Numeral Classifiers and Classifier Languages: Chinese, Japanese, and Korean [1 ed.] 1138054976, 9781138054974

Citation preview

Numeral Classifiers and Classifier Languages

Focusing mainly on classifiers, Numeral Classifiers and Classifier Languages offers a deep investigation of three major classifier languages: Chinese, Japanese, and Korean. This book provides detailed discussions well supported by empirical evidence and corpus analyses. Theoretical hypotheses regarding differences and commonalities between numeral classifier languages and other mainly article languages are tested to seek universals or typological characteristics. The essays collected here from leading scholars in different fields promise to be greatly significant in the field of linguistics for several reasons. First, it targets three representative classifier languages in Asia. It also provides critical clues and suggests solutions to syntactic, semantic, psychological, and philosophical issues about classifier constructions. Finally, it addresses ensuing debates that may arise in the field of linguistics in general and neighboring inter-disciplinary areas. This book should be of great interest to advanced students and scholars of East Asian languages. Chungmin Lee is Professor Emeritus in the Department of Linguistics at Seoul National University and a member of National Academy of Sciences, the Republic of Korea. Young-Wha Kim is Professor Emeritus in the Department of English Language & Literature at Hallym University in the Republic of Korea. Byeong-uk Yi is Professor in the Department of Philosophy at the University of Toronto in Canada.

Routledge Studies in East Asian Linguistics

The Grammar of Japanese Mimetics Perspectives from Structure, Acquisition and Translation Edited by Noriko Iwasaki, Peter Sells, and Kimi Akita Southern Min Comparative Phonology and Subgrouping Bit-Chee KWOK The Grammar of Chinese Characters Productive Knowledge of Formal Patterns in an Orthographic System James Myers Numeral Classifiers and Classifier Languages Chinese, Japanese, and Korean Chungmin Lee, Young-Wha Kim and Byeong-uk Yi

For more information about this series, please visit www.routledge.com/ Routledge-Studies-in-East-Asian-Linguistics/book-series/RSEAL

Numeral Classifiers and Classifier Languages Chinese, Japanese, and Korean

Edited by Chungmin Lee, Young-Wha Kim and Byeong-uk Yi

First published 2021 by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN and by Routledge 52 Vanderbilt Avenue, New York, NY 10017 Routledge is an imprint of the Taylor & Francis Group, an informa business © 2021 selection and editorial matter, Chungmin Lee, Young-Wha Kim and Byeong-uk Yi; individual chapters, the contributors The right of Chungmin Lee, Young-Wha Kim and Byeong-uk Yi to be identified as the authors of the editorial material, and of the authors for their individual chapters, has been asserted in accordance with sections 77 and 78 of the Copyright, Designs and Patents Act 1988. All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging-in-Publication Data A catalog record for this book has been requested ISBN: 978-1-138-05497-4 (hbk) ISBN: 978-0-367-63528-2 (pbk) ISBN: 978-1-315-16630-8 (ebk) Typeset in Times New Roman by Apex CoVantage, LLC

Contents

List of figures List of tables List of contributors Preface

1

ix x xi xiv

Introduction

1

Numeral classifiers and diversity of classifier systems

6

BYEONG-UK YI

1 Introduction 6 2 Classifiers and numeratives 8 3 Classifier languages 13 4 Classifiers and plural marking 26 Acknowledgments 29 Notes 29 References 35 2

Taxonomy of numeral classifiers: a formal semantic proposal J I U N - S H I U N G WU AND ONE - S OON HE R

1 Introduction 40 2 A confused state of affairs 42 3 A math-based taxonomy of C/M 47 4 A formal semantic account 50 5 Implications of the mass/count distinction 61 6 Conclusion 65 Acknowledgments 66 Notes 66 References 67

40

vi

Contents

3

Japanese semantics and the mass/count distinction

72

TA K A S H I I I DA

1 Introduction 72 2 Three kinds of numeral suffixes 74 3 Numeral phrase modifier bun 76 4 Individuated and non-individuated reference of noun occurrences 78 5 A simple test for identifying different kinds of numeral suffixes 80 6 Chopsticks, shoes, and twins 86 7 What does a sortal suffix contribute to the meaning of a sentence? 89 8 Conclusion 93 Notes 95 References 96 4

A continuum-based approach to the count-mass distinction in Korean

98

K WA N G - S U P KI M

1 Introduction 98 2 Syntactic differences between count and mass nouns 100 3 A continuum-based approach to Korean nouns 105 4 The so-called plural marker -tul as a distributivity marker 109 5 Conclusion 115 Notes 116 References 119 5

Definiteness, specificity, and genericity in numeral classifier languages

121

CHUNGMIN LEE

1 Introduction 121 2 Demonstrative anaphoric definite in discourse 121 3 Bare common N definites in non-subject position in Chinese, Korean, and Japanese 123 4 Demonstratives or second person pronouns in slur words 129 5 Specific indefinites 130 6 Dynamic genericity 130 7 Quantization, distributivity, and plurality in CL languages 132 8 Concluding remarks 133 Notes 134 References 134 6

Classifiers, articles, and bare nominals B Y E O N G - U K YI

1 Introduction: The DP hypothesis and bare nominals 137

137

Contents

vii

2 Longobardi’s account of bare nominals 140 3 Chierchia’s typology: mass nouns and kind-reference 146 4 Numeral classifiers and quasi-determiners 151 5 The subject constraint on Chinese nominals 155 6 Concluding remarks 159 Appendix 161 Acknowledgments 163 Notes 164 References 169 7

Japanese numeral quantifiers that count events

172

M A N A K O B U C HI - P HI L I P

1 Introduction 172 2 Previous studies 174 3 Japanese verbal event classifiers: survey 178 4 The NQV+suru construction 184 5 Concluding remarks 191 Notes 193 References 195 8

How classifiers affect the mental representation of entities

197

M U T S U M I I M AI AND JUNKO KANE RO

1 Introduction 197 2 Part I: influence of language on the ontological distinction and construal of physical entities 200 3 Part II: do classifier languages grammatically distinguish count nouns and mass nouns? 214 4 Conclusions and future research 225 Acknowledgments 227 Note 227 References 227 9

Descriptive function of numeral classifiers: a corpus-based analysis of numeral classifiers in Korean Y O U N G - W H A KI M

1 2 3 4 5 6 7

Introduction 231 Classifier vs. non-classifier languages 231 Numeral classifiers in Korean 232 Descriptive function of numeral classifiers 237 Head nouns and their numeral classifiers 242 Semantic taxonomy of numeral classifiers 245 Numeral classifiers frequently in use in Korean 249

231

viii Contents 8 Conclusion 253 Acknowledgments 254 Notes 254 References 256 Author index Language index Subject index

258 260 262

Figures

7.1 8.1 8.2 8.3 8.4 8.5 8.6a 8.6b 8.7a 8.7b 9.1 9.2 9.3 9.4 9.5 9.6

The basic structure for the Chinese classifier. Sample material sets for (a) a complex object trial, (b) a simple object trial, (c) a substance trial. Subject’s classification behavior in the no-word context in the word extension (neutral-syntax) tasks and non-lexical classification task. English speakers’ classification behavior in the (a) neutral-syntax condition, (b) count-syntax condition, and (c) mass-syntax condition. The stimuli of Saalbach and Imai’s study. Proportion of shape choice in each condition in (a) German, (b) Japanese, and (c) Chinese. Grand average ERPs for the trials involving object nouns in the violation conditions (dotted lines) in contrast to the matched condition (solid lines). Grand average ERPs for the trials involving substance nouns in the violation conditions (dotted lines) in contrast to the matched condition (solid lines). Grand average ERPs for the trials involving count classifiers in the violation conditions (dotted lines) in contrast to the matched condition (solid lines). Grand average ERPs for the trials involving mass classifiers in the violation conditions (dotted lines) in contrast to the matched condition (solid lines). Distributive readings of -tul vs. -ssik. Taxonomic structure of Kay (1971). The distinction of [+human] vs. [-human]. Hypernyms vs. hyponyms. Co-hyponyms as independent lexical items. The kinds of head nouns are co-hyponyms, sharing the same feature.

177 201 204 206 210 212 220 221 223 223 236 245 245 246 246 247

Tables

1.1 2.1 2.2 4.1 4.2 7.1 8.1 9.1 9.2 9.3

Standard forms of Korean numeral noun phrases Taxonomy of C/Ms derived from Zhang (2013) Types of C/M based on mathematical value The major properties of the five different types of nouns Four types of distributive operators Verbal event classifiers and countability in Donazzan (2013) Instructions used by Saalbach and Imai (in preparation) Structure types of NumPs Numeral classifiers used more than 10 times (3% in frequency) Numeral classifiers seldom in use

16 45 49 108 112 176 211 250 251 252

Contributors

One-Soon Her is a chair professor at the Department of Foreign Languages and Literature, Tunghai University, and the Graduate Institute of Linguistics, National Chengchi University in Taiwan. He received a Ph.D. in linguistics from the University of Hawaii in 1990. Before his return to Taiwan in 1992, he worked on machine translation in software industry in Utah for six years. In the last ten years, he has published more than 20 papers that center around numerals and numeral classifiers in the world’s languages. His most recent research interest is question types in the world’s languages. He is the founder and editor of Taiwan Journal of Linguistics, a diamond open access journal. Most of his more than 80 publications can be downloaded from his homepage at www3. nccu.edu.tw/~osh/. Takashi Iida is a professor emeritus of Keio University. He was educated at Tokyo University and the University of Michigan. He taught at Kumamoto University, Chiba University, Keio University, and Nihon University. He was the President of the Japan Association for Philosophy of Science from 2005 to 2011 and the President of the Philosophical Association of Japan from 2012 to 2015. He has written a four-volume introduction in Japanese to the contemporary philosophy of language, Gengo Tetsugaku Taizen (KeisoShobo, 1987–2002), a book on Wittgenstein (Kodansha, 1997), and a book on quantification in Japanese (NHK Publishing, 2019), as well as more than 50 papers. Mutsumi Imai is a developmental psychologist studying lexical and conceptual development. She received her Ph.D. from Northwestern University. Imai’s greatest interest is to specify the processes and mechanisms through which children build up the system of lexical as well as non-lexical knowledge. Her unique contribution to the field is her cross-linguistic perspective in the investigation of lexical acquisition. She has examined how universal and specific language factors affect lexical development, and she has also investigated the effect in the other direction, that is, how language and culture affect thought. She has established that the relation between cognitive development and language learning consists of a bi-directional bootstrapping process (Imai, Kanero, & Masuda, 2015; Imai & Masuda, 2013). Imai has been designated a fellow of the Cognitive Science Society and is also a governing board member of that society. She is also a fellow member of the Psychonomic Society and has served as Executive Board Member of the International Association for the Study of Child Language (IASCL).

xii

Contributors

Junko Kanero is an assistant professor of psychology at Sabancı University in Istanbul, Turkey. She received her Ph.D. in developmental psychology and neuroscience from Temple University in 2016. Her research interests include language development in childhood, the use of technology in language education, and interaction between language and non-linguistic cognition. In this book, she wrote a chapter with Mutsumi Imai discussing the roles of numeral classifiers in the mental representation of entities such as objects and substances. Kwang-sup Kim is a professor of the English Department, Hankuk University of Foreign Studies. He received his Ph.D. from the University of Maryland and Hankuk University of Foreign Studies. His main interests are in syntactic theory, comparative syntax, the syntax–semantics interface, and the syntax–phonology interface. He has recently published a book chapter titled ‘The Copula as the Nominative Case Marker’ (Oxford University Press, 2019), has written a book on last-resort strategies in minimalism, Minimalism and Last Resort (Hankook, 2018), and has published numerous articles in journals such as Linguistic Inquiry, Lingua, and Studies in Generative Grammar. He was President of the Korean Generative Grammar Circle in 2013–2015. Young-Wha Kim is now a professor emeritus, Department of English Language & Literature, Hallym University, Korea. She received her Ph.D. from Hankuk University of Foreign Studies, Seoul, Korea, in 1984, after doing dissertation research work at the University of Indiana, Bloomington, supported by a Fulbright research grant. She has published The Noun Phrases (Myengsagwu Chongnon written in Korean, Hankookmunhwasa, 2015) funded by NRF-2010812A00098. Other major books she published as a first editor are Plurality in Classifier Languages (2011), Negation & Negatives (2006), and Expletives (2002). Her main areas of research are in syntax, semantics, morphology, and language acquisition. Mana Kobuchi-Philip obtained her Ph.D. in linguistics from the Graduate Center, the City University of New York in 2003. Her thesis was on the semantic analysis of the numeral classifier, focusing on Japanese, and she has worked on various issues concerning quantification since, such as distributivity, plurality, atomicity, floating and binominal each (and Japanese zutsu), and partitive constructions. Her interest has also expanded to include pragmatic topics such as the focus particle mo ‘also’, shika ‘except for’, and ka (question particle) in Japanese. She is an adjunct assistant professor at Queens College (Queens, NY) and teaches linguistics and Japanese language. Chungmin Lee is now a professor emeritus at the Department of Linguistics, Seoul National University. Including Excellent Scholar Project, a five-year project, on semantics and cognitive science as PI, supported by the Korean National Research Foundation, he has worked on several projects. He edited (with Kiefer and Krifka) Contrastiveness in Information Structure, Alternatives and Scalar Implicatures (Springer, 2017) and is on the verge of publishing Evidentials and Modals (Brill) (co-edited with Park). His book entitled SemanticSyntactic Structures and Cognition (written in Korean) (Hankook) has just been published. He is scheduled to publish A Reference Grammar of Korean

Contributors

xiii

(Cambridge, with Heerak Chae and James Yoon). He is the editor of book series Language, Cognition, and Mind (LCAM) with Springer. He received his Ph.D. in linguistics from Indiana University, Bloomington, in 1973 and since taught semantics and cognitive science at Seoul National University. Jiun-Shiung Wu is a professor of linguistics and the Director, the Graduate Institute of Linguistics, National Chung Cheng University, in Chiayi, Taiwan. He received his Ph.D. in linguistics from University of Texas, Austin, in 2003. His research focuses on formal semantics, formal pragmatics, computational linguistics, and syntax–semantics interface. He has conducted research on tense and aspect, modality, temporal relations between sentences in a coherent discourse, discourse structure and discourse functions, classifiers and questions, and so on. He has published more than a dozen journal papers. He is the author of Intensification and Modal Necessity in Mandarin Chinese (2019, Routledge Studies in Chinese Linguistics) and Temporal and Atemporal Relations in Mandarin. Taiwan Journal of Linguistics Monograph Series No. 2 (2007). Byeong-uk Yi is a professor of philosophy at the University of Toronto. He received a Ph.D. in philosophy from UCLA and held academic appointments at University of Alberta, University of Queensland, University of Glasgow, University of Minnesota, and Kyung Hee University. He has published a book, Understanding the Many (Routledge, 2002), and articles in a variety of areas in logic, linguistics, and philosophy, including metaphysics, philosophical logic, philosophy of mathematics, philosophy of language, and semantics of classifier languages.

Preface

The project for this collection stems from the initiative of Young-Wha Kim for a cross-linguistic study of three major East Asian languages: Chinese, Japanese, and Korean. Unlike most European languages, all these languages have strong systems of numeral classifiers and are often given as representative examples of numeral classifier languages. But there are substantial differences among them, including considerable differences in their classifier systems. With the financial support of a 2008 research grant of Korea Research Foundation, Kim launched a project for examining similarities and differences of the three languages while comparing them with English and other familiar European languages (e.g., Germanic and Romance languages). She recruited an international team of scholars and, in 2009, organized a workshop where they presented works on Chinese, Japanese, and Korean classifier systems and noun phrase structures (e.g., presence or absence of plurality marking). And she edited a collection that includes developments of the works, Plurality in Classifier Languages (PCL), which was published in 2011. The present collection was planned as a sequel to PCL that reexamines numeral classifier systems of the three East Asian languages in light of the studies published therein to test major accounts of numeral classifiers. This project was initiated by Chungmin Lee. Invited by Samantha Phua to propose a book in Routledge Studies in East Asian Linguistics, he recruited the other editors (Kim and Yi) to propose a sequel to PCL. In 2017, the editors organized a workshop about numeral classifiers at Yonsei University, where an international and interdisciplinary group of scholars whose native languages are Chinese, Japanese, or Korean (five linguists, two philosophers, and two psychologists) presented works on classifier systems of the languages. Most of the participants in the workshop made contributions to this collection. And two Taiwanese linguists were invited to make a joint contribution about Chinese numeral classifiers. The workshop held at Yonsei University was made possible by the funding provided by an Insight Grant of the Social Sciences and Humanities Research Council of Canada (SSHRC) awarded to Byeong-uk Yi (Grant No. 435-2014-0592) and the logistic and financial support of Nikolaj J. Pedersen and Veritas Research Center of Underwood International College of Yonsei University. Their support is hereby gratefully acknowledged. One of the editors, Yi, would also like to acknowledge a Chancellor Jackman Research Fellowship of the University of Toronto, which helped him to organize the workshop and to write his contributions. Last but

Preface xv not least, the editors wish to thank Routledge for publishing this collection and its staff members (Christina Low, Samantha Phua, Jacy Hui, and Katie Peace among others) for their help and patience at various stages of the challenging project. Chungmin Lee Young-Wha Kim Byeong-uk Yi May 31, 2020

Introduction

Most languages have numeral noun phrases resulting from combining nouns directly with numerals (e.g., ‘three cows’). In some languages, however, usual translations of such phrases have special expressions accompanying the numerals or intervening between nouns and numerals (e.g., the Korean so sey mari ‘cow three CL’ / sey mari -uy so ‘three CL-GEN cow’). The special expressions (e.g., the Korean mari) are called numeral classifiers, and languages with a substantial system of numeral classifiers are called numeral classifier languages. Chinese, Japanese, and Korean are three representative classifier languages, compared to non-classifier languages, such as English and Italian. The purpose of this collection of nine chapters is to extensively explore and present a comprehensive analysis of the numeral classifier systems of the three representative classifier languages. The contributors to this publication consist of four Korean, four Japanese, and two Chinese scholars, including linguists, philosophers, and psychologists. Seven of the nine chapters comprise theoretical discussions about the syntactic, semantic, and pragmatic characteristics relative to the classifier systems of the three East Asian classifier languages, while two chapters are devoted to empirical studies concerning psychological aspects of the Japanese language and to theoretical discussions supported by a corpus analysis of Korean examples. And thus, this book elaborates on linguistic, philosophical, and psycholinguistic discussions, including empirical analyses about the classifier systems of Chinese, Japanese, and Korean. In “Numeral Classifiers and Diversity of Classifier Systems” (Chapter 1), Byeong-uk Yi gives an overview of various kinds of (numeral) classifiers and (numeral) classifier languages and discusses some common misconceptions about classifier languages. To do so, he begins by noting that the term ‘classifier’ is used ambiguously: it is sometimes used in the narrow sense only for special expressions occurring in numeral noun phrases of classifier languages (e.g., counterparts of ‘two students’) that have no counterparts in non-classifier languages and sometimes in a broad sense for a larger group of expressions that include not only classifiers in the narrow sense but also their syntactic cousins: measure words (e.g., ‘liter’, ‘cupful’), expressions he calls paranumerals (e.g., ‘pair’, ‘couple’, ‘dozen’), and so on. (He proposes to reserve the term ‘classifier’ for classifiers in the narrow sense while using ‘numerative’ for classifiers in the broad sense.) He defines classifier languages as languages with a substantial system of classifiers (in the narrow sense) and divides such languages

2

Introduction

into two kinds: (a) mandatory classifier languages, which require classifiers (or numeratives) in numeral noun phrases (e.g., counterparts of ‘two students’), and (b) non-mandatory classifier languages, where some nouns can directly combine with numerals. While most theoretical discussions of classifier languages assume that all classifier languages must be mandatory classifier languages, he argues that most of them are non-mandatory classifier languages. In doing so, he gives a close examination of classifier systems of Chinese (Mandarin), Japanese, and Korean to argue that even these languages fall short of being mandatory classifier languages. He also argues against the prevalent view that classifiers and plural markers are complementary. To do so, he discusses some languages (e.g., Itzaj Maya, Korean) that can have both classifiers and plural markers in the same numeral noun phrases. In “Taxonomy of Numeral Classifiers: A Formal Semantic Proposal” (Chapter 2), Jiun-Shiung Wu and One-Soon Her assume that in a numeral classifier language, a classifier (C) or measure word (M) is needed to link a noun (N) and its numerical quantifier (Num). They first illustrate the confusions surrounding the convergence and divergence between Cs and Ms and offer some clarifications. They argue, from the multiplicative mathematical concept proposed in Greenberg (1990[1972]: 172), Au Yeung (2007), and Her (2012a), that C and M converge as the multiplicand, with Num as the multiplier. Yet, crucially, it is assumed in this chapter that a C’s value is numerical and necessarily 1, while an M’s value can be anything but 1, thus ¬1. An M’s value can therefore be numerical, for example, 2, 6, 12, and so on, or non-numerical, for example, kilo, meter, liter, bottle, and so on. This chapter then proposes a new and comprehensive taxonomy based on the different kinds of mathematical values encoded in C/Ms and further proposes a formal semantic account of this taxonomy. In Chapter 3, “Japanese Semantics and the Mass/Count Distinction”, Takashi Iida argues that Japanese indeed has countable predicates and that they can be recognized by the kinds of numeral suffixes which can modify them. He claims that Japanese numeral suffixes are divided into three classes: (a) sortal suffixes, or classifiers; (b) unit-forming suffixes; and (c) measure suffixes. These classes can be distinguished from each other by some simple tests. He states that a sortal suffix’s contribution to the meaning of a sentence in which it occurs is not to its truth-conditional content but only to its conventional implicature and that a noun which typically occurs with a sortal suffix has an individuating force by itself. He then argues that plural logic is applicable only to countable predicates; it is not applicable to non-countable predicates. Thus, the first question that must be settled before applying plural logic to Japanese is to make sure that the language has countable predicates. In Chapter 4, “A Continuum-Based Approach to the Count-Mass Distinction in Korean”, Kwang-sup Kim provides a continuum-based approach to the countmass distinction. After rejecting the traditional binary view that [+mass] is [-count] and vice versa, he proposes that one type can be more countable or less countable than another type: that is, there are many different types in between the countable type and the non-countable type. There are three important features involved in deciding whether a noun is more or less countable than another: [+ individual member], [+distributable], [+countable]. The feature geometry of the three features is as follows: (a) [+countable] → [+distributable] and (b) [+distributable]

Introduction 3 → [+individual member]. The nouns considered in this chapter can be classified into five types on the basis of the three features: English Count Nouns: [+countable] > Korean Count Nouns: [+countable (small numbers)] > Korean Count-Mass Nouns: [+distributable] > English APs headed by the and British collective nouns [+individual member] > English/Korean Mass Nouns: [-individual member]. This chapter also shows that the so-called plurality marker -tul of Korean is a distributive marker, not a marker used for countable nouns. In Chapter 5, “Definiteness, Specificity and Genericity in Numeral Classifier Languages”, Chungmin Lee states that in a dialogic question-and-answer discourse situation between interlocutors, a fixed article-like demonstrative is employed for dynamic anaphoric definite use, which is independent of proper deictic use of demonstratives in article-less languages, particularly in Korean, Japanese, and Chinese. Countable bare common nouns in numeral classifier languages and other article-less languages freely occur in argument position for unique definite use (Jenks 2018) due to presupposition maximization (Heim 1991). They often occur as immediate-situation definites (Hawkins 1978). However, in this chapter, Lee will also show how bare common nouns can function as anaphoric definites as well, even in non-subject/topic position in a sentence preceded by an utterance with an indefinite DP referred to. He claims this happens in Chinese, Korean, and Japanese (contra Jenks 2018). In this chapter, it will also be shown that the numeral ‘one’ before a human noun develops into a specific, indefinite marker in Korean and sees specificity effects in other languages. Finally, the issue of genericity in terms of information structure in article-less vs. article languages will be discussed, highlighting its dynamicity. In Chapter 6, “Classifiers, Articles, and Bare Nominals”, Byeong-uk Yi discusses an analysis of nominal phrases of classifier languages. Nominal phrases (e.g., ‘the cow’, ‘a cow’) have been considered noun phrases, that is, phrases headed by nouns (e.g., ‘cow’). Recently, however, many linguists (e.g., Szabolcsi 1983, Abney 1987, Longobardi 1994) have proposed and developed an influential alternative view: nominal phrases are not headed by nouns but by determiners (e.g., articles, quantifiers, demonstrative adjectives) and thus should be called determiner phrases. Influential as this view (which is called the DP hypothesis) is, it has serious difficulties in dealing with nominal phrases of classifier languages, for most classifier languages (including Chinese, Japanese, and Korean) regularly use bare nominals (i.e., nominal phrases without determiners). To avoid this problem, some influential accounts of classifier languages (e.g., Chierchia 1998a & 1998b, Cheng and Sybesma 1999) reject the DP hypothesis while relating regular use of bare nominals in classifier languages to the presence of classifiers. Yi argues that even such accounts fail to do justice to the regularity of bare nominals in classifier languages because they accept a key assumption underlying the DP hypothesis: bare nominals must be exceptions to a general rule that are subject to serious constraints. To do so, he points out that most human languages (e.g., Latin, Classical Chinese, most classifier languages) do not have articles and use bare nominals with both definite and indefinite meanings, and argues that use of bare nominals is not an exception but a rule in such languages. He also argues that analyses of bare nominals based on the assumption (including those based on the DP hypothesis or its modifications) not only ignore languages liberal on bare nominals but also

4

Introduction

give incorrect accounts of constraints on bare nominals in the very languages they focus on (e.g., Germanic and Romance languages, modern dialects of Chinese). In doing so, he gives a close examination of a special feature of Chinese bare nominals that Cheng and Sybesma (1999) highlight to support their account of classifiers and presents an alternative account of the special feature while criticizing their account. In Chapter 7, “Japanese Numeral Quantifiers That Count Events”, Mana KobuchiPhilip asserts that an event classifier is a unit for counting events rather than objects. She discusses the Japanese event classifier that appears in the verbal domain. Since this topic has not been extensively discussed in the literature, she provides a survey of Japanese event classifiers, sorting them into several types. Initially, the survey examines two event classifiers, do ‘times’ and kai ‘times’, which are both frequently used due to their generality, both applying to virtually any type of event. The focus then shifts to one of the more productive constructions with Japanese event classifiers, namely, those involving the light verb suru ‘do’. This chapter concludes with some general comments on the linguistic phenomena of counting events and counting objects, with special attention to the flexible nature of both modes of counting. In Chapter 8, “How Classifiers Affect the Mental Representation of Entities”, Mutsumi Imai and Junko Kanero state that the count/mass distinction has been noted as one of the most fundamental conceptual distinctions, as it is directly relevant to the identity of entities in the world. Objects are individuated, whereas substances are non-individuated. A cup is no longer a “cup” when it breaks into pieces. In contrast, clay stays “clay” even if it is divided into small lumps. They claim that understanding the difference between objects and substances is fundamental to our cognition, and without it, we are unable to judge “the sameness” of an entity. In English and many other languages, objects are labeled with count nouns, which can take singular and plural forms, whereas substances are labeled with mass nouns that cannot be enumerated. Quine (1969) proposed that the grammatical distinction is necessary for children to learn the difference between objects and substances, and thus speakers of a language that seem to lack systematic count/ mass marking, such as Japanese, fail to understand the distinction. This chapter explores the Quinean claim in three sections. The first section addresses whether Japanese children have abilities to distinguish objects and substances. The second section challenges the very notion of the Japanese language lacking the count/ mass grammar and entertains the possibility that the numeral classifier system in Japanese grammatically marks objects and substances. In the final section, Imai and Kanero propose general conclusions concerning the linguistic representation of objects and substances in the mind of Japanese speakers and discuss the relation between language and cognition as well as possible future research. In Chapter 9, “Descriptive Function of Numeral Classifiers: A Corpus-Based Analysis of Numeral Classifiers in Korean”, Young-Wha Kim aims to define the grammatical function and the relevant meaning of numeral classifiers in Korean. She puts focus on the descriptive function of numeral classifiers in relation to their head nouns, especially advocating the semantic relevance between numeral classifiers and their possible head nouns. In support of claiming the descriptive function of classifiers, the formation of compound nouns from classifiers and their possible head nouns is examined. Those classifiers constructing compound nouns retain

Introduction 5 their descriptive meaning but lose their function as numeral classifiers. She claims numeral classifiers keep the descriptive function in numeral classifier phrases, by which classifiers leave the task of counting or measuring to the numerals that accompany the classifiers. In this chapter, by means of a corpus analysis, it is shown that the general numeral classifier kay ‘CL’ is normally prevalent in use and that numeral classifiers are decreasing in current Korean. It is also shown that numerals not accompanying classifiers are often preferred in current Korean and that classifiers may help describe the fundamental nature of their head nouns. As has been introduced thus far, the contributors to this book engage in deep and detailed discussions on the theoretical issues based on the preceding studies, including works in PCL (Plurality in Classifier Languages; Kim, Young-Wha et. al., 2011). This process launched inspiring inquires triggered by possible gaps between theory and reality and established theoretical hypotheses, which were then verified with empirical studies conducted from data collection and corpus analysis. All together, these nine chapters comprise theoretical discussions, experimental analyses, and perspective implications on practical details about the classifier systems. This volume is a collaboration between Chinese, Japanese, and Korean scholars discussing the characteristics of the classifier systems unique to their native languages. In examining each of their contributions as an entire work, critical clues for the syntactic and the semantic issues of the classifier constructions are clearly provided in the hope that ensuing debates may prompt and expand further stimulating discussions in the field of linguistics.

1

Numeral classifiers and diversity of classifier systems Byeong-uk Yi

1

Introduction

Many languages, including most Indo-European languages, have numeral noun phrases resulting from combining nouns directly with numerals. 1 For example, the English noun ‘cow’ can directly combine with numerals (‘one’, ‘two’, ‘three’, etc.) to yield phrases one can use to talk about numbers of cows, such as those in ‘Alice saw one cow in the barn’, ‘Alice saw two cows in the barn’, ‘Alice saw three cows in the barn’, and so on. In a wide variety of languages, however, usual translations of such phrases have special expressions intervening between nouns and numerals.2 The special expressions are called numeral classifiers (in short, classifiers),3 for they combine with numerals and usually accord with a limited group of nouns. And languages with a substantial system of numeral classifiers are called numeral classifier languages (in short, classifier languages). In this chapter, I give a more precise characterization of numeral classifiers and discuss classifier systems of various languages. In doing so, I elaborate on the classifier systems of Mandarin (a dialect of modern Chinese), Japanese, and Korean and argue that all these languages have nouns that can directly combine with numerals.4 I also discuss languages (e.g., Itzaj Maya, Korean) that can have both classifiers and plural markers5 in the same numeral noun phrases. Say that a noun is directly numerable if it can directly combine with numerals. Then we can call classifier languages with no directly numerable nouns mandatory classifier languages and those with directly numerable nouns non-mandatory classifier languages. Most accounts of the function of classifiers hold that all classifier languages are mandatory. We can state this thesis as follows: T1. Mandatory Classifier Language Thesis: No nouns of classifier languages are directly numerable. Some who reject this thesis might still hold its weaker cousin: T2. Obligatory Classifier Thesis: No nouns with matching classifiers are directly numerable. For T2 does not imply T1 because some classifier languages might have directly numerable nouns with no matching classifiers.

Diversity of classifier systems 7 The two theses are closely related to the assumption that classifiers conflict with plural markers, such as the English -s. The assumption dates back to the influential study of classifier and plural marking by Sanches and Slobin (1973), who proposed “the hypothesis that plural marking and numeral classifiers do not co-occur” (ibid., 1). We can formulate the hypothesis as follows: T3. The Exclusion Thesis: A numeral noun phrase featuring a noun cannot have both a matching classifier and a plural marker for the noun.6 They attempt to explain this by taking all nouns with matching classifiers to be (or to be similar to) mass nouns of languages with the singular/plural morphology (e.g., English) while taking classifiers to have the same semantic function as measure words (‘liter’, ‘drop’, etc.) have in numeral noun phrases involving them: ‘three liters of water’, ‘three drops of water’, and so on. Call nouns similar to mass nouns (in the relevant respects) para-mass nouns.7 We can then formulate the theses about classifier language nouns they hold to explain T3 as follows: T4. Mass Noun Thesis: a. All nouns with matching classifiers are mass (or para-mass) nouns. b. No mass (or para-mass) nouns are directly numerable. Now, T4 implies the obligatory classifier thesis (T2). And most accounts of classifiers in theoretical linguistics hold T2 by accepting versions of the explanation of T3 proposed by Sanches and Slobin. Most of them also accept a cousin of T4(a): T4. aʹ. All nouns of classifier languages are mass (or para-mass) nouns. This, together with T4(b), implies the mandatory classifier language thesis (T1).8 Most contemporary accounts of classifiers, as noted above, aim to explain T3 and imply T1 and T2. I think such accounts are fundamentally mistaken. There is a wide variety of classifier languages with directly numerable nouns and, moreover, classifier languages with directly numerable nouns with matching classifiers. Such languages include languages with highly developed classifier systems that are usually considered mandatory classifier languages, such as Chinese, Japanese, and Korean. Furthermore, the exclusion thesis (T3) is also false. Some languages with well-developed classifier systems (e.g., Itzaj Maya) have nouns that can take both classifiers and plural markers in the same phrases. So I reject usual accounts of classifiers and hold an account that is compatible with and helps to explain the existence of languages violating T1–T3: the paranumeral account. On this account, classifiers are semantic cousins of numerals. Consider the English ‘pair’, ‘dozen’, and ‘score’. They have close semantic ties to numerals (‘two’, ‘twelve’, and ‘twenty’) and can be considered their cousins. Call such expressions paranumerals. Then the paranumeral account holds that classifiers are paranumerals of a special kind, namely, cousins of numerals for 1. Classifiers, on the account, are like the English ‘pair’, ‘couple’, ‘brace’, and ‘yoke’ in relating to a number but differ from them in relating to 1 rather than 2. And

8

Byeong-uk Yi

they differ from each other while relating to the same number, 1, basically in the way that the English paranumerals for 2 differ from each other while relating to 2. I aim in this chapter to give a survey of classifiers and classifier languages to show problems of the usual accounts of classifiers.9 In the next section (§2), I distinguish classifiers from a larger group of expressions that include their syntactic peers: measure words (e.g., ‘liter’, ‘cup’), paranumerals (e.g., ‘pair’, ‘dozen’), and so on. I follow Edkins (1857, 119f; 1864, 127) to call expressions of the larger group numeratives and characterize classifiers as numeratives of a special kind. In §3, I argue against T1 and T2. In doing so, I discuss Mandarin (Chinese), Japanese, and Korean to show that all these languages have directly numerable nouns with matching classifiers. In §4, I discuss languages with numeral noun phrases with both classifiers and plural markers to argue against T3.

2

Classifiers and numeratives

To characterize classifiers, it is usual to compare them with measure words, which include (a) the English ‘pound’, ‘cup’, and so on (that occur in ‘one pound of meat’, ‘two cups of coffee’, etc.) and (b) their classifier language counterparts. The term classifier is used ambiguously. It is sometime used in a narrow sense to exclude measure words (e.g., ‘pound’ and its classifier language counterparts) but sometimes broadly for a larger group of expressions that include, among others, measure words as well as classifiers in the narrow sense.10 Our main concern is the function of classifiers in the narrow sense. For classifier languages (e.g., Chinese) are so called because they have classifiers in this sense, not because they have classifiers in the broad sense, which can be found in most languages, including those that are not considered classifier languages (e.g., English).11 To give a proper account of the syntax and semantics of classifiers in the narrow sense, it is crucial to distinguish the smaller group comprehending them from the larger group that includes measure words as well. So I use another term, numerative, for expressions belonging to the larger group while using classifier only for those belonging to the smaller group.12 This helps to discuss both features that classifiers (in the narrow sense) share with other expressions in the larger group (i.e., numeratives) and those specific to them.    and measure words 2.1  Classifiers In English, for example, there is a clear syntactic difference between two kinds of numeral noun phrases: (a) those relating to the number of individuals of a certain kind (e.g., cows); (b) those relating to the amount of some stuff or substance (e.g., milk, meat). Consider (1a)–(1c): (1) a. b. c.

three cows three pounds of meat three kilograms of meat

Diversity of classifier systems 9 In (1a), ‘three’ combines directly with ‘cow’ (which takes the plural form). In (1b)– (1c), by contrast, the measure words ‘pound’ and ‘kilogram’ intervene between nouns and numerals. But some languages have translations of (1a) and (1b) that draw syntactic parallels, for they have special expressions intervening between nouns and numerals. Consider the usual translations of (1a) into Mandarin (Chinese), Japanese, and Korean: (2) a. san tou 3 CL ‘three cows’ b. ushi san cow 3 ‘three cows’ c. so sey cow 3 ‘three cows’

niu cow

[Mandarin]

tou CL13

[Japanese]

mali CL

[Korean]

In addition to nouns and numerals, these phrases have classifiers (the italicized) that accord with some nouns, such as those for cows: niu (Mandarin), ushi (Japanese), and so (Korean).14 They are like measure words of those languages in following numerals. Thus (2a) draws a syntactic parallel with the Mandarin translation of (1b): (3) a. san bang rou 3 pound meat ‘three pounds of meat’15

[Mandarin]

Similarly, (2b) and (2c) draw syntactic parallels with Japanese and Korean translations of (1c): (3) b. niku san kiroguramu meat 3 kilogram ‘three kilograms of meat’ c. koki sam khilo16 meat 3 kilo ‘three kilograms of meat’17

[Japanese] [Korean]

In the Chinese (2a) and (3a), the numeral with the classifier or measure word occurs before the noun (pre-nominal). In the Japanese and Korean phrases mentioned above, by contrast, the numeral with the classifier or measure word occurs after the noun (post-nominal). But Japanese and Korean have alternative translations of (1a) where numeral-classifier compounds occur pre-nominally: (2) bʹ. san tou 3 CL ‘three cows’

-no ushi GEN cow

[Japanese]

10

Byeong-uk Yi cʹ. sey mali 3 CL ‘three cows’

-uy so GEN cow

[Korean]

Except for -no and -uy, these phrases have the same structure as the Chinese (2a).18 And they draw parallels with alternatives to (3b) and (3c): (3) bʹ. san kiroguramu -no 3 kilogram GEN ‘three kilograms of meat’ cʹ. sam khilo -uy 3 kilo GEN ‘three kilograms of meat’

niku meat

[Japanese]

koki meat

[Korean]

  2.2  Numeratives Some languages, we have seen, deploy classifiers as syntactic peers of measure words: they take the same syntactic position as measure words in numeral noun phrases, and the numeral noun phrases containing them draw syntactic parallels with measure constructions. In his Mandarin grammar, Edkins (1857/1864) recognizes a large class of expressions that can take the same position as measure words in numeral noun phrases and calls them “numeratives” (1857, 119f; 1864, 127).19 I follow him in calling measure words and their syntactic peers numeratives.20 Then some languages, we have seen, have numeratives of a special kind (viz., classifiers) that can rarely21 be found in other languages. If so, are all numeratives either measure words or classifiers? Most works on classifier languages divide numeratives into classifiers and measure words and assume that all numeratives (or all except classifiers) have essentially the same semantic function.22 But this assumption has serious problems. Consider some English numeral noun phrases: (4) a. b. c. d. e. f.

three {pairs, dozens} of cows three {kinds, species} of cows three groups of cows three slices of {beef, oranges} three cups of {coffee, beans} three pounds of {beef, oranges}

These phrases draw syntactic parallels and the italicized expressions (e.g., ‘pair’, ‘dozen’) have the same position in the phrases as ‘pound’ has in (4f). Thus numeratives of English include ‘pair’, ‘kind’, and ‘group’, as well as ‘pound’, ‘cup’, and ‘slice’. The same holds for classifier languages. For example, Chinese, Japanese, and Korean have counterparts of the English numeratives in (4a)–(4f) that take the same position in their translations.23 (These languages additionally have numeratives of another kind (viz., classifiers) that occur in counterparts of the likes of (1a).) Can we take all the numeratives in (4a)–(4f) to have the same semantic function? I think not. Although they have the same syntactic position in the phrases,

Diversity of classifier systems 11 this does not mean that they are semantically homogenous. This is clear, I think, if we compare ‘pair’, ‘dozen’, ‘score’, and so on with ‘pound’, ‘kilogram’, ‘liter’, and so on. The latter are paradigmatic examples of measure words, which indicate units for measuring amounts (e.g., weight, volume) and can match mass nouns (e.g., ‘milk’, ‘meat’) – ‘three pounds of meat’ and ‘three liters of milk’, for example, are well-formed.24 By contrast, ‘pair’, ‘dozen’, and ‘score’ do not pertain to amounts or their measurement and can match only count nouns. For example, ‘*three dozen milk’, ‘*dozens of meat’, ‘*three score milk’, ‘*a score of milk’, and ‘*scores of meat’ are all ill-formed, while ‘three dozen eggs’, ‘dozens of books’, ‘three score years’, ‘a score of men’, and ‘scores of letters’ are wellformed. Thus Pelletier and Schubert (2003) distinguish ‘score’ and the like from ‘liter’ and the like in formulating the usual distinction between mass and count nouns. They say, “the usual view claims . . . mass expressions . . . are susceptible to measurement phrases like litres of and amount of . . . [while] count expressions . . . are susceptible to counting phrases like five, a score of . . .” (ibid., 250). It would also be implausible to take ‘kind’, ‘species’, and ‘group’ to be measure words. Unlike ‘slice’ and ‘cup (or cupful)’, which might be taken to indicate rough units for measuring amounts, ‘kind’ and ‘group’ do not pertain to amounts at all.25 In his classic in Chinese linguistics, Chao (1968, 584–620) recognizes various kinds of numeratives, including classifiers.26 He classifies Mandarin numeratives into nine groups. Five of them are the core groups, which Peyraube (1998, 61) says are recognized by most Chinese linguists.27 It is useful to consider those five groups: (A) (B)

(C) (D) (E)

classifiers: tou (for humans), ben (for books), and so on group numeratives: (B1) qun ‘group, flock’, zu ‘section, group’, and so on (B2) zhong ‘kind, species’, lei ‘kind, category’, and so on (B3) dui ‘pair, couple’, shuang ‘pair, couple’, da ‘dozen’, and so on partitive numeratives: pian ‘slice’, di ‘drop’, and so on container numeratives: bei ‘cup’, xiang ‘box’, and so on standard numeratives: bang ‘(Chinese) pound’, li ‘(Chinese) mile’, and so on

It is straightforward to apply this classification of numeratives to other classifier languages. And we can see that Japanese and Korean also have numeratives of all five groups. As the list given above suggests, I think Chao’s group numeratives, (B), are quite heterogeneous. They include two subgroups, (B1) and (B2), that Edkins calls (a) “collectives” and (b) numeratives “expressive of kind or manner”, respectively (1864, 128; original italics). And it is useful to distinguish another subgroup, (B3), from the two.28 Numeratives belonging to (B3) are counterparts of the English paranumerals: ‘pair’, ‘dozen’, ‘score’, ‘gross’, and so on. These words have clear relations to numbers (2, 12, 20, etc.), and the same holds for their Chinese counterparts: dui and shuang relate to 2 and da to 12. They have clear semantic ties to numerals and can be considered their cousins. I call such numeratives paranumeral numeratives (in short, paranumerals).29

12

Byeong-uk Yi

   and paranumerals 2.3  Numerals English and Chinese alike, we have seen, have paranumeral numeratives. And they are prevalent in classifier languages. Japanese has tsui ‘pair/couple’ and daasu ‘dozen’; Korean has ssang ‘pair/couple’, khyeley ‘pair’, and tasu ‘dozen’; and Burmese has three paranumerals for 2, one for 12, and one for 144 (viz., the Burmese counterpart of ‘gross’) (Pe 1965, 180).30 Moreover, some languages might have a series of paranumerals one can hardly distinguish from (regular) numerals. They might be as close to numerals as the ‘hundred’ in ‘hundreds of books’ is to the ‘hundred’ in ‘three hundred books’. As Greenberg observes, “It is particularly common for classifiers not to occur with higher units of the numerical system and their multiples, e.g., 10, 20, 60, 100, 300” (1972, 168). In Mandarin (Chinese), for example, ba shi ‘eighty’ [8 ten] and wu bai ‘5 hundred’ [5 hundred] can directly combine with nouns, as in (5a)–(5b): ba shi ren 8 ten man ‘eighty people’ b. wu bai lohan 5 hundred arhat ‘5 hundred arhats’

(5) a.

[Mandarin; Chao (1968, 595)] [Mandarin; Chao (1968, 595)]

In Burmese, too, numerals for multiples of 10, 100, and so on can directly combine with nouns. In his article about Burmese classifiers, Burling says: Burmese is peculiar in its treatment of numbers divisible by ten (including 20, 30, 150, 3620, etc.). . . . One cannot say the equivalent of “cows twentyanimals,” as one would say . . . “cow three-animals,” but must instead say nwâ hnashé “cows two-tens.” (1965, 252) Pe recognizes this to include N – Num + T (where ‘T’ is for “ten or multiples of ten”) as a type of Burmese numeral noun phrases (1965, 165). Invoking the (apparent) numerative-free examples, one might conclude that in Mandarin and Burmese, some numerals can directly combine with nouns. So Cornyn says that “Tens, hundreds, and so on are not followed by a classifier” in Burmese (1944, 27). Chao (1968, 595f) and Burling (1965) give a different analysis. Chao (1968) takes shi and bai in (5a)–(5b) to serve as numeratives and lists numerals for powers of ten (shi ‘ten’, bai ‘hundred’, quian ‘thousand’, etc.) among group numeratives. Similarly, Burling includes their Burmese counterparts (she ‘ten’, ya ‘hundred’, etc.) among numeratives (1965, 262), and Pe says that they might be included among “set-numbers”, that is, paranumeral numeratives (1965, 180).31 Thus, Greenberg says “Many analysts consider words for ‘ten’, ‘hundred’, and so on in these languages [e.g., Burmese] as a subtype of classifiers [viz., numeratives]” (1972, 172).32 I do not think there is a convincing reason for either accepting or rejecting the Chao-Burling analysis. Chao and Burling support the analysis by invoking the

Diversity of classifier systems 13 mandatory classifier language thesis for Mandarin or Burmese, but one might plausibly take (5a)–(5b), for example, to lack numeratives to conclude that the thesis fails to hold for Mandarin. I think the analysis is as good as (if not clearly superior to) the usual Cornyn-style analysis. One cannot argue that the Chao-Burling analysis cannot be correct because numerals (or their identical twins) are too different from any kind of numeratives, including paranumeral numeratives. Moreover, we can imagine a small change in Mandarin or Burmese that would favor the analysis. Suppose that the pronunciation of shi in (5a) changes just a bit (e.g., an unusual tone) to make it distinguishable from when it is used as a (regular) numeral, as in ba shi ge ren ‘eighty people’ [8 10 CL person]. This would clearly support Chao’s analysis as one for the imagined variant of Mandarin. I think this helps to highlight the semantic diversity that numeratives have despite their syntactic affinity. Occurring as syntactic peers of measure words, some numeratives might have the same semantic function as (regular) numerals.33

3

Classifier languages

Numeral classifiers, we have seen, are numeratives of a special kind. And some languages have substantial systems of classifiers integrated into their numerative systems. Such languages are called numeral classifier languages (in short, classifier languages). They include many East and Southeast Asian languages (e.g., Chinese, Japanese, Korean, Thai, Burmese, Vietnamese, Malay), some American languages (e.g., some Mayan languages, Ojibway), and a small number of Indo-European languages (e.g., Bengali, Persian). According to Aikhenvald (2000, 101 & 121ff), classifier languages are widespread in East and Southeast Asia and Oceania, prevalent in South and Central America, and found in scattered pockets in North America but rare in Africa and absent from Australia. Gil (2008, §2) classifies 140 out of 400 languages in his data set as classifier languages.34 We can distinguish classifier languages into two kinds: the mandatory and the non-mandatory. Mandatory classifier languages (or languages with mandatory classifier systems) are those with no directly numerable nouns, and non-mandatory classifier languages (or languages with non-mandatory classifier systems) are those with directly numerable nouns. Most contemporary accounts of syntax and semantics of classifiers assume that all classifier languages are mandatory ones (T1). As is well known from early typological studies of classifiers, such as Greenberg (1972) and Allan (1977), however, this assumption is not correct. Moreover, many languages often considered mandatory classifier languages (e.g., Chinese, Japanese) fall short of being mandatory. 3.1   Non-mandatory classifier languages Burling says, “In many languages of Southeast Asia, a number is never used without being accompanied by one of the special morphemes, known as classifiers [viz., numeratives]” (1965, 244; my italics). And Cheng and Sybesma hold: “In

14

Byeong-uk Yi

Chinese . . . in order for a noun to be countable, a measure word or classifier is always necessary” (1998, 386; my italics).35 Some might take the same to hold for all classifier languages to accept T1: T1. Mandatory Classifier Language Thesis: No nouns of classifier languages are directly numerable. But this thesis does not follow from the characterization of classifier languages given above. It is one thing to say that a language has a substantial system of classifiers and quite another to say that the system allows no noun whatsoever to combine directly with numerals. And many languages with a substantial classifier system have directly numerable nouns. In his pioneering study of classifiers that compares a wide range of classifier languages, Greenberg (1972) points out several problems with T1.36 One of the main problems he points out is that “In some languages, always considered to be numeral classifier languages, the group of nouns which do not take classifiers is still more extensive (e.g., Vietnamese)” (ibid., 168). This statement, it seems, reflects Emeneau’s account of Vietnamese numeratives (1951, 84–109). Emeneau divides Vietnamese nouns into two kinds: the “classified” and the “nonclassified” (ibid., 84). Say that a noun is classified if it has a matching classifier and that it is nonclassified if it is directly numerable but has no matching classifier.37 Then he takes 178 of the 471 nouns in his basic Vietnamese vocabulary to be nonclassified (ibid., 93). They include Vietnamese counterparts of ‘girl’, ‘boy’, ‘artist’, ‘neck’, ‘soul’, ‘body’, ‘continent’, ‘story’, and so on (ibid., 94 and 100).38 Like Greenberg (1972), most typological studies of classifiers recognize nonmandatory classifier languages (see, e.g., Allan (1977, 286) and Aikhenvald (2000, 98)). And Gill (2008, §2) classifies 68 of the 140 languages he considers classifier languages as such languages. I think even this seriously underestimates the prevalence of non-mandatory classifier languages. Some of the remaining 78 languages in his list fall far short of being mandatory. They include Vietnamese but Vietnamese, we have seen, is by no means a mandatory classifier language. The same holds for Korean (see §3.3.1). And some languages with stronger classifier systems that are usually considered mandatory (e.g., Mandarin, Japanese) also fall short (see §3.3.2–3).    classifiers 3.2  Optional Allan recognizes non-mandatory classifier languages: “Burmese and Vietnamese have large numbers of nouns which do not occur with a classifier” (1977, 286). Still, he holds that “Numeral classifier languages . . . are so called because a classifier is obligatory in many expressions of quantity” (1977, 286; my italics).39 What does he mean by “obligatory” if he does not mean to rule out non-mandatory classifier languages? He might perhaps use it to suggest that in numeral noun phrases involving classifiers, the classifiers are obligatory because the nouns matching them are not directly numerable. Say that a classifier matching a noun is optional for the noun if the noun is directly numerable

Diversity of classifier systems 15 (otherwise, it is obligatory for the noun).40 Then we can formulate the suggested thesis as follows: T2. Obligatory Classifier Thesis: Classifiers matching nouns are not optional for the nouns (i.e., no classified nouns are directly numerable). Although T1 implies T2, T2 does not imply T1; some languages might have both classified and nonclassified nouns. So some who reject T1 might still hold T2. But T2 is also false. As Greenberg points out, “In many languages the classifiers are not compulsory even for the restricted set of nouns that have them” (1972, 168). For an example, he refers to Jacob’s study of Khmer: Khmer has nouns with optional classifiers, such as ko:n ‘child’ and sdac ‘king’ (Jacob 1965, 145). And Gil (2008, §2) takes all 62 languages he considers non-mandatory classifier languages to have optional classifiers. They include Minangkabau, Hungarian, and Armenian. Nomoto (2013, 12ff) presents Malay as a representative of such languages.41 The Malay classifier buah, he says, is optional in (6): (6) dua (buah) 2 (CL) ‘two books’

buku book

[Malay; Nomoto 2013, 12]

Moreover, many languages that Gil (2008) considers mandatory classifier languages42 have optional classifiers. He presents Vietnamese as a representative of such languages (ibid., 4), but Vietnamese also has optional classifiers, as Emeneau (1951, 95) points out. Thus, Simpson and Ngo divide Vietnamese classified nouns into two kinds: (a) “obligatory-classifier nouns” and (b) “optional-classifier nouns” (2018, 213). The latter, (b), include the Vietnamese counterparts of ‘house’ and ‘corner of mouth’ (Emeneau 1951, 95) and of ‘room’, ‘village’, ‘whale’, and ‘propeller’ (Simpson and Ngo 2018, 214). Korean, Japanese, and Mandarin Chinese also have nouns with optional classifiers, as is argued below.    of classifier systems 3.3  Diversity There are two ways, we have seen, that classifier systems and even substantial ones might not be mandatory. First, they might not be fully comprehensive and have nonclassified nouns. Second, they might not have strong control over all classified nouns and assign some of them optional classifiers. Clearly, the two factors (strength and comprehensiveness) do not have binary values but can be realized in different degrees. At one end of the scale would be the group of mandatory classifier languages. But not all classifier languages belong to this group. Many of them have nonclassified nouns or optional classifiers. Vietnamese has both, as we have seen. And I think most languages commonly considered mandatory classifier languages would turn out to fall short on close examination. To see this, it would be useful to give a close examination of classifier systems of Mandarin (Chinese), Japanese, and Korean. Gil (2008) takes all these to be mandatory. Simpson and Ngo (2018, 212f) suggest the same in contrasting these languages with Vietnamese. But we can see that they also have directly numerable nouns and optional classifiers.

Byeong-uk Yi

16 3.3.1

The Korean classifier system43

Korean has an extensive system of classifiers. The language has more than four dozen classifiers.44 All of them except one are specific classifiers, namely, classifiers matching a limited range of nouns (e.g., mali for nouns for animals). The one exception, kay, is a general classifier, a classifier that can be used with a wide variety of nouns, including many with no matching specific classifiers. This considerably increases the number and range of classified nouns. Still, the language has directly numerable nouns. Most of them have optional classifiers, but some are nonclassified nouns.45 Most Korean grammars recognize four standard forms of numeral noun phrases. Using ‘N’, ‘Num’, and ‘CL*’ for nouns, numerals, and numeratives, respectively, we can formulate the forms as follows:46 Table 1.1 Standard forms of Korean numeral noun phrases

Post-nominal Pre-nominal

With numerative

Without numerative

KN1. N – [Num – CL*] KN2. [Num – CL*] – -uy – N47

KN3. N – Num KN4. Num – N

KN1 and KN2 are forms involving classifiers or other numeratives. Instances of KN1 include (2c) and (3c), which have mali (classifier) and khilo ‘kilogram’, respectively; instances of KN2 include their pre-nominal cousins: (2cʹ) and (3cʹ). In addition to these forms, Korean has two numerative-free forms: KN3 and KN4. Among the four forms, KN1 is the most widely used in modern Korean. It is used with no special restrictions (Lee and Ramsey 2000, 98). By contrast, there are substantial constraints on the nouns and numerals that can yield instances of the numerative-free forms. Still, their instances are not unusual exceptions. KN3 (the numerative-free cousin of KN1) is regularly used and is a “natural” construction with nouns for humans: salam ‘human being’, haksayng ‘student’, and so on (ibid., 98). KN4 “was used extensively in Middle Korean, but now its use has diminished to include only native numerals with a limited number of nouns” (ibid., 98).48 Still, it is used quite often in modern Korean. Now, some nouns can participate in all four forms. For example, haksayng ‘student’ occurs in (7a)–(7d): (7) a.

haksayng twu student 2 ‘two students’ b. twu myeng-uy 2 CL-GEN ‘two students’ c. haksayng twul student 2 ‘two students’

myeng CL haksayng student

Diversity of classifier systems 17 d. twu haksayng 2 student ‘two students’49 The noun can take the classifier myeng for human nouns, as in (7a)–(7b), but can directly combine with some numerals, as in (7c)–(7d).50 (So the classifier is optional for the noun.) And a variety of nouns can occur in instances of KN3 or KN4: (8) a. b.

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

{tokkaypi, kwysin} twul {(Korean) hobgoblin, ghost} 2 ‘two {(Korean) hobgoblins, ghosts}’ pyel seys star 3 ‘three stars’51

(Chae 1996, 198)

twu {nwun, son, pal, tali} 2 {eye, hand, foot, leg} ‘two {eyes, hands, feet, legs}’ sey kacok 3 family ‘three families’ twu {mal, sori} 2 {remark/statement, saying/statement} ‘two {statements, statements}’ sey {kak, pyen} 3 {angle, side} ‘three {angles, sides}’ ilkop sayk.kkal 7 color ‘seven colors’ sam {hyengcey, nam.may} 3 {brothers, siblings} ‘3 {brothers, siblings}’ sam wen sayk 3 primary color ‘three primary colors’52

Most of the nouns in the above-mentioned instances of KN3–4 have matching classifiers, but some of them are nonclassified. For example, tokkaypi ‘(Korean) hobgoblin’ and kwysin ‘ghost’ have no matching classifiers (see (8a)). They can take no specific classifier, for Korean classifiers are for humans (e.g., myeng), animals (e.g., mali), or inanimate or abstract entities (e.g., tay for some machines, kaci for some abstract entities). Nor can they take the general classifier kay, for this is used primarily for concrete individuals with stable shapes and cannot match nouns for animates.53 Similarly, I think, Korean has no classifier matching kacok ‘family’.54

18

Byeong-uk Yi

Despite having a rich and quite comprehensive classifier system, we have seen, Korean has a variety of directly numerable nouns. This does not mean that they can directly combine with any numerals. Instances of KN3 must involve native Korean numerals. Most instances of KN4 have native Korean numerals, although some of them (e.g., (9f)–(9g)) have Sino-Korean numerals (e.g., sam ‘three’). And instances of both forms usually involve numerals for small numbers (e.g., those below 10).55 Their grammaticality or naturalness can decrease (and sometimes rapidly) as the numerals get higher.56 Let me conclude discussion of the Korean classifier system by noting two special numerative-free constructions. First, Korean has a subtype of KN4 that involves the Sino-Korean adjective tay ‘big, great’: KN4ʹ. NumSK – tay – N, where NumSK is for a Sino-Korean numeral.57 Instances of this form include (10a)–(10c): (10) a. chil tay pulkasa.uy 7 great wonder/miracle ‘(The) Seven Wonders’58 b. sa tay kise 4 great outstanding.book ‘Four Great Masterpieces’ c. sam tay tosi 3 great city ‘(the) 3 great/greatest cities’ These phrases involve no classifier while the nouns have matching classifiers. And one can get grammatical phrases by replacing the nouns in the phrases (e.g., tosi ‘city’ in (10c)) with most classified or nonclassified Sino-Korean nouns (e.g., phokpho ‘waterfall’). Second, in most compound words involving numeral noun phrases, the phrases have no classifiers. Consider the Sino-Korean translations of ‘triangle’, ‘quadrilateral’, and ‘triumvirate’: (11) a. b. c.

sam.kak.hyeng 3.angle.figure ‘triangle’ sa.pyen.hyeng 4.side.figure ‘quadrilateral’ sam.twu.cengchi 3.head.government ‘triumvirate’

These words contain instances of KN4: sam.kak ‘3 angles’ [3 angle], sa.pyen ‘4 sides’ [4 side], and sam.twu ‘3 heads’ [3 head]. Similarly, most numeral noun

Diversity of classifier systems 19 phrases in compound words are instances of the numerative-free KN4, not of KN2.59 Both kinds of constructions discussed above are Sino-Korean. We can see that both Chinese and Japanese have their cousins (§§3.3.2–3). 3.3.2

The Japanese classifier system60

Japanese has a rich classifier system that includes two general classifiers: the native tsu and Sino-Japanese ko. The system is often more demanding than the Korean system. In Japanese, one would add classifiers in a variety of cases where one might use numerative-free constructions in Korean. Still, Japanese also falls far short of being a mandatory classifier language.61 Consider the Japanese for triangles: (11) aʹ. san.kaku.kei 3.angle.figure ‘triangle’ Like its Korean cousin, (11aʹ) contains a phrase (san kaku) where a noun (kaku ‘angle’) and a numeral (san ‘three’) directly combine. Similarly, most (if not all) Japanese numeral noun phrases in compound words (like their Korean counterparts) have no classifiers. And Japanese has a cousin of KN4ʹ: JN4ʹ. NumSJ – dai – N, where NumSJ is for a Sino-Japanese numeral (dai ‘big, great’ is a Sino-Japanese adjective). Instances of JN4ʹ include the Japanese counterpart of (10c): (10) cʹ. san dai toshi 3 great city ‘(the) three greatest cities’62 Moreover, Japanese has usual numeral noun phrases where numerals and nouns (or noun phrases) directly combine. Consider (12)–(14): (12) a. b.

(13) a. b.

ni genri 2 principle ‘(the) two principles’ ichi bunya 1 area ‘one area’

(Downing 1996, 73) (Ibid., 73)

mizu-no izumi juuni (Shimojo 1997, 131) water-GEN spring 12 ‘twelve (water) springs’ nijuu-shi-no hitomi 24 GEN pupil (of eye) ‘twenty four (eye) pupils’ (roughly, ‘twenty four eyes’)

20

Byeong-uk Yi

(14) a. b. c. d.

futa kazoku 2 family ‘two families’ nana fushigi seven wonder/miracle ‘(The) Seven Wonders’ too -no hanashi 10 GEN story ‘ten stories’ too -no {hasi, oterasan, depaato} 10 GEN {bridge, temple, department.store} ‘ten {bridges, temples, department stores}’

These phrases can be considered instances of Japanese cousins of KN3 and KN4: JN3. (a) Num N (b) Num -no N (where -no is a genitive construction marker) JN4. N Num (13b) is an instance of JN4 and all the others are instances of JN3(a) or JN3(b). To discuss these numerative-free constructions, it would be useful to note some features of the Japanese classifier system related to numerative-free constructions. As Downing (1996, 72–75) points out, the system has limitations in covering abstract nouns (e.g., genri ‘principle’). It has only a few special classifiers for abstract nouns (e.g., ken for incidents). The paucity of special classifiers for abstract nouns is compensated by frequent use of the general classifier tsu with abstract nouns.63 Still, some abstract nouns can directly combine with numerals, as in (12a)–(12b) (which have Sino-Chinese numerals).64 Some concrete nouns are also directly numerable, as (13)–(14) show.65 Shimojo (1997) relates some such cases to the mismatch between the native classifier tsu and Sino-Japanese numerals:66 . . . the native-Japanese numeral system covers only from “one” to “ten” in present-day Japanese. It is common in present-day Japanese that nativeJapanese classifiers appear with Sino-Japanese numerals . . . however, tsu is among the few exceptions, consistently appearing with the native-Japanese numerals. . . . One strategy to express a large quantity of a tsu-like entity is simply to dispense with the classifier, and this is still fairly frequent in presentday written Japanese. (Shimojo 1997, 728; my italics) He calls Japanese numerative-free numeral noun phrases “absent classifier” constructions (ibid., 712)67 and gives (13a) as an example of such a construction (ibid., 737).68 (He also quotes a passage from the English Bible with several numerativefree constructions (e.g., ‘twelve young bulls’) to note that all three Japanese translations he examined69 use no classifier in translating the passage (ibid., 714).)70 One might give the same explanation for the absence of classifier in (13b).71

Diversity of classifier systems 21 While the phrases in (12)–(13) involve Sino-Japanese numerals, native Japanese numerals can also participate in numerative-free constructions. (14a)–(14b) are instances of JN3(a) involving native numerals: futa ‘two’ and nana ‘seven’. (14c)–(14d) are instances of JN3(b) involving the native numeral too for ten.72 And too is never supplemented with a classifier (Nomoto 2013, 87), although the other native numerals are usually supplemented with classifiers (see, e.g., (14aʹ) below). (14b) is a definite noun phrase for the Seven Wonders of the World.73 Similarly, a variety of instances of JN3(a) can be used as definite noun phrases. Consider (15a)–(15d): (15) a. b. c. d.

san juushi 3 musketeer ‘(the) three musketeers’ (kakushi toride -no) san akunin hidden fortress GEN 3 villain ‘the three villains (of the hidden fortress)’74 Tokyo roku daigaku Tokyo six university ‘Tokyo’s six [great] universities’ Nakamura san kyodai Nakamura 3 brother ‘(the) three Nakamua brothers’

(15a) is an instance of JN3(a), and (15b)–(15d) are phrases where instances of the same form are modified by noun adjuncts or adjective phrases.75 It would require further study to clarify the conditions under which instances of JN3(a) are well-formed.76 But there seems to be no denying that Japanese has a variety of such instances, which helps to show that it is not at all a mandatory classifier language. And the numerative-free phrases discussed above help to show that Japanese has optional classifiers. Most of the nouns in the phrases have matching classifiers, and one can add the classifiers to some of the phrases. For example, the classifier nin matches the human nouns in (15b) and (15d) and the general classifier tsu the nouns hanashi ‘story’ and kazoku ‘family’. And one can add tsu to (14a) and nin to (15b): (14) aʹ. futa tsu -no kazoku 2 CL GEN family ‘two families’ (15) bʹ. (kakushi toride -no) san nin -no hidden fortress GEN 3 CL GEN ‘the three villains (of the hidden fortress)’

akunin villain

This shows that tzu and nin are optional for kazoku ‘family’ and akunin ‘villain’, respectively.

