Count and Mass Across LanguagesOUP UNCORRECTED PROOF Revises,15/6/2012, SPiOXFORDSTUDI ESI NTHEORETI CALLI NGUI STI CSgeneraleditorsDavid Adger and Hagit Borer, Queen Mary, University of LondonadvisoryeditorsStephen Anderson, Yale University; Daniel Bring, University of California, Los Angeles; Nomi Erteschik-Shir, Ben-Gurion University; Donka Farkas, University of California, Santa Cruz; Angelika Kratzer,Universityof Massachusetts, Amherst; AndrewNevins, UniversityCollegeLondon; Christopher Potts,Stanford University, Amherst; Barry Schein, University of Southern California; Peter Svenonius, Universityof Troms; Moira Yip, University College LondonRecent titles23 A Derivational Syntax for Information Structureby Luis Lpez24 Quantication, Deniteness, and Nominalizationedited by Anastasia Giannakidou and Monika Rathert25 The Syntax of Sentential Stressby Arsalan Kahnemuyipour26 Tense, Aspect, and Indexicalityby James Higginbotham27 Lexical Semantics, Syntax, and Event Structureedited by Malka Rappaport Hovav, Edit Doron, and Ivy Sichel28 About the SpeakerTowards a Syntax of Indexicalityby Alessandra Giorgi29 The Sound Patterns of Syntaxedited by Nomi Erteschik-Shir and Lisa Rochman30 The Complementizer Phaseedited by E. Phoevos Panagiotidis31 Interfaces in LinguisticsNew Research Perspectivesedited by Raffaella Folli and Christiane Ulbrich32 Negative Indenitesby Doris Penka33 Events, Phrases, and Questionsby Robert Truswell34 Dissolving Binding Theoryby Johan Rooryck and Guido Vanden Wyngaerd35 The Logic of Pronominal Resumptionby Ash Asudeh36 Modals and Conditionalsby Angelika Kratzer37 The Theta SystemArgument Structure at the Interfaceedited by Martin Everaert, Marijana Marelj, and Tal Siloni38 SluicingCross-Linguistic Perspectivesedited by Jason Merchant and Andrew Simpson39 Telicity, Change, and StateA Cross-Categorial View of Event Structureedited by Violeta Demonte and Louise McNally40 Ways of Structure Buildingedited by Myriam Uribe-Etxebarria and Vidal Valmala41 The Morphology and Phonology of Exponenceedited by Jochen Trommer42 Count and Mass Across Languagesedited by Diane Massam43 Genericityedited by Alda Mari, Claire Beyssade, and Fabio Del PreteFor a complete list of titles published and in preparation for the series, see pp.3056.OUP UNCORRECTED PROOF Revises,15/6/2012, SPiCount and MassAcross LanguagesEdited byDI ANEMASSAM1OUP UNCORRECTED PROOF Revises,15/6/2012, SPi3Great Clarendon Street, Oxford,ox2 6dp,United KingdomOxford University Press is a department of the University of Oxford.It furthers the Universitys objective of excellence in research, scholarship,and education by publishing worldwide. Oxford is a registered trade mark ofOxford University Press in the UK and in certain other countries editorial matter and organization Diane Massam2012 the chapters their several authors2012The moral rights of the authors have been assertedFirst Edition published in2012Impression:1All rights reserved. No part of this publication may be reproduced, stored ina retrieval system, or transmitted, in any form or by any means, without theprior permission in writing of Oxford University Press, or as expressly permittedby law, by licence or under terms agreed with the appropriate reprographicsrights organization. Enquiries concerning reproduction outside the scope of theabove should be sent to the Rights Department, Oxford University Press, at theaddress aboveYou must not circulate this work in any other formand you must impose this same condition on any acquirerBritish Library Cataloguing in Publication DataData availableLibrary of Congress Cataloging in Publication DataData availableISBN9780199654277 (Hbk.)ISBN9780199654284 (Pbk.)Printed in Great Britainon acid-free paper byMPG Books Group, Bodmin and Kings LynnOUP UNCORRECTED PROOF Revises,15/6/2012, SPiContentsGeneral Preface ixThe Contributors xAbbreviations xv1. The count mass distinction: Issues and perspectives 1Jila Ghomeshi and Diane Massam1.1 Setting the stage 11.2 Are count and mass conceptually universal and are they mappedto the real world uniformly? 31.3 Are count and mass universally expressed in language and arethey always expressed in the same way? 41.4 If count and mass are not expressed in the same way universally,how are they expressed? 51.5 Conclusion 72. Lexical nouns are both mass and count, but they are neithermass nor count 9Francis Jeffry Pelletier2.1 Introduction: Informal accounts of mass and count 92.2 mass and count as syntax 112.3 mass and count as semantics 112.4 Problems with the syntactic approach 132.5 Problems with the semantic approach 152.6 Evaluation, and aaw in common 172.7 A different approach 172.8 Related proposals 212.9 Cross-linguistic comments 232.10Further advantages 242.11 Anal philosophical remark 263. Aspects of individuation 27Elizabeth Cowper and Daniel Currie Hall3.1 Introduction 273.2 Taxonomy of English nouns 303.3 Chinese 343.4 Plurality and classiers 36OUP UNCORRECTED PROOF Revises,15/6/2012, SPi3.5 Plural marking in Chinese? 373.6Apparent plural marking in Korean 423.7 Individuation in Persian 453.8 Conclusions 524. Collectives in the intersection of mass and count nouns:A cross-linguistic account 54Heike Wiese4.1 Introduction 544.2 Conceptual and morphosyntactic distinction in the mass/countdomain 574.3 Variation in syntactic-conceptual mass/count correspondences 664.4 Semantics as a mediator of syntactic and conceptual classications 714.5 Conclusions 735. Individuation and inverse number marking in Dagaare 75Scott Grimm5.1 Introduction 755.2 The semantic basis of inverse number marking in Dagaare 775.3 Language internal correlates 875.4 Cross-linguistic correlates 905.5 A formal account of -ri 945.6Conclusion 976. General number and the structure of DP 99Ileana Paul6.1 Introduction 996.2 Malagasy 1006.3 Cross-linguistic considerations 1076.4 The emerging typology 1107. Plural marking beyond count nouns 112Saeed Ghaniabadi7.1 Introduction 1127.2 Background on Persian noun phrases 1137.3 Data 1147.4 Deniteness/Number syncretism 1217.5 Categorial identity of Persian plural marker 1227.6 Analysis 1247.7 Conclusion 128OUP UNCORRECTED PROOF Revises,15/6/2012, SPivi Contents8. Aspectual effects of a pluractional sufx: Evidence from Lithuanian 129Solveiga Armoskaite8.1 The problem:-in- is not a dedicated aspectual morpheme 1298.2 Sufx-in- is Number 1308.3 The aspectual effects of in- are epiphenomenal 1398.4 Conclusions & further questions 1439. Decomposing the mass/count distinction: Evidence from languagesthat lack it 146Martina Wiltschko9.1 Introduction 1469.2 The mass/count distinction is not universally associated withcategorical properties 1479.3 The source of the categorical properties of the mass/countdistinction 1589.4 Variation in the content of the categorical properties:[ +bounded] versus [ +animate] 1629.5 Conclusion 16910. On the mass/count distinction in Ojibwe 172Eric Mathieu10.1 Introduction 17210.2 Number as an inectional category in Ojibwe 17410.3 Ojibwe pluralized mass nouns 18310.4 Basis for a solution 19010.5 Conclusion 19811. Counting and classiers 199Lisa Lai-Shen Cheng11.1 Introduction: Three puzzles 19911.2 Count/mass at the classier level 20211.3 Chinese classiers 20511.4 Do all classiers individuate or divide? 