n. old.man walk-dur-nfut.ptcp-acc man-acc shoot-nfut.3sg ‘The old man shot at the walking man.’ b. Tachin girka-d-di-w etiken bej-u like.this walk-dur-nfut.ptcp-acc old man man-acc nemke-n. shoot-nfut.3sg ‘The old man shot the man who/as he was walking like this.’ c. Gia koke-rikem d’uu-la-j tooren iisni. other die-cvb house-loc-refl voice sound:nfut.3sg ‘When the other one died a voice sounded from the house.’ In the Even case, the diachronic dimension is obvious. The participle started in attributive use, and the adverbial use must have come later. This is found in other languages, too, arguably with the participium coniunctum of Latin (Mu¨ller-Bardey 1990: 10) or with the adjectives and participles of Martuthunira and other Australian languages (Schultze-Berndt and

418

J. van der Auwera and A. Malchukov

Himmelmann 2004). Interestingly, the opposite directionality might also be attested. A possible example is the internally headed relative clause in Tungusic languages, illustrated below for Evenki, which is related to Even. What is clear is that the internally headed relative clauses in Tungusic have an adverbial origin (Malchukov 1996; and Dik 1997b: 82–92 for other languages). Consider (50). Evenki (Malchukov 1996: 367; data from L.M. Brodskaya) (50) Baka-ra-n eseNer-du Muriwul Nal-al-di-wi ﬁnd-nfut-3sg tree-dat Murivul hand-pl-ins-ref.poss d0 awacˇad0 a-ri-wa-n. hold-nfut.ptcp-acc-3sg ‘He found Murivul, [while] holding on the tree with his hands.’ The relative clause, an appositive one, is eseNer-du Murivul Nal-al-di-wi d0 awacˇad0 a-ri-wa-n. Its head does not occur as an antecedent, but as an internal constituent. The structure originates from a temporal clause such as (51). Evenki (Malchukov 1996: 367; data from L.M. Brodskaya) (51) Baka-ra-n Ø eseNer-du Muriwul Nal-al-di-wi ﬁnd-nfut-3sg Øi tree-dat Murivuli hand-pl-ins-ref.poss d0 awacˇad0 a-ri-wa-n. hold-nfut.ptcp-acc-3sg ‘He found Murivul, as he was holding on the tree with his hands.’ In (51) the main clause predicate has a zero object, coreferential with the subject of the adverbial. That a structure such as that of (51) is indeed the origin of structures such as (50) is still visible in the kind of agreement markers shown on the participle. The 3sg markers of (50) are not of the attributive type, showing case and number, but rather of the type found in adverbial constructions (‘converbs’). So we have a spread from adverbial morphology into the attributive domain. The question relevant to this chapter is whether this change went through the depictive stage. Semantically, (50) would indeed allow a depictive reading, with ‘holding on the tree with his hands’ being a secondary predication about a main-clause participant; and we have seen before that appositives and depictives are at least very close. Structurally, however, the structure is special, for the semantically main clause participant (Murivul) would occur within the depictive structure—remember that under the relative-clause interpretation, we are dealing with an internally headed relative clause. So either we posit a depictive in-between stage (but then we have to allow ‘internal depictives’) or we do not, but then we have to allow a direct line between attributives and adverbials.

A semantic map for depictive adjectivals

419

compl

pred

app

restr

dep

adv Figure 13.28 The present participles in Maithili

Finally, it is of course possible for a language to have a strategy for all the uses distinguished on the semantic map (Figure 13.28). Such a language may be the Indo-Aryan language Maithili. Its present participle is used for predicatives (in the periphrastic tenses) ((52a)), attributives ((52b)), depictives/complementatives ((52c)), and depictives/adverbials ((52d, e)). Maithili (Yadav 1996: 235–6) (52) a. Kha-it ch1j-l-ah. eat-prs.ptcp aux-pst-3.hon ‘He was speaking.’ chOra ke dekh-l-1hu-ø. b. H´s-´it laugh-prs.ptcp boy acc see-pst-3.nhon ‘I saw the laughing boy.’ d´ur-´it dekh-l-1hu-ø. c. ChOra ke boy acc run-prs.ptcp see-pst-3.nhon ‘I saw the boy running.’

1e-l-1hu. d. H1m d´ur-´it I run-prs.ptcp come-pst-1 ‘I came running.’ e. Kha-it b1Oj-l-ah. eat-prs.ptcp speak-pst-3.hon ‘He spoke while eating.’ As the translations into English show, the English -ing construction is very similar. But English is even more general, in that it allows nominal uses.

420

J. van der Auwera and A. Malchukov

But it is also less general, for it permits the adverbialization of the multifunctional -ing. (53) a. Everybody likes my singing. b. ‘Well yes,’ he answered grudgingly. Note that in both English and Maithili the predicative is still a little special: though it uses the same participial form, it also requires a copula.

13.4 Conclusions and further prospects In the map that we propose, depictives have direct semantic afﬁnities with four other semantic notions, viz. predicatives, complementatives, attributives, and adverbials. This semantic contiguity is reﬂected formally. Depictives may align themselves formally with each of their semantic neighbours, and with more than one. One language may also provide evidence of more than one expression strategy (e.g. Russian), and the expression strategy that characterizes the depictive may in fact be a mix of the expression strategies used for some of its neighbours (e.g. Mangarayi), or a neutralization of a neighbouring strategy (e.g. some dialects of Swiss German: Bucheli Berger, Ch. 4, this volume). The map allows the usual contiguity-based predictions. For instance, if a certain strategy is found in both restrictives and predicatives or in both complementatives and adverbials, it should normally also be found in depictives. Or if we ﬁnd attributive-style agreement in the adverbial domain (e.g. Martuthunira, as described by Dench 1988), we should also ﬁnd it in depictives. The map also allows the usual diachronic perspective. If a strategy disappears (the loss of agreement in Germanic) or spreads (the spread of the agreement morphology into Even adverbials), this happens along the paths of the map. The semantic map proposed is not the ﬁrst one offered for what are here called ‘depictives’. Earlier suggestions can be found in Mu¨ller-Bardey (1990) and in Ko¨nig and van der Auwera (1990). It should not be the last map either. In new versions complex notions, which are given only one location in the present map, might be taken apart. Perhaps the subtypes of appositives and complementatives are better given their own position (see sections 13.3.3 and 13.3.4). Perhaps it is worth representing subject and object depictives separately. The adverbial domain, too, needs to be analysed further, along the lines of Schultze-Berndt and Himmelmann (2004). And most obviously, the map is incomplete. It does not yet provide for resultatives or for subject complementatives.

A semantic map for depictive adjectivals

421

(54) a. I painted the door green. b. John seems intelligent. Clearly, the resultative is rather close to the depictive, and the formal encoding is often identical, as examples from e.g. Alamblak and Mandarin could illustrate. The subject complementative may have to be situated between the predicative and the object complementative, with ‘seem’ being intermediate between a copula and a true complement taking verb. Nevertheless, even at this stage our map makes a strong statement about both the semantic and the formal similarity between various uses of adjectives and, more generally, adjectivals, from both a synchronic and a diachronic perspective.

This page intentionally left blank

References Aarts, Bas (1995). ‘Secondary predicates in English’, in Bas Aarts and Charles F. Meyer (eds), The Verb in Contemporary English: Theory and Description. Cambridge: Cambridge University Press, 75–101. Adams, Bruce (1983). A Tagmemic Analysis of the Wolaitta Language. Ph.D. dissertation, University of London. Agbadja, Kokou Se´name´ (1983). ‘Pour une approche nouvelle des pre´tendus ‘‘adverbes’’ Eße’, Afrique et Langage 19: 32–51. Aikhenvald, Alexandra Y., and Dixon, Robert M. W. (eds) (2006). Serial Verb Constructions in Crosslinguistic Perspective. Oxford: Oxford University Press. Aissen, Judith (2000). ‘A serial verb construction in Tzotzil’, http://ling.ucsc.edu/ Jorge/aissen. Ameka, Felix K. (1991). Ewe: its Grammatical Constructions and Illocutionary Devices. Ph.D. dissertation, Australian National University. —— (1992). ‘Focus constructions in Ewe and Akan: a comparative perspective’, in Chris Collins and Victor Manfredi (eds), Proceedings of the Kwa Comparative Syntax Workshop MIT 1992. Cambridge, Mass.: MIT Working Papers in Linguistics 17, 1–25. —— (2001). ‘Ideophones and the nature of the adjective class in Ewe’, in Voeltz and Kilian-Hatz (2001: 25–48). —— (2002). ‘Constituent order and grammatical relations in Ewe in typological perspective’, in Kristin Davidse and Beatrice Lamiroy (eds), Nominative/Accusative. Amsterdam: Benjamins, 319–52. —— (2003a). ‘Prepositions and postpositions in Ewe (Gbe): empirical and theoretical considerations’, in Patrick Sauzet and Anne Zibri-Hertz (eds), Typologie des langues d’Afrique et universaux de la grammaire. Paris: L’Harmattan, 41–67. —— (2003b). ‘Today is far: situational anaphors in overlapping clause constructions in Ewe’, in Mary E. Kropp Dakubu and Emmanuel K. Osam (eds), Studies in the Languages of the Volta Basin 1. Legon: Department of Linguistics, University of Ghana, 9–22. —— (2006). ‘Ewe serial verb constructions in their grammatical context’, in Aikhenvald and Dixon (2006). —— and Schultze-Berndt, Eva (2000). ‘Adverbialiser or predicate(ion) marker? Ewe -i and its relatives in a typological perspective’, Third World Congress of African Linguistics, Universite´ du Benin, Lome´, Togo, August 2000. Amha, Azeb (2001). The Maale Language. Leiden: Centre for Non-Western Studies. —— and Dimmendaal, Gerrit J. (2006). ‘Verbal compounding in Wolaitta and selforganizing principles in languages’, in Aikhenvald and Dixon (2006: 319–337).

424

References

Amuzu, Evershed K. (1993). Three-place predicates in Ewe. MA thesis, University of Ghana, Legon. Andrews, Avery (1982). ‘A note on the constituent structure of adverbials and auxiliaries’, Linguistic Inquiry 13(3): 313–17. Ansre, Gilbert (1966). The Grammatical Units of Ewe: A Study of their Structure, Classes and Systems. Ph.D. dissertation, University of London. Apridoni Ze, Sˇukia (1986). Sitq˙vatganlageba Axal Kartulsˇi (XIX–XX sauk.. p_ rozis ˙ mixedvit. martivi c.inadadeba) (EI). Tbilisi: Mecniereba. ˙ Arnott, David W. (1970). The Nominal and Verbal Systems of Fula. Oxford: Clarendon Press. Asher, Nicholas (1993). Reference to Abstract Objects in Discourse. Dordrecht: Kluwer. Austin, Peter (1981a). A Grammar of Diyari, South Australia. Cambridge: Cambridge University Press. —— (1981b). ‘Switch-reference in Australia’, Language 57(2): 309–34. —— (ed.) (1988). Complex Sentence Constructions in Australian Languages. Amsterdam: Benjamins, 205–18. —— (1997). Texts in the Mantharta Languages, Western Australia. Tokyo: Institute for the Study of Languages and Cultures of Asia and Africa. Bartsch, Renate (1972). Adverbialsemantik. Frankfurt a. Main: Athena¨um. —— (1976). The Grammar of Adverbials. Amsterdam: North-Holland. Basilaia, Nik.andre (1966). ‘P_ redik. atuli gansazVvrebis sˇesaxeb Axal Kartulsˇi’, Soxumis ˙ P_ edagogiuri Institutis Sˇromebi 18–19: 257–71. ˙˙ ˙ Bayer, Samuel (1997). Confessions of a Lapsed Neo-Davidsonian. New York: Garland. Behaghel, Otto (1923). Deutsche Syntax. Eine geschichtliche Darstellung. Heidelberg: Winter. —— (1928). Geschichte der deutschen Sprache, 5th edn. Berlin: de Gruyter. Bhatt, Christa, and Lindlar, Markus (eds) (1998). Alles Ko¨lsch. Eine Dokumentation der aktuellen Stadtsprache in Ko¨ln. Bonn: Bouvier. Biblia (1989). Biblia. Tbilisi: Sakartvelos Sap_ atriarko. ˙ Bodomo, Adams (1997). Paths and Pathﬁnders: Exploring the Syntax and Semantics of Complex Verbal Predicates in Dagaare and Other Languages. Ph.D. dissertation, Norwegian University of Science and Technology, Trondheim. Boeder, Winfried (1995). ‘Sufﬁxaufnahme in Kartvelian’, in Frans Plank (ed.), Double Case: Agreement by Sufﬁxaufnahme. New York: Oxford University Press, 151–215. —— (1999a). ‘Attribution in Georgian’, in Helma van den Berg (ed.), Studies in Caucasian Linguistics: Selected Papers of the Eighth Caucasian Colloquium. Leiden: Universiteit Leiden, 50–9. —— (1999b). ‘Some notes on the Georgian resultative’, in Werner Abraham and Leonid Kulikov (eds), Tense-Aspect, Transitivity and Causativity. Amsterdam: Benjamins, 117–39. —— (2002). ‘Syntax and morphology of polysynthesis in the Georgian verb’, in Nicholas Evans and Hans-Ju¨rgen Sasse (eds), Problems of Polysynthesis. Berlin: Akademie, 87–111.