Byeong-uk Yi

22 3.3.3

The Chinese classifier system

Dialects of modern Chinese (e.g., Mandarin, Cantonese) have classifier systems stronger than the Korean and Japanese systems. They are usually considered mandatory. In his Mandarin grammar, Chao asserts that “a numeral cannot directly modify a noun except in wenyan” (1968, 585).77 Similarly, Cheng and Sybesma hold that “In Chinese . . . in order for a noun to be countable, a measure word or classifier is always necessary” (1998, 386). As Her (2012) argues, however, some nouns are directly numerable even in modern Chinese. Some might use some well-known examples to argue that Mandarin and probably all classifier languages fail to have mandatory classifier systems. Consider (16a)–(16b): (16) a. b.

ba shi ren (= (5a)) 8 ten person ‘eighty people’ san (*ge) tian 3 CL day ‘three days’

[Mandarin; Chao 1968, 595] [Mandarin; Li and Thompson 1981, 105]

In these phrases, the nouns (ren ‘person’ and tian ‘day’) seem to combine directly with numerals. But Chao takes shi ‘ten’ to occur as a numerative in (16a), as noted above (§2.3), and tian ‘day’ to belong to a special group of numeratives, which he calls “quasi-measures” (1968, 608ff).78 He justifies these analyses by invoking the mandatory classifier language thesis (T1) for Mandarin. Some might challenge this thesis and take (16a)–(16b) as exceptions to a general rule that holds usually but not without exceptions. Li and Thompson (1981, §4.2.1) might accept this view. They do not include numerals for powers of ten among group numeratives, which they call “aggregate measures” (ibid., 107f). And they take note of (16b) and its ilk to hold: If a noun itself denotes a measure, it does not take a classifier; examples include: kuai ‘piece’, li, ‘mile’, jin, ‘tael’, liang ‘ounce’, chi ‘foot’, and tian ‘day’. (1981, 105)79 Whether (16a)–(16b) can be considered instances of the numerative-free form Num – N, we have seen, depends on what numeratives are taken to include. I think Chao’s analyses of the phrases are viable, if not compelling, and as plausible as alternative analyses that conflict with the mandatory classifier language thesis. So I do not think Li and Thompson (1981) give a convincing argument against the thesis. But we can see that the thesis has other serious problems. Chao’s list of the so-called quasi-measures (ibid., 608–620) reveals some of the problems. The list includes 100 expressions (Mq1–Mq100). About half of them (roughly the second half) are Mandarin counterparts of English nouns relating to measurement: ‘day’, ‘month’, ‘dollar’, ‘degree’, and so on. The others (roughly the

Diversity of classifier systems 23 first half) include: guo ‘country’, mian ‘side, direction, page’, fangmian ‘respect, aspect’, dian ‘(theoretical) point’, yuan ‘main (government) branch, college (of a university)’, piao ‘vote’, ce ‘volume, pamphlet’, ci ‘word’, yan ‘word’, cisu ‘morpheme’, and so on. Chao includes them among numeratives because they can “follow numerals . . . directly” (ibid., 608), as yuan does in (17): (17) liang yuan two chambers (of legislature) ‘two chambers (of legislature)’ But they can also take matching classifiers (e.g., the general ge). If so, it is hard to see why one cannot simply classify them as nouns to conclude that Mandarin has directly numerable nouns. Moreover, the range of such nouns is much wider than Chao’s list of quasi-measures might suggest. In his recent article that distinguishes classifiers from measure words to give a semantic account of classifiers, Her (2012) denies that Mandarin is a mandatory classifier language. Although classifiers are “normally required” in the language, he argues that they (unlike measure words) are “semantically redundant” (ibid., 1679).80 In doing so, he holds that Mandarin classifiers “can be omitted in some contexts” (ibid., 1680). He gives two kinds of examples to show this: (18) a. b.

(19) a. b.

si (ge) ren 4 (CL) person ‘four persons’ wu (zhang) bing er 5 (CL) loaf two ‘5 loaves and 2 fish.’81

[Mandarin; Her 2012, 1676] (tiao) yu CL fish

liang ren san jiao de bisai 2 person 3 foot GEN race ‘two-person-three-foot race’82 yi ren yi xin yundong 1 person 1 letter campaign ‘one-man-one-letter campaign’

[Mandarin; ibid., 1669]

[Mandarin; ibid., 1679] [Mandarin; ibid., 1679]

And he points out that “some pedagogical grammarians” explain that classifiers differ from measure words in that they “can be omitted in some contexts” (ibid., 1680):83 because classifiers do not express quantity they can be omitted, “1 C cup” = “1 cup” [“yi ge beizi” = “yi beizi”]. (Ma 2011, 37; my boldface) Classifiers: 1 C bed (1 bed), 1 C ox (1 ox), 1 C person (1 person) [yi zhang chuang . . . yi tou niu . . . yi ge ren . . .], can be omitted without any change in meaning. (Wang 2004, 113; my boldface)84

24

Byeong-uk Yi

(19a)–(19b) are compound words that contain instances of the Num – N form. Such words include the Chinese counterparts of (11a)–(11c): (20) a. b. c.

san jiao xing 3 angle shape/figure ‘triangle’ si bian xing 4 side shape/figure ‘quadrilateral’ san tou tongmeng 3 head alliance ‘triumvirate’

(20a)–(20c) contain instances of the numerative-free form: san jiao ‘3 angles’, si bian ‘4 sides’, and san tou ‘3 heads’. And (19a) contains two instances of the same form: liang ren ‘two people’ and san jiao ‘three feet/legs’. Similarly, (19b) contains yi ren ‘one person’ and yi xin ‘one letter’. In Mandarin, as in Korean and Japanese, numeral noun phrases contained in compound words do not involve classifiers, as we have seen. In the usual numeral noun phrases, however, Her says that classifiers are “normally required” (2012, 1679), as noted above. While pointing out that they can be omitted, as in (18a)–(18b), he suggests that it is quite unusual to omit them. He says that classifiers are “stylistically allowed to be absent in certain formal literary contexts” (ibid., 1679; my italics) and that they “can be omitted if stylistically required” (ibid., 1669; my italics). But he observes that this constraint does not hold in all branches of Mandarin. He gives instances of Num – N used in the Beijing dialect of Mandarin: (21) a. b. c.

san gongsi 3 company ‘three companies’ liang qiguan 2 organ ‘two organs’ yi zhenghun guanggao 1 marriage-seeking ad ‘an ad seeking marriage’85

[Beijing Mandarin; ibid., 1680] [Beijing Mandarin; ibid., 1680] [Beijing Mandarin; ibid., 1680]

And he concludes that “In Beijing dialects, classifiers are more freely omitted than in Taiwan Mandarin” (ibid., 1680).86 Mandarin (including both Taiwan and Beijing dialects), we have seen, falls short of having a mandatory classifier system. Moreover, numerative-free constructions might not be as unusual as Her seems to suggest, even in Taiwan Mandarin and other dialects of Mandarin. It is normal, if not mandatory, as noted above, for numeral noun phrases in Mandarin compound words not to have classifiers.87 The same holds, I think,

Diversity of classifier systems 25 for instances of the Mandarin cousin of the Sino-Korean form KN4ʹ, such as (22a)–(22c): (22) a.

qi da qiji 7 great wonder/miracle ‘(The) Seven Wonders’ b. si da {qishu, gudu} 4 great {outstanding.book, ancient.capital} ‘(The) Four Great {Masterpieces, Ancient Capitals}’ c. wu da hu 5 great lake ‘(The) Five Great Lakes’

These phrases involve no classifiers, and this is normal for such phrases. Moreover, Mandarin seems to have a substantial range of directly numerable nouns. Consider, for example, (23a)–(23c): (23) a. b. c. d. e.

wu da.yang 5 ocean ‘(the) five oceans’ {wu, liu} da.zhou {5, 6} continent ‘(the) {five, six} continents’ qui hai 7 sea ‘(the) seven seas’ liang dang 2 party ‘two parties’ san dian 3 point (in space or time) ‘three points [of, e.g., a triangle]’

All these phrases are instances of Num – N.88 Chao might deny this by taking them to involve the so-called quasi-measures (i.e., numerative cousins of nouns). His list of quasi-measures includes dian ‘(theoretical) point’ (which one might use to translate ‘His answer had two points’). He might take the same word as used for spatial points to occur as a quasi-measure in (23e). Similarly, he might take (23d) to involve the quasi-measure use of dang ‘party’, as (16) involves the same use of yuan ‘chamber’. But I do not think one can plausibly hold that (23a)–(23c), too, have only numerative cousins of nouns. Moreover, holding this would not help to maintain the mandatory classifier thesis. The thesis, we have seen, fails to hold for Mandarin in any case. Once we see this, we have no further reason to inflate the list of numeratives to analyze the likes of (23a)–(23e). And we can conclude that Mandarin also has a substantial range of directly numerable nouns.89

26

4

Byeong-uk Yi

Classifiers and plural marking

A variety of classifier languages (including Chinese, Japanese, and Korean), we have seen, have directly numerable nouns and optional classifiers. But most accounts of classifiers in theoretical linguistics deny this. They hold that classifier languages cannot have directly numerable nouns (the mandatory classifier language thesis) and that no nouns with matching classifiers are directly numerable (the obligatory classifier thesis). These theses are closely related to the assumption that classifiers conflict with plural markers. To explain this assumption, usual accounts of classifiers take classifier language nouns or nouns with matching classifiers to be akin to mass nouns of languages with the singular/plural morphology (e.g., English). This view implies the mandatory classifier language thesis (T1) or the obligatory classifier thesis (T2). So we must reject the usual accounts of classifiers. Moreover, we can see that the assumption they aim to explain is false: classifiers and plural marking are not mutually exclusive. The assumption dates back to an influential study of classifier and plural marking: Sanches and Slobin (1973). They propose “the hypothesis that plural marking and numeral classifiers do not co-occur” (ibid., 1) and state the thesis as follows: . . . when in a quantification expression, there is an element which encodes semantic information in agreement with the item being enumerated, there will be no element in conjunction with the item being enumerated which marks number. Specifically, we will not observe syntactic structures of the type: */ Noun + plural/ + /Number + Numeral Classifier/. (1973, 6)90 We can formulate this thesis as follows: T3. The Exclusion Thesis: A numeral noun phrase featuring a noun cannot have both a matching classifier and a plural marker for the noun.91 Sanches and Slobin hold that all or most of the 60–70 languages they examine clearly conform to T3 (ibid., 1). And they argue for the thesis by comparing classifiers with the English ‘head’ in, for example, ‘seven head of cattle’: One might naively test this [i.e., the exclusion thesis] on oneself as a nativespeaker of English in the following way: although plural marking of nouns is the dominant form of quantification expression in ENGLISH, e.g. ‘seven cows’, there are some forms in ENGLISH equivalent to numeral classifiers in other languages and when we use these we do not use plurals. For example, ‘seven heads [sic, head] of cattle’. It seems to violate our intuitions as native-speakers of the language to say something like: ‘*seven heads of cows’, ‘*seven cattles’, or any of the other combinations of plural and classifiers possible. (Ibid., 6f)

Diversity of classifier systems 27 And they give an account of classifiers that implies T3: . . . if we consider again the concept of plurality versus classifiers as expressed in ENGLISH, it is obvious that those nouns for which we have classifier forms, usually called ‘mass’ as opposed to ‘count’ nouns, which we pluralize, express unindividuated quantity. These mass-nouns, when individuated, demand mention of semantic features of the individual member of the class, for example . . . ‘a grain of sand’ . . . ‘a drop of water’ . . . ‘a stick of candy’. (Ibid., 13; original italics) This account assumes that classified nouns are mass nouns and takes classifiers to have the same function as ‘drop’ in, for example, ‘three drops of water’ and ‘three water drops.’ Thus the account yields an explanation of T3 by accepting T4 and T5: T4. Mass Noun Thesis: a. All classified nouns are mass (or para-mass) nouns. b. No mass (or para-mass) nouns are directly numerable.92 T5. a. Nouns with plural markers are count nouns. b. Count nouns are directly numerable. These theses imply T3. But one cannot take this to yield a suitable explanation of T3. T4(a)–(b) imply T2, which we have seen is false.93 Moreover, T3 is also false. The English ‘head’ can co-occur with ‘-s’, and classifiers can co-occur with plural markers in some classifier languages. Consider (24a)–(24e): (24) a. b. c. d. e.

ten head of Jerseys94 600 head of longhorns 30 head of 1200 lb mother cows three head of cows twenty head of black men

(Judd 1886, 428) (Morton U.d.) (Schell 2012) (Schell 2012) (Olmsted 1904, 245)

In these phrases, ‘head’ matches the plural forms of ‘Jersey’, ‘longhorn’, ‘cow’, and ‘black man’. As Sanches and Slobin (1973, 6f) assume, ‘*three head of cattles’ is ill-formed. But the reason is that ‘*cattles’ is ill-formed,95 not that ‘head’ conflicts with ‘-s’. They assume that ‘seven head of cows’ is also ill-formed but it is not, as (24c)–(24d) show.96 We can now see that classifiers can co-occur with plural markers in some classifier languages. A variety of such languages have been discussed since Sanches and Slobin (1973), who discuss “a few exceptions” to T3 while proposing the thesis (ibid., 10).97 For reasons of space, let me focus on a well-documented Mayan language: Itzaj Maya. Itzaj Maya is an endangered Mayan language spoken by the Itzaj people in Guatemala. The language is studied and documented in Hofling (1991; 2000) and

28 Byeong-uk Yi Hofling and Tesucún (1997). It has “Mayan numerals . . . regularly used from ‘one’ to about ‘five’, and Spanish numerals . . . for higher numbers”, according to Hofling (2000, 141). While “Spanish numerals directly precede the nouns they modify”, Mayan numerals “must be followed by a numeral classifier [viz., numerative]” to yield numeral noun phrases of the form [Num – CL*] – N, where ‘CL*’ is for numeratives (ibid., 141). Some of the numeratives are classifiers: the general classifier pʹeel, tuul for animates, teek for plants, several shape-based classifiers (e.g., kuul for round objects), and so on (ibid., 142).98 The language also has a plural marker: -ooʹ (ibid, 179). And it can co-occur with classifiers. In numeral noun phrases involving classifiers, Hofling says, “Plurals may be additionally marked by nominal suffixes” (e.g., -ooʹ) and “this is more likely to happen when the NP refers to specific, definite information” (ibid., 219). He shows this with some definite noun phrases, such as (25a)–(25c): (25) a. b. c.

aʹ ʹox tuul winik-ooʹ [Itzaj Maya; Hofling 2000, 222] DET 3 CL man-PL ‘the three men’ aʹ ox tuul mejen paal-ooʹ [Itzaj Maya; ibid., 225] DET 3 CL small child-PL ‘the three small children’ aʹ kaʹ kuul naj-ooʹ [Itzaj Maya; ibid., 222] DET 2 CL house-PL ‘the two houses’99

Finally, Korean also has constructions violating the exclusion thesis. Like most classifier languages, Korean has no grammatical number system. Still the language has a plural marker, -tul, “a suffix . . . used to mark plurals” (Lee and Ramsey 2000, 88). Although it is not an obligatory plural marker (ibid., 100),100 this does not mean that it is always optional. Consider (26a)–(26b): (26) a. b.

ce chayk that/those book ‘those books’ ce chayk that/those book ‘that book’

-tul PL

[Korean] [Korean]

These do not have the same meaning. While (26a) amounts to the plural ‘those books’, (26b) amounts to the singular ‘that book’. The reason for this, I think, is the singular preference of Korean demonstrative pronouns: although ce ‘that/those’ might take the plural reading, the pronoun selects the singular reading unless there is an overriding element (e.g., -tul). So the plural marker -tul is more likely to be used in definite noun phrases.101 Now, consider (27a)–(27e): (27) a.

ce haksayng that/those student ‘those three students’

-tul PL

sey myeng 3 CL

[Korean]

Diversity of classifier systems 29 b. c. d. e.

ce i.sip.sam myeng -uy haksayng -tul that/those 23 CL GEN student PL ‘those 23 students’ ce yelsey myeng -uy elin haksayng -tul that/those 13 CL GEN young student PL ‘those thirteen young students’ elin haksayng -tul sey myeng young student PL three CL ‘three young students’ yelsey myeng -uy elin haksayng -tul 13 CL GEN young student PL ‘(the) thirteen young students’

[Korean] [Korean] [Korean] [Korean]

In all these phrases, haksayng ‘student’ takes both the classifier myeng and plural marker -tul.

Acknowledgments My work for this chapter was supported in part by a SSHRC Insight Grant [Grant No. 435-2014-0592] and a Chancellor Jackman Research Fellowship, which is hereby gratefully acknowledged. I would like to thank the many scholars who kindly helped me on various Chinese, Japanese, and Korean examples and especially Yasmine El Gebali, Eamon Darnell, Sarah Hamilton, Lu-Vada Dunford, One-Soon Her, Young-Wha Kim, and Takashi Iida for comments on ancestors of this chapter. Needless to say, I am solely responsible for any errors or infelicities.

Notes 1 In this chapter, I do not discuss proper nouns and use ‘noun’ as short for ‘common noun’. By ‘numeral noun phrases’, I mean roughly noun phrases that feature numerals and usual common nouns: ‘three cows’, ‘three liters of milk’, and so on (see, e.g., (1a)–(1c) and (4a)–(4f) in §2). 2 See, for example, the Mandarin (Chinese), Japanese, and Korean translations of ‘three cows’ given in §2.1: (2a)–(2c). Languages with classifiers belong to various language families (see §3). 3 Allan (1977) regards numeral classifiers as devices for classifying nouns, “classifiers for nouns” (ibid., 285), and distinguishes them from other kinds of devices for classifying nouns or predicates: concordial, predicate, and intra-locative classifiers. In this chapter, I do not discuss classifiers of other kinds and use ‘classifier’ as short for ‘numeral classifier’. (But I think this terminology is misleading. I do not think classification is their essential function.) See, for example, Craig (1994), Croft (1994), and Aikhenvald (2000) for other studies of numeral classifiers as noun classification devices. 4 By ‘Chinese’, ‘Japanese’, and ‘Korean’ in this chapter, I mean Modern Chinese, Japanese, and Korean, respectively, unless noted otherwise. Old Chinese is not a classifier language, and it is plausible to take the same to hold for ancient Japanese and Korean. But their modern descendants have highly developed classifier systems and are often considered mandatory classifier languages. 5 A particle or morpheme is said to be a plural marker for a noun if the particle or morpheme (like the English -s) indicates that the noun combined with it can denote more than one of the entities the noun can individually denote (e.g., three of all the cows). 6 See also Tʹsou (1976, 1216) and Borer (2005, Chapter 4), who attributes T3 to Tʹsou (1976).

30

Byeong-uk Yi

7 We can take proponents of the mass noun thesis (T4), stated below, to give different specifications of the relevant respects, yielding different versions of the thesis. 8 See note 92 (§4) for a list of proponents of versions of T4. 9 This chapter does not aim to elaborate on or defend the paranumeral account. For more on this account, see, for example, Yi (2009; 2011a; 2011b; unpublished). For similar accounts, see, for example, Greenberg (1972, 172f), Au Yeung (2005; 2007), Her (2012), and Her and Lai (2012). 10 For example, Greenberg (1972) uses ‘classifier’ in the narrow sense and Allan (1977) in the broad sense – Allan says “everyone agrees” that the English ‘pack’ and ‘dozen’, for example, “are classifiers” (ibid., 286). Some authors (e.g., Cheng and Sybesma (1999; 2005), Li and Thompson (1981, 104–112), and Gil (2008)) use the term in both senses in the same work. See also note 20. 11 Allan (who uses ‘classifier’ broadly) says “Perhaps all languages have classifiers” (1977, 285), and Croft says that if measure words are included among classifiers, “all languages are numeral classifier languages” (1994, 151ff). But not all languages have classifiers in the broad sense. Some native American languages have neither classifiers nor measure words (Greenberg 1972, 176), and some languages do not even have numerals. 12 Hundius and Kölver (1983) say that ‘(numeral) classifier’ is “now the most commonly used term” for classifiers in the narrow sense (ibid., 209). Edkins (1857, 119f; 1864, 127) introduces the term ‘numerative’ to cover all expressions belonging to the larger group (see §2.2). 13 Abbreviations used in glosses are as follows: CL: classifier; CL*: numerative; DET: determiner; LOC: locative marker; N: noun; NOM: nominative marker; Num: numeral; GEN: genitive (construction) marker; PL: plural marker; TOC: topic marker. 14 Incidentally, the italicized classifiers stem from nouns for heads. 15 In (3a), the postpositive linking particle -de (which amounts roughly to ‘of’) can be added to san bang. See note 18. 16 Korean has two series of numerals: the native Korean (e.g., sey(s) ‘three’) and the SinoKorean (e.g., sam ‘three’), which are loan words from Chinese (they match different ranges of classifiers). In (2c) and (3c), the native Korean sey(s) and the Sino-Korean sam are used to match mali and khilo, respectively. 17 Instead of the Japanese and Korean counterparts of (1b), I give the counterparts of (1c), for the Japanese niku san paundo ‘three pounds of meat’ [meat 3 pound] is obsolete (pound is no longer a customary unit for measuring the weight of meat in Japan). (In Korean, however, (traditional Korean) pound might still be used as a measurement unit and kun ‘(Korean) pound’ as a measure word, as in koki sey kun ‘three pounds of meat’ [meat 3 pound].) 18 The linking particles -no and -uy are postpositive markers for genitive constructions. They are similar to the English ‘of’, but there is an important difference: ‘of’ is a preposition, but -no and -uy are postpositive particles (or suffixes). So (3bʹ) and (3cʹ), listed below, do not have the same syntactic structure as the English pseudo-partitive construction given as their translation (‘three kilograms of meat’) and are closer in syntax to the attributive construction ‘three-kilogram meat’ (or ‘three kilograms worth of meat’). The same holds for their Chinese counterpart (3a), where the Chinese counterpart of -no and -uy (viz., the suffix -de) can be optionally added to san bang, as in (3aʹ): (3)

aʹ.

san bang -de rou [Mandarin] 3 pound GEN meat ‘three pounds of meat’

19 In an earlier work, Edkins (1853/1868) calls them “numeral particles”, “classifying particles”, or “auxiliary words” (1853, 88f; 1868, 81f). 20 Numeratives are usually called liangci (which is commonly translated as ‘measure word’ but literally means quantity word) in Chinese grammar and josuushi (which

Diversity of classifier systems 31

21 22 23 24 25 26 27

28 29 30 31 32 33 34 35 36 37 38 39 40

41 42 43

means adjuncts to numerals or numerary adjuncts) in Japanese grammar. For more on various terms for classifiers and numeratives, see Yi (2011a, §2). I think English (which is not a classifier language) has a couple of classifiers. See note 96 (§4). See, for example, Lyons (1977), Hundius and Kölver (1983, 168), Tai and Wang (1990), and Cheng and Sybesma (1998; 1999; 2005). So we can get Korean translations of, for example, ‘three pairs of cows’ by replacing the classifier mali in (2c) and (2cʹ) with the Korean counterpart of ‘pair’ (i.e., ssang). The same holds for Chinese and Japanese. They can match count nouns as well, as in ‘three {pounds, kilograms} of oranges’ (cf. (4f)). See, for example, Greenberg (1972, 171, no. 11) and Adams (1989, 9). Croft distinguishes numeratives similar to ‘group’ (e.g., ‘bunch’, ‘row’) from both classifiers and measure words and calls them “numeral pseudoclassifiers” (1994, 152). He calls them all “measure words” (or “measures”), but what he calls measure words include classifiers and other numeratives. And he says that only “standard measures” are measure words in the proper sense (ibid., 604). See note 20. The other four groups are for either numeratives matching verbs or nouns that are reclassified as numeratives because they are used in special ways to combine with numerals. Numeratives belonging to one of these peripheral groups (the so-called quasi-measures) are discussed in §3.3.3. For more on Chao’s account of classifiers and numeratives (or what he calls measure words), see Yi (2009; 2011a, 210ff). Edkins includes them among collective numeratives, (B1). I introduce the notion of paranumeral numerative in Yi (2011a; 2011b, §2). See also Yi (2009, §3). Numeratives belonging to (B1) and (B2) can be called collective and kind/manner numeratives, respectively. The Burmese, Chinese, Japanese, and Korean for dozen are loan words from English. Martin gives the same analysis of Korean numerals for powers of ten (1992, 171). See note 55 (§3.3). Burling includes them among what he calls classifiers. Unlike Greenberg, however, he uses ‘classifier’ in the broad sense for numeratives in general. On the paranumeral account, classifiers have the same semantic function as numerals for 1. Note that Simons and Fennig (2018) list about 7000 known living languages. They say that a noun is countable if it combines (directly or indirectly) with a numeral. And Greenberg suggests that no classifier languages are genuinely mandatory: if one incorporates T1 in defining classifier language, he says, “it is not excessive to say there are no numeral classifier languages” (1972, 168). These definitions are refinements of Emeneau’s. Because he uses ‘classifier’ for numeratives (ibid., 95), his definitions include among the so-called classified nouns mass nouns (i.e., nouns for “material things that are not discrete”), which have no matching classifiers. See also Nguyen (1957, §3, 131–132; 1997, 94), Simpson and Ngo (2018, 213–215), Bisang (1999), and Löbel (2000). Incidentally, Simpson and Ngo rightly take Vietnamese nonclassified nouns to be “count nouns” (ibid., 213). Allan uses ‘(numeral) classifier’ for all numeratives. And say that a noun has obligatory classifiers (or classifiers are obligatory for the noun), if the noun has a matching classifier and is not directly numerable. (Note that nouns with obligatory classifiers can take numeratives other than classifiers in numeral noun phrases.) Malay is not considered in Gil (2008). Gil defines mandatory classifier languages as languages where “a numeral cannot quantify a noun without . . . a classifier being present” (ibid., 4). In applying this criterion, however, he adds some qualifications (2008, 4f). Many thanks to Young-Wha Kim, Kwang-sup Kim, Chunghyoung Lee, Chungmin Lee, Jinsook Kim, and Jinho Kang for discussions about Korean examples. Although I have consulted them for judgments about Korean examples, I am solely responsible for claims about Korean made in this chapter.

32

Byeong-uk Yi

44 Martin (1992, 179–185) divides Korean numeratives into five groups and lists about 20 classifiers in the first group, which is meant to include classifiers (he includes about 35 numeratives in the group, but I think some of them are not classifiers). Unterbeck (1994) lists about 50 Korean classifiers (I think almost all in her list are classifiers). Note that this is comparable to the number of Mandarin classifiers. Chao’s extensive list of Mandarin numeratives (1968, 585–593) has 51 classifiers. Incidentally, according to Croft (1994, 152), only 9 of 400 Tzeltal numeratives listed in Berlin (1968, cf. 174f) are classifiers. 45 While considering Korean a mandatory classifier language, Gil (2008) cites no work that supports the classification. Nor do Simpson and Ngo (2018, 212f). In her study of Korean classifiers, Unterbeck says that classifiers “are obligatory elements in counting contexts, i.e., they are able to immediately precede (or follow) numerals” (1994, 367). But it is one thing to say that classifiers can immediately precede or follow numerals (as all Korean classifiers do) and quite another to say that they are obligatory. 46 See, for example, Chae (1982, 159), Martin (1992, 171ff), Sohn (1999, 352f), Lee and Ramsey (2000, 97ff), and Lee and Chae (2002, 143). See also Kang (1994) and Kim (2010). 47 KN2 involves the genitive construction marker -uy whether or not the numerative is a classifier. Some Korean numeral noun phrases (e.g., han song.i kwuk.hwa.kkoth ‘one chrysanthemum flower’ [1 CL chrysanthemum.flower], twu pun olapeni ‘(you) two brothers’ [2 CL brother]) have the variant of KN2 that results from omitting -uy. But they are somewhat unusual, and conditions for omitting the particle are not well understood. See Martin (1992, 172) and Chae (1996, 207). 48 But (9f)–(9g), below, are instances of KN4 involving Sino-Korean numerals. See Lee and Ramsey (2000, 97ff) for more on uses of and constraints on the forms. They point out that only (but not all) count nouns can occur in instances of KN3–KN4 (ibid., 98). 49 The native Korean numeral twu(l) has the adjectival and nominal forms: twu (adjectival) and twul (nominal). While (7a), (7b), and (7d) have the adjectival form, (7c) has the nominal form, which is comparable to the ‘two’ in ‘two of the students’. The nominal form has the same syntactic status in (7c) as the numeral-classifier compound twu myeng in (7a)–(7b). (The compound is a noun phrase. Korean numeratives (including classifiers) are considered nouns of a special kind, those called bound nouns.) 50 Most nouns that can participate in all four forms are human nouns, but some nonhuman nouns (e.g., pyel ‘star’) can also do so. See note 51. 51 The noun pyel ‘star’ participates in all four forms (KN1–KN4). It takes the general classifier kay, and twu pyel ‘two stars’ [star 2] is also well-formed. See also the examples involving kheph ‘cup’ in Park (2012, 392). 52 See also instances of KN4 given in Lee and Ramsey (2000, 98f). They include Korean counterparts of ‘{two, five, thirty} countries’, ‘three schools’, and ‘two rabbits’. 53 Chae gives tokkaypi and kwysin, among others, as examples of nonclassified nouns (1996, 198) and discusses restrictions on the Korean kay (ibid., 209ff) while contrasting it with the Chinese general classifier ge, from which it derives (ibid., 199ff). See also Park (2012). 54 I do not think kacok ‘family’ can take the general kay. But Google search yields sites with instances of KN2 amounting to ‘{two, three} families’ that involve kay. This might indicate dialectical, generational, or individual variation in the use of classifiers relating to gradual expansion of the scope of kay. 55 Exceptions are Sino-Korean instances of KN4 involving numerals for multiples of powers of 10, such as i chen.man tongpho ‘20 million compatriots’ [2 ten-million compatriot]. Martin applies the Chao-Burling analysis to take such numerals to involve a kind of paranumeral numeratives, those he calls “numeral counters” (1992, 171). See Yi (2011a, 226f). 56 For more on constraints on the forms, see Lee and Ramsey (2000, 99). 57 In instances of KN4ʹ, tay is commonly taken to be attached to the numerals. 58 See also the Japanese and Chinese counterparts of (10a): (14b) in §3.3.2 and (22a) in §3.3.3.

Diversity of classifier systems 33 59 The phrase for South Korea’s economic development plans in the sixties and seventies (o.kay.nyen kyeyhoyk ‘5-year plan’ [5.CL.year plan]) has a classifier (kay) for nyen ‘year’. Unlike its classifier-free cousin (o.nyen kyeyhoyk ‘5-year plan’ [5.year plan]), however, the compound sounds stilted and grandiose. 60 I learned about most of the Japanese examples discussed below from Jiro Abe, Takashi Iida, Youichi Matsusaka, Hiroki Nomoto, Katsuhiko Sano, Mitsuaki Shimojo, Masahiro Yamada, and Tomoyuki Yamada, mostly in private conversations. Thanks are due to them for the examples and discussions. I am solely responsible for errors in descriptions and explanations of the examples. 61 See, for example, Zubin and Shimojo (1993), Shimojo (1997), Kobuchi-Philip (2011, 298–303), and Nomoto (2013, 16–17) for discussions of Japanese numerative-free constructions. 62 For similar instances of JN4ʹ, see a site in Japanese Wikipedia: https://ja.wikipedia. org/wiki/%E4%B8%96%E7%95%8C%E4%B8%89%E5%A4%A7%E4%B8%80% E8%A6%A7. The site has a phrase that contains (10cʹ): sekai san dai toshi ‘the three greatest cities of the world’ [world 3 great city]. 63 Unlike its Chinese and Korean cognates (ge and kay), ko does not match abstract nouns. Downing says it is a “[g]eneral inanimate classifier for some speakers” (1996, 20). Shimojo says that in modern Japanese, ko “has been developing into a general classifier while retaining a semantic core” (1997, 705), which suggests its earlier use for “three-dimensional hand-sized manipulable objects” (ibid., 714f). 64 See Downing (1996, 73), who gives Japanese counterparts of ‘the two principles of individual liberty and fundamental equality of the two sexes’ and ‘Family problems are not treated as one area of society’s problems’ that include (12a) and (12b), respectively. See Kobuchi-Philip (2011, 300–301) for other examples involving abstract nouns. 65 Some might consider kazoku ‘family’ an abstract noun in a broad sense. 66 Like Korean, Japanese has two series of numerals: Sino-Japanese and native Japanese numerals. See, for example, Martin (2004, 767). 67 Instead of “absent classifier”, Zubin and Shimojo (1993) use “zero classifier”. Shimojo (1997) says that he avoids this term because the term suggests that Japanese has a mandatory classifier system (ibid., 731). 68 (13a) is a phrase in a 1955 Japanese translation of the Bible (Japan Bible Society 1955). 69 The translations are from different periods. They were published in 1887, 1955, and 1987. 70 Zubin and Shimojo give examples of Japanese constructions where “bare numerals are combined with nouns without an intervening classifier” (1993). But their examples differ from those in (12)–(14) in an important respect. Consider the italicized parts in (a): (a)

Sono hon-ni-wa hanashi-ga That book-LOC-TOP story-NOM ‘There are 12 stories in that book.’

jyuuni 12

aru. exist

(Ibid., example (6c))

In (a), the noun (which takes the case marker -ga) does not combine with jyuuni ‘twelve’, and the numeral (on a standard account) forms an adverbial construction (called the floating quantifier construction) comparable to ‘all’ in ‘They were all happy.’ Interestingly, the cousin of hanashi-ga jyuuni that has an instance of JN4 (i.e., *hanashi jyuuni-ga) is ungrammatical (Yamada and Iida, private conversations). This illustrates a delicate feature of the Japanese classifier system. (My discussion of their examples in Yi (2011a, 224f) mistakenly takes the cousin of (a) to be grammatical by failing to distinguish it from (a) itself.) Incidentally, some specific classifiers (e.g., wa, hen) can be added to (a) (Iida and Yamada, private conversations), while neither of the general classifiers (tsu and ko) can be added (Zubin and Shimojo 1993). See also Nomoto (2013, 17). 71 (13b) is the title of a famous novel by Sakae Tsuboi, published in 1952. Although hitomi is for pupils of eyes, the title is usually translated as ‘Twenty-four eyes’, as in Tsuboi (1963) (see also https://www.imdb.com/title/tt0047281/ for the film based on

34

Byeong-uk Yi

the novel that won a Golden Globe award in 1955). According to Y. Matsusaka (private conversation), hitomi in (13b) can be replaced by nouns for other body parts (e.g., me ‘eye’, te ‘hand’, ashi ‘foot, leg’). Examination of (13b) and related examples gives rise to interesting issues about the Japanese classifier system. 72 While the nouns in (14d) are concrete, hanashi ‘story’ is an abstract noun. The SinoJapanese numeral jyuu ‘ten’ can also combine directly with hanashi: Jyuu (pen) 10 CL ‘ten stories’

73 74 75

76

77 78

79 80 81 82 83

84

85

-no GEN

hanashi story

(Many thanks to Nomoto for the examples involving too and jyuu (private correspondence).) Its Korean counterpart, (10a), is an instance of KN4ʹ. (15b) is the title of a 1958 film directed by Akira Kurosawa. Consider also (chikurin -no) shichi kenjin ‘(the) Seven Sages (of the Bamboo Grove)’ [(bamboo.grove GEN) 7 sage] and dango san kyodai ‘(the) three dango brothers’ [dango 3 brother]. (The last example, which Sano noted, is the title of a song. See http://www.tokyowithkids.com/entertainment/dango.html.) See also Iida (2019, §1). Matsusaka (private conversation) points out that its instances (with Sino-Japanese numerals) are literary expressions, not common in spoken language. Yamada (private conversation) holds that they are “strongly definite” noun phrases used much like (plural) proper names for specific entities that are (supposedly) familiar. While Yamada and Abe hold that san gakusei ‘(the) three students’ [3 student], for example, is also grammatical (as used in a context satisfying a suitable definiteness condition), Nomoto considers the phrase ungrammatical. But he points out that ni gengo ‘(the) two languages’ [2 language] can be used both as a definite and as an indefinite phrase. Wenyan is the literary style of Chinese that retains the style of Classical Chinese, a written form of Old Chinese used in the literature from the Spring and Autumn period to the Han Dynasty. Like Classical Chinese, Wenyan is not a classifier language. Like (16a), (16b) illustrates a general phenomenon in classifier languages. As Greenberg says, classifier languages have “particular classes of nouns, e.g., measures, units of time, and the word ‘time’ in such phrases as ‘three times’ which hardly ever occur with classifiers” (1972, 168). But Tʹsou says: “In Modern Chinese, ‘month’, ‘week’, ‘hour’ and ‘second’ . . . take classifiers, whereas ‘year’, ‘day’, ‘night (monosyllable)’ and ‘minute’ . . . do not” (1976, 1240). Her includes among measure words all “non-classifier unit words”, that is, numeratives other than classifiers (2012, 1669). (18b) occurs in the Chinese translation of ‘5000 people were fed with 5 loaves and 2 fish.’ Her notes that in China, the game that (19a) refers to is called er ren san zu [2 person 3 foot]. This phrase, too, has two instances of Num –N: er ren ‘two people’ and san zu ‘three feet’. Her also says that in an article about teaching Chinese for technology students, Chu (1994) notes that “in a 1.67 million-character corpus of science textbooks, 1731 instances of [Num N] are found” (2012, 1680). But it is not clear from this statement how many of the nouns in those instances are ones relating to measurement that Chao would assign to the quasi-measure category. The passages in Ma (2011) and Wang (2004) are quoted in Her (2012, 1680), which has the original Chinese passages, full pinyin Romanizations, and English translations (where ‘C’ stands in for classifiers in the Chinese passages). The quotations given above have Her’s English translations and parts of the pinyin Romanizations in square brackets. Her says that predicate phrases containing (21a)–(21c) were “uttered by characters supposedly from Beijing” in “the popular movie . . . Fei Cheng Wu Rao (If You Are the One)” (2012, 1680).

Diversity of classifier systems 35 86 See also Di (2008), who argues that in Beijing Mandarin, a tonally specific version of yi ‘one’ can directly “precede a (count) noun” (ibid., §1). 87 It is implausible to take the numerative phrases in the compound words to result from omitting classifiers. I think Modern Chinese has inherited rules for forming such classifier-free compound words from earlier stages of Chinese that did not have a strong classifier system that involves classifiers as regular components of usual numeral noun phrases. (I think it is the same with (22a)–(22c) below.) 88 Some might take (23a)–(23b), which involve the character da ‘big, great’, to have the same structure as (23a)–(23c). In (23a)–(23b), however, the character does not occur as an adjective but as part of the nouns da.yang ‘big.sea (i.e., ocean)’ and da.zhou ‘big. land.mass (i.e., continent)’. 89 Most such nouns have matching classifiers. See, for example, (18a)–(18b). 90 By “element which encodes semantic information” and “element . . . which marks number”, they mean classifiers matching nouns and markers of grammatical number (which include plural markers), respectively. 91 Tʹsou (1976, 1216) independently proposes T3 (or a stronger thesis). See also Greenberg (1972, 177) and Borer (2005, Chapter 4), who attributes T3 to Tʹsou (1976). 92 In addition to Sanches and Slobin (1973), prominent proponents of T4 include Sharvy (1978), Hansen (1983, Chapter 2), Lucy (1992, 89), Krifka (1995), Link (1998, 214ff), and Borer (2005, 93f). Doetjes (1996; 1997), Chierchia (1998a; 1998b, 355; 2010), Cheng and Sybesma (1998, 410; 1999; 2005), and Zhang (2013) do not take classified nouns to refer to unindividuated stuff but still hold versions of T4. See also Quine (1969, 35ff), Greenberg (1972, 181ff), and Allan (1977, 293f). 93 See Greenberg (1972, 189). He holds a version of T4 but recognizes this problem. 94 See Judd (1886, 428) for other similar phrases. 95 For ‘cattle’ is a plural noun. Incidentally, “Three Cattle” is the title of a painting by James Burnett (National Galleries Scotland U.d.). 96 I agree with Sanches and Slobin (1973) to take the ‘head’ in ‘seven head of cattle’ to be a classifier (see also Greenberg (1972, 173)). (Although one cannot take English to have a classifier system, in my view, it has a couple of isolated classifiers, such as ‘head’ and the ‘sail’ in ‘three sail of ships’.) Pace Sanches and Slobin, however, taking ‘head’ as a classifier yields counterexamples to T3, as we have seen. 97 See, for example, Friedrich (1970, 382) on Tarasacan, Allan (1977, 294) on Ojibway and Yucatec Maya, Butler (2011, 32 and 42) on Yucatec Maya and other Mayan languages, Aikhenvald (2000, 249) on Tariana, Doetjes (2012, 2569) about Hausa, and Mache (2012, 39) about Persian. See also Yi (unpublished). 98 Hofling says “The Mayan numerals for ‘six’ to ‘ten’ are now fading from memory but are recoverable” and “When using the Mayan system from ‘five’ or ‘six’ to ‘ten’, only the all-purpose inanimate classifier pʹeel is used” (2000, 141). 99 Itzaj Maya is a counterexample to the usual assumption that classifier languages have no articles. The determiner aʹ in (25a)–(25c) is a definite article. See Hofling (2000, 247ff.). 100 For example, (2c) and (2cʹ) (which amount to ‘three cows’) have no plural marker. Moreover, one cannot add -tul to them (Lee and Ramsey 2000, 100). 101 See, for example, Song (1975; 1988b; 1988c) for more about -tul.

References Adams, K. 1989. Systems of Numeral Classification in the Mon-Khmer, Nicobarese and Aslian Subfamilies of Austroasiatic (Pacific Linguistics Series B. No. 101). Canberra: Australian National University. Aikhenvald, A. 2000. Classifiers: A Typology of Noun Categorization Devices. Oxford: Oxford University Press. Allan, K. 1977. Classifiers. Language, 53(2), pp. 285–311.

36

Byeong-uk Yi

Au Yeung, W. H. B. 2005. An Interface Program for Parametrization of Classifiers in Chinese. PhD Dissertation. Hong Kong University of Science and Technology. Au Yeung, W. H. B. 2007. Multiplication Basis of Emergence of Classifiers. Language and Linguistics, 8(4), pp. 835–861. Berlin, B. 1968. Tzeltal Numeral Classifiers: A Study in Ethnographic Semantics. The Hague: Mouton. Bisang, W. 1999. Classifiers in East and Southeast Asian Languages: Counting and Beyond. In: J. Gvozdanovi´c, ed., Numeral Types and Changes Worldwide. Berlin: Mouton de Gruyter, pp. 113–185. Borer, H. 2005. In Name Only. Oxford: Oxford University Press. Burling, R. 1965. How to Choose a Burmese Numeral Classifier. In: M. Spiro, ed., Context and Meaning in Cultural Anthropology, 1st ed. New York: The Free Press, pp. 243–264. Butler, L. K. 2011. The Morphosyntax and Processing of Number Marking in Yucatec Maya. PhD Dissertation. University of Arizona. Chae, Wan. 1982. Diachronic Study of Korean Classifiers [in Korean]. Chin-Tan Hakpo, 53–54, pp. 155–170. Chae, Wan. 1996. The Meaning and Loaning Process of the Korean Classifier ‘kay’ [in Korean]. Chin-Tan Hakpo, 82, pp. 193–215. Chao, Yuen Ren. 1968. A Grammar of Spoken Chinese. Berkeley, CA: University of California Press. Cheng, L. and Sybesma, R. 1998. Yi-wan Tang, Yi-ge Tang: Classifiers and Massifiers. Tsing Hua Journal of Chinese Studies, New Series, 28, pp. 385–412. Cheng, L. and Sybesma, R. 1999. Bare and Not-So-Bare Nouns and the Structure of NP. Linguistic Inquiry, 30(4), pp. 509–542. Cheng, L. and Sybesma, R. 2005. Classifiers in Four Varieties of Chinese. In: G. Cinque and R. Kayne, eds., The Oxford Handbook of Comparative Syntax, 1st ed. Oxford: Oxford University Press, pp. 259–292. Chierchia, G. 1998a. Plurality of Mass Nouns and the Notion of ‘Semantic Parameter’. In: S. Rothstein, ed., Events and Grammar, pp. 53–103. Dordrecht: Kluwer. Chierchia, G. 1998b. Reference to Kinds across Languages. Natural Language Semantics, 6, pp. 339–405. Chierchia, G. 2010. Mass Nouns, Vagueness and Semantic Variation. Synthese, 174, pp. 99–149. Chu, F. 1994. Keiji Hanyu Zhong de Shuming Jiegou [in Chinese: The Numeral–N Construction in Chinese for Technology]. Hanyu Xuexi (Learning Chinese), 3, pp. 40–41. Cornyn, W. 1944. Outline of Burmese Grammar (Language Dissertation No. 38). Language, 20(supplement), pp. 1–34. Craig, C. G. 1994. Classifier Languages. In: R. Asher and J. Simpson, eds., The Encyclopedia of Language and Linguistics, 1st ed. Oxford: Oxford University Press, pp. 565–569. Croft, W. 1994. Semantic Universals in Classifier Systems. Word, 45(2), pp. 145–171. Di, Xuan. 2008. Demonstratives, Numerals and Color Terms in (Beijing) Mandarin. Selected Papers from the 2006 Cyrus Syntax Fest. Available at: http://ling.auf.net/ lingBuzz/000609 (LingBuzz/000609). Doetjes, J. 1996. Mass and Count: Syntax or Semantics? Proceedings of Meaning on the HIL (HIL Occasional Papers in Linguistics 1). Holland Institute of Generative Linguistics (HIL), Leiden University, pp. 34–52. Doetjes, J. 1997. Quantifiers and Selection: On the Distribution of Quantifying Expressions in French, Dutch and English. PhD Dissertation. Leiden University.

Diversity of classifier systems 37 Doetjes, J. 2012. Count/Mass Distinctions across Languages. In: C. Maienborn, K. von Heusinger, and P. Portner, eds., Semantics: An International Handbook of Natural Language Meaning, Vol. 3. Berlin: De Gruyter Mouton, pp. 2559–2580. Downing, P. 1996. Numeral Classifier Systems: The Case of Japanese. Philadelphia, PA: John Benjamins. Edkins, J. 1853. A Grammar of Colloquial Chinese as Exhibited in the Shanghai Dialect. London: London Mission Society. Edkins, J. 1857. A Grammar of the Chinese Colloquial Language, Commonly Called the Mandarin Dialect. Shanghai: London Mission Society. Edkins, J. 1864. A Grammar of the Chinese Colloquial Language, Commonly Called the Mandarin Dialect, 2nd ed. Shanghai: Presbyterian Mission Press. Edkins, J. 1868. A Grammar of Colloquial Chinese as Exhibited in the Shanghai Dialect, 2nd ed. Shanghai: Presbyterian Mission Press. Emeneau, M. 1951. Studies in Vietnamese (Annamese) Grammar. Berkeley, CA: University of California Press. Friedrich, P. 1970. Shape in Grammar. Language, 46(2), pp. 379–407. Gil, D. R. 2008. Numeral Classifiers. In: M. Haspelmath, M. Dryer, D. Gil and B. Comrie, eds., The World Atlas of Language Structures Online. Munich: Max Planck Digital Library, Chapter 55. Available at: http://wals.info/feature/55. Accessed: 31 Jul. 2010. Greenberg, J. H. 1972. Numeral Classifiers and Substantival Number: Problems in the Genesis of a Linguistic Type. Working Papers on Language Universals, 9, pp. 1–39. Reprinted in: Greenberg (1990), pp. 166–193. Greenberg, J. H. 1990. On Language: Selected Writings of Joseph H. Greenberg. Edited by K. Denning and S. Kemmer. Stanford, CA: Stanford University Press. Hansen, C. D. 1983. Language and Logic in Ancient China. Ann Arbor, MA: University of Michigan Press. Her, One-Soon. 2012. Distinguishing Classifiers and Measure Words: A Mathematical Perspective and Implications. Lingua, 122(14), pp. 1668–1691. Her, One-Soon and Lai, Wan-Jun. 2012. Classifiers: The Many Ways to Profile ‘One’ – A Case Study of Taiwan Mandarin. International Journal of Computer Processing of Oriental Languages, 24(1), pp. 79–94. Hofling, C. A. 1991. Itzá Maya Texts, with a Grammatical Overview. Salt Lake City, UT: University of Utah Press. Hofling, C. A. 2000. Itzaj Maya Grammar. Salt Lake City, UT: University of Utah Press. Hofling, C. A. and Tesucún, F. F. 1997. Itzaj Maya-Spanish-English Dictionary. Salt Lake City, UT: University of Utah Press. Hundius, H. and Kölver, U. 1983. Syntax and Semantics of Numeral Classifiers in Thai. Studies in Language, 7(2), pp. 165–214. Iida, Takashi. 2021. Mass/Count Distinction and Japanese Semantics. In: Y.-W. Kim, C. Lee, and B.-U. Yi, eds., Numeral Classifiers and Classifier Languages. London and New York: Routledge. [Chapter 3 of this volume.] Jacob, J. 1965. Notes on the Numerals and Numeral Coefficients in Old, Middle and Modern Khmer. Lingua, 15, pp. 143–162. Judd, D. W. 1886. The American Agriculturist for the Farm, Garden & Household, Vol. 45. New York: O’Judd Co. Kang, Beom-Mo. 1994. Plurality and Other Semantic Aspects of Common Nouns of Korean. Journal of East Asian Linguistics, 3(1), pp. 1–24. Kim, Kwang-sup. 2010. On the Syntax and Semantics of Korean Classifier Constructions. Studies in Generative Grammar, 20(2), pp. 549–576. Reprinted in: Kim et al. (2011), pp. 141–168.

38

Byeong-uk Yi

Kim, Young-Wha et al., eds. 2011. Plurality in Classifier Languages. Seoul: Hankook. munhwasa. Kobuchi-Philip, Mana. 2011. The Mass Hypothesis and Japanese. In: Kim et al. (2011), pp. 283–322. Krifka, M. 1995. Common Nouns: A Contrastive Analysis of Chinese and English. In: G. Carlson and F. Pelletier, eds., The Generic Book, pp. 398–411. Chicago, IL: University of Chicago Press. Lee, Iksop and Chae, Wan. 2002. Kuk.e Mwunpeplon Kang.uy [in Korean: Lectures on Korean Grammar], 2nd ed. Seoul: Hakyeonsa. Lee, Iksop and Ramsey, S. 2000. The Korean Language. Albany, NY: State University of New York Press. Li, Charles N. and Thompson, S. A. 1981. Mandarin Chinese: A Functional Reference Grammar. Berkeley, CA: University of California Press. Link, G. 1998. Algebraic Semantics in Language and Philosophy. Stanford, CA: CSLI. Löbel, E. 2000. Classifiers versus Genders and Noun Classes: A Case Study in Vietnamese. In: B. Unterbeck, M. Rissanen, T. Nevalainen, and M. Saari, eds., Gender in Grammar and Cognition, pp. 259–320. Berlin: Mouton de Gruyter. Lucy, J. A. 1992. Grammatical Categories and Cognition: A Case Study of the Linguistic Relativity Thesis. Cambridge, MA: Cambridge University Press. Lyons, J. 1977. Semantics, Vol. 2. Cambridge: Cambridge University Press. Ma, A. 2011. Hanyu Jiaoxue Zhong Ying Zhuyi Minglianci de Fenlei Wenti [in Chinese: Nominal Classifiers Should Be Properly Classified for Pedagogical Purposes]. Xiandai Yuewi (Modern Chinese), 2011(3), pp. 37–39. Mache, A. 2012. Numeral Classifiers in Persian. München: Lincom Europa. Martin, S. E. 1992. A Reference Grammar of Korean. Tokyo: Tuttle Pub. Co. Martin, S. E. 2004. A Reference Grammar of Japanese, Rev. ed. Honolulu, HI: University of Hawaii Press. Morton, B. C. U.d. Longhorn Cattle. Available at: http://whozoo.org/Anlife99/brentmor/ longhornindex2.htm. Accessed: 04 Jan. 2018. National Galleries Scotland. U.d. James Burnet, Three Cattle. Available at: www.national galleries.org/art-and-artists/28958/three-cattle. Accessed: 04 Jan. 2018. Nguyen, Dinh Hoa. 1957. Classifiers in Vietnamese. Word, 13, pp. 124–152. Nomoto, H. 2013. Number in Classifier Languages. PhD Dissertation. University of Minnesota. [Available at: https://semanticsarchive.net/Archive/zBmYTg2Z/Nomoto2013_ diss.pdf. Accessed: 11 Sept. 2018.] Olmsted, F. L. 1904. A Journey of the Seaboard Slave States in the Years 1853–1854 with Remarks on Their Economy, Vol. 1. New York: G. P. Putnam’s Sons. Digital Print in Cambridge Library Collection, 2009. [Available at https://epdf.pub/a-journey-in-the-seaboardslave-states-volume-1-with-remarks-on-their-economy-ca.html. Accessed: 20 July 2019.] Park, Jung-Ku. 2012. The Evolution of Chinese Classifiers and its Relation to Korean Classifier System Viewed from a Typological Perspective [in Korean]. Journal of Korean Linguistics, 63, pp. 391–412. [Available at: www.dbpia.co.kr/Journal/ArticleDetail/ NODE01843081.] Pe, Hla. 1965. A Re-Examination of Burmese ‘Classifiers’. Lingua, 15, pp. 163–185. Pelletier, F. and Schubert, L. 2003. Mass Expressions. In: D. Gabbay and F. Guenthner, eds., Handbook of Philosophical Logic, Vol. 10, 2nd ed. Dordrecht: Reidel, pp. 249–336. Peyraube, A. 1998. On the History of Classifiers in Archaic and Medieval Chinese. In: B. K. Tʹsou, ed., Studia Linguistica Serica, Proceedings of the 3rd International Conference on Chinese Linguistics. Hong Kong: City University of Hong Kong Press, pp. 39–68. Quine, W. V. 1969. Ontological Relativity & Other Essays. New York: Columbia University Press.

Diversity of classifier systems 39 Sanches, M. and Slobin, L. 1973. Numeral Classifiers and Plural Marking: An implicational Universal. Working Papers on Language Universals, 11, pp. 1–22. Schell, K. 2012. Developing a Grazing Plan – Balancing Forage/Animal Inventory. SDSU iGrow. Available at: http://igrow.org/livestock/beef/developing-a-grazing-plan-balancingforage-animal-inventory/. Accessed: 4 Jan. 2018. Sharvy, R. 1978. Maybe English Has No Count Nouns: Notes on Chinese Semantics. Studies in Language, 2(3), pp. 345–365. Shimojo, Mitsuaki. 1997. The Role of the General Category in the Maintenance of Numeral-Classifier Systems: The Case of tsu and ko in Japanese. Linguistics, 35(4), pp. 705–733. Simons, G. and Fennig, C., eds. 2018. Ethnologue: Languages of the World, 21st ed. [Online]. Dallas: SIL International. Available at: www.ethnologue.com. Accessed: 30 Aug. 2018. Simpson, A. and Ngo, Binh. 2018. Classifier Syntax in Vietnamese. Journal of East Asian Linguistics, 27(3), pp. 211–246. Sohn, Ho-Min. 1999. The Korean Language. Cambridge: Cambridge University Press. Song, Seok Choong. 1975. Rare Plural Marking and Ubiquitous Plural Marker in Korean. Papers from the Eleventh Regional Meeting, Chicago Linguistic Society, 11, pp. 536–546. Song, Seok Choong. 1988a. Explorations in Korean Syntax and Semantics (Korea Research Monograph 14). Berkeley, CA: Institute of East Asian Studies, University of California. Song, Seok Choong. 1988b. A Ubiquitous Plural Marking. In: Song (1988a), pp. 19–23. Song, Seok Choong. 1988c. Nouns and Pronouns, Singular and Plural. In: Song (1988a), pp. 19–23. Tai, J. and Wang, Lianqing. 1990. A Semantic Study of the Classifier tiao. Journal of the Chinese Language Teachers Association, 25(1), pp. 35–56. Tʹsou, Benjamin K. 1976. The Structure of Nominal Classifier Systems. In: P. Jenner, L. Thompson, and S. Starosta, eds., Oceanic Linguistics Special Publications No. 13, Austroasiatic Studies, Part II. Honolulu, HI: The University Press of Hawaii, pp. 1215–1247. Tsuboi, Sakae. 1963. Twenty-Four Eyes. Translated by Akita Miura, 3rd ed. Tokyo: Kenkyusa. Unterbeck, B. 1994. Korean Classifiers. In: Young-Key Kim-Renaud, ed., Theoretical Issues in Korean Linguistics. Stanford, CA: CSLI, pp. 367–385. Wang, H. 2004. Liangci De Fenlei Yu Duwai Hanyu Liangci Jiaoxue [in Chinese: Classification of Classifiers and Teaching Classifiers in Chinese as a Second Language]. Jinan Xuebao (Journal of Jinan University: Humanities and Social Sciences Edition), 2014(2), pp. 113–116. Yi, Byeong-uk. 2009. Chinese Classifiers and Count Nouns. Journal of Cognitive Science, 10(2), pp. 209–225. Reprinted in: Kim et al. (2011), pp. 245–264. Yi, Byeong-uk. 2011a. What Is a Numeral Classifier? Philosophical Analysis, 23, pp. 195–258. Partially reprinted in: Kim et al. (2011), pp. 1–51. Yi, Byeong-uk. 2011b. Afterthoughts on Chinese Classifiers and Count Nouns. In: Kim et al. (2011), pp. 265–282. Yi, Byeong-uk. Unpublished. Numeral Classifiers and Plural Marking: The Paranumeral Account. Unpublished Manuscript. Zhang, Niina Ning. 2013. Classifier Structures in Mandarin Chinese. Berlin: de Gruyter Mouton. Zubin, D. and Shimojo, Mitsuaki. 1993. How ‘General’Are General Classifiers: With Special Reference to ko and tsu in Japanese. In: J. A. Guentner et al., eds., Proceedings of the Nineteenth Annual Meeting of the Berkeley Linguistics Society. Berkeley, CA: Berkeley Linguistics Society, pp. 490–502.

2

Taxonomy of numeral classifiers A formal semantic proposal Jiun-Shiung Wu and One-Soon Her

1

Introduction

In a numeral classifier language like Chinese, when a noun (N) is quantified by a numeral (Num), a classifier (C) or measure word (M) is often needed, as in (1a) and (1b), respectively. Note that C and M appear in exactly the same syntactic position and are in complementary distribution, as shown in (2). san tiao yu 3 C fish ‘3 fish’ b. san xiang yu 3 M-box fish ‘3 boxes of fish’ (2) a.* san tiao xiang yu b.* san xiang tiao yu (1) a.

This fact suggests that C and M in a classifier language form a single syntactic category, which we shall dub ‘C/M’. This C/M convergence view is shared by many linguists, for example, Hass (1942), Emeneau (1951), D. Nguyen (1957), Chao (1968), Becker (1975), Li and Thompson (1981), Huang (1982), Hundius and Kölver (1983), Craig (1994), Lin (1997), Cheng and Sybesma (1998, 1999), and Hsieh (2008), among many others, where various arguments have been constructed to support this view (see Her 2012b for a summary). Most standard dictionaries, grammar books, and language textbooks of classifier languages also do not make any formal distinction between the two. However, as likewise noted by numerous linguists, Cs and Ms are very different. This following informal characterization by Tai and Wang (1990: 38) is an oftencited example: A classifier categorizes a class of nouns by picking out some salient perceptual properties, either physically or functionally based, which are permanently associated with entities named by the class of nouns; a measure word does not categorize but denotes the quantity of the entity named by noun. (Tai and Wang 1990: 38)

Taxonomy of numeral classifiers

41

The undeniable fact that Ms quantify the head noun, but Cs must qualify the noun in terms of certain semantic features, has compelled many researchers to claim that C and M are two distinct semantic and/or syntactic categories, for example, Greenberg (1990[1972]), Tai and Wang (1990), Tai (1994), Her and Hsieh (2010), X. Li (2011, 2012), and Zhang (2011), among many others. T. Nguyen (2004: vii), for example, is emphatic that “classifiers and measure phrases, often treated as belonging to a single category in prior literature, are shown to be fundamentally distinct”. Likewise, various arguments have been constructed to support this divergence view (also see Her 2012b for a summary). The fact that C and M seem to converge and diverge at the same time has long put linguists in a quandary. In practice, most syntactic works, especially the more recent ones in the generative framework, such as Tang (2005), Hsieh (2008), Huang et al (2009), Her (2012b), and A. Li (2014), to name just a few, take C and M to be a single category, some explicitly and others implicitly, even though, as H. Zhang (2007: 45) observes, “Chinese linguists generally agree that a distinction between these two kinds should be made”. Wang (1994: 19–20) complains that “previous works in Chinese grammar treat classifiers and measure words on an equal footing” and proclaims that “it is essential to tell classifiers from measure words both semantically and syntactically”. N. Zhang (2009) also observes that “the relation between classifiers and measure words has baffled linguists for a long time” and subsequently proposes in N. Zhang (2011, 2013) that C and M constitute two different categories in both syntactic and semantic terms. Yet, even for those who explicitly claim that C and M should be distinguished, the criteria proposed are often informal and imprecise. The quote from Tai and Wang (1990: 38) given above is a good example. An added confusion in this already confused state of affairs is the issue of whether classifier languages like Chinese make a lexical distinction of mass and count. While many, such as Cheng and Sybesma (1998), Tang (2005), Hsieh (2008), Her and Hsieh (2010), Yi (2009, 2011a, 2011b), and Her (2012a), among others, assume such a distinction in Chinese and contend that C requires a lexical count noun but M does not, many others, such as Quine (1969), Allan (1977), Hansen (1983), Link (1991), Lucy (1992), Krifka (1995), Chierchia (1998a), Mou (1999), Toyota (2009), and Toyota et al (2012), insist that the fact that C/M is required for Num to quantify N means that N can only be mass in such languages. In this chapter, we aim to accomplish three things. First, we will follow Her (2012a) in interpreting the convergence and divergence between Cs and Ms from a mathematical perspective and further argue that C/M as a single syntactic category can be classified into different subtypes in terms of their mathematical values. Thus, Her and Lin’s (2015) formal taxonomy of various subtypes under C/M will be discussed and supported. Second, we shall offer a formal semantic account for the C/M unification as well as the various types in the C/M taxonomy. Third, we will demonstrate that in carrying out the above two goals, it is necessary to assume a lexical mass/count distinction in classifier languages. This position then leads to the conclusion that such a distinction is universal. The chapter is organized into six sections. Section 2 offers a brief description of the current confused state of affairs in the study of Cs and Ms and provides some

42

Jiun-Shiung Wu and One-Soon Her

clarification. Section 3 first reviews the similarities and difference between Cs and Ms and then accounts for such convergence and divergence with a new taxonomy of C/Ms based on an insight from a mathematical perspective, where C/Ms are viewed as multiplicands with various kinds of values. Section 4 proposes a formal semantic account of Cs and various subtypes of Ms. Section 5 demonstrates that the taxonomy and the semantics proposed in this chapter are evidence for a universal mass/count distinction at the lexical level. Section 6 consists of a summary and some concluding remarks.