20911.5 Plural classiers 21511.6 Conclusion 21812. Countability and numeral classiers in Mandarin Chinese 220Niina Ning Zhang12.1 Introduction 22012.2 Decomposing countability 221OUP UNCORRECTED PROOF Revises,15/6/2012, SPiContents vii12.3 The two features in nouns 22712.4 The two features in unit words 22912.5 Comparing with the dichotomous-contrast analysis 23112.6Reections on theories of the relation between CLs and countability 23312.7 Summary 23713. Semantic triggers, linguistic variation and the mass-count distinction 238Alan C. Bale and David Barner13.1 Introduction 23813.2 English, Mandarin, and the mass-count distinction 23913.3 English and Mandarin heuristics and the age of distinction 24813.4 Possible triggers for a parametric distinction 25113.5 Conclusion 25914. Classifying and massifying incrementally in Chinese languagecomprehension 261Natalie M. Klein, Greg N. Carlson, Renjie Li, T. Florian Jaeger,and Michael K. Tanenhaus14.1 Background 26114.2 Previous research 26314.3 Experiment One: English measure phrases 26514.4 Experiment Two: Chinese massiers 27114.5 Experiment Three: Chinese count classiers 27514.6Comparing across language and ontology 27814.6Conclusions 282References 283Index 304OUP UNCORRECTED PROOF Revises,15/6/2012, SPiviii ContentsGeneral PrefaceThetheoreticalfocusofthisseriesisontheinterfacesbetweensubcomponentsofthe human grammatical systemand the closely related area of the interfacesbetweenthe different subdisciplines of linguistics. The notionof interface hasbecome central in grammatical theory (for instance, in Chomskys recent MinimalistProgram) andinlinguisticpractice: workontheinterfaces betweensyntaxandsemantics, syntaxandmorphology, phonologyandphonetics, etc. has ledtoadeeperunderstandingof particularlinguisticphenomenaandof thearchitectureof the linguistic component of the mind/brain.Theseries covers interfaces betweencorecomponents of grammar, includingsyntax/morphology, syntax/semantics, syntax/phonology, syntax/pragmatics,morphology/phonology, phonology/phonetics, phonetics/speech processing, seman-tics/pragmatics, and intonation/discourse structure, as well as issues in the way thatthe systems of grammar involving these interface areas are acquired and deployed inuse(includinglanguageacquisition, languagedysfunction, andlanguageprocess-ing). Itdemonstrates, wehope, thatproperunderstandingsofparticularlinguisticphenomena, languages, languagegroups, or inter-languagevariations all requirereference to interfaces.The series is open to work by linguists of all theoretical persuasions and schools ofthought. A main requirement is that authors should write so as to be understood bycolleagues in related subelds of linguistics and by scholars in cognate disciplines.The countmass distinction is, as the introduction to this volume makes clear, aperfect workship for investigating major issues to do with the interaction betweenlanguage, cognitionandthe world, andhas beencentral inbothlinguisticandphilosophical discussions over the decades. The present volume brings togetherscholars andresearchers whouse cross-linguistic andexperimental methods toinvestigate the extent to which the countmass distinction is a fundamental linguis-tic category. The emerging consensus is that this category is indeed fundamental, butthat its expressionin languages of the world is tightly constrained by grammaticalfactors and by factors that lie at the interface of syntax and compositional semantics.David AdgerHagit BorerOUP UNCORRECTED PROOF Revises,15/6/2012, SPi12Countability and numeral classiersin Mandarin ChineseNI I NANI NGZHANG12.1 IntroductionThe word zhi in the Chinese example in (1a) is called a numeral classier (I will call itCL henceforth). A CL occurs with a numeral and an NP, such as, respectively, santhree and bi pen in (1a). In (1b), di is also a CL.(1) a. Yaoyao kanjian-le san zhi bi.Yaoyao see-perf three cl penYaoyao saw three pens.b. Yaoyao kanjian-le san di you.Yaoyao see-perf three cl oilYaoyao saw three drops of oil.Some languages have CLs and some do not. Some languages have the counterpartof theCLin(1b), but not that in(1a). FromtheEnglishtranslationsof thetwoexamples in (1) we can see that English has the word drop to correlate with the CL diin (1b), but does not have a counterpart to the CL ben in (1a). CLs like ben are calledindividual CLs in Chao (1968), and CLs like di are called individuating CLs in thischapter. Languagesthat havebothtypesof CLs, suchasChinese, arecalledCLlanguages.Why do classiers (CLs) exist in CL languages such as Mandarin Chinese? It hasbeen widely assumed that the obligatory occurrence of a CL with a numeral and anoun in CL languages is related to the contrast between count and mass nominals.Thegoal of thischapteristoshowthat thistraditional assumptionisnot ne-grainedenoughtocoverthesystematiccontrastsofnominalsineitherMandarinChineseor other languages. Instead, I arguethat twosyntagmaticproperties ofnominals are syntactically signicant: the ability of a nountocombine withanumeral directly, and the ability of a noun to be modied by a size or shape modier.OUP UNCORRECTED PROOF Revises,15/6/2012, SPiThe two newly recognized properties or features can be attested in the co-occurrencerestrictions of quantiers and CLs. It is the interactions between these two features,rather than the alleged binary contrast between count and mass, that explain varioussyntactic contrasts of countability, cross-linguistically. I argue that althoughthepositivevalueof the rst featurealoneisenoughtodenethecount statusof anominal, it is the combination of the negative values of both features that denes themassstatusof anominal. Thischaptershowsthat thepopularassertionthat allnouns in Chinese are mass nouns is not accurate. Instead, all nouns in Chinese arenon-count nouns, but they are further dividedinto mass and non-mass ones. Thechapter also claries the distinctive