Genders and Classifiers : A Cross-Linguistic Typology [1 ed.] 9780198842019

Table of contents :
Cover
Genders and Classifiers A Cross-Linguistic Typology Edited by Alexandra Y. Aikhenvald and Elena I. Mihas Language and Culture Research Centre James Cook University
Copyright
Contents
Preface
Notes on the contributors
Abbreviations
Chapter 1: Noun categorization devices
1 Noun categorization devices in their various guises
2 Several noun categorization devices in one language
3 How noun categorization devices differ and what they have in common
3.1 Interaction with other grammatical categories
3.2 Differences in use depending on contexts
3.3 Special semantic properties
3.4 Using more than one noun categorization device
3.5 Principles of assignment
3.6 Semantic parameters in noun categorization devices
4 Multiple classifier languages
5 The utility of noun categorization devices
6 The origins and development of noun categorization devices
7 About this volume
References
Chapter 2: Genders and classifiers in Kampa (Arawak) languages of Peru
1 Introduction
2 Relevant aspects of Kampa grammar
2.1 Linguistic profile
2.2 Kampa noun classes
2.3 Kampa N 1 +N 2 compounds and noun incorporation
2.3.1 N 1 +N 2 compounds
2.3.2 Noun incorporation
3 Gender system
3.1 Gender marking
3.2 Coding of animacy
3.3 Semantic bases of gender assignment
4 Core Kampa multiple classifier set
4.1 Overview
4.2 Frequency of classifier use
4.3 Semantics of core Kampa classifiers
4.4 Functions of Kampa classifiers
5 Multiple classifiers as an areal feature
6 Conclusions
Acknowledgements
References
Chapter 3: Classifiers in Shiwilu (Kawapanan)
1 Introduction
1.1 The Shiwilu people and their language
1.2 Selected features of Shiwilu grammar
1.3 Goal and organization of this chapter
2 Overview of the Shiwilu nominal classifying system
3 Morphosyntactic contexts and text frequency of Shiwilu classifiers
3.1 Morphosyntactic contexts of Shiwilu classifiers
3.2 Text frequency of Shiwilu classifiers
3.3 Functions of classifiers in the corpus
3.3.1 Derivational function
3.3.2 Concord-like function
3.3.3 Discourse functions
3.3.4 Adjective-like function
3.3.5 Disambiguating function
4 Classifier stacking and reduplication
4.1 Classifier stacking
4.2 Classifier reduplication
5 Classifier incorporation and argument reference
5.1 Incorporation of one classifier
5.2 Incorporation of two classifiers or a classifier and a noun
6 Conclusions
Acknowledgements
References
Chapter 4: A view from the North
1 The problem
2 Genders and classifiers in Arawak languages of north-west Amazonia: An illustration
3 Arawak languages and their grammatical features
4 Genders and classifiers in the Arawak languages of north-west Amazonia
4.1 Languages of set I: genders only
4.2 Languages of sets II, III, and IV
4.3 The resilience of genders and the emergence of classifiers
5 Where do classifiers come from?
6 What are classifiers good for?
7 To conclude
1 Set II languages: genders and with numeral classifiers
2 Languages of Set III: genders, and classifiers with number words and on some nouns
3 Gender distinctions with classifiers in multiple contexts
Acknowledgements
References
Chapter 5: Possessive classifiers in Zamucoan
1 Introduction
2 The Zamucoan family
3 The Zamucoan gender system
4 NP-internal possessive constructions: General structure
5 Ayoreo possessive classifiers
5.1 The inventory of classifiers
5.2 Innovations
6 Chamacoco possessive classifiers
7 Old Zamuco possessive classifiers
7.1 Old Zamuco documentation
7.2 Constructions with classifiers in Old Zamuco
8 Possessive classifiers in Zamucoan and beyond
9 Conclusions
Acknowledgements
References
Chapter 6: The elusive verbal classifiers in ‘Witoto’
1 Introduction
2 ‘Witoto’ typological profile
3 ‘Witoto’ multiple classifier system
4 Peculiar verbal formatives in Murui
5 Verbal classifiers in Mɨka
6 Summary
References
Chapter 7: Multifunctionality of deictic classifiers in the Toba language (Guaycuruan)
1 Introduction
1.1 The Toba language: General characteristics and typological profile
1.2 Data sources and methodology
2 Deictic classifiers in Toba
2.1 Semantic and morphosyntactic characteristics
2.2 The classifying function of deictic classifiers
2.3 Variability in classifier choice and reclassification
2.4 The function of deictic classifiers in signalling time, mood, and evidentiality
3 Functions of deictic classifiers in discourse
3.1 Deictic classifiers and information structure
3.2 Deictic classifiers and text genres in Toba
3.2.1 The trickster story
3.2.2 Advice ( nqataGak)
3.2.3 Prayers ( natamnaGak)
4 Conclusions
References
Chapter 8: Classifiers in Hmong
1 Introduction
1.1 Language background
1.2 Orthography and representation of tone
1.3 Wordhood
1.4 Structure of the noun phrase
1.5 What constitutes a classifier in Hmong?
1.6 Sources of current study
2 Classifiers in Hmong: The literature
2.1 When classifiers are mandatory
2.2 Subcategories of classifier
2.3 Syntactic differences between sortal and mensural classifiers
2.4 Alternations of form with cov
2.5 Discourse properties
2.6 Verbal modification of an action
2.7 Class nouns
3 Classifiers in Hmong: Findings from the data
3.1 Subcategories
3.2 Open/closed class status
3.3 Syntax
3.3.1 The apparent co-occurrence of a classifier with cov or tej
3.3.2 Gerner’s (2015) verbal classifier types as noun phrases
3.3.3 Quantifier + classifier to emphasize or indicate a group
3.3.4 Apportioning construction
3.3.5 Use of classifie
3.3.6 Classifiers with elaborate expressions and co-compounds
3.3.7 Omission of a classifier: the full range
3.3.8 Classifiers, quantifiers, and wordhood
3.4 Discourse
4 Conclusion and further research
Acknowledgements
References
Chapter 9: Numeral classifiers in Japanese
1 Introduction
1.1 Introduction to the Japanese language
2 Introduction to numerals and numeral classifiers in Japanese
2.1 Numerals
2.2 Numeral classifiers and other types of counters in counting expressions
2.3 Lack of clear distinctions between types
2.4 Obligatoriness and substitution of numeral classifiers
2.5 Inventory of numeral classifiers
3 Semantics of Japanese numeral classifiers
4 Origins of Japanese numeral classifiers
4.1 Historical origins and current developments
4.2 Lexical sources for numeral classifiers
5 Functions of Japanese numeral classifiers
5.1 Discourse pragmatic functions
5.1.1 Grammatical constructions in which numeral classifiers appear
5.1.2 Meanings and uses of the constructions
5.2 Numeral classifiers refining or distinguishing the meanings of nouns
5.3 Idiosyncratic uses of numeral classifiers: conveying metaphorical meaning
5.3.1 Standard and idiosyncratic uses
6 Conclusion
Acknowledgements
References
Chapter 10: Numeral classifiers in Munya, a Tibeto-Burman language
1 Introduction: The people and the language
2 Number words and classifiers in Munya
2.1 Two sets of number words
2.2 Classifiers
2.2.1 Sortal classifiers
2.2.2 Mensural classifiers
2.3 Pseudo classifiers
3 The functions of numeral classifiers
3.1 Discourse functions
3.2 Stacking number words to express an approximate meaning
3.3 Quantifiers
3.4 Adverbs
3.5 Nominal and adverbial manner demonstratives
3.6 Complementizing strategy
4 Verbal action classifiers
5 Summary
Acknowledgements
References
Index of authors
Index of languages, language families, and linguistic areas
Index of subjects

Citation preview

Genders and Classifiers

E X P L O R AT IO N S I N L I N G U I S T IC T Y P O L O G Y general editors: Alexandra Y. Aikhenvald and R. M. W. Dixon Language and Culture Research Centre, James Cook University This series focuses on aspects of language that are of current theoretical interest and for which there has not previously or recently been any full-scale cross-linguistic study. Its books are for typologists, fieldworkers, and theory developers, and designed for use in advanced seminars and courses. published 1 Adjective Classes edited by R. M. W. Dixon and Alexandra Y. Aikhenvald 2 Serial Verb Constructions edited by Alexandra Y. Aikhenvald and R. M. W. Dixon 3 Complementation edited by R. M. W. Dixon and Alexandra Y. Aikhenvald 4 Grammars in Contact edited by Alexandra Y. Aikhenvald and R. M. W. Dixon 5 The Semantics of Clause Linking edited by R. M. W. Dixon and Alexandra Y. Aikhenvald 6 Possession and Ownership edited by Alexandra Y. Aikhenvald and R. M. W. Dixon 7 The Grammar of Knowledge edited by Alexandra Y. Aikhenvald and R. M. W. Dixon 8 Commands edited by Alexandra Y. Aikhenvald and R. M. W. Dixon 9 Genders and Classifiers edited by Alexandra Y. Aikhenvald and Elena I. Mihas published in association with the series Areal Diffusion and Genetic Inheritance Problems in Comparative Linguistics edited by Alexandra Y. Aikhenvald and R. M. W. Dixon

Genders and Classifiers A Cross-Linguistic Typology

Edited by A L E X A N DR A   Y.   A I K H E N VA L D and E L E NA   I .   M I H A S

Language and Culture Research Centre James Cook University

1

1 Great Clarendon Street, Oxford, ox2 6dp, United Kingdom 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. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries © editorial matter and organization Alexandra Y. Aikhenvald and Elena I. Mihas 2019 © the chapters their several authors 2019 The moral rights of the authors have been asserted First Edition published in 2019 Impression: 1 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, by licence 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 work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2018965638 ISBN 978–0–19–884201–9 Printed in Great Britain by Clays Ltd, St Ives plc Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.

Contents Preface  viii Notes on the contributors  x Abbreviations  xiv 1 Noun categorization devices: A cross-linguistic perspective  1 Alexandra Y. Aikhenvald 1 Noun categorization devices in their various guises  1 2 Several noun categorization devices in one language  6 3 How noun categorization devices differ and what they have in common  9 4 Multiple classifier languages  17 5 The utility of noun categorization devices  17 6 The origins and development of noun categorization devices  21 7 About this volume  23 References 24 2 Genders and classifiers in Kampa (Arawak) languages of Peru  30 Elena I. Mihas 1 Introduction  30 2 Relevant aspects of Kampa grammar  32 3 Gender system  45 4 Core Kampa multiple classifier set  49 5 Multiple classifiers as an areal feature  57 6 Conclusions  58 Appendix: Core Kampa multiple classifiers  59 References 63 3 Classifiers in Shiwilu (Kawapanan): Exploring typologically salient properties  67 Pilar M. Valenzuela 1 Introduction  67 2 Overview of the Shiwilu nominal classifying system  72 3 Morphosyntactic contexts and text frequency of Shiwilu classifiers  78 4 Classifier stacking and reduplication  92 5 Classifier incorporation and argument reference  95 6 Conclusions  99 References 101

vi Contents 4 A view from the North: Genders and classifiers in Arawak languages of northwest Amazonia  103 Alexandra Y. Aikhenvald

1 The problem  103 2 Genders and classifiers in Arawak languages of north-west Amazonia: An illustration 104 3 Arawak languages and their grammatical features  106 4 Genders and classifiers in the Arawak languages of north-west Amazonia 110 5 Where do classifiers come from?  117 6 What are classifiers good for?  120 7 To conclude  123 Appendix 1: The Arawak languages of north-west Amazonia  124 Appendix 2: Data sets – classifier systems in the Arawak languages of north-west Amazonia  129 References 138 5 Possessive classifiers in Zamucoan  144 Luca Ciucci and Pier Marco Bertinetto 1 Introduction  144 2 The Zamucoan family  144 3 The Zamucoan gender system  146 4 NP-internal possessive constructions: General structure  149 5 Ayoreo possessive classifiers  152 6 Chamacoco possessive classifiers  158 7 Old Zamuco possessive classifiers  160 8 Possessive classifiers in Zamucoan and beyond  166 9 Conclusions  172 References 173 6 The elusive verbal classifiers in ‘Witoto’  176 Katarzyna I. Wojtylak 1 Introduction  176 2 ‘Witoto’ typological profile  178 3 ‘Witoto’ multiple classifier system  182 4 Peculiar verbal formatives in Murui  185 5 Verbal classifiers in Mɨka 189 6 Summary  193 References 195



Contents

vii

7 Multifunctionality of deictic classifiers in the Toba language (Guaycuruan)  197 Cristina Messineo and Paola Cúneo 1 Introduction  197 2 Deictic classifiers in Toba  199 3 Functions of deictic classifiers in discourse  209 4 Conclusions  216 References 218 8 Classifiers in Hmong  222 Nathan M. White 1 Introduction  222 2 Classifiers in Hmong: The literature  226 3 Classifiers in Hmong: Findings from the data  231 4 Conclusion and further research  246 References 247 9 Numeral classifiers in Japanese  249 Nerida Jarkey and Hiroko Komatsu 1 Introduction  249 2 Introduction to numerals and numeral classifiers in Japanese  252 3 Semantics of Japanese numeral classifiers  260 4 Origins of Japanese numeral classifiers  262 5 Functions of Japanese numeral classifiers  264 6 Conclusion  279 References 280 10 Numeral classifiers in Munya, a Tibeto-Burman language  282 Junwei Bai 1 Introduction: The people and the language  282 2 Number words and classifiers in Munya  283 3 The functions of numeral classifiers  289 4 Verbal action classifiers  295 5 Summary  297 References 298 Index of authors 299 Index of languages, language families, and linguistic areas 303 Index of subjects 306

Preface Every language has some means of categorizing objects into humans, or animates, or by their shape, form, size, and function. The most wide-spread are linguistic genders— grammatical classes of nouns based on core semantic properties such as sex (female and male), animacy, humanness, and also shape and size. Classifiers of several types also serve to categorize entities. Numeral classifiers occur with number words, possessive classifiers appear in the expressions of possession, and verbal classifiers are used on a verb, categorizing its argument. Genders and classifiers of varied types can occur together. Their meanings reflect beliefs and traditions, and in many respects mirror the ways in which speakers view the ever-changing reality. This volume elaborates on the expression, usage, history, and meanings of noun categorization devices of different kinds, exploring their various facets across the languages of South America and Asia, known for the diversity of their noun categorization. The volume starts with a typological introduction outlining the types of noun categorization devices, their expression, scope, and functions, in addition to the sociocultural aspects of their use, and their development. It is followed by revised versions of papers originally presented at the International Workshop ‘Genders and classifiers in Amazonia and beyond’ organized by the editors and held at the Language and Culture Research Centre, James Cook University, 9–10 August 2017. An earlier version of Chapter 1 was circulated to the contributors, with a list of issues to be addressed, so as to ensure that the studies of individual languages within this volume were cast in terms of a common set of parameters. This is the ninth monograph in the series Explorations in Linguistic Typology, devoted to volumes from International Workshops organized by the Language and Culture Research Centre and its predecessors. The Workshop, and subsequent discussions between the editors and the authors, were intellectually stimulating, with cross-fertilization of ideas and scholarly debates. Each author has undertaken intensive fieldwork and has firsthand in-depth ­knowledge of their languages, in addition to experience of working on linguistic typology, ­historical and comparative linguistics, and language contact and areal diffusion. The analysis is uniformly cast in terms of basic linguistic theory—the cumulative typological framework which provides the foundation for sound empirically-based descriptive and analytic works. We avoid formalisms (which provide reinterpretations rather than explanations, and come and go with such frequency that any statement made in terms of them is likely to soon become inaccessible). It is our hope that this volume will further contribute to a consolidated conceptual and analytic framework, primarily established in the previous work by Aikhenvald. Our aim is to cover, and explain, the newly established parameters of variation in



Preface

ix

noun categorization devices in a synchronic and in a diachronic perspective, opening new perspectives on classifiers of different kinds. We are grateful to all the participants in the Workshop and colleagues who took part in the discussion, providing feedback on presentations at various stages, particularly R. M. W. Dixon, Pema Wangdi, and Sihong Zhang. We owe a special debt of gratitude to Jolene Overall and Brigitta Flick, for helping us organize the Workshop in a most efficient manner. Brigitta and Jolene’s support and editorial assistance in preparing the volume were invaluable. The Workshop was made possible through the Australian Research Council Laureate Fellowship Project ‘How gender shapes the world: a linguistic perspective’ (to Alexandra Aikhenvald). We gratefully acknowledge the financial assistance from the Division of Research and Innovation at James Cook University.

Notes on the contributors Alexandra  Y.  Aikhenvald  is Distinguished Professor, Australian Laureate Fellow, and Director of the Language and Culture Research Centre at James Cook University. She is a major authority on languages of the Arawak family, from northern Amazonia, and has written grammars of Bare (1995) and Warekena (1998), plus A Grammar of Tariana, from Northwest Amazonia (CUP, 2003), and The Manambu language of East Sepik, Papua New Guinea (OUP, 2008) in a­ ddition to essays on various typological and areal topics. She is the editor of The Oxford Handbook of Evidentiality (OUP, 2018) and co-editor, with R. M. W. Dixon, of The Cambridge Handbook of Linguistic Typology (CUP, 2017). Her other major publications include Classifiers: a typology of noun categorization devices (OUP, 2000), Evidentiality (OUP, 2004), Imperatives and Commands (OUP, 2010), Languages of the Amazon (OUP, 2012), The Art of Grammar (OUP, 2014), How Gender Shapes the World (OUP, 2016), and Serial Verbs (OUP, 2018). Address: Language and Culture Research Centre, James Cook University, PO Box 6811, Cairns, N Qld 4780, Australia; e-mail: [email protected] Junwei Bai  is a PhD student of the Language and Culture Research Centre at James Cook University. He was awarded his Master’s degree from Nanjing University, with a thesis on serial verb constructions in Mandarin Chinese. His major research interests are Tibeto-Burman languages, Chinese grammar, Cognitive Linguistics, and language typology. He is currently working on a comprehensive grammar of Munya, a little-known Tibeto-Burman language spoken in western Sichuan Province in China. Address: Language and Culture Research Centre, James Cook University, PO Box 6811, Cairns, N Qld 4780, Australia; e-mail: junwei.bai@my. jcu.edu.au Pier Marco Bertinetto  completed his studies in Torino. He taught History of the Italian Language at the University of Torino (1975–9) and subsequently General Linguistics, first at the University of Roma I (1980) and then at Scuola Normale Superiore (from 1981 to his retirement in 2017), where he directed the Laboratorio di Linguistica ‘G. Nencioni’. He is editor-in-chief of the Italian Journal of Linguistics (since 1989). He is member of Academia Europaea, Suomalainen Tiedeakatemia (Finnish Academy of Sciences) and Österreichische Akademie der Wissenschaften. In 2009 he was president of Societas Linguistica Europaea.   He has supervised more than 35 PhD dissertations and organized more than 20 ­international conferences. His main research areas include experimental phonetics and phonology, experimental morphology, tense-aspect semantics, typological and descriptive linguistics. A list of his publications can be found at: http://linguistica.sns.it/PaginePersonali/Bertinetto.htm Address: via Matteotti 197, I-55049 Viareggio, Lucca, Italy; e-mail: [email protected], [email protected] Luca Ciucci  is Postdoctoral Research Fellow at the Language and Culture Research Centre at James Cook University. In 2007 he began his research on the Zamucoan languages (Ayoreo,



Notes on the contributors

xi

Chamacoco, and †Old Zamuco). During his PhD years, he discovered the earliest grammar of Ecuadorian Quechua, published in 2011. In 2013 he completed his PhD at Scuola Normale Superiore in Pisa. His monograph Inflectional morphology in the Zamucoan languages (CEADUC, 2016) is considered the most detailed morphological description of a small language cluster from South America. His research activities include the grammatical description of Ayoreo and Chamacoco, the reconstruction of Proto-Zamucoan, and the analysis of the historical data available for †Old Zamuco. In 2017 he began documenting Chiquitano (aka Bésɨro), an isolate. He is particularly interested in language contact between Chiquitano and the surrounding languages (such as Zamucoan), and in the comparison between the different Chiquitano varieties spoken nowadays plus the historical documents available on this language. Address: Language and Culture Research Centre, James Cook University, PO Box 6811, Cairns, N Qld 4780, Australia; e-mail: [email protected] Paola Cúneo is Research Associate at the National Scientific and Technical Research Council in Argentina and a teaching assistant of Linguistics at the University of Buenos Aires. Her research has focused on a descriptive and typological approach to word-formation and nominal classification in the Toba language (Guaycuruan, Gran Chaco, Argentina), with special reference to ethnobiological vocabulary. She has published Formación de palabras y clasificación nominal en el léxico etnobiológico en toba (guaycurú) (2013) and several articles on the language in i­nternational journals. She was involved in interdisciplinary projects related to indigenous peoples and languages from Argentina, funded by the University of Buenos Aires and the National Agency for Scientific Promotion and Research of Argentina (ANPCyT). She has also been carrying out academic and community ­collaborative labour for more than ten years with the Toba people. Address: Instituto de Lingüística, Facultad de Filosofía y Letras, Universidad de Buenos Aires, 25 de Mayo 217/221, 1°, (C1002ABE) Ciudad Autónoma de Buenos Aires, República Argentina; e-mail: [email protected] Nerida Jarkey  is Senior Lecturer in Japanese Studies at The University of Sydney. She has a particular interest in the semantics of grammar, with a focus on multi-verb constructions and on transitivity in Japanese and White Hmong. She also works on the expression of subjectivity and identity in Japanese. Nerida is the author of Serial Verbs in White Hmong (Brill’s Studies in Language, Cognition and Culture, 2015) and is Editor for Language and Socio-Linguistics for the journal Japanese Studies. Address: School of Languages and Cultures (A18), University of Sydney, NSW, 2006, Australia; e-mail: [email protected] Hiroko Komatsu  submitted her PhD thesis in Japanese Studies in the School of Languages and Cultures at The University of Sydney in 2018. As part of her wider interests in numeral classifiers in Japanese, Hiroko’s thesis examines the meaning of Japanese numeral classifiers, with a particular focus on the ways in which the range of referents of a single numeral ­classifier can be extended. The research for her contribution to this chapter was conducted in her capacity as a doctoral candidate at The University of Sydney. Hiroko’s current research is on the lexical sources of numeral classifiers. She works as a lecturer in the Center for Language Education and Research at Sophia University. Address: Center for Language Education and Research, 7–1 Kioi-cho, Chiyoda-ku, Tokyo 102–8554, Japan; e-mail: [email protected]

xii

Notes on the contributors

Cristina Messineo  is a Senior Researcher at the National Scientific and Technical Research Council in Argentina and an Associate Professor of Linguistics at the University of Buenos Aires. She has carried out research on indigenous languages of the Gran Chaco region since 1987, with studies on the Wichí, Toba (Qom), and Maká languages. Her interests include documentary and descriptive linguistics, syntactic complexity, typological studies, and verbal art. She has published several books and articles on languages of the Gran Chaco, such as: Arte Verbal Qom: consejos, rogativas y relatos (2014), Léxico, clasificación nominal y categorización etnobiológica en el Gran Chaco (2010), Lengua toba (guaycurú). Aspectos gramaticales y discursivos (2003), among others. She has also received John Simon Guggenheim Memorial Foundation Grant (2004) for a project of documentation of the Toba speaking genres and the Ken Hale Prize from SSILA (2004) for her collaborative revival project of language and culture in the Toba community of Derqui (Buenos Aires). Also, she has led several research projects funded by the University of Buenos Aires, CONICET, and the National Agency for Scientific Promotion and Research of Argentina (ANPCyT). Address: Instituto de Lingüística, Facultad de Filosofía y Letras, Universidad de Buenos Aires, 25 de Mayo 217/221, 1°, (C1002ABE) Ciudad Autónoma de Buenos Aires, República Argentina; e-mail: [email protected] Elena i. Mihas  (PhD 2010, University of Wisconsin, Milwaukee) has been studying Ashéninka and Ashaninka varieties of Kampa Arawak since 2008. She is the author of three monographs, Upper Perené Narratives of History, Landscape and Ritual (Nebraska University Press, 2014), A Grammar of Alto Perené (Arawak) (De Gruyter Mouton, 2015), and Conversational structures of Alto Perené (John Benjamins, 2017), and numerous papers on the grammatical aspects of Kampa languages. She is currently a Research Scholar at the Center for Latin American and Caribbean Studies, University of Wisconsin (Milwaukee). Address: College of Letters & Science Center for Latin American & Caribbean Studies, UW Milwaukee, USA; e-mail: [email protected] Pilar Valenzuela  (PhD 2003, University of Oregon) is Professor at Chapman University in Southern California. Her research focuses on the study and documentation of Amazonian languages, and the issues that these raise for linguistic theory, typology, and areality. Most of her publications deal with synchronic and diachronic morphosyntax, and language contact. Recent articles discuss transitivity harmony in the Panoan and Takanan families (Amerindia 2017, 39: 397–438), classifiers in Shiwilu in Northwestern Amazonian perspective (Anthropological Linguistics 2016, 58, 4: 333–80), the Shiwilu applicative system (Studies in Language 2016, 40, 3: 513–50), and Andean features in Western Amazonian languages as evidence for an intermediate linguistic subarea (Lexis 2015, 39, 1: 5–56) and in her recent paper ‘Difusión de rasgos andinos y elementos para una sub-área lingüística intermedia Andes-Amazonia en el norte del Perú’, in Alejandra Regúnaga, Silvia Spinelli, and María Emilia Orden (comps.), IV Encuentro Internacional de Lenguas Indígenas Americanas. Libro de Actas. Santa Rosa, Argentina: Universidad Nacional de la Pampa, 657–84. Address: Wilkinson College of Arts, Humanities, and Social Sciences; Department of World Languages and Cultures, Chapman University, Orange Campus, One University Drive Orange, CA 92866; e-mail: [email protected] Nathan M. White  is a current PhD student of the Language and Culture Research Centre at James Cook University. He studied Intercultural Studies at Biola University and completed an MA in Linguistics at Trinity Western University in 2014. He has taught courses in Linguistics



Notes on the contributors

xiii

at Fresno Pacific University and College of the Sequoias. His research interests include language typology, language documentation and revitalization, phonology of tone languages, East and Southeast Asian languages, Semitic languages, and indigenous languages of California. The topic of his PhD thesis is ‘A comprehensive reference grammar of Hmong’. Address: Language and Culture Research Centre, James Cook University, PO Box 6811, Cairns, N Qld 4780, Australia; e-mail: [email protected] Katarzyna (Kasia) I.  Wojtylak  received her PhD in December 2017 from James Cook University, with a thesis awarded summa cum laude for a reference grammar of Murui (Bue), a Witotoan language spoken in Northwest Amazonia. She is currently Adjunct Research Fellow at the Language and Culture Research Centre. Her research interests cover Amazonian languages. She has a strong background in linguistic description, language documentation, linguistic typology, and ethnographic research. Since 2010, she has been working with various indigenous groups from Northwest Amazonia (Murui, Mɨka, Mɨnɨka), focusing on various aspects of their language and culture (such as literature, language contact, and social ­organization). Address: Language and Culture Research Centre, James Cook University, PO Box 6811, Cairns, N Qld 4780, Australia; e-mail: [email protected]

Abbreviations 1

first person

1d one-dimensional 1pl.e

first person plural exclusive

1pl.i

first person plural inclusive

1sg.i

first person singular inclusive (also impersonal passive)

2

second person

2d two-dimensional 3

third person

3d three-dimensional A

subject of transitive verb; active pronominal marker

abl ablative abs absent absl absolutive abst abstract acc accusative add additive adj adjective adj.cl

adjectival classifier

adjz adjectivizer adp adposition adv adverb(ial) advl adverbializer affect affective agt

agent, agentive case

am

adverbial marker

an animate ana anaphoric andtv andative anim animate appl applicative approx approximative

Apu

Abbreviations Apurucayali Ajyninka

art article Asha Ashaninka Ashe Ashéninka assist assistive at postposition dɨne in Murui att

attainment marker

attr attributive aux auxiliary become1–3

‘become’ markers

ben benefactive bf

base form

Ca Caquinte caus causative cav

cavity (classifier)

cc

copula complement

cert certainty cl classifier cl.day

numeral classifier for day

cl.drop

numeral classifier for drop

cl.g

general classifier

cl.gen

generic classifier

cl.hit

verbal action classifier for hitting

cl.month

numeral classifier for months

cl.perform

numeral classifier for performance

cl.place

numeral classifier for place

cl.rep classifier-repeater cl.year

numeral classifier for years

cm

change of state marker

cn

class noun

cocl

complement clause

com command comit comitative comp complementizer compl completive

xv

xvi Abbreviations cond conditional conj conjunction conn connective cont continuative contrst contrast cop copula cp

completive marker

cs

copula subject

cts

close to speaker

D possessed dat

dative case

dc

deictic classifier

dclf

deictic classifier

def definite del delimitative dem demonstrative dem:inan

inanimate demonstrative

dem.id

demonstrative identifier

deprec deprecatory des desiderative dim diminutive dir directional dir.as

away-from-speaker direction

dir.down

downward direction

dir.ds

downstream direction

dir.nons

nonspecific direction

dir.up

upward direction

dist distal distr distributive dm

discourse marker

dp

discourse particle

dr ‘derivational’ dur durative e

event nominalizer

ego egophoric



Abbreviations

emph emphatic ep epenthetic erg

ergative case

ev evidential evi.dir

direct evidential marker

evi.vis

evidential marker, visual

evid evidential ex existential exc, excl exclusive exh exhaustive exist existential exp

experiential case

f, f feminine fem, fem feminine ff

full form

foc focus fp

feminine plural

frust frustrative fs

feminine singular

fut future g

generic (classifier)

gen genitive gener

general (classifier)

ger gerund gf

generic form

gnrl general gr group hab habitual honor honorific hor, hor horizontal hsy hearsay hum human icpl incompletive if

indeterminate form

imp imperative

xvii

xviii Abbreviations imper imperative impers impersonal impf imperfective inan inanimate incl inclusive incp inceptive infin infinitive ins instrumental inst instrumental int intent inter interrogative interj interjection intj interjection intns intensifier intr intransitive ip

illocutionary particle

ipfv imperfective irr irrealis kin

kinship (plural)

lim limitative lk linker loc locative locl localizer log logophoric m, m masculine m

middle pronominal marker

Ma Matsigenka masc, masc masculine mod modality mp

masculine plural

ms

masculine singular

n.cl

noun class marker

n.contr non-contrastive n.s/a.top

topical non-S/A subject

Na Nanti



Abbreviations

neg negative neg.exist

negative existential

neuter

neuter gender (no abbreviation)

neut

neutral (classifier)

nf non-feminine nfem non-feminine nfi

non-future indicative

nhum non-human NJ

Native Japanese

NK

Northern Kampa

nm non-masculine nmlz nominalization nmz nominalizer No Nomatsiguenga nom nominative nonfem non-feminine nouncl

noun classifier

np

noun phrase

npast non-past nsg nonsingular nsp non-specific num.cl

numeral classifier

o

object of transitive verb

obj object P patient Paj

Pajonal Ashéninka

pass passive past

past (no abbreviation)

pau paucal pclf

possessive classifier

Pe

Perené Alto

pfv perfective Pi

Pichis Ashéninka

pl, pl plural pl.exc

plural exclusive

xix

xx Abbreviations pl.inc

plural inclusive

pn

personal name

pol polite pos.pol

positive polarity

poss

possessive, possessor

poten potential pp

plural participants

pr ‘pronominal’ pred predicative prep preposition pres presential pres.vis

present visual

prog progressive pron

personal pronoun

prox proximal ptc participle purp purposive q

question marker

quant quantifier r possessor rc

relative clause

real realis rec recipient recip

associative-reciprocal marker

red

reduplicated verb

redup reduplication ref reflexive rel

relative clause marker

rem.p.rep

remote past reported

rep repetitive req request retr retrospective rfl reflexive rslt resultative s

subject of intransitive verb

s/a.top

Abbreviations topical S/A subject

seq sequential sfp

sentence-final particle

sg, sg singular sgf

singular feminine

sgnf

singular non-feminine

simil similative SJ Sino-Japanese SK

Southern Kampa

shon

subject honorific

sp specific Sp Spanish SS

same subject

stat stative sub subordinator subj subjunctive subs

substance (classifier)

T Theme term terminative th thematic top topic trid tridimensional ttl title uncert uncertainty vcc

verbless clause complement

vcl

verbal classifier; verbal action classifier in Chapter 10

vcs

verbless clause subject

verb.cl

verbal classifier

vert vertical vm

valency modifier

xxi

1 Noun categorization devices A cross-linguistic perspective A l e x a n dr a Y. A i k h e n va l d

A noun may refer to a man, a woman, an animal, or an inanimate object of varied shape, size, and function, or have abstract reference. Noun categorization devices vary in their expression, and the contexts in which they occur. Large sets of numeral classifiers in South-East Asian languages occur with number words and quantifying expressions. Small highly grammaticalized noun classes and gender systems in IndoEuropean and African languages, and the languages of the Americas are expressed with agreement markers on adjectives, demonstratives, and also on the noun itself. Further means include noun classifiers, classifiers in possessive constructions, verbal classifiers, and two lesser-known types: locative and deictic classifiers. All noun categorization devices are based on the universal semantic parameters of humanness, animacy, sex, shape, form, consistency, orientation in space, and function. They may reflect the value of the object, and the speaker’s attitude to it. Their meanings mirror socio-cultural parameters and beliefs, and may change if the society changes. Noun categorization devices offer a window into how speakers conceptualize the world they live in. Each contribution to this volume offers a novel study of noun categorization devices in regions known for their elaborate classifier systems. This introduction offers a general typological background, focusing on the issues particularly relevant for the languages discussed within the volume. We start with a brief overview of noun categorization devices in their various contexts.

1  Noun categorization devices in their various guises We distinguish the following seven types of noun categorization devices. This classification is primarily based on the contexts in which each device occurs.1 1  This typology of noun categorization devices is based on the general framework in Aikhenvald (2000), Aikhenvald (2012: 279–303) which is focused on Amazonian languages, and Aikhenvald (2004, 2016), with Genders and Classifiers. First edition. Alexandra Y. Aikhenvald and Elena I. Mihas (eds) This chapter © Alexandra Y. Aikhenvald 2019. First published in 2019 by Oxford University Press

2

Alexandra Y. Aikhenvald

I.  Gender systems—frequently, masculine and feminine—are realized through agreement within a noun phrase (on an adjective or a member of a closed class of determiners) or within a clause (e.g. on the predicate). (An alternative term for such agreement systems is ‘noun classes’.) Gender can be expressed covertly, as an obligatory marker on the agreeing constituent, and also overtly, on a noun itself. As a rule, every noun in the language belongs to a gender. There is always some semantic basis to the assignment of genders; but genders can also be assigned based on morphological and on phonological make-up of the noun (see Aikhenvald 2016, and references there). Warekena of Xié, a North Arawak language, has two genders—feminine and nonfeminine—marked on third person pronouns, and through agreement on demonstratives, adjectives, and verbs. Gender is ‘covert’ (that is, not marked on the noun itself). In (1a) and (1b), the gender-sensitive agreement forms of demonstratives are in bold. (1)  a. ayuta      neyawa  yu-tapa-pa   dem.dist.fem.sg  woman 3sgfem-come-redup   ‘That woman is coming.’ b. eta   dem.dist.nonfem.sg   ‘That man is coming.’

enami man

i-tapa-pa 3sgnf-come-redup

 

 

Warekena

Warekena

Small agreement gender systems are a common feature of Arawak languages (see Chapter  2 on genders in the Kampa subgroup of Arawak, and Chapter  4 on the Arawak languages of north-west Amazonia). Two genders are distinguished in Zamucoan languages of Bolivia and Paraguay (Chapter 5), and in Toba, a Guaycuruan language (Chapter 7). Genders typically interrelate with other nominal and verbal categories—in many languages, fewer genders are expressed in plural than in singular. This is what we find in Warekena of Xié and all other Arawak languages (see also Chapters 2 and 4), and in Toba (Chapter 7). Dependencies between gender and other grammatical systems are discussed in Aikhenvald and Dixon (2011). II. Numeral classifiers occur with number words and quantifiers, categorizing the noun in terms of its shape, animacy, and other inherent properties. Not every noun has to be assigned a classifier. In most Amazonian languages, they are expressed special focus on gender systems; an up-to-date summary with reference to other sources and approaches is in Aikhenvald (2017). See also Kilarski (2013) on the history of studies; some functions of classifiers are recapitulated by Contini-Morava and Kilarski (2013) (with more detail in Aikhenvald 2000: 307–36). See also Aikhenvald (2015) for an up-to-date bibliography on noun categorization devices. Additional issues in classifiers and conceptualization of linguistic entities are addressed in Bisang (2017); Jiang (2017) explores the ways in which classifiers in Chinese reflect speakers’ worldview.



1  Noun categorization devices

3

through bound morphemes attached to a number word, and are used just with low numbers. Palikur, a North Arawak language, has nineteen numeral classifiers—see (2a) and (2b) (Aikhenvald and Green 2011: 411–12). Classifiers are in bold face. (2)  a. paha-tra       ahin   one-num.cl:extended  path   ‘one path’ (extended) b. paha-t   one-num.cl:vertical   ‘one stick’ (vertical)

ah stick

Palikur   Palikur  

Besides number words, numeral classifiers can be used with the interrogative quantifier ‘how many’, as in Munya, a Tibeto-Burman language, Bahuana, an Arawak ­language of north-west Amazonia, White Hmong, and Japanese (Chapters  4,  8,  9, and 10). In Japanese, they also occur with approximate expressions of number such as suu- ‘several’ (§2.4 of Chapter 9; see also Aikhenvald 2017, and 2000: 98–124). Not every noun in a language has to be associated with a numeral classifier. In Japanese, numeral classifiers are obligatory for all nouns except those which ‘tend to denote abstract referents such as units of time, colours, kinds, grades, classes, ­geographical areas, social units’ (Downing 1996: 15), and also some nouns denoting events, particularly in loanwords (see §2.4 of Chapter 9). III. Noun classifiers are associated with the noun itself and are independent of any other element within a noun phrase or in a clause. They categorize the entity in terms of the generic type or class it belongs to. Not every noun in a language will have a noun classifier (see Dixon 1982, 2015). Classifiers may be independent words. Dâw, from north-west Amazonia, has a dozen noun classifiers; an example is at (3) (Martins 1994: 51) (see also (18), from Yidiny, an Australian language). tog (3)  dâw noun.cl.human girl ‘a girl’ (lit. human girl)

 

Dâw

In Witotoan, Kawapanan, and Arawak languages, noun classifiers are attached to the noun itself, and can be deployed as productive derivational devices (see §3.3.1 and §4.1 of Chapter 3, on Shiwilu, a Kawapanan language; Chapter 6 on Murui (Witotoan); Chapter 2 on Kampa; Chapter 4, on North Arawak languages). They maintain their classificatory functions, categorizing the noun or its root in terms of relevant semantic features. IV. Possessive classifiers occur in possessive constructions, characterizing the Possessee in terms of its inherent properties. Many Tupí-Guaraní languages of South

Alexandra Y. Aikhenvald

4

America have two possessive classifiers—one for pets and one for game animals (see Rodrigues  1997: 72–3, and also Kakumasu  1986: 371). Kadiwéu, a Guaycuruan language, has three classifier forms—two for domestic animals (one for males and one for females), and a further one for all other nouns (Sandalo 1995: 57; further examples in Aikhenvald 2000: 126–9). These are illustrated in (4), (5), and (6). (4)

l-wiGadi 3poss-cl:non.female.animal ‘his male horse’

(5)

l-wiqate 3poss-cl:female.animal ‘his female horse’

(6)

Gad:-neb:i 2poss-cl:general ‘your river’

aqi:di river

apolokGanGa horse

apolokGanGa horse

 

Kadiwéu   Kadiwéu

Kadiwéu

 

Palikur, a North Arawak language, has five possessive classifiers—see Table 1 (Aikhenvald and Green 2011: 436). Possessive classifiers tend to correlate with the type of noun: in many languages, including Palikur and Kadiwéu, possessive classifiers are only used with optionally, or alienably, possessed nouns.2 In Zamucoan and Mataguayan languages of the Chaco region, possessive classifiers show gender agreement with the possessed noun.

Table 1.  Possessive classifiers in Palikur Form

Semantics

Examples

-pig

‘pet’ (used with domesticated animals)

gi-pig pewru (3masc-pet dog) ‘his dog’; gi-pig mutom ‘his sheep’

-mana

‘food’ (used with fruit and vegetables)

pi-mana uwas (2sg-food orange) ‘your orange’

-mutra

‘plant’

n-amutra pilatno (1sg-plant banana) ‘my plant-banana’ (the one I planted)

-win

‘catch; animal caught to eat’

nu-win arudiki (1sg-catch tapir) ‘my catch-tapir’ (the tapir I caught)

-kamkayh

‘child’

nu-kamkayh awayg (1sg-child boy) ‘my son’

2  Relational classifiers are an additional type. They categorize the relationship between the possessor and the possessee and reflect the ways in which the possessor can be manipulated by the possessee—eat it, drink it, wear it, sell it, and so on (see Lichtenberk 1983, and a summary in Aikhenvald 2017: 375–7). This is an almost exclusive feature of Oceanic languages (see Guérin 2017).



1  Noun categorization devices

5

Here,  gender is integrated within the system of possessive classifiers (see further ­discussion in §8 of Chapter 5). V. Verbal classifiers occur as bound morphemes on verbs and characterize the S  (intransitive subject) or O (transitive object) in terms of shape, dimensionality, and function. Palikur has eleven verbal morphemes whose use is obligatory.3 In (7), a piece of cotton string in the object function is referred to with a verbal classifier for linear objects (Aikhenvald 2012: 293). (7)

nah watak-buk-e 1sg untie-verb.cl:linear-completive ‘I untied the string completely.’

ini this:neuter

mawru cotton

 

Palikur

Verbal classifiers can also categorize obliques: this is the case in Mɨka, a Witotoan language ((22)–(23) in Chapter 6). Along similar lines, verbal classifiers in Innu, an Algonquian language, may categorize the S (intransitive subject), the O (transitive object), and an oblique argument, instrument, or location (Drapeau and LambertBrétière 2011: 302–4) (similar examples from Motuna and Tarascan are mentioned in Aikhenvald 2000: 162–3). Just occasionally a verbal classifier can categorize the subject of a transitive verb (A): examples of this unusual phenomenon come from Shiwilu, a Kawapanan language (Valenzuela 2016: 368; §5.1 of Chapter 3, this volume). The object of a transitive verb or the subject of an intransitive verb can be ­categorized through using different classificatory verbs. Different verb stems are then regularly used to express handling, location, and existence of objects of different shapes and different arrangements. This is a feature of numerous languages of North and South America (see, for instance, Rushforth 1991 on Mescalero Apache, Fortescue 2006 on Wakashan languages, and also Frank  1990, on Ika, and Malone 2004 on Chimila, both Chibchan; and a summary in Aikhenvald 2017: 378–80). VI. Locative classifiers occur with adpositions or location markers, and are ­chosen based on the inherent physical properties of a noun which refers to a location. Palikur has twelve locative classifiers (Aikhenvald and Green 2011: 430–5). In (8), the choice of the locative classifier -hakwa- ‘into liquid’ is based on the physical property of the location—the noun un ‘water, waterway’. a-hakwa-t (8) wis-uh tarak-e-gu 1pl-excl push-completive-3fem 3neuter-into.liquid-dir ‘We push it (the canoe) into the water.’

un Palikur waterway  

3  Based on partial analysis of secondary data of about a dozen languages with verbal classifiers, Passer (2016) erroneously claims that verbal classifiers do not categorize the noun they refer to. He suggests, inter alia, that verbal classifiers do not occur with overt nouns (a statement easily shown to be incorrect for many languages, including Palikur in (7)).

6

Alexandra Y. Aikhenvald

Locative classifiers have been described for a number of languages of South America— including Palikur and Lokono, from the Arawak language family, and also Carib languages (see an overview in Aikhenvald 2017: 281). This type appears to be comparatively rare. Similar to numeral, possessive, verbal, and noun classifiers, locative classifiers do not have to subsume the whole nominal lexicon. VII. Deictic classifiers are a further, less frequently attested type of noun ­categorization devices. Deictic classifiers are obligatory with demonstratives, and typically refer to the position of the entity, including horizontal—or ‘lying’, and vertical—or ‘standing’. In Siouan languages of North America, they occur on demonstratives and definite articles (Rankin 2004, and references there). Classifiers with demonstratives in Mandan, an extinct Siouan language, are illustrated in (9) (Rankin 2004: 206). (9) a. re-wak   this-cl:hor   ‘this (lying)’ b. re-rak   this-cl:sitting   ‘this (sitting)’

Mandan     Mandan    

Deictic classifiers are a salient feature of Toba and other Guaycuruan languages (see Table 1 in §2.1 of Chapter 7). One language can have several noun categorization devices—this is what we turn to next.