References

425

—— and Schroeder, Christoph (1998). ‘Attribution und sekunda¨re Pra¨dikate im Sprachvergleich: Deutsch, Englisch, Kurdisch, Georgisch, Tu¨rkisch’, Sprachtypologie und Universalienforschung 51(3): 207–27. Bohnenberger, Karl (1913). Die Mundart der deutschen Walliser im Heimattal und in den Aussenorten. Frauenfeld: Huber. Bolinger, Dwight (1971). The Phrasal Verb in English. Cambridge, Mass.: Harvard University Press. Braune, Wilhelm (1987). Althochdeutsche Grammatik, 14th edn. Tu¨bingen: Niemeyer. Bresnan, Joan (1982). ‘Control and complementation’, Linguistic Inquiry 13(3): 343–434. —— and Mchombo, Sam (1986). ‘Grammatical and anaphoric agreement’, in Anne Farley, Peter Farley, and Karl-Erik McCullough (eds), Papers from the Parasession on Pragmatics and Grammatical Theory. CLS 22, Part 2. Chicago: Chicago Linguistic Society, 278–97. Breu, Walter (1994). ‘Interactions between lexical, temporal and aspectual meaning’, Studies in Language 18(1): 23–44. Bruce, Les (1984). The Alamblak Language of Papua New Guinea (East Sepik). Canberra: Paciﬁc Linguistics. Brun, Leo (1918). Die Mundart von Obersaxen im Kanton Graubu¨nden. Lautlehre und Flexion. Frauenfeld: Huber. Bucheli Berger, Claudia, and Glaser, Elvira (2002). ‘The syntactic atlas of Swiss German dialects: empirical and methodological problems’, in Sjef Barbiers, Leonie Cornips, and Susanne van der Kleij (eds), Syntactic Microvariation. http:// www.meertens.knaw.nl/books/synmic/ —— (2004). ‘Zur Morphologie des (ko)pra¨dikativen Adjektivs und Partizip II im Alemannischen und Bairischen’, in Franz Patocka and Peter Wiesinger (eds), Morphologie und Syntax deutscher Dialekte und Historische Dialektologie des Deutsch. Vienna: Praesens, 189–226. Burcˇ. ulaZe, Tea (2002). ‘P_ redik.atuli gansazVvrebis ‘‘orgvari’’ bunebis taobaze ˙ Kartulsˇi’/The double nature of the predicate complement in Georgian’ (Re´sume´ 55), Saenatmecniero Z iebani 13: 50–5. BWB: Bayerisches Wo¨rterbuch (1997). Herausgegeben von der Kommission fu¨r Mundartforschung. Heft 3. Munich. Carlson, Gregory N. (1977). Reference to Kinds in English. Bloomington: Indiana University Linguistics Club. —— and Pelletier, Francis J. (eds) (1995). The Generic Book. Chicago: University of Chicago Press. Carrier Duncan, Jill, and Randall, Janet H. (1992). ‘The argument structure and syntactic structure of resultatives’, Linguistic Inquiry 23(2): 173–234. Chierchia, Gennaro (1995a). Dynamics of Meaning. Chicago: University of Chicago Press. —— (1995b). ‘Individual-level predicates as inherent generics’, in Carlson and Pelletier (1995: 176–223).

426

References

Chomsky, Noam A. (1985). The Logical Structure of Linguistic Theory. Chicago: University of Chicago Press. —— (1999). Derivation by Phase. MIT: Occasional Working Papers in Linguistics 18. Cˇikobava, Arnold, et al. (eds) (1950–64). Kartuli enis ganmartebiti leksik.oni. Tbilisi: ˙ SSSR Mecnierebata Ak. ademiis gamomcemloba. Cinque, Guglielmo (1999). Adverbs and Functional Heads: A Cross-Linguistic Perspective. Oxford: Oxford University Press. Clark, Marybeth (1996). ‘Where do you feel? Stative verbs and body-part terms in mainland Southeast Asia’, in Hilary Chappell and William McGregor (eds), The Grammar of Inalienability. Berlin: de Gruyter, 529–64. Clauss, Walter (1929). Die Mundart von Uri. Laut- und Flexionslehre. Frauenfeld: Huber. Coates, Jennifer (1983). The Semantics of the Modal Auxiliaries. London: Croom Helm. Collins, Chris (1993). Topics in Ewe Syntax. Ph.D. dissertation, MIT. Cootes, Sandra (1989). ‘Grammatical and anaphoric agreement in Sesotho’. MS. Corbett, Greville (1998). ‘Morphology and agreement’, in Andrew Spencer and Arnold M. Zwicky (eds), The Handbook of Morphology. Oxford: Blackwell, 191–205. Craig, Colette G. (1977). The Structure of Jacaltec. Austin: University of Texas Press. Dale, Desmond (1972). Shona Companion. Gwelo, Zimbabwe: Mambo Press. Dauwalder, Hans (1992). Haslitiitsch. Wie mma s seid und cha schriiben. Eine haslideutsche Kurzgrammatik. Meiringen: Gemeinnu¨tziger Verein Meiringen. Davidson, Donald (1967). ‘The logical form of action sentences’, in Nicholas Rescher (ed.), The Logic of Decision and Action. Pittsburgh: University of Pittsburgh Press, 81–95. Reprinted in Davidson, Donald (1980), Essays on Actions and Events. Oxford: Clarendon Press, 105–22. Davitiani, Ak. ak.i (1973). Kartuli enis sintaksi I. Sintaksis zogadi sak.itxebi. Martivi ˙ ˙ ˙ c. inadadeba. Tbilisi: Ganatleba. —— Topuria, Varlam, and Kaldani, Maksime (1957). Svanuri p_ rozauli tekstebi II: ˙ ˙ Balskvemouri k. ilo/Svanskie prozaicˇeskie teksty II: Nizˇnebal’skoe narecˇie (¼ Masalebi Kartvelur enata sˇesc.avlisatvis III) (SMA, EI). Tbilisi: SMA gam-ba. Dechaine, Rose-Marie (1993). ‘Serial verb constructions’, in Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld, and Theo Venneman (eds), Syntax: An International Handbook of Contemporary Research, vol. i (Berlin: de Gruyter), 789–825. Delﬁtto, Denis, and Bertinetto, Pier Marco (2000). ‘Word order and quantiﬁcation over times’, in Higginbotham et al. (2000: 207–43). Dench, Alan C. (1988). ‘Complex sentences in Martuthunira’, in Austin (1998: 97–139). —— (1991). ‘Panyjima’, in Robert M. W. Dixon and Barry J. Blake (eds), Handbook of Australian Languages, vol. iv. Melbourne: Oxford University Press, 125–243. —— (1995). Martuthunira: A Language of the Pilbara Region of Western Australia. Canberra: Paciﬁc Linguistics. Diesing, Molly (1992). Indeﬁnites. Cambridge, Mass.: MIT Press.

References

427

Dieth, Eugen (1986). ‘Schwyzertu¨tschi Diala¨ktschrift’, in Christian Schmid-Gadalbert (ed.), Dieth-Schreibung. Aarau: Sauerla¨nder. Dik, Simon C. (1997a). The Theory of Functional Grammar, Part 1: The Structure of the Clause. Berlin: Mouton de Gruyter. —— (1997b). The Theory of Functional Grammar, Part 2: Complex and Derived Constructions. Berlin: Mouton de Gruyter —— Hengeveld, Kees, Vester, Elseline, and Vet, Co (1990). ‘The hierarchical structure of the clause and the typology of adverbial satellites’, in Jan Nuyts, A. Machtelt Bolkestein, and Co Vet (eds), Layers and Levels of Representation in Language Theory. Amsterdam: Benjamins, 25–70. Dimmendaal, Gerrit J. (1983a). The Turkana Language. Dordrecht: Foris. —— (1983b). ‘Turkana as a verb-initial language’, Journal of African Languages and Linguistics 5: 17–44. —— (2006). ‘Head marking, dependent marking and constituent order in the Nilotic area’, in F. K. Erhard Voeltz (ed.), Studies in African Linguistic Typology. Amsterdam: Benjamins, 71–92. Dixon, Robert M.W. (1979). ‘Ergativity’, Language 55: 59–138. —— (1994). Ergativity. Cambridge: Cambridge University Press. Doke, Clement M. (1992 [1927]). Text-book of Zulu Grammar. Cape Town: Maskew Miller Longman. Do¨lling, Johannes (2003). ‘Flexibility in adverbal modiﬁcation: reinterpretation as contextual enrichment’, in Lang et al. (2003: 511–52). Dowty, David R. (1972). ‘Temporally restrictive adjectives’, in John Kimball (ed.), Syntax and Semantics, vol. i. New York: Seminar Press, 51–62. —— (1979). Word Meaning and Montague Grammar: The Semantics of Verbs and Times in Generative Semantics and in Montague’s PTQ. Dordrecht: Reidel. —— (2003). ‘The dual analysis of adjuncts/complements in Categorial Grammar’, in Lang et al. (2003: 33–66). Du¨rscheid, Christa (2002). ‘‘‘Polemik satt und Wahlkampf pur’’: Das postnominale Adjektiv im Deutschen’, Zeitschrift fu¨r Sprachwissenschaft 21: 57–81. Duthie, Alan S. (1996). Introducing Ewe Linguistic Patterns. Accra: Ghana Universities Press. Eckardt, Regine (1997). Events, Adverbs and Other Things. Ph.D. dissertation, University of Stuttgart. Eisenberg, Peter (2002). ‘Morphologie und Distribution: Zur Morphosyntax von Adjektiv und Adverb im Deutschen’, in Friederike Schmo¨e (ed.), Das Adverb: Zentrum und Peripherie einer Wortklasse. Vienna: Praesens, 61–76. Enﬁeld, N. J. (2002a). ‘Cultural logic and syntactic productivity: associated posture constructions in Lao’, in N. J. Enﬁeld (ed.), Ethnosyntax: Explorations in Culture and Grammar. Oxford: Oxford University Press, 231–58. —— (2002b). ‘Semantics and combinatorics of ‘‘sit’’, ‘‘stand’’, and ‘‘lie’’ in Lao’, in John Newman (ed.), The Linguistics of Sitting, Standing, and Lying. Amsterdam: Benjamins, 25–41.

