You are on page 1of 256


Primes and Universals



New York


Oxford University Press, WaI~on Street, Oxford 0x2 1 6 ~ ~ OxJTord Mew York Athens Aucklmd Bangkok & d a y Caimlta Cape Town Dar es Sahorn DeWi iESorence Hong Kong Iseanbul Karachi K~ata Lumpr Madria Madrid Me&ourne Mexico City Nairnrbi Forb Sagepore Wipei Tokyo Toramto and associated companies k B e r h lbadan Oxford is a erode mark of Oxford University Press
O Anna Wierzbicka 1996


AN rigfigs reservtrd No past 0 1 rhb pwbticntiom may be reproduced, stored {n a retrieval system, or transmitted, b any form or by any means, without the prbr permisston h wrilfng of Oxford University Press. WizFtin I s UK, exceptions are aljowed k respect of any fair deating for the plrspose of research or private study, or criticism or review. as permared wder !he iL$pyrighf,Designs m d Paeenrs Act# 1198. or in t f ~ case e of reprographic ~epr0d~clid)FI in accurdance with the t e r m ofthe Ifcences issued by tk Cupyrigh~ Licensing Agency. Eaquiries camcernirsg reproduction outsfde these t e r m m d in other countrie~ should be Sent to the Rights Department. Oxford Umiversity Press, at !he address abave

Thk book b soM mbject to the condition l e t it shdl not, by WQJJ of8~11pde or olherwke, be !en#, re-soh$, faired on8 or ofjiwwfse circu!ale$ vvilhorrs r k e p u 5 l i s k r ~ ~ u conseat r in any form of binding or cover other than rhes in which it is pubEished and without a simifar condition inchtdtng lhEr condirfon being imposed on the subseqwent purchaser British Library Catab,,uing in Pub!ica?toti Data Dada awaiIabi'e Library . y a Congress C~taiogrhg h Publication Dara Senrmsltcs :primes and universab / Aana Mfierzbicka. JncIudes previow$ pubhhed materiaf rev. m d exprmded ior I h b pubficatbn. Jnizchdes bi5Siographkal references and index. 1. $emnnrfc.~. 2. Universa!~'s(Lingtais#ics]3. Grammar, campararive and genera!. 4 Language and d t u r e . 1. Titte. ~ 3 1 2 5 . 5 . ~ 5 4 ~ 5 419% ~ o I ~ . ~ ~ A c95-2fprm ~o

Prixrted in Great 8rirah on acid$ree paper by Bookcrojt (80th) Ltd, Mi&omer Norma

This book owes a great deal to my friend and colleague Cliff Goddard of the Universiky of New England, who read and ma& very detailed comments on the first draft of it. I; have revised all the chapters, same of them quite extensively, in response to Cliff's criticisms and suggestions. Over a number af years, CliU ;has been my principal partner in the search for semantic primes and semantic universals, and interminable tdephone Biscussions with him have k n an unfailing souce of insight and inte1lectud pleasure. 1arnn also very grateful to my old friend Andrzej Bogustawski of Warsaw TJmiversity, who three decades ago initiated the search for semantic primes, who has mntinued this search thoughout this period, and who, despite distance, has remained an invaluable interlocutor and colleague. I would like to thank the colleagues who read and mmemted on an earlier draft of the Introduction to this book and thus enabled me to improve it, in particular Sasha AikhenvaPd, Avery h h e w s , Jentcrnne Bmner, Bob Dixont, Mark h r i e , Ian Green, Jean Harkins, Randy Allen Harris, Helen WLoghlin, Andy Pawlley, and Jane Simpson. I am particularly grateful to my extremely able Research Assistant, Helen O'Loghlin, who went far beyond the call of duty in assisting me to prepare this book for publication, chasing references, tracking down iineansistcneies and errors, dis~ussing ideas, and suggesting possible ways of improvement. Her help was indispensable. I would like, too, to thank the Australian Research Council for a grant for research assistanw, which made this possible. I would also like to thank Tim Curnow, who worked as my Research Assistant at an earlier stage of the preparation of the book {also under an ARC grant), and whose help was also invaluable. It is also a pleasure to express my heartfelt gratitude and appreciation to Ellalene Seymour, for her expert, patient, and good-humowlred typing and editing of the sruwessive drafts of this book. Finally, I would like to thank my students at the Australian Nationaie University, and in particular, the participants of my Seminar on Semantics, who have contributed both valuable data and ideas to the project. Some portions of this book first appeared, in different f o m , as articles in journals or as chapters in colllective volumes. I thank the publishers far permission to include revised and expanded versions of the following publications or parts thereoE

'Prototypes Save? On the Uses and Abuses of the Notion of 'Prototype' in Linguistics and Relaked Fields. In Savas L. Tsohactzidis (ed.). Meanings md P~o~obypes: Studies in Li~gwisfk Cafegorizatiem.London: Routfdge & Kegan Pau1. 19910. 347-3167. Semantic Primitives and Semantic Fidds. In Adrienne Lekres and Eva F d e r Kittag (eds,). Frames" Fie[&, a d Contrwfs: New Essays EjJ Semantic and LexScaE Organization. Hillsdale, NJ: Lawrence Esfbalum. 1992. 209-27. Semantic Complexity: Conwptual Primitives and the Principle of Substitutability. Theoretical ICimg~lis~ies, E 7 . 1991. 75-97, Semantic Uniwzrsals and Trlmitlve Thought? The Question ofthe Psychk Unity of Humankind. .?~u~aal ~JLi~~guEsfic A~nthropafogy. U1. 1994. 1-27, l667. Ostensive Definitions and Verbal Definitions: Innate Conceptual Primitives tand the Acquisition of Concepts, In Maciej Grochowski and Daniel Weiss (eds.]. Worth are Physieiamjor an AiEing Mhd Sagners Slavistische Sammlung, xvii. Munich: Otto Sagner. 1991.46740. Back tra Definitions: tagnition, Semantics, and Lexicography. Lexfcograph[ca.8. 1992. 14W4. (Published in 1994.1 What we the Uses of Theorebid Lexicography? Didbnaries. 14. 1992-3. 44-7%. Replies to Discussmts. DLfimwries. 14, 1992-93. 139-59. The Meaning of Conour Terms: Semantics, CnBture, and Cognition. Cognitive Lingwtstics. l$1. 1990. 99-P-nTO. Dictionaries versus Encyclopaedias: How to Draw the Line. In Philip Davis i(ed.1. DescrQative wd Thearetical Modes in the Alternative LinisguLrics. Philadelphia! Amsterdam: John B e n j d s . Forthcoming. What is a Life Form? Conceptual Issues in Ebhnolbiologg. Journal of Linguistic AmrhropoIagp. 21. 1992. 3-21). Semantic Rules Know no Exoeptions. Studies in htagttage. 15B. 1991. 37 1-98. The Semantics of Grammar: A Reply to Professor Palmer. Journal a f Lingwhtics.
27/2. 1991. 495-8.

1. Language and Meanring 2. SemantL Primitives (or Primes) 3. Lexical Universals 4. Innate C o n a p t s and Language Acquisition 5. The Universal Syntax of Meaning 6. The iC\Baturrul Semantic Metalanguage m S M ) 7. Semantic Invariants 8. Metlhodologi@d Issues 9. Past, Present, and Future of NSM Semantic Theory

2. A Smrvey af Semantic Prirnitivlt?~

A. OLD PRPMITIVES 11. Introduction 2. Substantives: I, YOU, SOMEONE, SOMETHING, PEOPLE 3. Determiners: THIS, T H E SAME, OTHER 4. Quantifiers: ONE, W Q , MANY (MUCH)l, A L L 5 . Mental Predicates: THINK, KNOW, WANT, FEEL 6 . Speech: SAY 7, Actions and Events: D O and HAPPEN 8. Evaluators: G O O D and S A D 9. Descriptors: BIG and SMALL 10. T i m : WHEN, BEFORE, AFTER 11. Space: WHERE, UNDER, ABOVE 12. Partonomy a n d Taxonomy: BART (OF) and KIND (OF) 13, Metapredicates: NOT, CAN, VERY 84. Zntercllawsal Linkers: IF,BECAUSE, LIKE

k Semantic Basis far Grammatical Typdogg. In Werner Abraham, Talmy Givon, aad Sandra Thompson (eds.). Discourse. G r a m n r and Typolagye Complementary Series of Studies in Language. Amsterdam: John Bemjamins. l79-209. Semamftics and Epistemology: The Meaning of 'Evidentials' in a Cross-linguistic Perspective. Language Sciences. 16,'I. 1994. 8 1-137.

15. 16. 17. 18.

Introduction Determiners and Quantifiers: SOME and M O R E Mental Predicates: SEE and HEAR M~~~vememt, Exisllenoe, Life: MOVE, T H E R E IS, LIVE

19. Space: FAR and NEAR, SIDE, INSIDE, HERE 20. T h e : A LONG TIME, A SHORT TIME, NOW 21. Imagination and Possibility: IF . . . WOULD, CAN, MAYBE 22. WORD 23. General Discussion: Opposites and Converses 24. Conclusion

89 97
101 107 108 110

2. Natural Kinds and Cultural Kinds 3. Speech Act Verbs 4. Emotion Concepts 5. Conclusion
6. Semantics and 'Trimi~ve a?.ougStW 1. Introduction 2. The Universality of BECAUSE 3. The Universality of IF 4, The Universality of SOMEONE 5. The Universality of ALL 6 . The Universality of KNOW and THINK 7, General Discussian

3. Universal Grammar: The Syntax m f Udxersal Semantic Primitives 1. Introduction 2. Preliminary Discussion 3. Substantives: YOU, I; SOMEONE, PEOPLE; SOMETHING 4. Mental Predicates: THINK, KNOW, WANT, FEEL, SEE, HEAR 5. Sglm3Al: SAY 6. Actions, Events, and Movement: DO, HAFPEN, MOVE 7. Existence and Life: BE (THERE ISARE) and LIVE 8. Determiners and Qluanltifiers: THIS, THE SAME, OTHER; ONE, TWO, h i Z A W (MUCH], SOME, ALL 9. Augmentor: MORE 10. Evaluators: GOOD and DAD I I. Descriptors: BIG and SMALL 12. Time: WHEN, AFTER, BEFORE, A LONG TIME, A SNORT TIME, NOW 13. Space: WHERE; FAR, NEAR; UNDER, ABOVE; SIDE; INSIDE; HERE 14. 1ntercUausa;mlLinkers: BECAUSE, IF, IF . . . WOULD 15. Clause Operators: NOT and MAYBE 1 1 6 . Metapredicate CAN 11 7. Intensifier: VERY 18. Taxonomy, Partonomy: KIND OF, PART OF 14. Similarity: LIKE 20. General Discussion
4. Prototypes amd Invmiiants P . Intwducticrn 3 2 . Abuses of "'Prototypes'9n Semantics: Some Illustraitions -b3. Uses of 'Tro~totypes"in Semantics: Some Illustrations 4. Conclusion

112 1112 P1 1 3

P 14
119 120 122 3 24

126 129 130 131

131 133 137 139 140 141 1141 143 144

7. Semantic CompIexty and the Wob IcDf Ostemsian in the Acquisitian of Cancepb 1. Introduction 2. Complex Concepts as Configurations of Simple Ones 3. Abstract Concepts: Words For Emotions 4. Relatively Simple Concrete Concepts: Body Parts and the Natural Environment 5. Temperature Terms and the Conoept of 'Fire' 6. Cultural and Naturall Kinds: 'Breadknd Water' 7. Plugging Concepts In 8. Conclusion

211 2 1 11 2112 2114

2118 22 1 225 232 233 235


I. 2. 3. 4. 5. 6. 7. 8. 9. 10.
Linguistic Meaning Research Definitions as a Tool far Cross-cuIturaQ The Conoept of a Semantic Invariant Determinacy of Meaning Problems of Polysemy Lexical Mearming and Illocutionary Rhetorical Devices Family Resemblances Dictionary Definitions Fodor on Definitions Clsnclusion

3 1 4 8 148 149 160 167

5. Semrrwdc RimiitiExes and Semantic Fields 1. Introduction

9. Semawtics and kxicogrspby P. Introduction 2. Scope versus Adequag and Truth

3. Saying Something that is not True 4. Saying Something that is Superfluous 5. Confusing Meaning with Knowledge 6. Definition8 which are too Broad 7. Capturing the Invariant 8. Standing Firmly on the Cirollmd of Discreteness 9. Distinguishing Polysemy from Vagueness 10. Avoiding Circularity 11. Relying on Indefinables 12. Using Simple Language 13. Exploring New Models of Definition 14. Conclusion
8, introduclion 2, Mcaniny and Scientific Kncawllcdgc 3, Meaning and Coilour Charts 4. Meaning and Psychological Reality 5 . Colour Terms as Quotations 6. "Black" awd 'White'" 'Dark"' and "Lighlt" 7. Green, gwyrdd welsh), btuy (Hanunlliio) 8. Blue, rniebfaki (Polish), goSuboj and 8 h Q (Russian), aoi (Japanese), and f& flhsui) 9. ""Red" and "Yell~w" 10. Macro-white and Macro-Mack 11. Macro-red and Grue 12. Names of Mixed Coiours 13. caBrown" 84. Names of Specific (Locally Salient) Referents 15. Condusion: Chromatolrogy, Cognition, and Culture
1 1 . The Semantics of Nalhrral Kinds 1. Iaatroduction 2, Abstract Comcepts and Concrete Concepts

Types of Linguistic Evidenm "Life F o m s Y 3 nEnglish Folk Zoology Are there Monogeneric ""LiEe Forms"? "Life Foms'7n English Folk Botany Polflypic Genera "'Ciestalts" a d "Distinctive Features'" "Hidden Namres" and ""Proper Names'" 101. Living Things and Artefacts: Similar or Radically Different? 11, Conrclwsion
3. 4. 5. 6. 7. 8. 9.


358 360 361 365 366 367 370 374


aP. THE S E W l C S OF G R R
1 1 1 3 . Semntie Rules i e Grammar I . Introduction 2. Semantic Rules: The Past Practiw 3. 'Wual Nouns" and Absolute Predictiveness 4. Evidence for Different ConceptualFzaEions 5. The Mystery of Scaks 6 . Predictiveness and Different Languages 7. Different Cultures, Different Conceptualizatims 8. The Semantics of Gender 9. The Unconscious Character of Semantic Rules 10. Conclnasi~n
14. A Semantit Basis far Grannmattsl Deseriptiau and Typailagy: Tramsithity and Reflexives 1. Introduction 2. The Uniqueness d Grammatical and Semantic Systems 3. Typology and Semantics 4. Reflexive Constructions 5. Transitive Constructions 6. Concliwsiosr 15. CompEurimg Grammaticral Categories across Lmguag;es: The Semantics loif Elridemtialls 1. 111tr~du~tion 2. Kashaya 3. Quechua 4. Wlntu 5. Maricopa 6. Bulgarian and Maoedonian 7. Conclusion 8. A Summary of the FormuIae

402 404 4 x 0 1 7 409 420 425

3. 4. 5. 6. 7.

Scientific Knowledge versus Everyday Knowledge An IPlwskration: Folk Mice versus Scientific Mice The Evidenoe for the Folk Concept General Discussion Concllusion

12. Semantics and Ethuobiloliagy 11. Introdunction 2. Ethnobiological AnaByds: Tools and Methods

General Issues


1. Language andl Meaning


Language is an instrument -- . - roranveying meaning. The structure of this reflects its function, and it ;an onlf be I;roperly understood in items of its fmctionr. To study language without reference to meaning is like studying road signs from the point of view of their physical properties (haw much they weigh, what kind off paint are tOiey painted with, and so on], or like studying the structure of the cye without any rcfercncc to seeing. Curiously, however, this is precisely h w many linguists study language. A scien~e of language in which meaning has at best a very marginal place is an anomaly and an aberration (which in itself will present an absorbing topic of study for the future historians of linguistics)); and of course not all present-day linguists approach the study of language in that spirit. Yet in university curricula currently adopted in many linguistics departments throughout the world, "formal syntax" still occupies a far more central place than semantics (the study of meaning), and semantics is still often treated as marginal. Two twentieth-century American linguists have k e n particularly influential in shaping a '"nnguistics without meaning"Ae0nad BBoomfield and Noam Gluomsky B1oomfield (unlike his great contemporary and co-founder of American linguistics, Edward Sapir) was afraid of meaning, and was eager to relegate the study of meaning to other disciplines such as sociology or psychology. The reason he was afraid of it was that he wanted to estabiish linguistics as a science and that he thought that meaning couldn't be studied with the same rigour as linguistic sounds and Foms. BloomfieEd% bbeavioousism made him find all references to ideas, concepts, thoughts, or mind unscientific; "'mentalism" was used by him, and by mamy other influential linguists of his generation, as ;a dirty word.' As Randy Allen Harris, the author of The Lhg-uiCsrfcs Wars (1993: 27-81, put It: "Bllaomfreld" ideas defined the temper of the Binguisltic times: that it [linguistics] was a descriptive and

, -

I As a close oollabonatar of Sapir, Morris Saadesh (1941: 5 9 ) , painted out, another eonfinned behaviourist, Twaddell, "criticized Sapir as a mentallist dealing with an 'urnknown and unknowable mind"".

taxonomic science, like zoology, geology, m d astronomy; that mental speculations werle tantamount to mysticism, an abndonmenr of science; that questions (learning, knowing, and using a Ianall the rebvant psyclluotogi~l guage) would be answered by behaviorism; that meaning was outside the scope of scientific inquiry." It has often been said, in F3looirmfielld's defence, that it wasn't BloomfieOd himself but the "Bloomfieldians" or "post-Bloomfieldians" [and especially Chomsky" mentor Zeillig Harris) who sought to banish meaning from linguistics. For exampb, Matthew (1943: 8 114) points out that even "in one of his last general papers he [Bloomfield] continued to make dear that 'in language, f o m s cannot [be separated from meanings"' (1943; In Hockett 11970: 401). But it is not unreasonable to say that what the PostBloomlie1dians did was to take BEoodield's largely (though not consistently) anti-semantic stand to its logical ccmclusion. Matthews tries to explain why lBloodeldyssuccessors "came to believe that forms couId and should be described without reference to meaning'" and 'khy, in adoptiug a theory in which the separation af form and meaning was axiomatic, they were so sure they were continuing his work". He notes that the usual explanation given is '"hat however oentral meaning may have been and however important its investigation, BBoodeldys account of how it should be described effectively closed the door to scientific studyy' (1993: 115). Matthews seeks to distance himself from this conclusion but in my view it is inescapable. Bloomfield didn't '?ejectyheaning in the sense of avoiding any mention of it In Pinguistic description but he did want to exclude semantic considerations from linguistic analysis. For example, he ridiculed the idea that the grammatical catelgolly of number [singular wrsas plural) has a semantic basis and could be defined with reference to meaning: '"hod grammar defines the class of plural nouns by its meaning "ore than one' (person, place, or thing), but who could gather from this that o m is a plmal while whe& is a singular? Class-meanings, like a1 other meanings, ellude the linguist" power of definition." "933/1935: 26612 Bbofie11d himself denied that he had ever wanted to "undertake to study language without meaning, simply as meaningless sound" (letter to Fries; quoted in Hymes and Fought 1975: 1QO9); but the message of Longwage was none the less 11oud and clear: there was no room for semantics within tlne "linguistic science", at least not for the foreseeable future.
We have definued the meaning oF a linguistic form as the situation i n u which the speaker utters L and the response which it calls forth from the hearer. . . . The sitCuriously, Bloomfield didn? pay any attention to the Fact Ghat oafs is no1 a "plural" contrasting Gfih a singular (Uke, for example, dogs wntrasts with 6%) and that It dawsn'8 really belong to the same " f ~ m class" as dogs does. The "Som cllass"io which oats belongs, and its iruvariaol meaning, is discussed in Chapter 13. (See also Wierzbicka 1988.)

oations which prompt people to atter speech include every abject and happening in their universe. IN order to give a scientifically accurafe definition of meaning for every form of a language, we should have L o have a scientificaily accurate knowledge of everythimg in the speakers' world. The actual extent af human knowledge is very small, compared to this. We can define the meaning of a speech-form accnrately when this meaning has to do with some matter of which 1 ~ possess e scientific knowledge. We can define the names of minerals, for example, in terms oE chemistry and mineralogy, as when we say that the ordinary meaning of the English word salt is csodium chloride CNaCl)', and we can define the names of plants or animals by means o l the technical t e m s of botany or zoology, but we have no precise way of defining words like lave or hate, which cancern situations that have not been aaocurately classified-andl these latter are in the great majority. . . . The statement of meaning is therefore the weak point in language study, and w i 1 3 remain so until human knowledge advances very far beyond its present state. Pn practice, we d e h e the meaning of a linguistic. form, wherever we can, in t e m s of some other scienoe. Where this is impossible, we resort to makeshift device^.^ (BloomfieBd 193331935: 159-40)

Thus, for IBloonnfUeld meaning codd be referred to, but not studied, and given his "anti-mentalistic"', behaviouristic conception of rmeaninrg, it could scarmly have been otherwise. As Hymes and Fought (1975: BO1Q) put it, "'Boomfield included meaning in his conoeption of language structure but not in his short-term linguistic theory. . . . scepticism as to the practical pos;sibility of incorporating meaoring explicitly in linguistic analysis led to shifts . . . to reliance ern distributional patterning . . . among the Bloomfieldians." The ''cognitive revolution" of the late fifties and the sixties banished (or 3Z E % m e E e ghost 6T6ehiviourisii1, and made mind, ariid meanfig; 5 so % ~ e f i c n r bf n human scie-noes in general, and of linguistics in par6ci-i'&K:TQ icgaote one 02" the main actors of the 'tognithe revdu1ion"Jerome G n e r (1990: 8)l: 'That revollution was intended to bring k i n d ' back into & kh@-&an sciences after a long cold winter of objectivhsm."For Bruner, B "nimind"Vs dosely related to '%neaningW:"Now let me tell you first what I and my friends thought the revolution was about back there in the Bate 1950s. It was, we thought, an all-out effort to establish meaning as the mntral conoept of psychology-nd stimuli and responses, not overtly observable behavior, not biological d k s and their transformation, but meaning" (Ila. 2). But, in his own words, Bruner's is not '"he uswai account d progress marching ever forware I@. 1); for in his view, "that revolution has now been diverted into issues that are marginal to the impulse that brought it
Bloomfield's reference to "NaCl" as "the ardinary meaning of the English word sali3' highlights his failure to dislinguish scientific knowledge From "ordinary meaning"',a a s da also P v j s remarks on the names o f plants. and animals.. For detailed discussion of these matters see Chapters 11 and 12. As for the meaning of emotion terms (such az b v a and hotel, see Chapter 5.

into being. Indeed, it has been technicalized in a manner that even undermines the original irnpulsel"(p. 11. What has becn lost sigE~tor is meaaing. Very early on, for example, emphasis began shifting from "meaning" to 'cinfomation," "om the cuvssrruclbn OF meaning to the processing raf information. These are profaundlly different mattcrs. The key factor in the shift was the introduation of computation as the ruling metaphor and of computability as a necessary criterion of a good theoretical modell. lnformation is indimerent with respect to meaning. (P.41 'Very soon, computing became the model of the mind, and in place of the concept d meaning there emerged the concept of cornpuhBaWlity, i@. 6) It was inevitable that with computation as the metaphor of the new cognitive science and with computability as the necessary if not sufficient criterion of a workable theory withim the new science, the old malaise about mentallism wadd =-emerge. @. 8) Brumer decries the ""cgnitive revolution" for abandoning meaning as its central cmcern and for "'opting far Ynfcvrmation processing' and computation instead" (1371; and he urges "'that psychology stop trying to be ' m a n ing-free' in its system of explanation" (20). But if psychology has been betrayed by the "cognitive revolution", with its escape from meaning, what is one to say of linguistics, in which the promising early refeaences b "'mnd" (as in Chomsky's Language and Mind), have led to a premmpation with formalisms, and in which "meaning-freeJ' syntax has for decades usurped the place rightfully belonghg to the study of meaning? Oliver Sacks (1993: 48) smmarizes the c'hijacking'y of the "cognitive revolution" as follows: "Bauner describes how this oridnal impetus was subverted, and replaced by notions of computation, information processing, eltc., and by the computational (and Chomskyanr) notion that the syntax of a language could be separated from its semantics."Sacks strongly endorses Bruner's position, and comments: "From Boole, with his 'Laws of Thought' in the 1850s, to tlne pioneers of Artificial Intelligence at the present day, there has Z r e e n a persistent notion that one may have an intelligence or a language based on pure logic, without anything so messy as heaning9eing involved." Unfortunately, as noted by Sacks, this persistent notion was shared by the main spirifus mavens of the c'c~ga$Sive revolution" in Ilinguistics, Noarn Chomsky, whose influgncc on the filcld can hardly bc ovcrestimaled. is mentalist, anti-btoornfieldian stand, in his attitude to meanskr remained (and still remains) a Bloodeldian. Like : Bfo@mmfield,~""h .., . had a deep mdhodologicail aversion t'o meaning, and REWo*-rmPorccd onrey.Fthk key @lemm<s s f the BEEEoomfieldian policy ToGar'iI'meiining: it had to be avoided in formal analysis" (R. A. Hamis rl mT$9p. -

1 agee with Harris (1993: 252) that while some "prefer to look at Chomsky's impact on linguistics as Ilnc last gasp or U3loomficldianismu'~ such a view is "far too narrow". But one also has to agree with Chomsky's critics that although he broke, in a way, Bloomfield's taboo on mind, Chomsky's professed mentalism proved to be as inimical to the study of meaning as was Bloom;fieldlysbehaviourism. To quote one critic (Edelman 11992: 243):

One of the most pervasive and inflwemtial approaches to these critical questions
[af how language and thought are connected] was pioneered by Chamsky. I n u his

formal systems approach, the principal assumption is that the rules of syntax are independent or wmmtim. Language, in this view, Is independent c u f the rest o o F mgnition. I must take issue with this notiom. The set of rules formulated under the idea that a grammar is a formal system are essentially algorithmic. In such a system, no use i s made of meaning. Chomskyk sacalled generative grammar . . assumes that syntax is independent of semantics anud that the language faculty is independent of external cognitive capabilities. This definition of grammar is impnriolrs to any attempt to d i s o o n h it by referring to facts a b u t cognition in general, A language dehed as a set of strings of uninterpretd symbols gemcrated by praduction rules is like a computer language.


This brings us back to Brunar" remarks quoted earlier. As ha points out (1990: I), "the new cognitive science, the child of the [cognitive] revolution, has gained in technical wccessles at the price of dehumanizing the very concept it had sougEEt to reestablish in psychology, and . . . has thereby estranged much of psychology from the other human sciences and the humanities"".The same can be said about linguistics. In 'talking abut a "linguistics without meaning" I do not wish to underestimate the work done in linguistic semantics over the last several decades. Nor would I question the significance of the other trends in linguistics that sought to transcend the limitations imposed upon the discipline by generative grammar. Harris (1993) and others are right to rejoice in the "greening of linguisticsn"ofthe last decade or two, with the dynamic development of functional linguistics, icognitive linguistics, pragmatics, and so on. At the same time, however, E think that the Bloodeldian and Chomskyan antisemantic bias is still hanging over linguistics like a dark shadow. The fact that "formal syntax" still occupies ;a prominent place in the curricula of many linguistic departments, at the expense of the study of language as an instrument for conveying meaning, gives suflicient subslance to this daim. In the latest version of Chomskyan linguistics references to meaning are apparently n a langer disallowed. But this does not change its basically antisemantic orientation, Chomsky no longer asserts that 'Yf It can be shown that meaning and related notions do play a role in linguistic analysis, then . . . a serious Mow is struck at the foundations af linguistic theory'' (1955:

141). But he none the less r~emains what he has always been: "a deep and syntactic Lndamentalistw (R.A. Harris 1993: 139). Matthnvs 5 h (1993: 245) sums up his commenls on tP~cplace of meaning in Chomskyss $ y T Y c r ; ; ~ e e n t work as follows: "Where did that leave an aiclcounk of meaning? ~ a ~ f i F . ' ~ h o m s as ~ ,dways, is primarily a student of syntax, or of "ranmar' in * a traditional sense. Therefore we can expect, as always, little more than programmatic statements and passing remarks.'" Nor has; the semantic void created by the "syntadic fundamentallism" of Chomskyan Igmmar been filled by the so-called 'Tomal semantics", which also features prominently in the teaching p r o g a m e s of m y Iingwistics departments. Despite its name, "formal semanticsY"or "model-theoretical semantics"") doesn't seek to reveal and describe the meanings encoded in natural Ianguage, or to compare meanings across languages and cultures. Rather, it sees its god as that of translating certain carefully selected types of sentences into a l o g i d calculus. 32 is interested not in meaning (in the sense of conicleptual structures encoded in language] but in the logical progarties of sentences such as entdment, contradiction, or logical equivalence or, as Chierchia and McConnell-Gin (1490: 1 B] put it, in "infannational significance", not in "cognitive signlficmceJ'. (Cf- Bruner's (1990: 4) comments on the shill from "meaning" to "information"', quoted earlier.) To quote one noted formal semanticist (of the ""Wonta$ue grammar" schiooPl), "the model theoretic intension of a word has. in principle xsothhg wharsocver to do with what goes on h a person's head when he uses that word" (Dowty 1978: 3751). Having explained that in modd-theoretical semantics the meaning of a sentence is seen as "a set of possible worldsy', Dowty acknowledges that "one may reasonably doubt whether sets of wssible worlds have mything at all to do with the psychological process of sentence comprehension'", and he admits that "there is no sense in which a person mentally has access to 'all the possible worlds that there are"v1376). Thus, Chornskyans like to tajk about "mind'" but do not wish to study meaning, and "formal semanticists" like to talk about "meaning'3ut only i n the mnse of possible worlds ar truth conditions, not in the sense of conwpruaE structures. One thing that both schools share is the great emphasis they place on being formal. This emphasis on formal models, at the expense csf a search for meaning and understanding, brings tor mind, again, Bmner's (11940: 65) remarks about psychology: ""Isimply will not do to reject the theoretical centrality of meaning for psychology on the grounds that it is "ague" Its vaggvenss was In the eye of yesterday's farmallistic logician. We are beyond that now." Despite all tine promises of the '"cognitive revolution" in h m a n scienoes in general and of the "Chomkyan revo1ution'~n linguistics, now, at the close of the cemtury, meaning (not the 1ogician"s"meaning" h t lthe mean-

%@SKY abiding


ing which underlies human cognition, communication, and culture) is still regarded by many linguislts as messy and as "the weak point of language .studyY"(Blboomfield 193311935: 140). This book hopes to demonstrate tbat it doesn't have to be so.

2. Semantic Primitives (or Primes)

To put it brinefly, in human speech, diKferent sounds have diflerent mearrmingls. To study this co-ordination o f certain soumuds with certain meanuings, is to study lamguage. Leonard B;Laodield (193311935: 27)

How is it possible to admit that to study language is to study the oorrelations between sound and meaning and, at the same time, to try to keep linguistics maximally ''meaning-free"? Bloondield's own reason for this contradictory position is quite clear: he wanted linguistics to be a serious and rigorous discipline-"a science"; and it was not clear at the time how, if at all, meaning could be studied in a rigorous and '%cientific"manner. In fact, even today, many defenders of the central role of meaning in linguistics don" sean to mind if meaning is spoken of in a loose, vague3~d hoc way, without any coherent methodology. On this point, I must say that I agree with Bloodeld: if we really want to study, in a ~ g o r o u s way, correlations between sounds and meanings (or between f o m s and meanings), our standards of rigour and coherence in talking about meaning shoulld Ire just as hi& arid exacting as in talking about sounds and forms. As I have tried to demonstrate for a quarter of a century, the key to a rigorous and yet insightful talk about meaning lies in the notion of smantic primitives (or semantic primes)~. To take an exmple. Two prominent researchers into child language and the authors of a very valuable study on the acquisition of meaning, Lwcia French and Katherine Nelson (1985: 381, start their discussion of the concept ' i f ' 'by saying: "it is dilrficult to provida a prccixc definition oS the word $'" Then, after some discussion, they conclude: "The fundarnenttal meaning of 6 in both logic and ordinary language, is one of implication.'" Two common assumptions are reflected in these statements. First, that it is possible to define all words-including +and second, that if a word seems diKcw8t to define, one had better reach for a scientificsounding word of Latin origin [such as impSicatioion). In my view, these assumptions are not only false, but jointly constitute a stumbling-block for semantic analysis. One cannot define all words, because the very idea of 'defining' implies that there is not only something to be defined [a

defmiendum) but also s o m e ~ n to g define it with (a deffiniens, or rather, a set of ""dlfinienses").

semantic practice, presented in this book: meaning cannot be described without a set of semantic primitives; one can pmport to describe meaning by translating unknowns into unknowns (as in Pascal's (l667!1954: 580) mock-definition "'Light is the luminary movement of luminous bodies'"), but nothing is really achieved thereby. Without a set of primitives all descriptions of meaning are actually or potentially circular (as when, for example, to demand is defined as 'to request h l y " and to request as 'to demand gently" see Wierzbicka 1987~:

more than two millennia ago by Aristofle (1937: 1141"): First of all, see if he [the anallyst] has failed to make the definition through terns that are prior and more intelligible, For R h e reason why the definition is rendered is to make known the tern stated, al'd we make things known [by taking not any random terms, ]but suclr as are prior and more inte1ligible . . . accordingly, it is dear that a man who does not define through ,hems of this kind has not defirmed at all.

It could be argued that what is clear to one person may not be clear to another, and that therefore no absolute order of semantic simplicity cam be established. To this, however, Aristode had an answer: what matters is mot what i mare intelligible to particular individuals, but what is semantically more basic and thus inherently more intelligible:
For, as it happens, dinerent things are more intelligible to difFerent people, not llhe same things to all . . . Moreover, to the same people different things am more i n d lligisble at different times . . . so chat those who hold that a definition ought to ha rendered through what is more intelligible to particular individuals wouM not haw to render the same definition at s l l l l ltimes evenu to the same person. It is clear, then, that the night way to define is not through terns of that kind, [but through what is absolutely moue intelligible: for only in this way could the definition come always to be one and the same. The "absolute order oF understanding" depends on semantic complexity. For example, one cannot understand the concepts of ' p r o m i s e h r 'denounce%ithout f i s t understanding the concept of "say', for 'proenlse" and Udenance' are buiEt upon %sayy. Similarly, one cannot understand the concepts of 'deixis1% 'demonstration', or 'ostenslon'wwithout first understanding the concept of "his" on which they are built; and one cannot understand the concept of "mpllication' wilthout first understanding the semantically mare basic concept of 'iP,

d out in my Semantic Primitives (Wierzbicka 1472 3), constudents of artificial languages often glace great emphasis on the arbitrariness of "primitive terns", For example, Nelson Goodman (11951: 57) wrote: "It is not bscawe a term is indefinable that it is chosen as primitive; rather, it is because ;a term has been chosen as primitive for a system that it is indefinable . . . In general, the t e m s adopted as primitives of a given system are readily definable in some other system. These is no ;absolute primitive, no one correct selection of primitives." Bunt the idea that the same applies to the semantics of natural language is a fallacy, and a recipe far stagnation in semantic research. There is of course no reason why linguists shouldn? invent arbitrary sets of primitives and "defineY%hatever they like in terms of such sets. Bunt it will do little to advance our ulnderstanding of human cornm~mimtionand cognition. To quote Leibniz:
If nothing could be comprehended in, Itsdf nothing at all muld ever be comprehended, k u s e what can only be comprehended via something else can be cornpreheadedl only to the extelolt to which that other thing can be comprehended, and sol on; acaordindy, we can say that we have understood something only when we have broken it dowo into parts which can be understood In themselves. [Leibniz 1903/19611; 430; my translation)

Semantics crvn have an expIamratory value only if it manages to '3dene'"or explicate) complex and obscure meanings in items of simple and selfexplanatory ones. If a human being can understand any uttesanoes at all

(someone else's or their own) it is only because these utteranices are built, so to speak, out of simple elemenlts which can be understood by themselves. This basic noint. which modern lin~uisticshas lost sight of. was made

Further 1declare that there are certain things which we render more obscure by tpyimg to define them, because, since they are very simple and clear, we annot know and perceive them better than by themselves. Nay, we must place in the number of

those chief errors that can be co&tted in the sciences, the mistakes committed by those who would try to define what ought only to be conceived, and who cannot distinguish the clear from the obscure, nor discriminate between what, in order to be known, requires and deserves to be defined, from what a n be best known by itself. (170111931: 3241 For Descartes, then, as for Leibniz, there was no question of "choosingy' some arbitrary set of primitives. What mattered was to establish which concepts am so clear that they cannot be understood better than by themselves; and to explain everything else i n terns of these. This basic principle was applied first of all to lexical semantics, and was phrased in terms of the definability of words. For example, Pascal wrote:
It is dear that there are words which cannot be defined; and i f nature hadn't ppso-

vided far this by giving all people the same idea a H our expressions wauRd be obscure; but in f a t we can use those words with the same confidence and certainty as if they had been explained in the clearest possible way; because nature itself has given us, without additional words, an understanding of them better than what our art could give through our explanations, (1667iP9541: %Oi] Similarly, Arnauld: Our first observation is that no attempt should be made to define all words; such an attempt would be useless, even impossible, to achieve. To define a word which already expresses a distinct idea unambiguously would be useless; for the goal of definition-to j~ointo a word one dear and distinct idea-has already been attained. Words which express ideas of simple things are understad by all and require no definition . . . Further, it i s impossible b define all words. In defining we employ a definition to express the idea which we want to join to the defined word; and IF w e them wanted to define "the definition,'%till other words would be needed-and so on to infinity. Hence, it is necessary to stop at some primitive words, which are nat defined. To define too much is just as great a failing as to define too little: Either way we would fill into the confusion that we claim L o avoid. (l6162/1964: 86-T; emphasis added) Chomsky, despite his claims that generative grammar was a continuation of "Cartesian linguistics" (see Chamsky 11966), has always omitted any mention of this central thread in the Cartesian [as well as the Leibnizian]

theory of language and mind. (See also the references to the "Cartesian conception" of language and cognition in Chomsky's s o r e recent writings, e.g, in Chamsky 199113). My o m interest i n the pursuit of non-arbitrary semantic primitives was triggered by a lecture on this subject given at Warsaw University by the B"cPPish linguist Andrzej Bogusiawski in 1965. The 'gooldle dream" of the seventeenth-century thhkers, which couldn't be realized wiltbin the framework of philosophy and which was therefore genera-ally abandoned as a utopia, could be realized, Boguslawski maintained, if it was approached from a linguistic rather than from a purdy philosophical paint of view. The experience and achievements of modern linguistics (both empirical and theoretical] nude it possible to approach the problem of conceptual primitives in a novel way; and to put it on the agenda of an empirical science. &elbnizYs theory of an "alphakt of human thoughts" (l903!1961: 435) o d d be dismissed as a utopia laecause he never proposed anylthing like a campfete list of hypotheticaf primitives (although in his unpublished work he left several partial drafts, see kilbniz 11903). As one modern commentator wrote, having pointed out the difficulties involved in the proposed swrck " h thew circumstanm it is understandable that bibnbniz should consistently avdd the obvious question as to the number and type of funmental concepts. The approach would be more convincing if" one could at least gain same clue as to what the tabb of fimdamental concepts might ok like" (Marltin 3964: 25). The best clues as to what the table of fundameaatal concepts might ioak like come from the study of languages. Pn this sense linguistics has a chance of succeeding where philosophical speculation has failed. This book, which is based on linguistic research undertaken (by lcolleagues and myself) over three decades, does propose a complete (if hypothetical) table of fumdamental human concepts capable of generating all other concepts (see Chapter 2). Crucially, this Pist purports ranso to be a table of laical universals--a point which will be discussed in the next section.