2  Several noun categorization devices in one language A language can combine more than one kind of noun categorization device. A small gender system may be used with demonstratives and other modifiers, alongside numeral classifiers with quantifying expressions and number words. This is a feature of numerous Indic and Dravidian languages (see, for instance, Rastorgueva et al. 1978 and Emeneau 1964), and a number of North Arawak languages (Chapter 4). Noun classifiers coexist with gender systems in a number of languages from Northern Australia (see, for instance, Reid  1997, on Ngan.gityemerri; Sands  1995: 281–2; Dixon 2002: 454–507). A number of Micronesian languages (e.g. Truquese: Benton 1968) have numeral classifiers as well as possessive classifiers. Noun classifiers and numeral classifiers as independent systems are attested in Akatek, a Mayan language (Zavala 2000), and Minangkabau, a Western Austronesian language (Marnita 2016). Two genders (­feminine and non-feminine) are distinguished within the system of numeral classifiers in a number of North Arawak languages of the Upper Rio Negro region (see Chapter 4).



1  Noun categorization devices

7

Possessive classifiers in Zamucoan languages occur together with the markers of two genders—feminine and masculine (Chapter 5). Three kinds of noun categorization devices—numeral classifiers, genders, and noun classifiers—have been described for Chinantec languages of Mexico (Foris 1993). In ’Dongo-ko, a Mba language from the Ubangi branch of Niger-Congo, possessive classifiers coexist with genders and with noun classifiers (Pasch  1985). Kadiwéu, a Guaycuruan language from Brazil, combines possessive classifiers, genders, and deictic classifiers as independent systems (see Sandalo and Micheloudakis 2016; and more on the co-occurrence of different means of noun categorization in one language in Aikhenvald 2000: 184–92). Palikur has as many as five independent systems of noun categorization devices. The language distinguishes: (I) three genders (feminine, masculine, and neuter) marked on demonstratives, free and bound personal pronouns, and two genders (feminine and masculine) marked on verbs; (II) nineteen numeral classifiers with number words; (III) eleven verbal classifiers; (IV) five possessive classifiers (listed in Table 1); and (V) twelve locative classifiers. In addition, a two-fold gender distinction—feminine and masculine—in Palikur is integrated within the system of numeral classifiers just for the number word ‘one’. The feminine and masculine gender markers obligatorily co-occur with the animate classifier -p-, e.g. paha-p-ru tino (one-num.cl:anim-fem woman) ‘one woman’, paha-p-ri awayg (one-num.cl:anim-masc man) ‘one man’ (see the discussion in Aikhenvald and Green 2011). This is reminiscent of the gender distinctions made within numeral classifiers in a number of related North Arawak languages—including Achagua (Chapter 4). Different kinds of noun categorization devices allow the speaker to highlight distinct facets of an object. In (10), from Minangkabau (Marnita 2016: 64), three pieces of Toona Sinensis (also known as Chinese mahogany) are categorized as long rigid objects through the numeral classifier batang. The noun classifier surian with a generic meaning ‘wood’ indicates that we are dealing with wood as material (and not with a tree). The noun classifier surian ‘wood’ helps disambiguate the two meanings of the noun surian ‘Toona Sinensis wood or tree’. The classifier itself comes from this same noun. batang surian surian (10)  tigo three num.cl:long.rigid noun.cl:wood Toona.sinensis ‘three (pieces) of Chinese mahogany wood’

Minangkabau  

In (11) and (12), from Palikur, the numeral classifier -tra categorizes the noun akati ‘cord’ as a linear extended object (also shown in (2a)). The cord is inanimate, and belongs to the neuter gender: this is shown by the third person neuter cross-referencing prefix

8

Alexandra Y. Aikhenvald

a- on the locative classifier. The locative classifier in (11), -buhkumna ‘on.linear’, reflects the horizontal, or linear orientation of the cord along which the birds are sitting. The position of the referents ‘along’ the cord in (11) implies a plural reading for the subject, ‘crow’. bat a-buhkumna paha-tra (11) yu crow sit 3neuter-on.linear one-num.cl:linear ‘Crows sat on/along a (horizontal) cord.’

akati cord

Palikur  

The locative classifier -min ‘on.vertical’ in (12) indicates that the cord is positioned vertically; it is understood that ‘crow’ has a singular referent. bat a-min paha-tra (12) yu crow sit 3neuter-on.vert one-num.cl:linear ‘A crow sat on a (vertical) cord.’

akati cord

Palikur  

The gender prefix reflects the animacy of the object. The choice of a numeral classifier is based on its dimensionality. The locative classifier shows its position and orientation, and the number of entities involved. Noun categorization devices vary in size. Gender systems range from two in Semitic languages, or Portuguese and French, to ten, as in Bantu, or even to several dozen, as in some languages of South America. Systems of numeral classifiers can be small: for instance, Telugu has two (Krishnamurti and Gwynn  1985: 106–7), and Warekena of Xié has six (Chapter  4), while Munya has over fifteen (Chapter  10). Korean and Japanese have more than 150 classifiers (Lee 2014; Downing 1996; §2.5 of Chapter 9 of this volume). Having a large set of numeral classifiers is a feature of the languages of mainland South-East Asia (see, for instance, Enfield 2017). The number of possessive classifiers ranges from two or three, as in Kadiwéu, Chamacoco (§6 of Chapter 5), and many Tupí-Guaraní languages, to five as in Palikur (Table 1), and twenty in Panare, a Carib language (Mattéi-Müller 1974; Carlson and Payne 1989; Payne and Payne 2013: 82–6). The number of verbal classifiers is typically small, e.g. eight in Innu, or eleven in Palikur; and so is the number of locative and deictic classifiers (see the discussion of Toba in Chapter 7). The set of noun categorization devices can be open. In Macushi and Apalaí, two Carib languages, any noun with generic reference can be used as possessive classifier (Koehn  1994; Koehn and Koehn  1986: 85; Abbott  1991: 85–6). This is reminiscent of  Ayoreo where many generic nouns can be used as possessive classifiers (§5 of Chapter 5), making it hard to name the exact number of classifiers. In Hmong, a wide range of nouns can be used as classifiers, making them an open class (§3.2 of Chapter 8). In many South-East Asian languages, including Thai and Lao, a noun itself can be fully or partially repeated in the numeral classifier slot to refer to it being counted (a similar principle has been described for Truquese, a Micronesian language, by



1  Noun categorization devices

9

Benton  1968). Such ‘repeaters’ are typically used if a noun cannot be subsumed under any of the existing classifiers. The presence of repeaters (sometimes also called auto-classifiers or self-classifiers) makes the set of classifier morphemes almost ­open-­ended.4 Grammaticalized repeaters often give rise to classifiers as a closed or semi-closed class of affixes to a numeral. Repeater technique is usually restricted to numeral classifiers and classifiers in multiple environments (see, for instance, the discussion of Murui in Chapter 6).

3  How noun categorization devices differ and what they have in common Each of the noun categorization devices have their special properties, challenging the idea of a putative binary division between genders on the one hand (I in §1) and classifiers on the other (II–VII in §1). Noun categorization devices differ in terms of their scope and the nominal constituent which they classify. Gender categorizes the noun within an attributive noun phrase or a clause. A numeral classifier categorizes the noun within a noun phrase which contains a number word or a quantifier. A noun classifier categorizes a noun on its own. A possessive classifier categorizes the possessed noun within a possessive noun phrase. A locative classifier categorizes the argument of an adposition, and a deictic classifier categorizes the head of a noun phrase modified by a deictic. These differences are summarized in Table 2. We now turn to a few further special features of each type. Table 2.  Noun categorization devices and their scope Type

Scope

What is categorized

I Gender

Attributive NP or clause

Head noun, A/S or S/O; oblique

II Numeral classifiers

Numeral/quantifier NP

Head noun

III Noun classifiers

Noun

Head noun

IV Possessive classifiers

Possessive NP

Possessed noun

V Verbal classifiers

Clause

S/O or oblique

VI Locative classifiers

Adpositional NP

Noun referring to a location

VII Deictic classifiers

Attributive NP

Head noun

4  This is similar to alliterative agreement described as one of the agreement techniques for a number of varieties of Baïnounk, a West Atlantic language spoken in Senegal and Guinea-Bissau (Sauvageot  1967; Cobbinah  2010). For a number of nouns, agreement is marked by repeating the first CV sequence of the stem on the agreeing constituent, making the number of agreement classes virtually open.

10

Alexandra Y. Aikhenvald

3.1  Interaction with other grammatical categories Noun categorization devices differ in how they interact with other grammatical categories within the language (see Aikhenvald  2000: 242–70, and Table 10.7 there). A marker of gender can be fused with that of number. Genders can be neutralized in plural number, as we saw at I in §1 (further discussion of dependencies between gender, number, person, and case is in Aikhenvald and Dixon 2011 and Aikhenvald 2000: 252–7). The use of possessive classifiers often correlate with whether or not the object has to be possessed. Some nouns are obligatorily, or ‘inalienably’ possessed: they can never occur without a possessor. These typically include body parts, kinship terms, and culturally important items (including ‘name’ or ‘house’) (see, for instance, Chapter 2 for a discussion of possession classes of nouns in the Kampa languages of Peru). Other, alienably possessed, nouns can occur on their own. They are optionally, or alienably possessed. Possessive classifiers typically occur just with alienably possessed nouns. We saw in (IV) in §1 that this is the case in Palikur, Zamucoan languages, and Kadiwéu. In White Hmong, classifiers in possessive constructions are obligatory just for alienably possessed nouns (§2.1 of Chapter 8; Bisang 1993: 30). 3.2  Differences in use depending on contexts Noun categorization devices can be used differently depending on semantic or other properties of the forms they occur with. The use of numeral classifiers may correlate with the origin of the classifier, and of the number word. In Kolami, a Dravidian language, all number words above five are borrowed from Marathi, an Indic language; they require a classifier (also borrowed), while number words from one to four distinguish three genders (Subrahmanyam 2006: 306–7). The Native Japanese generic classifier -tsu is used with the native number words below ten. No classifier is used for counting the same referents above ten with Sino-Japanese number words (§2.4 of Chapter  9). Only native number words in Munya require a numeral classifier (Chapter 10). Numeral classifiers may be restricted to small numbers. In Nung, a Tai language, classifiers are optional with multiples of ten (Saul and Wilson 1980: 27; see also Marnita 2016: 68–9, on numeral classifiers in Minangkabau; and Haas 1978: 73 and Bhaskararao and Joshi 1985: 20 on Burmese and Newari). In Warekena of Xié, classifiers are only used with number words ‘one’ and ‘two’. When counting large numbers in casual speech in Japanese, classifiers can be omitted (§2.4 of Chapter 9). The correlation between the value of number word and the use of classifier may have to do with the size of the number word system. In Munya (Chapter 10), native number words which have to be used with a classifier cover the values from one to twenty only. Loanwords from Tibetan—which optionally occur with classifiers—are used to count beyond twenty.



1  Noun categorization devices

11

Verbal classifiers are typically used on verbs of selected semantic types. In many languages, classificatory verbs and verbal classifiers tend to cover the semantic groups of handling, motion, location, and existence. This is the case in Athabaskan and in Wakashan languages, and also Ika and Chimila, Chibchan languages from Colombia.5 Verbal classifiers in Palikur are used just with transitive verbs of affect and handling, and stative verbs of dimension, physical property, and colour (Aikhenvald and Green 2011: 420–2). Cherokee, a Southern Iroquoian language, has five verbal classifiers, covering objects with the meanings of animate, liquid, flexible, long and rigid, and neutral/ compact (see Blankenship  1997: 92; Mithun  2017: 767–8, and references there). Examples from Oklahoma Cherokee are in (13)–(17) (Blankenship 1997: 92). (13) Wèésa gà-káà-nèè’a cat 3sg>3sg-animate-give.pres.prog ‘She is giving him a cat.’

Oklahoma Cherokee  

(14) Àma gà-nèèh-néé’a water 3sg>3sg-liquid-give.pres.prog ‘She is giving him water.’

Oklahoma Cherokee  

(15) Àhnàwo gà-nv́v́-nèè’a shirt 3sg>3sg-flexible-give.pres.prog ‘She is giving him a shirt.’

 

Oklahoma Cherokee

(16) Gànsda àa-d-éé’a stick 3sg>3sg-long-give.pres.prog ‘She is giving him a stick.’

Oklahoma Cherokee  

(17) Kwàna àa-h-nèè’a peach 3sg>3sg-compact-give.pres.prog ‘She is giving him a peach.’

Oklahoma Cherokee  

Classifiers in Cherokee are employed with transitive verbs involving handling of physical objects, such as ‘hold’, ‘handle’, ‘break’, ‘drop’, ‘carry’, and also with the verb of transfer, ‘give’ (categorizing the ‘gift’). Classifiers also occur with intransitive verbs of motion or position. Along similar lines, verbal classifiers in Mɨka, a Witotoan language (§5 of Chapter 6), are used just with verbs which imply direct physical contact with the object, e.g. ‘cut’, ‘throw away’, and ‘pick, pluck’. Classifiers which frequently occur with a verb may form conventionalized ­combinations with it. The form may then get lexicalized. This has been documented for a number of verbs of motion, transportation, and affect in Innu, an Algonquian 5  See Fortescue (2006) on Wakashan languages, Poser (2005), Krauss (1968), Rice and de Reuse (2017) on Athabaskan, Frank (1990) on Ika, and Malone (2004) on Chimila.

12

Alexandra Y. Aikhenvald

language with a large set of verbal classifiers categorizing S, O, or obliques (see V in §1). High-frequency combinations of a verb stem with a classifier referring to the typical object or a typical location can now be treated as one lexical item, e.g. matâpê- ‘arrive at the shore’ (matâ- ‘arrive at’, -pê- ‘classifier:free-flowing liquid’) (Drapeau and Lambert-Brétière  2011: 301). In Chapter  6, Wojtylak describes a similar process in Murui, a Witotoan language closely related to Mɨka. We return to this in §6. 3.3  Special semantic properties Numeral classifiers stand apart from other noun categorization devices in that they fall into two categories—sortal and mensural. Sortal classifiers characterize a referent in terms of its features, which include humanness, animacy, shape, and function. Mensural classifiers provide information about the properties of the referent and combine reference to the ‘natural or inherent quantum, configuration, or boundary of the referent’ (as put by Jarkey and Komatsu, Chapter  9). Examples from Japanese are in (1) in Chapter 9; Table 2 in Chapter 10 contains a list of sortal and mensural classifiers in Munya. Sortal and mensural classifiers may differ in their properties. In Korean, a sortal classifier may occasionally be omitted with the number word ‘one’ in its indefinite meaning (Lee  2014: 38). A mensural classifier can never be omitted. Only sortal classifiers in Japanese can be used in counting events (§2.3 of Chapter 9)—we return to this in §5. Numeral classifiers of mensural type tend to share similarities with measure terms. The choice of numeral classifier always correlates with the properties of a referent, while the choice of a measure does not have to do so (see the discussion in §2.2 of Chapter 9, on Japanese, and Chapter 10, on Munya). The marker of arrangement, -tsa ‘storey (of a house)’, in Munya occurs in the same slot as a classifier, but does not have the same syntactic properties as classifier morphemes: it is described as a ‘pseudoclassifier’ in §2.3 of Chapter 10. Two consecutive number words can occur together accompanied by one numeral classifier, with an approximate reading, e.g. Mandarin Chinese liang-san-ge ren (twothree-cl:gen person) ‘several people’ (the discussion of this phenomenon in Chinese and in Munya is in §3.2 of Chapter 10; see also §2.4 of Chapter 9, for similar examples in Japanese, and Haas 1978: 72–3 on similar constructions in Burmese). 3.4  Using more than one noun categorization device Some noun categorization devices can occur more than once. Several noun classifiers can be used within one noun phrase. Two classifiers referring to humans can occur together in one noun phrase in Yidiny, an Australian language. In (18), the classifier bama ‘cl:person’ co-occurs with waguuja ‘cl:man’, which form one noun phrase with the noun wurgun ‘a teenage boy’, forming a noun phrase which translates, literally, as cl:person cl:man boy. The classifiers are in a generic-specific relationship to each other and to the noun to which they refer (Dixon 1977: 484; 2015: 49).



1  Noun categorization devices

13

bama waguuja Yidiny (18) ŋanyji we+nom cl:person+abs cl:man+abs   wurgun muyŋga gunda-alna   pubescent.boy+abs cicatrice+abs cut-purp   ‘We must cut tribal marks [on] the teenage boy.’ (lit. person man pubescent boy) Several noun classifiers can occur on the same root in Shiwilu, e.g. dadapu-dek-lu’ (white-cl:liquid-cl:soil) ‘white clay (obtained from white/clear waters and used to paint ceramic’ ((47), Chapter 3). In Kampa languages, several classifiers can occur on one noun to achieve precision in the description of the object, e.g. Satipo Ashaninka no-ma-kamara-peta (1sg.poss-robe-cl:cinnamon.colour-cl:devalued) ‘my dirty useless robe’. The classifier -kamara usually refers to a dirty and stinky piece of clothing. The shape classifier -peta ‘flattened flaccid fruit; deflated hanging pouch’ adds a negative value to the noun (§4.1 of Chapter 2). A noun in Tariana can contain up to three classifiers, in a part-whole relationship  to each other, e.g. kara-ka-whya-puna-kha (rel+fly-th-cl:canoe-cl:roadcl:curved) ‘curved part of an airstrip’ (lit. curved of road of canoe-like flyer) (see (15a-c)-(16a,b) of Chapter 4). This is similar to the instances of double marking of genders (or ‘noun classes’) on the noun itself, and on agreeing constituents, in Kikiyu, Ndali, and Nyanja, all Bantu languages (Vail 1974: 42; Stump 1993: 173–5), and a few languages of Northern Australia, including Nungali (Bolt et al. 1971: 70) and Yanyuwa (Kirton 1971; Evans 1994).6 The occurrence of more than one noun categorization device of other kinds within the same context is uncommon (see (36a-b) from White Hmong in Chapter 8, which illustrate this point). Kadiwéu appears to be unique in allowing its two possessive classifiers to occur with the same noun (Sandalo 1995: 57), as shown in (19). i-neb:i (19) i-wiGadi 1poss-cl:animal 1poss-cl:generic w-akipe niy:Godi 3sg.subject-drink water ‘The horse of mine drinks water.’

apolikGanGa horse    

Kadiwéu      

Two possessive classifiers in one noun phrase are said to express ‘emphasis’. 3.5  Principles of assignment Noun categorization devices differ in how they are assigned to a noun. Gender systems stand apart from the rest. Gender assignment can be based on semantic principles. A semantically based division of nouns into feminine and masculine, 6  A modifier in a noun phrase in Tariana can contain up to three classifiers marking agreement with multiple targets (Aikhenvald 2003: 99–100).

14

Alexandra Y. Aikhenvald

or non-feminine, is a feature of North Arawak and Witotoan languages (Chapters 4 and 6). Gender choice in Kampa languages and in Ayoreo (§3.3 of Chapter 2 and §3 of Chapter 5) can be partly accounted for by mythological ­associations, along the lines of the Myth-and-Belief principle put forward by Dixon (2015: 29) in his discussion of gender choice in Dyirbal, an Australian language. Celestial bodies in Kampa are believed to be mythical people, and are assigned to masculine or feminine gender according to the sex of the entity. For instance, kashiri ‘moon’ in Perené is a mythical man, and the word belongs to masculine gender. Alternatively, gender can be assigned on the basis of morphological and phonological features of the noun, and only partly its semantics (see Aikhenvald 2004, and an earlier study in Corbett 1991, for examples from German and other European languages). The degree of semantic motivation for gender choice varies. In quite a few IndoEuropean and Bantu languages, gender choice is semantically opaque, leading some scholars to claim that there is no semantic basis to it at all (cf. Meillet  1964, Trudgill 2007). The choice of gender in German is frequently cited as an example of a gender system where semantics plays no role. In a classic study, Zubin and Köpcke (1986) showed that this is not the case. There is indeed a complex semantic rationale for gender choice which depends on the semantic field the noun belongs to. Male and female adults of domestic and game animals are assigned masculine and feminine gender respectively—this demonstrates the correlation between the natural gender, or sex, and linguistic gender. Neuter gender is assigned to juvenile terms, and those not specified for sex. Masculine gender includes types of cloth, precipitation, and minerals. Disciplines and types of knowledge are assigned feminine genders; metals (except alloys) are neuter (further discussion is in Zubin and Köpcke 2009). Semantic opacity is not an exclusive feature of gender. The choice of classifiers of other kinds can also be far from straightforward. The choice of most verbal classifiers in Cherokee is straightforward. The verbal classifier for ‘compact items’ is an ­exception: it subsumes abstract nouns and items not covered by other classes (Blankenship 1997: 93). The choice of deictic classifier in Toba is largely transparent. It is determined by typical posture or orientation of the object (§2.2 of Chapter 7): humans and trees are classified as ‘vertical’, beds and snakes as ‘horizontal’, and smaller animals (such as capybara) and pots as ‘sitting’. Stars, houses, and internal body parts are also ­categorized as ‘sitting’, possibly, based on an extension from ‘sitting’ position to permanence of the entity. Elements of the natural environment—forest or lake—are ­categorized as extended in space and thus ‘horizontal’. The choice of a classifier can be partially explained—but is not fully predictable. The classifier -hon in Japanese is another case in point (Matsumoto 1993: 676–8; §2.3 of Chapter  9, this volume). In its most common use, it covers saliently onedimensional objects, e.g. long, thin, rigid objects such as sticks, canes, pencils, candles, trees, and processed fish. It also covers martial arts contests with swords (which



1  Noun categorization devices

15

are long and rigid), hits in baseball, shots in basketball, judo matches, rolls of tape, telephone calls, radio and TV programmes, letters, movies, medical injections, bananas, carrots, pants, guitars, and teeth. It is never used with terms for snakes and worms. They are referred to with the classifier -hiki for living animates (§5.2–3 of Chapter 9). Such heterogeneity of meanings comes about through various processes of semantic extension and metonymy, possible to explain, but hard to predict. Differences in the specific features of noun categorization devices point towards the fact that genders, on the one hand, and classifiers, on the other, do not form a dichotomy. This was previously demonstrated, and argued for, in Aikhenvald (2000, 2015, 2017). Gender is just one of seven types of categorization devices established so far, on an equal footing with each of the other six—numeral classifiers, noun classifiers, possessive classifiers, verbal classifiers, locative classifiers, and deictic classifiers. 3.6  Semantic parameters in noun categorization devices What all the noun categorization devices share is their semantic basis. The universal parameters include humanness, animacy, shape, dimensionality, consistency, and directionality. At the same time, different devices have their semantic preferences. Gender systems typically have sex, humanness, and animacy as their major semantic features. Noun classifiers tend to have generic meanings categorizing the referent in terms of its social status and function. Numeral and verbal classifiers tend to ­categorize the referent by shape and directionality. Possessive classifiers tend to reflect generic-specific relations and are chosen based on the nature of the possessed noun. Locative classifiers combine reference to physical properties and orientation of the referent (and may include animacy). Deictic classifiers tend to be chosen based on orientation, directionality, and shape, and also animacy of the referent. These preferences are summarized in Table 3 (adapted from Table 11.4 in Aikhenvald 2017). Table 3.  Preferred semantic parameters in noun categorization devices Device

Typical semantics

genders

animacy, humanness, physical properties, rarely nature or function

noun classifiers

social status, functional properties, nature

numeral classifiers

animacy, humanness, physical properties, nature, rarely functional properties

verbal classifiers

physical properties, orientation, rarely animacy, nature

possessive classifiers

physical properties, nature, animacy, functional properties

locative classifiers

physical properties, orientation, rarely animacy

deictic classifiers

orientation, dimensionality, shape, animacy

16

Alexandra Y. Aikhenvald

Physical properties recurrent in noun categorization devices cover shape, size, consistency, and dimensionality. Temperature as a physical property is never used as a parameter in noun categorization. Categorization by colour appears to be extremely rare. A possible exception is a colour-based classifier in Satipo Ashaninka -kamara ‘classifier for cinnamon-coloured objects’ ((23) in Chapter 2). Value is rarely encoded within the system of noun categorization devices. Northern Kampa languages are unusual in having a classifier with purely evaluative meanings (used in multiple contexts: see §4), e.g. the classifier ‑shiteki ‘devalued, rag-like’ in Alto Perené (in (22) of Chapter 2). Some Oceanic languages have a relational or a possessive classifier with the meaning of ‘valuable possession’ (e.g. Raga: Lichtenberk 1983: 154). The Lolovoli dialect of the Northeast Ambae has a special possessive classifier bula-, covering animals, crops, and especially valuable possessions, e.g. objects of adornment (Hyslop 2001: 178–80). Shape-based classifiers can develop negative overtones. For instance, the shape classifier ‑peta ‘flattened, flaccid fruit’, ‘deflated hanging pouch’ in Satipo Ashaninka has a negative meaning of ‘devalued, useless’ ((23) and (25) in Chapter 2). The classifier -tek ‘skin’ in Shiwilu has developed deprecatory meaning and can refer to something ugly and bad in quality (§3.3.4 of Chapter 3). Similarly, value, speaker’s attitude, and importance can be encoded through linguistic gender. In Palikur, males belong to the masculine gender and females to the feminine gender. In addition, masculine gender is associated with ugly things, and feminine gender with small and cute ones. Rats are assigned to the masculine gender because they are looked upon as dirty and harmful. But a cute little baby rat will be referred to as feminine (Diana Green, p.c.; further examples in Aikhenvald  2016: 44–51). In its standard use, the Japanese numeral classifier -hiki refers to small animate beings, including fish, insects, and worms (see Jarkey and Komatsu, §5.2 of Chapter 9). The classifier was metaphorically applied to actors performing the traditional Japanese art of kabuki in the early Edo period (1603–73). Referring to a member of the profession as yakusha ip-piki (actor one-num.cl:small.animate) reflected the lowly status of kabuki actors, and became standard usage in official documents at the time. This use of -hiki is now standard. In the modern language, this classifier can be used in non-standard ways with reference to humans, giving the impression of disdain and lack of value ((31)–(32) in Chapter 9): three people who do not live up to the expectations of adult behaviour are referred to with -hiki. The classifier effectively downgrades people to the status of non-human animates, and has overtones of ‘not intelligent’, ‘small’ and ‘not civilized’. Depending on the context, however, the same classifier may have an opposite meaning of endearment, pride, and affection. This is what we see in  a description of three delightful croissants rising together in the oven ((30) in Chapter 9). The exact overtones can only be understood within the context—this is what makes classifiers a flexible and creative device.



1  Noun categorization devices

17

All noun categorization devices can be used creatively, to highlight special features of a referent, and the speaker’s attitude to it. We return to this in §5.

4  Multiple classifier languages The same set of forms for noun categorization devices may appear in several contexts, creating a multiple classifier system. In Tariana, a North Arawak language, we find the same classifier morpheme used with (i) a demonstrative modifier, (ii) a number word, (iii) an adjective, (iv) a noun itself, and (v) on a possessive modifier. Example (20) illustrates all these contexts. It comes from literacy materials compiled by speakers of Tariana (see also Table 3 in Chapter 4). pa-dapana di-tape-dapana (20) ha-dapana that-cl:house one-cl:house 3sgnf-medicine-cl:house hanu-dapana wa-ya-dapana-nuka   big-cl:house 1pl-poss-cl:house-present.visual  ‘That one hospital (lit. medicine house) is ours.’

     

Tariana

Multiple classifier systems vary in the contexts where classifiers are required. Many Kampa (Arawak) languages of Peru employ the same classifier forms with number words, verbs, adjectives, and on nouns themselves, as noun classifiers (Chapter 2 of this volume, Mihas 2017, and also Michael 2008: 332, on Nanti). In White Hmong, classifiers are used with demonstratives, number words, and in possessive constructions (Chapter 8). In Shiwilu, the same set of classifiers is used with number words, demonstratives, nouns, adjectives, and verbs (§3 of Chapter 3). In Murui, a Witotoan language, classifiers are used on number words, adjectives, nouns, pronominal possessives, and demonstratives; Mɨka, from the same family, also uses classifiers on verbs (§3 of Chapter 6). One classifier environment in a multiple classifier system can be historically older than another one—we return to this in §6. Multiple classifier systems can involve further contexts. In Murui, Tariana, Baniwa of Icana, and a few other languages (e.g. Anamuxra: Ingram  2003, and Omaha-Ponca: Rankin  2004: 216–17) classifiers also occur on interrogative modifiers. Classifiers in Shiwilu can occur with personal pronouns ((26)–(28) in Chapter 3; and similar ­examples from Tariana there). However, no language has been found so far with a special kind of noun categorization restricted to just interrogatives, or just personal pronouns. The existence of languages with ‘multiple’ classifiers in many contexts points towards the intrinsic unity of noun categorization devices as a linguistic p ­ henomenon.

5  The utility of noun categorization devices Noun categorization devices are never semantically redundant. They often serve to tease apart distinct meanings of a polysemous noun. Anything to do with water, or a waterway, in Tariana is referred to with the polysemous word uni ‘water, waterway’.

18

Alexandra Y. Aikhenvald

Classifiers help distinguish a lake, uni hanu-nai (water big-cl:lake) ‘big lake’, from a bay, uni hanu-dawa (water big-cl:corner) ‘big bay’ (see Table 7 in Chapter 4). Along similar lines, ‘river’ in Burmese can be viewed as a place, as a line on a map, or a sacred object: each of these meanings is expressed through using different classifiers on a number word (see Becker 1975). Deictic classifiers in Toba also help distinguish different meanings of the same form. The noun waGayaGa ‘fox’ occurs with the deictic classifier ñi ‘tridimensional’ used to refer to animals. The same form waGayGa is applied to the kind of fish Charex leticiae whose mouth is similar to that of a fox. But, unlike the fox, the fish is referred to with the deictic classifier zi ‘horizontal’ ((27a-b) of Chapter 7; similar examples from Baniwa of Içana are in (11)-(12) in Chapter 4, and §3.3.5 of Chapter 3). Using different numeral classifiers in Munya helps highlight different features of the noun, and thus has a pragmatic effect. Crops can be referred to with the classifier for plants—the connotation then is that the crops are growing fine. Or they can be referred to with the classifier for long objects—the connotation is that they are not growing as well as expected and may not yield a good harvest (§2.2.1 of Chapter 10). Classifiers do not necessarily provide information additional to the noun (contrary to Bisang 2017: 224). Rather, they may foreground a property of the object. A classifier can be used creatively, to highlight a property of a person or an object. This is how the classifier -hiki ‘small living animates’ can be used in Japanese to refer, in a somewhat jocular way, to people who do not live up to the normal expectations (§5.1–2 of Chapter 9, and §3.6). Noun classifiers help expand the lexicon. The noun yera ‘tobacco’ in Murui can combine with a dozen classifiers, creating terms for various items associated with tobacco. So, yera-fo (tobacco-cl:cavity) means ‘tobacco container with a hole’, yerako (tobacco-cl:cover) means ‘round tobacco container’, and yera-rui (tobaccocl:day) means ‘the day when tobacco is sent out’ (see (9) in Chapter 6, and similar examples in §4.3 of Chapter 2, §3.3.1 of Chapter 3, and §6 of Chapter 4). The presence of classifiers may correlate with the lack of a semantic subclass of adjectives. Tariana and Baniwa of Içana have multiple classifiers many of which are based on shape and form. There are no adjectives referring to shape: classifiers do the job. Tariana has no adjective meaning ‘long’: instead, one uses the classifier -pi ‘long object’, e.g. panisi hanu-pi (house big-cl:long) ‘a big long house’ (§6 of Chapter 4). Classifiers in Shiwilu encode shape and size; for instance, the classifier -la ‘seed’ can indicate small roundish shape (further examples of such ‘adjective-like’ functions of classifiers are in §3.3.4 of Chapter 3). Classifiers can individuate parts of an object. In (24), from Alto Perené, the classifier ‑tsa ‘curvilinear objects’ refers to individual threads of a piece of nylon material (§4.1 of Chapter 2). Similar examples are in §6 of Chapter §4. All noun categorization devices can be used anaphorically, to refer to something or someone mentioned previously, or obvious from the context. This has been described



1  Noun categorization devices

19

for numeral classifiers in Vietnamese (e.g. Daley 1998: 60–3) and Japanese (Downing 1996: 159–91 and Chapter  9 of this volume), and many other languages with other systems of classifiers (see (28), from Ashaninka, in Chapter 2; §3.3.3 of Chapter 3, on Shiwilu; §6 of Chapter  4 on Tariana and Baniwa of Içana; §7.2 of  Chapter  5, on Old  Zamuco; §5 of Chapter  6, on verbal classifiers in Mɨka, and §3 of Chapter  7, on Toba). If the context is clear enough, there may be no need to mention the noun at all: a classifier will help keep track of what is being talked about. Verbal classifiers in Innu can replace an argument or an oblique which is recoverable from the context (see Drapeau and Lambert-Brétière 2011: 306–11; see also §2.2 of Chapter 2, and §3.3.3 of Chapter 3). Classifiers in Murui often occur on their own, without the mention of an overt noun (§2 of Chapter 6; a similar usage in White Hmong is described in §1.4 of Chapter  8). Boat journeys to the Tariana-speaking villages used to last for several days. In our discussions of placenames and names for landmarks, encountered classifiers outnumbered full nouns in a ratio of about twenty to one. Classifiers can have further pragmatic overtones. In White Hmong, and also in Vietnamese they correlate with definiteness and referential salience of an object (see Chapter 8). In Vietnamese, if a referent is to be further deployed in the narrative, it has to occur with a classifier at its first mention (Daley 1998; Thurgood 2001). The utility of classifiers goes hand in hand with their frequency in discourse. Thiesen and Weber (2012: 163) state that, in Bora, approximately four out of every ten words bear a classifier. According to Mihas (Chapter 2), on average there is a classifier every 4.72 sentences in Matsigenka, and one every 8.5 sentences in Alto Perené Ashéninka, both from the Kampa subgroup of Arawak. The ratio 1/10 in terms of token frequency for nouns compared with classifiers has been calculated for a corpus of narratives in Tariana and Baniwa of Içana. Noun categorization devices are amenable to language engineering, and may reflect social changes and attitudes. Following an order of King Mongkut issued in 1854 with regard to classifiers in Thai, ‘noble’ animals such as elephants and horses should be counted without any classifier; the classifier tua could only be used for animals of a ‘lower’ status (Juntanamalaga 1988). The choice of numeral classifiers in Maonan, a Tai-Kadai language spoken in China, reflects the changing place of women within the society. Similarly to a number of languages of South-east Asia, women used to be counted with the classifier tɔ2, which also subsumes animals and children. Respected women were typically counted using the human classifier ʔai1 (num.cl:human). At present, all women who have a professional status are referred to with the ‘human classifier’. The choice of the human classifier can also be used as a token of a person’s status, and their maturity. Children are typically referred to with the animal classifier tɔ2. This applies to primary school pupils, as in sa:m tɔ2 la:k8fia:k8 (three num.cl:animal pupil) ‘three pupils’. But college students will be referred to with ʔai1, e.g. ŋɔ4 ʔai1

20

Alexandra Y. Aikhenvald

ta4jɔŋ5 (five num.cl:human college.student) ‘five college students’ (Lu 2012: 83–4, 101–2, 119–21). Classifiers can reflect the ways in which people live, and how they relate to each  other. River-dwelling peoples, such as the Tariana and Baniwa of north-west Amazonia, have classifiers for canoes. Classifiers in Korean and many South-East Asian languages reflect kinship relationships and social hierarchies. The meanings, and the uses, of nominal classification systems provide a unique insight into how the world is categorized through language. All classifiers discussed above relate to a property of nouns. Verbal action markers— sometimes referred to as verbal action classifiers or even verbal classifiers—are rather different phenomena. In a number of Tibeto-Burman and South-East Asian languages, verbal action markers occur with number words and categorize the action in terms of its frequency, time duration, and manner. Verbal action markers in Newari, a Tibeto-Burman language, appear on a number word in the same slot as classifiers used to categorize nouns (Bhaskararao and Joshi 1985: 18). In (21), the number word ni- ‘two’ is used with the numeral classifier -pu ‘thin and long object’ (Bhaskararao and Joshi 1985: 18; 25). (21)

ni-pu two-cl:long.thin ‘two belts’

peti belt

Newari

 

In (22), the verbal action marker dhu: ‘times’ occurs with the same number word, indicating the number of times the event has occurred. (22) ni-dhu: two-verbal.action.marker:times ‘(Somebody) came twice.’

wAlA came

 

Newari

Along similar lines, verbal action markers in Munya occur in the same slot on a number word as a classifier would. But rather than categorizing a noun, they refer to the number of times the activity is performed (§4 of Chapter 10). Verbal action markers in Ersu, a Tibeto-Burman language from China, cover frequency, degree, and ­duration; they occupy the same slot as a numeral classifier (Zhang 2016: 408–44; see also Wang and Zheng 1993: 49–50 on Mulao, a Tai-Kadai language, and Haas 1942: 205 on Thai, and §3.3.2 of Chapter 8 of this volume, on White Hmong).7 7  An overview of the phenomenon in a selection of languages from East Asia is in Gerner (2014). Gerner makes a distinction between verbal action markers (which he calls ‘verb classifiers’) and classifiers used on verbs to categorize nominal arguments which he misleadingly confuses with noun incorporation. Incorporated nouns can historically develop into verbal classifiers, but not all verbal classifiers come from incorporated nouns (as has been demonstrated in Aikhenvald 2000: 149–71, Aikhenvald and Green 2011, and Chapters 2 and 4 of this volume).



1  Noun categorization devices

21

The number words da- ‘one’ and mena- ‘two’ in Murui, a language with classifiers in multiple contexts, occur with the frequency marker -kaño ‘times’ in the classifier slot, forming da-kaño ‘once’ and mena-kaño ‘twice’ (Wojtylak 2017: 148–9). The form -piu ‘times’ in Tariana and -pia in Kurripako (Bezerra 2012: 57) behave in a similar fashion: they accompany a number word or a quantifier in the slot normally occupied by a classifier. Examples from Kurripako are apa-pia (one-time) ‘once’, yama-pia (two-time) ‘twice’, madali-pia (three-time) ‘thrice’, kadali-pia? (how.many-time) ‘how many times?’. A classifier in the same slot with the quantifier ‘how many’ is illustrated by kadali-ma inaithepe (how.many-cl:fem women) ‘how many women?’ (Bezerra 2012: 31; examples of classifiers in Baniwa of Içana-Kurripako are in §3 of Appendix 2 to Chapter 4, this volume). ‘Event counters’ in Japanese, such as -kai ‘times’ in san-kai ‘three-times’, are similar to verbal action markers: they supply information about the number of actions and occur in the same slot as a numeral classifier. And numeral classifiers in Japanese can go beyond categorizing nouns. The classifiers ‑hon ‘long thin things or events that involve a trajectory (such as baseball hits, telephone calls, and movies)’ and ‑hatsu ‘explosive events, bullets, or bombs’ categorize both objects and actions (§2.3 of Chapter 9). Classifiers such as -ken ‘incidents and cases’, -rei ‘bows (involving bowing during worship at a shrine)’, and ‑han ‘crimes’ are used to categorize just actions or events (§2.3 of Chapter 9). The overlap between numeral classifiers for actions and for objects points towards a common basis for categorization across word classes. It appears, however, that this overlap is limited just to numeral classifiers in large systems (such as Japanese). So far, no system has been described in which noun classifiers or possessive classifiers would categorize a noun and an action within the same language.

6  The origins and development of noun categorization devices Noun categorization devices may be archaic, or may have developed relatively recently. The two genders, masculine and feminine, in Arawak languages are of fair antiquity and can be reconstructed to the proto-language (see Chapters 2 and 4). In some subgroups of North Arawak languages, classifiers developed independently; in others they did not develop at all. Low frequency of classifiers in Shiwilu texts points towards their relatively recent origin (§6 of Chapter 3). Some numeral classifiers in Munya (Chapter 10) originate in nouns. So do classifiers in multiple functions in Murui, Shiwilu, and North Arawak languages (see also (10), for a noun classifier with an obviously nominal origin in Minangkabau). However, nouns are not the only source of classifiers (see Aikhenvald 2000: 362–3). Deictic classifiers in Guaycuruan languages, including Toba (Chapter 7), come from positional verbs (Ceria and Sandalo 1995). Classifiers in multiple functions in Kampa languages come from a variety of bound nominal and verbal roots (§4.1 of Chapter 2). This qualifies Mithun’s (1986: 388) suggestion that ‘all classificatory stems begin life as nouns’.

22

Alexandra Y. Aikhenvald

Genders stand apart from other noun categorization devices in that their exponents can originate from pronouns (see Aikhenvald 2016: 76–82, and further references there). Bound noun roots—used as agreement markers within a noun phrase or incorporated into the verb—appear to have given rise to classifiers in Arawak languages (including North Arawak and Kampa, discussed within this volume) and also in Shiwilu. If a language utilizes the same set of classifiers in more than one environment, one classifier environment can be historically older than another. Classifiers in possessive constructions in Cantonese may be considered a later development as a result of Hmong-Mien influence (Matthews 2006: 231–2). Classifiers with demonstratives in Tariana developed relatively recently due to Tucanoan influence as the people moved to the Vaupés River Basin linguistic area. This environment is absent from closely related North Arawak languages (Aikhenvald 2007, Chapter 4 of this volume). In the history of languages, a frequently used combination of a classifier with a noun can acquire a conventionalized meaning. Temporal expressions in Murui often consist of a fixed combination of a demonstrative with a classifier, e.g. bi-rui (thiscl:day) ‘today’, jiai-mona (other-cl:season ) ‘next year’ (Wojtylak 2017: 149). Noun classifiers in Shiwilu form part of established placenames and terms for body parts, e.g. Milek-pi-lu’-dek (yarina.palm-cl:fruit-cl:soil-cl:river) ‘Yarinayacu river’ (lit. ‘the river on whose banks the fallen fruits of the Yarina palm abound’ ((48) in §4.1 of Chapter 3) (similar examples from Tariana are in §6 of Chapter 4). We can recall from §1 that verbal classifiers of all sorts tend to occur with transitive verbs of affect and induced position, categorizing the noun in S or O function in terms of its shape and other inherent properties. Combinations of a verb stem with a classifier referring to the typical object or a typical location may come to be frequently used, so much so that, over time, they form one lexical items. This process has been described for some verb-classifier combinations in Innu, an Algonquian language (see §1, and Drapeau and Lambert-Brétière 2011: 301). Verbal classifiers in Murui followed a similar path. But in contrast to Innu, all combinations of verbs with classifiers in Murui have become fully conventionalized. As a result, the erstwhile verbal classifiers now occur as optional formatives on a few verbs implying direct effect on the object. Verbal classifiers are still productive in the related Mɨka—which reflects an earlier stage. None of the neighbouring languages have verbal classifiers. Language contact may have played a role in the demise of verbal classifiers as a productive noun categorization device in Murui. Numeral classifiers may develop additional, non-classificatory functions. In Munya, the number word ‘one’ in combination with the general classifier can be used as a degree adverb ‘a little’ (§3.4 of Chapter 10), or as a complementizer (§3.6 of Chapter 10). Combinations of ‘one’ with erstwhile classifiers gave rise to indefinite pronouns (§3.5 of Chapter 10). The spread and development of noun categorization devices is often due to areal  diffusion. Classifiers in the languages of Lowland Amazonia are found in



1  Noun categorization devices

23

g­ eographically contiguous zones (see Aikhenvald  2012: 300–3). We find languages with complex systems of noun categorization devices in north-west Amazonia spanning adjacent regions of Brazil, Colombia, and adjacent regions of Venezuela, and from north-eastern to central Peru. Many of the languages with multiple classifiers discussed in this volume are spoken within these regions—Murui in Colombia, Tariana in Brazil, Baniwa of Içana-Kurripako in Brazil, Colombia, and Venezuela, Shiwilu and Kampa languages in Peru (see also §5 of Chapter 2). Possessive classifiers are shared by the languages of the Chaco region, among them Zamucoan. Agreement in gender on possessive classifiers is an unusual feature Zamucoan languages share with some of their neighbours (§8 of Chapter 5). Complex histories of noun categorization devices in various languages reflect the histories of their speakers. The meanings, and the uses, of genders and of classifiers provide a unique insight into how the world is categorized through language.