428

References

Enﬁeld, N. J. (2003). Linguistic Epidemiology: Semantics and Grammar of Language Contact in Mainland Southeast Asia. London: Routledge Curzon. —— (2004). ‘Adjectives in Lao’, in Robert M. W. Dixon and Alexandra Y. Aikhenvald (eds), Adjective Classes: A Cross-Linguistic Typology, Oxford: Oxford University Press. 323–47. —— (to appear). ‘Multi-verb sequences in Lao’, in Anthony V. N. Diller and Jerold A. Edmondson (eds), The Tai-Kadai Languages. London: Routledge Curzon. Enukasˇvili, Ruben (1977). ‘P_ redik.atuli gasazVvrebis sak. itxi Zvel Kartulsˇi’, Z veli ˙ Kartuli Enis K. atedris Sˇromebi 20: 151–64. Ernst, Thomas (1984). Towards an Integrated Theory of Adverb Positions. Bloomington: Indiana University Linguistics Club. —— (2000a). ‘Semantic features and the distribution of adverbs’, in FabriciusHansen et al. (2000: 79–97). —— (2000b). ‘Manners and events’, in Tenny and Pustejovsky (2000: 335–58). Essegbey, James (1999). Inherent Complement Verbs Revisited: Towards an Understanding of Argument Structure in Ewe. Ph.D. dissertation, Leiden University. Evans, Nicholas D. (1995). A Grammar of Kayardild, with Historical-Comparative Notes on Tangkic. Berlin: Mouton de Gruyter. Fabricius-Hansen, Cathrine, Lang, Ewald, and Maienborn, Claudia (eds) (2000). Approaching the Grammar of Adjuncts. Berlin: ZAS Papers in Linguistics 17. Fanselow, Gisbert (2001). ‘Features, theta-roles, and free constituent order’, Linguistic Inquiry 32: 405–37. Ferguson, Charles A. (1959). ‘Diglossia’, Word 15: 325–40. Filipenko, Marina V. (2000). ‘Adverbials with ﬂoating and ﬁxed semantic scope’, in Fabricius-Hansen et al. (2000: 99–106). Fillmore, Charles J. (1988). ‘The mechanisms of Construction Grammar’, Berkeley Linguistic Society 14: 35–55. Fintel, Kai von (1994). Restrictions on Quantiﬁer Domains. Ph.D. dissertation, MIT. Fischer, Ludwig (1960). Luzerndeutsche Grammatik und Wegweiser zur guten Mundart. Zu¨rich: Guggenbu¨hl & Huber. Fleck, David (2003). A Grammar of Matses. Ph.D. dissertation, Rice University. Fortune, George (1955). An Analytical Grammar of Shona. London : Longmans Green. —— (1984). Shona Grammatical Constructions, part ii. Harare: Mercury. Fox, Danny (2000). Economy and Semantic Interpretation. Cambridge, Mass.: MIT Press. Frey, Werner (2000). ‘Syntactic requirements on adjuncts’, in Fabricius-Hansen et al. (2000: 107–134). —— and Pittner, Karin (1998). ‘Zur Positionierung von Adverbialen im deutschen Mittelfeld’, Linguistische Berichte 176: 489–534. Fuchs, Gabriela (1993). ‘Das pra¨dikative Adjektiv und Partizip im Walliserdeutschen’, in Helen Christen (ed.), Variationslinguistik und Dialektologie. Ergebnisse aus studienabschliessenden Arbeiten an der Universita¨t Freiburg/Schweiz. Freiburg/ Schweiz: Germanistica Friburgensia 15, 65–79. Gerber, Gustav (1871). Die Sprache als Kunst. Bromberg: Mittler’sche Buchhandlung.

References

429

Geuder, Wilhem (2000). Oriented Adverbs: Issues in the Lexical Semantics of Event Adverbs. Ph.D. dissertation, Universita¨t Konstanz. Gigineisˇvili, Ivane, Topuria, Varlam, and KavtaraZe, Ivane (1961). Kartuli dialektologia I: Kartuli enis k.ilota mok.le ganxilva. T.ekstebi. Leksik. oni/Gruzinskaja ˙ ˙ dialektologija I: Kratkij obzor dialektov gruzinskogo jazyka, Teksty. Slovar’ ˇ (¼Axali Kartuli Enis K.atedris Sromebi 3). Tbilisi TU gam-ba. Glonti, Aleksandre (1955). ‘Pilologiuri nark. vevebi. 2. Atributiviani sˇemasmeneli ˙ ˙ ˙ Kartulsˇi’, Staliniris [Cxinvalis] P_ edagogiuri Institutis Sˇromebi 2: 254–8 (¼Aleksandre ˙ ˙˙ ˙ Glonti (1980), Pilologis ˇcanac. erebi. Tbilisi: Sabcˇ.ota Sakartvelo, 138–43). ˙ —— (1978). Kartuli c. inadadeba. SavarZ#isˇo masala sintakssˇi. Tbilisi: Ganatleba. ˙ Goddard, Cliff (1985). A Grammar of Yankunytjatjara. Alice Springs: Institute for Aboriginal Development. —— (1996). Pitjantjatjara/Yankunytjatjara to English Dictionary. Alice Springs: Institute for Aboriginal Development. Goldberg, Adele E. (1995). Constructions: A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press. Green, Georgia (1970). ‘How abstract is surface structure?’. Chicago Linguistics Society. Chicago: Chicago Linguistics Society, 270–81. —— (1973). ‘A syntactic syncretism in English and French’, in Bray B. Kachru et al. (eds), Issues in Linguistics: Papers in Honor of Henry and Renee Kahane. Urbana: University of Illinois Press, 257–78. Green, Tom (1987). ‘The Semantics of Temporal Clitics in Warlpiri’. MS, University of Sydney/MIT. Gu¨ldemann, Tom (1996). Verbalmorphologie und Nebenpra¨dikationen im Bantu. Eine Studie zur funktional motivierten Genese eines konjugationalen Subsystems. Bochum: Universita¨tsverlag Dr. N. Brockmeyer. —— (1997a). Prosodische Markierung als sprachliche Strategie zur Hierarchisierung verknu¨pfter Pra¨dikationen am Beispiel des Shona. Leipzig: Institut fu¨r Afrikanistik, Universita¨t Leipzig. —— (1997b). ‘Prosodic subordination as a strategy for complex sentence construction in Shona: Bantu moods revisited’, in Robert K. Herbert (ed.), African Linguistics at the Crossroads: Papers from Kwaluseni. Cologne: Ru¨diger Ko¨ppe, 75–98. —— (1998). ‘The relation between imperfective and simultaneous taxis in Bantu: late stages of grammaticalization’, in Ines Fiedler, Catherine Griefenow-Mewis, and Brigitte Reineke (eds), Afrikanische Sprachen im Brennpunkt der Forschung. Linguistische Beitra¨ge zum 12. Afrikanistentag Berlin, 3.–6. Oktober 1996. Cologne: Ru¨diger Ko¨ppe, 157–77. Gussenhoven, Carlos (1992). ‘Sentence accents and argument structure’, in Iggy M. Roca (ed.), Thematic Structure: Its Role in Grammar. Berlin: Foris, 79–105. Haider, Hubert (2001). ‘Adverb placement: convergence of structure and licensing’, Linguistics in Potsdam 6: 50–77. Haiman, John (1983). ‘On some origins of switch reference marking’, in Haiman and Munro (1983: 105–28).

430

References

Haiman, John and Munro, Pamela (eds) (1983). Switch-Reference and Universal Grammar. Amsterdam: Benjamins. Hale, Kenneth (1976). ‘The adjoined relative clause in Australia’, in Robert M. W. Dixon (ed.), Grammatical Categories in Australian Languages. Canberra: Australian Institute of Aboriginal Studies, 78–105. —— (1981). ‘Preliminary remarks on the grammar of part–whole relations in Warlpiri’, in Jim Hollyman and Andrew Pawley (eds), Studies in Paciﬁc Languages and Cultures: In Honour of Bruce Biggs. Auckland: Linguistics Society of New Zealand, 333–44. —— (1982). ‘Some essential features of Warlpiri verbal clauses’, in Swartz (1982: 217–315). —— (1983). ‘Warlpiri and the grammar of non-conﬁgurational languages’, Natural Language and Linguistic Theory 1: 5–47. —— (1994). ‘Core structures and adjuncts in Walpiri syntax’, in Norbert Corver and Henk van Riemsdijk (eds), Studies on Scrambling: Movement and Non-movement Approaches to Free Word-Order Phenomena. Berlin: Mouton de Gruyter, 185–219. —— Laughren, Mary, and Simpson, Jane (1995). ‘Warlpiri syntax’, in Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld, and Theo Vennemann (eds), Syntax: An International Handbook of Contemporary Research, vol. ii (Berlin: de Gruyter), 1430–51. —— and Keyser, Jay (2001). The Basic Elements of Argument Structure. Cambridge, Mass.: MIT. Halliday, Michael A. K. (1967). ‘Notes on transitivity and theme in English, Part 1’, Journal of Linguistics 3: 37–81. Harada, Shin-Ichi (1975). ‘The functional uniqueness principle’, Attempts in Linguistics and Literature 2: 17–24. Harvey, Mark (2002). A Grammar of Gaagudju. Berlin: Mouton de Gruyter. Haspelmath, Martin (1994). ‘Passive participles across languages’, in Barbara Fox and Paul J. Hopper (eds), Voice: Form and Function. Amsterdam: Benjamins, 151–78. —— (1995). ‘The converb as a crosslinguistically valid category’, in Haspelmath and Ko¨nig (1995: 1–56). —— (2003). ‘The geometry of grammatical meaning: semantic maps and crosslinguistic comparison’, in Michael Tomasello (ed.), The New Psychology of Language: Cognitive and Functional Approaches to Language Structure, vol. ii. Mahwah, NJ: Erlbaum, 211–42. —— and Buchholz, Oda (1998). ‘Equative and similative constructions in the languages of Europe’, in van der Auwera (1998: 277–334). —— and Ko¨nig, Ekkehard (eds) (1995). Converbs in Crosslinguistic Perspective: Structure and Meaning of Adverbial Verb Forms. Berlin: de Gruyter. Hayward, Richard J. (ed.) (1990). Omotic Language Studies. London: School of Oriental and African Studies. Heath, Jeffrey (1984). Functional Grammar of Nunggubuyu. Canberra: Australian Institute of Aboriginal Studies.

References

431

Heim, Irene (1982). The Semantics of Deﬁnite and Indeﬁnite Noun Phrases. Ph.D. dissertation, University of Massachusetts. Helbig, Gerhard, and Buscha, Joachim (1994). Deutsche Grammatik. Ein Handbuch fu¨r den Ausla¨nderunterricht, 16th edn. Berlin: Langenscheidt. Hengeveld, Kees (1992a). Nonverbal Predication: Theory, Typology, Diachrony. Berlin: Mouton de Gruyter. —— (1992b). ‘Parts of speech’, in Michael Fortescue, Peter Harder, and Lars Kristoffersen (eds), Layered Structure and Reference in a Functional Perspective: Papers from the Functional Grammar Conference in Copenhagen 1990. Amsterdam: Benjamins, 29–55. Henzen, Walter (1927). Die deutsche Freiburger Mundart im Sense- und su¨do¨stlichen Seebezirk. Frauenfeld: Huber. Herburger, Elena (2000). What Counts: Focus and Quantiﬁcation. Cambridge, Mass.: MIT Press. Higginbotham, James, Pianesi, Fabio, and Varzi, Achille C. (eds) (2000). Speaking of Events. Oxford: Oxford University Press. Himmelmann, Nikolaus P. (1997). Deiktikon, Artikel, Nominalphrase. Zur Emergenz syntaktischer Struktur. Tu¨bingen: Niemeyer. Hinton, Leanne, Nichols, Johanna, and Ohala, John (eds) (1994). Sound Symbolism. Cambridge: Cambridge University Press. Hotzenko¨cherle, Rudolf (1934). Die Mundart von Mutten. Laut- und Flexionslehre. Frauenfeld: Huber. —— (1961). ‘Die Raumstruktur des Schweizerdeutschen’, Zeitschrift fu¨r Mundartforschung 3: 207–27. Hudson, Joyce (1978). The Core of Walmatjarri Grammar. Canberra: Australian Institute of Aboriginal Studies. Hyman, Larry M., and Watters, John R. (1984). ‘Auxiliary focus’, Studies in African Linguistics 15: 233–73. Idiotikon: Schweizerisches Idiotikon. Wo¨rterbuch der schweizerdeutschen Sprache ed. Friedrich Staub and Ludwig Tobler. Frauenfeld: Huber, vol. i (1881). Imnaisˇvili, Ivane (1948). ‘Sˇedgenili sˇemasmeneli Kartulsˇi’/Sostavnoe skazuemoe v gruzinskom jazyke (Re´sume´ 214), Pusˇk.inis saxelobis Saxelmc.ipo P_ edagogiuri Institutis Sˇromebi 5: 187–216. ˙˙ ˙ —— (1957). Saxelta bruneba da brunvata punkciebi Z vel Kartulsˇi (¼Zveli Kartuli Enis K.atedris Sˇromebi 4). Tbilisi: Tbilisis universtetis gamomcemloba. ˙ ˙ Imnaisˇvili, Vaxtang (1997). ‘Zu einer Art des ‘‘einfachen Objekts’’ im Georgischen’, ˙ Georgica 20: 83–8. Jackendoff, Ray S. (1972). Semantic Interpretation in Generative Grammar. Cambridge, Mass.: MIT Press. Jacobs, Joachim, (1997). ‘I-Topikalisierung’, Linguistische Berichte 168: 91–133. —— von Stechow, Arnim, Sternefeld, Wolfgang and Vennemann, Theo (eds) (1995). Syntax: An International Handbook of Contemporary Research. Berlin: de Gruyter.