In tine: theory presented in this book it was hypothesized, from the start, that conceptual primitives can be found through in-depth analysis of any natural language; but also, that the sets of primitives identified in this way would "makch"",nd that in fact each such set is just one language-specific manifestation of a universal set of fumdamend human concepts. For example, it was expected that the concepts 'someone', 'something', and 'want', which are indefinable in English, would also prove to be inde-

1 4 General 13sues

finable in other languages; m d that other lmguages, too, will have words [or bound morphemes) to express these conoepts.

in all languages'" But it is precisely this strongest universalist hypothesis which was tested in Semamfic and Lexical E$niver$aissand which also underIks the present book. While h e theory presented in this book is radically universalist, two provisos must be entered: fi

native speakers of different languages. Since the indefinable concepts-the primitivew-are the fundament m which the semantic system of a language is built, if this fundament were in each case dflerent, speakers of different languages wodd be imprisoned in difierent and incommensurabie mnceptual systems, without any possibility of ever reaching anyone outside one's awn prison. This is mntrary to human experience, which points, rather, to We existence of both differences and similarities in the human conceptualization of the wodd; and which tells us that while cross-cultural communication is difficult, and has ilts limitations, it is not altogether impossible. The assumption that all languages, however different, are based on iso- ,/momplhic sets of semantiixJmitives is comiktent with that experience. n u d l recently, this assumption was based largely ton theoreticdGnsidrQWGd.F'"'&atiorrs; ratheb than on e&pirical studies of &&rent languages of the world. This situation, h~wever,has changed with the Gb6mtion of Scmanrie and Lexical Universab (Goddard and Wierzbicka 19948ba cotleetian in which conceptual primitives posited initially on the basis of a mere handful of languages were subjected to a systematic study across a wide range of languages from diFEerenr fmilies and dilMjerent continents. The languages investigated in this volume included: Ewe (of the Niger-Congo family in West Africa), Mandarin Chinese, Thai, Japanese, the Australian languages Uankunytjatjara, Arrernte (Aramda), and Kayardild, three Misumalpan languages of Nicaragua, the Austronesian languages Acehnese (of Indonesia], Longgu (of the Solomon Islands), Samoan, and MangapMbulla (of New Guinea), the Papuan language Kdam, and-the only European language beside English-French. This first large-scale attempt tto test hypothetical conoeptuall primitives cross-Pingunistically did not answer all the questions, but except ;for one or two grey areas requiring further investigation, the studies induded in the volume did strongly support the hypothesized set of primitives. In most cases, words (or bound mophems] for the proposed primitives (e.8. 'Iy and cyouy,'someone' and %somethingy,%herey and 'when', 'bigJ and 'small', 'good' and %a&, or 'do'and 'happen') could be readily identified. 1x1 his discussion of "'universalism'9n semantics, John Lyons (11977: 331-2) stated that as far as he could see, no one advocates the most extreme f o m of "semantic universalism", that is, the position that ""there is a &xed set of semantic components, which are umniversal in that they are lexicalized
p m

M u 1

A s all itpanslatom know to their cost, every language has words which m a p a v e no semantic equivalents in other languages, and every language draws -emantic distinctions which o t h e ~ languages dlio not. For example, tramnsjaking the dassk texts of the Hindu cultural tradition into European languages one must face the fact that these languages do not b e words coming even near in meaning to key Sanskrit terns such as nirvma, brahman, atman, or kmma (see BoDe 1W7: 219-583. But even comparing languages which are genetically, geogairphically, and culturally very dow, for example French and Endish, one constantly encounters examples of profound lexical differences. For example, the French word mathem- has no counterpart in English, as pointed out by the English translator of Sirnone Neil" meditations on this concept, who finally in desperation decided tan use, throughout his translation, the totally inadequate English word c'a~iction" (Vi e ' il 1972: 1633. In a sense, most words ian all languages are like the French maCheus, that is, unrenderahle (without distortion) in some other languages. More than that, every language has words which are intimately bound up with one particular culture and which have no equivalents in any other languages. (See e.g. Wierzbicka 1991b, 1992a). At the same time, a 1 3 1 languages also have words which-unlike msaJ'hesss-do appear to have semantic counterparts in all other languages. The hypothesis explored in this book (and in the work which led to it) is that in every Ilanguage the set of such readily '"anslatable" words coincides with the set of this l a = g l _ s & < ~ ~ ~ a ~ e s . nt belongs to a unique netwoZ "in a u6ijuFnelfurorkof -. rela--* rtmguages we cannot expect to z E ~ x s ~ ~ % X " ~ W ~&- n W. i~5~i i z s i ' i $ ! &~ - ~ ---of indehables. -.---- *-marphism in the lexicon (and, as we shall see, also in grammar) that gives substance to the motion of universal semantic primitives. For exmple, the English words big and d l correspond in meaning to the Russian words boSr$oj and maien%g, even though in English, 3mali has

also a special relationship with fibfie, and even though in Russian, rna/enlkjk-fomally a diminutive-has a special relationship with diminutive adjectives such as belenrkV ('white' + DD~M) or JElb~glen'kij('round' + DIM). Whatever the diflerences in "resonance" (see Section 8.7) between small and malenrkg are, these differences cannot be shown through definitions; and so, from a definitional point of view, they constitute a "perfect" match (in the systems of English and Russian indefinables, they omupy the same slot). Similarly, regardless of any diflerences in, 'kesonance'~anduse), the Japanese words ookii and fiigai oonstitute a perfect semantic match for big and small,and the Japanese words d i and wasui, for good and bad (See Onishi 1994.) Furthermore, it is only the postulated isomorphism of exponents of conptual primitives which allows us to oompare different semantic systems am. For any comparison requires a berfiurn compcrratfoni8, a common maswe. The hypothesized set of wiversal semantic prlrniPives offers us such a common measure and thus makes it possible to study the extent of mantic differences between languages. So the theory presented here combines. in a sense, radical universalism

4. Innate: Ccmoegrts and Language Acquisition

Acquiring language consists in large part of learning how to map or translate fram one representatiomal system @he child's prelinguistic conmpaual notitions).into another language). (Bowerman 1976: IOi]

dently of language), particularlly those which are universal (e.g, obj~ect permanence)." Bowenman quotes with approval Macnamara" (1972: 5) sltatement that "it is inconceivable that the hearing of a logicall term (by which Ihe mums wards such as "ndl', 'or', 'morey, 'all', and 'same') should generate for tihe first time the appropriate logical operator in a child's mind, Indeed the only possibility of his learning such a word would seem to 'tae if he experienced the need for it in his o w thinking and looked far it in the linguistic usage about him." m a t is particularly interesting in Boweman's (1976) discussion of the gsablem of innateness is her clear perception of the link between a child's first concepts, language universals, and semantic primitives. The view that a mntrall process in language acquisitlm is the child'g search for links between cogariltive and linguistic;concepts and linguistic f o m amdl aperations has been strengthend and encouraged by recent developments in linguistics. Many !inguists now argue, on pounds quite independent of child language, that the most basic elements o l language are not abstracl syntactic comfiguratians like grammati~ arelations, l but ralher a unilrersai set oF prime semantic concepts that combine wording to general and language-speciti~ camstraints to yield botb words and sentences. [IOIZ] The linguists to whom Bowerman refers at this point are generative semanticists, that is, representatives of a school which flourished briefly in the late sixties and early seventies but has now long oeased to exist (see e.g. It. A. Harris 1993). But the idea of a universal set of semantic primes was neither due to that school, nor linked in any way with its fate. (an the contrary: as I argued at the time (e.g. Wierzbi~ka1967a,b, 1972, 1976b), it was a lack of a strong commitment to that idea which made the positionr of the gemerative semantics school-suspended in mid-air between Chamskyam "meaning-free" syntax and genuine semantics-mntenablee


As mentioned earlier, the idea that fundamental human concepts (sementic primes) a z universal is closely l i n k -__-Ffiib 1 t h n o r i s J n p t s are&ate. It is heartening to see, therefore, !thatover the last twenty years, child language acquiition stirudies have not only increasingly viewed language learning as, above all, a quest for meaning, but have also iancreasingly assumed that the child embarks on this quest not as a passive t a b u l a rma but as an acno$oqujpped with some innate basic concepts. y o quote Boweman (11976: 112-131, ""te%iPa is now commonly viewed as coming to the language-learning task well equipped with a stock of basic concepts that he has built up through his interactions with the world . . . Some early concepts undoubtedly develop autonamously (Len indepen-

Grammar) hypothesis, Boweman (1985: 1284) writes: H argue that the BCG hypothesis does

contain a fundamental insight into early language development: that chil&en% starting semantic space is not a E ; Q ~ u S MSQ, ~ passively awaiting the imprint of the language being learned before t a h g on structure, Rather, children are conceptuallly prepared for language learning." At the same time, Bowerman (1985) argues Chat ""the initial organization of semantic spa= is not fixed but flexible", that the child's "semantic space" does not 'Veiime a single, privileged set of semantic notions that strongly attracts the grammatical fmmis o f the input", and that "one i r n p ~ ~ ~fackar ~ b n t that a n influence the meanings children adopt is the gemmfic strwfiwe of she input J'r2ngu~ge"( 1284). But there is no reason why the initial organization of the child's '%emantic spaceY%hould not be flexible in the way Bowerman describes it and yet fixed in its minimum *ore of "absoludy essential concepts"', as stipulated by Sapir. There is also no conflict between thc tenet (which 1will defend in

acquisition and the linguistically based search for innate and universal semantic primitives is perhaps best expressed by Brucrer (1990: 72): " h e case for how we "enter language' must rest upon a sePective set of prelimguistic 'readiness for meaning'. That is to say, there are certain classes of meaning to which human beings are innately tuned and for which they actively search. Prior to language, these exist in primitive fcmm as protolinguistic representations of the world whose full realization depends upon the cultmal tool of language." Given the attention that Chomsky" writings on language continue to receive in the world market of ideas, it is perhaps worth mentioning here @homskyYs mw theory on the acquisition of concepts, aocording to which most ooncepts (including, for example, 'chase', "persuade" "murder', or 'liable', and perhaps even "ureaucralt' and 'carburettor') are innate. Speaking asif the semantic complexity of most concecpts, Chomsky (1441b: 291 writes: '"arring miracles, this means that the ooncecpts must be essentially available prior to experienoe, in something like their full intricacy. Children anuse be basically acquiring labels far concepts they already have, a view advanced mosit stron~ly by Jerw Fodor."

e meanings of most words are innate rather than construed within a culture out of innate primitives, is used in Clhomsky's writings (as weill as in Fodor's; see Chapter 73, as an argument against Pexid semantics: w a d s are very difficult to define, but there is no need for linguists G o try to define them, because they are simply labels for lananalysable innate concepts. "'Ordinary dictionary definitions do not come close to characterizing the meaning of words" "hornsky 1487: 21); none the less, they '%an be sufficient for their purpose [because the basic principles of ward meaning (whatever they are) are known to the dictionary user, as they am to the language learner, independently of any instruction or experience" [ilrid.). This effecrively absolves the leiinmist from the need to study the meaning of words or to take an interest in lexicography. Even the general principles of word meaning ("'whatever they areW")re clearly too hard to study. Here ain, Chomsky's mentalism is as inimical to the study of meaning as was Bomfielld's hhawiourism.

5 . The Unjlwlerssrl Syntax of Meaning

In what has been said sa far, the emphasis was very much on the elements: the primitive concepts, the indefinable words. But to say anything meaningful we need more than words: we need sentences in which words are meaningfixlly put together. SimiQarly,to think something we need more than "conoepts": we need meaningful wmbinations of conoepts. Despite its abvious limitations, Leibniz's old metaphor off an "alphabet of human thoughts" is stilll q ~ useful k here: conceptuaf. primitives are components which have to be mmbined in certain ways to be able to express meaning. For example, the indefinable word w m t makes sense only if it is put in a certain syntactic frame, such as ""Iant ta do this'" In positing the elements I, WANT, DO, and THIS as innate and universal conceptual primitives, I am also positing certain innate and miversa! rules of syntax-not in the sense of some intuitively unvepifirable formal syntax h IQ Ghomsky, but i n tbe sense of intuitively verifiable patterns determining possible combinations, of primitive concepts. For example, the meaning of the sentence ""Iant to do this" is intuitively clear to any native speaker of English, and cannot be made any

clearly cross-linguistic semantic investigation.Vreliminaq evidence suggests, for example, that patterns such as "I want to do something", "I know this", "'Where are you?", or "I can't mow'' are universal (that is, attestable in all languages). Facts of this kind are as important for the study of the innate conceptual system (or the "prelinguisfic readiness for meaning"'; Bruner 1990: 721) as the presenoe in all languages of words for TI', 'you', 'where" 'want', 'think', or 'know'. ust as attempts to separate syntax from meaning, and to absolwtiu: sythave failed as a path to understanding how natural language wworks, it is used, and how it is acquired, so too any attempts to separate alaing from syntax and to absolutize the lexicon would lead nowhere, for tax and meaning are inextricably bound. To quote Oliver Sacks (1993: : "it is increasingb dear, from studying the natural acquisition of lanage in the child, and, equally, from the persistent failure of oomputers to derstand' language . . . that syntax cannot be separated from semantics. s precisely through the medium of beanhgs'that natmal language and atmal intelligence are built up."

6 . The Naltwal Semantic Metalanguage @SM]

I believe that the strangest support for the hypothesis of a language-like

innate conceptual system corns from Its proven merits as a working tool in the investigation of languages and cultures. As pointed out earlier, any meaningful comparison requires a tertim comparalionb, that is, a common measure. If by investigating as many diverse languages as possible we can establish a hypothetical shared core of all natural languages, we can them treat this shared core as a language-independent metalanguage for the description and comparison of all languages and cdtures. Without such a language-independent metalanguage, we would be for ever condemned to dhnocentrism, for we could only describe nother languages and cultures through the prism of our o m language (whether colloquial or technical) (see e.g. Lutz 19851. But if we can identify tlne _ _ shared c p ~ $*all e nratutalJanguagles and build ;pn-tET%a~.i~ a "natural m a n & metalanguage", we can &en des;hr%e*Th& : meanings conveyed in any language, as if from inside, while at t%e same * t h e ushg sentences from our o m language, which-if at times unidiomatic-are none the less directly intelligible to us. Ta put it differently, the shared core of all languages can be seen as a set of isomorphic mimil


l ) ^ r


languages, which can be used as language-specific vers~ versal Natural Semantic Metalanguage CI'JSM). If we try to explain the meaning of Russian or Japane, ply providing them with ard hoe English glosses (using fi we inevitably distort their meamring and impose on, the1 spctive inherent to the English language. On the other kn full-blown Englbh glosses we were to provide a doss in th , that is, in the English version of the Natural Semantic ~m%Eiij~uage, no such distortion would be necessary, for the English version of WSM can matoh exactly the Russian or the Japanese versions. For enramplle, as pointed out earlier, the Russian NSM formula ja x n c l o E u t ~ sdda"L o matches semantically the English NSM formula I want to $0 this. The idea that all languages share an identifiable core is by no means new. Wilhelrn HumboEdt; emphasized that in both lexicon and grammar, there is a ""midpoint around which all languages revolve" "903-36, v. 4: 21). Nor is it a novel idea that for semantic descriptions of different languages a special "inltermediary language" is needed-and not just an artificial system of abstract features (like the Markerewe of Katz and Fodor 8963) but a more language-like semantic metalanguage. The notion of "jazyk posrednik"", 'language-intermediary', of tlne Moscow semantic school ((seeZokowskij 19641, is particularly relevant here. What is new in the present theory is the assumption that an effective metalanpage for the description and comparison of meanings can be found in the common core d natural languages, and that it can be, so to speak, oawed out of them. Incorporating this assumption, the NSM theory oombines the phibsophical and logical tradition in the study of meaning with a typologid approach to the study of language, and with broadly based empirical crass-linguistic investigations. Unlike various artificial languages used for the representation of meaning, !the Natnrd Semantic Metalanguage, cawed out of natural language, can be understood without fwther explanations (which would necessitate the use of some other metalanguage, and so on, ad in#aifcm), and thus offms a fim basis for a genuine elucidation of meaning. AS h a Agud (1980: 457) put it in her Historia y tearia de Jos cmes, "ninguna l e n w formal pntede ser, en filltima instancia, m h precisa que d bnguaje natural que es su dtimo metailenguaje", i.e. ' k o formal language can be, in the hst instance, more precise &an the natural language which is its ultimate metalang~age".~

In priociple, data from language aquisition studies are very important to m a n t i c theony. The dilRiczahty is that to be directly relevant these stzai8ies should Ibe cuonducted within fhe hrmework o f a coherent semantic theory, and should be so devised as to test specific smantic hypotheses. lm thc past. this usuallly hasn't k e n ilmc case.

Sae also the following r m n t statement by Ham6 and GiPlet (1994:27-8): "Another important wnsaqnuenoe of the w n d cognitive revolution is the priority that must be given to ordinary languages In dekiog what am the phenomena for a scientific psychology. We will endeavor as Tar as possible to prwnl and understand oognition in Items of the ordinary languages through which we think, rather than looking ror abstract represlentalions of them. That

The need for a universally based metalanguage in human sciences has been well ilUlnslraled by the recent interdisciplinary debates on the nature of human emotions. (For detailed discussioa, see e.g. Wierzbkka 1992c, 1994h). For example, it has been repeatedly pointed out that if we try to explain key emotion terns of other languages (such as the Ilongot Iigel, or the IFaluk fago and s& by using English words and combinations of "eQove/sadlnesdco~nparison"~ or "juswords such as 'canger~passio~energy", tified anger", we are imposing ;an Anglo cultural perspctive ocn o?her c d tmes. For from an Ifaluk p i n t of view, fago is a wniiied concept; not a mixture of the concepts encoded in the Bngllisb words mger, hwe, sa&ess (8ios which Ifaluk has no equivalents). The uncritical use of culturally shaped English words (such as anger, slsorme, depression, emotiom, miad, or selA as 'kdture-free'balytical tods, and the reificalion of the concepts encoded in them, has bean strongly criticized (in my view, with goad reason) in rmmt anthropologicd literature (see e.g. Rosaldo 1980; Luh 1988; Komdo 1990; see also Wierzisicka 1993b). But to move from "dlecomstmction" to constructive rebuilding d the mietalanguage of human sciences, we need to go beyond conceptual relativism and reach for conceptual universals.

basis oEcnommwnication,and the mainstay of culture; to a large extent, they are also the vehicles by which culture is transmitted. : ]It s h u l d go witbout saying that to be able to fully understand cultures caiflerent from our own, we must be able to grasp the meaning of words encoding culture-spedfic mnmpts. For example, to understand Japanese cdtwre, andl to interpret it to wltusd autsiders, we nwd to grasp the meaning of key Japanese words such as omae, on, or wa (see Wierzbicka 119916; allso Chapter 8); and to be able to understand Malay culture, we need to be able ta g a s p the meaning of key Malay words such as maim, hahs or Iah (see Goddard 1994c, forthcoming c]. The use of the Natural Semantic Metalanguage allows us to state such meanings in a pmcise and illurninating way. It allows us t o go beyond the vicissitudes of language use and ta capture, and reveal, the semantic invariant of a word.

7. Semantic Invariants
In recent decades, semantics haw slnffered at the hands not only of its enemies but also of some of its fkiends. As I wfi1 argue in detail later (see in particular Chapter 41, especially h a d u l to its progress has been the d o e trine of 'Tamuily resemblances" and the associated attacks on the noltion of semantic invariant-a corner-stone of effective semantic analysis. One of the main tenets of this book is that words do have meanings, ;and that these meanrings can be articulated. If they haven" been successfuilly articulated in the past, for example, by the proponents of semantic "features" and "'markers"', it is not bemuse words do not have any constant meanings but because the methodollogy was inappropriate. Qf loanzrse, meanings can change, and bey may vary from one dialect, sociolect, or "generatioIQect"toanother. But semantic change as suvch is not gradual; only the spread of semantic change is. [One meaning may gradually disappear, another m y gadw1iy spread, but both meanings are determinate, and the difference between them is discrete.) In any given speech community, meanings are shared. These shared meanings constitute the
is rsrdIcaO because It resists the idea that a m e w formal EallcuJus must Ibe devised Ilo represent thought. Such calcbli U e i:at the heart oS U h e artificial iatellignnce project, the metludo0ogical princides of Chomsky and the tramsfonnational grammarians, and the assumption of formagists af aB kikjods."

Summarizing the results of the cross-linguistic investigations reported in ~emaaaerslfic afld ~ e x i c a Uxsiwersah l [Goddard and Wierzbicka 1994b1, 1 Wrote (Wierzbicka 3994b: 445): '"unting for semantic and lexical universals is not like pearl-fishing. Primitives do not present themselves glittering and unmistakable. Identifying them is an empirical endeavwr but one that calls for much interpretative effort." h this sectiom, I will briefly survey the main methodological problems arising in the pmeess d identifying universal semantic, prinnitjves and building a Natural Semantic Metalanguage. (For more detailed discussion, see Goddard 1994sr; Goddard and! Wierzbicka n 994a.31

8.1. Polysemy
Polysemy is extremely widespread in natural language, and c o r n o n everyday words--including indefinables-are particularly Ilikdy to be involved in it. A semantic primitive cannot be identified, therefore, simply by pointing to an indefinable word. Rather, it must be identGed with reference to some illustrative sentences. Far example, tine English word wanr has at least two meanings, as illustrated below:
[A) I want y w to do something. (TT] This house wants painting.

Of these two meanings only A is proposed as a semantic primitive. The NSM theory does not dab that for every semantic primitive there will be, in every language, a separate ward-as long as the absence of a

26 General lsssres
separate word For a given primnitive can R x convinrcingly explained [in a n t e r n of polysemy. The notion of diflerent principled and coherent way) E grammatical frames plays a particularly important roie in this regard. For example, If in the Australian language Yankunyltjatara (see Goddard 1494b) both the conceplts THINK and HEAR, posited here as primitives, are expressed by means af the same verb, kdini, this is mot seen as a counter-example, because [as Goddard shows) tlhesc two nzcanimgs of kuIini are associated v&h diKerent grammatical frames, and so this verb is demonstrably polysemous. Of course polysemy must never Ire posltdated lightly, but neither should its presence be denied on dogmatically a priori grounds: each case has to be examined on its merits, with reference to some generd methodologiml principles. (For detailed discussion, w Chapter 6; also Goddard 11994-2, 199'lla).

aster now, they are in complemntary distribution andl can be seen as allolexes of the same primitive, DO. This is why NSM sentences can be said to match, semantically, across languages, even though the inflexional categories in these languages differ. Far example, the Chinese NSM sentence adapted from Chappell i(E9a41: 138)
Chli-shl hdu, w E S shllio-]la xiE shknme happen after I say-pa;v c n . something 'After this happened, I said something.' can be matched with the English NSM sentence: After this happened, I said something. even though the English word for HAPPEN, in contrast to the Chinese one, is marked for past tense: when combined with after, the form happened can be seen as an aUolex of HAPPEN, on a par with happen.

If one word (or morpheme) can be associated with two diflerent meanings, one meaning can often have two on. more different lexical exponents. By analogy with ccd~omorphs)' and "allophones",such different exponents of the same primitive are calkd in NSM theory "dlolexes". To start with some relatively trivial examples, in EngIish, I and me are allollexes of the same primitive concept (in Latin, EGO, in Russian, JA). Often, the allolexes d a primitive are in cornpleaentary distribution; for example, in Latin the three f o m s hic, koec, hoc are all exponents of the same primitive THIS, and the choice between them depends on the gender of the head noun. Omen, the cornbinatican with anloither primitjve form5 the choice of one of a set of allolexes. FOPexample, in English, a combination of the primitives SOMEONE and ALL is realized as everyone or everybody3and a corn--\ bination of ALL with SOMETHING is realized as ewerytfirhg. In tlnesne particular loontexts, -one and -body can be seen as allaiexes of SOMEONE, on a par with someone; and -thing can 'be seen as an allolex of SOMETHING, on a pas with somethhg. The notion d ailolexy plays a particularly important role in the NSM approach to inflexional categories (first articulated by Cliff Goddard at the 1992 Semantics Symposium held in Canberra). Consider, for example, the following sentences:

8.3. Obligatory or Semi-obligaltary Portmanteaus

The notion of alolexy is ciosely linked with that of semantic portmanteaus, which I will illustrate w i t h a simple example from Russian. The expression like thiss c o r n o n E n both everyday English and in English NSM sentenoes, is normally rendered in Russian by means of the word tak, which expresses n lcombination off the two primitives LIKE and THIS. Ja sdelal &to tak I did this like-this Since, however, Russian does have separate exponents for both LIKE and THIS (kt& and Or@), the use of an obligatory, or semi-obligatory, portmanteau for their combination does not present a problem for the NSM theory. It would present a problem if the postulated primitives did not have their own expnents usable in other contexts.

8.4. Valenq Options

f i e notion of valency options (developed in Chapter 3) refers to different combinability patterns available to the same primitive. For example, the primitive DO can occur in the fol!owing combinations:

(A) P am doing it now. (B) I did it before now (earlier). (C) T will do It after now (later). By themse19ves, the forms am doing, did and will 40 oonvey different meanings, but when combined with the temporal adjuncts now, befire nnow, and

(A) X &dl samething* (B) X did something to penon 3Y. (C) X did something with thing 2.
Obviously, "doing something to someone", or "doing something with something" implies ""doing something". None the less, sentences B and C

cannot be analysed in terms of A and something else. It has L o be recognized, therefore, tlval in each case the difference in meaning is due to the sentence as a whole, not Ito the predicate as such, and that the three sentences share in fact the m e predicate (DO), albeit they realize different valency options of this predicate.

The second relationship is reflected in the colloquial phrase "one and the same", and in the apparent paraphrase relation between sentences such as A and B below:

(A) These two shoes belong to one pair. = (B) Thewe two shoes Mow%to thr: same pair.
But dose as the elements within each pair may be, neither THE SAME and LIKE nor THE SAME and ONE cam be identified or defined in terns of each other. For example, in the sentence

8.5. Non-compositianal Relationships

Semantic primitives are, by definition, indefinable: they are Leibniz" uultimate "simplesy" histotle% ""piora", in terms of which all the mmplex meanings can be articuliated, but which cannot be deoomposed themselves. They can, of course, be represented as bundles of some artificial features, such as "+ Speaker, - Hesrrer" far 'I" but this is not the kind of decomposition which leads from complex to simple and from obscure to clear. As pointed out earlier, the meaning of a sentenoe like "1 h o w this" cannot be clarified by any further deccnmpositbn-not even by decomposition into some other mamingful sentences; and "featlares"",hich have no syntax and which are not part of n a h d language, have no meaning at all: they have to be assigned meaning by sentences in natural languages, rather than the other way around. This m a n s that, from a compositional point of view, elements such as 'Hbnd "ouhre semantically simple and have no identifiable part in comman. At the same time, intuitively, these two elements are cleady related. Their relationship, however, 3s non-compositiond. A semantic system is not like a bag full of marbles, each of them perfectly round, self-contained, and independent of the others. Rather, it is a sysltem 'khtout se tient", to invoke (in a new context) Saussure's famous, phrase. IPP this system, there are elements which "be1lon;g together" and which have the same combinatorial properties, such as 'I' and 'you', or 'good' and 'bad" Elements of this kind are intuitively related, but this doesn" mean, that one of them can be defined in terms of the other. E n the universal semantic system there are many different kinds off moncompositimal relationsbips. For example, the elements I, YOU, THIS, HERE, and NOW, are all muiluaily related, although they do not all have the same combinatadall properties. We can acknowledge this relationship by putting on them all one label, "&eicticic",but doing this-while udulhas nothing G o do w i t h semantic decomposition. The primitive THE SAME has a non-compositional relationship with the primitive LIKE, and also with the primitive ONE. The first is highlighted in sentenoes such as the following one: This fish is like that other fish, but it is not the same fish.

I have one son and two daughters.

h e ' has clearly nothing to do with U e same" and in the sentence: They came at the same time. 'the same' has nothing to do with 'like'. Mon-compositional semantic relations of different kinds ante real and hportamt, and they oEer an interesting geld for remarch (we Goddard and Wiembicka 1994.01.But they must not be conhsed with composiltional relations, which mn be revealed by definiltions (such as, for example, that between a8Ieep and awake, or between dead and alive].

8.6. Recurrent Polysemies

Non-compositionall semantic relations are often reflected in recurring polywmic patterns involving two, or more, different primitives. Of course, a n 0 natural language will ever be found in which uhe word for 'I' will be the same as the word for 'you', or the word For 'big" the same as the word far 'smalll': since tlhe combinatorial possibilities of both elements within each pair are the m e , polysemy of their exponents woulld lead to intolerable confusion. Other nom-compositional mlations, however, are often reflected in recurring polysemic patterns. For example, in some languages the word far THE SAME is lthe same as the word for ONE, icar the word for THIS is the same as the word for HERE; there are also languages in which the word for WANT is the same as the word for SAY, or where the word for DO is the same as the word far HAPPEN. This doesn? team, however, that in those languages people do not distinguish the concept ONE from the concept THE SAME, or the concept WANT from the concept SAY; or that they have no words to express some of these concepts, They do have words for all of them, and if some of these words aane polysemous (and mean, for exmple, f l) 'one" (21 'the same', or (1) 'want', (2) 'say'), the different meanings of such polysemous words can be easily distinguished on the basis of distinct grammatical frames assodated with each of them. (For examples and discussion, see Wierzbicka 19946).

8.7. Resonance
Since every language embodies a unique semantic system and reflects a unique culture, the exponents of universal semantic primitives in different languages often "feel" different (to both native speakers and to linguistic experts on these languages). For example, it is easy to believe that in the Papuan language Kalam, where the words for KNOW, THINK, SEE, and HEAR all share the same verball formative ng (Pawley 19941, these words "feel'9ornehow different in meaning from the corresponding English words [which are formally mrelaited to each other). Or i f the word for FEEL is polysemous between 'feelhand %tlomach"as is the case with the word o'mi in the Australian language Yankunytjatjara, see Goddard 1994b), it is easy to believe that this ward "feelsYVifferentfrom the English word feel, or from the Acehese word rasa (a borrowing from Sansktit; Durie et al. 1994). Elifferenoes of this kind are real! and important, and they are acknowledged in ]theNSM notion of "resonance" (first arhculated 'by Goddard at the 1992 Semantics Symposium in Canberm). They must not be confused, however, with semantic differences semsrr strictol.
8.8. Canonical Sentences

some which are not mmposed exclusively of pdmitives. For example, if we want to check whether a language has words for the primitives QlNE and TWO, it is practical to use sentences like the fchlIowing:

I have two sons and one daughter.

and 'daughterhre not universal, and the even though the conwpts of %ony words glossed as 'sonhand 'daughterkay not match semantically across language boundaries (for s m e languages may distinguish a man's son or daughter from a woman's son or daughter). The notion of a canonical sentence both in the strict sense Cppimitives only) and in the broader sense (primitives with a controlled admixture of non-primitives) has proved to be a valuable tool in cross-linguistic semantic research (see Goddard and Wiembicka 19946131. In the future, this notion may also prove useful in the cross-cultural study of language acquisition and cognitive development; and may answer, in some measure, the call frequently voiced by child language researchers "for a mare powerful crosslinguistic methadology" (Johnston 1985: 496).

4. Past, Present, and Future cpf NSM Semantic Theory

Most sentences uttered in any one language cannot be translated into other languages without some loss, and/or addition, of meaning. The NSM theory hypothesizes, however, that there are ako some kinds of sentence which can be translated-without loss andfor addition of meaning-into any language whatsoever. These are sentences fornulatied in "local representatives'kf universal semantic primitives, according to the universal syntactic rules (that is, rules for combining the primitives). Sentences of this kind include, for example, the following ones: You did something bad. I know when it happened. 1 want to see this. These people didn't say anything about this. If yoru do this, P will do the same. This person can" move. Sentences of this kind are regarded in NSM research as "canonical sentences", which can be used to test the validity of the Natural Semantic Metalanguage (as developed until now), and to seek any weak points which may need revision. For practical reasons (Lo make the tlesting more effective in working with native speakers) it is often useful to include in the set of canonical sentences Since its inception imn the mid-sixties, the basic assumptions and goals of the NSM theory have remained unchanged: the search for universal! semantic primitives, the avoidanoe of artificial "features" m d "markers'" the rejection of logical systems of representation, the reliance on anatural language as the only self-exp1anatory system for the representation of meaning. At the sarme time, the theory has not stood still; on the contrary, it has been constantly developing. These devetopments could be said to have gone in six main directions: the proposed set of primitives has considerably increased; the search for primitives came to be identified with a search far Ilexicall universals; 3. the search for l e x i d primitives came to be combined with a search far universal syntactic patterns (that is, for universallly available combinations of primitilres); 4. the pursuit of, first, primitives and then their combinations grew into a broader programme of building a ful-scde "natural semantic metalanguage"; 5. the theoretical underpinnings of the whole enterprise became g r d u ally m r e and more clearly arlticulated (as discussed im Section 83; and 6. the range of domains, languages, and cultures to which NSM theory was applied, and against which it was tested, expanded substantially.

These developments cannot be discussed here comprehensively; a few brief comments on each of them, however, are in order. 1. NSM theory started as a search for lexically embodied indefinable concepts, or semantic primes, identified as such by trial and error, within one language [ m ylanguage)^. The fiat tentative Pist of primitives identified in 1972. It in this search was published in my book Semamric P~imSfive.~ included fourteen elements. As the proposed primitives were tested against m increasing range of semantic domains, most of them [on present count, deven of the fourteen) proved themselves effwtive tools in semantic analysis. But at the same time it k c m e increasingly clear that the minimal set of fourteen was insufficient. (See Wierzbicka 119846.) A major impulse for their expansion was the Semantic Workshop held in Adelaide in 19815, and organized by Cliff G o d b r d and David Wilkins, where Goddard proposed a number of new primitives for further investigation. (See Goddard 1986a, I989a.) As the consmtive expanded sets were tested in semantic analysis, the process repeated itself, and expansion contimed. (For the current head count, see Chapter 2.) The prowss of e x p s i o n greaitlly facilitated semantic analysis of numerous semantic domains and made it possible to formulate semantic explications that were much more readable and intuitively intelligible than those based on earlier, leaner sets. The theoretical "cost" of this expansion lay in the need to abandon the kibnizian principle of mutual independence d primitives. E n the early versions of the NSM theory, if the dements appeared to be semantically related (as, for example, 'good'and "ant', or 'the wme'and 'other?, it was assumed that at least one of them must be semantically complex (on the grounds that if two elements share a common part they must have parts, and therefore cannot be mantically simple). This assumption was never strictly adhered to, however. For example, I, YOU, and SOMEONE were regarded as primes from the outset, even though they are intuitively related (for every "I", and every "you'" is a "someone"pl. in h e , the assumption of mutual independence of primitives was rejected altogether, and it was recognbed that primitives can be intuitively related (as " I s b a d ""smeon~e" are), without being composttkonallly related and without being decomposable [that Is, definable). 2. The first proposed primitives were identified, by trial and error, on the basis of a handful of European languages. With t h e , through the work of experts on many diverse languages, the empirid basis grew consider;ab11y, including, among others, languages as diverse as Chinese [see e.g. Chappel1 1985, 1986a,ib), Ewe [Ameka 19816, 11987, 1990, 19911, lapan~ese('Travis 1992; Hasada 1W4), N d a y [Goddard 1994~1, the Austronesian language Mangap-Mbwla (Bugenhagen 199011, ar the Australian languages Yankunytjatjara (Goddard 1990, 19920,b) and Arrernte (Wilkins 19815;

Harkhs 19921. This expansion culminated in Semantic and Lexfcal UniversaJrs (Goddard and Wierzbicka 1994b1, mentiocred earlier. A priori, one might have expected that the process of testing a hypothetical set of primitives across a wider range d languages would bad to a redu~tianof the proposed set (as one proposed primitive after mother would fail to show up in this or that language). On the whole, however, this has not h a p p e d . On the contrary, the list of primitives has shown a tendency towards gradual expansion. 3. For a long time, research into the syntax of the proposed primitives lagged behind that into the primitives themsellves-a point commented on by several reviewers I(e.g, McCawley 1983). This delay, though unfortunate, was dictated by the nature of things: one can hardly investigate the patterns of combination of primitives before one has some idea of what the primitives are. The first articie devoted primarily to the syntax of the primitives was my "'Lexical Universals and Universals of G r m a r " "ierzbicka 1991~).The Symposium on the Universal Syntax of Meaning held im Canberra in July I994 (organized by Goddard and myself) launched a major programme of research in this area across a number of languages. 4. The building of the Natural Semantic Metalanguage was, and continues to be, a gradual process. In contrast to more spculative semantic theories, NSM constantly swks cmfirmation-or disconhation-in large-scale descriptive projects. For example, in my E'ng/fsh Speech Act Verbs ~(Wiereszbicka1987aS I attempted t o analyse ithe meaning of more than 200 English verbs; and more recently, in a series .sf articles rn another conceptual domain (see e.g. Wierzbicka 1990c, 199k, 39414~)I have similarly sought to analyse at Ieaslt 1 0 1 0 : English emotion terns. It is through descriptive projects of this kind that the inadequacies [as well as the strengths) @If successive versions of NSM b e m e apparent, irnd that future directions of development could be seen more clearly. Perhaps the mast important direction had to do with the growing simplification and standardization of the syntax of explications, linked directly with the search for universal syntactic patterns. 5. The theoretical underpinnings of NSN research were gradually wtic lulated more clearly, and its methodology fomulated more explicitly, as important theoretical concepts like "polysemy", 'klEolexy", "valency option", '%on-compositianal relationship", and 'kreonance"were gradually clarified and more rigorously articulatd (we Smtian 8; all= Goddard 19941; Goddard and Wiembicka 11994b). The Symposium on Semantic and Lexi~alUniversals held in Canberra in February 1992 and organized by CliB Goddard and myself played an important role in this regard. 6. Over the years, the range of domains to which NSM research addressed itself has continued to expand, including mot only l e x i d semantics (as in, for example, Goddard 1990, 199la; Travis 1992; Hasada 1994;

34 General issues Ameka 1990; Wierzbicka 1985, f987a1, but aiso the semantics of grammar (e.g. Pumeka 19910; Chappell 1986(r,b, 1991; Wieinbicka 1988) and pragmatics [e.g. b e k a 1987; Coddad 1986b; Harkins 1986; Wierzbicka 199la; Willkins 1986). Furthemore, this resear~hhas expmded into more dirmt comparison of cultures, via their lexicon, grammar, conversational routines, and discourse structure ((PI.%.h e k a 1987; Goddard 19926, forthcoming c; Harkins 1994; Wierzbicka B991a, 11492a; Wilkins 1992). Most recently, NSM r e m c h has moved into yet another direction, leading to the devePopmemt of a '"heory of cultural scripts:', which offers a framework for comparing cultural moms operating in different. cultures, a framework based on universal semantic primitives and universal syntactic patterns [e.g. Wierzbicka 11993e, 1994Ez,d,e, forithcoming c; Goddard forthcoming b; Goddud and Wierzbicka forthcoming). But while all these developments are (as it seems to those involved) significant, NSM theory still1 has a long way to go. The pursuit of semantic primitives needs tca be fiIYa~kd, the study of the syntax of primitives, n d s to be more. fully developed, the smpe of cross-llinguistic testing of both primitives and thdr syntax needs to be substantially widened, language-specific versions of the Na,tuaal Sernmtic Metalanguage need to be buiit, the NSM-based analysis of culture and cognition needs to be exltended to new areas, h e theory of cultural scripts needs to be further fleshed out, and so on. This book therefore constitutes an open invitation.

A Survey of Semantic Primitives


The set of primitives presented and discussed in this chapter has evdved in the course of nearly three decades of research by myself and collleaguesand it is still evolving. Some of the primitives proposed here are better established than orhen. Of the fourteen primitives posited in my Semmbic Primitives (1972) ten have survived nearly a quarter of a, century of critical assaults [by myself and others), and (with one exception: PART] the position of these original members of the set can be regarded as particularly strong. This old guard includes the "substantives" I, YOU, SOMEONE, and SOMETHING, tlne ""mental predicates" THINK, WANT, FEEL, and SAY, and the demonstrative THIS. But the main divide mns between ~thohclse elements which were tested across a wide range of languages in the project reported in Sem:mtic and Lexkal UnsivermEs (Goddard and Wierzbicka 1994b), and those which were not included in that project, and which must, therefore, be regarded as less well established. Accordingly, the present chapter, surveying the primitives, is divided into two parts, called, for convenience' sake, "Old Primitivesy' and "Mew Primitives". The set of old primitives includes the following elements:

I, YOU, SOMEONE, SOMETHING, PEOPLE crdete~~r~y' THIS, THE SANE, OTHER cGq2tantifiersy' ONE, TWO, MANY (MUCH), ALL ccmentalpredicatesy' THINK, KNOW, WANT, FEEL SAY "actiions and eventsp' DO, HAPPEN 6cevaI~atorsyy GOOD, BAD 6cdescriptors'y BIG, SMALL 4ct~e" WHEN, BEFORE, AFTER 4cspacepy WHERE, UNDER, ABOVE "partonomy and taxonomy" PART (OF), KIND (On

ccmetapredicates" "interclausal linkers'"



2.1. I and YOU Joseph Bradsky's (19941 poignant recent poem entitled "Infinitive" 'arts as fo81ows: D c a savages, though I've never mastered your tongue, Free of pronouns and gerunds, li've learned to bake mackerel wrapped in palm leaves aad Favor raw turtle bgs, w i l h their flavor or sllovmess. For talc poet, the factual inaccuracy of the image does not matter, but the b~l! of the matiler is tbaE there. are no tongues in the world (no matter how ""svagem")hich would be 'Tree of pronouns and gerunds'" O.F gerunds, yes, but of pronouns, no. E n particular, there are no languaps in the world which would be "Tree" of words for ' I or YOU (in the sense THOU). This is not to say that claims have never been made-not only in poetry, but alss in s~holarlly lileraitme-that languages "free of personal pronoums" do exist, but notions of this kind have never been substantiated and they must 'be regarded as fanciful. (See Wierzbicka 119941b.)1 It is true that many languages, especially South-east Asian languages, have developed a nlwnlmber of ciaborate substitutes for " y o u ' b d ''1'"' and tlhalt irta many circumstances it is more appropriate to use some such substitute than the barest, the most basic, pronoun. For exampie, in a polite conversation in Thai, the use of the basic words for ""yom" and "I"' would sound crude and inappropriate. Instead, variouls self-deprecating expmssions w u l d be used for ' T b d various deferential expressions for 6cyouyy. Many o f the expressions which stand for refer to the speaker" hair, crown of the head, top of tlne head, and the Like, and many of the expressions which stand for "you" refer to the addressee's feet, soles of the k t , or even to the dust underneath his feet, the idea being that the speaker is putting the most valued m d mspcted part of his own body, the head, at the same level as the lowest, the lmslt horuowable part of the addressee's body (see Cook 1968). Hut this does not mean that Thai bas no personal pronouns, no basic words for ""you" and "Iyy.
Maw& (1943) claim to have shown that the concept 'I' has no lexica2 exponents i m Winlzn, Kaei, and Japanese; far detailed aefuaeth d these claims see Goddaad (1995).