2 A confused state of affairs As noted by some researchers, such as H. Zhang (2007: 45) and Her (2012a: 1669), the uncertain status of Cs and Ms is reflected in the often confusing terminology used in the literature. Terms used for C include ‘classifier’, ‘count-classifier’, ‘count-noun classifier’, ‘individual classifier’, ‘qualifying classifier’, and ‘sortal classifier’; those for M include ‘measure word’, ‘mass-classifier’, ‘mass-noun classifier’, ‘massifier’, ‘mensural classifier’, ‘measural classifier’, and ‘quantifier’. This list is not meant to be exhaustive. Furthering the confusion, many also use the term ‘classifier’, ‘numeral classifier’, or ‘quantifier’ for C/M as a single category, and others use the term ‘measure word’, ‘measure’, ‘unit word’, or ‘numerative’ instead for the same purpose. Most of the papers or books published in Chinese use the term 量詞 liangci ‘measure word’ to refer to the category of C/M, while a small number of them do clearly distinguish between 分類詞 fenleici or 類別 詞 leibieci ‘classifier’ and liangci ‘measure word’. In this chapter, we use the term ‘C/M’ to refer to the unified category of the elements between Num and N; the construction is referred to as [Num C/M N]. ‘C’ is strictly for ‘classifiers’ (or ‘sortal classifiers’) and ‘M’ strictly for ‘measure words’ (or ‘mensural classifiers’ or ‘massifiers’). Explicit criteria for this distinction will be discussed in section 3. Another symptom of the confused state of affairs, as noted by Her and Hsieh (2010: 528), is the drastically different inventories of ‘classifiers’, or linagci, compiled by different researchers for Mandarin Chinese, ranging from as few as 51 (Chao 1968) and several dozen (Erbaugh 2002), to 126 (Gao and Malt 2009), to 200 (Hung 1996), to 427 (Huang and Ahrens 2003), to as many as many as 600 (Hu 1993). These different numbers no doubt reflect very different ideas about what belongs to this category, and even for those who do make a clear distinction between Cs and Ms, there is confusion over what exactly counts as a C. To the best of our knowledge, Lai (2011) and Her and Lai (2012) on Mandarin, Chen (2013) and Chen et al (2020) on Taiwanese southern Min, and Liao (2014) on Taiwanese Hakka are the only three works in the literature that offer a comprehensive list of Cs in a classifier language that is based on explicit and testable criteria. The problem is thus two-fold: how to precisely and insightfully distinguish C and M and yet also unify C/M at the same time. Previous accounts fall short one way or another. Accounts that treat C and M as one formal category typically lack precise or formal distinction between the two subcategories, for example, Hsieh (2008), Huang et al (2009), and A. Li (2014), to name just a few. Li and Thompson (1981: 106) famously claim that “any measure word can be a classifier”. On the

Taxonomy of numeral classifiers

43

other hand, in works that do offer a distinction between C and M, the distinction is typically stated informally, and there is also a lack of accounting for C and M’s formal convergence. Examples abound; here we discuss several examples from prominent works, starting with a classic, Chao (1968: 584–620), where Cs are referred to as ‘classifiers’, ‘individual measures’, or ‘Mc’ for short and considered one out of eight types within the category of nominal measure words.1 (3) Chao’s (1968: 584–620) Classification of Measure Words 1

Mc (Classifiers, or Individual Measures): for example, 一個人 yige ren (1 C person) ‘one person’

2

Mc’ (Classifiers Associated with V-O): for example, 惹一場禍 re yichang huo (cause 1 C disaster) ‘cause a disaster’

3

Mg (Group Measures): for example, 一行字 yihang zi (1 line character) ‘one line of characters’

4

Mp (Partitive Measures): for example, 一堆土 yidui tu (1 pile dirt) ‘one pile of dirt’

5

Mo (Container Measures): for example, 一鍋麵 yiguo mian (1 pot noodle) ‘one pot of noodles’

6

Mt (Temporary Measures): for example, 一頭百髮 yitou baifa (1 head white-hair) ‘a headful of white hairs’

7

Mm (Standard Measures): for example, 一尺布 yichi bu ‘one meter of cloth’

8

Mq (Quasi-Measures, or Autonomous Measures): for example, 一釐 yili or 一趴 yipa ‘one percent’

The fundamental disadvantage associated with this taxonomy is that the eight types are all disjoint and allow no intersecting natural classes. For example, a group measure, a container measure, and a quasi-measure that has a standardized value, for example, 打 da ‘dozen’, 茶匙 chichi ‘teaspoon’, and 釐 li ‘%’, respectively, can of course also be seen as a standard measure. Further, any measure word of any type currently without a standardized value can easily take on a standardized value, be it permanently or temporarily, among a small population or a large population, and thus become a standard measure. The fact that the different types intersect means that having them as disjoint misses the essential purpose of taxonomy. The same weakness is seen in Aikhenvald (2000: 115–117), a seminal work on classifiers, where she distinguishes between ‘(numeral) classifiers’ (a term for both ‘sortal classifiers’ and ‘measural classifiers’) and ‘measure words’ (a term

44 Jiun-Shiung Wu and One-Soon Her exchangeable with ‘quantifiers’ and ‘quantifying expressions’) in classifier languages based on the observation that classifiers use the unit provided by a noun, while quantifiers, or quantifying expressions, establish the unit to be counted. Under her view, ‘classifiers’, ‘sortal’and ‘mensural’, must co-occur with count nouns, while ‘measure words’ appear with count or mass nouns. Consider the examples in (4). (4) san zhi/shuang/xiang/jin xie 3 C/M-pair/M-box/M-kilo shoe ‘3 shoes/3 pairs/boxes/kilos of shoes’ In spite of their identical syntactic position, zhi ‘C’, shuang ‘pair’, xiang ‘box’, and jin ‘kilo’ are seen as a sortal classifier, mensural classifier, quantifier, and quantifier, respectively. The former two are ‘classifiers’, but the latter two are excluded. Such exclusion misses important generalizations of C/M as a single category and is also in conflict with the view shared by most grammarians, that shuang ‘pair’, xiang ‘box’, and jin ‘kilo’ are all Ms, or mensural classifiers. S. Huang (2013), in an important recent work on Chinese grammar, where an entire chapter is devoted to numeral classifiers, claims to follow Aikhenvald’s (2000) distinction of ‘sortal’ and ‘mensural’ classifiers; furthermore, his distinction is also essentially the same as that of Tai and Wang (1990: 38): In the following discussion we will ignore Chao’s types (7) and (8), since they do not represent what I take to be true classifiers in the strict sense of the term. I will for the time being distinguish just two basic types of NC for nominal classifiers, more or less following Aikhenvald (2000), sortal classifiers and mensural classifiers. A sortal classifier is one which individuates a referent in terms of its more inherent properties such as animacy, shape, or consistency. Mensural classifiers are used for measuring more temporary state of an object, its quantity, or the arrangement it occurs in. . . . Note that sortal classifiers and mensural classifiers are to be distinguished from ordinary measure words such as jin (斤) in yijin tang. (S. Huang 2013: 167–168) S. Huang’s (2013) rejection of Chao’s type 7, or standard measures like jin ‘catty, kilo’ and 尺 chi ‘meter’, as mensural classifiers is puzzling. Standard measures, after all, occupy exactly the same syntactic position as other C/Ms, for example, zhi ‘C’, shuang ‘pair’, and xiang ‘box’ in (4), and serve the same function to measure the noun. The exclusion of Chao’s type 8, that is, quasi-measures, is also unnecessary, as these are simply standard measures of portion that often appear without a head noun. The final taxonomy we shall briefly review is found in N. Zhang (2013), where seven types of ‘unit words’, or C/Ms in our term, are recognized: individual (e.g., 枝 zhi), individuating (e.g., 滴 di), kind (e.g., 種 zhong), container (e.g., 瓶 ping), standard (e.g., 斤 jin), partitive (e.g., 段 duan), and collective (e.g., 群 qun). In spite of the apparent resemblance with Chao’s taxonomy, there are several noteworthy differences. First, N. Zhang rightly ignores Chao’s type 2, classifiers associated with V-O, as such classifiers easily appear in non-object positions, and

Taxonomy of numeral classifiers

45

Chao’s type 8, quasi-measures, which, as shown earlier, can be seen as standard measures. Second, N. Zhang (2013: 212) contends that Chao’s type 6, or temporary measures, can be seen as container measures or nouns. Third, N. Zhang includes kind classifiers as a separate type, which is rather reasonable. The fourth difference is more important: Chao’s type 4, partitive measures, is divided into two types, partitive and individuating. An individuating unit word must select nouns that are not delimitable, but a partitive unit word can select either delimitable or undelimitable nouns. A delimitable noun is atomic, while an undelimitable noun has no intrinsic shape, size, or boundaries. Note that N. Zhang makes an extraordinary, and likely to be controversial, claim that container, standard, partitive, and collective unit words have a left-branching structure, but individual, individuating, and kind unit words have a right-branching structure. The seven types thus in effect form two formally distinct categories, as shown in Table 2.1. This dichotomy has serious implications. Assigning the individual Cs and the individuating Cs to the same category is inconsistent with the fact that the former requires delimitable nouns and the latter undelimitable nouns. On the other hand, assigning the individual Cs and collective Ms to two different categories misses the generalization that they both require delimitable nouns. More importantly, this dichotomy predicts that the left-branching measures and right-branching classifiers need not follow the same word order in a language. We should thus expect to find that the two distinct categories deviate in word order in some languages. Yet, Greenberg (1990[1975]: 228) claims that Cs and Ms do not deviate in word order in any classifier language. In an ongoing research project headed by the second author, in a database of over 500 classifier languages, no evidence is found that Cs and Ms differ in word order in any of them. In addition, in all these languages, in a nominal phrase that consists of a Num and an N, either a C or an M is used, never both. C and M are always in complimentary distribution. These facts are disconfirming to N. Zhang’s dichotomy and suggest strongly that C and M belong to the same category and occupy the same structural position.2 Last but not least, it is crucial to distinguish between the formal syntactic category of C/M, as defined by the construction [Num C/M N], and the function they serve, for there is unfortunately a prevailing confusion over whether certain types of Cs or Ms are in fact universally available in all languages. The simple fact is that

Table 2.1 Taxonomy of C/Ms derived from Zhang (2013) Category

Type

Examples

Measures (left-branching)

Container Standard Partitive Collective Individual Individuating Kind

瓶 ping, 箱 xiang 斤 jin, 尺 chi 段 duan, 節 jie 群 qun, 打 da 枝 zhi, 朵 duo 滴 di, 灘 tan 種 zhong, 類 lei

Classifiers (right-branching)

46 Jiun-Shiung Wu and One-Soon Her classifier languages like Chinese employ C/M as a syntactic category, while nonclassifier languages like English have the functional equivalent of such words but do not distinguish a C/M category formally. Many linguists make the misleading claim that measure words, or mass-classifiers, are a mundane part of all natural languages, and only sortal classifiers, or count-classifiers, are unique to classifier languages like Chinese and Japanese and are not a part of English grammar except in rare cases such as five head of lettuce (e.g., Allan 1977: 285–286; Croft 1994: 151–152; Tai 1994: 481; Wang 1994: 1; Aikhenvald 2000: 115; Her and Hsieh 2010: 528, among many others). N. Zhang (2013: 246) and Croft (1994: 151–152) claim specifically that standard measures (e.g., kilo in five kilos of apples), container measures (e.g., bottle in three bottles of milk), kind CL (kind in three kinds of chocolate), partitive CLs (e.g., section in three sections of orange), and collective CLs (e.g., group in three groups of students) are universally available, but individuating CLs and individual CLs exist only in classifier languages. Such a view seriously confuses function with form. Worse still is Toyota’s (2009: 120) use of the term classifier in his claim that English has numerous classifiers. In the following quote, PDE stands for present-day English.3 The use of classifiers with the mass noun is quite common in PDE, as in a piece of furniture, an item of clothing, etc. There may be numerous classifies in PDE, but typical examples used for the analysis in this paper are listed in (5). (5)

an article of clothing; a blade of grass; a block of ice; a bit of information; a bunch of grapes; a cake of soap; a cut of meat; a drop of water; an ear of corn; a grain of corn; an item of clothing; a leaf of sage; a loaf of bread; a lot of water; a piece of information; a sheet of paper; a slice of bacon; a speck of dust; a stick of chalk; a strip of land; a suit of clothing. (Toyota 2009: 120)

None of the words of measure in the above quote is the Chinese-style M or C. In a non-classifier language, there can of course be words of measure, for example, kilo, bottle, section, group, and so on in English, which have the semantic function identical to that of Ms in a classifier language, for example, the corresponding jin, ping, jie, qun, and so on in Chinese (e.g., Lyons 1977: 464; Croft 1994: 152; Löbel 2000: 223). However, formally, since there is no (overt) structural position for C in English or other non-classifier languages, there can be no such position for M.4 There can thus be no Ms in English, not in the same formal sense as in Chinese or other classifier languages. Words of measure in English in fact behave as Ns, and nothing like Ms in Chinese (e.g., Lehrer 1986; Van Riemsdijk 1998; Aikhenvald 2000; Löbel 2000; Borer 2005; Kayne 2005). As shown in (5), the word of measure box must pluralize like a noun, and its complement must be introduced by a PP, not a bare NP. Its syntactic category and structural position in (5) are exactly identical to those in (6) and (7), where box behaves as a mundane common noun. (5)

a. three boxes of chocolate b.* three box of chocolate c.* three boxes chocolate

Taxonomy of numeral classifiers (6)

a. three boxes of excellent quality b.* three box of excellent quality c.* three boxes excellent quality

(7)

a. three boxes of theirs b.* three box of theirs c.* three boxes theirs

47

In short, while a classifier language typically has both Cs and Ms,5 a non-classifier language has neither, though it does have a class of words that share the same semantic function as Ms in classifier languages.

3 A math-based taxonomy of C/M Any characterization of C and M must reconcile the dilemma that the two converge structurally in [Num C/M N] and yet diverge semantically. A long-neglected mathematical perspective offers a unique insight to the relation between Num and C/M, as well as the precise distinction between C and M. The earliest source of such a view we know of is Greenberg (1990[1972]: 172), where he states with unmistakable clarity that “all the classifiers are from the referential point of view merely so many ways of saying ‘one’, or more accurately, ‘times one’”. Thus, san tiao yu [three C fish] of (1a) is seen as [[3 × 1] fish]. Seeing sortal classifiers as a multiplicand with the numerical value of ‘one’ implies the entity being quantified must have natural boundaries and thus appear as single units (Croft 1994: 148; Bisang 1999: 113–121). This view has been pursued in some of the more recent works, for example, Bender and Beller (2006), Au Yeung (2005, 2007), and Yi (2009). Her (2012a) takes this view most seriously in coming up with a precise formulation for the unification of, and distinction between, C and M. (8) C/M Distinction in Mathematical Terms (Her 2012a: 1679) [Num × N] = [[Num × X] N], where X = C iff X = 1, otherwise X = M. Under this precise formulation, C and M converge in entering the same multiplicative relation as the multiplicand, with Num as the multiplier. Thus, crucially, while Cs and Ms converge in denoting one unit, the one unit denoted by Cs has the precise numerical value of one, but the one unit denoted by Ms does not, for example, da ‘dozen’ denotes the precise numerical value of twelve. C/M as a multiplicand denoting one unit thus is naturally identically coded structurally in a language. This convergence explains why C/M consists of a single syntactic category and in many respects behaves in a uniform manner. Yet C and M also diverge in their respective values: all Cs as the multiplicand must necessarily have the exact numerical value 1, while an M’s value can be anything except 1 and can thus be numerical or non-numerical. More simply, an M’s value is never necessarily 1. This divergence in mathematical value accounts for the semantic differences and consequent behavioral differences observed between Cs and Ms.

48

Jiun-Shiung Wu and One-Soon Her

The most significant difference is M is semantically substantive, but C is semantically redundant, in the sense that M does, but C does not, contribute additional information to the head noun (e.g., Greenberg 1990[1974]: 201; W. Li 2000: 1117). Her and Hsieh (2010) apply Aristotle’s essential vs. accidental property to C/M’s semantic distinction:6 while a C denotes one or more essential properties of the head noun, an M denotes accidental properties. This means that a C is semantically redundant in the nominal expression, but an M is not. For example, the following three expressions, 三隻魚 san zhi yu, 三條魚 san tiao yu, and 三尾魚 san wei yu, have exactly the same connotation, that is, 3 fish, in spite of the different Cs; yet, if the three different Cs are replaced with three different Ms, then each expression will have a different connotation, for example, 三磅魚 san bang yu ‘3 pounds of fish’, 三箱魚 san xiang yu ‘3 boxes of fish’, and 三籃魚 san lan yu ‘3 baskets of fish’. This difference between C and M also receives a plausible explanation in the multiplicative theory: a multiplicand in a multiplicative relation is redundant if and only if its value is 1; a C is thus redundant in a nominal expression, but M is not (Her 2012a). Such redundancy also manifests itself in C’s being transparent to adjectival modification. Thus, (9a) and (9b) have the same meaning, but (10a) and (10b) do not. (9) a.

一大顆蘋果 = yi  da ke pingguo = one big C apple ‘one big apple’

b. 一顆大蘋果 (Her and Hsieh 2010 (13a)) yi ke da  pingguo one C big apple ‘one big apple’

(10) a. 一大箱蘋果 ≠ b. 一箱大蘋果 (Her and Hsieh 2010 (13a)) yi da  xiang pingguo ≠ yi xiang da pingguo one big M-box apple one M-box big apple ‘one big box of apples’ ‘one box of big apples’ Note that an M can, of course, be represented redundantly as 1M; however, crucially, the value of 1M is entirely different from the numerical value of any C, which is exactly 1. For example, san da [3×dozen] can be expressed redundantly as [3×1dozen]; however, both expressions are different from [3×1]. Such a misunderstanding is found in N. Zhang (2013: 37), where she claims that Greenberg’s ‘times one’ interpretation of all Cs is true of all measure words as well: Greenberg (1972: 10) points out that “all the classifiers are from the referential point of view merely so many ways of saying . . . ‘times one’.” This is also true of all measure words. All types of CLs and measure words are used in counting, telling us what counts as one in the context, i.e., the unit of counting. (see Croft 1994: 152 and Allan 1977: 293). (N. Zhang (2013: 37) Importantly, in Greenberg’s works on numeral classifiers (Greenberg 1972, 1974, 1975), he has consistently made the distinction between the classifier construction and the measure construction. It is thus clear that Greenberg only intends for the ‘times one’ interpretation to apply to Cs, not Ms. In fact, in the same passage, Greenberg (1990[1972]: 172) further applies multiplication to two instances of

Taxonomy of numeral classifiers

49

measure words, two dozen and twelve pairs, and acknowledges that their ‘identity of final numerical result’, which is unmistakably 36, is drastically different from the result of [2×1] or [12×1]. C/Ms are indeed all units of counting, but the crucial difference between C and M is their inherent mathematical value. There is thus another caveat to heed: the actual value of an M may accidentally be precisely 1, but that is not the same as necessarily 1. For example, san bang yu ‘three pounds of fish’ may just happen to be exactly three fish, each weighing exactly one pound. This sheer accident does not make pound a C, not even temporarily. However precise or imprecise the value of an M, be it number, weight, volume, size, time, height, length, or monetary value, the crucial point is that as a multiplicand, its mathematical value is not necessarily 1. The value of a C is necessarily 1, but it can also have an accidental non-numerical value. For example, san tiao yu ‘3 fish’ may just happen to be exactly three pounds of fish, again each weighing exactly one pound. This sheer accident likewise does not make tiao an M, not even temporarily. C thus acknowledges a single unit of an entity with natural boundaries, whereas M creates boundaries for an entity which may or may not have natural boundaries. Some Ms thus also encode numerical values, such as 2, 12, or a vague number, and create boundaries for entities with natural boundaries only. Other Ms thus encode non-numerical values and create boundaries for all entities. Her and Lin (2015) thus put forth the innovative concept that a insightful taxonomy of C/M types can be achieved according to the different types of mathematical values encoded by C/M, as shown in Table 2.2. The two most important criteria for this classification scheme are: first, whether the value is exactly 1 and, second, whether the underlying mathematical value of a C/M as a multiplicand is numerical. The first criterion is crucial because a multiplicand is redundant only when its value is 1; this unique property sets Cs apart from Ms. The second criterion is crucial because C/Ms with a numerical value must select nouns that are delimitable, or atomic, and thus countable, while those with a non-numerical value can appear with either delimitable or undelimitable nouns. Non-numerical measure words thus come in various subtypes, for example, volume, container, weight, length, area, time, money, portion, and so on. Any C/M with a fixed or standardized value can be seen as standard, thus including C, M1, and M3. It is thus possible for a C/M to be crossed-listed, if its value is ambiguous; for example, 茶匙 chachi ‘teaspoon’ in Chinese can be used as a formal standard measure in cookbooks as well as a causal container elsewhere

Table 2.2 Types of C/M based on mathematical value Numerical or not

Fixed or not

Examples

Numerical

Fixed

NonNumerical

Variable Fixed Variable

1 ¬1 >1 (¬1 ) ¬n(¬1 ) ¬n(¬1 )

C/M Type 個 ge, 隻 zhi, 條 tiao, 本 ben, 朵 duo 2 雙 shuang, 對 dui; 6 手 shou; 12 打 da 排 pai ‘row, 群 qun ‘group, 幫 bang ‘gang’ 斤 jin ‘catty’, 升 sheng ‘liter’, 碼 ma ‘yard’ 滴 di ‘drop’, 節 jie ‘section’, 杯 bei ‘cup’

C M1 M2 M3 M4

50

Jiun-Shiung Wu and One-Soon Her

without a fixed value. It is thus possible for a container measure to be M1, if its value is numerically fixed, for example, the container measure 籠 long ‘steamer, cage’, when referring to the small steamed pork buns in a Chinese restaurant specializing in northern cuisine, has a standardized value of 10. However, elsewhere, it is an M2 or less commonly M4, with a numerical or non-numerical but variable value. The so-called partitive measure, and also what Zhang (2013) calls ‘individuating’ unit words, are now part of M3 and M4, depending on whether the value is fixed, for example, 半 ban ‘half’, or variable, for example, 節 jie ‘section’. A collective measure is now part of M1 and M2, again depending on whether its value is fixed, for example, 打 da ‘dozen’, or variable, for example, 群 qun ‘group, herd’. Some linguists also consider event classifiers, for example, 場 chang, a separate type among C/Ms. For example, Huang and Ahrens (2003) distinguish three types: individual, event, and kind. However, an event, by definition, is delimitable and must have a beginning and an end. Event classifiers are thus merely a subtype of Cs (Lai 2011; Her and Lai 2013). Finally, we treat the kind unit words, for example, 種 zhong, 類 lei, 樣 yang, and 式 shi, as measure words, more specifically as a subtype of the non-numerical M4. This classification is confirmed by a process of elimination. A kind unit word is not restricted to delimitable nouns, which means it can only be an M3 or M4. Given that what constitutes a kind is hardly standardized or fixed, it cannot be an M3. The C/M taxonomy thus offers comprehensive coverage of all the elements that appear as C/M in the constituent formed by Num, C/M, and N.

4 A formal semantic account In this section, we shall first briefly review several important accounts of formal semantics, Chierchia (1998a, 1998b), Krifka (1995), Jiang (2012), X. Li (2011), and Rothstein (2010), and demonstrate that they cannot model the semantic distinction between C and M argued for in Section 3, before we propose our own formal semantic account. 4.1   Review of previous accounts The semantics of a C/M is closely related to that of the head noun, as a fundamental criterion for identifying different types of C/M is the semantics of the nouns they select. Since Chinese nouns are not marked for plurality, there are two potential theoretical hypotheses for the semantics of Chinese nouns. The first is that Chinese nouns denote kind, referred to as ‘kind analysis’ in this chapter, for example, in Chierchi (1998a), Krifka (1995), and so on, and the other is that Chinese nouns denote mass, that is, the famous mass noun hypothesis, for example, in Chierchia (1998b), Hansen (1976), and so on. We start from the kind analysis. Carlson (1980), observing that bare plurals in English can denote kind, proposes that kind is a (special type of) individual, which is distinguished from an individual described a singular count noun, and that kind can be mapped to object (i.e., individual), which, in turn, can be mapped to stage (Carlson 1980: 69). McNally (2017) looks into the semantics of kind from another perspective, based on the idea that common nouns denote distributional

Taxonomy of numeral classifiers

51

representations. However, McNally also proposes a type shifter KO’ to achieve the same purpose of mapping kind to an object. There is another type of research on the semantics of nouns that seems relevant here. Grimm (2012) examines nouns in languages such as Welsh or Maltese and describes a third distinction among nouns, in addition to singular/plural and count/ mass, that is, collective/entity. Nouns in these languages thus have singular and plural morphological markings, demonstrate the count-mass distinction, and also denote collectivity and entity. Morphologically, collective nouns are not marked, but they are marked when denoting an entity of that collective (Grimm 2012: 53). Grimm (2012: 68) proposes a hierarchy of individualization: substances < granular aggregates < collectives < individual entities and suggests that the grammatical number system for nouns relies on different parts of the hierarchy; for example, languages that have a collective/entity distinction use the middle part of the hierarchy. Chierchia (1998a: 354) explicitly proposes that nouns denote kind in Chinese. Chierchia bases his idea on the fact that a bare noun can appear in an argument position in Chinese. A constituent that can appear in an argument position must be of type e. If Chinese nouns were of type , like English count nouns, then it would be difficult to explain why a constituent of type can appear in a position where a constituent of type e is required. Although not providing formal details, Chierchia (ibid.) identifies as one of the distinctive features for languages whose nouns are of type e that such languages utilize classifiers. If Chinese nouns denote kind, then some device is required to map kind to countable objects, that is, individuals, so that such nouns can be quantitized. Krifka (1995) proposes that classifiers perform type-shifting, in a sense similar to Carlson (1980) and McNally (2017), and maps a kind to countable objects. For example, xiong ‘bear’ is a name referring to kind, that is, Ursus, and the other usages, including san zhi xiong [three C bear], are derived from the kind reading. To account for the [Num C/M N] construction, Krifka proposes an operator that takes a noun denoting a kind and returns the number of specimens of that kind. See the following example. (11) a. [zhi] = λnλyλiλx[RTi(x, y) ^ OUi(y)(x) = n] b. [san zhi] = λyλiλx[RTi(x, y) ^ OUi(y)(x) = 3] c. [san zhi xiong] = λiλx[RTi(x, Ursus) ^ OUi(Ursus)(x) = n] (Krifka 1995: 401) In (11), RT applies to a kind and gives us the property of being a specimen or a subspecies of the kind. OU, object unit, is a function that takes a noun expressing a kind and gives back the number of the specimens of that kind. The subscripted i represents a possible world. The number san ‘three’ instantiates the n argument of the classifier zhi, that is, OU specifies that the number of the specimens of a kind is three. The semantics of xiong is Ursus, a kind, which instantiates the y argument in the semantics of san zhi ‘three C’. Therefore, the semantics of san zhi xiong ‘three Cl Ursus’ is: RT maps the kind, Ursus, to specimens, OU counts the specimens of the kind Ursus, and the number of the specimens is three.

52

Jiun-Shiung Wu and One-Soon Her

For san qun xiong [three herd Ursus], Krifka analyzes qun as a counterpart of English herd and its semantics is: λnλyλiλx[RTi(x, y) ^ herdi(x) = n]. For san zhong xiong ‘three kind Ursus’, the semantics of zhong is λnλyλiλx[RTi(x, y) ^ KUi(y)(x) = n]. As can be seen from the semantics of zhi and qun, the difference is that qun does not require an OU operator because its core semantics is herd. On the other hand, the semantic difference between zhi and zhong is that the former takes an OU operator, which returns the number of specimens of a kind, while the latter has a KU (kind unit) operator, which gives back the number of kind. The kind analysis of nouns cannot apply to Chinese nouns for two reasons.7 First, under Krifka’s (1995) idea that a classifier serves as a type shifter that maps a kind to an individual, a noun without a classifier should receive a kind reading. While Chierchia (1998a) does not discuss the function of classifiers, in this chapter, Chinese nouns are proposed to express kind. But, standing alone without a classifier, a Chinese noun does not (necessarily) express kind. For example, the phrase wu bing er yu (five loaf two fish), gets an individual reading only, that is, five loaves and two fish. In examples without a numeral, such as shu, wo dou nianwan le ‘book, I all read-finish Pfv’, shu ‘book’ here does not refer to kind as well, since it is impossible to finish reading books as a kind. Second, it is unclear how exactly the type shifter can pick out a coherent individual from a kind. While it seems natural to think of a kind as consisting of individuals, Carlson’s (1980) theory treats kind as a type of individual. Then, the question is how the type shifter can identify a discrete, natural or man-made unit from a kind, for example, a banana, a car, and a trillion. This question has not been addressed in works that support the kind-to-individual theory.8 To resolve this problem, one may think of Grimm’s (2012: 134) definition of ‘wholes’, “any two parts that make up the whole of x [that thing] are connected to each other.” Yet this definition is too restrictive for classifiers in Chinese. For example, in the phrase yi ben zhilipuosui de shu ‘one C in-tatters book’, the classifier and zhilipuosui ‘in tatters’ are still compatible. Such examples suggest that Grimm’s definition of wholes cannot be applied to Chinese classifiers. Hence, the second problem remains. On the other hand, Chierchia (2010) suggests that classifiers, which are of type , are needed to type shift kind (of type k) to type so that numbers, which are of type , can go with such a noun. He (1998a: 349) further proposes that kind is a function from worlds w to pluralities. Chierchia (1998b, 2010) suggests that the semantics of a plural is the set of all the atoms and of all the possible sum individuals of the atoms and that the semantics of mass equals that of plurals, except that the boundaries among the atoms for mass are vague. Chierchia (1998b: 74) further suggests that classifiers map a mass noun to atoms. Therefore, while technically a classifier can shift the semantic type of a kind (or mass), it is not clear how the vague boundaries of atoms for a mass noun can be made distinctive so that individual atoms can be separated from each other. That is, even with Chierchia’s (1998a, 1998b, 2010) ideas, the second problem is still in effect. Next, let’s look at the mass noun hypothesis. Hansen (1976) suggests that Chinese nouns are mass. Chierchia (1998b) also has a similar proposal.9 As pointed out above, Chierchia (1998b, 2010) proposes that the semantics of a plural is the set of all the atoms and all the sum individuals of the atoms and that the semantics of a mass noun is like that of a plural, except that the boundaries between the atoms

Taxonomy of numeral classifiers

53

for a mass noun are vague. Rothstein (2010), on the other hand, proposes the following semantics for mass nouns, count singular nouns, and count plural nouns, based on a complete atomic Boolean algebra M: (12) a. Root nouns: Nroot ⊆ M: Root nouns denote a Boolean algebra of mass entities, the closure of a set of atoms in M under the sum operations  m. b. Mass nouns: Nmass = Nroot c. Singular count nouns: COUNTk(Nroot): A singular count noun denotes a set of ordered pairs of which the first projection is Nroot ∩ k, a subset of Nroot whose members do not (generally) overlap, and the second projection is the context k. d. Plural count nouns: in a default context k, PL(Nk) ⊆ M × {k}, where the first projection is the closure of Nroot ∩ k under sum and the second project is k. COUNTk takes the root noun meaning and gives back entities which, in the given context k, are qualified as atoms and thus can be counted. Based on the semantics in (12), Rothstein also assumes that C/Ms in Chinese are the overt realization of the COUNTk operator, which applies to the root meaning (= the mass meaning) of a noun. X. Li’s (2011) formal semantic analysis of C/Ms is based on Rothstein’s (2010) idea. He first distinguishes four types: (a) counting classifiers, which identify the natural counting unit inherent to sets of discrete entities; (b) classifiers with a measuring reading, which “simply measure the quantity of entities along a certain dimension, for example weight, length” (X. Li 2011: 126); (c) classifiers with both counting and measuring readings; and (4) kind classifiers. Note that X. Li uses the term ‘classifiers’ for all C/Ms. He also assumes that Chinese nouns are mass. The semantics of the four types are shown in (13) and the semantics of numbers in (14): (13) a. b. c. d.

[duo] = λkλx.π1(x)ϵ(ᴗk∩k)^Blossom-form(π1(x))^π2(x) = k [ping] = λkλx.π1(x)ϵ(BOTTLE∩k)^CONTAIN(π1(x), k)^p2(x)=k [jin] = λnλkλx.xϵᴗk ^ POUND(x) = n The kind classifier is an overt lexical realization of the operator , which takes a kind and gives back the set of subkinds according to a partition.

(14) a. [wu] = λPλx.P(x)^|π1(x)|k = 5 b. [yi] = 1 (13b) is an instance of C/Ms with both counting and measuring readings. Since in phrases such as yi ping shui ‘one bottle water’, the number expresses the quantity of bottle, the atom, π1(x), is a member of the intersection of bottle and the context variable k. CONTAIN is a relation for containers on the counting reading. (13c) is an example of standard measures with a measuring reading. The first lambda-bound variable of (13c) is different from those of (13a) and (13b) because X. Li (2011: 151) claims that a C/M with a measuring reading forms a constituent with Num first, while those with a count reading, such as (13a) and (13b), merge with a kind N first. (13d) should be self evident.

54

Jiun-Shiung Wu and One-Soon Her

The semantics of numbers in (14) are proposed in X. Li (2011: 149–151). It is important to note that numbers have different semantics for different readings: on the one hand, there are C/Ms with a measure reading only, and on the other hand, there are those with a count reading and those with both count and measure readings. Let’s see the diverse semantic derivations below. (15) a. [yun] = ∩CLOUD b. [duo yun] = λx.π1(x)Î(ᴗ∩CLOUD∩k)^Blossom-form(π1(x))^π2(x)=k c. [wu duo yun] = lx.π1(x)ϵ*(ᴗ∩CLOUD∩k)^*Blossom-form(π1(x))^|π1(x)|k=5 ^π2(x)=k (16) a. [shui] = ∩WATER b. [wu ping shui] = x.p1(x)ϵ(BOTTLE∩k) ^ CONTAIN(π1(x), ∩WATER) Ù|π1(x)|=5 ^π2(x)=k (17) a. [yi jin] = λkλx.xϵᴗk^POUND(x)=1 b. [mi] = ∩RICE c. [yi jin mi] = λx.xϵᴗ∩RICE^POUND(x)=1 As shown in (15)–(17), numbers play different roles and hence have different semantics. In (15) and (17), wu ‘five’ takes the phrase consisting of a C/M and a kind noun as its argument, while in (17), yi ‘one’ instantiates the n argument in jin ‘pound’, and the phrase composed of yi and jin takes a noun as its argument. Jiang (2012) also adopts the idea that Chinese nouns are kind-referring and a C/M transforms a kind-denoting noun into a set, that is, , so that the atoms in this set can be dealt with. She claims that this provides a unified account for C/ Ms in Chinese because standard measures classifiers, like bang ‘pound’, measure parts under sum; kind classifiers, like zhong ‘kind, measure sub-kinds; container classifiers, like wan bowl’, measure parts via a fill-in-relation with respect to the noun; group classifier like qun ‘group’ measure sets formed as groups; partitive classifiers, such as ceng ‘layer’ measure partitions. (Jiang 2012: 139) Note that Jiang also uses the term ‘classifiers’ for all C/Ms. While all of the works above are enlightening in deciphering the semantics of C/Ms in Chinese, they face some challenges. First, they all assume that Chinese nouns are either kind or mass. As pointed out above, Chinese nouns cannot express kind, because, standing alone without a classifier or as a bare noun, a Chinese noun does not (necessarily) have a kind interpretation. Moreover, the hypothesis that Chinese nouns are mass has been seriously challenged. For example, Yi (2009, 2011a, 2011b), adopting a similar mathematical perspective on C/Ms as we do, argues specifically that a lexical distinction of mass/count is needed in Chinese, as well as other classifier languages. Kuo and Wu (2010) propose specific syntactic tests for mass/count distinction in Chinese.10

Taxonomy of numeral classifiers

55

Particularly, Yi (Unpublished) argues against what he refers to as a (revised) individualizer theory, where classifiers are considered some type of ‘individualizer’, which functions to ‘individualize’ a kind or mass and transforms the kind/ mass to atoms, because such an analysis leads to inconsistencies in both linguistic facts and the syntactic interactions between numeral, classifiers, and head nouns. Furthermore, Her and Hsieh (2010) and Her (2012a, b) demonstrate that it is not possible for a C to serve as an ‘individualizer’, and thus that count nouns must be recognized lexically. Her and Hsieh (2010: 533) use yi gen xiangjiao ‘one C banana’ as an example and argue that: Under this view [all Chinese nouns are mass], xiangjiao ‘banana’ can only refer to the banana mass, and the reading of a natural unit of banana with peel is only accidental and due to the classifier gēn, which ‘carves out’ an elongated discrete unit. This view thus predicts that (5) [yì gēn xiangjiao ‘one C banana’], besides this natural reading, can also mean an elongated unit of bits or pieces of the banana substance or mashed banana. Such a reading is simply impossible. The second challenge is related to the first. Krifka (1995) utilizes an OU to take a kind as an input and the specimens of the kind as output. Rothstein (2010) uses a ᴗ operator for a similar purpose. Chierchia (2010) suggests that classifiers perform such a function. Yet it is not clear how these operators can distinguish nouns that allow this operation from those that do not. There are nouns in Chinese that behave in a way parallel to mass nouns in English, for example, shui ‘water’ and mianfen ‘flour’, in that they do not have atoms in their semantics. Obviously, OU or ᴗ cannot apply to such nouns, as no Cs can go with such nouns, which allow Ms only. In X. Li’s (2011) account, numbers come in two different semantics, depending on the different readings of a [Num C/M N] expression. A unified semantics for numbers is, of course, preferred, which is the case under our proposal for the semantics of C/Ms. In order to capture the semantic distinction between C and M that we have argued for, we utilize Link’s (2003[1983]) semantics of mass nouns and plural nouns,11 where the semantics of a singular count noun is a set of entities denoted by the noun. He proposes a * operator, which takes a set of entities and produces all of the individual sums of the entities. The sum operation is represented as ⊕. If α is an individual and β is another individual, then α⊕β is also an individual (i.e., individual sum) composed of α and β, which can be considered a plural individual. Here is an example. (18)

a. [gou] = {Xiaobái, Xiaohuáng, Xiaohēi} b. Xi obái Xi ohuáng Xi oh i

Xi obái Xi ohuáng

Xi obái Xi oh i Xi ohuáng Xi oh i

56

Jiun-Shiung Wu and One-Soon Her

(18a) is a set of three dogs: Xiaobai, Xiaohuang, and Xiaohei. This is the semantics of singular count noun gou ‘dog’. The * operator applies to gou ‘dog’ and forms the semantics of plural gou: all of the individual sums of the three dogs, that is, (18b). Under this definition, the starred noun now has the same cumulative property as a mass noun. A plural noun and a mass noun are both closed under sum formation. Any sum of parts of a plural noun is again the plural noun, and any sum of a mass noun is again the mass noun. In (18), for example, if we perform sum operation on Xiaobai⊕Xiaohuang and Xiaobai⊕Xiaohei, we get Xiaobai⊕Xiaohuang⊕ Xiaobai⊕Xiaohei, reducible to Xiaobai⊕Xiaohuang⊕Xiaohei. If we execute sum operation on Xiaohuang⊕Xiaohei and Xiaobai⊕Xiaohuang⊕Xiaohei, we get Xiaohuang⊕Xiaohei⊕ Xiaobai⊕Xiaohuang⊕Xiaohei, which, again, can be reduced to Xiaobai⊕Xiaohuang⊕Xiaohei. Pick any two individuals in (18b) and perform sum operation, you get an individual in (18b). This is the cumulative property for plural nouns and mass nouns. Combining (18a) and (18b), we get (19): (19)

Xiaobai Xiaohuang Xiaohei

Xiaobai Xiaohuang

Xiaobai

Xiaobai Xiaohei

Xiaohuang

Xiaohuang Xiaohei

Xiaohei

In (19), the bottom level is the atomic level, where the elements in this level are atoms, that is, individual dogs. The other levels are non-atomic individual (sum) level, where the elements in these levels are plural individuals, that is, an individual composed of two or three dogs. Based on (19) and following Krifka (1998), suppose that [gou] is the semantics of singular gou ‘dog’ and gou represents singular gou and plural gou. Then gou can be defined as follows: (20)

a.

gou

is the smallest function that satisfies the following conditions:

(i) ∀xϵDe [[gou](x) → gou(x)] (ii) ∀x, yϵDe [(gou(x) ^ gou(y))→ gou(x⊕y)] b. [*gou] = λxϵ(De–A) [gou(x)] (20a) is the definition for a lattice for gou ‘dog’, singular and plural, in Chinese, represented as gou. (20a-i) is the baseline. All singular [gou] is gou. (20a-ii) is a recursive definition for gou. If x is gou and y is gou, then x⊕y is also gou. Hence, if both x and y are singular [gou], then x⊕y is gou. If z is also singular [gou], then x⊕y⊕z is gou, as well. We can build (19) if there are three dogs. We can build a more complicated lattice if we have more than three dogs. (20b) is the semantics of plural [gou], that is, [*gou] in Link’s (2003[1983]) terms. A represents the atomic level of a lattice. A lattice is type e because everything in a lattice is an individual – either an atomic individual or an

Taxonomy of numeral classifiers

57

individual sum. De – A stands for a lattice except for the bottom (atomic) level. (20b) says that the semantics of plural gou is one of the individual sums in the lattice for gou. Furthermore, according to Link’s idea, the semantics of a plural count noun is the same as that of a mass noun; that is, the semantics of a mass noun such as shui ‘water’, tu ‘earth’, mianfen ‘flour’, and so on, is a lattice like (19) without the bottom (atomic) level. 4.2   Proposal of a formal semantic account Given the semantics of a singular count noun, a plural count noun, and a mass noun as discussed above, we can provide semantics for the taxonomy of C/Ms proposed in section 3: (21)

a. Numerical – Fixed – × 1: [C]= λnϵR+ λPϵD λxÎ(DA or De–DA) [P(x) ^ atom(x) = n × C], where C = 1 and C profiles the head noun P in the sense of Her (2012a).12 b. Numerical – fixed – × m (where m is a positive real number, which is fixed and greater than 1): [m1] = λnϵR+ λPϵD λxϵ De–DA [P(x) Ù atom (x) = n × m], where m stands for fixed number greater than 1, and m1 has selectional restrictions on the head noun P. c. Numerical – variable – × m (where m is a positive real number): [m2] = λnϵR+ λPϵD λxϵ De–DA [P(x) Ù atom (x) = n × m ^ m > 1], where m2 has selectional restrictions on the head noun P. d. Non-numerical – Fixed: [m3] = λnϵR+ λPϵD λxϵ(De–DA) [P(x) Ù M3(x) = n], where M3 represents a specific measure of P. e. Non-numerical – Variable: [m4] = λnϵR+ λPϵD λxϵ(De–DA) [P(x) Ù M4(x) = n], where M4 stands for a vague measure of P. f. ∀xϵDe, atom(x) = #{y|y ≤ x} (Krifka 1998) g. y ≤ x iff y⊕x = x, ∀x, yϵDe

An explanation for the notations in (21) is needed. R+ represents a positive real number. A lattice such as (19) is of type . DA stands for the atomic level of a lattice such as (19). De–DA represents the parts of a lattice that stand for the semantics of a plural noun or of a mass noun. Following Krifka (1998), atom returns the number of all of the atoms that are parts of x. If x itself is an atom, atom(x) = 1, because by definition, an atom is a part of itself. If x is an individual sum, for example, u⊕v⊕w, then atom(u⊕v⊕w)=3. The definition of ‘part of’ (≤) is given in (21g). m is an operator that gives back the number of a certain measurement.

58

Jiun-Shiung Wu and One-Soon Her

(21a) is the semantics of C as defined in this chapter, such as zhi (C for animals), zhang (C for objects with a flat surface), tiáo (C for entities with a long shape), ben (C for bound printed volumes), and so on. C’s in our definition refer to one complete, discrete atom of a count noun. The exact number of atoms expressed by a nominal phrase is determined by the number in the phrase. Because the atom of the noun that goes with C can be 1 or greater than 1, x is either a member of the atomic level, that is, DA, or a member of individual sum levels, that is, De–A. (21b) is the semantics of m1, including shuang ‘pair’, duì ‘pair’, da ‘dozen’, dao ‘100 sheets (of paper)’, and so on. These measure words give us the exact number of atoms of the count noun they select, but the number must be greater than one: shuang and duì denote two, da expresses twelve, and dao means one hundred. Let’s look at the semantic derivations of shi ke pingguo [ten C apple], shi shuang kuaizi [ten M-pair chopsticks] and shi da meigui [ten M-dozen rose] below. (22)

a. [shi] = 10 b. [shi ke] = [ke] ([shi]) = λnϵR+ λPϵD λxϵDe–DA [P(x) ^ atom(x) = n × 1] (10) = λPϵD λxϵ De–DA [P(x) ^ atom(x) = 10 × 1] c. [shi ke pingguo] = [shi ke] ([pingguo]) = λPϵD λxϵ De–DA [P(x) ^ atom(x) = 10 × 1] ([apple]) = λxϵ De–DA [[apple](x) ^ atom(x) = 10 × 1] = λxϵ De–DA [[apple](x) ^ atom(x) = 10]

In (22), because ke is a C, when it combines with the number 10, we get 10 × 1. Then, it goes with [apple], a cover term for singular apple and for plural apple, just like (20a). Then, we get the number of the atoms of [apple] as 10. Here, x is a member of De–DA because the number 10 rules out the possibility that the atomic level of the semantics of apple. Next, let’s look at the semantic derivation of shi shuang kuaizi [ten M chopsticks]. (23)

a. [shi shuang] = [shuang]([shi]) = λnϵR+ λPϵD λxϵ De–DA [P(x) ^ atom (x) = n × 2] (10) = λPϵD λxϵDe–DA [P(x) ^ atom (x) = 10 × 2] b. [shi shuang kuaizi] = [shi shuang] ([kuaizi]) = λPϵD λxϵ De–DA [P(x) ^ atom (x) = 10 × 2] ([ChopstiCks]) = λxϵ De–DA [[ChopstiCks](x) ^ atom (x) = 10 × 2] = λxϵ De–DA [[ChopstiCks](x) ^ atom (x) = 20]

Shuang in (23) is a numerical M with a fixed value 2. The semantic derivation here is very similar to that in (22), except that ke represents 1 but shuang 2. Because shuang denotes a fixed number 2, the atom of ChopstiCks is 20.

Taxonomy of numeral classifiers

59

Finally, let’s look at the semantic computation of shi da meigui [ten M rose]. (24)

[shi da meigui] = [shi da] ([meigui]) = λPϵD λxϵ De–DA [P(x) ^ atom (x) = 10 × 12] ([rose]) = λxϵ De–DA [[rose](x) ^ atom (x) = 10 × 12] = λxϵ De–DA [[rose](x) ^ atom (x) = 120]

(24) is just like (23), except that da expresses the numerical value of 12. This is why we get the reading where the number of the atoms of rose is 120. The “times one” in the semantic of a classifier may seem vacuous. However, the semantic derivations of shi ke pingguo [ten C apple], shi shuang kuaizi [ten M-pair chopsticks] and shi da meigui [ten M-dozen rose] in (22), (23), and (24) prove that this semantics is not vacuous but significant. While ke, shuang, and da refer to the number of atoms, ke itself simply means 1, while shuang denotes 2 and da 12. This is why shi ke pinguo [ten C apple] refers to ten (ten times one) apples, whereas shi shuang kuaizi [ten M chopsticks] describes twenty (ten times two) chopsticks, and shi da meigui [ten M rose] one hundred and twenty (ten times twelve) roses. m1 has selectional restrictions on the head nouns it takes. For example, yi da meiguo ‘a dozen roses’, san da beizi ‘three dozen glasses’, and wu da qianbi ‘five dozen pencils’ are good, but yi da xuesheng ‘one dozen student’ and liang da gongche siji ‘two dozen bus driver’ are not good. It seems that da prefers inanimate nouns. Of course, this observation is not fine grained enough and can be further refined. But these examples show the selectional restriction of da. Shuang ‘pair’ is similar. It goes only with entities that come in pairs, such as gloves, footwear, and so on. But yi shuang xuesheng ‘one pair students’ is very bad and hard to make sense of. We will not go into the details of the selectional restrictions between m1 and its head noun and leave this issue for future studies. (21c) is the semantics of m2 as defined in section 3, for example pai, lie, wo, and so on. This type of measure words goes with count nouns and gives us an unspecified number of the atoms of the nouns. Let’s take san wo xiaogou [three M-nest puppy] as an example. Wo does not denote a specific number, unlike ke, shuang, or da. But this phrase does refer to the number of atoms of xiaogou ‘puppy’, although the number is not specified. The semantic derivation of this phrase is as follows: (25)

[san wo xiaogou] = [san wo] ([xiaogou]) = λPϵD λxϵDe–DA [P(x) ^ atom(x) = 3 × m ^ m > 1] ([puppy]) = λxϵDe–DA [[puppy](x) ^ atom(x) = 3 × m ^ m > 1]

m 2,

just like m1, has selectional restrictions on the head noun it takes. For example, yi wo shitou ‘one nestful stones’ and liang wo shu ‘two nestful trees’ are bad. So, wo is not compatible with inanimate entities or plants. But yi wo laoshi ‘one nestful teachers’ and san wo junren ‘three nestful soldiers’ are not good either. We will not go into the details of the selectional restrictions in this chapter. But it should be apparent that there are selectional restrictions between m2 and its head noun.

60

Jiun-Shiung Wu and One-Soon Her

These three types discussed above are referred to as numerical because these words themselves express a (specified or unspecified) number of atoms. The number in the nominal phrase multiplies the number indicated by the C/m, and this multiplication yields the total number of atoms expressed. The reason C, M1, and M2, as in (21a), (21b), and (21c), respectively, access the atoms of a count noun is that nominal phrases, such as shi ke pingguo [ten C apple], shi shuang kuaizi [ten M1 chopsticks], and san wo xiaogou [three M3 puppy], refer to a discrete entity as a whole. Moltmann (1997: 20) states that “singular count nouns differ from mass nouns in that they characterize an entity as an integrated whole. . . . Then, singular count nouns, but not mass nouns, must express wholeproperties as part of their lexical meaning.” Since in Link’s analysis, plural nouns and mass nouns have the same type of semantics, Moltmann’s statement also applies to plural nouns. This is why the semantics of C, M1, and M2 rely on the atomic level of the semantics of a count noun. (21d) is the semantics of M3, measure words that go with either a count noun or a mass noun and which themselves express a specific quantity of a certain unit of measurement, for example, jin ‘catty or kilo’, chi ’meter’, miao ‘second’, nian ‘year’, and so on. Let’s compare san jin weiyu [three M tuna] and san jin shadingyu [three M sardine]. These two nominal phrases do not access the atoms of tuna or sardine. San jin weiyu [three M tuna] is actually only a small part of a whole tuna, but san jin shadingyu [three M sardine] refers to a (possibly large) number of sardines. This is why in (21d), the variable x is a member of De–A, to exclude the atomic level of the lattice for weiyu ‘tuna’ and shadingyu ‘sardine’. Let’s perform the semantic derivation for these two nominal phrases. (26)

a. [san jin weiyu] = [san jin] ([weiyu]) = λPϵD λxϵDe–DA [P(x) ^ jin(x) = 3] ([tuna]) = λxϵDe–DA [[tuna](x) ^ jin(x) = 3] b. [san jin shadingyu] = [san jin] ([shadingyu]) = λPϵD λxϵDe–DA [P(x) ^ jin(x) = 3] ([sardine]) = λPϵD λxϵDe–DA [[sardine](x) ^ jin(x) = 3]

In (26), weiyu ‘tuna’ and shadingyu ‘sardine’ are guaranteed to be plural or mass because they apply to x, a variable which is of the type De–DA. san jin weiyu [3 M tuna] and san jin shadingyu [3 M sardine] refer to the weight (jin here) of tuna or sardines but do not access the atoms. Hence, this phrase does not refer to tuna or sardines as a whole. Finally, (21e) is the semantics of M4, measure words that go with a count noun or a mass noun but which themselves do not denote a specific quantity of a measurement, for example, di ‘drop’, dai ‘bag’, bei ‘glass/cup’, and so on. wu di shui ‘five drops of water’ expresses quantity of water, but the quantity is not specific. And this phrase measures the quantity of water in terms of di ‘drop’ but does not refer to the number of atoms. Let’s see how (21e) works.

Taxonomy of numeral classifiers (27)

61

a. [wu di] = [di] ([wu]) = λnϵR+ λPϵD λxϵ De–DA [P(x) ^ di(x) = 5] (5) = λPϵD λxϵDe–DA [P(x) ^ di(x) = 5] b. [wu di shui] = [wu di] ([shui]) = λPϵD λxϵDe–DA [P(x) ^ di(x) = 5] ([water]) = λxϵDe–DA [[water](x) ^ di(x) = 5]

The five semantics in (21) can be reduced to two, one for C/Ms with a numerical value and the other for those with a non-numerical value: (28)

a. [C/m] = λnϵR+ λPϵD λx ϵ(DA or De–DA) [P(x) ^ atom(x) = n × m (^ m > 1)], where C/m is distinguished whether m stands for a fixed number and, if it does, whether it equals or is greater than 1 and where C/m either profiles the head noun or has selectional restrictions on the head noun. b. [m] = λnϵR+ λPϵD λxÎDe–DA [P(x) ^ M(x) = n], where M represents a certain unit of measurement. M3 and M4 are distinguished in terms of whether M is a specific or vague measurement.

The semantics of C/Ms proposed here avoid all the problems in the formal semantic accounts reviewed earlier and enjoy several advantages. First, it does not assume that all nouns in Chinese are mass, an assumption that has been seriously challenged. Second, whether a particular type of C/M accesses the atoms in the semantics of a noun is made explicitly clear. Third, numerals receive a unified semantics. In addition, the semantics proposed in (21) and reduced in (28) capture the significant semantic distinctions among the C/m proposed in this chapter. Numerical C/m, that is, C and m1, both deal with the number of atoms of a noun, and this is why they subcategorize for a count noun. On the other hand, non-numerical m’s, that is, m3 and m4, do not address the number of atoms, and instead they simply provide a (fixed or variable) measurement for a noun. This is why m3 and m4 can go with either a count noun or a mass noun.

5

Implications of the mass/count distinction

The mainstream view in the literature on the mass/count distinction and numeral classifiers is the mass noun hypothesis, that is, classifier languages do not make this distinction in the lexicon (e.g., Yi 2009, 2011a, 2011b). Yet an important implication of the taxonomy and its semantics we have proposed in the chapter is that a lexical mass/count distinction is necessary in classifier languages, much the same as in non-classifier languages. A number of researchers have reached the same conclusion that the dominant view is mistaken, for example, Cheng and Sybesma (1998, 1999), Tang (2005), Watanabe (2006), Hsieh (2008), Yi (2009, 2011a, 2011b), Kuo

62

Jiun-Shiung Wu and One-Soon Her

and Wu (2010), Her and Hsieh (2010), Her (2012a), and N. Zhang (2013). We will discuss some of the arguments put forth in such works and offer an additional piece of evidence derived from the often-overlooked fact that Cs may be optional, but Ms are obligatory in all classifier languages. An important motivation for the mass noun hypothesis is that the use of a C is obligatory in Chinese for putative count nouns to be quantified by Num. This misconception is very common in the literature (Her 2012a: 1686), as seen in the examples below, most quoted from prominent works on Chinese. To a speaker of English, one of the most striking features of the Mandarin noun phrase is the classifier. A classifier is a word that must occur with a number . . . and/or a demonstrative . . ., or certain quantifiers . . . before the noun. (Li and Thompson 1981: 104, emphasis in original) While the use of a measure word is occasionally required in English, its presence is obligatory at all times in Chinese when a number is placed before a noun. (Zhang et al 2002: 59–60, emphasis added) In counting, all concrete nouns in standard Chinese must be used with a numeral classifier construction. (Sun 2006: 164, emphasis added) In Chinese, it is not possible to directly quantify a noun through the addition of a numeral. Instead, a classifier must intervene between the numeral and the noun to be quantified, whether the noun is conceptually a count or mass noun. (S. Huang 2013: 164, emphasis added) Chinese nouns behave like mass nouns in the Indo-European languages, as they always require the presence of classifiers in the enumeration. (Cheung 2016: 243, emphasis added) A depiction closer to the real picture is found in N. Zhang’s (2013) book devoted to Mandarin classifiers, where, in spite of the use of the word obligatory, several exceptions are recognized. Mandarin Chinese is a typical CL language. This is because, first, in a numerical expression . . . the occurrence of a CL is obligatory in the language (except in idiomatic expressions, compounds, or certain list contexts). (N. Zhang 2013: 1) The environments where the use of a C is optional in the language certainly go far beyond idiomatic expressions, compounds, and list contexts. To start with, Cs are often allowed to be omitted, motivated by economy or prosody (e.g.,

Taxonomy of numeral classifiers

63

Wei 2000: 213). The extent of optional Cs also varies in terms of genres and discourse purposes (e.g., Ding 2005; Yang 2009). For example, in a study of a 1.67-million-character corpus of science textbooks, Chu (1994) finds a whopping 1731 instances of nouns quantified directly by a numeral. As pointed out in Her (2012a: 1680), though largely overlooked in the formalist literature, optional Cs have been duly noted by some pedagogical grammarians. Here are two examples. 因為個體量詞不表量，故可省略，“一個杯子”=“一杯子” yinwei getiliangci bu biao liang, gu ke shenglue, “yi ge beizi”=“yi beizi” (because classifiers do not express quantity, they can be omitted, “1 C cup”=“1 cup”) (Ma 2011) 個體量詞：一張床（一床）、一頭牛（一牛）、一個人（一人），省略 後語意不變。 Getiliangci: yi zhang chuang (yi chuang), yi tou niu (yi niu), yi ge ren (yi ren), shenglue hou yuyi bu bian. (Classifiers: 1 C bed (1 bed), 1 C ox (1 ox), 1 C person (1 person), C can be omitted without any change in meaning.) (Wang 2004: 113) It has been noted (e.g., Yue 2003: 85) that the Sinitic languages in the south tend to use more classifiers than those in the north. Conforming to a similar pattern, Mandarin dialects likewise exhibit different degrees of optional Cs. Beijing Mandarin thus uses fewer Cs and allows C to be dropped much more freely than Taiwan Mandarin. In Her and Chen’s (2013) survey of the dialogues in the popular movie from China, 非誠勿擾 Fei Cheng Wu Rao (English title: If You Are the One) and its sequel, out of 282 noun phrases where a C is required prescriptively, more than a quarter, or 74 tokens, to be precise, are without Cs. This misconception of obligatory Cs is certainly not restricted to Chinese. Burling (1965: 244) claims that “In many languages of Southeast Asia, a number is never used without being accompanied by one of the special morphemes known as classifiers”. Greenberg (1990[1972]: 168) thus complains “On such a view, it is not excessive to state that there are no numeral classifier languages”. He offers four observations. First, there are certain classes of nouns, for example, units of time and money, which universally do not require Cs.13 Second, some classifier languages, for example, Vietnamese, allow more extensive classes of such nouns. Third, in some languages, Cs do not occur with certain kinds of numerals, for example, in Khasi, Cs do not occur with the numeral 1, and it is common for Cs not to occur with multiples of 10 or 20, depending on whether their numeral system is decimal or vigesimal. Noonan (2003: 321) observes that in Chantyal, the human classifier is dropped in casual speech, while the non-human numeral classifier occurs only with the numbers 1 and 2. Finally, there are languages, for example, Khmer, where the use of Cs is entirely optional. In a survey of 140 classifier languages, Gil (2013) identifies 78 as having obligatory Cs, while he fully acknowledges the concessions noted above in using the rather loosely

64

Jiun-Shiung Wu and One-Soon Her

defined term ‘obligatory’. With that understanding, the existence of 62 languages of optional Cs in the survey is highly significant. We can thus conclude with confidence that classifier languages universally allow a noun to be directly quantified by a numeral under certain circumstances. This in turn means a mass/count distinction must be made in classifier languages, just like non-classifier languages. Specific to our account, where C/Ms are seen as multiplicands, is the fact that Cs’ redundancy and optionality are consequences of Cs having the value of 1, as a multiplicand is redundant and thus optional if and only if its value is 1. This is important because our account also requires a lexical mass/count distinction, as not only Cs, but also Ms with numerical values, must select count nouns. If our account is on the right track, then so is the lexical mass/ count distinction in classifier languages and vice versa. Given the conventional view that a lexical distinction of mass/count is made in non-classifier languages that mark plurality, such as English and French, now we only need to establish this distinction in languages with neither Cs nor plural markers (PMs hereafter) in order to claim the universality of this distinction. According to Her and Chen (2013), out of the 114 languages covered by both Gil’s (2013) survey of 400 languages for Cs and Haspelmath’s (2013) survey of 291 languages for PMs, only 8 have neither Cs nor PMs: Chimariko (an indigenous language of California, now extinct), Imonda (Papuan), Kombai (Papuan), Mapudungun (Araucanian), Pirahã (Amazonian isolate), Salt-Yui (Papuan), Yidiny (Australian, nearly extinct), and Yingkarta (Australian). We offer two arguments for the mass/count distinction in such languages. Our first argument is logically oriented. Recall that the conventional rationale for the mass/count distinction is whether a noun is quantified by numerals directly or indirectly by means of a C. Note that there are plural-marking languages that allow only the singular form of N in [Num N] when the value of Num is above one, for example, Georgian (Karen Chung, p.c.), Finnish (Chierchia 2010: 104), Hungarian (Csirmaz and Dékány 2014: 154), Welsh (Mittendorf and Sadler 2005: 7), and Turkish (Ionin and Matushansky 2006: 326). Thus, if the [Num N] construction with PMs on N is seen as direct quantification of N by numerals, then surely the same [Num N] construction in PM-less and C-less languages must be seen as such as well. In other words, regardless of whether the noun is marked for plurality, only count nouns are selected by numerals for direct quantification. Our second argument is empirically oriented, as there is evidence in linguistic facts. We shall use Pirahã as an example, which is anumeric and thus without exact numerals and also does not mark plurality (Everett 2005). Behavioral experiments demonstrate that monolingual Pirahã speakers are unable to conceptualize exact numerical quantity larger than three (Gordon 2004; Everett and Madora 2012). However, Pirahã has two distinct quantifiers for larger quantities, which manifest a mass/count distinction (Nevins et al 2009). (29)

a. xaíbái ‘many’ (used with putative count nouns only) b. xapagí ‘much’ (used with putative mass nouns only)

Extending the view that the concept of count only requires the notion of individual, or one (e.g., Yi 2009; Her 2012a), we contend that the latter notion

Taxonomy of numeral classifiers

65

is both a necessary condition and a sufficient condition for the former. The concept of individual, or one, is universal in human cognition, as even prelinguistic infants have been shown to be capable of representing small but precise numbers (1–3) (Feigenson et al 2004). A similar view is found in Watanabe’s (2006) syntactic account, where a #P is suggested for all noun phrases, and the feature [number] in the # head is the universal locus of the mass/count distinction: “The # head is [+number] in the case of count nouns, whereas it is [-number] in the case of mass nouns” Watanabe (2006: 271). The lack of (the concept of) exact numerals above three is therefore inconsequential to the mass/count distinction, and so is the (lack of) overt marking of plurality. Given the arguments put forth, we contend that a lexical mass/count distinction is made in all human languages.

6

Conclusion

In this chapter, based on the clear distinction between classifiers (Cs) and measures (Ms) in classifier languages like Chinese, we re-classify such unit words into two types: numerical and non-numerical, according to the mathematical value they denote. The numerical type addresses the number of atoms in the semantics of a noun, while the non-numerical type offers a measurement for a noun rather than referring to the number of atoms. For the numerical type, one subtype represents a fixed number and the other a variable number. The fixed numerical type is further divided into two groups, depending on whether the value is precisely 1. The ones that have a numerical value of 1, in addition to profiling a noun in the sense, as discussed in Her (2012a), for example 隻 zhi, 條 tiao, 本 ben, and so on, are grouped under C (classifier). Note that m (measure word) stands for all those unit words whose value is not 1. Those that have a precise, fixed numerical value greater than 1, for example 雙 shuang ‘pair’,對 dui ‘pair’, 打 da ‘dozen’, and so on, are referred to as m1. The subtype referred to as m2 consists of those with numerical values that are not precise or fixed, for example 排 pai ‘row’, 群 qun ‘group’, 幫 bang ‘gang’, and so on. The non-numerical type comes in two subtypes. One subtype represents a measurement of fixed quantity, including 斤 jin ‘catty’, 升 sheng ‘litter’, 碼 ma ‘yard’, and so on. Jin represents a measurement of fixed weight and sheng a measurement of fixed volume of liquid. Ma represents a measurement of fixed length. This subtype is called m3. The other subtype, m4, refers to those with a measurement of variable quantity, for example 滴 di ‘drop’, 節 jie ‘section’, 杯 bei ‘cup’, and so on. Formally, the semantics of numerical C/m is distinguished from that of nonnumerical C/m in the following way: the former addresses the number of atoms in the semantics of a noun, while the latter describes a measurement of a noun rather than referring to the number of atoms. Moreover, numerical C/m either profiles or has selectional restrictions on the noun it goes with, but non-numerical C/m does not. One of the most interesting implications of our study is that the distinction between numerical C/m and non-numerical C/m relies on the mass/count distinction in the lexicon. We further contend that the mass/count distinction is universal.

66

Jiun-Shiung Wu and One-Soon Her

Acknowledgments We thank the anonymous reviewers for their constructive comments, which made this chapter better. We are also grateful to the editors for their hard work. Moreover, the second author, O.-S. Her, gratefully acknowledges the financial support of the two MOST research grants, 104-2410-H-004-164-MY3 and 106-2410-H-004106-MY3. Remaining errors are ours.

Notes 1 One more type is recognized, that is, Mv, or measures for verbs of action, which are usually referred to as ‘verbal classifiers’. These are very different from nominal classifiers and are excluded from discussion in this chapter. 2 This means that X. Li’s (2011: 118) proposal that C/M involves both left- and rightbranching structures, the former needed for the quantity reading and the latter the counting reading, is likewise problematic. 3 Ironically, Toyota (2009) concludes, rather reasonably, that PDE is not a non-classifier language due to the common use of such words of measure with mass nouns and not count nouns, while earlier English is a classifier language, since such words of measure are scarce and thus a strict mass/count distinction is not made. 4 Some researchers view plural markers in non-classifier languages like English and Cs in classifier languages like Chinese as identical grammatical elements, formally and functionally (e.g., Borer 2005; Her and Chen 2013). However, even under such a view, words of measure in non-classifier languages cannot be seen formally as Cs or plural markers. In English, for examples, they must be nouns in the formal sense. 5 There are rare cases where a language has Chinese-style Ms but has no Chinese-style Cs. Canglo Monpa, a Tibeto-Burman language in Tibet (Jiang 2006: 50), and Hindi (e.g., McGregor 1995: 69–70) are two examples. 6 According to Robertson (2008), this distinction is characterized as follows: “P is an essential property of an object o just in case it is necessary that o has P, whereas P is an accidental property of an object o just in case o has P but it is possible that o lacks P”. 7 In the sense of Grimm (2012), a kind can also be seen as a collective. 8 The same criticism applies to the mass noun hypothesis, as pointed out by Her and Hsieh (2010: 533). 9 Please note that, while Chierchia (1998a) suggests that Chinese nouns denote kind, he (ibid: 353) also states that all Chinese nouns are mass. 10 Please note that Chierchia (2010: 355) says that “[. . .] saying that all members of category NP are mass-like does not mean saying that something resembling the mass/ count distinction cannot be found in such languages [. . .].” We are not certain whether Chierchia is talking about the grammar of Chinese here. It would be contradictory to claim that in the grammar of Chinese all nouns are mass and at the same time suggest that in the same grammar count/mass are distinguished. 11 Please note that, although we use Link’s (2003[1983]) proposal, the semantics of C/M proposed in this chapter can be easily stated in terms of Chierchia’s (1998a, b) and Rothstein’s (2010) semantics of mass/count nouns, with minor modification. 12 The semantics presented here assumes a left-branching structure, where that Num and C/M form a constituent first, which then merges with N. The best evidence for the left-branching account is the fact that N never comes in between Num and C/M cross-linguistically (e.g., Greenberg 1990[1975]: 227). Also see Her (2012b) for more evidence. 13 Adopting Her and Tsai’s (2015) framework that allows silent elements in syntax, Her et al (2015) demonstrate that some of such nouns are in fact Ms which select a silent noun.

Taxonomy of numeral classifiers

67

References Aikhenvald, A. 2000. Classifiers: A Typology of Noun Categorization Devices. Oxford Studies in Typology and Linguistic Theory. Oxford: Oxford University Press. Allan, K. 1977. Classifiers. Language, 53(2), pp. 285–311. Au Yeung, Wai Hoo Ben. 2005. An Interface Program for Parameterization of Classifiers in Chinese. PhD dissertation. Hong Kong: Hong Kong University. Au Yeung, Wai Hoo Ben. 2007. Multiplication basis of emergence of classifier. Language and Linguistics, 8(4), pp. 835–861. Becker, A. 1975. A linguistic image of nature: The Burmese numerative classifier system. Linguistics, 165, pp. 109–121. Bender, A. and Beller, S. 2006. “Fanciful” or genuine? Bases and high numerals in Polynesian number systems. Journal of the Polynesian Society, 115(1), pp. 7–46. Bisang, W. 1999. Classifiers in East and Southeast Asian languages: Counting and beyond. In: J. Gvozdanovic, ed., Numeral Types and Changes Worldwide. Berlin: Mouton de Gruyter, pp. 113–185. Borer, H. 2005. Structuring Sense, Vol. 1: In Name Only. Oxford: Oxford University Press. Burling, R. 1965. How to Choose a Burmese Numeral Classifier. In: M. Spiro, ed., Context and Meaning in Cultural Anthropology. New York: The Free Press, pp. 243–264. Carlson, G. 1980. Reference to Kinds in English. New York: Garland. Chao, Yuen-ren. 1968. A Grammar of Spoken Chinese. Berkeley: University of California Press. Chen, Meng-Ying. 2013. Identifying True Classifiers in Mandarin Chinese. MA thesis. Taipei: National Chengchi University. Chen, Meng-Ying, Hsieh, Miao-Ling, and Her, One-Soon. 2020. Distinguishing classifiers in Taiwan Southern Min. Language and Linguistics, 21(3) (July 2020). Cheng, Lisa L.-S., and Sybesma, R. 1998. Yi-wan Tang, yi-ge tang: Classifier and massifiers. Tsing Hua Journal of Chinese Studies, New Series, 28(3), pp. 385–412. Cheng, Lisa L-S., and Sybesma, R. 1999. Bare and not-so-bare nouns and the structure of NP. Linguistic Inquiry, 30(4), pp. 509–542. Cheung, Chi-Hang Candice. 2016. Chinese: Parts of speech. In: Sin-Wai Chan and J. Minnett, eds., The Routledge Encyclopedia of the Chinese Language. London: Routledge, pp. 242–294. Chierchia, G. 1998a. Reference to kinds across languages. Natural Language Semantics, 6, pp. 339–405. Chierchia, G. 1998b. Plurality of mass nouns and the notion of “semantic parameter”. In: Susan Rothstein, ed., Events and Grammar. Dordrecht: Kluwer, pp. 53–103. Chierchia, G. 2010. Mass nouns, vagueness and semantic variation. Synthese, 174(1), pp. 99–149. Chu, Fuzhang. 1994. Keji hanyu zhong de shuming jiegou (The numeral-noun construction in Chinese for technology). Hanyu Xuexi (Learning Chinese), 1994(3), pp. 40–41. Craig, C. 1994. Classifier languages. In R. Asher and J. Y. Simpson, eds., The Encyclopedia of Language and Linguistics, Vol. II. Oxford: Pergamon Press, pp. 565–569. Croft, William. 1994. Semantic universals in classifier systems. Word, 45, 145–171. Csirmaz, A. and Dékány, É. 2014. Hungarian is a classifier language. In Simone Raffaele and Francesca Masini, eds., Word Classes. Amsterdam: John Benjamin, pp. 41–160. Ding, Jiayong. 2005. Xiandai hanyu shuming jiegou de pianzhang gongneng (The discourse functions of the modern numeral-noun construction). Yuyan Yanjiu (Studies in Language and Linguistics), 25(1), pp. 19–25. Dryer, M. 2008. Word order in Tibeto-Burman languages. Linguistics of the Tibeto-Burman Area, 31, pp. 1–88.