7  About this volume The aim of this volume is to learn more about diverse means of noun categorization, their forms, meanings, and development. The nine contributions to the volume address noun categorization devices in the regions of their major concentration and diversity, for languages of different affiliations and different typological profiles. The first six chapters focus on languages of South America. We start with the ­languages of Lowland Amazonia, known for their complex systems of genders and classifiers. Chapter 2 ‘Genders and classifiers in Kampa (Arawak) languages of Peru’, by Elena Mihas, provides a comprehensive analysis of noun categorization devices in this subgroup, focusing on the small gender system and a large system of classifiers in multiple contexts. Shiwilu is a member of the small Kawapanan language family, from Peruvian Amazonia. Its closed set of about twenty classifiers is the focus of Chapter  3, ‘Classifiers in Shiwilu (Kawapanan): Exploring typologically salient properties’, by Pilar Valenzuela. In Chapter 4, ‘A view from the north: Genders and classifiers in Arawak languages of north-west Amazonia’, Alexandra Aikhenvald discusses co-existing systems of genders and classifiers of various types in a number of Arawak languages from the Upper Rio Negro basin—the major locus of linguistic diversity within the family. The next three chapters focus on individual types of classifiers. In Chapter  5, ‘Possessive classifiers in Zamucoan’, Luca Ciucci and Pier Marco Bertinetto analyse a complex system of possessive classifiers and a system of two genders across the three languages of the family—Ayoreo, Chamacoco, and the extinct Old Zamuco. The fate of verbal classifiers in the languages of the ‘Witoto’ dialect continuum in Colombian Amazonia, with classifiers in multiple contexts, is the topic of Chapter 6, ‘The elusive verbal classifiers in “Witoto” ’, by Katarzyna Wojtylak. In Chapter 7, ‘Multifunctionality of deictic classifiers in the Toba language (Guaycuruan)’, Cristina Messineo and Paola

24

Alexandra Y. Aikhenvald

Cúneo offer an analysis of semantic and morphosyntactic properties of deictic classifiers in the language and the ways in which they are used in different genres of ­discourse. The last three chapters focus on classifier systems in the languages of Asia. In Chapter 8, ‘Classifiers in Hmong’, Nathan M. White discusses classifiers in multiple environments in White Hmong, exploring the differences in their use with nouns, number words, and in possessive constructions. In Chapter 9, ‘Numeral classifiers in Japanese’, Nerida Jarkey and Hiroko Komatsu offer a comprehensive overview of numeral classifiers in Japanese, addressing their generally obligatory character, the options for their omission in specified environments, and their metaphorical and creative uses which reflect subjective attitudes and social values. In Chapter  10, ‘Numeral Classifiers in Munya, a Tibeto-Burman language’, Junwei Bai explores noun categorization in a previously undescribed Tibeto-Burman language spoken in Sichuan province in China, with a special focus on how classifiers can develop other, non classificatory, uses.

All noun categorization devices—despite their differences—reflect a single ­phenomenon: the ways humans classify objects through language. Noun categorization devices reflect common cognitive mechanisms and common semantic features, mirroring different cultural experience of their speakers. This volume offers new insights into  the range of noun categorization devices in their diversity, exploring their ­historical development, histories, areal diffusion, manipulations in discourse, and creative use. The studies within this volume, all based on extensive firsthand data and analyses, will further refine our understanding of noun categorization devices across the world.

References Abbott, M. 1991. ‘Macushi’, in D. C. Derbyshire and G. K. Pullum (eds), Handbook of Amazonian Languages, vol. 3. Berlin: Mouton de Gruyter, 23–160. Aikhenvald, A. Y. 2000. Classifiers: A typology of noun categorization devices. Oxford: Oxford University Press. Aikhenvald, A. Y. 2003. A grammar of Tariana, from northwest Amazonia. Cambridge: Cambridge University Press. Aikhenvald, A. Y. 2004. ‘Gender’, in G. E. Booij, C. Lehmann, and J. Mugdan (eds), Article 98 of Morfologie/Morphology. Ein Handbuch zur Flexion und Wortbildung/A Handbook on Inflection and Word Formation, 2 Halbband. Berlin: Mouton de Gruyter, 1031–45. Aikhenvald, A. Y. 2007. ‘Classifiers in multiple environments: Baniwa of Içana/Kurripako: a North Arawak perspective’. International Journal of American Linguistics 27: 475–500. Aikhenvald, A. Y. 2012. The languages of the Amazon. Oxford: Oxford University Press.



1  Noun categorization devices

25

Aikhenvald, A. Y. 2015. ‘Oxford Bibliography Online: Classifiers’ (refereed updateable resource with summaries and evaluation for each entry), General Editor: Mark Aronoff. New York: Oxford University Press (updated 2015). Aikhenvald, A. Y. 2016. How gender shapes the world. Oxford: Oxford University Press. Aikhenvald, A. Y. 2017. ‘A typology of noun categorization devices’, in A. Y. Aikhenvald and R.  M.  W.  Dixon (eds), The Cambridge Handbook of Linguistic Typology. Cambridge: Cambridge University Press, 361–404. Aikhenvald, A. Y. and R. M. W. Dixon. 2011. ‘Dependencies between grammatical systems’, in A. Y. Aikhenvald and R. M. W. Dixon (eds), Language at large. Essays on syntax and semantics. Leiden: Brill, 170–204. Aikhenvald, A. Y. and D. Green. 2011. ‘Palikur and the typology of classifiers’, in A. Y. Aikhenvald and R. M. W. Dixon (eds), Language at large. Essays on syntax and semantics. Leiden: Brill, 394–450. Becker, A.  J. 1975. ‘A linguistic image of nature: the Burmese numerative classifier system’, Linguistics 165: 109–21. Benton, R. A. 1968. ‘Numeral and attributive classifiers in Truquese’, Oceanic Linguistics 7: 104–46. Bezerra, Z. A. 2012. Dicionário Kuripako/Português. Mimeographed. Missão Novas Tribos do Brasil. Bhaskararao, P. and S.  K.  Joshi. 1985. ‘A Study of Newari Classifiers’, Bulletin of the Deccan College Research Institute 44: 17–31. Bisang, W. 1993. ‘Classifiers, Quantifiers and Class Nouns in Hmong’, Studies in Language 17: 1–51. Bisang, W. 2017. ‘Classification between Grammar and Culture. A Cross-Linguistic Perspective’, in T.  Pommerrening and W.  Bisang (eds), Classification from Antiquity to Modern Times. Berlin: De Gruyter, 199–230. Blankenship, B. 1997. ‘Classificatory verbs in Cherokee’, Anthropological Linguistics 39: 92–110. Bolt, J. E., W. G. Hoddinott, and F. M. Kofod. 1971. An elementary grammar of the Nungali language of the Northern Territory. Armidale: Mimeo. Carlson, R. and D. Payne. 1989. ‘Genitive classifiers’, in Proceedings of the 4th Annual Pacific Linguistics Conference. Eugene: University of Oregon, 89–119. Ceria, V. G. and F. Sandalo. 1995. ‘A Preliminary Reconstruction of Proto-Waikurúan with Special Reference to Pronominals and Demonstratives’, Anthropological Linguistics 37: 169–91. Cobbinah, A. 2010. ‘The Casamance as an area of intense language contact: the case of Baïnounk Gubaher’, Journal of Language Contact 3: 175–201. Cobbinah, A.  Y. 2017. ‘Suffixed plurals in Baïno(u)nk languages: agreement patterns and ­diachronic development’, Journal of African Languages and Linguistics 38: 145–85. Contini-Morava, E. and M.  Kilarski. 2013. ‘Functions of nominal classification’, Language Sciences 40: 263–99. Corbett, G. 1991. Gender. Cambridge: Cambridge University Press. Daley, K.  A. 1998. Vietnamese classifiers in narrative texts. Arlington: Summer Institute of Linguistics and the University of Texas in Arlington. Dixon, R. M. W. 1977. A Grammar of Yidiny. Cambridge: Cambridge University Press.

26

Alexandra Y. Aikhenvald

Dixon, R. M. W. 1982. Where Have All the Adjectives Gone? And Other Essays in Semantics and Syntax. Berlin: Mouton. Dixon, R.  M.  W. 2002. Australian Languages: their origin and development. Cambridge: Cambridge University Press. Dixon, R. M. W. 2015. Edible gender, mother-in-law style & other grammatical wonders. Studies in Dyirbal, Yidiñ, & Warrgamay. Oxford: Oxford University Press. Downing, P. 1996. Numeral Classifier Systems: The Case of Japanese. Amsterdam: John Benjamins. Drapeau, L. and R.  Lambert-Brétière. 2011. ‘Verbal classifiers in Innu’, Anthropological Linguistics 53: 293–22. Emeneau, M. B. 1964. ‘India as a linguistic area’, in Hymes, D. (ed.), Language in Culture and Society: a Reader in Linguistics and Anthropology. New York: Harper and Row, 642–53. Enfield, N.  J. 2017. ‘Language in the Mainland Southeast Asia Area’, in A.  Y.  Aikhenvald and  R.  M.  W.  Dixon (eds), The Cambridge Handbook of Linguistic Typology. Cambridge: Cambridge University Press, 601–23. Evans, N. 1994. ‘The problem of body parts and noun class membership in Australian languages’, University of Melbourne Working Papers in Linguistics 14: 1–8. Foris, D. 1993. A Grammar of Sochiapan Chinantec. Arlington: Summer Institute of Linguistics and the University of Texas in Arlington. Fortescue, M. 2006. ‘The origins of the Wakashan classificatory verbs of location and handling’, Anthropological Linguistics 48: 266–87. Frank, P. 1990. Ika Syntax. Studies in the Languages of Colombia 1. Dallas: Summer Institute of Linguistics and the University of Texas at Arlington. Gerner, M. 2014. ‘Verb classifiers in East Asia’, Functions of Language 21: 267–96. Guérin, V. 2017. ‘The Oceanic subgroup of the Austronesian language family’, in A. Y. Aikhenvald and R.  M.  W.  Dixon (eds), The Cambridge Handbook of Linguistic Typology. Cambridge: Cambridge University Press, 911–41. Haas, M. R. 1942. ‘The use of numeral classifier in Thai’, Language 18: 201–5. Haas, M. R. 1978. ‘The use of numeral classifiers in Burmese’, in Language, culture, and history. Essays by Mary  R.  Haas, selected and introduced by Anwar  S.  Dil. Stanford, California: Stanford University Press, 65–81. Hyslop, C. 2001. The Lolovoli dialect of the North-east Ambae language, Vanuatu. Canberra: Pacific Linguistics. Ingram, A. 2003. ‘The morphosyntax of numeral classifiers in Anamuxra: details of a multiple classifier system’, Anthropological Linguistics 45: 129–68. Jiang, S. 2017. The semantics of Chinese classifiers and linguistic relativity. London: Routledge. Juntanamalaga, P. 1988. ‘Social issues in Thai Classifier Usage’. Language Sciences 10: 313–30. Kakumasu, J. 1986. ‘Urubu-Kaapor’, in D. C. Derbyshire and G. K. Pullum (eds), Handbook of Amazonian Languages, vol. 1. Berlin: Mouton de Gruyter, 326–406. Kilarski, M. 2013. Nominal Classification: A history of its study from the classical period to the present. Amsterdam: John Benjamins. Kirton, J.  F. 1971. Papers in Australian Linguistics. No. 5. Australian National University, Canberra: Pacific Linguistics. Koehn, E. and S. Koehn. 1986. ‘Apalai’, in D. C. Derbyshire and G. K. Pullum (eds), Handbook of Amazonian Languages, vol. 1. Berlin: Mouton de Gruyter, 33–127.



1  Noun categorization devices

27

Koehn, S. 1994. ‘The use of generic terms in Apalaí genitive constructions’, Revista Latinoamericana de estudiosn etnolingüisticos VIII: 39–48. Krauss, M. 1968. ‘Noun classification systems in Athabascan, Eyak, Tlingit and Haida Verbs’, International Journal of American Linguistics 34(3): 194–203. Krishnamurti, Bh. and J.  P.  L.  Gwynn. 1985. A Grammar of Modern Telugu. Delhi: Oxford University Press. Lee, Y. 2014. Classifiers in Korean. Munich: Lincom Europa. Lichtenberk, F. 1983. ‘Relational Classifiers’, Lingua 60: 147–76. Lu, T.-Q. 2012. Classifiers in Kam-Tai languages. A cognitive and cultural perspective. Boca Raton: Universal Publishers. Malone, T. 2004. ‘Classifiers in Chimila (Chibchan)’, in A. Y. Aikhenvald (ed.), Nominal classification, Special issue of Sprachtypologie und Universalienforschung 57 2/3, 144–201. Marnita, R. A. S. 2016. Classifiers in Minangkabau: a typological study. Munich: Lincom Europa. Martins, S. A. 1994. ‘Análise da morfosintaxe da língua Dâw (Makú-Kamã) e sua classificação tipológica’. MA thesis, Universidade Federal de Santa Catarina, Florianópolis. Matsumoto, Y. 1993. ‘Japanese numeral classifiers: a study on semantic categories and lexical organisation’, Linguistics 31: 667–713. Mattéi-Müller, M.-C. 1974. ‘El sistema de posesión en la lengua panare’, Antropologica 38: 3–14. Matthews, S. 2006. ‘Cantonese grammar in areal perspective’, in A.  Y.  Aikhenvald and R.  M.  W.  Dixon (eds), Grammars in contact. A cross-linguistic typology. Oxford: Oxford University Press, 220–36. Meillet, A. 1964. ‘The feminine gender in the Indo-European languages’, in D.  Hymes (ed.), Language in Culture and Society: a reader in linguistics and anthropology. New York: Harper and Row, 124. Michael, L. D. 2008. ‘Nanti evidential practice: Language, knowledge, and social action in an Amazonian society’. PhD dissertation, University of Texas at Austin. Mihas, E. 2017. ‘The Kampa subgroup of the Arawak language family’, in A. Y. Aikhenvald and R.  M.  W.  Dixon (eds), The Cambridge Handbook of Linguistic Typology. Cambridge: Cambridge University Press, 782–814. Mithun, M. 1986. ‘The Convergence of Noun Classification systems’, in C. G. Craig (ed.), Noun Classes and Categorisation. Proceedings of a Symposium on Categorisation and Noun Classification, Eugene, Oregon, October 1983. Typological Studies in Language 7. Amsterdam: John Benjamins, 379–98. Mithun, M. 2017. ‘The Iroquoian language family’, in A. Y. Aikhenvald and R. M. W. Dixon (eds), The Cambridge Handbook of Linguistic Typology. Cambridge: Cambridge University Press, 747–81. Pasch, H. 1985. ‘Possession and possessive classifiers in ’Dongo-ko’, Afrika und Übersee 68: 69–85. Passer, M. B. 2016. ‘(What) Do verb classifiers classify?’ Lingua 174: 16–44. Payne, T.  E. and D.  L.  Payne. 2013. A typological grammar of Panare, a Cariban language of Venezuela. Leiden: Brill. Poser, W. J. 2005. ‘Noun classification in Carrier’, Anthropological Linguistics 47: 143–68. Rankin, R. L. 2004. ‘The history and development of Siouan positionals with special attention to polygrammaticalization in Dhegiha’, in A.  Y.  Aikhenvald (ed.), Nominal classification, Special issue of Sprachtypologie und Universalienforschung 57 2/3, 202–27.

28

Alexandra Y. Aikhenvald

Rastorgueva, V. S., V. A. Efimov, and A. A. Kerimova. 1978. ‘Iranskie jazyki’ (‘Iranian languages’), in N.  I.  Konrad (ed.), Jazyki Azii i Afriki. tom 2. Indoevropejskie jazyki. Moscow: Nauka, 7–253. Reid, N. 1997. ‘Class and Classifier in Ngan’gityemerri’, in M. Harvey and N. Reid (eds), Nominal Classification in Aboriginal Australia. Amsterdam: John Benjamins, 165–228. Rice, K and W. de Reuse. 2017. ‘The Athabaskan (Dene) language family’, in A. Y. Aikhenvald and R.  M.  W.  Dixon (eds), The Cambridge Handbook of Linguistic Typology. Cambridge: Cambridge University Press, 707­­–46. Rodrigues, A. D. 1997. ‘Nominal Classification in Kariri’, Opción 13: 65–79. Rushforth, S. 1991. ‘Uses of Bearlake and Mescalero (Athapaskan) Classificatory Verbs’, International Journal of American Linguistics 57: 251–66. Sandalo, M. F. 1995. ‘A grammar of Kadiwéu’. PhD dissertation, University of Pittsburgh. Sandalo, F. and D. Micheloudakis. 2016. ‘Classifiers and plurality: evidence from a deictic classifier language’, The Baltic International Yearbook of Cognition, Logic, and Communication Volume 11. Number: cognitive, semantic and crosslinguistic approaches 1–39. Sands, A. K. 1995. ‘Nominal Classification in Australia’, Anthropological Linguistics 37: 247–346. Saul, J. E. and N. F. Wilson. 1980. Nung Grammar. Dallas: Summer Institute of Linguistics and the University of Texas at Arlington. Sauvageot, S. 1967. ‘Note sur la classification nominale en baïnouk’, La Classification nominale dans les langues négro-africaines. Paris: CNRS, 225–36. Stump, G.  T. 1993. ‘Reconstituting morphology: the case of Bantu preprefixation’, Linguistic Analysis 23: 169–204. Subrahmanyam, P. S. 2006. ‘Kolami’, in S. B. Steever (ed.), The Dravidian languages. London: Routledge, 301–27. Thiesen, W. and D.  Weber. 2012. A Grammar of Bora, with special attention to tone. Dallas, Texas: SIL International Publications in Linguistics 148. Thurgood, G. 2001. ‘Review of Vietnamese classifiers in narrative texts, by Karen Ann Daley’, Language 77: 198. Trudgill, P. 2007. ‘On grammatical gender as an arbitrary and redundant category’, in D. Kibbee (ed.), History of Linguistics 2005. Selected papers from the Third International Conference on the History of Language Sciences, Urbana-Champaign, Illinois, 1–5 September 2005. Amsterdam: John Benjamins, 24–36. Vail, L. 1974. ‘The noun classes in Ndali’, Journal of African Languages 11: 21–47. Valenzuela, P.  M. 2016. ‘Classifiers in Shiwilu (Kawapanan) in Northwestern Amazonian Perspective’, Anthropological Linguistics 58: 333–80. Wang, J. and G. Zheng. 1993. An Outline Grammar of Mulao. National Thai Studies Centre, Australian National University, Canberra. Wojtylak, K. I. 2017. ‘A grammar of Murui (Bue), a Witotoan language of Northwest Amazonia’. PhD dissertation, James Cook University. Zavala, R. 2000. ‘Multiple classifier systems in Akatek Mayan’, in G.  Senft (ed.), Systems of Nominal Classification. Cambridge: Cambridge University Press, 114–46. Zhang, S. 2014. ‘Numeral classifiers in Ersu’, Language & Linguistics 16: 883–915. Zhang, S. 2016. A reference grammar of Ersu, a Tibeto-Burman language of China. Munich: Lincom Europa.



1  Noun categorization devices

29

Zubin, D. and K.  M.  Köpcke. 1986. ‘Gender and Folk Taxonomy: The Indexical Relation Between Grammatical and Lexical Categorization’, in C. G. Craig (ed.), Noun Classes and Categorisation. Proceedings of a Symposium on Categorisation and Noun Classification, Eugene, Oregon, October 1983. Amsterdam: John Benjamins, 139–80. Zubin, D. and K. M. Köpcke. 2009. ‘Gender control: lexical or conceptual’, in P. O. Stenkrüger and M. Krifka (eds), Trends in Linguistics: Studies and Monographs. On inflection. In memory of Wolfgang U. Wurzel. Berlin: Mouton de Gruyter, 237–62.

2 Genders and classifiers in Kampa (Arawak) languages of Peru E l e na I . M i h a s

1 Introduction The chapter investigates the semantics, morphosyntax, and functions of the gender and classifier systems of Kampa languages of Peru. The aims are two-fold: to establish the paths of the historical development of the two systems, and identify synchronic interdependencies, if any, between them. The data come from my fieldwork on three Kampa varieties of Alto Perené (a.k.a. Ashéninka Perené1), Pichis Ashéninka, and Satipo Ashaninka carried out in the Chanchamayo and Satipo Provinces of Peru in 2009–16. I also draw on secondary sources, such as the authoritative dictionaries of Matsigenka (Snell 2011) and Ashéninka (Payne 1980), Bible texts ‘Irineane Tasorentsi’ (2008) and ‘Kameethari Ñaantsi’ (2008), and text collections compiled by Anderson (1985, 1986), Michael et al. (2013), and Weiss (1975). The Kampa Arawak subgroup comprises Ashéninka Pichis (Pi), Alto Perené (Pe), Ashéninka Pajonal (Paj), Ucayali-Yurua Ashéninka (U-Y), Ajyininka (or Ashéninka) Apurucayali (Apu), Ashaninka (Asha),2 Caquinte (Ca), Matsigenka/Machiguenga (Ma), 1 The language name is surrounded with controversy. The Spanish-language term Alto Perené (its English equivalent is Upper Perené) was recently brought into circulation due to the indigenous leadership’s insistence on its use. Perené is the name of the main watercourse of the region, and Alto Perené is perceived by the speakers to be a close counterpart of the auto-denomination katonkosatzi ‘those from upriver’. The alternative name, Ashéninka Perené, is strongly dispreferred by the native community. Speakers cite two reasons for rejecting the name Ashéninka: it was imposed by outsiders, and it is a mis­ nomer because the native population uses it to refer to the Kampa speakers of the Gran Pajonal highlands. 2  The proposal about Ashaninka dialect continuum spoken across a vast geographical area was first put forth by Sylvester Dirks in his analysis of the lower Tambo variety’s phonemic inventory (1953: 302). It was later reaffirmed by the prominent SIL linguists Kenneth Pike and Willard Kindberg (1956: 415), Willard Kindberg (1961: 519, footnote 1; 1975: 2), and Lee Kindberg (1961: 505, footnote 1). The dialect spoken in the valleys of the Satipo and Lower Perené Rivers and along the Panga and Mazamari Rivers of the Satipo province of Peru is called here Satipo Ashaninka (coded by the abbreviation Sa). The term Satipo is used as a convenient geographical reference point, because Satipo is the main trading town of the area. Genders and Classifiers. First edition. Alexandra Y. Aikhenvald and Elena I. Mihas (eds) This chapter © Elena I. Mihas 2019. First published in 2019 by Oxford University Press



2  Genders and classifiers in Kampa (Arawak) languages of Peru

31

Kampa Southern

Nomatsigenga

Northern

Nanti Matsigenga Caquinte Ashéninka Ashaninka

Figure 1  Internal classification of Kampa (Michael 2008: 218) Kampa Nomatsiguenga Matsigenka

Nanti

Caquinte

Ashéninka

Ashaninka

Figure 2  Internal classification of Kampa (Michael 2011)

Nanti (Na), and Nomatsiguenga (No). A putative member of the Kampa unit is South Ucayali Ashéninka. Within the group, the non-definitive basic division into Northern Kampa (Ashe, Asha, and Ca, henceforth NK) and Southern Kampa (Ma, No, Na, henceforth SK) is motivated by the shared innovations (Michael  2008: 218–19). The proposed division shown in Figure 1 also reflects the geographical clustering of the languages. The most recent internal classification of the Kampa languages (Michael  2011; Lawrence  2012), shown in Figure  2, distinguishes between No and the rest of the group (Asha-Ashe-Ca-Ma-Na). The classification is proposed on the basis of the shared sound changes and innovations which do not include No. Both classifications in Figures 1–2 are non-definitive due to the lack of empirically-based comprehensive grammars and dictionaries of Kampa languages. For  this reason, in this chapter the terms Northern Kampa (NK) and Southern Kampa (SK) should be understood to refer to the geographical locations of these languages. Speakers of Kampa languages occupy a vast area of the central-eastern lowland Peru and the state of Acre in Brazil. Geographically, Kampa-speaking populations are found in the region covering the eastern foothills of the Andes and the western fringe of the Amazon basin. On the basis of the 2007 National Census data, the current ethnic population of the Ashaninka and Ashéninka is estimated to be 97,477 people (Perú: Análisis Etnosociodemográfico 2010). Populations of Ma and No are less numerous, with 11,279 and 8,016 speakers, respectively; 439 people identified themselves as Caquinte; for Nanti, there are no official statistical data. A number of Kampa language varieties are highly endangered, such as Pi and Pe, whose speakers ceased to transmit their mother tongues to the children.

32

Elena I. Mihas

2  Relevant aspects of Kampa grammar 2.1  Linguistic profile A general outline of Kampa languages is found in Mihas (2017b). Here I provide a more detailed account of the relevant aspects of Kampa grammar. Kampa languages are highly synthetic, agglutinating, head-marking, incorporating, have both gender classes and classifiers, with no case marking of core arguments. The nominative-­ accusative alignment is found in transitive constructions, evidenced in the patterns of verbal person marking. Only A/S arguments are obligatorily marked on the verb, whereas the indexation of objects depends on the referent’s topicality. The constituent order is VS/VO, with A occurring either after or before the direct object due to ­discoursepragmatic considerations. Kampa languages have open word classes of verbs, nouns, and ideophones, a semiclosed class of adverbs, and closed classes of underived adjectives, pronouns, and interrogative words. The structural properties and behaviour of adjectives vary across Kampa. Two sets of adjectives are usually identified. The set of underived (inherent) adjectives is small, with no more than a dozen members. The adjectives share many morphosyntactic properties with nouns, but unlike nouns they cannot be possessed; neither can they inflect for nominal tense. Kampa adjectives show gender agreement with head nouns in third persons, as in Sa (1). In SK, some adjectives agree in both gender and animacy with the controller noun (see §3.2 for details). antearo (1) a. ir-oki 3m.poss-eye.nm big.nm ‘His eyes are big.’

Satipo Ashaninka

b. anteari  omani big.m   fish.species.m ‘a big fish’ The class of number words is small, constituted by a basic set of a few number terms with the meanings ‘one’, ‘two’, ‘three’, and ‘five’ (Table 1). In some languages, there is an additional term for ‘four’. The available evidence suggests that the number words come from verb roots, except for ‘five’ which combines the body part noun root pako~bako ‘hand’ with the number root apa ‘one’. Speakers have a basic ability to count on their fingers, but the counting routine is hardly ever used. Kampa number words do not show gender agreement with controller nouns in noun phrases, as illustrated by Pe in Table 1. To be integrated into discourse, number words are not required to take classifiers. Neither is the plural number inflection required by the number word construction. Kampa languages are both prefixing and suffixing. Kampa noun morphology is mainly suffixal. Prefixal morphology is limited to the pertensive possessor marker. The noun template for Pe suffixal morphology is given in Scheme 1.



2  Genders and classifiers in Kampa (Arawak) languages of Peru

33

Table 1.  Pe number words Number words

Animate noun

Inanimate noun

aparoni ‘one’

aparoni mantsiyari ‘a/one sick person’

aparoni mapi ‘a/one rock’

apite ‘two’

apite itomi ‘his two sons’

apite ishechani ‘his two axes’

mava ‘three’

mava tsinani ‘three girls’

mava kovitzi ‘three pots’

apapakoroni ‘five’

apapakoroni jevari ‘five chiefs’

apapakoroni kitaiteri ‘five days’

Stem

Possession -ni -te, -ri

Spatial Case -ki

Plural Number -paye

Past-oriented Tense -ni

Scheme 1  Pe noun template

Nominal inflectional morphology (spatial case, plural number, and tense) is non-­ essential for the referent’s identification. The reason is the diffuse spatial semantics of the formative -ki~-ku describing spatial relations between the figure and the ground, and the transnumeral nature of Kampa nouns. Tense marking is also of minor importance because of the specific semantics of its meaning. The nominal temporal suffix -ni indicates a ceased existence of humans and high animates, or ceased possession of inanimate entities, e.g. Pe jevari-ni ‘late chief ’ and novanko-ni ‘my former plot of land’ (i.e. it belongs to somebody else) (see Mihas 2015: 356–9 for details). Possession is the only critical inflection needed by bound nouns to be incorporated into discourse (see §2.2 for details). Verbs exhibit complex suffixal morphology coding plural number, valence-­adjusting devices, manner, direction, degree, modality, aspect, and reality status. Verbs divide into two classes on the basis of the reality status inflection they take, namely Conjugation Classes I and A. The verbal category of tense appears to be of minor importance; the past tense distinction has been found so far only in a few NK ­languages (e.g. in Pe and Sa). In verbs, affixes occupy particular positions within certain zones. The identification of zones is made on the basis of broad generalizations relating the formative class to its semantics and function. Placement in the same zone does not preclude affixes from co-occurring in a verb. The sequence of the Pe verbal affixal zones is given in Scheme 2. Person (A/S) Irrealis Causative

STEM

Applicative Reversative Plural number Adverbial (degree, time, manner) Modality Direction

Aspect Reality status

Scheme 2  Pe verbal affixal zones

Person Nominalizer Relativizer Plural number Remote Past Tense Negation

34

Elena I. Mihas

2.2  Kampa noun classes The marking of possessive relations occupies a central place in the grammar of Kampa nouns. Following the pan-Arawak pattern, nouns are divided into possessable and non-possessable (Figure  3). Possession is obligatorily head-marked in the form of affixal morphology. Non-possessable nouns do not normally take possessive inflection. They are constituted by unbound (free, self-standing) roots which refer to the entities from the domains of natural elements, supernatural benevolent and malevolent forces, and astral entities. They also include proper names, vocative/address kinship terms, and some orientation and geomorphic terms. However, in some contexts, they are inflected on the pattern of optionally possessed nouns. For example, when the Pe term kamari ‘demonic spirit’ is used to refer to the person’s own demons, it takes possessive affixes, as in i-yamari-te (3m.poss-demonic.spirit-poss) ‘his demons’. Optionally possessed nouns belong to the semantic domains of flora, fauna, kinship, household items, and landscape. They are constituted by unbound roots which could be integrated into discourse on their own, e.g. Pe mapi ‘rock’, nija ‘water’, and ani ‘brother-in-law’. The optionally possessed nouns take possessor prefixes (which are gender-sensitive in third persons) and possessive suffixes -ni~-ne, -te, and -ri~-re. The choice of the suffixes -ni~-ne and -te is largely motivated by the word’s syllable count. As illustrated in Pe (2a-b), -ni is used with mono- and disyllabic roots and -te with trisyllabic roots. -ri~-re tends to be used with artefacts, plant parts, and, as illustrated in Pe (2c), kin terms. (2) a. i-mapi-ni 3m.poss-stone-poss ‘his stone’

Alto Perené

b. a-nija-te 1pl.poss-water-poss ‘our water (supply)’ c. n-ani-ri 1sg.poss-brother-in-law-poss ‘my brother-in-law’ Noun classes Possessable Obligatorily possessed

Non-possessable

Optionally possessed

Figure 3  Kampa noun classes



2  Genders and classifiers in Kampa (Arawak) languages of Peru

35

Obligatorily possessed nouns are constituted by bound roots, which are found in the domains of kinship, body parts, parts of wholes, human, animal, and plant excretions, and valuable possessions. For the purpose of comparison, this subclass of obligatorily possessed nouns is called here Regular Bound Nouns. They are inflected for the ­‘specific, definite R’ or ‘unspecified R’, i.e. they are either marked by the ‘specific, ­definite R’ possessor prefix, which shows gender distinctions (masculine vs. feminine/ non-masculine) in third persons, or by the gender-insensitive ‘unspecified R’ possessor suffix -(n)tsi. In NK, kin terms can take the unspecified possessor suffix -(n)tsi, e.g., Apu tomitsi ‘son’, sintotsi ‘daughter’, sarintsi ‘grandson (Female Ego)’, anirontsi ‘daughter-in-law (Male Ego)’ (Payne 1981: 50). However, in SK languages, e.g. in Ma, kin terms may not receive the unspecified possessor marker (Farmer and O’Hagan 2014: 12). In Na, human, animal, and plant part terms cannot occur in an unpossessed form (Michael 2012: 154). Among bound nouns which are obligatorily marked for possession, there is a ­special set of nouns with basic meanings pertaining to the domains of plant parts, body parts, and spatial and geomorphic features. It is labelled here Special Bound Noun Set 1. Its boundaries are fuzzy. It partially overlaps with the Regular Bound Nouns because both subclasses include plant and body parts terms. The Special Bound Noun Set 1 is an open class of bound nouns estimated to comprise a few dozen members. Nearly all core Kampa multiple classifiers belong to this set (see §4 for a discussion of core Kampa classifiers). The cognate forms exhibit similar semantics (see Appendix) and morphosyntactic behaviour. A summary of the characteristics of Set 1 nouns is given below. Set 1 nouns are inflected on the pattern of obligatorily possessed nouns. They show gender agreement with third person possessors, exemplified by the Pe o-tapi (3nm. poss-lower.side) ‘its bottom (of a hill)’ and i-chee (3m.poss-horn/spiny.thorn) ‘its horn (of a deer)’. Some Set 1 nouns are morphologically constrained. They do not take the unspecified possessor suffix -(n)tsi, exemplified by the Ma bound roots -tsego ‘branch’, -poa ‘trunk’, -tsa ‘liana’, -a ‘water’, -tishi ‘mountain’, and -nanpina ‘side’ (Farmer and O’Hagan 2014: 13). However, at least in Pe and Sa, the constraint does not apply stringently. For instance, the Pe and Sa counterparts -chevo and -chao of the Ma bound noun -tsego ‘branch’ could be either inflected for a specific possessor, ochevo (Pe) ~ ochao (SA) ‘its branch’, or for an unspecified possessor, chevotsi (Pe) ~ chaotsi (Sa) ‘a branch’. Set 1 nouns occur either as constituents of possessive phrases or as free-standing nouns, as exemplified by the Asha plant part terms otsoaki ‘its shoot’, ominka ‘its trunk/stalk’, and opa ‘its sheath-like leaf/pod’ in (3).

36

Elena I. Mihas

o-tsoaki] Ashaninka (3) [o-shiok-ak-e     kipatsi-ki] [o-jiva-t-i        3nm.s-grow-pfv-real  earth-loc  3nm.s-lead-ep-real  3nm.poss-shoot [impoiji o-minka]   [impoiji    o-pa] conn  3nm.poss-stalk  conn     3nm.poss-pod/sheath-like.leaf ‘It grows from the ground: the tip of the shoot comes first, then the stalk, then the leaf.’ (Irineane Tasorentsi 2008: 38) Set 1 nouns take other inflectional nominal morphology, e.g. plural number and case markers. In Sa (4), the plant part term -minka ‘trunk/stalk’ bears the marker of the  spatial case -ki. The polysemous bound noun -minka exhibits the substance/­ consistency meaning ‘wooden material’. o-minka-ki Satipo Ashaninka (4) n-ojok-an-ak-i-ro 1sg.a-abandon-dir-pfv-real-3nm.o 3nm.poss-wooden.material-loc pitotsi canoe ‘I left it (the cargo) in the wooden canoe.’ Set 1 nouns are modified like other nouns, e.g. by adjectives, number words, and demonstratives. In Sa (5), the bound noun onaki ‘its interior space’ is modified by the derived adjective kameetsari ‘good’. (5) y-am-ak-e-ro           {aparoni o-naki      kameetsa-ri}NP  3m.a-bring-pfv-real-3nm.o one 3nm.poss-interior.space be.good-nmz Satipo Ashaninka ‘He brought a nice jar.’ Set 1 nouns regularly occur in the morphologically simple unpossessed form as compound heads, as exemplified by the Pi -naki ‘interior space’ in (6). It functions as the compound head occupying the second compound part slot (see §2.3.1 for details on the directionality of head-branching in N1+N2 compounds). (6) apiteroite i-m-pary-ee      o-moro-naki both    3m.s-irr-fall-irr 3nm.poss-hole-interior.space ‘Both will fall into a pit (lit. the interior space of a hole).’ (Kameethari Ñaantsi 2008: 33)

Pichis Ashéninka

Set 1 nouns regularly incorporate into verbs, as illustrated by the Ma opa ‘its pod’, ‘its  sheath-like leaf ’ in (7) (for details on Set 1 INs, see §2.3.2). The INs occur in ­unpossessed form. ina maroro (7) ara-pa-t-ak-e-ro 3f.a.cook-pod-ep-pfv-real-3f.o mother beans ‘Mother cooked beans in their pods.’ (Snell 2011: 329)

Matsingenka

There is another special set of nouns called here Unbound Noun Set 2. It is constituted by unbound, morphologically complex lexical units. They are usually trisyllabic



2  Genders and classifiers in Kampa (Arawak) languages of Peru

37

nouns of shape semantics, illustrated by the Sa morinte ‘oval-shaped convexity’ in (8). Example (8b) illustrates classificatory noun incorporation (see §2.3.2 for details on classificatory noun incorporation). (8) a. pi-n-tik-ashi-t-ea      morinte 2s-irr-build-appl.int-ep-irr  oval.shaped.convexity ‘Build a wall.’

Satipo Ashaninka

b. pi-n-tik-a-morinte-shi-t-ea               impaneki 2s-irr-build-ep-oval.shaped.convexity-appl.int-ep-irr  sand ‘Build a wall from sand (to play).’ c. pi-betsik-e-ro     pi-banko-morinte 2a-make-irr-3nm.o 2poss-house-oval.shaped.convexity ‘Make your house with an oval-shaped roof.’ The Set 2 nouns have opaque etymology and are usually language-specific. Some Set 2 nouns could be possessed, exemplified by the Pe and Sa tsaranka ‘a flexible covering attached to a node’. When possessed, they take possessor prefixes, exemplified by the Pe i-tsaranka shaveta (3m.poss-covering butterfly) ‘butterfly wings’. When occurring in compounds, nouns from Sets 1 and 2 often display a wide range of semantic extensions, or meanings associated with a single linguistic form. In Pi, the bound noun -patha ‘dirt, mass, semi-solid substance’ refers to the dough in (9a) and the soil soaked in blood in (9b). The Pi free forms are o-patha (3nm.poss-dirt) ‘its dirt’, ‘its mass’, and patha-tsi (dirt-unsp.poss) ‘dirt’, ‘mass’. (9) a. o-konov-ak-e-ro      tanta-patha 3nm.a-mix-pfv-real-3nm.o  bread-mass ‘She kneaded the bread dough.’

Pichis Ashéninka

b. i-pai-t-ak-i-ro kipatsi=ka iraantsi-patha 3m.a-call-ep-pfv-real-3nm.o land=dem blood-mass ‘They called this land the blood-soaked soil.’ (Kameethari Ñaantsi 2008: 193) The Unbound Set 2 and Bound Set 1 nouns with the shape and substance semantics, as illustrated by (8c) and (9a–b) respectively, are regularly combined with other nouns to produce novel words with the purpose of filling a gap in the lexicon (see §4.4 for an overview of the functions of the noun sets). Table  2 brings to the fore the most significant finding about the overlap of the ­criterial features of the two subclasses of bound nouns inflected on the pattern of obligatorily possessed nouns. The overall patterns of their morphosyntactic behaviour are essentially identical. Some noteworthy differences between the two subclasses of Regular Bound Nouns and Special Bound Noun Set 1 are manifest in N1+N2 compounding and incorporation processes. These differences are addressed next, in §2.3.

38

Elena I. Mihas

Table 2.  Morphosyntactic behaviour of Kampa noun classes Noun class/ Morphological and syntactic behaviour

Agreement gender

Possession Optionally possessed (unbound)

Obligatorily possessed Regular Bound Nouns

Special Bound Noun Set 1

Unbound Noun Set 2

Yes

Yes

Yes

Optional

Animacy

No

No

No

No

Spatial case

Yes

Yes

Yes

Yes

Plural number

Yes

Yes

Yes

Yes

‘Unspecified R’ -(n)tsi

No

varies

varies

No

Compounding

Non-productive

Yes

Yes

Yes

Incorporation

Non-productive

Yes

Yes

Yes

Modification

Yes

Yes

Yes

Yes

2.3  Kampa N1+N2 compounds and noun incorporation 2.3.1 N1+N2 compounds  This outline of Kampa compounding patterns is highly relevant to the analysis of the bound forms used as noun classification devices in Kampa. Compounding is known to involve two or more lexical units of semantic substance, exemplified by bound noun roots which can function as free forms in other contexts (Lieber and Štekauer 2009: 4–5).3 Following Lieber and Štekauer (2009), I treat Kampa compound units as free forms if they occur in discourse on their own after being inflected for possession. The N1+N2 compounds are characterized by a particular internal constituency. Endocentricity (i.e. the existence of the internal head) and right-headedness are prominent structural properties of compounds. Noun compounds N1+N2 are composed of bound and unbound roots. Any compound component could be either unbound or bound. Overall, there is a clear preference for a productive compound scheme to have a bound noun in the second compound part slot. I consider a ­compounding scheme unproductive if it is not used in novel forms referring to ­non-native artefacts, and vice versa. The attributive compound schemes in which an unbound noun occupies the second part slot N2 are generally unproductive, i.e. the constructional scheme is generally not applied to novel expressions. 3  The ongoing discussion of compounding phenomena in contemporary typologically-oriented linguistic literature acknowledges the elusiveness of the reliable criteria for defining compounding as a linguistic category (see ten Hacken 2017 for details).



2  Genders and classifiers in Kampa (Arawak) languages of Peru

39

The attributive Unbound Adj+Bound N scheme is productive, and equally productive are the schemes of Bound N1+Bound N2, expressing the ‘parts of wholes’ semantic relation, and Unbound N1+Bound N2, featuring a ‘kind of ’semantic relation. The compounding scheme Unbound N1+Bound N2 expresses a classificatory relation (specific-generic), X is a kind of Y between the compound parts. The scheme is used in the contemporary coinages of new words, exemplified by the Sa bametantsipanko (teaching-house) ‘school’(i.e. a house of teaching) and Pi tasorentsi-panko ­(deity-house) ‘temple’ (i.e. a house of the deity) in Table 3.4 An important characteristic of N1+N2 compounding is recursivity, when components compound themselves. If recursivity is taken to be an indicator of productivity (Lieber and Štekauer 2009: 820), only two compound schemes are truly productive: Unbound Adj+Bound N and Bound N1 +Bound N2. These recursive compound subtypes are exemplified by the Pe antaro-jenka-moro (big-odour-hole) ‘the hole which stinks a lot’ (about a person’s anus), and Sa i-tsenko-tsara-peta (3m.poss-pants-clothes-devalued) ‘his old pants’. Table 3 summarizes the main characteristics of Kampa N1+N2 compounds, such as the boundedness of compound parts, internal constituency, semantics, and recursivity. The compound heads in Table  3 are bolded. The abbreviation RCS stands for recursivity. As shown in Table 3, the adjectives in Adj+N compounds are restricted to the first compound part slot. The adjectives are normally unreduced, exemplified by pashine ‘other’ in Pi (10). In the attributive compound pashine ‘other’ +-poro ‘face’, the bilabial stop /p/ in -poro ‘face’ undergoes gliding, /p→w/. (10) pi-shine-t-a-i-ne-ri      eejatzi pashine-voro-ki . . . Pichis Ashéninka 2a-allow-ep-term-irr-3t-3r  also     other-face-loc ‘Allow it to him (to slap you on), the other cheek, too.’ (Kameethari Ñaantsi 2008:12) The number word plus noun construction Num+N is not included in Table 3 due to the low frequency of its tokens (see §4.2 for details). The number word construction exhibits similar preferences for the directionality of head-branching. In particular, the compound scheme Num + N is right-headed, with the head being either a bound or unbound noun. The productivity of this compound scheme is not correlated with the boundedness of the noun components. The scheme is unproductive as it is not applied to novel expressions. In Sa (11a), the second compound part kitaiteri ‘day’ is an unbound noun, a deverbal nominalization derived from the verb root kitaite 4  The semantic relation between the parts of the Sa compound bametantsi-panko (teaching-house) ‘school’ resembles that of ‘purpose’ rather than ‘a kind of ’. The argument against treating it as the purpose relation comes from a very productive Kampa-wide construction used for describing the affordances of places. This construction is formed with the instrumental applicative -ant. A place for teaching in Pe will be described as i-yome-t-ant-ai-tz-i=nta (3m.s-teach-ep-appl.inst-impers-ep-real=adv) ‘where they teach’.