A Imgmge may not make a disthcttioa which would corr.-,spond to that between the words "he" and "she", and in fact many languages, for example Turkish, have just one word for '%heyhand "she"', undifferentiated for sex. But no known language fails to make a distinction between the speaker and the addressee, i.e. between "yaw" and '7I"'. This does not mean that the range of use of the words for '""yo"arud is the s m e , in d l languages. F , Thai-English dictionaries gloss For example, in Thai, the word C ~ ~ Fwhich has a range of use incomparably more narrow than a n t i s English as "Iy'", equivaIent. When used by women, it is restricted to intimates, and it signals a high degree of informality and closeness; when used by men, it signals superiority, rudeness, disrespect (Treerat 1986; Cook 11968). But since there ane no invariant semantic components which could be always attributed to d 6 n , other than "E"" the heavy restrictions on its use must be attributed to cultural rather than semanltic factors. In a society where references to oneself are in many situations expecbd to be aocompanied by expressions of h u d i l t y or inferiority, a bare ""I' becomes pragmatilcalBy marked, and it must be interpreted as either very intimate or very rude. But this pragmatic markedness should not be confused with demonstrable semantic complexity (see Diller 1994). The universality of I and YOU (brilliantly guessed by Wilhelm HumboPdt, in the early days af typdlogical linguistic investigations, and reasserted by Boas (1911; see also lngram 1978)) taillies well with the indefinable nature of these two concepts: while attempts to define them (e.g, in, terns of "speaker" m d "addressee"'; see e.g. Reichenbach 1948: 81 1 3 have often been undertaken, these attempts have never h e n successful. Words such as "~peaker"and "addressee'hase neither universal nor semantically simple. Roughrly speaking, '"he spcakcr (of some w o r d s ) ' ~ s""te person who says these words"; and "the addressee (of some words)" is "the person to whom these words are said". Furthermore, If ""H"oesnYt mean '"he speaker", ""thespeaker" doesn't mean "'I"". For example, if I whisper to the person nexk to me "1 don? like the speaker", I mean neither that "The speaker doesn? like the speakery', nor that ""n:onytlike myself"". Similarly, if someone asks me ""Who are you speaking to?'hnd I reply "I'm speaking to you'" 1 can hardly m a n that 'cIym speaking to the person to whom I am speaking'" or that "The speaker is spaking to the person to whom the speaker is speaking" (see Setensen 1963: 961. Finally* the idea of "I" is not necessarily tied to speech: we rely on the concept of "'I" in our thoughts, as well as in our speech. For example, if I think to myself '" want to do something ( X i ) today", I do not think of myself as "the speaker""ut simply as 'TI"'. F r m this point of view, Russell's attempted definitions of ' l ' 1["I-the person experiencing this' (1964: 851, and 'I"-the biography to which this

2. A Survey o f Semrsnr tic Primitives 39 belongs'l965: 107) ) axe perhaps more plausible than Reichenba~h's(since they do not refer to speech), but on ittire whole they are hardly convincing either: whether uttered aloud or thought silently, the sentence "I want to go now" can hardly mean 'the peaon experiencing this wants to go now', or 'the person to whose biography this belongs wants to go now" As pointed out by Ssrensen (1963: 961," 'I' and "ouhre sigtvs i n the first and the second person respectively . . . Now, whatever the difference between fist, second, and third person signs may be, there i s a dineren%, a difference of meaning . . . Therefore, a sign in the third person cannot be semantically identical with a sign in the first or second person". (See also Castafieda 1988.) Of course "Iyy m sometimes be intended as referentially identical to a "third person sign'" such as, for axamplc, "tire author o f ihese lines", ,or as Ithe expression "this person" accompanied by a self-directed gesture; but cllearly, neither '"he author of these lines" nor "this person" mean the same as 'I,, and even &om a referentid paint d view, expressions such as "the author of' these limes" or "this person" are not always equivatent to '1'2.2. SOMEONE and SOMETHING

I know this about someone: this someone (this penon) did It. I know this about something: you see this something (this thing). Vhe expr~essims this someone and this s~mebhiq, which sound awkward in English, will be discussed in Chapter 3.) In linguistic literature, the distinction between SOMEONE and SOMETHING, which plays an important role in the grammar af many languages, is often represented in terms of binary features, such as + HUMAM, or + ANIMATE, or +- PERSONAL. But amaunts of this kind are a good example of pseudo-analysis, since sthe features which are invented to awount for the dinerence between SOMEONE and SOMETHING themselves need to be d e b e d (or explained) in tems of SOMEONE and SOMETHING. For example, the sentences I met someone nice. I saw something interesting. can hardly be paraphrased (except in jest) along the following lines: I met a nice human [animate, personal), thing [entity). I saw an interesting nun-human (inanimate, impersonal) entity. As I argued in L h p a MemtrrIis fifteen years ago, to substitute ""ientityYVor mrneone and something is to avoid the categorization embedded in all natural languages and to try to replace it with an artificial device alien to them. The disltlncticcn Ihertween "persms" and ""non-personsyy is quite f u n h e n k;al to human conceptualization of the world. Natural languages diner in this respect sharply from artificial languages relying on the abstract notion of "referential indices"'. Linguists who have assumed that the language of symbolic logic is a suitable tool for analysing naturas language have somerimes taken for granted that the notion of "semantic prime" c m be identified with the notion of "atomic predimte", because what logicians describe as "arguments" can be simply thought of as indices. (This applies, in particular, to '"enerative semanticists"'; see e.g. McCawley 3973: 334.) But natural languages don't work like that. The distinction between SOMEONE and SOMETHING is basic for them and cannat be reduced to any differa c e between predicates. To put it differenltily, the concept of SOMEONE (a 'person" is essential P P the differences in to human conmptuaEization of the world, and despite a d t u r a l context and cultural interpretation (see e.g. Shweder and Bourne 1984), it bas a s;tab;le, irredwible loore across all languages and ~uEtwe8 (see Spiro 1993;Wierzbicka 1993b); no language and no culture blurs the fundamental divide between SOMEONE and SOMETHING.

A11 languages have words for WHO and WHAT, and can distinguish lexically between the questions "What is this?" and "'Who is this?" The distinction beltween %ho'and "hat', 'soameone'and 'something', ' p r s o n k n d 'thingbrovides the most fundamental f o m d human categorization (for while YOU and I are also fundamenltal to human thimking,,they do not a t egorize the contents of the worlld)~. It is hpclssible to define 'someone' or 'something' in m y simpler terns. In English, the apparent morphoiogical complexity of the wards sameone and something may suggest the idea that these words are in fact demmposabHe (into 'some' + 'one" and kame' + '"Itinfl").But af mwse someone doesn't mean the same as "ome on& and something doesn't mean the same as %onrething'. When the words who and what am used in questions, the mnoelpts 'someone' and 'somethinghare combined with an interrogative meaning (" wmt to know something', 'I want yau to say something')),but this interrogative meaning is not an inherent part of the words W ~ and Q what as such. For example, when used in so-called "embedded questions",

know (don't h o w ) who did it. I know (don't know) what you see.

who and who$ do not express an interrogative meaning at all. Rather, senternaces of this kind can be interpreted as folllows:

40 Geflerd Iswes

As the evidence reported in Semantic and Lexical Umi~ersa!s (GOadard and Wierzbicka 1994b) illustrartres, all languages appear to distinguish, in one way or another, between a m r e general notion of SOMEONE, or BEING (hurnsm or non-human), and a notion of PEOPLE (necessarily human). For example, someone (and who) can refer to God: Who created the world?--God. God is someone inhitely g o d and merciful. But the word peapk (whi~his inherently "plura1"")amot refer to any group of beings other than human being (not even to the very human-like gods of the Greek Olppus). The well-known fact that i n many languages, the word for cpeople',is also used as a tribal name is (as pointed out by Greenberg 191616a: 261 dearly a case of polysemy, mnzlparable to the polysemy of the English ward man (I] a male human being, (2) a human being, and not evidence that a word for 'pipeople' may be missing. It is a striking fact that in many languages the word for people has a different stem from the word for an individual human being, and doesn't look like an ordinary plural. The English word people is a good case in paint. Similarly, in Geman, French, and Russian the words Leufe, gens, and @udi, 'people', are all different from the words for an individual human being (Mensch, hornme, and Eelavek, respectively). By itself, this formal difference doesn't prove anything about the semantic relationship, but it is certainly highly suggestive, and it tallies well with the semantic fact that it is impssible to define people in terns of someone and something else (or i n d d in any other way]. On the other hand, if we aocept that both someone and peopk are irreducible semantic primitives, numerous other concepts can be explicated via these two. For example, every language Eras a large number of words referring to "cultural kindsya (see Lyons 1981), \that is, to human artefacts, such as, for example, q , mtrg, bottle, boomerang, chair, and so on. A11 these words make references (in their semantic skmctureie)to people, because they designate objects " m d e by people"', 'bused by people'" and physically defined with reference to the human body. (For example, cups are made by people, for people to drink from; they are made in such a way tErat people can hold them in one hand, and so on. For discussion, see Wierzbicka 1985.) In addition to names of material artefacts, there are also numerous words referring to social life (e.g. society, tribe, fami!y, committee, and so on), to human emotions Qe.g. sh~me, embar~il~ssmnt', pride), to language (e.g. ha$wage, d i ~ k c lshag), , and so on, which refer in their meaning to 'people'.

The status of PEOPLE as a fundamental element of human thought is reflected in various ways in the grammar of innumerable lmguages. For example, in Palish normind declension, masculine nouns referring to p o pie have an accusative plural identical in form with the genitive, whereas nouns referring to things and animals have ;an accusative plural identical in form with the nominative. QCf. also the so-called "hierarchy of agentivityz' discussed, for exahnple, by Silverstein 1976, Dixon 1979, Comrie 1989, Mallinson and Blake 11981, and Wierzbicka 1981, in which the category "human" plays a prominent role.) of PEOPLE is a conceptual primitive is The hypothesis that the ~onoept consistent with the results of reoent studies of language acquisition. For example, as pointed out by McShane (1991: 197), Carey (1985) 'Tomnd t h t children initially organize biological knowledge amund hulrnruns as a prototype. Inferences about biological properties of other species are based both on what children believe about humans and haw similar the other species is to humans."JackendoK (1992) talks in this connection abauk "a faculty of social cognition"'. Referring to Kaitz, Baker, and Mamamara's (1974) h d ing that children as young as 17 months know that proper names can be applied to people and people-like objects such as dolls but not to inmimate objects such as boxes, he c o m m t s : "That is, ~ e seem y predisposed to make a cognitive distinction between persons and everything else-the distinction 1 anol claiming is pertinent to social cognition-and they are predisposed as well to find a linguistic distinction that encodes this difference." The evidence from cross-linguistic semantic investigations points in the same direction. Finally, the hypothesis that the concept of PEOPLE is, in all probability, a semantic universal and a conceptual primitive offers a solution to the old and apparently insoPub1e problem of how the notion of % m a n beingv can be defined. Is a human being a 'featherless biped', as the cynics maintained in ancient Greece? Or is it a 'rational animal', as medieval philosophers wed to daim? Or is it perhaps, as the French writer Vercors (1956) once maintained, 'a being endowed with a religious sense" "rl these and other definitions are dearly deficient, and it is a relief to be able to go back to Pascal's (1667/1963: 579) view that the notion is basic and that all attempts to define it must fail. r corLinguistic evidence suggests, however, that Pascal's view requires ; rection. It is not the notion of an individual human being, l % a m e ,which a p p r s to be universal and indefinable, but the notion of PEOPLE, a social, rather than biological, category. Given the universal presence of the concepts PEOPLE, SOMECINE, and ONE, the notion of an individual human being does not need to be regarded as grimiltive. But it is impossible to d e h e both 'hwman being'ami 'people" and cross-linguisticevidenoe suggests that it is the latter, not the former, conoept which is indeed wiversal (see Goddani and Wierzbicka 1994l1)~

42 General Issues If we think of universal semantic primitives as innalte concepts, the idea that a social category of PEOPLE may be innate is unexpected, and it wrtainly gives food for thought. If we are "ational animals' (with the notions of THINK and KNOW being part of our genetic endowment), we are also 'social animals', so much so that the idea of PEOPLE as a smiat category is also a part of this endowment, In fact, of course, according to the innate 9 9 0 ) ,we are not 'animals'aat all: we and universal folk model (see Bmner 1 are PEOPLE, evcry single one of wlaulm is also SOMEONE and 7'-all irreducible and apparently universal human conmplts. alent implies 'vidbiEty'as well as Yhisness'is refuted by Crowley's (114778: 721 authoritative study of Bandjalang, which makes it clear that the socalled '"isible" demonstrative gaya is i n fact unmarked, and that the closest Bandjaimg equivalent of this doesn't mean 'this, which I can see" but simply 'this'. What is controversial here is not the statement that the Eidabal word gay0 has a range of use somewhat different from that of the English word fJsSs, but llsc claim that this dillicrcncc in usr: is due to a sp~ifiabhc: semanrtic difference: "isyin English] versus 'this, whilch I can see"[in Gidaball). Differenoes in the range of use can sometimes be explained in tems of facton other than <thesemantic. But the presence of zu specijiabb semantic. difference could not be reconciled with llnc claim that two lexical items have the same meaning. Experience shows, however, that reports concerning alleged semanltic differences cannot be accepted at face value. 3.2, THE SAME and OTHER The 'Veteminers'THE SAME and OTHER appear to be used, universally, in sentences such as the following ones: It happened in the same place, not in another place. It happened at the same time, not at another time. She did (saidr'thoughtlwanted] the s a m . I saw it, and two other people saw it. It was not the same fish, but it was the same kind of fish. (English sentences such as 'Give me another beer!" illustrate a languagespecific, not a universal, use of the word other.) The element THE SAME carresponds to a fundamenltal Pogicall relatian of "identity", which occupies a prominent place i n the philosophical and logical literature on thought, knowledge, and logical relations in general. OTHER may seem to be no more than a negated version of identity (as I argued myself in Wierzbicka 1989c), but in fact a phrase such as "I and two other people" cannot be reduced to 'the same9and negation. Both concepts, THE SAME and OTHER, are lexical universals, and they both play a considerable role in the lexicon and in grammar. For examplel they are needed to account for the meaning of many wonj,unctions and particles, such as also, loo, or otherwise (see Eoddard 1986b; Wieazlbicka 1 9 8 6 ~ 1991~1, ~ as well as for that of grammatical 'keferenceWacKmg" "vices, such as, for example, "switch referen~e". Exponents of both THE SAME and OTHER are often involved in comrnon polysemic patterns, in particular, one linking THE SAME with ONE, and another, linking OTHER with SOMEONE (see Goddard and Wierzbiicka 19946); in each case, however, the polysemic knots can be

3. Determilversr TWLS, THE SAME, OTHER

3.1. THIS The word this and its counterparts in other languages provide a basic means for identifying what we are talking about; and in any language, as in English, one o z w point to an object or a person and say something that means 'this" "this thing" or 'this person'* It is impossible to define the concept of THIS in t e m s of any simpler concepts; and while technical labels such as "deictic" or "demonstrative" are sometimes mistaken for a statement of meaning, presumably nobody vvould argue that, for example, #hisdog means Veictic d o g k r Yemonstrative dog'. What linguists do sometimes argue is that rhix means bear the speaker'. But this is an iQPusion, too-first of all, because near ((innon-metaphorical use) always refers to spatial relations, whereas u tF S $ is not restricted to space (compare e.g. fuFSi8 day, or 5h& song], and second, because if 1 point to one of my own toeth m d say "'This tooth hurts", this can hardly be interpreted as "itlire tooth near the speaker" (see Fillmore 1975a). Despite occasional claims to the contrary, careful examination of the available cross-linguistic evidence suggests that all1 languages have a clear and unproblematic exponent for THIS. The other demonstrative pronouns often do not match semantically across language boundaries [for example, mna, and am, neither of the three Japanese demonstrative pronouns kl0~10, sono nor a m corsesplclnds exactly to English that, but this can be matched semantically with kona]. In particular, it is not the case that (as has sometimes been claimed] some languages lexidly distinguish "his, wwhh I cazr see' from 'this, which I cannot see', without having a basic, unmarked term for THIS. For example, Cecil Brow's (1985: 287) suggestion that the Eidabal dialect of the Australian language Bandjallang doesn't have an exact semamtic equivalent for the English word skis because the nearest Gidabal equiv-

2. A Survey of5"enaantic Psimithe8 45

disentangl~ed and the distinctness of the two concepts in a given language can be upheld.

4. Quantifiers: ONE,TWO,M A W (MUCH), ALL

4.1. ONE, TWO, MANY (MUCH) Fantasizing about his "UP" (Universal People), Donatd Drown (1991: 139) writes: "UP language wntains both proper nouns and pronouns. The 1stter include at least three persons and two categories of number. Their language contains numerals, though they may be as few as 'one, two, and many'." Empirical evidenm surveyed in Semsrntic and Lexical Universals (Goddard and Wierzbicka 19946) supparts the view that a11 languages have words for at least these three quantifiers: ONE, TWO, and MANY (as well as for ALL, and, probably, SOME OF; see Part 33 of this ~ h a p t e r ) . ~ The element ONE can appear, universally, in sentenoes such as the follawing oms: They have two sons and one daughter. They don" have two sons; they have one son.

In grammar, ONE plays am important role in the widespread (though of course not universal) @ategory of "singular"'. In kct, the category of "plural", too, is based semantically on the concept ONE Csinoe in its prototypical uses it m e a s , usually, 'more than one". The element 'TWO' is needed ta account for the meaning of body part words such as eyes, ears, or hands, as we11 as for the meaning of numerals; and it pilays a significant role in g r m a r in the widespread category of dualis (see Humboldt 1827!1973). Turning now to the thbd quantifier listed by Donald Brown, that is, to MANY, I will note, first of all, that while all languages do appear to have a word to translate the English word many, this word doasn" have to make any overt distinction between 'many'and %ucEnY.Unlike the words for ONE and TWO,the counterparts of mamy do not necessarily imply disc creteness.
Aocording ta Popjes and Polqjes (1986;as reported io Diana Green 1993: 13, 'We Canela language, a member of the JE: language family fin Brazil], bas no numerals auY all; it is limited to poeral terns like "alone", "a couple", "few", and ""many"to express quanliities". 1 e x p f however, lthat if the words in question are carefully tested for polysemy, it will tramspire that the word glossed as '%lone'bc;un also mean 'oone; and that gloss& as "a couple" can also mean 'two'. The words for ONE and TWO can of m r s e have many a0lolexes. For example, Diana Green (1993: 1) notes dhar the Palikfir language of Brazil has "twnly two words to express the concept of one", and that "'nineteen OF these words have dewln dimerent Fonns, making at lead 209 ways lo say (be nwmber one, all spoken in everyday conversationu".

For example, in Polish, the word for both "any' and 'much' is du20, in Russian, mogo, in French, beaucoup, and in Japanese, fakussm. It is true that in those languages which have an obligatory category of nominal number, the distinction between 'muchkand hanykwjlll bc reflected in the number of Ithe head noun; for example, in Russian, in, the phrase mnogo vody, 'much water'@ lot of water), the word for water is used in the genitive singular, whereas in the phrase mnogo sob&, 'many dogs', the word far dogs n the genitive plural. But in Japanese, which doesn't have an oblligis used i atory category of number, there is no corresponding formal distimction, and the word Eakman covers both 'mwchhnd "any'. From the point of view of a speaker oh English, it may seem that a word Pike sdasan must be polysemous m d have two distinct meanings ('muchy and 'many'). But from the paint of view of a speaker of Japanese, such a conclusion would seem counter-intwitive, and it would seem more natural to say thaf the English words much and many are simply twa allolexes (i.e. alternative exponents] of one primitive concept Cctakusan'), and all1 things considered, this "Japanese" point of view appears to be more justified. It is true that to a speaker of English the words much and many appear to have different meanings, and that, ffor example, the phrase msrny chickems means something different from much chickem, but this difference (to do with discreteness) can be attributed to the head noun (with the assumption that, for example, chicken is polysemous in English), and so it can be argued that the two words (much and mmy) are in fact in complementary distribution. Leaving aside, then, the issue of the distinction between much and mamy, Bet us turn to the qlvestion of the semantic simplicity (or otherwise) of the conoept in question, Mere, the only plausible approach ta semantic decomposition would presumably be to try to reduce m;r*rcMmmy to 'morey9along the lines envisaged by Sapb (1949: 125):
It is very important to realize thaf psychologically all comparatives are primary in rellaltion G o thein wrrespmdhg absolutes ("positives'~. Just as more men precedes both some men and many men, so better precedes both good amd very good Linguistic usage tends to start Biom the graded concept, e.g. good (= better &an iradgferenr), bad I(=worse thm irad@eretlt], large I= larger &an o~overage size), much i(= more than ofair m ~ u n f )Jew i , I[= Iess ban afair numbed.

Quoting this passage a quarter of a century ago Fierzbicka 19711, I argued that Sapir's thesis that all comparatives are primary (more basic) in relation to their corresponding absdutes 3s correct with respect to size, number, m o u n t , and dimensions, though wrong with respect to good and bad9 as w e 1 1 1 as to other "qualitative" kkids of aldljectlves. However, in the course of a quarter of a century of research into this and other related mattem, 1 have come to the concllusionr that, attractive as

Sapit's thesis was, it was not partially wrong but altogether wrong. In particular, it no bnger seems plausible that muds and many are semantically based on 'more', whether this relationship is conceived of as 'more than a fair amowntlnuber', "ma than one expects', "ore thm the norm', or in any other way. Intuitively, the idea of 'many people' seems more basic than is, apparently, a lexthat of b o r e people" and the fact that 'm;;my(much~' ical universal (see Goddard and Wierzbicka 19946) supports this intuition. The idea that ' a m a u n t h r "umber' may be more basic concepts than 'mwclELEmany'is entirely implausible: words such as amount and ncrmber are: not universal, they are of course acquired by children much later than much or a lot, and they constitute abstractilolns built on the basis of simple ideas 'onev, and 'two', rather than offering a foundation for them. such as cmuchy, (Far discussion of this question, see Sections b and 7; see also Sedons 16 and 23.) Finally, is 'littler'few' (as in "little butter" or "few people"') also a conceptual primitive and a kxical universal, on a par with MANYSMWCH? Since 'fewfiittle~ppears to be an opposite of 'many!much" jjut as 'small' is an opposite of 'big" and since the latter two concepts have both been put forward as primitive: a d universal (see Section 95, it may seem quite clear that Yewflittle'shonitd also be proposed as such. But the not so sinnpl~e,and preliminary investigations suggest that the claim of 'littlejrfew" to the status of a conceptual primitive and lexical universal is weaker than that of its apparent opposite 'mucbfmany" as well as that d the two related opposites 'bighnd 'small" It is possible, in other words, that 'httldfew' may prove to be reducible to a combination of 'mucWmamy"nd negation ('not mucb'not mamy'), whereas h a l l J m n o t be reduoed to 'nat big' (see Section 9 besow). Data from child language are certainly suggestive in this respect; while (in English) both the adjactives big and little appear very early and are used very frequently in child language (see e.g. Braine 1976; Section 8) and while the words a b t and b t s also frequentlly occur in tranwripts of child language (see e.g. French and Nelson 1985) the same does not seem to be true ofthe wordsjfew and iftble (as an opposite of a lof]~.But the matter requires further investigation.
4.2. ALL

As pointed out by Godderd, however, this kind of analysis may seem convincing from a purely logical point of view, but not from the point of view of psychollolgical plausibility. The suggested analysis seems particularly unconvincing when applied to volitive or expressive utterances, such as "'Regards to all!" or "To hell with it all!"; it seems hardly plausible to paraphrase them as follows: Regards to all. = one can" say thinking of someone: I don't [send) regards to this person To hell with it all. = one can't say thingring of something: I don't want this to go to hell Similarly, the expression 'that's all", which frequently occurs in the transcripts of YoUgr children's speech (see e.g. French and Nelson 19851, can hardly be paraphrased along those lines, The ffact that the ward a// (and, apparently, the concept ALL) appears very early in children's speech [see e.g. Brdne 1976) also supports the view that the analysi attempted in Lingwa Menrralb was psychologicaEly implamsible. ALL plays an important rob in both lexicon and grammar. For example, English particles and conjunctions such as as all, almost, altogether, aithough, arll the same, ai.o, already, al/ sight by their very form hint at the presence of the semantic oamponent 'all' in their meaning, as do also adver'l, bial and pronominal expressions such as always, ail aver, ~ ~ e r a J everywhere, everyone, whatever, whenever, and so on; and i d a r examples could be quoted ltiom other langua~es.(The evidence for ALL as a lexical miversa1 will be: dis~ussed in Chapter 6; see aQsoGoddard and Wierzbicka 1994b.l Like negation and existence, ALL is accepted as one of the fundamental conmpts in logic (as the so-called ' k i v e n a i quantifier'", and from a logical point of view the need far ALL as a universal semmtic primitive will no doubt seem olbvious and overdue rather than controversial. But semant b has its o m point of view, and its o m internal logic, which is dinerent from the logic dlogical systems. Above all, it requires an anchoring in language universals, which have to be confirmed empirically and not only on the basis of inteliwtual spelculation. For h i s reason (among others) the concepts 'and' and 'or" indispensable to the logician, have: not been proposed in the present system as semantic primitives. But ALL--like NOTis one of the points where logic and semantics shake hands.

ALL was proposed as a semantic primitive by Goddard (19898). E n Lingua Mentalis mierzbicka 1980) I had argued that this concept was not indefinable because it could be analysed along the following lines:
All dogs are faithful. = one can't say thinking of a dog: this (dog) is not faithfull

2 A Survey ox Sematic Primisiwes 49

5 . Mental Pr~edicaltes: THINK, KNOW, WANT, FEEL

A s argued elegantly by Bauner (11990: 351, 'k1E cultures have as one of their most powerful constitutive instruments a folk psychology . We learn our culture's folk psychology ear1y", as we learn language early, and the turo processes are inextricably linked. A good illustration of Bruner's thesis comes from the area of emotions: not only is the wry conapt of %moltion" (and its counterparts in other languages) cdturally shaped and determined, to some extent, by language, but every language includes its own taxonomy of emotions, which offers "a set of more or less connected, more or less normative descriptions about how human beings 'tick' " [B~unrer 19910: 35). But in addition to all hose culture-spific systems of folk psychobgy there Is, Bruner suggests, a universal, Innate folk psychology which constitutes a basis and a starting-pdnt of all furlther developments: "We come into the world already equipped with a primitive form of folk psychologf' (1990: 73). Cross-linguistic investigations reported in Semmrlc and Lexical Unirersais [Goddard and Wlerzbicka 1994Er) support and substantiate Bmner's clims; in particular, they allow us to state that the innate and universal "thhary of mindl"%cludes the follbwing major r;onstituenlk: THINK, KNOW, WANT, and E E E L a finding which converges with recent data on language acquisition, as reported, for example, in Welhan 11990, and, especially, 1911411. The results of the research on lexical ulmnfwrsals and on child language acquisition tally well with the results of purely philosophid reflection and conceptual analysis. The iindefinabiiitg of THINK, of the basic 'cogito', was proclaimed three centuries ago by the Cartesians, and nobody has ever h e n able to refute their stand ion this point. In fact, THINK was for tbe Cartesians iu prime example of an indefinable word (Arnauld 166331964: 36): "'Obviously, we conceive nothing more distinctly than we conceive o w own I;Erought. Nor is there a clearer proposition than 'I think; therefore, 1 am'. We can be certain of this proposition only if we am conceive distinctly mean. We require no explanation of these what 'to be' and what 'to thin&;" words, since they are words so well understood that in explaining them we only obscure them." A look at dictionary "definiltions"of think conlinns the validity of this view. For example, The American Herifage Dictianary of she EngItJS hngwage (AHDOTEjC 1973):


and phrases such as "cogitation", "cognition", "cognitive processes"",'conaptions", and the like are almost caricatural examples of the old and persistent practioe of "defining" something that is clear via something that is obscure, and something that i simple via something that is complex. (For further discussion of THINK,see Wierzbicka forthcoming$) What holds for THINK holds allso for KNOW, WANT, and FEEL. For example, Longman's Dictionary o f heE m g I L D S Language (LD0TE.L P 984) defines know via cognition, and cognition via know:

- ta haw direct oognition cognition - the aGt or process ob knowing that invoives the processing oC sensory infomation and includes perception, awareness, and judgehow

Not only are these definitions circular, but also they offer a p o d example of '"regress" "om simpb to complex and from clear to obscure. The same applies to the definitions offeel and wan$ (in the same dictionary]:
reell want

- to have one's sensibilities markedly afFected - to have a desire

to think - to have a thought thought - the act or prmss of thinking; cogitation cogitation - 1, tlnolughtbl consideration; 2. a serious thought The circularity of these definitions hardly requires a comment. Generally speaking, "definitions'hhich try to analyse 'thinking' in terns of words

(with semibl being defined as "capable of being fejh or perceived'" desire as 'Yo wish for, want", and w&h as "'to have as a desire"). The universality of "mental predicates" has sometimes been disputed. For a m p l e , Hallpike (1979) has claimed that some languages lack exponents of THINK and KNOW, and that their speakers, like children at the 'preoperational sfage", have no clear concepts of 'thinking' or 'knowing'. As I argue in detail in Chapter 6, this claim is untenable (as is also the notion that pre-school children don't have concepts of 'think' and 'Loow3; siee Wellman 1494). It has also been sometimes suggested that Borne languages don't have a word for EEEL, or don't distinguish lexically between FEEL and THINK. On closer inspection, however, these reports, too, turn out to be unfounded. (See Wierzbicka 11994jb, 1994h.) The fundamental status of B e concepts THINK, Kbl(EW, WANT, and FEEL is manifested h the important role they play in grammar. For example, KNOW plays an essential role in the systems ef mood-with the 'Veclaratives'Qbeig based on the semantic component 'I know', and the ""interrogariw" on the components 'I don? know-B want to know'. Clearly, KNOW-as well as THINK-is also the basis of "evidentials" ('1 know bsause I see', 'I h o w bemuse I h w ' , 'I Link, I don't say: I know', and so on; see Chapter IS, Section 8). F E U underlies exclamatory constmctions, diminutives, "experienmr constructions" of different kinds, and so on, whereas WANT f o r m the basis of the imperative.

6. Speech: SAY
The universal concept of SAY can be illustrated with the following canonical sentences: I said something to you. People say somlething bad a b u t you. I want to say something now. Like the indefinability of mental predicates (e.g. THINK), the indefinability of SAY can best be appreciated by looking at contortions and vicious circles in the attempted dictionary definitions of this word. The conwpt d SAY plays an important role in speech as a basis of difm ta say ferent il;lawcutionaryfarces (e.g. in questions which imply: want y something'), in the thematic organization of utterances ['I want to say something about this'), and in the basic "subjac~t-predicatey'structure of sentences ('I'm thinking about JC I say: Y). In the lexicon, its most Important function lies in the categorization of discourse, since the distin~tions between diKerent "speech acts" and "speech genres" shape, to a considerable extent, our interpretation of human interaction. (Sae Wierzbicka 19870.1

remgnine that (as argued by Bogusiawski l99ll), both DO and HAPPEN have to be aweped as irreducible semantic primitives. Paraphrases such as You did something bad. = something bad happened bemuse you wanted it are clearly incomoct, and even if it is true that Voinrg'aalways implies some 'wanthg'and some 'happening" these implications cannot be stated satisfactorily in tems of puaphrases, (It is also worth noting that in children's speech, do appears very early, and is used very widely; set: e.g. Clark 1983: $22.) Both the concepts DO and HAPPEN play an important sole in the gramn so-called "a~ctive"languages, such mar of many languages. For example, i as Dakota, the case of the subject depends an whether the predicate refers to Voiing' or to %appeningY. In fact, the tems "agent'hnd "patient" (is. 'the person who does somethingknd 'the person to whom something happens" are widely used in the description of most languages, and fundamental grammatical phenomena such as transitivity, passives, or reflexives are defined if not in terns of 'doingband 'happening' then at least with reference to these eonoepb [sm Chapter 14).


7. Actions and Events: DO and HAPPEN

The concepts of 'action' and 'event" that is, Voing' and 'happening', play an extremely important role in human discourse. Essentially, that's what all sbries are about: what happened, and what this or that person did. To L 9 1 1 0 : 77): "one of the most ubiquitous and powerful disquote Bruner ( course f o m s in human communication is nrrrmrirve . . . Narrative requires . . . a means for emphasizing human action or %gentivity'--action directed toward goals controlled by agents." The future, too, is mostly talked about in terms of future events and actions: what will happen to me (or to some other people)? What will I (or somebody else] do? Indeed, it is hard to imagine a language in which people couldn't ask questions of this kind, and to my knowledge no such language has ever been reported. good) things hapIn human reflections about lire the thought of "bad pening to peopb" wcupies a central place [see e.g, the characteristic title of a papular work When Bod Things Happefi to Good People, Kushner 1982); and t b notion of %someone doing something badyor, to a lesser extent, 'something good" is at the heart of both ethics and the law. In earlier work [Wlembicka 1972, 1980) I tried to define 'happeny in terms of cbecoming',and Voing' in terms of a combination of 'happening' and I have come )to and 'wanting', bur these attempts were not suc~essful,

8. Evaluators: GOOD and BAD

Over the centuries, there have been many attempts to d e h e away the conoepts cgoodyand 'bad'. In most cases, in these attempts "good' was linked with %wantingy,and 'bd", with "not wanting': 'good' is what someme [I, people, God . . .) wants, "ad', what someone doesn't want [see e.g. Schliek 1962 l@ll)e With t h e , however, it has become increasingly dear that attempts of this kind do not and cannot succeed, sinoe it is always possible to present "good' and 'wanting"or 'bady and 'not wanting') in opposition to one another. For example:


I know it is b d to do it, but I still want to do it. 4. know it would be good to do it, but I don? want to do it.
It is also possible to juxtapose 'good'and cwant" as in St Paul's famous statement:
FOP I do not do the goad I want, but the evil I do not want is what 1do. [Romans 7: 19).

Both by contrasting and by juxtaposing 'goodknd "antbe show that we conceive of them as two separate ccmcepts: if I can say that I can want !to do what 1 think is good, and that I can also want to do what I think is bad,

2. A Survey ~~JSrej~lenrfc 'crif~itives 53

then it would make no sense to try to reduce 'good20 %antJ, or %ada to &not wanty. The (non-compositional] relationship between %anthand "oodhan be compared to that between 'think'and %now', in so far as both k a n l k n d Yhink'imply a subjective, individual perspective* whereas "ood' and %nowq imply an objective and inherently valid one. What someone wants may be bad, and may be diffferent from what somebody else wants; sirnilarly, what somoue. thinks may 1Pe wrong, and may be different from what somebody else thinks. But what is " g a d " (or '"ad"') is good [or bad) regardless of individual differences in the point of view, m d what is %mwn"to anybody) most be true. O f course people argue about what is 'good' and what is %add',but this very k t indicates that by lasing these words they lay a claim to some objective validity. I n n other words, people regard EBiflerent things [and diuerent actions) a5 good or bad, but they all agree that some things or actions (no matter which ones) can be validly regarded as "oodd" or 'bad' (although they may not agree which things or actions). What applies to the speakers of English the users of the English words good and ha4 applies also to all other languages and cultures: everywhere in the world, people may disagree whether something is 'good" or 'bad', but in doing so, they rely on the mnwpts 'gooda and "ad', The fact that-as far as we know-all languages have words for GOOD and BAD (see Hill 1487; Goddard and Wierzbicka 19942) strongly supports tlne hypothesis that these two mncepts are innate and fundamental elements of human thought (experience can teach us to regard certain things as 'good' or 'bad', but it cannot temh us the very concepts of 'good' and 'bad'). But while the fundamental nature of the concepts GOOD and BAD appears to be well established and well supported by linguistic evidence, one puzzle remains: why is it that (as pointed out by Greenberg 89660: 52) in many languages the word for BAD looks, from a mnpholo@cd point of view, like a combination of negation and the word for GOOD, whereas the word for GOOD never looks like a combination of negatives and the word h r BAD? Donald Brown (1991: 131) writes this about his imaginary "'Universal People" ('UP): However much grammar varies from language to language, some things are always present. For example, UP language includes a series of contrasting terms hat theoretically could be phrased in three different ways, but that are only p h r a d two ways. To illustrate, they muld talk about the '"ood" and the "bad" [two contrasting terms, neither with a marker d d e d to express negation); or they could talk about the "goodY>nd the ""not good" &e., not having the ward " " b a a a ' k a t a11 brut expressing its meauing with a marked version of its opposite, the marking in this case to negate), or they coluid talk about the "bad" and the "mat bad" (i.e., not hay-

ing the word '"ooodl," etc.). Logically, these alternatives are identical: each arrangement conveys the same inrannation. But . . . the third possibility never omrmrs as the obligatory or common way of taEking. But although, apparently, in some sense GOOD is "unmarked", whereas BAD is (perhaps] perceived as rm absence, distortion, or perversion of GOOD, it appears that all languages have a word for BAD,as welt as a word for GOOD. The word for BAD may or may not look like a negated version of the word for GOOD, but if the only "opposite'kf GOOD to be found in a language looks like a negated vewion of the word for GOOD then it seems that this word means 'bad'rather than 'not good'. Where culltwres differ is in their willingness to contrast 6006) with BAD: clearly, in some cultures people preffer, in many contexts, to contrast 'good%with 'not good-ather than with 'bad'i&resumably, to avoid giving offence). If this is the case, then the word for HAD n a y seem to be somehow 'cstronger" in meaning than the English word bad. For example, Ghappell (1994: 142) unrites this about Mandarin Chinese:
GOOD and BAD are senantantically asylnmetricaP in Mandarin, hvdi 'bad' being

semantiaily narrower in its range of application. In this case, the use of simple mgalicrn of the moupheme Sslio GOOD which gives b G hiio might, in fact, Bbe preferable since Ssdi is more semanticaPly specialized at its end of the sale to mean 'immoral', 'nasty'or "ell' than hsjo is on the 'saintly'end of the scale.