68

Jiun-Shiung Wu and One-Soon Her

Emeneau, M. 1951. Studies in Vietnamese (Annamese) Grammar. Berkeley: University of California Press. Erbaugh, M. 2002. Classifiers are for specification: Complementary functions for sortal and general classifiers in Cantonese and Mandarin. Cahiers de linguistique-Asie orientale, 31(1), pp. 33–69. Everett, C. and Madora, K. 2012. Quantity recognition among speakers of an anumeric language. Cognitive Science, 36(1), pp. 130–141. Everett, D. 2005. Cultural constraints on grammar and cognition in Pirahã: Another look at the design features of human language. Current Anthropology, 46, pp. 621–646. Feigenson, L., Dehaene, S., and Spelke, E. 2004. Core systems of number. Trends in Cognitive Science, 8(7), pp. 307–314. Gao, Ming and Malt, B. 2009. Mental representation and cognitive consequences of Chinese individual classifiers. Language and Cognitive Processes, 24(7), pp. 1124–1179. Gil, D. 2013. Numeral classifiers. In: M. Haspelmath, M. Dryer, D. Gil, and B. Comrie, eds., The World Atlas of Language Structures Online. Max Planck Digital Library. Available online at http://wals.info/feature/55. Accessed on 2015/2/21. Gordon, P. 2004. Numerical cognition without words: Evidence from Amazonia. Science, 306, pp. 496–499. Greenberg, J. 1990[1972]. Numerical classifiers and substantival number: Problems in the genesis of a linguistic type. In: K. Denning and S. Kemmer, eds., On Language. Selected Writings of Joseph H. Greenberg. Stanford, CA: Stanford University Press, pp. 166–193. [First published 1972 in Working Papers on Language Universals 9: 1–39. Stanford, CA: Department of Linguistics, Stanford University.] Greenberg, J. 1990[1974]. Studies in numerical system: Double numeral system. In: K. Denning and S. Kemmer, eds., On Language. Selected Writings of Joseph H. Greenberg. Stanford, CA: Stanford University Press, pp. 194–206. [First published 1974 in Working Papers on Language Universals 14: 75–89. Stanford, CA: Department of Linguistics, Stanford University.] Greenberg, J. 1990[1975]. Dynamic aspects of word order in numeral classifier. In: K. Denning and S. Kemmer, eds., On language. Selected Writings of Joseph H. Greenberg. Stanford, CA: Stanford University Press, pp. 227–240. [First published 1974 in Word Order and Word Order Change, ed. C. Li, 27–46. Austin, University of Texas Press.] Grimm, S. 2012. Number and Individualization. PhD Dissertation. Stanford: Stanford University. Hansen, C. 1976. Mass nouns and ‘a white horse is not a horse.’ Philosophy East & West, 26(20), pp. 189–209. Hansen, C. 1983. Language and Logic in Ancient China. Ann Arbor: University of Michigan Press. Haspelmath, M. 2013. Occurrence of nominal plurality. In: M. Haspelmath, M. Dryer, D. Gil, and B. Comrie, eds., The World Atlas of Language Structures Online. Max Planck Digital Library. Available online at http://wals.info/feature/55. Accessed on 2015/2/21. Hass, M. 1942. The use of numeral classifiers in Thai. Language, 18(3), pp. 201–205. Her, One-Soon. 2012a. Distinguishing classifiers and measure words: A mathematical perspective and implications. Lingua, 122(14), pp. 1668–1691. Her, One-Soon. 2012b. Structure of classifiers and measure words: A lexical functional Account. Language and Linguistics, 13(6), pp. 1211–1251. Her, One-Soon, Chen, Jing-Perng, and Tsai, Hui-Chin. 2015. Justifying silent elements in syntax: The case of a silent numeral, a silent classifier, and two silent nouns in Mandarin Chinese. International Journal of Chinese Linguistics 2(2), pp. 193–226. Her, One-Soon and Chen, Yun-Ru. 2013. Unification of numeral classifiers and plural markers: Empirical facts and implications. Presented at the 27th Pacific Asia Conference

Taxonomy of numeral classifiers

69

on Language, Information, and Computation (PACLIC 27), November 22–24, 2013, National Chengchi University, Taipei, Taiwan. Her, One-Soon and Hsieh, Chen-Tian. 2010. On the semantic distinction between classifiers and measure words in Chinese. Language and Linguistics, 11(3), pp. 527–551. Her, One-Soon and Lai, Wan-Jun. 2012. Classifiers: The many ways to profile ‘one’, a case study of Taiwan Mandarin. International Journal of Computer Processing of Oriental Languages, 24(1), pp. 79–94. Her, One-Soon and Lin, Kun-Han. 2015. Fenleici yu liangci de qufen: yi Taiwan Huayu weili (Distinguishing classifiers and measure words: A case study of Taiwan Mandarin). Chinese Linguistics, 4, pp. 56–68. Her, One-Soon and Tsai, Hui-Chin. 2015. On silent elements: A case study of grand and its silent entourage. Natural Language and Linguistic Theory 33(2), pp. 575–605. Hsieh, Miao-Ling. 2008. The Internal Structure of Noun Phrases in Chinese. Taiwan Journal of Linguistics: Book Series in Chinese Linguistics. No. 2. Taipei: Crane Publishing. Hu, Qian. 1993. The Acquisition of Chinese Classifiers by Young Mandarin-Speaking Children. PhD dissertation. Boston: Boston University. Huang, C.-T. James. 1982. Logical Relations in Chinese and the Theory of Grammar. PhD dissertation. Boston: MIT. Huang, C.-T. James, Li, Audrey Y.-H., and Li, Yafei. 2009. The Syntax of Chinese. Cambridge: Cambridge University Press. Huang, Chu-Ren and Ahrens, K. 2003. Individuals, kinds and events: Classifier coercion of nouns. Language Sciences, 25, pp. 353–373. Huang, Shuanfan. 2013. Chinese Grammar at Work. Amsterdam: John Benjamin. Hundius, H. and Kölver, U. 1983. Syntax and semantics of numeral classifiers in Thai. Studies in Language, 7(2), pp. 165–214. Hung, Feng-Sheng. 1996. Prosody and the Acquisition of Grammatical Morphemes in Chinese Languages. Bloomington: Indiana University Linguistics Club. Ionin, T. and Matushansky, O. 2006. The composition of complex cardinals. Journal of Semantics, 23, pp. 315–360. Jiang, Li. 2012. Nominal Arguments and Language Variation. PhD dissertation. Boston: Harvard University. Kayne, R. 2005. Movement and Silence. New York: Oxford University Press. Krifka, M. 1995. Common nouns: A contrastive analysis of Chinese and English. In: G. Carlson and F. Pelletier, eds., The Generic Book. Chicago: University of Chicago, pp. 398–411. Krifka, M. 1998. Class Notes for Semantics I. Austin: University of Texas. Kuo, Jenny Yi-Chun and Jiun-Shiung Wu. 2010. Countability in English and Mandarin. In: Dingfang Shu and K. Turner, eds., Contrastive Meaning in Languages of the East and West. Oxford: Peter Lang, pp. 494–515. Lai, Wan-Jun. 2011. Classifiers in Taiwan Southern Min. MA thesis. Taipei: National Chengchi University. Lehrer, A. 1986. English classifier constructions. Lingua, 68, pp. 109–148. Li, Charles N. and Thompson, S. 1981. Mandarin Chinese: A Functional Reference Grammar. Berkeley: University of California Press. Li, Wendan. 2000. The pragmatic function of numeral-classifiers in Mandarin Chinese. Journal of Pragmatics, 32(8), pp. 1113–1133. Li, Xu-Ping. 2011. On the Semantics of Classifiers in Chinese. Ramat Gan, Israel: bar-Ilan University. Li, Xu-Ping. 2012. Measure readings of Mandarin classifier phrases and the particle de. Language and Linguistics, 13(4), pp. 693–741. Li, Y.-H. Audrey. 2014. Structure of noun phrases – left or right? Taiwan Journal of Linguistics 12.2:1–31.

70

Jiun-Shiung Wu and One-Soon Her

Liao, Jia-Yu. 2014. Identifying Classifiers in Hakka. MA thesis. Taipei: National Chengchi University. Lin, Jo-Wang. 1997. Noun phrase structure in Mandarin Chinese: DP or NP? Language and Linguistics, 3, pp. 401–434. Link, G. 1991. Quantity and number. In: D. Zaefferer, ed., Semantic Universals and Universal Semantics. Berlin: Foris, pp. 133–149. Link, G. 2003[1983]. The logical analysis of plurals and mass terms. In: J. GutierrezRexach, ed., Semantics: Critical Concepts in Linguistics, Vol. III. London: Routledge, pp. 327–349. Löbel, E. 2000. Classifiers versus genders and noun classes: A case study in Vietnamese. In: B. Unterbeck, M. Rissanen, T. Nevalainen, and M. Saari, eds., Gender in Grammar and Cognition. Berlin: Mouton de Gruyter, pp. 259–320. Lucy, J. 1992. Grammatical Categories and Cognition: A Case Study of the Linguistic Relativity Thesis. Cambridge: Cambridge University Press. Lyons, J. 1977. Semantics, 2 vols. Cambridge, UK: Cambridge University Press. Ma, Aimin. 2011. Hanyu jiaoxue zhong ying zhuyi mingliangci de fenlei wenti (Nominal classifiers in Chinese should be properly classified for pedagogical purposes). Xiandai Yuwen (Modern Languages), 2011(3), pp. 37–39. McGregor, R. 1995. Outline of Hindi Grammar, 3rd ed. Oxford: Oxford University Press. McNally, L. 2017. Kinds, descriptions of kinds, concepts, and distributions. In: K. Balogh and W. Petersen, eds., Bridging Formal and Conceptual Semantics. Selected Papers of BRIDGE-14. Düsseldorf: dup, pp. 39–61. Mittendorf, I. and Sadler, L. 2005. Numerals, nouns, and number in Welsh NPs. In: M. Butt and T. King, eds., Proceedings of the LFG05 Conference. Stanford: CSLI, pp. 294–312. Moltmann, F. 1997. Parts and Wholes in Semantics. Oxford: Oxford University. Mou, Bo. 1999. The structure of Chinese language and ontological insights: A collectivenoun hypothesis. Philosophy East and West, 49(1), pp. 45–62. Nevins, A., Pesetsky, D., and Rodrigues, C. 2009. Evidence and argumentation: A reply to Everett (2009). Language, 85(3), pp. 671–681. Nguyen, Dinhhoa. 1957. Classifiers in Vietnamese. Word, 3, pp. 124–152. Nguyen, Tuong Hung. 2004. The Structure of the Vietnamese Noun Phrase. PhD dissertation. Boston: Boston University. Noonan, M. 2003. Chantyal. In: G. Thurgood and L. Randy, eds., The Sino-Tibetan Languages. London and New York: Routledge, pp. 315–371. Partee, B., ter Meulen, A. and Wall, R. 1993. Mathematical Methods in Linguistics. Dordrecht: Kluwer Academic. Quine, W. 1969. Ontological Relativity & Other Essays. Columbia University Press, New York. Robertson, T. 2008. Essential vs. accidental properties. In: Edward Zalta, ed., The Stanford Encyclopedia of Philosophy. Available online at: http://plato.stanford.edu/entries/essential-accidental. Accessed on 2009/10/1. Rothstein, S. 2010. Counting and mass/count distinction. Journal of Semantics, 27, pp. 343–397. Sun, Chaofen. 2006. Chinese: A Linguistic Introduction. Cambridge: Cambridge University Press. Tai, H.-Y. James. and Wang, Lianqing. 1990. A semantic study of the classifier Tiao. Journal of the Chinese Language Teachers Association, 25(1), pp. 35–56. Tai, James H.-Y. 1994. Chinese classifier systems and human categorization. In: William S.-Y. Wang, M. Y. Chen, and Ovid J. L. Tzeng, eds., In Honor of William S.-Y. Wang: Interdisciplinary Studies on Language and Language Change. Taipei: Pyramid Press, pp. 479–494.

Taxonomy of numeral classifiers

71

Tang, Chih-Chen Jane. 2005. Nouns or classifiers: A non-movement analysis of classifiers in Chinese. Language and Linguistics, 6(3), pp. 431–472. Toyota, J. 2009. When the mass was counted: English as classifier and non-classifier language. SKASE Journal of Theoretical Linguistics, 6, pp. 118–130. Toyota, Junichi, Hallonsten, P., and Kovačević, B. 2012. Counting system and classifier/ nonclassifier languages. In: Junichi Toyota, P. Hallonsten, and M. Shchepetunina, eds., Sense of Emptiness: An Interdisciplinary Perspective. Newcastle upon Tyne: Cambridge Scholars Publishing, pp. 128–145. van Riemsdijk, H. 1998. Categorical feature magnetism: The endocentricity and distribution of projections. Journal of Comparative Germanic Linguistics, 2, pp. 1–48. Wang, Hanwei. 2004. Liangci de fenlei yu duiwai hanyu liangci jiaoxue (Classification of classifiers and teaching classifiers in Chinese as a second language). Jinan Xuebao (Journal of Jinan University, Humanities and Social Sciences Edition), 2004(2), pp. 113–116. Wang, Lianqing. 1994. Origin and development of classifiers in Chinese. PhD dissertation. Columbus: Ohio State University. Watanabe, A. 2006. Functional projections of nominals in Japanese: Syntax of classifiers. Natural Language and Linguistic Theory, 24, pp. 241–306. Wei, Pei-Chuan. 2000. Donghan wejin nanbeichao zai yufashi-shang de diwei (The position of the Eastern Han and Six Dynasties in the history of Chinese grammar). Hanxue Yanjiu (Chinese Studies), 18, pp. 199–230. Yang, Xiaoying. 2009. Shuming jiegou de duojiaodu yanjiu (On the Numeral-Noun Construction from Multiple Perspectives). MA thesis. Guangdong, China: Jinan University. Yi, Byeong-Uk. 2009. Chinese classifiers and count nouns. Journal of Cognitive Science, 10, pp. 209–225. Yi, Byeong-Uk. 2011a. What is a numeral classifier? Philosophical Analysis, 23, pp. 195–258. Yi, Byeong-Uk. 2011b. Afterthoughts on Chinese classifiers and count nouns. In: Youngwha Kim et al., eds., Plurality in Classifier Languages. Seoul: Hankookmunhwasa, pp. 265–282. Yi, Byeong-Uk. Unpublished. Numeral classifiers and plural marking: The para-numeral account. Unpublished Manuscript. Department of Philosophy, University of Toronto. Yue, Ann. 2003. Chinese dialects: Grammar. In: G. Thurgood and R. LaPolla, eds., The Sino-Tibetan Languages. London and New York: Routledge, pp. 84–125. Zhang, Hong. 2007. Numeral classifiers in Mandarin Chinese. Journal of East Asian Linguistics, 16(1), pp. 43–59. Zhang, Hongnian, Cheung, Samuel Hung-nin, Liu, Sze-yun, Shih, Li-lin. 2002. A Practical Chinese Grammar. Hong Kong: The Chinese University Press. Zhang, Niina Ning. 2009. Syntactic properties of numeral classifiers in Mandarin Chinese. Talk given on April 10, 2009, at the Graduate Institute of Linguistics, National Chung Cheng University. Zhang, Niina Ning. 2011. The constituency of classifier constructions in Mandarin Chinese. Taiwan Journal of Linguistics, 9(1), pp. 1–50. Zhang, Niina Ning. 2013. Classifier Structures in Mandarin Chinese. Berlin: Mouton de Gruyter.

3

Japanese semantics and the mass/count distinction1 Takashi Iida2

1

Introduction

Recent development of plural logic3 is good news for students of the semantics of languages that have no systematic distinction between singular and plural, such as Japanese. Plural logic gives us a semantical framework in which reference and predication need not be semantically singular. The traditional assumption that they should be semantically singular or reducible to singular constructions has precluded a natural account of the expressions of number-neutral languages, such as Japanese. For example, in the Japanese sentence4 (1)

Kodomo ga waratta. child(ren) NOM laughed ‘A child/The child/Children/The children laughed.’

The noun kodomo may denote a number of children, as well as a single child.5 But there is no expression that corresponds to the noun in symbolic languages standardly used in formal semantics for expressing truth conditions of natural language sentences such as languages of predicate logic, in which all predicates should be singular; that is, they can be true of single things only separately, not jointly. If we stick to such metalanguages, we have to turn kodomo somehow to a singular predicate. The usual way to do this is to interpret kodomo as referring to a set consisting of a number of children or a mereological sum of them.6 It could be argued that this way of proceeding is unsatisfactory.7 Instead of introducing ‘plural objects’ like sets or mereological sums, we can change the logic of our metalanguage and allow plural reference and predication. If we do so, there is no need to introduce any special objects in our ontology; kodomo is not a singular predicate that denotes each of some ‘plural objects’ but a number-neutral predicate that may be true of a number of children as well as a single child. Thus, it seems that plural logic provides a desirable framework for a semantic account of Japanese. There is a question, however, we must settle before we can apply plural logic to Japanese expressions. Plural logic is applicable only to countable predicates, not to non-countable predicates. Consider the following sentence: (2)

Mizu ga koboreta. water NOM spilled ‘Water spilled.’

Japanese and mass/count distinction 73 Mizu (water) does not refer to a single countable object or a number of countable objects. Like the English ‘water’, it is a mass noun. Plural logic by itself does not give us any hint about how to handle such non-countable expressions. In order to apply plural logic to (1), for example, we should first justify the assumption that kodomo in (1), unlike mizu in (2), is a countable predicate. But it has been claimed for several reasons that there is no mass/count distinction in Japanese. If this is a plausible claim, we should give up the hope that plural logic gives us a key to the semantics of Japanese. Three sorts of reasons have been given for denying the mass/count distinction for Japanese nouns: i. As Japanese has no singular/plural distinction and does not have an indefinite article, there is no way of telling the difference between nouns like kodomo (child/children) and mizu (water) without invoking their meaning. ii. In Japanese, the same quantifier expression (e.g., takusan ‘many, much’) are used to express count and mass quantification; for example, takusan no kodomo means ‘many children’, and takusan no mizu means ‘much water’. iii. In Japanese, usually numerals cannot modify nouns without the help of numeral suffixes (NumSuf).8 Thus, the Japanese counterpart ‘three children’, for example, is (3)

san nin three NumSuf ‘three children’

no GEN

kodomo child(ren)

and this has just the same syntactic form as the Japanese expression that means ‘three cans of petroleum’. (4)

san kan no sekiyu three NumSuf GEN petroleum ‘three cans of petroleum’

Thus, even when you wish to assign a number to countable objects (e.g., children), you should use the construction of the form shared by expressions that specify quantities of mass objects: (*) Numeral + NumSuf + no + Noun. I think the last one of these three reasons is the most important. Still it does not give a good reason to deny the existence of a mass/count distinction for Japanese. Thesis (ii) is not correct.9 Not all Japanese quantifier expressions can serve as both count and mass quantifiers. There is an important class of Japanese quantifier expressions that are constructed with the help of so-called indeterminate phrases such as dare (who), dore (which), and dono N (which N). These are applicable only

74

Takashi Iida

to count nouns. Moreover, even among quantity nouns, which include takusan, mentioned above, tasuu (many) and shousuu (a few) are used only for count quantification, whereas tairyou (huge amount) and shouryou (small amount) are used only for mass quantification. As for (i), although the existence of plural forms and an indefinite article is crucial for distinguishing count nouns from mass nouns in English, there may be a totally different way to tell the former from the latter in other languages. And I will argue that the kinds of numeral suffixes matching Japanese nouns in instances of (*) mentioned in (iii) yield criteria for distinguishing Japanese count nouns from mass nouns. This is the reason (iii) is important, although many scholars have cited it as evidence against the existence of a mass/count distinction in Japanese. It provides a clue to the mass/count distinction in Japanese. Though (3) and (4) apparently have a common form, there exists a big difference between them, because the numeral suffixes in them are of different kinds. Lisa Cheng and Rint Sybesma argue that Chinese has two kinds of classifiers, individual classifiers and non-individual classifiers (massifiers), and that this fact shows the existence of mass/count distinction in Chinese (Cheng and Sybesma, 1998). Similarly, I am going to argue that the existence of different kinds of classifiers, or as I call them, ‘numeral suffixes’, in Japanese shows that mass/count distinction exists also in Japanese. But my account of Japanese numeral suffixes differs from their account of Chinese classifiers in two important respects. First, I distinguish three kinds of Japanese numeral suffixes, not two. Second, although Cheng and Sybesma hold that in Chinese, the mass/count distinction exists only at the level of “the type of classifiers used, not the type of noun”,10 I claim that we can single out a class of count nouns in Japanese by attending to the kind of numeral suffix that goes with them and hence that the mass/count distinction exists among nouns as lexical items.

2 Three kinds of numeral suffixes I divide Japanese numeral suffixes into three classes.11 1

Sortal suffixes (classifiers) (A)

for individuals nin tou hon mai

(B)

人 頭 本 枚

(for persons). (for big animals). (for stick-like objects). (for sheet-like objects). ...

for pluralities of individuals kumi 組 soku 足

(for sets or groups of individuals). (for pairs of footwear).

Japanese and mass/count distinction 75 2

Unit-forming suffixes (A)

container type bin kan hako pakku

(B)

partitive type kire teki tsubu

3

瓶 (bottle). 缶 (can). 箱 (box). パック (pack). 切れ 滴 粒

(slice). (drop). (grain).

Measure suffixes kiro meetoru en baito

キロ メートル 円 バイト

(kilogram or kilometer). (meter). (yen). (byte).

It is relatively clear which numeral suffix belongs to the class of measure suffixes. They are nouns that express unit of measurement; kiro is a unit of weight or distance (kiro is ambiguous between the two), meetoru (meter) is a unit of length, and so on. This class is open ended, and a new measure suffix is added to the language whenever some new measurement system is introduced into society. A good example is baito (byte), which became part of common Japanese only in the 1980s. Unit-forming suffixes, especially those of the container type, are also openended. Container-type suffixes stem from common nouns for some kinds of container and can be used independently without being associated with numerals, as the following example shows. (5)

Mizu ga bin ni Water NOM bottle(s) LOC ‘Water is in the bottle.’

haitte-iru. is in

Just like measure suffixes, a new suffix of this kind will be added when a new form of container becomes popular in society. Pakku (pack) is just such a case, and this is shown by the fact that it is written in katakana, which is reserved for a word of foreign origin. Interestingly, nouns for containers that figure as unit-forming suffixes can even be modified by numeral phrases with numeral suffixes of the first kind. In the following sentence, the numeral phrase san ko modifies the preceding container noun hako. (6)

Hako san ko no hon box(es) three NumSuf GEN book(s) ‘There are three boxes of books.’

ga aru. NOM be there

76

Takashi Iida

Unit-forming suffixes of the partitive type may not occur by themselves. But they can be parts of compound nouns such as kami-kire (紙切れ, piece of paper), suiteki (水滴, water drop), and oo-tsubu (大粒, big grain). These compound nouns can also be modified by numeral phrases. (7)

Oo-tsubu san teki no namida ga ochita. Large grain three NumSuf GEN tear NOM rolled down ‘Literally: Three large drops of tears rolled down.’

It is rather rare for a new partitive type unit-forming suffix to emerge, but it does happen. Piisu, a transliteration of English ‘piece’, is a case in point. In contrast to suffixes of the second and third kinds, numeral suffixes of the first kind, which I call ‘sortal suffixes’ or ‘classifiers’,12 form a closed class. They occur only with numerals, either with definite numerals, such as san nin (three persons) in (3), or with indefinite numerals, like suu nin (a few persons), as in the following sentence. (8)

Suu nin no a few NumSuf GEN ‘A few students came.’

gakusei ga student(s) NOM

kita. came

They cannot be a part of a compound noun, unlike unit-forming suffixes. Though nuno kire (piece of cloth) and ko bin (small bottle) are all right, *nuno mai (intention: for sheets of cloth) or *ko nin (intention: for small persons) cannot be recognized as Japanese.13 Suffixes of this kind should be learned one by one; in contrast to unit-forming suffixes, many of which come from nouns with independent meanings, a learner cannot guess what a given sortal suffix is for. Some hold that there is a recent tendency to use the general-purpose suffixes tsu (つ) and ko (個) for a wide variety of nouns.14 If this is correct, we may expect that the number of sortal suffixes in use will decrease in the future, for it seems to be extremely rare for a new sortal suffix to emerge. Nevertheless, sortal suffixes are important to Japanese as a language. Although I claim that a sortal suffix by itself does not contribute to the truth condition of a sentence in which it occurs, this does not mean that the suffix makes no contribution to the meaning of the sentence. A sortal suffix sometimes disambiguates homonyms and almost always conveys an important piece of information through its conventional implicature, or so I will argue later.

3

Numeral phrase modifier bun

Last section characterizes three kinds of numeral suffixes by invoking more or less syntactical features, such as possible contexts of occurrences and productivity. Kobuchi-Philip claims that there is another syntactic feature that distinguishes sortal suffixes from other kinds of numeral suffixes.15 She proposes

Japanese and mass/count distinction 77 a criterion that involves the numerical phrase modifier bun (worth). Consider (9), with the sortal classifier for person(s) nin : (9)

roku six

nin NumSuf

bun worth

no GEN

gakusei student(s)

Compare this with (10) with the unit-forming suffix hako (box) and (11) with the measure suffix pondo (pound): (10) ni hako bun no two NumSuf worth GEN ‘books enough to fill two boxes’

hon book(s)

(11) san pondo bun no niku three NumSuf worth GEN meat ‘meat that amounts to three pounds’ Kobuchi-Philip holds that (9) is ungrammatical, in contrast to (10) and (11), and that this gives us a criterion for distinguishing sortal suffixes from other numeral suffixes. In evaluating Kobuchi-Philip’s claim, we should note that there are two different constructions involving numeral noun phrases and the modifier bun. They are (α)

Num + NumSuf + bun + no + N

and (β)

N′ + Num + NumSuf + bun + no + N,

where N and N′ are common nouns. As N′ may be dropped in the construction (β) if it is clear from the context, it sometimes happens that an expression is ambiguous whether it is an instance of (α) or (β). In fact, though (9)–(11) seem to be of form (α), all of them can be interpreted as having form (β) with some contextually given N′. Take (9). If you are trying to figure out how many teachers are necessary to teach a certain group of students, you might use (9) to mean ‘students that require six teachers to teach’. Although it is a little more difficult to come up with a suitable context, it is possible to interpret (10) or (11) to be of form (β), not (α). For example, the easiest way to think of (10) as an instance of (β) is to interpret it to mean ‘books whose monetary value is equal to that of two boxes of something’. A similar interpretation works for (11), too. If (9)–(11) are thought to exhibit construction (β), then all of them are grammatical and may be meaningfully interpreted. If we think that (9)–(11) exhibit the construction (α), then it is clear that (10) and (11) are grammatical, while (9) is not. Hence, Kobuchi-Philip’s claim is correct, and we may single out sortal classifiers among numeral suffixes by her criterion, as long as we are sure that we are concerned with the construction (α).

78

Takashi Iida

Considering the use of bun also helps to distinguish between the unit-forming numeral suffixes and the measure suffixes. This time, our consideration is semantic in character, because it is concerned with what entailment each kind of numeral suffixes has. Before going into this, it is necessary to look at the different ways a phrase of the form Numeral + NumSuf is positioned with respect to the noun phrase it modifies. It is because the position of a phrase of this form, which I call ‘numeral phrase’,16 as well as that followed by bun in a sentence, may make a difference in the way it should be interpreted. A numeral phrase can occur in three different positions in a sentence. The constructions involving a numeral phrase can be classified according to the position of its occurrence in a sentence. In the following, let Q be a numeral phrase, N a common noun that Q modifies, and cp a case particle ga (NOM) or o (ACC). There are three different ways Q may occur in a sentence. I list them, with one example sentence for each. They all mean that three students came. (I)

‘Q no N’ type

(12) San three (II)

nin NumSuf

no GEN

gakusei student(s)

ga NOM

nin NumSuf

ga NOM

kita. came

nin NumSuf

kita. came

kita. came

‘N Q’ type

(13) Gakusei san student(s) three

(III) ‘N cp Q’ type, or ‘Q N cp’ type (14) Gakusei ga student(s) NOM

san three

(15) San three

gakusei ga student(s) NOM

nin NumSuf

kita. came

I will mostly work with type (III) occurrences, because here we can most clearly see that sortal suffixes (classifiers) differ from the other two kinds of numeral suffixes.

4

Individuated and non-individuated reference of noun occurrences

Unlike the mass/count distinction, the distinction between individuated and nonindividuated references is supposed to apply to particular occurrences of nouns, not nouns as lexical items. An occurrence of a noun N in a context has an individuated reference if its extension in the context of its use is given as consisting of single Ns; in other words, questions like the following should be answerable in principle: What is it to be one N? When is one N the same as another N?

Japanese and mass/count distinction 79 If an occurrence of N has an individuated reference, N has an individuated domain. By contrast, an occurrence of a noun N in a context has a non-individuated reference when its extension in the context of its use is given only as those to which N applies; the only question that is supposed to be answerable is Is/are this/these N or not? If an occurrence of N has a non-individuated reference, N need not have an individuated domain; it may have an individuated domain if N is a noun like hon [book(s)], or it may not have if N is a noun like mizu (water). Let us consider the following sentences: (16) Hon ga san satsu book(s) NOM three NumSuf ‘There are three books.’

aru. be there

(17) Hon ga san hako aru. book(s) NOM three NumSuf (box) be there ‘There are three boxes of books.’ (18) Hon ga san kiro book(s) NOM three NumSuf (kg) ‘There are three kg of books.’

aru. be there

Satsu in (16) is a sortal suffix for book-like objects, hako in (17) is a unit-forming suffix, and kiro in (18) is a measure suffix. There is no need to know the exact number of individual books to decide whether (17) or (18) is true; in the case of (17), you only need to know that each box contains hon, namely, a book or books, and in the case of (18), you only need to know that what weighs 3 kg are hon, that is, a book or books. Thus, in the case of (17) and (18), if you know whether hon applies to a given object or objects, it is enough to decide their truth values. By contrast, in order to know the truth value of (16), you have to know how many books there are, and this requires you to know which object constitutes a single book. In short, the occurrence of hon in (16) is individuated, while it is not so in (17) and (18). In short, if you need to know what it is to be one N in order to understand a sentence in which N occurs, its occurrence has an individuated reference; otherwise, it has a non-individuated reference. This characterization, however, is given in English, and, if N is a Japanese common noun, this does not work, because Japanese expressions corresponding to ‘one N’ would be nothing but gibberish. In general, a numeral cannot come immediately before a noun in Japanese, and for that reason, we don’t have a simple formula like ‘one N’. Although there is a way to state the distinction directly for a Japanese noun, we will be able to explain it only after we have developed some relevant material. For the time being, I should ask you to tolerate the present explanation given in English. The phrase ‘Q bun’ can modify a noun N only when N has a non-individuated reference. This is also the case with a noun-involving bun phrase ‘N´ Q bun’ like gakusei san nin bun (just enough for three students).

80

Takashi Iida

If Q has a unit-forming or a measure suffix, then N which Q applies always has a non-individuated reference. Hence, the occurrences of N in S and S(bun) may have the same non-individuated reference. But if Q has a sortal suffix, the occurrence of N in S must have an individuated reference, whereas in S(bun), it must have a non-individuated reference because ‘Q bun’ can only modify a noun with a non-individuated reference. There are two possibilities: either (1) S(bun) with N does not make sense because the occurrence of N has an individuated reference, or (2) the occurrence of N in S(bun) is turned into that of a non-individuated reference. The latter is what is achieved by a noun-involving bun phrase that occurs in construction (β). Consider the following expressions: (19) gakusei san nin bun no student(s) three NumSuf worth GEN ‘textbooks that are just enough for three students’

kyoukasho textbook(s)

(20) hon san hako bun book(s) three NumSuf worth ‘the weight equal to three boxes of books’

no GEN

omosa weight

bun worth

no GEN

karorii calorie

(21) ringo ni-hyaku guramu apple(s) two hundred NumSuf ‘the calorie equal to 200g of apple’

A noun-involving bun phrase can modify a noun N only when N has a nonindividuated reference, but the noun N' that is a part of the modifying phrase may have an individuated reference as well as a non-individuated one. This means that Q may have a unit-forming or measure suffix as well as a sortal one. In the above examples, (19) has a sortal suffix, (20) a unit-forming one, and (21) a measure one.

5 A simple test for identifying different kinds of numeral suffixes Now I would like to present a test to see which kind a given numeral suffix belongs to. Let S be a sentence of the form Noun + ga + Num + NumSuf + aru/iru.17 and S(bun) a sentence that is just like S except that it has bun right after a numeral suffix that occurs in S. Hence, S(bun) is of the form Noun + ga + Num + NumSuf + bun + aru/iru. You may recognize that S has a type (III) construction, more specifically, an ‘N cp Q’–type construction. For example, if (22) is S, then S(bun) is (22bun). (22)

Ringo ga ni-hyaku guramu apple(s) NOM two hundred NumSuf ‘There are two hundred grams of apple.’

aru. be there

Japanese and mass/count distinction 81 (22bun) Ringo ga ni-hyaku guramu bun apple(s) NOM two hundred NumSuf worth ‘There is apple that weighs two hundred grams in all.’

aru. be there

It must be noted that in some contexts, (22bun) may be an instance of a different construction that is similar to (β) in section 3. N + ga + N′ + Num + NumSuf + bun + aru/iru, meaning that there are apples which are equal in worth with 200 grams of something that is specified in the context. We distinguish the two interpretations of (22bun) as (α) and (β), just as we did in section 3. I will examine for each kind of numeral suffix how S and S(bun) are logically related to each other. My claim is that each kind of suffix gives rise to a specific entailment pattern. 5.1   Unit-forming suffixes I start with unit-forming suffixes. Consider the following pair of sentences. (23)

Ringo ga san hako apple(s) NOM three NumSuf (box) ‘There are three boxes of apples.’

aru. be there

(23bun) Ringo ga san hako bun apple(s) NOM three NumSuf (box) worth ‘There are apples that are enough to fill three boxes.’

aru. be there

(23) is an instance of an S of the specified form with the unit-forming suffix hako, and (23bun) is the bun-inserted variant of S for this S, namely, S(bun). As the English translation makes clear, (23) entails the existence of three boxes as containers of apples, while there is no such entailment with (23bun), provided that (23bun) is not construed as (β) and there is no hidden occurrence of another noun phrase before san hako. Hence, under this assumption, (23bun) does not entail (23). The converse entailment, however, holds. That is, if there are three boxes of apples, then there are apples which are enough to fill three boxes. If (23bun) is construed as (β), it is obvious that there is no entailment in this direction, either. Now, let us consider the case of a partitive suffix. (24)

Ringo ga san kire apple(s) NOM three NumSuf (slice) ‘There are three slices of apple.’

(24bun) Ringo ga san kire apple(s) NOM three NumSuf (slice) ‘There is enough apple for three slices.’

aru. be there bun worth

aru. be there

Suppose that (24bun) is construed as (α), not (β). Then, it is obvious that (24) entails (24bun); if (24) is true, then how can (24bun) be false? Does (24bun) entail (24)? The

82

Takashi Iida

answer is ‘no’: for (24) to be true, there must be actually three separate slices of apple, but (24bun) does not imply that. If (24bun) is construed as (β), then there is no entailment between (24) and (24bun) in either direction. Thus, we can conclude that the following holds for any unit-forming suffix π and any common noun N. If S is a sentence of the specified form with a unit-forming suffix π and a common noun N, then S entails S(bun), but S(bun) does not entail S. 5.2   Measure suffixes Let us turn to measure suffixes. This time, consider the following pair of sentences. (25)

Ringo ga ni-hyaku guramu apple(s) NOM two hundred NumSuf (gram) ‘There are two hundred grams of apples.’

aru. be there

(25bun) Ringo ga ni-hyaku guramu bun apple(s) NOM two hundred NumSuf (gram) worth ‘There are apples that amount to two hundred grams.’

aru. be there

How are (25) and (25bun) logically related to each other? There seems to be no difference between them in truth condition; if (25) is true, then (25 bun) must also be true and vice versa. The same seems to hold with other measure suffixes listed in §2, such as kiro (キロ, either kilogram or kilometer) and en (円, yen). It is essential here that a noun phrase in S not be of type (I) of §3. If S had a noun phrase of type (I), S and S(bun) would not be equivalent. To see this, consider the following pair of sentences. (26)

Ni-hyaku guramu no ringo ga aru. two hundred NumSuf (gram) GEN apple(s) NOM be there ‘There are two hundred grams of apples, or There is/are an apple/apples that weigh(s) two hundred grams each.’

(26bun) Ni-hyaku guramu bun no ringo ga aru. two hundred NumSuf (gram) worth GEN apple(s) NOM be there ‘There are apples that amount to two hundred grams.’ As the English translation of (26) shows, (26) is ambiguous between two readings. (26) under the first reading is equivalent to (26bun), but (26) under the second reading is not. According to this reading, there is an apple which weighs two hundred grams or there are apples each of which weighs two hundred grams. Hence, if there are more than one apple of two hundred grams, then (26) is true, while (26bun) is false. There is no ambiguity in (25) which is similar to (26). In a sentence of type (III), if a noun occurs with a measure or unit-forming suffix, it always has a nonindividuated reference. It is true that we could have chosen S with a noun phrase of type (II) such as the following.

Japanese and mass/count distinction 83 (27)

Ringo ni-hyaku guramu apple(s) two hundred NumSuf (gram) ‘There are two hundred grams of apples.’

ga aru. NOM be there

There is no ambiguity in (27), and it is equivalent to its bun-inserted variant. I have chosen the type (III) construction, because it is suitable for all sorts of numeral noun phrases and does not yield sentences that sound awkward, while some specific cases of the type (II) sound awkward. Let us summarize our result for measure suffixes. If S is a sentence of the specified form with a measure suffix π and a common noun N, then S and S(bun) are logically equivalent. It should be added that, just as in the case of unit-forming suffixes, if (25bun) is construed as an instance of (β), then there is no entailment between (25) and (25bun) in either direction.    suffixes 5.3  Sortal Finally, we consider the case of sortal suffixes, or classifiers. You might expect that this case must be obvious, because I agree with Kobuchi-Philip in holding that bun cannot modify sortal suffixes. The situation is not so simple, because of the existence of a class of nouns with thing/stuff ambiguity. Such a noun sometimes refers to a number of individuals (including just one individual) and sometimes to the stuff that constitutes (parts of) such individuals. Examples are ringo (apple), tamago (egg), maguro (tuna), and matsu (pine). We may call them ‘nouns of thing/ stuff ambiguity’. In most cases, they are names of things found in nature and variously processed by us for our convenience: for our food, clothes, furniture, and so on. You will find many examples of these in your home. If a noun occurring in S has a thing/stuff ambiguity, then S(bun), the bun-inserted variant of S, makes perfect sense; this seems to be impossible if bun cannot modify sortal suffixes without resulting in ungrammaticality. An explanation can be sought in the very fact that a combination of a sortal suffix with bun results in ungrammaticality; if you encounter such a combination, you will try to make sense of it and find a suitable interpretation; such an interpretation can be found in most cases, because it can be construed as an instance of the construction (β), not (α). In the cases of unit-forming or measure suffixes, combinations of numeral suffixes and bun make sense as they are, and it is not necessary to seek another interpretation unless there is a special reason for doing so in the context. We are going to see how such an alternative interpretation is sought for combinations of sortal suffixes and bun, first for nouns without thing/stuff ambiguity and then for nouns with thing/stuff ambiguity. (a) Here is an example of S with a noun that has no thing/stuff ambiguity. (28)

Koin ga san coin(s) NOM three ‘There are three coins.’

ko NumSuf

aru. be there

84

Takashi Iida

(28bun) Koin coin(s)

ga NOM

san three

ko NumSuf

bun aru. worth be there

If we construe (28bun) as an instance of the construction (α), it is ungrammatical. However, (28bun) will not be judged ungrammatical in most cases; instead, it will be regarded as an instance of (β) and interpreted as saying that there are coins which are enough for some three things, because the numeral suffix ko is used for individual things in general. What these three things are should be determined from the context of the utterance; maybe there are enough coins to fill three piggy banks or to buy three rolls. In this example, it is easy to find a suitable interpretation for a combination of a sortal suffix and bun, because the sortal suffix in question is a general-purpose one, namely, ko.18 But if we have a more specialized classifier, then it will not be so easy to find an interpretation. Consider the following. (29)

Ushi ga go cow(s) NOM five ‘There are five cows.’

(29bun) Ushi cow(s)

ga NOM

go five

tou NumSuf

iru. be there

tou NumSuf

bun worth

iru. be there

The classifier tou is used for big animals like cows and horses. In order to make sense of (29bun), we have to interpret it as an instance of the (β) construction and find some big animals such that the cow or cows given in the context are equal in worth in a certain way with five of them. Though it is not impossible to find such an interpretation, it is a little difficult to imagine the contexts in which the utterance of (29bun) is natural. As it is clear that there is no logical entailment in either direction between (28) and (28bun), or, if (29bun) makes sense, between (29) and (29bun) either, we may say the following. If S is a sentence of the specified form with a sortal suffix π and a common noun N that does not have thing/stuff ambiguity, then either (i) S(bun) does not make sense, or (ii) no entailment holds between S and S(bun). (b) Now, consider the sentences like (28) and (28bun), which have tamago (egg) instead of koin (coin). (30)

Tamago ga san egg(s) NOM three ‘There are three eggs.’

ko NumSuf

aru. be there

(30bun) Tamago ga san ko bun aru. egg(s) NOM three NumSuf worth be there ‘There is egg which amounts to (the edible parts of) three individual eggs.’ This time, there is no difficulty in understanding what (30bun) says. Anybody who hears (30bun) will imagine a bowl into which three eggs are broken. Moreover, (30)

Japanese and mass/count distinction 85 entails (30bun); if there are three individual eggs, it is also true that there is some egg stuff which amounts to (the edible parts of) three individual eggs. It is worth noticing that there is another reading of (30bun). According to it, san ko (three things) does not refer to three eggs, but some other three things that can be determined from the context of the utterance. If these three things are cakes, (30bun) says that there is/are enough egg/eggs to make three pieces of cake. If (30bun) is read in this way, it is not entailed by (30). It is clear that in the latter reading, (30bun) is construed as an instance of the construction (β) just as it was in the case of (31). I claim that the former reading is also an instance of (β). In general, if N is a noun having thing/stuff ambiguity, then the sentence S(bun), that is, N + ga + Q + bun + aru is most naturally interpreted as an abbreviated form of Nstuff + ga + Nthing + Q + bun + aru. Thus, the natural interpretation of (30bun) is to read it as (31) Tamagostuff ga tamagothing san ko egg stuff NOM individual egg(s) three NumSuf

bun aru. worth be there

Clearly, this is an instance of (β), and it means that there is egg stuff which amounts to (the edible parts of) three individual eggs. If S has a noun that has thing/stuff ambiguity, its bun-inserted variant S(bun) will be naturally interpreted as an instance of (β) and read like (31). This is not the only interpretation, as we saw in the case of (30bun), but it is always available for a noun with thing/ stuff ambiguity. And in this interpretation, S entails S(bun) because the stuff denoted by the noun occurring in S constitutes the individuals denoted by the same noun.    simple test 5.4  A Finally, let us put together all the considerations discussed above. Let N be a common noun, ν a numeral, π a numeral suffix, and S a sentence of the form N + ga + ν + π + aru/iru Further suppose that S(bun) is a sentence that results from S by inserting bun immediately after π. If we can be sure that S(bun) is not an instance of (β), the three kinds of numeral suffixes are distinguished by the following properties. 1 2 3

If it is sortal, then S(bun) does not make sense. If it is unit-forming, then S entails S(bun) but not vice versa. If it is measure, then S and S(bun) are equivalent.

But if S(bun) can be interpreted as an instance of (β), then there is no such property which distinguishes between the three kinds. In particular, if N that occurs in S is a

Takashi Iida

86

noun with thing/stuff ambiguity, an interpretation of S is always available according to which S is an instance of (β). Thus, I propose the following as a simple test for identifying the different kinds of numeral suffixes. 1

If there is a construal of S(bun) according to which it does not make sense, then π is sortal. If there is a reading of S(bun) according to which S entails S(bun) but not vice versa, then π is unit forming. If there is a reading of S(bun) according to which S and S(bun) are equivalent, then π is measure.

2 3

This test is not a decisive one, because the reason entailment between S and S(bun) holds might have nothing to do with the meaning of the modifier bun; it might be something to do with the meaning of the particular noun N that occurs in S. But I believe this test serves well, at least as a provisional one.

6

Chopsticks, shoes, and twins

Numeral suffixes like zen (膳) and soku (足) seem to offer counterexamples to our test. Consider (32) and (33), with the bun-inserted variants of them: (32)

Hashi ga go zen chopstick(s) NOM five NumSuf ‘There are five sets of chopsticks.’

aru. be there

(32bun) Hashi ga go zen bun aru. chopstick(s) NOM five NumSuf worth be there ‘There are chopsticks just enough for five sets.’ (33)

Kutsu ga go soku shoe(s) NOM five NumSuf ‘There are five pairs of shoes.’

aru. be there

(33bun) Kutsu ga go soku bun aru. shoe(s) NOM five NumSuf worth be there ‘There are shoes just enough for five pairs.’ How are (32) and (32bun) or (33) and (33bun) logically related to each other? It is not unreasonable to think that they are logically equivalent; that is, they have the same truth conditions: if one is true, then the other must also be true. At the same time, it is almost certain that they have different implications; even if they have the same truth condition, there must be some reason why a speaker chooses to utter one and not the other. In what situation is it natural to utter (32bun) instead of (32)? We may imagine someone who counts a bunch of chopsticks that are alike and utter (32bun) after having found out that there are ten of them. But, do we think (32) is not true in this situation? I suppose not. If there are ten chopsticks that can be paired to each other to form five sets of them, there is no denying that there are five sets of chopsticks.

Japanese and mass/count distinction 87 If we hear the utterance of (32), on the other hand, we just think of five sets of chopsticks that may be different from each other in various ways. Still it is obvious that (32bun) is also true when (32) is true. If (32bun) may sound strange in this situation, it is because (32bun) lacks an extra piece of information that (32) suggests, namely, that the chopsticks are not only enough to form five sets but are already arranged as those sets; hence, uttering (32bun) after (32) violates one of the Gricean maxims of conversation (the Maxim of Quantity). But this does not mean that there is a difference in truth condition between them. Thus, we must conclude that (32) and (32bun) are logically equivalent. Similarly, we can see that (33) and (33bun) are logically equivalent. Then, in our test, numeral suffixes zen and soku must be measure ones. Can we accept this conclusion? Is there any reason to think that zen and soku are measure suffixes? One might defend this view in the following way. For one thing, we have numeral suffixes for individual chopsticks and individual shoes. For individual chopsticks, we use a sortal suffix hon (本), which generally applies to long and narrow things. Similarly, for individual shoes, we use a sortal suffix ko (個), which applies to material objects in general. Thus, hashi ichi zen chopstick(s) one NumSuf ‘one set of chopsticks’ consists of hashi ni chopstick(s) two ‘two chopsticks’

hon NumSuf

If we already have the sortal suffix hon for chopsticks, then why do we need another one for sets of chopsticks? Hence, zen must express a measure. But there is some reason not to regard zen and soku as measure suffixes. Although zen (膳) can be used by itself, not as a part of a numeral noun phrase, in such uses, it is a common noun for a tray of food, not a sort of measure. This is even more obvious with soku (足). It never appears outside of a numeral noun phrase. Now I am going to argue that zen and soku are sortal suffixes and that the logical equivalence between (32) and (32bun), or (33) and (33bun), does not affect the validity of our test for different kinds of numeral suffixes. First, let us remember that what we count is not limited to single things; we also count combinations of things like coffee sets or groups of people like couples and families. Kumi (組) is a general-purpose sortal suffix for such combinations or pluralities. Thus, we say: (34)

Fuufu ga san couple(s) NOM three ‘There are three couples.’

kumi NumSuf

iru. be there

Kumi is a sortal suffix according to our test, and this is shown by the fact that its bun-inserted variant (34bun) does not make sense, or there is no entailment

88

Takashi Iida

relation between (34) and (34bun) when the latter is interpreted as an instance of (β). (34bun) Fuufu couple(s)

ga NOM

san three

kumi NumSuf

bun worth

iru. be there

In what sort of circumstances does an utterance of (34bun) make sense? One possibility is the situation in which we are trying to form three groups of people, and a couple or couples are somehow necessary for them. There can be many cases. It might be stipulated that each of the three groups should have at least two couples. In another, it might be stipulated that each group should have at least one person who belongs to a couple. In the former case, the truth of (34) is not sufficient for that of (34bun). In the latter, (34) may not be true even though (34bun) is true, for the existence of two couples would be enough to satisfy the stipulated condition. Second, some nouns may refer to either a plurality or its member. Futago [twin(s)] is a case in point. Whether a noun refers to a plurality or its member is marked by the presence or absence of plural endings in English. In languages whose nouns do not have singular or plural forms, such as Japanese, however, the number difference must be marked in some other ways. And one of the ways is to use different sortal suffixes, as in the following sentences. (35)

Futago ga san kumi twin(s) NOM three NumSuf ‘There are three pairs of twins.’

iru. be there

(36)

Futago ga roku twin(s) NOM six ‘There are six twins’

iru. be there

nin NumSuf

Third, hashi [chopstick(s)] and kutsu [shoe(s)] are like futago [twin(s)] in that they may refer to either a plurality or its member. Hence, considering a parallel example with futago [twin(s)] might help us to clarify the situation. Consider the following sentence, which is the bun-inserted variant of (35). (35bun) Futago twin(s)

ga NOM

san three

kumi NumSuf

bun worth

iru. be there

As futago may refer either to pairs of twins or individual twins, (35bun) has two interpretations. If futago refers to pairs of twins, (35bun) means something similar to what (34bun) means, namely, that there are pairs of twins, which are just enough for three combinations of some sort. In this interpretation, it is obvious that (35bun) is construed as an instance of (β). If futago refers to individual twins, it means that there are individual twins who are just enough for three combinations of some sort. The obvious candidate for such a combination is a pair of twins; hence, (35bun) means in this interpretation that there are enough individual twins to form three pairs of twins. It is obvious again that this interpretation construes (35bun) as an instance of (β).

Japanese and mass/count distinction 89 We may now see what is responsible for the apparent contradiction. It is not our test but the ambiguity of the noun futago [twin(s)]. Those who think that (35) and (35bun) are logically equivalent must interpret futago in (35bun) as referring to individual persons, whereas futago in (35) refers to pairs. But (35bun) has in fact two readings: the one in which futago refers to individual persons and the one in which it refers to twins as pairs, just as fuufu [couple(s)] refers to couples as pairs in (34bun). As futago in (35) refers to pairs, the second interpretation is the right interpretation for our test, and either (35bun) does not make sense if it is construed as an instance of (α), or it is logically independent from (35) if it is construed as an instance of (β). Now it must be obvious why our test seemed to fail with zen and soku. Like kumi, they are sortal suffixes that apply to pluralities only. Moreover, the common nouns to which these suffixes are attached can refer to members of a plurality as well as pluralities themselves. Hashi [chopstick(s)] may refer to individual chopsticks as well as matched pairs of them; similarly, kutsu [shoe(s)] may refer to individual shoes as well as matched pairs of them. Hence, if (32) and (32bun) are thought to be logically equivalent, then the occurrences of hashi in two sentences refer to different things; in (32), it refers to matched pairs of chopsticks, while in (32bun), it refers to individual chopsticks. If we do not want to have different references in the two sentences, then we should interpret (32bun) in such a way that hashi refers to matched pairs of chopsticks, not individual chopsticks. Then, either (32bun) does not make sense, or it can be construed as an instance of (β); if the latter is the case, it is not difficult to imagine a scenario in which (32) and (32bun) are not equivalent. The situation is very similar to the case of nouns that have thing/stuff ambiguity. Let N be a noun like hashi and futago that may refer to pluralities or their members and Q a numeral noun phrase; then the sentence of the form N + ga + Q + bun + aru/iru is most naturally interpreted as an instance of (β) of the form Nindividual + ga + Nplural + Q + bun + aru/iru. Thus, the natural interpretation of (32bun) is Hashiindividual ga

Hashiplural

go

zen

bun

aru,

which means there are individual chopsticks that are enough to form five matched pairs of chopsticks.

7 What does a sortal suffix contribute to the meaning of a sentence? We may conclude that zen and soku are sortal suffixes. This conclusion does not make our test for different kinds of suffixes invalid. It is a hidden ambiguity in such nouns as hashi [chopstick(s)] and kutsu [shoe(s)] that makes it seem that our

90

Takashi Iida

test gave a wrong verdict. These nouns may refer to either individuals or pluralities of them. If so, an interesting question arises. As we have noticed, different sortal suffixes are used depending on whether a noun refers to individuals or pluralities. The noun hashi requires hon when it refers to individual chopsticks, while it requires zen when it refers to matched pairs of them. The question is whether a sortal suffix like hon and zen has a truth conditional content. It is important to address this question, because it is frequently assumed that sortal suffixes, unlike other kinds of numeral suffixes, do not contribute to the truth conditions of sentences in which they occur. Eric McCready (2012, p. 148) considers the question, with special emphasis on the case of zen. And he concludes that “classifiers make a dual contribution to meaning: A conventionally implicated domain restriction, and an individuation of objects for quantification that takes in the truth-conditional domain”. I believe that there is an alternative way to construe the function of a sortal suffix; a sortal suffix does not directly contribute to the truth condition of a sentence in which it occurs but helps to disambiguate the noun it attaches to through its conventional implicature. In this view, we do not need to suppose that a sortal suffix (a classifier) has any truth-conditional content. I defend this alternative in this section. There are some clear cases in which a particular choice of a sortal suffix helps to disambiguate an expression. For example, kami might mean either paper (紙) or hair (髪). Although the two senses can be easily distinguished in written Japanese, as different Chinese characters (kanji) are used for them, it may be difficult to distinguish them in spoken language in spite of the difference in accent. But if they are accompanied by a sortal suffix as in the following, there will be little chance of misunderstanding them. (a)

kami san mai paper three NumSuf ‘three sheets of paper’

(b)

kami san hair three ‘three hairs’

bon NumSuf

The case of kami is that of pure homonymy; kami in the sense of paper and kami in the sense of hair are different words that have similar pronunciations by a pure accident. There is no reason to think that there is a word that has a single extension that encompasses both paper and hair and that different sortal suffixes restrict the word to one or the other. When kami is used in a sentence, it is used by itself to refer to either paper or hair, and a sortal suffix is chosen accordingly. A sortal suffix indicates which word occurs in a sentence. The cases of pure homonymy are rare compared to those in which a single word has different senses that are connected to each other in a certain way. Such cases are sometimes called ‘paronymy’. In Japanese, a sortal suffix also helps to identify one particular sense among those that are not only associated with one single word

Japanese and mass/count distinction 91 but also related to each other. For example, ringo, just like the English ‘apple’, may mean either fruits or trees. Sometimes it is the accompanying sortal suffix that makes it clear which sense is at issue. In (c), ringo refers to fruits, while it refers to trees in (d). (c)

ringo san apple(s) three ‘three apples’

ko NumSuf

(d)

ringo san bon apple(s) three NumSuf ‘three apple trees’

In this case, too, I think it is better to construe an occurrence of ringo as referring to either fruits only or trees only rather than referring to both of them indiscriminately; a sortal suffix is, as it were, chosen after the preceding decision about the noun’s reference. Consider the following sentence. (37) Ringo ga tasuu aru. apple(s) NOM many be there ‘There are many apples/apple trees.’ Because of the occurrence of tasuu (many), ringo in (37) must have an individuated reference.19 Still, (37) is ambiguous; it may mean that there are many apple fruits, or it may mean that there are many apple trees. But it cannot mean that there are many things that are either apple fruits or apple trees.20 The same holds for hashi [chopstick(s)] and kutsu [shoe(s)]. It is not that a sortal suffix zen individuates a domain that is not yet fully individuated but that its presence helps to determine which one of the two fully individuated domains is the right one for the occurrence of hashi at issue. In general, I think, when a sortal suffix is attached to a noun, the noun has an associated domain that is already fully individuated. A sortal suffix is chosen according to this individuated domain, and it gives the hearer a hint for that domain, which sometimes has the effect of disambiguation. In order to explain how a sortal suffix can do so, we must be clear about its contribution to the meaning of a sentence in which it occurs. McCready argues quite successfully that the contribution should be a conventional implicature. He invites us to consider the following sentence.21 (38) # Otoko man/men

ga ni satsu NOM two NumSuf

haittekita. entered

As McCready tells us, this sentence is weird, but not false; instead, it is just inappropriate, because the classifier satsu is used for books. In such a case, we may first suspect that it might be a case of presupposition failure. The standard behavior of presuppositions are (i) its failure induces truth value gaps, (ii) escaping from presupposition ‘holes’ such as negation and modal operators, and (iii) exhibiting

92

Takashi Iida

‘binding’ behavior: if a sentence S carries presupposition P, and S´ entails P, then no presupposition is projected in ‘If S´, then S’. It is easily seen that the case of a sortal suffix satisfies (i) and (ii) but not (iii). Hence, it cannot be a case of presupposition failures.22 It cannot be a case of conversational implicature, either. For one thing, what the sortal suffix satsu indicates cannot be canceled.23 There is one remaining option, the case of conventional implicature. Our case satisfies all three of its main properties: (i) it is triggered by particular lexical items or constructions, (ii) it is scopeless, and (iii) it cannot be ‘bound’ by presupposition. 24 Thus, we may say that the ‘meaning’ of (38) has two components: truth-conditional content: that two men entered, and conventional implicature: that men are book-like objects. Now we can see that the weirdness of (38) comes from the conventional implicature it has: its falsity makes (38) inappropriate. All this is very clear, and I am in complete agreement with it. Moreover, I want to claim that, once we recognize the ambiguity in nouns like hashi [chopstick(s)] and kutsu [shoe(s)], this is enough to explain how sortal suffixes like zen and soku work, and, pace McCready, there is no need to suppose that they “make a dual contribution to meaning”. Consider the sentence (39) Ringo apple(s)

ga NOM

go five

hon NumSuf

aru. be there

As remarked above, the occurrence of ringo must have an individuated domain because of the presence of the sortal suffix hon. But ringo may mean either fruits or trees. Thus, (39) is ambiguous, at least25 between (39a) Ringofruit

ga

go

hon

aru.

ga

go

hon

aru.

and (39b) Ringotree

Because the sortal suffix hon is used for long and slender things, (39a) and (39b) have, respectively, the following conventional implicatures. (CI: 39a) Apple fruits are long and slender things. (CI: 39b) Apple trees are long and slender things. (CI: 39a) is obviously false. This means that if the occurrence of ringo in (39) is interpreted as in (39a), then (39) will be inappropriate, at least. Hence, we may conclude that (39) should be interpreted as (39b).26

Japanese and mass/count distinction 93 Just as the conventional implicature of the sortal suffix hon contributes to disambiguate ringo [apple(s)], the conventional implicature of zen does the same with respect to hashi [chopstick(s)]. Consider again (32): (32) Hashi chopstick(s)

ga NOM

go five

zen NumSuf

aru. be there

Just as ringo is ambiguous about whether it refers to apple fruits or apple trees, hashi is ambiguous about whether it refers to individual chopsticks or matched pairs of them. Thus, theoretically, (32) has two readings. (32a) Hashiindividual (32b) Hashimatched_pair

ga ga

go zen go zen

aru. aru.

Just as before, each has the following conventional implicatures. (CI: 32a) Each chopstick is a pair. (CI: 32b) Each pair of matched chopsticks is a pair. As the former is obviously false and the latter is obviously true, it is immediately apparent that (32b) is the right interpretation. In this way, the conventional implicature of the sortal suffix zen makes it possible to find the individuated domain for hashi in (32). We do not need to suppose that the sortal suffix directly operates on a not yet completely individuated domain for hashi. Thus, I conclude that the semantic contribution of a sortal suffix to the meaning of a sentence does not extend to its truth condition but is confined to its conventional implicatures. This does not mean, however, that a sortal suffix’s semantic contribution is not important in interpreting a sentence in which it occurs. We have just seen that exactly the opposite is true.

8

Conclusion

Our discussion so far allows us to make three claims: I. II. III.

We can distinguish a class of sortal suffixes from other kinds of numeral suffixes with a simple test. A noun with a matching sortal suffix should have an individuated domain. A sortal suffix does not have truth-conditional content.

If they are true, then they make it possible for us to do two things. First, we can draw the individuated/non-individuated distinction in such a way that uses only the resources available in Japanese. If S is a sentence in which a common noun N occurs, and there is a sortal suffix π that is appropriate for N, then it makes sense to ask what it is that is called

94

Takashi Iida (%)

ichi/hito27 one

π

no N. GEN

For example, we can ask what it is that is called (e) it- tou no one NumSuf (for big animals) GEN

ushi, cow

or (f) hitoone

ri no NumSuf (for persons) GEN

gakusei. student

By III, we know that sortal suffix π has no truth-conditional content. This means that (%) corresponds to English ‘one N’ and that (e) and (f) correspond to ‘one cow’ and ‘one student’, respectively. Hence, we can go metalinguistic and, instead of asking what it is to be one N, ask what it is called ‘ichi/hito π no N’ for any noun N and a sortal suffix π appropriate for N. Thus, we can characterize the individuated/non-individuated reference for an occurrence of a Japanese common noun N in a sentence S in this way: if there is a sortal suffix π appropriate for N, and we need to know what it is to be called ‘ichi/hito π no N’ in order to understand S, then the occurrence of N has an individuated reference; otherwise, it has a nonindividuated reference. This means that plural logic, which is just an extension of the standard logic, can be applied to those noun occurrences with individuated references. Second, our results can serve as a basis for singling out count nouns from common nouns in Japanese and regarding the rest of them as non-count nouns. We have been working with a distinction between individuated and non-individuated references. It is a particular occurrence of a noun that has an individuated or nonindividuated reference. The distinction can be also drawn among nouns as lexical items. By II and III, the individuation of the domain must be achieved by the modified noun itself. This means that a noun that typically occurs with a sortal suffix has an individuating force by itself. It is the main characteristic of a count noun. Moreover, by I, we can single out a class of such count nouns with a simple test. However, we cannot define a count noun simply as a common noun that can be modified by a numeral phrase with a sortal suffix. If this were the right definition, then almost all of the Japanese common nouns would be count nouns. The main reason such a simple definition does not work lies in the existence of sentences like the following. (40) Biiru ga san bon Beer NOM three NumSuf ‘There are three bottles of beer.’

aru. be there

(41) Jyamu ga san ko Jam NOM three NumSuf ‘There are three cans/jars of jam.’

aru. be there

Japanese and mass/count distinction 95 Though a numeral phrase san bon (san ko) with the sortal suffix hon (ko) seems to modify biiru (jyamu) that immediately precedes it, in reality, it modifies some noun for a container that is implicitly understood in the context, and the occurrence of biiru in (40) [jyamu in (41)] does not have an individuated reference. Hence, such cases should be excluded if we wish to characterize a count noun as a common noun taking a certain kind of numeral phrase. Moreover, the sortal suffix tsu, which is called a ‘general classifier’ along with ko, which occurs in (41), can be used with a wide variety of common nouns; some of its occurrences cannot be regarded as having an individuated reference. We may argue that the ‘general classifier’ tsu is sometimes used as a container suffix like hon/bon and ko and that the noun occurrences that are modified by such uses of tsu do not have an individuated reference. Hence, I propose the following as a characterization of a count noun in Japanese. A common noun N is a count noun if and only if N can be modified by a numeral noun phrase with a sortal suffix that is not used as a container suffix.

Notes 1 I gave two talks that were based on previous versions of this chapter: one at Workshop on the Semantics of Nouns, Quantifiers, and Classifiers, held at Kyung Hee University, Seoul, on July 18th of 2015, and another at Peking University, Beijing, on September 15th of 2015. On both occasions, I learned much from comments and questions from the audience. I thank the audiences of the talks. In particular, I would like to thank Professor Byeong-uk Yi of the University of Toronto for organizing the wonderful workshop and Professor Chen Bo of Peking University for inviting me to give the talk. I thank Professor Yi also for detailed comments and discussions, to which the present version owes a great deal. 2 Professor Emeritus of Keio University, Tokyo, Japan. E-mail: [email protected] 3 (McKay, 2006), (Oliver and Smiley, 2001), (Oliver and Smiley, 2013), (Yi, 1999), (Yi, 2005), and (Yi, 2006). 4 Here is a list of abbreviations used in this article. NOM: nominative, GEN: genitive, ACC: accusative, LOC: locative, NumSuf: numeral suffix. 5 The occurrence of kodomo here might be a definite one as well. Nothing depends on whether kodomo is definite or indefinite, however, as far as the point in discussion is concerned. 6 See, for example, (Landman, 2000), (Lasersohn, 1996), and (Schwarzschild, 1996). 7 See the works cited in note 3. 8 When a definite collection of individuals is intended, a numeral can directly modify a noun without any numeral suffix, as is shown by san baka (three fools), roku daigaku (six universities), and shichi kenjin (seven sages). 9 Also see (Kobuchi-Philip, 2011, pp. 306–309). 10 (Cheng and Sybesma, 1998, sec. 2.3). 11 There are at least two Japanese terms for numeral suffixes. Masuoka and Takubo use jyo-suu-ji (助数辞) in (Masuoka and Takubo, 1993), while Iida, Asako uses jyo-suu-shi (助数詞) in (Iida, 2004) and (Iida, 2005). And scholars use these terms with different extensions. Iida, Asako excludes measure suffixes from jyo-suu-shi, and Masuoka and Takubo include them among jyo-suu-ji. Moreover, they do not explicitly distinguish between sortal and unit-forming suffixes; unit-forming suffixes are listed in (Iida, Asako, 2004), but they are not distinguished from sortal ones, while they are not encountered among the examples of jyo-suu-ji listed in (Masuoka and Takubo, 1993). For various terminologies for numeral suffixes, see (Yi, 2011).

96

Takashi Iida

12 It frequently happens that numeral suffixes in general are called ‘classifiers’. But this name is appropriate only for sortal suffixes; hence, I use ‘classifiers’ only for them. 13 (Cheng and Sybesma, 1998) make a similar observation about Chinese classifiers (sec. 2.2). 14 See, for example (Iida, Asako, 2005, p. 36). But see (Onodera, 2014) for a different view. 15 See example (46a) of (Kobuchi-Philip, 2011). (Watanabe, 2006) also contains the claim that bun cannot co-occur with a classifier; see his example sentence (40). 16 In Japanese grammar, an expression of the form ‘Numeral + Numeral Suffix’ such as san tou is regarded as a noun phrase, because (i) like san tou no ushi (three cows), it forms a complex noun phrase with a common noun connected by the case particle no, which is used for connecting two noun phrases, and (ii) like san tou da in the sentence Ushi wa san tou da (There are three cows), it forms a predicate with a copula da. See (Masuoka and Takubo, 1993, pp. 34–36). 17 Japanese has two verbs, aru and iru, for expressing existence. Very roughly speaking, iru is used for persons and animals, whereas aru is used for inanimate things like desks and stones; there is much discussion about the exact distribution of them; for the purpose of this chapter, however, there is no need to go into this controversial issue. 18 Strictly speaking, this is not true; there are things like trees and sheets to which ko does not apply. See (Iida, Asako, 2004, pp. 345f.) and (Iida, Asako, 2005, Ch. 4). 19 If tasuu (many) is replaced by takusan (many, much), then ringo may be used in its stuff sense as well; hence, the resulting sentence would be ambiguous in four ways, for ringo may refer to apple wood as well as a kind. 20 Pelletier (2012) suggests that every noun has one comprehensive domain that includes any kind of things it may refer to. But if it were so, I wonder why (37) could not mean the numerousness of both apple fruits and apple trees. (37) means either there are many apple fruits or there are many apple trees, not both. 21 (McCready, 2009, p. 199) and (McCready, 2012, p. 144). I have changed the number of the example sentences. 22 (McCready, 2009, pp. 199–201) and (McCready, 2012, pp. 144–146). 23 (McCready, 2012, p. 146). 24 (McCready, 2009, pp. 201f). and (McCready, 2012, pp. 146f). 25 There is another possible reading of (39). It is to interpret hon as a container suffix. Then (39) will mean something like there are five bottles of apple juice. This reading may be safely ignored here. 26 Even though (39a) is wrong in that its conventional implicature is false, we may make sense of its assertion in some circumstances. For example, if it is found that there are five apple fruits at the place that is relevant to the conversational context, we may suppose that hon was just a slip of the tongue and the speaker wanted to say that there were five apple fruits. But, of course, this does not make (39a) appropriate. 27 For numbers less than ten, there are two kinds of numerals in Japanese, native ones and Chinese ones; hito and ichi are both numerals for one, the former a native one and the latter a Chinese one. Only native ones co-occur with the classifier tsu.