40

Elena I. Mihas

Table 3.  Properties of the N1+N2 compound types N Boundedness

Internal Structure

Semantics

Examples

RCS

1

Unbound N1 + Unbound N2

endocentric, left-headed

attributive

mapi-teroina (stone-solid. no spherical.object) ‘a ball-like rock’ (Pe)

2

•B  ound N1 + Unbound N2 • Unbound Adj + Unbound N • Unbound Adj + Bound N

endocentric, right-headed

attributive

inka-nija (rain-water) ‘rain water’(Pe) antaro-tsomonte (big-belly) ‘big belly’ (Pe) antaro-naki (big-interior space) ‘big basket/pot/ gourd’ (Sa)

yes

3

Bound N1 + Bound N2

endocentric, right-headed

parts of wholes

incha-taki (plant-bark) ‘the bark of a tree’ (Pe)

yes

4

Unbound N1 + Bound N2

endocentric, right-headed

classificatory ‘a kind of ’

bametantsi-panko (teaching-house) ‘school’ (Sa) tasorentsi-panko ‘deity-house) ‘temple’ (Pi)

no

‘be light’. Bound nouns, exemplified by the bound noun -kota ‘split plank-like’, ‘piece’ in (11b), either occur after the number word, as in apite-kota (two-split.plank.like) ‘two pieces’, or are directly adjoined to the number word root and are followed by the nominalizer -ro, as in apa-kota-ro (one-split.plank.like-nmz) ‘one piece’. (11)

a. aparo-kitaiteri  i-ja-t-aj-e         i-ba-ni  Satipo Ashaninka one-day      3m.s-go-ep-term-real  3m.poss-plot-poss ‘One day he went to his plot of land.’ b. p-am-ak-e-na     apite-kota /  apa-kota-ro  2a-bring-pfv-irr-1sg.o  two-splt.plank.like/one-split.plank.like-nmz kaniri manioc ‘Bring me two/one piece(s) of manioc.’

In summary, the most significant finding concerns the interrelation between the boundedness of compound parts, their distribution, and productivity of the compounding types. The productive compound schemes listed in Table  3 require a bound  noun in the second component slot N2. The productive compounds are all endocentric and right-headed. The productive schemes tend to make full use of



2  Genders and classifiers in Kampa (Arawak) languages of Peru

41

recursivity. The bound nouns which occur in the second compound part slot N2 ­usually come from the Regular Bound Nouns subclass and the Special Bound Noun Set 1 (see §2.2, Table 2). 2.3.2  Noun incorporation  In this analysis, I will compare the distinct features of incorporated nouns (henceforth IN) which come from the the Regular Bound Nouns subclass and the Special Bound Noun Set 1 (see Table 2, §2.2). The illustrative examples come from my Ashaninka field data and Ashaninka examples extracted from secondary sources (unless specified otherwise). Overall, noun incorporation is very productive in Kampa. Incorporating verbs do not exhibit constraints due to their semantic class, although the inherent selectional restrictions are involved in that the IN is expected to be compatible with the semantic frame of the verb. There are very few constraints on INs either. Incorporated nouns come from all noun classes: non-possessables (unbound), optionally possessed (unbound), and obligatorily possessed (bound) nouns (see §2.2 for details on noun classes). Unbound incorporating nouns are exemplified by (12a–b), where the nonpossessable unbound noun kamari ‘demonic spirit’ is used to refer to a specific definite referent, a demonic vulture, the main protagonist in the narrative. In (12a–b), the syntactic functions of the INs are the intransitive subject S and transitive subject A, respectively. (12)

Unbound INs a. ari    i-kis-a-kamari-t-an-ak-a pos.pol  3m.s-ep-be.angry-demon-ep-dir-pfv-real ‘He, the demon, got angry.’ (Weiss 1975: 340) b. i-jetsi-kamari-t-an-ak-e-ro 3m.a-drag.one’s.anus-demon-ep-dir-pfv-real-3nm.o ‘He, the demon, dragged his anus against it (the tree’s surface) in order to clean the anus.’ (Weiss 1975: 341)

As Table 4 summarizes, the incorporating patterns of the bound nouns subclasses are overlapping. In what follows, I will examine the properties of bound INs which come from the Regular and Special Bound Noun subclasses. In Table 4, bound nouns from either subclass are shown to refer to animate and inanimate, indefinite and definite discourse participants. In Asha, the ranges of the participant’s semantic roles (Experiencer, Agent, Patient, Theme, Companion, Instrument, and Location) and syntactic functions of INs (A, S, O, and obliques) overlap in both subclasses. Both bound noun subclasses trigger a reduction in the argument structure of the verb when the topic is backgrounded (see Mihas 2015: 617–18 for details on noun incorporation as a topic demotion strategy in Pe).

42

Elena I. Mihas

Table 4.  Incorporating patterns of two subclasses of bound nouns Parameters

Regular Special Bound Bound Nouns Noun Set 1

Restrictions on the semantic classes of verbs

No

No

Restrictions on the semantic classes of INs

Yes

Yes

Restrictions on the semantics of noun referents (animacy, definiteness)

No

No

Change of meaning from specific to generic in INs

No

Yes

Semantic roles of INs

Overlapping

Syntactic functions of INs Recursivity

Yes

Yes

Reduction of the verb’s argument structure (as a topic demotion strategy)

Yes

Yes

Classificatory noun incorporation

No

Yes

As Table 4 shows, members of both subclasses are used recursively. The recursive use of incorporation, when two Regular Bound Nouns, -vatsa ‘flesh’, ‘body’ and -pori ‘thigh’, ‘leg’ are incorporated, is illustrated in Pe (13a). The Pe (13b) exemplifies the recursivity of incorporation with two INs which come from the Special Bound Noun Set 1, -chee ‘horn’, ‘crescent-like’ and -kii ‘stick-like’, ‘pointed’. (13)

a. ari    i-to-vatsa-t-e-ro       {i-vori}O pos.pol  3m.a-cut-flesh-ep-irr-3nm.o  3m.poss-thigh It was the case that he would cut his thigh’s flesh. i-piyo-t-ak-e-ro {osheki i-vatsa}O 3m.a-pile.up-ep-pfv-real-3nm.o a.lot   3m.poss-flesh He would pile up a lot of his flesh. airo   osamani-tz-i     i-ñ-aty-e-ro neg.irr  3nm.s.delay-ep-real 3m.a-see-prog-irr-3nm.o It wouldn’t take long for him to see that shirink-a-vatsa-vori-t-an-aj-i-Ø contract.in.size-ep-flesh-thigh-ep-dir-term-real-3s the thigh’s flesh shrank. b. a-vishiri-ako-t-a-ri        ooriatsiri [.  .  .] kashiri 1pl.a-dance-appl-ep-real-3m.o sun     moon ovakira  i-yov-a-chee-kii-t-ap-aj-i conn   3m.s-come.out-ep-crescent.like-stick.like-dir-term-real ‘We danced at the sight of the sun [..and] the moon, when the moon was new.’ (Mihas 2014: 278)



2  Genders and classifiers in Kampa (Arawak) languages of Peru

43

The backgrounding of old or established information is illustrated by (13a); it is a cross-linguistically common function of noun incorporation (Mithun 1984). In the first two sentences, the body parts ivori ‘his thigh’ and ivatsa ‘his flesh’ are introduced as independent topical NPs in transitive object function. The NPs are co-indexed on the verbs by the third person non-masculine suffixes -ro. In the last sentence, both body part nouns are treated as old information, evidenced in their incorporation into the verb. When it happens, the verb becomes formally intransitive. Argument reduction is often observed in transitive constructions, when a bound NP in O function is incorporated. As the result of this topic demotion strategy, the verb occurs in its intransitive frame. However, in classificatory compounding, the verb remains formally transitive. In (13a), the noun-incorporated transitive verb to ‘cut’ in i-to-vatsa-t-e-ro i-vori (3m.a-cut-flesh-ep-irr-3nm.o 3m.poss-thigh) preserves its two-place argument structure due to the full expression of the ‘specific’ NP ivori ‘his thigh’ in O function. The NP ivori ‘his thigh’ has a ‘classificatory’ relation to the ‘generic’ IN -vatsa ‘flesh’, i.e. the incorporated noun describes the property of the NP referent. In classificatory noun incorporation, an incorporated ‘general N stem . . . can be accompanied by a more specific external NP, which identifies the a­ rgument implied by the IN’ (Mithun 1984: 863). In Kampa, the ‘generic’-‘specific’ classificatory relation usually requires the presence of the substance or shape bound nouns in the incorporated noun slot, as exemplified by (8b), (13a), and (17). Lexical compounding, another basic function of noun incorporation, is exemplified by (13b). The compounded IN -chee-kii- (crescent.like-stick.like) is paired with the verb root to form a lexicalized combination V+N1+N2 referring to the new moon stage. Each bound noun subclass possesses distinct features with regard to the scope of their semantic domains, range of semantic extensions, and referential functions. In particular, the subclass of Regular Bound Nouns is constituted by body and plant part terms, human and animal excretions, kinship/social terms, and terms referring to valuable possessions. These INs refer to entities, rather than to the entities’ properties. In a compound, they exhibit their specific meanings. As illustrated in (14a-b), the Sa body part terms -pori ‘thigh’, ‘leg’ and -shempa ‘shoulder’ retain their basic meanings. Example (14a) illustrates the possessor separation construction (Aikhenvald 2013: 36), when the possessor R is indexed on the verb as a core grammatical relation by the suffix -ri ‘3m.o’ and is also expressed as a constituent separate from that, yora ‘that one’. Regular Bound INs (14) a. i-karaja-vori-t-ak-e-ri yoraO 3m.a-break-thigh/leg-ep-pfv-real-3m.o/r dem.nom ‘They broke legs of that one . . .’ (Irineane Tasorentsi 2008: 104) b. i-kita-shempa-t-i        i-n-i 3m.s-be.white-shoulder-ep-real 3m.s-be-real ‘He has white (stripes) on his shoulder (about an anteater).’

44

Elena I. Mihas

The Special Bound Noun Set 1 consists of nouns from the semantic domains of body and plant parts, and spatial and geomorphic terms. It partially overlaps with Regular Bound Nouns in that both subclasses include body and plant parts terms. In contrast to the INs from the Regular Bound Nouns, those of the Special Noun Set 1 are characterized by unique features such as (i) systematic polysemy of lexemes covering both narrow, specific and broad, general meanings, (ii) frequent referencing of properties of noun referents, and (iii) propensity to form classificatory compounds. I will briefly discuss each feature of the Set 1 INs below. The Set 1 nouns are notorious for their polysemy, featuring broad, general meanings in verb incorporation. The extensions of the polysemous bound noun -jenka with the basic meaning ‘body odour’ are illustrated by the Asha INs in (15). The incorp­orated noun -jenka ‘body odour’ refers to various immaterial entities such as time in (15a), smoke clouds in (15b), contaminated air in (15c), and speech sound in (15d). The syntactic functions of the INs range from S in (15a), O in (15b–c), and Oblique in (15d). (15)

Special Bound Noun Set 1 INs a. ari    tsiteni-jenca-ite-an-aqu-e-Ø pos.pol be.dark-immaterial-compl-dir-pfv-real-3s ‘It was the case that the time (of the day) changed to complete darkness.’ (Irineane Tasorentsi 2008: 31) b. oshequi o-n-cachaa-jenca-t-an-aqu-e a.lot   3nm.s-irr-smoke-immaterial-ep-dir-pfv-real ‘It will produce a lot of smoke clouds.’ (Irineane Tasorentsi 2008: 108) c. i-p-a-jenka-ak-e-mpi-ro 3a-give-ep-immaterial-pfv-real-2o-3nm.o ‘He will infect you (lit. he will give the ‘bad air’ to you).’ d. te     i-n-comi-jenca-t-ea-ro-ji neg.real 3m.a-irr-confuse-immaterial-ep-irr-3nm.o-neg yora     ashi-t-a-ro-ri      iri-nea-ne dem.nom own-ep-real-3nm.o-rel 3m.poss-word-poss ‘The owner of his language does not get confused by the speech sound.’ (Irineane Tasorentsi 2008: 153)

Next, INs formed from Set 1 often describe an entity’s property in intransitive and  transitive constructions, as in (16) and (17), respectively. In Asha (16), the incorp­orated spatial term -tsempeki ‘outer frame’, ‘skeleton-like’ describes the property of a human referent as being ‘naked’. The human participant is indexed by the subject index i- ‘3m.s’ on the verb. The noun referent can alternatively be interpreted as ‘emaciated’ or ‘defeathered’ in other morphosyntactic and pragmatic contexts.



2  Genders and classifiers in Kampa (Arawak) languages of Peru

45

(16) i-shi-a-tsempequi-t-an-ac-a 3m.s-flee-ep-outer.frame/skeleton-like-ep-dir-pfv-real ‘He fled naked.’ (Irineane Tasorentsi 2008: 50) The Special Bound Noun Set 1 members have a proclivity to form classificatory compounds, as illustrated in Sa (8b), Pe (13a), and Asha (17). A classificatory compound includes a verb root combined with a ‘generic’ noun used to describe a property of the ‘specific’ referent of an optional free-standing NP, which functions as the verb’s argument. (17)

i-quiv-a-tsempequi-yee-t-aqu-e-ro 3m.a-wash-ep-outer.frame/skeleton-like-distr-ep-pfv-real-3nm.o o-vatsa-ni 3nm.poss-flesh-ceased.exist ‘They washed her entire dead body.’ (Irineane Tasorentsi 2008: 115)

In summary, the V+Bound N scheme is very productive. When incorporated, members of the Special Bound Noun Set 1 display a number of unique properties which Regular Bound Nouns do not seem to possess. In particular, the Set 1 INs tend to exhibit general rather than basic narrow meanings, refer to noun referent’s properties rather to the noun referent itself, and are capable of forming classificatory compounds.

3  Gender system 3.1  Gender marking In this section, I will examine gender marking on agreement targets, controller nouns, and in pronouns. The Kampa way of marking linguistic gender is an old gender agreement system dating back to Proto-Arawak (Aikhenvald 1999: 84). The key function of gender marking is to accelerate cognitive processing of speech. The syntactic analyses of gender-marked grammatical constituents are facilitated through the identification of inflectional forms and syntactic relations. Kampa nouns divide into two strongly grammaticalized gender-based classes, masculine and feminine (the latter is termed ‘non-masculine’ in Na, No, and Pe grammars). Linguistic genders are distinguished in third persons via obligatory gender agreement with controller nouns, expressed formally on a wide range of agreement targets, and in some cases, on controller nouns. In other words, both ‘covert’ and ‘overt’ ways of gender marking are present. The agreement markers are ‘uniform’ in  Arawak languages, and ‘go back to the Proto-Arawak third person singular cross-referencing forms’ (Aikhenvald 1999: 84). The proto-forms are *ri-, *i-, *-ri, *-i and *thu-, *u-, *-thu,*-u (Aikhenvald 1999: 88). Gender is also expressed in third person pronouns (Table  5). Gender distinctions show no neutralization in nominals inflected for plural number.

46

Elena I. Mihas

‘Covert’ gender agreement is expressed on head nouns (via possessor formatives) in possessive NPs and noun modifiers (adjectives and nominal demonstratives), possessive pronouns, demonstrative identifiers (‘here/there it is’), verbs (via person indexes), and in some languages, on coordinating operators. The ‘covert’ gender agreement marking on various targets is illustrated by Caquinte (18a–b), Matsigenka (18c), and Pichis Ashéninka (18d–e). (18)

a. iri-nta     ir-aniritsori   i-katsima-t-ak-i-ri 3m-dem:dist 3m.p-cousin 3m.a-hate-ep-pfv-real.i-3m.o ‘That cousin of his hated him.’ (Castillo Ramirez 2017: 96)

Caquinte

b. i-shig-an-ah-a        imoroiroki-haniki   matsari Caquinte 3m.s-escape-abl-reg-real.a collared.peccary-dim thin.m The thin little offspring of peccary escaped. (Castillo Ramirez 2017: 102) c. aityo    irashi   pi-tomi  kotsiro? Matsigenka exist.inan 3m.poss 2poss-son knife ‘Does your son have a knife? (lit. is there his (knife), your son’s knife?)’ (Snell 2011: 822) d. jirika    pi-ireki-te dem.id.m 2poss-money.m-poss ‘Here it is, the money.’

Pichis Ashéninka

e. tzimatsi 5 tanta ipoña apite shima exist five bread conn.m two fish.m ‘There are 5 loaves of bread and two fishes.’ (Kameethari Ñaantsi 2008: 71)

Pichis Ashéninka

Overt gender marking on Kampa nouns is detectable in a small set of nouns with opaque semantics.5 For example, Pe kashiri ‘moon.m’ and kosaniri ‘caiman.m’ are of masculine linguistic gender. But many animate nouns of masculine linguistic gender bear the suffix -ro, as exemplified by the Pe maniro ‘deer’, sharo ‘agouti’, saavaro ‘guacamayo’, and kontsaro ‘dove’. The second group of nouns with the overtly marked gender is formed via the productive nominalizing process. The participant nominalizing forms -ri/-ro and -nto~nt(z)i are suffixed to verbal roots and -sato~sat(z)i to predominantly nominal roots. The nominalizations usually refer to human entities, e.g. Pe pinkatsari ‘chief.m’ and pinkatsaro ‘chief.nm’, sheripiari ‘shaman.m’ and sheripiaro ‘shaman.nm’, shimariantzi

5 In his analysis of the overt gender marking of animates and inanimates in the Alto Urubamba Matsigenka variety of Kampa, Emlen (2014: 324) suggests that it might be worthwhile to examine ‘a loose correlation between the gender of each species and the presence of now non-productive gender classifying suffixes (masc. -ri, fem. -ro) (Michael  2008: 295); this will provide an interesting source of data for ­diachronic comparison with other Kampan languages.’



2  Genders and classifiers in Kampa (Arawak) languages of Peru

47

Table 5.  Sa personal pronouns sets Person Number Gender

Continuous topic

Contrastive additive focus

Reflexiveintensive

Contrastive topic

Contrastive exhaustive focus

1sg

naro

narori

narosati

nainti

narotake

2

abiro

abirori

abisati

abinti

abirotake

3m 3nm

iri(o) iro

iriori~iriyori irori

irisati irosati

irinti irointi

iritake irotake

1pl

aro

arori

arosati

arointi

arotake

‘fisherman.m’ and shimarianto ‘fisherman.nm’, kipatsisatzi ‘underground dweller.m’ and kipatsisato ‘underground dweller.nm’. Pronominal gender is observed in multiple sets of Kampa personal pronouns. It is distinguished in third person singular only. There are no special free pronominal forms for plural number. In Sa pronouns in Table 5, plural number is expressed by simply adding the plural suffix -jei to the singular forms. (The contrastive exhaustive focus set cannot be pluralized.) The choice of the form for pronominal gender is obviously motivated by the meaning of the basic forms iri ‘third person singular.m (he)’ and iro ‘third person singular.nm (she/it)’. 3.2  Coding of animacy Some Southern Kampa languages formally mark a distinction between animate and ­inanimate forms on a range of NP constituents in third person singular. The animacy agreement is coded by the suffixes -ni ‘animate’ and -ti ‘inanimate’ in Ma (Snell 2011: 818–19, 824) and No (Shaver 1996: 30).6 In particular, in Ma, the animacy agreement targets include some adjectives, number words, quantifiers, interrogatives, and existential verbs (O’Hagan and Michael 2015). Northern Kampa languages lack the animacy-coding system. The Ma animacy and gender marking systems minimally intersect. Most adjectives in Ma agree with controller nouns only in gender. Few adjectives take both agreement gender and agreement animacy markers, as in (19a–b). (19) a. i-tyomia-ni 3m-small-anim ‘her small son’

o-tomi 3f.poss-son 

Matsigenka (O’Hagan and Michael 2015: 4)

b. o-tyomia-ti   koviti 3f-small-inan pot ‘small pot’ 6  The nontransparent animacy assignment and marking is illustrated by the Alto Urubamba Matsigenka chemically potent or sticky substances, trees that produce them, and new substances like gasoline and soap, which are conceived of as animate entities of the feminine linguistic gender (Emlen p.c., November 1 2017).

48

Elena I. Mihas

Examples of other animacy-coding targets in Ma include (O’Hagan and Michael 2015: 4):

• number words, e.g. mavani ‘three.anim’~mava-ti ‘three-inan’ • quantifier, e.g. maganiro ‘all.anim~magatiro ‘all.inan’ • interrogative pronouns, e.g. tyani ‘who/which.anim~tyati ‘what/which.inan’ • positive existential verb, aiño and aityo, which exhibits two forms, animate and inanimate, respectively.

3.3  Semantic bases of gender assignment Kampa nouns are allotted to the genders based on a relatively transparent gender assignment system. The assignment criteria include biological sex for human nouns (males are masculine and females are feminine/non-masculine) and animacy for non-human nouns (inanimate entities are usually non-masculine). Non-human animates are masculine by default (with a few exceptions). The gender assignment in Ashe and Asha is reinforced by the speakers’ cultural assumptions about the historical development of their universe (see also Aikhenvald 2012: 284–5 on the role of myths in gender choices in Ashe and Asha, and Emlen  2014 on semantic motivations of gender assignment among Matsigenkas). Based on the ideas of agentive and nonagentive transformations of humans which took place in primordial times, nonhuman animate and inanimate entities are assigned to a gender class depending on what sex they were in human form in the past. The gender assignment system is exemplified by Pe in Figure 4. In Pe, the myth-based motivation of gender assignment is as follows: • masculine (believed to be males in their human life): wild animals, birds, ­reptiles, fish, flying and non-flying insects, limbless invertebrates (worms), and astral bodies

Gender assignment in Pe

Biological sex

Animacy

Human males (MASC) Animate Inanimate Human females (FEM/NMASC) Animals (MASC) FEM/NMASC

MASC

Plants Celestial bodies Household items Money Geomorphic features Milk Natural elements

Figure 4  Alto Perené gender assigment



2  Genders and classifiers in Kampa (Arawak) languages of Peru

49

• non-masculine (believed to be females): plant life-forms, geomorphic features, and natural elements. Tools and household items tend to be non-masculine (but for some, kitsapi ‘needle’ is of masculine gender), as well as the newly introduced means of transportation, like cars and bikes, and technological devices, such as mobile phones, television sets, computers, DVD players, and (video) cameras. In Ma (O’Hagan and Michael 2015: 3), the assignment is motivated by the salience of the species’ biological sex: • typically masculine: mammals, snakes, fish, and social insects • typically feminine: non-snake reptiles, amphibians, molluscs, crabs, non-social insects. In Pe, Pi, and Sa, the default gender is masculine. It is used to refer to the fetus in a woman’s womb and to a crowd of people, as in Pi (20). maaroni Pichis Ashéninka (20) j-ama-ye-t-apa-ak-e-ne-ri 3m.a-bring-distr-ep-dir-pfv-real-3t-3m.r all mantsiya-ye-t-atsi-ri     i-sheninka be.sick-distr-ep-stat-rel 3m.poss-fellowman ‘They brought him all their sick fellowmen.’ (Kameethari Ñaantsi 2008: 10) There is no gender resolution. When nouns of different genders are coordinated in NK, the first conjunct adjacent to the verb will trigger the corresponding verbal person marking of linguistic gender.

4  Core Kampa multiple classifier set 4.1  Overview Kampa gender and animacy-marking systems coexist with a classifier system.7 Kampa languages have each a set of classifying forms which occur in different morphosyntactic loci.8 It means that the same form is used in a range of morphosyntactic ­contexts: 7  The membership of cognate forms in the core multiple classifiers set has been identified on the basis of quantization and qualitative analysis of items harvested from the comprehensive Ma dictionary (Snell 2011), the classic work on Ma classifiers by Shepard (1997), and other significant secondary sources (e.g. David Payne 1980, 1981; Doris Payne 1987; Michael 2008; Michael et al. 2013). For Northern Kampa data, I relied on my fieldwork on Perené, Satipo, and Pichis Kampa, and available text collections (Weiss 1975; Anderson 1985, 1986; Kameethari Ñaantsi 2008; Irineane Tasorentsi 2008; Mihas 2011, 2014, 2017a). 8  In individual languages, the distribution of some classifiers is constrained, that is not all classifiers listed in Appendix occur in various morphosyntactic contexts. For example, the Pe -pi ‘long, slender, hollow’, -pa ‘sheath-like, long, slender’ and some others do not incorporate into verbs. It could be argued that because these forms are hardly ever used on their own as obligatorily possessed nouns in Pe everyday discourse, they have lost a verb-incorporating propensity.

50

Elena I. Mihas

with a noun, an adjective, a number word, a quantifier, or a verb. The propensity of such forms for a fairly unrestricted distribution earned them a label ‘multiple classifiers’ (Aikhenvald 2000). Across Kampa, multiple classifier systems of individual languages have a shared group of cognate forms with nearly identical basic meanings, and in many cases, overlapping semantic extensions (see Appendix). The core Kampa classifiers predominantly categorize noun referents in terms of their shape (19 out of 29 forms). Other classification parameters include substance (5 forms) and arrangement (5 forms). The majority of multiple classifier forms come from the Special Bound Noun Set 1 of the obligatorily possessed bound nouns (see §2.2, Table 2). In NK, they are characterized by the following features (see §2.3 for details and examples): • association with multiple semantic extensions • occurrence in the second compound part slot in the N1+N2, Adj+N, Num+N, and V+N compound types • tendency to serve as heads of compounds (except the V+N type, where the verb is the head) • recursivity of the Bound N1+Bound N2 and Adj+Bound N compound types • application of all compound types to novel expressions • as INs, they often exhibit broad, general meanings, tend to describe a referent’s property, and possess a proclivity to form classificatory compounds. To indicate the classifying function of the Special Bound Noun Set 1, I will use the abbrevation cl ‘classifier’ in the glosses of the examples in the remainder of the chapter. The membership in the multiple classifier set is not restricted. As comparison across individual Kampa languages indicates, it is an open class. For example, the classifying form -pio~-vio ‘pile’ (#27, Appendix) appears to be used only in SK languages. NK languages employ -poroki ‘assortment’ (#28, Appendix) to convey an idea of a pile. Some mulitple classifiers originate from verbs. In particular, the verb root -na ‘fit’ is hypothesized to be the source of the Set 1 bound noun onaki ‘its interior space’ (see Example (5)). The evidence comes from Ma and Sa data. In Ma, the verb naki(tagantsi) means ‘to have the form of empty space within something’; o-naki-t-ak-e (3f.s-have.empty. space-ep-pfv-real) ‘it has the form of empty space’ (about a room, a hollow tree, a cave, the interior of a pot or drum) (Snell 2011: 290). In Sa, the verb na means ‘to fit’, exemplified by o-na-ja-t-ant-ea-ri (3nm.s-fit-cl:liquid-ep-appl.inst-irr-rel) ‘a vessel for water’ (Kindberg 1980: 198). Another pan-Kampa verb root, porok ‘break into small pieces’, has served as a source of the Set 1 bound noun -poroki with the basic meaning ‘crumbs’. For instance, the NP i-poroki shima (3m.poss-particles fish) ‘fish particles’ refers to the crumbs left after the smoked fish is consumed. The bound noun -poroki has a variety of semantic extensions, referring to acne, coagulated blood drops, and sand particles. The term is reported in Ma (Snell 2011: 398), and is widely used in Pe and Pi discourse. Example



2  Genders and classifiers in Kampa (Arawak) languages of Peru

51

(21) illustrates various morphosyntactic loci of the classifier -poroki ‘assortment’: (21a) is an example of the use of the lexical verb porok ‘break into pieces’; in (21b), the Bound N1+Bound N2 compound type is shown, followed by (21c) and (21d), exemplifying the Adj+Bound N and V+Bound N compound types, respectively. (21)

a. aña eeroka   porok-i-ro-ri Pichis Ashéninka dp   2top   break.into.pieces-real-3nm.o-rel  tasorentsi-panko                 deity-house ‘Wow, you are the one who is going to destroy the temple.’ (Kameethari Ñaantsi 2008: 58) b. i-tonki-poroki       kaminkari janta 3m.poss-bone-cl:assortment  corpse   dem.adv ‘A group of the corpse’s bones is over there.’ c. ari    j-am-ai-t-ak-e-ne-ri pos.pol 3m.a-bring-impers-ep-pfv-real-3t-3m.r kiso-poroki-ri join.tightly-cl:assortment-nmz ‘They brought for him a paralysed man.’ (Kameethari Ñaantsi 2008: 18) d. te      i-n-kov-e      iri-ñe-e-ro neg.real 3m.s-irr-want-irr 3m.irr-see-irr-3nm.o o-m-pitsi-poroki-vai-t-e 3nm.s-irr-be.dirty-cl:assortment-dur-ep-irr ‘He does not want to see dirty spots on her.’ (Kameethari Ñaantsi 2008: 18)

Multiple classifiers show semantic agreement with the controller noun. The use of classifiers is restricted by the salient physical properties of the noun referent. Shepard (1997: 37) illustrates this point with Ma examples in which classifiers are combined with the number word paniro ‘one.anim’: • -tsa ‘liana-like’,‘curvilinear’ is selected to describe a snake, pa-tsa-ni-ro maranki (one-cl:liana-like-anim-nmz snake) ‘one snake’ • -shitsa ‘vein’, ‘string-like’ describes a squirrel in pa-shitsa-ni-ro tsigeri (onecl:string-like-anim-nmz squirrel) ‘one squirrel’ • -empe ‘branch’, ‘limb’ classifies a spider monkey as an animal which grabs tree branches with its prehensile tail, pa-empe-ni-ro osheto (one-cl:branch-animnmz) ‘one spider monkey’. Unique classifiers classify particular features of language speakers’ material culture and spiritual beliefs. The NK classifier -shiteki ‘devalued, rag-like’ comes from from oshite(ki) ‘its wood shavings’, as in Pe (22). It is sometimes used as a racial slur referring to Andean colonists. Its use reflects the lower social standing of outsiders in Kampa society.

52

Elena I. Mihas

a-ak-i-ro=ra chora-shiteki Alto Perené (22) o-kov-i 3ms-want-real 3nm.a.take-pfv-real-3nm.o=sub colonist.nm-cl:devalued ‘She wants it when she is married (to an Ashaninka man), a useless Andean female colonist.’ Kampa multiple classifiers are not gender-sensitive. Neither are special classifying forms reported for making a distinction in (in)animacy. Classifiers are sometimes stacked up to achieve precision in the object’s description. Classifier strings are found in plant names, place names, and names of artefacts. The ordering of classifiers usually follows the putative order of their acquisition: the ones which are reported to be universally acquired by the children early on are characteristic of the referent’s shape; others are  acquired at a later stage of the children’s development (Clark 1973; Tran  2011). Those acquired early are usually placed further away from the root, whereas the ones acquired later normally occupy the slot closer to the root. In Sa (23), the Set 1 noun -peta describes the contour of the speaker’s robe. The noun has the meaning ‘flattened, flaccid fruit’, exemplified by a crushed plantain, or ‘deflated hanging pouch’, illustrated by a breast. The Sa Set 1 noun -peta has another extension with arrangement semantics, namely ‘stack’, ‘roll’, ‘bundle’. The Set 2 noun kamara denotes its brown colour (see §2.2 for details on the Set 1 and Set 2 noun subclasses). The Sa term kamara ‘cinnamon colour’ has a negative value connotation, usually referring to a dirty and stinky piece of clothing. The shape term -peta ‘flattened, flaccid fruit’, ‘deflated hanging pouch’ features a negative value extension ‘devalued’, ‘useless’. The classifier kamara ‘brown colour’ is placed closer to the root, whereas -peta ‘devalued’ occupies the slot away from the root. no-ma-kamara-peta      Satipo Ashaninka (23) pi-m-p-e-na 2a-irr-give-irr-1sg.o 1sg.poss-robe-cl:dirty-cl:devalued ‘Give me my dirty useless robe.’ Not all Kampa lexicon is classified. In Pe, abstract nouns tend to be unclassified, exemplified by yo-t-antsi (know-ep-nmz) ‘wisdom’, ‘knowledge’ and kam-a-ni-t-antsi (die-ep-intns-ep-nmz) ‘death’. These nouns are all deverbal nominalizations. Classifiers have a basic individuating function when used with number words. The Num+N number word construction is employed in direct ostension situations when a physical demonstrable target is covered by the term that is being pointed at. For example, in (24), the speaker is pointing to the nylon material while asking the addressee to use three nylon threads for making a clothes line to hang the laundry. Because of the shared background knowledge, the addressee does not require a full noun phrase to disambiguate the topical referent. mava-tsa               Alto Perené (24) pi-vetsik-e-ro 2a-make-irr-3nm.o three-cl:curvilinear ‘Make three (threads).’



2  Genders and classifiers in Kampa (Arawak) languages of Peru

53

4.2  Frequency of classifier use This analysis focuses on the frequency of classifiers in discourse, and association of frequent classifier use with the productive compound schemes and referent topicality. The assessment of classifier use was conducted on the basis of the count of tokens in the Pe and Ma texts. Over 406 classifier tokens were harvested from 3,473 sentences of the Pe texts (Mihas  2011,  2014). The Ma text collection contains 1,113 classifier tokens in 5,261 sentences of text (Michael et al. 2013). The density of classifier use is higher in Ma discourse: on average, there is a classifier every 4.72 sentences; in Pe, it is every 8.5 sentences. In the Pe sources, the Bound Noun1 + Boun Noun2 scheme has produced 62% of all Pe classifier tokens (252 tokens) (see §2.3.1 for details on productive compound types). The V+Bound N scheme has generated approximately 38% of the total number of tokens (154 tokens). The most frequently encountered classifiers in the Pe texts were -aa ‘liquid’, -mashi ‘wide, flat (space)’, and -shi ‘leafy substance’. In the Ma text (Michael et al. 2013), the most frequent classifier is -a ‘fluid’, followed by -gite ‘sky’, ‘environment’, and -se ‘mass’. Pe classifiers are commonly encountered in introductions and reactivations of inanimate topical referents. The exuberant use of the productive scheme Bound Noun1 + Bound Noun2 is associated with the topicality status of the noun referent (for details on the functions of classifiers, see §5). An illustrative example comes from a conversation between two Pe speakers discussing old agricultural techniques and tools, cited in the Pe grammar (Mihas 2015: 670–2). In the 23-sentence text, which contains 29 noun phrases, four classifier tokens are attested. They all are generated by the Bound N1+Bound N2 compound scheme. The first token occurs in Sentence 1 of the conversation, when the spatial term -mashi ‘wide, flat (space)’ in i-yaniri-mashi (3m.poss-manioc-cl:wide.flat) ‘their manioc garden’ is used to introduce a new topic. In the subsequent text, the argument index -ro ‘3nm.o’ is employed for referent-tracking, but no other explicit references are made to the manioc garden via a full NP. In Sentence 16 (reproduced in Example 25), while describing the practices of manioc cultivation, Elias introduces a new topic, the cutting tool used by his kinfolk for ­making indentations in the ground in the old times. He uses the Pe shape classifier -peta ‘flattened, flaccid fruit’, ‘deflated hanging pouch’, which also has an extension ‘devalued’, in the full NP kotsiro-peta (knife-cl:devalued) ‘old knife’. The token is repeated two more times, when Gregorio asks for a clarification, and Elias responds affimatively. (25) E

poña y-a-ak-i-ro o-tzim-ak-i=ra conn 3m.a-take-pfv-irr-3nm.o 3nm.s-exist-pfv-real=adv Then they would take ‒ as it existed [then] ‒ → kotsiro-peta knife-cl:devalued an old knife.

54

Elena I. Mihas G → kotsiro-peta? knife-cl:devalued An old knife? E → kotsiro-peta=ja knife-cl:devalued=n.contr The old knife.

Constructions with number words are sometimes employed for low animacy referent introductions, especially if these referents are edible entities.9 The construction is based on the unproductive compound scheme, Num+Bound N. Based on the available limited evidence, introductory constructions with number words appear to be more common in SK than in NK. The low numbers of the illustrative tokens (less than half a dozen) in the surveyed Pe and Ma texts point to a highly restricted set of specific discourse situations when the Num+Bound N construction is employed.10 In Ma (26a), the speaker is introducing an animate topical referent, an edible caterpillar species, while reporting on his visual discovery. In Sa (26b), the speaker begins her report about a fishing outing with an introduction of a new topic, the fish that she caught. pa-tsa-ni ro shigopa (26) a. no-ne-ak-i 1sg.s-see-real.i one-cl:liana-like-anim-nmz caterpillar.sp ‘I saw one (liana-like) shigopa.’ (O’Hagan and Michael 2015: 6) b. n-a-ak-e      apa-sama-ro   shima    Satipo Ashaninka 1sg.s-take-pfv-real one-CL:unit-nmz fish ‘I caught one (unit of) fish.’

9  Mary Ruth Wise, a highly regarded Kampa scholar, notes that in Matsigenka and Nomatsigenka, c­ lassifiers ‘are used with numbers referring to inanimate things [rather than animate entities-EM] much more frequently’ (p.c., September 27 2017). 10  The Kampa culture-specific practices might serve as a basic premise towards the explanation of the ‘invisibility’ of the Kampa Num+N construction. The lacuna of the Num+N tokens in the old sources (Delgado 1896, Sala 1905, Touchaux 1909, and Pio Aza 1924) and their low count in the recently collected corpus could be associated with the community-wide linguistic-cognitive practice of the infrequent employment of counting routines and imprecision and approximation in counting discrete items (see Aikhenvald  2012: 355–9 for details on the absence of the counting routine among the speakers of most Amazonian languages). The published works cite multiple lexemes which illustrate three morphosyntactic loci of the classifier use—noun, adjective, and verb—except number words and quantifiers. Mary Ruth Wise notes that classifiers are used more frequently with nouns than with number words in SK discourse (p.c., September 27 2017). A focused study of counting practices among the Mundurukú, an Amazonian people of Brazil (Pica et al. 2004), has revealed that the Mundurukú have words for one, two, three, and four, and ‘hand’ is used for five or more items; they have a rudimentary ability to count on their fingers. However, counting is rarely used. The scholars conclude that ‘by requiring an exact one-to-one pairing of objects with the sequence of number words, counting may promote a conceptual integration of approximate number representations, discrete object representations, and the verbal code. [. . .] This “crystallization” of discrete numbers out of an initially approximate continuum of numerical magnitudes does not seem to occur in the Mundurukú’ (2004: 503). The absence of the counting routine and crystallized number terms in Kampa languages might explain the low token numbers of the Num+N construction in Kampa discourse.



2  Genders and classifiers in Kampa (Arawak) languages of Peru

55

Human referents are infrequently introduced in the constructions comprising classifying forms. It happens when individuals are involved in joking or antagonistic relationships. In Pe (27), the classifier -poroki ‘assortment’ is optional, being used for sheer fun. It is used twice, being combined with a noun, noimi ‘my husband’ and incorporated into the verb saik ‘be at’. ant-e-ro-ni no-imi-poroki=ra     Alto Perené (27) iri 3m.top make-irr-3nm.o-rel.irr 1sg.poss-husband-cl:assortment=dem jiri i-saik-a- poroki-tz-i=ra dem.id 3m.s-be.at-ep-cl:assortment-ep-real=adv ‘He is the one who will make it, my skeleton-like husband. Here he is, he sits there, a bag of bones.’ In summary, the most important finding concerns the employment of classificatory forms in introductions and reactivations of inanimate topical referents. The use of the productive scheme Bound Noun1 + Bound Noun2 is correlated with the topicality status of the noun referent. In addition, the unproductive Num+N construction is sometimes employed for low animacy referent introductions, especially if the referents are edible entities. Human referents are classified in introductions, when the speaker and the addressee are in a joking (or antagonistic) relationship. 4.3  Semantics of core Kampa classifiers Three basic semantic parameters of noun classification are selected for the analysis of the core Kampa classifier set. Core Kampa classifiers are divided into three groups on the basis of shape, substance, and arrangement parameters, as summarized in the Appendix. The referent’s shape is classified on the basis of its geometry and size (length, width, and breadth, curved or linear boundaries, hollow or solid spaces, and pointedness). The object’s substance is a salient physical property as referent types are described with regard to their liquid, semi-liquid, semi-solid, powdery, leafy, or immaterial nature. The arrangement parameter characterizes an internal composition of the referent in question, i.e. whether the object is branching from one node, split, assorted, or unitized. Most classifiers have transparent etymologies. The semantic profile of the core classifier system shows no commitment to a particular semantic parameter. The Kampa multiple classifier system could be argued to have a semantically diffuse profile exhibiting no clearly defined semantic orientation. Overall, most core Kampa multiple classifiers possess a preference for describing inanimate referents, but some also cover animate referents such as insects and ­animals. Classification of human referents is infrequent. 4.4  Functions of Kampa classifiers Classifiers matter a great deal in Kampa social interactions. The main purpose of their use is pragmatic, not syntactic. Table 6 summarizes the main functions of classifiers.

56

Elena I. Mihas

Table 6.  Functions of classifiers Domain

Function

Scheme

Examples

Discourse/

Topic introduction/reactivation

N1+N2

25

Pragmatics

Reference-tracking

V+N

28

Individuation

Num+N

24, 26a-b

Disambiguation

any

23

Vividness

any

27

Lexicon

Formation of new words

N1+N2; V+N

8c, 21c

Grammar

No agreement with head noun

29a-b

Significant correlations are found with regard to the compound schemes (see §2.3.1) and their functions. In particular, the Bound N1+Bound Noun2 scheme is associated with topic introductions and reactivations, the Num+Bound N scheme with the individuating function, and the V+Bound N scheme with reference-tracking. In everyday discourse, classifiers are often used as an economical means of introduction of an inanimate topical participant, as illustrated in (25). Another important function is keeping track of referents mentioned in prior discourse. In Asha (28b), the shape classifier tsantsa ‘length’ is an anaphor referring back to the antecedent iitsaare ‘their clothes’. (28) a. iri-sheme-t-aca-ye-ro 3m.a-be.proud-ep-appl.soc.caus-irr-3nm.o   ‘They will be proud of their clothes.’

i-itsaa-re    Ashaninka 3m.poss-clothes-poss

b. i-n~quenaquena~tsantsa-t-aca-ye-t-e-ro 3m.a-irr~walk.redup~cl:length-ep-appl.soc.caus-ep-distr-ep-irr-3nm.o ‘They will walk around in long robes.’ (Irineane Tasorentsi 2006: 94) Classifiers have an individuating function when used with number words, as in (24) and (26a-b). They are optionally used in joking and antagonistic discourses for ­rhetorical purposes and better listener engagement, to kick it up a notch, as in Pe (27). Classifiers are also paired with verbs in classificatory compounds to specify the property of the specific noun referent. This technique is used for disambiguation. The classificatory relation pertains between the ‘specific’ free-standing noun, which is the verb argument, and a ‘generic’ IN. For example, in Pi (17), the incorporated ‘generic’ noun -tsempeki ‘outer frame’, ‘skeleton-like’ describes the shape property of the ‘specific’ noun referent ovatsani ‘her dead body’. Formation of new words is another task assigned to classifying forms. In Sa, a computer mouse might be called ashiro-kitso (metal-cl:oval-shaped), and a laptop ashiro-pata (metal-cl:flat.oblong). A significant function is assistance with organizing the speakers’ world into parametrized systems



2  Genders and classifiers in Kampa (Arawak) languages of Peru

57

and identifying objects as particular entities (Appendix). The existence of unique classifiers provides a window into the language group-specific ways of categorization of the universe, as in Pe (22). Kampa multiple classifiers do not have a grammatical agreement function.11 The evidence in support of this statement comes from NK verbless clauses which display non-obligatory classifier use. In verbless clauses, there is no restriction on the locus of classifier marking. In contrast, in Murui (Witotan), where the agreement is mandatory, the classifier form is ‘copied’ onto the adjective in the VCC (verbless clause complement) function (Wojtylak 2017). In Pe (29a–b), the classifier -taki ‘peel’, ‘covering’ occurs on either constituent, VCS (verbless clause subject) or VCC (verbless clause complement). (29) a. pariantzi-takiVCS kitamaroVCC             Alto Perené plantain-cl:peel white ‘The plantain’s peel is white.’ b. pariantziVCS kitamaro-takiVCC plantain   white-cl:peel ‘The plantain has white peel.’

5  Multiple classifiers as an areal feature This analysis argues that the multiple classifier system is an areal feature, manifested in some circum-Kampa languages. Evidence comes from the geographically adjacent Arawak languages, such as Piro/Yine, Amuesha/Yanesha’, Apurinã, and some others. The close neighbours Chamicuro (Arawak) and Shipibo-Conibo (Panoan) have not been reported to have classifiers. Neither Piro/Yine nor Apurinã possesses a multiple classifier system, but Piro/Yine has productive classificatory noun incorporation (Hanson 2005), and Apurinã has productive Noun1 + Noun2 compounding (Facundes 2000: 166–96). Among the circum-Kampa languages, Amuesha/Yanesha’ (Ya) is a strong contender for being counted as a member of the multiple classifier area. It is the closest neighbour of the Pe language/ethnic group. Ya has no grammaticalized gender distinction in third person singular (Adelaar and Muysken 2006: 425), but it has productive compounding of the types N1+N2, Adj+N, and V+N (Adelaar with Muysken 2006: 425–6). Ya texts (Duff-Tripp 1997, 1998) are suggestive of the existence of a multiple classifier system, with a single set of classifier forms used in various loci. Ya classifying forms are cognate with those of Kampa, as seen in the Appendix. The forms occur in nouns, adjectives, number words, and verbs, as in (30). Unlike Kampa 11  The absence of the syntactic agreement rules governing the behaviour of classifiers could be argued to point to the shallow history of the Kampa multiple classifier system. Further detailed synchronic and diachronic comparative studies of Kampa languages are required to fully investigate the grammaticalization cline of the Kampa classifier system.