I would suspect, however, that diRerences of this kind (interesting as they are) are due to culltural rather than strictly semantic reasons, and that BAD, like GOOD,is indeed a universal semantic primitive. (For further discussion, see Wicmbicka 1994b: 4916-7.) The idea of a %ba deed", a %ad person', or 'bad peoplebay play a greater role in some cultures than in others; for example, it is no doubt more prominent in the Judaeo-Christian culture than, say, in Japanese culture, but this doesn't mean that in Japanese culture one cannot speak at ali about %ad actions' or 'bad people'(~ ~ Qnishi 1994). T o say that %ad3means the same as 'not good' is a bit like saying that "lalack' is the same as 'not white'. If not good may sometimes be used as a euphemism f o bad, ~ it is precisely because the two do not mean the same, and to say to a child "It is bad to lie" is not the same thing as to say "To lie is not good". To see this irreducible diflerence between "adhnd h o t gooda it is usef d to consider "stronger" words such as evil, v i c i w s (as applied to actions or people) and words such as terrible or horr$c [as applied to events). It mems hardly neoessary to argue at length that an "evil deed" or a "terrible disaster" is not simply "something that is not good". As for Sapir's idea that, psychologicaliy, %etter3 precedes 'good' (and that, by implication, kwae"recxdes %ad"), it is inconsistent with both

54 General Issrses moss-linguistic evidence and evidence from child language: it is 'good" and 'bad" not 'betterband 'worse', which emerge as lexical universals and which commonly owur in transcripts in the data ofconversations with young children. Thus, data from cross-linguistic investigations and fmm child language research converge on this point wilth in-depth semantic analysis of natural language, pointing to the fmdamental, irreducible character of the twin concepts GOQD and BAD. words for 'bigkand 'sma11bnd a well-known non-universality of comparatives (see e.g. Longacre 11985: 243). Swond, it has become cllear that the 'kelati~e"character of \the conoepts %bigband " d l ' can be awoun"yjed for without any comparatives, along the following lines:
This is a big dog. = when I t h k of dogs, I think: this is a big dog

9. Descriptors: BIG and SMALL

The concepts %ig' and 'small'are particularly easy to identify cross-lingu.istically: words for 'big' and 'small" are frequently used, easily acquired, and easy to identify (see Goddard and Wierzbicka U1"394b),As frequently noted i n the literature, the lexical exponents of these concepts are nomally not symmetricall, with the word far BIG being treated as, in some yay, more basic. For example, the question "How big is it?"' does not imply that the object in question is big, whereas 'Wow small is it?" does imply that it is small. (See e.g. Greenberg 1966a: 52-3-31 It is quite tempting, therefore, to try to d e h e SMPILL via BIG as 'not big" But 'not big-oes not mean the same as %mally,and one can easily say off something (for example, a dog, or an apple) that Yt is neither big nor small'. Both B 1 0 and SMALL are of course ""relative" terns, in the sense that a small elephant is still quite a big animal, whereas a big mouse is not. As argued by Aristotle in his Categarim: Things are net great or small absolutely, they are so cdled rather as the result of an act of comparison. For instanoe, a mountain is called small, a grain large, in virtue of the fact that the latm is greater than others of its kind, the former kss. Thus there is a reference here lo an extcrnal standard, for iF the terms "greatknd 'small' were used absolutely, a mountain would never be called small or a grain
large. (1437: 5 3

Aristotb" argument is of course persuasive, and it was E ty h kind of reasoning which led me to posit in earlier work (Wiembicka 1971) an analysis of 'bighand 'small' based on the comparative: This dog is big. = this dog is bigger than one would expect (See alsohkovskij 1964b.)B have now repudiated such analyses, however, and for a number of reasons. First, there is the question about the meaning of the comparative itselfi if we define 'big\vla 'biggerke couldn't define 'bigger' via 'big', and we would probably have to accept 'bigger'as a conceptual puimiltive-a very dubious move, given the apparent universality of

As mentioned earlier, the words big and little appear very early in children" sspeech, and are used very frequently. For exarnpne, Braine (1976: 32) draws attention to a productive pattern of two-word combinations with the words big and littk in the speech of his son, Jonathan, before his second birthday. Bnterestingly, Bratne points out that ''Jonathan often contrasted two objects i n consecutive utterances, for example, bfg stick followed i m e dhtely by little ~ f i c kindicating , the relative sizes of the two s t i c k s ' k d he comments: "This sort of behavior seems suflicient evidence for the productivity of these size-attribution fomu11ae." But while the kind of behaviour exemplified by Jonathan shows that the are felt to be sernanticallly linked, and that their juxtawords big and 1ft;tsL position is indeed usled for oomparison, it does not mean that the conoepts in question are semantically "relative'% the sense suggested by SSapir [as discussed in Section 4). The idea that brge (or big) means 'larger than of average s k yseems completely incompakible with the frequent and competent use of the words big and little by infants in the second year of life.s The pattern d u s e dewribed by Braine supports the hypothesis that a comparison of size YX is bigger than Y 3 ' is based on agiuxtaaposition of opposites [next to TT, Xis big; next to X, Y is little). It is also consistent with the of size (as in ""This is a big dog") hypothesis that an: "abso~uta'~ssessment refers molt only to '?his'~uulk also to "dogs" in general ["For a dog, this is a big dog"]. St is paPliicularly interesting in this connection to compare Braine's comments on little Jonathan's way of handling comparison and that described by Longacre with respect Ito the languages of Papuw New Guinea: Comparison in Papw New Guinea is not expressed within a single sentence, but by a pair d sentences within a paragraph. It is, furthennore, really not comparison, but contrast. In Safeyoka (a dialect of Waj~okeso), for example, we find pairs of sentences such as 'The black man's bbots are small. The white man%boats are huge'. There is no direct wag of saying *Theblack mm's boats are smaller ahan the white man's boats'or 'The white mm's boats are bigger than the black m a n ' s ' .
As Johnstom (1985: 98OJpoints out (from a child's point or view], "Haw can one small sboe Be bigger than another small shoe, or a single ohjjlieca Be both bigger and sml!erY What tiis observation shows is that the idea of a "small shoe" or a "big shoe" i s bate Sapir) psycluolagiwlcally simpler ahan that oTa '"smaller shoe" or e "bigger shoe".

The fmdamental nature of the concepts %bigyand %mall?s reflected in the rote that they play in the grammar of many languages, i n particular in the categories of so-called ""'diminutives"and ""rawgmentatives"". la. Time: WHEN, BEFQRE, AFTER As pointed out by Keesing (19941, time tends to be exoticized in Westem aaounts of non-Western languages in cultures. The best example of this exoticization is the account of the Hopi language given by Whorf, who claimed that the Hopi conception of h e is radically different from that reflected in European languages. "After long and carefd study and amalysis, the Hopi language is seen to loonltain no words, grammatical forms, constructions or expressions that refer directly to what we call 'time', or to 31956: 57). past, or future, or to enduring or lasting" 'horf But Whorf's ideas about Hopi have now been refuted in a careful study of the Hopi language by Malotki (19831, whose overall conclusion is that " W a d % claim about Hopi time conception being radically different from a w s does . . . not hold" (5301. In a similar %in, Keesing argues that the Kwaio language d the Solemon Islands, which he has studied in detail, has the same basic ternporal categories as Western languages do, and that "Kwaio talk about duration and temporality in everyday life much as we do" (1994: 5). In particular, Kwalo has a w o m a n word (ahfa) "referring either to points in Q used as well as equivalent to English time ar periods in time. . . A ~ E is %heny or French 'quandy, to introduce temporal clauses, as in: d a t a maim migi i 'Aoke, W e n we gel to Azlku'." Keesing concPudes (6): ""A1 the evidence available on everyday talk in non-Western languages would indicate: that other 'exoricy peoples, like the Kwaio, situate events precisely i n time in complex ways, are wncemed with duration, and have intricate lhguistic devices for wordhating plans and activities." The cross-linguistic inwestigations reported in Slemmssleic and Lexfcal ICPnIversaJlr;(Goddam-d and Wierzbicka 1994b) point in the same direction: despite all the differences in the conmptualizatim of, and alititudew to, time, discussed, for example, by Gees& (1966) or Hall (l983), and despite the considerable diffierences in the ljnguistic encoding of temporal notions (see e.g. Hopper 1982; Bybee and Dahl 19891, the fundamental temlpord conoepts encoded in languages of the wodd appear to be the same. They indude: W E N , AFTER, and BEFORE. (Far dimssiora of durational concepts, see Part LE of this chapter.) Cross-linguistic investigations suggest that, for example, questions such as the following ones are readily available and frequently used in all lanpages:

When did it (will it) happen? When did you [will you) do it? For example, Kwaio "abounds with terns . . . that mark times in the diurnal cycle. . . . Kwaio talk is sprinkled with these markers of the daily cycle, based on angles of solar deviations diurnally (and such phenmena as dusk and subsequent insect noises, such as kemlrS "rickets cry", which allow p e cise planning and looordination of work, rendezvous, and trave8'~K~esing 1994: 6). E t would Iw: quite futile to try to reduoe such when-phrases to something simpler; and it wodd be equally futile to try to reduce phrases of temporal succession ('after' and 'before9) to something simpler. If it has often been claimed that a sense of 'whenness3s lacking in "primitive thought" "ncluding, for example, medieval European mentality; see Le Pan 1989: 1133, even stronger claim have been made about the alleged absence of the idea of 'temporaI suocession' in many languages and GUItures, and a linear conception of time has often beem contrasted with a cyclical one (see e.g. Le Pan 1989: 89). Cross-linguistic investigations suggest, however, that whatever the differences in cdtural emphasis and elaboration might be, all languages have words for AFTER and BEFORE, and that in any language one can easily make statements such as ""A was barn after W a n d ""B was born before A", or " A died after B" and "'B died before ASS. What is particularly interesting about these findings is the apparent redundancy of the exponents of temporary succession; for why should all languages have words for both BEFORE and AETER, rather than for just one d these concepts? Aren't they reducible to one another? A o m a logical point of view, indeed, they might be reducible to one another. But the fact that natural languages have lexical resources for expressing both of them suggests that from the point dview d human conceptualization of reality, '" happened after X'means something diflerent from, and irreducible to, "X happened before I " " Clearly, what matters is a different perspective on the events in questian, a different point of view, and, as pointed out by Slobin (11985a: lE81), "the ability to view scenes from different perspectives" is a salient feature of human cogmition, cllearly reflected in all languages. It is particularly interesting to see in this connection the early emergence and kfter'in child language. of both 'beabre"sometirrmes realized as For exarnpb (French and Nelson 1985: 110-11 1 1 ; see also Cami and French
1984): AJter the birthday, they go home. [age: 3 years, 11 month] Well, $rst I1 didn't t a w how to, but now, when P get dressed, 1 can put on m y pants. [age: 4 years, 2 months]

And my daddy jlllst wants to eat them. L i b chocolate cookies. B had one befare we came here. [age: 3 yeas, 1 0 manths]

In a narrative, AFTER is realized as a simple portmanteau ""Lenn"" or "and then" [i.e. Wa8r this'). As Bruner (1990: 79) notes, "Childmn early start mastering grammatical and lexical f o m s for 'bincBing2he sequences they remunt-by the use of temporals Eike "henhand 'later', and eventually by the use of causals."


What applies to the temporal triad WHEN, A S E R , and BEFORE ER, and ADQvE. appears to apply also to the spatial triad WHERE, First of all, evidence suggests that all languages have a word for WHERE [SOMEWHERE, PLACE) distinct from the word for WHEN. (See Goddard and Wierzbicka, L994b.E This word lclan apply to both entities and events; for example: Where are y0u1 (Where is this thing?) Where did it happen7 Since for entities in contrast to "happenings") 'whereys interpreted as %being somewhere', in the system posited here 'being somewhere' I s regarded as an alllolex of 'somewhere'. As the universality of the words for WHERE confirms, WHERE Is a fundamental human concept, incapable of being defined. In earlier work (Wierzbicka 19721, I have tried to analyse WHERE via PART, but this attempt lad L o bizarre and intuitively unaaptable results; and to my howledge no other, more successful attempts to analyse this concept lnave ever k e n proposed. As for ABOVE and UNDER, they present the same apparent redundancy as AFTER and BEFlORE do, for if A is above B then B must be under A. By itself, this redundancy would not be a reason for not positing them both as primitives: since human minds are not disembodied wmputers (see Jlohnson 1987; Edelmaru 19921, our conceptualization of the world reflects our "embodiment", and also our position on the ground: since we normally walk with the head up, the contrast between ABOVE and UNDER may not be conceived of as reversible. None the less, it must be admitted that the case for positing both 'under' and 'above' as semantic primitives (as it was done jm Goddard and Wierzbicka 1994b) is not nearly as strong as that for 'beforey and 'after'. Intriguing data bearing on this question come from research into child language. In particullar, this research shows that while both Ithe conuoepts

' b f o r e b d 'afterhamerge early [apparenltly before the end of the third year; see Cami and French 11984), the spatial pair ~("under'and 'above'] behaves in a curiously asymmetrical way, with "nder' apparently emerging much earlier than ' a b ~ v e ~ m not d only in English but also isn a number of other languages [see Johnston and Slobin 1979). Another relevant fact emerging from child language research is that the mncepts linked in English with the words up and down appear very early in children's speech, and are used very frequently [so much so that, for example, in the speech of the 16-nnonth-o8d Allhon Bbom they were among tlnr: seven most frequent one-word utterances; see Bloom and Lahey 1978: 114; also Clark 1985: 746). At first, these words apply predominantly to the movememts of the child's body, as he or she is being Eifted or put back down by the parrent (see Boweman 1976: l67]. By contrast, the earliest uses of d~nder (and its equivalents in other languages) are no doubt stative, and presumably refer to manipulable objects located temporarily under large objects with more or less fixed location, such as a table or a bed. ('The idea d looking for things under the table or ulnder the bed is no doubt more relevant to a small child than that of looking for things "above" something-presumably, parlilly bemuse of the child's o m size and partly because things fall down, rather than rise.) It is interesting to note in this connection Clark's (11985: 7441 observation , , being in contraslt in about the prepositions sous, "udder', and s ~ r "on' French children's s p h . It makes a lot of sense to assume that from a child" point of view the likely choice is between interesting objects being under the table or OM the table, rather than under the table and hove the table. Considerations of this kind suggest that while 'under'can be seen as a well-established semantic primitive, % b o w b n n o t ; and also that in future research the motion of ban' (as in "the box is on the table"') should be scmtinized ~ Y Sa p s s i b b canwptual primitive ~(presumably,along with the notions 'uphand Uown'). From a purely logical point of view, "bove'and ' u n d e r b a y seem to be relahd in the same way as 'before' and %fterY, that is, as converses (see e.g. Apresjan 1974, 8992; Cruse 1986). But since in human experience the spontaneous movement of things is unidirectional (because they fall to the ground, unless supported on some stable surface), the contrast between 'on' and "under' m q be psychologica1Py more real (as far as location of things is concerned) than that between 'under' and "above'. On the other the movement bpp" may indeed hand, as far as human action is con~erned, seem to be directly related to the movement Vown', and at this stage it is dinaicult to see how one of these concepts could be defined in k m s of' the other. [For discussion of 'in' and 'inside', see Section B of this chapter,)

12. Partanlomy and Taxonomy: PART (OF) and KIND [OF)

12.1. PART PART is a controversial primitive, putty (no pun intended) because many languages don't have a word with a range of use similar to that of the English noun part, and partlly bemuse some languages don't seem to have a word for part at all. In propsing PART as a universal semantic primitive, th~erefore,it is important, k s t of dl, to chrify what uses of the English part are meant to illlustrate the paskuhted primitive; and second, to examine how the meaning in question is expressed in a language which doesn" seem to have a word corresponding to the English part at all. Pn English, part cam be used in (at Least) three different ways. First of dl, it can refer to "things" identifiable, SO to speak, within larger things, as in the fallowing sentences: The foot is a part of the leg. A knife has two parts: a blade and a handle. A petal is a part of a flower. Second, pma can refer to a '"iece'kf something, that is, to something which cannot be thought of as a m identifiable thing before in gets detached from a larger thing [see Cruse 19816: l 571. For example: He ate part af the melon (not the whole melon). Third, part a n be used to ~refferto a subset of a group of discrete entities, including people: Part of them went to the right, and part went b the lek. In the system of primitives proposed here, the semnd and third uses illustrated above, which can both be linked with the traditional label "partilive", exemplify the newly proposed primitive SOME (OF) (see Part 24 of this chapter). It is only the first use, then, which is regarded as corresponding to the primitive clement PART. It is inconceivable how the word part, as used, Ifor example, with reference to body parts, couEd be defined in simpler concepts. It Is also inconceivable that a language would fail to provide its speakers with some means for referring to body parts and for saying that, for example, the head is a part of the body, or that the foot is a part of the Peg. (See Andersen 19783; C. Brown 1976; Chappell and McGregor 1995). How does one express such thoughts in a language which does not have a noun cclrresponding to the English noun part? Different languages provide different solutions. For example, in the Australian language

Yankunytjatjara, PART is expressed by means of the so-callled %avinghuffix -@t;rr (Goddard 19946: 254-6): Yunpa mulya-tj~ara, tjaa-tjara, k u ~ ukutjara-tjara. face nose-]WIVING mouth-HAVING eye two-HAVING % face has a nose, mouth and two eyes (as parts).' Puntu kutju, palu kutjupa-kutjupa tjuja-tjara. body one but something many-HAVING '(It is) one body, but with many parts.' (Romans P2:4) As noted by Goddard, the swfix -tjara is polysemous, but in certain contexts it can only mean "art', In particular, Goddard points out that a sentence whose '"word-for-word calque rendition . . . could be read as 'This thing has two whats?'has a very clear meaning in Yankunytjatjara, and refers unambiguously to the number of Bmu nyangaltja nyaa kutjara-tj~ma? thing this what two-HAVING 'What two parts does this thing have?'

It is undoubtedly true !that cultures differ in ihe ainount of interest they show in the concept of "art3. As argued in Goddard 1989a, modern Western culture places a great emphasis on viewing various aspects of reality in terns of complexes analysable into 'parts" whereas, for example, Australian Aboriginal culture does not. But cultural difierences of this kind should not obscure the f a d that the conoept of PART cam also be expressed in those languages whose speakers are less inclined to talk about '"arts" in the abstract (in contrast to heads, feet, handles, and other s p i f i c kinds of "parks") than are speakers of technologically complex modern societies. (For further discussion, see Wierzbiclca 1994b: 488-92.) The concept of PART plays a m important role in the grammar of many languages, mainlly because it underlies so-called "possessive constructions'" of various kinds. The label "possessive", frequently used in grammatical dm.criptions, has no constant semantic content, but it is usually u s d with respect to constructions whose meaning involves the concept of' PART. For example, the so-called '?nalienable possession" is usually based on the notion of % part of a person'ar 'a part of a person's body' (often extended to things which are seen as 'like a part of a person" see Chappall and McGregor 1995). On the other hand, "alienabEe possessionY~s based on a .combination af PART md, so to speak, Visponibillity'. The conceptual links between 'parthood" and 'ownership' can be represented along the following lines (partial explications only):
my hand - a part of my body if I want, I can do many things with it

62 General Issues rmycar - a c a r if I want, I can do many things with it like I can do many things with a part of my body other people can't do the same with it It is not clear when the concept PART first appears in children's sspeiech, although utteramax such as '%ommy hand" or "caw taily?re attested at a very early stage (see Braine 1974: 15, 19). Interestingly, Braine (1976: 7) includes also the combination 'bother part" in his record of one child's earliest two-word combinations. Apparently, 3- or 3-and-a-half-year-olds can already be quite compebnt in talking about "parts'" For exmp1e (French and Nelson 1985: 1091:
Eat the paen part first. ['"green part" refers to icing on the cake.]

2 A Swwey of Semantic Primitives


82.2. MIND The concept of "inrd' is at the heart of the human mtegorization of the "contents of the world", The lexicon of every language is fidl of taxonomic concepts which rely crucially on this concept. For example, in English, a sol3.s is 'a kind of flower', an oak, 'a kind of tme', and a parrot, 'a kind of birdy. The important role of taxonomic (l.e. %kiddi'-based) classification in all languages and cultures has oftiten been denied in the past. B n particular, it has ofken been claimed that in traditional non-Western societies ethnobiological classification is predominantly non-taxonomic-unlike the Western scientific dassification, which is based on a hierarchy of kinds. In a cllassic statement of Ithis position Uvy-Hruhl (1926: 176) wrote:
Iln spite of appearances, then, these minds, which evidently haw no idea of genera, have none of species, families, or varieties either, although they are able to dellineate them in their language.

To illustrate this claim, Uvy-Bruhl (1925: 170) repeated with approval remarks made by a traveller about Australian Aboriginal languages:
He states that generic Items such as tree, fish, bird eihc. were lacking, although specific terns were applied to every variety of tree, fish or bird. . . . The Tasmanians had no words to represent abstract ideas, and though they could denote every variety aE gum-tree or bush by name, they had no word for tree.

But while trustworthy recent investigations by anathropollogists and Bingwists have shown that general terns such as 'tree', 'bird', or 'fish" not to mention 'animal' m d 'plant', may indeed tPe scant in a language while more specirfic words for creatures and plants may be present in abundanoe (see e.g. C. Brown 1984; Berlin 1992), the idea that a hierarchy of kinds is either absent from, or marginal to, hlk-biological classifications has not stood the )test of time. (See in particular the evidence and discussion in Bedin 1992.) First, the apparently universal presence of at least some hierarchical categorization reflected i n the iexicon (e.g. beeJ-'oak" or %bird"-kuckoo') does support the view that taxonomies play an important role in the conceptualization of living kinds, despite Hallpike's (19791: 202) and others' unsupported assertion to the contrary. The semantic relation between h m s such as 'tree'and 'oak' can be verified by a variety of llinguistic tests, and wherever such tests have been aplpllied they support the view that class inclusion is indeed involved. (See Chapters 11 and 12.) Second, as pointed out by Bedin (1992: 52-31, in every known language there is a set of words regarded as the "real names'bf certain classes of living things. When asked "What is this called?'?nfonnants might reply with a folk generic term, with a '%orizontal extension" of such a tern [e.g. "'Et is like lilac"], or might say "1 don't knowM",but they will not say, for example, "It is called a bird", or ''It is called a bush"'. The presence of such "rea1 names" establishes beyond any reasonable doubt the psychological reality of the notion of biological species (or 'Tolk genera'". Third, and most importantly [from the present point of view), linguistic evidence suggests that the concept of 'kindyor 'kinds') is a lexical uniwrsal. Ellen (8986: 88) mocks the idea that informants may in the course of their ordinary lives use senltences such as '"8 X a kind of I"?' or "'How many kinds of Yare there?", but in fact senknces referring to "'kinds" of living things are widely attested im traditional non-Western languages. (See Goddad and Wierzbicka, B994b.J This applies, in particular, to the following types of sentence: There are three kinds of bat @am, mango, etc.). This is not the same fish, but it is the same kind of fish. It is particularly important that cross-linguistic evidence supports the universality of the distinction between KIND and LIKE (or betweha categorization and similarity), as illustrated in the following sentence:"
It seems natural to think Ithat t h i n g which arc ""afthe same kind" [e.g. cabs; or oaks; or da@"a;@ldils] are also "'like each other'" Yet recent work i n cognitive psycluolagy as we61 as linguistics has k d to the growing mnvidion that hwman categorization cannot be reduced to notions such as "'likeness" or "similarity'" [See e.g. Atcan 1990; Carey 1985; Gelman and Coleg 1991; Keil 1986; Medin and Ootony i989;Rips 1989.)

Similarly, of American Indians L6vy-Bmhl (1926: 1171) says the following (again quoting, with approval, travellers' reports):
almost every species has its panticullar Indian name. But It would be in vak to seek among them words for the abstract ideas of plant,animal, and the abstract notions colour, tone, sex, s p ~ i e s etc. , . . . [n California, "there are m a genera, no species: every oak, pine, or grass has ils separate name.'"

This tree doesn't look llke that other tree, but they are the same kind of tree. But if the concept of KIND cannot be reduced b the concept of LIKE, it can hardly be reduced to any other conoept or concepts, and, to my knowledge, no viable decomposition of KIND has ever been proposed. Qn present evidence, therefore, KIND must be regarded as a universal semantic primitive. (See Wierzbicka forthcoming d) Finally, it is worth noting that, in children" language, the question ""What kind of?" is attested as appearing relatively early, though after 'what" 'wherea, or 'who' [see e.g. Clancy 1985 on Japanese; SmiE 1975 on Serbo-Croatian)

113. Metapredicataw: NOT, CANI VERY

13.1. Negation: NOT
The nos-relation is ane of the simplest and mast fundamental relations knawn t o the human m i n d . Wayce 1917: 265; quoted in Ham 11989: 1)

Negation is probably the least controversial of all the lexical universals which have ever been proposed. Nobody has ever reported coming across a language without negation, and exponents of negation-unlike those of most other conceptual primitives posited here-are routinely reported in all descriptive grammars. But while the question of negation as a lexical universal is quite straightforward, the same m n o t be said about its status as a conceptual primitive. In fact, to many readershurprise and even dismay, negation was missing from the iist of primitives which I postulated in Semantic Primitives (Wierzbicka 1972) and in Liaguca Mentalis mierzbicka 1980); instead, my 1972 and 1980 lists included the elements 'don't want' ("dliswant') or 'I don't want' I('" Iswant"), .(Seealso Wierzbicka 1967.) In postdating 'diswanting' rather than negation as a semantic primitive, I was trying to come to grips with the fact that tlne semantic relation between the phrases '" want" and "I don? tant" seems to be different fmm that between, say, "I know" and "I don't t o w ' " or "I do" and "I don't do"- "1 don't knowY"or '" don't do"] means, roughly speaking, that Tt is don't want", however (on one readhg not the case that I know (or do)" ''I at least), does nolt seem to mean that 'it is not the case that I wantJ [as h ""Ion'f particularly want"); rather, it seems to mean that I positively 'diswanthomething. It is also true that the interjectim No! can be used to express a strong 'di~want~~rejection"), rather than merely a denial of

wanting. By assuming that Uswantinghas sernanticakly simpler than negation, I seemed to be able to explain such facts, What was more diffic d t to explain in that approach was the use of negation in declarative sentences-a point to which I will return be low.)^ But although my analysis of negation as "rejection'kas consistent with a long phiBosopIniieta1 tradition of thinking about negation (see erg.Bergson 1911) and although them is considerable empirical evidence which appears to support it, I have now came to doubt whether it is tenable. Undoubtedly, acts of "rejection" ('I don't want this!') play an important rob in human life, and it is not surprising that "rejection" should have specia8 lexical exponents in many languages (e.g. in Acehnese, Longgu, Samoan, ;and Kayardild; see Goddard and Wierzbicka 1494b). But althauglh these exponents of "rcjcction'~awoficlr identical with tkpc GXPOnents of negation, this does not necessarily mean that "rejection" is a simple semantic notion, which can be said to underlie all negation. After all, many languages (e.g. Samoan) have also special negative imperatives ('donPt?"espib the fact that Uon't!' is not a simple con~ept but a sernantic molecule analysablle as 7 don%want you to do this'. While accepting the old philosophical notion that all negation implies "rejmticm", I experimented with a whole series of analyses of declarative sentences, such as, foa example, A, By and C below (for " 'This is not black"):
(A): I don't want to say ""tis is black". (B) I don't want someone to think: "his is black". [C] I don't want to say: I can think: "this is Mack".

But none of these analyses seemed quite right, and the "details" seemed always impossible to work out. In particular, it has always seemed dficullt to see how mmplex sentences Qe.g. with negation embedded in an $-dause) c o l d be pllausibly analysed via bolt wanting" What has finally convinoed me that this whole approach to the semantics of negation was probably misguided was a closer examination of data from child language aquisition. Since negation appears very early in children's speech (in tlne second year of life; see e.g. Braine 19761, it is very hard to believe that spontaneous utterances such as "no wetY"meaning 'I'm nat wet'; see Braine 1976: 7) can be somehow based on the idea d k o t wanting someane to think or say something'. It is true that i n early child utterances "0' frequently means 'I don't l don't want to want' (e.g. "no mama" is interpreted by Braine 1976: 7 as ' go to mama'], but it seems mom plausible to analyse "'no mama" as 'I don't want . . . mama' than "no wet" as '1 don? t a n t someone to sayithink that I am wet'. I have come to accept, then, that 'notYs simply 'not', and that it cannot

66 C;esreral Ssmes be reduced to anything else-not even to the intuiltively appealing notion af 'reje~tion'. Hawing reached this conclusion, I would now interpret the three eadiest uses of negation identified in child language as '"on-e~stence"",'rejection", and "denial" (see Bloom 1991) along the following lines: (I) 'there isn't an X (here)', (2) 'I don't want this', and (3) 'this is not an X', assuming that aJl these uses involve the use of the same semantic primitive NOT. Far example (Bloom 19911: 143):

(Kathryn not finding a packet in Mommy's shirt, which had no pocket:) KATHRYN. no pmket [i.e. "here is atp pocket herell

and %eaial'bave b a l l y oanvinced me t h t solution B is more justified, after all. What remains to be explained (on the assumption that NOT is always NOT) Is why the semantic relation between the phases "I don" twantY'and "1 want" dodoes not seem to Be the same as that between "I don" know" and "I knowD',or "I don't d o ' h n d '" do", and why "I don't want to do A ' " often appears to imply that I positively wmt 'boot to do P. At present, I do not knave a fully satisfactory answer to this question; but the possibility of an analysis in terns of invited inferences (perhaps along Gricean lines e.g. Orice 11975) is no dnoublt worth exploring.

[Kathryn pushing away a sliver ofworn soap inu the [bathtub, wanttatg to be washed with new pink soap:) KAT~I~EVN to . di~hr soap time. 3 don't want dirty soap']

13.2. CAN
CAN is a relatively recent addition to the list of primitives. From a crosslinguistic point of view, "an' is particularly difficult to identify, partlly because It is often involved in complex patterns of polysemy, and partlly RKncausr: its exponents oftemr appear to be bound morphemes rather tban distinct words. From the point of view of decomposition, there is allso a temptaliion to try to treat "an' as complex, because of its intuitively "fly' character (discussed, for example, by Austin 1961). Yet all attempts to define away 'can'(inc1uding my own) have proved unsatisfactory (for general dscussian, referenaes, and for m y earlier analysis of 'cm', see Wierzbicka 11487b). The conclusion that 'can' is semantically elementary, despite its apparent intuitive links with 'if', had a significant liberating eKect on subsequent analyses across numerous semantic domains (as had the earlier conclusion that 'because' was semanticaaly elementary, despite its intuitive Pinks with "if" see Section 14). Because of' the close links oetween CAN and the newly proposed primitive MAYBE, the two elements will be discussed jointly in Part B off this chapter.

(Kathryn, Mommy, and Lois looking for the truck:] Where's the truck? (Mommy picking up the car,giving it Ito Kathryn:] Were it is. There's the truck. KATHRYN, no truck [i.e. 'this is nab the truck'] Horn (1989: 863) comments on Bloom's 'krejection'btegory of negation as foPlows:
Bloom" rejection category corresponds to what pK~losophers-at least since Peirce-have long identified as the S I U B J E ~ l V Eor PRELOGICAL negative. Heinemann (1944:1138) glasses this "rdagical use of negation"^ 'l da not wish [will, desire, etc.1 thatkr 'It is not in my interests that" aellngside the 'Uogical' negation oE'It is not itrue that'. On this view, the rejection category should antedate both nomuexistence and denial; that it does not (at least in Blam's data] may reflect the diKerenoe between possessirng a concept amd expressing it syritacticaIlg.

From a "semantic primitives" " p o i n t of view, the question boils down to choosing betmen two soiutions (A and B) to the problem of negation: should the 'kejwtion" use of negation be seen as based on a simple mmantic element don" want' (in Latin, nolo) and the "non-existesrce" or 'cdmial" use be seen as based an a combination of T don't want' and 'say' (solution A); or should rather the 'non-existence' and 'denialkse be seen as based on a simple semantic element 'not', and the 'rejectionbse, as based on a combination of three primitives: 'I + not + want' (solution BIT In earlier work wierzbiclca 1967, 1972, 1198h) 1 opted for solution A [despite strong opposition from my colleague Andrzej Boguslawski, who for a long time has argued for negation as a semantic primitive against my attempts to reduce it to a simpler notion of 'rejection"); but the language acquisition data suggesting that Yejie@tionWoesnYt antedate 'non-existence"


'The concept of W R Y might be seen as dispensable in, a universal system uf semantic primitives, as it is inherently subjective and "hg~recise". But evidence suggests that all natural languages have a word colrresponding to the English word wery and that despite (and perhaps because ofl this subjoctivemss and imprecision this concept is not dispensable at all. The area where VERY seems most relevant is that of expressive evaluations. For example, expressions such as wlanderfid, mrarveiiow, terr$c, cfwfil, and horrible seem to rely crucially on the combinations of the elements GOOD and BAD with VERY ('very good" 'very bad')" In earlier work (Wierzbicka 19721 I tried to d e h e %eq' away via 'morey, along the following lines:

68 Genera/ Issues This is very good (big). = this is more than good (big) and later (Wierzlbicka 1980) I tried to link this interpretation to a performative analysis, as follows: This is very good (big]. = I say: this is good (big) P want to say more than this This anaiysis was questioned at the time by Dwight BoEinger (personal letter), who argued that very good does not mean the same as more than good, and that the expanded version ' Iwant to say mare than g a d does not ensure the desired interpretation. In chilldren%speech (in English), VERY is often realized as so, for example (French m d Nelson 8985: 119):

als "combines material implication with the relevance of a causal relaltion from the protasis to the apodosis" (1986: 96). This amounts to an attempt to define $-sentenoes via a combination of concepts such as 'or" 'and', 't~ue','false' ('not true"), and 'because" along the following lines: E f it rains, I will stay at home. = Either it will rain and bemuse of this I will slay at home or it wilt not rain and because of this I will! not stay at home or it will mot rain and I will stay at home (despite his?). This analysis may seem like an improvement on a purely truth-functional definition, but in my view, it Is not tenable either, If only because it is not the case that "if always implies "because" It is true that 'if' implies some sort of connection between two propositions, and also that a causal link is often involved, too; 1 claim, hiowever, that the 'if" connection is sui generk, and cannot be reduced to anything else; and that a link with 'becauser is not always present. For example, the sentence P f he insults me, I will forgive him. does not imply that I will forgive him bemuse he has insulted me: it is true that I can forgive him only if he has done something bad to me (e.g. if he has insulted me), but it is not true that the insult will be the "cause" of my forgiveness. Similarly, the sentence If he invites me to dinner I will1 not ga does not mean that I will1 not go because he has invited me: if he doesn't invite me I will not go either. It hardly needs to be pointed out that a truth-functional analysis of $sentenoes is highly counter-intuitive, as well as inadequate from the point ef -view of natural language (because of the lack of the requirement that the two clauses should be somehow connected). Since attempts to make if less inadequate by adding to it a causal component do not work, the conclusion suggests itself that-&om the point of view of natural language-this analysis is simply irrelevant and should be abandoned altogether. Insltead, we must conclude that the PF-relation is fundamental, irreducible to anything else; in other words, that it is a conceptual primitive. It is worth noting in this connection that--contrary to what one might expect-the concept of IF appears relatively early in child language, although apparently later than BECAUSE. Here are some examples of $ sentences from the speech of American 4-year-~Ids(French and Nelson 1985: 11&15):
What do you do if you wanina make oatmieal cookies? . . . well, you see, aFter, if you eat your faad up, ya eat dessert.

Yeah, somaltimes ifyooruke s a hungry, go to a restaurant that gives yttl a lot of stuff. . . and then we go to the next door prking lot and it is so cold there and every-

thing. Examples of this kind make it particularly clear that an analysis of 'very" via 3nore"s intuitive/y unteaable ("sometimes if you're more khan hungry .. ."). Since no othier, more plausible, analyses of %.veryyhave been proposed, m d since awumdated cross-linguistic evidence points strongly to the universality of this concept, it seem seasomable at this stage to accept it as a universal semantic primitive.

114. Interclausal Linkers: IF, BECAUSE, LIKE

14.1. IF In logic, c'condjltionds'"@sentenms) are defined in terns of truth conditicams: "ifp then q" is taken to mean that either p md g are both Itrue, o r p and q are both false, or p is false and g is true. It has often been pointed out, however, that this definition does not corresponnd to the use of +sentences in natural languages. As Comrie (1986: 80) notes, according to the logical definition "the only relation that need hold between protasis and apodosis is that expressed in the truth table, so that otherwise totally laarelated propositions may appear as protasis and apodosis, subject only to the condition that they have appropriate truth values, as in P f Paris is the: capital of France, two is an even number." According to Comrie, in natural language (inn contrast to the artificial language of Logic) sentences of this kind are anamdous because In natural language $-sentences require a causal connection between the two propositions in question. Comrie" own definition proposed for condition-

70 General Ismes
(%%'bat do you do at a birthday party?) d then yon go. you do a movie, and then if you have time, yau play, m (For a discussion of the universality of IF, see Chapter 6.) 14.2. BECAUSE According to Kant, causation-with time and spac~onsltitrutes one of the basic categories of human cognition; it is not a category that \ye learn from experience but one af the categories which underlie our interpretation of experience. Data from language acquisition, as well as from cross-cultural semantics, are consistent with Kant's view. The finding that apparently all languages have a lexical exponent of causation (whether it is a conjunction Pike because, a noun like causej or an "ablative" suffix) is particularly significant in this regard, (For discussion, see Chapter 6; see also Wierzbicka 1934b.) From the point of view of language acquisition, too, it is significant that despite the highly abstract and "non-eapirical'kcharacter of the conoept of causality, becawe-sentences appear quite early in children" speech. Mere are a few characteristic emampbs from the speech of American Zyear-olds [Bloom 1991): 1 was crying k a u s e I didn't want to wake up, because i d . was dark, so dark.

But whib the results of studies such as Bloom (1991) do indeed appear to support the view of Searle (1983) and others t h b "we discover causality by experiencing it through our actions and perceptions" (Bloom 1991: 3781, &is is fully consistent with Kamtysview that causality is an innate form of h m a n perception of the world. In is also consistent with the view that causality [or, more precisely, the notion af BECAUSE) is a simple concept, rooted in our subjective experience of %antingy and Voing', and not in any thleoreltical sp~culationsabout "might-have-beens'", along the limes proposed in my Semnfic Primithe8 (Wierzbicka 1972: 17):

X happened because Y happened. = if P hadn't happened X wouldn't tave happened

As I have pointed out elsewhere (Wierzbicka l989b: 3211, while it may be tme that "']iff Mary hadn't met John, she wauldn" tave married him", it doesn't tollow from that that "Mary married John because she had met him", All the evidence leads, then, to the conclusion that BECAUSE is indeed a universal semantic primitive, an irlreducible category of human language and cognition. (For a discussion of the universality of BECAUSE see Cbystpter 6 . 1

114.3. EEKE
The coneept of LIKE can be illustrated with the fallowing sentenms:

Edny blue barrel is inside other barrels) You can't see it cause it's way inside. C384] (going towards disks] Get them cause I want it. (2703 (telling and demonstrating haw she slmps on the sofa) Cause 1 was tired but now I ' m not tired. (271) (regarding TY, which is on) I left it open because I wamna watch it. (339) Bloom (8991) oomments on the results of the study of causality in young children's speech as follows: The mncept of causality attributable to these children's thinking, from the evidenoe of what they talked about, mphasbed the aetions, feelings, and perceptions of persons im everyday causal events, or intentional causality. They discovered leausall connsctions through their own and others' actions or heard thema in everyday discourse about everyday events. Causality for them was neither the 'cement of the universe" that provides the structure of reality nor an inmate quality of the mind that determines reaEity [as per the theories of Hume and Kant, respectively]. Rather, the construction of a theory of causality begins in infancy with the emergence of an understanding a l the regularities in the relation between change and Ithe actjams of oriescll' amd others llnat briang about change.

I did it Iike this: . . . I am not like these people. I think af people Iike you.
The impartanlcie of the conwpt LIKE in the human conceptualization of the world was justly emphasiaed by J. L.Austin in his Senm md Sensibilia (19162b: 7.11, where he wrote:
Like is the great adjuster-word, or alltematively put, the maim flexibility-device, by whose aid, in spite of the limited scope of our vocalbdary, we can allways avoid being left completely speechless.

Conmenting a n Austin's s o d s , T a m r Sovran (1992: 342) writes: The concepts of similarity and its operators seem to have the same function in Banguage and in thought, in the process d croncepts, and in the promss d scientific growth. They help us to l a v e the sak ground of known, Uabeled, cabgorixd terms, and to expand our knowledge and language to newly discmered areas.

E agree entirely with the spirit of these remarks (both Austin's and
Sowanqs). As for the phrasing, however, l would insist that it is Awskim's

2. A Survey o f Sem

"Pike' rather than Sovran's 'similarity' wl~ichis the great "asdjusbr-word'" and the main 'YexibiQity-devicc'" in English, and which a n be matched with other such w d s and devices in other languages. For lexample, a medieval Latin hymn ("Regina coeli laetare") includes the following line: ltcsurrcxit sicut dixil. '14e has risarr as he said."'lile has risen like he said he would.')


15. Introduction
In the last two years, the sysbern of semantic primitives h expanded, from 37 to as many as 55. Despite this sapid expansion, the new primitives offered here for the reader" consrideration: Rave not been proposed lightly. Indirectly, they are the product of many years'tthnking, searching, and experimenting. More directly, they have been born out of careful reconsideration of the whole system and lengthy discussions, particularly those with my co-editor Cliff Goddard, foillowing our collective work on the wdume testing the earlier, m r e restricted set of primitives (Goddad and Wierzbicka 1994b).5: As mentioned earlier, the new primitives haven't yet been extensively tesbd, and cross-linguistic evidentx is vital for deciding their future fate. Their present status must be regarded as qdk different from that of the old primitives, which have already been subjated to extensive cross-linguistic 'Itesting. The order of presentation of the "new primitives" will follow, roughly, that of the old primitives (except, of course, for the areas to which nothing has been added). Thus, I will skut with determiners and quantiiiers, following an with mental predicates and with a section on movement, existence, and life, which can be seen as roughly corresponding to Section 7, "ktions and Eventsm",n Part A. This will be followed by sections on space and time (this t h e , first spaw, then time), and by a section entitled "Imagination and Possibility", which corresponds to the final section of Part A, "Interclausal lnkers"",inally, the most recent and the most tentative of all, the concept of WORD will be briefly discussed-the least solid (at this stage) of all the proposed primitives. The chapter will close with a general discussion and a brief conclusion. The new primitives tentatively posited in this part include the follcrwing:


I imagine that the word sicut represents here the universal semantic p r h i tive LIKE, and yet it could hardly be said to indicate 'sirnilarlty', either in the logical or, for that matter, colloquial sense of the word. Furthermore, I would claim that among "similarity operators" listed by Sovran I(" 'like', 'the same', 'as', and others"), one-'the same'-is not an exponent of the same conmpt "likekt d l , but an irreducible concreptud primitive in its own right. LIKE does have a number of exponents (aQ1olexes) in English, as it does in other languages, and as (in some of its uses] is indeed s good example, but the same is not. In natural language, it is essential for people to be able to make distinctions such as the following one: This fish is like that other fish, but it is not the same fish and cross-linguistic evidence suggests that in all languages people have Bexical resources for making such distinctions. (See Goddard and Wierzbicka E 9946.) The addition of 'like' to the list of primitives (proposed in Goddard E989a) has simplified semantic analysis of numerous aspects d language. n semantic primitives to In particular;, it has allowed explications couched i account, in a simple and natural way, fair the role of prototypes in human language and cognition. In my work, I have tried to use tlne notion of prototype from the start, and, for example, my 1972 analyses of emotion c w wpts or 1480 aaralyses of kinship and colom concrepts were based on this notion (see e.g. McCawley's (1983: 656) comment: "Wierzbicka makes extensive use of prototype analyses")~.But my early lists of primitives were lacking an element which wodd allow me to phrase these anallyses in a simple and natural way. The addition of Yike' to the list changed this. As pointed out by Goddard, 'like' was "a semantically primitive hedge, built into NSM [Natural Semantic Metalanguage], with obvious benefits In t e r n of reducing the length of Lingb~amentali8 style explications, which lean heavily on expressions such as 'can be thought of as'and 'in the same way as' " (l989a: 53). Thus, the addiltion of 'like' facilitated a radical simplification of the syntax of the explications, as we11 as making the semantic account of prototypes, hedges, metaphors, and vagueness more accurate and intuitively satisfying.