functions of the CLs of CL languages.In addition to this introduction and the summary section at the end, this chapteris composed ofve main parts. Section 12.2 proposes my new approach to the count-masscontrast, basedonthetwofeatures. Section12.3studiesthetwofeaturesinChinese nouns and section12.4 investigates the feature numerability of unit words.Section12.5 further argues that the count and non-count contrast is syntactic, andsection12.6 shows the problems of certain current syntactic analyses of CLs.12.2 Decomposing countability12.2.1 Identifying two new features syntagmaticallySince de Saussure (1916), two kinds of relationshipbetween linguistic elements arerecognized: paradigmaticandsyntagmatic. Aparadigmaticrelationshipis estab-lished by a substitution test. For instance, the three wordsof, by, andfor establishaparadigmaticrelationinthestringgovernment{of/by/for}thepeople, sinceeachof themcansubstituteanother. Theymayoccurinthesamesyntacticposition.A syntagmatic relationship is dened by the compatibility of co-occurring elementsin the same construction, e.g. the relationship between the and people in the stringthepeople. Paradigmaticandsyntagmaticrelationshipshavebeenmetaphoricallyviewed as vertical and horizontal ones, respectively.Many formal features such as tense and aspect of verbal expressions, gender andpersonofnominalexpressionsaredenedparadigmatically. Selectionfeaturesaretypicallysyntagmatic features. For instance, thetransitiveverbdrinkc-selects anominal, because it needs to occur with a nominal; and it s-selects a liquid-denotingnominal, because it needs to combine with this type of nominal.Different kinds of syntagmatic relations exhibit different properties. In selection,the occurrence of the selected element is obligatory. But there are other syntagmaticrelations that do not exhibit this kind of obligatoriness. For instance, gradability ofadjectival expressions is dened by the possibility to occur with a degreeword. In(2a), the adjective nice is gradable since it may occur with the degree word quite. (2b)tells us that the adjective next is not gradable, since it may not occur with quite.OUP UNCORRECTED PROOF Revises,15/6/2012, SPiCountability and numeral classiers in Mandarin 221(2) a. the quite nice book b. the (*quite) next bookAnother example of a non-obligatory co-occurrence relation is seen in the featureofagentivity. Agentivityof averbal expressionisdenedbythepossibilitytobemodiedby anagent-orientedadverb. For instance, the VPshoutedin(3a) isagentivesinceit mayoccurwiththeagent-orientedadverbdeliberately, andtheVP arrived in (3b) is not agentive, since it may not occur with deliberately.(3) a. Kim shouted deliberately. b. Kim arrived (*deliberately).In both the gradability and agentivity cases, a feature of an element is dened simplyinthe way that it allows X. Allowingdoes not meanrequiring. Therefore thepresence of X is not obligatory.Withthis backgroundinmind, I nowintroducetwofeatures whicharealsodened syntagmatically, in order to analyze the count-mass contrast of nominals.Somenominalsmaycombinewithacardinal numeral directly, andsomemaynot. In (4a), for instance, the noun apple combines with the numeral one directly. In(5a), however, the noun oil may not combine with the numeral.(4) a. one apple b. ve apples c. zero applesd. 0.5 apples e. 1.0 apples f. ve nouns(5) a. (*one) oil b. (*one) furnitureThe contrast can also be seen in predication (adapted from Chierchia2010:104):(6) a. The boys are at least thirty. b. *The gold is at least thirty.c. The gold is at least thirty pounds.The numeral thirty is the predicate of the nominal the boys in (6a), whereas it maynot be a predicate of the nominal the gold in (6b). Comparing (6b) and (6c), we seethat the numeral needs the support of the measure word pounds to function as thepredicate of thestringthe gold. Followingtheassumptionthat thecopulainanominal predicate constructionis a tense-bearer and the surface order of thesubject-copula string is derived by the raising of the subject from its base-position,Iassumethat in(6a), thesubjectnominal ismergedwiththenumeral predicatedirectly in its base-position, whereas this is impossible in (6b). The contrast is relatedto the nominal type of boy and that of gold.I usethefeaturenumerabilitytorepresentthe contrastbetweennominalsthatmaycombinewithanumeraldirectlyandnominalsthatmaynotdoso. Accord-ingly, [ numerable] means allowinganumeral, and[ numerable] means dis-allowing a numeral. Therefore, the nominals in (4) and (6a) are [ numerable] andthose in (5) and (6b) are [ numerable].The numerals in the nominals in (4), which are all [ numerable], are different.In this analysis, numerability cares about the ability to occur with a numeral only,OUP UNCORRECTED PROOF Revises,15/6/2012, SPi222 Niina Ning Zhangandnospecialstatusisgiventothecontrastbetweensingularityandplurality, oramong integer, zero, and other numerals.The feature of numerability is attested in the fact that certain elements intrinsic-ally bring about a relevant effect. In Dutch, the presence of a collective afx such as-werk makes the noun unable to occur with any numeral (de Belder 2011a: 218) andthustheafxisamarkerof [ numerable]inmyanalysis. In(7a), thenominalsuikersugar has[ numerable], sinceitoccurswiththenumeral driethree. Inboth (7b) and (7c), -werk occurs. In the presence of the numeral drie three, (7b) isnotacceptable. Theacceptabilitycontrastindicatesthatitisthesufxthatbringsabout the feature [ numerable] to the nominal (col collective).(7) a. drie suiker-en b. *drie suiker-werk-enthree sugar-pl three sugar-col-plthree sugarsc. suiker-werksugar-colconfectioneryInadditiontonumerability, wealsoidentifythefeaturedimensionality. Somewordsmaybemodiedbyasize-denotingexpression(e.g. big, small)orshape-denotingexpression(e.g. long, round, square, thin)andsomemaynot. FollowingDixon(1982), Icall thetwotypesof expressionsdimensionexpressions. In(8a),(8b), and(8c), thedimensionadjectivesbig, large, andsquaremodifythenounsapple, furniture, and watermelon, respectively. In (8d), (8e), (8f ), and (8g), however,large, big, and square may not modify oil, music, noun, and wine (see Jespersen 1924:198, Quine1960:104, McCawley1979 [1975]:170, Bunt1985a:199).