432

References

Jaggar, Philip (1992). An Advanced Hausa Reader with Grammatical Notes and Exercises. London: School of Oriental and African Studies. Jakobson, Roman (1936). ‘Beitrag zur allgemeinen Kasuslehre. Gesamtbedeutungen der russischen Kasus’, in Roman Jakobson, Selected Writings, vol. ii. The Hague: Mouton de Gruyter, 23–71. —— (1971 [1957]). ‘Shifters, verbal categories, and the Russian verb’, in Roman Jakobson, Selected Writings vol. ii. The Hague: Mouton de Gruyter, 130–47. Janda, Laura A. (1993). A Geography of Case Semantics: The Czech Dative and the Russian Instrumental. Berlin: Mouton de Gruyter. Jespersen, Otto (1965). A Modern English Grammar on Historical Principles, part iii: Syntax. 2nd vol. London: Allen & Unwin. Johanson, Lars (1995). ‘On Turkic converb clauses’, in Haspelmath and Ko¨nig (1995: 313–47). Johnston, Michael (1995). ‘When-clauses, adverbs of quantiﬁcation, and focus’, in Susanne Preuss, Raul Aranovich, William Byrne, and Martha Senturia (eds), Proceedings of the 13th West Coast Conference in Formal Linguistics. Cambridge: Cambridge University Press, 549–64. Katz, Graham (1993). The Semantics of Free Adjuncts: Deriving the Weak/Strong Distinction. Cambridge Mass.: MIT Working Papers in Linguistics 20. —— (2000). ‘Anti neo-Davidsonianism: against a Davidsonian semantics for state sentences’, in Tenny and Pustejovsky (2000: 393–416). Kay, Paul, and Fillmore, Charles J. (1999). ‘Grammatical constructions and linguistic generalizations: the what’s X doing Y? construction’, Language 75: 1–33. Keenan, Edward L., and Comrie, Bernard (1977). ‘Noun phrase accessibility and universal grammar’, Linguistic Inquiry 8: 63–99. Kilham, Christine, Pamulkan, Mabel, Pootchemunka, Jennifer, and Wolmby, Topsy (1986). Dictionary and Source Book of the Wik-Munkan Language. Darwin: Summer Institute of Linguistics. K. iziria, Anton (1977). ‘Gavrcobili sˇemasmeneli Kartvelur enebsˇi’, Sakartvelos SSR ˙ Mecnierebata Ak.ademiis Enatmecnierebis Instituti, Tezisebi 33: 9–11. ˙˙ ˙ Klimov, Georgij (1974). ‘On the character of active languages’, Linguistics 131: 11–23. Kniffka, Gabriele (1996). NP-Aufspaltung im Deutschen. Hu¨rth: Gabel. Kofod, Frances M. (n.d.). ‘Introduction to Kija grammar’. MS. Koizumi, Masatoshi (1994). ‘Secondary predicates’, Journal of East Asian Linguistics 3: 25–59. Ko¨nig, Ekkehard, and Van der Auwera, Johan (1990). ‘Adverbial participles, gerunds and absolute constructions in the languages of Europe’, in Johannes Bechert, Giuliano Bernini, and Claude Buridant (eds), Toward a Typology of European Languages. Berlin: Mouton de Gruyter, 337–55. Koster, Jan, and Zwart, Jan-Wouter (2000). ‘Transitive expletive constructions and the object shift parameter’. MS, University of Groningen. Kratzer, Angelika (1995). ‘Stage-level and individual-level predicates’, in Carlson and Pelletier (1995: 125–75).

References

433

Kroeger, Paul R. (1993). ‘Another look at subjecthood in Tagalog’, Philippine Journal of Linguistics 24(2): 1–15. Kuno, Susumu, and Takami, Ken-Ichi (1993). Grammar and Discourse Principles: Functional Syntax and GB Theory. Chicago: University of Chicago Press. Kutsch Lojenga, Constance (1994). Ngiti: A Central-Sudanic Language of Zaire. Cologne: Ru¨diger Ko¨ppe. Kutscher, Silvia (2001). Nomen und nominales Syntagma im Lasischen. Eine deskriptive Analyse des Dialekts von Ardes,en. Munich: Lincom Europa. —— and Genc¸, Nuran Sevim (1998). Ardes,en narrates—Ardes, eni na isinapinenpe. A Collection of Laz Spoken Texts with Glosses and Translations into English, German and Turkish. Munich: Lincom Europa. —— Mattissen, Johanna, and Wodarg, Anke (eds) (1995). Das Mut’aﬁ-Lazische. Cologne: Department of Linguistics, University of Cologne. K.vacˇ. antiraZe, Tamaz (1978). Msgavsebiti ˇsedareba Kartulsˇi. Tbilisi: Tbilisis uni˙ verstetis gamomcemloba. ˙ ˙ K.vacˇ. aZe, Leo (1957). ‘MimVeobisa da saxelis zogierti pormis sintaksuri k. valipik. a˙ ciisatvis’, K. omunisturi AVzrdisatvis 12: 27–36. ˙ —— (1996). Tanamedrove Kartuli enis sintaksi. Meotxe sˇevsebuli gamocema. Tbilisi: ˙ Rubik.oni. —— (1999). ‘P_ redik. atuli gansazVvreba Zvel Kartulsˇi’, Enatmecnierebis Sak. itxebi ˙ 1999(4): 31–40. —— (2000a). ‘P_ redik.atuli gansazVvreba Kartuli enis dialektebsˇi’, Enatmecnierebis ˙ ˙ Sak.itxebi 2000(3): 3–9. —— (2000b). ‘P_ redik. atuli gansazVvreba Svanursˇi’, Sulxan-Saba Orbelianis saxelobis ˙ Tbilisis Saxelmc.ipo P_ edagogiuri Universitetis Kartuli Enis K. atedris Sˇromebi 5: 5–12. ˙˙ Lamberti, Marcello, and Sottile, Roberto (1997). The Wolaitta Language. Cologne: Ru¨diger Ko¨ppe. Landman, Fred (2001). Events and Plurality. Dordrecht: Kluwer. Lang, Ewald, Maienborn, Claudia, and Fabricius-Hansen, Cathrine (eds) (2003). Modifying Adjuncts. Berlin: de Gruyter. Larson, Richard K. (1988). ‘On the double object construction’, Linguistic Inquiry 19: 335–91. —— (1991). ‘Some issues in verb serialization’, in Claire Lefebvre (ed.), Serial Verbs: Grammatical, Comparative and Cognitive Approaches. Amsterdam: Benjamins, 185–210. Laughren, Mary (1992). ‘Secondary predication as a diagnostic of underlying structure in Pama-Nyungan languages’, in Iggy Roca (ed.), Thematic Structure: Its Role in Grammar. Berlin: Mouton de Gruyter, 199–246. —— (2002). ‘Syntactic constraints in a free word order language’, in Mengistu Amberber and Peter Collins (eds), Language Universals and Variation. Westport, Conn.: Ablex Greenwood, 83–130. —— (in press). ‘Some Warlpiri non-ﬁnite clauses reviewed’, in Rachel Nordlinger (ed.), Australian Journal of Linguistics, Special Issue.

434

References

Lazog˘lu, Fahri, and Feurstein, Wolfgang (1984). Lazuri Alfabe. Lazca Alfabe. Entwurf eines lazischen Alphabetes. Parpali 1. Lazuri Carelepe. Laz dili ve ku¨ltu¨ru¨ yay¸nlar¸. Progrom. Lechner, Winfried (2001). ‘Phrase structure paradoxes, movement and ellipsis’. MS, University of Tu¨bingen. Legendre, Ge´raldine (1997). ‘Secondary predication and functional projections in French’, Natural Language and Linguistic Theory 15(1): 43–87. Lehmann, Christian (1988). ‘Towards a typology of clause linkage’, in John Haiman and Sandra A. Thompson (eds), Clause Combining in Grammar and Discourse. Amsterdam: Benjamins, 181–225. Lehmann, Thomas (1993). A Grammar of Modern Tamil. Pondicherry: Pondicherry Institute of Linguistics. Lewis, David (1975). ‘Adverbs of quantiﬁcation’, in Edward Keenan (ed.), Formal Semantics. Cambridge: Cambridge University Press, 3–15. Lewis, Marshall (1985). Ewe Relativization, NP Accessibility and Universal Grammar. Bloomington: Department of Linguistics, Indiana University. Li, Charles N., and Thompson, Sandra A. (1981). Mandarin Chinese: A Functional Reference Grammar. Berkeley: University of California Press. Loos, Eugene E. (1999). ‘Pano’, in Robert M. W. Dixon and Alexandra Y. Aikhenvald (eds), The Amazonian Languages. Cambridge: Cambridge University Press, 227–50. Lu¨bker, Friedrich (1837). Grammatische Studien. Studien zur Syntax des Adjectivums und Adverbiums in den alten Sprachen. Parchim und Ludwigslust: Hinstorff’sche Hofbuchhandlung. McConnell-Ginet, Sally (1982). ‘Adverbs and logical form’, Language 58: 144–84. McGregor, William B. (1989). ‘Phrase fracturing in Gooniyandi’, in La´szlo´ Mara´cz and Pieter Muysken (eds), Conﬁgurationality: The Typology of Asymmetries. Dordrecht: Foris, 207–22. —— (1990). A Functional Grammar of Gooniyandi. Amsterdam: Benjamins. —— (1992a). ‘The noun phrase in (some) Australian languages: a reply to Mark Harvey’, Australian Journal of Linguistics 12: 315–19. —— (1992b). ‘The semantics of ergative marking in Gooniyandi’, Linguistics 30: 275–318. —— (1997a). ‘Functions of noun phrase discontinuity in Gooniyandi’, Functions of Language 4: 83–114. —— (1997b). ‘Indeﬁnite and comparative determiners in Australian languages’. Paper presented at the Second International Workshop on Australian Linguistics, University of Melbourne, December 1997. —— (1997c). Semiotic Grammar. Oxford: Clarendon Press. —— (1998a). ‘Numeral and Quantifying Words in Australian Languages’. Paper presented to Institut fu¨r Sprachwissenschaft and Institut fu¨r Didaktik der Naturwissenschaft, Universita¨t Salzburg. —— (1998b). ‘Optional ergative marking in Gooniyandi revisited: implications to the theory of marking’, Leuvense Bijdragen 87: 491–534.