Elena I. Mihas

58

bound nouns which exhibit zero formal reduction in compounding schemes, Ya classifiers normally appear in reduced form when adjoined to noun and verb roots. (30) a. aser-n-o’     ? metal-big-cl:tube  ‘big metal tube’

Yanesha’ (Payne 1987: 33–5)

b. ne-rray-at     arrap̃ta’met ̃ ‘flat’ my-grate-cl:flat  ‘my grater’ c. toor-rr-tsatse’teen     arrorot ̃ ‘hollow’ open-cl:hollow-aspect  ‘He opens the kettle (when he shouldn’t).’

d. ma’p-pan   aspan ‘leaf ’ three-cl:leaf  ‘three leaves’

6 Conclusions All Kampa languages have genders and classifiers. SK languages have three systems of noun categorization, with an additional animacy-marking system. The systems do not intersect, as summarized in Table 7.

Table 7.  Summary of gender and classifier systems Parameters

Gender system

Multiple classifier system

Origin

Proto-Arawak third person singular verbal person markers

• Special set of (obligatorily possessed) bound nouns • Unbound nouns • Verbs

Distribution

Non-exhaustive Exhaustive (adjectives, (nouns, verbs, number words, demonstratives, possessor NP, possessive pronouns, demonstrative adjectives) identifiers, personal pronouns, verbs, and coordinating operators)

Degree of Mandatory syntactic agreement grammaticalization

No agreement

Diachronic evolution

• Open system (new members are added in each language) • Relatively stable (the nouns have retained their basic meanings)

• Closed system • Stable system



2  Genders and classifiers in Kampa (Arawak) languages of Peru

59

The origin and diachronic development of genders and classifiers are different. Gender agreement morphology comes from the pan-Kampa verbal person markers. The sources of multiple classifiers are bound nouns inflected on the pattern of obligatorily possessed nouns, unbound nouns, and bound verb roots. The employment of bound and unbound nouns as noun classification devices is associated with their exuberant polysemy. Many nouns from Sets 1 and 2 (Table  2) exhibit semantically ‘generic’ senses when they occur as free-standing NPs. For example, in the Sa ­examples (4) and (5), the NPs ominka and onaki refer to the wooden substance of the canoe and to the interior space of the jar, respectively. Yet they continue to display their basic meanings, as exemplified by the plant part term ominka ‘its trunk/stalk’ in Sa (3). Overall, the etymologies of the sources are usually transparent. The transparency of the sources is suggestive of the system’s innovative nature (see also footnote 11). The gender system is strongly grammaticalized. Gender marking is mandatory and exhaustive, being reflected in the agreement marking on noun modifiers (adjectives, demonstratives, possessor NP), possessive pronouns, demonstrative identifiers, personal pronouns, verbs, and coordinating operators. Multiple classifiers show less ­exuberant distribution, occurring on nouns, verbs, number words, and adjectives. Classifiers are neither sensitive to gender nor animacy. Classifiers are semantically motivated, showing semantic agreement with controller nouns. The multiple classifier system does not participate in the syntactic parsing of constituents via morphological agreement. The main purpose of their use is pragmatic (Table 6). Both systems are stable. The Kampa gender marking shows no sign of erosion. The multiple classifier system is stable, because most nouns continue to be used in their basic sense and occur as free-standing constituents in a clause. They occur in unreduced form in all morphosyntactic environments. The gender system is closed. In contrast, the membership in the multiple classifier system is open, as Kampa languages innovate and introduce new classifying devices. The pan-Kampa system of multiple classifiers is an areal feature.

Appendix: Core Kampa multiple classifiers Form

Gloss

Source/ Objects Classified Independent form Inanimate

Animate

Shape *aki ‘seed’

1

-ki~-gi (Ma) -ki (Sa, Pe, Pi)

seed-like, small, hard, rounded

2

-kii~gii (Ma) -kii (Sa, Pe, Pi) -kiji (Ca)

i/okii ‘its stick’ stick-like, (Ma), with a Ø (Pe, Pi) sharp, pointed end

small fruit, grain, stone, tooth, coin bone, stick, stalk, long leg, corpse, spindle

small bird, insect

Elena I. Mihas

60 Form

Gloss

Source/ Objects Classified Independent form Inanimate

Shape 3

oval-kitho shaped, (Pi, Paj) grain-like -kitso (Sa, Pe) -kicho~-gicho (Ma)

okitho ‘its grain’ (Pi)

grain, small fruit, pebbles

4

-mashi

wide, flat

mashitsi ‘skin’, imashi ‘its hide, skin’ (No), omashi ‘its dry leaf of maize’ (Pe)  

flat, wide area of land, lancet blade; wide nose or face of a human or animal; flattened dead snake

5

-meni

wedge-like, flat, slender

omeni (Ma, Sa, Pe) ‘its blade’, menitsi (Paj) ‘a blade’

blade of machete, knife, oar

6

-menta

flat, slender; with an inclined plane (Pe)

imenta ‘its comb’ (of a bird) (Pe), omenta ‘its buttress root’ (Paj), omenta ‘flat and thin object’(Ma)

door, table, knife; hill crest, stone, buttress root of a tree; acquiline nose, rooster comb

7

-minka

cylinderlike, elongated

ominka ‘its ear of corn’, ‘its trunk’ (Pe); iminka ‘his neck’ (Ma)

stalk, tree trunk, pole, ear of corn, neck; wooden substance

8

-monki

rounded, concave; suspended (Pe);

omonki ‘its rounded fruit’ (Ma), omonki ‘its bulge’ (Pe)

small dirt pile, backpack, pregnant woman’s belly

9

-naki

convexity, interior space

onaki (Ma) ‘its hole’, onaki (Paj) ‘its empty container’, inaki ‘its shell’(Paj)

space inside a pot, tambour, wasp nest, cavity in the ground; snail, armadillo shell; human skull

10

-nampi

side, boundary

onampi ‘its boundary’

boundaries of trails, settlements, plots of land

Animate



2  Genders and classifiers in Kampa (Arawak) languages of Peru Form

Gloss

Source/ Objects Classified Independent form Inanimate

61

Animate

Shape 11

-pa~va (Ma) -pa (Sa, Pe, Pi, Paj)

sheath-like, long and slender

12

-pana~-vana wide green leaf-like, (Ma, Sa) -pana~wana flat, broad (Paj) -pana (Pi, Pe) -pan~Vpn (Ya)

13

-pi~-vi (Ma) -pi (Sa, Pe, Pi) -Vph (Ya)

14

opa ‘its sheath, skin, corncob, avocado, plantain with its skin, husk’ (Ma, Sa), communal house opa ‘its shoot’ (Pe, Paj) opana ‘wide green leaf ’ (Ma, Sa, Pe, Pi)

leaf, piece of paper, paper money; types of green leaves used for wrapping food

long, slender, hollow long, thin

opi ‘its stem of cane’, ‘its shoot’(Sa)

stems of sugar cane, reed; arrow shaft, needle; vein

-poa~-voa (SK only?)

trunk-like, large, cylindrical

opoa ‘its trunk’ (Ma)

trunk, pole, manioc tuber

15

-teni, -tenka

convexity, depression (curved inwards)

oteni ‘its interior’ (Ma, Sa), otenkana ‘its interior’(Pe)

space inside a house, floor of a valley, ravine

16

-taki

covering

otaki ‘its bark’ (No), itaki ‘beetle’ (Ma)

tree bark, shell of nuts, peel of fruit, shell of insects and animals

beetles, armadillos, turtles

17

-tsa (Ma, Sa, Pe) -tha (Pi, Paj) -Vxh (Ya) rope, root

curvilinear

otsa ‘small root’ (Ma), itsa ‘edible worm’ (Pe)

roots, vines, cords, strings, rope, thread, power line, shoulder strap

snakes, worms, maggots

18

tsantsa~ thantha

length

tsantsa ‘(its) length’ long nail, watercourse, (Pe, Sa, Ma) trail, robe, house

19

-tsei (Ma) -tsei~tsee~chee (Sa) -chee (Pe, Pi)

spiny, thorny, sharp, pointed

i/otsei ‘its sharp end’ (Ma), ochee ‘its spiny thorn’ (plant) (Sa), ichee ‘its horn’ (Pe), cheentzi ‘a thorn’ (Paj)

thorn, cutting edge of machete, pencil, new moon; animal horn, porcupine quills

small animal body, fish

Elena I. Mihas

62 Form

Gloss

Source/ Objects Classified Independent form Inanimate

Animate

Substance 20

-a~-aa~-ja -aa (Paj, Pe, Pi)

liquid

oja ‘liquid’(No), ija pitsi (3m.poss. liquid honey bee) ‘bee honey’ (Sa)

river outlets; types of drinks, honey; tears, dew; human/animal eyeball (Ma)

21

-jenka~enka

immaterial

o/ijenka ‘his/her/ its body odour’ (Sa)

stench, gas, vapour, air, sound, daylight, fire

22

-pane~-vane (Ma, Sa); -pani-~vani (Pe, Pi); -poeñ~Vpñe (Ya)

powdery substance

i/opane ‘its dry grainy edible mass’ (Ma), opani ‘its grainy mass’ (Pe), panetsi ‘dust’ (Pe, Paj) pañe’te’‘dust’ (Ya)

types of flour; small grainy things, dust, sand

23

-patsa~-vatsa semi-solid substance (Ma, Sa); -patsa (Pe), -patha (Pi, Paj); -paats (Ya)

opatsa ‘its soil’ (Sa), opatha ‘its dirt’ (Paj)

tar, clay, mass of pounded barbasco, mashed potato, manioc; brain matter of humans and animals

24

-shi

leafy substance

oshi ‘its leaf ’(Sa)

vegetation, hunting blind, weeds, garbage

branching from one node

ijempeki ‘its limb’ (Sa), oimpiki ‘tree limb’ (Pe)

plant branch; crab limbs; human fingers

Arrangement monkey with a prehensile tail

25

-empe (Ma) -jempeki (Sa) -impiki (Pe) -empiki (Pi)

26

-kota (Pe, Sa) split -kota~-gota plank-like, (Ma) small, flat, thin

okota ‘its split plank’ (Paj) okota ‘its piece’ (Ma)

sole of a foot, small piece of solid food

27

-pio~vio (Sa, Ma)

pile

ivio ‘anthill’ (Sa)

pile of dirt, stones, manioc

28

-poroki~voroki~ -veroki (Ma) -poroki (Sa, Pe, Pi)

a group of tiny objects; assortment of objects

iporoki shima (3m. poss.crumbs fish) ‘crumbs of fish meat’(Ma, Sa, Pe)

smoked fish and bread bony person crumbs, acne, blood clots, grains of sand; breakable stones, vegetables, words, any assorted discrete items

group of animals, people



2  Genders and classifiers in Kampa (Arawak) languages of Peru Form

Gloss

Source/ Objects Classified Independent form Inanimate

63

Animate

Arrangement 29

-sama

oval, extended; unit of something

osama ‘its body’(about manioc) (Sa)

manioc, nests of wasps and termites, wood log, flute, canoe

body of a little girl, small animal, bird

Acknowledgements I express my profound gratitude to the speakers of Perené, Pichis, and Satipo Kampa languages who contributed to this research in 2009–16. The fieldwork was supported by the National Science Foundation Grant BCS-0901196, the Endangered Languages Documentation Program Small Grant 0002, the Firebird Foundation for Anthropological Research Grant, the Foundation for Endangered Languages Grant, Jacobs Research Funds Grant, two James Cook University Faculty Grants, and by the Language and Culture Research Centre of James Cook University through the Australian Laureate Fellowship awarded to Alexandra Y. Aikhenvald by the Australian Research Council. I thank Sasha Aikhenvald for the useful comments on the drafts, and Mary Ruth Wise for sharing her analyses of Yanesha’ data. I am also appreciative of the helpful comments made by Nick Emlen, Zachary O’Hagan, Tom Durand, and Richard Alexander Castro Mamani on the final draft.

References Adelaar, W. and P.  Muysken. 2004. The languages of the Andes. Cambridge: Cambridge University Press. Aikhenvald, A.  Y. 1999. ‘The Arawak language family’, in R.  M.  W.  Dixon and A.  Y. Aikhenvald (eds), The Amazonian languages. Cambridge: Cambridge University Press, 65–106. Aikhenvald, A. Y. 2000. Classifiers: A typology of noun categorization devices. Oxford: Oxford University Press. Aikhenvald, A. Y. 2012. The languages of the Amazon. Oxford: Oxford University Press. Aikehnvald, A.  Y. 2013. ‘Possession and ownership: A cross-linguistic perspective’, in A.  Y.  Aikhenvald and R.  M.  W.  Dixon (eds), Possession and ownership: A cross-linguistic typology. Oxford: Oxford University Press, 1–64. Anderson, R. J. 1985. Cuentos folklóricos de los asheninca (Comunidades y Culturas Peruanas 18, 19). Vol. 1–2. Yarinacocha: Instituto Lingüístico de Verano. Anderson, R. J. 1986. Cuentos folklóricos de los asheninca (Comunidades y Culturas Peruanas 20). Vol. 3. Yarinacocha: Instituto Lingüístico de Verano.

64

Elena I. Mihas

Anderson, R. J. 2002. Historias de cambio de los ashéninka (Comunidades y Culturas Peruanas 29). Lima: Instituto Lingüístico de Verano. Castillo Ramirez, A. 2017. ‘Aspectos de la frase nominal en caquinte (campa-arawak)’. Tesis de licenciatura, Universidad Nacional Mayor de San Marcos. Delgado, E. 1896. Vocabulario del idioma de las tribus campas. Lima: Imprenta liberal. Dirks, S. 1953. ‘Campa (Arawak) phonemes’. International Journal of American Linguistics 19(4): 302–4. Duff-Tripp, M. 1997. Gramatica del idioma yanesha’ (amuesha). Lima: Ministerio de Educación, Insitituto Lingüístico de Verano. Duff-Tripp, M. 1998. Diccionario yanesha’ (amuesha)-castellano. Lima: Ministerio de Educación, Insitituto Lingüístico de Verano. Emlen, N. Q. 2014. ‘Language and coffee in a trilingual Matsigenka-Quechua-Spanish frontier community on the Andean-Amazonian Borderland of Southern Peru’. Ann Arbor: University of Michigan dissertation. Facundes da Silva, S. 2000. ‘The language of the Apurinã people of Brazil (Maipure/Arawak)’. Buffalo, NY: University of New York dissertation. Farmer, S. and Z. O’Hagan. 2014. ‘Nominal reference in a tiered lexicon: A semantic account of noun classifiers in two Amazonian languages’, paper presented at the Syntax and Semantics Circle, University of California, Berkeley, 7 November 2014. Hacken, P. ten (2017-03-29). ‘Compounding in morphology’. Oxford Research Encyclopedia of Linguistics. Retrieved on 28 August 2017. http://linguistics.oxfordre.com DOI: 10.1093/­acrefore /9780199384655.013.251 Hanson, R. 2005. ‘Incorporation in Yine’, talk given at the Research Centre of La Trobe University, 31 August 2005. Irineane Tasorentsi. Oquenquetsatacotaqueri Avincatsarite Jesoquirishito. 2008. El Nuevo Testamento de nuestro Señor Jesucristo en el idioma ashaninca (campa). 2nd ed. Lima: Editorial Sagradas Escrituras Para Todos, S.A. Kameethari Ñaantsi. Iñaaventaitzirira Avinkatharite Jesucristo. Ovakerari Antzimotavakaantsi. 2008. El Nuevo Testamento de nuestro Señor Jesucristo en el idioma asheninka (del Pichis). 2nd ed. Lima: La Liga Bíblica. Kindberg, L. 1961. ‘Independent and dependent sentence types in Campa (Arawak)’, in B. F. Elson and J. Comas (eds), A William Cameron Townsend en el vigésimoquinto aniversario del Instituto Lingüístico de Verano. Mexico: Instituto Lingüístico de Verano, 505–18. Kindberg, W. 1961. ‘Campa (Arawak) morphology’, in B. F. Elson and J. Comas (eds), A William Cameron Townsend en el vigésimoquinto aniversario del Instituto Lingüístico de Verano. Mexico: Instituto Lingüístico de Verano, 519–53. Kindberg, W. 1975. ‘Parentesco en campa ashaninca’, Datos Etno-Lingüísticos 11. 1–5. Lima: Instituto Lingüístico de Verano y Ministerio de Educación. Lawrence, A. 2012. ‘Reconstruction of Proto-Kampa verbal morphology’, paper presented at the 38th Annual Meeting of Berkeley Linguistics Society, Berkeley, California, 11–12 February 2012. Michael, L. 2008. ‘Nanti evidential practice: Language, knowledge and social action in an Amazonian society’. Austin: University of Texas dissertation. Michael, L. 2011. ‘La reconstrucción y la clasificación interna de la rama Kampa de la familia Arawak’, paper presented at the Congreso de Idiomas Indígenas de Latinoamerica V, University of Texas, Austin, 7 October 2011.



2  Genders and classifiers in Kampa (Arawak) languages of Peru

65

Michael, L. 2012. ‘Possession in Nanti’, in A. Y. Aikhenvald and R. M. W. Dixon (eds), Possession and ownership: A cross-linguistic typology. Oxford: Oxford University Press, 149–66. Michael, L., C. Beier, and Z. O’Hagan. 2013. Matsigenka texts. ms. Mihas, E. 2011. Añaani katonkosatzi parenini. El idioma del Alto Perené. Milwaukee: Clark Graphics. Mihas, E. 2014. Upper Perene Arawak narratives of history, landscape and ritual. Lincoln, NE: University of Nebraska Press. Mihas, E. 2015. A grammar of Alto Perené (Arawak). Berlin: Mouton de Gruyter. Mihas, E. 2017a. Alto Perené-Español-English. A trilingual illustrated thematic dictionary of Alto Perené (Arawak) of Peru. Munich: Lincom Europa. Mihas, E. 2017b. ‘The Kampa languages of the Arawak language family’, in A. Y. Aikhenvald and R. M. W.  Dixon (eds), The Cambridge Handbook of Linguistic Typology. Cambridge: Cambridge University Press, 782–814. Mithun, M. 1984. ‘The evolution of noun incorporation’, Language 60(4), 847–94. O’Hagan, Z. and L. Michael. 2015. ‘On exuberant noun categorization in Matsigenka (Arawak): Gender, animacy, and a multiple classifier system’, talk given at the Gender and Classifiers: Areal and Genealogical Perspectives Workshop, Max Planck Institute for Psycholinguistics, Nijmegen, 27 January 2015. Retrieved on 5 August 2017. http://linguistics.berkeley.edu/~zjohagan/ presentations Payne, D. 1980. Diccionario ashéninca-castellano (Documento de Trabajo 18). Lima: Instituto Lingüístico de Verano. Payne, D. 1981. The phonology and morphology of Axininca Campa (Summer Institute of Linguistics Publications in Linguistics 66). Arlington, TX: University of Texas Press. Payne, D. 1991. ‘A classification of Maipuran (Arawakan) languages based on shared lexical retentions’, in D. Derbyshire and G. K. Pullum (eds), Handbook of Amazonian Languages. Berlin: Mouton de Gruyter, 355–499. Payne, D. 1987. ‘Noun classification in Western Amazon’. Language Sciences 9(1): 21–44. Perú: Análisis Etnosociodemográfico de las Comunidades Nativas de la Amazonía, 1993 y 2007. 2010. Lima: Dirección Técnica de Demografía e Indicadores Sociales, Instituto Nacional de Estadística e Informática. Retrieved on 5 August 2017. http://proyectos.inei.gob.pe/web/­ biblioineipub/bancopub/Est/Lib0902/index.htm Pica, P., C. Lemer, V. Izard, and S. Dehaene. 2004. ‘Exact and approximate arithmetic in an Amazonian indigene group’. Science 306: 499–503. www.sciencemag.org. Retrieved on 25 September 2017. Pike, K. and W. Kindberg. 1956. ‘A problem in multiple stresses’, Word 12: 415‒28. Pío Aza, J. 1924. Estudio sobre la lengua machiguenga por el misionero. Lima: Casa editora ‘La Opinion nacional’. Sala, P. G., de la. 1905. Diccionario, gramática y catecismo: Castellano, Inga, Amueixa y Campa. Lima: F. Barrionuevo. Shaver, H. 1996. Diccionario nomatsiguenga-castellano, castellano-nomatsiguenga. Yarinacocha: Ministerio de Educación and Instituto Lingüístico de Verano. Shepard Jr., G. 1997. ‘Noun classification and ethnozoological classification in Machigenga’. The Journal of Amazonian Languages 1(1): 30–59. Snell, B. 2011. Diccionario matsigenka-castellano. Con índice castellano, notas enciclopédicas, y apuntes gramaticales (Serie Lingüística Peruana 56). Lima: Instituto Lingüístico de Verano.

66

Elena I. Mihas

Touchaux, M. P. 1909. Apuntes sobre la gramática y el diccionario del idioma campa o lengua de los antis tal como se usa en el río Apurimac. Lima: Imp. Nacional de Federico Barrionuevo. Tran, J. 2011. ‘The acquisition of Vietnamese classifiers’. PhD dissertation, University of Hawai’i at Mānoa. Weiss, G. 1975. ‘The world of a forest tribe in South America’, Anthropological Papers of the American Museum of Natural History 52(5): 219‒588. Wojtylak, K. I. 2017. ‘Classifiers as derivational markers in Murui (Northwest Amazonia)’, in L. Körtvélyessy, P. Štekauer, and S. Valera (eds), Word-formation across languages. Newcastle upon Tyne: Cambridge Scholars Publishing, 393–425.

3 Classifiers in Shiwilu (Kawapanan) Exploring typologically salient properties Pi l a r M . Va l e n z u e l a

1 Introduction 1.1  The Shiwilu people and their language The Shiwilu1 are an autochthonous people from Peruvian Amazonia. Their present-day and probably also traditional territory is located on the banks of the Aipena river, a tributary of the Huallaga close to its confluence with the Marañon. The number of people who nowadays call themselves Shiwilu might reach a few hundreds. Most of them inhabit the town of Jeberos (Jeberos District, Alto Amazonas Province, Loreto Region) and nearby villages on the Aipena and Marañon rivers. Their heritage language, likewise called Shiwilu (ISO 639-3 jeb), belongs to the Kawapanan family alongside Shawi (ISO 639-3 cbt). Although in the 1950s Shiwilu had approximately 1,500 speakers (Bendor-Samuel 1981[1958]: 1), today only a few elders of the grandparental and greatgrandparental generations exhibit native competency in the language. These individuals are fully bilingual in Spanish and predominantly employ this language in their daily life. Documentation and description of Shiwilu is underway.

1  Spanish speakers generally use the term ‘Jebero’ as both ethnonym and glossonym, while ethnic unit members prefer ‘Shiwilu’ for both senses. An analogous relationship holds between ‘Chayahuita’ and ‘Shawi’. As in previous works, I adopt the terms Shiwilu and Shawi in this chapter. Furthermore, favouring the orthographies of the corresponding Native languages, the name of the linguistic family is spelt ‘Kawapanan’ instead of the Spanish-based ‘Cahuapanan’.  The Shiwilu data are also given in the practical orthography, which includes the following conventions: represents the central vowel /ə/; as for consonants, stands for a voiced denti-alveolar approximant /ð̞/, for /ʎ/, for /ɲ/, for /tʃ/, and for /j/. The diacritic indicates glottal stop /ʔ/ after a vowel, but syllable boundary following a consonant. Stress is not phonological in Shiwilu. As a general rule, primary stress falls on the first syllable of disyllablic words but on the second syllable of words with three or more syllables; addition of certain affixes results in modifications (Bendor-Samuel 1981[1958]: 35–6, 73–4; Valenzuela and Gussenhoven 2013). Genders and Classifiers. First edition. Alexandra Y. Aikhenvald and Elena I. Mihas (eds) This chapter © Pilar M. Valenzuela 2019. First published in 2019 by Oxford University Press

68

Pilar M. Valenzuela

1.2  Selected features of Shiwilu grammar The present section outlines central features of Shiwilu grammar by way of providing the necessary context for the discussion of its classifier system, particularly certain cross-linguistically unusual morphosyntactic properties. Shiwilu is agglutinating with some fusion, predominantly head-marking, incorp­ orating, and polysynthetic. It has more suffixes than prefixes. The order of major clause constituents is grammatically free in that AVO, AOV, and OVA are frequent in text. Nevertheless, certain fixed orders are compatible with the OV pattern (Dryer 2007); e.g. with one exception Shiwilu has postpositions rather than prepositions, and the possessor precedes the possessum in the noun phrase. Bendor-Samuel (1981[1958]: 74–81) identifies four major word classes in Shiwilu: nominal, verb, adverb, and particle. In turn, nominal subdivides into four subclasses: noun (pidek ‘house’), adjective (a’llupi ‘large’, kala ‘three’, asu’ ‘this’), pronoun (kua ‘I’), and relative (ka’a’su’ ‘that which s/he ate’). Noun, adjective, and relative are open classes; pronoun is closed. In discussing the Shiwilu classifier system, it is important to treat these subclasses separately, since not all classifiers occur in every nominal context. The reference system of Shiwilu distinguishes four persons (1st exclusive, 1st inclusive,2 2nd, 3rd) and two numbers (minimal and augmented, called ‘singular’ and ‘plural’ for simplicity). As can be observed in Table 1, this system is realized in the personal pronouns, subject enclitics on non-verbal predicates, possessive pronominal modifiers, copula subject suffixes, and verbal inflection (represented by the subject suffixes from the non-future indicative (nfi) paradigm). Within the noun phrase, adjectives and certain quantifiers may precede or follow their head nominal, while demonstratives always precede the nominal they modify. Relative clauses are postnominal or internally headed. Table 1.  Shiwilu four-person system  

Personal Non-verbal Possessive Copula Verb subjects, pronouns predicate subjects modifiers subjects non-future indicative

1sg(.e) kua

=ku

=wek

-ka

=lek

1sg.i

kenmu’

=kenmu’

=mapu’

-ka

=lek

2sg

kenma

=ken

=pen

-ma

=la

3sg

nana

=ø

=nen

-a

=lli

1pl.e

kuda

=kudek

=widek

-kudek

=llidek

1pl.i

kenmu’wa’ =kenmu’wa’

=mapu’wa’

-kawa’

=lekwa’

2pl

kenmama’ =kenma’

=penma’

-mama’

=lama’

3pl

nawa’

=ø=lusa’

=nenna’

-erka

=llina’

2  The first person singular inclusive refers to ‘I and you’. In procedural speech, it functions as impersonal passive, equivalent to Spanish se. This can be observed in examples (37) and (44).



3  Classifiers in Shiwilu (Kawapanan)

69

Shiwilu is a classifying language, but lacks noun classes or a gender system (cf. certain Arawak languages where classifiers coexist with gender markers; Aikhenvald this volume, Mihas, this volume). Nouns, including those denoting kinship relations and plant/body parts, can appear by themselves as a free word and take possessive marking directly: itekla ‘hand’ > iteklawek ‘my hand’, dinlupi ‘heart’ > dinlupiwek ‘my heart’, awa ‘mother’ > awawek ‘my mother’, mekshi ‘man’s father-/brother-in-law’ > mekshiwek ‘my father-/brother-in-law (man speaking)’, etc.3 A handful of kinship terms exhibit some irregularity when possessed by the 3rd person: wila ‘child’, wila=wek ‘my child’, wila=pen ‘your child’, willin ‘his/her child’ (*wila=nen). Example (1) presents two body part nouns occurring by themselves, without possessive marking. katu’ (1) Ñiñi’wa i’na a’-ñi-lli dog foc caus-exist-nfi.3sg>3sg two ‘A dog has two front legs and two back legs.’ (Valenzuela et al. 2013; entry: tanpa’)

tanpa’, arm

katu’ two

lantek. foot

Shiwilu also has a set of obligatorily possessed nouns that denote the body, or parts of a plant or body, and require possessive marking to work as an independent word: pi‘body’ > piwek ‘my body’, tek- ‘skin/hide/bark’ > teknen ‘his/her/its skin/hide/bark’, llin- ‘vine, tail’ > llinen ‘its vine/tail’, etc. (see Table 3). Once these inalienable nouns take possessive marking, they display independent noun properties such as preceding or following an adjective, serving as head of a relative clause construction, and functioning as postpositional complement or host of a case-marker. Consider (2) involving the possessed noun teknen (see also Valenzuela 2016a: 356). ker-a’kasu’] awa=wek=ki.        / *tek (2) [[Tek=nen] skin=poss.3sg bring-nmlz.1sg>3sg mother=poss.1sg=gen  ‘Its bark (of the bamboo) which I brought (from the forest) is for my mother.’ Shiwilu independent nouns commonly take part in compounds, where they can work either as the first or the second component. Observe the role of nala in nalatanpa’ [nala ‘tree (trunk)’ + tanpa’ ‘arm, branch’] ‘tree branch’ and kenñala [ker’ ‘manioc’ + nala ‘tree (trunk)’] ‘manioc stem cutting’. Unlike nouns, classifiers are only attested as the second compound element: chipenñan [chiper ‘moriche palm’ + -nan cl. trunk] ‘stem of moriche palm’.4 Aikhenvald (2012: 198) observes that exclusive or special incorporation of body parts is a salient feature of Amazonian languages. Consistent with this, in Shiwilu 3  Shiwilu and Shawi differ in this respect. In Shawi, a noun may take possessive marking directly or it may first require the relational -ne. These two noun groups have been respectively labelled ‘inalienable’ and ‘alienable’, albeit there is not a clear semantic basis for this formal distinction (Valenzuela forthcoming). 4  In Baure, a Southern Arawak language from Bolivia, two classifying forms may combine to yield a complex, compound classifier with a more restricted usage than that of a simple classifier (Admiraal and Danielsen 2014: 95–7). The mandatory co-occurrence of -lu’-si [cl.meat-cl.cane], illustrated in (14) and (28), may be seen as a possible initial stage of a compound classifier in Shiwilu.

70

Pilar M. Valenzuela

classifiers and certain nouns, especially those denoting the body or plant/body parts, may incorporate in the verb: kitek-tanpa’-lli [bite-arm-nfi.3sg>3sg] ‘he/she/it ­arm-bit him/her/it’. The incorporation of classifiers is analysed in §5. The Shiwilu verb is minimally composed of a stem and an inflectional suffix. However, verbs may be morphologically very elaborate and constitute whole sentences by themselves. (3)

Na-per-wipu-i-tu-llen. recently-get.tired.of.doing-hold.baby.in.the.arms/lap-ben.appl-vmnfi.1sg>2sg ‘Just now I got tired of holding it (the baby) in my arms for you.’

Most Shiwilu verbs are inherently intransitive, transitive, or ditransitive; ambitransitives are few. The valency modifier -tu (vm) is particularly interesting (Valenzuela 2016b, 2017). It may increase or diminish the verb valency by one, by manipulating the object: di’- ‘kill/cut (without explicit object)’ > di’-tu- ‘kill/cut sth. or sb.’ (see (13), (57)); lli’- ‘see’ > lli’-tu ‘see (without specific object), appear’ (see (43)). The suffix -tu additionally works as verbalizer (compare (31a) and (31h)) and locative applicative ((58)). Shiwilu inflectional suffixes occupy the final position in the verb and, simultaneously, encode (tense-)mood and up to two arguments. They divide into ‘unipersonal’ and ‘bipersonal;’ the former attach to intransitive verbs or transitive verbs with a 3rd ­person object (as illustrated in (4)–(5)), while the latter mark transitive verbs with a 1st or 2nd person object (see (7)). When the object is 3rd person plural, -dek is added to the verb (see (6)). Core arguments can additionally be expressed by unmarked NPs. (4) (Kenma) tekka’-la. 2sg run-nfi.2sg ‘You ran.’ (5)

(Kenma) lli’-la 2sg see-nfi.2sg>3sg ‘You saw him/her/it.’

(nana). 3sg

(6) Wa’an peklu’-pa-dek-lli chief call-cont-3pl.o-nfi.3sg>3sg ‘The chief is calling the men.’ (7) Wa’an peklu’-pa-llenma’ chief call-cont-nfi.3sg>2pl ‘The chief is calling you (pl.).’

enmu’pinen=lusa’. man=pl

(kenmama’). 2pl

When the verb is ditransitive, the bipersonal suffix indexes the subject and the nonpatient object. The patient may be expressed by a free NP or be entirely omitted.



3  Classifiers in Shiwilu (Kawapanan)

enka’ -lun. (8) a. Ker’   manioc give-nfi.2sg>1sg   ‘You gave me manioc.’

b. Enka’-lun.   give-nfi.2sg>1sg   ‘You gave it to me.’

71

Two other inherently ditransitive verbs appear in (11) and (20). Although only two arguments can be indexed in the bipersonal suffix, a third, patient argument may be realized by -dek when it is 3rd person plural. (9) A’-du’-pa’-dek -llen caus-sit-approximative.appl-3pl.o-nfi.1sg>2sg ‘I seated the children next to you.’

wila=lusa’. child=pl

As shown above, argument NPs do not take case-marking in Shiwilu. However, =ler may encliticize on NPs in A function when the object is 3rd person, for discoursepragmatic or disambiguation purposes (Valenzuela 2011). For example, =ler is mandatory when both arguments are 3rd person and O precedes A in the clause. (10) a. Krawila(=ler) wa’an  dunker’-apa-lli.   Gabriela(=A) chief  look.for-cont-nfi.3sg>3sg   ‘Gabriela is looking for the chief.’ b. Wa’an  dunker’-apa-lli   chief    look.for-cont-nfi.3sg>3sg   ‘Gabriela is looking for the chief.’

Krawila=ler. / *Krawila Gabriela=A

It has been argued that Shiwilu is a good representative of the Andes-Amazonia transitional area, in that it combines structural properties typical of Western Amazonia (such as a classifier system) with features specifically associated to the Central Andean families Quechuan and Aymaran (like verbal suffixes simultaneously encoding the A>O interaction) (Valenzuela 2015, 2018). 1.3  Goal and organization of this chapter In a recent publication (Valenzuela 2016a) I provided an overall, detailed account of the Shiwilu nominal classifying system.5 The main goal of this chapter is to delve into a few cross-linguistically unusual grammatical properties of Shiwilu classifiers in  order to contribute to a general typology of nominal categorization devices. Furthermore, after showing that Shiwilu classifiers are relatively uncommon in text, it is suggested that considering the discourse frequency of classifying forms may be relevant in the study of classifier systems. 5  This article is the first detailed description of a Kawapanan classifier system. Interesting observations on Shawi classifiers can be found in Derbyshire and Payne’s (1990) comparative study. However, these are not based on primary work on the language or a secondary source focusing on its classifier system.

72

Pilar M. Valenzuela

The remainder of this chapter is structured as follows. §2 provides an overview of the Shiwilu classifier system. Although this section is largely based on Valenzuela (2016a), it includes previously unidentified classifiers which additionally introduce new criteria for the internal organization of the system. §3 begins with a presentation of the various morphosyntactic loci where classifiers appear in Shiwilu, and continues with an analysis of the overall text frequency of classifier morphemes as well as their frequency in regard to derivational and non-derivational functions. It is shown that the profile of Shiwilu classifiers is chiefly derivational, while other usages are uncommon. §4 deals with classifier stacking and reduplication, while §5 is devoted to incorp­ orated classifiers and their ability to encode O, S, and A arguments. §6 summarizes the conclusions and offers final remarks.

2  Overview of the Shiwilu nominal classifying system Shiwilu has a closed set of at least twenty-one classifying forms. These are bound, mostly monosyllabic roots, that attach to the right of their host. They encode the kind of properties typical of classifiers such as sex, shape, size, constitution, etc. As  shown in Table  2, Shiwilu establishes a first distinction between animate and ­inanimate classifiers. The former divide into female and male natural gender, whereas the latter are structured according to salient dimensionality, constitution, function, a combination of dimensionality and constitution, and arrangement. Not all nouns fall under the scope of a classifier, and there is no default or residual classifier (Valenzuela 2016a). As will become evident to the reader, the choice of a classifier is semantically fairly transparent in Shiwilu, a characteristic that reveals a moderate degree of grammaticalization of the system, and possibly its relatively young age. Note that Shiwilu has three pairs of homonymous classifiers: -la (cl.bone) and -la (cl.seed), -pi (cl.fruit) and -pi (cl.body), -lu’ (cl.meat) and -lu’ (cl.soil). Two of these formally identical pairs appear to have originated from the phonological reduction of different independent nouns: lu’lun ‘meat’ vs. lupa’ ‘soil’, lansi’ ‘skeleton’ / lanla’ ‘stem of a palm tree’ vs. lada ‘seed, eye, face’. In addition, two classifiers are formally identical to associated independent nouns: -dek (cl.liquid) and dek ‘water’, ‑lanser (cl.skeleton) and lanser ‘skeleton, bare branches of a dead tree’. Valenzuela (2016a) compiles seventeen Shiwilu classifiers, but anticipates that additional such morphemes might be found with further research. In fact, Table 2 includes four previously unidentified classifiers: -u’pi (cl.bract), -lanser (cl.skeleton), ‑lek (cl.penis), and -luwa (cl.shicana). In glossing the latter classifier, I have opted to maintain the local Spanish term shicana, given the difficulty in finding an appropriate concise expression in English. Shicana designates a dish consisting of salted small fish which are wrapped in leaf, pressed so as to extract the liquid, tied with vine



3  Classifiers in Shiwilu (Kawapanan)

73

Table 2.  Shiwilu classifying morphemes Parameter A n i m a t e

S e x

Classifier Category (Gloss in parenthesis) Female

-lun ~ -llun

female (FEM) Shiwilulun ‘Shiwilu woman’  

Male

-pen

male (MALE) Shiwilupen ‘Shiwilu man’

One-na(n) dimensional ~ -ña(n) (long) -la ~ -lla -llin

A n i m a c y

I n a n i m a t e

D i m e n s i o n a l i t y

long, rigid, wooden (TRUNK) pentunan ‘log used as a bridge’ long, not large, non-wooden (BONE) pintella ‘cigarette’ long, thin, flexible (VINE) dullin ‘intestine’

Two-mek dimensional (flat) -tek ~ -chek

flat, thin, extended (LEAF) ker’mek ‘manioc leaf ’

-la Threedimensional ~ -lla (round, -pi tubular)

round, small (SEED) wisekla ‘fruit of the Amazon grape’

Constitution

Function

flat, to cover, to wrap (SKIN) ketchek ‘manioc peel’

round, not small (FRUIT) lawa’pi ‘jungle cacao’

-si

tubular (CANE) illapasi ‘barrel of a shotgun’

-pi

body, bulky object (BODY) wa’napi ‘metal object (car, aeroplane, radio)’

-du’ ~ -yu’

bulky w/bumpy surface (CORNCOB) kaladu’ ‘three corncobs, ice-cream bean fruits, war grenades’

-u’pi

elongated, oval, bulky object like the bract of a banana tree (with an overhanging, open petal) (BRACT)

-dek ~ -yek

liquid (LIQUID) chiter’dek ‘corn drink’

-lu’ ~ -llu’

powdery, pasty, land (SOIL) kasetllu’ ‘sugar’

-lu’ ~ -llu’

meat, flesh (MEAT) lamanlu’ ‘troop-peccary meat’ 

-dan

elongated, thick tuber (MANIOC) utekdantu- ‘harvest manioc’ (Continued )

Pilar M. Valenzuela

74

Table 2. Continued Parameter

Classifier Category (Gloss in parenthesis)

 

 

-dun ~ -yun

clothing for the body (CLOTHES) pi’pi’yun- ‘sew clothes’

 

 

-lek

long, made of flesh (PENIS) kalalek ‘three snakes, three eels’

Dimensionality and ­constitution

-lanser

skinny animal or person; bare branches of a dead tree (SKELETON) widunalanser ‘old broom (made from a bush that has lost its leaves)’

Arrangement

-luwa

elongated, wrapped in leaf, pressed, and tied w/vine fibre (SHICANA) chinluwa ‘dish made of salted small fish wrapped in leaf, pressed, tied w/vine fibre, and cooked in the fire’

 

 

 

fibre, and cooked in the fire. The classifier -luwa applies to elongated objects arranged in a similar fashion.6 (11)

Uku’la-u ala’-luwa sell-imp.2sg>1sg one-cl.shicana ‘Sell me one shicana!’

chin-luwa=pen! smoke-cl.shicana=poss.2sg

Elongated, oval shaped, bulky objects may be referred to by -u’pi (cl.bract). In addition to the bract of a banana tree and the shapaja palm tree (Attalea moorei), this classifier applies to certain armoured catfish due to their body shape and their ventral fins which are compared to the overhanging petals of the banana bract (Mr Fidel Lomas, p. c., January 2018). (12)

Walatek i’na a’llupi-u’pi. kind.of.fish foc large-cl.bract ‘The walatek armoured catfish is large.’

The classifier -lek (cl.penis) applies to elongated animals or body parts. It cannot refer to objects or plants, since it implies that the classified element is constituted of flesh. Hence, -lek classifies different species of snakes and eels, and may extend to iguanas, large fish like pirarucu or barred surubim, and an alligator tail. All these referents can also be classified by the sequence ‑lu’-si (cl.meat-cl.cane) (Valenzuela 2016a: 346). The difference between the two alternatives lies in that an entity ­designated by -lu’-si is necessarily bigger than one designated by -lek. Consider the next pair of examples. 6 Including dunallin alekpi, a kind of torch made with the dry resin of the copal tree which is ground, wrapped in leaf, pressed, and tied tightly.



3  Classifiers in Shiwilu (Kawapanan)

(13)

Katu’-lek ellek di’-tu-lek nala=lek. two-cl.penis afaninga.snake kill-vm-nfi.1sg>3sg pole=inst ‘I killed two afaninga snakes (Chironious exoletus) with a pole.’

(14)

Tanak tana=k jungle=loc

pa’apilamu pa’-apa-ila-amu go-cont-outward-ptc.ss.1sg

75

mañilu’ mañilu’ boa.constrictor

a’llupillu’si kananek. a’llupi-lu’-si kanan-lek large-cl.meat-cl.cane find-nfi.1sg>3sg ‘While walking in the jungle, I came upon a very large and thick boa constrictor.’ The noun lanser ‘skeleton, bare branches of a dead tree’ has grammaticalized into a classifier referring to elongated, skinny animals such as a skinny dog, various kinds of monkeys, or wading birds like the grey-winged trumpeter (Psophia crepitans) or different heron species, purportedly because of their very thin legs. This classifier likewise applies to skinny people and the bare branches of a dead tree. In (15), ‑lanser appears incorporated in the verb. (15)

Chin-lanser-apa-lek. smoke-cl.skeleton-cont-nfi.1sg>3sg ‘I am smoking it (an elongated skinny animal, e.g. a skinny monkey).’

The classifier -dan is exclusively used with maniocs and yams (Dioscorea trifida) with an elongated shape; roundish yams are classified by -pi cl.fruit). (This information will become relevant in §3.3.4, when discussing the adjective-like function of classifiers.) Valenzuela (2016a) compares -dan to the dedicated classifier for ‘corn’ found in Mayan languages (Aikhenvald  2000: 285, 2017: 369). Both have restricted semantics and represent crops of high cultural importance for the respective speech communities. It must be added that Shiwilu has a couple of verbs that only apply to the process of growing manioc: inyu’- ‘plant manioc cuttings’ and utekdantu- ‘harvest manioc’. Note that the latter contains the classifier -dan; however, *utek- by itself is not attested in Shiwilu today. The next sentence involves the verb inyu’-. (16)

Sadawek inyu’tapalli sada=wek inyu’-tu-apa-lli wife=poss.1sg plant.manioc-vm-cont-nfi.3sg ‘My wife is planting manioc in her field.’

menmiñik. menmi=nen=k field=poss.3sg=loc

Although sowing other plants (such as barbasco, pineapple, and sugarcane) also involves introducing a single cutting into a hole, employing inyu-’ in these instances produces an infelicitous utterance.