(8) a. a big apple b. large furniture c. square watermelond. *large oil e. *large music f. *big noung. *square wineThe contrast is also found in predication, as seen in (9) (from Chierchia2010:110):(9) a. The violets are small.b. The furniture is small.c. *The snow is small.In (9a), the dimension adjective small is the predicate of the violets. Similarly, in(9b), theadjectiveisthepredicateofthefurniture. In(9c), however, theadjectivemay not be the predicate of the snow.I use the feature dimensionality to represent the contrast between nominals thatmaybemodiedbyadimensionmodierandnominalsthat maynot be. Thus,[ dimension] meansallowingadimensionmodier, and[ dimension]meansOUP UNCORRECTED PROOF Revises,15/6/2012, SPiCountability and numeral classiers in Mandarin 223disallowing a dimension modier. The nominals in (8a), (8b), (8c), (9a), and (9b) are[ dimension] and other nominals in (8) and (9) are [ dimension].Whenanominalhas[ dimension], weknowthatitsdenotationmusthaveacertain shape or precise limits (Jespersen 1924: 198). The shape or limits are denableor measurable incertaindimensions (e.g. length, size, volume andshape), andtherefore, atomicity is exhibited. In contrast, a nominal with [ dimension] denoteseither material, which in itself is independent of shape or size, such as silver, water,butter, gas, air, orimmaterial notions, suchasleisure, music, trafc, success, tact,common sense (Jespersen 1924: 198). In my understanding, the former group of nounscan occur with a standard or container measure, as seen in (10a) and (11a), whereas thelatter group cannot, as seen in the rest of the examples in (10) and (11).(10) a. a kilo of butter b. *a kilo of leisure c. *a kilo of nouns(11) a. a bowl of butter b. *a bowl of leisure c. *a bowl of nounsSimilartonumerability, dimensionalityisalsoattestedinthefact that certainelements intrinsically bring about a relevant effect. For instance, shuiwater alonemay not be modied by xiao small, as seen in (12a); but if it is followed by a CL suchas di, the whole compoundshui-di canbe modiedbyxiao, as seenin(12a).Similarly, nimud alone may not be modied by xiao, as seen in (12b); but if it isfollowed by a CL such as kuai, the whole compound ni-kuai can be modied by xiao,as seen in (12b) (this issue is further discussed in Zhang2011b).(12) a. *xiao shui a. xiao shui-dismall water small water-clsmall drop(s) of waterb. *xiao ni b. xiao ni-kuaismall mud small mud-clsmall chunk(s) of mudc. *da yun c. da yun-duobig cloud big cloud-clbig piece(s) of cloudIt is important to pointout that words such as big and small and their Chinesecounterparts haveanintensifyingusage. As statedinMorzycki (2009: 176), anadjective that normally expresses size characterizes the degree to which the gradablepredicate holds, as shown in the examples in (13).(13) a. big idiot b. big smokerc. da hao xingshi d. xiao xian shenshoubig good situation small show skillvery good situation show the skill a little bitOUP UNCORRECTED PROOF Revises,15/6/2012, SPi224 Niina Ning ZhangThe intensifying readings are not size readings, and thus the adjectives in such ause are not dimension adjectives.12.2.2 Dening count and mass by the two featuresTraditionally, the notionof count is indirect contrast tothe notionof mass.Different from this binary analysis, I use the two values of the two features, numer-ability and dimensionality, to describe the count-mass contrast.The feature numerability alone may distinguish a count noun from a non-countnoun. If anominal maycombinewithanumeral directlyinthecontext, it has[ numerable] and thus is a count nominal in that context. Otherwise, it is a non-count one. According to Chierchia (1998a: 353, 2010: 104), being able to combine witha numeral is the signature property of a count nominal.But numerability alone is not enough to identify whether a noun is a mass noun.A non-count noun is not necessarily a mass noun. On the one hand, well-recognizedmass nouns, such as the word oil, may be neither combined with a numeral directly(see(5a)), normodiedbyadimensionadjective(see(8d)). Ontheotherhand,words suchas furniture may be modiedby a dimensionmodier (see (8b)),althoughtheycannotcombinewithanumeraldirectly(see(5b)). Suchnounsarenon-countandnon-mass. Iclaimthat althoughthefeature[ numerable]aloneisenoughtodenethecount statusof anominal, it isthecombinationof both[ numerable] and [ dimension] that denes the mass status of a nominal.The four possible combinations of the two values of the features are summarizedin Table12.1:Among the four possibilities in Table 12.1, (a) and (b) are both count, (d) is mass,and(c)isnon-countandnon-mass. Althoughcountisnotmassandmassisnotcount, what is new in this analysis is the independent status of (c). The existence ofthisgroupof nounsindicatesthat non-count nominalsdonot havetobemassnouns. Also, from a different perspective, having the feature [ dimension] meansthatthenounisnotamassnoun, butitdoesnotmeanthatthenounmustbeacount noun(contraWiltschko2005a, amongothers). DucklingandtheGermanwordEichhrnchensquirrel canoccurasnon-countnouns, inadditiontocountTable12.1. Possible combinations of [ +numerable] and [ +dimension][numerable] [dimension] examplea. apple in (4a), (8a) countb. noun in (4f ), (8f ) countc. furniture in (5b), (8b) non-count, non-massd. oil in (5a), (8d) massOUP UNCORRECTED PROOF Revises,15/6/2012, SPiCountability and numeral classiers in Mandarin 225nouns(seedeBelder2011b, fn. 13), althoughtheycanbemodiedbydimensionmodiers(e.g. small duckling). Moreover, theindependent statusof (b)indicatesthat not all count nouns denote entities that have physical dimensions, since not allcount nouns may be modied by a dimension modier.Iclaimthatthetwofeatures, numerabilityanddimensionality, areuniversal indeningcountandmassnouns, assumingnumeralsanddimensionmodiersareavailable inall languages. Also, they are the only criteria tobe consideredinanalyzing the count-mass contrast.12.2.3 Attesting the two features in co-occurrence restrictionsThe linguistic reality of numerability and dimensionality is independently attested inco-occurrence restrictions of articles, quantiers, and classiers (CLs).Itiswell-knownthat indenitearticlesandsomequantiersoccurwithcountnouns in English. For instance, every and many occur with nouns that have[ numerable], and much occurs with nouns that have [ numerable].(14) a. {every} apple b. {many/*much} apples(15) a. *{every} oil b. {*many/much} oilc. {*many/much} furnitureIn Chinese, some CLsare sensitive to the dimension featureof the noun. SomeCLstakenounswith[ dimension]only. Forinstance, noliquid-denotingnounmay be modied by a dimension adjective, as seenin(16a). Such a nounis[ dimension]. It can occur with the CL di, as seen in (16b).