References

435

McKay, Graham (2000). ‘Ndje´bbana’, in Robert M. W. Dixon and Barry Blake (eds), The Handbook of Australian Languages, vol. 5. Melbourne: Oxford University Press Australia, 154–354. McNulty, Elaine (1988). The Syntax of Adjunct Predicates. Ph.D. dissertation, University of Connecticut. Maienborn, Claudia (2000). ‘Modiﬁcation and underspeciﬁcation: a free variable account of locative modiﬁers’, in Fabricius-Hansen et al. (2000: 153–76). —— (2001). ‘On the position and interpretation of locative modiﬁers’, Natural Language Semantics 9: 191–240. Malchukov, Andrej (1996). ‘Internal relative clauses in Tungusic languages in a synchronic and diachronic perspective’, Sprachtypologie und Universalienforschung 4: 358–82. Manser, Joe (2001). Innerrhoder Dialekt. Mundartwo¨rter und Redewendungen aus Appenzell Innerrhoden. Appenzell: Appenzeller Volksfreund. Maslov, Jurij S. (1988). ‘Resultative, perfect, and aspect’, in Vladimir P. Nedjalkov (ed.), Typology of Resultative Constructions. Amsterdam: Benjamins, 257–67. Mattissen, Johanna (1995). ‘Verbmorphologie’, in Kutscher et al. (1995: 45–82). Menge, Hermann (2000). Lehrbuch der lateinischen Syntax und Semantik. Completely new edition by Thorsten Burkard. Darmstadt: Wissenschaftliche Buchgesellschaft. Merkle, Ludwig (1975). Bairische Grammatik. Munich: Heimeran. Merlan, Francesca (1982). Mangarayi. Amsterdam: North-Holland. Meeussen, Achille E. (1967). ‘Bantu grammatical reconstructions’, Africana Linguistica 3: 79–121. Milsark, Gary L. (1974). Existential Sentences in English. Ph.D. dissertation, MIT. Moulton, William G. (1941). Swiss German Dialect and Romance Patois. Ph.D. dissertation, Yale University. Mu¨ller, Stefan (2002). Complex Predicates: Verbal Complexes, Resultative Constructions, and Particle Verbs in German. Stanford, Calif.: CSLI. Mu¨ller-Bardey, Thomas (1990). ‘Kopra¨dikation als grundlegende syntaktische Operation’, in Hansjakob Seiler with Elﬁe Konrad and Birgit Schwarze (eds), Internationales interdisziplina¨res Kolloquium ‘Sprache und Denken: Variation und Invarianz in Linguistik und Nachbardisziplinen’, Lenzburg/Schweiz, May 1989. Cologne: Arbeiten des Ko¨lner Universalien-Projekts (akup) 81, 1–20. Nakajima, Heizo (1990). ‘Secondary predication’, Linguistic Review 7(3): 275–309. Napoli, Donna Jo (1975). ‘A global agreement phenomenon’, Linguistic Inquiry 3: 413–35. —— (1989). Predication Theory: A Case Study for Indexing Theory. Cambridge: Cambridge University Press. Nash, David (1982). ‘Warlpiri preverbs and verb roots’, in Swartz (1982: 165–216). —— (1986). Topics in Warlpiri Grammar. New York: Garland. Nedjalkov, Igor V. (1997). Evenki. London: Routledge. Nedjalkov, Vladimir P. (1995). ‘Some typological parameters of converbs’, in Haspelmath and Ko¨nig (1995: 97–136).

436

References

Neeleman, Ad (1994). Complex Predicates. Utrecht: Led. Nekes, Herman, and Worms, Ernest A. (1953). Australian Languages. Fribourg: Anthropos Institut. Newman, Paul (2000). The Hausa Language: An Encyclopedic Reference Grammar. New Haven, Conn.: Yale University Press. Nichols, Joanna (1978a). ‘Secondary predicates’, Berkeley Linguistics Society 4: 114–27. —— (1978b). ‘Double dependency?’, Chicago Linguistics Society 14: 326–39. —— (1981). Predicate Nominals: A Partial Surface Syntax of Russian. Berkeley: University of California Press. —— (1982). ‘Prominence, cohesion, and control: object-controlled predicate nominals in Russian’, in Paul Hopper and Sandra A. Thompson (eds), Studies in Transitivity. New York: Academic Press, 317–50. —— (1983). ‘Switch reference in the Northeast Caucasus’, in Haiman and Munro (1983: 245–65). —— (1986). ‘Head-marking and dependent-marking grammar’, Language 62: 56–119. Ogawa, Yoshiki (2001). ‘The stage/individual distinction and (in)alienable possession’, Language 77(1): 1–25. Paul, Hermann (1919). Deutsche Grammatik, vol. iii. Halle (Saale): VEB Max Niemeyer. —— (1998). Mittelhochdeutsche Grammatik. 24th edn. Tu¨bingen: Niemeyer. Pianesi, Fabio, and Varzi, Achille C. (2000). ‘Events and event talk: an introduction’, in Higginbotham et al. (2000: 3–47). Pinkster, Harm (1988). Lateinische Syntax und Semantik. Tu¨bingen: Francke. Pittner, Karin (1999). Adverbiale im Deutschen. Untersuchungen zu ihrer Stellung und Interpretation. Tu¨bingen: Stauffenburg. Plank, Frans (1985). ‘Pra¨dikativ und Kopra¨dikativ’, Zeitschrift fu¨r Germanistische Linguistik 13(2): 154–85. Platt, John T., and Platt, Heidi K. (1972). ‘Orientation of manner adverbials’, Papers in Linguistics: 227–49. [Q˙ipsˇiZe, Ioseb] (1914). I. Kipsˇidze: Grammatika mingrel’skago (iverskago) jazyka s’’ chrestomatieju i slovarem’’ (¼ Materialy po jafeticˇeskomu jazykoznaniju 7). Sankt Peterburg: Tipograﬁja Imperatorskoj Akademii Nauk’’. Quirk, Randolph, Greenbaum, Sidney, Leech, Geoffrey, and Svartvik, Jan (1985). A Comprehensive Grammar of the English Language. London: Longman. Rapoport, Tova R. (1991). ‘Adjunct predicate licensing and D-structure’, in Susan Rothstein (ed.), Perspectives of Phrase Structure: Heads and Licensing. San Diego, Calif.: Academic Press, 159–87. —— (1993). ‘Verbs in depictives and resultatives’, in James Pustejovsky (ed.), Semantics and the Lexicon. Dordrecht: Kluwer, 163–84. —— (1999). ‘Structure, aspect, and the predicate’, Language 75: 653–77.

References

437

Reinhart, Tanya (1995). ‘Interface economy and markedness’, in Chris Wilder, Hans-Martin Ga¨rtner, and Manfred Bierwisch (eds), The Role of Economy Principles in Linguistic Theory. Berlin: Akademie, 146–69. Roberts, Ian (1988). ‘Predicative APs’, Linguistic Inquiry 19(4): 703–10. Rooth, Mats (1995). ‘Indeﬁnites, adverbs of quantiﬁcation, and focus semantics’, in Carlson and Pelletier (1995: 265–99). ¨ ber die Sprache der Lazen. Lemgo, Detmold: Meyersche Rosen, Georg (1844). U Hofbuchhandlung. Rothstein, Susan D. (1985). The Syntactic Forms of Predication. Bloomington: Indiana University Linguistics Club. —— (2000). ‘Secondary predication and aspectual structure’, in Fabricius-Hansen et al. (2000: 241–64). —— (2001). Predicates and their Subjects. Dordrecht: Kluwer Academic. Rottland, Franz (1982). Die su¨dnilotischen Sprachen. Beschreibung, Vergleichung und Rekonstruktion. Berlin: Reimer. SADS: Syntaktischer Atlas der Deutschen Schweiz (ongoing project). Saltarelli, Mario (1988). Basque. London: Routledge. SˇaniZe, Ak. ak.i, and Topuria, Varlam (eds) (1939). Svanuri p_ rozauli tekstebi I. Balsze˙ ˙ mouri k. ilo. T.ekstebi sˇek.ribes A. SˇaniZem da V. Topuriam (Masalebi Kartvelur enata ˙ sˇesc.avlisatvis. I.) (SSRK. Mecnierebata Ak. ademiis Sakartvelos Piliali. ENIMK.I.) Tbilisi: SSRK. Mecnierebata Ak. ademiis Sakartvelos Pilialis gamomcemloba. —— and Kaldani, Maksime (eds.) (1978). Svanuri enis krestomatia. tekstebi sˇek.ribes ˙ ˙ ˙ A. SˇaniZem, M. Kaldanma da Z. Cˇ. umburiZem (¼Zveli Kartuli enis k.atedris sˇromebi 21). Tbilisi: Tbilisis universtetis gamomcemloba. ˙ ˙ Sapir, David J. (1977). ‘The anatomy of metaphor’, in David J. Sapir and Jon C. Crooker (eds), The Social Use of Metaphor. Philadelphia: University of Pennsylvania Press, 3–32. Schein, Barry (1995). ‘Small clauses and predication’, in Anna Cardinaletti and Maria Teresa Guasti (eds), Small Clauses. New York: Academic Press, 49–76. Schroeder, Christoph (2003). Depiktive im Sprachvergleich Deutsch-Tu¨rkisch. Eine kontrastiv-typlogische Analyse. Habilitationsschrift, Universita¨t Osnabru¨ck. Schultze-Berndt, Eva (2000). Simple and Complex Verbs in Jaminjung: A Study of Event Categorisation in an Australian Language. Ph.D. dissertation, Catholic University of Nijmegen. —— (2001). ‘The cognitive basis of allative marking of secondary predicates: evidence from Australian languages’. Paper presented at the International Cognitive Linguistics Conference, Santa Barbara, Calif., 24 July 2001. —— (2002). ‘Grammaticalized restrictive clitics on adverbials and secondary predicates: evidence from Australian languages’, Australian Journal of Linguistics 22(2): 231–64. —— (2006). ‘Secondary Predicates in Australian Languages’, in Martin Everaert and Henk van Riemsdijk (eds), The Blackwell Companion to Syntax, Volume IV. Malden: Blackwell.

438

References

Schultze-Berndt, Eva and Dawuda, Carmen (2001). ‘German depictive secondary predicates in a typological perspective’. MS, Ruhr-Universita¨t Bochum. —— and Himmelmann, Nikolaus P. (2004). ‘Depictive secondary predicates in crosslinguistic perspective’, Linguistic Typology 8(1): 59–131. SDS: Sprachatlas der deutschen Schweiz (1975). Band iii, Formengeographie, bearbeitet von Doris Handschuh, Rudolf Hotzenko¨cherle, Rudolf Tru¨b u.a., Bern. Sicard, Harald von (1965). Ngano dze CikaraNga (Karangama¨rchen). Uppsala: Institutionen fo¨r Allma¨n och Ja¨mfo¨rende Etnograﬁ vid Uppsala Universitet. Sievers, Eduard (1892). Tatian. Lateinisch und altdeutsch mit ausfu¨hrlichem Glossar, 2nd ed. Paderborn: Scho¨ningh. Simpson, Jane (1988). ‘Case and complementiser sufﬁxes in Warlpiri’, in Austin (1988: 205–18). —— (1991). Warlpiri Morphosyntax: A Lexicalist Approach. Dordrecht: Kluwer. —— and Bresnan, Joan (1983). ‘Control and obviation in Warlpiri’, Natural Language and Linguistic Theory 1: 49–64. Soi Rawa (Guillermo Ramı´rez Guimaraes) (1995). Soi Rawa: joni benatian non moabo ipaoni. Programa de Formacio´n de Maestros Bilingu¨es de la Amazonı´a Peruana. Lima: Asociacio´n Intere´tnica de Desarrollo de la Selva Peruana and Instituto Superior Pedago´gico ‘Loreto’. Sonderegger, Stefan, and Gadmer, Thomas (1999). Appenzeller Sprachbuch. Der Appenzeller Dialekt in seiner Vielfalt. Appenzell & Herisau: Erziehungsdirektionen AR und AI. Spagnolo, L.M. (1933). Bari Grammar. Verona: Missioni Africane. —— (1960). Bari English Italian Dictionary. Verona: Missioni Africane. Sparing-Cha´vez, Margarethe (1998). ‘Interclausal reference in Amahuaca’, in Desmond C. Derbyshire and Geoffrey K. Pullum (eds), Handbook of Amazonian Languages, vol. iv. Berlin: Mouton de Gruyter, 443–85. Steube, Anita (1994). ‘Syntaktische und semantische Eigenschaften sekunda¨rer Pra¨dikate’, in Anita Steube and Gerhild Zybatow (eds), Zur Satzwertigkeit von Inﬁnitiven und Small Clauses. Tu¨bingen: Niemeyer, 243–64. Stowell, Timothy (1981). Origins of Phrase Structure. Ph.D. dissertation, MIT. Stucki, Karl (1917). Die Mundart von Jaun im Kanton Freiburg. Lautlehre und Flexion. Frauenfeld: Huber. Stump, Gregory T. (1985). The Semantic Variability of Absolute Constructions. Dordrecht: Reidel. Swartz, Stephen M. (ed.) (1982). Papers in Warlpiri Grammar, in Memory of Lothar Jagst. Berrimah, NT: Summer Institute of Linguistics. —— (1991). Constraints on Zero Anaphora and Word Order in Warlpiri Narrative Text. Darwin, NT: Summer Institute of Linguistics. Szadrowsky, Manfred (1936). ‘Zur hochalemannischen Syntax IV’, Beitra¨ge zur Geschichte der Deutschen Sprache und Literatur 60: 445–58. Takezawa, Koichi (1993). ‘Secondary predication and locative/goal phrases’, in Nobuko Hasegawa (ed.), Japanese Syntax in Comparative Grammar. Tokyo: Kurosio, 45–77.