76

Pilar M. Valenzuela

Three classifiers have grammaticalized as productive nominalizers. These are ‑nan (instrumental, concretizer), -tek (habitual agent),7 and -pi (resultative) (Valenzuela 2016a: 350–1). Consider the following data (generally, transitive verbs are intransitivized before undergoing nominalization): dektunan ‘scissors’< dektun- ‘cut (without explicit object)’, lulentutek ‘healer’ < lulentu- ‘heal (without explicit object)’, chimipi ‘cadaver’ < chimin- ‘die’. Other classifiers perform a nominalizing function only ­marginally. For example, inyu’- ‘plant manioc cutting’ + -dan (cl.manioc) > inyu’dan ‘first planting of manioc in a new garden’. As mentioned in §1.2, most inanimate classifiers are formally identical to a set of obliga­ torily possessed nouns that denote the body or a plant/body part.8 Additionally, there are associated, segmentally longer independent nouns in the language. When an obligatorily possessed noun and a segmentally longer independent noun are available for roughly the same referent, the latter seems to be more common in spontaneous text. The following extract pertains to a story about the first woman who learned how to manufacture clothes with the help of forest spirits. Note the use of the full noun lada instead of its inalienable counterpart la- (see also ‘meat’ in (32) where lu’lun is used instead of lu’-). (17)

–Iñer  dek-ker’          lada=nen,     si’yektu-ker’ all    throw.away-imp.2sg>3sg  seed=poss.3sg  gather-imp.2sg>3sg lada=nen,     ter’a-ter’a      a’tu-ker’      lada=nen– seed=poss.3sg  plant-plant  aux-imp.2sg>3sg  seed=poss.3sg itu-llima. tell-nfi.3sg>3sg:hsy ‘–Take out all its seeds, gather its seeds, and plant its seeds– she (the forest spirit) told her.’ (jeb-26-02-10-pvb-1)

In (17), lada takes possessive marking. However, this is not a grammatical requirement (recall (1) where tanpa’ ‘arm’ occurs unmarked). In addition to the seed of a plant, lada designates the eyes or face of a person or animal (see the discussion in §3.3.4 and §3.3.5). Table 3 lists Shiwilu classifiers along with their associated inalienable and independent nouns. Blank cells indicate that no associated noun has been identified. The relationship between classifiers and the independent nouns assigned to them in Table 3 is sometimes only speculative. This is especially the case of -dun (cl.clothes) and dunan ‘waist’, as well as -du’ (cl.corncob) and du’lupi ‘spine’. Further research might reveal culturally specific associations that can help us better understand these relationships. For example, native speaker Mr Meneleo Careajano explained that corncobs sprout from the stem, which is like the backbone of the plant. He added that corncobs are conceived as sitting on the lap of their stems and offered the following possible exchange (the verb wipu- was earlier introduced in (3)). 7  An additional non-classifying function of -tek can be found in §3.3.4. 8  In some Preandine Arawak languages of Peru (a.k.a. Kampa) almost all core multiple classifiers come from a set of obligatorily possessed nouns (Mihas, this volume).



3  Classifiers in Shiwilu (Kawapanan)

77

Table 3.  Classifiers and associated inalienable and independent nouns in Shiwilu Classifier

Inalienable noun + =nen poss.3sg

Independent noun

-lun cl.fem

 

 

-pen cl.male

 

 

-na(n) cl.trunk

nanen ‘its tree trunk’

nala ‘tree, trunk’

-la cl.seed

lanen ‘his/her/its seed/eye/face

lada ‘seed, eye, face’

-si cl.cane

sinen ‘its cane, tube’

silu ‘cane’

-la cl.bone

lanen ‘its bone, stem’

lansi’ ‘bone,’ lanla’ ‘stem of a palm’

-llin cl.vine

llinen ‘its vine, tail’

lullin ‘vine’, llintek ‘tail’

-tek cl.skin

teknen ‘his/her/its skin, hide, bark’

chipitek ‘bark, hide, skin’

-mek cl.leaf

meknen ‘its leaf ’

lalumek ‘leaf ’

-lu’ cl.meat

lu’nen ‘his/her/its meat, flesh’

lu’lun ‘meat’

-lu’ cl.soil

lu’nen ‘its powder, paste’

lupa’ ‘soil’

-pi cl.body

pinen ‘his/her/its body’

 

-pi cl.fruit

pinen ‘its fruit’

 

-u’pi cl.bract

u’pinen ‘its bract’

 

-dan cl.manioc

danen ‘its elongated tuber’

(ker’ ‘manioc’, mama’ ‘yam’)

-lek cl.penis

leknen ‘his/its penis’

 

-du’ cl.corncob

du’nen ‘his/her/its stomach, its corncob’

du’lupi ‘backbone’

-dek cl.liquid

deknen ‘his/her/its liquid’

dek ‘water’

-lanser’ cl.skinny

 

lanser’ ‘skeleton, bare branches of a dead tree’

-dun cl.clothes

 

dunan ‘waist’

-luwa cl.shicana

 

chinluwa

78 (18)

Pilar M. Valenzuela a.      

–Ma’pu’su’  chitetllu’pen? ma’pu’su’  chiter-lu’=pen how   corn-cl.soil=poss.2sg ‘How is your corn field?’

b. –Ipa’ wipu-tu-lli.   already hold.baby.in.the.arms/lap-vm-nfi.3sg   ‘It (the stem) is already holding (corncobs) in its arms or lap.’ In Shiwilu the same classifier morphemes can appear in multiple morphosyntactic environments (i.e. it is a ‘multiple classifier’ language in terms of Aikhenvald 2000). Nonetheless, classifiers are not grammatically required in any of these contexts. This topic, along with the functions and text frequency of classifiers, are treated in the next section.

3  Morphosyntactic contexts and text frequency of Shiwilu classifiers 3.1  Morphosyntactic contexts of Shiwilu classifiers A typologically noteworthy characteristic of the Shiwilu system of nominal classification is the ability of the same set of classifiers to appear in a range of morphosyntactic contexts. Thus, classifiers may append to the right of number words and other quantifiers, demonstratives, question words, nouns, adjectives, formatives that require a classifier to derive a noun or adjective, verb stems, and even personal pronouns. The following sentences illustrate the presence of classifiers in these environments. (19) Demonstrative Nana-la enka’-u that-cl.bone give-imp.2sg>1sg ‘Give me that candle!’

(kanchila’)! candle

(20) Numeral quantifier Kaladu’ panuer’! kala-du’ panu-ker’ three-cl.corncob give.as.present-imp.2sg>3sg ‘Give him/her three corncobs (as present)!’ (21)

Non-numeral quantifier Wapu’si kella pinterpen? wapu’-si ker-la pinter=pen many-cl.cane bring-nfi.2sg>3sg tobacco=poss.2sg ‘Did you bring many bundles of tobacco?’



3  Classifiers in Shiwilu (Kawapanan)

(22) Noun Wilawek wila=wek child=poss.1sg

wa’dan wa’dan non.Indian

ilulenna ilulenna medicine

79

uwan u-w-an drink-ep-ptc.ss.3sg>3sg

iñer’  kuwetllinnenlusa’      dinketlli. iñer’  kuwer-llin=nen=lusa’    dinker’-lli all  worm-cl.vine=poss.3sg=pl  evacuate-nfi.3sg>3sg ‘After drinking the Western medicine, my child evacuated all his worms.’ (23) Adjective Asu’ serenpa a’llupi-pi. this pineapple large-cl.fruit ‘This pineapple is large.’ (24) Question word Ma’nen-mek i’na nana? what-cl.leaf foc that ‘What (kind of) leaf is that? (25) Verb Pekki’yunek. peksa’-ker’-dun-lek wash-completely-cl.clothes-nfi.1sg>3sg ‘I washed all the clothes.’ (26) Personal pronoun Kualun / Kuapen kua-lun kua-pen 1sg-cl.fem 1sg.cl.male

enpu’nipa’ enpu’niipa’ very.muchfoc

pata’tutekku pata’-tu-tek-ku touch-vm-agt.hab.nmlz-1sg

nuka’ka nuka’-ka cop-1sg ‘I (woman) / I (man) am very hardworking.’ The classifier -dun (cl.clothes) has distributional constraints. It only occurs on low ­numbers and a handful of verbs whose meanings are closely associated with clothes: peksa’dun- ‘wash clothes’ (illustrated in (25)), a’wiñiyun- ‘put clothes in the sun to dry’, pi’pi’yun‘sew clothes’, a’pasu-dun- ‘mend clothes’, and a’si’yun- ‘dye clothes’. Valenzuela (2016a) suggests that -dun may not have yet fully grammaticalized into a classifier. This would imply that one of the two contexts where it is attested may be diachronically primary.9 9  Derbyshire and Payne (1990: 258) state that Shawi classifiers appear to be ‘primarily a verb-­incorporated system that has developed characteristics of a numeral system’ (but cf. Aikhenvald 2000: 227 and footnote 6 in this work).

80

Pilar M. Valenzuela

As stated in §2, classifiers are not grammatically required in any of the contexts illustrated in (20)–(26). Unlike most classifier languages, nominal modifiers, including demonstratives and low numbers, usually occur without a classifier as can be observed in (1) and (50).10 On the other hand, the ability of classifiers to occur in such a variety of morphosyntactic environments, including personal pronouns, is typologically noteworthy. The data in (26) suggest that pronouns only combine with the animate classifiers -lun (cl.fem) and -pen (cl.male). This is not the case, however, as shown in (27)–(29). Note that these sentences present the same structure, which provides a suitable context for a classifier marked pronoun. ñichu (27) Kenmalansetchekpu’si’pa’ kenma-lanser-tek=pu’=sa’=i’pa ñi-etchun 2pl-cl.skeleton-deprec=simil=del=speculation exist-fut.3sg

   

wilapen. wila=pen child=poss.3sg ‘Maybe your child will be extremely skinny just like you.’ (28) Kenmalu’sipu’si’pa’ kenma-lu’-si=pu’=sa’=i’pa 2pl-cl.meat-cl.cane=simil=del=speculation

ñichu ñi-etchun exist-fut.3sg

   

wilapen. wila=pen child=poss.3sg ‘Maybe your child will have a large penis just like you.’ Shiwilu speakers are somewhat familiar with cartoons, either because television has become available in Jeberos in the last few years or because of their experience with television while visiting or living in other towns or cities. Taking advantage of this, hypothetical situations were posited where plants and objects were assigned human characteristics. The context of (29) is that of two trees talking about a third such entity. (29) Kenmananshapu’si’pa’ kenma-nan=sha=pu’=sa’=ipa’ 2sg-cl.trunk=dim=simil=del=speculation

saka’nansha saka’-nan=sha rough-cl.trunk=dim

   

nuka’a. nuka’-a cop-3sg ‘Maybe it has a rough trunk just like you.’ 10  In Shawi low numbers tend to co-occur with a classifier (Hart 1988: 491). For other languages in the region that also differ from Shiwilu in this respect see Payne (1986, 1987), Derbyshire and Payne (1990), Barnes (1990), Aikhenvald (1994), Gómez-Imbert (2007), Seifart and Payne (2007: 382–3), Thiesen and Weber (2012), Wojtylak (2016), among others. See also the discussion in Aikhenvald (2012: 279–303).



3  Classifiers in Shiwilu (Kawapanan)

81

Interestingly, Aikhenvald (2002: 64) reports that in the Arawak language Tariana the  feminine classifier -ma may attach to plural personal pronouns: waha-ma-pe (1pl-cl:fem-pl), ihya-ma-pe (2pl-cl:fem-pl), and naha-ma-pe (3pl-cl:fem-pl). Most probably, continues Aikhenvald, this characteristic arose due to contact with East Tucanoan languages where a term for ‘woman’ occurs with non-singular pronouns to disambiguate their gender reference. However, the pronoun + classifier combination in Tariana has very limited application and therefore is different from the more ­general ability of Shiwilu classifiers to combine with personal pronouns. In fact, this feature of Shiwilu classifiers is cross-linguistically unusual and maybe even unique. 3.2  Text frequency of Shiwilu classifiers A question that has been overlooked in the study of classifier systems concerns the frequency with which classifying forms occur in discourse. Addressing this question can help us better characterize classifier systems and their functions, assess their relative vitality, and even ascertain aspects that may prove relevant in understanding their diachronic development. It seems to be the case that classifier languages can differ significantly with regard to the relative text frequency of their classifying forms. Considering the texts in the appendix to their grammar of Bora (a Boran language spoken on both sides of the Peru-Colombia border), Thiesen and Weber (2012: 163) state that approximately four out of every ten words bear a classifier; i.e. on average this language features a classifier every 2.5 words. In Shiwilu, the fact that classifiers can appear in an exceptional variety of morphosyntactic environments might lead one to assume that their presence is pervasive in discourse. But contrary to this expectation, it has been suggested that classifiers may be relatively infrequent in this language, especially when a non-derivational function is involved (Valenzuela 2016a: 373). The goal of this section is to estimate the overall text frequency of Shiwilu classifiers as well as their frequency relative to derivational and non-derivational functions. In order to address this question, I have analysed a corpus of 15 texts (12 narrative, 2 procedural, and 1 conversation11). Eight were collected by Bendor-Samuel in the 1950s, when Shiwilu speakers counted ca. 1,500; the remaining seven were obtained by myself between 2005 and 2012, at a time when Shiwilu had already become critically endangered. Table 4 presents the number of classifier tokens found in the corpus and the frequency with which they occur calculated in terms of the ratio between words Table 4.  Text frequency of Shiwilu classifiers  

Texts

Bendor-Samuel

8

Words

Classifier tokens

Words : Classifiers ratio

2,657

110

24.15

Valenzuela

7

6,806

233

29.21

Total

15

9,463

343

27.59

11  No significant difference in the use of classifiers has been observed based on text type.

82

Pilar M. Valenzuela

and classifiers. It can be said that the texts from the 1950s and those compiled in the present century do not differ very much in regard to classifier frequency. The information in Table 4 confirms that Shiwilu classifiers are in fact relatively infrequent in text (at least when compared with Bora classifiers); on average, there is a classifier every 27.6 words. Moreover, Bendor-Samuel collected a short text ­composed of 128 words that contains only two tokens of the classifier -pi (cl.body), while a second text of 386 words collected by Valenzuela has only three tokens of the classifier -lun (cl.fem). The only conversation included in the corpus was collected by Valenzuela and contains 801 words and 27 classifier tokens, i.e. a ratio of one classifier every 29.7 words. Therefore, the types of texts considered in the corpus do not seem to significantly affect classifier frequency.12 Analysis of the corpus also confirms that deriving new lexemes is by far the most common function of classifiers in Shiwilu, representing 82.5% of their occurrences, whereas all other functions only reach 17.5%. This is summarized in Table 5. The very low usage of classifiers in non-derivational functions in Shiwilu is directly related to the fact that they are not mandatory with numerals, demonstratives, or any kind of nominal modifier. Therefore, it is expected that languages that differ from Shiwilu in this respect present a much lower words-classifiers ratio. An additional conclusion concerns function. It has been argued that the primary function of classifier systems may be discourse related, having to do with participant reference (Payne 1987: 39). Our data do not support this hypothesis for Shiwilu. Fourteen of the twenty-one classifiers identified in the language are represented in the corpus. These are listed in Table 6, from more to less commonly used. In addition, this table records the number of derivational and non-derivational usages of each classifier. Table 5.  Frequency of derivational and other functions of Shiwilu classifiers  

Derivational function

Other functions

Bendor-Samuel

 92 (83.6%)

18 (16.4%)

Valenzuela

191 (82%)

42 (18%)

Total

283 (82.5%)

60 (17.5%)

Table 6.  Classifiers represented in the corpus, overall tokens and derivational versus non-derivational usages Classifier

Bendor-Samuel (1955–6)

Valenzuela (2005–12)

Total

1)  -lu (cl.fem)

12 deriv. 12 non-deriv. 0

49 deriv. 42 non-deriv. 7

61

12  According to Mihas (this volume), on average there is a classifier every 4.72 sentences in Matsigenka, and one every 8.5 sentences in Alto Perené Ashéninka.



3  Classifiers in Shiwilu (Kawapanan)

83

Classifier

Bendor-Samuel (1955–6)

Valenzuela (2005–12)

Total

  2)  -lu’ (cl.soil)

18 deriv. 12 non-deriv. 6

41 deriv. 37 non-deriv. 4

59

  3)  -nan (cl.trunk)

34 deriv. 33 non-deriv. 1

24 deriv. 22 non-deriv. 2

58

  4)  -dek (cl.liquid)

14 deriv. 14 non-deriv. 0

40 deriv. 31 non-deriv. 9

54

  5)  -tek (cl.skin)

9 deriv. 9 non-deriv. 0

19 deriv. 19 non-deriv. 0

28

  6)  -llin (cl.vine)

5 deriv. 5 non-deriv. 0

17 deriv. 16 non-deriv. 1

22

  7)  -lu’ (cl.meat)

5 deriv. 0 non-deriv. 5

8 deriv. 2 non-deriv. 6

13

  8)  -pi (cl.fruit)

2 deriv. 0 non-deriv. 2

8 deriv. 5 non-deriv. 3

10

  9)  -pi (cl.body)

3 deriv. 0 non-deriv. 3

7 deriv. 4 non-deriv. 3

10

10)  -mek (cl.leaf)

0 deriv. 0 non-deriv. 0

10 deriv. 6 non-deriv. 4

10

11)  -la (cl.seed)

7 deriv. 6 non-deriv. 1

2 deriv. 2 non-deriv. 0

9

12)  -pen (cl.male)

0 deriv. 0 non-deriv. 0

6 deriv. 5 non-deriv. 1

6

13)  -dan (cl.manioc) 0 deriv. 0 non-deriv. 0

2 deriv. 0 non-deriv. 2

2

14)  -la (cl.bone)

1 deriv. 1 non-deriv. 0

0 deriv. 0 non-deriv. 0

1

Total

110 (32%) deriv. 92 non-deriv. 18

233 (68%) deriv. 191 non-deriv. 42

343

84

Pilar M. Valenzuela

The classifiers in Table 6 can be said to be the most commonly used in the language (and, with a single exception, coincide with the group of classifying morphemes ­initially identified by the present author).13 3.3  Functions of classifiers in the corpus Classifiers in the corpus perform various types of functions. A primary distinction is made between derivational (68%) and non-derivational (32%) usages. The morpheme ‑lu’ (cl.meat) deserves special mention. It is the only classifier attested in both text collections that shows more tokens in non-derivational (11) than derivational (2) functions. Some of its non-derivational usages can be observed in examples (32) (concord-like), (38) (adjective-like), and (43)–(44) (disambiguating). 3.3.1  Derivational function  By far, the most frequent function performed by classifier morphemes in Shiwilu is the derivation of new vocabulary; this represents 68% of  all occurrences. The significant difference in the text frequency of the female and male classifiers reported in Table  6 (61 tokens of the former as opposed to only 6 tokens of the latter) is mainly due to the fact that in a couple of commonly used noun pairs the female versions involve derivation through ‑lun (cl.fem), while the male counterparts resort to other strategies. As shown in (30a), the term referring to a young woman or young girl is obtained by attaching ‑lun to the noun wila ‘child’, whereas there is a completely different lexeme available to refer to a young man. As for the noun pair in (30b), taser- ‘old’ does not constitute a word by itself. ‘Old woman’ takes the classifier ‑lun (cl.fem) but, unexpectedly, ‘old man’ takes -pi rather than ‑pen (cl.male). As treated in §2, synchronically -pi is analysed as a resultative nominalizer or a classifier meaning ‘body, entity taken as a whole’.14 (30) a. wila-lun [wila ‘child’ + -lun] ‘young woman’ b. tasellun [taser- ‘old’+ -lun] ‘old woman’

nadi’nek ‘young man’ taser-pi ‘old man’

Other instances of the derivational function of classifiers in the corpus follow. (31)

a.  sek-llin [sek- ‘with the hands’ + -llin cl.vine] ‘path’ b.  punka-pi [punka’- ‘float’ + -pi cl.fruit] ‘shirui (k. small armoured catfish)’ c.  mutu’-tek- [mutu’ ‘head’ + -tek cl.skin] ‘wig’ d.  mudi-dek [mudin ‘breast’ + -dek cl.liquid] ‘milk’

13  The additional seven classifiers included in Table 2 were obtained from other texts, the illustrative examples in the Shiwilu Dictionary (Valenzuela et al. 2013), or during elicitation sessions with different native speakers. These are ‑si (cl.cane), -u’pi (cl.bract), ‑dun (cl.clothes), -du’ (cl.corncob), -luwa (cl.shicana), ‑lanser (cl.skeleton), and -lek (cl.penis). 14 However, -pi and -pen may be diachronically linked. Valenzuela (2016a: 340) suggests that ‘taserpi ‘old man’ might represent an instance of lexicalization, where an older version of the classifier-pen was preserved in ­frozen form’.



3  Classifiers in Shiwilu (Kawapanan) e. f. g.   h. i. j. k. l. m.

85

kadu’-la [kadu’ ‘egg’ + -la cl.seed] ‘testicle’ nala-nan [nala ‘tree’ + -nan cl.trunk] ‘fallen tree’ pentu-nan [pentu- ‘cross’ + -nan cl.trunk] ‘bridge’ (horizontally placed log used to cross a stream or another obstacle) sek-llin-tu- [sek- ‘with the hands’ + -llin cl.vine + -tu vm] ‘clear a path’ ma-lu’-tu- [ma- ‘catch,’ -lu’ cl.soil + -tu vm] ‘obtain/extract clay’ anu’-lu’-tu [anu’- ‘fall’ + -lu’ cl.soil + -tu vm] ‘fall to the ground’ pa’-mek-tu- [pa’- ‘go’ + -mek cl.leaf + -tu vm] ‘walk in the forest’ pa’-dek-tu- [pa’- ‘go’ + -dek cl.liquid + -tu vm] ‘walk in the water’ ala’-lu’ [ala’ + -lu’ cl.soil] ‘once’

The nouns in (31a)–(31g) above are formed by adding a classifier to a manner prefix (sek- ‘with the hands’), a verb stem (punka’- ‘float’, pentu- ‘cross’), or a noun (mutu’ ‘head’, mudi ‘breast’, kadu’ ‘egg’, and nala ‘tree’). On the other hand, in (31h) the noun sekllin ‘path’ takes the suffix -tu which works as verbalizer thus yielding ‘clear a path (in the jungle)’. In (31i)–(31l), suffixation of the applicative -tu allows for the incorp­ oration of a classifier referring to the O argument; this O is semantically a patient in (31i), but a location in (31j)–(31l). Lastly, in (31m) the number ala’ ‘one’ receives the classifier -lu’ (cl.soil) which here indicates number of times. Derived words involving two or more classifiers can be found in §4.1. 3.3.2  Concord-like function  Although uncommon in text, Shiwilu classifiers may realize a concord-like function. In these instances, the referent at hand is conveyed by a full noun, and thus the presence of one or more concording classifiers in the same utterance seems redundant. Sentence (32), taken from the corpus, contains three references to the pork: through the possessive construction kusher lu’lunen ‘pig’s meat’, and the morpheme -lu’ (cl.meat) on the verb and the noun kusher. (32) Kusherunta’ kusher=unta’ pig=add

lu’lunen lu’lun=nen meat=poss.3sg

iñer iñer all

uku’lalu’tullima uku’la-lu’-tu-lli=ima sell-cl.meat-vm-nfi.3sg>3sg=hsy

kushellu’. kusher-lu’ pig-cl.meat ‘They also sold the pig’s meat, all the pork.’ (Bendor-Samuel, 1955–6) Example (32) was taken from a text collected in the 1950s, but similar utterances are attested in Shiwilu today (Valenzuela 2016a: 361–2). Interestingly, native speakers not only readily accept constructed sentences exhibiting two or three tokens of the same classifier (often in addition to the corresponding full noun), but even consider these

86

Pilar M. Valenzuela

expressions instances of ‘very good/nice Shiwilu’ (Mrs Emérita Guerra and Mrs Julia Inuma, p. c. 2012, 2013, 2014).15 3.3.3 Discourse functions  Instances of classifiers realizing a discourse related function can also be found in the corpus. The extract in (33) is part of a text describing how to make a drinking bowl. Note that the classifier -lu’ (cl.soil), which applies to  pasty or powdery substances, cataphorically refers to the ground bark of the apacharama tree, pidatek.16 (33) Sinanpachi  yanu’tanna’         musha’su’, sinan-pachi  ya-nu’-tu-anna’       mu-sha-a’su’ plate-bowl  des-be.so-vm-ptc.ss.3pl>3sg  be.good-dim-nmlz.3sg peksanlu’-tu-llina’ peksan-lu’-tu-llina’ grind-cl.soil-vm-nfi.3pl>3sg

damellu’shasu’, damer-lu’-sha-a’su’ be.fine-cl.soil-dim-nmlz.3sg

pidatek        peksanna’. pida-tek      peksan-anna’ apacharama-cl.skin  grind-ptc.ss-3pl>3sg ‘To make very nice drinking bowls, the apacharama bark is ground very finely.’ As a child, native speaker Mr Meneleo Careajano used to observe his mother, Mrs Soledad Chota, make pottery. Hence, he is very familiar with this activity. While analysing (33) with me, Mr Meneleo volunteered (34), which is a more complete description of this part of the process. (34) Sinanpachi  yanu’tanna’         peksanlu’tullina’ sinan-pachi  ya-nu’-tu-anna’       peksan-lu’-tu-llina’ plate-bowl  des-be.so-vm-ptc.ss.3pl>3sg  grind-cl.soil-vm-nfi.3pl>3sg damellu’shasu’, damer-lu’-sha-a’su’ be.fine-cl.soil-dim-nmlz.3sg

sulalu’llina’, sula-lu’-llina’ sift-cl.soil-nfi.3pl>3sg

nu’tanna’ nu’-tu-anna’ be.so-vm-ptc.ss.3pl>3sg

aka’lu’llina’, aka’-lu’-llina’ mix?-cl.soil-nfi.3pl>3sg

15  One such constructed example is the following. i.  Asu’-dan ker’ a’llupi-dan   this-cl.manioc manioc large-cl.manioc    ‘I am craving this large manioc’

luya-dan-[l]ek. crave-cl.manioc-nfi.1sg>3sg

16 The apacharama bark is burned, ground, and mixed with the clay in order to make pieces of pottery more resistant and impermeable.



3  Classifiers in Shiwilu (Kawapanan)

87

nu’tanna’         piwellintullina’. nu’-tu-anna’       piwer-llin-tu-llina’ be.so-vm-ptc.ss.3pl>3sg  form.roll?-cl.vine-vm-nfi.3pl>3sg Nu’su’    nu’tullina’        sinanpachi. nu’-su’   nu’-tu-llina’        sinan-pachi be.so-advl  be.so-vm-nfi.3pl>3sg  plate-bowl ‘To make drinking bowls one grinds (the burned apacharama bark) very finely, sifts it, then mixes it with clay, and forms strips with it. This is how one makes drinking bowls.’ Mr Meneleo’s description features four tokens of -lu’ (cl.soil) and further introduces the classifier -llin (cl.vine) in conveying the modelling of clay strips. Nevertheless, he omits full nouns designating the apacharama bark and/or the clay. Since the potterymaking process is part of this community’s common knowledge, the speaker may evaluate that exclusively resorting to classifiers suffices to make himself understood in this context. In the corpus there are a number of instances where a classifier constitutes the only mention of a referent that is considered understood (e.g. anu’-lu’-tu- [fall + cl.soil + vm] ‘fall to the ground’ in (31j)), or an entity that does not play a central role in the discourse. In the following text extract there are three references to the skeins elaborated by the protagonist, all through the classifier -pi (cl.fruit). The full noun piterpi ‘skein’ is completely absent from the text. (35)

Ipi’ma ipa’=ima already=hsy

asu’ asu’ this

pa’anpanpu’si’ma pa’-apa-an-pu’=sa’=ima go-cont-ptc.ss.2/3sg-simil=del=hsy

pitetchapilalli,            ipi’ma    u’neipi pitertu-apa-ila-lli         ipa’=ima    u'ni-pi spin-cont-pass.outward-nfi.3sg  already=hsy  very.large=cl.fruit kalapipu’su’       nu’tulli.       Nu’an      ima kala-pi-pu’-su’      nu'-tu-lli       nu’-an      ima three-cl.fruit-simil-advl  be.so-vm-nfi.3sg>3sg  be.so-ptc.ss.2/3sg  hsy ukllinantununta’n                ima  ipa’ uklli-nan-tu-nunta’-an             ima  ipa’ be(come).day-pres.appl-vm-rep-ptc.ss.2/3sg>3sg  hsy  already a’leiteklun-pi nu’tulli. a’leiteklun-pi nu’-tu-lli five-cl.fruit  be.so-vm-nfi.3sg>3sg ‘As she was walking she spun, she had already made some three very large skeins. In this way, when it dawned again, she had already made five skeins.’

88

Pilar M. Valenzuela

In the next couple of examples classifiers play an anaphoric function. In (36a), the wife of Arakayu, spiritual owner of the fish and main character of the story, is reintroduced in the discourse by the full noun phrase ‘Arakayu’s wife;’ the predicate nominal contains the classifier -lun (cl.fem) in semantic agreement with the subject referent. The participant is then tracked through additional uses of -lun in (36c) and (36e). (36) a.      

Arakayu sadinta’ u’cha’payalunsha. Arakayu sadin=nta’ u’cha’paya-lun=sha Arakayu his.wife=add beautiful-cl.fem=dim ‘Arakayu’s wife was also a beautiful woman.’

b.      

Iñetller tantena’su’,     luya’su’, iñer=ler tanten-a’su’     luya’-a’su’ all=a admire-nmlz.3sg>3sg  covet-nmlz.3sg>3sg ‘Everyone admired her, coveted her,

c.      

“u’cha’payalunsha!”    tullina’. u’cha’paya-lun=sha   tu-llina’ beautiful-cl.fem=dim  say-nfi.3pl “what a beautiful woman!” people said.’

d.      

Nu’sik       shinenchek,   ker’enchek, nu’-sik      shin-enchek  ker’-enchek be.so-ptc.ds.3sg  long-hair  black-hair ‘Then, she had long, black hair,

e.      

munkuladalunsha . . . munku-lada-lun=sha round-face-cl.fem=dim and a pretty round face . . .’ (Valenzuela 2012: 135; Valenzuela 2016a: 363)

Example (37) is part of a procedural text on how to make chicha punta, a traditional Shiwilu drink made of corn and manioc. The referent ‘manioc’ is introduced by the noun ker’. Next, the classifier dan (cl.manioc) which is selected in semantic agreement with ker’ attaches to the numeral and serves an anaphoric function so that ker’ is not repeated. (37) Peksantapamu peksan-tu-apa-amu grind-vm-cont-ptc.ss.1sg.i

nana nana that

inkaluner inkaluner simultaneously

ker’unta ker’=unta’ manioc=add

a’ler’-apa-lek, alei’teklun-dan. roast-cont-nfi.1sg.i>3sg five-cl.manioc ‘While one grinds (the corn), one grills the manioc simultaneously, five whole maniocs.’ (Valenzuela 2012: 87; Valenzuela 2016a: 364)



3  Classifiers in Shiwilu (Kawapanan)

89

3.3.4  Adjective-like function  Yet another usage of Shiwilu classifiers represented in the corpus is that of helping to encode or intensify a quality or property, that is, an adjective-like function. For example, the animate classifiers -lun (cl.fem) and -pen (cl.male) may indicate that a person is good-looking or robust: ku’aper ‘woman’ > ku’aperlun ‘good-looking/robust woman.’ Another example involves the classifier -lu’ (cl.meat) which can convey that a body part is thick or fat, as in the following corpus example. (38) Tanpen Tanpen Fire.Spirit

mudei’ma muda’=ima person=hsy

shipen, shin-pen long-cl.male

piper’unkellu’, pipelladasu’. piper’-unker-lu’ piper’-lada-a’su’ (be)red-neck-cl.meat (be)red-face/eye-nmlz.3sg ‘The Fire Spirit was a tall man with a reddish thick neck and a reddish face.’ (Valenzuela 2012: 128) When applied to an animate entity, lada may designate either the face or the eyes. In  (38), this noun appears incorporated in the verb and is interpreted as referring to the Fire Spirit’s face. Text examples where lada has a different reading are shown in §3.3.5. The next two text extracts belong to a narrative about a woman who had two sonsin-law, Centipede and Vulture. The woman was very fond of Vulture, but despised Centipede. Therefore, when distributing the crop she would give Vulture’s wife the best, largest yams, while Centipede’s wife would get the worst, smallest ones. As indicated in §2, the classifier -pi (cf.fruit) applies to roundish yams, and this is what we find in (39) where this morpheme attaches to u’ni ‘(be) very large’. However, in (40) the reduced size of the yams given to Centipede’s wife is emphasized by resorting to the classifier -la (cl.seed), which occurs inserted in u’ñisha ‘(be) small’ (formed by u’ni ‘(be) very large’ + =sha dim). (39) U’nipisik u’ni-pi-sik (be)very.large-cl.fruit-ptc.ds.3sg

ima ima hsy

mama’ mama’ yam

ku’tin ku’tin her.daughter

enka’kuñi enka’-kun-lli give-going-nfi.3sg>3sg ‘She went to give the largest yams to her daughter (Vulture’s wife).’

90

Pilar M. Valenzuela

sadimi’na lli’la’ka’sui’ma, (40) ….nana Takuntekku’ ….nana Takuntek-ku’ sadin=ima=i’na lli’-la’-a’kasu’=ima that Centipede-dead his.wife=hsy=foc see-just-nmlz.1sg17>3sg=hsy u’ñillashala          mama’  enka’llima. u’ni-la=sha=la         mama’  enka’-lli=ima (be)very.large-cl.seed=dim=just yam  give-nfi.3sg>3sg=hsy ‘…to Centipede’s wife she gave the leftovers, the very small yams.’ (Valenzuela 2012: 177; Valenzuela 2016a: 367) The classifier -tek (cl.skin) has grammaticalized into a deprecatory marker in Shiwilu.18 For example, the expression ketllatek [ker’ ‘manioc’ + -la cl. bone + ‑tek deprec] can be used in reference to a small, thin, bad quality manioc. Note that -la (cl.bone) is  employed here instead of the expected -dan (cl.manioc) (§2); in addition, ‑tek conveys contempt on the part of the speaker (the deprecatory -tek is also found in (27)). The opposite meaning can be achieved by the diminutive morpheme =sha. The next pair of examples, taken from the narrative about Iguana and Tortoise, ­illustrate this contrast. In mythical times, Iguana and Tortoise were beautiful young girls. One day, they agreed to embellish each other before going to a party. First, Tortoise painted small, beautiful patterns on Iguana’s body. However, when switching positions Iguana painted large, ugly patterns on Tortoise’s back. The extracts in (41) and (42) appear towards the end of the story. (41) Nanamalek nana=malek that=reason

ipa’lalek ipa’la=walek now=lim

ekpa'palli ekpa’-apa-lli carry-cont-nfi.3sg>3sg

madu madu tortoise

u’nei     llinsetcheksa'…. u’ni     llinser-tek=sa’ (be)very.large pattern-deprec=del ‘In this way, even today, the tortoise bears very large, ugly patterns . . .’ (42) Nanamalek nana=malek that=reason

nana nana that

ipa’lalek ipa’la=walek now=lim

llinsetchanen llinser=sha=nen pattern=dim=poss.3sg

a’ñapalli          u’chimusha. a’-ñi-apa-lli         u’chimu=sha caus-exist-cont-nfi.3sg>3sg  beautiful=dim ‘Therefore, even today she (the iguana) has very beautiful, fine patterns.’ 17  The speaker mistakenly selected the nominalizer corresponding to the 1st person singular. The right suffix in this context is -a’su’ nmlz.3sg. 18  This process is also attested in Shawi with respect to the cognate classifier -te’.



3  Classifiers in Shiwilu (Kawapanan)

91

3.3.5  Disambiguating function  Recall that with animates the body part noun lada may translate either as ‘eye’ or ‘face’. In fact, in (38), verb-incorporated lada referred to a character’s face. Fortuitously, the corpus also comprises two texts where lada appears in the verb, this time making reference to someone’s eyes. Crucially, in these instances lada is preceded by -lu’, which I analyse as the classifier for meat or flesh. Before introducing the next two text examples it must be mentioned that the Shiwilu used to treat certain sight problems with breast milk or the morning dew. (43) …mudidekpen mudi-dek=pen milk-cl.liquid=poss.2sg

iyalek’amu iya-lek-amu des-ask-ptc.ss.1sg>3sg

weklek, wek-lek come-nfi.1sg

shaya’.        Nana’ka shayá’,       nana=a’ka man’s.older.sister:voc  that=contrst aku’lu’ladatupachinku aku’-lu’-lada-tu-pachinku put-cl.meat-seed/eye/face-vm-subj.2sg>1sg kua  musha’su’       lli’tapateku…. kua  mu-sha-a’su’     lli’-tu-apateku 1sg  be.nice-dim-nmlz.3sg  see-vm-subj.1sg ‘I have come to ask for your breast milk, lady. If you applied that into my eye I would be able to see well.’ (Bendor-Samuel 1981: 163) Example (44), taken from the Shiwilu Dictionary (Valenzuela et al. 2013), confirms the use of the sequence -lu’-lada in expressing a very similar circumstance. Alternative versions omitting the classifier -lu’ were rejected by language consultants. (44) Pulunker’ dew

wila child

dapi-sik have.pain.in.the eye-ptc.ds.3sg

aku’-lu’-lada-lek.            / * aku’-lada-lek put-cl.meat-eye/face-nfi.1sg.i>3sg  ‘When the child’s vision hurts, dew is put in his eye.’ (Valenzuela et al. 2013, entry: dapipalli) Finally, let us consider the following extract from a text narrating the Shiwilu’s first  journey to the town of Moyobamba, where they encountered the Spaniards. The latter are described as white people having blue, light-coloured eyes. The noun lada occurs twice, incorporated into descriptive predicates. In the first reference to the Spaniards’ eyes, lada is preceded by -lu’ (evidently the colour blue also contributes to the interpretation of lada).

92

Pilar M. Valenzuela

(45) Nanalupa’ nana=lupa’ that=approx.loc

Dadapu dadapu white

Muda’ muda’ person

nanpipalli, nanpi-pa-lli live-cont-nfi.3sg

kañellu’ladatusu’,         adawa’ladatusu’. kañer-lu’-lada-tu-su’        adawa’-lada-tu-su’ (be)blue-cl.meat-eye/face-vm-adjz  (be)light.colored-eye/face-vm-adjz ‘Over there lived white people, with blue eyes, light-coloured eyes.’ In conclusion, when lada incorporates in the verb, preposing the classifier -lu’ (cl. meat) to it might be a strategy to guide speakers into interpreting this noun as ‘eye’ rather than ‘face’. With minor differences, the discussion in this section confirms the functions of Shiwilu classifiers identified in Valenzuela (2016a: §4). The individuating usage described in that work is not attested in the corpus on which this chapter is based. Conversely, the disambiguating function of -lu’ (cl.meat) discussed here is not included in Valenzuela (2016a). Classifier incorporation and argument-reference is dealt with in §5.

4  Classifier stacking and reduplication 4.1  Classifier stacking As shown in §3.2, the prevalent function of Shiwilu classifiers is the productive derivation of new vocabulary, as illustrated by inya’- ‘urinate’ + -dek’ cl.liquid > inya’dek ‘urine,’ ilantu- ‘shoot’ + -lu’ cl.soil > ilantulu’ ‘gunpowder’, panpek ‘irapai palm19’ + -la cl.bone > panpekla ‘stem of the irapai palm’, etc. This word forming process can be recursive, so that a base already containing a classifier may take an additional such morpheme. In all instances, the last classifier serves as head. Consider the lexemes in (46). (46) a. inya’-dek-pi   [inya’- ‘urinate’ + -dek cl.liquid + -pi cl.fruit]   ‘bladder’ b. inya’-dek-llin   [inya’- ‘urinate’ + -dek cl.liquid + -llin cl.vine]   ‘urethra’ c. ilantu-lu’-la   [ilantu- ‘shoot’ + -lu’ cl.soil + -la cl.bone]   ‘cartridge’ d. iñinun panpek-la-si   [iñinun ‘bee hive’ + panpek ‘irapai palm’ + -la cl.bone + -si cl.cane]   ‘bee species’ (the entry to its hive is said to resemble the stem of the irapai palm) 19  Lepidocaryum gracile Martius.



3  Classifiers in Shiwilu (Kawapanan)

93

In (47a-b) the classifiers -dek [cl.liquid] and -lu’ (cl.soil) appear in the opposite order. (47) a. dadapu-dek-lu’   [dadapu ‘white’ + -dek cl.liquid + -lu’ cl.soil]   ‘white clay’ (obtained from white/clear waters and used to paint ceramic) b.      

lu’lek-lu’-dek [lu’lek ‘ischimi20 ant’ + -lu’ cl.soil + -dek cl.liquid] ‘Ischimi yacu’ in the local Spanish (lit. Black Ant River) (i.e. river on whose banks ischimi ants abound)

Lu’leklu’dek represents a familiar pattern found in hydronyms. Often, rivers are named after a plant, animal, or something else that abounds in the surrounding area, or according to an event said to have taken place on their banks. Hence hydronyms often involve the classifiers -lu’ (cl.soil, ‘area where something abounds or where something happened’) and -dek (cl.liquid, ‘river’). Furthermore, a hydronym may potentially exhibit a sequence of three classifiers. (48) a.       b.      

Milek-pi-lu’-dek yarina.palm-cl.fruit-cl.soil-cl.liquid ‘Yarinayacu’ in the local Spanish (lit. Yarina Palm River) (i.e. river on whose banks the fallen fruits of the yarina palm (Phytelephas macrocarpa) abound) Chipenñalu’dek chiper-nan-lu’-dek moriche.palm-cl.trunk-cl.soil-cl.liquid (i.e. river on whose banks there are fallen moriche palm trees (Mauritia flexuosa)).

Different derivational layers are also attested in (49c). Interestingly, the classifier -dek (cl.liquid) occurs twice, before and after -lu (cl.soil). (49) a. waka’mudi-dek   udder-cl.liquid  ‘milk’ b. waka’mudi-dek-lu’   udder-cl.liquid-cl.soil  ‘butter, cheese’ c. waka’mudi-dek-lu’-dek   udder-cl.liquid-cl.soil-cl.liquid

‘butter soup’

The instances of classifier stacking given above show right-headedness; i.e. the classifier to the right indicates the type of entity we are dealing with. Thus, in (47a) dadapu-dek-lu’ 20  Species of black ant.

94

Pilar M. Valenzuela

(ending in -lu’ cl.soil) refers to the clay, whereas in (47b) Lu’lek-lu’-dek (ending in -dek cl.liquid) is the name of a river. In (46a-c) the last classifier describes the shape and size of the referent (bladder, urethra, cartridge), and the preceding classifier corresponds to the consistency of the substance contained in/conducted by the noun referent (urine, gunpowder). Similarly, in (49c) waka’mudi-dek-lu’-dek, ending in -dek (cl.liquid), is the name of a soup, and the preceding morpheme -lu’ (cl.soil) refers to the consistency of the defining ingredient. 4.2  Classifier reduplication An additional characteristic of Shiwilu classifiers that, to my knowledge, has not yet been attested in any other classifying language is their ability to reduplicate. Let us return to the story about Iguana and Tortoise, where these characters are depicted as two very pretty girls. In the text extract in (50), reduplication of -lun (cl.fem) combines with the adjective u’chimu to intensify the characters’ beauty. This morphological process may apply when the referent is constituted of two or more entities. (50) Napi’ napi’ long.ago

ima ima hsy

Pekkuala Pekkuala Iguana

Madulek Madu=lek Tortoise=com

muda’ muda’ people

ñanna’ ñi-anna’ exist-ptc.ss.3pl

katu’ta katu’ta two

wilalunlusa’   u’chimulunlunshala        ñiwiñina’. wila-lun=lusa’  u’chimu-lun-lun=sha=la      ñi-win-llina’ child-cl.fem=pl  good-cl.fem-cl.fem=dim=affect  exist-frust21-nfi.3pl ‘It is said that long time ago, when Iguana and Tortoise were people, they were two very pretty girls (with endearment).’ (jeb-15-01-13-pvb-1, Valenzuela 2016a: 366)

Reduplication of Shiwilu classifiers was first reported in Valenzuela (2016a: 366), where this strategy is described as a peculiarity of -lun (cl.fem). Nonetheless, further research has revealed that other classifiers may also undergo this process. Consider the following sentence involving the male classifier -pen. (51)

Nadi’nek u’chapayapenpenchala nadi’nek u’chapaya-pen-pen=sha=la young.man good-looking-cl.male-cl.male=dim=affect ‘Only very good-looking young men passed by.’

naku’llina’. naku’-llina’ pass-nfi.3pl

As in the preceding example (50), the reduplicated classifier in (51) follows an adjective. Furthermore, the ending =sha=la ~ =cha=la22 adds a sense of intensification, 21  Since the suffix -llina’ may convey present or past tense, -win is added to force a past reading. 22  The ending =sha=la results from the combination of the diminutive =sha ~ =cha which has also affective and intensifying meanings, and =la, analysed as a ‘diminutive with affection or esteem’ in Bendor-Samuel (1981: 140).