(16) a. *chang {you/shui/xue/niao/yanlei}long oil/water/blood/urine/tearb. san di {you/shui/xue/niao/yanlei/*putao}three cl oil/water/blood/urine/tear/grapethree drops of {oil/water/blood/urine/tear/*grape}In contrast, putao grape can be modied by a dimension adjective, as seen in (17a)below. Suchanounis[ dimension]. Itmaynotoccurwithdi, asseenin(16b)above. OtherCLsthat reject nominalswith[ dimension] includeji (forliquidmedicine), pao (for urine), and tan (for any liquid). I call such CLs (part of Chaos1968 partitive CLs) individuating CLs, which select [ dimension].(17) a. da putaobig grapebig grapeb. san ke {putao/*you/*xue/*yanlei}three cl grape/oil//blood/tearOUP UNCORRECTED PROOF Revises,15/6/2012, SPi226 Niina Ning ZhangWords like putao can be selected by another kind of CL, calledindividual CLs inChao (1968). The CL ke in (17b) is such a CL. It selects nouns with [ dimension].OtherCLssuchasben, tou, andzhiarealsoindividualCLs. Moreover, collectiveCLs, suchaszugroup,quncrowd, dadozen,shuangpair, andduipair,andpartitive CLs, such as ye page, duan paragraph, and zhang chapter (they are alsoChaos1968 partitive CLs), also occur with nouns with [ dimension].12.3 The two features in nouns12.3.1 Numerability of nounsOccasionally, we see people claim that numeral CL languages do have count nouns,orpeoplefeel reluctant toadmit that thereisnocount nouninsuchlanguages.However, westill needtoconsider[i]f weassumethat classierlanguageshavecount nouns (similar toEnglishsilverware, cf. constructions likethree pieces ofsilverware), then it is unclear what necessitates the use of classiers (Krifka 2008: 5).If weput unit wordssuchasdui pile aside, itisundeniablethatnonouninChinese is able to combine with a numeral directly, as shown in (18). Therefore, allnouns in the language have the feature [ numerable]. This means that no noun inthe language is a count noun.(18) a. *san xianglian b. *san youthree necklace three oilThe occurrence of a unit word such as a CL is obligatory between a numeral and anounin Chinese. IncontrasttoChinese, insome languagessuchas Hopi(Whorf1956 [1941]: 141, Greenberg 1990 [1972]: 176), Halkomelem Salish (Wilhelm 2008: 64),and Yudja (an indigenous language spoken in Brazil; see Lima 2010b), all nouns cancombine with a numeral unconditionally. (19) is an example (Lima2010b:7).(19) txaba apeta [Yudja]three bloodthree units of blood(the unit is identied in the context: drops, puddles, or containers)In Yudja, there are neither CLs nor plural markers. When a numeral and a nounarecombined, theexactunitofcountingdependsonthediscoursecontext(Lima2010b: 13). Limareportsthenaturalnessof datalike(19)intheabsenceof eitherUniversal SorterorUniversalPackagereffects. Wecanseethatall nounsinsuchlanguages may have the feature [ numerable].Between the above two patterns, in languages such as English, some nouns maycombinewithnumeralsdirectly(e.g. apple)andsomemaynot (e.g. oil), asseenbefore.OUP UNCORRECTED PROOF Revises,15/6/2012, SPiCountability and numeral classiers in Mandarin 22712.3.2 Dimensionality of nounsAlthough all nouns in Chinese are non-count nouns, they are not all the same withrespect todimensionality. In12.2.3, we see that nouns with[ dimension] areselected by individual CLs, and nouns with [ dimension] are selected by individu-ating CLs. As shown in (20), nouns such as he river can be modied by a dimensionadjective such as changchang long. In contrast, material nouns such as you oil in(21a) and immaterial nouns suchas minzhu democracy in(22a), reject suchadjectives (putting aside the intensifying reading of such adjectives; see the discus-sion of (13)).(20) a. changchang de he b. da qi-qiu c. fang xigualong de river big air-ball square watermelonlong river big balloon square watermelon(21) a. *changchang (de) you b. *da (de) zheng-qilong de oil big de steam-aic. *fang de mianfensquare de our(22) a. *changchang (de) minzhu b. *bo (de) zibenzhuyilong de democracy thinde capitalismThe constraint is shownnot only inmodication, but also inpredication.Thestringhenchangverylong maynotbethepredicateofthemassnounyouoil in (23a), but it can be the predicateof the non-massnoun heriverin (24a).Thestringhendaverybig maynotbethepredicateofthemassnounzheng-qisteam-air in(23b), but it canbe the predicate of the non-mass nounqi-qiuballoon in (24b).(23) a. *You hen chang. b. *Zheng-qi hen da.oil very long steam-air very big(24) a. He hen chang. b. Qi-qiu hen da.river very long air-ball very bigThe river is very long. The balloon is very big.This contrast shows that the feature dimensionality can divide Chinese non-countnouns into the mass-type, which has [ dimension], and the non-mass-type, whichhas [ dimension].Greenberg (1972: 26) claims that nouns in CL languages have the characteristics ofa mass noun. The idea is also seen in Hansen (1972), Krifka (1995), Doetjes (1996),Chierchia (1998a), among many others. According to our new analysis of the count-mass contrast, however, not all nouns in Chinese are mass nouns.OUP UNCORRECTED PROOF Revises,15/6/2012, SPi228 Niina Ning Zhang12.4The two features in unit words12.4.1 Classication of unit wordsAll unit words tell us what counts as one in counting. Unit words include CLs andmeasure words. The latter group is composed of standard measures such as kilo andcontainer measures such as cup in three cups of tea.Amongvarioustypesof CLs, kindCLshavenooccurrencerestrictions. Theyoccurwithalltypesofnouns. Standardandcontainermeasuresoccurwitheither[ dimension]nounsormaterial typeof [ dimension]nouns, butreject nounsdenoting immaterial notions (see (10) and (11)). So these three types of unit wordsare not sensitive to the contrast between [ dimension] and [ dimension]. In thefollowingdata, thenounsinthea-examplesare[ dimension]andthoseintheb-examples are [ dimension].(25) a. shi zhong luobo b. shi zhong mianfen [kindcl]ten cl carrot ten cl ourten types of carrot ten types ofour(26) a. shi gongjin luobo b. shi gongjin mianfen [standard measure]ten kilo carrot ten kilo ourten kilos of carrots ten kilos ofour(27) a. shi xiang luobo b. shi xiang mianfen [container measure]ten box carrot ten box ourten boxes of carrots ten boxes ofourWhen these three types of unit words occur with nouns of [ dimension], they donot represent the natural units of the elements encoded by the nouns.Unit words that select [ dimension] nouns are individuating CLs (section 12.2.3),as shown in (28). Such CLs occur with mass nouns (e.g. Croft 1994: 162). Semantic-ally, individuating CLs are associated with the idea that the noun refers to some kindof mass and the CL gives a unit of this mass (Aikhenvald2003:318).