References

439

Tenny, Carol (2000). ‘Core events and adverbial modiﬁcation’, in Tenny and Pustejovsky (2000: 285–334). —— and Pustejovsky, James (eds) (2000). Events as Grammatical Objects: The Converging Perspectives of Lexical Semantics and Syntax. Stanford Calif.: CSLI. Tobler, Titus (1837). Appenzellischer Sprachschatz. Zu¨rich: Orell Fu¨ssli. Toweett, Taaitta (1979). A Study of Kalenjin Linguistics. Nairobi: Kenya Literature Bureau. UturgaiZe, Tedo (1960). Tusˇuri k. ilo (¼Zveli Kartuli Enis K.atedris Sˇromebi 6). Tbilisi: Tbilisis universtetis gamomcemloba. ˙ ˙ Valenzuela, Pilar M. (1997). ‘Basic verb types and argument structures in ShipiboConibo’. MA thesis, University of Oregon. —— (1998). ‘El morfema de ergatividad en el Shipibo-Conibo’, in Luis Miranda and Amanda Orellana (eds), Actas del II Congreso Nacional de Investigaciones Lingu¨´ıstico-Filolo´gicas. Lima, Peru: Universidad Ricardo Palma, 217–45. —— (1999). ‘Adverbials, transitivity and switch-reference in Shipibo-Konibo (Panoan)’, in Sabrina J. Billings, John P. Boyle, and Aaron M. Grifﬁth (eds), Chicago Linguistic Society 35: Theory and Linguistic Diversity. Chicago: Chicago Linguistic Society, 355–71. —— (2002). ‘Causativization and transitivity in Shipibo-Konibo’, in Masayoshi Shibatani (ed.), The Grammar of Causation and Interpersonal Manipulation. Amsterdam: Benjamins, 417–83. —— (2003). Transitivity in Shipibo-Konibo Grammar. Ph.D. dissertation, University of Oregon. Valiquette, Hilaire (1993). A Basic Kukatja to English Dictionary. Wirrimanu (Balgo): Luurnpa Catholic School. Van der Auwera, Johan (ed.) (1998). Adverbial Constructions in the Languages of Europe. Berlin: Mouton de Gruyter. —— and Plungian, Vladimir (1998). ‘Modality’s semantic map’, Linguistic Typology 2: 79–124. Van Valin, Robert, and LaPolla, Randy J. (1997). Syntax: Structure, Meaning and Function. Cambridge: Cambridge University Press. Vendler, Zeno (1967). ‘Verbs and times’, in Zeno Vendler, Linguistics in Philosophy. Ithaca, NY: Cornell University Press, 97–121. Voeltz, F. K. Erhard, and Kilian-Hatz, Christa (eds) (2001). Ideophones. Amsterdam: Benjamins. Vogel, Petra M. (1997). ‘Unﬂektierte Adjektive im Deutschen. Zum Verha¨ltnis von semantischer Struktur und syntaktischer Funktion und ein Vergleich mit ﬂektierten Adjektiven’, Sprachwissenschaft 22: 403–33. Walch, Maria (2005). ‘Bsufﬁg mit einem gstohligen Auto fahren: Zur Morphosyntax des Partizips II starker Verben im Lechrainer Dialekt’, in Sabine Kra¨mer-Neubert and Norbert Richard Wolf (eds), Bayerische Dialektologie. Aken der Internationalen Dialektologischen Konferenz 26–28. Feburary 2002. Heidelberg: Winter, 381–395 (Schriften zum Bayerischen Sprachatlas 8).

440

References

Watkins, Mark H. (1937). A Grammar of Chichewa: A Bantu Language of British Central Africa. Philadelphia: Linguistic Society of America and University of Pennsylvania. Westermann, Diedrich (1930). A Study of the Ewe Language. Oxford: Oxford University Press. Wilkins, David P. (1988). ‘Switch reference in Mparntwe Arrernte (Aranda): form, function, and problems of identity’, in Austin (1988: 141–76). —— (1989). Mparntwe Arrernte (Aranda): Studies in the Structure and Semantics of Grammar. Ph.D. dissertation, Australian National University. Wilkinson, Melanie (1991). Djambarrpuyngu: A Yolngu Variety of Northern Australia. Ph.D. dissertation, University of Sydney. Williams, Edwin (1980). ‘Predication’, Linguistic Inquiry 11: 203–38. —— (1983). ‘Against small clauses’, Linguistic Inquiry 14: 287–300. Wilson, Stephen (1999). Coverbs and complex predicates in Wagiman. Stanford, Calif.: CSLI. Winkler, Susanne (1997). Focus and Secondary Predication. Berlin: Mouton de Gruyter. Wipf, Elsa (1910). Die Mundart von Visperterminen im Wallis. Frauenfeld: Huber. ¨ B: Wo¨rterbuch der bairischen Mundarten in O ¨ sterreich. 1963–1999. Herausgegeben WO ¨ im Auftrag der Osterreichischen Akademie der Wissenschaften, bearb. von Viktor Dollmayr und Eberhard Kranzmayer. Vienna. Bd. i–iii. Wodarg, Anke (1995). ‘Textsammlung’, in Kutscher et al. (1995: 113–32). Wyner, Adam Z. (1994). Boolean Event Lattices and Thematic Roles in the Syntax and the Semantics of Adverbial Modiﬁcation. Ph.D. dissertation, Cornell University. Wyngaerd, Guido Vanden (2001). ‘Measuring events’, Language 77: 61–90 Xubua, Mak.ar (1937). Megruli tekstebi/Megrel’skie teksty/Textes me´gre´liens ˙ ˙ (SSRK. Mecnierebata Ak.ademiis Sakartevelos Piliali, ENIMK.I). Tbilisi: SSRK. MA Sakartvelos Pilialis gamomcemloba. Yadav, Ramamatar (1996). A Reference Grammar of Maithili. Berlin: Mouton de Gruyter. Yatsushiro, Kazuko (1999). Case Licensing and VP Structure. Ph.D. dissertation, University of Connecticut (distributed by MIT: Working Papers in Linguistics). Zavala, Roberto (2006). ‘Serial verb constructions in Olutec’, in Aikhenvald and Dixon (2006). Zehetner, Ludwig (1985). Das bairische Dialektbuch. Munich: Beck. Zhang, Niina (2000). ‘The structure of depictive and resultative constructions in Chinese’, ZAS Papers in Linguistics 22: 191–221. Zwicky, Arnold (1987). ‘Constructions in monostratal syntax’, Chicago Linguistic Society 23: 389–401.

Index of languages Alamblak 410, 421 Alemannic see German Amawaka 259, 298 Amharic 299, 319, 320 Anindilyakwa 182, 191 Anlo see Ewe Arabic Classical Arabic 319 Egyptian Arabic 299, 320 Moroccan Arabic 400, 401 Arrernte (Mparntwe Arrernte) 192–5, 197, 338 Australian languages 12, 31, 33, 35, 40, 42, 50, 54, 59, 60, 62, 63, 65, 69, 173–6, 178, 191, 194, 196, 197, 199, 256, 292, 298, 338, 349, 401, 414, 417 Bantu languages 302, 323–5, 331, 332, 340, 341, 346, 347, 353 Bari 64, 308, 309, 321 Basque 32, 415, 416 Bavarian see German Berber 64 Bininj Gun-wok see Mayali Chechen 213 Chewa 346 Chinese see Mandarin Chinese Djambarrpuyngu 181, 192, 196, 197 Dutch 11, 394, 411–13 English 2–6, 8, 10, 12, 14, 15, 18, 19, 22, 24–7, 30, 31, 34, 37, 51, 52, 54–6, 58, 61, 64, chapter 2, 109, 113–15, 117, 119, 128, 134, 136, 140, 141, 143, 146, 147, 163, 175, 191, 201, 214, 218, 256, 265, 272, 280, 344, 351, 352, 355, 371, 377, 379, 382, 390, 391, 395, 397–9, 403, 404, 406–8, 411, 412, 419, 420 Even 32, 417, 418, 420 Evenki 395

Ewe 11, 24, 32–4, 52, 57, 59, 64, 65, chapter 11, 408 Anlo 372 Kpele 365, 370 Fula 408, 409, 412, 413 Gaagudju 181, 184 Georgian 12, 14, 28, 31–5, 37, 39, 41, 42, 44, 46, 47, 53, 54, 62, 65, chapter 6, 237, 238, 240, 241 German 2–4, 8, 10, 11, 13, 14, 18, 34, 36, 38, 53–8, 109, 113, 115–17, 123, 127, 128, 131, 138, 140–3, 145, 148, 153, 155, 161–5, 171, 201, 218, 235, 255, 349, 351, 352, 377, 403–5 Alemannic 142, 143, 148, 156, 158, 165 Bavarian 142, 158, 160 Dialect of Cologne 58 Dialect of Diepoldsau 65, 159, 170 Dialects of Appenzell 28, 32, 40, 44, 45, 62, 146, 147, 153–65, 169–71 Middle High German 142, 156–9 Old High German 142, 148, 151–3, 156, 159, 171 Swabian 142 Swiss German 25, 28, 30, 32, 40, 41, 44, 45, 51, 53, 54, 60, 62, 65, chapter 4, 403, 404, 420 Germanic languages 59, 145, 403, 420 Gooniyandi 33, 173–5, 178–80, 183, 185–7, 189, 190, 192, 196, 197 Greek (Old Greek) 40, 212 Gunwinjguan languages 181 Ilokano 53, 351, 352 Indo-Aryan languages 419 Italian 32, 142 Jabirrjabirr 189 Jacaltec 350 Jaminjung 59, 63, 65, 180, 183, 187, 190, 193 Jiwarli 174, 192

442

Index of Languages

Kalenjin 305, 306 Kapanawa 259 Kartvelian languages chapter 6 Kashibo-Kakataibo 298 Kayardild 60, 182 Kija 177 Kipsikiis 305–8, 321 Kpele see Ewe Kukatja 191, 193 Kwa languages 355, 408 Lao 34, 52, 61, 64, chapter 12 Latin 14, 27, 31, 36, 40, 42, 50, 60, 232, 352, 417 Laz 34, 39, 51, 65, 202, 211, chapter 7 Maale 41, 61, 63, 309, 310, 317–19, 321 Maithili 419, 420 Mandarin Chinese 52, 405, 406, 410, 421 Mangarayi 180, 401–3, 414, 420 Martuthunira 12–14, 32, 42, 50, 54, 60, 63, 173, 179, 187, 417, 420 Matses 259, 273, 292 Mayali 181, 351 Mayan languages 350 Mingrelian 202, 211 Mparntwe Arrernte see Arrernte Ndje´bbana 181 Ngalakgan 191, 192 Ngiti 401, 402 Nguni 350 Nilotic languages 52, 55, 59, 64, chapter 9, 337 Norwegian 145, 393, 403, 404 Nunggubuyu 53, 191, 415, 416 Nyamal 182 Nyikina 180 Nyulnyul 182, 185, 192 Nyulnyulan languages 182, 189, 191