References Cheng, L. and Sybesma, R. 1998. Yi-wang Tang, Yi-ge Tang: Classifiers and Massifiers. Tsing Hua Journal of Chinese Studies 28. pp. 385–412. Iida, Asako. 2004. Kazoe-kata No Jiten (Dictionary of Ways of Counting). Tokyo: Shogakukan. Iida, Asako. 2005. Kazoe-kata De Migaku Nihon-go (Brushing Up Your Japanese by Mastering Ways of Counting). Tokyo: Chikuma Shobo. Kim, Young-Wha et al. 2011. Plurality in Classifier Languages. Plurality, Mass/Kind, Classifiers and the DPs. Seoul: Hankookmunhawasa Publishing Company. Kobuchi-Philip, M. 2011. The Mass Hypothesis in Japanese. In: Kim 2011. pp. 283–321. Landman, F. 2000. Events and Plurality. Dordrecht: Kluwer Academics.

Japanese and mass/count distinction 97 Lasersohn, P. 1996. Plurality, Conjunction and Events. Dordrecht: Kluwer Academics. Masuoka, Takashi and Takubo, Yukinori. 1993. Kiso Nihongo Bunpou (Basic Japanese Grammar). 2nd ed. Tokyo: Kuroshio Publishing. McCready, E. 2009. Conventional Implicature and Classifiers. Journal of Cognitive Science 10. pp. 195–208. McCready, E. 2012. Classification without Assertion. Proceedings of FAJL 5: Formal Approaches to Japanese Linguistics (MIT Working Papers in Linguistics #64). pp. 141–154. McKay, T. 2006. Plural Predication. Oxford: Oxford University Press. Oliver, A. and Smiley, T. 2001. Strategies for a Logic of Plurals. Philosophical Quarterly 51. pp. 289–306. Oliver, A. and Smiley, T. 2013. Plural Logic. Oxford: Oxford University Press. Onodera, Jyuri. 2014. Jyo-Suushi Tsu No Shiyou Han-I Ni Kansuru Ichi Kousatsu (On the Range of Use of the Numeral Suffix tsu). Gengo to Bunmei (Reitaku University) 12. pp. 91–111. Pelletier, F. 2012. Lexical Nouns Are Both +MASS and +COUNT, but They Are Neither +MASS nor +COUNT. In: D. Massam, ed., Count and Mass Across Languages. Oxford: Oxford University Press. pp. 9–26. Schwarzschild, R. 1996. Pluralities. Dordrecht: Kluwer Academics. Watanabe, Akira. 2006. Functional Projections of Nominals in Japanese: Syntax of Classifiers. Natural Language & Linguistic Theory 24. pp. 241–306. Yi, Byeong-Uk. 1999. Is Two a Property? Journal of Philosophy 95. pp. 163–190. Yi, Byeong-Uk. 2005. The Logic and Meaning of Plurals. Part I. Journal of Philosophical Logic 34. pp. 459–506. Yi, Byeong-Uk. 2006. The Logic and Meaning of Plurals. Part II. Journal of Philosophical Logic 35. pp. 239–288. Yi, Byeong-Uk. 2011. What Is a Numeral Classifier? In Kim 2011. pp. 1–51.

4

A continuum-based approach to the count-mass distinction in Korean* Kwang-sup Kim

1

Introduction

Korean nominal expressions can be classified into two types, depending on the compatibility with numerals and the so-called plural marker -tul.1 The first type is compatible with both of them, as shown in (1a–b),2 whereas the second type is compatible with neither of them, as in (2–3). (1)

a. b.

(2)

a. b.

(3)

a. b.

sey haksayng three student Kim kyoswu-ka haksayng-tul-ul Kim professor-NOM student-Tul-ACC ‘Prof Kim met his/her students’ *sey sakwa three apple *Susie-ka sakwa-tul-ul Susie-NOM apple-Tul-ACC ‘(intended) Susie ate apples’ *sey mwul three water *Susie-ka mwul-tul-ul Susie-NOM water-Tul-ACC ‘(intended) Susie drank water’

manna-ss-ta. meet-PAST-DEC3

mek-ess-ta. eat-PAST-DEC

masi-ess-ta. drink-PAST-DEC

The first type – the student-type – is by and large similar to English count nouns in the sense that they can co-occur with cardinal numbers without the help of classifiers. However, they are quite peculiar in that (i) they are not compatible with big numbers, and (ii) they can co-occur with classifiers, as shown in (4–5). (4)

a. b.

(5)

a.

??payk hundred ??i two

haksayng student payk haksayng hundred student

haksayng sey student three

myeng CL

‘a hundred students’ ‘two hundred students’

‘three students’

The count-mass distinction in Korean b. c.

haksayng student haksayng student

payk hundred i two

myeng CL payk hundred

myeng CL

99

‘a hundred students’ ‘two hundred students’

The second type, just like English mass nouns, obligatorily requires classifiers when they co-occur with cardinal numbers, as in (2a), (3a) and (6a–b). (6)

a. b.

sakwa apple mwul water

sey three sey three

kay CL can glass

Interestingly, this type is further divided into two sub-types. The apple-type is distinguished from the water-type with regard to the compatibility with the socalled plural marker -tul. As shown in (2b) and (3b), both the apple-type and the water-type produce ill-formed sentences when bare nouns with -tul occur in the object position. However, the grammaticality of the apple-type can be ameliorated via modification, whereas that of the water-type cannot, as illustrated in (7a–b). (7)

a. Susie-ka

[yeki iss-te-n] Susie-NOM [here be-Retro-Adnominal]

sakwa-tul-ul (ta) mek-ess-ta. apple-Tul-ACC (all) eat-PAST-DEC

‘Susie ate (all) the apples that were here’ b. *Susie-ka [yeki iss-te-n] mwul-tul-ul

(ta) masi-ess-ta. Susie-NOM [here be-Retro-Adnominal] water-Tul-ACC (all) drink-PAST-DEC

‘(intended) Susie drank (all) the water that was here’

In fact, the apple-type, unlike the water-type, can be compatible with -tul even if they are not modified, as long as they co-occur with locative phrases.4 (8)

a. b.

[Yeki ceki-ey] sakwa-tul-i [Here there-at] apple-Tul-NOM ‘Apples were fallen here and there’ [Yeki ceki-ey] mwul(*-tul-)i [Here there-at] water(*-Tul-)NOM ‘Water was sprinkled here and there’

ttelece-iss-ess-ta. fallen-be-PAST-DEC ppulyece-iss-ess-ta. sprinkled-be-PAST-DEC

This chapter attempts to explore the possibility of explaining the peculiarities of the three types of nominals. In this chapter, I provide a continuum-based approach to the count-mass distinction. After rejecting the traditional view that if a noun is not a count noun, it is a mass noun and vice versa, I propose that one type can be more or less countable than the other type: that is, there are many different types between the countable type and the non-countable type. More specifically, it will be claimed that the apple-type is more countable than the water-type and less countable than the student-type, which is in turn less countable than English count nouns.

100

2

Kwang-sup Kim

Syntactic differences between count and mass nouns

There are two possible approaches to the count-mass distinction. One is to assume that nouns, regardless of whether they are count nouns or mass nouns, enter the derivation as an undivided mass, as Borer (2005) proposes.5 The other is to assume that count nouns, unlike mass nouns, enter the derivation as a set of individual members. Whatever approach we may adopt, we need to explain the selectional differences between count nouns and mass nouns.6 The former type must be merged with at least one functional category, whereas the latter may be merged with no functional category. (9) a. *Dog is in the yard. b. Dogs are in the yard. c. A dog is in the yard. (10) Water is in the fridge. This section attempts to explain this contrast while assuming that the count noun denotes a set of individual members, whereas the mass noun denotes undivided stuff. In this section, I claim that this semantic difference gives rise to syntactic differences in that the former is projected into NumberP and DP, whereas the latter is not. This claim will be shown to explain the contrast between (9) and (10), and furthermore, it will be shown to provide a clue to the puzzles introduced at the outset of this chapter. 2.1  Number    phrase and the count-mass distinction I propose that the nominals that denote sets of individual members have an unvalued number feature; hence they need to be merged with a lexical item that can fix the value of the number feature. As illustrated in (11–12), the count noun dog has an unvalued number feature – [x number]; if it is merged with the plural morpheme -s, it is valued as [+PL]; and if combined with the zero singular morpheme ø-pl, it is valued as [-PL]. (11) a. [dog [x number]]: Merger with s, and Valuation b. [Num(ber)P S[+PL] [dog [+PL number]]]: Lowering of -s onto dog at PF c. [NumP [dog [+PL number]] S[+PL]] (12) a. [dog [x number]]: Merger with Ø-PL, and Valuation b. [NumP Ø-PL [dog [-PL number]]]: Lowering of Ø-PL onto dog at PF c. [NumP [dog [-PL number] Ø-PL]] As shown in (11–12), the two morphemes -s or ø-PL are the heads of Num(ber)P, and being suffixes, they undergo lowering onto N at PF. In short, count nouns must be combined with a number-denoting functional head.7 By contrast, mass nouns denote an undivided mass, possessing no unvalued number feature, so they don’t need to be merged with a number-determining expression.

The count-mass distinction in Korean

101

   and the count-mass distinction 2.2  DP We have seen that count nouns are different from mass nouns in that the former has an uninterpretable number feature, whereas the latter does not. There is another important structural difference between count and mass nouns. According to the DP-analysis, NPs are projected into DPs when they are used as arguments.8 It will be shown that it is not the case that every nominal is projected into DP; count nouns must be projected into DPs, whereas mass nouns are not unless they are definite. 2.2.1  DP and count nouns As mentioned above, count nouns are known to denote sets of discrete individuals. For example, (13) can be interpreted as ‘this is a member of the set of dinosaurs in the park’. (13) There is a dinosaur in the park. The denotation of the count noun dinosaur contains two important concepts: set and its individual member(s). In Kim (2013), I propose that the determiner the can bear a relation with either ‘set’ or ‘the individual member(s) of the set’. In (14a), the dinosaur denotes ‘the species called “dinosaur”’, because the makes a relation with the set, not with its individual member, making the set of dinosaurs definite. (14) a. The dinosaur is extinct. (= the species or group called dinosaur is extinct) b. The dinosaurs got angry. [= (all) the members of a presupposed set of dinosaurs got angry] In (14b), on the other hand, it makes a relation with the individual members of the set: it does not make the set of dinosaurs definite, but instead it makes the members of the set definite. In short, the can be a determiner either for the set or for the individual member(s) of the set.9 On the other hand, quantifiers bear a relation with the individual members only, although they take a set of individuals as their complement. Sentence (15a), for example, is interpreted as ‘every member of a presupposed set of dinosaurs got angry’.10 (15) a. Every dinosaur got angry. b. Most dinosaurs got angry. c. Three dinosaurs got angry. To recapitulate, (i) the count noun dinosaur denotes a set of dinosaurs, and the determiner the can bear a relation with either the set or its individual member(s), while quantifiers bear a relation with the individual member(s) only. We have seen from 2.1 that count nouns must be projected into NumP. They also need to be further projected into DP; unless there is a D-projection, there is no way to provide information about whether it is about the set of individuals or its individual member(s). Therefore, the D-projection is obligatorily

102

Kwang-sup Kim

required for count nouns. In (14a), the is a determiner for the set of dinosaurs. In (14b) and (15a–c), by contrast, the, every, and most bear a relation with the individual member(s) of the set, and their subjects are represented as (16a–d), respectively, provided that both determiners and quantifiers are the heads of DPs.11 (16) a. [DP the [NumP -s [NP dinosaur]]:

‘the member of a presupposed set of dinosaurs’ b. [DP every [NumP Ø-PL [NP dinosaur]]] ‘every member of a presupposed set of dinosaurs’ c. [DP most [NumP -s [NP dinosaur]] ‘most members of a presupposed set of dinosaurs’ d. [DP three [NumP -s [NP dinosaur]]12 ‘three members of a presupposed set of dinosaurs’

Thus far, we have considered overt determiners/quantifiers. Determiners can be covert as well as overt. Sentences (17a–b) share the same subject, dogs, but the subject is interpreted in a different way: they are interpreted as either (18a) or (18b). (17) a. b. (18) a. b.

Dogs are barking. Dogs are faithful. There are dogs which are barking. Dogs are generally faithful.

I propose that this difference arises from the fact that the null determiner D can quantify over either the set of dogs or the individual members of the set. (19) [D [NumP -s [NP dog]] If D quantifies over the individual members of the set, we get the reading ‘there are members of an indefinite set of dogs barking’. If, on the other hand, it quantifies over the set of dogs, the group or the species that takes a dog as its member has the property of being faithful. Of course, the interpretations of (17a–b) are dependent on whether the predicate is individual level or stage level (Diesing 1992; Kratzer 1995). If the predicate is individual level, D is forced to quantify over the set, and if it is stage level, D is forced to quantify over the individual members of the set. 2.2.2

DP and mass nouns

Let us now turn to mass nouns. They do not denote a set of individual members, but they simply denote an undivided mass. As they have no individual member, they have no unvalued number feature; hence, there is no need for NumP. Let us recall that the D-projection is required for count nouns, because there is a need to provide information about whether it is about the set or its individual member(s). Mass nouns do not need the D-projection for that purpose, because it denotes an

The count-mass distinction in Korean

103

object with no internal structure. Hence, the subject water in (20a) is represented as (20b), which has no functional projection at all. (20) a. Water is expensive these days. b. [NP water] Of course, classifiers can divide an undivided mass into smaller objects, and cardinal numbers quantify over the smaller objects.13 (21) a. two cups of water b. three gallons of gasoline Furthermore, it is possible for a mass noun like water to be merged with the determiner the, if it bears a relation with the object denoted by water: that is, mass nouns can be projected into DP projections if they are definite. (22) [DP the water in this glass] In sum, mass nouns are not projected into NumP and can be optionally projected into DP if they are definite. We are now in a position to explain the contrast between (9a) and (10), which are rewritten as (23a–b).14 (23) a. *Dog is in the yard. b. Water is in the fridge. Let us first consider (23b), of which the subject is represented as (24). The mass noun water does not require a number projection nor a determiner projection. So it can function as an argument even if it is not headed by any functional category. Furthermore, water has the inherent number feature [-PL]. Therefore, the operation ‘Agree’ can take place between the subject and T (Chomsky 2000, 2001, 2004). (24) [water[-PL]] (25) a. T[x number] be water[-PL] in the fridge: Agree b. T[-PL number] be water[-PL] in the fridge: Subject Raising c. [water[-PL] T[-PL number] be water[-PL] in the fridge] In (23a), by contrast, the count noun dog requires two functional projections: one for the number feature and the other one for specifying whether it is about the set or its individual member(s). As there is neither an overt determiner nor an overt pluralityrelated morpheme, we need to posit two zero morphemes, as illustrated in (26). (26) *[DP

Ø

[NumP

Ø-PL

[Dog[x number]]]]

Positing two zero morphemes, however, leads to a violation of Myers’ (1984) generalization, which prohibits a zero-derived word from being a host for another zero morpheme.

104

Kwang-sup Kim

(27) Zero-derived words do not permit the affixation of further derivational morphemes. In (26), there are two zero morphemes, which is a violation of Myers’ generalization. The problem is that if we delete one of the zero morphemes, an ill-formed nominal is generated. There is no way to generate a well-formed nominal if count nouns enter the derivation in bare form. Therefore, (23a) is ill-formed. 2.2.3

The type between count and mass nouns

Before closing this section, it is worthwhile to note that there is a type between the count type and the non-count type. Interestingly, adjectives such as intelligent, poor, and rich are peculiar in that (i) they must be merged with the if they are used to be arguments, and (ii) they must be treated as plural expressions. (28) a. The poor are happier than the rich. b. [DP the [AP poor[+PL]]] This makes sense if we consider that the denotation of poor is a set of poor men. In (28a), the plays a role as a determiner for the set of poor men, not for the members of the set. The set denoted by poor must be distinguished from other sets, like the set of rich men. So they need to be attached by the definite article the. This type has a mass noun-like property in that they are not countable and simultaneously has a count noun-like property in that they denote a set with plural members. In British English, collective nouns like team and government can be treated as plural expressions. This is another instance in which determiners are used for the sets, not for the individual members, although they denote sets with individual members. As illustrated by (29a–b), the collective noun team has plural individual members in British English, but quantifiers must quantify over the set, not over the individual members: in (29b), three teams means not ‘three members of a team’but ‘three different teams’. (29) a. The team are playing tonight. b. Three teams are playing tonight. Interestingly, (30a–b) show that the individual members of collective nouns as well as adjectives cannot be distributive. (30) a. *The team won a thousand pounds each. b. *The rich eat three apples each a day, while the poor eat one apple each. British collective nouns, just like adjectives like poor, have an inherent plural number feature, so that they are not projected into NumP but projected into DP, as in (31). (31) [DP the [NP team[+PL]]] To summarize this section, (i) count nouns must be projected into NumP and DP, (ii) mass nouns cannot be projected into NumP but can be optionally projected into

The count-mass distinction in Korean

105

DP, and (iii) adjectives like poor and British collective nouns cannot be projected into NumP but must be projected into DP.

3 A continuum-based approach to Korean nouns One of the important claims made in the previous section is that there are expressions that are neither count nor mass nouns. This point will be made clearer if we consider Korean nominals. First of all, this section shows that in Korean, there are several types that can be located between the genuine count nouns and the genuine mass nouns on the count-mass scale and then proposes that the notion ‘the binary count-mass distinction’ must be replaced by ‘the count-mass continuum’. 3.1   Korean count nouns The type in (1) – the student-type – are count nouns because they denote a set of individual members, but they are slightly different from their English counterparts in that they do not have an unvalued number feature, which amounts to saying that there is no NumP. They are unspecified with regard to plurality and so can be interpreted either as singular or a plural expressions. For instance, (32) can be construed as either ‘Is there a student in the classroom?’ or ‘Are there students in the classroom?’, and it can be continued by the answer either in (33a) or in (33b). (32) Kyosil-ey haksyang-i iss-ci? Classroom-in student-Nom exist-Q ‘Are there students in the classroom?’ ‘Is there a student in the classroom?’ (33) a. Haksyang-i han myeng iss-e. Student-NOM One Cl-NOM exist-Q ‘There is one student in the classroom’ b. Haksyang-i yel myeng iss-e.15 Student-NOM 10 Cl-NOM exist-Q ‘There are ten students in the classroom’ Just like English count nouns, Korean count nouns denote a set of discrete members, and determiners and quantifiers can bear a relation with the individual members of the set. Strong quantifiers like motwu ‘all’ and taypwupwun ‘most’ take a set of individual members as their complement, quantifying over the individual members of the set.16 (34) a. [DP [NP Haksayng(tul)] motwu]-ka Kim [DP [NP student(Tul)] all]-NOM Kim ‘All students like Prof. Kim’ b. [DP [NP Haksayng(tul)] taypwupwun]-i Kim [DP [DP student(Tul)] most]-NOM Kim ‘Most students like Prof. Kim’

kyoswu-lul cohahan-ta. professor-ACC like-Ind kyoswu-lul cohahan-ta. professor-ACC like-Ind

106

Kwang-sup Kim

(35) a. [DP [NP haksayng] student b. [DP [NP haksayng] student

motu] all taypwupwun] most

In addition, the type can co-occur with cardinal numbers like tases ‘five’ without classifiers: that is, cardinal numbers can quantify over the individual members without the help of classifiers. As shown in (36a–b), cardinal numbers can either precede or follow this type of nominal. (36) a. [tases [five b. [haksayng [student

haksayng] ‘five students’ student] tases] ‘five students’17 five]

If we assume that the cardinal number tases ‘five’ is the head of DP in both constructions,18 the syntactic representations of (36a–b) would look like (37a–b), respectively. (37) a. [DP tases [DP five b. [DP [NP haksayng] [DP [NP student]

[NP haksayng]] [NP student]] tases] five]

In short, the student-type is quite similar to English count nouns. As mentioned above, however, they are different from their English counterparts in that they have no number projection. Furthermore, they are incompatible with big numbers. The expressions in (38a–c) are all awkward. (38) a. ?? b. ?? c. ??

chen thousand man ten thousand payk man million

haksayng student haksayng student haksayng student

‘a thousand students’ ‘ten thousand students’ ‘a million students’

This shows that Korean count nouns are less countable than English count nouns, which is corroborated by the fact that they are compatible with classifiers, as illustrated in (39).19 (39) a. haksayng Student b. haksayng Student c. haksayng Student

payk hundred chen thousand man ten thousand

meyng Cl meyng Cl meyng Cl

‘a hundred students’ ‘a thousand students’ ‘ten thousand students’

The count-mass distinction in Korean

107

   count-mass nouns 3.2  Korean Let us now turn to type 2 in (2–3), which can be further divided into the appletype and the water-type. As mentioned in section 1, the apple-type permits -tul in limited contexts, whereas the latter type does not permit it in any contexts. The nouns that belong to the water-type are genuine mass nouns: that is, they are not countable. They denote an object with no individual members. On the other hand, the apple-type nouns are between count nouns and mass nouns in that (i) they denote objects that consist of an individual member or individual members, but (ii) determiners and quantifiers cannot bear a relation with the individual member(s): that is, the individual members are not countable. Thus, let us refer to the appletype nouns as count-mass nouns. What mass nouns and count-mass nouns have in common is that they do not permit quantifiers to quantify over the individual members. As a consequence, they are not compatible with numerals and need classifiers for counting. (40) a. *[sey [three b. *[sey [three

sakwa] apple] mwul] water]

(41) a. sakwa apple b. mwul water

sey three sey three

kay Cl khep Cl

‘three apples’ ‘three cups of water’

However, the apple-type differs from the water-type in many respects. First of all, the classifier kay denotes an individual member of apples. By contrast, khep ‘cup’ cannot denote an individual member of water; water does not have an individual member or members. Second, the individual members of the apple-type can be involved in distributivity. Sentences (42–43) show that the individual members of sakwa ‘apple’, unlike mwul ‘water’, can be distributed by hana ssik ‘one each’. (42) Sakwa-ka hana-ssik senmwul phocangtoye iss-ta Sakwa-NOM one-each gift-wrapped be-Ind ‘the apples were gift-wrapped one by one (43) *Mwul-i hana-ssik senmwul phocangtoye iss-ta Water-NOM one-each gift-wrapped be-Ind ‘(intended) the water was gift-wrapped one by one’ To sum up, the individual members of the apple-type can be distributive, although they cannot be quantified by determiners/quantifiers. Thus far, we have seen five different types of nouns: (i) English count nouns, (ii) English/Korean mass nouns, (iii) English adjective and British English collective nouns, (iv) Korean count nouns, (v) Korean count-mass nouns. English

108

Kwang-sup Kim

count nouns denote sets of individual members, and the individual members are countable and distributable. Korean count nouns have the same denotation, and they are both countable and distributable. However, they are different from English count nouns in that they are compatible with small numbers only, and they can co-occur with classifiers. Korean count-mass nouns denote objects with individual members, and the individual members are not countable, but they are distributable. Some adjectives like poor and British English collective nouns are quite similar to Korean count-mass nouns: they denote sets with individual members, but the individual members are not countable.20 They are slightly different from Korean count-mass nouns in that they are not distributable. In short, they have the feature [+individual member], which is evidenced by the fact that they are treated as plural, but they are neither countable nor distributable. Mass nouns denote objects with no individual members. So they are neither plural nor countable nor distributable. The major properties of the five different types can be summarized as follows. (44)

Table 4.1 The major properties of the five different types of nouns

English count nouns Korean count nouns Korean count-mass nouns Some English adjectives and British English collective nouns Korean/English Mass nouns

Individual member

Distributable

Countable

+ + + +

+ + + –

+ + (small numbers only) – –

–

–

–

The above table shows that if x is countable, x is distributive but not vice versa, and if x is distributable, x has individual members but not vice versa. This means that ‘countable’ is higher than ‘distributable’, which is in turn higher than ‘+individual member’ on the countability hierarchy. (45) The Hierarchy in Countability [+countable] > [+distributable] > [+individual member] > [-individual member] According to the hierarchy in (45), English count nouns are more countable than Korean count nouns, which are more countable than Korean count-mass nouns, which are more countable than English adjectives and British English collective nouns, which are more countable than mass nouns. It is noteworthy that the claim advocated here, contra Borer (2005), predicts that classifiers and plural markers like -s may not be in complementary distribution. The plural morpheme -s shows that its host noun has the feature [+individual member], which does not necessarily entail that it has the feature [+countable]. If a noun does not have the feature [+countable], it must co-occur with classifiers for counting. This means that classifiers and plural markers can co-occur with each other. Aikhenvald (2003) argues that they are not in complementary distribution in

The count-mass distinction in Korean

109

many languages, including Yik, Nootka, Tlingit, Tucano, North Arawak, and South Dravidian. In fact, expressions like a pair of glasses also run counter to Borer’s claim, supporting the continuum-based approach. They have a plural morpheme -s, but they require a classifier like pair. (46) a. two pairs of scissors b. ??two scissors To conclude, the binary approach to the count-mass distinction must be replaced by the continuum-based approach.

4 The so-called plural marker -tul as a distributivity marker One of the claims made above is that Korean count-mass nouns can be distributive, although they are not countable. This provides a clue to the puzzle of ‘why Korean count-mass nouns are compatible with the so-called plural marker -tul in certain contexts, even though they are not countable’. It will be shown that the phenomena introduced in (1–3) and (7) follow if (i) -tul is a distributivity marker, not a plural marker, and (ii) count nouns must be projected into DPs, whereas indefinite noncount nouns must not. 4.1   Types of distributive operators As proposed by many linguists (Choe 1987; Gill 1988), distributive operators are two-place predicates. In (47), for instance, each takes two apples and three students as its argument. In (47), the number of two apples is dependent on that of three students: that is, 6 apples (2 × 3) were eaten according to the situation described by (47). The number dependency is usually expressed by the notion ‘scope’: two apples is under the scope of three students. So let us refer to two apples as a narrow scope argument and three students as a wide scope argument. (47) [Three students] ate [[two apples] each] [wide scope argument] [narrow scope argument] In (47), the distributive operator each co-occurs with the narrow scope argument, not with the wide scope argument. 4.1.1

Internal vs. external distribution

Gill (1988) proposes that distributivity relations can be divided into two types: external distribution and internal distribution. Sentence (47) is an example of external distribution. It is external distribution because the wide scope argument is outside of the projection of the narrow scope argument. Sentence (48a) is a Korean counterpart of (47): just like each, -ssik ‘each’ requires two arguments – a wide scope argument and a narrow scope argument – and sakwa twu kay and haksayng

110

Kwang-sup Kim

sey myeng serve as a narrow scope argument and a wide scope argument, respectively. This is also an example of external distribution because the two arguments do not belong to the same nominal projection: that is, the two arguments co-occur with two different Case markers, which means that they belong to two different nominal projections. (48)

[Haksayng sey myeng]-i [sakwa twu kay ssik]-ul mek-ess-ta. Student three Cl-NOM [apple two Cl each]-ACC eat-PAST-DEC ‘Three students ate two apples each’

Let us now turn to (49), which is an instance of internal distribution. In this sentence, haksayng ‘student’ is a wide scope argument, sey myeng ‘three classifier’ is a narrow scope argument, and both arguments belong to the same DP, which is marked by the nominative Case marker -i. (49) [Haksayng sey myeng-ssik]-i han team-i-ta. [Student three Cl-each]-NOM one-team-copula-ind ‘Each team consists of three students’ Internal distribution is a member-group relation. Sentence (49) is interpreted as ‘there is a group of students consisting of smaller groups, and each of the smaller groups has three students as its member’. In (49), ssik takes sey myeng ‘three Cl’ as its narrow scope argument, and [haksayng D] ‘students’ as its wide scope argument, as shown in (50). (50) [ClP [DP haksayng D] [Cl’ sey myeng-ssik]] In (50) the DP [DP haksayng D] denotes a group of students, and sey meyng ‘three classifier’ distributes over the members of the group of students. 4.1.2

The locus of distributive operators

As distributive operators are two-place predicates, there are two possible positions for them: they co-occur with either a narrow scope argument or a wide scope argument. In (47) and (51a), the distributive operator each co-occurs with the narrow scope argument. This type of each is often referred to as binominal each, because it requires two nominals. Although each is the sister of three apples in (51a), it needs to be associated with the two students as well. In an attempt to capture this distributive relation, Heim, Lasnik, and May (1990) propose that each undergoes QR to the wide scope argument the two students at LF. (51) a. The two students bought three apples each. b. each [the two students] bought three apples each In fact, each can co-occur with the wide scope argument in the overt syntax, as in (52). (52) Each of the two students bought three apples.

The count-mass distinction in Korean

111

In this sentence, each is used as a determiner, and so it is usually referred to as determiner each. In fact, determiner each as well as binominal each is binominal in nature, for it obligatorily requires a narrow scope argument. (53) a. ??Each student arrived. b. ??Susie ate each apple. In short, the distributive operator each can co-occur with either a narrow scope argument or a wide scope argument. To sum up, distributive operators can be classified into two types, depending on whether they can co-occur with either a wide scope argument or a narrow scope argument, and they can also be divided into two types, depending on whether they require internal distribution or external distribution. If so, there are four possible combinations, as shown in (54a–d). (54) a. Types of Distribution: External Distribution, Locus of the Operator: Narrow Scope Argument b. Types of Distribution: External Distribution, Locus of the Operator: Wide Scope Argument c. Types of Distribution: Internal Distribution, Locus of the Operator: Narrow Scope Argument d. Types of Distribution: Internal Distribution, Locus of the Operator: Wide Scope Argument Sentence (51a) is an example of (54a), and (52) is an example of (54b). On the other hand, (49) is an example of (54c); both the narrow scope argument and the wide scope argument belong to the same nominal projection, and the distributive operator -ssik co-occurs with the narrow scope argument. The question is now whether the type in (54d) is empirically attested. The answer is positive. The operator per is a case in point. In (55a–b) per, which is the sister of person, takes person as a wide scope argument and three dollars as a narrow scope argument. (55) a. The charge for guide is three dollars per person. b. Three dollars per person is too expensive. (56) [DP

three dollars Narrow scope argument

[per person]] Wide scope argument

Sentence (57) is a Korean counterpart of (55b). Just like per, the distributive operator -tang co-occurs with a wide scope argument, and its narrow scope argument is the nominal that hosts the tang-phrase. (57)

[Il in-tang sam pwul]-un nemwu pissa-ta. One person-per three dollars-Top too expensive-Ind ‘Three dollars per person is too expensive’

(58)

[DP

il in tang [sam pwul]] Wide scope argument Narrow scope argument

112

Kwang-sup Kim

So -tang is another operator that co-occurs with a wide scope argument and permits internal distribution. To recapitulate, distributive operators can be classified into four types, depending on their locus and whether their two arguments belong to the same nominal projection. (59)

Table 4.2 Four types of distributive operators Types of D

Loci of D

External Distribution Internal Distribution

4.2

Narrow Scope Argument

Wide Scope Argument

Binominal each in (51a), Binominal ssik in (47) ssik in (49)

Determiner each in (52) per in (55a–b), tang in (57)

-Tul as a distributivity marker

We have seen from (59) that binominal -ssik is attached to the narrow scope argument, and it permits both external distribution and internal distribution. This section claims that -tul patterns like -ssik: (i) it is a distributive operator: that is, it produces a well-formed sentence only if it can take two arguments: a narrow scope argument and a wide scope argument; (ii) it is attached to the narrow scope argument; and (iii) it permits both internal distribution and external distribution. This claim is shown to explain all the phenomena introduced in section 1, along with the claim that count nouns are projected into DP, whereas indefinite non-count nouns are not. 4.2.1

DP-internal distribution and -tul

Let us recall that count nouns are projected into DPs, whereas mass nouns are not. This sheds light on the contrast illustrated in (1b), (2b), and (3b), which are rewritten here as (60a–c), respectively. (60) a. Kim kyoswu-ka haksayng-tul-ul manna-ss-ta. Kim professor-NOM student-Tul-ACC meet-PAST-DEC ‘Prof Kim met his/her students’ b. *Susie-ka sakwa-tul-ul mek-ess-ta. Susie-NOM apple-Tul-ACC eat-past-ind ‘(intended) Susie ate apples’ c. *Susie-ka mwul-tul-ul masi-ess-ta. Susie-NOM water-Tul-ACC drink-PAST-DEC ‘(intended) Susie drank water’ As mentioned above, -tul is a distributivity marker, which requires two arguments – a narrow scope argument and a wide scope argument. If count nouns are projected into DPs, it is not surprising that count nouns are compatible with the distributivity marker -tul. It is because the distributivity marker can take NP as its narrow scope argument and DP as its wide scope argument. In (60a), for instance,

The count-mass distinction in Korean

113

-tul takes haksayng ‘student’ and the whole DP as its narrow scope argument and wide scope argument, respectively. (61)

[DP [haksayng-tul] D]

On the other hand, (60b–c) are ill-formed because non-countable bare nouns like sakwa ‘apple’ and mwul ‘water’ cannot project into DPs. The internal distributivity relation is not possible; the nouns sakwa and mwul and their projections are identical, and distribution over oneself is not permitted. (62) a. [NP [N sakwa]-tul] b. [NP [N mwul]-tul] Let us now turn to the contrast between (7a–b), rewritten here as (63a–b). (63) a. Susie-ka

[yeki iss-te-n]

sakwa-tul-ul (ta) mek-ess-ta.

Susie-NOM [here were-Retro-Adnominal] apple-Tul-ACC (all) eat-PAST-DEC

‘Susie ate the apples that were here’ b. *Susie-ka [yeki iss-te-n] mwul-tul-ul (ta) masi-ess-ta.

Susie-NOM [here were-Retro-Adnominal] water-Tul-ACC (all) drink-PAST-DEC

‘(intended) Susie drank the water that was here’

If an NP is modified by a relative clause, the whole expression can be taken to be definite/specific. Let us recall that definite/specific nominals can be further projected into DP even if they are not count nouns. If this is so, the object can be analyzed as DP in (63a), as in (64a–b). In (64a–b), [NP sakwa] is a narrow scope argument, and the whole DP is a wide scope argument. (64) [DP [NP [yeki iss-te-n] a. b.

[NP sakwa]-tul]] D]

Wide Scope Argument (group): the whole DP Narrow Scope Argument (an object with a single individual member): [NP sakwa]]

What is noteworthy is that sakwa ‘apple’ can denote either an object with individual members or the individual members. The NP sakwa denotes an object with a single individual member, and the DP [DP [NP [yeki iss-te-n] [NP sakwa]] D] denotes an object that consists of objects, each of which has a single individual member. Therefore, the former can be used as a narrow scope argument and the latter as a wide scope argument. On the other hand, mwul ‘water’ denotes an object with no individual member. So it cannot function as a narrow scope argument; therefore, (63b) is not grammatical, although it is modified by a relative clause. 4.2.2

External distribution and -tul

Thus far, we have seen that internal distribution is possible when nominals are projected into DP, since DPs can provide both the narrow scope argument and

114

Kwang-sup Kim

the wide scope argument within the same nominal projection. This section is concerned with external distribution, in which the narrow scope argument and the wide scope argument belong to a different projection. 4.2.2.1

LOCATIVES AS A WIDE SCOPE ARGUMENT

The wide scope argument can be outside of the nominal projection. In (8a), repeated here as (65), -tul takes sakwa ‘apple’ as a narrow scope argument and yeki ceki-ey ‘here and there’ as its wide scope argument. (65) [Yeki ceki-ey] sakwa-tul-i ttelece-iss-ess-ta. [Here there-at] apple-Tul-NOM fallen-PAST-DEC ‘Apples were fallen here and there’ This is an instance of NP-external distribution; the wide scope argument is outside the projection of sakwa ‘apple’. The bare NP sakwa cannot be NPinternally distributive, but it can serve as a narrow scope argument. However, the mass noun mwul ‘water’ cannot be a narrow scope argument; the narrow scope argument as well as the wide scope argument must denote an individual member, but mwul ‘water’ cannot. Hence, (8b), rewritten here as (66), is not grammatical. (66) [Yeki ceki-ey] mwul(*-tul-)i ppulyece-iss-ess-ta. [Here there-at] water(*-Tul-)NOM sprinkled-PAST-DEC ‘Water was sprinkled here and there’ 4.2.2.2

EVENT AS A NARROW SCOPE ARGUMENT

Interestingly, however, (67a) is grammatical if the implicit subject is taken to be plural. (67) a. Pro[+PL] mwul-tul com masie. Pro water-Tul please drink ‘Why don’t you drink water?’ b. *Susie-ya, mwul-tul com masie. Susie-vocative water-Tul please drink (intended) ‘Susie, why don’t you drink water?’ The ungrammaticality of (67b) shows that a well-formed sentence can be generated as long as the subject is plural. This suggests that the plural subject plays a role as a wide scope argument. The question is now what serves as the narrow scope argument in (67a), given that the mass noun mwul ‘water’ cannot function as a narrow scope argument. I propose that in (67a), -tul does not attach to mwul ‘water’ but to the vP [mwul-tul com masie] ‘drink water please’. In the sentence, -tul takes the drinking event as a narrow scope argument and an implicit plural subject as a wide scope argument.

The count-mass distinction in Korean (68) Pro[+PL] [vP [vP mwul com masi ‘drink water please’] Wide scope argument narrow scope argument

115 tul]

As illustrated in (69a–b), -tul takes the VP [vP mwul com masi] as its complement and then lowers onto the closest constituent. (69) a. [vP [vP mwul com masi] tul]: Lowering of -Tul onto mwul b. [vP [vP mwul-tul com masi] tul] In (69a), mwul ‘water’ asymmetrically c-commands all the other constituents within vP, which means that it is the highest constituent in vP, which in turn means that it is closest to -tul. Hence, it can be a host for -tul. In fact, -tul can be attached to almost every projection, including AdvP and CP. (70) a. Pro[+PL] ese-tul com Pro[+PL] quickly-Tul please ‘Everybody, go quickly please’ b. Pro[+PL] ese com Pro[+PL] quickly please ‘Everybody, go quickly please’

ka-la. go-IMP ka-la-tul. go-IMP-Tul

In (70a), -tul is the sister of the vP [ese com ka] ‘go quickly please’, and in (70b), it is the sister of the CP [CP [TP ese com ka]-la] ‘go please’. (71) a. b.

[vP [vP ese com ka] tul]: Lowering of -tul onto ese [vP ese-tul com ka]

(72) a. b.

[CP Pro[+PL] [CP [TP ese com ka]-la]-tul]: Lowering of -tul onto ese [CP Pro[+PL] [CP [TP ese-tul com ka]-la-tul]]

In (71a) and (72a), vP and CP are the first arguments of -tul, respectively, and the plural subject pro is its second argument. It is well known that vP denotes an event, and CP also denotes an event in imperatives. So they can function as narrow scope arguments. To sum up, the implicit event argument can play a role as a narrow scope argument21 when (i) an event-denoting expression like VP is attached by the distributive operator -tul and (ii) there is a plural subject, which can function as a wide scope argument.

5

Conclusion

Nouns have traditionally been classified into two types: count nouns and mass nouns. If a noun is countable, it is a count noun, and if not, it is a mass noun. This chapter, however, provided a continuum-based approach to the count-mass distinction, claiming that there are many types of nouns between the count noun and the mass noun. According to this view, one type can be more or less countable than the other type. There are three important features involved in deciding whether a noun

Kwang-sup Kim

116

is more or less countable than the other: [+ individual member], [+distributable], [+countable]. The feature geometry of the three features is as follows: (73) a. [+countable] → [+distributable], not vice versa b. [+distributable] → [+individual member], not vice versa. We have seen that the nouns considered in this chapter can be classified into five types on the basis of the three features, as follows. (74) a. English count nouns are countable. b. Korean count nouns are countable but compatible with only small numbers. c. Korean count-mass nouns are not countable but distributable. d. English APs and British English collective nouns have the feature [+individual member] but are not distributable. e. Mass nouns do not have the feature [+individual (member)]. If we combine (73) with (74), we are led to the conclusion summarized in (75). (75) More Countable English Count Nouns: [+countable] Korean Count Nouns: [+countable (small numbers)] Korean Count-Mass Nouns: [+distributable] English APs headed by the: [+individual member] English/Korean Mass Nouns: [-individual member] Less Countable

Notes * This work was supported by the research fund (2021) of Hankuk University of Foreign Studies. 1 In section 4, it will be argued that -tul is not a plural marker but a distributivity marker. 2 Human-denoting nominals usually belong to this type. However, non-human-denoting expressions like mwuncang ‘sentence’ are also compatible with numerals and -tul. (i)

sey mwuncang b. mwuncang sey kay three sentence sentence three CL (ii) I chayk-ey-nun elyewun mwuncang-tul-i This book-at-TOP difficult sentence-Tul-NOM ‘This book contains many difficult sentences’

manh-ta. many-DEC

3 NOM: nominative, ACC: accusative, GEN: genitive, TOP: topic, CL: classifier, DEC: declarative, PL: plural, PAST: past, NumP: number phrase, Retro: retrospect. 4 To make matters complicated, even the water-type nominals can be attached by -tul if the subject is plural. (i)

(Nehi-tul, Ne,) mwul-tul (You-Tul, *you) water-Tul ‘(You guys) drink water’

com masi-e. please drink-Imperative

This supports the claim that -tul is not a plural marker but a distributivity marker.

The count-mass distinction in Korean

117

5 Borer (2005) claims that in order to be counted, they need to be merged with a functional head, which carries out the function of dividing the mass into countable units. Count nouns and mass nouns are different in how they are divided into countable units. The former is headed by a functional head called Div(ision), which is realized as -s. On the other hand, mass nouns can be counted with the help of classifiers. One of the predictions made by this approach is that the plural marker -s cannot co-occur with classifiers: they are in complementary distribution. However, Aikhenvald (2003) claims that the prediction is not borne out in many languages, including Yik, Nootka, Tlingit, Tucano, North Arawak, and South Dravidian. 6 Wiltschko (2012) argues that the count/mass distinction is not grammaticized in languages like Halkomelem. This proposal is compatible with Chierchia’s (1998) claim that languages differ with respect to the denotation of nouns. 7 As will be discussed in 2.2, NumP is further projected into DP. 8 It is Abney (1987) who first proposed the DP analysis. Later the DP analysis was elaborated by such scholars as Valois (1991), Cinque (1994), Longobardi (1994, 2001), Carstens (2000, 2001), and Bernstein (2001). 9 Generally speaking, if a predicate is individual level, the determiner the quantifies over the set, and if it is stage level, it quantifies over the individual member(s). 10 In fact, strong quantifiers can also select an of-P that consists of of and a definite DP. In (ia–b), students denotes a presupposed set of students, the student denotes a definite set of students, and the quantifiers most and all bear a relation with the members of the definite set. (i)

a. b.

[DP most [PP of [DP the students]]] ‘most members of the definite set of students’ [DP all [PP of [DP the students]]] ‘all members of the definite set of students’

11 Quantifiers and determiners as well as -s and a usually have a number feature, and they agree with the number feature of NumP. (i)

a. b. c. (ii) a. b. c.

[DP every-pl [NumP ø-pl [NP dinosaur]]] [DP most+pl [NumP -s [NP dinosaur]]] [DP three+pl [NumP -s [NP dinosaur]]] *every dinosaurs *most dinosaur *three dinosaur

However, the definite article the and some have no number feature. It seems that they have an unvalued number feature, which is valued by the number feature of NumP. The partitive constructions show this point in a clear way. Verbs agree with partitive quantifiers in partitive constructions. (iii) a. b.

One of the computers was broken. Two of the computers were broken.

Interestingly, some of the computers must be treated as a plural quantifier when it takes a plural restriction, as shown in (iva–b), (iv) a. b.

Some of the computers were broken. *Some of the computers was broken.

This suggests that the number feature of some is valued by the closest constituent with a number feature. (v) a. b.

some[x number] of [DP the [NumP -s[+pl [NP dinosaur]]: Valuation some[+pl number] of [DP the [NumP -s[+pl [NP dinosaur]]

12 Milsark (1974, 77) argues that some quantifiers, including cardinal numbers, are ambiguous between a strong and weak reading. I assume that both strong quantifiers and weak quantifiers take NP/NumP as their complement. One important difference between them lies in the fact that strong quantifiers take a presupposed set as their argument, whereas weak quantifiers select a non-presupposed set.

Kwang-sup Kim

118 (i)

a. b.

[DP three [NP students]] [DP three [NP D students]]

I also assume that DP can be recursive in that DP can be selected by another D. (ii) a. b. c.

[DP the [DP three [NP students]]] [DP all [DP the [DP three [NP students]]]] [DP all [PP of [DP the [DP three [NP students]]]]

13 Interestingly, classifiers are only compatible with the weak cardinal numbers. This suggests that Classifier Phrase cannot denote a presupposed set. (i)

a. b.

*all/most cups of water *all/most gallons of gasoline

14 Titles or names are just like mass nouns in the sense that they do not denote the individual members of a set. Thus, they require neither an overt number-denoting expression nor a determiner. (i)

a. b. c.

Do all syntacticians suffer from asteriskitis, doctor? Dr DoLittle is head of department. Poor fool! He thought he’d passed the syntax exam. (Radford 2004, 144)

15 The so-called plural marker -tul can be optionally present if there are more than one students. (i)

Haksyang-tul-i yel myeng iss-e. Student-NOM 10 Cl-NOM exist-Q ‘There are ten students in the classroom’

16 In Korean, as in English, strong quantifiers can take as their complement a PP as well as an NP. In (i), tapwupwun ‘most’ takes the PP haksayng cwung ey ‘out of students’ as its complement. (i)

a. b.

Haksayng-cwung ey taypwupwun-i Kim kyoswu-lul student-out of most-NOM Kim professor-ACC ‘Most of the students like Prof Kim’ [DP [PP Haksayng-cwung ey] taypwupwun-i]] [DP [PP student-out of] most-NOM]]

cohahan-ta. like-DEC

17 Strong quantifiers as well as cardinal numbers can occur in either the pre-nominal position or the post-nominal position. (i)

a. b.

motun haksayng all student taypwupwun-uy haksayng most student

There are two possible analyses of this construction. One is to assume that quantifiers are the heads of DP, and the other is to assume that there is a null D and quantifiers are its specifier. (ii) a. b. (iii) a. b.

[DP motun [DP taypwupwun-uy [DP motun [DP taypwupwun-uy

[NP haksayng]] [NP haksayng]] [NP haksayng] D] [NP haksayng] D]

I leave for further research which analysis is correct. 18 As mentioned in note 16, the prenominal cardinal number tases ‘five’ in (36a) may not be the head of D but the specifier of a null D, as illustrated in (ia–b). (i)

a. b.

[DP tases [NP haksayng] D] [DP five [NP student] D]

The count-mass distinction in Korean

119

19 Just as strong quantifiers can occur in either the pre-nominal position or the post-nominal position, the cardinal number-classifier complex can occur in both positions. (i)

a. b. c.

payk hundred chen thousand man ten thousand

meyng-uy CL-GEN meyng-uy CL-GEN meyng-uy CL-GEN

haksayng ‘a hundred students’ student haksayng ‘a thousand students’ student haksayng ‘ten thousand students’ student

See Kim (2010) for a discussion about whether the two constructions in (39) and (i) are syntactically and semantically different. 20 Some collective nouns, like cattle, are countable. (i) Cattle are grazing in the field. (ii) Thousands of cattle are stranded in Texas after the hurricane. This lends support to the claim that the continuum-based approach is on the right track. 21 The implicit event argument can function as a wide scope argument as well. In (i), -ssik takes sakwa twu kay ‘two apples’ as a narrow scope argument and plural repeated events as a wide scope argument. (i)

Susie-ka sakwa-lul twu kay-ssik Susie-Nom apple-ACC two CL-each ‘Susie ate two apples at one time’

mek-ess-ta. eat-PAST-DEC

References Abney, S. P. 1987. The English Noun Phrase in Its Sentential Aspect. PhD dissertation, MIT. Aikhenvald, A. Y. 2003. Classifiers: A Typology of Noun Classification Devices. Oxford: Oxford University Press. Bernstein, J. B. 2001. The DP hypothesis: Identifying the clausal properties in the nominal domain. In M. Baltin and C. Collins, eds., The Handbook of Syntactic Theory. Oxford: Blackwell, pp. 536–561. Borer, H. 2005. Structuring Sense. Vol 1: In Name Only. Oxford: Oxford University Press. Carstens, V. 2000. Concord in minimalist theory. Linguistic Inquiry, 31, pp. 319–355. Carstens, V. 2001. Multiple agreement and case deletion. Syntax, 4, pp. 147–163. Chierchia, G. 1998. Reference to kinds across languages. Natural Language Semantics, 6, pp. 339–405. Choe, Jae-woong. 1987. Anti-Quantifiers and a Theory of Distributivity. PhD dissertation, University of Massachusetts, Amherst. Chomsky, N. 2000. Minimalist inquiries: The framework. In R. Martin, D. Michaels, and J. Uriagereka, eds., Step by Step: Essays on Minimalism in Honor of Howard Lasnik. Cambridge, MA: MIT Press, pp. 89–155. Chomsky, N. 2001. Derivation by phase. In M. Kenstowicz, ed., Ken Hale: A Life in Language. Cambridge, MA: MIT Press, pp. 1–52. Chomsky, N. 2004. Beyond explanatory adequacy. In A. Belletti, ed., The Cartography of Syntactic Structures. Vol. 3, Structures and Beyond. Oxford: Oxford University Press, pp. 104–131. Cinque, G. 1994. On the evidence for partial N-movement in the romance DP. In paths towards universal grammar. In G. Cinque, J. Koster, J.-Y. Pollock, L. Rizzi, and R. Zan,

120

Kwang-sup Kim

eds., Studies in Honor of Richard S. Kayne. Washington DC: Georgetown University Press, pp. 85–110. Diesing, M. 1992. Indefinites. Cambridge, MA: MIT Press. Gill, D. 1988. Georgian reduplication and the domain of distributivity. Linguistics, 26, pp. 1039–1065. Heim I., H. Lasnik, and R. May. 1990. Reciprocity and plurality. Linguistic Inquiry, 22, pp. 63–101. Kim, Kwang-sup. 2010. On the syntax and semantics of Korean classifier constrictions. Studies in Generative Grammar, 20, pp. 549–576. Kim, Kwang-sup. 2013. On two types of operators in English. Proceedings of 2013 International Conference on English Linguistics. Kratzer, A. 1995. Stage-level and individual-level predicates. In G. N. Carlson and F. J. Pelletier, eds., The Generic Book. Chicago: University of Chicago Press, pp. 125–175. Longobardi, G. 1994. Reference and proper names. Linguistic Inquiry 25, pp. 609–665. Longobardi, G. 2001. The structure of DPs: Some principles, parameters and problems. In M. Baltin and C. Collins, eds., The Handbook of Syntactic Theory. Oxford: Blackwell, pp. 562–603. Milsark, G. 1974. Existential Sentences in English. PhD dissertation, MIT. Milsark, G. 1977. Toward an explanation of certain peculiarities of the existential construction in English. Linguistic Analysis, 3, pp. 1–29. Myers, S. 1984. Zero derivation and inflection. MIT Working Papers in Linguistics, 7, pp. 53–59. Radford, A. 2004. Minimalist Syntax: Exploring the Structure of English. Cambridge: Cambridge University Press. Valois, D. 1991. The Syntax of DP. PhD dissertation, UCLA. Wiltschko, M. 2012. Decomposing the mass/count distinction. Evidence from languages that lack it. In D. Massam, ed., Count and Mass across Languages. Oxford: Oxford University Press, pp. 120–146.

5

Definiteness, specificity, and genericity in numeral classifier languages Chungmin Lee

1

Introduction1

In a dialogic question-and-answer discourse situation between interlocutors, a fixed article-like demonstrative is employed for dynamic anaphoric definite use, which is independent of proper deictic use of demonstratives in article-less languages. Countable bare common nouns in numeral classifier languages and other article-less languages freely occur in argument position for unique definite use (Jenks 2018) due to presupposition maximization (Heim 1991). However, this chapter will also show how bare common nouns can easily function as anaphoric definites, even in non-subject/topic position in a sentence preceded by an utterance with an indefinite DP referred to. It will also be shown that the numeral ‘one’ before a human noun develops into a specific indefinite marker in Korean and sees specificity effects in other languages. Finally, the issue of genericity in terms of information structure in article-less vs. article languages will be discussed, highlighting its dynamicity.

2

Demonstrative anaphoric definite in discourse

To see the anaphoric definite use of the particular fixed familiar medial DEMonstrative ku in Korean (K), the medial sono in Japanese (J), and the distal na in Chinese (C) (like the distal that in English in the two-way system), consider the dialogue in (1) and (2). Those anaphoric definites from demonstratives in (2) in the three numeral classifier (NumCL) languages are contrasted with the anaphoric definite by the article in (3) in English, an article language. Chinese uniquely has a classifier in the prenominal position even after the anaphoric definite demonstrative, as in (2c)[na‘that’*(benCL) shu‘book’], unlike in K [ku (*kwonCL-uy)chayk ‘that copy of book’] and in J [sono(*satuCL-no)hon], as illustrated in (2a,b).2 An optional ACC(usative) marker can be attached to the object noun chayk ‘book’ or to the classifier kwon ‘volume’ in K (2). (1)

na [caymi-iss-nun chelhak chayk han kwon] sa-ss-e (K) I interesting philosophy book one CL buy-PAST-DEC ‘I bought an interesting philosophy book.’

(2)

a.

ku chayk-uy ceca-ka nwukwu-y-a? DEMmed book –of author-NOM who-is-Q

(K) [medial]

122

Chungmin Lee b. sono hon–no tyosya-wa dare-desu-ka? (J) DEMdist book-of author-TOP who-is-Q (a&b) ‘Who is the author of the book?’ c. 那本书的作者是谁？(C) na ben shu de zuozhe shi shui? that CL book-of author is who (a, b, c) ‘Who is the author of ku book, of sono book, and of na ben[CL] book?’

(3)

a. Really? Who is the author of the book? b. Really? Who is the author of ??that book?

The definite article in (3a) in English is a default anaphoric definite, and that in its place is odd (because the article is ‘strong’, to be discussed); unless it is used in at least covert contrast: time of speaking and the event yesterday. It was visible yesterday. So it’s not near in time (Darcy Sperlich p.c.), as in (4). (4)

That book you showed me yesterday, it was complete rubbish.

Or (3a) can be continued as (5), where previously they had talked about some authors, and Darcy was one to be avoided in distinction. (5)

A: Darcy wrote it! B: Didn’t I say to avoid books by that author?

In Chinese, another language of the demonstrative system of two-way distinction like English, if the speaker is holding the book, says, ‘I bought an interesting philosophy book’, and refers back to the book, 这 zhe should be used, but if the speaker isn’t holding the book, 那 na should be used. We noticed that anaphoric definites are realized with a demonstrative in CL languages, whereas they are typically realized with a definite article in English, an article language. On the other hand, in all the CL languages, K, J, and C, the anaphoric demonstratives used in (2) in response, ku, sono, and na, can all be deleted and replaced by the zero anaphora instead, as in (6a) in K (in J as well) and in (6b) in C. It is not the case that only unique definites are realized with a bare noun. (6)

a. [Φ chayk-uy] ceca-ka nwukwu-y-a? book-of author-NOM who-be-Q b. [Φ shu de] zuozhe shi shui? book-of author be who (a&b) ‘Who is the author of the book?’

(K) (C)

In the respondent’s utterance, ‘of the book’ with the anaphoric definite demonstratives ku chayk-uy ‘that book-of’ deleted (or in parentheses), approximated to definite articles the/das, can easily be deleted by maximized presuppositions. Then, the definite article in ‘the author/der autor (German)’ remains for the respondent’s question in article languages but not in article-less languages, as in (7); definiteness is predicted from the author (book) functional relationship bridging. The bare N

Definiteness, specificity, and genericity 123 definite here may be local but is referring to the book in the previous speaker’s utterance anaphorically. The uniqueness of the author of a book is presupposed (even for the [optional] plurality of co-authors). Therefore, any demonstrative marking may lead to oddity in the given context, as in (8). The maximally presupposed, non-redundant, and therefore simple, form (7) is rather preferred in this immediate context. (7)

(8)

nwukwu-y-a? a. Φ1Φ2 ceca-ka author-NOM who-be-Q b. Φ1Φ2 tyosya-wa dare-desu-ka? author-TOP who-is-Q c. Φ1Φ2 zuozhe shi shui? author be-who (a&b&c) ‘Who is the author?’ a.

(K) (J) (C)

ku ceca-ka nwukwu-i-ya? DIST author-NOM who-be-Q b. ??sono tyosya-wa dare-desu-ka? (J) DIST author-TOP who-is-Q c. ??na zuozhe shi shui? (a&b&c) [Intended] ‘Who is the author?’ ??

However, the product–producer relation (Schwarz 2009) is typically a ‘strong’ definite article in German and Chinese, requiring either a strong definite article (von dem, not vom) in German or a demonstrative in Chinese, as in Jenks’ [2018, (14b)] example of na shou shi ‘the piece of poem’ (product) and its poet: #(na  wei) shiren ‘the poet’ (producer). This distinction is somewhat valid in Korean, too: Mia-nun ku kurim-ul acwu cohaha-n-ta. Haciman *(ku (kurim-uy) hwaka-rul moru-n-ta ‘Mia admires the picture. But she doesn’t know the painter’. In contrast, ‘I entered a house and turned on the lights in the living-room’ has the part-whole relation between ‘house vs. lights/kitchen’ in Korean: na-nun enu cip-ey ture ka #(ku) kesil-uy pwul-ul khi-ess-ta. This semantic distinction of part–whole vs. product–producer is universal. Typically, the definiteness of a book applies to its type if its content rather than its physical form matters. For instance, if a student asks others in the first class: A: Ni mai shu le ma? ‘Did you buy the (text)book?’ B: Wo mai shu le ‘I bought the book’, pointing at her copy. Not a demonstrative N #na/#zhe ge shu is appropriate. But looking at a book on the floor, if a student asks 谁的书 丢了？(Shéi de shū diūle?) ‘Whose book dropped?’, my answer (那)是我的 书((Nà) Shì wǒ de shū.) has the same token identity, and a demonstrative may apply, regarding the copy on the floor.

3

Bare common N definites in non-subject position in Chinese, Korean, and Japanese

The maximally bare N definite in common Ns is not limited to the topic/subject position even in Chinese, contra Jenks (2018). Examine (9). The bare common N

124

Chungmin Lee

mao in the object position is definite and refers back to the indefinite CL mao in the previous utterance. Similarly, in (10), any kind of indefinite common nouns can appear in the preceding clause to be definitized in the second clause anaphorically. Examine. (9)

a. you ge mao jin lai le. 来了一只猫。 exist CL cat in came lai le i zhimao ‘A cat came in.’ came a CL cat b. Guo meiyou kan mao. Ta jixu gongzuo. 郭没看见[Φ]还在继续工作。 G

didn’t

see cat

he CNT work

G not see PRF CJN CNT work

‘Guo didn't see the cat. He kept working.’ [PRF=perfective; CJN=conjunctive; CNT=continuative]

(10) 前面有一个动物/植物/人/东西, 他没看见动物/植物/人/东西, 径直 走了过去。 qianmian you yige dòngwù/zhíwù/rén/dōngxī, ta meikanjian dòngwù/–, jingzhi zou le guoqu. (Mengyang Jia p.c.) ‘In front was an animal/plant/man/thing, (but) he didn’t see the animal/–; it passed running away.’ (In video in some contexts.) Even in the subject/topic position, either indefinite or definite bare common Ns can appear in Chinese. The subject in (11) is regarded as a bare indefinite N by many natives (or it may be viewed as definite because of its topicality and the universal quantifier). The bare definite subject in (12) is possible when the speaker heard there are not many women in country A, but then he went there, and found so many women there. (11) 女人都喜欢购物。 nuren dou xihuan gouwu. woman all like shopping ‘All women like shopping (12) 女人也不少啊 nuren ye bu shao a woman even not a few ‘Even the women are not small in number.’ A bare common N in subject/topic position is by default definite in Chinese, as in (13b). Korean and Japanese are different in this respect; they can have an indefinite bare noun in the subject position easily as far as its predicate is eventive or stage level, as in Saram-i param-ey ssure-ci-ess-e ‘A man fell in the wind’. K and J don’t need a preceding existential you yi ge N ‘There is a –’. This may differ a little from Jenks’ (2018) generalization. It is not the case, however, that immediate-situation definites (Hawkins 1978) alone can appear in a nonsubject position, as in (13a). Anaphoric bare common N definites are common against Jenks’ (1918) claim and example. Let’s examine his dialogue example (16a, b, c), changed to (14a, b, c) here. After (14a), according to his claim,

Definiteness, specificity, and genericity 125 nuren in non-subject position in (14c) must be indefinite, unlike na  ge  nuren, an anaphoric demonstrative CL common N definite in (14b). Contrarily, the bare N nuren here readily functions as an anaphoric definite N, referring back to the indefinite common N you ge nuren in the previous utterance of (14a).To make nuren an indefinite independent of the indefinite N in (14a), you (yi) ge nuren, not the bare form nuren, must be used (Mandarin speaking Fey Yuan and Mengyang Jia p.c.). The bare common N nuren may be interpreted as immediate-situation definite, if it’s qizi ‘wife’, but not as indefinite. In Korean, the same happens; in (15b), the demonstrative ku is optional, and the bare common N definite can be anaphoric to refer to (15a). Exactly the same happens in Japanese, as in (16b) (Yasunari Harada p.c.). (13) a. Hufei he-wan-le tang. H drink-finish-LE soup ‘Hufei finished the soup.’ b. Gou yao guo malu. dog want cross road ‘The dog(s) want to cross the road.’ (Jenks 2018 from Cheng and Sybesma 1999: 510) (14) a. You ge nuren sha le Lisi. have CLF woman kill LE Lisi ‘A woman killed Lisi.’ b. Jingcha huaiyi na ge nuren nashihou shou le shang. police suspect that CLF woman at.that.moment suffer LE injury ‘Police suspect that the woman suffered an injury.’ c. Jingcha huaiyi nuren nashihou shou le shang. police suspect woman at.that.moment suffer LE injury (i) ‘Police suspect that a woman suffered an injury.’ (Jenks’ translation) (ii) ‘Police suspect that the woman suffered an injury.’(Yuan’s translation) (15) a. Han yeca-ka Lisi-rul cwuk-i-ess-ta. a woman-NOM L-ACC die-CAUSE-PST-DEC ‘A woman killed Lisi.’ b. Kyengchal-un tangsi (ku) yecha-ka pwusang-tang-ha-n kes-uro uysimha-n-ta police-TOP

at.that.moment that woman-NOM injury suffer-ADNpst N-as suspect

‘Police suspect that the woman suffered an injury.’

(16) a. On'na-ga Lisi-o korosi-ta. woman-NOM L-ACC kill-PST ‘A woman killed Lisi.’ b. Keisatsu-wa (sono) on'na-ga kega-o sita-to kangaete-iru police-TOP that woman-NOM injury-ACC do(incur) C think-ST ‘Police suspect that the woman suffered an injury.’[C: complementizer, ST: stative] Those bare common Ns in non-subject position are neither indefinites nor immediatesituation definites like tang ‘the soup’ in (15) above, which do not require any

126

Chungmin Lee

antecedents to be linked to. Bare Ns such as xiàozhǎng (C)/kyocang (K)/kōchō (J) ‘the principal’, xiaotou ‘the thief’, chushi ‘the cook’, or lǎobǎn ‘the boss’, of a representative or salient type, though possible as anaphoric definites, are more likely to be immediate-situation definites than those bare common Ns such as nuren ‘woman’, nu zhengren ‘female witness’, or nanhai ‘boy’, which is a common generic type. The latter type are more likely to become anaphoric definites either with a demonstrative or without, generally in Chinese, Korean, and Japanese. In the case of Jenks (2018) (15), which is put as (17) here, the initial utterance in the narrative has a conjoined DP in (a), of which just one conjunct is referred back to by #(na ge) nansheng. Indeed, nansheng here is preferred with the demonstrative na ge, presumably because there are two individuals in the previous context. If, however, there involves a sense of contrast, a bare N can be anaphorically definite: if (17a) is followed by (Tamen zhengzai liaotian) ‘they are chatting’. Wo juede nansheng bi nusheng da/youqu ‘I think the boy is taller/is more interesting than the girl’. If followed by a gapped structure Wo zuotian yudao le nansheng, jintian (yudao le) nusheng ‘Yesterday I met the boy, today (I met) the girl’, again the bare nouns become anaphoric definites. If (17a) is followed simply by #Wo juede nansheng hen youqu ‘I think the boy is interesting’, it would be odd. In (17c), if a stress is put on NANSHENG to contrast it with ‘that girl didn’t get a gift from me’, then the bare common N can be an anaphoric definite. The kind of verbs and nouns involved also affect anaphoricity; yudao ‘meet (somebody)’ compared against dai gei ‘bring and give (to somebody)’ can be more accidental (the object being nonspecific) than ‘give something to some specific person’. If the verb happens to be ‘call’, the bare noun becomes an anaphoric definite easily: Wo zuotian da dianhua gei nansheng ‘I called the boy yesterday’. If the pair of common Ns nansheng/ nusheng in (17a) is replaced by laoshi ‘teacher’/xuesheng/ ‘student’, the anaphoric definiteness of one of the following bare Ns is more likely (I owe this observation to Fei Yuan p.c.). Even in Korean, the same tendency of preferring a demonstrative DP holds, but in Korean and Japanese, the bare N can be an anaphoric definite a little more easily than in Chinese in proper contexts. (17) Mandarin narrative sequences (Jenks (2018) (15)) a. Jiaoshi li zuo-zhe yi ge nansheng he yi ge nusheng. classroom inside sit-ST one CL boy and one CL girl ‘There are a boy and a girl sitting in the classroom.’ [This applies to (18a, 19a)] b. Wo zuotian yudao #(na ge) nansheng. I yesterday met that CL boy ‘I met the boy yesterday.’ [ST=stative] c. Wo dai gei #(na ge) nansheng yi ge liwu. I bring give that CL boy one CL gift ‘I’m bringing a gift for the boy.’ (18) a. Kyosil-ey classroom in

namca boy

ai-wa and

yeca ai-ka girl-NOM

anc-a iss-ta sit-ST-DEC

Definiteness, specificity, and genericity 127 b. Na-nun ecey (ku) I-TOP yesterday DEMmed ‘I met the boy yesterday.’

namca ai-lul male child-ACC

manna-ass-ta meet-PST-DEC

(19) a. kyoshitsu-ni-wa otokonoko-toon'nanoko-ga iru classroom in-TOP boy and girl-NOM exist b. watashi-wa kino (sono) otokonoko-ni at-ta I-TOP yesterday DEMmed male child-with meet-PST ‘I met the boy yesterday.’ Likewise, Jenks’ claim “Anaphoric definites in Mandarin must include demonstrative determiners even when the identity of the referent is unknown. In other words, demonstrative descriptions can refer de dicto given an appropriate context” [emphasis mine] in its ‘must’ form cannot hold. If we further consider a non-subject bare common N gou in (20), although it is given as indefinite in Huang et al (2009, 283), which can be interpreted as a numeral CL phrase, as in (20b), it is only one restricted use. The bare N can be an immediate-situation definite, as given in (21). A further possibility is a bare N’s anaphoric definite reading with le attached at the end of the sentence (with an immediate occurrence; Danqing Liu p.c.), as in (22). Moreover, if the verb happens to be a change of state verb or a Theme-affecting verb such as 毆打 (ōudǎ), because of the Theme’s existence presupposition (Lee 1973), the bare N Theme object is more likely to be an anaphoric definite. Another simplest possible anaphoric definite expression can be null, as in Wo kandao Φ le. However, this anaphoric definite reading of a bare N is less preferable than the independent indefinite reading of gou. This preference seems to be a little stronger in Chinese than in Korean and a little stronger in Korean than in Japanese, in that order [C>K>J] [see (18) and (19) for the optionality of the demonstrative and consult native speakers’ reactions for the degree]. For Japanese, otoko, on’na and other bare common nouns are used as anaphoric definites (Yasunari Harada p.c.). In Korean novels, heroes/heroines together with their metonymic nouns are bare nouns, although all are linked as definite referents. Another important use of a bare N should be generic. A generic is indefinite, but I would call it ‘quasi-definite’ in the sense that it is kind-referring even in the object position, as shown in (23). In English, it is typically bare plurals. Incidentally, Google Translate produces ‘I see dog’, as the translation of (20a). This tells much about a bare noun’s variability in article-less languages. (20) a. Wo kandao gou. I see.ASP dog ‘I saw a dog/dogs.’ [Italics mine] b. Wo kandao yi zhi/yixie I see.ASP one CL ‘I saw a/some dog(s).’

gou. some dog

(21) Wo kandao gou (le). – An immediate-situation definite, as in the hairdresser’s saying: Kan jingzi! ‘Look at the mirror!’ Likewise, Kan gou! ‘Look at the dog!’