3  Classifiers in Shiwilu (Kawapanan)

95

e­ ndearment, exclusiveness. But reduplication is not limited to -lun (cl.fem) and -pen (cl.male), as shown by (52)–(53) where inanimate classifiers also exhibit this behaviour. ka’-pa’-tu-llidek. (52) Mu-lu’-lu’=sha=la (be)good-cl.meat-cl.meat=dim=affect eat-assist.appl-vm-nfi.1pl.e>3sg ‘We were invited to eat only very good meats.’ (lit. we helped to eat only very good meats) (53) Nanta-llin-llin=sha=la be.hard-cl.vine-cl.vine=dim=affect ‘These are all very resistant ropes.’

nuka’-a. be-3sg

In sum, the data above show that classifier reduplication is productive in Shiwilu albeit in a specific construction. The reduplicated element attaches to an adjective or descriptive predicate to intensify a characteristic of a plural referent. Further, addition of =sha=la yields a sense of intensification, endearment, or exclusiveness.

5  Classifier incorporation and argument reference 5.1  Incorporation of one classifier Cross-linguistically, when incorporated in the verb, classifiers usually exhibit an absolutive configuration; i.e. they tend to signal some characteristic of the (inactive) intransitive subject or the object (Derbyshire and Payne 1990: 245; Aikhenvald 2000: 162). A less common pattern is attested in Bora/Miraña (Boran, Peru, and Colombia), where incorporated classifiers make reference to the subject, be it intransitive or transitive, but not the object (Grinevald and Seifart  2004: 269; Thiesen and Weber 2012: 127–9, 173–4). Again, Shiwilu is noteworthy in this respect since incorp­ orated classifiers may realize an intransitive subject, an object, and even a transitive subject (cf. Aikhenvald’s (2017: 378) statement that ‘Verbal classifiers never categorize transitive subject’). The data in (54) and (55) illustrate, respectively, animate and inanimate classifiers in intransitive verbs. In (a) the base verb is inactive, whereas in (b) it is active. (54) a.      

Chiminluñi. chimin-lun-lli die-cl.fem-nfi.3sg ‘The woman died.’

b.      

Tekka’luñi. tekka’-lun-lli run-cl.fem-nfi.3sg ‘The woman ran.’

96 (55)

Pilar M. Valenzuela a.      

Chiminañi chimin-nan-lli die-cl.trunk-nfi.3sg ‘The tree died.’

(nala). nala tree

b.      

Tekka’deklli (La’pir’). tekka’-dek-lli La’pir’ run-cl.liquid-nfi.3sg Rumiyacu.river ‘(The Rumiyacu river) has a torrential flow.’

When transitive verbs are involved, all other things being equal, animate classifiers tend to be interpreted as subject, and inanimate classifiers as object. In the next e­ xamples, the base verbs are uklu- ‘spread sth.’, di’- ‘kill (without an expressed, specific object)’, and di’-tu- ‘kill sb./sth.’. (56) Uklu-mek-lli. spread-cl.leaf-nfi.3sg>3sg ‘S/he spread the sheets.’ (57) a.         b.        

Di’luñi. di’-lun-lli kill-cl.fem-nfi.3sg ‘The woman killed (someone)’ *‘Someone killed the woman.’ Di’luntullen. di’-lun-tu-llen kill-cl.fem-vm-nfi.1sg>2sg ‘I (woman) killed you’ (talking to the deceased). *‘I killed you (woman).’

One function of the suffix -tu is to serve as applicative bringing a location into the core. This is what we observe in (58) where the tree trunk, which constitutes the location, is coded as object through the incorporated classifier -nan (cl.trunk) (see also (31j, k, l)). (58) Pa’-nan-tu-lli. walk-cl.trunk-vm-nfi.3sg>3sg ‘She/He walked on the tree trunk.’ As indicated in Valenzuela (2016a: 368), in addition to animacy ‘other factors such as  linguistic and non-linguistic contexts, verb semantics, and speech participants’ knowledge of the world also affect the interpretation’ of an incorporated classifier as either subject or object of a transitive verb. In the following causative construction, the



3  Classifiers in Shiwilu (Kawapanan)

97

intransitive base verb requires an animate subject, hence the inanimate classifier ‑lu’ (cl.soil) is read as causer and subject of the whole construction. (59) A’-chinku’-lu’-pa-lli. caus-cough-cl.soil-cont-nfi.3sg>3sg ‘The dust is making him/her cough.’ In (60) the inanimate classifier -mek may be interpreted as transitive subject or object, depending on the non-linguistic context. (60) Sekwa’-mek-lli. scratch-cl.leaf-nfi.3sg>3sg ‘She/He scratched the leaf.’ / ‘She/He was scratched by the leaf.’ (e.g, the leaf has tiny thorns) (This example was first given in Valenzuela (2016a: 368); the difference in pronunciation is due to the participation of a different speaker) The ambiguity in (60) can be resolved if the verb takes a bipersonal suffix. Recall from §1.2 that these suffixes code the subject and a speech-act participant object, and thus the incorporated inanimate classifier is interpreted as transitive subject. (61) Sekwa’-mek-llen. scratch-cl.leaf-nfi.3sg/1sg>2sg ‘You were scratched by the leaf.’ (lit. ‘The leaf scratched you.’) (Valenzuela, 2016a: 368; pronunciation corresponding to a different speaker) The data in (54)–(61) were obtained through elicitation, and independently confirmed by three different native speakers. However, the text corpus also contains an instance of an incorporated classifier that refers to the transitive subject. The next example is taken from a story about a boy who manages to kill a black puma with a spear. The inanimate classifier -nan (cl.trunk) is selected in semantic agreement with rejón ‘spear’, which plays the A syntactic role (rejón is a Spanish loan). rejón. (62) Da’suketñantapilalli da’-suker’-nan-tu-apa-ila-lli rejón entering-pierce-cl.trunk-vm-cont-outward-nfi.3sg>3sg spear ‘The spear pierced (the black puma’s throat) and entered it completely.’ (Bendor-Samuel 1981[1958]: 155, lines 27–8) From the story context it is clear that the spear is in fact the transitive subject in (62). But again, resorting to a bipersonal suffix can help us see the argument structure more clearly. In the modified sentence in (63), the bipersonal suffix -llun indicates that the object is the speaker and, therefore, the classifier -nan (cl.trunk) can only

98

Pilar M. Valenzuela

refer to the subject. (The language consultant preferred to use the manner prefix pa- ‘do sth. by stabbing’ instead of da’- ‘entering’.) (63) Pasuketñantullun pa-suker’-nan-tu-llun stabbing-pierce-cl.trunk-vm-nfi.3sg>1sg ‘(The spear) speared me in the upper leg.’

tulawekkek. tula=wek=kek upper.leg=poss.1sg=loc

5.2  Incorporation of two classifiers or a classifier and a noun A Shiwilu verb may incorporate up to two classifiers, or a classifier and a noun. When two classifiers are incorporated in the verb, the first one must be interpreted as referring to the object and the second one to the subject. (64) a.      

Di’penluñi. di’-pen-lun-lli kill-cl.male-cl.fem-nfi.3sg>3sg ‘The woman killed the man.’ / * ‘The man killed the woman.’

b.      

Di’lunpenñi. di’-lun-pen-lli kill-cl.fem-cl.male-nfi.3sg>3sg ‘The man killed the woman.’/ * ‘The woman killed the man.’

In (65) the inherently intransitive verb tekkua- ‘be afraid’ increases its valency after taking the suffix -tu, thus yielding the transitive tekkua-tu- ‘fear’. (65) Tekkuatektuluñina’ tekkua-tek-tu-lun-llina’ fear-cl.skin-vm-cl.fem-nfi.3pl>3sg ‘These young girls fear vampires.’

ishektek ishek-tek bat-cl.skin

asu’ asu’ this

wilalunlusa’. wila-lun=lusa’ child-cl.fem=pl

The position of incorporated elements in the verb relative to the valency modifier suffix ‑tu is an interesting topic that requires detailed treatment. Here I limit myself to a few preliminary observations. Note in (65) that the classifier corresponding to the object precedes -tu while the one corresponding to the subject follows it. Example (66) presents the same morpheme order, with the difference that the first incorp­ orated element is a noun and the second one a classifier. The verb base is composed of the intransitive di’ser- ‘burn (without explicit object)’ and the suffix -tu which increases its valency so that an object is required. (66) Di’setpidektupenchañina’. di’ser’-pidek-tu-pen-sha-llina’ burn-house-vm-cl.male-dim-nfi.3pl>3sg ‘The little men burned the house.’ (‘little men’ with mocking overtone)



3  Classifiers in Shiwilu (Kawapanan)

99

There are, nonetheless, instances where the two incorporated classifiers occur after the suffix -tu. Consider (67). (67) Apetchuteklunshallina’. aper’-tu-tek-lun-sha-llina’ be.spoiled-vm-cl.skin-cl.fem-dim-nfi.3pl>3sg ‘The little women stole the fur.’ (‘little women’ with mocking overtone) The morpheme order in (67) may be motivated by the fact that the semantics of aper’‘be spoiled’ is not the intransitive version of aper’tu- ‘steal’, e.g. it does not mean ‘steal (without explicit object)’. This suggests that synchronically the predicate at hand might be best analysed as aper’tu-, that is, without segmentation of ‑tu. This proposal would account for the position of the incorporated classifiers in (67), but not for the morpheme order in (43), where both incorporated elements precede the suffix -tu.23 This example is repeated as (68), for convenience. (68) …mudidekpen mudi-dek=pen milk-cl.liquid=poss.2sg shaya’. shayá’ , man’s.older.sister:voc

iyalek’amu iya-lek-amu des-ask-ptc.ss.1sg>3sg

weklek, wek-lek come-nfi.1sg

Nana’ka nana=a’ka that=contrst

aku’lu’ladatupachinku aku’-lu’-lada-tu-pachinku put-cl.meat-seed/eye/face-vm-subj.2sg>1sg kua  musha’su’       lli’tapateku…. kua  mu-sha-a’su’     lli’-tu-apateku 1sg  be.nice-dim-nmlz.3sg  see-vm-subj.1sg ‘I have come to ask for your breast milk, lady. If you applied that into my eye I would be able to see well.’ (Bendor-Samuel 1981: 163) A peculiarity of (68) is that both incorporated elements refer to the speaker’s eyes. In fact, it has been argued that the function of the classifier -lu’ (cl.meat) in this context is to disambiguate the semantics of the noun lada. Therefore, it can be concluded that the classifier-noun sequence at hand, encoding the same referent, cannot be interrupted by -tu or any other morpheme since it encodes a single referent.

6 Conclusions This chapter has explored a number of grammatical properties exhibited by the Shiwilu nominal classifying system that are typologically unusual (or maybe even unique) 23  Note that (44), a similar sentence taken from the Shiwilu Dictionary, does not include the suffix -tu.

100

Pilar M. Valenzuela

and/or constitute counterexamples to proposed cross-linguistic generalizations. First, Shiwilu classifiers can appear in a remarkable number of morphosyntactic contexts. In fact, these morphemes can attach to the right of numeral and non-numeral quantifiers, demonstratives, nouns, adjectives, formatives that take a classifier to derive a noun or adjective, question words, verb stems, and even personal pronouns. However, unlike what is attested in other classifying languages, Shiwilu classifiers are not required on any type of nominal modifier; even low numbers and demonstratives are usually attested without a classifier. In addition, it has been shown that Shiwilu classifiers have the ability to reduplicate after an adjective or descriptive predicate in order to intensify a characteristic exhibited by a plural entity. This has been attested in a specific construction. To my knowledge, classifiers on personal pronouns and productive classifier reduplication have not been reported in the literature for any other language. Finally, when incorporated in the verb Shiwilu classifiers can refer not only to the O and S arguments, but also to the A. Again, this makes Shiwilu unusual since incorporated classifiers (and nouns) rarely realize the transitive subject. A question that has been overlooked in the study of classifier systems concerns the frequency with which classifier morphemes occur in discourse. The present study has focused on this issue. By analysing a corpus of spontaneous texts it has been shown that classifiers are relatively infrequent in Shiwilu (at least when compared with languages such as Bora). A question that arises is why a language having at its disposal a set of over 20 classifiers, that can potentially appear in such a range of morphosyntactic environments, presents low text frequency of classifying forms. A related issue concerns the non-obligatoriness of classifiers in the language. A possible answer to these interrogations may be that the system is fairly recent and did not have the time to fully grammaticalize. This hypothesis is consistent with the observation that classifier choice is fairly transparent. Alternatively, it may be the case that the system is falling out of use, possibly as a result of language obsolescence. Crucially, it has been possible to compare texts from present day Shiwilu with texts collected in the 1950s when the language was used in everyday life. No significant difference with respect to classifier frequency was observed between the two datasets. Nonetheless, it is possible that even in the 1950s classifiers had already lost prominence in the language. This work has also examined the relative frequency of derivational vs. non-derivational functions of Shiwilu classifiers. The productive derivation of new lexemes is by far their most frequent function, representing 82.5% of occurrences. Therefore, Shiwilu does not lend support to the hypothesis that the main function of classifier systems has to do with participant reference in discourse. Finally, it has been suggested that classifying languages may differ significantly with respect to the text frequency of their classifier morphemes. It is hoped that researchers investigating classifier systems address this question in the future so that comparison across languages becomes possible. The results can help us better understand the profile of classifier systems, assess their relative vitality, and perhaps ascertain aspects that may prove relevant in understanding their diachronic development.



3  Classifiers in Shiwilu (Kawapanan)

101

Acknowledgements I would like to express my deepest gratitude to my long-term Shiwilu collaborators for their generosity and patience in helping me understand the workings and nuances of their native language’s classifier system: Mrs Emérita Guerra Acho, Mrs Julia Inuma Inuma, Mr Meneleo Careajano Chota, and Mr Fidel Lomas Chota. Of course, any shortcomings or mistakes are exclusively my own. This chapter has benefited from valuable comments by Alexandra Aikhenvald to whom I am very grateful. I would also like to thank the organizers and other participants in the International workshop on classifiers (Language and Culture Research Centre, James Cook University, Cairns, August 2017) who motivated me to further delve into the Shiwilu classifier system.

References Admiraal, F. and S.  Danielsen. 2014. ‘Productive compounding in Baure (Arawakan)’, in S. Danielsen, K. Hannss, and F. Zúñiga (eds), Word Formation in South American Languages. Amsterdam: John Benjamins, 79–112. Aikhenvald, A. Y. 1994. ‘Classifiers in Tariana’. Anthroplogical Linguistics 36: 407–65. Aikhenvald, A. Y. 2000. Classifiers: A typology of noun categorization devices. Oxford: Oxford University Press. Aikhenvald, A. Y. 2002. Language contact in Amazonia. Oxford: Oxford University Press. Aikhenvald, A. Y. 2012. The languages of the Amazon. Oxford: Oxford University Press. Aikhenvald, A. Y. 2017. ‘A typology of noun categorization devices’, in A. Y. Aikhenvald and R.  M.  W.  Dixon (eds), The Cambridge Handbook of Linguistic Typology. Cambridge: Cambridge University Press, 361–404. Aikhenvald, A. Y. (this vol.) ‘A view from the north: Genders and classifiers in Arawak languages of north-west Amazonia’. Barnes, J. 1990. ‘Classifiers in Tuyuca’, in D. L. Payne (ed.), Amazonian Linguistics: Studies in Lowland South American Languages. Austin: University of Texas Press, 273–92. Bendor-Samuel, J. 1981[1958]. The structure and function of the verbal piece in the Jebero language. Perú: Summer Institute of Linguistics. Derbyshire, D. C. and D. L. Payne. 1990. ‘Noun classification systems of Amazonian languages’, in D. L. Payne (ed.), Amazonian Linguistics. Studies in Lowland South American Languages. Austin: University of Texas Press, 243–72. Dryer, M. 2007. ‘Word order’, in T.  Shopen (ed.), Clause Structure, Language Typology and Syntactic Description, Vol. 1. Second Edition. Cambridge: Cambridge University Press, 61–131. Gómez-Imbert, E. 2007. ‘Tukanoan nominal classification, the Tatuyo system’, in W. L. Wetzels (ed.), Language endangerment and endangered languages. Linguistic and anthropological studies with special emphasis on the languages and cultures of the Andean-Amazonian border area. Leiden: CNWS Publications, 401–28. Grinevald, C. 2000. ‘A morphosyntactic typology of classifiers’, in G.  Senft (ed.), Systems of Nominal Classification. New York: Cambridge University Press, 50–92. Grinevald, C. and F. Seifart. 2004. ‘Noun classes in African and Amazonian languages: Towards a comparison’. Linguistic Typology 8, 2: 243–85.

102

Pilar M. Valenzuela

Hart, H. 1988. Diccionario chayahuita-castellano (Canponanquë nisha nisha nonacasoʼ). Serie Lingüística Peruana 29. Yarinacocha, Peru: Ministerio de Educación and Instituto Lingüístico de Verano. Mihas, E. (this vol.) Genders and classifiers in Kampa (Arawak) languages of Peru. Payne, D. L. 1986. ‘Noun classification in Yagua’, in C. Craig (ed.), Noun Classes and Categorization: Proceedings of a Symposium on Categorization and Noun Classification, Eugene, Oregon, October 1983. Amsterdam: John Benjamins, 113–31. Payne, D. L. 1987. ‘Noun classification in the Western Amazon’. Language Sciences 9, 1: 21–44. Seifart, F. and D.  L.  Payne 2007. ‘Nominal Classification in the North West Amazon: Issues in  Areal Diffusion and Typological Characterization’. International Journal of American Linguistics 73, 4: 381–7. Thiesen, W. and D.  Weber. 2012. A Grammar of Bora, with special attention to tone. Dallas, Texas: SIL International Publications in Linguistics 148. Valenzuela (Bismarck), P.  M. 2011. ‘Argument Encoding and Pragmatic Marking of the Transitive Subject in Shiwilu (Kawapanan)’. International Journal of American Linguistics 77: 91–120. Valenzuela (Bismarck), P. M. 2012. Voces Shiwilu: 400 Años de Resistencia Lingüística en Jeberos. Lima: Fondo Editorial Pontificia Universidad Católica del Perú. Valenzuela (Bismarck), P. M. 2015. ‘¿Qué tan “amazónicas” son las lenguas kawapana? Rasgos centro-andinos y evidencia para una posible sub-área lingüística’. Lexis 39 (1): 5–56. Valenzuela (Bismarck), P.  M. 2016a. ‘Classifiers in Shiwilu (Kawapanan) in Northwestern Amazonian Perspective’. Anthropological Linguistics 58 (4): 333–80. Valenzuela (Bismarck), P.  M. 2016b. ‘ “Simple” and “Double” Applicatives in Shiwilu (Kawapanan)’. Studies in Language 40 (3): 513–50. Valenzuela (Bismarck), P. M. 2017. ‘The Valency Modifier -tu in Shiwilu: Applicative-Antipassive Polysemy and Beyond’, Symposium on Linguistic Complexity, Universidad de Sonora, Mexico, November 2017. Valenzuela (Bismarck), P. M. 2018. ‘Difusión de rasgos andinos y elementos para una sub-área lingüística intermedia Andes-Amazonia en el norte del Perú’, in A. Regúnaga, S. Spinelli, and M. E. Orden (comps.), IV Encuentro Internacional de Lenguas Indígenas Americanas. Libro de Actas. Santa Rosa, Argentina: Universidad Nacional de la Pampa. 657–84. Valenzuela (Bismarck), P.  M. Forthcoming. ‘Kawapanan Family’, in P.  Epps and L.  Michael (eds), Handbook of Amazonian Languages. Berlin: Mouton de Gruyter. Valenzuela, P. M. and C. Gussenhoven. 2013 ‘Illustration of the IPA: Shiwilu (Jebero)’. Journal of the International Phonetic Association 43 (1): 97–106. Valenzuela, P., M. Careajano, E. Guerra, J. Inuma, and F, Lachuma. 2013. Kirka’ Llawer’lla’la’ Ñak: Diccionario Shiwilu-Castellano. Lima: Federación de Comunidades Nativas de Jeberos (FECONAJE). Wojtylak, K. I. 2016. ‘Classifiers as derivational markers in Murui (Northwest Amazonia)’, in L. Körtvélyessy, P. Štekauer, and S. Valera (eds), Word formation across languages. Newcastle upon Tyne: Cambridge Scholars Publishing, 393–425.

4 A view from the North Genders and classifiers in Arawak languages of north-west Amazonia A l e x a n dr a Y. A i k h e n va l d

1  The problem Several types of noun categorization devices can co-exist in one language (see Aikhenvald 2017 and references there, for examples). Genders and other noun ­categorization devices—be it numeral classifiers, or classifier of other kinds—are generally thought of as being relatively independent from one another (see further examples in Chapter 2, on Kampa languages; Chapter 5, on Zamucoan languages; and Chapter 7, on Toba, a Guaycuruan language). Gender will be marked on some modifiers, on verbs, and also on nouns themselves. Numeral classifiers will be associated with number words, possessive classifiers with possessive constructions, and so on.1 Co-existing and overlapping systems of genders and classifiers are cross-linguistically uncommon. In a number of Arawak languages from north-west Amazonia, two genders—feminine and non-feminine—are obligatorily marked on verbs and nouns, and demonstratives and other modifiers within a noun phrase. Classifiers used on number words, and in a variety of other contexts, categorize the noun in terms of its physical properties, and also distinguish gender. Gender is thus integrated within the system of classifiers. Gender markers may co-occur with classifiers in one word. Gender distinctions and gender markers are uniform across the Arawak language family, and can be reconstructed for the proto-language. Classifiers vary, in terms of

1  For instance, gender classes and numeral classifiers co-exist as relatively independent systems in a few Dravidian and Indic languages. Malto, a South Dravidian language, has thirty numeral classifiers—each categorizes the noun in terms of its animacy, form, shape, and also size. This is in addition to two genders: male human versus the rest in the singular, and human-non-human in plural (Mahapatra 1979; see also Emeneau 1956; see also Aikhenvald 2000: 184–201). Genders and Classifiers. First edition. Alexandra Y. Aikhenvald and Elena I. Mihas (eds) This chapter © Alexandra Y. Aikhenvald 2019. First published in 2019 by Oxford University Press

104

Alexandra Y. Aikhenvald

the contexts in which they occur and in terms of their forms. They are likely to have developed later and perhaps independently in individual subgroups. Two major questions arise: 1. Firstly, how do gender distinctions fit in within a system of classifiers? 2. Secondly, which classifier contexts can be considered historically older, and which are more innovative? In §2, I start with some examples of genders and classifiers as co-existing systems in Arawak languages of north-west Amazonia. In §3, I outline some relevant features of Arawak languages, and then briefly turn to the languages of north-west Amazonia, a locus of considerable diversity even between closely related languages (and the putative home-land of Proto-Arawak: see Aikhenvald 2013b). In §4, I offer an overview of gender and classifier contexts in the Arawak languages of north-west Amazonia and attempt to address questions 1 and 2. In §5, I look at the origins of classifier forms. Some functions of classifiers are discussed in §6. The last section contains a summary.

2  Genders and classifiers in Arawak languages of north-west Amazonia: An illustration Arawak languages of north-west Amazonia are grammatically quite diverse, ­especially in their noun categorization devices. Warekena of Xié is an example of a language with numeral classifiers and gender as relatively independent systems. Example (1) (Aikhenvald 1998: 298, author’s fieldwork) illustrates a numeral classifier for humans on the number word ‘one’ and agreement gender on the adjective and the verb. Gender-sensitive forms and classifiers are in bold face. ne-yawa (1) peya one+num.cl:human human-fem yu-tapapa 3sgf-go ‘One good woman went off . . .’

weduana-ɾi-yawa good-adj-fem

Warekena of Xié

Gender in Achagua is partly independent of classifiers, and partly integrated within the classifier system. Gender agreement on adjectives is marked with suffixes ‑i ‘non-­feminine’, -u ‘feminine’ (Wilson 1992: 34; see also Meléndez 1998: 75), as shown in (2) and (3). (2) waʃ iálikua-ezi man-masc.sg ‘strong man’

(3) iina-etoo woman-fem.sg ‘strong girl’

ka-dánani-i attr-strength-masc.sg ka-dánani-u attr-strength-fem.sg

Achagua

Achagua



4  Genders and classifiers in Arawak languages of north-west Amazonia

105

Suffixes -ezi ‘masculine or non-feminine’ and -etoo ‘feminine’ occur on nouns as nonword-class changing suffixes and on verbs as markers of word-class-changing derivations. In addition, the language has circa twelve classifiers used with number words. Markers of agreement gender occur together with the classifier -na for animate mammals (Wilson 1992: 63). (4) áaba-na-i one-num.cl:mammal-masc ‘one male dog’ (5) áaba-na-u one-num.cl:mammal-fem ‘one female dog’

Achagua

auli dog

Achagua

auli dog

A number of languages have the same, or almost the same, classifier set used in several morphological contexts, creating a multiple classifier system. Baniwa of Içana-Kurripako dialect continuum has forty to fifty classifiers used with number words, adjectives, as derivational suffixes on nouns, and in predicative possessive constructions. Gender agreement appears in demonstratives and cross-referencing prefixes. Example (6) comes from the Hohôdene dialect of Baniwa of Içana (Aikhenvald 2007: 488 and author’s fieldwork).2 The classifier -pi ‘long objects’ used with an inanimate referent has the same form in all the contexts. apa-pi (6) lhiahi dem.prox.nonfem one-cl:long ‘this one long bone’

iñapi bone

maka-pi big-cl:long

Hohôdene Baniwa

If we are dealing with an animate and gender-differentiable referent, the adjective will take a classifier and a gender agreement marker. The demonstrative will agree with the noun in gender (but will not take any classifiers). The number word ‘one’ contains just the classifier for flying animates. (7)

ru-kapa-ni-pida Hohôdene Baniwa 3sgfemA/Sa-see-3sgnonfemO/So-reported rhuahi apa-apa karaka maka-apa-ru dem.prox.fem one-cl:flying.anim chicken big-cl:flying.anim-fem.sg ‘This one big hen reportedly saw him.’

Tariana employs feminine and non-feminine genders with personal pronouns, and cross-referencing prefixes; classifiers are used in other contexts, including demonstratives, adjectives, and number words. Example (8) illustrates the classifier for females (-ma). 2  All the examples from Baniwa of Içana in this chapter are from Hohôdene dialect (unless indicated otherwise).

106

Alexandra Y. Aikhenvald

inaru pa-ma   du-nu-pidana (8) ha-ma dem.inan-cl:fem woman one-cl:fem  3sgfem-come-rem.p.rep di-na    matsia-ma-pu-pidana     duha 3sgnf-obj  good-cl:fem-very-rem.p.rep  she ‘This one woman reportedly came to him, she was very beautiful.’

Tariana

Example (9) illustrates the classifier for round objects (-da). hipa-da (9) ha-da dem.inan-cl:round stone-cl:round aı-se    di-swa-naka there-loc  3sgnf-stay-pres.vis ‘This big stone is there.’

hanu-da big-cl:round

Tariana

What the examples of all these languages have in common is the presence of gender distinctions in at least one of the following contexts: on verbs, personal pronouns, and also on demonstratives and adjectives. This reflects a notable feature of Arawak languages reconstructable to the proto-language. We now turn to some relevant grammatical information on the family.

3  Arawak languages and their grammatical features The Arawak language family is the largest in South America in terms of its geographical spread. There are over forty extant languages and a few dozen extinct ones. Arawak languages are spoken in at least ten locations north of the River Amazon, and at least ten to the south of it. The distribution of extant and recently extinct languages is shown in Map 1.3 The genetic unity of Arawak languages was first recognized by Father Filippo Salvadore Gilij as early as 1782. The recognition of the family was based on a comparison of pronominal cross-referencing prefixes in Maipure, a now extinct language

3  The internal classification of Arawak languages remains a matter of some debate. A number of grammatical and lexical traits distinguish languages spoken to the north of the River Amazon from those ­spoken to the south. Well-established subgroups include Kampa in Peru, South Arawak languages in Brazil and Bolivia, Xinguan Arawak languages, and a few small North Arawak groupings in Brazil, Colombia, and Venezuela. These include the Ta-Arawak subgroup, with Guajiro (or Wayuu-naiki), Añun (or Parauhano) in the region of Peninsula Guajira in Venezuela and Colombia, Garifuna (or Black Carib) in Central America, and Lokono (alternatively known as Dian, or Arawak) in Guyana, French Guyana, and Suriname (and also the long-extinct Taino, the language of the first indigenous group encountered by Columbus), and the Uapuí subgroup in the Upper Rio Negro Basin. An alternative name for the family is Maipuran. The term Arawakan was earlier used for the combination of a well-established genetic subgrouping (known as Maipuran) and a number of other groups not demonstrably related (see, e.g. Noble 1965, Matteson 1972) and is hence avoided by the majority of experts in the field. A detailed discussion of the family is in Aikhenvald (1999, 2002, 2012a: 32–6, 2018a; a comprehensive bibliography and an up-todate classification is in Aikhenvald 2015).

90°W

80°W

70°W

60°W

†Tanio

20°N †Island Carib (or Iñeri)

Garifuna (Black Carib) 0

1,000

km

Central America

10°N

GUA SHE

CAQ Guyana LAR BG

BWC BR

TN

MN

R.

Ne

MW

mazon R. A

g ro

a

eir

ad

R.

M

R.

CM

R.A

rag MO

uap

oré

MO BA MO

PAU Bolivia

MO

WA YW MH

PA SL

Brazil

20°S

TE KI

Brazil

a R . P TE

30°S

0

500

70°W

km

Pa R.

Argentina

ra g

uay

Paraguay

ná

R.G

ra

PI CHO PI Peru PI CA CA Iñ CA

Dio s

AM

uai

AP AP

R. M adre de

i

ayal

R.Uc

10°S

WP

WP

French Guiana

LAR PR PR

a

oco

o

rin

jõs

R.V CB YC au 0° R R.Caq pés uetá .Pu tu m Ecuador ayo RS

R.O

ranc

PIA

R.B

Colombia

Suriname

LAR

R . Xi n g u

Venezuela PIA

pa

PAR

Ta

10°N

1,000

60°W

Map 1  Extant Arawak languages Source: Aikhenvald (2012a: 34–5)

50°W

108

Alexandra Y. Aikhenvald Map 1 Continued Key: AM Amuesha (or Yanesha’) AP Apurinã BA Baure BG Baniva of Guainia (with closely related †Yavitero and Warekena of Xié) BR †Baré BWC Baniwa of Içana/Kurripako CA Kampa languages (Machiguenga, Nanti, Nomatsiguenga, Ashaninka, Ashéninka, Pichis, Alto Perené, Pajonal) CAQ Caquetio CB Cabiyari (or Kawyari) CHO Chontaquiro CM Chamicuro GUA Guajiro (or Wayuu-naiki) Iñ Iñapari KI Kinikinao LAR Lokono Arawak MH Mehinaku (Xinguan Arawak)

MN †Manao MO Mojo (Ignaciano, Trinitário) MW Mawayana PA Paresi (or Haliti) PAR Parauhano (or Añun) PAU Paiconeca (and the extinct Paunaca) PI Piro (or Yine) PIA Piapoco (or Dzase) PR Palikur RS Resígaro SHE †Shebayo SL Salumã (or Enawenê-Nawê) TE Terêna TN Tariana WA Waura or Waujá (Xinguan Arawak) WP Wapishana YC Yucuna YW Yawalapiti (Xinguan Arawak)

from the Orinoco Valley, and in Moxo from Bolivia. The limits of the family were established by the early twentieth century. All Arawak languages are synthetic, predominantly head-marking and suffixing with a closed set of prefixes (see Aikhenvald 1999, 2002: 288–95, 2018a–b, forthcoming, for a summary of Proto-Arawak grammar). Most of the person-marking prefixes are uniform and stable across the family. A common feature of Arawak languages is marking ­participants—including subjects and in many instances, also objects and other grammatical roles—on the verb with bound morphemes (see Aikhenvald 1999: 88–9, forthcoming; Mihas 2017). Personal prefixes typically mark the subject (A/S) on verbs and also the possessor on nouns. In at least two thirds of the languages, personal suffixes or enclitics express the object (O), and the subject of stative verbs (So) and/or the subject of non-verbal predicates (see Aikhenvald forthcoming for further details). The majority of Arawak languages (with few exceptions) do not employ cases for marking grammatical relations. The reconstructed system in Table  1 distinguishes two numbers (singular and ­plural) and three persons (first, second, and third, with feminine and non-feminine gender distinctions in third person singular in the majority of languages). Additional prefixes which go back to Proto-Arawak include *ka- ‘attributive, relativizer’ and *ma- ‘privative’. Arawak languages north of the Amazon have one prefix position and can take only one pronominal suffix (or enclitic) on the verb. This is in contrast to Kampa languages which allow more than one suffixed pronominal marker. The obligatory presence of bound pronouns on verbs and possessed nouns in most Arawak languages correlates with the optional character and limited use of free



4  Genders and classifiers in Arawak languages of north-west Amazonia

109

Table 1.  A composite statement of A/Sa/possessor prefixes, O/So suffixes/ enclitics in Arawak languages Prefixes

Suffixes

Person

sg

pl

sg

pl

1

*nu- or *ta-

*wa-

*-n or *-te

*-wa

2

*pi-

*(h)i

*-pi

*-hi

3non-fem

*ri-, i-

*na-

*-ri, -i

*-na

3fem

*thu-, ru-, u-

*na-

*-thu, -ru, -u

*-na

impersonal/reflexive

*pa-

 

 

non-focused A/Sa

*i-, a- (?)

 

 

 

dummy So/O

-

-

*-ni

(Aikhenvald 1999: 83; 2002: 289)4

independent pronouns (cf. along the lines of the discussion in Dixon 2010: 208–12). Across the Arawak family, independent pronouns transparently consist of a crossreferencing prefix plus a one-syllable emphatic particle, e.g. Baré nu-ni, Warekena of Xié nu-ya, Tariana nu-ha, Kurripako nhoa (from *no-haa), Bahuana nu-i ‘I’. Independent pronouns are used sparingly—mostly to express focused arguments, copula subjects, and copula complements (more on this in Aikhenvald 2018a). Third person pronouns often overlap with demonstratives, and can be used as specifier articles. Not every bound pronoun has a corresponding independent pronoun. For instance, no Arawak language has a free pronoun correlate of the ‘indefinite’ or the reflexive-impersonal prefix. Nouns divide into obligatorily possessed and optionally possessed; some nouns cannot be used in possessive constructions and are not ‘possessible’. Number words form a closed class: only ‘one’, ‘two’, and ‘three’ can be reconstructed for the proto-language. The root meaning ‘one’ is also used in the meaning of ‘other’. Having a small system of underived number words is a typical feature of Amazonian languages (see Dixon and Aikhenvald 1999; Aikhenvald 2012a: 351–4). In the history of Arawak languages, some categories are particularly stable. Gender is one of these. Two genders, feminine and non-feminine, can be reconstructed to the proto-language, as shown in Table 1. Many languages add to these further separate systems of classifiers in different environments—numeral classifiers, noun classifiers (used as derivational markers on nouns), possessive classifiers, classifiers on verbs, 4  The person marking system includes additional person values. A number of Arawak languages (see Table 1 and Aikhenvald 1999: 88; 1995a) have a further vocalic prefix *i- or *a- which marks a non-specified or focused subject (A/S) and unspecified, or indefinite, possessor. There is also some evidence in favour of a generic or dummy suffix (or enclitic) *‑ni marking So/O, whose origin and development is a matter for further study. 

110

Alexandra Y. Aikhenvald

and also classifiers on adpositions or location markers (as in Palikur: Aikhenvald and Green 1998). Classifiers and gender markers are bound morphemes. Gender-sensitive suffixes are also used on nouns (typically, those with human referents) and as agreement markers on modifiers (as in (1)–(5)). Demonstrative pronouns can be based on personal prefixes in combination with distance markers (as in (6)–(7), from Baniwa of Içana). Here I concentrate on the systems in the area of the Orinoco-Rio Negro basins and the adjacent areas north of the Amazon, encompassing the Rivers Meta and Putumayo (see Map 1). This region is known for its higher concentration of structurally divergent languages than any other Arawak-speaking region. It is also thought of as the putative homeland of Arawak peoples (see references and discussion in Aikhenvald 2002: 14; 2013b; 2015, and references there). Many of the groups share cultural features, including the cult of the Yurupary spirit and magic flutes which women are not supposed to see, associated with male initiation and fertility rites, and common origin myths.5 The list of languages and established low-level groupings are in Appendix 1.6 Even within established groupings, grammatical differences between languages can be striking—this especially concerns classifier contexts and use.

4  Genders and classifiers in the Arawak languages of north-west Amazonia Table 2 shows the distribution of genders and classifiers in the Arawak languages of north-west Amazonia.7 The system of two genders is pervasive in the following contexts: (a) on the verb, possessed nouns, and in independent pronouns; (b) as wordclass preserving derivations on nouns; (c) as word-class changing derivations; (d) as agreement markers within a noun phrase on adjectives; (e) as agreement markers within a noun phrase on demonstratives; and also (f) as agreement markers within a noun phrase on a number word. The feminine gender applies to females; the non-feminine gender applies to everything else and can be considered a default and a functionally unmarked choice. Languages vary in terms of which gender contexts they employ. 5  See, for instance, Brüzzi (1977: 314–15), Hill (2001: 56–9) on the Kuwai (or Kowai) cult of the BaniwaKurripako; Hugh-Jones (1979) and Reichel-Dolmatoff (1996), among others, Schackt (1990, 2013) on the Yucuna; Bourgue (1976) and Bonnemère (2001: 31–8) on the Cabiyarí. 6  The only other Arawak languages north of the Amazon spoken outside this area with classifiers are Palikur (for a detailed description of the system and arguments in favour of its independent development see Aikhenvald and Green 1998), Wapishana (Gomes dos Santos 2006) and possibly Mawayana (Eithne Carlin, p.c.). 7  No information is available on genders or classifiers in a number of extinct languages documented only in short word lists, such as Mandawaka, Yabaana, Yumana, Passe, Amarizana, and Guinau. Additional information on poorly known languages, including now extinct Kaisana, Manao, Maipure, and the highly endangered Cabiyari, is summarized in Table A in Appendix 1.



4  Genders and classifiers in Arawak languages of north-west Amazonia

111

We can distinguish four sets of languages, in terms of the morphological loci of gender and classifier realization. Set I languages have only a two-fold gender distinction of feminine versus nonfeminine in singular only. Set II languages have a two-fold gender distinction of feminine versus non-feminine in singular only, and classifiers on low number words (‘one’ and ‘two’, and, just in Bahuana, the quantifier ‘how many?’). Set III languages have a two-fold gender distinction of feminine versus non-feminine in singular only and classifiers on low number words and some nouns (in addition to the modifier ‘other’ in Yucuna). Set IV languages have a two-fold gender distinction of feminine versus non-feminine in singular only and classifiers in multiple environments. The four types are summarized in Table 2. 4.1  Languages of set I: genders only Piapoco, Guarequena, Baré, and Yavitero are examples of set I languages. The ­feminine and non-feminine genders are distinguished in third person prefixes (used to mark A/Sa on verbs and the possessor on nouns), in personal pronouns, on agreeing constituents in a noun phrase, and in derivation. The number word ‘one’ is the only number word to have two gender agreement forms (synchronically, this word patterns with adjectives). Yavitero stands apart from the other three languages in that it does not have gender agreement within a noun on adjectives nor number words (Mosonyi and Largo 2000: 623, 637; Mosonyi 1987: 140). A noun phrase with two agreeing modifiers from Piapoco is at (10a-b) (Klumpp 1990: 86, 114–15). (10) a. yáí dem.prox.nonfem.sg ‘this big buck deer’

néri deer

manuí-iri-ca big-nonfem.sg-art

b. úáí         néri  manuí-íchúa-ca dem.prox.fem.sg deer big-fem.sg-art ‘this big doe deer’

Piapoco

Piapoco

The number word ‘one’ requires agreement in gender with the noun it modifies. The same set of forms derives relativized verb forms used as modifiers (Klumpp 1990: 114, 149, 200). c. abé-iri asiali i-culúa-íri-wa one-nonfem.sg man 3sg-sicken-nonfem.sg-intr ‘one man who sickened’

Piapoco

d. abé-íchúa   inanái  i-culúa-íchúa-wa one-fem.sg woman 3sg-sicken-fem.sg-intr ‘one woman who sickened’

Piapoco

Table 2.  Gender and classifier contexts in Arawak languages of north-west Amazonia Set

Language

(a) Pronouns and crossreferencing affixes

I

Piapoco (U), G Guarequena (U), †Baré

(b) Wordclass preserving derivations

(c) Wordclass changing derivations

G*

G (some)

†Yavitero (GX)

(d) Adjectives (e) Demonstratives

G

(f) Number words

Classifiers

G

G (‘one’ only)

—

—

—

II

!Warekena of Xié, Baniva of Guainia (GX) †Bahuana

G

G*

—

G

G

G within CL

CL with ‘1’, ‘2’

III

Achagua, Yucuna

G

G

G

G

G

G within CL in ‘1’

CL with ‘1’, ‘2’, ‘3’, some nouns

IV

Baniwa of Içana-Kurripako (U)

G  

G; CL  

G; CL  

G; CL  

G

G, CL  

CL in multiple contexts

IV

Tariana (U)

G - feminine/non-feminine gender; G* - non-productive use of gender-sensitive markers; CL – classifier GX – Guainia-Xié subgroup; U – Uapuí subgroup; ! – critically endangered; † - extinct

CL



4  Genders and classifiers in Arawak languages of north-west Amazonia

113

Non-productive gender-sensitive derivational suffixes -i ‘masculine sg’, -u ‘feminine sg’ occur on kinship terms and a few terms denoting human beings, e.g. quìrasi-ì (infantmasc.sg) ‘infant boy’, quìrasi-ù (infant-fem.sg) ‘infant girl’. Further derivational suffixes are -ri/-li ‘masculine sg’ and -ru ‘feminine singular’, e.g. bé-ri (adult-masc.sg) ‘man (mature)’, bée-ru (adult-fem.sg) ‘woman (mature)’, nu-cuí-ri (1sg-parent’s.sibling-masc.sg) ‘my uncle (mother’s brother)’, nu-cuí-ru (1sg-parent’s. sibling-fem.sg) ‘my aunt (father’s sister)’ (Klumpp 1990: 67–8) (cf. also nu-átúa ‘my mother’ and nu-ániri ‘my father’). 4.2  Languages of sets II, III, and IV Table  3 shows the contexts for the use of classifiers in the languages belonging to Sets II–IV. All the languages have gender-sensitive classifiers. The forms are listed in Table 4. The forms cognate with Proto-Arawak gender markers (listed in Table 1) are in bold. Classifier sets are listed in Appendix 2. The Proto-Arawak feminine versus non-feminine distinctions are integrated within the system of classifiers. Every language contains reflexes of Proto-Arawak gender markers non-feminine –ri, -i, and/or feminine -ru, -u. Tariana is unusual for the family in that the reflexes of the Proto-Arawak feminine and non-feminine gender markers -(r)u and -(r)i are only marginally integrated within the classifier system. The non-feminine form is discernible only in the special adjectival form of the animate classifier -ite (from -ita-y). The reflexes of the ProtoArawak gender markers are used with personal pronouns, cross-referencing markers, and on nouns themselves as derivational affixes. An unusual feature of Baniwa of Içana-Kurripako is the existence of special classifiers with adjectives which distinguish gender for animates. This is reminiscent of Achagua (Examples (4) and (5)) where numeral classifiers distinguish genders for some animate referents. 4.3  The resilience of genders and the emergence of classifiers In most Arawak languages of north-west Amazonia, gender agreement is maintained on pronominal affixes and derivations, on adjectives and demonstratives, and also on number words (especially ‘one’). Even in those languages which have innovated classifiers, gender distinctions are embedded into classifier system. The contexts of gender agreement widespread in the family include number words, adjectives, and demonstratives. The expression of gender—one of the most archaic and resilient features of Arawak languages—is integrated within the systems of classifiers. Unlike genders, classifiers in their many contexts are not shared by all the Arawak languages, suggesting that they are a relatively recent innovation. We now turn to the question of their histories—which classifier context could be considered historically the oldest. Two hypotheses can be suggested.