Pitjantjatjara 177, 191 Rhaeto-Romance 142 Romance languages 62, 142, 171 Russian 24, 33, 34, 54, 64, 65, 73, 88, 93, 99, 305, 399, 400, 408, 409, 420 Semitic languages 299, 319, 320 Sharanawa 273 Shipibo-Konibo 28, 31, 32, 35, 36, 38, 39, 41, 42, 44, 46, 49, 61, 63, 66, 256, 259, 261, 264, 275, 285 Shona 32, 34, 37, 41, 44, 53, 59, 61, 63, 64, 302, chapter 10 Slavic languages 62 South Caucasian languages see Kartvelian languages Svan 201–4, 206–9, 211, 212, 217, 219–26, 229, 234, 235 Swedish 145, 403, 404 Swiss German see German Tamil 393, 413, 414 Tungusic languages 32, 395, 417, 418 Turkana 32, 34, 61, 300, 302–6, 308, 309 Turkish 53, 64, 201, 237, 255, 351 Tush 39, 46, 47, 201, 202, 208, 212, 213, 214, 216, 234, 235 Tzotzil 376

Omotic languages 32, 53, 55, 59, 61, chapter 9, 337, 338

Wagiman 179 Wangkangurru 180 Warlpiri 4, 25–8, 31, 32, 37–42, 44, 46, 48, 51, 52, 54, 56, 58–60, 63, chapter 2, 174, 187, 200, 292 Warrgamay 182 Warrwa 177, 188, 191, 195 Wik Mungkan 191 Wolaitta 41, 61, 63–5, 299, 300, 309–11, 313, 315–17, 319, 338

Panoan languages 36, 38, 39, 44, 46, 50, 63, 256, chapter 8 Panyjima 194, 195

Yankunytjatjara 40, 50, 60 Yir Yoront 182, 191 Yoruba 355

Index of terms ablative (case) 38, 98, 196, 202, 212–16, 235, 264, 271–8, 280, 281, 293–8 absolutive (case) 76, 80–2, 181, 261, 263, 265, 270, 274, 275, 298, 302 active language 239 adjunct 1, 2, 4, 9, 12–16, 19–28, 30, 32–7, 39, 40, 43, 44, 46, 48, 50–67, 71–5, 78, 83, 87, 88, 90–3, 97, 98, 100, 105, 106, 108, 120, 125, 128, 131, 134–7, 140, 160, 202, 211, 216, 219, 227, 228, 231, 232, 237, 238, 240–4, 248–60, 264, 269–74, 276–8, 280–5, 287–98, 307–9, 312, 319, 323, 324, 332, 335–9, 342, 345–7, 353, 359–61, 363, 364, 367, 368, 370–8, 380, 381, 387–91 coordination of 228, 232 see also complement vs. adjunct adornment (e.g. expression of clothing) 37, 73, 74, 97, 209, 220–1, 230, 362, 385–6 adpositional complement see complement adverbial(is) case 46, 65, 202, 219–33, 238 adverbial complement see complement agreement 14, 15, 30, 32, 34–40, 42, 46, 47, 49, 50, 57, 60, 62, 63, 66, 79, 88, 92, 141, 142, 145, 146, 148, 149, 151–4, 156–8, 161, 170, 171, 184, 203, 205, 206, 209, 211, 212, 215, 218, 219, 220, 221, 223, 227, 231–5, 238, 240, 256, 259, 260, 264–7, 269, 271–98, 311–13, 318, 319, 324, 327, 333, 341, 376, 378, 380, 383, 391, 400, 401, 403–5, 420 case agreement 12, 33, 35–7, 39–41, 44, 46, 48, 60, 62, 63, 67, 69, 71, 72, 76, 80–4, 95, 98, 148, 149, 173, 178, 181, 186, 197, 202, 203, 207, 212–16, 218, 223, 225, 226, 229, 230, 235, 258, 260, 298, 415, 417, 418

gender agreement 60, 62, 88, 141, 143, 148–50, 152, 154, 155, 324, 326, 333, 334, 400–3, 415 number agreement 33, 36, 44, 60, 62, 88, 141, 143, 148, 150, 152, 154, 155, 158, 184, 203–5, 216, 226, 227, 233–5, 326, 334, 400–3, 415, 417, 418 person agreement 36, 64, 180, 181, 183, 184, 206, 238, 239, 305, 306, 308, 310, 311, 313, 317, 326, 334, 341, 346, 402, 415 allative (case) 59, 98, 259, 262, 264, 269, 271, 273, 294, 295, 298, 377 anterior clause/converb see converb apposition 90, 93, 98, 99, 206, 215, 398, 402, 404–6, 410, 417, 418, 420 argument structure 112 assertive focus see focus benefactive (construction; expression) 17, 29, 45–50, 265, 277, 278 biclausal (construction; expression) 11, 34, 51, 61, 189, 238, 256, 335, 363, 375, 377 cardinal see numeral case agreement see agreement categorical statement 304 circumstantial/conditional 3, 15–26, 34, 56, 67, 86, 88–90, 108, 131, 134, 142, 223, 227, 229, 232, 233, 265, 278, 279, 309 cleft (sentence/construction) 95, 137, 350 clitic (markers) 65, 69–71, 78–80, 90–2, 98, 100–4, 168, 180, 183, 187, 191, 193, 259, 263, 282, 296, 307, 356–8, 364, 366, 367, 369–71, 373, 375, 377 cognate object 200, 307 collective see quantiﬁcation, collective comitative (case) 3, 36–8, 62, 64, 133, 134, 291, 292, 295–7, 359, 375, 377

444

Index of terms

comparison see similative complement 24–5, 71, 84–5, 91, 92, 95, 100, 102, 105, 152, 153, 155, 157, 217, 219, 231, 257, 264, 276, 296, 299, 308, 330, 338, 342, 356, 357, 360–2, 364–6, 368, 370, 371, 373–8, 380, 388, 390, 407, 409, 421 vs. adjunct 24–5, chapter 2, 335, 361, 364, 368, 371, 373–5, 378, 380 adpositional complement 72, 87 adverbial complement 218 predicative complement/complementative 3, 24, 25, 62, 64, 105, 153, 157, 202, 204, 209, 212, 215–17, 235, 338, 364, 390, 402, 406–14, 419–21 complex predicate/predication 4, 27, 46, 50–2, 59, 71, 96, 105, 184, 185, 193, 276, 317, 332, 342, 351, 379 complex sentence 188, 189, 330, 344 compound(ing of) verb see verbal compound concession/concessive see circumstantial concomitance 27, 29, 36–8, 43–50, 221–3, 225, 227–34, 336, 411 condition(al) (attribute) see circumstantial Construction Grammar 174 controller (of secondary predicates) 1, 4, 9, 12, 14, 30, 32, 33, 35–8, 40–2, 50, 54–7, 61–4, 67, 88, 92, 105, 147, 150, 152, 153, 155–7, 160, 161, 165–71, 173, 178, 181, 183–6, 188, 190, 193–5, 197, 198, 200, 202, 204–6, 209, 211, 223, 226, 230, 234, 235, 241, 243, 252, 256–8, 266, 295, 298, 311, 365, 366, 368, 372, 376, 403 object/O 51, 54, 67, 147, 153, 160, 161, 166, 168, 195, 200, 209, 296 indirect object 54, 67, 96, 153, 160, 200 other non-subject 54, 55, 153, 160, 200 conventionalisation 51–3, 352, 384–5 converb 32, 37, 40, 60, 61, 63, 64, 67, 257, 265, 311–21, 340, 342, 343, 346, 417, 418 anterior 29, 40–9, 209, 210, 264, 267, 268, 285, 294, 295, 309, 311, 313–20, 328, 329, 343 different subject 61, 267–9, 279, 280, 295, 309–13, 316–19 free subject 61, 327, 328, 330–47

same/restricted subject 40, 41, 43, 44, 60, 61, 63, 264–9, 284, 285, 287, 293–7, 309–19, 346, 373, 377 simultaneous 29, 40–3, 45–9, 85, 165, 210, 257, 266, 267, 269, 284, 293–5, 302, 309–14, 316–20, 329, 330, 332, 339, 343, 346 subsequent 41–3, 46, 264, 266, 293–5 co-participation 35, 176, 178, 187, 188, 190, 191, 195, 197, 198, 256, 375 copular construction 34, 52, 58, 61, 64, 78, 95, 189, 202, 216, 217, 219, 256, 264, 282, 304, 335, 337, 338, 381, 386–8, 399, 401, 402, 407, 415, 420, 421 dative (case) 54, 77, 80, 90, 96, 105, 149, 160, 162, 197, 203, 206–9, 213, 215, 216, 226, 230, 302 dependency (grammar) 1, 56, 184, 188, 190, 194, 198, 231, 265, 324, 361 dependent marking 299, 302 depictive (secondary predicate) deﬁnition of 14–15, 25–7, 56–7, 65, 98, 242, 244, 324, 339, 343–6, 348, 355–6, 367, 378 detached adjunct 27, 72–3, 78, 92, 93, 216, 417 different-subject clauses see converb discontinuous noun phrase 79, 186, 199, 204, 205, 218–19, 236 discourse prominence see focus distributive see quantiﬁcation, distributive emphatic pronoun 29, 36, 43, 45, 47–50, 66, 183, 190, 197, 264, 269, 281–3, 293, 295–7 enclitic see clitic ergative (case) 33, 37, 41, 69, 76, 79–82, 95, 178, 185, 194, 202, 203, 206–8, 212, 215, 226, 259, 261–4, 270, 274, 279, 298 essive 34, 64, 220 event orientation 4, 7, 10, 12–15, 19, 26, 28, 31–4, 38–40, 43, 46, 50, 58, 65–7, 69, 109, 224–5, 234, 237, 238, 242, 244, 249–55, 258, 260, 289, 291–3, 295–7, 306–8, 312, 318, 359, 361, 368, 375, 377, 411

Index of terms event variable 10, 108–11, 113, 120, 121, 129–31 ﬁnite(ness) 21, 52, 61, 74, 84, 85, 88, 91, 92, 105, 128, 160, 185, 193, 238, 256, 257, 265–7, 311, 343, 353, 379–81, 391, 407 focus 2, 8, 15, 18, 19, 23–6, 55, 56, 65, 67, 91, 109, 114, 116, 117, 119, 120, 122, 123, 125, 127, 130–3, 218, 242, 324, 342, 347–53, 357, 358, 362, 366, 378 fractured noun phrase see discontinuous noun phrase frequency adverb(ial) 21, 22, 29, 36, 43, 45–9, 110, 111, 175–7, 191, 255 function expression see role general adjunct construction 15, 26, 28, 30, 34, 37, 43, 44, 46, 61, 65, 66, 237, 244, 248, 251, 252, 254, 257, 307, 312, 319, 338, 339, 342, 346, 347, 353, 381 deﬁnition of 15, 26, 66, 338–9 generic (sentence) 21, 86–9, 96, 100, 105, 107, 111, 113, 114, 120, 176, 226 genitive (case) 44, 62, 149, 203, 211, 215, 216, 252, 253, 259, 261, 262, 333 gerund see converb gerundive 42, 46, 210, 313 goal (role) 96–8, 240, 251, 252, 303 grammatical relation(s) 1, 54–5 67, 97, 153, 160, 170, 175, 186, 189, 198, 200, 259, 264, 296, 299, 345, 356, 379 grammaticalization 32, 66, 156–8, 170, 273, 332, 342, 346, 377, 379 habitual 21, 86, 88, 107, 110, 111, 304, 314, 351 head marking 238, 299, 302 ideophone 52, 57–9, 67, 320, 321, 323, 337, 356–8, 360, 361, 364, 367, 372, 377 indeﬁnite (noun phrase/expression) 107–10, 112–15, 117, 119, 139, 143, 176 individual-level predicate 2, 10, 20, 21, 27, 34, 57, 58, 93, 95, 101, 120 instrumental (case) 5, 24, 33, 34, 37, 44, 64, 65, 128, 131–4, 140, 194, 202, 203, 215,