The whole new set has been extensively discussd with Cliff Goddard, and has undergone a wmber of ofisioms$Follomiing his suggestions. 14. large part of the new set has also been discussed d d P u Jean Waskina, and my understanding a some asgects of the new system has greatly profited from these diacussiorus. Apart from Jean, I am also indebted lo a nlumber of other collleagues in Canberra, in parthlar, Tim Curnow, lBob Dixon, Nick Enfield, David NasSa, Helen O%oghlin, and Tim Shopen, f am also deeply indebted to my colleague Andmej Boguslawski in Warsaw, despite the: distance, a m unFailling critic, debater, and co-thinker. Last but not least B would like to scknowledlge my indebtedness to my daughter Clare, who discussed most of the new primitives with me ;and who orered many helpful criticism and suggmtioos.



MOVE, THERE IS, (BE) ALIVE 'Won-mental predicates'" FAR, NEAR; SlDE; INSIDE; HERE "Space" A LONG TIME, A SHORT TIME; NOW "Time" "'imagination and possibility'VF . . WOULD, MAYBE "Words" WORD

nize that they couldn't b z fully cxpPicated with a set of primitives which didn't include SOME. In works on languages and logic, the Englnska word some is frequently linked with the so-called '%xistentiall quantifier"'. Far example, McCawley (1981: 101-2) writes:
The so-called ntniversai quantifier corresponds to several difFerent English words: dl, every, my, each; the existential quantifier corresponds to certain uses of the words scrrrre md dm... . The existential quantifier, henceforth represented by the symbol 3, is also the c o m o n element in a number of things that natural languages often distinguish, h r example, various uses of o h m a end some in English. McCawley goes on to point out that the English language distinguishes obligatorily between singular and plural, but that the logi~al concept of existential quantifier is indifferent to this distinction: The formula 0:man x)Xadmire x Hider) is supposed to be true if at last one man admires Hitter and false if no man admires Mitler. It is non-committal on whether exactly one man or more than one admires Hitier. However, English sentences snlust draw the distinction between one and more than one: (a] Some man admires Mitler. (6) Same mcn admire KiUer.

16. Determiners and Quantifiers: SOME and MORE

16.l. Determiner: SOME The English word some is polysemous. The sense posited here as primitive is that of "inderteminate number"^^ illustrated in the fallowing passage from a itext in the Australian language Guugw-Yimidhirr (Haviland 1979: 1 1 6 3 ;Havilland's numbers): (321 Dhana gada-y waguurt-nganh, 3sg + MOM oome-PAST outside-AEL They m e from the outside [i.e. from inland]; (331 gurra buurraay-nganh dhalun-nganh gdmlna gade-y, and water-ABL See-ABL also come-FAST and they also came from the water, from the sea; (34) mundal bubu-wi badi = badiimbarr gada-gr , some + aes ground-~ocunder = underneath come-PAST some came underneath the burface of] the earth; (35) mundal wanggaar = nggarr bubu-wi gada-y, ground-~oc C#Jm&PAw some + ABS above = REDUP others came above the ground; (36) mundal birri wanggaar gada-y. come-PAST some + AES river- above and others came up the rivers. The primitive proposed here is realized in Guugu-Yimidhirr as rntrndal. As Havilamd's glosses show, in English this sense can sametimes be rendered as ~oscrmre,and sometimes as cotker~, E t is important to stress, however, that it is a concept which functions as a quantifier, situated somewhere between ONE and ALL. The quantifier SOME was proposed as a utriverszlll semantic primitive by Boguslawski (19165: 581, alongside two other quantifiers: ONE and ALL. B confess that it took me more than twenty years to come to believe in ONE and ALL [uvhicl~ wcrc also indcpeazdcr~klyposilcd by Goddaod s t ithe Adelaide Woxkshop in 19186); and i t has taken me close to another decade to come to bclieve in SOME. But looking at scnlenccs such as those in the Guugu-Uimidhirr passage (which, as Jean Harkins, personal communication, points out, are very common in Aboriginal stories), one has to recog-

But the clonlsept SOME proposed here as a semantic primitive corresponds only &a sentence [b)above, not to sentence [a]. What is meant codd perhaps be better expressed as SOME OF, as in the fol8owing sentenloes:
Some of them admire Hitler. Some of tihem turned right, and some (of them) turned left. In Polish, the ward cz@C 'part" can sometimes be used as an exponent of this notion, far example: Czgsh z nich paszla na prawo, a czg% na Bewo. Part of them turned right, and part turned left. Similarly, in many other European languages the exponent of the primitive PART can be used as a quantifier, to express the primitive SOME. Curiously, as ClifE Goddard (pzrsosaal alommwnicatioan) notes, the same is true of ehc Australian Ilanguagc Yankunyljatjara. But in many other languages lthere is no lexical lravedap between SOME and'PART. (Moreover, even in English sentences SOME cannot always be paraphrased in terns of the word part. For example, sometimes (of same times) cannot be defined as 'part of the lime'.) B t is important to note that while in works on language and logic some may be regarded as an equivalent o l the verb exist (or the expressiora there is), from a semantic point of view SOME and THERE IS are two different nations, which cannot be reduced to one (see Section 16.2 below). For

examp~e, it seems clear that the Gwugu-Yimidhirr passage adduced at the outset cannot be paraphrascd in tern~s of exBf or rS?erejlr. One could of course say (in English), "There were people among thlem who went under the ground; there were people among them who went above the ground"; but a paraphrase of this kind relies on a relative clause-a stmcture which is not universally available (as well1 as on the expwssion "among them", L c another way of saying ""wmei"')). which Imks swspiicious8y I Similarly, it is true that a sentence such as "Some people admire Nitler* does seem to be paraphrasable (in English) as 'There are people who admire Hitler'". ]But if we try to define away the complex and language-specific relativizer 'kho", on which this paraphrase depends, we have to hll back on some (23orne people admire Hitler'": Same peopUe admire Hitler. # there are people these peopie admire Hitler Finally, it might be suggested that some @nthe relevant sense) can be defined away as '"ot all", along the lines of "Some people dlnire Hitler = Not all people admire Hitler". But a paraphrase of this kind is not vdid, since "not all"-in contrast to some-impilks something like "most". 16.2. Augmentor: MORE The element MORE, included in one of his tentative lists of indefinables by Leibniz, appears on the Eist of semantic primitives not for the first time. 1 tentatively included it in one of the 1989 lists @Vieirzlbicka 1989b: 105), only to replaoe it later with the element MUCH (EUPANQ, proposed at the time and convincingly argued for by Goddard. To have both MUCH (MANY) and MORE on the list seemed intolerably un~conolsnical~ given the close semantic links between the two concepts, and so it seemed imperative to try to define 'much' via 'morey or the other way round. Given the intuitive closeness of the two concepts it is oertainly worth trying n o rcduncc tlrcun n o om. In my judgcunct~t, hrawcwcr, none of thc aiLempLs undertaken in the paslt were really convincing. If we want to try to define 'mnucla'via "more' the obvious way to go is to refer to some expectations, along the following lines: much (many] = more than one couIdlmrowPd expect But this approach, reasonable as it may seem at first, is not always convincing. Far example, in the sentenoe Many people came w( see the Pope), but not as many as expected. the word snmy can hardly mean h m o than expeckd'.

In the Moscow semantic tradition, the key word used in this and many other similar contexts was ""norm" (see e.g. ~olkovskij 1964h):
much (mamy) = more than the norm

But the word norm doesn't always make sense in sentenoes with much or many. For example, the sentence Many people are afraid of lightning. could hardly be paraphrased in terms of the phrase ' k o r e than the norm1'. Of wurse it could be argued that what. was meant was not the ordinary Russian word norma @om') but an artificial word with a different rneaning, but it is mot dear what exactly such a statement would mean or how it could be verified. On tlne other hand, if we try to define b o r e ' via 'rnucUmanyq we run into other difficulties. At first sight, the approach which appears to work with other comparatives seems to work here as well (see Section 9 above):6
A is bigger than B. = if someone thinks of these two thing at the same time this person can think: " A is big, B is not big"

But there are many situations when a paraphrase of this kind would not work for "more''. For example, if I say that I want more to eat, a paraphrase dong the lines proposed above doesn" seem to make sense. Similarly, the sentences:

H want to say more,

I want to see more. 1 want to h o w more about this.
can hardly be paraphrased in the '"these two things" format. It is not a comparative kore', then (a converse of "less')), which ]I am positing here as a universal semantic primitive, but, so to speak, an "aaugmentative" one, illustrated in canonicaE sentenoes such as

E want mow. Give me more. I want to s~eknowhear more.

An arnalysis along these lines, which was proposed by CBiflGaddard (personal c o m u n i cation), is simpler, and, I think,bdter, than the fo1lowing one, which I proposed in Wiembicka (i971): A is bigger than B. = if peoplle can say about B "it is big" they cam say the same about A fc a n ' t say: "if people can sag this ahoolpt A they lcanu say the same about Bn'

When the presence of lexical exponents d MQRE is cross-linguistid1y tested, it is probably worth including questions about "less' as well. At this stage, however, 'less' seems to be a much less likely candidate for a lexical universal than MORE. I would expect that many languages will be found which have a word for MORE [in an augmentative sense) but not for 'less'. The study of language acquisition strengthens this expectation (based on internal semantic grounds], since first, children start using the word more, in contrast to less, very early (see e.g. BSralne 197163, and second, those early uses of more are augmentative, not relative. As, for example, Johnston (1985: 9743 put it: 'caltihough we think of more as expressing judgments of relative quantityiextent, the child's more is at first non-quantitative and non-comparative"".As shown by Braie, a combination of more and a word designating an objject of desire r(e.g. '"ore juice!"")is in fact among the most common early two-word utterances in child language (Brahe 1976; McShane 19911, whereas r'rt~s does not appear on the list of the early'twoword patterns a t all. Boweman (8976: 128) notes that her daughters Christy and Eva initially used Ithe word more "'in connection with a restricted set of objects at first-food and drink" and that '%loom's (11973) daughter Allison likewise first p r o d u d . . . 5noreks a request for an additional serving of food or drink, although within only a few days she began to use these words across a range of more varied contexts." Reflecting on the apparent aspmeltry between the concepts 'more'and Yessklooe is tempted to think that perhaps there is indeed some special psychological link between the concepts MORE and WANT. AS we know them, human beings are perhaps more indined to think, and to say:

want want want want

(to have, to eat, to drink) more. b see more. D o know more. to say more.

The concepts 'see'and 'hear-play a fundamental role in human communiation. As pointed out by Eoweman (1976: 1381, in transcriptions of two-word chiid utterances from diverse languages one frequently finds sentences such "here (there) ball", or "see man". Generally speakas "this (that] doggiey*, ing, the word see (alongside this, that, here, and there] appears to be one of the basic cromwicatiw tools in early interaction between children and adullts. At that stage, the word hear doesn't seem to be nearly as important as see, but appasemtly, before long, it too begins to play a special role, alongside bok, listen, and watch (see e.g. Bloom er at 1975). The concepts 'look', "isten', anal! %watchy are complex, and involve 'wanting' as well as 'seeing' or 'hearing' (I"wantingto see' and "anting to hear"; but 'seie'and %ear9themselves cannot be similarly decomposed into simpler concept^.^ Admittedly, in earlier work (see Wiembicka 1981011 B have argued that SEE and HEAR (as well as 'rinmellknd 'taste'] can be defined via the corresponding body parts: eyes and ears, nose and mouth. To see and to hear, E claimed, means to h o w something about something because of one" eyes or one's ears. But there are problems with this account. First of all, it presents the motions of 'see' and 'hear' as very complex, and this is hard to a m p i in view of tbe role these dements play in many areas of lexicon and grammar of many languages (such as, in particular, "evidentials"; see Chapter 15). Second, if we define 'see' and 'hear'via 'eyes' and "ears', we cannot define 'eyes' and 'ears' via 'see' and 'hear', and we have to adopt purely aamtomiml definitions, along the following lines: eyes - two pasts of the face in the upper part of ltlw face ears - two parts of the head, on both sides of the head But although in the past 1 tried to justify such definitions mysclF (Wienbi~ka19XO), my waders and listeners always Found them unsatisfactory, because they felt that 'seeing' and 'hearing' was an integral part of their meaning. I would mow propose, then, defining 'eyes' via SEE, and 'ears' via HEAR, as follows: eyes - twa parts of the face these parts are alike because of these two parts a person can see

than to use tlze corresponding sentences with "less'. It is also worth noting that VERY-another quasi-quantitative concept-luas no universal opposite either. Finally, I would Bike to suggest that the augmentative element MQRE plays a crucial role in o w understanding of numbers. For what is 'Vhree'" if not "one more than two"?

17. Mental Predicates: SEE and HEAR

See babyr'See prettyJSee train. (from a 2-year-old's first word-combination list; Braine 19%: 70)

his own list of universal sernanlia: primitives, he would include in it "see" ';ad "hear'" 1 have now reached h e same conclwinm.

' At a Lhnguiskhc Forum held at #heANU in 198%4,Bob Dixon said that if he were to pro-

&O General Issues

ears - two parts of a person's head these parts are alike because of these two parts, a person can hear Third, while "eebnd % e a r b a y seem to be notions derived from sensory experience, and therefore unlikely to be either universal or innate (because experienw is miable), in fact they do not have to be viewed in that way. A person born blind may still "sae'homething (images or colours) in his or her mind, and may therefore have an innate nation of 'seeing'. Similarly, people deaf from birth may still "hmr" something i n their heads. It is interesting to note, for example, frequent references to "hearing' (sometimes with, and somethes without inverted w m s ) in the autobiography of a man who describes himself as totally deaf (Wright 1943: aiei 1):
I do not live r im a world of complete sitenm. There is no such thing as absolute deafness. Coming from one whose aural nerve is extinct, this statement may be taken as authoritative. . . . If I stand on a wooden floor E a m 'hear' footsteps behind me, but not when standing o w a floor made of some less resonant suhstanoe-for examy own voicle. This is not sutprisple stone or concrete. I caw even partially 'hear' m ing, for people hear themsekes talk mainly by bone-conduction inside their heads . .. Likewise, 1'bear' a piano if B place a finger o m it while it is being played. . . . I cannot hear wind-iwstnzhnnents (flute, bagpipes, oboe). . I itniave a passion for military bands, thou& haring little except the drumtaps, a sad boom-thud from the big dmm and a clattering exhilaration from the kettledrums.

Dehitions linking seeing and Rearing with knowledge, along the lines of: to see - to know something about something because of one's eyes to hear - to know something about something because of one" sears may seem plausible because they are consistent with a wide range of contexts these words occur, but they cannot be said to capture their semantic invariant. For example, when one sees s mirage in the desert this could hardly be interpreted in terns of gaining knowledge about something. Of course all auditive and visual experiences [including ringing in one's ears and seeing colourful dreams) can lead, indirectly, ito some knowledge i(e.g. about one's health, or about omne's sconscious desires), but this is not what sentences about such experiences mean. Furthemore, an analysis of a 2-year-old's utterance "see pretty" as '1 want yodsomeone to know because of your eyes \that there is something pretty here' is hardly convincing: apart from its precocious complexity8the baby wants someone to k e y ,not to 'know'. It is important to add that the approach to SEE and HEAW proposed here does not extend to the other senses. The supposed symmetry between the human "five senses" is spurious, and from a universal perspective there is no such thing as the human '"ve senses". Beyond '%edng" and "hearing", different languages draw their distinctions in different ways. As Classen (1993;: 4-21 points out, '"n the West we sure aaocustomed to thinking of peroeption as a physicaj rather than cultural act. The five senses simply gather data about the world. Yet even our time-honoured notion of here being dive senses is itself a cultural construction. Some cultures recognize more senses, and other cultures fewer.'" Admittedly, Classen goes on to say that "the Wausa of Nigeria divide the senses into two, with one term for sight and one for all the other senses"". This doesn't mean, however, that the Hausa word which stands for 'hear" as well as a l the other senses except %eev is not polysemous. PoQysemic patterns of this kind we common. For example, in Russian the word sSy$otP can stand both for 'bear' and >melll"see the Academy dictionary of Russian: Akademija Nauk SSSR 11961, iv. 204). Whak matkrs, however, is mot the term as such, but the term combined with a particular grammatical frame. The hypothesis that SEE is a universal semantic primitive is consistent with the view widespread across cultures that there is a special relationship between seeing and knowing, and that eyewitness evidence is more reliable than any other kind of evidence, and the hypothesized status of HEAR as a universal semantic primitive tallies wlP with thet special role of vocal speclu in human communi~ation:while SAY applies to both vocal and other signs, audible messages play a more important role in human


Whether or not all1 languages have separate words for SEE and HEAR is not always self-evident, bemuse in a number of languages both SEE and HEAR share their lexical exponents with other concepts, notably wililh KNOW and THl[NK. But (as argued in Chapter di), lexica! overlaps of this kind can be shown to be due b polysemy, Apart from such common patterns of polysemy, to my knowledge no language without words for SEE and HEAR has ever been r e p ~ r t e d . ~ The common poiysemic patterns involving SEE or HEAW on the one hand and KNOW on the other are of course not accidental. They point to conceptual Pinks, but I would argue that these lli~ks are not compositional.
A particularly interesting case OF pdysem involving SEE and WEAR lhirs k e n repnted by Sasha AikhenvaPd (personal urmnnunimtion). In the Tariana language (from the Arawak family, spoken in Brazil), the same verb is us& for both SEE and HEAR, but in the HEAR sense it requires an objacl which implies an 'kudikor)r'~Dlbjeck ('words', "sounds', 'l~armguage', ekc.) In Lhis language, the sentence of the foron '1 Verb(se&eau] a bird' can mean either ' I s s a bird' or '1 hear n bird*.But it is also possible to say, using the s a m verb, the eqalwivatent of V hear a bird but I. don" see iinhor "I see a bird but I don't hear it'. To do this, m e woukd use sentences of the fallowing Som: 'I Verb[seeihear) C h e voice of a bird, but I don't Verb(seen7ear) it7,or 'I Veub(seefiear) a bird but 1don't Uerb[seerZleau) its voice'. Pn my view this fact shoes that the verb in question is polysemous [unlike, for example, C h e English verb

82 General Issues societies than other kinds of messages; and spoken languages are not on a par with other semiotic systems. Tll~c fact ltliat 0 3 . all tlrc scl~scs ouly S83E :toad I-IEAR strc grsl~~mnaticaPiizcd in the category of "evidentids" (see Chapter IS) is another reflation of their suecia1 status in human cognition, as is also the fact h a t 'Lvisibility.yE' ---..-r is often encoded in the systems of demonstratives. One way to characterize illis dillerc~vz bclwecn Llm mnccpu 'see' and 'hear' on the one hand, and 'smell', 'taste" and 'touchhn the other, is to say that 'see' and 'hear' are, essentially, mental predicates, referring to events m d processes which do not rely crucially on the body, whereas 'smell', 'taste', and 'touch' are, essentially, "sensory" predicates, referring to experiences which do rely, crucially, on the body. This difference is reflected in the fact that it is perfectly natural b attribute 'seeing' rundl "earing'-but not smelling, tasting, or touching-to God. For example, it is perfectly natural to say that God hears our prayen, or that he sees our hearts and indeed o w actions; but it would sound ludicrous to say that he "smetlsy'something. Thus, we can conceive of 'seeing' and 'hearing' in a more abstract, less physical way than we can mnccive of 'smelling' (or 'tasting' or 'touching'). This is consistent with the 'hypothesis that %seeQ and 'bear', ,in contrast to 'smell', 'taste', or 'touch', are conceptual primitives. caUy related to 'change of place', and therefore muld not be regarded as elementary. If, however, we allow elementary concepts to be mutually related (in non-mmpositional ways), then my argument can no longer be regarded as valid: if both 'I' and 'someone' ('person') can be regarded as elementary, despite beinl intuitively related, so earn 'movementa and 'place'. Furthemore, the notiom MOVE, which I am now positing as a semantic primitive, is not necessarily linked with a pssage of some object n pers m from one place to another. The prototypical examples of MOVE in the intended sense can be found in sentences such as the following ones:

I see something is moving (in this place). I can't move. Samething moved inside me.
ntenccs of this kind, the idea of 'change of place' is not necessarily relat all (wen if it is true that whenwer something moues, something anges place, if only momentarily). Similarly, if we wanted to say that someone shivered, or that someone's lips trembled, it would seem rather ludicrous to try to paraphrase such sentences in terms of a repeated change of place. On L e other hand, mnccpts such as 'go' or 'walk' do imply a change af place, but they also h p l y movement, and their explications would have to include both PLACE (WHERE) and MOVE.

18. Movement, Existence, Life: MOVE, THERE IS, LIVE

18.1. MOVE The idea of 'movement' or 'motion' was put forward as indefinable by John, Locke, who mocked attempts to reduce it to other concepts: Nor have the madern philosophers, who have endeavoured b throw oB the jargon of the schoolls, and speak intelligibly, much better succeeded in defining simple ideas, whether by explaining their causes, or any otherwise. Tk atomists, who define motion to be 'a passage from one place to another,' what do they do mare than put one synomymous word lor another? For what is pmsoge other than motion? And if they were asked what passage was, how would they 'better define it than by motion? For Is it not at least as proper and sipifiwnt to say, Passage is a moltion from one plam ta another, as to say, Motion is a passage, kc.? This is to translate, and not to define, when we change two words; of the same significallion one for o dis~ower another; which, when one is better understood than the other, may serve L what idea the unknown stands for; but is very far From a definition, unless we will say every English word in the dictionaw is the definition of the Latin word it answers, and that motion Is a definition of molus. [Locke t64W1959:35)


18.2. THERE IS artesians regarded it as self-evident that 'existence' fin French IJexQ1ence) as among those ideas which are so clear that no definitions could make any dearer. h r many years, I have rejected this view, in the belief 'existence' could be defined in terms of, so to speak, "possible refer: ' that is, along the following lines (see Wierzbicka 1972, 1980): There are no unicorns (ghosts, Mack swans). = ane can? say about something: "'this is a unicorn (ghost, black swanl" ses of this kind never seemed to me quite right, but I bdieved that e they could be amended in some minor way and thus be rendered

In Lingua Menlalis wienbicka 1980: 51, 1 rejected Locke's claim and argued (following Leibniz 11651198L: 297) that 'movement' was semanti-

But after more than two decades of trying, and failing, to find the necary "minor amendments", 1now believe that the time has come to give any attempts to define 'existenoe' away and to recognize that the Lesians were right on this point as well (see e.g. Amauld 16631964: 66). erhaps the only qualification which I would make is that both the noun &fence (French exiFrenee) and the verb lo ex& ( F m c h exhter) belong to sophical, not everyday, vocabulary, and that it is more justified to nate the expression there idare (in French B y a)l as the basic lexical

$4 General &sues

iwas 85 xistence' is expressed by means of the definite article (in a verbless senena), as in the following examples (C stands far connective palrlticle): Mala-na-mulmulm, season-c-hunger 'There was famine." Pata taina, a tava parika. no s d t ART water at1 "ere was no salt, only water."

exponent of the primitive in question. The point is not trivial, because the difference is not merely stylistic. Apart from stylistic difkrenoes, in contexts, exLt m d here islare may seem to be interchangeable, b others they are not: (1) "&ere are no unicorns. = Unicorns don? exht. (2) There are no cockroaches here. + ?Cockroaches don" exist here. As sentence 2 illustrates, the verb exist does not co-occur with phrase. It is a verb usled to make absolute statements, staiarements world as a whole, or about whole classes of entities rather than a viduals or groups. By contrast, the concept THERE IS (ARE) ca both to the world as a whole and to specific individuals in spwific

Ma arrslana a vaden parika. and Ebmerly ART women all 'In former times only women (existed).' e fact that in other types of sentence the dement a means, roughly ' [&at is, 'I think you know which one I an;r talking about') idate the observation that in verbless sentences of the kind above it means 'there is'. Nor does the grammatical status of the eletituent of the noun phrase) diisqualify it as a lexical expoimitive THERE IS: there is no reason to expect that this tual element should always be lexically encoded as a verb or a verb
rn language acquisition are highly relevant in this regad. Far te devdopment, kxistence"in the sense postulated here as fact one of the first concepts emerging in infants' speech. st early realization of this conoegt comes in the f o m oF one-word ining 'existenw' with negation, such as Walllgone" and, at a stage, with two-word combinations such as "'milk alllgone" (Boweman mom an adult point of view, it might seem that utterances such are even more complex, and that they express disappearan~e' ' rather than simply 'there isn"' (Boweman 1976: B2$], but i("disappearance' and 'cessatioao"~could, arguably, be ideas implicit in tuation as interpreted by the adult. But the meaning of 'non-existence" 't') is clearly there. In any case, whatever the meaning of dlinfancy, coniextualized sentences such as "no pocket" (said by ng a pocket i n Mmmy's shirt, which had no pocket; leave lilttle room far doubt. arly emergence of cnno-existence' sentences in child laneage matches, in an interesting way, clear marking of "negative exislnceYn those languages which don't have a verb, or a verbal phrase, for 'there is' as such (such as Tolai; cf. the "no salt" example cited learlier). I am not suggesting that 'non-existence"IC"lilnere isn't') is expressed in child language earlier than %xistence"I("there is')), but only that it is expressed at a very early stage more cleady [because it is not open to

(3) There are no ghosts (ghosts don't exist)^. (4) There are no ghosts in this place (?ghosts don't exist in this place).
In acoepting THERE BSIARE as a semantic primitive, and in h e English expression there hiare (rather than be or exist) as it exponent in English, I am folllowing an idea put f ~ r ~ aby ~Cliff d at the Adelaide Workshop in 19816 (Goddard 1986a). The posed here corresponds to what Goddard called a "presentativ ~onstmction~'. If the proposed primitive ALL approximates logicians"'"mnivers tifierv'",he proposed primitive THERE ISiARE approxi "existential quantifier"". As we will discuss in detail In C have sometimes been expressed as to the availability of these in all the languages of the wodd. While I will leave discussion of the a p t ALL for Chapter 6, I would like to suggest here that in fact have a lexical counterpart of the English fhere i8hre counterpart may of course be homophonous with the expo other meaning or meanings, but if so, then we can confid basis for establishing polysemy (such as different g r a m linked with the different meanings). For example, in Polis other languages; see Verhaarr, 19166-73) the comept THERE ISI(kR expressed by means of the same verb which (in a different grammatical c stru@rtion)i serves also as a copula. The hypothesis that a11 languages have a lexical exponent of the con THERE IS should not be misconstrued as a claim that every language a existential verb, or a verb phrase comparable to lthe English phase t is. For example, in the Austronesian language Tolai CMlosel 1984:

pears, however, that this is not universal. Qn the other hand, the commonth-old Julie Bates (Bates e E arl. 1988: 252): There's a cleaning lady there. k t an earlier stage, 'there isvis frequentjy expressed in infants' spe means of a two-word combination, with the words there or here inr t nd, there are conwpts relating to death, such as klie', 'kill', 'murder', ny', 'l~eswrection~, 'imartality', 'reincarnation', "corpse; 'stillborny, so on, all referring in their meaning to 'living'. For example: Turtles live for a Bong t h e .

hypothesis that the pattern '%ere/there + X' was used to show or to d alttention to things, indicating their presenoe or existence"'. Thus, far from being the philosopher's brain-child, "existence' (t 'there is') is in fact something that "comes out of the mouths of bab

At this time he died. = at this time something happened to him before this, he lived (was ative] after this, he didn't live (wasn't alive)

118.3. LIVE (ALIVE)
anvw. Do you think the fire w~wlldlisten? CHILD. NN Fires aren" alive, silly. (Kulczaj and Daly 1979: 575)

m an animal as "vvliens sentiens" ('a living thing, a feeling thing" they k what they were doing: one cannot d e h e (in simpler terns) the concept

To begin with, there are conmpts relating to buman age, such as '01 'young', or kbild': These people are old. = 'Fhrese people have lived far a long time. In English, the word live takes also adjuncts which describe places and camtitions of life; for example: These people live in the desert. When I was young, I lived alone. For many yeays, t lived in poverty (in constant pain).
RtececalP also McCawley's (1973: 1571 dehition '90 kill = to cause to become not alive". In C i n g ~MenraSis l~ (l980: 168-91 I argued, contra McCaaley, that 'dead' is semantically simpler than 'alive', but 1 now think that he was closer to the truth, rm this point, than I.

C'anhals", in the al-inclusive sense) - living things these things can fwl something these things can do something - living things these things can't feel anything these things can" t o anything - thhgs, not living things people make these thingsPr" these things have many parts when people do something to these things, some of these parB can move
In the present s y s t e m of primitives, 'hake" is not regard4 as e primitive, and so it is k r e as a semantic molecule. The relationship between "make'hnd DO requires further


2. A Survey o f Semmtic Primitive8 $9

1. The march fly is a big fly that sucks blood from people and from some other


- things, not living things

these things are like many things these things can do things as if they could think these things can't feel

living things. The rainbow snake swallows men, women or children alive. . PI green tree is one that is still alive. . The life of a butterfly (a book title)^. 5. This is the story of my life. 6. That old man's heart stopped. Those women are thumping (his chest] to bring him back to lire.

119. Space: FAR and NEAR, SIDE, INSIDE, HERE


Fourth, there is the mystery of people who are in a coma, o or other arliiicial life-support systems, and so on. These peo don't do anything, and, as h r as we know, don't think, don't feel, know, and don't tvmt anything; and yet hey are considered "a Clearly, from lay people's point of view there is a mystery here: couldn? explain what they mean by "alive" i n such cases, and yet the that they sonnebow know what they mean. In hct, fmm a scientific of view, too, a mystery is involved here, a mystery which has minds d many scientists over centuries. The constantly changing criteria of life and death point in the w e direction. Last but not least, in many cultures people talk a great dea the lives of individual people, and human life in general. Li life-is one of the: main subjects of folk philosophy in a wide variety tuses. It is d f i c d t to see how this important area of h u m n disc be understood if we didn" t a l l have a b a i c concept of 'living'. To qu one example of such discourse [from Ewlesia~tes 41: 3-5):
3. This is an evil among all things idhat are done under the event unto dl: yea, also the heart of the sons of men is ness is in their heart while they live, and after that they 4. For to hiun that is joined to all the living there is hope: For a living dog is be

ter than a dead lichn. 5. For the living know that they shalll die: but the dead know not any th ther have they any more a reward; for the memory of them is fogotten.

mulct one try to define away tbe concept of NEAR? e first temptation would probabiy be to try to reduce NEAR to FAR, f "near (close) =not far". ]But this won't d o c i f only ruse NEAR can combine with VERY, whereas "not h r Y y cannot: (very wry close, "very not hrjr. (One can say, ofcourse, "not very far", but oesn't mean the same as "very ciose".)~l would be to try to represent both NEAR and EAR etical primitive "distanceY"near = small distance, h r ~ " an abstract sense, covering both FAR But ' L & ~ l t a n(in irersal Iconcept. We emnot be sure, without serious or not a11 languages have words for FAR and te sure, however, that we d 1 not find in dl languages ing to dklmce. In fact, even in English distance is a fairly md, which is not found in the everyday speech d a 1 From the point d view of everyday Pangwage, distance artificial creation, forgad (so to speak) on p u p w e to cover two ay concepts: FAR and NEAR. re, the word distance is probably chiefly necessary for talkkg rs and measures, But the wordsfar and mear (or C ~ Q S Q are ] not arily, with numbers and measures. They are fairly vague wpression, not an accurate assessment. Consider, for h g words from a Chistmu carol [Horrobin and

The whole range of the universal concept LIVE (ALIW] as envisag here can be illustrated with the examples adduoed in the r a n t dictionary of the Australian language h r e m t e (Henderson and Dobson 1994], where the word itethe is glossed as "1. alive, living'hd "2. lifeq".I will only q u ~ f the English glosses of tbe Anemte sentences:

omre languages, the word far NEAR may look like a combination of negation and for FAR Cwt as the wardl for BAD may look Bike a amhioation af megation and for GCIOD). It is always possible, however, that mmbinabiPiry tests will show that pathetical 'not-Far' word wjB1 m a m NEAR, ratheher than NOT + FAR Qust as a "netrd may mean BAD rather than MOT + GOOD).

Be near me, Lard Jesus; I ask You to stay Close by me for ever kmd love me, I pray.

It seems unlikely that anyone would want to paraphrase phrases like '%e near me" and "close by me" via "smaEl distance". h rclarcd point is that botln FAR and NEAR appear io crnbody a mrtain point of view: normally, '"it is h r fram this place to that other placep' rather than "between two places'" by contrast, the word distil~meeimplies that the speaker is not mentally associated with one place more than with the other, and so one speaks, normally, about the distance '%batween" two places rather than ''from" one plaae "toY\nother. Furthermore, FAR and NEAR appear to suggest a different perspective: while both refer to two places it is far from A to B, it is close from A to B), FAR seems to be more particularly "far fromm,mdNEAR, "near W"'. Thus, while one m a say both: L it far from A to IUP"P ?Is it close fmrn A to BT one can only say: She lives far from us She lives near (to) us, not *She lives far to us. %he lives near from us.

comparatives can be deffinad in the same way as ali the other comparatives Section 11 16.2). Finally, why can't FAR be defined via "long'>ilher as "a long way" or via "a long Lime"? h e n %tthe sentences A, B, and C below quite close in meming?


(A) P s it far from here to Tipywerary? (($1 Is it a long way from here to Tipprary? r[C) Would it take a long time to get from here to Tipperary?
But k s t , tlne expression "a long way9%as, so to speak, an "allalive" perspective, not an "ablative" one, as the phrase "a long way toq9hows: It is a long way to Tipperary. ?It is far to Tipperary. phrase "a long way to" reflects the point of view of someone who is bout the destination, not about the point of departure. zu. decomposition d f a r into 'lmg' and 'way' wouldn't take us ecause it only generates two further questions: what is "long" and at n s way"? Without independent definitions of "long" and way^', even if we managed to convince ourselves thatJir means the same as a hng way, this codd mean only that the expression a bmg woy is an urnanallysable lexeme (an allolex offas). An analysis cl?f Tar' via 'a long time'cannot be accepted either. Under certain conditions, journeying-time may provide a satisfactory answer to the question "how far?" but generally speaking, the two questions "how far is ilt?"'aasd "how long would it take to get there?" do not mean the same. For example, the sentence

It seems, then, that in English at least, EAR has an inherently ''ablative" perspective (FAR FROM), and NEAR, an inherently "a1Plative" m e (NEAWCLOSE TO). Furhemore, it seems lihly that FAR is, k s t of dl, "far from here'"cTar from this plaoe"") whereas NEAR is, fint of all, "near toyE (see "be near me, Lord Jesus"). This differennoe in perspective i("alblativealversus "alllativeYe'") offers additional suprprort for the view that l?AR and NEAR mnnrot be reduaed to some unitary concept of 'djlstanae'. A third possible approach would be to try to reduce both FAR and NEAR to their respective comparatives, along the following Iines:
far = further than one would expect near = nearer than one would expect But how would one then define the comparatives "further" and "nearer"? lPresumabR4ly, in terns of ""greater or smaller distance"".But this would bring us back to the solution which we have already considered, and rejected. Om. the other hand, if we accept FAR and NEAR as semantic primitives, the

It would take a long time to get fmm A to B. say really imply that it is far fmm A to B (e.g, paaces A and 5 codd be arated by monnitains). Similarly, the sentence How far w i the sun from the earth? doesn't mean the same as How long wodd it take to get from the earth to the sun?

T conclude^ then, that both NEAR and FAR are, in all probability, miemantic primitives. The fact that in many languages the conciept of appears to play an important role in demonstrative systems provides d~tional evidence for the importance and ""bsicness"d this cclncept. T would add, however, that by tentatively positing both these elements as primitives I do not mean to suggest that they are hlly symmetrical, and, in some sense, perfect "opposites"'.

2. A Survey

01 Setnacr~fic Prirtsifivcrs


In the literature on language acquisition it has often been mentioned that the spatial notion of 'proxirnity'(see e.g. Johnston 1985: 969) or kontiguIly"Slobin IgtJSa: I Iffa) emcrgcs very early in child lanyuagc; and the terms "proximity" and 'kontiguity"', as well as "beside" (e.g. ibid.), are used interchangeably in those discussions. The generallimtion that the basic develop ment order of "locative notions"is "in/onyy c <'under" <"beside"ibid.) is based on research In which the label "beside" represents {for English] a series of expmssims including beside, by, b ear, nexf toJ and the to (Johnston and Slobin, 1979: 5341. For Italian, the label 'beside' represent8 the expression vicino q 'close to' (Clark 1485: 145). This means, however, a m that what has been described as "the notion 'besidey" may in k t c sgond better to the proposed primitive NEAR. On the other hand, &en the early emergem and the w e of the conoepili 'nearibeside9n child language>the proposed primitive NEAR should perhaps be seen as referring, primarily, ta a relatioar between people and thiags (Xis near tobext to I?) rather than to a d i s b o e between places [Bike FAR). Slabin's (19&5a: 1180) observation that "MI crosslingIlistie aoqaisition data point to an initial salience of toplogical nations of containment, support amd contiguity" a n perhaps tre related to the semantic evidence supporting the notions INSIDE Q"kontainment"")arrd NEAR TQ ("lcontiguity"") and, perhaps, also to a possible primitive 'an' or 'touch' C%swpportJ surfaoe"), not i n d d e d in the present system. 19.2. SIDE (ON WHAT SIDE) The concept of SIDE [suggested as a possible primitive by Goddard, personal co~lwnunication) is crucial for people's spatiat "@rI"iientatlonY" A kundamental frame of referenoe for spatial orientation is provided by the human body, with its basic four sides, organized, conoeg~itudly, in the farm of two pairs:
[l) on the right-hand side ("'on the right side"') on the left-hand side ("'on the left side") (2) "in front*' "behind"

[B) i n fmnt of me = on one side of me

1 can see things on this sick behind ime = on one side of me I can%see things on this side


The idea that the concepts 'front' and "behind' we based on the notion of 'seeingYis consistent with the finding (Johnston 1 1 9 8 5 ) that '"he first uses of behim! i n English refer only to a smaller object totally hidden From view by a lager object" (Slobin 19856: 11&I@ES. Qf the two body-centric orientirtional pairs cfront" versus "on the right" versus "on the Ieft"'), the first one, "frontY'versus" b a ~ k ' ~ ~ ~ appears to play an important role in ail languages and cultures, whereas the second one, "on the rightJy versus "on the 1eRmJ ih s ,more restricted as a frame of orientation (see kvinson and Brown 1992].83 If it is true, however, that the "body-oentric"' ideas of 'frontband 'backbre universal, this fact by itself supports b n t ' s (and Vice's) tenet that the human body provides an important frame of reference for human interpretation of spae.14 Another important frame of reference is provided by the natural environment, and, i n patrticuilar, by the sun. On analogy with the four sides of' the body, the natural environment, too, appears to be almost universally interpreted in terns of four sides. If in the human body the four sides are distinguished with reference to the right hand and the face, in the natural environment the basic reference-point is provided by the sun.15 Here, too,
a T o s p k d a person's front as a "side" may seem wunter-intuitive, even absurd, because with mfenmoe to a pepson's body we normally speak of only two sides: the right side and Ithe 3eR side. But this is d m to the polysemy of n P 1 e ward side in English. For example, in Polish, the two U ~ i d e of ~ "a prson's body, that is, the right side and the left side, are called bok& whenas side as a semantic primitive is expwssecY by the word srrona [more precisely: ON SIDE X = PO STRONlE X , where stroaa k used in the locative case, and p c p is a preposition). (For a definition of 'face', see a a p t e r 3.) l3 Levimam aad Brown [1991] question the importance of the conoepts 'in front of and 'behind' in the cai~eeptualizationof space, referring, in particular, to the Australian language GUM&-Yimidhirr:"Instead of notions like 'in front of', 'behind", 'to the left of'%'opposite', ek., which coocepts am uncoded in the language, Gwgu Ilirnidhirr speakms must specify locations as l C j n rough English gloss) 'la the North of, 'to the S o u t h OF,'to the East o r , etc." However, Mavilmd's (1979: 179) bask vocabulary List af this language d m include a ward (in fact, two] glossed as 'in front'. l4 For a different view, we Levinson and Brown (IW2). The importance d the four sides of the body dy a basis for spatial orientation was m t l y disputed by Levinson and Brown (1992). who write: "Kant was wrong to think that (he strc~ctwre aTspaaial regions Sounded on the human frame, and in particular the distinctions based on left and right* are in some sense essential human intuitions." A ~ounter-exampleto Kmt's theory is provided, a m d i n g to Levinson and Brown, by the Mayan language Tzeltal, of which they say: "It i s true that they [the Tenejapans, i.e. the speakers of Tselltal] have names for the IeR hand and the right hand, and also a dean for handhrm in general. But they do not generalize the dislinction to spatiat regions-there is no linguistic expression glossing as 'to t b left' ar 'on the left-hand side' or the like" (1992: 3. If this statement was accurate then SlDE couldn't be a universa0 semantic primitive; for if Tzelnal bas words For the right hand and K l h e left hand [as Levinson and Brown tell IBS), and if it also had an expression meaning


For the concepts of 'in front' and 'behind'ttwo alternative analyses (see PI and IP1 below) can be proposed, both slnpporlled by widespread patterns of pdysemy: one based on the concepts of 'faoe' and 'back'or 'khi~ld'(as names of body pasts) and the other based on the concepts of %efosehnd 'after': (A) in front of me = an the same side as my face behind me = on the same side as my bacWbehind

44 General Iss~ex the Four sides are divided, conoeptually, into two pairs. Thus, for "east" and "west" explications along the following lines can be proposed: I G the east side =every day people can see the sun on this sidic: before they @an see it above them the west side = every day people see the sun an this side after they see it above them

2. A Survey o f Semantic Primitives 95

(B) your left hand is on one side of you she is sitting on the same side, very near [to you]

Like the four sides of the body, the four sides of the world, too, are widely used as a frame of reference for 'brientation"To illustrate (from the Australian language Yir-Yorormt; Mphw 1991: 1.654):
An kawa nhilin. "she is sitting just to the east here' Ejilnst off to your left] This can bNe explicated along the following; lines (where A refers to the YirYoront sentence, and B, to the English gloss):

As this example iflwstrates, in Australian languages ((asin many other Banguilges of the world; see e.g. Levinson and Brown 1991; Hadland 19911, the natural environment [especiallly the sun) plays a more important role as a f r m e of reference for spatial orientation than it does in English and other European languages. At the same rim, the universal or near-universal division of the world into 'Tour sides"(two lpairs of two sides) is undoubtedly modelled on the "four sides" (two plus two) of the humm body.
19.3. INSIDE
YOUcan" t it cause it's say imside. (from a 2-yam-old; Bloom 1991: 3841

(4 every day people can see the sun on one side before they can see it
above them she is sitting on the same side, very near [to here]
b n [this, one, etc.] side o f . . .'" t h e m presuma"olg there would be no di%dty in putting the two together and constructing expressions meaning 'on the side of the tjght hand' and 'no the side of the bft hand'. 1awpnect that tlah indeed is the case, akthough, nlveedkss L o say, tlne matter rquires verifficabion. I do not doubt the accuracy or the impontmm of Levinson and Brown's findings that the mncepts of *uight'and 'left' play a relatively minor role in the Tzeltal system of spatial orientation. But perhaps they go too h u when they say that in Tzeltal 'dght' and "eft' don't have "'regional ex&nsions"at all, espiaIly since ithey themselves p r o d m two Tm1taP sentences glm& as ' m e man is standing at the weman's uigY111hand'kand "The man is standing at her IeR hand". The interpretation of such sentences su@atPd by lfurinson and ' s hand' (rather than Brown along the lines of 'The man is standing NEAR the w ~ ~ m m].light V O M I THE SIDE of the woman's right hand') seems lio me unwnvinning. One muld say L k l : a b u t t d y was howdng near a woman's right hand but not that a man was standhg near her sight hand (unless he was a Lilliputian standing on a chair). The twoTzeltal sentences in question king to mind a Pine from the A~plDStks'C~eed:"et sedet ad dextecam Patris", 'and is seated at the right hand of the Father'. Surely, the idea its not that Christ is sitting near the right hand of Lhe Fallner, blun that h.c is sillinlg acnr thc Fnthca. on his right-l~ondside, One can imagine a language in which the word for 'nose' is polysemous and means also 'front'. On k t , Mary Laupllmren, personal comlrmication, informs me that Warlpiri is a case in paint.] In a language like that, to say 'the man was standing In front of the woman' one wnuEd have to aay somebPljng homophonous to 'the man was standing at the nme ofthe woman'. On this t sentence in question really means 'the man was slanding basis, someone might argue t h ~ the near k Y ~ enose o l the woman'. But would a~vgibsldy ever want L o say a thing like that [speakm the ground]? The same, I think, applies to the ing of normal-siw people, both standing o righi hand. B concllude, then, that interesting as the TzsPtal examples mag be, theve isno mason b o regard it as a counter-example no the hypolhesis that SkDE [ONSIDE X of Y) ts a le%iml and semantic universal. 16. P am not assuming that the words for the "four sidles of the world" mean exactly lbe same in different languages. On the contrary, I expect that the details of the conceptualhation---especially ror "north' and kouth'-may well differ from ane hrugulags, or one group of languaga, to another.