128 Chungmin Lee (22) Wo kandao gou le. – An anaphoric definite, as in the discourse where the addressee previously said gou already, and gou in the immediately following sentence by the speaker may refer back to the same antecedent dog 一 只狗 (yī zhǐ gǒu) in 一只狗偷偷溜进来。 (Yī zhǐgǒu tōutōu liū jìnlái. ‘A dog sneaked in’). (23) a. Wo xihuan gou. I like dogs b. Ta zai mai he ASP selling ‘He is selling dogs.’

gou. dogs

To conclude, several native speakers of Chinese I consulted with feel uncomfortable about the given form of (21a). They want to see le at its end. Then it can easily become an anaphoric definite, referring to an immediately preceding gou. It can also function as an immediate-situation definite, as in a request. Otherwise, bare nouns in the object position (without le at the end) may be either indefinite or generic. Typologically, because Korean and Japanese subject bare Ns are not necessarily definite, unlike in Chinese, and their object bare Ns are flexible between definites, indefinites, and anaphoric definites, rather like those in Vietnamese (V), Cantonese, and so on, though with no CL+N as definites, we can come up with three types – C, K/J, and V. We can assume individual-denoting expressions of type e (Heim 1982) and adopt Trinh’s (2011) and Jenks’ (2018) semantics for common nouns (with numeral classifiers), which builds on work by Krifka (1995) and Chierchia (1998). Three distinct nominal projections for noun phrases may be set up: DP>CLP>NP in the tree structure. For a unique bare N definite, just an adjoined NP for ι is needed above an NP. With demonstrative D, CLP is still required in Chinese, but its sister becomes NP without its higher CLP in Korean and Japanese. Let’s rely on Jenks’ (2018) formal model of anaphoric definite articles that take one argument more than their unique counterparts, which, he thinks, is satisfied by an index. His proposed denotation for the anaphoric definite article may be applied to anaphoric demonstrative definite in C, K, and J, our three classifier languages.

[[ιx]] = λsr.λP.λQ.: ∃!x[P(x)(sr)∧Q(x)].ιx[P(x)(sr)] Jenks regards that the denotation of ιx has its indexical argument as a property of type , following Nowak’s (2014) analysis of English demonstratives and the idea of domain restrictions of presuppositional determiners (e.g., von Fintel 1994). For the definiteness of bare Ns in classifier languages, the definite type-shifter ι is needed, following Chierchia (1988). Article languages, on the other hand, undergo the Blocking Principle of not doing covertly what you can do overtly, and the projection of DP is required for a definite article. Therefore, for unique definite bare Ns, we need only NPs and ι, whereas we need a D that accommodates the demonstrative in the DEM (CL) NP anaphoric demonstrative definite structure in classifier languages. The additional structure in is presuppositional.

Definiteness, specificity, and genericity 129

4

Demonstratives or second person pronouns in slur words

If the medial demonstrative ku in Korean and sono in Japanese are replaced by the distal one ce (K) or ano (J), a pointing gesture is needed in a visual context, not referring back. This seems to be a special feature of the third distal demonstrative in a three-way system. The distal demonstrative na (ge) in Chinese and that in English in two-way systems do not require pointing. On the other hand, the proximal demonstratives i (K), kono (J), and zhe (C) can be employed both for referring back rather emphatically and for pointing. Incidentally, the proximal demonstratives i (K), kono (J), and zhe (C) are often employed for expressing intense hatred to the addressee. (24a) is an example of a slogan at an anti-Soviet Union demonstration after its downing by Su-15 of civilian Korean Airline flight KAL007 crossing the Pacific Ocean, killing 269 passengers and crew aboard in 1983.That happens in Japanese as in (24b) and also in Chinese (but the proximal demonstrative CL is optional in Chinese). This is to draw the addressee’s attention (as in a real fight where the accuser tries to draw the accused closer to him/ her to hit the person as hard as the accuser wishes). In Korean and Japanese, the proximal demonstratives are used to address the hearer, but in Chinese, the second person 你 (nǐ) is employed to curse the addressee, as in (24c). This happens in English as well. In (c iv), the subject is the second person, and the proximal demonstrative is used in the predicate with the predicative nominal. If the proximal demonstrative is used without 你 (nǐ), it is not usually addressed to the hearer. In English, as in Chinese, the second person pronoun is used to curse the addressee directly. I aktokha-n soryen-nom-tul-a! (Korean) this atrocious ADN Soviet Union-rascal-PL-VOCative ‘These atrocious Soviet Union rascals!’ b. Kono baka-yaro! (or Kono!) (Japanese) this fool-INTensifier ‘This moron/idiot!’ (ba ‘horse’, ka ‘deer’, yaro [familiar/vulgar form of suru(?)] c. i. 笨蛋 Bèndàn! (Chinese) ‘A moron!’ ii. 你笨蛋/白痴 Nǐ bèndàn!/Báichī you moron idiot ‘You a moron/An idiot.’ iii. 你是个笨蛋 Nǐ shì gèbèndàn you are CL moron ‘You are a moron!’ iv. 你這个笨蛋 Nǐ zhègèbèndàn! you this CL moron ‘You this moron!’ (Mengyang Jia and Darcy Sperlich p.c.) v. 這個白癡 Zhège báichī this CL idiot

(24) a.

130

Chungmin Lee

Intensified slur/curse uses of proximal demonstrative in Korean and Japanese SOV languages vs. the same uses of second person pronouns in Chinese and English SVO languages are interesting.

5

Specific indefinites

In Korean, the numeral ‘one’ before a human noun becomes a specific indefinite marker, lengthening its vowel with slight stress in han (nongpwu) ‘a certain farmer’ (Lee 1989). It is distinct from a numeral quantifier in quantity use han, whose vowel is short. Bare nouns function as indefinite nouns at times as well. A NumCL phrase has two forms: post-nominal, for example, haksayng sey myeng [student + three + CL], and pre-nominal, for example, sey myeng-ey haksayng [three + CL + student]. The latter tends to be specific (Lee 1989). These indefinite NumCL phrases can occur in the sentence head (subject/topic) position freely in Korean (and Japanese).3 In Chinese, indefinite NumCL/bare nouns rarely occur in the topic position, as in (25a) except in a modal sense S in quantity reading (25b). With an existential predicate you, indefinite NumCL Ns must be specific in pre-modal topic position (25c) and nonspecific (with arbitrary persons) post-modally (25d). (25) a. *san-ge ren zuo-le yi-bu che. [individual] three-Cl person sit-Prf one-Cl car ‘Three persons were seated in one car.’ b. san-ge ren zuo yi-bu che. [quantity] three-CL person ride one-CL car ‘Three persons, instead of four persons, should (ought to) ride a car.’ c. you san-ge ren yinggaiDeo zuo yi-bu che. [pre-modal] be three-CL person should ride one-CL car ‘Three particular persons should ride a car.’ d. yinggaiDeo you san-ge ren zuo yi-bu che. [post-modal subject] (Tsai et al. 2017) In Chinese, in the subject position, an indefinite NumCL N phrase is specific with a modal following. However, the same indefinite NumCL N phrase at a post-modal subject position receives a non-specific reading. In Korean, fuzzy numerals such as twu se ne kay ‘2, 3, 4 items’ (as in Sakwa twu se ne kay cwu-sey-yo ‘Give me two, three or four apples!’) are frequently used in any position of a sentence including the subject position, and the nominal involved in fuzzy numerals is not specific.

6

Dynamic genericity

The default type of generic is a ‘quasi-universal’ generic. Generic sentences of this type are topic constructions cross-linguistically (in line with Lee [1996] and his refuting Cohen et al’s (2002) claim that focused bare plurals are interpreted existentially). Because of the kind-reference tendency, generics in NumCl languages are typically bare singular nouns. In a shifted context, Contrastive Topic and Contrastive Focus can apply to generic phrases. In focus-sensitive cases such as Even mammals lay eggs, an update of context is needed where the predicate ‘lay eggs’

Definiteness, specificity, and genericity 131 has previously been a potential topic to form a new non-default generic statement. In contrast to characterizing or generalizing in generic utterances, NumCL-marked or PL-marked utterances in Korean and Japanese tend to be anti-generic (existential) and distributive in nature. However, in a special context of a salient, dramatic exemplar, ‘One + CL + of Noun’ in K can serve as a generic statement exceptionally with individual-level predicates, as in (26). Contrast it with the English counterpart with an indefinite article plus a noun, which has the same effect. It has no modal and simply occurs in the present tense with an atemporal reading. However, it receives a specific reading if it occurs with episodic predicates. This genericity seems to be impossible in Chinese. A one + CL + N phrase in Chinese must be in modal context to be kind denoting, as in (27). Otherwise, a bare N in topic position can be generic, as in (30). In Korean and English, a NumCL N phrase or an indefinite NP, respectively, can be generic, with a single typical token representing its type or kind. ‘A book’ in A book is a window into a new world is a generic noun referring not to a single book but books in general. A definite article + N can also become generic. In English, a bare plural is more popular as a generic topic. (26) han phyen –uy shi-ka seysang-ul kamtongshikhi-n-ta one piece of poem-NOM world-ACC move-PRES-DEC ‘A poem touches (moves) the world.’ (27) 一个 学生 就 应当 刻苦学习。 Yi-ge xuesheng jiu yingdang keku xuexi. a CL student just should hard study ‘A student should study hard.’ Liu Danqing (2002) (28) Liu Danqing’s comments (e-mail May 30, 2020): “Such sentences are numerous in daily communication, particularly in an educational context. But notice, in such sentences, a modal verb such as yingdang ‘should’ is necessary to get the generic reading of the sentence, though the adverb jiu ‘just’ is optional. Thus, some formalists like to say that it is the modal verb that licenses the indefinite subject, while a functional statement could be that the sentence is a metonymy that makes an individual refer to a kind.” (29) 学生 就 应当 刻苦 学习。 Xuesheng jiu yingdang keku xuexi student just should hard study ‘Students should study hard.’ Generics interact with information structure, suggesting the need for dynamic genericity. The generic operator GEN in (30) poorly represents real-world analogs, permitting gross exceptions. It is a case of non-monotonic reasoning. (30) GENx [bird(x)][fly(x)] for say-nun nal-n-ta (‘Birds fly’). bird-TOP fly-PRES-DEC (in K) We have attempted to characterize the dynamic anaphoric definiteness of a particularized demonstrative, specificity effects of (NumCL) indefinites, and dynamic aspects of genericity in NumCL languages, as opposed to article languages.

132

7

Chungmin Lee

Quantization, distributivity, and plurality in CL languages

7.1   CL common N harmony and super-ordinate category  plurals in Korean In Korean, the plural marker -tul can be attached to kakwu ‘furniture’, a collective/ mass N in English, or a super-ordinate category label, as in kakwu-tul ‘furnitures (Lit.)’. The plural N must denote a set of (typically different) lower basic category objects under furniture, such as a table, a bookshelf, a TV set, and a chair. So, kakwu sek cem ‘three pieces of furniture’ may consist of a table, a bookshelf, and a TV set rather than three tables, which is possible in a special context where other sub-kinds of furniture are in mind. English animal-s is similar in that it typically denotes a set of lower basic category animals such as tigers, salmon, penguins, and bacteria. In Korean, kacok ‘family’ + -tul, kwukmin ‘nation’ + -tul are also possible, denoting the sums of the members of the family or nation. But the classifier of animals -mari typically composes with subkind animals, rather than with the higher category tongmwul ‘animal’ (?tongmwul sey mari ‘three animals’). It can roughly be represented as (31) mari:= λn {x | Card(x) = n ˄ animalsubkind(x)} (modified from Krifka 1997) ‘Card(x)’ here refers to the number of atomic individuals belonging to some subkind animal, say, mal ‘horse’, forming mal sey mari [horse 3 CL] ‘three horses’. NumCL must compose with its Common Noun (CN) harmoniously; *mal sey tay [horse 3 CLmachine], ?*mal sey kay [horse 3 CLsmall-object] (young children say this), kkangphay sey myeng/??mari [gangster 3 CLhuman/CLsubkind-animal] ‘three gangsters’. This is analogous to selection restriction. The PL marker -tul cannot occur with a NumCL phrase directly: *mal sey mari-tul (a minor set of younger people may accept this). A classifier cannot occur alone: *yeki mari-ka iss-ta ‘Here is mari’. A bare CN can have a group reading in its plural meaning, as Joh (2008) indicated, blocking distributive particles. But that reading only occurs in stage-level sentences. A bare CN has an important Generic (G) or Kind reading, with TOP-marking, which is distributive in generic (modal) sentences, as in (32), but not in (33): (32)

haksayng-un kakca/nwukwuna/motwu cilmwun-ul sey pen –ssik ha-l-swu iss-ta student-TOP each/whoever/all question-ACC 3 CLevent -DISTR do-can-DEC

‘STUDENTs each/whoever/all can ask a question three times.’

(33) #haksayng-i kakca cilmwun-ul sey pen –ssik hay-ss-ta student-NOM each question-ACC 3 CLevent -DISTR do-can-DEC ‘Lit. student each asked a question three times.’ In sum, a mass-like bare CN has various readings in contexts, and NumCL composes with its restrictive CN to be divided into a set of atomic individuals, with the whole becoming distributive or non-distributive depending on predicate types. Classifiers classify objects and events and are crucially related to quantization effect in aspect, distributivity, and plurality. This chapter explores how basic mass-like bare Common Noun is, particularly in classifier languages, and how

Definiteness, specificity, and genericity 133 it composes with numeral classifiers (NumCl) for various effects. This may be generalized to all other types of languages. 7.2   Quantization effect in aspect A bare CN, in its mass-like form in mass/count nouns, is most commonly used in contexts in its underspecified interpretation (34a), getting a quantized reading (as an individuated whole[s]) when NumCL composes with it (34b), as analyzed in (34c): (34) a. nan-nun sakwa-rul sip pun tongang/man-ey mek-ess-ta I –TOP apple-ACC 10 minute for/in eat-PAST-DEC ‘I ate APPLE for/in ten minutes.’APPLE: an (certain) apple, apples, the apple, the apples. b. nan-nun sakwa sey kay-rul sip pun man-ey/*tongan mek-ess-ta I –TOP apple 3 CL-ACC 10 minute in/*for eat-PAST-DEC

‘I ate three apples in ten minutes.’ c. A Predicate P is quantized iff ∀x,y[P(x)˄P(y)]→¬(y.” (Ibid., xiv) He holds that these assumptions imply T2. To show this, he argues as follows: In a N → eK language, all nouns are kind-referring. Kinds are of an argumental type. Hence, kind denoting Ns will be amenable to being directly merged with verbs. (Ibid., xv) Notice, however, that this argument uses neither A1 nor A2. Instead it invokes another assumption: A3. A noun that refers to a kind is an argumental noun (i.e., a noun that can form an argumental bare nominal). And in giving the argument, he presupposes another thesis: A4. All classifier languages are N → eK languages (i.e., [+arg, –pred] languages). Now, A3 and A4 imply T2, for A3 implies a cousin of T2: T2ʹ. All N → eK languages are bare argument languages. So basic assumptions of the NMP account include not only A1–A2 but also A3– A4, which Chierchia implicitly assumes in his derivation of T2. Now, A2 and A4 imply T1.80 And A2 by itself implies a cousin of T1: T3. Obligatory Classifier Thesis: All classifiers are obligatory. That is, no nouns with matching classifiers are directly numerable.81 But both T1 and T3 are false. Many classifier languages, as we have seen, have directly numerable nouns with matching classifiers. Such nouns include the italicized Chinese, Japanese, and Korean nouns in (26)–(28): (26) a. wu bing er yu 5 loaf 2 fish ‘5 loaves and 2 fish.’ b. wu zhang bing er tiao 5 CL loaf 2 CL ‘5 loaves and 2 fish.

[cf., (22a); Mandarin] yu fish

[cf., (22a); Mandarin]

Classifiers, articles, and bare nominals (27) a. futa kazoku 2 family ‘two families’ b. futa tsu -no 2 CL GEN ‘two families’

163

[=(21a); Japanese] kazoku family

(28) a. twu haksayng 2 student ‘two students’ b. twu myeng -uy 2 CL GEN ‘two students’

[Japanese]

[=(20b); Korean] haksayng student

[=(20bʹ); Korean]

Some might defend Chierchia by arguing that he simply uses ‘classifier language’ not for all languages commonly called classifier languages but only for a smaller group of languages, namely, mandatory classifier languages, “where the presence of a classifier is basically obligatory whenever a numeral expression is to be combined with a noun” (2011, xiii). On this proposal, T1 (as he understands it) is simply an analytic truth, which requires no explanation, and the thesis does not relate at all to Chinese, Japanese, and Korean, which turn out not to be classifier languages in his sense. But this is not how he uses the term ‘classifier language’. He gives (modern) Chinese and Japanese as paradigmatic members of the group of languages he calls “classifier languages” (1998a, 92; 1998b, 354; 2010, 107) and includes Korean, too, among those languages in his recent work here in question (2011, xxvii). Moreover, one cannot use the proposed stipulation of the term ‘classifier language’ to rule that the Chinese, Japanese, and Korean classifiers (e.g., the expressions in boldface in (26)–(28)) are not genuine classifiers (in his sense of the term). Thus his analysis of the semantics of classifiers, A2(b), applies to them whether or not the languages turn out to be classifier languages in the proposed sense. And the analysis implies the obligatory classifier thesis (T3), which is also false, as is shown in (26)–(28). Like most contemporary syntactic or semantic accounts of classifiers, Chierchia’s account assumes T1 and T3 and purports to explain their truth. But such accounts cannot be correct. Ever since Greenberg (1972), it has been well known in typological studies of classifier languages that both theses are false. Moreover, as we have seen, they do not even hold for languages with rich and strong classifier systems commonly considered mandatory classifier languages, including modern Korean, Japanese, and Chinese.82

Acknowledgments The work for this chapter was supported in part by a SSHRC Insight Grant [Grant No. 435-2014-0592] and University of Toronto’s Chancellor Jackman Research Fellowship, which is hereby gratefully acknowledged. I presented the ancestor of the chapter in the 13th International Symposium on Chinese Language and

164

Byeong-uk Yi

Linguistics (IsCLL-13). I would like to thank One-Soon Her, James Huang, Lindsey Chen, and other participants of the symposium for useful comments and discussions. I would also like to thank Peter Master for discussions about English bare nominals, and Sarah Hamilton, Lu-Vada Dunford, Yasmine El Gebali, and Eamon Darnell for comments on drafts of this chapter. Needless to say, I am solely responsible for any errors or infelicities.

Notes 1 This chapter is an updated and expanded version of an article presented in the 13th International Symposium on Chinese Language and Linguistics (IsCLL-13). The article, “Numeral Classifiers and Bare Nominals” (Yi 2012), was published in the Proceedings of the symposium. 2 He regards determiners as adjectives of a special kind, “limiting” adjectives (ibid., 202f). In modern grammar, however, they are not considered adjectives. 3 I use the term ‘nominal phrase’ (in short, ‘nominal’) for phrases traditionally considered noun phrases without implying a view about their correct analysis. 4 See, for example, Szabolcsi (1983), Abney (1987), Stowell (1989), Longobardi (1994), and Poole (2011, 72–79). On this proposal, noun phrases in the proper sense (in short, NPs) occur as complements of determiners. 5 Similarly, Poole says “We are assuming that the determiner system . . . form[s] part of UG” (2011, 78). 6 See, for example, Lyons (1999, xv and 48f) and Dryer (2011a; 2011b). 7 Abbreviations used in glosses are as follows: ACC: accusative; CL: classifier; ClP: classifier phrase; CL.PL: plural classifier; D: determiner; DP: determiner phrase; DAT: dative; DEF: definite article; FP: sentence final particle; GEN: genitive; IMP: imperative; LOC: locative; N: noun; NOM: nominative; Num: numeral; NumP: numeral phrase; PL: plural; SG: singular. 8 All Korean examples are mine unless noted otherwise. 9 For it is in the pre-verbal subject position (the same nominal in the object position can take an indefinite meaning as well). See note 21. 10 By ‘Chinese’, ‘Japanese’, and ‘Korean’ in this chapter, I mean Modern Chinese, Japanese, and Korean, respectively, unless noted otherwise. Classical (or Old) Chinese is not a classifier language (see note 44), and it is plausible to take the same to hold for ancient Japanese and Korean, but their modern descendants are languages with rich and strong classifier systems. 11 The term ‘(numeral) classifier’ is ambiguous. It is sometimes used in the narrow sense for special expressions whose counterparts are not (usually) available in non-classifier languages and sometimes in the broad sense for a larger group of expressions that also include measure words (e.g., ‘liter’) among others (see, e.g., Yi 2011a). I use the term in the narrow (and strict) sense while using ‘numerative’ for expressions belonging to the larger group. For discussions of various classifier systems, see Greenberg (1972), Allan (1977), Croft (1994), Aikhenvald (2000, Chapter 4), Gil (2011), and Yi (2011a; 2021). 12 Some classifier languages have articles. For example, Itzaj Maya has the definite article á. (The language also has a plural marker.) See note 45. 13 I focus on nominals featuring common nouns (especially count nouns) and use ‘noun’ as short for ‘common noun’. 14 Simpson (2005) and Wu and Bodomo (2009) modify Longobardi’s account for classifier languages. Like Longobardi’s, however, their accounts are inapplicable to languages with neither articles nor classifiers (e.g., Latin, Classical Chinese). See §3.2 for discussions of such languages and §5.2 for problems of their accounts. 15 The noun has plural forms (e.g., poētae ‘poet.PL.NOM’) and can directly combine with numerals.

Classifiers, articles, and bare nominals

165

16 Note that all the bare nominals in the examples given below feature count nouns. I do not discuss bare nominals featuring mass nouns, partly because English (unlike Italian) regularly uses bare nominals featuring mass nouns, as in (a): (a)

17 18 19

20

21

22 23 24 25 26 27

28

29

30

Fresh water is often drinkable. (Longobardi 1994, 642)

Longobardi treats them like bare nominals featuring plural count nouns, that is, plural forms of count nouns (e.g., big beavers and dogs in (7a)–(7b)). See note 19. Longobardi (1994, 628) attributes this modification of the DP hypothesis to Stowell (1989). By ‘the DP hypothesis’, I will henceforth mean the modified version that applies only to argumental nominals. Longobardi proposes to “assume” that the first two of the above-mentioned constraints are “universal principles” and holds that “It would probably be surprising if [they] were not part of . . . UG” (1994, 641). Longobardi gives the same analysis of bare nominals featuring mass nouns modified by adjectives (e.g., ‘fresh water’ in ‘Fresh water is often drinkable’), and holds that mass or plural count nouns can occur in bare nominals because they are both “kindreferring” (1994, 634). This relates to Chierchia’s account of bare nominals (1998a; 1998b). See §3. This does not solve a similar problem: existential bare nominals can occur as prepositional objects (which are not controlled by predicates), as in ‘I spoke with young children’ (cf., (6c) in Longobardi (1994, 612)). Longobardi suggests an alternative solution: in English, the default existential reading constraint optionally applies to the S-structure while the lexical government requirement applies to the logical form (ibid., 645f). Chinese is a topic-prominent language where the pre-verbal subject position has a strong tendency to avoid indefinite nominals. See Chao (1968, 76ff), Li and Thompson (1981, 85f), and Matthews and Yip (2011, §4.2). This constraint is discussed in §5. Note that Chinese bare nominals can take indefinite meanings in the object position. See, for example, Li and Thompson (1981, 131), Matthews and Yip (2011, §4.2.5), and Cheng and Sybesma (1999, 509f). Smythe (who cites Plato’s use of (8) in the Phaedrus) renders the sentence as “The man is mad” and adds the gloss “a definite person, distinguished from other men” (1956, 287). (9) is quoted in Smythe (1956, 291). Hodler says, “we may well assume with certainty that the article was absent in the proto-Indo-European” (1954, 9; my translation). See also Hewson (1972, 11), who says that proto-Indo-European had no article. The indefinite article ‘a(n)’ stems from the old English numeral ān ‘one’. See, for example, Jespersen (1954, 405ff), Hewson (1972, Ch. 1), and Sommerer (2011, 291). For such examples, see, for example, Hewson (1972, 18f) and Sommerer (2011, 14 and 26). Sommerer elaborates on this statement: “In Present Day English, singular count, plural count and non-count nouns are not supposed to occur ‘bare’ when they are used in a context where the speaker or hearer knows the entities from previous discourse, the intermediate or larger situation or general world knowledge; in other words, in all contexts that make it clear that the noun refers to a unique, identifiable entity” (2011, 13). In this connection, note that the Latin bare nominal poēta in (1) and the Greek nominals in (8)–(9) feature singular count nouns that do not shift to take mass noun meanings. The same holds for the Germanic and Romance singular nominals in the examples discussed below (e.g., (11)–(15), (17), (19)). So some linguists distinguish two forms of the zero article: the one used with mass or plural count nouns and the one used with singular count or proper nouns. Master calls the former “the zero article” and the latter “the null article” (1997, 221). See also Chesterman (1991, 16f). For the article, Jespersen says, “is used more sparingly in English than in many other languages; it is used chiefly when the word without it would not be easily understood as sufficiently specialized” (1933, 163).

166

Byeong-uk Yi

31 See also, for example, Jespersen (1933, 163–169) and Quirk et al. (1985, 276–281). 32 See Jespersen (1933, 166), who uses as representative examples ‘next Sunday’ and ‘last Sunday’ while noting that ‘on Sunday’ means either ‘next Sunday’ or ‘last Sunday’. For more on bare nominals with ‘last’ or ‘next’, see Jespersen (1954, 519ff). 33 See also (18a) below and note 56. 34 Moreover, in Old or Middle English, ‘riche’, for example, can (without coordination) form a bare nominal for the rich people. See Jespersen (1961, 272). 35 See also (19a) below. I think the italicized nouns in (17c) below can be used with indefinite meanings: (17) c.

for three days horse and foot, cannon and tumbril, drum and standard, kept pouring downward past the mill. (Ibid., 464)

(But Peter Master thinks they can be considered definite (private conversation).) 36 See also Roodenburg (2004), who gives a variant of Heycock and Zamparelli’s analysis while criticizing some aspects thereof. The objections to Heycock and Zamparelli’s analysis given below apply to Roodenburg’s as well. 37 He identifies a kind (e.g., the “dog-kind”) with “the totality of its instances” (or the mereological sum of its members) (1998b, 349f). 38 In such languages, Chierchia says “Plural marking, of the kind familiar from many Western languages, will be absent” (1998b, 353). 39 Chierchia says that all nouns of such languages are “in some sense, mass” because “the predicate counterpart of a kind is a mass denotation” (1998b, 353). On his account, the predicate version of a kind-referring noun has in its extension both atomic individuals and all their mereological sums, which he holds is the feature characterizing mass nouns (1998b, 348–351). 40 Chierchia suggests that this is empirically correct. He holds that there is no classifier language with nouns that cannot form bare arguments and argues that this supports his account (1998b, 357f and 402). 41 See Chierchia’s discussion of Slavic languages, which have no articles (1998b, 360ff). 42 See, for example, Gill (2011) and Schachter and Otanes (1972, 111–113) about Tagalog. Gill says “in Tagalog, the noun phrase mansana ‘apple’ may be understood as ‘apple’ (mass), ‘an apple’ (singular), or ‘some applies’ (plural); however, it does not require a classifier in order to occur with a numeral, as in tatlong mansana ‘three apples’” (2011). 43 See Schachter and Otanes (1972, 112). Note also that Tagalog has a non-mandatory plural marker: mga (ibid., 111–113). Schachter and Otanes use it to draw a mass/count distinction for Tagalog nouns (ibid., 112). 44 See, for example, Greenberg (1975, 229–231), who gives a summary of Wang (1958, Vol. 2, Ch. 34), and Peyraube (1998). Greenberg says that Classical Chinese “does not use classifiers and the usual construction is Q–N [i.e., Num–N]” (1974, 231). 45 But some classifier languages have both articles and plural markers. Itzaj Maya is a classifier language but has both the plural marker -ooʹ and the definite article á (Hofling 2000, 141ff, 219, and 247ff). (Hofling says that á is a pro-clictic marker for definiteness (ibid., 247ff  ).) Moreover, they can all co-occur, as in (a): a.

aʹ kaʹ kuul DEF 2 CL ‘the two houses’

naj-ooʹ [Itzaj Maya; ibid., 222] house-PL

(The plural marker is not always mandatory.) For more about Itzaj Maya, see Hofling (2000) and Yi (2021, §4), who discuss plural markers of classifier languages. 46 Note that directly numerable nouns might not directly combine with all numerals. 47 Note that the noun can still take the classifier myeng (for human nouns): (20)

aʹ. haksayng

twu

student 2 ‘two students’

myeng CL

[Korean; cf., Lee and Ramsey 2000, 97–99]

Classifiers, articles, and bare nominals bʹ. twu myeng 2 CL ‘two students’

48 49 50 51 52 53 54

55 56

57

58 59 60 61 62 63 64 65

-uy GEN

haksayng student

167

[Korean; cf., ibid., 97–99]

Classifiers that, like myeng, can be omitted in some cases are called optional classifiers. For more about optional classifiers, see Yi (2021, §3.2). Both nouns in (21a)–(21b) have a matching classifier: the general classifier tsu. (22a) (without the classifiers in parenthesis) is part of the Chinese translation of a passage of the Bible (‘5000 people were fed with 5 loaves and 2 fish’) that Her (2012, 1669) gives as an example. For more about non-mandatory classifier languages, see Yi (2011a; 2021, §3) and the works cited therein. See also Yi (2021, §3.3.3), who discusses classifier systems of modern Korean, Japanese, and Chinese. See also, for example, Doetjes (1996, 44) and Cheng and Sybesma (1999, 539), who call such nouns “count mass nouns”. See also Huang (2011, xxii). See Yi (2014, §§1–2) for a related discussion of the mass noun thesis and features commonly attributed to classifier languages. See also Huang (2011, xvii–xx). Chierchia (2011) makes a substantial revision to his original account of classifier languages in Chierchia (1998a; 1998b), qualifying most of his theses about classifier languages in the original account. Still, the revised account implies that no classifier language noun is directly numerable. See Appendix. Chierchia does not distinguish count nouns from their singular forms, contrasting “singular common count nouns” (e.g., ‘dog’) with “plural common count nouns” (e.g., ‘dogs’) (1998b, 345). Chierchia uses this principle to explain why singular count nouns cannot occur as definite or indefinite arguments in English (which has definite and indefinite articles) (1998b, 360). Using the principle, he also argues that bare substantivized adjectives cannot occur at all as arguments, which he holds is empirically confirmed: ‘*Rich are greedy’ is illformed. But even modern English, as noted above, does not completely ban them (see, e.g., (16a)–(16c)). Cheng and Sybesma use the term ‘classifier’ ambiguously, both narrowly for what they call “count-classifiers” (i.e., classifiers in the strict sense) and broadly for what I call numeratives (i.e., classifiers in the broad sense), as they make explicit (2005, 273). While I use the term only in the strict sense, I keep the same term in discussing their view because the discussion and objection focus on cases that (covertly or overtly) involve classifiers (in the strict sense), except for (24)–(25), which involve what they consider plural classifiers (1999, 534f) (I do not think they are classifiers in the strict sense). Cheng and Sybesma hold that the N-to-Cl movement is covert (1999, 522). This chapter does not discuss expressions in post-verbal subject positions. So I will henceforth use ‘subject’ (or ‘subject position’) as short for ‘pre-verbal subject’ (or ‘preverbal subject position’). Some other positions (e.g., the object position, post-verbal subject position) can have nominals (e.g., bare nominals) with indefinite meanings. See note 21. See also Cheng and Sybesma (1999, 518 and 539). And some languages (e.g., Itzaj Maya) have both articles and classifiers (see note 45). They would have to take ni-hiki [2 CL, ‘two’] to have a covert noun or an empty noun position. Note also that they hold that “noun phrases with overt numerals necessarily yield an indefinite interpretation” (1999, 528). The translation of (23c) is Downing’s. Cheng and Sybesma argue that the nominal in (23e) is “strictly speaking not the equivalent of ‘the three students’” because the sentence “is interpreted as ‘for all x, x is a member of three students, x came’” (i.e., ‘Every one of three students came’) (1999, 535). On the reading of (23e) explained above, which I think is the prominent reading, however, (23e) is not equivalent to ‘Every one of three students came’ (where ‘three students’ is given the indefinite (or existential) reading). This sentence (in the reading) is logically equivalent to ‘There are at least three students each of whom came’, and can

168

66

67

68

69 70

Byeong-uk Yi

be true while (23e) is false — the three students who came might not be those in question when one says (23e). Note also that their analysis violates syntactic constituency of classifiers. In Chinese (as in other classifier languages), “the numeral and the classifier form one constituent”, as they point out (1999, 536) while referring to Croft (1994, 151) and Tang (1990, 403). (See also Huang 2011, xxvi.) But in (4aʹ), which they propose for (4a), the classifier ben does not form a syntactic constituent with the numeral san. Cheng and Sybesma say that “‘A book’ in Chinese languages is literally [one [the book]]” (2005, 284) and that “the classifier singles out one entity from the plurality of entities provided by the semantic representation of the noun in the lexicon; it picks out one instance of what is denoted by N” (2005, 276; my italics). That is, Cl–N phrases with no accompanying numerals, such as the Mandarin ben shu [CL book, ‘a book’] and the Cantonese bun syu [CL book, ‘a/the book’] (Cheng and Sybesma 2005, 261 and 270). Note that not all classifier languages have such phrases. In Korean, numerals (e.g., sey ‘three’) cannot be omitted in phrases involving classifiers (e.g., (4a)–(4b)). The same holds for Japanese. See also note 72. I use ‘CL.PL’ as short for ‘plural classifier’ in the glosses for (24)–(25) for convenience of exposition, but this is not meant to assume Cheng and Sybesma’s account of xie, di, and so on, which I do not think are classifiers (in the strict sense). The only numerals the so-called plural classifiers can occur with are those for one. For example, the Mandarin xie can occur with yi ‘one’, as in (25c): (25)

71

72 73 74

c. yi xie shu [Mandarin] 1 CL.PL book {some, a few (or several)} books

But this does not amount to ‘one book’ (or ‘one of the books’), but to ‘some books’ (or ‘a few (or several) books’). See, for example, Li and Thompson (1981, 112f). Note that Tang’s “Classifier Phrases” differ from Cheng and Sybesma’s “classifier phrases” (CL1). They are meant to capture numeral noun phrases (e.g., (4a)), not bare CL–N phrases, and are to taken to be headed not by classifiers but by Num–CL compounds. See Tang (1990a, 343; 1990b, 403f). Simpson says that among Southeast Asian languages, Vietnamese, Hmong, and Nung have bare CL–N nominals while Thai, Khmer, Burmese, and Indonesian do not (2005, 823). See also Matthews and Yip (2011, §4.2). James Huang points out that specific indefinites can be (pre-verbal) subjects, as in lang lai ye [wolf come FP, ‘Wolves are coming’], where lang ‘wolf’ forms an indefinite nominal for specific wolves (comments on Yi 2012). If so, we can modify the subject constraint as follows: C2ʹ.

Non-specific indefinite nominals cannot occur in the pre-verbal subject position.

And we can revise the traditional account to explain this constraint: the Chinese preverbal subject position is for topics, but a non-specific indefinite cannot introduce a topic (while a specific indefinite might). 75 We have an independent reason to regard Korean as such a language. Korean has a topic marker ‘-(n)un’, which differs from the subject marker ‘-i/ka’, which occurs in (2). And the result of replacing the subject marker in (2) with the topic marker cannot take indefinite meaning. 76 Moreover, not all languages with articles use them as syntactic peers of counterparts of other prominent determiners of Romance or Germanic languages (e.g., demonstrative adjectives). Kana, for example, uses articles and demonstratives in different positions, as in (a): (a)

ló bárí DEF fish ‘this fish’

āmā this

[Kana; Ikoro 1996, 70]

Classifiers, articles, and bare nominals

77

78 79 80 81 82

169

So Dryer concludes that although the determiner category “is well-motivated for English” and its ilk, it is not so for other languages (2011c, §3). See also Dryer (2011b, §3). Greenberg notes that “inherently determined” nouns (e.g., kinship terms) tend to resist the definite article in Stage 2 languages (ibid., 254ff). To illustrate this, he says that Italian “has mio padre and not *il mio padre” (ibid., 256). Note that mio padre ‘my father’ is a bare nominal (the Italian possessive pronoun mio is not a determiner) and that it cannot be considered an N-to-D move DP – cf., Longobardi (1994, 623). Chierchia says, “By classifier languages one usually refers to systems where the presence of a classifier is basically obligatory whenever a numeral expression is to be combined with a noun” (2011, xiii). See also Chierchia (2005, 107). By ‘bare argument language’, he means languages in which (common) nouns can form argumental bare nominals (or, as he puts it, “the verb can directly merge with the bare noun”) (ibid., xiii). Proof: A classifier language noun must be kind-referring (eK), by A4, and thus cannot directly combine with numerals, by A2(a). Proof: If a noun has a matching classifier, then the noun must be kind-referring (eK), by A2(b), and thus cannot directly combine with numerals, by A2(a). For more on non-obligatory classifiers and non-mandatory class languages, see Yi (2021). For an account of classifiers compatible with the existence of such classifiers and languages, see Yi (2009; 2011a; 2011b), who holds that classifiers are para-numeral numeratives (i.e., numeratives that are cousins of numerals for 1). See also Her (2012).

References Abney, S. 1987. The English Noun Phrase in Its Sentential Aspect. Ph.D. Dissertation. MIT. Aikhenvald, A. 2000. Classifiers: A Typology of Noun Categorization Devices. Oxford: Oxford University Press. Allan, K. 1977. Classifiers. Language, 53(2), pp. 285–311. Bloomfield, L. 1933. Language. New York, NY: Holt, Rinehart and Winston. Reprinted. Chicago, IL: The University of Chicago Press, 1984. Chao, Yuen Ren. 1968. A Grammar of Spoken Chinese. Berkeley, CA: University of California Press. Cheng, Lisa Lai-Shen and Sybesma, R. 1999. Bare and Not-So-Bare Nouns and the Structure of NP. Linguistic Inquiry, 30(4), pp. 509–542. Cheng, Lisa Lai-Shen and Sybesma, R. 2005. Classifiers in Four Varieties of Chinese. In: Cinque and Kayne (2005), pp. 259–292. Cheng, Lisa Lai-Shen and Sybesma, R. 2012. Classifiers and DP. Linguistic Inquiry 43(4), pp. 634–650. Chesterman, A. 1991. On Definiteness: A Study with Special Reference to English and Finnish. Cambridge, MA: Cambridge University Press. Chierchia, G. 1998a. Plurality of Mass Nouns and the Notion of ‘Semantic Parameter’. In: S. Rothstein, ed., Events and Grammar. Dordrecht: Kluwer, pp. 53–103. Chierchia, G. 1998b. Reference to Kinds across Languages. Natural Language Semantics, 6(4), pp. 339–405. Chierchia, G. 2010. Mass Nouns, Vagueness and Semantic Variation. Synthese, 174(1), pp. 99–149. Chierchia, G. 2011. Comments. In: Kim et al. (2011), pp. xiii–xvi. Cinque, G. and Kayne, R. 2005. The Oxford Handbook of Comparative Syntax. Oxford: Oxford University Press. Croft, W. 1994. Semantic Universals in Classifier Systems. Word, 45(2), pp. 145–171. Doetjes, J. 1996. Mass and Count: Syntax or Semantics? Proceedings of Meaning on the HIL (HIL Occasional Papers in Linguistics), 1, pp. 34–52.

170

Byeong-uk Yi

Downing, P. 1996. Numeral Classifier Systems: The Case of Japanese. Amsterdam and Philadelphia, PA: John Benjamins. Dryer, M. S. 2011a. Definite Articles. In: Dryer and Haspelmath (2011), Chapter 37. Available at: http://wals.info/chapter/37. Dryer, M. S. 2011b. Indefinite Articles. In: Dryer and Haspelmath (2011), Chapter 38. Available at: http://wals.info/chapter/38. Dryer, M. S. 2011c. Order of Demonstrative and Noun. In: Dryer and Haspelmath (2011), Chapter 88. Available at: http://wals.info/chapter/88. Dryer, M. S. and Haspelmath, M. Eds. 2011. The World Atlas of Language Structures Online. Munich: Max Planck Digital Library. Available at: http://wals.info. Gil, D. 2011. Numeral Classifiers. In: Dryer and Haspelmath (2011), Chapter 55. Available at: http://wals.info/chapter/55. Greenberg, J. H. 1972. Numeral Classifiers and Substantival Number: Problems in the Genesis of a Linguistic Type. Working Papers on Language Universals, 9, pp. 1–39. Reprinted in: Greenberg (1990), pp. 166–193. Greenberg, J. H. 1975. Dynamic Aspects of Word Order in the Numeral Classifier. In: Charles Li, ed., Word Order and Word Order Change. Austin, TX: University of Texas Press, pp. 27–43. Reprinted in: Greenberg (1990), pp. 227–240. Greenberg, J. H. 1978. How Does a Language Acquire Gender Markers? In: J. H. Greenberg et al., eds., Universals of Human Language, Vol. 3. Stanford, CA: Stanford University Press, pp. 47–82. Reprinted in: Greenberg (1990), pp. 241–270. Greenberg, J. H. 1990. On Language: Selected Writings of Joseph H. Greenberg. Edited by K. Denning and S. Kemmer. Stanford, CA: Stanford University Press. Her, One-Soon. 2012. Distinguishing Classifiers and Measure Words: A Mathematical Perspective and Implications. Lingua, 122(14), pp. 1668–1691. Hewson, J. 1972. Article and Noun in English. The Hague: Mouton. Heycock, C. and Zamparelli, R. 2003. Coordinated Bare Definites. Linguistic Inquiry, 34(3), pp. 443–469. Hodler, W. 1954. Grundzüge Einer Germanischen Artikellehre. Heidelberg: Carl Winters Universitätsverlag. Hofling, C. A. 2000. Itzaj Maya Grammar. Salt Lake City, UT: University of Utah Press. Huang, James C.-T. 2011. Comments. In: Kim et al. (2011), pp. xvii–xxvii. Jespersen, O. 1933. Essentials of English Grammar. London: George Allen & Unwin. Reprinted. London and Tuscaloosa, AB: The University of Alabama Press, 1964. Jespersen, O. 1954. A Modern English Grammar on Historical Principles, Part VII: Syntax. Completed and edited by N. Haislund. London: George Allen & Unwin. Jespersen, O. 1961. A Modern English Grammar on Historical Principles, Part II: Syntax. London: George Allen & Unwin. Kim, Young-Wha et al., eds. 2011. Plurality in Classifier Languages. Seoul: Hankook. munhwasa. Lee, Iksop and Ramsey, S. 2000. The Korean Language. Albany, NY: State University of New York Press. Li, Charles N. and Thompson, S. A. 1981. Mandarin Chinese: A Functional Reference Grammar. Berkeley, CA: University of California Press. Longobardi, G. 1994. Reference and Proper Names: A Theory of N-Movement in Syntax and Logical Form. Linguistic Inquiry, 25(4), pp. 609–665. Lyons, C. 1999. Definiteness. Cambridge, MA: Cambridge University Press. Master, P. 1997. The English Article System: Acquisition, Function, and Pedagogy. System, 25(2), pp. 215–232. Matthews, S. and Yip, Virginia. 2011. Cantonese: A Comprehensive Grammar, 2nd ed. London and New York, NY: Routledge.

Classifiers, articles, and bare nominals

171

Peyraube, A. 1998. On the History of Classifiers in Archaic and Medieval Chinese. In: B. K. Tʹsou, ed., Studia Linguistica Serica, Proceedings of the 3rd International Conference on Chinese Linguistics. Hong Kong: City University of Hong Kong Press, pp. 39–68. Phinney, E. et al. 1988. Cambridge Latin Course, Unit I. North American 3rd ed. Cambridge, MA: Cambridge University Press. Poole, G. 2011. Syntactic Theory, 2nd ed. New York, NY: Palgrave Macmillan. Quirk, R. et al. 1985. A Comprehensive Grammar of the English Language. Essex: Longman. Roodenburg, J. 2004. French Bare Arguments Are Not Extinct: The Case of Coordinated Bare Nouns. Linguistic Inquiry, 35(2), pp. 301–313. Schachter, P. and Otanes, F. T. 1972. Tagalog Reference Grammar. Berkeley, CA: University of California Press. Simpson, A. 2005. Classifiers and DP Structure in Southeast Asia. In: Cinque and Kayne (2005), pp. 806–838. Smythe, H. W. 1956. Greek Grammar. Cambridge, MA: Harvard University Press. Sommerer, L. 2011. Old English se: From Demonstrative to Article. Ph.D. Dissertation. University of Vienna. Stowell, T. 1989. Subjects, Specifiers, and X-Bar Theory. In: M. Baltin and A. Kroch, eds., Alternative Conceptions of Phrase Structure. Chicago: University of Chicago Press, pp. 232–262. Szabolcsi, A. 1983. The Possessor that Ran Away from Home. Linguistic Review, 3(1), pp. 89–102. Tang, Chih-Chen J. 1990a. A Note on the DP Analysis of the Chinese Noun Phrase. Linguistics, 28(2), pp. 337–354. Tang, Chih-Chen J. 1990b. Chinese Phrase Structure and the Extended Xʹ-Theory. Ph.D. Dissertation. Cornell University. Wang, Li. 1958. Han Yu Shi Gao [in Chinese: Outline of the History of Chinese Language]. 3 Vols. Beijing: Ke xue chu ban she. Wu, Yicheng and Bodomo, A. 2009. Classifiers ≠ Determiners. Linguistic Inquiry, 40(3), pp. 487–503. Yi, Byeong-uk. 2009. Chinese Classifiers and Count Nouns. Journal of Cognitive Science, 10(2), pp. 209–225. Reprinted in: Kim et al. (2011), pp. 245–264. Yi, Byeong-uk. 2011a. What Is a Numeral Classifier? Philosophical Analysis, 23, pp. 195– 258. Partially reprinted in: Kim et al. (2011), pp. 1–52. Yi, Byeong-uk. 2011b. Afterthoughts on Chinese Classifiers and Count Nouns. In: Kim et al. (2011), pp. 265–282. Yi, Byeong-uk. 2012. Numeral Classifiers and Bare Nominals. In: Yung-O Biq and Lindsey Chen, eds., Proceedings of the 13th International Symposium on Chinese Languages and Linguistics (IsCLL-13). Taipei: National Taiwan Normal University, pp. 138–153. Yi, Byeong-uk. 2014. Numeral Classifiers and the White Horse Paradox. Frontiers of Philosophy in China, 9(4), pp. 498–522. Yi, Byeong-uk. 2021. Numeral Classifiers and Diversity of Classifier Systems. In: Y.-W. Kim, C. Lee, and B.-U. Yi, eds., Numeral Classifiers and Classifier Languages, pp. 6–39. London and New York: Routledge. [Chapter 1 of this volume.]

7

Japanese numeral quantifiers that count events Mana Kobuchi-Philip

1

Introduction

The distinction between an object classifier and an event classifier can be seen, for example, in the following Japanese sentences: (1)

a. b.

John-wa senshuu bentoo-o san-ko John-TOP last week lunch box-ACC 3-CL ‘John bought three lunch boxes last week.’ John-wa senshuu bentoo-o san-kai John-TOP last week lunch box-ACC 3-CL ‘John bought a lunch box three times last week.’

katta. bought katta. bought

Ko in (1a) is a numeral classifier, that is, a classifier that occurs with a numeral. In (1a), it is used for counting individual lunch boxes. That is, it is an OBJECT CLASSIFIER. In contrast, the classifier kai in (1b) is used for counting events; specifically, John’s buying a lunch box. Thus, it is an EVENT CLASSIFIER (e.g., Huang and Ahrens 2003). 1 In Japanese in particular, a numeral classifier is always preceded by a numeral, to form a word of the form [Num CL] whether the classifier is the object-type or the event-type. We will call this combination a NUMERAL QUANTIFIER (e.g., Miyagawa 1989a), or an NQ for short. An event classifier that occurs in the verbal domain is sometimes called a ‘verbal classifier’ in the literature (e.g., Zhang 2002, Paris 2011). However, we will be generally talking about not just the classifier but the NQ, as shown in (1b). Thus, let us use the term VERBAL EVENT NQ for this type of NQ. (2) shows our definition of the verbal event NQ in this chapter. (2)

a. Verbal Event NQ = The direct combination of a numeral and an event classifier in that order, which occurs in the verbal domain. b. [TP. . . . [VP . . . [NQ Num+CLevent ]NQ . . V ]VP ]TP

In order to avoid possible confusion, let us consider a few specific NQs. First, the object classifier ko in (1a) occurs in the verbal domain. Thus, san-ko ‘three

Japanese numeral quantifiers

173

pieces’ is a verbal NQ, but not a verbal event NQ. Such an NQ, that is, an NQ containing an object classifier occurring in the verbal domain, has been a topic of much discussion in linguistics literature for decades and is sometimes referred to as the ‘floated/floating NQ’ both in the syntax and the semantics literature (e.g., Harada 1976; Miyagawa 1989a; Fukushima 1991; Fujita 1994; Nakanishi 2007; Kobuchi-Philip 2007; Miyagawa and Arikawa 2007; Kitaoka 2014, to name a few). However, we will not be discussing this type of verbal NQ in this chapter. Now consider the following sentences: (3)

a. b.

kyonen [san-kai-no shujutsu]-o uketa. last year 3-CL-GEN operation-ACC received ‘I received three operations last year.’ John-wa atama-o go-hari nutta. John-TOP head-ACC 5-CL sewed ‘John got five stitches on his head.’

In (3a), the classifier kai ‘times’ is an event classifier. However, the NQ san-kai ‘three times’ forms an NP with an event-denoting noun shujutsu ‘operation’. Given that the NQ occurs in the nominal domain, it is not a verbal NQ, despite the fact that kai is an event classifier. It is a NOMINAL EVENT CLASSIFIER. Finally, consider the classifier hari in (3b). Hari means ‘needle’ in Japanese; it is an independent noun when it is used without a numeral. In that case, it refers to an object. However, hari functioning as a classifier, as in (3b), is a distinct lexical entry; it does not mean ‘needle’, but rather ‘stitch’, a counting unit for sewing. Since hari in (3b) is an event classifier, and the NQ go-hari occurs in the verbal domain, this NQ is a verbal event NQ. The NQs in (1b) and (3b) are examples of the NQs we will discuss in the current chapter. The classifiers in these NQs are VERBAL EVENT CLASSIFIERS. In the literature, verbal event classifiers/NQs have been discussed mainly with respect to Chinese data, in which linguists discuss various different verbal event classifiers. In Japanese, on the other hand, they have rarely received any attention (for a notable exception, see Tamamura 2004).2 For this reason, in this chapter, we have two goals. One is to give a survey of Japanese verbal event classifiers to see what the general picture is. The second is to make a first attempt at examining one particularly interesting construction with the Japanese verbal event NQ, namely, the construction in which a verbal event NQ is immediately followed by the light verb suru ‘do’, such as ni-haku suru ‘(to) lodge two nights’. As a first report of the work in progress, what we present here is largely a survey and observations rather than a formal analysis. The organization of this chapter is as follows. In the next section, section 2, we will take a look at some issues in the literature surrounding verbal event classifiers in Chinese. In section 3, we try to organize the data of Japanese verbal event classifiers/NQs. In section 4, we focus on Japanese verbal event NQs co-occurring with the light verb suru ‘do’. Section 5 will be concluding remarks.

174

2

Mana Kobuchi-Philip

Previous studies

The linguistic research on verbal event classifiers is not exactly abundant crosslinguistically, and most of it seems to come from Chinese data.3 In his book on Chinese grammar, Chao (1968) offers a long list of as many as 40 Chinese verbal event classifiers. He groups them together as ‘measures for verbs of action’ which “express the number of times an action takes place”. Furthermore, he classifies them into three types: (4)

“It may be a cognate object expressing the action of the verb (Mv1-18), or the part of the body which performs the action (Mv19-25), or the instrument with which the action is performed (Mv26-40).”4 (Chao 1968, pp. 615–616)

The following are examples of each: (5)

a. (Mv1) kann ‘look’, which may co-occur with, for example, kann ‘to look’ b. (Mv22) chyuan ‘fist’, which may co-occur with, for example, daa ‘to strike’ c. (Mv38) chiang ‘spear’, which may co-occur with, for example, chuo ‘to stab’

While Chao’s work stays descriptive, event classifiers in Chinese have increasingly been subject to linguistic analyses in recent literature, in which new observations and new analyses are presented. In this section, we briefly take a look at three recent papers to see the issues surrounding event classifiers. 2.1

Paris (2013)

Paris (2013) distinguishes two classes of verbal event classifiers in Chinese. One type is those that can combine only with the numeral yi ‘one’; the other is those that can combine with the whole range of numbers, including non-numeral quantitydenoting lexical elements such as ji ‘some’. Consider the examples below: (6)

(7)

a. Ta kan-le ni yi yan. he look-SFX you one eye. ‘He cast a glance at you.’ b. Ta kan-le ni ji yan. he look-SFX you several eye. ‘He cast a glance at you several times.’

(Paris 2013, p. 271, 55) (Paris 2013, p. 271, 57)

a. Ta ma-le Zhangsan yi dun. (Paris 2013, p. 272, 64) he curse-SFX Zhangsan one a while ‘He gave Zhangsan a good scolding.’ b. *Ta ma-le Zhangsan lian dun. (Paris 2013, p. 272, 64) he curse-SFX Zhangsan two a while ‘He gave Zhangsan two good scoldings’

Paris calls the examples in (6a) and (7a) ‘weak Q Clvs’ and those in (6b) and (7b) ‘strong Q Clvs’, where Q stands for a numeral and Clv stands for a verbal (event)

Japanese numeral quantifiers

175

classifier, the combination of which is our verbal event NQ. She points out that strong NQs are markers of frequency, indicating the number of times an event takes place; on the other hand, the weak NQs indicate an occurrence of just a single event, as exemplified in (6). However, Paris shows further that the opposition between weak and strong NQs is similar to the opposition between the full verb and the light verb counterparts in languages such as English and French. In English, specifically, it is the distinction between verbs such as to chat, to swim, to gasp, and so on and their light verb counterparts such as to have a chat, to take a swim, to give a gasp, and so on. The two groups of verbs differ in aspectuality. While the former pattern indicates unbounded/homogeneous predications, the latter indicates bounded predications, parallel to the mass/count distinction. To be concrete, the verb kan-le ‘looked’ in (6a) is interpreted as an iterative event, while the verb ma-le ‘scolded’ is interpreted as a durative event. Consequently, the former can be associated with various numerals, while the latter cannot. The ill-formedness of a Chinese sentence like (7b) is considered comparable to that of an English sentence, such as *John took two swims for an hour (cf. John took a swim for an hour). In sum, what Paris addresses in her paper are the aspectual restrictions of certain Chinese verbal event classifiers, a phenomenon we will see again in Japanese (see subsection 3.1 below). 2.2   Donazzan (2013) While Paris (2013) shows that there are two different types of verbal event classifiers, which interact differently with aktionsart, Donazzan (2013) argues that a more ontologically fine-grained analysis can be made.5 Consider an event of knocking on a door. The interpretation of the English sentence (8a) is actually ambiguous, as shown in (8b–c): (8)

a b. c.

John knocked at the door twice. (Donazzan 2013, p. 227, 19) John gave two knocks at the door. John knocked at the door twice, each time for a few minutes.

The ambiguity is due to the interpretation of ‘knocking twice’. It might be the case that John’s knocking once is one blow to the door, as in (8b), or it might be the case that John’s knocking once involves an event of pounding on the door – hitting it several times – that happened on two different occasions, as in (8c). Donazzan calls the former the ‘event-related’ and the latter the ‘occasion-related’ readings (i.e., event-counting vs. occasion-counting) and shows that Chinese verbal event classifiers subcategorize for these two readings. Take a look at the following examples in Chinese: (9)

a. Wo ti-le liang jiao. (event-related) (Donazzan 2013, p. 232, 27b) I kicked two CL ‘I gave two kicks.’ b. Wo qiao-le san ci men (occasion-related) (Donazzan 2013, p. 231, 26b) I knocked 3 CL door ‘I knocked three times at the door, each time striking it several times.’

176

Mana Kobuchi-Philip

In (9a), the two kicks occur consecutively in one certain occasion. On the other hand, in (9b), three occasions of door-knocking take place within a larger interval of time. The distinct readings in (9) are specified by the lexical meaning of the given verbal event classifier; jiao in the former case and ci in the latter. Note that both are iterative events. Donazzan offers a formal definition of the ‘occasion’ (a term borrowed from Cusic 1981) and further defines the event-related classifier and the occasion-related classifier formally. Thus, Chinese ‘strong’ classifiers are divided into two categories, and Donazzan labels the ‘weak’ verbal event classifiers a subcase of event-related classifiers. She offers a summary in the following table: (10)

Table 7.1 Verbal event classifiers and countability in Donazzan (2013) events

occasions countable O

non-countable O

e-Related VCl 𝓌-Related VCl

STRONG countable

xia1, jiao, bu

xia2 WEAK uncountable

In this way, Donazzan organizes the verbal event classifiers with respect to countability in addition to occasion/event-relatedness. 2.3   Zhang (2017) While Donazzan (2013) discusses the event-related and occasion-related readings in semantic terms, Zhang (2017) concentrates on the syntactic significance of what she calls the event-internal and event-external classifiers, which correspond to event- and occasion-related verbal event classifiers in Donazzan’s discussion.6 Zhang demonstrates that the two types of classifiers have distinct syntax. While the NQ with an event-external classifier can grammatically take different positions, as shown in (11), the NQ with an event-internal classifier can take only a single position, as shown in (12): (11)

a. Dalin cengjing kan-guo na bu dianying san ci. (Zhang 2017, p. 14, 35) Dalin before see-PRF that CL movie 3 CL b. Dalin cengjing san ci kan-guo na bu dianying. Dalin before 3 CL see-PRF that CL movie c. Dalin cengjing kan-guo san ci na bu diaying. Dalin before see-PRF 3 CL that CL movie ‘Dalin saw that movie three times before.’ (3 occasions)

(12) a. Dalin cengjing da-guo Yuru san quan. (Zhang 2017, p. 15, 37) Dalin before beat-EXP Yuru 3 CL ‘Dalin punched Yuru three times before.’ (3 punches in one event) b. *Dalin cengjing san quan da-guo Yuru. c. *Dalin cengjing da-guo san quan Yuru.

Japanese numeral quantifiers

177

An adequate syntactic analysis to explain (11)–(12) also has to cover more basic properties of the Chinese event classifier, which Zhang summarizes as follows: (13) a. The numeral-CL sequential generalization: The numeral regularly precedes the classifier. b. The s-selection generalization: Like an object classifier, an event classifier also has its specification as to what kind of event it is a unit of. c. The non-adjunct generalization: The Chinese NQ does not behave like an adverbial element, and an event classifier behaves like a head element, rather than an adverbial. d. The non-imperfective generalization: An event classifier is incompatible with an imperfective aspect. In order to explain these syntactic data all together, Zhang proposes that the basic structure for the Chinese classifier is (14), that the distinct base positions for the event-internal and event-external classifiers is as shown in (15), and that (16) is the special property of the sentence in which the event classifier appears: (14) Figure 7.1 The basic structure for the Chinese classifier (Zhang 2017).

UnitP Numeral

Unit' Unit

XP[+V, -N]

CLv

(15) a. [viewpointAspP [UnitPexternal [vP [VP]]]] b. [viewpointAspP [vP [UnitPinternal [VP]]]]

(Zhang 2017, p. 16)

(16) [viewpointAspP [UnitP CLv]] [-imperfective]

(Zhang 2017, p. 14)

The proposal explains the generalizations in (13) and the contrast between (11) and (12), given the claim that the event-external and event-internal classifiers are base generated in distinct positions, resulting in possible or impossible derivations. In sum, Zhang contributes not only detailed descriptions of the event-external and event-internal classifiers but also a formal account of their subtle syntactic properties. 2.4  Summary   In this section, we have reviewed three articles on the Chinese verbal event classifiers. They all discuss different types. Paris (2013) draws a distinction between strong and weak verbal event classifiers, connecting it to aspectuality. Donazzan

178

Mana Kobuchi-Philip

(2013) offers a refined analysis of the strong event classifier, making a sharp distinction between event and occasion plurality. Zhang (2017) addresses the syntactic distinction between the event-related classifier and the occasion-related classifier, offering a formal syntactic analysis. Given this as our background, in the next section, we will turn to the Japanese verbal event classifiers.

3

Japanese verbal event classifiers: survey

Unlike the case of Chinese, Japanese verbal event classifiers have never been formally scrutinized in the linguistics literature. For this reason, we start with a general survey of the raw data. Martin (2004; originally published in 1975) provides a small section about Japanese verbal event classifiers in his seminal reference book of Japanese. In this nearly 1200-page book, the entire discussion of verbal event classifiers takes up only one 16-line paragraph, and there is no analysis. And yet it is an important early mention of the Japanese verbal event classifier. The following are some examples of what he collected from various sources:7 (17) a. tabako-o ip-puku suru cigarette-ACC 1-CLtake do ‘have one take of cigarette’ c. furo-o hito-furo abiru bath-ACC 1-CLbath bathe ‘take one-bath of bath’ e. hora-o hito-fuki fuku boast-ACC 1-CLblow blow ‘give one blow of boast’ g. ame-ga hito-furi kuru rain-NOM 1-CLfall come ‘one fall of rain comes’

b. ie-o ip-po deru house-ACC 1-CLstep leave ‘leave the house one step’ d. shawaa-o hito-abi suru shower-ACC 1-CLbathe do ‘take one bathing of shower’ f. onando-o hito-katazuke katazukeru closet-ACC 1-CLtidy-up tidy up ‘tidy up one-tidying of the storeroom’ h. kushi-o hito-nade suru comb-ACC 1-CLstroke do ‘give one stroke of comb.’

This small collection of examples already exhibits some interesting linguistic patterns. In what follows, we will attempt to make some descriptive generalizations. 3.1   Occurrence only with the numeral ‘one’ We have seen in Paris (2013) and Donazzan (2013) that some Chinese verbal event classifiers co-occur only with the numeral ‘1’. This is applicable to Japanese as well. Many NQs of the form [1-CL] are almost lexicalized and quite idiomatic, often including the verb that follows: (18) a. ichi-danraku suru b. ? ni-danraku suru 1-CLparagraph do 2-CLparagraph do ‘a period of turmoil/work/etc. ends’

Japanese numeral quantifiers b. ? ni-ken 2-CLlook

suru do

(20) a. hito-aruki suru8 1-CLwalk do ‘have some walk’

b. ? futa-aruki 2-CLwalk

suru do

(21) a. hito-hada nugu 1-CLskin strip ‘pitch in to help’

b. ? futa-hada 2-CLskin

nugu strip

(19) a. ik-ken 1-CLlook ‘take a look’

suru do

179

The verbal event classifiers that occur only with the numeral ‘one’ are the ‘weak’ classifiers of Paris (2013) and event-related classifiers of Donazzan (2013). They give rise to a reading without actual counting but only loosely indicate ‘once’, with the sense of ‘a short while’ or ‘a little bit’. That is the interpretation of the Japanese examples given above as well. 3.2   Instrumental nominals as verbal event classifiers There are a number of verbal event classifiers that are nouns, often denoting objects but also used as the unit of some events. This phenomenon in Chinese is discussed in Chao (1968) (also see Zhang 2002). The following are examples from Japanese:9 (22) a. hito-kuchi taberu 1-CLmouth eat ‘to eat one bite’ c. hito-koe kakeru 1-CLvoice call ‘to call out once’

b. is-seki enzuru 1-CLseat perform ‘to perform a stage’ d. ip-pitsu kaku 1-CLbrush write ‘to write a letter’

In (22a), kuchi means ‘mouth’, but it is used as the unit of biting (and eating). In (22b), seki means ‘seat’, and it is used as the unit of performance. Koe in (22c) means ‘voice’ and functions as event of calling out (or briefly greeting) someone. In (22d), the classifier is hitsu, which is actually not an independent noun, but the Chinese character means ‘brush’.10 This functions as a unit of a writing event, usually a letter to someone. Employing a noun which refers to a tool for a verbal event classifier is not very productive. In fact, all the cases I have found seem to be quite limited and idiomatic. 3.3   Verbal nouns as verbal event classifiers There are some VERBAL NOUNS (VNs) which function as verbal event classifiers. Consider:

180

Mana Kobuchi-Philip

(23) a. san-shiai suru 3-CLgame do ‘to have three games’ c. san-kaiten suru 3-CLrotate do ‘to make three rotations’

b. d.

san-oofuku suru 3-CLround trip do ‘to make three round trips’ san-kooen suru 3-CLperform do ‘to give three performances’

The verbal event classifiers in (23a)–(23d) can all be independent VNs, which form the light verb construction of the form [VN-suru], as shown below: (24) a. shiai suru game do ‘to play a game’ c. kaiten suru rotation do ‘to rotate’

b. d.

oofuku suru round_trip do ‘to make round trips’ kooen suru performance do ‘to perform’

Suru ‘do’ in this construction is referred to as a light verb (e.g., Miyagawa 1989b; Uchida and Nakayama 1993, among others). A light verb apparently provides grammatical support to make a verb from a VN, which is a noun that carries verbal meaning. The contrast between (23) and (24) and, more generally, the similarity between the verbal NQs and VNs, both forming the light verb construction, is quite an intriguing linguistic phenomenon, and it calls for close scrutiny. We will come back to this in section 4. 3.4   Verbs as verbal event classifiers A number of action verbs themselves act as verbal event classifiers: (25) a. hito-keri ireru 1-kick put in ‘give a kick’ c. hito-fuki suru 1-swipe do ‘give a swipe’

b. d.

hito-maki suru 1-roll do ‘roll once’ hito-aruki suru 1-walk do ‘walk a (short) while’

The verb that functions as a verbal event classifier takes a verb itself in the socalled renyoo-form. The renyoo-form is the nominal form of a verb, as the following data demonstrate: (26) a. John-no keri-wa tsuyo-sugi-ta. John-GEN kick-TOP strong-too-PAST ‘John’s kick was too strong.’ b. John-wa nihongo-no yomi-ga tadotadoshii. John-TOP Japanese-GEN reading-NOM clumsy ‘John’s Japanese reading is clumsy.’