Table 3.  Classifiers in Arawak languages of north-west Amazonia: forms and contexts of use Set

Language

Form

Number word

Noun

Quantifier

Adjective

Possessive construction

II

!Warekena of Xié (Table I)

suffixes to ‘one’ prefixes to ‘two’

√

—

—

—

—

6

II

†Bahuana (Table II)

infixes to number words; suffixes to quantifiers

√

—

√

—

—

26

III

Achagua(Table III) Yucuna (Table IV)

suffixes

√

(√)

—

—

—

c. 12 c. 12

IV

Baniwa of IçanaKurripako (§4.3, Tables V–VIII), Tariana (§4.3, Tables IX–X)

suffixes

√

√

√

√

√

45 80 +

√

Nominalized verb

Number of classifier forms



4  Genders and classifiers in Arawak languages of north-west Amazonia

115

Table 4.  Gender-sensitive classifiers and their forms: languages of sets II, III, and IV Set  

Nonfeminine

Feminine

Context

Data: Appendix 2

II

Warekena of Xié8

e-

tuwa-

Number word ‘two’

Table I

II

Bahuana

-tura

-ruwa

‘one’, ‘two’, ‘how many’

Table II

III

Achagua

-iri, -i

-ito, -u

‘one’

Table III

III

Yucuna

-ja -u

-jaru -a’awelo

‘one’, ‘two’ ‘other’

Table IV

IV

Baniwa of Içana-Kurripako  

-hipa, -ita -ma -ita -ma -ita-y, dari  -daru ‘human female’ -pari -paru ‘flying animate female’

Number words Other contexts Adjectives

Tables V–VIII

IV

Tariana

-hipa, -ita -ita -ita-y

Number words Other contexts Adjectives

Table IX

-ma

Hypothesis 1. All the languages under consideration which have classifiers use them with number words (see Table 3). This may suggest that this could be the most archaic and historically primary context for classifier use. The only living and well-described language from set I, Piapoco, appears to offer some support for this hypothesis. Number words take gender agreement (see §4.1). Just one form, abé-émàa wa-cáapi (one-side our-hand) ‘five’ (lit. one side of our hand: Klumpp 1990: 149), could contain a remnant of an older classifier with quantifying semantics. Reinoso Galindo (2002: 191) considers the morpheme ‘side’ in this form (presented, in its underlying form, as abé-ema wa-kápi ‘one-half 1pl-hand’) a unique classifier. This form was reported by Klumpp (1990: 149) as being in use only by older speakers. No other forms with possible classifier functions have been attested in the language. Hypothesis 2. A set of bound inalienably possessed nouns with reference to ­quantification or shape such as ‘side of ’, ‘path of ’, ‘oblong part of ’ formed compounds as ­second components. They gave rise to agreement markers on number words and other targets via alliterative agreement. In languages of set III, they developed into what can be synchronically described as numeral classifiers. We can recall, from Table 2, that number words tend to be a wide-spread target of gender 8  No information on gender distinctions in classifiers in Baniwa of Guainia is available (Mosonyi and Camico 1996, 2000); see Appendix 2.

116

Alexandra Y. Aikhenvald Bound nouns with reference to shape/quantification to alliterative agreement on number words (set III) to markers on adjectives, and other environments (set IV).

Scheme 1  Hypothetical development of classifiers in languages of set III and set IV

a­ greement in Arawak languages of the region. The agreement markers then spread to multiple contexts (in languages of set IV). The hypothetical development is set out in Scheme 1. The fact that classifiers are used on number words and also on some nouns as noun classifiers in set III languages (Achagua and Yucuna) corroborates Hypothesis 2. Piapoco (set I) also has a few Noun-Noun compounds whose second component is a bound noun, reminiscent of noun classifiers, such as i-ùwi-ápuná (3masc.sg-ear-path) ‘ear canal’, càiná-wàa-íri (sand-expanse-masc) ‘expanse of sand, desert’ (Klumpp 1990: 166–7), i-íwita-pu (3masc.sg-head-extension?) ‘opening’ (Reinoso Galindo 2002: 123), íida-kua (canoe-flat.extent?) ‘prow of canoe’ (Reinoso Galindo et al. 1994: 18). The form ‘side’ (which we discussed with regard to Hypothesis 1) also appears in such contexts, e.g. i-càcu-(i)máa (3masc.sg-cheek-side) ‘side of a house’ (cf. yémaami ‘half and a remaining part’: Klumpp 1995: 138; yema-lewa ‘next to, at the side of ’: Reinoso Galindo et al. 1994; y-éma ‘its half ’: Reinoso Galindo 2002: 191). A further suggestive example comes from Cabiyari, a language from the Uapuí group. Information on this language is limited, which is why it has not been included in Table 2 (see Table A in Appendix 1). Cabiyari has two genders in all the agreement contexts (demonstratives, adjectives, and number word ‘one’. Reinoso-Galindo (n.d.: 9–10) mentions the existence of classifiers which attach to nouns as derivational affixes: ‑khá ‘liana, extended and cylindrical objects’, -phe ‘palms’, -kua ‘surfaces’, -phu ‘containers’, -pi ‘trees’, ‑phi ‘hills’. Set IV languages include noun classifiers as a highly productive device—more on this in §§ 6 and 7. Difficulties with tracing the history of any category in Arawak languages of northwest Amazonia have to do with the availability of full grammars and language obsolescence. All languages of set II and most languages of set I are obsolescent or extinct. Data in Table 2 are based on traditional varieties of languages. Obsolescence of gender forms has been attested in Baré (Aikhenvald 1995a) (set I) and Warekena of Xié (Aikhenvald 1998) (set II): the non-feminine forms of adjectives and demonstratives were used instead of feminine forms by the remaining speakers. It is possible that Yavitero, a highly endangered language at the time when Jorge Mosonyi conducted his fieldwork, had lost gender agreement in a number of environments due to language obsolescence. Most dialects of Warekena of Xié (set II) have lost numeral classifiers (the human form is used for all referents). Numeral classifiers in Achagua (set III, Meléndez 1998, Wilson 1992) are falling out of use; then the non-feminine singular form is being used



4  Genders and classifiers in Arawak languages of north-west Amazonia

117

instead of other forms. Younger speakers of Tariana (set IV) are losing the non-feminine numeral classifier ‑hipa (thus simplifying the system).

5  Where do classifiers come from? Table 5 shows the number of established cognates in classifiers in the languages considered here (leaving aside gender markers). With the exception of Tariana and Baniwa of Içana-Kurripako, the number of cognates is low—this may indicate that classifiers have developed fairly recently. Classifier forms shared by more than two languages are listed in Table 6. Some are clear cognates of Proto-Arawak forms reconstructed by Payne (1991). In terms of their semantics, these classifiers appear to be quite basic—they categorize nouns in terms of their shape and form. Cross-linguistic investigations (see a survey in Aikhenvald 2000, 2017, and references there) show that languages with small sets of classifiers tend to develop classifiers with these meanings. It may well be the case that these classifiers go back to archaic bound forms which developed into classifiers in just some languages. A similar path for the development of classifiers in Kampa languages, a separate subgroup of Arawak, was suggested by Mihas (Chapter 2; see also Mihas 2015, 2017). Along similar lines, a strong connection between bound nouns and classifiers has been demonstrated for Shiwilu, a Kawapanan language, by Valenzuela (Chapter 3), and Weber (2002), for Bora, from the Boran family. A shared feature of the two languages with multiple classifier contexts (set IV) is that classifiers are often cognate with nouns. Examples of nouns which have grammaticalized as established classifiers include: Tariana, Baniwa of Içana -ya ‘classifier for skin’, -ya ‘skin’ Tariana, Baniwa of Içana -hiwi ‘classifier for thin objects’, -hiwi ‘needle’ Tariana, Baniwa of Içana -hipani ‘classifier for waterfall’, hipa(ni) ‘waterfall’ Tariana, Baniwa of Içana -dapana ‘classifier for houses’: Tariana pani-si, Baniwa pan-tti ‘house’, nu-pana ‘my house’ (in combination with -da- ‘round things; general classifier).

Table 5.  Number of cognates in classifiers in North Arawak Tariana (c. 80 classifiers) 31

Baniwa of Içana (c. 45)

5

5

Yucuna (12–13)

5

5

2

Achagua (12)

0

0

0

0

Warekena (6)

4

3

1

2

0

Bahuana (26)

Table 6.  Shared classifiers in the Arawak languages of north-west Amazonia and their cognates elsewhere in the family Baniwa of Içana

Tariana

Achagua

Yucuna

Bahuana

Other Arawak languages

-na; -ne, nay (with adjectives) ‘vertical objects; mammals’

na ‘vertical objects’

-na ‘mammals’

-na‘mammal, tree’

-na ‘long’ (e.g. bottles, pineapples)

Yawalapiti -na ‘vertical objects’  

-pi ‘long thinnish objects’

—

—

—

Cabiyari -pi ‘trees’; Waurá -pi ‘linear’, Terêna, Pareci -hi ‘long, thin objects’, Baure -pi ‘long objects’, Ignaciano -pi ‘long and thin objects’, Amuesha -py ‘long objects’

-apu ‘long thinnish stick-like’

‑ahu ‘extended and elongated, cylindrical’

—

-ɸua ‘gardens’ (?)

-da ‘round; generic’, -da ‘round, -daɾi (with adjectives) generic’

Resígaro -ʔapo ‘shortcut on a river’; Piapoco -pu ‘extension’ (as a suffix on nouns) Bare -bu as a suffix in dina-bu ‘road’ (Aikhenvald 1995a) Yawalapiti, Waurá -apu ‘classifier: path’; Proto-Arawak *(a)pu ‘road’ (Payne 1991)

—

-ta ‘flat objects’

Yawalapiti -ta ‘spherical’, Apurinã -ta ‘roundish edge’ -da ‘objects (class marker) with large surface, hollow or rounded’

-kwa ‘flat, extended’, kwe (with adjectives)

-kwa, -kwa-na ‘plain, flat’

-kua ‘oblong’

-a’aku ‘concave’

—

-wa ‘small holes’

-wa ‘hole, stretch’

-iwa ‘big, roundish’

‑wa ‘hammocks; hollow objects’

-wa ‘hammock’ Piapoco -wàa ‘expanse’ (as a suffix on nouns)

-pu ‘long thinnish stick-like’

Cabiyari –kua ‘surfaces’; Yawalapiti, Waurá –kana ‘hollow objects’, Apurinã –kɨ ‘rounded’, nominalizer -koa ‘round hollow containers’

-apa, masc apa-ɾi, fem apa-ʒu ‘flying animate beings; masc: semi-oval objects, e.g. banana’

-apa ‘largish — long objects’, e.g. bananas

-a’apa ‘line-like objects’

—

Apurinã, Nomatsiguenga -pa ‘cylindrical or pod-like shape’

‑ta’ama ‘one side’

—

Piapoco –(i)máa ‘side’

-ima ‘side’, with adjectives -ima-ɾi

-ima ‘side; paired object’

—

-ku ‘folded cloth, hollow’

-ku ‘folded cloth’

-iku ‘recipient’ — (e.g. bottle)

—

Palikur -iku, -eku, -rik, -ik ‘the inside of; extended object with boundaries’; Ignaciano –kuʔa ‘cavities’; Resígaro -ko (as in –iiʔsa-ko ‘intestines’, from –iiʔsáú ‘belly’: Allin 1975: 409); Piapoco -kua ‘flat expanse’ (as a suffix on nouns)

-ita ‘human’, -ite with adjectives

-ta ‘human’

—

—

—

-maka ‘stretchable extended objects, cloth’

 

 

 

Apurinã -maka ‘class term for soft, flat items’, mãka ‘clothes’ (Facundes 2000), Paresi-Haliti maka ‘hammock’ (Brandão 2014: 176), Bahuana hɨmaka ‘hammock’ (Ramirez 1992). Yucuna -maka ‘formative in the noun a’aru-maka-ʃi (skin-clothnonpossessed) ‘clothing’ (Schauer et al. 2005: 300),

120

Alexandra Y. Aikhenvald

The reflexes of the independent root *-maka ‘hammock’ grammaticalized as a classifier in Baniwa of Içana-Kurripako, Tariana -maka ‘classifier for stretchable extended objects, cloth’ (Aikhenvald 2002: 307–8), cf. Yucuna -maka ‘formative in the noun a’aru-maka-ʃi (skin-cloth-nonpossessed) ‘clothing’ (Schauer et al. 2005: 300) Apurinã -maka ‘class term for soft, flat items’, mãka ‘clothes’ (Facundes 2000) Paresi-Haliti maka ‘hammock’ (Brandão 2014: 176) Bahuana hɨmaka ‘hammock’ (Ramirez 1992) Lokono Arawak hamaka ‘hammock (unpossessed form)’ (Patte 2011: 21). The classifier for clothing -mada in Bahuana (Ramirez 1992: 55; Table II in Appendix 2) possibly reflects an independent process of grammaticalization. The form goes back to Proto-Arawak *mata ‘skin’ (with cognates in Palikur and a number of other languages outside our area of study: Payne 1991: 418). This suggests the possible development of individual and especially specific classifiers at a later stage and at the level of individual groupings (if not languages). Members of word classes other than nouns do not grammaticalize as classifiers (similarly to many languages of the world: Aikhenvald 2000: 401–5, 442–6), but unlike Kampa languages where some classifiers go back to verbal roots (Mihas Chapter 2).

6  What are classifiers good for? We now turn to a brief outline of the functions of classifiers in the Arawak languages of north-west Amazonia. These functions are consistent with cross-linguistically established tendencies for classifier usage (see Aikhenvald 2017 and references there, and §5 of Chapter 1). Firstly, classifiers help disambiguate polysemous nouns. This is illustrated with (11) and (12), from Baniwa of Içana. The form tsinu refers to a dog and is classified as a vertical entity in (11). In (12), the same form refers to an eel-like fish, which is classified as a curved object. (11)

apa-na one-cl:vert ‘one dog’

tsinu dog

(12)

apa-kha tsinu one-cl:curved ‘dog’ ‘one sarapó fish (Apteronotus bonaparti, eel-like)’

Baniwa of Içana

Baniwa of Içana

Classifiers can also help focus on a shape property of a referent. A man can be referred to with the classifier for a ‘human non-feminine’ being, as in (13), also from Baniwa of Içana. An excessively tall man will be referred to with the classifier for a vertical entity, ‘as if ’ he were a tree.



4  Genders and classifiers in Arawak languages of north-west Amazonia

(13)

121

Baniwa of Içana

atsiari maka-dari man big-adj.cl:hum ‘a big man’

Baniwa of Içana

(14) atsiari maka-ne man big-adj.cl:vert ‘an excessively tall man (like a tree)’

Table  7 illustrates the ways in which the many meanings of polysemous noun uni ‘water, liquid, waterway’ can be disambiguated with the help of classifiers. Prolific use of classifiers can account for the absence of a semantic subclass of adjectives. In languages with multiple classifiers based on shape and form, such as Baniwa of Içana and Tariana, there are no adjectives referring to shape. For instance, Tariana has no adjective meaning ‘round’: instead, one uses the classifier -kwema ‘round object’, e.g. panisi hanu-kwema (house big-cl:round) ‘a big round shaped house’. Secondly, classifiers can be used to expand the lexicon. In Tariana, a classifier can attach to a dummy root maka- forming a noun which would correspond to a classifier and reflect notions of shape, e.g. maka-pukwi (dummy-cl:round.hollow) ‘a ring-like object’, maka-kwema (dummy-cl:round) ‘a round object’. In this way, classifiers help introduce new words for new concepts, thus avoiding borrowing a term form a contact language (in agreement with the general inhibition against borrowing forms in the language: Aikhenvald 2002). Using classifiers on nouns are a highly productive means of forming new words in Tariana. Several classifiers can occur on a noun. Examples (15a–c) and (16a–b) illustrate the use of classifiers in their derivational Table 7.  How classifiers may disambiguate polysemous referents in Tariana uni hanu-pua (big-cl:waterway)

‘big river’

uni hanipa (big+cl:large.space)

‘big, large river, large pool of water’

uni hanu-kha (big-cl:curved)

‘big, curved river’

uni hanu-wani (big-cl:abstract.places)

‘big river (as a location)’

uni hanu-nai (big-cl:lake)

‘big, lake-like river’

uni hanu-dawa (big-cl:corner)

‘big bay in a river’

uni hanu-pina (big-cl:swamp)

‘big, swampy river’

uni hanu-puna (big-cl:road)

‘big river (road-like) for canoe to travel’

uni hanu-kwa (big-cl:flat.surface)

‘big river (with flat surface)’

uni hanu-peku (big-cl:thin.stretch)

‘narrow stretch of a river’

uni hanu-thiwa (big-cl:hollow)

‘ravine-like waterway’

uni pumeni-peri (sweet-collective)

‘sweet water, juice, soft drink’

122

Alexandra Y. Aikhenvald

function to create lexical items referring to recently introduced objects, such as an ‘airplane’ in (15a) and a ‘car’ in (16a). Classifiers reflect part-whole relationships between components of the entity: an airstrip in (15b) is referred to as a ‘road-like part of a canoe-like flyer’, and the term for a curved part of an airstrip in (15c) is literally ‘curved part of a road-like part of a canoe-like flyer’. The term for a street, or a highway, in (16b) is literally ‘a road-like part of a canoe-like roller’. (15)

a. kara-ka-whya rel+fly-th-cl:canoe ‘airplane’

Tariana

b. kara-ka-whya-puna rel+fly-th-cl:canoe-cl:road ‘airstrip’

Tariana

c. kara-ka-whya-puna-kha rel+fly-th-cl:canoe-cl:road-cl:curved ‘curved part of an airstrip’

Tariana

(16) a. ka-kolo-ka-whya rel-roll-th-cl:canoe ‘car’ b. ka-kolo-ka-whya-puna rel-roll-th-cl:canoe:cl:road ‘street, highway’

Tariana

Tariana

Classifiers in Tariana share their derivational functions with classifiers in Shiwilu (§4  of Chapter  3) and in Kampa languages (§4.1 of Chapter  2). The part-whole ­relationship within a classifier construction in Nanti, a Kampa language from Peru, was noted by Michael (2013: 164–5). However, using several classifiers—each in a part-whole relationship to the preceding one—on one noun appears to be a rare ­feature of Tariana. Thirdly, classifiers individuate the noun referents. For instance, Tariana deri means ‘banana’ (in general). This form cannot be pluralized. In order for a plural marker to be added, a classifier has to attach to a noun, individuating the referent and making it into a singular object, e.g. deri-pi (banana-cl:palm.tree) ‘banana palm’, deri-pi-pe ‘banana palms’. Fourthly, classifiers are often used anaphorically to refer to a previously mentioned entity. As the head noun in discourse is often omitted, classifiers are more frequent in discourse than full nouns. A noun is typically stated once (at the start); and then referred to with classifiers (ratio 1/10 in terms of token frequency calculated for a corpus of Tariana and of Hohôdene Baniwa).



4  Genders and classifiers in Arawak languages of north-west Amazonia

123

7  To conclude We have seen that several Arawak languages of north-west Amazonia have co-­existing systems of noun categorization—small systems of genders based on sex and a­ nimacy, and classifiers based on physical properties. The distinction between feminine and non-feminine genders appears in cross-­ referencing markers on nouns and verbs (as shown in Table 1), on nouns as derivational suffixes, and on demonstratives and also adjectives as agreement markers. The Arawak languages considered here divide into four sets with regard to how they combine genders with classifiers (see Table 2). Set I languages—Piapoco, Guarequena, Baré, and Yavitero—have just a two-fold gender distinction of feminine versus non-feminine in singular, and no classifiers of any sort. Set II languages—Warekena of Xié, Baniva of Guainia, and Bahuana—have a twofold gender distinction of feminine versus non-feminine in singular, and classifiers on number words. Set III languages—Achagua and Yucuna—have a two-fold gender distinction of feminine versus non-feminine in singular, and classifiers on number words and some nouns (in addition to the modifier ‘other’ in Yucuna). Set IV languages have a two-fold gender distinction of feminine versus non-­ feminine in singular only and classifiers in multiple environments. Baniwa of Içana (set IVa) employs genders with demonstratives, where Tariana (set IVb) employs classifiers. Languages with classifiers (sets II–IV) differ as to the contexts in which classifiers occur. In all of the languages, classifiers are used with number words. Bahuana adds quantifiers to this. In Achagua and Yucuna, classifiers are used on some nouns. In Baniwa of Içana and Tariana, classifiers are used on nouns and nominalized verbs, on quantifiers, on adjectives, and in possessive constructions (see Table 3). Both genders and classifiers have been affected by language obsolescence: this could be the reason for gender variation in Baré and other highly endangered languages. Of the languages discussed, Baniwa of Içana-Kurripako has the most complex ­system of classifiers, with four sets in different environments (including a special set of classifiers with adjectives, some of which distinguish additional gender forms). Tariana has a reduced set of contexts in which gender is marked, due to areal diffusional impact from neighbouring East Tucanoan languages: this is the most aberrant language of all. The distinction between feminine and non-feminine genders is integrated into the classifier systems in all languages (see Table  4). Gender marking has been inherited from the proto-language. Not so for classifiers. The number of cognates in classifiers

124

Alexandra Y. Aikhenvald

across the languages discussed here is low (see Table  5; the classifier cognates across the family are listed in Table 6). Classifiers appear to have evolved as a result of independent development in low-level subgroups or even individual languages (see §5). Using classifiers with number words is a feature of all the languages with classifiers discussed here. This may suggest that classifiers developed as numeral classifiers in the first place (hypothesis 1 in §4.3). One can offer an alternative pathway. Bound noun roots with meanings related to quantification or shape—as parts of compounds—may have developed into agreement markers via alliterative agreement (hypothesis 2 in §4.3). Data from extant languages point towards this path of development. This is corroborated by the ways in which bound nouns gave rise to classifiers in Kampa languages (another subgroup of Arawak languages) and Amazonian ­languages from other regions. Numeral classifiers and classifiers in multiple environments in Arawak languages have a variety of functions—they help disambiguate polysemous nouns, focussing on their individual properties, and help expand the lexicon by creating new nouns. They help maintain anaphoric reference, and, once the referent has been established, can be  more frequent in discourse than the noun itself (as we saw in §6). Cross-linguistically, the presence of numeral classifiers in a language is often associated with a large set of number words (see Aikhenvald 2000: 99–100, and discussion and references there). Similar to many other Amazonian languages, Arawak languages defy this statement—all of them have a small system of number words (see Aikhenvald 2012a: 351–8, on number words in Amazonian languages, and Johnson 2003: 153, for a restricted use of number words and the absence of counting routine in Machiguenga, a Kampa language).

Appendix 1:  The Arawak languages of north-west Amazonia The Arawak languages of north-west Amazonia do not form one subgroup. The following subgroups are firmly established. • The Uapuí group consists of Baniwa of Içana-Kurripako dialect continuum, Tariana (formerly a dialect continuum), Guarequena, Piapoco, and Cabiyari. These peoples share a number of grammatical and phonological innovations, a high percentage of shared lexicon, and origin myths (they are said to have emerged from the Uapuí waterfall on the Içana river). Resígaro, a moribund Arawak language spoken in northeast Peru, Loreto province: see Seifart 2011,  Aikhenvald 2001), appears to be closely related to this group. The language is spoken within the multilingual complex of the ‘People of the Centre’



4  Genders and classifiers in Arawak languages of north-west Amazonia

125

(Wojtylak:  Chapter  6), and has undergone massive restructuring under the influence of unrelated Bora (apart from a multiple classifier system with numerous borrowings from Bora, it developed dual number and also genders in dual pronouns: see Aikhenvald 2001, 2012a on the mechanisms involved). This is an aberrant Arawak language (in the sense of Grace 1990) and has not been included in this discussion. Baniwa of Içana-Kurripako is spoken by 3–4,000 people in the basin of the Içana River and its tributaries in Brazil and the adjacent areas of Colombia and Venezuela, stretching into the basin of the Middle Vaupés. Baniwa has at least twenty dialectal varieties, all mutually intelligible to varying degrees (sharing 90–96% of their lexicons) (see lists of dialects in Nimuendajú 1950/5; Rodrigues 1986); pace Ramirez 2001b (since he worked with two Tariana speakers who mixed their language with Baniwa), there is no mutual intelligibility between Tariana and Baniwa of Içana-Kurripako; see also Aikhenvald 2014). Partial descriptions of Baniwa of Içana-Kurripako grammar are in Taylor (1991) and in Ramirez (2001a,c). All the data quoted here come from A.  Y.  Aikhenvald’s work unless otherwise specified (this includes c. 300 pp of transcribed texts; in addition to 100 pp texts by Fontes 2014). Other sources on Kurripako were taken into account (e.g. Bezerra 2005, 2012; Taylor 1991; and the information in Valadares 1993 and Granadillo 2006). A full description of the system of classifiers in Baniwa of Içana is in Aikhenvald (2007); further forms from the Kumandene Kurripako dialect (Bezerra 2005) are quoted when appropriate (see also Hill 1988 on some classifiers, and González-Ñáñez 1985 on classifiers with number words, both sources focusing on the varieties of Kurripako spoken in Venezuela). Tariana is spoken by less than 100 people in the Vaupés area. The language is in contact with dominant East Tucanoan languages in the multi­lingual Vaupés area. Information on Tariana comes from Aikhenvald’s work, based on c. 30 hours of transcribed recordings (covering narratives, conversations, etc.); a comprehensive grammar is in Aikhenvald (2003); additional information on contact induced change in the language is in Aikhenvald (2002, 2006). There are minor dialectal ­differences. The language is endangered and no longer fully learnt by children. Guarequena is currently spoken by about 300 people in Guzmán Blanco (Municipio Autonomo Guainía of the state of Amazonas in Venezuela) on the  left shore of the Guainia river, and two speakers in one Warekena of Xié-speaking community in Brazil (where it is called Warena, or ‘the old Warekena’). Apart from short word lists, the main source on Guarequena is a basic grammar sketch by González-Ñanez (1997) with little in-depth analysis

126

Alexandra Y. Aikhenvald

of the patterns of the language. González-Ñanez (1997: 32) notes a close proximity between Kurripako and Guarequena. See Aikhenvald (1998) for a reconstruction of historical migrations of the Guarequena; Aikhenvald (2012b) for the word list of Warena. The spelling ‘Guarequena’ is maintained here (just as in Aikhenvald 1998) following the sources and to avoid further confusion with Warekena of Xié. Piapoco is spoken by 5–6,000 people in the region of the Upper Vichada, middle Guaviare, Upper Orinoco and Atabapo in Colombia, and adjacent areas of Venezuela. There are two grammars, Klumpp (1990) and Reinoso Galindo (2002), and two dictionaries, Klumpp (1995) and Reinoso Galindo et al. (1994), in many ways complementary to each other. Cabiyari is a poorly documented and highly endangered North Arawak ­language spoken by about twenty elders within the Vaupés Linguistic area on the Apaporis River and its tributaries Cananari and Pirá-paraná. Available ­materials include a very brief sketch by Reinoso Galindo n.d. (some information in Ramirez 2001b: 387, and word lists in Huber and Reed 1992, KochGrünberg 1911, Matteson 1972: 231–4, and Meléndez Lozano 2000. Data on the vitality of Cabiyari come from Katherine Bolaños Quiñónez, who has recently started salvage work on the language. • Warekena of Xié, Baniva of Guainia, and Yavitero form a set of very closely related languages. Warekena of Xié is said to be mutually intelligible with Baniva  of Guainia by the speakers. Warekena of Xié is a moribund language spoken in nine communities on the Xié river, a tributary of the Upper Rio Negro. Information of Warekena is based on Aikhenvald (1998, 2012b,c), consisting of ten hours of transcribed recordings of the language. The language is currently spoken by no more than fourteen elderly people (in Brazil). It is possible that the Warekena of Xié are later migrants from the adjacent Venezuela; however, their presence at the current location was documented as early as 1831, by Natterer (1831). Its close dialect Baniva of Guiania is spoken by about 200 people in the community of Maroa and a number of other settlements in the Department of Casiquiare of the state of Amazonas in Venezuela. Alternative names for the language are Baniva or Baniva of Maroa. A partial grammar of the language is in Socorro Sánchez (2005); Mosonyi and Camico (1996, 2000) is a brief sketch of the language with few examples; additional information on the number of the speakers and a few features of the language are mentioned in Álvarez and Socorro (2002); some features are in an earlier sketch by de la Grasserie (1892). Yavitero, or Baniva of Yavita, was formerly spoken in the vicinity of Atabapo (close to Baniva of Guainia); Mosonyi (1987) is a competent



4  Genders and classifiers in Arawak languages of north-west Amazonia

127

grammar of the language, accompanied by a word list (a brief sketch is in Mosonyi and Largo 2000). • Baré is an extinct Arawak language, formerly spoken in an extensive area within the Upper Rio Negro region along the Baria river and the Casiquiare channel and into the Orinoco basin, extending into the basin of the river Xié and Upper Guainia up to the Atabapo. See Aikhenvald (1995a), a map there for the spread of Baré and an overview of existing sources; examples here are from Aikhenvald (1995a) and a collection of c. 150 pp. of texts, based on fieldwork with the last fluent speaker of the language; there is a partial sketch in Cunha de Oliveira (1994) and Lopez Sanz (1972). Baré shares lexical isoglosses with Warekena of Xié; its exact relations with other languages of the area require further study. • Yucuna is spoken by about 6–700 people in the basins of the Miritiparaná and the lower Caquetá in the Department of Amazonas in Colombia (see Fontaine 2008: 45–6 on the origins of the group). The language shares a number of features with the Uapuí subgroup of North Arawak (especially Piapoco). The main sources on the language are grammar sketches by Schauer and Schauer (1978, 2000) and a further sketch in the dictionary by Schauer et al. (2005). • Achagua is spoken by about 200 people in the north-east of the Vichada department in Colombia and adjacent areas of Venezuela. The language shows affinity with Piapoco, and also with Yucuna; the exact nature of the relationships requires further investigation. There is a dictionary of the language (Meléndez 2011); the grammatical descriptions by Wilson (1992) and Mélendez (1998; also 1989) complement each other. The first documentation of this language goes back to Neira and Ribeiro (1762), making it one of the earliest attested languages. • Bahuana, formerly spoken on the Demini river (Middle Rio Negro) in the north of the state of Amazonas in Brazil, is unusual for the family in that it does not distinguish gender in its pronominal prefixes or independent pronouns. Ramirez (1992) is the only source on the language, based on the author’s work with the last and obsolescent speaker. Numerous Arawak languages of north-west Amazonia have become extinct during the past two centuries. For many of them there is not enough grammatical information (e.g. Mandawaka, Yabaana, Yumana, Passe, Amarizana, Uainuma, and Guinau (presumed to be closely related to Baré)). Table A summarizes the facts of four ­languages for which some grammatical information is available.

Table A.  Gender and classifiers in extinct and/or poorly known languages of the area Language Pronouns and cross-referencing affixes

Word-class preserving ­derivations

Word-class changing derivations

Adjectives Demonstratives Number words

Classifiers

!Cabiyari

G

G (limited)

?

G

G

G (only ‘one’)

noun classifiers affixed to nouns

†Manao

G

?

?

?

?

?

?

—

G (?)

?

?

?

?

?

G

G (limited)

—

—

?

—

number words ‘1’, ‘2’, ‘3’

†Kaisana

†Maipure



4  Genders and classifiers in Arawak languages of north-west Amazonia

129

Table B.  Shared vocabulary percentages between some Arawak languages Tariana 77–80

Baniwa Hohôdene-Kurripako

55

52–6

56

53

Piapoco

43

50

47

Yucuna

53

53

72

43

Achagua

41

41

30

30

28

Baré

32

36

25

25

24

36

Warekena of Xié

33

32

30

36

29

39

39

Bahuana

33

30

30

30

25

29

22

33

Palikur

35

35

32

24

30

35

29

31

43

Wapishana

Sources are: Hanke (1960) on Kaisana; for Manao see the ‘Christian doctrine’ composed around 1740 analysed in Joyce (1951), Brinton (1892: 38–44), and then de Goeje (1948) on Manao; Gilij (1781: 333–5, 1782: 187–8) and Zamponi (2003: 23) on Maipure; Reinoso Galindo (n.d.) on Cabiyari. Table B shows shared vocabulary percentages between some of the languages ­discussed her- (with some comparisons with languages further to the north-east). (The percentages are based on a 300 word list; Guarequena, Yavitero, and Baniva of Guainia have not been included, as there is not sufficient information).

Appendix 2: Data sets – classifier systems in the Arawak languages of north-west Amazonia Gender-sensitive markers are in bold face.

1  Set II languages: genders and with numeral classifiers Classifiers are infixes to number words a-…-riñi ‘one’ and suffixes to kɨra-…-’(a) ‘two’ and suffixes to yaɸa- ‘how many?’. Examples (Ramirez 1992: 55) are in (A.1a–c). a-da-riñi karaka one-cl:obj.with.large.surface.bird-one     chicken ‘one chicken’

Bahuana

(A.1b) kɨra-da-‘a                karaka two-cl:obj.with.large.surface.bird-two  chicken ‘two chickens’

Bahuana

 karaka yaɸa-da how.many-cl:obj.with.large.surface.bird  chicken ‘How many chickens?’

Bahuana

(A.1a)

(A.1c)

130

Alexandra Y. Aikhenvald

Table Ia.  Numeral classifiers in Warekena of Xié (Anamoim dialect) Semantics

‘One’

‘Two’

human male

Peya

e-naba

human female

Peya

tuwa-naba

animals

Pamiña

pamiña-naba

curvilinear objects

peɾeyaɾu

e-naba

periods of time

papuɾiaɾuni babuya

bu-naba

fish

eɾe-naba

Source: Aikhenvald (1998).

Table Ib.  Numeral classifiers in Baniva of Guainia Semantics

‘One’

‘Two’

Example

human and anthropomorphized beings

patsià

éenábà

patsià niâmali ‘one person’

elongated and cylindrical objects

pawíyà túwanábà pawíyà êeli ‘one tobacco’

all the rest

péyá

búunaba

péyá papêra ‘one (piece of) paper, one book’, búunabà wíminali ‘two stars’

Sources: Mosonyi and Camico (1996: 33, 2000: 204); it is not clear whether the list is complete.

Table II.  Numeral classifiers in Bahuana (Ramirez 1992: 55) Cl

Translation

Example

-tura

animate male

a-tura-riñi kuriña-nawɨ ‘one child, boy’

-ruwa

animate female

a-ruwa-riñi kuriña-nawɨ ‘one child, girl’

-ɨda

pointed or granular objects

a-ɨda-riñi nia-tsɨ ‘one louse’

-daɸɨa

paired organs

kɨRa-daɸɨ’ kaɨda ‘two hands’

-tɨa

buildings

a-tɨa-riñi ɸanici ‘one house’

-na

bottle, pineapple

yaɸa-na iɸa ‘how many ­pineapples?’

-da

objects with large surface, concave or rounded, also birds

a-da-riñi ɸicatɨ ‘one toucan’

-tsuɸa

pointed sharp objects

a-tsufa-riñi akusa ‘one needle’

-wa

hammock

kɨRa-wa’ hamakɨRa ‘two hammocks’

-kuda

big mammals, crocodiles, maniocs

a-kuda-riñi kema ‘one tapir’

-ha

period of time: night, day

kɨRa-ha’ ‘two days, two nights’



4  Genders and classifiers in Arawak languages of north-west Amazonia Cl

Translation

Example

-mada

clothing

a-mada-riñi kamica ‘one shirt’

131

-bara

roads

a-bara-riñi waɸu ‘one road’

-kanua

long objects

a-kanua-riñi watapa ‘one blow gun’

-ca

swamp

a-ca-riñi hamadi ‘one swamp’

-bia

banana, food

a-bia-riñi kuratana ‘one banana’

-batɨa

implements such as knife, spoon, tooth

a-batɨa-riñi maria ‘one knife’

-tsa

rope-like objects

a-tsa-riñi imaci ‘earthworm’

-ɸua

gardens

a-ɸua-rinni nu-kaunɨ ‘my only garden’

-nuba

door

a-nuba-riñi numada-tsɨ ‘one door’

-bua

monkey

kɨRa-bua’ a huawa ‘two capucchin monkeys’

-ta

canoes, palm flowers

a-ta-riñi hica ‘one canoe’

-katɨa

leaves

a-katɨa-riñi nɨ-katɨ ‘its only leaf ’

-mɨna

trees

a-mɨna-riñi atamɨna ‘one tree’

-biuna

beaches

a-biuna-riñi ni-biuna ‘his only beach’

2  Languages of Set III: genders, and classifiers with number words and on some nouns The set of classifiers in Achagua is in Table III (Wilson 1992: 23–4, 62; Meléndez 1998: 91 contains a shorter list; see also Perez de Vega 1963: 21). Gender of humans and sex-differentiable animals is integrated into the system of classifiers: 1. The classifier -ta is not used with the number word ‘one’. Instead, the language uses a masculine and a feminine form (the same as the one which appears in derivational non-word class changing gender). (Neira and Ribeiro 1762: 52 ­documented the form abata ‘other’, no longer used.) 2. The classifier for ‘mammals’ distinguishes two gender agreement forms (Wilson 1992: 63), e.g. áaba-na-i (one-cl-masc) auli ‘one male dog’; áaba-na-u (one-clfem) auli ‘one female dog’ (as we saw in (4) and (5). Some of the classifiers appear on the noun itself as non-productive derivational devices reminiscent of noun classifiers (Meléndez 1998: 92–3), e.g. -kua ‘cl:oblong’ in li-numa-kua (3masc.sg-mouth-cl:oblong) ‘coast, edge, rim’, sarusaru-kua (rot-cl:oblong) ‘type of snake’ -tui ‘cl:rounded’, in ʃipanulitui (sp. of turtle+cl:rounded) ‘a constellation’, -ahi ‘cl:container’, kara-ahi (pot-cl:container) ‘pot’.

132

Alexandra Y. Aikhenvald

Table III.  Classifiers in Achagua one

two, three

-ahi ‘recipient, container’ -ahu ‘extended and elongated, cylindrical’ -bai ‘cutting instrument, extended lance-shaped’ -dani ‘jungle’ -hira ‘elongated’ (e.g. paper) -iku ‘container, bottle’ -iwa ‘big and rounded’ -kua ‘oblong’ -ʃiida, ʃida ‘rectangular, cuboid’ -tui ‘rounded’ ba-ka-iri (one-suf-masc.sg) ‘one (masculine)’ ba-ka-ito (one-suf-fem.sg) ‘one (feminine)’

-ta ‘human’

áaba-na-i (one-cl-masc) ‘one (male mammal)’ áaba-na-u (one-cl-fem) ‘one (female mammal)’

-na-i ‘mammal’

Table IV.  Classifiers in Yucuna pajluwa ‘one’ (in general)

apú ‘other’ (in general); masculine

pajluwa-ja ‘one person’ (masculine) pajluwá-jaru ‘one person (feminine)’

apa’awelo ‘other (feminine)’

pajluwa-ta ‘one (flat)’

apata ‘other (flat)’

pajluwe-la ‘one (extended)’

aphela ‘other (extended)’

pajlúwa-a’ala ‘one (round)’

apa’alá ‘other (round)’

pajlúwa-a’aku ‘one (concave)’

apa’akú ‘other (concave)’

paj’luwa-a’apa ‘one part of something’

?

paj’luwa-e’ete ‘a heap’ (e.g. fish)

?

pajluwa-a’ama ‘one all together, all at once’

?

pajluwa-epi ‘one herd of animals’

?

pajluwa-na ‘one (tree or a big animal)’

apu-a’na ‘other (tree or big animal)’

Classifiers in Yucuna are in Table  IV. Schauer et al. (2005: 305) and Schauer and Schauer (2000: 521). Schauer and Schauer (1978: 50) add to this -ta’ama ‘one side’ (with number word ‘one’). Some classifiers can appear on nouns themselves, e.g. ri-numa-’alá (3sgmascmouth-cl:round) ‘the area around his mouth’, ke’ra-la-ni (red-cl:liquid-adjz) ‘red water’ (Schauer et al. 2005: 305).



4  Genders and classifiers in Arawak languages of north-west Amazonia

133

3  Gender distinctions with classifiers in multiple contexts Baniwa of Içana-Kurripako has two genders in cross-referencing and demonstratives (see example (6)). There are also about 45 classifiers. These occur in several morphosyntactic environments (see Aikhenvald 2007, for the Hohôdene dialect; a brief ­statement in Ramirez 2001a, and Bezerra 2005 for a different dialect): (I) as derivational suffixes on nouns, predicative possessive constructions, and purposive and relativized verbs in nonmain clauses—examples (A.2–4). (A.2) tʃipaɾa-api metal.object-cl:hollow ‘pan’ (A.3) lhiahi prox.dem.nonfem.sg

derivational affix on a noun tʃipaɾa-api metal.object-cl:hollow

possessive construction

nu-dza-api 1sg-poss-cl:hollow ‘This pan is mine’

tʃipaɾa-api (A.4) lhiahi prox.dem.nonfem.sg  metal.object-cl:hollow pi-dzana-kaɾu-api 2sg-cook-purp-cl:hollow ‘This pan is for you to cook’

purposive form

(II) with number words (one to four) and interrogative quantifiers—(A.5–6). (A.5) apa-api one-cl:hollow ‘one pan’

tʃipaɾa-api metal.object-cl:hollow

number word

tʃipaɾa-api (A.6) kadali-api how.many-cl:hollow  metal.object-cl:hollow ‘How many pans?’

quantifier

(III) with adjectives and interrogative modifier ‘which’, as in (A.7–8). maka-api (A.7) tʃipaɾa-api metal.object-cl:hollow  big-cl:hollow ‘a big pan’

(A.8) kwama:pi (underlying form: kwame-api) pi-pedzu-pha-ɾi  which+cl:hollow 2sg-like-int-rel ‘Which hollow object (that is, pan) is the one you like best?’

adjective

‘which’

134

Alexandra Y. Aikhenvald

Classifiers divide into four sets, according to the forms distinguished in each of the environments (I), (II), and (III). The main difference between the groups lies in the existence of special gender-sensitive forms for humans and animates; the existence of special forms for adjectives (gender-sensitive with regard to gender-differentiable nouns), and special forms with number words. Table V.  Classifiers with special gender-based sets for humans and high animates Classifier

(i) Derivational suffixes, (ii) On number (iii) On adjectives in predicative possessive words constructions

1. Humans and their attributes

-ita

-hipa (A.10)

2. Male humans 3. Feminine humans

-ita (A.9)

-ma (A.11)

-ite (underlying form ‑ita-y); -da-ɾi (A.9), (A.10) -da-ɾu (A.11)

Baniwa of Içana

(A.9) apa-ita     nawiki maka-dari one-cl:human person big-cl.adj:human ‘one big person’ (A.10) aphepa (apa-hipa) nawiki maka-dari one-cl:male    person big-cl.adj:human ‘one big man’ inaru (A.11) apa-ma one-cl:fem woman ‘one beautiful woman’

maʧa-daru beautiful-cl.adj.fem

About 17 classifiers distinguish special forms used with adjectives; one (referring to flying animals) has a gender distinction—see (7). A selection is in Table VI. About 22 classifiers have the same form in all the contexts—as in (6). Table VI.  Subset B - Classifiers which distinguish two forms in contexts (i–ii) and (iii) Classifier

Contexts (i) Context (iii) and (ii)

1. Round objects and natural phenomena also used as generic classifier, e.g. kuphe ‘fish’, hipada (rapidcl:generic) ‘stone’, idza ‘rain’

-da

-da-ɾi

-apa

-apa-ɾi

2. Non-human animate and inanimate objects of crescent shape, e.g. kepiɾeni ‘bird’, paɾana ‘banana’, extended to any flying being, e.g. pitiʒi ‘bat’, ainidzu ‘mosquito’



4  Genders and classifiers in Arawak languages of north-west Amazonia

135

Classifier

Contexts (i) Context (iii) and (ii)

3. Feminine flying animate, e.g. kaɾaka ‘hen’

-apa

-apa-ɾu (7)

4. Flat, round, extended objects, e.g. kanaɾi ‘looking glass’, kaida ‘beach’

-kwa

-kwe < -kwa-y

5. Curvilinear objects, e.g. a:pi ‘snake’, hinipu ‘road’

-kha

-khay < -kha-y

6. Vertical or upright and standing objects, e.g. tsi:nu ‘dog’, haiku ‘tree’, dzawi ‘jaguar’

-na

-ne,-nay