The concept of INSIDE (like SIDE, put forward as a possible primitive by Goddard, personal communication) is rdevaant to a P P natural and humanmade "containers". b o n g the natural "containers", the most sdient is perhaps the mouth, presumably conceptualized a l over th~e world as a part of the body meant for, roughly speaking, "putting something in" @ s i well as speaking; see Chapter 7 ' ) . But presumably the whole body can be seen, ancrass cltures, as something INSIDE which there are various interesting and important "things" (or "partsym). In the natural environment, the concept of INSIDE is clearly relevant ta caves and also to armimd dweI1ings: burrows, tree-holes, nests, and the like. Among the humman-made '"containers"",Ltihe most important ones are no doubt human dwellings (houses, huts, and so on), and d s o wntainers for food and drink hots, cups, bowls, and so on). In English, Slue word isfde often appears to be interchangeable wiith j i ~ : inside thk house = in the hause inside the mve = in the cave inside the jar = in the jar But this is not always the case, either because a substitution of inside for in changes the (A) or because the resulting phrase is unacceptable:
(A) in the garden ~t inside the garden in ithe walls ;e inside the walls (of the city) (El) in the milk # *inside the milk in the d r ; f +inside the air

Clwly, the EngJish preposition in [like its closest translation equivalents in many other languages) is polysemous; and this is not the place to try to sort out its different meanings. Jnsidc, too, has more than one meaning:

2. A Survey of ,Yemantic Primitives 97

(I) People don't know what happens inside a volcano. (2) I was inside when it happened. (3) I went inside. (4) Outside China, people don" talk about it m c h , but inside the country, people don't seem to be talking about anything else.
Jasii$@4 (as in sentence 4) has a contrastive meanhg, built upon the notion

89.4. HERE

of 'outsidey (inside4 = mot outside]. Insit& is adverbial and refers spcifically to a Vwe1ling"umam or animal). d5asii$e3 is directional and refers to a sequence of times (roughly: before I moved, I was not insidel the dwelling; dter I moved, I was insidel it). l~&'el, however, appears to be indefinable. rOutsidem is mot a candidate for a primitive, as it is clearly composed of INSIDE and NOT: outside X = not insidel X . ) As mentioned earlier, the prepositions Sm, on, and under [and their doses&equivalenlts in other ]languages) emerge particularly early in child language, and in a particular order (in 4 on ~ s mder; s Johnston and Slobin 1979; Slobin h985a); end cognitive development has often been linked with the order of acquisition of Iowtive prepositions, For example, Mills Q198.5: 237) writes: "'If prepositions are classified according to the complexity of the conceptual reliatiotnships encoded iro them, the resulting order of @.omplexity coldd predict the order of acquisition . . . this cllassification will predict, for example, that h e preposition expmssing the notion ' i n w l be leaned before that expressing 'between'."" But It is not cllear what exactly is meant here by "the notion 'h3 ".As mentioned earlier, Eia is a pdysemous word and if one simply counts the occurrence of in in child language, one cannot be sure what concept or concepts are being expressed. For example, in the phrase in the cup, in means 'inside', but in the phrase in fhi3 place, in is only a part of the phrase ire a place (which as a whole means %omewhere9. It is possible that the data on the early use of the prepsition in in English cmflate the use d t w o different semantic primitives i n child language, both supported by independent evidence: the early emergence of where-questions (see e-g. Ervin-Tripp 11970; Tyack and Ingram 1977) and \the early mmprehension of the prepositbn ini~sid~ (see Bates el oL 1988: 190). For example, in a sentence such as "1 no make duty in Ithe potty?" "loom 1991: 1198), in presumably means 'inside'. On the other hand, if a child takes his guineapig home from school "cause they don't belong in school"' [Bloom 1991: 3851, it is likely that k means 'where' rather than 'inside'. 1 presume thait Slobin's ((P985a: Pl8O)r observation (quoted earlier] about the initial salience of "the notion of coatainrment"refers to the notion INSIDE, not G o the notion PLACE.

HERB is a spatid counterpart of NOW, more d which will be said later [Section 20.2). Both these elements were put forward as possible primitives by Cliff Goddard bersonal cornmication). On the hoe of it, 'here' is not a semantic primitive, because it seems to be clearly decomposable into 'this' and 'plaice', aloug the fallowing lines: "here = in this place". The fact tbat in many languages (em$. in Samoan; see Mosel 1994: 339) the three concepts %ere" '"now" and 'this' share the same lexical exponent appears to support this analysis. But if we identi$ 'here' with 'this place', then we cannot use the expression "this p1aw'in explications with reference to any other plaoe that we may wish, to taEk about (for example, ''this other place"). On the other hand, if we tried to link the concept of 'herehore tightly with the concept of 'I'as ""the place where I am", this would solve some difficulties but it would ereate others. For example, if I refer to two small objects lying in my open hand, such as two rings, as "this one here" and '"his one", the phrase ""this one here'km hardly m m 'the one which is in the place where 1 am'. The problems which wim in the attempts to decompose HERE are similar to those which arise in the case of NOW. In both cases, the condusion suggests itsdf that in fact the attempts at decomposition are futile, and that MERE and NOW are semantidlly simple, as the 'Veictic substmtives" 1 and YOU are simple. (For further discussion, see Section 20.2.)




Waving once tried to reduce a11 temporal concepts to non-temporal ones, (Wierzbicka 1972, 19XO), to have three temporal primitives (WHEN, AFTER, and BEFORE] seems a lot, and yet over the years it has become inareasingly clew that even this set is not sufficient to deal with all the aspects of time. In particular, it is not sufficient to deal with duration and, more genemlly, "passage of time'" Simple sentences such as:

I did it for a long time. It happened a long time ago.

could simply not be paraphrased in terms of the available temporal concepts; and yet sentences of this kind appear to be very common in everyday dismwse, in all languages. As Keesing (1994: 6) notes about the Kwaio, "'Talk about the passage d time ([inreference to how long garden work will be done, or when the pork will be cooked, or how long someone wijl be gone)

98 General I s s w is, for the Kwaio as with us, a constant theme of quotidian experienm and communication." So finally it became clear that, in addition to the three basic "temporal" primitives WHEN, AFTER, and BEFORE, something else was needed to account for the '"assage of time"; and the smswer came in the h m of two "duration" primitives, A LONG TIME and A SHORT TIME, analogous to the two "'dlstancel~rimitives FAR and NEAR. By introducing these two primitives, I am fdlowing, once more, in the footsteps of the Cartesians, who regarded both 'Yime" @emp~, presumably, in the sense of 'when" although they never bothered to make it clear) and 'duration" (darPe] as irreducible, clear ideas, which cannot be further defined. (See e.g. Arnauld 316621964: 66, 86.1 Since we aim always at a minimal set of primitiws, ilt would be preferable, of course, to introduce one 'Vwational" primitive rather tinan two, and, at first sight, this does seem possible: why not posit a neutral primitive, "some time", and then generate the meanings k long time' and 51 short t h e ' by combiabp this neutral primitive with the available elements MUCH and NOT MUCH? But there are strong arguments a g b s t such a move. First of all, experience and preliminary inquiries suggest that languages are more likely to haw words, or phrases, for the concepts A LONG TIME and A SHORT TIME than for h e putative concept ''some timey'. Having two "extreme"durationa1 primitives rather than a singlle one covcri~lg L ~ whole C ranga: Elsay sccrrn L o crcatc .uarnccessarydillilicwPtics In the area of measures. For example, how could omle analyse in these terms a question sluch as "'How long was he there?''? But the semantics ~f measures is notoriously wrnp8ex. In everyday discourse across cultures people are no doubt more likely to ask: "was he there (for] a long tiine?'?han ""hw Bong was he tlinere?"~nandit wouId be wrong to treat the latter question as semantically simpler than the former. Can all "'durational" conoepts be explicated in terms of the primitives A LONG TIME and A SHORT TIME? 1 do not claim t h t . It seems likely that some concepts which might be called "duratianal" would call for a diE ferent approach. For example, wry tentatively: He did it from sunrise to sunset. = (Cf. He did it alE day.) he didn? do it before the sunrise he didmy do it after the sunset he did it at all times after the sunrise before the sunset But I do not think that w can explicate alll references ta time without s o m explicitly durational primitives, such as the two primitives proposed here: A LONG TIME and A SHORT TIME. It should, however, be pointed out that of the two primitives proposed here one (A LQMG TIME) is a strongex candidate than the other (A SHORT TIME). Unlike in the case of NEAR and FAR, there are perhaps no compelling arguments against an analysis of one primitive via the negation of the other fa short time = not a long time]. If we consider, for example, references to 'a short time' in transcripts af children's speech, Uhey commonly take the form of the expression "for a little while", or "a little bit"', and these could perhaps 'beparaphrased, without a loss of meaning, via 'bat long'" Far example (French and Nelson 89185: 106-7):
First I wake up and wake my mom and dad up them slmp with them far a little

Andl we wait for a liltthe whiQe, but not too long, then we go back in the school and CU~~Y. . . . md we go m d wait for a little while and then the waibr comas and gives us the little sltufif with the dimem on it, and then we wait for a little bit . . .

Oin the other hand, it is not clear how words such moment or expressions such as ol once could be defined via negation and 'a Bong time'. H n a crass-cultural perspective, too, the evidence is mixed. Preliminary investigations suggest that h d i n g matching equivalents far '% short time'" is not as unproblematic as it is for "'a long tlirne"'. But the matter req~uites further investigation.

Another temporal concept, NOW, first proposed as a semantic primitive by Cliff Gaddard at the Adelaide Semantic Workshop (Cioddard 11986a3, belongs to the 'deictic" subset of the primitives, which includes also the "substantive" dements I and YOU, the "determiner" THIS, and the spatial element HERE. For a long time, this element was not included i n the proposed set of primitives because it seemed to be decomposable into other primitives. In fact, there seemed to be not ane but: two plausible ways of decomposing bow" which for ease of reference I will designate as A and B:

[A) now = at h i s time [B) now = at the time when I say Ithis
The fact that, as mentioned earlier, in some languages (e.g. in Samoan; see Mosel 8994:332) the same word is used not only far 'this' and %ere"ut also for %owyappears to support this analysis. As pointed out by Goddad, however, neither of these two analyses is really satisfactow.

2. A Survey o f Semanrk Primitives 101

Analysis A makes it very difficult, if not impossible, to refer to more thaa one time as 'this time', For example, if we say in a semantic fornufa: this happened a long time ago at this time . . . it is not dear whether the expression "at this time"refms to the t h e of reference or It@l the time d speech, a d it seems impossible to difirentlate between the two. Given that ago means 'before now', if now meant 'at this time" ago would mew %before this h e ' , and one couldn't refer to a time 'before this time" 'this time9. n malysis B, where " 'e time when This particular difficulty is resolved i I saw this" can be clearly distinguished from "a time before the time when I saw this". But as pointed out by Goddard, the word now can be used, and is n o m l l y used, with reference to a period much longer than tihe moment of speech (cf. e.g. the English expression nowadays). Furthemore, the three tenses distinguished in many languages of the world, present, past, and future, are normally used with reference to a period much more extensive, or extendible, than the very moment of speech. For example, if I say:
Iif yau would haw eated all that turkey, your tummy would have kcr-

21. Imagination and Possibility: IF MAYBE

. . . WOULD, CAN,


remark by a 4-year-old; Kuczaj and DaPy 8974:575)

e discontinuous expression if- . . wouMcan be used in English in two difrent senses (and in two different grammatical frames): as a hypothetical referring to a real possibility or as a counterfactual, as in the motto to this It is this second, not the first, use which is posited here as a uniernantic primitive. The hypothetical if. . . would can be defined in terns of "f', as follows: If X happened, I woulld do Y. = if X happens, I will do Y I don't say: I think: it will happen the counterfactual ij" . . . woufd cannot be similarly defined; and pmnnary evidence suggests that the counterfactual ifjT. . . would (in contrast to the hypothetical i f . . . would) may well be a linguistic universal. (See Wiembicka, forthcoming g.) I n English, and in many other languages, the counterfactual ( i f . . . overlaps formally with the conditionai (ii'. In many other lanhowever, the two elements in question do have distinct lexical nts. For example, in Samoan the word for IF (the I F of real p ~ s sibnlity) is @fa& whereas the word for the counterFactuaB IF . . . WOULD Mosel and Hovdhaugen 1992: 654; GENR stands far general tensemood particle, DIR for directional particle and w s for ergative suf(1) "Ansa 'e le sau, seman'ii 'ou te alu atu. if(irr.) 2 . s ~ not come, probably GENIE go DIR "ad you not come, I was probably going to go and see you.' (2) Afai ua lapoa se puaa, 0 !e a f m a = ina loa if PERF large aa~(wspl.s~.) pig ELIT shoot = ES then tino. . . ma hatau atu 1 = o = na and sell DIR ART = POSS = 3 . ~ body 0 'If a pig gets fat, then it will be shot and its body will be sold . . .'

I now live in Canberra, but before, I lived in Ballarat in Victoria. I do not mean ta suggest that Z have moved to Canberra immediately
before the moment of speech. The hypothesis that NOW is in fact a universal semantic primitive simplifies enomwsly semantic analysis of the various tense systems and alllows us to operate with very simple distinctions such as 'cnow9y, "before now", '(after now", "a long time before now'" "before now, not a long timale before now", and so on. At the same time, this hypothesis frees the expression '"this tirne" from its links with the present time and allows us to use it freely with reference to any time that we wish to spa& about. What may =em objectionable about .the present analysis is that it m u l d present the word new and the expression rhb time as semantically unrelated, and would not allow us to account explicitly for the "decictic" nature of the conoept 'now', But the same olbjwtion could be raised with respect to 3 and YOU, both primitives of long-standing and apparently unshakeable status; aria it would fdl on the same grounds: I and YOU, too, have often been lcallled 'cdeicticconcepts" and they do have a semantic link with the expression "this person", but this link is not compositional. If neither I nor YOU can be equated with "this penon", NOW cannot be equated witb "this time"; and ifI and YOU cannot be equated with '"he person who says !hisn and ""the person to whom I say this"', respectiveby, NOW cannot be equated with "the time when E say this"'.

A lexical overlap analogous to the English one m u r s also in the Australian language Yankunytjatjara (Goddard 19946: 2491, where the word tjing~mby itself (used in a biclausal sentence) means IF, whereas

Sldarly, in English:

What would happen if people were fish? calm. Then if a wbde came, they would get eated. But if they hided, the maybe couldln" find 'em. And, then tbey wouldn't get aated.

can't do itll can do d o t . mles dso remy on the notion of CAN: ou can't do this. You m y tsay this. You can't do it like tKs. You can't t o things like this. Yaw can do it like this. concept of GOOD and BAD, and social rules, which are based on

The fact that, despite lexical overlaps, the concept MAYBE Is r e a l k both English and Polish as a particle, whereas the concept CAN is reslr

(62) Ma-pitja-ku-na? away-come-~uP-1 'Cm T carnie in?' (63) Punkal-ku-n. fall-r;uT-you 'You could fall.'
In the list of primitives tested in Semrrsrtfc and Lexical Univ (Goddard and Wiembicka 11994b1, the elements distinguished here as as CAN. Further work in this area kLsrs shown, however, that it is neoes

CAN at our disposal. be suggested that CAN is linked in a special way with 'wantso much SO that one may even be temptd to explicate one in terns he other (as I did in Wierzbicka 1 9 8 9 ~318-19). : For example: God can do everything. = If Cod wants to do something, he does it.

d m g these lines seems also to have the virtue of accounting for ly fell: llnk &tween CAN and IF (as J. L .Austin (1961: 1531 put iffy'", Hut neither of the above two pmaphsases is really satisrrs neither of them conveys the intended idea of "onmipotenw". ider aim sentences such as the following ones:

Her stomach cannot ~ g e s fatty t foods. Ilfter we built a d m here, the water coddn't tow in this direction any
(I) I can't do it now; maybe someone dse can. (2) I can't do it here; maybe I can do it samewhere dse. (3) You can" see. it now; maybe you wm see it later. ( 4 1 1 I can't say it is far, I can't say it is near.

ich can hardy be paraphrased in terms of WANT: When her stomach wants to digest fatty foods it does not do it. When the water wmts to tow in this direction it doesn? t o it.

use of these ideas i s in the context of movement:

I can't moveirH can move.

(perhaps an archetypal experience of a baby, tightly held, or wapped a d then given the freedom 4.0 movie).

106 General lsmm

2. A Survey

ioJT Semaantic PrirnSti~~"~~ 107

*I h o w she will1 come, but maybe she will not come. *E know she will not come, but maybe she will come.
Ncvertlietass, MAYBE cannot be reduced lo KNOW (or NOT KNOW any more than CAN can be reduced to WANT. The sentence "'maybe h~e did itY3mplies that '7 ddoqliknow whether he did it", but cannot be reducedl to it. In a sense, then, the notions CAN and MAY BE can be regarded as clear and indispmrsabb. On the other hand, in many contexts the distinction between these two concepts appears to be blurred. For example: This can break. Something bad could happen to them. Bad things can happen to everybody. They can" know about this. In terns of the traditionail distinction between cpossibility' and 'ability' ithe sentences above would no doubt be linked ~6th 'possibility'; but it is di&cult to see how these sentences oouId be paraphrased (without a change d meaning) with MAYBE and without CAN. In e a r l b work Pierzbicka 19721, I tried to account for the semantic links between CAN mntenlws with personal subjects (such as "H mn moue~dolsedhearIsay"')m d those with inanimate subjects (e.g. "this ean break", 4'somethinrg bad can happen"') in terns of a "hidden" personal predicate, along the following lines: This can break = I can think: this will break. (A similar analysis was also proposed by Antinucci and Parisi 1976.) But I no longer regard this analysis as valid. To begin with, one could equally well1 tlry to 'reoaverPa hidden 'think9n all other CAN sentences:
I can't do it. = I can't think: I will do it. I can'lt m u e . = I can't think: I will mow.

To amount for the fact that, on the face of it, two concepts posited as distinct and irreducible semantic primitives can both be replaced with one concept of 'possibility" 1 would suggest that the notion of 'possibility?~ more complex than either MAYBE or CAN, and, in a sense, spans over lboth of them, rather like the abstract notion of 'distancebspans over both FAR and NEAR, or the notion of 'sizey spans over both BIG and SMALL. The analogy is not perfect because FAR and NEAR, or DIG and SMALL, intuitively felt to be 'opposites', whereas CAN and MAYBE are related ome other way. But maybe even this imperf~ctanalogy can help us to pt that, related as t h y are, MAYBE and CAM may none the less be distinct and irreducible semantic primitives. should also be pointed out that while the concepts CAN and MAYBE share their lexical exponents (in dinerent grammatical forms), this is means universal, (See for example the cantrast between the verb , %anz, and the p t i c l d a d ~ e r b~ieiieichs, haybe', in Geman). %wring lexical overlap are a common feature of many primiltives (e.g. EONE and OTHER, or MAYBE and IF), and while they are dearly ccidentd, there is n n o reason to assume that they must be due to comnand semantics. (For discussion, see Wierzbicka 1994b.)

22. WORD
RDI (or %wordsy> may seem redudant in the lexicon of because it is intuitively relalied to the concept of SAW that one does, a word is something that one says. 5nly partially valid: while ween 'cword" and "deedYYs word, or some words, one can also say something "in" e words or: in some other words. (In fact, this is what BI about: ways of saying the same thing in other words.) we can talk about speech [about "saying" things) c'word~";in some cases, however, w reference to e essential to the intended meaning, as the following
(A) You said something bad about this person. @) You said some bad words to this person. ce A can refer to a criticism, whereas sentence B is more likely to to some swear-words. e notion of "swear-words"is a good example or the apparent indisnsability of the concept 'words': one cannot "swear'hithout saying what regarded in a given society as some ""bd woods" (for a hller discussion '%wearin$;",see Wierzbicka 1987~).

But this is counter-intuitive and unconvincing, and sometimes can bad to a b w d results: God can do everything. = I can think: God will do everything. Undoubtedly, the distinction between CAN and MAYBE in sentences with inanimate subjects requires further investigation. Rut in sentences with human subjects the distinction between these two concepts seems well established, despite the faclts that both CAN sentences and MAYBE sentences can be paraphrased [approximately] i n terns of 'possibility': Maybe she will do it. = It is possibie that she will do it. God can do cvesything. = Everything is possible for God.

B0 1 9
Other categories of speech which appear to involve crucially the notiom o l b a r d ' ~ L Y G nanlnes, ~ U J pcrforinakive verbs [ne.g. pramiJi~~), speech fommlae @.g. Goodbye), and magical formulae (see Goddard forthcoming a) . For example, a "name1?s a word (or words) generally used to identify someone or something; "coonting"invo1ves saying a word that means; 'orre', aner that a word that means 'two', then a word that means Yhree', and so on; magical speech involves saying certain words to cause certain things to happen, and so on. ([These are of coulrse not Full definitiwns, but only partial characterizations of tlne phenomena inr question.) "In the beginning was the word" (John 1: 1). Distant ias this sentme is from normal everyday life, Bible translators seem to find less difficulty in translating it into numerous languages of the world khan many much more prosaic and down-to-earth sentences (see e.g. Nida 1949. If the conmpts BEFORE, ALL, SOMETHING, OTHER, HAPPEN, THERE IS, MOW, SOMEONE, SAY, and WORD are aall universal human concepts, the relatively easy cross-translatability of this sentence would be easy to understand: BEFORE NOW BEFORE ALL THINGS HAFFENED THERE WAS A WORD (or: SOMEONE SAID PI WORD) E to cEduration'y, with the addition of the two recurring Watuces "b ry)e" and "-POL". Following this line of analysis, we would be able to account, in cornprositional ems, for the fact that within each pair the "+ me" m stand for both members of the pair: How big is it? -Very bidvery small. HOWfar is it? --Very fadvery near. How long? -A very jong timeja very short time. But attractive as such an analysis may seem, it has to be rejected-if we require that the basic units of semantic analysis are linked with "red (intejIiigible] words"rather than with analyticall fictions, which have no meaning unless and until they are defined in tems of intuitively intelligible real words. E n real (natural) languages there are simply no such words as "+ pon."Qr "-POL"'. In real speech, therefore, we cannot paraphrase a sentence such as "This dog is big" with sometlning like "The size of this dog is It might be suggested, of oourse, that the artificial words "+ ~u>z'bnd could be replaced with the "rea8'bords big and $maSI, along the foNowbg lines:

+ WL".

23. General Dis~ussion:Opposites and Converses

The present system of senrantic primitives m y seem merconomical hso far as it includes some pairs of "opposites". Tw begin with, there is the pair of evaluators GOOD and BAD, the oldest and the best-established one. AIthough these two elements have, intuitively, something in comtnm, I have rejected ithe temptation to extract from them some semantic common core ('%aiue"), and to distinguish the two as a "positive" and a "negative" member of the pair. Rather, I have zussumd that concepts such as 'value', 'positive', and 'negative'aare more complex than the basic elements GOOD and BAD, and that although these two elements do form a coherent a i o l l system apart from all the other dements, none the less none of them a m be further d~composed. U n addition to GOOD and BAD, the present system of primitives includes now two (and possibly three) further pairs .sf "opposites": 816 and SMALL, and FAR and NEAR (and possibly also A LONG THME and A SHORT TIME). Aiming at a minimall system of primitives one must be tempted, of course, to reduce such elements further, extraclting for each pair a common core, and positing for a!! pairs two "poles": + POL and - POL (along the lines suggested by Bierwisch 1967). One cmld try, then, to reduce BIG and SMALL to ""size", EAR and NEAR to "dlstance'~and A

far - big distanae near - small distance But this would be analogous to saying that b i g reallly means %ig shey,and lsmli, 'small size'. In fact, 'size' is not a simpler concept than 'big' or %mall" bunt a more complex one, a d so it is %bey which has to be defined in t e m s of 'bighad 'small" not vice versa. Similarly, it is not 'far'which should be ddened i n tems of Vdistancnce', or 'good" in tems of "value', but b e other way around. In natural language we simply cannot go beyond words such asfar and near, z u s we cannot go beyond gaod and bad Replacing good and bad with "positive value" and "negative value" we would engage in pseudo-analysis, not in real semantic decampositim; and the same applies to attempts to replace big and s m a l l with "'size", orfir and near with! "distance"". In saying this, I am not denying the reallity of the structural relations within the area of "opposites", On the contrary, T would like to add one further observation highlighting the dose affinity between the two members of each pair, namelly that within each pair of "opposites" both members appear to have similar combinatorial possibilities. Since this is an obsewatiom concerning all the pairs of "opposites", it provides also a further argum n t for the reality of this group as a distinct subsystem within the whole system of primitives. In arguing that the relations between the "opposites" within each pair are

not compositional (that is, Ithat they cannot be accounted Tor in terms of further definitions), I argue that these "opposites" are more deeply ingrained in human cogmition than they would be if they were reducible to some other elements. It is well known that "opposites" play an important role in all known human languages (see e.g. Apresjan 1974, 1992; Cruse: 1986; John Lyons 1977; Lehret 11974). I am suggesting that this is mot a surface phenomenon but something that is rooted in the underlying system of primitives. What applies to "opposites"aaplpes d s o to some extent to converses: BEFORE and AFTER, and possibly to ABOVE and UNDER. The system of primitives would be more economical if each pair of converses were replaced with just one element, because from a logicall p i n t of view " A happened before B"" is equivalent to "I3 happened after A"",nd "A is above B" is equivalent to " B is under A". But a natural language has its own natural") logic, and in this "natural Iogicy'BEFORE and AFTER, or ABOVE and UNDER, are not multually equivalent. In both pairs, each dement is linked with one particular point of view, and in huumiaar comunication ai difference in point of view m y be as important as a diEerence between two predicates. For example, while both past tenses and future tenses situate events in time with reference ta the present moment, past tenses do so in terns of the concept BEFORE, whereas future tenses do so in terns of the concept AFTER: (A) It happened before now.

(IB) It will happen after now.

Looking at the expanded list of primitives and comparing it with the older one, one is bound to notice that Ithe new list is less austere nor only m ir its size but also in its composition, The new spatial elements, of which there are as many as five (ON PHIS] SIDE, INSIDE, HERE, FAR, and NEAR), bring the set of primitives down b earth (from its pwvious heights of abstraction). At the same time, the element THERE IS links the system more closely to reality, as do also the new "deictic" dements HERE and NOW. Tine pabed opposites FAR-NEAR, and A LONG TIME-A SHORT TIME, strengthen the element of subjectivity and add an anthropocentric, experiential perspective (as do the old primitives BIG and SMALL)l. SEE and HEAR bring colom and sounds to the system, and, if I may venture to say so, MOVE brings movement, and ALIVE brings life. MAYBE brings an dement of uncertainty, linked with a human, psychological perspective (quire different from the logical perspective of "gossible"'), and I F . . . WOULD brings, or rather restores, the element of h t a s y , which was once brought to the system by the ex-primitive "imagine''. W1 in all, then, the changes can be seen as being all for the better. It must be remembered, however, that--quite apart from the obvious requirements razor-we are Iooking for the shared Eexical and grammatical of Oo@mYs core of all languages; and that given the tremendous diversity of languages as we know them, this shared core is bound to be small. It is imperative, therefore, to continue to subje~t every proposed primitive, and every proposed ~ ~ a tframe, i ~ to relentless 1 scrutiny, so that only those remain which are truly indispensable and truly universal.

The set of proposed universal semantic primitives has expanded from 14 (in 1972) and 13 (in 1977) to 37 In 11993, and ~owidlramatidly-to 55. The question imposes itself: how many more primitives (or hypothetical p e tlives) are likely to emerge from future work? For once, 31 Feel that humble agnosticism is in order. I would Bike b recall, however, that when Bogustawski launched the search for semantic primitives in 1965, he mentioned the figure 100 ("airnost certainly less than a hundred", he said, as I recall). Although I still expect that the ultimate figure will be doser to 50 than to 100, f now acknowledge that (as argued by Goddard at the 1986 Adelaide Workshop), my original sets of 14 and P 3 were quite mrealistid1y small. As the set of primitives expands, and as their grammar takes shape, the Natural Semantic Metalanguage grows in flexibility and in expressive power. In principle, them, the expansion of the semantic system is a positive, not a negativ~e,development.

Universal Grammar: The Syntax of Universal Semantic Primitives

Most gmmalllcal patterns In any language am language-specific, but there may also be some patterns which are universal. In fact, if cross-cultural undenltanding is possible at all, despite the colossal variation in language a common core of "human understanding" relystructures, there muslt ing not only on some s h a r d or matching lexical items but also on some shared or matching grammatical patterns in whi& shared lexical items can be used. Arguably, this m m o n core defines a set of "basic sentences" which can be said in m y Smguage, and which can be matched across Innp a g e boundaries, and the grammar of these basic sentences consists in the possible distribution patterns of the "atomic elements" (that is, the lexical jirrdefinables]. To dixover those patterns we have to look at the lexical indefinable~themselves, to see what their possibilities of co-occurnenoe are. Therefore, in searching for universal g m m a t i d patterns, llse should not be looking for universals of farm, but rather far universals of comlsinablility. This chapter is a tentative and preliminary attempt to do just that. Trying to write a sketch grammar of the universai semantic primitives is a daunting task. First of all, such a grammar must make a large number of predictions conoerning all the languages of the world. These predictions c a m p be empirically tested, and it is high& likely, not to say inevitable, that upon further testing some of them will turn out to be inmrrect. Second, this grammar amounts to ai hypothesis about the innate grammar of human cognition, As a hypothesis about human cognition, the system developed here is not as readily testable as it is as a hypothesis about linguistic u~niversds.ExpIoration of limguistic evidence may prove t o be, at least for some time, the main avenue off access ta the grammar of human cognition. But no matter how dificullt it may be ltcl verify-or to falsifyThis chapter owes a great deal to Bong discussions with Cliff Gddard, who has ~ontribukd to it many important ideas. CBiYlFs cdtidsms and suggestions led to a substantid revision of the eadkr draR

the hypothesis about human cognition advanced here, the questions raised are disconoertingly heavyweight. Third, this grammar has, to be wen as the '9eaP grammar of PIASM, the hnguage of semantic description. It is proposed, therefore, as a set of constraints on NSM explicaiations and paraphrases. These constraints will not always be adhered to in practice, but they will always have to be kept i n mind, so that any departures from the mles of combinatorial semantics outlined here will be allowed only as short-cuts and compromises Justified by practical considerations such as increased brevity or readability. It hardly needs to be added that the word "always" in the last paragraph is not meant to predude future changes in the proposed system. Despite the present attempt at coldlification, many areas of NSM grammar are still1 in a state of flux. The grammar proposed here is nleither complete nor "hal". It is put forward as a starting-point for testing and discussion.

2. Preliminary Discussion
the survey of primitives and their combinatoriat possiloiler of general points should be made. rammar owitlined below assumes a radically expanded set of instead of 37 as in Geddard and Wierzbiicka 1994b). Since d some of the new primitives is still somewhat uncertain, so is, course, the status of their grammatical characteristics. Second, the meta-terminology of NSM grammar is still evolving. Terms such as "valency", "linkers", "swbj~ect",'~objeet", "cle~mplement"~ and "aldjunct" are used in the present chapter on a somewhat provisional basis. Nopefily, however, for the present purposes their intended meaning will pter makes a m p extensive use of a new theoretical concept: "valency of semantic prhitives"".or exampnlle, it is assumed that the predicate WOlD has two different valency options: it may combine with one "substmtive"(which may be called a "sulbje~t"),as in sentence A kIow, or with two c~sub~tantives'~ (a "subjectlYl and a "complement"), as in sentenoe

(A) This is good. [BE This is good for rne/yodthese people.

Strict adherence to the rules of NSM syntax, as sketched in this chapter, is not a l w p dlesisable, as bng as all the departures from the NSM mules can be regarded as convenient abb~eviations, that is, as long as we have a clear idea of how the '%nggraanmaticaB"car "semigrmatica1'" segments of the: exp!ications could be replaced w i t h fully "grammatical" ones.

114 General fssrres

Some: predicates-for example DO: m d THPNK-my even opema three Lcsl~ts" for ccsubstantives"(a first slot for a "subject",a second, for a "camplement", and a third, for an "object") someone (11 did something (2) to someone (3) someone (I] thought something (2) about something (3) But although both DO and THINK can be said to open three slots, heir valency options are different: DO bas two vaalency options (A and B)l whereas THINK has three (A, B, and @): DO (A) someone did something (B) someone did something b someone THINK (A] someone thought something (B) someone thought something about something [C) someonre thought about something Fourth, it should be pointed out that the grammar sketched in this chapter allows for several types of complex sentences, and thus glees far beyond simple clauses offered as examples of NSM semrtenoes in earlier work (kt.$. irr my Liaguo MessSSSaiis (1980) or Semtmfics o j Grummar (1988)). A key rolle belongs in this respect to the primitives which function, or can function, as '7nterclausal linkers'? BECAUSE, IF, I F . . . WOULD, LIKE, WHEN, AFTER, and BEFORE. These linkers ]providea mechanism for combining Wo or even three clauses into one complex sentence. IFimaBly, the theoreticai concept of "ailolexy" (analogous to "ailomarphy") should 'Ix: mentioned here, too, h r allthough it is not a new concept in MSM theory, it is one which raises important questions for NSM grammar. For example, the account of the combiruatoriaP possibilities of the primitives SQMEQNE and SOIMETWlNG proposed here depends on t k assumptions that in English, person (in same of its uses] can be seen as an alkolex of SOMEONE, and thing (in some of its uses), as an allolex of SQMETHING.

with the mental predicates THINK, KNOW, W A W , FEEL, SEE, and NEAR: I thinmnowiwanUf~Useehear. You t h i n M h o w / w a n ~ f e e l / s ~ e a r r s is not to say that there are no restrictions on these combinations. In dar, in some languages (e.g, in Japanese, see e.g. Inrowe 1979; or in , see Haimim 19951, only 1can freely co-omur with mental predicates eclarative sentences, whereas YOU normally combines with them only rions [and third person subjects require the presenoe of spcial "mi"markers, highlighting the limited character of our kno7FFledge of ther people's internal states). YOU and I can also universally owm in combination with the action predicate DO, in a role which may be conveniently labelIed as that of an "agent'" for example: YouA did something bad.

They can also owur in the rote of a 'cmover", i n combination with the edicate MOVE: YoulrI moved.

, YOU and II a n combine with the predicates of description and ion, as in the foilowing sentences:
You are a goodhad person. I am a badigood person.

U and I @an be used i n the r d e of either of the arguments of a 'Yelain relationai sentences such as the following ones:
You are llike this other person/me. I am like other peopleiyou. They can also co-owur with spatial (though not with temporal] predimte phrases: I am in this @am;you are in another place. I am under this thing; you are above this other person. I am here; you are here. I am far from you. Furthermore, YOU and I occur as "patients" in combination with the universal predicates HAPPEN and DO, for example: Something bad happened to meiyou. This person did something bad to me.


YOU and I have a wide range of wniversat syntactic roles. Perhaps the most important one among them is the role of "psyclkologicd subject". What I mean by this is )that YOU and I can universally owur in combination

116 GeiteralI3mes In combination with SAY, they can also occur in the role of an eraddre~~ae": This person said something to meirgrou. Finally, YOU anldl I can be used in the role of a "psylclhoEogica1 object'" clear, given the uniqueness of every "'I"' and every "thouu"(YOU-Sg), and the non-uniqueness of "persons"and "people~'.2 It should be wted h a t in many languages the basic word for SOMEi t h 'Veteminers" either. For example, irr ONE doesn't readilly wmbine w English ane doesn? tomalily speak of thb someone, one someone, or the s m e someone; and in Russian the phrase Pdot kto-to ( " s someone') is even less aweptab8e than its English counterpart. Usually, however, SOMEONE has alPoUexes which can readily caimbine with determiners 4e.g. peaan i n English); and of course words more complex than SOMEONE (but including SOMEONE in their meaning) are widely used in combination with determiners @.goSht man). Perhaps the main wason for the awkwardmess of expressians such as "this someone" lies in the fact that their meaning is usudallly encoded in a special portmanteau, that is, in the third person pronouns such as he m d JAB in English. The fact that most languages have such portmanteaus [sametimes with, and sometimes without, an added reference to gender]~ highlights the importance of the combination of THIS and SOMEONE in human discourse. Do both SOMEONE and PEOPLE combine with all the elements included here in the list of '"determiners" and ""qantiffiew7? I would suggest that while both these elements combine with THIS, THE SAME, OTHER, TWO, SOME, MANY, and ALL, perhaps only SOMEONE mbinew -with ONE (as in one person). In fact (as mentioned earlier), English itsellf provides a good example of the asymmetuy between PEOPLE (plural) and its non-existent singular equivalent (with the word person not hieing as strictly restricted to humans as the word people is]. (See Chaplter

in sentences swh as the following ones:

1thinlk about you. This person knows everything about y o d m .