445

216, 221, 225–7, 234, 238, 249–51, 302, 305, 399, 400, 408, 409 intention 29, 42, 45–50, 335 interrogative 12, 38, 91, 129, 234, 242, 244, 261, 264, 284, 329, 350, 362, 378, 384 intonation see prosody isolating (language) 52, 356, 379, 391 lexicalization 221, 282, 283, 307, 335, 370–3, 377 life-stage expression 16, 29, 30, 34–6, 43, 45–50, 64–6, 161, 211, 228, 234, 238, 256–7, 264, 269, 279–81, 290, 293, 295–7, 305 location (expression) see also place 6, 27, 29, 38–9, 45–50, 80, 82, 84, 92, 129, 136, 212, 214, 234, 235, 252–4, 269–76, 337, 347, 359 locational nominal 80, 238, 252–4 locative (case) 3, 34, 35, 38, 39, 43, 46, 64, 65, 80, 92, 98, 129, 131, 134, 136, 140, 181, 183, 185, 190, 194, 197, 215, 235, 252–4, 259, 261, 264, 265, 269–77, 292–5, 298, 302, 315, 345, 357, 360, 370, 377, 401 manner (adjunct, adverb, adverbial, expression) 3, 5–15, 17, 25, 27, 29, 31–4, 36, 38, 43, 45–50, 53, 56, 58–61, 64, 66, 69, 72, 78, 82, 83, 87, 92, 95, 105, 108, 109, 123–31, 134, 136–40, 189, 201, 205, 218, 224–34, 237, 238, 242–8, 254, 258, 264, 269, 283–5, 287, 290–7, 303–9, 318, 321, 336, 340, 347, 360, 361, 363, 368–70, 381, 382, 385, 389, 390, 411–13, 415 mass quantiﬁer 175, 176, 192, 196, 197 mental state adverb 3, 29–31, 109, 124, 127, 134, 140, 250 metonymic shift 7, 13, 14, 40, 170, 225, 228, 231, 232, 374 motative case 238, 249, 251, 252, 256, 257 negation/negator 17–19, 22, 27, 43, 53, 73, 75, 76, 88, 89, 97, 104, 147, 162, 165, 230–3, 345, 356, 362, 363, 379, 381, 387, 400, 415 nominative (case) 149, 202, 209, 212, 214, 215, 226, 229, 232, 282, 283, 298, 302, 310

446

Index of terms

nuclear scope 22–4, 112, 113, 117, 119, 120, 123, 128, 137 null subject 60, 63 numeral 35–6, 44, 173–83, 186, 187, 191–3, 195–7, 203, 212, 237, 238, 254–6, 257, 290, 291, 293–7 distributive 35, 46, 237, 238, 254–7 ordinal 36, 176, 191 ornative (e.g. expression of clothing) see adornment participant orientation 1, 3, 4, 7–15, 17, 19, 21–44, 46, 50–67, 69, 79, 82–4, 92, 189–91, 201, 202, 209, 219, 222, 223, 225–7, 232–8, 240–4, 249–58, 260, 270, 271, 279, 280, 282, 289, 291, 293, 295–7, 307, 308, 310, 312, 315–19, 323, 324, 332, 334, 336–8, 340, 342, 343, 347, 353, 356, 358, 360, 361, 363, 366–8, 371–8, 381, 385, 410, 411 participial mood 325–7, 330–2, 342, 343 participle see also converb 21, 26, 40, 41, 44, 46, 60, 67, 142, 160–5, 203, 210, 216, 218, 220, 223–5, 226, 231, 235, 242, 245, 257, 325–9, 331, 342–5, 347, 350, 394, 403, 413, 415–20 past participle 41, 44, 46, 142, 143, 147, 149–52, 154–6, 160–5, 167–70, 209, 210, 313, 403 partition of the clause 22, 23, 109, 112, 119, 121, 127, 129 part-whole syntax 81 passive 54, 89, 96, 149, 207, 302, 406 perception predicate 27, 94, 95, 97, 99, 102, 103, 338, 342, 364, 366, 372, 374, 377, 407, 408 periphrastic predicate see also complex predicate 52, 184, 185, 193, 336, 342, 346, 347, 419 person marking 36, 64, 180, 181, 183, 184, 205, 206, 238–9, 305, 306, 308, 310, 311, 313, 317, 324, 326, 334, 341, 346, 402, 415 phasal adverb 121 phase (expression) 211, 228, 234, 393, 394 place (expression) see also location 80, 98, 212, 235, 253, 273, 285, 293, 295–7 polypersonal predicate see person marking

posture (expression) 29, 31, 32, 38, 44–9, 60, 161, 209, 218, 220, 221, 315, 358, 361, 362, 377, 385, 386 pragmatic restriction/inference 19, 53, 90–2, 109, 124, 127, 147, 169, 178, 189, 242, 351–2, 384–5, 405 predicate of events 108, 110, 117, 119, 122–6, 139, 140 predicative 4, 30, 35, 56, 58, 59, 61, 80, 87, 98, 125, 141–3, 145, 146, 148–50, 152–8, 171, 183, 184, 188, 189, 193, 204, 205, 215–17, 219, 228, 256, 304, 312, 320, 321, 335, 344, 352, 357, 358, 360, 377, 381, 383, 387, 391, 398–408, 410, 412, 414, 415, 419–21 complement 3, 24, 25, 62, 64, 202, 204, 209, 212, 215–17, 235, 338, 364, 390, 398, 402, 406–5, 419–21 marker 24, 32–4, 58, 64, 308, 309, 356, 366, 367, 369, 371, 375 of becoming 83 of being 83 presupposition 18, 19, 22–4, 67, 114–16, 123, 124, 139, 386 preverb 46, 59, 71, 77, 85, 189, 206, 252, 253, 379, 388 process orientation see event orientation process-related adverbial 109, 127, 128, 136, 137, 224, 234 pronoun see emphatic pronoun property expression 31, 82, 124, 125, 188, 231, 233, 282, 303, 308, 357, 360, 363, 377, 379, 398, 411 prosody/prosodic break 5, 19, 21, 22, 27, 70, 76, 81, 88, 100, 124, 136, 173, 184, 185, 193, 194, 199, 216, 217, 235, 238, 240, 244–6, 248, 254, 258, 323–9, 342, 344, 345, 347–9, 405 purpos(iv)e 29, 42, 45–9, 223, 249, 250, 265, 335 quantiﬁcation 3, 35, 36, 50, 60, 65, 107, 109–11, 112, 114, 116, 118, 121, 125, 127, 129, 137–40, 173–8, 182, 183, 186, 188, 190, 191, 193–200, 218, 225, 254, 255, 290, 291, 359, 360, 372, 377 collective 29, 35, 45–9, 227, 254, 261, 374 distributive 35, 46, 237, 238, 254–7

Index of terms expression of quantity 27, 35, 43, 174–6, 181, 184, 190, 196, 211, 225, 228, 234, 264, 269, 283, 290, 291, 296, 347, 359 quantiﬁcational adverb 3, 35, 58, 109–11, 114, 118, 121, 125, 137, 139, chapter 5 quantiﬁer ﬂoating 194, 218, 219 recipient (role) 34, 96, 97, 195 reduplication 35, 191–5, 255–7, 317–19, 356, 357, 379 relative clause 165, 303, 326–8, 330, 332, 335, 336, 340, 342, 346, 394, 405, 409, 418 relative tense see taxis restricted subject reference (see also converb) 44, 60, 62, 267, 327, 329, 342 restrictive (clitic; marker) 65, 66, 154, 158–60, 170, 180, 183, 352 restrictor 19, 22–4, 107–9, 112, 113, 115, 119, 122, 125, 128, 130, 134, 137, 139, 140 resultant state see converb, anterior resultative (attribute, secondary predicate) 4, 6, 15–17, 41, 44, 71, 83–6, 91–3, 95, 103–6, 130, 149–51, 155, 209, 210, 304, 306, 343, 348, 349, 351, 381, 382, 387–9, 391, 420, 421 role expression 29, 30, 33, 34, 36, 42, 43, 45–9, 51, 64, 161, 238, 256, 305, 390, 400 same-subject clauses see converb scrambling 113–19, 124, 131, 132, 139 Semiotic Grammar 174 similative/similitive (comparison) 27, 29, 33, 34, 45–9, 62, 176, 201, 202, 226, 233–5, 265, 286–7, 289, 307, 323, 358, 360, 377 simultaneous clause/converb see converb split noun phrase see discontinuous noun phrase stage-level predicate 2, 20, 21, 27, 34, 57, 58, 93, 95, 100, 101 state 3, 4, 9, 10, 20, 25, 28–31, 33, 34, 40–2, 45, 47–9, 51, 67, 73, 78–80, 84, 85, 88, 90, 92, 93, 97, 99–102, 104, 105, 109, 110, 121, 124, 127,

447

134, 140, 147, 149, 151, 160, 161, 163, 167–9, 174, 196, 201, 209, 210, 220, 223, 225, 227–35, 237, 242, 243, 250, 276, 284, 302–4 306–8, 312, 315, 316, 318, 321, 328, 334–6, 341, 343, 358, 361, 362, 366, 367, 375, 377, 382, 386, 388–91, 393, 395 stative predicate 20, 97, 99–102, 105, 111, 303, 308, 335, 340, 383, 386, 387 strong adjunct 3, 20–2, 25, 27, 72, 73, 92, 108, 120, 128 subcategorisation 105 subject-oriented adverbs/subject orientation 7, 40, 56, 86, 89, 124–8, 130, 138, 225, 336, 337 subordination 40, 41, 51, 57, 61, 63, 64, 67, 151, 256, 257, 268, 271, 302, 309, 323–5, 330, 332, 333, 340, 344–6, 360, 382, 386, 390, 402, 407, 409, 414 asyndetic 323, 324, 332, 346 prosodic 323, 324, 353 subsequent clause see converb, subsequent switch-reference 40, 43, 44, 61, 63, 259, 265, 267, 313, 316, 317 syntactic function see grammatical relation(s) taxis 41, 42, 318, 324, 328, 330–2, 339, 340, 342, 343, 346, 350, 353, 409 temporal (adjunct, adverbial, marker) 15–18, 23, 35, 39, 40, 43, 46, 67, 71, 78, 90–2, 95, 98, 100–4, 121, 122, 128, 134–6, 139, 187, 198, 257, 261, 262, 265, 268, 279, 280, 291, 292, 295–7, 317, 360, 402, 417, 418 thetic statement 304 topic/topicalization 54, 108, 114, 115, 117, 119, 121, 124, 127, 132, 133, 138, 218, 243, 344, 349, 385, 387 translative (case) 84, 93 variation 25, 31–3, 35, 44, 46, 55, 66, 67, 83, 108, 113, 115–17, 125, 131, 148, 181, 203, 226, 229–33, 244, 300, 350, 353 intragenetic 32, 203, 229–30, 291, 300, 319, 370–2

448

Index of terms

variation (cont.) intralinguistic 31–2, 35, 54–5, 143–6, 229, 370 verbal compound 84, 189, 314, 380, 386 weak adjunct 3, 20–3, 25, 26, 72, 73, 108 word order 54, 56, 71, 72, 91, 92, 100, 104, 107, 202, 236, 244, 319–20, 352, 356, 382, 383

order of adverbial 11, 77, 86, 87, 107, 128, 131, 132, 136, 138, 140, 205, 242, 352, 386 verb-ﬁnal language 265, 299, 300, 319, 320 verb-initial language 299, 300, 302, 305, 319, 320