Japanese numeral quantifiers

181

As shown in (26), the verb in the renyoo-form can follow N-no. This indicates that the verb in the renyoo-form is a nominal element, since -no is a particle of genitive case that regularly combines two nominal elements. In addition, the verbs in the renyoo-form in (26) are followed by a topic marker -wa or nominative case marker -ga, demonstrating that they are the core elements of the NP. Thus, assuming that the renyoo-form of a verb is syntactically nominal, we can regard it as a type of VN. Depending on the meaning of a verb, its interpretation can be either iterative or durative.11 For example, (25a–b) above are likely to be interpreted as iterative, when kicking and rolling in the context assume telic events. On the other hand, (25c–d) are likely to be interpreted as durative, with swiping and walking taken as continuous events. 3.5   Verbal event classifiers on the basis of Chinese-origin  morphemes The most stereotypical verbal event classifiers in Japanese are formed by a single character of Chinese origin, such as the following: (27) a. san-shuu suru (三周) 3-CLgo-round do ‘to make three rounds’ c. san-shoo suru (三勝) 3-CLwin do ‘to win three games’ e. san-patsu utsu (三発) 3-CLblast do ‘to make three shots’

b. san-paku suru(三泊) 3-CLlodge do ‘to lodge three nights’ d. san-pai suru (三敗) 3-CLlose do ‘to lose three games’ f. san-po aruku (三歩) 3-CLstep do ‘to walk three steps’

The verbal event classifiers in these examples are not independent words but bound morphemes with a certain eventive meaning derived from the given Chinese character. They all denote some event or action, as shown in the English gloss in the subscript. These verbal event classifiers are ‘strong’ in the sense of Paris (2013); that is, they go with any numeral. And they count the number of events or actions.    generic event classifier kai and do ‘times’ 3.6  The Like object classifiers, event classifiers are generally bound to the contextual/ pragmatic selectional restrictions (Zhang 2002, 2017), but this does not apply to kai and do.12 (28) a. John-wa juu-o san-patsu John-TOP gun-ACC 3-CLblast ‘John shot three shots of a gun.’ b. *John-wa juu-o san-paku John-TOP gun-ACC 3-CLlodge

utta. shot utta. shot

182

Mana Kobuchi-Philip c. John-wa juu-o san-kai/do John-TOP gun-ACC 3-CLtime/CLtime ‘John shot a gun three times.’

utta. shot

As shown here, shots might be counted with hatsu ‘blast’, but haku ‘lodge’ is impossible. On the other hand, replacing it with kai or do, which both mean ‘time’, is totally acceptable. Between kai and do, the interpretation of the sentences seems to be quite interchangeable at first blush, and this holds for (28c). However, there is a difference between the two that emerges when we consider more data. For example, while kai can occur with any numeral, do occurs virtually only with numerals representing small numbers (Tamamura 2004) as the following sentences show: (29) a.

kongetsu 10-kai jimu-ni itta. this month 10-CL gym-to went ‘I went to the gym 10 times this month.’ b. ? kongetsu 10-do jimu-ni itta. thismonth 10-CL gym-to went ‘I went to the gym 10 times this month.’

In many cases, kai and do might easily be exchanged. However, the native speaker intuition tells us that do is used for events which are each time significant in some sense. For example, one tries to accomplish something important, or what has happened had a significant effect. For this reason, repetition seems strange. In other words, each time this type of event occurs, one might assume that it is over. In contrast, kai does not imply such an expectation; rather, kai is used for events whose repetition is perhaps even expected. This accords with the observation that do prefers to be associated with numerals for small numbers. Thus, (29a) with kai sounds more natural than (29b) with do, since we assume that visiting a gym can occur many times. Similar observations can be made about the contrasts in (30) and (31): (30) a.

ni-do-to kuruna. 2-CL-COMP come-not ‘Don’t come twice. (Don’t come back ever again.)’ b. ?? ni-kai-to kuruna. 2-CL-COMP come-not

(31) a.

kyooju-wa san-kai-no renzoku koogi-o professor-TOP 3-CL-GEN continuous lecture-ACC ‘The professor gave three continuous lectures.’ b. ?kyooju-wa san-do-no renzoku koogi-o professor-TOP 3-CL-GEN continuous lecture-ACC ‘The professor gave three continuous lectures.’

shita. did shita. did

The phrase ni-do-to ‘2-CL-COMP’ in (30a) is one of the negative polarity items, regularly associated with a negative predicate. Here, replacing do with kai makes

Japanese numeral quantifiers

183

the sentence quite awkward. With do, the speaker of (30a) commands the addressee not to come again. Aside from the fact that ni-kai-to is not the right phrase, with kai, (30b) seems to totally lack the sense of sternness. (31) involves a set of three lectures by a professor. With kai in (31a), it sounds like the lectures have been planned as a series. On the other hand, (31b) with do raises the question why the professor had to give the same lecture three times. The Chinese character of kai (回) means rotation, while that of do (度) means ‘degree’. We are not certain as to exactly how this difference is connected to our observation. However, while rotations can be repetitions of the same movement, the concept of degree does not seem to indicate repetition. Rather, each degree in temperature, for example, marks another step either upward or downward on a scale. We speculate the significance of the event marked by do may come from this lexical property.   3.7  Summary In this section, we have seen some characteristics of the Japanese verbal NQs, which are immediately visible. We made the following observations: (32) i. Many Japanese verbal event classifiers occur only with the numeral ‘1’ and give rise to the ‘a little while’ or ‘a little bit’ reading rather than the iterative reading. ii. Some verbal event classifiers are instrumental. As nouns, they may refer to some objects related to the relevant event, but as classifiers, they function as units of the event. iii. Some verbal event classifiers are verbal nouns. As nouns, they refer to some events and as classifiers, they function as a unit of events. Incidentally, verbal NQs seem to follow the pattern of verbal nouns, followed by the light verb suru. iv. Quite a few Japanese verbs themselves function as verbal event classifiers. They take the nominal form of the verb (‘renyoo’ form), and they directly combine with a numeral. v. Many verbal event classifiers are of Chinese origin functioning as bound morphemes that denote an event or action. vi. There are two generic event classifiers, kai and do. Unlike other event classifiers, these can be the unit of measure for a wide range of events and actions. However, the use of do seems to be more limited to a small number of occurrences. Each of these properties is intriguing and worth investigating in detail. However, in the next section, we will turn to one phenomenon we have seen throughout the above discussion, namely, the fact that the verbal NQ combines with suru ‘do’. Here we should make a general remark on distribution. Compared to object classifiers, event classifiers are much more limited in distribution in Japanese. That is, it seems that counting of events is not as prevalent as the counting of

184

Mana Kobuchi-Philip

objects. On this point, Tamamura (2004) notes that it is easy for the human mind to recognize and perceive individual objects in the environment, while actions and events change in state, which makes the recognition and perception harder for us. It is true that objects have their shape and permanence in space, while actions and events are abstract, and they disappear from sight. Thus, Tamamura claims that counting eventuality is a higher-level psychological activity. (Tamamura 2004, p. 4) On the other hand, Bisang (2018) claims that verbal classification is more limited than nominal classification from a more general logical perspective. From his investigations in Southeast Asian languages (Bisang 1999), he claims that classification shares similar initial conditions both in the nominal and the verbal domains. However, later grammatical development is constrained by the following two conditions: (33) i.

The grammatical category for which classification is used should not impair the initial classificational system. ii. The individual elements used for classification should not interfere with other processes of grammaticalization that are also related to their concrete meaning.

Bisang claims that for the nominal classifier, individuation and identification do not conflict with these two conditions, but for the event classifier, they do. In sum, one might get the impression that event classifiers are generally less prominent than object classifiers, and some of them might be lexicalized and/or become archaic over time. That said, we still consider the Japanese verbal event NQ a noteworthy linguistic phenomenon.

4 The NQV+suru construction In the previous section, we saw that verbal event NQs in Japanese frequently occur with suru ‘do’. It is a pattern which occurs regularly enough. There are two basic syntactic and semantic questions: (34) a. It seems that suru ‘do’ of [NQV+suru] construction is light, like the [VN+suru] construction, but what is the syntactic status of, and relation between, the NQ and suru? b. How do we interpret sentences with [NQV+suru]? What is the significant semantic effect of this construction? 4.1

Suru as a light verb and the structure of [NQV+suru]

The Japanese light verb construction (also called the ‘support verb construction’ or the ‘complex predicate’) has been a topic of steady discussion since the late 80s (Kageyama 1977, 1991; Grimshaw and Mester 1988; Miyagawa 1989b; Uchida and Nakayama 1993; Matsumoto 1996; Miyamoto 1999; Saito and Hoshi 2000; Kamiya 2006; Uegaki and Fujii 2008, among others). One of the typical light verb

Japanese numeral quantifiers

185

constructions in Japanese involves the combination of a verbal noun and the verb suru, as shown below: (35) a. keisatsu-wa otoko-o taiho-shita. police-TOP man-ACC arrest-did ‘The police arrested the man.’ b. sono kyooju-wa choo-ni tsuite kenkyuu shiteiru. the professor-TOP butterfly-to about research doing ‘The professor is doing research on butterflies.’ The verb suru ‘do’ in these sentences is ‘light’ in the sense that it does not carry the literal meaning of ‘do’, but it only supports the verbal property of the preceding VN, syntactically forming a complex verb with the VN. In this sense, [NQV+suru] is similar to the light verb construction with a VN, since a verbal NQ denotes an event just like a VN does, and suru provides an aid to form a complex verb. However, unlike some VNs, an NQ can never co-occur with the accusative Case marker -o when followed by the verb suru, as shown below: (36) a.

Taro-wa Berlin-de san-paku Taro-TOP Berlin-in 3-CLlodge ‘Taro stayed three nights in Berlin.’ b. *Taro-wa Berlin-de san-paku-o Taro-TOP Berlin-in 3-CLlodge-ACC

(37) a.

Taro-wa toonamento-de san-shoo Taro-TOP tournament-in 3-CLwin ‘Taro won three games in the tournament.’ b. *Taro-wa toonamento-de san-shoo-o Taro-TOP tournament-in 3-CLwin-ACC

shita. did shita. did shita. did shita. did

This contrasts with the VN data, where some VNs may co-occur with the accusative Case marker, while others may not: (38) a. sono kyooju-wa choo-ni tsuite kenkyuu the professor-TOP butterfly-on about research ‘The professor is researching on butterflies.’ b. sono kyooju-wa choo-ni tsuite kenkyuu-o the professor-TOP butterfly-on about research-ACC ‘The professor is researching on butterflies.’ (39)

a.

ya-ga mato-ni meechuu arrow-NOM target-at strike-ACC ‘The arrow struck the target.’ b. *ya-ga mato-ni meechuu-o arrow-NOM target-at strike-ACC

shita. (=35b) did shita. did

shita. (Uchida and Nakayama 1993) did shita. did

186

Mana Kobuchi-Philip

(40)

a.

densha-ga eki-ni toochaku train-NOM station-at arrival ‘The train arrived at the station.’ b. * densha-ga eki-ni toochaku-o train-NOM station-at arrival-ACC

shita. did shita. did

When a VN can co-occur with an accusative Case marker, as in (39b), the verb suru can be regarded as a heavy verb rather than a light verb. It is necessary to exercise caution when the lexical item for the classifier can also function as a VN. Consider the following set of data: (41) a.

Taro-wa Taro-TOP ‘Taro b. Taro-wa Taro-TOP ‘Taro c. *Taro-wa Taro-TOP

toonamento-de san-shiai shita tournament-in 3-CLgame-ACC did had three games in the tournament.’ toonamento-de shiai-o shita tournament-in game-ACC did had games in the tournament.’ toonamento-de san-shiai-o shita tournament-in 3-CLgame-ACC did

Generally, an NQ combines with suru without an accusative marker -o, as in (41a). In (41b), shiai ‘game’ without a numeral is a VN instead of an NQ, and it is perfect when followed by an accusative Case marker -o and the verb suru. However, once a numeral is associated and the VN becomes a classifier, then it cannot tolerate an accusative marker, as shown in (41c). For the sake of completeness, let us also note that the grammaticality of (42b) below should not be mistaken for a counter-example. Consider: (42) a. Taro-wa soko-de san-kaiten Taro-TOP there-at 3-CLrotate ‘Taro turned three times there.’ b. Taro-wa soko-de san-kaiten-o Taro-TOP there-at 3-CLrotate-ACC ‘Taro made a triple axel there.’

shita. did shita. did

While (42a) is a typical [NQV+suru] construction, (42b) is not. In (42b), san-kaiten is used as the name of a particular technique in a sport such as figure skating or gymnastics. It is a technique in which one makes three rotations at once in the air. In figure skating, this is called ‘triple axel’. Thus, here, the verb suru is a heavy verb, and san-kaiten is its direct object, associated with the accusative Case marker -o. There is presence/absence alternation of the accusative -o with the light verb construction with the VN. Some VNs can function as the direct object of the heavy verb suru in association with the accusative -o, forming the heavy verb construction [VN-o suru], while others are rather awkward in the same construction: (43) a.

shiai-o suru game-ACC do ‘to play a game’

b. *oofuku-o suru round_trip-ACC do (intended) ‘to make round trips’

Japanese numeral quantifiers c. *kaiten-o suru rotation-ACC do (intended) ‘to rotate’

d.

kooen-o performance-ACC ‘to perform’

187

suru do

Such data have generated discussions with respect to the condition of the presence of accusative -o (e.g., Grimshaw and Mester 1988; Miyagawa 1989b; Uchida and Nakayama 1993). However, in the current context with respect to NQs, we consider it rather obvious that the NQ does not co-occur with the accusative -o. This is because the NQ is, after all, a quantifier rather than any sort of argument. Now, let us consider the syntactic relation between the NQ and suru. Actually, they stand in quite a tight relation. We find that they cannot be separated easily. Consider the following data: (44) a.

Taro-wa ike-o san-shuu shita. Taro-TOP park-ACC 3-CLround did ‘Taro made three rounds of the pond.’ b. *Taro-wa san-shuu ike-o t shita. Taro-TOP 3-CLround pond-ACC t did c. *san-shuu Taro-wa ike-o t shita. 3-CLround Taro-TOP pond-ACC t did d. *Taro-wa ike-o san-shuu yukkuri shita. Taro-TOP pond-ACC 3-CLround slowly did

In (44a), the NQ and suru are directly adjacent to each other, and the sentence is well-formed. On the other hand, in (44b) and (44c), the NQ is scrambled to another position, and these sentences are ungrammatical. Furthermore, even if the NQ stays in the position lower than the direct object, inserting another lexical element between the NQ and suru makes the sentence ungrammatical, as shown in (44d). This shows that the combination of [NQV+suru] is syntactically quite tight, and NQV and suru must be adjacent. This contrasts with the case of an NQ that combines with a heavy verb. Observe the following data: (45) a. Taro-wa ike-o san-shuu hashitta. Taro-TOP pond-ACC 3-CL ran ‘Taro ran three rounds around the pond.’ b. Taro-wa san-shuu ike (-no mawari)-o Taro-TOP 3-CLround pond (-GEN circumference)-ACC ‘Taro ran three rounds around the pond.’ c. san-shuu Taro-wa ike (-no mawari)-o 3-CL Taro-TOP pond (-GEN circumference)-ACC ‘Taro ran three rounds in the park.’ d. Taro-wa ike-o san-shuu yukkuri hashitta. Taro-TOP pond-ACC 3-CLround slowly ran ‘Taro ran three rounds in the park slowly.’

hashitta.13 ran hashitta. ran

All these sentences are grammatical. In (45a), the NQ and the verb are directly adjacent to each other. However, the NQ can independently be scrambled to another

188

Mana Kobuchi-Philip

position, as shown in (45b) and (45c). Furthermore, another lexical element can be inserted between the NQ and the verb, as shown in (45d). This set of data strongly suggests that this NQ is an adverbial element, like an ordinary floating NQ. For completeness, let us also note that the light verb suru cannot survive without the NQ: (46) a. Taro-wa ike-o #(san-shuu) Taro-TOP pond-ACC (3-CLround) ‘Taro made (three circles) of the pond.’ b. Taro-wa ike-o (san-shuu) Taro-TOP pond-ACC 3-CL ‘Taro ran (three circles) of the pond.’

shita. did hashitta. ran

(46a) is a [NQV+suru] construction, and (46b) contains the same NQ but a full verb: hashitta ‘ran’. (46a) becomes ill-formed if the NQ is removed. The combination of ike ‘pond’ and shita ‘did’ does not mean anything sensible. On the other hand, in (46b), the NQ can easily be omitted. This shows that while the NQ in (46b) is likely to be an adverbial element, the NQ of [NQV+suru] is an essential part of the verbal predicate. Let us also note that the combination [NQV+suru] is not any kind of a compound verb, since focus particles such as mo ‘even’ and shika ‘except’ can follow the NQ before suru: (47) a. Taro-wa kooen-o san-shuu-mo shita. Taro-TOP park-ACC 3-CLround-even did ‘Taro made as many as three rounds in the park.’ b. Taro-wa kooen-o san-shuu-shika shi-nakat-ta. Taro-TOP park-ACC 3-CLround-except do-NEG-PAST ‘Taro did not make more than three rounds in the park.’ Thus, we must assume a structure that potentially provides a position for a focus particle between the NQ and suru. That is, the NQ and suru at least have to take distinct syntactic positions. In sum, on the one hand, the verbal NQ associated with suru cannot move independently away from suru; on the other hand, the two must occupy syntactically distinct positions. Without going into technical details, we assume that the basic structure of NQ-suru within a VP would look like (48a), which contains an NQPhrase, as shown in (48b): (48) a.

VP

b. V'

DP

NQP NQ'

V

NQ

NQP

V

n-CL

suru

san-shuu 3 CL

Japanese numeral quantifiers

189

The combination of the NQ and the light verb suru ‘do’ should form a syntactic unit as a verb. Thus, this unit could enter into a thematic relation with DPs in its maximal projection. As for the morphological status of the NQ itself, we assume that it is a single word, a compound consisting of a numeral and a classifier. This assumption comes from (i) that a numeral-classifier combination in Japanese undergoes the phonological change typical for a compound in Japanese, as shown in (49) below, and (ii) there is categorically no possibility of any lexical insertion between the numeral and the classifier:14 (49)

a. /ichi/ + /shuu/ = /isshuu/ (shuu = classifier for events of circular movement) b. /ichi/ + /hatsu/ = /ippatsu/ (hatsu = classifier for blasting event)

   interpretation 4.2  The In this section, we will briefly look at the interpretation of an [NQV+suru] sentence. We refer readers who are interested in the latest formal semantic analyses to recent work on this topic such as Li (2019). Here we would like to stay with the actual Japanese data and try to describe its characteristics with respect to the interpretation. First, the significance of this construction is its availability only with an event classifier. An object classifier is never productively associated with the light verb suru ‘do’ to function as a complex verb: (50) a. kimono-no sentaku-o san-mai shita. kimono-GEN laundry-ACC 3-CL did ‘I washed three kimonos.’ b. naze nekutai-o ni-hon shimashita-ka. why necktie-ACC 2-CL did-Q ‘Why did you wear two neckties?’ These sentences are quite acceptable for native speakers of Japanese. However, they are simply sentences with a floating NQ in which the NQ takes the position just before the verb suru ‘do’. It is not the case that [NQ suru] forms a complex verb. The initial question we posed earlier is the following, repeated here from (34b): (51) How do we interpret sentences with [NQV+suru]? What is the significant semantic effect of this construction? As in Chinese (Zhang 2017), Japanese sentences with an NQV always yield the event-related reading. This is not very surprising in the case of [NQV+suru], since the NQ is inside a complex verb, and the scope of the numeral is the event classifier itself. So the interpretation would be that there are a number of events, specified by the classifier, on one occasion. Consider the following examples: (52) a. Taro-wa sono katana-o mi-furi Taro-TOP that sword-ACC 3-CLswing ‘Taro made three swings of the sword.’

shita. did

190

Mana Kobuchi-Philip b. booru-wa yuka-o san-kaiten shita. ball-TOP floor-ACC 3-CLrotate did ‘The ball made three rolls over the floor.’ c. Taro-wa ike-o san-shuu shita. Taro-TOP pond-ACC 3-CLround did ‘Taro made three rounds of the pond.’

(=44a)

In (52a), ‘Swinging’ is what the classifier -furi specifies, and Taro’s three swings of the sword occur consecutively as a single event. Likewise, in (52b), the classifier is a unit for rolling/rotating, and three rolls are counted, but they are regarded as a single continuous event. Also in (52c), Taro is moving around the circumference of the pond (without specifying how, e.g., walking, running, or riding a bike) and three cycles are counted. Again, Taro’s moving about the pond is regarded as a single event. Thus, in every case, a single eventuality is expressed, that is, event-internal reading. This contrasts interestingly with an NQ in the adverbial position, since the sentence with an adverbial NQ is ambiguous between event-external and event-internal readings if the aktionsart allows it. Thus, contrast (52a), repeated here, and (53): (52) a. Taro-wa sono katana-o mi-furi Taro-TOP that sword-ACC 3-CLswing ‘Taro made three swings of the sword.’ (53) Taro-wa sono katana-o san-kai Taro-TOP that sword-ACC 3-CLtime ‘Taro swung the sword three times.’

shita. did futta. swung

(52a) only refers to a single event, while (53) has the same ambiguity as the English gloss: either Taro swung the sword three times in one occasion, or he swung it on three different occasions. If the numeral should be treated as any sort of quantifier, its being embedded inside the verbal cluster gives the impression that the formal interpretation calculation might be quite complicated. However, for the moment, we might make use of neo-Davidsonean event semantics and arrive at the following rough representation (54a) for the sentence (52a), which might be translated in ordinary English as shown in (54b): (54) a. ∃e'∃e [ e'∈e ∧ SWING(e') ∧ Ag(e')=t ∧ Th(e')=s ∧ | e' | =3 ] b. There is an event e, which has subevents e', whose cardinality is three, and each of which is a swinging event, whose agent is Taro and whose theme is the sword. This represents the interpretation that there is an event consisting of three subevents of swinging of the sword by Taro. However, it has to be further specified that the event e occurs at some specific time. This part would guarantee the single eventuality. Now also consider a sentence such as the following: (55) John-wa konshuu ni-kai ike-o san-shuu shita. John-TOP this week 2-CL pond-ACC 3-CL did ‘John made three rounds of the pond two times this week.’

Japanese numeral quantifiers

191

Such a sentence is certainly possible, even if it may sound slightly clumsy. It means that there are two events, each of which consists of John’s making three circles in the park. Specifically, the first NQ ni-kai ‘two times’ is occasion-related and has wider scope, and the second NQ san-shuu ‘three circles’ is event-related, with a narrower scope. In this case, the logical representation would differ from (54a) in that the higher-order event e has the cardinality two, while the subevent e' has the cardinality three. Syntactic restriction of the quantifier meaning is often evidence of quantifier scope interactions. It might be worth pursuing an analysis in which an adverbial NQ interacts scopally with an implicit tense operator, whereas an NQV is unable to move from its base-generated position and therefore can only have narrow scope. We leave this matter to future research.

5

Concluding remarks

In this chapter, we first looked at some issues surrounding verbal event classifiers in Chinese, a language in which verbal event classifiers are abundant. Then we explored verbal event classifiers in Japanese, which was a survey and the first step of classifying the relevant data. Finally, we discussed the [NQV+suru] construction, a rather productive Japanese construction with an event classifier, with respect to its syntactic and semantic characteristics as a preliminary step to a fullfledged formal investigation. In the linguistics literature, the vast majority of research on numeral classifiers in the past has been on object classifiers. In particular, the floating numeral quantifier construction in Japanese has been much discussed, and some authors have claimed its relation to events (Nakanishi 2007; Kobuchi-Philip 2007). For example, Kobuchi-Philip (2007) discussed the mechanism in which the numeral on the floating numeral quantifier (the NQ in the preverbal, adverbial position) indicates the number of the object represented by the classifier, each of which also has the property represented by the predicate. Thus, it is a mechanism that indirectly counts events. This analysis stems from, among others, the observation that sentences of the floating numeral quantifier construction regularly yield a distributive reading. Consider the following examples: (56) a. san-nin-no otoko-ga kooen-de hana-o ktta. 3-CLperson-GEN men-NOM park-at flower-ACC bought i ‘Three men together bought some flowers in the park.’ ii ‘Three men each bought some flowers in the park.’ b. san-nin-no otoko-ga kooen-de Tanaka-o koroshita. 3-CLperson-GEN men-NOM park-at Tanaka-ACC killed i ‘Three men together killed Tanaka in the park.’ ii ‘Three men each killed Tanaka in the park.’ c. # otoko-ga kooen-de san-nin Tanaka-o koroshita. man-NOM park-at 3-CLperson Tanaka-ACC killed ‘Three men each killed Tanaka in the park.’ In (56a), the NQ is associated with the NP otoko ‘man’ within the DP. This sentence has two interpretations: the collective interpretation (i) and the distributive

192

Mana Kobuchi-Philip

interpretation (ii). It is ambiguous. In (56b), the NQ is in the same syntactic position, but the predicate is a so-called once-only predicate. That is, the event referred to by the predicate can occur only once in the actual world we live in. Thus, even though the sentence yields both a collective reading and a distributive reading, the latter is eliminated from a possible interpretation, and it is interpreted in the collective reading. Now, consider (56c), which has a floating NQ. This sentence is not ambiguous but has only the reading that three men each killed Tanaka. Since this is an impossible state of affairs in our world, this sentence is taken as semantically ill-formed. Data like (56) show that only a distributive reading is available for the floating NQ construction, while both the distributive and collective readings are available for the non-floating NQ construction. The distributive reading indicates a one-to-one relationship between individuals and events. Thus, the numeral in the floating NQ construction is, in a way, counting events, even when its classifier is an object classifier. Counting events in a rather hidden context can be found in the ‘event-related reading’ of a sentence containing a large number (Krifka 1990): (57) Four thousand ships passed through this lock last year. This is an English sentence and has no classifier. The numeral is bare and directly associated with the noun ships. Yet, clearly one can interpret this sentence under an event-related reading as well as under an object-related reading. The object-related reading is straightforward; that is, there were actually four thousand ships which passed through the lock last year, some of which might have passed through the lock more than once. The other, event-related, reading is such that there were four thousand ship-passing events through the lock last year. In this case, some of the ships might have passed through the lock more than once, and in that case, the actual number of ships would be less than four thousand. The latter reading is available in Japanese sentences, both with a floating NQ and a non-floating NQ: (58)

a.

kyonen 4000-soo-no fune-ga kono seki-o last year 4000-CL-GEN ship-NOM this lock-ACC ‘Four thousand ships passed through this lock last year.’ b. kyonen fune-ga kono seki-o 4000-soo last year ship-NOM this lock-ACC 4000-CL ‘Four thousand ships passed through this lock last year.’

tootta. passed tootta. passed

This kind of event-related reading tends to be more easily available when the numeral signifies a very large number. In any case, the point we are making here is that counting events is possible without an event classifier, with an object classifier, without a floating NQ, and even without any classifier. Furthermore, there are also cases when we count objects despite the presence of an event classifier. Consider the examples here: (59)

a.

John-wa jibun-no hoho-o hito-tsumami shita. John-TOP self-GEN cheek-ACC 1-pinch did ‘John gave a pinch on his cheek.’

Japanese numeral quantifiers

(60)

193

b. John-wa nabe-ni shio-o hito-tsumami ireta. John-TOP pot-in salt-ACC 1-pinch put ‘John put a pinch of salt in the pot.’ a. John-wa kyuuri-ni shoyu-o san-teki tarashita. John-TOP cucumber-on soy sauce-ACC 3-drop dropped ‘John dropped three drops of soy sauce on the cucumber.’ b. kona-ni san-teki-no mizu-o kuwaeta. powder-to 3-CL-GEN water-ACC added ‘I added three drops of water into the powder.

Tsumami in (59a) functions as an event classifier of the sort we discussed in subsection 3.4 above. It is a nominal form of a Japanese verb tsumamu ‘to pinch’. Exactly the same NQ functions as an object classifier in (59b), in which tsumami is used as a unit for measurement for salt. Meanwhile, in (60), the classifier is teki, which literally means ‘drop’. In both (60a) and (60b), the NQ san-teki ‘3-drops’ refers to the amount of the liquid, that is, soy sauce in (60a) and water in (60b). Both tsumami and teki have an equivalent lexical item in English, pinch and drop, respectively, and they are used as units for measurement, specifically a unit for certain liquid objects. It is noteworthy since these unit expressions come from certain verbs, to pinch and to drop. Thus, counting events and counting objects seem to be achieved without necessarily depending on a classifier of one or another type. Rather, we make use of the characteristics of each classifier in a creative manner. Having made this remark, the solid presence of the event classifier is even more intriguing. Its formal semantic properties must capture the individualization of the event in question, so one of the questions to investigate is exactly which classes of Japanese verbs are licensed to be used as event classifiers. They must be classes of verbs related to aspectuality and aktionsart. We have remained descriptive throughout this chapter, but we hope this will be the first step toward a comprehensive study of Japanese verbal event classifiers.

Notes 1 Huang and Ahrens (2003) use the term ‘individual classifiers’ for those classifiers that count non-eventual, concrete individuals. However, we find this slightly misleading, since an event can be considered an individual as well. Thus, in this chapter, we use the term ‘object classifier’. However, we use this term only in this chapter to contrast with the event classifier. 2 Some Japanese scholars have commented on verbal event NQs when discussing floating NQs (e.g., Fukushima 1991; Kobuchi-Philip 2007). However, the verbal event NQ was never at the center of the discussion. 3 For discussions of verbal event classifiers in other languages, see, for example, Nguyen (1997) on Vietnamese, Matthews and Leung (2004) on Thai, and Gerner (2009) for various Asian languages. 4 Mv here is Chao’s abbreviation of ‘measure for verbs’. 5 Donazzan (2013) refers to Paris (2011), which was published in Taiwan and which we do not have access to. However, it has exactly the same title as Paris (2013). Although the two seem to be separate papers, we take the liberty of assuming that the main thesis of the two papers is the same.

194

Mana Kobuchi-Philip

6 Here we focus on Zhang (2017); however, Zhang has made other contributions to the study of verbal event classifiers in Chinese, starting with Zhang (2002). 7 Our presentation here of Martin’s (2004) data is sometimes abridged or sometimes simplified. For the original examples and their specific sources, see Martin (2004, pp. 776–777). 8 Here and elsewhere, the numeral ‘one’ in English might correspond to ichi- or hito- in Japanese. The presence of the two is due to the fact that Japanese adopts two systems in counting numbers; one is borrowed from Chinese, which prevails in Modern Japanese, and the other is the original from the Old Japanese, which survives sporadically throughout the use of the numeral in the Modern Japanese. Ichi- belongs to the Chinese system, and hito- belongs to the original Japanese. They can be considered allophones. The original Japanese numerals run only up to 10 and correspond to Chinese in the following way: 1: hito- 2: futa- 3: mi- 4: yo- 5: itsu- 6: mu- 7: nana- 8: ya- 9: koko- 10: too 9 (21a) is an example of this as well. 10 The Chinese character hitsu (筆) can function as another classifier with the Japanese reading fude. However, when it is read as fude and associated with the numeral ‘1’, it means ‘one stroke’, a completely different usage. (i) watashi-ga ip-pitsu kakimasu. (ii) kono ji-o hito-fude-de kakimasu. I NOM 1-brush write this character-ACC 1-brush-with write ‘I will write a letter.’ ‘I will write this character in one stroke.’ 11 Consider more examples of this form: (a) hito-naki suru 1-cry do ‘to cry a while’ (d) hito-yasumi suru 1-rest do ‘to rest a while’ (g) hito-oyogi suru 1-swim do ‘to have a swim’ (j) hito-suberi suru 1-glide do ‘to ski a while’

(b) hito-susugi suru 1-rinse do ‘to give a rinse’ (e) hito-abi suru 1-bathe do ‘to take a shower/bath’ (h) hito-hataraki suru 1-work do ‘to work a while’ (k) hito-katazuke suru 1-tidy_up do ‘to tidy up briefly’

(c) (f) (i) (l)

hito-aruki suru 1-walk do ‘to walk a little’ hito-mooke suru 1-earn do ‘to make money’ hito-mawari suru 1-round do ‘to go around once’ hito-sawagi suru 1-bustle do ‘to have a bustle’

These mostly give rise to the durative reading. 12 We should point out that this selection is not exactly part of the formal property of the nouns but rather pragmatic or contextual. For example, in the world we are familiar with, beer comes in a bottle or a can. Since such containers come in a long shape, we count beer with -hon ‘long objects’ as its unit. However, if John were in space and if beer were supplied in the shape of layered sheets which John consumed like eating a slice of cheese, then the unit would change to -mai ‘flat sheet-like object’. Admittedly, such flexibility seems less likely to occur with event classifiers. 13 In (45b) and (45c), it would be more natural to change ike ‘pond’ to ike-no mawari ‘the circumference of the pond’. The particle -o here is not exactly the same as the -o that is used for the direct object of a transitive verb, but it is the one used for the path/area of moving action, referred to by a motion verb, as in (i)–(ii). (i) (ii)

hikooki-ga sora-o airplane-NOM sky-ACC ‘Airplanes fly in the Taro-wa gobangai-o Taro-TOP Fifth Ave.-ACC ‘Taro walks on Fifth Avenue.’

tobu. flies sky.’ aruku. walk

Air planes fly in the sky, and one walks on a street. That is, the moving activity referred to by the motion verb takes place directly where the o-marked noun refers to. Thus,

Japanese numeral quantifiers

195

ike-o hashiru ‘pond-ACC run’ sounds like one runs in/on a pond. Since we do not run in/ on a pond but rather around a pond in normal circumstances, it is somewhat unnatural. However, (46a) and (46d) seem more acceptable without the specification with ‘circumference’. We speculate that this is the effect of the fact that ike is immediately followed by san-shuu; that is, san-shuu right after ike makes it easy to interpret ike ‘pond’ as ‘the circumference of a pond’. Furthermore, note that such a specification is not necessary in (44), though san-shuu suru ‘make three rounds’ is a motion verb. However, we assume that it does not require the specification with ‘circumference’, since n-shuu suru lexically specifies that the moving action takes place on a circumference. 14 It is well known that Chinese NQs behave differently on this point. In Chinese, an adjective can easily be inserted when the classifier is the mass type (‘massifier’), as in yi da zhang zhi ‘one large sheet of paper’ (Cheng and Sybesma 1999). Assuming that the count type classifier and the mass type classifier in a single language take common structural properties, a Chinese NQ is not likely to be a compound. Accordingly, Zhang (2017) proposes the structure (13) presented in subsection 2.3.

References Bisang, W. 1999. ‘Classifiers in East and and Southeast Asian languages: counting and beyond’, in Gvozdanovic, J., (ed.), Numeral types and changes worldwide, 113–185, Mouton de Gruyter, Berlin. Bisang, W. 2018. ‘Nominal and verbal classification—a comparative perspective’, in McGregor, W. & Wichmann, S., The Diachrony of Classification Systems, 241–282, John Benjamins, Amsterdam. Chao, Y. R. 1968. A grammar of spoken Chinese, University of Southern California Press, Los Angeles. Cheng, L. L.-S. & Sybesma, R. 1999. ‘Bare and Not-So-Bare Nouns and the Structure of NP’, Linguistic Inquiry 30, 509–542. Cusic, D. 1981. Verbal plurality and aspect. PhD thesis, Stanford University. Donazzan, M. 2013. ‘On counting and measuring events’, in E. Chemla, Homer, V., & Winterstein, G. (eds.), Proceedings of Sinn und Bedeutung 17, 219–236, École Normale Supérieure, Paris. Fujita, N. 1994. On the Nature of Modification: A Study of Floating Quantifiers and Related Constructions, PhD thesis, University of Rochester. Fukushima, K. 1991. Generalized floating quantifiers, PhD thesis, University of Arizona. Gerner, M. 2009. ‘Instruments as verb classifiers in Kam (Dong)’, Linguistics 47, 697–742. Grimshaw, J. & Mester, A. 1988. ‘Light verbs and theta-marking’ Linguistic Inquiry 19, 205–232. Harada, S. 1976. ‘Quantifier Float as a Relational Rule’, in Metropolitan Linguistics 1. Linguistic Circle of Tokyo Metropolitan University, 44–49. Huang, C. R. & Ahrens, K. 2003. ‘Individuals, kinds, and events: classifier coercion of nouns’, Language Sciences 25, 353–373. Kageyama, T. 1977. ‘Incorporation and Sino-Japanese verbs’, Papers in Japanese Linguistics 5, 117–156. Kageyama, T. 1991. ‘Light verb constructions and the syntax-morphology interface’, in Nakajima, H., (ed.), Current English Linguistics in Japan, 169–204, Mouton de Gruyter, Berlin. Kamiya, M. 2006. The implausibility of LF-reconstruction in Japanese light verb construction’, University of Pennsylvania Working Papers in Linguistics 12, 149–160. Kitaoka, D. 2014. (Non-) floating numeral quantifiers in Japanese, MA thesis, Memorial University.

196

Mana Kobuchi-Philip

Kobuchi-Philip, M. 2007. ‘Individual-denoting classifiers’, Natural Language Semantics 15, 95–130. Krifka, M. 1990. ‘Four thousand ships passed through the lock: Object-induced measure functions on events,’ Linguistics and Philosophy 13, 487–520. Li, A. 2019. ‘On the Semantics of Verbal Classifiers’, in Ronai, E, Stigliano, L., & Sun, Y. Proceedings of the 54th Annual Meeting of the Chicago Linguistic Society. 261–270. CLS, Chicago. Martin, S. 2004. A reference grammar of Japanese, University of Hawai’i Press, Honolulu. Matsumoto, Y. 1996. ‘A syntactic account of light verb phenomena in Japanese’, Journal of East Asian Linguistics 5, 107–149. Matthews, S. & Leung, T. T.-C. 2004. ‘Verbal vs. nominal classifier constructions in Cantonese and Thai’, in Burusphat, S. (ed.), Papers from the Eleventh Annual Meeting of the Southeast Asian, 445–459. Program for Southeast Asian Studies, Arizona State University. Miyagawa, S. 1989a. Syntax and Semantics 22, Structure and Case Marking in Japanese, Academic Press, New York. Miyagawa, S. 1989b. ‘Light verbs and the ergative hypotheses’, Linguistic Inquiry 20, 659–668. Miyagawa, S. & Arikawa, K. 2007. ‘Locality in syntax and floating numeral quantifiers’, Linguistic Inquiry 38, 645–670. Miyamoto, T. 1999. The light verb construction in Japanese: The role of the verbal noun, John Benjamins, Amsterdam. Nakanishi, K. 2004. ‘Measurement in the nominal and the verbal domains’, Linguistics and Philosophy 30, 235–276. Nakanishi, K. 2007. Formal properties of measuremant constructions, Mouton de Gruyter, Berlin. Nguyen, D. H. 1997. Vietnamese, John Benjamins, Amsterdam. Paris, M.-C. 2013. ‘Verbal reduplication and verbal classifiers in Chinese’, in Cao, G., Chappell, H., Djamouri, R., & Wiebusch, T., Breaking Down the Barriers: interdisciplinary studies in Chinese linguistics and beyond, Vol.1, 257–278. Academia Sinica, Taipei. Partee, B., ter Meulen, A., & Wall, R. 1990. Mathematical Methods in Linguistics. Kluwer, Dordrecht. Sadanobu, T. 1997. ‘Reikai nihongo bunpo; Kaisuu hyoogen-to jutsugo-no kankei. (The relationship beween the iterative expression and the predicate)’, Gengo 26, 38–43, Taishuukan, Tokyo. Saito, M. & Hoshi, H. 2000 ‘Japanese light verb constructions and the minimalist program’, in Martin, R., Michaels, D., & Uriagereka, J. (eds.), Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 261–295. MIT Press, Cambridge, MA. Tamamura, F. 2004. ‘Mono-o Kazoeru・Koto-o Kazoeru (Counting Objects, Counting Events)’, Nihon Gengo Bunka Kenkyuu 6, 1–9. Uchida, Y. & Nakayama, M. 1993. ‘Japanese Verbal Noun Constructions’, Linguistics 31, 623–666. Uegaki, W. & Fujii, S. 2008. ‘Nihongo Shien Dooshi Koobun-ni okeru Idiom-see-to Kisoku-see. (The Idiomaticity and regularity in Japanese support verb construction)’, Proceedings of Natural Language Processing 14, 845–848. Zhang, N. 2002. ‘Counting and Classifying Eventualities in Chinese’, ms., Leibniz-Zentrum Allgemeine Sprachwissenschaft (ZAS), Berlin. Zhang, N. 2017. ‘The syntax of event-internal and event-external verbal classifiers’, Studia Linguistica 71, 266–300.

8

How classifiers affect the mental representation of entities Mutsumi Imai and Junko Kanero

1

Introduction

The count/mass distinction has been noted as one of the most fundamental conceptual distinctions, as it is directly relevant to the identity of entities in the world. Objects are individuated, whereas substances are non-individuated. When we say that two objects are ‘identical’ or ‘the same’, we are referring to ‘two objects in their entirety’ and not to ‘two distinctive parts of a single object’. In contrast, when we talk about the ‘identity’or ‘the sameness’of substances, there is no notion of wholeness. Substances are of ‘scattered existence’, and there is no such thing as ‘whole sand’, ‘whole water’, or ‘whole clay’ (cf. Quine, 1969). This portion of sand is identical to that portion of sand, as long as the two portions consist of the same physical constituents. This difference in identity or sameness between objects and substances leads to fundamentally different extension principles for determination of category membership across the two ontological kinds. For example, the label ‘cup’ is applied to any whole object of a similar ‘cup’ shape that can potentially contain liquid, regardless of its color and material components. If a ‘cup’ is broken into pieces, each piece no longer constitutes a ‘cup’. In contrast, the word ‘clay’ is extended to any portion of clay, irrelevant of shape. One can divide a portion of clay into many small pieces, and each piece is still clay. 1.1  The    gavagai problem How do infants come to know this ontological constraint in word learning? Infants must learn meanings of words by inference from a single or very limited number of examples. In so doing, they face a well-known problem of induction, that is, the “gavagai” problem posed by Quine (1960, 1969): When someone goes to a place where he does not know a single word of the language spoken there, he must guess meanings of words, and the only clue he has is what he can observe in the situation in which a given word is uttered. However, it is virtually impossible to determine the referent of the word in the situation, let alone the meaning of it. If the traveler sees a rabbit and hears the word gavagai spoken at the same time, how could he know whether the word refers to the whole white fluffy animal itself or to an unbounded substance, such as the animal’s fur or its meat?

198 Mutsumi Imai and Junko Kanero Quine (1960, 1969) argues that this indeterminacy of word meanings is a real problem for learners of any language, using the example of how the English words ox and cattle should be translated into Japanese. In Japanese, whenever a noun is enumerated, it has to be accompanied by a classifier. For example, the English phrase five oxen is translated as go too no usi (five CL GEN cattle), in which too, a classifier for big animals, functions as a unit of quantification roughly equivalent to the English classifier head. Thus, a direct translation of five oxen is not possible in Japanese. The closest translation of ‘five oxen’ in Japanese might be ‘five heads of cattle’. Coming back to the problem faced by infants learning their first language, it is critical that they know whether the word in question refers to a bounded thing or to an unbounded thing, because even if an infant could correctly identify the small white animal in front of him as the referent of the word usagi ‘rabbit’, generalization to other referents is only possible when he also knows the ontological principle of word meaning extension – that names for bounded things should be generalized on the basis of the sameness of the thing as a whole, where names for unbounded things should be generalized on the basis of the sameness of its material constituent. How do infants come to know this principle before they start learning the meanings of words? One clue they may have is the difference in the word forms that codify bounded things and unbounded things. For example, in English, infants may come to know the form difference between count nouns and mass nouns and then that the form difference corresponds to a conceptual distinction between bounded things and unbounded things. In fact, Quine (1969) conjectured that this is how (English-speaking) infants acquire the ontological difference between object kinds and substance kinds. 1.2  Count/mass    distinction in language and its  psychological consequences Quine’s conjecture (1969) evoked many questions, and there are at least three directions in which the issue concerning the relation between language and the ontological concept has been addressed. In the first direction, Quine’s thesis about the relation between acquisition of count/mass grammar and the acquisition of the ontological distinction between objects and substances has been questioned. In the second direction, Quine’s assumption that Japanese and/or other classifier languages lack the count/mass distinction has been questioned. The third direction has to do with linguistic relativity (i.e., the Whorfian hypothesis) – the idea that speakers of different languages think differently (Whorf, 1956). If we take Quine’s thesis to an extreme, we must predict that speakers of English-type languages with obligatory count/mass marking and speakers of Japanese-type languages with classifier grammar must have drastically different world views. The former would think that entities in the world are divided into two kinds of things, those that are bounded and individuated and those that are unbounded and non-individuated, whereas the latter would not care about this distinction or perhaps not even notice it. Researchers have asked whether this hypothesis, which can be formulated as a linguistic relativity hypothesis, is tenable.

Classifiers and mental representation 199 1.3   Overview of the chapter This chapter is divided into three sections. In the first part, we address whether Japanese children, whose ambient language does not have transparent and systematic count/mass marking, are able to infer word meanings using the ontological distinction between objects and substances. If Quine’s conjecture is correct, Japanese infants should be unable to learn the ontology-based extension principles for object names and substance names, because they lack the necessary clues in their linguistic input. We review our previous work examining this question (Imai and Gentner, 1997; Imai and Mazuka, 2007; cf. Soja, Carey and Spelke, 1991). We conclude that Quine is incorrect and that children DO acquire the ontological distinction even when their language does not have apparent syntactic marking of count nouns and mass nouns. While finding a cross-linguistically shared appreciation of the ontological distinction, Imai and Gentner (1997) also found a substantial difference across English speakers and Japanese speakers on their construal of boundedness in various types of physical entities. We explore whether the difference between English and Japanese in marking count nouns and mass nouns penetrates into people’s notion of ‘sameness’ in a non-linguistic context and, if so, how the cross-linguistic difference arises. We further review a more recent study that compares the construal of perceptually ambiguous solid entities across speakers of the Chinese and Japanese languages – both with a classifier system but whose ways of syntactic implementation are fairly different (Saalbach and Imai, in preparation) – to discuss how language-specific construal of entities interacts with grammatical properties of the classifier system. In Part II, we ask whether Quine’s linguistic analysis of classifiers (see also Chierchia, 1998; and Lucy, 1992) – Japanese and other classifier languages do not have a syntactic distinction between count nouns and mass nouns – itself is correct in light of the processing in the brain. Several theorists argue against this view and have proposed different analyses of the count/mass status of classifier languages. In particular, some researchers have argued that classifier languages do indeed have count nouns and mass nouns, and the distinction is grammatically realized through the use of distinct classifiers for count nouns and for mass nouns (Cheng and Sybesma, 1998, 1999; Yi, 2009, 2010). We report on experiments that measured event-related potentials (ERPs) to test whether and how the count/mass distinction is processed in Japanese speakers’ brains. We did not find evidence that the count/mass distinction is specifically made in the brain by the use of classifiers. Instead, the results suggest that processing of a classifier phrase (a noun + numeral + classifier; e.g., enpitu ni hon [pencil two long-thing-CL] ‘two pencils’) in general is semantic based, although it may involve some syntactic aspects; the relative weight placed on semantic and syntactic processes varied across different types of nouns paired with different types of classifiers, and the neural process seems to rely more on semantic information when object names were paired with mass classifiers (e.g., enpitsu huta sazi [pencil two spoon] ‘two spoons of pencil’). In the final section, we integrate the two parts to draw general conclusions concerning the linguistic representation of object names and substance names in

200

Mutsumi Imai and Junko Kanero

the minds of Japanese speakers and concerning the relation between language and cognition. We close the chapter by addressing unsolved problems and directions for future research.

2

Part I: influence of language on the ontological distinction and construal of physical entities

2.1   Is Japanese children’s word learning constrained by the  ontological principles? This section addresses the issue of whether Japanese children are able to infer word meanings based on the ontological distinction or whether they get stuck on Quine’s (1960, 1969) “gavagai problem” described earlier. Imai and Gentner (1997) tested this question by comparing Japanese- and English-reared children of three age groups (early 2-year-olds, late 2-year-olds, 4-year-olds) and adults. Imai and Gentner devised a word extension task in which the experimenter introduced a novel word (e.g., dax) in association with an unfamiliar physical entity that the children had never seen before. Participants were presented with a target entity and taught a label for it. They were then shown two test items and were asked to judge to which of the two alternative entities the label should be applied. One of the test items shared the same shape as the target but differed in material. The other alternative entity was the same as the target with respect to material composition but different in shape. A child’s choice of either the same-shape or the same-material alternative was to reveal which of the two dimensions he or she used to generalize the novel label. For English speakers, the novel words were carefully introduced in such a way that participants could not know whether the entity was syntactically seen as a count or a mass noun, e.g., (1) (this dax and the dax could be used for either an object or a substance, in contrast to some dax which could only be used with a substance). (1)

Look at this dax. Can you point to the tray that also has the dax on it?

Because the grammatical structure of Japanese does not reveal the noun’s individuation status, sentences in Japanese naturally did not provide countability information about the target entity, e.g., (2). (The use of bare noun such as dax is a very natural way to refer to either substances or objects in Japanese.) (2)

Kore wa dax desu. Dotira no sara ni this TOP dax is which GEN tray LOC ‘This is a dax. Which tray is a dax on?’

dax ga aru? dax NOM exist

Imai and Gentner (1997) then tested three different types of physical entities. The first type, the complex objects, were real artifact objects that had fairly complex shapes and distinct functions. For example, a T-joint pipe made of plastic (target) was presented along with a metal T-joint pipe (shape test) and broken

Classifiers and mental representation 201 pieces of the target (material test). If the participant pointed to the metal pipe, it was assumed to be an indicator that he or she construed the target entity as a countable object. In contrast, if the participant pointed to the plastic pieces, it would indicate that he or she saw the target entity as an uncountable substance (Figure 8.1a). The second type of entity, the simple objects, had very simple structures with no distinct parts (Figure 8.1b). They were made of a solid substance, such as wax, and were formed into a very simple shape. For example, a kidney-shaped piece of wax (target) was presented together with a kidneyshaped piece of plaster (shape test) and some wax pieces (material test). The third type of entity, the substances, were nonsolid substances, such as sand or hair-setting gel, that were arranged into distinct, interesting shapes when presented. For example, a target of wood chips formed into a U-shape was presented together with tiny leather pieces configured into a U-shape (shape test) and piles of wood chips (material test) (Figure 8.1c). Here, Imai and Gentner hypothesized that solid entities with complex and cohesive structures would be more naturally (and perceptually) individuated than entities with simple structures. They also hypothesized that entities with simple structures would be more naturally individuated than nonsolid substances.

Figure 8.1 Sample material sets for (a) a complex object trial, (b) a simple object trial, (c) a substance trial.

202

Mutsumi Imai and Junko Kanero

2.2   Understanding of the ontological principles  in Japanese children Children and adults in both language groups showed similar classification behavior relying on the entities’ appearance. In the complex object trials, all participants tended to show an object construal and to extend the labels by shape. They showed a substance construal in the substance trials. It seemed that even 2-year-old Japanese children, whose ambient language does not provide a systematic and easily perceptible syntactic distinction between object names and substance names, applied different rules for determining the identity of complex objects and of substances and extended novel words accordingly. 2.3  Cross-linguistic    differences in the construal of entities Although English- and Japanese-reared children both understood the ontological principles for extending object names and substance names, there was a marked difference in how English and Japanese speakers construed the simple objects and the substances. For example, in the simple object trials, English speakers treated the simple-shaped discrete entities in the same way as the complex objects and showed a clear object construal bias, whereas Japanese children did not show any systematic tendency in their classification. In fact, Japanese adults tended to see the simple objects more as lumps of uncountable substances, choosing the material alternative more often than the shape alternative. And in the substance trials, whereas Japanese speakers almost always generalized novel words based on the material identity, English speakers did not show any preference between the shape identity and the material identity. Imai and Gentner’s (1997) results suggested that the ontological distinction between objects (e.g., pipe) and substances (e.g., wax) is understood even among children whose language has no apparent count/mass syntax. This result refutes the strong version of linguistic relativity (e.g., Quine, 1969) and suggests that the ontological distinction is shared universally. At the same time, their results also uncovered noteworthy cross-linguistic differences between the two language groups in a way that was consistent with linguistic relativity (Whorf, 1956; see also Lucy, 1992). Lucy (1992), an anthropological linguist, also advanced the linguistic relativity position. Lucy also assumed that classifier languages treat all nouns as mass nouns. He predicted that speakers of a classifier language should show stronger attention to the material constitution of entities than English-type languages that have the obligatory count/mass grammatical distinction and that they are more likely to construe entities in light of their material composition even outside the realm of language, that is, when simply perceiving and categorizing things. Does the cross-linguistic difference found in the word extension task transfer to classification in a non-linguistic context? Many studies have reported that children are more likely to form adult-like, consistent categories when asked to find new referents of the label given to a target entity) than when asked to find the ‘same’ object without using any labels (e.g., Imai, Gentner and Uchida, 1994; Landau, Smith and Jones, 1988; Markman and Hutchinson, 1984; Waxman and Gelman, 1986; Waxman and Kosowski, 1990). One possibility is that the effect of language is weakened in a no-word classification task.

Classifiers and mental representation 203 To evaluate this possibility, Imai and Mazuka (2007) tested Japanese-speaking and English-speaking 4-year-olds and adults in a no-word classification task. The stimuli and the procedure were the same as in the word extension task used by Imai and Gentner (1997), except that no label was provided. The participants were presented with a target entity and two alternatives and were asked to select an alternative which was the same as the target entity. The English instruction was (3a) and the Japanese instruction was (3b). (3)

a. b.

Show me what’s the same as this. Kore to onazi no wa dotti desuka. this with same one TOP which COP Q ‘Which one is the same as this one?’

The results in general indicated that, across the three trial types, Japanese speakers put more weight on the material than shape in determining the identity of entities (material bias), whereas English speakers put more weight on shape (shape bias). Thus, the cross-linguistic difference found in the word extension task (Imai and Gentner, 1997) was replicated in the no-word categorization task. The detailed analysis revealed that the adults’ performance in this no-word classification task was virtually identical to that observed in the word extension task, as shown in Figure 8.2b (English-speaking adults) and Figure 8.2d (Japanesespeaking adults). In the simple object trials, for example, adult English speakers and adult Japanese speakers showed the opposite classification patterns. But in contrast to adults, children’s classification styles in the no-word classification task were very different from the styles they showed in the word extension task. This discrepancy between the word extension and no-word classification tasks was particularly prominent in English-speaking children (see Figure 8.2a). Whereas in the word extension task, the English-speaking children showed virtually the same response patterns as the adult English speakers, their performance in the no-word categorization task was at a chance level in all three trial types. 2.4  How    do language-specific biases arise? What can be concluded so far from the studies by Imai and Gentner (1997) and Imai and Mazuka (2007)? First, participants’ classification between objects and substances is universally constrained by the ontological distinction, regardless of whether the speaker’s native language grammatically marks this distinction. However, at the same time, it appears that language-specific syntactic structures can influence the construal of entities (like those used in the simple object trials) that are located around the boundary of the two ontological kinds. The structure of the English language seems to bias English speakers toward the object construal (i.e., there is a bias to classify perceptually ambiguous entities based on shape, such as the kidney shape), whereas the structure of the Japanese language seems to bias Japanese speakers toward the substance construal (i.e., there is a bias to classify perceptually ambiguous entities based on material, such as wax). Furthermore, it seems that the language-specific construal of entities first becomes evident in the context of word learning and gradually develops into a general default construal that manifests itself without the invocation of labels.

204

Mutsumi Imai and Junko Kanero

Figure 8.2 Subject’s classification behavior in the no-word context in the word extension (neutral-syntax) tasks and non-lexical classification task: (a) American 4-yearolds, (b) American adults, (c) Japanese 4-year-olds, and (d) Japanese adults. (Adapted from Imai and Mazuka, 2007).

How does the shape bias arise in English speakers? Perhaps because the count/ mass distinction is obligatory, even though Imai and Gentner (1997) presented a novel noun without marking its count/mass status, the English speakers did not encode the noun as having a ‘neutral’ or ‘indeterminate’ syntactic status. In assigning either count or mass syntactic status to the nouns, the children may have assumed by default that the nouns were count nouns rather than mass nouns,

Classifiers and mental representation 205 because the count interpretation is more common for ‘the/this/that X’ (cf. Samuelson and Smith, 1999). The results of a control study by Imai and Mazuka (2007, Experiment 3) supported this possibility. In this experiment, the stimuli and the procedure were exactly the same as those used in the Imai and Gentner (1997) study of word extension with the ambiguous syntax, with one exception: Each novel noun was presented either in the count noun or the mass noun syntactic frame. The participants in the count noun condition heard novel nouns in the count noun syntax throughout across the three trial types (complex object trials, simple object trials, and substance trials) (4a). Likewise, for those in the mass noun condition, the novel nouns were presented in the mass noun syntax in all the trials (4b). (4)

a. b.

Look! This is an X (pointing at the target entity). Can you point to another X? Look! This is X. Can you point to some more X?

As shown in Figure 8.3, when novel nouns were presented in the mass noun syntactic frame, the default classification pattern (i.e., the pattern in the ambiguous syntax case in Imai and Gentner’s 1997 study) was drastically changed by the syntactic markers. The English-speaking adults’ response pattern in the mass syntax condition showed a random response in the complex object trials (48%), presumably because the complex objects invite the object construal very strongly, and it is extremely difficult to construe it as a piece of substance to be consistent with the syntax. In contrast, they showed a material bias in the simple object trials (85% material response). Thus, despite a strong bias toward construing a simple-shaped solid lump of substance as an individuated object, they were capable of mapping a novel label to the material of the entity when prompted to do so by syntactic cues. In the substance trials, again, they selected the material alternative highly above chance level (87%). The response pattern of the English-speaking children in the mass noun condition was overall very similar to that of the adults, showing a random response pattern in the complex object trials and a high rate of material responses in the substance trials (59% and 19.6% shape response, respectively). However, in contrast to the adults, the 4-year-olds’ performance in the simple object trials was at the chance level (46% shape response). Recall that in the word extension task using the ambiguous syntax, the English-speaking children’s shape response level was 91%, almost as high as that for the complex objects (95%). Even though their performance in the simple object trials was still at the chance level with the mass syntax frame, their shape-based responses decreased by 45% from those in the ambiguous syntax case. Therefore, English-speaking 4-year-olds definitely knew that mass noun syntax flags a target entity as a substance (see also Subrahmanyam, Landau and Gelman, 1999, for similar findings). However, their strong bias toward construing any discrete entities as individuated objects (Bloom, 1994; Shipley and Shepperson, 1990) must have made it difficult for them to construe the entities used in the simple object trials as portions of substances. The English speakers’ response pattern in the count syntax condition was intriguing. Their performance here was almost identical to that in the ambiguous syntax word extension task, showing a very high rate of shape responses. This is not surprising for the complex and simple object trials, because the rates of shape

206

Mutsumi Imai and Junko Kanero

Figure 8.3 English speakers’ classification behavior in the (a) neutral-syntax condition (from Imai and Gentner, 1997), (b) count-syntax condition, and (c) mass-syntax condition (from Imai and Mazuka, 2007).

Classifiers and mental representation 207 responses in these two trial types were already at ceiling even in the ambiguous syntax case. For the substance trials, however, both the children and the adults responded randomly, just as in the ambiguous syntax case. This pattern supports the idea that English speakers had assumed that the novel nouns presented in the ambiguous syntactic frame were actually count nouns. Together with the result that English-reared children showed the language-specific shape bias just like adults in the word extension task but not in the no-word classification task, the results support the idea that English speakers’ strong shape bias originates in childhood, when they learn to think that nouns are by default count nouns, and this bias gradually becomes a general cognitive bias that is applied outside the contexts of label extension.    theories pertaining to count/mass distinction  2.5  Revising in classifier languages The studies reviewed thus far (Imai and Gentner, 1997; Imai and Mazuka, 2007) were based on (but not committed to) the linguistic analysis proposed by Quine and other theorists – that classifier languages do not distinguish count nouns and mass nouns and that the grammar of classifier languages treats all nouns as mass nouns (the mass noun hypothesis, Chierchia, 1998; Lucy, 1992; Quine, 1960). However, more recently, this view has been challenged by a group of philosophers and linguists. Their theories differ in details, but they all note that numeral classifiers are divided into two types of classifiers – count classifiers and mass classifiers – and this difference functions as a syntactic marker for distinguishing count nouns and mass nouns in classifier languages. Count classifiers are used for bounded objects and provide a semantic basis for object classification, whereas mass classifiers simply provide units for quantification for unbounded things (usually, but not always, substances). Cheng and Sybesma (1998, 1999) and Doetjes (1997) claimed that the distinction between the two kinds of classifiers is reflected in their syntactic behavior. According to them, Chinese syntactically marks countability not at the level of NP, as is the case with English, but at the level of CLP (classifier phrase). Mizuguchi (2004) applied a similar idea to Japanese and claimed that Japanese also has different types of classifiers, including count classifiers and mass classifiers. Whereas count classifiers such as hon (a classifier mainly used for long objects) quantify countable objects only (see 5a), mass classifiers such as hai (‘cupful’) quantify substances, which must be individuated by containers (see 5b). (5)

a. b.

Pen go hon Koohii go hai

‘5 pens’ ‘5 cups of coffee’

*Koohii go hon *Pen go hai

Yi (2009, 2010) also argues that classifier languages have count nouns and mass nouns, which classifiers morpho-syntactically distinguish. Yi’s count noun hypothesis maintains that classifier languages have robust count nouns that are syntactically distinguished from mass nouns. For instance, in Chinese, nouns that can take the general classifier ge are count nouns. This rule also applies to Japanese and Korean with cognate classifiers ko (or -tu) and kay, respectively.

208

Mutsumi Imai and Junko Kanero

Zhang (2013) also suggests that classifiers in numeral classifier languages are divided into two classes: individuating classifiers and individual classifiers. According to Zhang, nouns in numeral classifier languages are all non-count nouns but differ in [Delimitability] – “the ability of a noun to be modified by a delimitive (size, shape, or boundary) modifier” (Zhang, 2012: 1). Nouns that reject a delimitive modifier (e.g., *big sand, *small water) are [-Delimitable] and mass nouns. Individuating classifiers provide a unit of individuation to these [-Delimitable] nouns. In Mandarin Chinese, for example, shui ‘water’ can be individuated by classifiers such as di ‘drop’, tan ‘puddle’, or bei ‘cup’, as in san di/tan/bei shui (three CL drop/puddle/cupful of water). In contrast, nouns that take a delimitive modifier (e.g., big dog, big apple) are non-mass nouns. These [+Delimitable] nonmass nouns are quantified using individual classifiers, which take the entirety of the referred entity as the unit of individuation (e.g., san ke xigua [three CL-forfruits watermelon] ‘three watermelons’). Individuating classifiers can be used for [+Delimitable] nouns too because, for example, the individuating classifier pian (piece, slice) be used to quantify watermelon (xigua), as in san pian xigua [three CL-pieces watermelon] ‘three pieces of watermelon.’ Words like watermelon change their count/mass status depending on the type of the classifier used for individuation, although in general, [-Delimitable] mass nouns and [+Delimitable] non-mass nouns are quantified by individuating classifiers and individual classifiers, respectively. The two sets of studies by Imai and colleagues (Imai and Gentner, 1997; Imai and Mazuka, 2007) reviewed in sections 2.1–2.3 suggest that differences in the structure of language – whether language obligatorily marks countability of nouns by grammar – do lead to differences in the construal of physical entities, if not the ontological distinction per se. It is important to explore how this conclusion can be generalized to a broader set of languages. In particular, would the object construal bias shown by English speakers also be observed in other languages that have the count/mass syntax? In many languages that mark count/mass distinctions, nouns are syntactically marked not only by count/mass but also by gender. It is worthwhile to examine whether the object construal bias found in English speakers would be extended to speakers of such a language. Second, we may also ask whether the pattern of construals for the complexshaped, simple-shaped, and non-solid substance stimuli shown by Japanese speakers would be generalized to speakers of other classifier languages. The way a classifier system divides nouns can be fairly different across different classifier languages, despite some universally observed semantic features (Croft, 1994). In particular, classifier languages differ in the degree to which classifiers are used in natural discourse. For example, Chinese classifiers must be used not only in numeral phrases (e.g., [numeral + classifier] table) but also in phrases with demonstratives (e.g., this [numeral + classifier] table) (however, the numeral after the demonstrative is often dropped). In contrast, Japanese classifiers are only used with numerals. Furthermore, in Chinese, a classifier functions as a rough equivalent to an indefinite article, while in Japanese, classifiers are only used when it is pragmatically important to specify the number of things in discourse. For example, as an equivalent to the English phrase: I have a cat, Chinese speakers are most likely to say (6a). In contrast, Japanese speakers are most likely to say (6b).

Classifiers and mental representation 209 (6)

a. b.

wo yang yi zhi I raise one small-animal-CL watasi wa neko wo I TOP cat ACC ‘I have a cat/some cats.’

mao. cat katte-iru. raise-state.