31.2. SOMEONE and PEOPLE By and large, SOMEONE and PEOPLE have the same combinatariial prossibilities as YOU and I: This persodthese people think(s)Jknow(s)~want~s)iFeeIQs~, see(s>iElhmr(s) something. Someone did something (bad) to this persodthese people. Something bad happened lto this persodthese people. This persodthese people moved. I said something to this persodthese people. This person is a gooctll'bad person. These people are goodhad people. This person is (not] like other people. These people are (not) like other peo@e. This person is in another plaoe. Them people are in this place. I think about this persodthese people. None the less, there are some systematic difTerences beltwen YOU a on the one hand, and SOMEONE and PEOPLE on the the restrictions on combinability with mental predicates SOMEONE and PEOPLE, in contrast to YOU m d I, can co-o wide range of ""dteenniuners", for example: this persodthese people (*this I, "this you) the same person, the same people (*the same Ilyou) another person, other people ("another you.0) one person ("one 1, lone you) two persons, two people (*two I-s) many perlaons, rnmy people (*many I-Q d l (these) persons, all peoiple ("all I-s) The reason why YOU m d I, on the whole, don%combine with "datenn ers'bnll '~uaatifiers"whereas SOMEONE and PEOPLE do is af mu

What I am suggesting, then, is that perhaps the semantic element FEOPLE doesn't really combine with the ""dtermjlnerkquantiffier"" ONE, or er, that it can onily combine with ONE in the '~artitilre""r "sdective"] ency option ONE OF? 'one people (in the relevant sense of people) one of these people Finally, whille YOU and I, as well as SOMEONE and PEOPLE, can be described as "good" or "'lbadY'(e.g."you are good"") only the latter pair can comlbine with these evaluators as attributes: someone goodsomeone bad (a good personfa bad person) good peopldbad people
The range of quan~ltjers with which YOU (that is, THOU') combines, mag be diEerent from that of 1. The matter requires further investigation.

SOMETHING (with am allolex ccthing'p)l has a wide, and, one might add, remarkably heterageneous, range of syntactic roles. There is a l a g overlap with the roles of SOMEONE, but not d l the roles of SOMEONE asc: equally applicable to SOMETHING, and some rdes of SOMETI-ILRrlG arc: not applicable to SOMEONE at all. Nomtallly, SOMETHING doesn't occur in the roles of a psy&aPogiic;ll sub:je.ct: ?This thing wants (*thinks) something. ?This )thingdid something.

But perhaps the most important, and unique, role of SOMETHING is that of a complement, clovering the range lof a "'psychological ~;lomplement",a "speech complementy',a m "action connplemenrt'~ and a m "event complment":

I wanr%rncaw/&i&feePseekeari~omethiang~
JI said somethhg.

E did something.
Something happened to me.

4. Mental Predicates: THINK, KNOW, WANT, FEEL, SEE,

Mental predicates (THINK, KNOW* WANT, FEEL, SEE, and HEAR) ombinle, first of all, with 'cpsycholagicd subjects" (I, YOU, SOMEONE, FEOPLE]: Uyou thinkjrptnowIwmtJfeek'se&ear somet thing)^. Someone thinksflnowdwantsIFeeBsJseesJlkears(something). People thinkflrnowiwan~feeliseeirhear ~(sornething). Furthermore, they a11 take a "psychotlogical complernenliY~IISOMETHING, THIS), for example: You want something. This person knows this. These people feel something bad. I seehear something. The range of possible "complements" is no doubt different in each case. For THINK and KNOW* the complement slot is likely to be Mled by a whole proposition [e.g. "'I think that. . . X ".', know that . . .") Far THINK, it can also be fiIled by direct discourse (e.g. ''I thought: gee? how strange!").. Far WANT, the most Iikely complement probably takes the form of an "quiclause" (e.g. "I wanted to goY3, and in any case, even if Ithe "com'plement" slot is dilled by a "substantive" (THIS or SOMETHING), this "substantive" has to stand for a proposition (since sentences such as "I want an apple'%ave to be interpreted as an abbreviated form of solnbnms about having or getting an object, "1 want to havelget an applen). SEE differs from the other mental predicates in its ability to take SOMEn its "mmpleaent" slot, m d also to cembine with & ONE end PEOPLE i place adjunct:

?I said this to this thing.

but under certain circunrstanms it can appear [at least semi-felicitouslyjl in an these robs. ('%is applies, itu particular, ta animals and hmm mllwtives.)l In addition, SOMETHING [thing) can occur in the role of a "patient" or a cGmover'y: You did this ta this thing. Something happened to this thing. This thing mowed. It a n aho o m r as a ccsubjed''of evaluation, description, relation, o r lwation: This thing is gaodfbad. This thing is bigsmall. This thing is like this other thing. This thing is udedabove all these other things. This thing is an the other side. This thing is fw (from this place]. Unlike YOU and I, but Like SOMEONE, SOMETHING can also be combined with an attribute: something goolcllibad (a goodhad thing) something bighimdl (a bidsmall thing) Furthemore, SOMETHING-like SOMEONE-can combine with a wide range of determiners (with the same d1Qlexicrestrictions as SOMEONE): this thing; another thing (someone else); the same thing; one thing two things; many things; few things; all things (everything]

I saw someone. You saw many people. I sae someone in this place.

120 General Issues

3. Universal Grammar 12 1

(In Endish the verb hear, too, has these combinatorial possibilities, but this is unlikely to be universal.) Perhaps the least clear of all is the semantic syntax of FEEL. In many languages [including English), sentences with FEEL and cccomplements'N such as somethilgg or s ~ are B not fully aweptable. In English, sentences such as '" feel goad"",'E feel bad", and "'P Fecl like this" "sound of course better than "Heel something like this". The syntax and semantics of FEEL senhnces in other languages require careful investigation before m y firm hypatheses concerning the cambinatorial possibilities of FEEL can be wnfidently put forward. In addition to the '%complements", some mental predicates open a slat for what we might call a "psychological topic"; for example: "to tthlnuk about smething" and 'Yo know about something". These "psychobgid topicsY"an co-occur with the "psychological ccmmplements" "to think something about something", ""toknow something ahout somettuing"'. (It Is by no means clear, however, that in all languages the exponents of THINK and K;N.O'W have as many ;as throe dierent slots; the picture presented here is at this stage only a matter of conjecture.) h e In the explications of various emotion concepts, I have often used t phrase "to feel something (good or bad) towards someonre", which seems very useful in modelling similarities and diflerences between different m o tion terns (see e.g. Wierzbicka 1992ejl. One can doubt, however, whether FEEL really has such a valency. It is m r e probable that both FEEL and WANT have, universally, only two slots: "psychological subject" and "psychob@call complement" (e.g. want someitlking" ""I don't feel angthing'". Finally, it must be noted that all mental predicates have also a time slot vcat that time, I itlhough~knew~wante&f"elu'~awflneard . . ."),although the exact rmge of possibilities may be in each case different.

You said something. 1said this. This person said the same. If you say one word . . . There is also an obligatory time slot (tlrougl~ not a place slat). P n fact, SAYING,like DOING, requires a fairly specific location in time (whereas mental predicates can Rave a more indefinite time span). Importantly, SAY opens a slot for an addressee. I have often argued gagaimst, for example, Ross 19711or MclCawley 19731 that SAY doesn't necessarily presuppose an addressee [see e.g. Wierzbicka 19TQ, I980), and I by this. But while there is no obligatory addressee slat, there is ceran optional one. Thus, while the sentence
ad Eiod said: let there be light! (Genesis 1: 33

not elliptical far 'mod said to someone: let there be light!'" the sentence angel said )to her: "Don't t afraid, Mary'' m n o t be anallysed into a "manotransitive SAY" and something else, along e angel said somethhg; the angel wanted Mary to hear this. it may be true that "saying something TO someone" hpPks wantperson to hear what we are saying, this doesn" differentiate the essee from lather people whom the speaker may also want to hem the age. For example, one can say: He said it to Mary, not to me, but I know that he wanted me to hear means that just as DO has two irreducible valency options [DO and TO), so has SAY [SAY and SAY TO). fact, I would suggest that SAY has one extra valency option (a third 1 , realized in English in the frame of SAY ABOUT, as in the sentence


5. Speech: SAY
The primitive SAT occupies, one might say, an intermediate place between mental predicates and the action preldlicate DO. In a sense, "saying something" can bue seen as a form of "doing something"', and so the "subjmt" of SAYING can be seen as an agent. Since, hawever, SAYING can allso 'i3e done in one" head, the '%ubject'kf SATING can also be seen irs a "psy' chological subject", andogous to the "subjwt'kof THINK or WANT. Like DO, SAY has also an obligatory slot for a "comp1ement"":ne "says somethingyy, as one "does something"', (Mental predicates, too, have a slot for a "mmplmenX'", but not necessarily an oMigatorqr"me; see e.g. "I thought about pu"..) For example:

I want to say something about these people.

f THINK and KNOW, realized ABOUT, KNOW ABOUT), the ould be called (for convenience) slot, the "locutionary topic'"51ot

AS I have proposed in Wienbicka (151764.

6. Actions, Ewnts, and Movement: DO, HAPPEN, MOVE

6.8. DO

Obviously, the action predicate [DO] opens, mniversalPy, an "agntY'slot: I/yodsomone/5?aople did (something). As the illustrative sentence above suggests, it opens also a slot for art "action mmplementyq:

"patient" is an additional (optional) syntactic slot in the structure of D 0 sentences, Finally, I foreshadowed in Chapter 1 the possibility of yet another vsulency option for DO, namely, an "'instrumental optionYWOWITH ("someone did something with something'"), as in the folllowing sentences: This person did this with one hand. This person did this with somielthing of this kind (a knife, a hanunerr, a boomerang, erc.). At this stage, however, this valency option is proposed only very kntatively.

I did something (bad). You did this. I did the same.

Clearly, there is also am (obligatory) time slot: k t t h t t h e , Vyou did this. as well as an (optional) place slot:

Taming to the element HAPPEN, we m d a similar valency structure except for the absence of an "agent" (and sw. "instawment") slot and far the cenbra1 position of the "patient" role. The ""patient slot'kf HAPPEN TO senbnces 6e.g. "something bad happened to me'" corresponds to the optional "patlent'hlot in DO senfences @.g. "'you did something bad to me"'). An obligatory "evmt compRement" (e.g. somethingkttnidthe same happened to me] corresponds to the 'kction complement" in H)Q sentences (e.g. "'I did $omethin@this/the'same"). The time slot is inherent in both DO and HAPPEN sentences [e.g. "at that rime, I did the same", "at that time, the sane happened to me"). Both DO md HAPPEN sentences have allso place slots [e.g. " I d i d it in this place", '*smething happened to me in this place"). But in HAPPEN sentences, a place slot can also be an alternative to the "patient'blot. In a ccpatientless"sentence such as Something bad h a p p e d in this place.

I did it in the same place.

The difference between the time and place slots is that time is relevant to all DO sentenms, whereas place is relevant only to some of them. [Sentences with mental predicates do have a time slot, bat they don't have ptaoe slots at all.] What is perhaps less clear about the semantic syntax of DO is the presence of a ccpatient"slot. In earlier publications (see e.g. Wierzbicka 1981) I attempted to analyse sentences such as '"Thlis person did something [bad) to me" via "do" md 'chappen to", dong the following lines: This person did something bad to me. = this person did something @ad] because of this, something bad happened to me (at the same time)
I have come to recognize, however, that (as argued i n Boguslawski 19911 this kind of analysis is untenable, if only because "DO TOYhentences vbw the situation as a single event, whereas a 'WO + HAPPEN TO" sentenoe views the situation as two causaily linked events.' H m n o t accept, however, Bogustawski" furher suggestion that DO md DO TO aw two different semantic primitives. Since the alleged two e1ernents IiEO and DO TO appear to be realized, universally, by means of the same lexical exponents, H think it is more justified to conclude that they represent a m diflerent valency options of the same primitive; and that
This is a point that Nick Enffield rightly insisted vpon during a seminar discwssicm at the ANU, thus helping to clarify the difference between DO TO and DO + HAPPEN TO.

phrase "'in tiis place" is an alternative to a "patient phrase" (e.g. 'Yo rather than an ;adjunct (as in "something happened to me in this "") This means that just as the element DO has two alternative pat,A and B:

(A) X did something (36) X did something to Y

element HAPPEN, too, has two alternative patterns (C and D):

6.3. WOVE

MOVE occurs as a predicate: in combination with all the substantives:

Ilyoulsomeonefsomct[hin~this moved.
These two people moved.
It is possible that MOVE can also occur, universally, in combination with place (WHERE):

a ilmalionall phrase. It is doublfui whether, for example, the Cartesian "sum", 'I am', '1 exist' (in "'Cogito ergo sum'') couPd be satisfacroriJy ren&red, and make sense, in! all the languages of the wodd. Existential sentences which can be expected in all cultures include probably the ifolIowing kinds: There are three kinds af bat. There is no such thing (as this). There are many people 1Be that (of this kind). There are no trees in this place. There is a lot of water here. There is no water here. There is someone in the garden.

Something is moving over there. Something moved here. Possibly, this additional dot ( h e WHERE slot) is avalhble ody in sentences with an "indehite'hubstantive, above all, with SOMETHING. E n an earlier draft of this chapte~ (distributed to the participants of the 1994 Canberra Symposium on the Universal Syntax of Meaning) 1 have suggested that MOVE can combine with "direction* (TOWARDS), as in the following sentence: The dog was running towards me. They were going south, She t m e d right. She was walking away from me. It is quite impossible to paraphrase such sentences in a way which would dissociate "direction" "om "mc~vement". In fact, the two notions ("movement" and "directioaa'" appear to be linked so dosely that I have even suggested that MOVE and MOVE TOWARDS should perhaps be considered as two valency options of the same Iconmpt (rather like DO and DO TO, ar SAY and SAY TO). But cross-linguistic evidence does not support this suggestion. In many languages (including Pdish) the verb ~omspooding to the proposed primitive MOVE does not combine with directional elements like forvwrds, although more compllex verbs of movemermt (such as 'goymd 'comeli) do. Clearly, the matter requires further inlrestigatbm-as does also one fmther possible valenq option of MOVE, discussed at the Semandcs Symposium in Ga~~bcrra in 1 5 5 1 5 5 1 4 : ""person X moved body-part Y"".

As mentilaanled in Chapter 2, sentences of this kind do not have to include any "existential verWyand they may convey their existential meaning iq different ways, but it can be expected that some lexico-grammatical means for conveying that meaning will always be avaiiabb. Sometimes, existence may seem dificult to distinguish Dom locatio~ (to be SOMEWHERE), and aften BE and W E R E share their lexical expoents (see e.g. I. Lyons 1977, ii. 723a; Clark i19709:
( 4 4 ) There are two people in the garden. [B) (Where is everyone?)
Two people are in the garden [and three in the house next door], ut the very fact that existence and location can co-occur (as in the sennce: "there is a lot of water here") %hawsthat they are not different pects of the same notion, and I would hypothesize that, despite overlaps, aU hnguages the two concepts ] i n question can be overtly distinguished. It should be noted, however, that (despite the abundant litmature on the bject) the relationship between existence and location requires a great al of further study. 7.2. LIVE (ALIVE]
VE is a wcry rcocfilt addition to the stet of primitives, and at this stage kitis known about its grammar. One could venture to say, however, that it a predicate, and that it opens a slot or slots for temporal adjuncts:

7. Existence and Life: BE (THERE ESjARE] and LIVE

7.1. BE
?"he predicate BE co-owurs with the "substantiwsY3SOMETPIING,

These people l i d for a long time. These two people Eved at the same time. This person was alive at that time. uPd appear that for the purposes of classification, LIVE cam also bz thout any adjuncts:

SOMEONE, and PEOlPLE, and with the "classifiers" KIND amd PART, usually in combination with same further determiners and frequently with

These things are living things (= Piwe?). On the other hand, co-occurrenm of LIVE with spatial adjuncts is p r o k bly not universal, alithough two separate types of sentences need to be distinguished here: those referring to temporary residence (e.g. '2 live in Canberra'" and those referring to permanent living conditions (e.g. "Fish n the first of them types live in water"). The nee of the expomnts of LIVE i is mrtainly not universal; their use in the second type relqluires further investigation.

8. Determiners anid Quantifiers; THIS, THE SAME, OTHER; ONE, TWO, MANY (MUCH], SOME, ALL
8.1. THIS
THIS has a wide range of roles, because it has, so to speak, a double claws membership: it can function as both a "deteminer" and a c'substantive'y. (Sometimes two different f o m s have to be used in these two different roles, e.g, kore (substantive) and k o n ~ (determiner) in Japanese, but as 1 have argued in Wiembicka (19911b), these two f o m s can be regarded as two allolexes of the same semantic primitive.) As a "substantive"",THIS can occur in the role of a "subject'bf evaluation, de~ription,relation, or lcmtion (that is, as an "evaluatum"', "descrlpturn", "relatarnn", or 'ciocatum"): This is goo&bad. This is big~'smaP1. This is like this other thing. This is abavdunder this other thing. This is not in the s m e place. Like all the other "substantives", THIS can also oocur in the: '"atient" role: Something bad happened to this. Like one other "substantive", SOMETHING, it can also occur as a "cornpleanent'" with DO, HAPPEN, and SAY, and with at least four mental predicates (KNOW, WANT, SEE, and NEAR): I did thislthis happened to me. I said thidI b o w thidI want t h i d see this/I hear this. It can also occur in a "predicate nominal" role in relational sentences:

At the m e time, THIS can owur i n the r o b of a "'determiwer'Yia combination with other "substantiws": this thing, this person, these people, (in] this plaoe, [alt) this time, this kind, this part. Furthermore, THIS am comibinne with some of the other "determiners", notably with OTHER, ONE, TWO, and ALL: this other person, this one person, these two people, all these people. It can also combine with the element LIKE, forming with it a quasi-determiner ""le this", e.g. someone Pike this, something like this. ""Lke this" is is important semantic molecule, often realized as a single portmanteau morpheme like so [or such) in English. This molecule can combine with all the "opposites" in the semmtic system: so good, so bad; so big, so small; so far, so close; so long, such a short time.


m e universal syntax of THE SAME is not clear at this stage. It appears, however, that it can, mniversallly, function as a "determiner" and can combine with the "substantivm" SOMETHING, SOMEONE, and PEOPLE (though not with YOU or I): the m e person; the same peopb; *the same you; *the same I It can also combine with ""pace" and "time": at the same t h e ; in the same place and with the classifiers: the same part; the same kind It is also likely that THE SAME can, universally, play the role of a "cornplemenb"s and so function as a quasi-substantive, for example:

I did the sanrejEI said the same. I though~wante~feQ.tiE(?)knew the same. The same happened b me.
Like THIS, it can also occur in the role of a "psychological topic":
I thought about the same.

Last but not least, it should be said that presumably in all languages THE SAME e m be used not only anafiorically but also cataphorically, and that in the latter case it opens a syntactic slot for the second member of the equation: THE SAME AS. For example:

A 1 these other things are under/abovnsde this.

All these things are like this.

I ~d the sarne as you. This thing is of the sarne kind as this other thing.

130 General Issuers Like the 'cdeteminers'WTHER and THE SAME, MORE may @haps) open a dot for a complement of its o m , for exampler more than two thhgs of the same kind Apparently, MORE can also combine with several determiners: much more, one more, two more But I don" think that this analysis is valid. From a moral point of view, it may be imporaa~lltto distinguish something that is "good for a person" from "something good that has h a m e n d to a person'". For example, for many marall teachers it may be important to be able to say things such as: When something bad happens to you, it may be good for you. If good things always happen to a person it may be bad for that prson.
A language which wouldn't be capable of expressing such ideas could be regarded as impaverished, and we can hypothesize that alh: ]languages are capable of expressing them.

Perhaps the central role ccattributes":


ithe evaluators GOOD and BAD is that of

11. Descriptors: BIG and SMALL

The descriptors B E G and SMALL have, primarily, an "attributive" role:
(a] bi@small thing (see Jso: something big)

someone goodhad (a goodjibad person) something goodhad (a goodjibad thing) g o o a a d people (Whether or not, or to what extent, such an attributive use can be extended to times, places, parts, and kinds remains to be investigated.) It is not entirely dear whether GOOD and BAD can be used (in nonelliptical sentenoes) as predlicates, as in: You are good. This is good. or whether such sentences shodd be regarded as non-elliptical versions of sentences with attributive p h ~ a s e s : ~ You are a good person. This k a good thhg.
It appears, however, that at least in some contexts GOOD and BAD c a ~ be used predicatively; i n paticular, that they can be so used with respct to 'kieElausal subjects":

(a] bigismall person (see also: someone big) bigkmall people (a) biglsrnaS11 place (a) bi@srnalI part On the face of it, they cam also be used predicatively; far example: These people are bi@small. As in the case of GOOD and BAD, however, it is not dear whether such a "predicative'bse shodd not be regarded as a csypto-attributive use: These people are b i ~ s m d pheopje. l Sentences such as: This is bi@small. may or may not be universallly available, but even if they were, a case could perhaps be made for regarding them as dlipticd, since the very notions of BIG and SMALL imply a reference to s o w standard of cornparisom.

1f someone does something like this, this is ba&good. Perhaps the most interesting question which arises in connection with the evaluators is that conwrning the relation batween GOOD and GOOD FOR or between: BAD and BAD FOR. My suggestion is that GOOD and G W D FOR repesent two different valency options of the same primitive Ojust like DO and DO TO or SAY and SAY'TO do). Admittedly, one could try to reduce GOOD FOR to G W D along the Falllowing lines: This was good For m. = bwatase of this, something good happened to me

WHEN [or AT A TIME) can be used, above all, as a 'kclause adjunct"". It is obligatory in DO, HAPPEN, and MOVE sentences, and possibly m i n SAY sentences:

* This mattes was 'brou~ht to my attention by Nick Enfield (gewond communication).

At that time, you did something. At the same time, something happened to me. At s o w other time, this thing moved. At that time, E said something to this person. To same extent, temporal adjuncts can no doubt also combine with mental predicates: At that time, I thought that he was a good person. At that time, I wanted to do it. At that time, I felt something bad. At h a t time, II didn't know ananything about i t At that time, I sawheard something. But the exact nature, and cxtcnt, of lhcse co-accurrcnccr;.lrcqunircs further iinvestigationr. In many languages, the exponent of WHEN can also be used in a biclausal construction, in which it Ewnctlons as an interclausal linker (cf. When E did these things, I felt something bad. It can be argued, bowever (as suggested by Goddard, personal c o m r r l cation), that when used as an interclausal linker, the English word whm, and its counterparts in other Ira~nguages, stand not just for the primitive WHEN 0.e. AT A TIME] but for a semantic molecule combining WHEN

In past and future tenses, the elements BEFORE and AFTER are COWBined semantically witrith the element NOW ("before now', '"after now'). But i f the basic temporal element WHEN [AT A TIME) operates, primarily, as a clause adjunct, the two time "modifiers" BEFORE and AFTER, which serve to establish the temporall sequence of two events, are often used as dause linkers (because the two events in question may well be referred to in different clauses). For example: You were born before P was born. I was born after you were born. Something bad happened after you did this. This happened before H saw you. Cross-iinguistic;allyl however, the most common use of thc ellcrncwt ER is probably in a narrative, where phrases meaning Wter tRisTCand ') arc used to introduce a i~ew event. It is likely that BEFHOIRE, too, is primarily in comlsinaitiorn with THIS ybefsre this"") It is important to however, that in phrases such as "after this'' and '"before this"' *'thisy' o refer to the content of an entire clause. urning now to the two dumdonal concepts A LONG TIME and A TIME, un;e will note that they combine, first of alll, with the pred0, HAPPEN, MOVE, and LIVE, and also with all the menta! I was doing it for a long tirne. It happened a long time ago. I felt something bad for a short time. He lived for a long time.

and THIS:
At some time before now, I did these things. At this time, I feet something bad. The element NOW cannot serve as an interclausal linker, and it cannot take "deetdners" (compare, for example, at ski$ time versus *this now, and a 8 the same time versus *"the same now), but otherwise the syntax of NOW appears to be similar to the s p t a of WHEN: NQW can combine with the mental predicates, with DO, HAPPEN, MOVE, and LIVE, with spatial predicates, and so on. For example:

3.1. WHERE and HERE

P now thinWknowiwanltSfeeYseer'hearisay. . .
This penon is moving naw. These people live now. The elements BEFORE and AFTER can perhaps be regarded as spwial modifiers (determiners) of time adjuncts, cormparable to the universal determher THE SAME at the same time; before ebis time; after this time

WHEN, WHERE (IN A PLACE), too, can be a clause adjunct, but ange of predicates with which it GO-OGCUPS is of course difterent: esseny, it w-occurs only with HAPPEN and DO (and possibly MOVE and
l a happened in this place. I did it in another place. Something moved in this place.

. Sentences such as:

s with mental predicates do not have a place slot, exapt perhaps

71 thoughtknewiwantd it in another place.

if aoceptabfe at d l , must be regarded as elliptical. As b r as FEEL is concerned, the situation is unclear. The relation between SEE and WHERE (IN Jar. PLACE), too, is at this stage far from clear. Can sentences such as I see something over there. be reduced to a combination of sentences about seeing and senltences about existence and location, along the lines of I see something this something is over there (there is something over there)? It could be argued that they cannot. For example, sun oasis that I can see in, the distanoe may be simply a mirage (and so may not be there at all, in the phce where I see it), and an apparition which someone can see in a place may not be really there. (Recall Berkeley's CLJ13) point that the sentence "1 see a silver speckin the sky" doesn" imply that "there is" a silver s p k in the sky.) Unlike WHEN phrases, however, WHERE phrases can also be predicates in their own right [as BE IN A PLACE or BE SOMEWHERE). For example: This thing is in this place. I know where it is. I was somewhere else.
As predicates, WHERE phrases can be ~ombinedwith ail the Ccsu'bstaa-

cept: 'Wistance in space" and "distanoe in time"" I would reject such an argument, however, because the notion of "distance" is, i n my view, inherently spatial, and phrases such as "distance in time" ';are metaphorical. Furithemore, while it could be mgued that the idea of "'boundaries'" w i f e s t e d in synrirctic frames such as 'Tram-to" is relevant to both 'WsItanwy3 and 'cdurationy',
[PI) I did it for a long t h e (from 9 to 5).

CIE) It is very far (from this plaae to this other place).

it should be pointed out that the 'Tram-toYVrame plays a different role in, wch mse. In fact, the notion of EAR is inseparable from that of c'from-toyQ; but the notion of A LONG TIME does make sense even without any explicit or implicit reference to "boundaries"":

I did it for a long time (never mind from when to when]. It is far away (?never mind from where to where).

E submit, then, that the idea of "distance" ((EAR and NEAR) makes sense
only with reference to two sped& places, whereas the idea of "duration" [A LONGE'SHOWT TIME) does not have to refer to two specific times. Conselquently,the "from-to" syntactic frame (or an alternative frame discussed below) is obligatory in (non-elliptical) sentenoes about distance, but ornly optional in (non-elliptical] sentenoes about duration. The alternative frame mentioned above takes in English the form of the prepositionfiom, without an anocompamrying $0;for example:
This thing is too far from m e H can" teach it. You me too far from mie-come closer.
hat I I [ suggesting, in effect, is that while the predicate FAR always quires two referenoe-points, these refereme-points don't have to be aces, but can d s o be people or things:

tilres" (YOU, I, SOMEONE, SOMETHING, PEOPLE, and THIS]. Finally, the syntax of HERE appears to be similar to that of WHERE, as the syntax of NOW is similar to the syntax of WHEN. Natiplrarllly, HERE (like NOW) does not combine with "determiners" (in another place versus *amother here). In contrast to the relatianship between NOW and WHEN, however, the two spatial primitives can sometimes co-occur, as in the phrase ""smewhere here". Like WHERE, HERE can allso be used priedicativeiy, as BE HERE. For example: I am here. 13.2. FAR and NEAR At first sight, the concept of "distance'~clhow far?") appears to be parallel to the temporall concept of "duration" ('for how long?'), In fact, it codd even be argped that the twa represent two different faces of l t k same can-

(A] (10 is far from place A to place B. ( 0 ) Thindperson A is far from thindperson B.

or example, I suggest that when we say af an object that it is long, what mean is that (what is ~anoeptualizedas] the first part of this objcxt is From (what is comnceptualized as) its Bast part. In other words, I suggest t perhaps we conceptualize length in terns of a distance between things a&) rather than between piams.
3.3. UNDER and ABOVE
Turning now to UNDER and ABOVE phrases we will note that at first sight they may seem to be similar in their functioning to the temporal

3. CPFFa'xpessal G r a m a s 113 7
modifiers BEFORE and AFTER. If the latter we to be interpreted as modi h r s of the temporal notion AT A TIME the former can be sets as modifiers of the spatial nation IN A PLACE WHERE):
ThL thing is wnderPabova this oll~cr thing. = this thing is in the place undedabove this other thing

Ebr some ccdeterminer'L, for example, "this'" whereas Y stands for some "substantive"-prototy~,ically# a person).

This interpretation, however, is not without some problems. To begin with, if we want ta say that, for example, The sky is above everything. do we mean that the sky is in a phm above everything? Or if we want to say that The head is above all the other parts of a person's body.
do we m a n that the head is in a gi8zloe [which is] above the places where all the other parts of a person" baddy are? An interpretation abng these lines seems counter-intuitive. It may be more j,ustified, therefore, to regard UNDER and ABOVE as relational, rather than strictly locational, notions, This approach would also sdve another probllem, namely that of the metaphorical ;appLatiorz of the concept ABOVE (if not UNDER) to peopie (with reference to their status, podtion, power, and so on). H t is quirte likely that the metaphorical use of the notion AB07JE with referenm b people is universal and that, for example, the idea that God is "above all people" can be rendered, and be wndewtoad, in a4 languages [sele Wierzbicka forthcming b). But this meltaphar would make little sense if it were to k interpreted in terms of places. It is true that "heaven" (166od"s place) is, metnphoricalllg speaking, a place which is "aboveY'the earlth @eaple's slam). But the metaphor of "heaven" is distinct from the metaphor of between Gad and "father'" and the Uatlter has dearly to do with the relatio~ people, and not Ibetwtxn two plaoes. 13.4. SIDE @INWHAT SIDE)

The concept of JMSIDE may seem to be related to thalt of SIDE C u O N WHAT SIDE), but H believe that the links between the two are not comitional: a sentence such as "A is inside iB'3oes not mean that '2is on inner side of B'" h a u s e SIDE (ON nHIS) SIDE) implies an adjacent oeation, and the notions of 'insideknd "djacencybre not mutually comThe syntax of the two concepts CON WHAT SIDE and INSIDE) is not identical either, since one of them opens three slots, and the other, two: A is on side Xof B* A is inside B. s not clear at this stage how many predicates can combine with INSIDE. ly, WHERE (be SOMEWHERE] an-so much so that we might be to view HNSDB as a spiaE case of "being somewhad"' But things "happen" ininside something (a house, a cave, a womb), and so we ouPd allow, perhaps, that WSIDE can combine directly with HAPPEN.

14. Interclausal Linkers: BECAUSE, IF, IF

. . . WOULD

terclausal Ilinken constitute a powerful device for building complex antic structures out of simple propositions. One such linker, W E N , already k e n introduced in Section 12, on time. But the primary funcof WHEN is, arguably, that af a temporal adjunct. By mntrast, the e elements discussed in the present section are primarily, or even ex&iwly, interclausal linkers. BECAUSE ECAUSE @an function either as an interclausal linker or as a dause unet. Arguably, its primary rale is that of a linker, as in the folbwing tenoe:

The concept of SIDB is used to indicate location (WHERE) of people, things, and even places: This person (thing] is on this side now; before, it was on the other side. In addition to the pewqn, thing, or place whose location is being descrilxd, SIDE requires also a point of reference. Often, this poimt of reference need not be mentioned explicitiy, being provided by the person of the speaker3 (PP the person spoken of. The most important point about the grammar of SIDE is that it is not a "substantive", and that its full frame is "on side X of l'"(where X stands,

o n the head (not because of someThe dog died because the man hit it c thing else).
s a clause adjunct, BECAUSE commonly occurs in phrases such as of this'" in dauses whicch present an event on. a state of affairs as from that described in the preceding sentence; for example:

3. Utriversal (C;F;CSJMII~EJT 139

The man hit the dog on the head; because of this, the dog died. Frorrv a logic01 point or vicw, onc would expect that BECAUSE always links events, and therefore that it has to connect a clause with another clause or with a clause substitute (a "substantive"-THIS or perhaps SOMETHING-referring to the content of another clause). In natural langwg, however, the role of BECAUSE does not seem to be similarly restricted, and phrases such as "because of me" or "because of you" may in fact be uniwersalfy available. If they are, then there is perhaps no need to regard them as elliptical or polysemous. (See Jackendoff 1983: 176-81.

He said it as if he didn" know anything. We can hypothesize that in the unmarked order ofclauses connected by IF, the IPc8ause comes first; but the IF . . . WOULD-clauses can come second, notably when they occur in conjunction with LIKBAS. (See Section 2 1 1 in Chapter 2.) W y is it that a sentence combining the elements WANT md NOT can be interpreted in two different ways? Far examplie, why can the sentence " I don't want to go" be interpreted not only as denying that "I want to go" but also as affirming that "']I want not to go"? And is this phenmenon universal? The matter requires further investigation. Other than raising this question, however, I wilI not dismss the semantic syntax of negation amy further in the present context. (For a wealth of relevant observations and ideas, see in particular Jesprsen 1917 and Horn 1989.)

IF is another iterc8ausd linker. Pm contrast to BECAUSE, however, it can only combine with a clause (as a part of a complex sentence); it cannot combine with a substantive THIS or SOMETHING substituting for a clause: If you do this, people will say something bad about you. *If something, people will say something bad about you. The phrase "if not" may seem to provide a munter-example to this claim, but it is probably not universal. {It cannot be universal if it is true tbait in some languages negation is realized only as a verbal suffix; and R. M.W Dixon @ e r s o d communication) informs me that this is the w e in the Amazonian language Jarawara,) In English (and in m n y other languages) the IF-clause can follow the ather ("main") clause of an IF-sentence, but this option doesm" t e r n to be univelvalIy available. [For example, according to Tien (1994) it is not available in Chinese.)

15. Qause Operators: MOT and MAYBE

15.1. Negaltion: NOT

egation is, universally, a "clause operator"'. Remarkably, it seems to be otally unrestricted: apparently any clause, of any kind, in any language# a m be negated. AS mentioned earlier, one difficult problem which arises in ction with negation is its relationship with "'wanting'" The matter es further investigation.

15.2. Possibility: MAYBE

sibility is commonly realized by means of a particle ear sentence adverb, h can allso be regarded as a "clause operator": Maybe it will1 rain tomorrow. Maybe this person did something bad. 3, the same meaning can be realized by means of an auxverb, or by means of a bound grammatical morpheme, but its sernanntax seems to be always the same (a "clause operator"). ere don't seem to be any restrictions on the combinabillity of MAYBE different types of predicates. It combines with action, event, and Maybe you will do it. Maybe it happened. Maybe someone else said it.
with mental predicates:

. .WOULD The primitive PF . . . WOULD, too, is primarily an inlterclausal linker:

14.3. I F .
If you had been hese, sir, my brother would mot have died. (The New Englkh Bib[@, John 11: 2131

P f I w r e you E wouldn't do it. In many languages it can also introduce a wish c1ause (e.g. "if only I were there . . ."')I, but these can probably be regarded as elliptical, and, in any case, they are not universaI (far example, they are not available in the Austronesian ianguage Mangaaba-Mbula; see Bugenhagen forthcoming]. Unlike IF, PF . . . WOULD can also be used as part of a complex interclausal linker "as iB*"that is, I F . . . WOULD plus LIKE]. For example:

Maybe this person thimnkcsSknowsJwant~1/fee1dseesShea~s tbe same.

1411 Geflcral13.rwes

and also with predicates of description, evaluation, relation, and Eomtiomt

Maybe this is gaodlbad. Maybe this thing was a biglsnrall thing. Maybe this thing was like time other things. Maybe all these people were in the same plam at this time.
In this respect-its unconstrained combinabllity-MAYBE is like negation; for this reason, it may indeed be better to view it as a ""cause operator'9ather than; a "'dause adjunct". None tEre lea, MAYBE is not quite as unconstrained as negation. In "surface syntaxy' it doesn't combine with the imperative: *Maybe don't do it! and in the "semantic syntax'9t doesn? combine with "mental predicates" In first parson (present tense) sentences, except in jocular or playful usage: ?Maybe I don't want to do it. ?Maybe I think about something else. ?Maybe I feel something bad. Tlre reason is that MAYBE implies tbat I don't h o w sornlet21ing, and nord l y lone is expected to know one's own current mental states.

This thing cam move. When something bad happens to a person, it a n be good for this person. I know: something bad mn happen to me.
It is not clear at this stage whether there are predicates which cannot combinre with CAN at all.

17. Intensifier: VERY

The intensifier VERY combines, first of all, with the evaluators GOOD and B A D and the descriptor BIG: very goodivery badsvery big Presumably, it also combines, universally, with the "'"determiners" Elrl'OJCWAW*and also with the "distance primitive5TAR,and with the durartional primitive A LONG TIME: ve;ryr mucMvery manylvery faria very long time Curiously, preliminary cross-linguistic bsting suggests that VERY does not always combine as readily ;and as freely with the, so t c a speak, "small" primitives SMALL, NEAR, and A SHORT TIME as it does with their opposites. Bath the scope and the nature of these restrictions need to be investigated. In many languages, W R Y (or some allonex such as very much in English] combines also with the mental predicate WANT:

16. Meitapredicate CAM

The mnnapredicate CAN mmbines, first of all, with the action predicate DO, and perhaps, groltotypically, with "'I" as an agent. Sentences such as:
I cm't t o Id1 c a m do it. I can't do it n w / I could do it before.

I want it very much

It is daubtfd, however, whether this is universal, if only bemuue in some languages [e.g. in Kayardild, see N. Evans 1994; also Harkins 1994) 'WANT is realized as a suffix, not as a full verib or adjective. In most languages there are of course onher "intensifiers'" such as ar all, redy, rear',proper, and trw in English, and their combinatorial possibilities may go far beyond those Ested here far VERY, brat the meaning of these 'ciintens&ers" may be different from the meaning of VERY.

reflect w r experience of our own limitations, and also, of our freedom (within certain limits). Next to action, movement is perhaps another prototypical area within which CAN, and CANNBIT, is most salient: I can" move. But-phaps by extension fmm these experiential prototypes-other bicates, and other "substantives", too, can combine with CAN: Yodsarneone/people can do it. Hi'yod'khisperson can say the same. I can" think about it for a long time. But CAN can also occur in sentences with non-peaonail ""sbjats'" for example: prd-

18. Taxonomy, Partonomy: KIND OF, PART OF

The notion KIND co-occurs, as a rule, with a ccdeteminer'y: this kind; mother kind; the same kind one kind; two kinds; many kinds; all kinds

In English, and in many other languages, one can also use KIND in sentences without a determiner, for example: An oak is a kind of tree. but this usage doesn" seem to be universal (see e.g. Dmie at al. 1994; or N. Evans 1994). On the other hand, evidenlw suggests (see Goddad and Wierzitpicka 1994b) that in all languages one can say things such as the following: There are two kinds of bat. This is not the same fish, but it is the same kind d fish. Presumably, it is also possible to say, in m y language, the structural equivalent of the sentence This fish [bird, tree, eltc.1 is of the same kind as this other fish (bid, tree, etc.1. (Of course the concepts of 'fish", "irdl", or Yree' as such are far from miversal. See e.g. C . Brown 1977, 1979.) The notion of PART is at present more problematic than most of the other oms, and little is known at this stage about its syntax. On the basis n metof the data available, one wodd expect to find PART, universally, i alingulistic sentences such as the foilowing ones: A blade is a past of a knife. A stump is a part of a tree. A foot is a part of someone" leg.
I w u l d also expect that. the concept of PART can be found, uniwrsally, in sentences of the following kind:

with the concept of THERE I[XiARE; and also with ccdeteminers'a(e.g. TWO, MANY, SOME]: The elephant has a lomg nose. = when people want to say sometlying about things of this kind [ELEIPIIAJkITS] they can say something like ttnis: this part lJ+JOSEIis long 19. Similarity: LIKE The combinatotial possibilities of LIKE are probably quite varied. To begin with, it can act as a predicative ""nnker",linkinng two "substantives": You are like me. I am not like other people.

It can also act as an "attributive linker" of substantives:

someone like me; something like this; people like you But Were are restrictions here:

like you; *you like someone else *someone like someone ellse; *people like someone else

In some languages, combimatjons d this kind are realized in the f o m of obligatory or almost ohligaltory portmanteaus. For example, in Polish they would normally be rendered as follows: ktok taki jab: ja someone such as I cot5 takiego something such-GEW. ludzie tacy jakc ty people such as you But difirences of this kind can be regarded as superficial. As an 'kttributive linkery'L1LIKEcan also apply to time and place: at a time like this; in a place like this P n attributive phrases, then, the "head" [the compared member] can be one of the following ('5ndehite") set: SOMEONE, SOMETHING, PEOiPLE, TIME, PLACE; whereas the point of referenae appears )to be restricted to the fol10wing (''definite'? %set: THIS, ME, YOU, THIS PERSON,THESE PEOPLE.