(28) a. shi dui tu b. wu gu zheng-qitencl earth ve cl steam-airten piles of earth ve puffs of steamc. wu zhang zhi d. wu di youve cl paper ve cl oilve pieces of paper ve drops of oile. wu tan you f. wu pao niaove cl oil ve cl urineve puddles of oil ve units of urineOUP UNCORRECTED PROOF Revises,15/6/2012, SPiCountability and numeral classiers in Mandarin 229g. liu ji yao-shuisix cl medicine-liquidsix units of medicine liquidUnit words that occur with [ dimension] nouns are divided into three types: (1)What counts asoneis bigger thanthe natural unitof theelementdenotedbythenon-massnoun. Inthiscase, acollectiveCLisused, asin(29a). CollectiveCLs(called groupmeasuresinChao1968: 595)includetheso-calledarrangementCLs,such as pai row and luo stack, and number set CLs, such as shuang pair, dui pair,and da dozen. (2) What counts as one is smaller than the natural unit. In this case,a partitive CL is used, as in (29b). (3) What counts as one matches the natural unit.In this case, an individual CL is used, as in (29c).(29) a. shi dui luobo [collectivecl]ten cl carrotten piles of carrotsb. shi pian luobo [partitivecl]ten cl carrotten slices of carrotc. shi gen luobo [individualcl]ten cl carrotten carrotsGenerally speaking, the same form of a unit word can belong to different types,depending on the type of the associated noun, and the semantic function of the unit.In(28a), the CLdui occurs withthe mass nountuearth, andit is thus anindividuatingCL. However, in(29a), dui occurs withthenon-mass nounluobocarrot, and it is thus a collective CL.12.4.2 Unit words as the unique numerability bearers in ChineseAll unit words may combine witha numeral directly, althoughunder certainconditions, the numeral yi one can be implicit.TheconditionsofsilentyiarestudiedbyYang(1996). Silentyicanfollowmeieach and the demonstrative zhe this or na that. When yi is covert, its occurrencein syntax can be attested by the singular reading of the whole nominal.(30) Shufen xiang mai zhe (yi) ben shu.Shufen want buy this one cl bookShufen wants to buy this book.Therefore all unit words have the feature [ numerable].OUP UNCORRECTED PROOF Revises,15/6/2012, SPi230 Niina Ning ZhangSince no noun in Chinese has the feature [ numerable] and all unit words havethefeature, thelatteraretheuniquenumerabilitybearerinthelanguage. Wecansee that numerability does not have to be anchored to lexical or root elements (foranon-lexical approachtothe count-mass contrast, see Allan1980, Borer 2005,Pelletier2009; for a lexical approach to the contrast, see Doetjes, to appear).Theanalytical realizationof numerabilityisparallel tothesituationthat tenseinformation can be realized by either verbs or auxiliaries in English. Formal featuresin general can be distributed in various types of elements.12.5 Comparing with the dichotomous-contrast analysisThestudyofthecontrastbetweencountandmassnounsdatesbacktoAristotle.Developingtheinsightsofmanypreviousstudies, Ihavemadethefollowingtwomain claims with respect to the contrast.First, a count noun is dened exclusively by [ numerable], i.e. the possibility tocombine witha numeral directly. It has beengenerallyrecognizedthat suchacombinationpossibility is the signature grammatical property of count nouns(e.g. Chierchia2010: 104). I havenowfurther arguedthatthisis theonlydeninggrammatical property of a count noun, cross-linguistically. This syntagmatic den-itionmeansthatthecount/non-countdistinctionisclearlylinguistic, ratherthanextra-linguistic. It isthusnot surprisingthat countabilityisexpressedinvariousways, cross-linguistically and within the same language. In Chinese, generallyspeaking, nonounmaycombinewithanumeraldirectly, andtherefore, nonounis acount noun. Numerabilityis insteadrepresentedexclusivelybyunit words,includingCLsandmeasurewords. InlanguagessuchasYudja(Lima2010b)andHalkomelem Salish (Wilhelm 2008: 64), no CL exists, and every noun can combinewith a numeral directly. Thus every noun can be a count noun. Between these twopatterns, in languages such as English and Dne (Wilhelm2008), in an unmarkedsituation(i.e. without a shift), some words are [ numerable], andothers are[ numerable].Second, the notion of mass is not the direct negation of count. Instead, it is thecombination of the two syntagmatic properties: [ numerable] and [ dimension].Words such as oil in English and their counterparts in Chinese are mass nouns. Thisrened analysis makes it possible to precisely identify elements that may notcombine with a numeral directly but may allow a dimension adjective, e.g. furnitureinEnglishandpingguo apple inChinese. Suchwords donot denote massiveobjects. AsChierchia(2010: 144)putit, weknowrightoffthebatthatfurniturecannotbe treated on a par withwater. Such words have beenidentied ascountmassnouns (Doetjes1996: 44, 2010: 44), objectmassnouns (Barner&Snedeker2005), andfakemass nouns (Chierchia2010: 110). ThesimilaritybetweensuchOUP UNCORRECTED PROOF Revises,15/6/2012, SPiCountability and numeral classiers in Mandarin 231words and the Chinese counterparts of English count nouns has been mentioned inDoetjes (1996: 35), Krifka (2008: section6.5), Cowper andHall (2009a: 1), andChierchia (2010: 111, fn. 12), among others. InDoetjes (1996: 34), count massnouns arethewordswhicharesemanticallycountbutbehavelikeamassnounsyntactically. If one assumes that there is a binary contrast and then gives a name tothe elements that do nott in the contrast, I do not think the analysis is complete.Although it has been widely believed that all nouns are mass nouns in Chinese,thedifferencebetweenEnglishtypical massnounsandChinesenon-massnouns,with respect to dimensionality, has been noted in Gil (2008: 8). Hends that unlikethe former, the latter can be modied by size and shape adjectives. In my analysis,bothfurnitureandpingguoare[ numerable]and[ dimension]. Therefore, dapinguo big apple is as natural as big furniture.Thecloseinteractionbetweenthenotionsnumerabilityanddimensionalityhaslong been realized in the literature, but the nature of the relation between them hasnot been claried (see Jespersen1924: 198). Quine (1960: 104) notes that unaccept-ability of *spherical water and *spherical wine. On the other hand, it is obvious thatcount nouns such as suggestion also reject spherical. Bunt (1985a: 199) also points outthat massnounssuchaswatermaynot bemodiedbyadjectivessuchaslarge.However, McCawley (1975: 170)nds that furnitureand footwear, which have alsobeen treated as mass nouns, admit size modication much more readily than hard-core mass nouns such as rice.The most recent and thorough discussion of the relation between the count-masscontrast andsizeadjectivesisdeBelder(2011a,b). Herdiscussiondoesnot coverother dimension modiers such as thick, thin, round, though. Crucially, she claimsthatifsomethingacquiresthe[size]feature, itautomaticallybecomescountable.