Here, the information ‘one’ is not verbalized by Japanese speakers unless this information is pragmatically important, for example, when saying: I have only one cat, but not two, in response to the question: Do you have two cats? It would be expected that these differences would result in a much higher frequency of classifier use in Chinese than in Japanese. To confirm this observation, Saalbach and Imai (2012) compared the frequency of classifier use in a Japanese novel and in its Chinese translation, using the Chinese-Japanese parallel corpus1 of the novel Bottyan (Master Daring) written by Soseki Natsume (1906/1964). In the original Japanese text, there were 111 instances of classifiers, while in the Chinese translation, there were 405 instances. Thus, 294 classifier tokens were added in the course of translating the original Japanese text to Chinese. On closer examination, there were 58 cases in which a classifier was used with ‘one’ (iti) in the Japanese original. In the Chinese translation, there were 156 cases of ‘one’ (yi) with a classifier construction. When the number was ‘two’ or ‘three’, there were 21 classifier instances in Japanese and 53 in Chinese. In the Chinese translation, 175 classifier instances were of the ‘demonstrative + classifier + noun’ construction (e.g., Zhe zhang weirenzhuang [this CL document]). However, in the original Japanese text, all these cases were simple ‘demonstrative + noun’ constructions without a classifier. This study thus revealed that classifiers are used roughly four times as often in Chinese as in Japanese, which is consistent with our structural analysis of the Chinese and Japanese classifier systems. Thus, even though the classifier systems in Japanese and Chinese are organized around similar semantic features, the ways in which classifiers are used in natural discourse are substantially different. Specifically, the classifier system is more deeply integrated in the syntactic system in Chinese than in Japanese. Previous research examining the influence of grammatical system on the construal of the world has focused on the presence/lack of a given grammatical categorization and has not considered how consistency, or frequency, and transparency of the grammatical marking affect the strength of the effect. 2.6   Further exploration of the language-specific biases:  influence of the syntactic implementation in the  construal of physical entities We compared the construals of entities across adult speakers of German, Japanese, and Chinese (Saalbach and Imai, in preparation). Nine stimulus sets were prepared; each consisted of a target and four choice items. The four choice items were: a same-shape item, a same-material item, and two filler items (Figure 8.4). In all sets, the target and the choice items were all solid and simple shaped, which can be seen either as an object or as a piece of a solid substance and induced the

210

Mutsumi Imai and Junko Kanero

Figure 8.4 The stimuli of Saalbach and Imai’s study. Top: Target; Bottom row, from left to right: Same-shape choice, Same-material choice, Same-dimension filler, Different-dimension filler.

greatest cross-linguistic difference between English and Japanese speakers in the previous research by Imai and Gentner (1997). As in Imai and Gentner (1997), a nonsense label was given to the target, and participants were asked to choose the choice item that could also be the referent of the label. Nonsense labels were presented in three different linguistic contexts: (1) count/mass ambiguous (Baseline), (2) count-cue, (3) mass-cue. In the baseline condition, as in Imai and Gentner (1997), the noun’s count/mass status was ambiguous (see Table 8.1 for the instructions). In the second and third conditions, the sentence frame indicated that the label was a count noun or the mass noun, respectively. The instructions for the three conditions in the three languages are given in Table 8.1. The results are shown in Figure 8.5. When the baseline condition was compared, the proportion of the object construal (i.e., choice of the shape item) was largely different across the speakers of the three languages (German: 56.3%, Japanese: 47.4%, Chinese: 29.00%). Importantly, while Japanese speakers were neutral between the object construal and the material construal for these ambiguous stimuli, Chinese speakers showed a strong bias toward the material construal. When the count or mass cue was given, we can see which construal (object or material) would be the default construal for speakers of each language. German speakers were greatly influenced by the mass cue (28.5% decrease of the shape choice) but not so much by the count cue (6.8 % increase of the shape choice). In contrast, for Chinese speakers, while the mass cue (i.e., use of mass classifier)

Classifiers and mental representation 211 Table 8.1 Instructions used by Saalbach and Imai (in preparation). COUNT CL = count classifier. MASS CL = mass classifier or quantifier Count/mass ambiguous (Baseline) English translation German Japanese Chinese Count-cue condition English translation German Japanese Chinese Mass-cue condition English translation German Japanese Chinese

This is called FEP. Can you point to the plate which also has the FEP on it? Das wird FEP genannt. Auf welchem Teller oder auf welchen Tellern ist auch das FEP? Kore wa FEP desu. Kono dochira ga FEP desu ka? Zhe shi FEP. Na yi ge huo zhe na yi xie pan zi shang hai you FEP? This is one [COUNT CL] FEP. Can you point to the plate which also has one [COUNT CL] FEP on it? Das ist ein FEP. Interessant, nicht wahr? Auf welchem Teller oder auf welchen Tellern ist auch das FEP? Koko ni i kko no FEP ga ari masu. FEP wo mou i kko aruiwa ikutsuka erande kudasai. Zhe shi yi ge FEP. Na yi ge huo zhe na yi xie pan zi shang hai you yi ge FEP? This is one [MASS CL] FEP. Can you point to the plate which also has one [MASS CL] FEP on it? Das ist ein Stueck FEP. Auf welchem Teller oder auf welchen Tellern ist auch das FEP? Koko ni hito katamari no FEP ga ari masu. Mou hito katamari no FEP wo erande kudasai. Zhe shi yi kuai FEP. Na yi ge huo zhe na yi xie pan zi shang hai you yi kuai FEP?

increased the dominant mass construal in the baseline condition only by 9%, the count cue (i.e., use of the count classifier) raised the shape choice by 46%. These results suggest that the object construal is the default for German speakers, while the material construal is the default for Chinese speakers. However, the default construal can be flexibly changed by the grammatical cue indicating the nondefault construal for Chinese and German speakers. The story was different for Japanese speakers. They did not show a bias toward either the object or the material construal when a noun was presented without an explicit count/mass cue by a classifier. Furthermore, unlike the case of Chinese speakers, the linguistic cue of the noun’s count/mass status brought only a modest change from the baseline in either direction. Their shape choice increased by 13.4% by the count cue and decreased by 11.50% by the mass cue. These results suggest that, supporting the conclusion by Imai and Mazuka (2007), the influence of the classifier system on the habitual construal of individuation is much less in

212

Mutsumi Imai and Junko Kanero

Figure 8.5 Proportion of shape choice in each condition in (a) German, (b) Japanese, and (c) Chinese.

Classifiers and mental representation 213 Japanese speakers than in Chinese speakers, presumably because the classifier categories are not used frequently enough to induce the material bias (see also Saalbach and Imai, 2012 for converging results on the influence of classifier categories on similarity judgments between objects). 2.7  Discussion    and conclusion of Part I What can we conclude from the series of research reviewed in Part I? First, the extreme linguistic relativity position conjectured by Quine (1969) is incorrect. Speakers of a classifier language do draw the ontological distinction between objects and substances prior to language acquisition. However, the presence of universal appreciation of the ontological concepts does not mean that the structure of language does not affect our construal of the world. Obligatory and transparent marking of the count/mass status of all nouns in English leads its speakers to develop a tendency to construe things, especially when their perceptual affordance is weak, as inherently individuated existence rather than a portion of undividable mass. German speakers have this tendency but in a smaller degree as compared to English speakers, presumably because the syntactic marking of count/mass distinction is less salient in German than in English. As for the influence of classifier system on the object-substance construal of entities, Lucy (1992), who claimed that speakers of classifiers tend to construe physical entities in terms of their material constituents, might be supported after all, given the results from Chinese speakers. However, the material bias does not seem to develop automatically just by having a classifier system in the language. Consistent and frequent use of classifiers in discourse is critical. The finding by Saalbach and Imai (in preparation) in conjunction with earlier studies by Imai and Gentner (1997) and Imai and Mazuka (2007) echoes the suggestion that the influence of language on cognition should not be assumed just by the presence of a given grammatical categorization system (Imai and Masuda, 2013; Imai, Kanero and Masuda, 2016; Saalbach and Imai, 2007, 2012). The scope of the influence should be examined empirically with careful linguistic analyses not only in light of how the grammatical categorization system conceptually divides the world but also how the linguistic device is implemented in natural discourse. The conclusion of Part I also speaks to how we should approach our question to another assumption by Quine (1969) – classifier languages do not have a grammatical count/mass distinction that allows speakers to learn the ontological distinction between objects and substances. Several linguists have argued that classifier languages have two sets of classifiers – one that generally quantifies objects and another that generally quantifies substances (Cheng and Sybesma, 1998, 1999; Doetjes, 1997; Mizuguchi, 2004; Yi, 2009, 2010; Zhang, 2013). Importantly, to discuss how language may affect thought, we additionally need to examine the use of language – whether speakers of the classifier languages indeed recognize and use the grammatical count/mass distinction to identify the ontological status of a noun. Therefore, Part II of this chapter discusses whether the count/mass status of nouns is processed grammatically or conceptually in the mind of speakers of a classifier language.

214

3

Mutsumi Imai and Junko Kanero

Part II: do classifier languages grammatically distinguish count nouns and mass nouns?

3.1  Are    Japanese classifiers processed syntactically or semantically? Empirical examinations As reviewed in Section 2.5 in Part I, there are two different types of classifiers in Japanese. Each usually accompanies an object name or a substance name (Mizuguchi, 2004). Importantly, whether speakers of a classifier language rely on the syntactic rule to differentiate between objects and substances is a separate question. First, as we review in the following section, it is unclear whether classifiers in general are processed at the syntactic level rather than the semantic level. Classifiers play a role in syntax, as the lack of a classifier after a numeral clearly produces an ungrammatical sentence, as in (7). (7)

*imooto wa mikan wo ni tabeta my young sister TOP orange ACC 2 ate ‘My sister ate two oranges.’

The use of an improper classifier [e.g., hon, the classifier for long thin things, as in (8a)] is regarded as anomalous by native speakers of Japanese [compare the grammatical (8b) in which the numeral ni is followed by the classifier for small three-dimensional objects, ko]. However, it is not clear whether Japanese speakers regard this violation as a semantic violation or a syntactic violation. (8)

a. *imooto wa mikan wo ni hon tabeta my young sister TOP orange ACC 2 CL ate b. imooto wa mikan wo ni ko tabeta

If nouns and classifiers are divided into the count and mass categories in the minds of speakers of a classifier language, mismatch of a noun and a classifier that crosses over the count/mass boundary (i.e., a count classifier used with a mass noun or a mass classifier used with a count noun) should be regarded as a syntactic violation, whereas mismatch within the count or mass category (e.g., use of hon, a count classifier, in the place of mai, another count classifier) may be understood as a semantic violation (Cheng and Sybesma, 1998, 1999; Mizuguchi, 2004). To understand whether the count/mass distinction is represented at the level of semantics or of syntax in the minds of classifier language speakers, event-related potentials in the brain can be analyzed. 3.2  Semantic    vs. syntactic processing using event-related  potential technique ERPs are transient electrical signals of the brain that appear in response to external stimulation. An ERP is identified by averaging the electronic activity elicited by a specific type of stimulation across several dozens of trials. Accumulation of

Classifiers and mental representation 215 previous research has identified various ERP components, each of which is considered to reliably reflect specific types of neural processing. The most well-known ERP component in psycholinguistics may be the N400. The N400 is characterized as a negative deflection that appears around the centroparietal region of the scalp approximately 400 ms after a semantically anomalous word or phrase is presented. The N400 was first identified by Kutas and Hillyard (1980) with the stimuli of words within sentences that are semantically anomalous (e.g., I take coffee with cream and dog) or improbable (e.g., He planted string beans in his car) in the context. Since its first discovery, the N400 has been widely recognized as the neural response to a semantic violation (see Kutas and Federmeier, 2011, for a review). Multiple ERP components have been suggested to reflect syntactic processing. First, the left anterior negativity (LAN) has been identified when morphosyntactic violations such as number disagreement, gender disagreement, and incorrect verb inflection are detected (e.g., Osterhout and Mobley, 1995). The temporal pattern of the LAN often overlaps that of the N400. However, the two components are topographically distinctive. The N400 is most pronounced at the center to posterior region of the scalp, whereas the LAN is prominent at the left anterior region. Another signature associated with syntactic violation, the early left anterior negativity (ELAN), is topographically similar to the LAN, but the ELAN appears earlier, around 150–350 ms (e.g., Friederici, Pfeifer and Hahne, 1993; Neville, Nicol, Barss, Forster and Garrett, 1991; Hahne and Friederici, 1999). The ELAN is thought to be associated with violations of word category or phrase structure rules (e.g., Der Freund wurde im besucht. ‘The friend was in the visited’; Friederici et al., 1993). Last, another widely reported ERP component is the P600, which is a relatively long-lasting positivity that appears about 600 ms after the onset of the target stimulus (see Hagoort et al., 1999, for a review). This component was first reported as another index of syntactic processing; however, now it is most commonly recognized as the indication of a more general process of reanalysis, because both semantic and syntactic violations cause the P600. For example, Osterhout and Holcomb (1992) first reported the P600 with incorrect use of transitive verbs, such as (9a), but van Herten, Kolk and Chwilla (2005) observed similar effects with semantic anomalies in Dutch sentences, for example, (9b). (9)

a. b.

The woman persuaded to answer the door. De vos die op de stropers joeg sloop door het bos. the fox[sg] that at the poacher[sg] hunted[sg] stalked through the wood ‘The fox that hunted the poachers stalked through the woods.’

3.3  Previous    event-related potential research on noun  categorization systems Although ERP research on classifiers is limited, ERP has been used to examine the neural processing of other noun categorization systems, such as gender grammar. Barber and Carreiras (2005) observed the LAN and P600 in response to grammatical gender

216

Mutsumi Imai and Junko Kanero

disagreements between an article and a noun in Spanish. This LAN-then-P600 pattern was also found in other studies (Gunter, Friederici and Schriefers, 2000; Barber, Salillas and Carreiras, 2004). Similarly, Münte and Heinze (1994) observed a frontally distributed negativity associated with violation of the gender grammar in German word pairs consisting of an article and a noun or of a pronoun and a verb; see (10). The authors concluded that the negativity reflected syntactic processing (i.e., LAN), suggesting that grammatical gender is processed primarily on a syntactic basis. (10) a. Das Haus The (neutral)-House b. *Der-Haus The (masculine)-House Although some studies failed to observe LAN effects (Barber and Carreiras, 2003; Hagoort and Brown, 1999), the general consensus of the field seems to be that gender disagreement elicits syntactic ERP components, except for in a few special cases (see Barber and Carreiras, 2003). A few ERP studies have examined the neural processing of Japanese numeral classifiers, although none addresses the count/mass distinction in the use of classifiers. Mueller and his colleagues (2005) investigated neural responses to classifier violation in auditorily presented sentences, for example, (11a). In the example, the classifier for birds wa was used to count cats. Their experiment tested two more kinds of violations, that is, word category violation and case violation. In the word category violation condition, the sentence had a missing noun, resulting in an impossible syntactic phrase structure in which the particle no was followed directly by a verb when a noun was expected (11b), where a noun was missing between the particle no and the verb tobikoeru. The case violation involved the misuse of a case-marking particle, such as the use of a nominative particle when an accusative particle would be required, for example, (11c), where the accusative particle wo should follow the second noun phrase ni hiki no neko. (11)

a. Iti wa

no kamo ga ni wa no *neko wo tobikoeru tokoro desu one bird-CL GEN duck NOM two bird-CL GEN cat ACC jump over about to COP ‘A duck is about to jump over two cats.’ b. Iti wa no kamo ga ni hiki no *tobikoeru tokoro desu one bird-CL GEN duck NOM two small-animal-CL GEN jump over about to COP c. Iti wa no kamo ga ni hiki no neko *ga tobikoeru tokoro desu one bird-CLGEN duck NOM two small-animal-CLGEN cat NOM jump over about to COP

Mueller et al. (2005) found that the classifier violation elicited the negativity with left frontal lateralization, whereas word category and case violations resulted in the negativity but with no left lateralization. The authors thus concluded that the negativity observed in the classifier violation condition was the LAN and that Japanese classifiers are processed syntactically rather than semantically. Sakai and her colleagues (2006), however, reached a different conclusion. Their study examined the ERPs elicited by visually presented word pairs of a noun and a classifier. In contrast to congruent pairs (12a)−(12c), incongruent pairs of a noun

Classifiers and mental representation 217 and a classifier (12d)−(12f) showed a strong negativity around 250–550 ms after the onset of the presentation of the classifier. (12) a. enpitu pencil b. Tomodati friend c kami paper d. ki tree e. sensei teacher f. megusuri eye drop

san three san three san three san three san three san three

bon long-object-CL nin human-CL mai flat-object-CL nin human-CL ko small-object-CL tyaku clothes-CL

As the negativity was not lateralized to the left side of the scalp, the authors interpreted the response as the N400 and concluded that Japanese numeral classifiers are processed at the semantic level. Whereas Mueller et al. (2005) argued that the neural processing of classifiers is primarily syntactic based, Sakai et al. (2006) maintained that it is semantic based. It is difficult to draw a clear conclusion as to whether the brain treats classifier violation as syntactic violation or semantic violation. One easy way to resolve this discrepancy is to assume that either or both studies misinterpreted their ERP data. The negativity found by Sakai et al. (2006) may not be the N400, as it was shifted more to the front part of the scalp than the typical N400. On the other hand, the negative deflection observed by Mueller et al. (2005) may not be the LAN, as it lasted much longer than typical LAN effects. The prolonged negativity found by Mueller et al. may reflect increased working memory load rather than syntactic processing (e.g., Marítn-Loeches et al., 2005; Yasunaga and Sakamoto, 2007). There is also a third possibility. The two different results by Mueller et al. (2005) and Sakai et al. (2006) may indicate that the Japanese classifier system is a semantically oriented grammatical system. Classifiers are grammatical morphemes that must accompany nouns with numerals. At the same time, classifiers semantically classify nouns. This differs from the case of grammatical gender, in which assignment of gender class to each noun is in most cases semantically arbitrary (i.e., gender assignment does not reflect the biological sex of the referent of the noun). In summary, no conclusion can be drawn about whether the processing of classifiers recruits a semantic network or a syntactic network, or both, as is also the case for whether classifiers make a grammatical count/mass distinction. The neural processing of classifiers may involve both syntactic and semantic processes, and a classifier violation for a given noun may elicit both semantic and syntactic ERP signatures regardless of the noun’s count/mass status. However, the relative weight on the semantic and syntactic components may differ across different types of classifier violations. In particular, if the count/mass distinction is realized by count

218

Mutsumi Imai and Junko Kanero

classifiers and non-countable mass classifiers, then noun-classifier mismatches that go across the ontological boundary (i.e., an object name combined with a mass classifier or a substance name combined with a count classifier) may invoke stronger syntactic responses in ERPs as compared to those made within the count/mass category boundary (i.e., an object name combined with an inappropriate count classifier or a substance name combined with an inappropriate mass classifier).    event-related potential experiments 3.4  Our To investigate whether speakers of a classifier language syntactically process the count/mass distinction at CLP, we conducted two ERP experiments (Kanero, Imai, Okada and Hoshino, 2015). Experiment 1 was similar to the study by Sakai et al. (2006) and examined ERP responses to mismatching NP-CLP pairs but compared mismatches within and across the count-mass boundary. 3.4.1 3.4.1.1

Experiment 1 STIMULI AND PROCEDURE

Experiment 1 tested four different conditions: a matched (control) condition, a violation within the count/mass category condition (the within-count/masscategory violation condition), a violation across the count/mass category boundary condition (the across-count/mass-category condition), and the animal-non-animal violation condition. Half of the nouns used were names of objects, and the other half were names of substances. The same set of nouns was used across all four conditions, but they were accompanied by different CLPs depending on the conditions. Thus, each object or substance name appeared four times throughout the whole experiment. Japanese classifiers are always used with a numeral, and thus all classifiers were presented in CLP with the number ‘two’ (e.g., ni hon [two + CL for long things]). In the animal-non-animal violation condition, an object name or a substance name was paired with a classifier for counting animals (13d and 13h). Animals are never counted using classifiers that are associated with non-animals, and nonanimals are never counted by animal classifiers. The animal-non-animal violation condition is a very strong and clear mismatch between an NP and a CLP, and we used the condition as a reference to evaluate the amplitude of EPRs in the withinor across-count/mass-category violation. In the within-count/mass-category violation, either an object name was followed by an incongruent count (individual) classifier or a substance name was followed by an incongruent mass (individuating) classifier. In the across-count/mass-category violation, either an object name was followed by an incongruent mass classifier or a substance name was followed by an incongruent count classifier. For instance, the object name hude ‘brush’ was paired with ken (count classifier for houses and other buildings) in the within-category-violation condition (13b), kire ‘a piece/slice of’ in the across-category-violation condition (13c), and hiki (small animal classifier) in the animal-non-animal violation condition (13d). Likewise, the substance name sio ‘salt’ was paired with kire in the within-category-violation condition (13f), dai

Classifiers and mental representation 219 (classifier for machines and functional artifacts) in the across-category-violation condition (13g), and wa (bird classifier) in the animal-non-animal violation condition (13h). In the matched condition, the target object and substance nouns were accompanied by their proper classifiers (13a) and (13e). (13) a. b. c. d. e. f. g. h.

object matched condition: hude ni hon object within-count-category violation: hude ni ken object across-count-category violation: hude ni kire object animal-non-animal violation: hude ni hiki substance matched condition: sio huta sazi substance within-mass-category violation: sio huta kire substance across-mass-category violation: sio ni dai substance animal-non-animal violation: sio ni wa

All nouns and classifiers were selected from commonly used vocabulary to create different combinations of nouns and CLPs. Thirty native Japanese speakers who did not participate in the main ERP experiment rated each noun-classifier pair for their degree of match on a scale of 1–5 (1: highly mismatched, 2: somewhat mismatched, 3: neither matched nor mismatched, 4: somewhat matched, 5: highly matched). All congruent noun-classifier pairs used in the matched condition were rated as highly matched, whereas all incongruent pairs were rated as highly mismatched (mean rating score: 4.60 for the matched, 1.31 for the within-count/ mass-category violation, 1.13 for the across-count/mass-category violation, and 1.04 for the animal-non-animal violation). A noun was considered an ‘object noun’ when the entity denoted by the noun had a clear boundary and its identity would be lost if it is broken into pieces, for example, (14a). A noun was considered a ‘substance noun’ when the denoted entity had no boundary and passed the universal grinder test suggested by Pelletier (1979), for example, (14b). Labeling classifiers as ‘count classifiers’ and ‘mass classifiers’ was not as straightforward, as researchers used different criteria for this classification (Zhang, 2013). In our study, we considered classifiers whose unit of counting coincides with the whole of an object, that is, individual classifiers by Zhang’s term (15a), to be ‘count classifiers,’ and classifiers that provide a unit of counting for a segment of a thing (15b) or liquid (15c) to be ‘mass classifiers’. (14) a. object noun: koppu ‘cup’, zitensya ‘bicycle’ b. substance noun: abura ‘oil’, hatimitu ‘honey’ (15) a. count classifiers: hon, mai, dai b. mass classifiers (thing): kire ‘piece/slice’, kakera ‘piece/chunk’, katamari ‘lump’ c. mass classifiers (liquid): hai ‘cup’, hukuro ‘package’, bin ‘bottle’ A noun (e.g., hude ‘brush’) and a CLP (e.g., ni hiki [2 + classifier for small animals]) were presented consecutively on a computer screen, each for 800 ms. Twenty-five native Japanese speakers were asked to press a ‘correct’ or ‘incorrect’ button to indicate if the classifier was appropriate to count the noun. EEGs were recorded from 32 electrodes placed on the scalp, and ERP signals after the presentation of the CLP were examined.

220

Mutsumi Imai and Junko Kanero

3.4.1.2

RESULTS AND DISCUSSION

As did Sakai et al. (2006), we found a negative deflection around 300–500 ms after the onset of CLP presentation in all three violation conditions involving both object and substance nouns. The effect was widespread but the strongest at the center of the scalp, as shown in Figures 8.6a and 8.6b. A slight shift of the pattern toward the anterior region was observed for all violation conditions except for the across-category violation between object noun and mass classifier (see below for the discussion of this result). Thus, the observed negativity was not a typical N400, for which the greatest deflection is found in the central-parietal part of the scalp, nor was it a typical LAN, as the effect was not limited to the left anterior region. However, the negativity was more similar to an N400, because the lateralization was very weak. Alternatively, the negative deflection may reflect both N400 and LAN. Some researchers suggest that, when semantic and morphosyntactic processes simultaneously take place, a hybrid of N400 and LAN effects can be observed (de Vega, Urrutia and Dominguez, 2010; Thierry, Cardebat and Demonet, 2003). Although it is difficult to conclude whether the negativity observed in the study is an N400 or an N400-LAN hybrid, our results clearly show that the violation of noun-classifier matching largely elicits the same ERP responses regardless of whether the violation went across the count/mass boundary. The incongruent pairs of non-animal object nouns and animal classifiers resulted in a larger ERP than the (non-animal) within- or across-count/mass-category violation pairs (see Figure 8.6a). This result demonstrated that the clear separation of

Figure 8.6a Grand average ERPs for the trials involving object nouns in the violation conditions (dotted lines) in contrast to the matched condition (solid lines).

Classifiers and mental representation 221

Figure 8.6b Grand average ERPs for the trials involving substance nouns in the violation conditions (dotted lines) in contrast to the matched condition (solid lines).

animal and non-animal classifiers in the classifier system in Japanese invokes the strongest effects when processing classifier phrases. Experiment 1 did not find evidence supporting the idea that classifier systems grammatically mark the count/mass distinction (Cheng and Sybesma, 1998, 1999; Mizuguchi, 2004). However, there remains a concern about the ecological validity of Experiment 1, because classifiers are usually embedded in a sentence rather than in a simple phrase consisting of an NP and a CLP. The discrepancy between our data and those found in Mueller et al. (2005), who reported the LAN for noun– classifier mismatches, could be due to the fact they embedded classifier phrases in a sentence, whereas we presented NPs and CLPs in isolation. To evaluate the possibility, we conducted Experiment 2 in which NPs and CLPs were embedded in short sentences. In addition, semantically implausible sentences [e.g., Enpitu wo ni-hon tabeta; ‘(Someone) ate 2 pencils’] were also added to ensure that participants paid attention to the entire sentence rather than focusing only on the noun and the corresponding CLP in the sentence. 3.4.2 3.4.2.1

Experiment 2 STIMULI AND PROCEDURE

We again contrasted the matched trials against violation trials but only included the within- and across-count/mass-category violations. Each sentence (e.g., Hagaki-wo

222

Mutsumi Imai and Junko Kanero

itimai katta [postcard-ACC 1-CL bought]) was presented in three segments: NP (a noun + a particle), CLP (a numeral + a classifier), and a verb (16). Semantically incongruent sentences were additionally included as filler trials. In the semantically incongruent trials, the congruency of a sentence could not be judged until the last word. For example, one of the sentences ‘Mansyon wo ni-tou nonda [apartment ACC 2-CL drank]’ which roughly means ‘(someone) drank two apartments.’ (16) a. Count classifier matched trials: Hagaki-wo iti-mai katta postcard-ACC 1-[count classifier] bought b. Count classifier within-count-category violation: Sentakuki-wo iti-mai* katta washing machine-ACC 1-[mismatched count classifier] bought c. Count classifier across-count/mass-category violation: Komugiko-wo iti-mai* katta flour-ACC 1-[mismatched count classifier] bought d. Mass classifier matched trials: Zyamu-wo huta-bin ageta jam-ACC 2-[mass classifier] gave e. Mass classifier within-mass-category violation: Piza-wo huta-bin* ageta pizza-ACC 2-[mismatched mass classifier] gave f. Mass classifier across-count/mass-category violation: Sukaahu-wo huta-bin* ageta scarf-ACC 2-[mismatched mass classifier] gave The stimuli were presented in the same manner as in Experiment 1, except that each segment of a sentence was presented for 200 ms, with 500 ms intervals between segments. In 25% of the trials, participants pressed a button to indicate whether the sentence was grammatically or semantically ‘correct’ or ‘incorrect’. Thirty-six native Japanese speakers participated in Experiment 2. The EEGs were recorded from the same 32 electrodes and processed in the same manner as in Experiment 1. 3.4.2.2

RESULTS AND DISCUSSION

Experiment 2 extended Experiment 1 by embedding nouns and classifiers into sentences. As shown in Figures 8.7a and 8.7b, a negative deflection was observed in both within-count/mass-category violation trials and across-count/mass-category violation trials around 400 ms after the onset of classifier presentation. This negativity appeared to be the typical N400 with a centro-parietal distribution. Consistent with Experiment 1, the within- and across-count/mass-category violations showed almost identical ERP signatures, again disconfirming the claim that speakers of classifier languages grammatically distinguish objects and substances through the selective use of classifiers (Cheng and Sybesma, 1998, 1999; Mizuguchi, 2004). The negative deflection was relatively weaker for the pairings of object

Figure 8.7a Grand average ERPs for the trials involving count classifiers in the violation conditions (dotted lines) in contrast to the matched condition (solid lines).

Figure 8.7b Grand average ERPs for the trials involving mass classifiers in the violation conditions (dotted lines) in contrast to the matched condition (solid lines).

224

Mutsumi Imai and Junko Kanero

nouns and incongruent mass classifiers, which was also similar to the results of Experiment 1. In addition, a positive deflection resembling the P600 was observed in the ERP, although it did not reach the level of significance. As stated earlier, the P600 is another widely reported ERP component that appears about 500 ms after the onset of the target stimulus (see Hagoort, Brown and Osterhout, 1999, for a review), which reflects the integration cost of expectations for upcoming materials in the processing of complex sequences (Christiansen, Conway and Onnis, 2012). The P600 effects were not strong in this study, possibly because the sentences we used had a relatively simple structure with only three segments. Importantly, the P600like effects were not different between the within-category and across-category violation conditions. 3.4.3

Discussion of Part II

The goal of our ERP experiments was to empirically investigate the proposal that classifier languages grammatically mark the conceptual distinction between objects and substances through the selective use of count classifiers and mass classifiers (Cheng and Sybesma, 1998, 1999, Mizuguchi, 2004). We hypothesized that, if the count/mass distinction is realized at the level of CLP, and if this linguistic device bears psychological reality, this distinction would be revealed in ERP responses. Specifically, we expected to see the LAN, as opposed to the N400, when the noun-classifier mismatch went across the count/mass boundary as opposed to when the mismatch occurred within the count/mass boundary. The results did not support this hypothesis, and we found N400 effects regardless of whether nouns and classifiers were presented in isolation (Experiment 1) or embedded in sentences (Experiment 2) and, more critically, whether the violation of noun-classifier pairing was made within or across the count/mass ontological boundary. These results suggest that (a) the processing of classifier phrases is in general semantically oriented, and (b) detection of anomaly of the countability status in CLP is not more ‘syntactic’ than detection of other types of anomaly in the noun-classifier pairings. The finding challenges the hypothesis that the count/ mass distinction is syntactically made at the level of CLP in Japanese. Our results indicate that, in classifier languages, the ontological object/substance distinction is realized at the lexico-semantic level, unlike languages with count/mass syntax like English. The present study demonstrated that Japanese-speaking adults do not rely on the classifier system to distinguish between count nouns and mass nouns. As mentioned earlier, Imai and Gentner (1997) demonstrated that Japanese children distinguish objects and substances and generalize the meaning of a novel word accordingly. Our ERP experiments suggest that perceptual properties of entities and experience with conventional word use may be sufficient for Japanese-speaking children to realize whether the novel noun refers to an object or substance (see Yoshida and Smith, 2003 for a similar view). It is, however, still possible that children’s awareness of the count/mass distinction is strengthened to some degree by the use of numeral classifiers (Li, Barner and Huang, 2008). For instance, young children may be sensitive to the probability of co-occurrence between a particular noun

Classifiers and mental representation 225 and a classifier and infer that nouns that are always accompanied by the same classifiers are countable, whereas nouns that are accompanied by multiple classifiers across situations are uncountable – a possibility that should be further examined in the future (see the next section for further discussion). It is also informative to conduct an ERP experiment in another classifier language such as Chinese, which has more strict and frequent use of classifiers. Although the Japanese language has 150 numeral classifiers, only around 80 classifiers are found in daily use (Downing, 1996). In Chinese, on the other hand, the classifier system seems to play a more important role. For example, as discussed in Part I, classifiers are used not only in numeral phrases (e.g., [numeral + classifier] table), but also in phrases with demonstratives (e.g., this [numeral + classifier] table). Therefore, compared to Japanese, Chinese uses four times as many classifiers and shows a stronger influence of classifier categories on non-linguistic cognition (Saalbach and Imai, 2012). Our experiments clearly demonstrated that Japanese numeral classifiers are processed primarily based on their semantic characteristics, and the grammatical count/mass distinction is not realized at the level of classifier phrase. However, it is still possible that the processing of classifiers qualitatively differs across languages. Further research is required to fully understand the role of the classifier system in the count/mass distinction.

4

Conclusions and future research

This chapter has explored how adult Japanese speakers and Japanese-reared young children represent the ontological distinction with respect to individuation. The ontological distinction between object kinds and substance kinds is appreciated and is used as a constraint for word learning by Japanese-reared children at 24 months of age. Thus, it is likely that children come to appreciate the ontological distinction very early, even when their native language does not grammatically mark the distinction, and in this sense, syntactic bootstrapping is not necessary for children to acquire object names and substance names. One may be concerned that it is too early to conclude that classifiers do not contribute to children’s awareness of the ontological difference between objects and substances. In fact, our results suggest that the neural processing of mass classifiers is distinctive. It is thus theoretically possible to think that young children use classifiers to acquire an awareness of the count/mass distinction: children may know that whereas object names usually each have only one matching classifier, substance names can be paired with various classifiers depending on the situation. Making use of the probability of co-occurrence of particular nouns and particular classifiers, they may realize that a noun that is always accompanied by a single classifier belongs to the category of object kinds, while a noun that is accompanied by various classifiers across different situations belongs to the category of substance kinds. This possibility would be supported if children’s awareness of the distinction between objects and substances changed after they master the use of classifiers. However, we do not think that this possibility is likely. The acquisition of classifiers is much slower than the acquisition of nouns, and Japanese children as old as age 5½ are still not fully familiar with the use of classifiers (Uchida and Imai,

226

Mutsumi Imai and Junko Kanero

1999). Considering this, it is more probable that the direction is the other way around: The ontological distinction, which exists from an early age, is used to learn the matching between nouns and classifiers (see Sato and Haryu, 2006, for some evidence for this possibility). Of course, it is possible that children do use classifier information after acquiring the meanings of classifiers for inference of noun meanings. For example, when hearing a novel noun neke with the classifier teki (‘drop’) in a context where no other cues are available, the child may infer that neke is a kind of liquid if she knows the meaning of this classifier. Chien, Lust and Chiang (2003) in fact argue that classifiers could bootstrap noun learning. However, in our view, classifier information, if used at all, can serve only as a weak and secondary cue, which is used in rare situations when the classifier at hand is specific and known by the child but perceptual and social cues are unavailable (Brandone et al., 2007). In actuality, the classifiers children frequently hear in everyday settings such as (17a) are very broad in their applications and do not sufficiently constrain the meaning of the noun associated with them. More specific, category-based classifiers such as (17b) are usually acquired only through formal school education, so by the time children learn them, they are likely to have already learned the nouns for which these classifiers are used. (17) a. ko (for small three-dimensional objects), hon (for long and thin things), mai (for flat things), hiki (for small animals), tou (for large or important animals), dai (for machines and some functional artifacts) b. satu (for books and other bounded reading materials), wa (for birds), sao (for clothes cabinets) Even though classifiers are not so useful for the purpose of constraining the meaning of nouns, asking whether the ontological distinction between object nouns and substance nouns is syntactically processed by adult Japanese speakers is still worthwhile. The results of our experiment suggest that count classifiers and mass classifiers are processed differently. When the participants were simply matching a noun and a classifier, the classifier phrase seemed to be processed both semantically and syntactically. However, to match object nouns and mass classifiers, a more flexible approach is required, and thus the semantic nature of classifiers seems to be considered more heavily. These results provide insights into the issue of how we should characterize grammatical categorization systems in general. Traditionally, researchers tend to draw a clear distinction between semantics and syntax. However, the way the classifier system is processed in the brain suggests that such a binary categorization does not properly reflect reality and that we should explore how semantics and syntax are integrated in the brain. Given this linguistic difference between Japanese and Chinese and its psychological consequence on the construal of physical entities, as reported in section 2.5, it is critical to extend our ERP study to Chinese and other numeral classifier languages. Close examination of the way the semantic and syntactic nature of nouns and classifiers in a given classifier language correlates with ERP responses of its speakers will inform us whether and how the count/mass distinction is represented

Classifiers and mental representation 227 in the mind. This, in turn, might further help us understand the interaction between semantic and syntactic processing. The lack of a grammatical count/mass distinction and the use of a numeral classifier system are prominent features that make Japanese distinctive from many other languages such as English. The uniqueness of the Japanese classifier system enriches research on how Japanese treats a range of fundamental concepts such as number, animacy, and the object/substance distinction. Further, it is a window into the bigger picture of how language systems and the ontological understanding of the world are interrelated. Developmental and neurophysiological research suggests that the numeral classifier system does not serve as a primary basis for the ontological object/substance distinction in Japanese speakers. However, it also demonstrates that Japanese and English speakers rely on different cues to judge the ‘sameness’ of entities, suggesting some influence of having the classifier system to non-linguistic concepts. Our ERP studies indicate that the Japanese classifier system is not a set of strictly grammatical systems but a complex system that integrates both semantic and syntactic information. Future research including examination of other classifier languages is needed to further reveal the universal nature of classifier systems as well as the unique nature of the Japanese classifier system.

Acknowledgments This chapter was written by updating another article by the authors published in the Handbook of Japanese Psycholinguistics (Imai and Kanero, 2015), and some texts and images are borrowed from the original article. The preparation of this chapter was supported by MEXT/JSPS KAKENHI Grant Numbers 16K13224, 16H01928, 18H05084 to MI.

Note 1 Chinese-Japanese parallel corpus by the Beijing Center for Japanese Studies was used.

References Barber, H. and M. Carreiras. 2003. Integrating gender and number information in Spanish word pairs: An ERP study. Cortex 39(3). 465–482. Barber, H. and M. Carreiras. 2005. Grammatical gender and number agreement in Spanish: An ERP comparison. Journal of Cognitive Neuroscience 17(1). 137–153. Barber, H., E. Salillas and M. Carreiras. 2004. Gender or genders agreement? In M. Carreiras and C. Clifton (eds.), On-line Study of Sentence Comprehension: Eye-tracking, ERP and Beyond. 309–327. Brighton, UK: Psychology Press. Bloom, P. 1994. Possible names: The role of syntax-semantics mappings in the acquisition of nominals. Lingua 92. 297–329. Brandone, A. C., K. Pence, R. M. Golinkoff and K. Hirsh-Pasek. 2007. Action speaks louder than words: Young children differentially weight perceptual, social, and linguistic cues to learn verbs. Child Development 78(4). 1322–1342. Cheng, Lisa Lai-Shen and Rint Sybesma. 1998. Yi-wan tang, yi-ge tang: Dan wei ci yu wu hua ci. [Yi-wan tang, yi-ge tang: Classifiers and massifiers.] Tsing-Hua Journal of Chinese Studies 28(3). 385–412.

228

Mutsumi Imai and Junko Kanero

Cheng, Lisa Lai-Shen and Rint Sybesma. 1999. Bare and not so bare nouns and the structure of NP. Linguistic Inquiry 30(4). 509–542. Chien, Yu-Chin, B. Lust and Chi-Pang Chiang. 2003. Chinese children’s comprehension of count-classifiers and mass-classifiers. Journal of East Asian Linguistics 12(2). 91–120. Chierchia, G. 1998. Reference to kinds across languages. Natural Language Semantics 6(4). 339–405. Christiansen, M. H., C. M. Conway, and L. Onnis. 2012. Similar neural correlates for language and sequential learning: Evidence from event-related brain potentials. Language and Cognitive Processes 27(2). 231–256. Croft, W. 1994. Semantic universals in classifier systems. Word 45(2). 145–171. de Vega, M., M. Urrutia and A. Dominguez. 2010. Tracking lexical and syntactic processes of verb morphology with ERP. Journal of Neurolinguistics 23(4). 400–415. Doetjes, J. 1997. Quantifiers and Selection: On the Distribution of Quantifying Expressions in French, Dutch and English. Hague, HAG: Leiden University Dissertation. Downing, P. A. 1996. Numeral Classifier Systems: The Case of Japanese. Vol. 4. Amsterdam, Netherlands: John Benjamins Publishing. Friederici, A. D., E. Pfeifer and A. Hahne. 1993. Event-related brain potentials during natural speech processing: Effects of semantic, morphological and syntactic violations. Cognitive Brain Research 1(3). 183–192. Gunter, T. C., A. D. Friederici and H. Schriefers. 2000. Syntactic gender and semantic expectancy: ERPs reveal early autonomy and late interaction. Journal of Cognitive Neuroscience 12(4). 556–568. Hagoort, P. and C. M. Brown. 1999. Gender electrified: ERP evidence on the syntactic nature of gender processing. Journal of Psycholinguistic Research 28(6). 715–728. Hagoort, P., C. M. Brown and L. Osterhout. 1999. The neurocognition of syntactic processing. In C. M. Brown and P. Hagoort (eds.), The Neurocognition of Language, 273–316. Oxford, UK: Oxford University Press. Hahne, A. and A. D. Friederici. 1999. Electrophysiological evidence for two steps in syntactic analysis: Early automatic and late controlled processes. Journal of Cognitive Neuroscience 11(2). 194–205. Imai, Mutsumi and Dedre Gentner. 1997. A crosslinguistic study of early word meaning: Universal ontology and linguistic influence. Cognition 62(2). 169–200. Imai, Mutsumi, Dedre Gentner and Nobuko Uchida. 1994. Children’s theories of word meaning: The role of shape similarity in early acquisition. Cognitive Development 9(1). 45–75. Imai, Mutsumi, Junko Kanero, and Takahiko Masuda. 2016. The relation between language, culture, and thought. Current Opinion in Psychology 8. 70–77. Imai, Mutsumi, and Takahiko Masuda. 2013. The role of language and culture in universality and diversity of human concepts. Advances in Culture and Psychology 3. 1–65. Imai, Mutsumi, and Junko Kanero. 2015. The nature of the count/mass distinction in Japanese. In. J.J. Nakayama (ed.), Handbook of Japanese Psycholinguistics (In Handbooks of Japanese Language and Linguistics Series), pp. 49–79. Berlin, Germany: De Gruyter Mouton. Imai, Mutsumi, Junko Kanero, and Takahiko Masuda. 2016. The relation between language, culture, and thought. Current Opinion in Psychology 8. 70–77. Imai, Mutsumi and Reiko Mazuka. 2007. Revisiting language universals and linguistic relativity: Language-relative construal of individuation constrained by universal ontology. Cognitive Science 31(3). 385–414. Kanero, Junko, Mutsumi Imai, Hiroyuki Okada and Noriko Hoshino. 2015. Do classifiers make the syntactic count/mass distinction? Insights from ERPs in classifier processing in Japanese. Journal of Memory and Language 83. 20–52.

Classifiers and mental representation 229 Kutas, M. and Kara D. Federmeier. 2011. Thirty years and counting: Finding meaning in the N400 component of the event-related brain potential (ERP). Annual Review of Psychology 62. 621–647. Kutas, M. and S. A. Hillyard. 1980. Reading senseless sentences: Brain potentials reflect semantic incongruity. Science 207(4427). 203–208. Landau, B., L. B. Smith and S. S. Jones. 1988. The importance of shape in early lexical learning. Cognitive Development 3(3). 299–321. Li, Peggy, David Barner, and Becky H. Huang. 2008. Classifiers as count syntax: Individuation and measurement in the acquisition of Mandarin Chinese. Language Learning and Development 4(4). 249–290. Lucy, J. A. 1992. Language Diversity and Thought: A Reformulation of the Linguistic Relativity Hypothesis. New York, NY: Cambridge University Press. Markman, E. M. and J. E. Hutchinson. 1984. Children’s sensitivity to constraints on word meaning: Taxonomic versus thematic relations. Cognitive Psychology 16(1). 1–27. Martín-Loeches, M., F. Muñoz, P. Casado, A. Melcón and C. Fernández-Frías. 2005. Are the anterior negativities to grammatical violations indexing working memory? Psychophysiology 42(5). 508–519. Mizuguchi, Shinobu. 2004. Individuation in Numeral Classifier Languages: A Case of Japanese Classifiers and Plurals. Tokyo: Shohakusha. Mueller, J. L., A. Hahne, Yugo Fujii and A. D. Friederici. 2005. Native and non-native speakers’ processing of a miniature version of Japanese as revealed by ERPs. Journal of Cognitive Neuroscience 17(8). 1229–1244. Münte, T. F. and H.-J. Heinze. 1994. ERP negativities during syntactic processing of written words. In H.-J. Heinze, T. F. Münte and G. R. Mangun (eds.), Cognitive Electrophysiology, 211–238. Boston, MA: Birkhauser. Natsume, Soseki. 1906/1964. Bottyan [Master Daring]. Tokyo: Kaiseisha. Neville, H., J. L. Nicol, A. Barss, K. I. Forster and M. F. Garrett. 1991. Syntactically based sentence processing classes: Evidence from event-related brain potentials. Journal of Cognitive Neuroscience 3(2). 151–165. Osterhout, L. and P. J. Holcomb. 1992. Event-related brain potentials elicited by syntactic anomaly. Journal of Memory and Language 31(6). 785–806. Osterhout, L. and L. A. Mobley. 1995. Event-related brain potentials elicited by failure to agree. Journal of Memory and Language 34(6). 739–773. Pelletier, F. J. 1979. Non-singular reference. In F. J. Pelletier (eds.), Mass Terms: Some Philosophical Problems, 1–14. Dordrecht: Kluwer Academic Publishers. Quine, W. V. O. 1960. Word and Object. Cambridge, MA: MIT Press. Quine, W. V. O. 1969. Ontological Relativity and Other Essays. New York, NY: Columbia University Press. Saalbach, H. and Mutsumi Imai. 2012. The relation between linguistic categories and cognition: The case of numeral classifiers. Language and Cognitive Processes 27. 381–428. Saalbach, H. and Mutsumi Imai. In preparation. How do grammatical cues of count/mass in German, Chinese and Japanese affect the construal of entities? Saalbach, H. and Mutsumi Imai. 2007. Scope of linguistic influence: Does a classifier system alter object concepts? Journal of Experimental Psychology: General 136(3). 485–501. Sakai, Yumi, Kazuki Iwata, J. Riera, Xiaohong Wan, Satoru Yokoyama, Yoshiteru Shimoda, Ryuta Kawashima, Kei Yoshimoto and Masatoshi Koizumi. 2006. Jishō kanren den-i de miru meishi to josūshi no shōgō purosesu – imi-teki shori ka bunpōteki shori ka. [An ERP study of the integration process between a noun and a

230

Mutsumi Imai and Junko Kanero

numeral classifier: Semantic or morpho-syntactic?] Cognitive Studies 13(3). 443–454. Sakamoto, Tsutomu. 2015. Processing of syntactic and semantic information in the human brain: Evidence from ERP studies in Japanese. In Mineharu Nakayama (ed.), Handbook of Japanese Psycholinguistics, 457–509. Boston, MA: Mouton de Gruyter. Samuelson, L. K. and L. B. Smith. 1999. Early noun vocabularies: Do ontology, category organization and syntax correspond? Cognition 73(1). 1–33. Sato, Kensuke and Etsuko Haryu. 2006. Yoji ni okeru josūshi no rikai: sonzairon-teki kategori ni chūmokushite. [Acquisition of numerical classifiers by Japanese preschoolers: Does ontological knowledge of the animate/inanimate distinction help them?] Japanese Journal of Developmental Psychology 17(3). 272–281. Shipley, E. F. and B. Shepperson. 1990. Countable entities: Developmental changes. Cognition 34(2). 109–136. Soja, N. N., S. Carey and E. S. Spelke. 1991. Ontological categories guide young children’s inductions of word meaning: Object terms and substance terms. Cognition 38(2). 179–211. Subrahmanyam, K., B. Landau and R. Gelman. 1999. Shape, material, and syntax: Interacting forces in children’s learning in novel words for objects and substances. Language and Cognitive Processes 14(3). 249–281. Thierry, G., D. Cardebat and J.-F. Demonet. 2003. Electrophysiological comparison of grammatical processing and semantic processing of single spoken nouns. Cognitive Brain Research 17. 535–547. Uchida, Nobuko and Mutsumi Imai. 1999. Heuristics in learning classifiers: The acquisition of the classifier system and its implications for the nature of lexical acquisition. Japanese Psychological Research 41(1). 50–69. van Herten, M., H. H. J. Kolk and D. J. Chwilla. 2005. An ERP study of P600 effects elicited by semantic anomalies. Cognitive Brain Research 22(2). 241–255. Waxman, S. R. and R. Gelman. 1986. Preschoolers’ use of superordinate relations in classification and language. Cognitive Development 1(2). 139–156. Waxman, S. R. and T. D. Kosowski. 1990. Nouns mark category relations: Toddlers’ and preschoolers’ word learning biases. Child Development 61(5). 1461–1473. Whorf, B. L. 1956. Language, Thought and Reality: Selected Writings. J. B. Carroll (ed.). Cambridge, MA: MIT Press. Yasunaga, Daichi and Tsutomu Sakamoto. 2007. On-line processing of floating quantifier constructions in Japanese: Using event-related brain potentials. Journal of Japanese Linguistics 23. 21–34. Yi, Byeong-uk. 2009. Classifiers and count nouns of Korean. Proceedings of the 2009 Summer Conference of the International Society for Chomskyan Studies. 125–140. Yi, Byeong-uk. 2010. Numeral classifiers and the mass/count distinction. Ms. University of Toronto. 6 October. Yoshida, Hanako, and Linda B. Smith. 2003. Known and novel noun extensions: Attention at two levels of abstraction. Child Development 74(2). 564–577. Zhang, Niina Ning. 2012. De and the functional expansion of classifiers. Language and Linguistics 13(3). 569–582. Zhang, Niina Ning. 2013. Classifier structures in Mandarin Chinese. Berlin, Germany: Mouton de Gruyter.

9

Descriptive function of numeral classifiers A corpus-based analysis of numeral classifiers in Korean Young-Wha Kim

1

Introduction

This chapter consists of five topics on numeral classifiers in Korean: (a) three typical structures of counting, (b) numeral classifiers as dependent nominal, (c) the syntax of numeral classifiers when with the plural marker -tul or the distributive marker -ssik, (d) the semantic relations between the numeral classifiers and their head nouns, and (e) the use of the general neutral classifier kay and the decrease in number of the numeral classifiers shown by a corpus analysis. First, in section 2, the characteristic distinction between classifier vs. nonclassifier languages will be explained. In section 3, three conventional structures of counting in Korean will be introduced, and the morphological status of Korean numeral classifiers will be defined. Discussions on the morpho-syntactically dependent status of numeral classifiers will include close comparison between the plural marker -tul and the distributive particle -ssik. In section 4, the descriptive function of numeral classifiers will be verified, and the effect of compound noun formation out of numeral classifiers will also be considered with regard to the loss of descriptive function. The emotional impact of numeral classifiers will be illustrated with examples, and thus the function of the head noun in compound nouns will be described as a subclass of the feature of classifiers. In section 5, head nouns will be described as subclass items representing the features of numeral classifiers. In section 6, semantic taxonomy and feature distinction will be discussed in relation to the subclass features of classifiers. In section 7, frequency statistics of numeral classifiers in current Korean will be introduced based on corpus analysis. The discussion will proceed to suggest the reasons for the decrease in frequency of numeral classifiers in current Korean. Based on the subclass specification of features, current phenomena of simplification or standardization of numeral classifiers will also be interpreted. Section 8 is the conclusion.

2

Classifier vs. non-classifier languages

Classifier languages, like Korean, Chinese, and Japanese, are distinguished from non-classifier languages, like English and Italian, in that they may require a numeral classifier within the noun phrase to count the number or measure the quantity of the head noun, that is, the head of the NP/DP. One linguistic point

232

Young-Wha Kim

generally assumed to be a difference between classifier and non-classifier languages can be explained with the following examples in (1) and (2). (1a,b) are the English examples, and (2a,b) are the Korean ones. (1)

a. b.

[How many apples] did you eat? I ate [three apples].

(2)

a.

ne-nun [sakwa]-lul [myech kay] mek-ess-na? (Korean) you-NOM apple-ACC how many CL eat-PAST-QUES ‘How many apples did you eat?’ na-nun [sakwa]-lul [sey kay] mek-ess-ta I-NOM apple-ACC three CL eat-PAST-DEC ‘I ate three apples.’

b.

(English)

In the English examples (1a,b), the two noun heads ‘apples’ are pluralized; that is, plurality is marked morphologically with the affixation of the plural morpheme ‘-s’ to ‘apple’. On the contrary, in the Korean examples (2a,b), the noun head sakwa ‘apple’ takes a bare form with no plural morpheme marked. Chierchia (1998a,b) suggested that classifier languages are of the type [+arg, -pred] and that bare nouns may convey plurality by means of classifiers. We agree with Chierchia in that bare nouns may convey plurality; however, it should be clarified that plurality is conveyed not by means of classifiers, but by means of numeral classifier phrases (NCLPs). In other words, it is not the classifiers themselves but the numerals accompanying the classifiers that convey plurality. In Korean, classifiers are always accompanied by numerals, from one to more, but they have no singular/plural distinction. Classifiers always come in their bare form, like mass nouns. By the term ‘mass’, we mean a collective group of individuals, and by ‘classifier’, we can also refer to a number of or a collective group of individuals. Mass nouns in the non-classifier languages and classifiers in the classifier languages seem to suggest a possible solution to the inquiry of the syntactic realization of plurality. Classifiers share the same feature with their relevant head nouns, but the classifiers can never be morphologically pluralized.1

3

Numeral classifiers in Korean

3.1  Three    conventional structures of counting in Korean In Korean, we can distinguish three structures of counting or measuring an item: post-nominal classifier phrases (CLPs), pre-nominal CLPs, and cardinal numerations, as can be seen in (3a–c). Some of the nominals are required to accompany a relevant classifier in counting or measuring items, but there are cases where the use of classifiers can be regarded as optional, as can be seen in (3c). (3)

a. (na-nun) chinkwu-lul sey-myeng chodayha-yess-ta (post-nominal CLP) (I-NOM) friend-ACC three-CL invite-PAST-DEC ‘I invited three friends.’

Descriptive function of numeral classifiers 233 b. (na-nun) sey-myeng-uy chinkwu-lul chodayha-yess-ta (pre-nominalCLP) (I-NOM) three-CL-GEN friend-ACC invite-PAST-DEC ‘I invited three friends.’ c. (na-nun) chinkwu seys-ul chodayha-yess-ta (cardinal numeration) (I-NOM) friend three-ACC invite-PAST-DEC ‘I invited three friends.’ (3a) is a post-nominal classifier phrase, and (3b) is a pre-nominal classifier phrase. In (3b), the classifier phrase sey-myeng ‘three-CL’ has undergone leftward movement to the position interior to its head noun chinkwu ‘friend’. The genitive case marker -uy is marked to the classifier phrase as a result of the movement. (3c) shows that a numeral classifier is not obligatory in counting or measuring an item and proves that cardinal numeration by itself can stand for the classifier phrase. In (3c), the cardinal numation seys ‘three’ behaves like a pronominal noun and plays the role of the numeral classifier phrase sey-myeng ‘three-CL’.    bare form 3.2  Morphologically We can use numeral classifiers for the numeral expression of singularity hana/han ‘one’; as in (4a–c): (4)

a. cungke han kaci evidence one CL ‘one piece of evidence’ b. han-kaci-uy cungke one-CL-GEN evidence ‘one evidence’ c. haksayng han myeng student one CL ‘one student’ d. haksayng yel myeng student ten CL ‘ten students’

Plurality information can be determined from the numeral expressions that accompany the classifiers, as in (4d). In (4d), it is the numeral yel ‘ten’ that informs the plurality of the head noun haksayng ‘student’ but not the numeral classifier myeng ‘CL’. A classifier itself cannot be morphologically pluralized. For instance, in the following example (5b), the ungrammaticality is caused by the affixation of the plural morpheme -tul right next to the classifier kay: (5)

a. b.

na-nun [sakwa]-lul I-NOM apple-ACC ‘I ate three apples.’ *na-nun [sakwa]-lul I-NOM apple-ACC ‘I ate three apples.’

[sey kay] three CL

mek-ess-ta eat-PAST-DEC

[sey kay]-tul three CL-PL

mek-ess-ta eat-PAST-DEC

234

Young-Wha Kim

(5a) is a grammatical sentence, but (5b) is not: in (5b), the classifier kay is followed by the plurality marker -tul, and the phrase results in ungrammaticality. The NumP sey kay ‘three CL’ does not allow the affixation of the plural marker -tul next to the classifier kay. Therefore, we can understand that the chief role of numeral classifiers is to help count the number or the quantity of head nouns but not to express plurality.2    classifiers as dependent nominals 3.3  Numeral Numeral classifiers (NCLs) in Korean behave like independent lexical items in keeping word spacing, but their grammatical category is classified into dependent nominals in Korean traditional grammar (e.g., Chae 1990, 1996; Im 1991; Liwu 2016; Oh 1995; Wu 2001, 2006).3 Numeral classifiers are dependent not only because they are required to accompany numeral expressions (Num), either in numbers or in quantifiers to convey full meaning as an independent unit, constructing NumPs, but also because they are not able to constitute a lexically independent word by themselves.4 Numeral classifiers and other dependent nominals, such as ppwun ‘only’, taelo ‘as it is’, mankum ‘as much as’, man ‘just’, pakkye ‘not any more than’, kkaci ‘up to’, and so on, keep word spacing. These dependent nominals should be differentiated from other affixational dependent particles, like the plural marker -tul and the case markers, such as nominative -ka/-i, possessive -uy, and accusative -ul/-lul. Yang (1993) categorized the dependent nominals as delimiters and stated the possibility that they could occur consecutively, forming clusters like the example in (6).5 (6)

(pap-ul) twu kongki mankhum-kkaci-man-(ul) kacyeka-si-o (rice-ACC) two CLbowl as much as-up to-just-(ACC) take-HON-DEC ‘Take (rice) just up to as much as two bowls.’

In (6), the dependent nominals mankhum, kkaci, and man occur consecutively, though they should keep a certain order in their linearity.6 The plural marker -tul is never allowed to attach to a numeral classifier unless the subject noun phrase is marked with plurality. This is because the plural marker -tul is a subject-oriented particle. In case the subject is plural, the plural marker -tul can be spread to any kind of grammatical constituents, including adverbials, like ese-tul ‘quickly-PL’ or cal-tul ‘well-PL’.7 3.4   Plurality and the distributive particle -ssik Plurality can also be expressed by means of the distributive particle -ssik, which can be attached to the classifier. In other words, the particle -ssik retains a distributive meaning, and when the particle -ssik is attached to the classifier, the head noun of the maximal projection NP/DP, which may be either subject or object, is required to be marked with the plural marker -tul.

Descriptive function of numeral classifiers 235 (7)

a. b.

wuli-nun chayk-ul sey kwuen-ssik sa-ss-ta (plural subject NP) we-PL-TOP book-ACC three CL-PRT buy-PAST-DEC ‘Each of us bought three books.’ na-nun kwahak-kwa miswul chayk-ul (kakkak) sey kwuen-ssik sa-ss-ta I-SG-TOP science-and art book-ACC (each) three CL-ssikbuy-PAST-DEC

‘I bought three books of both science and art.’

(plural object NP)

Instead of the plural marker -tul, in (7a), the subject NP takes the plural form of the first person as wuli ‘we’, which is regarded to give influence upon taking the particle -ssik. In (7b), the conjunctive particle -kwa ‘and’ connects the two nouns kwahak ‘science’ and miswul ‘art’, constructing the plural object NP kwahak-kwa miswul chayk-ul ‘science-and-art book-ACC’, and the effect of the particle -ssik attached to the classifier kwuen refers to the plural meaning as ‘three science books and three art books’. Adding the post-nominal adjective kakkak ‘each’ in (7b) helps the distributive meaning of the particle -ssik.8 3.5  Complementary    distribution between -tul and -ssik The plural marker -tul is sometimes discussed for its distributive function, for example, by Park (2006) and Kim Kwang-sup (2021; see Chapter 4 of this volume); however, its primary function is to inform the plurality of the subject noun of the clause. (8)

a. b.

nehui-(tul) motwu-(tul) ese-(tul) woase-(tul) anc-ala you-(PL) all-(PL) quickly-(PL) come-and-(PL) sit-IMP ‘(All of you) come and sit quickly.’ nehui-tul motwu ese woase anc-ala you-PL all quickly come-and sit-IMP ‘(All of you) come and sit quickly.’

In (8a), the plural marker -tul is possibly affixed to the adverbials, motwu ‘all’ and ese ‘quickly’, and the verb woa ‘come’, including the subject noun phrase nehui ‘you-PL’, which is a plural pronominal form of a singular ne ‘you-SG’. By the addition of -tul to the adverbials, as in motwu-tul ‘all-PL’ and ese-tul ‘quickly-PL’, the effect of honorary respect expands toward each of the members of the subject NP, nehui ‘you-PL’, the visitors; the speaker greets and shows his/her hospitality or concerns to each one of the visitors. In other words, the plural marker -tul in (8a) has a distributive function, pragmatically conveying the speaker’s sincere respect to each of the members of the hearers, the visitors in the case of (8a). In (8b), -tul is marked superficially only to the subject pronoun as in nehui-tul ‘you-PL’, and the sentence retains the same meaning as (8a). However, there lies a slight difference between the two examples (8a) and (8b), in that (8a) puts emphasis on each of the members of nehui ‘you-PL’, while (8b) tells all of them as a group. Contrary to the behavior of -tul, -ssik never attaches to subject noun phrases. (9)

a.

nehui-tul chayk-(ul) han kwuen-ssik you-PL book-(ACC) one CL-ssikDIST ‘(Each of you) take one book.’

kacye-la take-IMP

236

Young-Wha Kim b. c.

nehui-tul chayk-(ul) han you-PL book-(ACC) one ‘(Each of you) take one book.’ *nehui-ssik-tul chayk-(ul) you-ssikDIST-PL book-(ACC) ‘(Each of you) take one book.’

kwuen-ssik-tul kacye-la CL-ssikDIST-PL take-IMP han one

kwuen-ssik CL-ssikDIST

kacye-la take-IMP

As in (9a), the distributive marker -ssik is attached right after the numeral classifier and conveys the distributive reading referring to each of the subject noun phrase nehui-tul ‘you-PL’: in other words, if there are five of the students in the group referred to as nehui ‘you-PL’, each and every one of the five students will get one book. (9b) is grammatical, though the distributive marker -ssik and the plural marker -tul are attached consecutively to the numeral classifier phrase han kwuen ‘one CL’ and construct han kwuen-ssik-tul. However, (9c) is ungrammatical, since -ssik comes right next to the plural pronominal nehui ‘you-PL’. nehui-tul ‘you-PL’ is grammatical, but nehui-ssik-tul or nehui-tul-ssik is ungrammatical. In contrast to (9c), as in (9b) the consecutive affixation to the numeral classifier han kwuen-ssik-tul ‘one CL-ssik-tul’ doesn’t cause any problem in its grammaticality but delivers the meaning of plurality of the subject noun with the plural marker -tul. The distributive effect is conveyed by means of the distributive marker -ssik. -ssik conveys distributive reading about the numeral classifier kwuen in (9a,b) and tells about the head noun of the classifier chayk ‘book’ that it should be distributed to each of the subject nehui-tul ‘you-PL’. The distributive characteristics conveyed in (9a,b) are illustrated in (10a,b), correspondingly. (10)

Figure 9.1 Distributive readings of -tul vs. -ssik. a. plural marker -tul, for example, a group of five members, A, B, C, D, E, called by nehui-tul: A B C D E

nehui-tul ese-tul ‘take a seat’

anc-ala

b. distributive marker -ssik, for example, each of the head noun chayk ‘book’, 1, 2, 3, 4, 5, distributed by means of specifying the classifier with the affix -ssik: han kwuen-ssik ‘one book each’

chayk 1 chayk 2 chayk 3 chayk 4 chayk 5

When there is no distributive marker -ssik attached to the classifier kwuen, as in (11) below, the sentence should be interpreted as denoting only one book that should be delivered to the subject noun nehui.

Descriptive function of numeral classifiers 237 (11) nehui-tul chayk-(ul) han you-PL book-(ACC) one ‘(All of you) take one book.’

kwuen kacye-la CL take-IMP

In (11), the speaker may refer to each member of the group nehui-tul, but with no distributive marker -ssik specified to the numeral classifier -kwuen, it should be interpreted that the group as a whole should be given only one book but not one book each.    classifiers obligatory in counting mass nouns 3.6  Numeral For mass nouns, we always need numeral classifiers to measure the amount or quantity of an item.9 For example, as can be seen in the following examples (12a,b), to count the count noun sakwa ‘apple’, we can either use a classifier ‘kay’ or substitute the cardinal pronominal numeration hana ‘one’ for the classifier phrase han kay ‘one CL’. For the same numeration ‘one’, we have two forms, han and hana, and the former han is syntactically incomplete and needs a numeral classifier, like kay (for counting things) or salam (for counting people), and so on, but the latter hana is complete and syntactically independent to count the number of an item. For non-countable mass nouns, for example, ssal ‘rice’ as in (13a,b), we do not allow the substitution of a cardinal pronominal numeration hana ‘one’ for the classifier phrase han doe ‘one CL’. (12) a. sakwa han kay apple one CL ‘one apple’

b. sakwa hana (countable noun) apple one ‘one apple’

(13) a. ssal han doe rice one CL ‘one doe of rice’

b. *ssal rice

hana (non-countable mass noun) one

Numeral classifiers used in counting the amount of mass nouns are in fact the measuring units that can hold the mass materials. So, the counting units mostly in use in current Korean can be the standardized western measuring units, like kilograms and liters, or the kinds of containers, like cups, glasses, bowls, or pints, while we refer to the Korean containers as kongki ‘bowl’, can ‘glass’, or thong ‘bucket’, and so on. In other words, we can easily measure the amount of mass nouns by counting the number of the containers that fit the matter put in them.

4

Descriptive function of numeral classifiers

As was pointed out by Wu (2006, 158), numeral classifiers convey metaphoric function to help understand the meaning of the preceding noun heads. In other words, the semantic significance of numeral classifiers can be regarded as describing the property of their head nouns. A numeral classifier phrase consists of a numeral and a numeral classifier, and the function of the phrase is to count the number of items denoted by the head noun. The main function

238

Young-Wha Kim

of numeral classifiers in numeral classifier phrases is not to count the number or quantity of their head nouns but to satisfy the syntactic structure of the whole noun phrase, as can be seen in the structures of sakwa sey kay (apple three CL) ‘three apples’ vs. *sakwa sey ___ (apple three ____) ‘three apples’. The structure *sakwa sey ___ is incomplete and ungrammatical, since the numeral classifier kay is missing. As has been explained in regard to the example (12a) above, when the numeral classifier phrase sey kay is turned into a pronominal seys, then it works as grammatical, as in sakwa seys ‘apple three-pronominal’.10    numeral classifier kay 4.1  Neutral The numeral classifier kay is regarded as neutral in that it conveys no specific descriptive meaning about the head noun. It is clear that the neutral classifier kay never fails to accompany a numeral and that its function is restricted to help count the number of the head noun, making the structure complete and grammatical. Furthermore, kay is not allowed to construct a compound noun. The neutral classifier kay can be used to count anything but [+animate]. It can be used to count an abstract item, like somang ‘wish’, as in the example of (14): (14) yeki nay cakun somang han kay-lul tama ponay-pni-ta here my small wish one CL-ACC pack send-HON-DEC ‘Here, I am sending my small wish packed in it.’ The neutral classifier kay is necessary only to make the structure grammatically full and complete, but it does not give any information about the head noun except that the head noun to be counted is [-human]. 4.2   Numeral classifiers describing the containers that  hold their head nouns Unlike the case of the neutral numeral classifier kay, which is used for counting [-human] count nouns, numeral classifiers for counting mass nouns may describe the containers that can hold the non-countable mass nouns, like kongki, sapal, congci or ttwukpayki. We could imagine that kancang ‘soybean sauce’ or toencang ‘soybean soup’ can easily be contained in bowls and thus can understand the formation of the compound nouns kancang-congci and toencang-ttukpayki. These numeral classifiers describe the containers that could hold the materials of their head nouns. These four types of bowls, kongki, sapal, congci, or ttwukpayki, differ from each other in size and thus are used to hold the food appropriate to their sizes. Among the four types of bowls, congci is the smallest one and is used to serve sauce, like kancang ‘soybean sauce’; kongki is a medium-sized bowl for containing rice; sapal is a bowl a little bit larger than kongki and is often used to contain soup or rice; and ttwukpayki is the largest of the four bowls and is often used for boiling Korean soybean soup. The bowls in (15a–d) can be distinguished by their sizes as (15a) < (15b) < (15c) ≤ (15d), from the smallest to the largest.

Descriptive function of numeral classifiers 239 (15) a. congci