(2011b: 183) So for her analysis, size features entail the count status. This is differentfrom my analysis, which gives an equal status to numerability and dimensionality:neither entails the other, and thus there are four possibilities. One empirical conse-quence of her analysis is that she fails to capture the fact that non-count nouns suchas furniture may have the size feature. This kind of noun is predicted to be illicit inher theory (de Belder2011a:83 (34),2011b:180), contrary to the fact.Theideathatcountandmassisnotadichotomouscontrastandthusweneedmorefeaturestorepresent themis also seenin Muromatsu(2003)andAcquaviva(2010). However, in the absence of syntactic criteria to analyze the empirical issues,the idea is somewhat undeveloped. But it does pave the way for the research in thischapter.The proposed two features, numerability and dimensionality, are different, but arebothrelatedtothecountabilityofnominals. Thisisparalleltoourunderstandingthat tenseandaspect featuresaredifferent, but arebothrelatedtothetemporalproperties of linguistic elements.OUP UNCORRECTED PROOF Revises,15/6/2012, SPi232 Niina Ning Zhang12.6Reections on theories of the relation between CLs and countability12.6.1 The syntactic foundations of the presence of CLsThe novel analysis of the count-mass contrast proposed in this chapter opens a newwindow to see the syntactic foundations of individual CLs in CL languages.Counting is possible in the presence of a unit. The unit tells us what counts as onein the context. The general function of a unit word is to specify the unit for counting.Such a word is [ numerable]. In Chinese, when a noun occurs with a CL, it is theCL rather than the noun that is the bearer of numerability.Individual CLs are syntactically different from nouns. Thus the fact that numer-ability is realized on CLs rather than nouns is a syntactic issue. The syntactic natureof the existence of individual CLs can be seen in another fact: the occurrence of suchCLsissensitivetosyntacticcategoriesinEnglish. Countinginverbal phrasesinEnglish requires the occurrence of CLs (Krifka 2007: 39), but not in nominals, as seenin (31). There is no CL in the nominal counting construction three trips to Paris in(31a), but the CL times is obligatory in the verbal counting construction in (31b). LikenominalsinChinese, verbal phrasesinEnglisharenotnumerabilitybearers, andthusneedCLsincounting. If weconsidertherepresentationof numerabilityinverbal phrases, English should be treated as a CL language.(31) a. Bill made three trips to Paris.b. Bill traveled to Paris three *(times).This numerability-bearer analysis of CLs calls for a review of our current under-standing of CLs in CL languages.12.6.2 How special are the CLs of CL languages?All seventypesof unitwordslistedin12.4.1arecloselyrelatedingrammarandfunction (Croft 1994: 152). Like measure words, CLs are also counting units or unitcounters (Allan1977:293).It has been widely believed that all nouns in CL languages are mass nouns, andtherefore, thebasicfunctionofCLsistodividemassintounits(e.g. Quine1969,Greenberg 1972: 26, Link 1991, Borer 2005: 101, Krifka 2008: section 6.3). Accordingly,the syntactic projection headed by a CL has been called DivP (Borer2005). In theprevious discussion, I have shown that CL languages can distinguish mass from non-mass nouns. As aconsequence, the general functionof CLs is not dividingorindividuating.Let us examine how the dividing assumption misrepresents the basic function ofCLs. We have introducedve types of CLs in12.4.1:OUP UNCORRECTED PROOF Revises,15/6/2012, SPiCountability and numeral classiers in Mandarin 233(32) a. san zhong yang [kindcl]three cl sheepthree kinds of sheepb. san di shui [individuatingcl]three cl waterthree drops of waterc. san qun yang [collectivecl]three cl sheepthree groups of sheepd. san pian xigua [partitivecl]three cl watermelonthree slices of watermelone. san ben shu [individualcl]three cl bookthree booksFrom the translations of (32a), (32b), (32c) and (32d), we can see that English alsohas kind CLs such as kind, individuating CLs such as di drop, collective CLs such asgroup, and partitive CLs such as slice.Amongthe vetypes of CLs, theindividuatingCLin(32b) indeeddivides amassiveobjectintounits(alsoseetheexamplesin(28) ). CLslikethisarecalledpartitive measures in Chao (1968), and classiers for massive objects in Gerner andBisang (2010: 606). Such CLs are also found in non-CL languages such as English, asseenintheworddropinthetranslationof (32b). Obviously, individuatingCLscannot distinguish CL languages from other languages.What English does not have is individual CLs. There is no English counterpart forbenin(32e). It isthistypeof CLthat distinguishesCLlanguagesfromnon-CLlanguages suchas English. Innon-CLlanguages, individual CLs arenot overtlyrepresented by linguistic expressions. In such languages, it has been assumed (Quine1969:36) that the semantics of an individual CL is integrated either in the numeral(see Wilhelm2008:55) or the noun (see Chierchia1998a).Crucially, individual CLs donot divideor individuateanything. Theydonotoccur withmass nouns. As pointedout byBaleandBarner (2009b: 7), defaultclassiers [such as the individual CL ge in Mandarin Chinese] often combine withnouns that alreadyare interpretedas containingindividuals. SuchCLs neitherindividuate anythingnor createnewunits for the individuals anymore, unlikecollective or partitive CLs. Therefore, the popular belief that it is the individuating(discrete set-creating) function of CLs that is special in CL languages needs recon-sideration. A more precise generalization is that in addition to the various ways ofspecifying a unit for counting, CLs in CL languages may also represent the naturalunit of entities that show atomicity, whereas the CLs of other languages do not haveOUP UNCORRECTED PROOF Revises,15/6/2012, SPi234 Niina Ning Zhangthis semantic function. In other words, the CLs in CL languages are special in theirability to represent the natural units of the entities denoted by non-mass nouns.If the general function of CLs is not dividing, we need to reconsider Borers (2005)syntactic analysis of the count-mass contrast. In her analysis, the absence of dividingstructure (DivP) derives mass readings, andthepresenceof dividingstructurederives count readings. The two features proposed in this chapter call for a richerstructure to represent the count-mass contrast.12.6.3 The unreliability of the de and pre-CL adjective argumentsInthissection, I argueagainst theassumedcorrelationbetweenthecount-masscontrast and two phenomena in Mandarin Chinese: the occurrence of an adjective tothe left of a unit word and the occurrence of the functional word de to the right of aunit word (Cheng and Sybesma1998,1999).I have argued that in Chinese, neither nouns nor CLs make a distinction betweencount and non-count themselves, since all nouns are non-count elements (12.3.1) andallCLsarecountelements(12.4.2). ButtheselectionofCLsmaydistinguishmassnouns from non-mass nouns. Individual, collective, and partitive CLs occur with non-mass nouns and individuating CLs occur with mass nouns (other unit words, i.e. kindCLs, standard and container measures, occur with both mass and non-mass nouns).Cheng and Sybesma (1998, 1999) try to make a distinction between count CLs andmass CLs (calledmassiers). Thenames areusedtoshowthat inChinese, thecontrast between count and mass nouns can be distinguished at the level of CLs, ifnot at the level of nouns.In Cheng (2009b: 3), it seems that count CLs are equivalent to individual CLs andall other kinds of unit words are mass CLs. Cheng andSybesma (1988, 1999)formalize the following two criteria:Criterion A: A pre-CL adjective may occur with a mass noun, as seen in (33a), butnotwithacount noun, asseenin(33b)(Cheng&Sybesma1998: 390, 1999: 516).The term count noun in their analysis is called non-mass and non-count noun inthis chapter.(33) a. yi da zhang zhi b. *yi da wei laoshione big cl paper one big cl teacherone big piece of paperc. yi da tiao hao-han d. san da zhi laohuone big cl good-guy three big cl tigerone big good guy three big tigerse. san chang tiao xianglianthree long cl necklacethree long necklacesOUP UNCORRECTED PROOF Revises,15/6/2012, SPiCountability and numeral classiers in Mandarin 235It istruethat (33b)isnot acceptable. But isolatedcaseslikethisdonot affecttheobservationthatotherexamplesofthesametypeareacceptable, asshownin(33c,d,e) (seeChengandSybesma 1998: 390fn. 4for their acknowledgement ofcounter-examples). Tang(2005), Hsieh(2008), andLi (2011: 34), amongothers,all present manycounter-examples tothis claimabout pre-CLadjectives. Moreexamples can be found in Zhu (1982: 52), Lu (1987), and Luo (1988). Therefore, theadjective criterion is empirically problematic.CriterionB: Demayoccur betweenameasurewordandamass noun, but notbetweenaCLandacount noun(Chao1968: 555, 588; Zhu1982: 51; ChengandSybesma1998:388,1999:515). A typical pair of examples is (34):(34) a. san wan de tang b. *san ge de laoshithree bowl de soup three cl de teacherthree bowls of soup three teachersAgain, the unacceptability of (34b) is one of a few isolated cases. In fact, all typesof CLs can be followed by de in an appropriate context. The choice of the context hasnothing to do with the count-mass contrast. Instead, it has to do with the syntacticpositionof de. InZhang(2011a), I showthat there are twosources of de: oneintroducesaconstituent directlyandtheothersurfacesinacomparativeellipsisconstruction. Constructions of individual, individuating, and kind CL host the latterde only, whereas those of the other types of unit words (partitive and collective CLs,standardandcontainermeasures)hostdeofeithersource. Notethatthedivisionhere does not match with Cheng and Sybesmas distinction between count and massCLs. If one just considers the phonological formof de without consideringitsstructural position, then, de mayoccur withall types of CLs or unit words, asshown in (35).(35) Shufenchi-le yi-bai {ge/gongjin/bao/pian/dui/zhong} de pingguo.Shufeneat-perf one-hundred cl/kilo/bag/slice/pile/kind deappleShufen ate100 apples or100 {kilos/bags/slices/piles/kinds} of apples.Therefore, ChengandSybesmas (1998, 1999) claimthat onetypeof CL(thecount type) may not be modied by an adjective, and may not be followed by de,whereas the other type (the mass type) can, is descriptively inadequate.As mentioned above, several works, including Tang (2005: 432, 440446), Hsieh(2008: 34), Li (2011), etc. have already presented many counter-examples to falsify thealleged distinction. Wu and Bodomo (2009: 489) point out that the two alleged typesof CL can occur with the same NP (See also Borer 2005: 98), as shown in (36). Ben in(36a) and li in (36b) are count CLs, and xiang in (36a) and wan in (36b) are massCLs, in Cheng and Sybesmas system.OUP UNCORRECTED PROOF Revises,15/6/2012, SPi236 Niina Ning Zhang(36) a. san ben shu a. san xiang shuthree cl book three box bookthree books three boxes of booksb. san li mi b. san wan mithree cl rice three bowl ricethree grains of rice three bowls of riceHer and Hsiehs (2010: 541) following examples show that the two constraints onthe so-called count CLs (i.e. individual CLs) can even be violated at the same time.The CLs ke in (37a) and tiao in (37b) are typical individual CLs, but they are bothpreceded by a modier and followed by de.(37) a. yi da ke de gaolicai b. yi da tiao de yuone big clde cabbage one big cl deshone big cabbage one bigshInconclusion, theallegedtwocriteriacannot makeanydistinctioninCLsinChinese, regardlessofwhethertheassumeddistinctioncorrelateswiththecount-mass contrast.12.7 SummaryBasedonMandarinChinese, inthis chapter I have arguedfor amore renedsyntactic analysis of the count-mass contrast. I list my main conclusions as follows.The count-mass contrast of linguistic elements is decomposed into two features:[numerable]and[dimension]. [ numerable]meansanouncancombinewithanumeral directly, and thus it is a count noun. [ numerable] nouns are non-countnouns. Nouns in Chinese are non-count nouns in general. [ dimension] means anoun can combine with a shape or size modier, and thus it is a non-mass noun. Amass noun is dened by both [ numerable] and [ dimension]. Not all nouns inChinese are mass nouns.Some CLs select mass nouns and some select non-mass nouns. The latter type ofCL, i.e. individual CLs, distinguishes CL languages from other languages. Like othertypes of unit words, all CLs specify units and tell us what counts as one in counting.Thesemanticfunctionof individual CLsistorepresent thenatural unitsof theelements denoted by non-mass nouns. Such a CL has no dividing function at all.The feature [ numerable] maybe distributedindifferent types of elementscross-linguistically and within the same language. In Mandarin Chinese, nouns donot havethisfeature, whereas unit words do. Inlanguages suchasEnglish, thefeature is found in both count nouns and unit words such as measure words.OUP UNCORRECTED PROOF Revises,15/6/2012, SPiCountability and numeral classiers in Mandarin 237