An axe has a handle and a blade. A knife has two parts: a handle and a blde. A Bower has many parts which look alike. Needless to say, I am not suggesting that all languages have a verb corresponding to the English verb have. Rather, I am suggesting that the English word hve, when combined with a noun including in its meaning the concept of PART (such as blnde, handle, stump, or foot), can perhaps be viewed as an allolex of the ""eistentid" predicate THERE ISIARE. The fact that in many languages this is precisely how the 'c"prt-~hole"relation is expressed seems to support this idea:

PL knife has two parts. =

'% knife, there are two parts (in iuto it]''

I arm suggesting, then, that the concept of PART can combine, universally,

144 General Iswas

But LIKE phrases can allso function as "clause adjuncts", at least in combination with the predicates DO, HAPPEN, and SAY: It happened Pike this: . . I did it like this: . . . He said it like this: . . .

[description] [be] BIG, SMALL Same d these predicates can combine into a unit with the "'metapedicate" CAN. Different predicates take different types of "'ciomplements" and "abjects" [as discussed under individual! predicates). In addition to predicate complements, some types of clauses (depending on the nature of the predicate) @doeaQso clause adjuncts: temporal, spatial, and causal (e.g. at this timeJim this plaodbecause of this). A11 types of clauses cam combine with the two universal "cPause operators": negation and possibility (MAYBE). These two operators can cooaur, with NOT being within the scope of MAYBE (but not the other way rounrd): Maybe they didn" do it. "They didn't maybe do it. Clauses can be combined into comp3iex sentences by means of 'c1"8ikem'*: IF, IF . . . WOULD, BECAUSE, AFTER, BEFORE, WHEN, and LIKE. A clause can be turned into an adjiunct to another dause and thus 'Yncorporated" inin t o t if It is replaced ( i m p discourse) with the ""sbstfilmliiveY"THIS (accompanied by the clause Pinker): after thishfore thishemuse of thisflike this. The only clause linkers which don't aHow "cdause incorporation" and which are, universally, strictly inbrclausal, are IF and IF . . . WOULD. En addition to )the use of linkers (IF, IF . . . WOULD, BECAUSE, AFTER, BEFORE, LIKE) clauses can also be combined by 'hesting" (to use Weinreich's (1963) term), in the sense that they can be used as complements of certain predicates. Thus, the predicates SAY and THINK can take, universally, c'q~otative'~complements; for example: You said: I didn't do it. I thought: this will never happen to me. The predicates KNOW and WANT, to^, can take '>ropositionaU compllennents"";~ example:
I know: people sag something b d about me. PeopEe know: you didn" do it. I want you to do it (= I want: you do it]. T wmt to know it (= I want: I know it).

It appears that as a dause adjunct, and also as a time and place linker, LIKE is restricted to a combination with THIS. It seems likely &at LIKE can also function, universalllg, as an interclausal linker, as h the following sentences: Forgive us as (like) we forgive other people. You want to do good things for me, as (like] we want to do goad things for our children.

20, &nerd Discussion

A s c m be seen from the above survey, the syntax of the natural semantic metalanguage can be characterized as andogous in some respects to, but much simpler than, the syntax of natural languages. The basic unit of the WSM syntax is a "clause", which is constituted by a "ssulbstantive" and "a predicate"', and some additional elements determined by the nature of the predicate. In addition to this major type of clause (to be discussed betow) here is dso one minor type, which can be regarded as an analogue of ccsubj~Eess sentences" of the traditional grammar, and which includes 'kxisbntial sentencesn~entred ion the predicate THERE ESI'PLRE [e.g. '"here are many kinds of birds"; '?!here is p9ienty of water here"). The set of "sunbstantives" hcPudes the elements I, YOU, THIS, SOMEONE, SOMETHING, and PEOPLE. Some of these (the last three) can be combined into a wnit with "determiners" (THIS, THE SAME, OTHER, ONE, TWO, SOME, MUCHlMANY, LETTLWFBW, and PILL] and "attributes" (G00131, BAD,BlG, m d SMALL]. The list of possible predicates includes the following elements:

[mental predicates]: THINK, W A m , KNOW, FEEL, SEE, HEAR [speech] SAY [action and e~emrsj DO, HAPPEN, MOVE [existence and life] BE, LIVE [relation] LIKE; be] a PART of in a QLACR [ k ] UNDER, [kl ABOVE; [Ruej ON [this] [spam] SIDE; [be] INSIDE; [he] FEAR; be] NEAR; [be] HERE [evaluation] [be] GQQD, BAD


One element (THIS) has, universally, a double status and can fumction either as a "substantive" or as a "determiner". The evaluators and descriptors (GOOD and BAD, BIG and SMALL] can fiunction, universally, as "attributes" (to the 'csubstantives" SOMETHING, SOMEONE, and PEOPLE] as well as predicates.

146 Geflmal Issues

As can be gleaned from the above discussion, NSM has a rudimentary parts-oCspeech system. Elements which cam occur only as predicates (THINK, KNOW, SAT, WANT; SEE, HEAR, DO, HAPPEN, MOVE] ~han be regarded as analogues of verbs; and those which can be used either as predicates or as attributes (CXIOD, BAD; BIG, SMALL) can be regarded as analogues d adjectives. Elements which a n function as "subjwts" and which can rake attributes (SOMEONE, PEOPLE, SOMETHING] can be seen as analogues of nouns, and tihose which can function as '%ubjecitsM without behg able to Lake attributes (I, YOU, THIS) can be seen as malogues of pronouns. The elements which can combine with "substantives" but which cannot owur predicatively {THIS, OTHER, ONE, TWO, SOME; THE SAME) can be regarded as an analope of determiners, and those which can only owur in conjunction with substantives (PART and KIND) can be seen as aun analogue of classifiers. Elements which can link clauses (BECAUSE, IF, IF . . . WOULD, LIKE, AFTER, BEFORE) can be regarded as analogues af conjunctions, and those which can turn a substantive into a predicative phrase (UNDER, ABOVE, ON (ONE) SIDE OF) can be seen as an analogue of prepositions. The universal intensifier VERY, which combines with the attributes GOOD, BAD, BIG, and SMALL, with the "determiner" MUCWMAW and with the spatial and temporal predicates FAR and NEAR, A LONG TIME and A SHQWT TIME, can be seen as an analogue of one type of adverb, whereas temporal and spatial clause adjuncts [WHEN, WHERE, A LONGBHORT TIME] can be regarded as an analogue of another type of adverb. Finally, the "clause operators" negation (NOT) and possibility (MAYBE) can be regarded as an analogue of sentence particles. In addition, NSM has powerful iconic and indexical maehaniisms, extending its grammar far beyond the boundaries determined by the combhability rules sketched above. Consider, for example, a typical NSM explication such as the following one (see Chapter 5):
I feel sad. = (a) sometimes a person thinks s o m e t b g Ilie this: (b) sometl~ing bad happened if H didmy know that it happened I would say: P don't want it to [c) happen Q d ) I don't say this now Ife) because H know: I can't do anything (fl because of this, this person feels something bad xCgS I fael samething like this

In a sense, the whole fornula could be seen as one complex unit [an malogue of a '"entenm"") This "sentence" indudes as marry as seven clauses, which jointly form an integrated whole. But the grammar off this whole goes beyond the links established by the linkers BECAUSE and IF, and by the complement structure of the verbs T H I M and WANT. One important linker bindisng the clauses of this sentence imto a whole is {el, {A> or catthe element THIS, referring either anapholricalily [lines ( c ) ~ a p h o r i d y {line [EF)] to whole darmses. In the written form of NSM, there are also other devices which play an important auxiliary role. These include special spacing and indentation. The fact tbat a part of an explication is placed in a separate line: indicates that this part f m s a distinct semantic component, and a g o u p of such components identically indented under a component including the dement THIS (e.g. (b),{3,(4'and (e) in the explication above) form a larger unit. In om1 speech, NSM formulae would no doubt be often difficult to fob law, but in principle a system of pauses a d some rudimentary intonational contrasts could perhaps achieve the same effect. As mentioned earlier, since NSM is intended to be a model of the innate and universal ditag~a metatok, the NSM grammar sketched here mn be seen of human cognitionM".ut there is no s a hypothesis a b u t the ccgraanmaa reason why one could not mmain sceptical about the status of NSM (both in its lexicon and in its grammar) as a model d the language of human cognitian and yet recognize its value as an eiffective universal system of semantic-and cultural-notation, And, to repeat: the sketch grammar of NSM proposed here is highly tentative and is oflered o d y as a necessary starting-point for testing and

Prototypes and Invariants

2. Abuses of ""Prototypes" i n Semantics: Solme Illustrations

2.1. The Meaning of Boat
Discussing the meaning of the English word boat, Verschueren (1985: 4831 says: In trying to determine the meaning of the word BOAT, one could come up with a definition such as. a 'man-made object that can the u s 4 far taaveBPing en waier'. A defender of the checklist approach, coming across a boat with a hole in it and deciding that heishe stiilll wants to call it a BOAT (though it camot be used For travelling on wabr mymore), would have to revise hisfher definition: 'a man-made object that can normally be used ;far travelling on water, but in which there can aPso be a hole'. Further, h&he would have to determine how big the h d e can be before the object i m question is not a BOAT anymore, ibut simply a WRECK.The impracticality of the checklist appl-oach is sucUn that nuol even iils praponents would want to be guilty of the absurditties mentioned. A defender of the alternative theory could simply stick to hisher definition and describe a boat with a hole in terns of deviations from the prototypical boat. But instead of appealing t o prototypes, cauldn't we simply rephrase the fornula just a little? Coddn'li we say, in the first place, that boats are a ind of thing ma& for 'travelling on water'rather than able ro, "travd on "? TIt is quite true that a boat with a very big hole can? "travel on but why phrase the definition in terns of abiifty rather than dfunaioy3 anyway?

The role that the concept af pratetype has played in contemporary seman. tics is analogous to that which the mnlcnept of Gricem maxims has played i n generative grammar. A well-placed witness, James McCawley (1981: 2151, identified this role with the excellent slogan "Grice savesm".n grammar, if them was a conflict beiween gosdilnlaled rules and m h c aetwdl usage, it was assumed that "Grl.c&kowld rescue the grammarian: the usage could be accaunted far in k m s of Gricean maxims. (See Bach and Warmish 1982; for a critical discussion, see G . Green 11983;Wierzblcka 114191a.) Similarly in semantics. Just as the failure of gramanalticaN rules to work has often been presented as evidence of progress in linguistics (because It onty illustrates the importance of G r i m maxims), the failure of semantic. formulae t o work has often been presented as evidenloe d grogess in semantics. Semantic f o m d a e should not "works'; t o expect them to work means not to understand blire role of 'cprototypes" in langwage and oognition. Freqnendy, appeals t o prototypes have been combined witb a claim that t h r e are two approaches t o h u m categorization: the "classid" approach (linked with Plristotle) and the "prototype" approach Oinked, in partiadzas with Rasch and Witbgenstein). When these two approaches were contrasted, it was usually argued that the "classid approach" was wrong and the "prototype approach" was right. In this chapter I argue that the idea of contrasting these two approaches m semantic investigations, and that what in this way has proved unhelpful i is needed is a synthesis of the two traditions, not a choice of one over the other. There is a place for prototypes in semantic anagysis, but there is also a place for imvariants: one does not exclude the other. kmorditngly, in what folllows, I will discuss two sets of examples. The first set d l illustrate the t m d e n q t o abuse the concept of poliotype (the "pratoiliypes save" attitude); the secand set of examples will illlustrate the usefuhess of this concept when it is used as a spedfic analytical tool and not as a universal thought-saving device.


2.2. The Meaning of Bk~~chelFor ExLaUing ' e f ~ i n e ~ sand ' ' "protatypes"in writes:

language, Lakcoff (1986: 43-4)

iness may also arim from non-graded concepts-comceplis defined by models

at ;have no scales ltruilt into them. Fillmore (19821 gives as an example the timered u s e d backelor. He ohwies that bachelor is defined relative to an idep is monogamous, and between people of opposite sexes. .

model of the world, one in which there is a sclcial institution of marmilap,


idealized rnadel fits the classical theory of categories. Wilthim the madel, r is a very cllearly defined &riskatelian category. But this idealived cognitive odel, os ICM, does not fit the world as we know it very well. When this mlotlel h placed within the context of the rest of our knowledge, fuzziness arises-not because 01 what is in the model but because of dliscrepancies between the background assumptions of the modd and the rmt d o u r knowledge. Here are some cases where the background conditions fail, amd as a result it is difficult to give clear, unequivIs Tgrzan 'a bachelor? ]Is the P o p a bachelor? .


15 1 0 General Issues
The answers to such questions are not clearcut, and the reason is that the idedheed model with resplect to which bachelor is defined may not fit well with the rest i the of our knowledge. The sourwe d fuzziness here is not within the model, but m hteractiom of the model with olther models characterizing other aspects oF ow knowledge. Fuzziness of the above sort leads to prototype leEects--~ases of better and worse examples 0F hchelolrs. Thus the perennial bacfrebr turns up again in a new role. Thirty years ago, the most fashionable semantic theory of the time-Katz and Fodorys C1963) "new semantic theoryyy-made its triumphant entry into linguisti~ perched prewriousEy on this same example; today, the theory of prototypes finds the bachelor example equally serviceable. But if the fornula '%achelor-an unmarried (adult) male personYVoesnn?work, couldn" we perhaps revise it slightly, to make it work-muldnY we, to wit, replace It with Lhe following definition: "bachelor-a man who has never married thought of as a man who cam m r y if he wants to"? (Mone precisely: "people think of Ithis man like this: this man can marry someone if he wants to'".) What cases such as this make char Is that discussions off 'necessary andl sufficienlt features' typically kcus on physical features and ignore mental ones. Yet natural language concepts often constitute amalgams of both kinds of component. Far 'bachebr" k i n g thought of as someone who can marry is as necessary as being male and having never married. cognize that semantic caltegariies are TuzzyY-a point which in his view s been established in Eleanor Reschys work. For example, he wrote koff 8973: 458-9): r Rosch Heider [I9731 took up the question of whether people perceive catembership as e clear-cut issue or a matter of degree. For exampje, do p o embers of a given species as being simply birds or non-birds, as do r them birds to a oertain degree? Heider" results mnsistenlily showed a tlrc dcgrcc 06 their bidimess, that asked subjects to rank birds as L egree L o which they matched the ideal of a bird. If category membership ly a yes-or-no matter, one would have expected the sutrjects either to balk or to produoe random results. Instead, a fairly well-defined hierarchy of

1 Birdiness hierarchy
robins eagies chickens, ducks, geese penguins, pehans bats bins are typical oF birds. Eag,les, being predators, are less typical. Chickens, ucks, and geese somewhat less still. Bats hardly at all. And cows not at all.

2.3. The Meaning of Congratsr/ale According to Yerschueren (1985: 4731, "a typical congratulation is an expression of the speaker's being pleased about the hearer's sulocess i n doing or obtaining something important. The first a s p c t [in. !the speaker's pleasure] of this prototypical meaning is compllete~y absent Bkom many formal acts of congratulating. The second asped [i.e. the hearer's suaessj is beimg tampered with in the fallowing huedline from the InternatiomaI Herd4 Tribwe: 'Begin congratulates Sadat on their Nobel prize].""" But in fact, it is not true that the expression of pleasure 'is completely absent from many formal acts of congratulating'. Apparently, the expression of pleasure (is. saying that one is pleased) is being confused here with the experience of pleasure {is, with be@ pleased). Of course in many acts of congratulating, the experience of pleasure is absent; but if one doesn't say (or otherwise cmvey) that one is pleased, there is no act of congmtuIating. Smely, an expression of pkasrnre is part of the invariant of the coocept 'congratulate', not just part of its prototype?
2.4. The Meaning af B i d In a number of publications, George kakoff has accused other linguists of dealing in various 'convenient fictions', and castigated h e m for failing to

It is hard to see, however, how this reasoning can he reconciled with speaken' i h m intuition that whereas a bat is definitely nor a bird at ostrich is a bird-a 'Tunny" bird, an atypical bird, but a bird. This ld seem to support a concUusion opposite to LAaff's: bats, which have s and no beaks and don't lay eggs, are disqualified, because feath,and eggs are thought of as necessary (rather than merely protocomponents of the concept "bird" [see Wierzbicka 1985: 1 1 8 0 ;for er discussion of 'bird' see Section 3.5. course, if informants are specifically instructed to rank a set of given cmes terns on a "sale of birdiness", and if the set hey are given indudes th bats and caws, one lcan understand why they might dedde to place ove cows, but does this really establish that bats are thought of as any degree of "birdiness", and that it is impssibie to draw a line een words for birds and words for things other than birds?

speak d the conmpt %bird" mean the cencepl enwaled in the English word bird. may OF course have no word for 'hkd', having Lexically w a d e d slightly difexample, the closeat counteupact of bird in the Australian La~guavge lude bats, as well as grasshoppers [Heath 1 9 7 8 ~ 41). : The closest equivtrdian language Warlpiri excludes baits, but it also excludes emus (Hale prototype may well be the same in all these languages, but the boundkly. An adeguafle semantic analysis should reflect this.

152 General Issues

4. ;Prototypes an$Im~"arimts 153 bers'(some add numbers being rated by thers, e.g. 3 behg rated as odder than 501; strong et d 1983). er (1987: 62) goes even further than tion of "prototypical reductian", and claims simply a false statement". She realizes, of course, that the use be fully predicted from this simple definition, but, she claims now from bitter experience how readily the compbxities of meanthe redoctianist formal malysis"(91987: 63Flthat is, how dificult ich would make the right predictions. atotype theory can save us from the trouble ing to do so. ]In the case of fie, it is enough as a "'Tase sr;rtenreniln"he lack of fit between the definition and then be explained in terns of our cultural models of wllevant however, when one realizes that a Jan-

2.5. The Meaning of Lie

Amording to Coleman and Kay (C19&1), whether or not an u is a matter or degaec, and tlicrc is oro set of necessary and sufficient panents characterizing the concept 'lie', This conclusion, which h 3 been accepted and endorsed in wunllless linguistic articles and b based partly on so-mlled saciai lies and white lies and galrtly on deoeption by evasion. For example, ilvrzinrcere utterances such as " lovely dress!" or 'Wow nice to see you!" or "Drop in any time!" are to be partial lies, rather than either lies or non-lies. Sirnil surances given to terminally ill patients are regarded as p than either hes or non-lies. Findly, answers which are literally true b which are intended to mislleed or deceive khe addressee Qe.g going?" """1Jlle're out of papsika""gre dso mtegoriued as partial lies. It is very interesting to see that many informants are pr c'sociaQ liesm,""chmitabb IiesW,andevasions as "partial lies'" Howver, semanticists are not obliged to take inform ants"^ faoe value. Coleman and Kay's methodology-hke Roslch"s--te durn results expected and desired by the researchers. Since the were given a seven-paint s a l e from 1 C'very sure not-lie"') to lie"), they acted as expected and arranged d l the instances offered somewhere along the scale. In any mse, Coleman and Kay" aim intend to challenge the very notion of the discrete semantic featwe" hmdly b e said to h u e been ahieved. The word lie can be given a pe vdid definition in terms of '"discrete semantic features" [see W i e r ~
1985: 341-2):

, words designating "false statements", and

d differently. For example, Russian Eras two words car-

r guide-lines concerning heir use should be

Btural model", how will they know how to e the uses of ~ r a ~ h n /gar" d On the other hand, carefully hitions e m guide ithe students i n their use, and in their interortant indeed, but they are not "another imporCultural models are reflected in the encoded in the meaning of vrat3s someloaded in the meaning of lgat'; and both of these eL are somewhat different from that encoded in the meaning of It be dificdt to articuiab these meanings adequately (that is, in a way ch wodd ensure f d predictive power), but it is not impossible ta do so. edictive definitions of speech act verbs see

X lied to F,=
X said something to P X knew it was not true2 X said it because X wanted Y to think it was true [people would say: iff someomre does this, it is bad]

@Ifcwrse there are similarities between lies a d insincere or evasi ances, as there are similarities between birds and bats, and i aware of that. But this does not demonstrate that the notion of h e disc semantic feature is not valid (see Tsobatzidls 1990; Je The fac~t that infomantskrreponses are often graded is int Amstrong, Gleitman, and Gleitman (1983: 284) put it, it fact about something other than the structure of conceptsn-parti t graded responses are also triggesod by view of the f a ~ that
2 The conmpt of "rue' is neither simple nor universal, but it is simple the present purposes, it doesn't need to ble decompsed further.

ange (1982: 1551 ddeees vrm'a (a noun carrespondling to rrorl) as "creative 19011g to m&e the liar appear interesting and important", and he calis it "a particular

154 General J.T,PE~~X

4. Prsrrofypes ond f ~ v a r i a n f ~ 155

2.6. The Meaning of Mother According lo George Eakoff (1986: 371, tlne concept of holber'cannot liYe given an invariant definition, because it is an "expediential cluster" and because no definition "will cover the full range of cases". The range of cases conling under this concept is, according to Lakoff* very wide, and cannot be reduced to any common core (such as, for instance, '% woman who hag given birth to a child"") because the word mother refers not only to "'biological mothers" but also to adoptive mothers, "donor mothersY'(wPno provide eggs but nolt wombs), "surrogate mothers" (who provide wombs but not eggs), and so on. LakofFs argument is so idiosyncratic that if one is not to be swspected of misrepresenting it, it is best to quote it verbatim: This phenomenon is beyond the scope of the dassical theory. The concept mother is not cbarly defined, once and for all, in terms of common nlecessary and sufficiemt ocmdiriions. There need be no necessary and sufficient conditions for motherhood shared by normal biological mothers, donor mothers (who domabe an egg), surrogate mothers (who bear the child, but may not have donated t mothers, unwed mothers who give their children up for doption, T h y are nil mothers by virtue oh their relation to the ideal lcas models converge. That ideal case is one of the many kinds of cas prototype efrects. (Lakorff 1986: 39; see also Lslkoff 1987: 831 From a semamtic point d view, however, Lakcoff's claims carry little eonviction. The cmcial p i n t which W o f f overlooks is that foster mothers adoptive mothers, ""genletic mothersy'% '""surrogate mothers", and so not "mothers" on a par with 'biological mothers' (see Bogusiawski Without a modifier, the ward mother ('X is Y s mother" refers cle birth-givers, not to the donors of eggs, providers of wombs, caretakers, fathers' spouses. Lakoff points out that the expression real miorher may refer to a taker as well as to a birth-giver ("She raised me and I called her mo but she is not m y real mmher'" "She gave birth to me, but she was a real mother to me"), but he overlooks the syntactic-and hen semantic-differenw between my real mother [either birth-giver or taker) and a real mother to ma (careltaker only). Furthermore, he overl the h c t that the test with real is nat semantically reliable. For example tences such 8s "he is a real m a n y h r"she is a real woman" may refer t speaker's views or prejudices about mcn and worncn which havc no bml in the semantics of the words man and woman. He doesn't appreciat implications of the fact that the expression biological molder would be only in a contrastive context, and that normally (without a contrastive co text] one would not say '%he is his biological mother"', whereas expressia such as foster mother, adoprive ~norker, or S U T ~ O ~ Bmother ~P are nr restricted to contrastive contexts.


T e treat "biological mothers" as being on a par with "surrogate mothers" or 'Toster rnothersvVs a little like saying that ~lzcrc are two kinds of hones: ibiological horses and rocking-horses [or that there are two diverging "models of horsehood": a biological model and an artefact model); and that we cannot define home as 'a kind of animal . . .' because a rackinghorse is not a kind af animal at all. I arm not saying that the meaning of the word mother can be wholly reduced to that of 'birth-giver" arguably, a social and psychological camponent is also present:

X is Y's mother. = (a) at one time, before mow, X was very small (b) at that time, P was inside X (c) at that time, IT was like a part of X (4 because of this, people can think something like this about X: "Xwants to do good things for Y X doesn't want bad things to happen to I"" he social and psychological component (4 has to be formulated in of expectations (thoughts], not in terms of actual events; by contrast, iologiwi components (a), (b),and (c) have to be formulated as actual Wierzbicka 1980: 46-91),
2.7. The Meaning of Furnitto-e
a paper entitlieid ''Cognitive Representations of Semantic Categoriesy', h ( 1 9 7 5 ~193) : wrote: we hear a category word in a natural language such as furniture or bird and stand its meaning, what sort ohcognitive repreentabion do we generate?A list tures ne:essary a d sufficient for rn item to belong to the category? k cone Image which represents the category? A list of category members? An ability se the category term with no attendant mental representation at all? Or some r, less easily specified, form of representation? This passage contains an implicit assumption that bird and furniture are e same sort of '"ategory words". Following Rosch, many psychologists d, m r e surprisingly, linguists, adopted this assumption as self-evidently rrect. There are, however, clear grammatical indications (as well as antic evidence] to show that L I m c Lwo words embody cornplelely dilrerof concept. Bird is a taxonomic concept, standing for a ppartiwof creature'. But furniture is not a taxonomic concept at all: it is we concept (see Wierzbicka 1984, 1985, 1988; Zubin and Kltipcke which stands Bbr a heterogeneous collection el" things of difierent One can't talk of '"three furnitures" as one can of "three birds'" and an" imagine or draw an unspecified piece. ol" furniture, as one can

156 General h u e s draw an unspecified bird. For birds, one crnn draw a line between birds and no[-birds (bats being clearly In the latter cahegary)~, Forfurallure, one cannot draw a line between kinds of things which are induded in this supercategory and things which are mot-because by virtue of its meaning, the word jtdrraiture doesn't aim at identifying any particular kind of thing. People may argue whether or not a radio is "fumiturey"(see Abelson 1981: 7251, but not whether or not a pelican is a bird (see Amstrong et al. 11983: 268). The concept 'furniture' may indeed be said to be "fuzzyyy---like those encoded in all the other collective nouns designating heterogeneous collleoltions of things (kitchenware, crockery, ckthing, and so on]. But it is hard to see how the study of such collective nouns (mistaken for words of the same kind as clluntables such as bir4 may constitute anytlnmg like "a refutation of the psychological reality of an Aristotelian view of categories" in general (posch 1975a: 225). Bolinger (1992) has argued that both furniture and bird require 'Teatme analysis" as well as '"rototypes", and 1agree with this. None the less, widence suggests that they are fundamentally different in some respects, because a bird is, semantkally, "a kind of cmature"+whiereas furniture is concegtualizled as "things af different kinds", not as "a kind of thingyy. The fact that bird is a "count noun" (e.g. three bk&) whereas furniture is a "mass noun" (e,gm *fhreefurnitures) is not accidental, but reflects and provides evidence for this difference in the conceptualization. (For further discussion, see Wierzbicka 1992EP)i." 2.8. The Meaning of Toy According to George Lakoff (1973) (who bases his claims on Rosch's investigations), baa and doll are among the ''central members" of thc category
Bolinger says that f i ~ l r u r eis, in some respts, Uke sguoslh, and that one can say, Bbr example, ''a crookneck Is a kind of squash". He also pints out that bird should be cornpad not with fwmiture but with 0 piece off~rntrwe. But note Ithe following contr;lds: a sparrow is a kind of bird 3 crookneck is a kind of sqn~asla * a chair is a kilndl of fucailvlre * a clnair is a kind of piece of furniture Contrasts of this kind suggest that the conceptual structures involved are dimerent. These dl& Ferences are quite systemtic: a rose is a kind oS flower an oak is a kind of tree * a shirt is a kind OF clothing * a fork is a kind ofcutBey a shirt is a God of p i w of clothing * a Fork is a End of piece of cutieuy What applies tojuraifure, then, applies to a 3 1 nouns which s t a d for heterogeneous collectiorr of fihings ((clfasl~fng, cutlery, Ritchenwore, and so on). What applies to bird applies to all n o m which stand For par[icular kinds or things, or cvealures I(rrce,jowcr,$sh, and so on).

4. ProroIypes and Jn wrfmiry; 157

"toy', just as rob& and sparrow are among the ""r;ntraP rnembers'bf the category bird. Swhg and skates are among the "'peripheral members" of the category 'toy', just as chickera and duck are among the "peripheral members'bf the category bird. Consequently, j~useas one cannot say whether chickens a d ducks (and bats) are birds or not-birds, one cannot say whether swings aad skates are toys or not-toys. All one can say is that they are toys to a certain degree (Pess than balls or dolls). But the analogy between bird and 60y is just as spurious as that between bird andfurniture. W i l e bird is a taxonomic concept which stands for a particular kind of thing, toy is no more a taxonomic concept thanfirnfture is. It is a purely ffmbionall concept, which stands for things of any kind made for children to play with, One cannot draw an unspeciiied toy, just ae one cannot draw an unspecified piece of furniture. The category 'toy3s "it~zzy'~-beeause, by virtue of its semantic structure (entirely different fmm the semantic structure of 'bird" it does not aim at idenltifying any parliculm kind of thing. Words such as sparrow, chicken, and oxtrick can be shown to contain in their meaning the component 'bird' (see Wierzbilicka 11985), and it is quite legitimate to start their definitions with the phrase a kind o f bCd But words such as ball or doll do not contain in their meaning the component 'toy".They may be seen as "central mernbers'bf the category Yay', but this is quite irrellevant from the point of view of their s m m t i c structure. It would be completely unjustified to open the definitions of the words bajl a d doll with the phrase a kind o f toy. There are many halls used in various sports (rugby, sower, cricket, etc.) which are not thought of as 'toys' at all; and there are dolls (e.g. china dolls kept on the mantelpiece) which are not thought of as toys. Whatever we discover about the sitructurre of purely functional concepts such as %oy'(or 'vehicle', or 'weapony,or 'tooly),it cannot be transferred to taxonomic supercategories such as 'birdy, Yower', or "tree" The semantic relation between sparrow and bird is entirely different from! that between baU and toy. (See Wierzbicka 1984, 1985.)

2.9. The Meaning of Game

cept of ' g m e 3 a s no doubt been the most influential example of ed "fuzziness" of human concepts which has been offered in the litIt was brought up by Ludwig Wittgenstein, in a famous passage of Philosophic b~esligatbns. Wittgenstein didn't appeal to the concept f prototype, but he appealed to-and indeed inti-occllumd-the related otion of "family resemblance" between concepts. The underlying assumpon was the same: conoepts cannot be given clear definitions in items of te semantic components; it is impossible to capture the semantic ant of a concept such as, for example, 'game'-because all that

different instanoes share is a vague "family resemblmce", not a specifiable set off features. Wittgenstein" idea of "hmi8y resernblanm'%as played a cobssal role in the development of ""gototype semantics", and the popularity of this school!of thought is no doubt due substantially to his intellectual charisma. In my view, Wittgenstein's writings mntain some of the deepest and the most insightful observations on semantic matters to Ise found anywhere. But despite my gratitude to Wittgenstein I thinkc the time has come to meexamine his doctrine of ''family resemblances", which has acquired the status of unchallengeable dogma in much of the current literature on meaning (see e.g. Jackendog 11983; Baker and Hacker 1980; Lakoff 11987). Wittgenstein (1953: 3 1-2) wrote: Consider for example the proceedings that we call 'games'. I mean board-;games, card-games, Olympic games, and so on. What is common to them all? Don%say: 'There must be something common, or they would not be called "games" '-but look und see whether there is anything common to all. For if you look at them yaw will not see something that is common, to at!, but similarities, relationships, and a whole series of them at that. To repeat: don" think, but look! Look for example at board-games, with their multifarious relationships. Now pass to card-games; hene you find many correspondences with the h s t group, but many common features drop out, and others appear. When we pass next to ball-games, m u ~ h that is common is retained, but much is lost. Are they all 'amusing'? Compare chess with noughts and crosses. Or is there always winning and losing, or competition between players? Think of patience. In ball games there is winnlnug and lasing; lbur when a child throws his lball at the wall and catches it again, this Feature has disappeared. Look at the parts played by skill and luck; and at the difference Between skill in chess and skill in tennis. Think now of games like ring-a-ring-a-roses; here is &le element of amusement, h t haw many other characteristic features have disap peared! And we can go through the many, many other groups of games in the saw way; can see how similarities crop up and disappear. And the result of this examination is: we see a complicated network of similarities overlapping and criss-crossing: sometimes overall1 similarities, sometimes shilarities of detail. I can think of no thetter expressionu to clraractcsiac these similarities than ' E m i U g resemblances'; for the various resemblawces between members of a family: build, features, colowrr of eyes, gait, temperament, etc. eb. overflap awd criss-cross in the same way. And I shall say: 'games' Form a family. Passages like these have a hypnotic force, and it is not surprising that they have exercised a great influenm on coontless philosophers, psycholagists, and linguists. But are Wittgenstein" claims really true? Is It indeed1 impossible to say what all games have in common, i.e. impossible to capture the invariant of the ooncept kame'? The only valid ifom of challenge in a case like this is to try to do the '7rnpossibIe", to try to define the concept of 'game" 1 would suggest that

the fdlowing components are essential to this concept: (I) human action (animals can play, but they don" play games); (2) duration (a game can't be momentary); (33 aim: pleasure; (4) 'csuspension of reaIityy"the participants imagine that they are in a world apart from the real world); (5) welldefined goals (the participants know what they are trying to achieve); 46) well-defined rules (the participants know what they can do and what they cannot do); (7) the Gourse of events is unpredictable (nobody can know what exactly is going to happen). Accordingly, It propose the fallowing definition:
@] mmy kinds of things that people do

(b) for some time (el "for pleasure" (is. because they want to feel somathing good) (dl when people do these things' one cam say these things about these people: (.$ they want some things to happen ( f j ~ if they were not doing these things, they woddn" want these things to happen QgE they don" know what will happen [A) they know what they can do (9 they know what &ey cannot do
Component (a] indicates that "games'he human activities and that there are many kinds of them, [B) that "games'be not instantaneous but have duration, (c) that "games" are undertaken for pleasure, (dl that "gamesyL have oertain constant characteristics: (e) games have some goals, these goals have no meaning or value outside the game, (g)i the course of a game is unpredietable, (h) and (Ep games require certain rules, and the participants know what these rules are. I believe that this definition5 applies satisfactorily to board-games, cardgames, ball-games, and countless other kinds off activity called "games'" It does not apply to a situation when a child idly throws his ball at the wall and catches it again, but in English this activity would not be called a game. En Geman the word Spiel has a wider range of use, corresponding soughlly to the English playing. But this very fact contradicts Wittgenstein's (1953: 33) claim that ""we do not know the boundaries because none have been drawn". Boundaries do exist, and they have been drawn differently in difThe definition of games propo~ed here is not meant to cover cases ~Fmetaphoricalextension, ironic or lnumorous use, and the like, as, for example, in the case OF the phrase "the games people play'" or in the case of' "games" played by mathematicians, generative grammarians, or other scholars who enjoy solving dificwlt problems for their own sake. Here as elsewhere in semantics, playrul extensions have to be distinguished from the ba& meaning [which explains bo& the '%normal" use of the word and any extensions from that use].

ferent languages, and native speakers subconsciously know them and respect them. One feature which separates the concept of 'game'lexically encoded in English from the concept of 'SpielYexically encoded in Gemam is the idea of rules: of knowing; beforehand what one can do and what one cannot do. Another dimerence has to do with the idea of a well-dehed god, which may or may not be attained. If features like these are not identified and clearly stated, cross-linguistic lexical research cannot succeed. It is not snnrprisinag, therefore, that advocates of the theory of "family resemblancesYhswal!ydo not emgage in such research.

Jackendo@ 41983: 11 131, among others, has tried to use colour terns as evidence that natural language concepts cannot be exhaustively defined into primitives. He w o k : "onoe the marker cobr Is removed from the reading of V d ' , what is left to decmpose further? How can one make sense of redness minus coloration?" I hope that the f o m d a e adduced above provide an answer to these questions (for fuller discussion, see also Chapter 10). 3.2. The Meaning of Words for Emotions Inr a sense, one cannot convey to a blind penon what the word red stands for (see Locke 11981: 38); or to someone who has never experienced envy what the word enivy stands for. None the less, it is possible to define envy in terns of a prototypical situation, along the following lines (see Wierzbicka 1972, 1980, 19866):

3. Uses of ""fototypes"


Semantics: Some IUzustratians

So far, the discussion has k e n focused primarny on what I see as the abuses and misuses of the notion of ""prototype"'. 1t is time to turn lo the more positive aspects of the idea of "prrototype". "Prototypes" can" save us, but thley mn help if they are treated with caution m d with care, and, above all, if they are combinled with verbal definitions, instead of being treated as an excuse for mot ever defining anything. Lexicographic practice suggests tbat the notion of ""gototype" can Be utilized in a number d diGerent ways. Below, 1 will il;lustrate this with a bird's-eye survey of a number of different examples.

X feels envy. = sometimes a person thinks something like his: something good happened to this other person it didn" happen to me 1want things like this to happen to me because of this, this person feelis something bad X feels something Eke this
Definitions of this kind demonstrate, I think, the spruriolwsness of the dilemma of whether emotions are better thought of as prototypes or as kelasically definable"(see Ortony, Clore, and Foss 1987: 344). It has often been argued h a t emotion concepts cannot be defined because nobody has managed to define them. But, as pointed out by Ortony ef ~l.,"theobservation that philosophers and psychologists have so far failed to specify adequate definitions of emodon(s) does not establish that the goal is impossible"". Whether or not definitions of the kind proposed abave for emy constitute a "classicall" account is a matter for discussion. They do establish, however, that emotions are definable; and that they are definable in tems of a prototypical situation, and a prototypical reaction to it. Without definitions of this kind, it would be impossible to account for the relationships between concepts such as knvy" ,ealousy', 'hatred', 'contempt', "pity', 'admiration', and so on. It would also be impossible to compare, and to interpret, emotion concepts cross-linguistically [see Wierzbicka 19864. If the study of emotion concepts encoded in different languages is ever to get off the ground, it is crucial to understand that there is no conflict between prototypes and definitions. (For further discussion, see Chapter 5, section 4.)

3.1. The Meaning of Coiolur Terms

As I have argued in Wierzlbicka (1980, 11985), the meaning of words such as green or blue can be dafined along the &bllowituglines: green - colour thought of as the colour of grass blue - colour thought of as the colrour of the sky Since this analysis was first proposed, a number of critics have questioned the use of the phrase thaughr of a3 in these definitions, and one critic (Goddard L989a3 has proposed the addition aT tlla concept 'likc'to my proposed list of universal m a n t i c primitives. 'Faking this into a m u n t , one t e m s roughly speaking along the could rephrase the explications of colo~ri following lines:

X is green Xis blue

- the colour of Xis like the lciolour of grass - the colour of X is like the colour of the sky

Whib simple formulae of this kind do not s ~ fully m satisfactory either (see Chapter L O ) , there is a wide range of evidenoe to suggest that, in principle, the use of "prototypes" such as grass or sky in the explications of oolour tems is well jusltified.

3.3. The Meaning of C u p l According to Hersch and C a r m w a (1976: 2741, " h b o v (31973) has shorn that attempts to give well-defined characterizations in terns of traditional componentid analysis of the semmlti@stmchre of a common concept such as 'cup' are inadequate." " S X n i G t l y spaking, however, h b o v has only shown that definitions of nrp ogered by conventional dictionaries, such as Webster's Third, are inadequate. This is hardly surprising, but does it maUy lestabkh that no "well-dehed chharrasferizationu; . . of a m m o n concept such as "up' " arc. possible? Questions of this kind are best answered by simply doing what allegedly cannot be done. For b p y land for a host of relabd concepts, I believe I have done it in Wiembicka C1985). The dedinitions provided in that work distinguish betwoen characteristic components which are not part of the invarimt and components which are absolutely necessary.

context is that 34 is not semantically equivalent to 33 either, and that it would be wrong to regard 34 as m explication of 33. Expression 34 treats the mother's or father" brother i n the same way as ar, mother's or father's sister's husband, and therefore it distorts the meaning of 33. If a mother's (or a mother" mother") sister's husband is categorized as 'uncle' at a11 it Is done by analogy with the focal, prototypical uncle. A definition which would exclude marginal uncles completely (such as "X's uncle = a brother d X's mother or father"")ould be empirically inadequate, but a disjunction which makes no difference between focal and marginal members is also inadequate. In my view, a satisfactory: definition should account for both, the invariant and the prototype. For wncle, the invaiant consists in a certain type of human relationship; and the quality of this relationship is conveyed by the reference to the prototype. I propose (roughly) the following: X i s Y's uncle. = if someone were a brother of my mother or father I codd say about this person: "this is my uncle" P can think about X like I could think about this p e r s ~ n This definililon leaves the denotational range of mcle vague, as I think it should, pointing at the same time ctearly to the prototype, as I also think it should. (See Wierzbicka 19%: 348-9.)

For example, a Chinese cup, smaEl, thin, dainlt)r,haindbless and saue;erbss,can stall be recognized as a cup--as long as it is clearly adequate for drinking hot tea from, in a formal setlling (at a table), being able to raise it to the mouth with one hand. This means that while a saucer and a handle am definitely included in the prototype of a cup @n 5dm8"cup m a w r havc a kdndlc, and a saumli) they are not imclurlled in what might be called the essential part of the concept, On the other Etaid, the components 'made to drink hot liquids hamhand 'small enough far people to be able to raise them easily to the mouth with one hand' have to be included im it. (Wierzbickil 1985: 59) In that sense, these definitions cannot be criticized ""for treating all components as contributing equally to the definition of a term" (Hersch and Caramazza 1976: 274). At the same time, they do contradict the assertion that "no subset of these components G z a n conclusively be said to be awessary and sufficient to define a term" "bid.); and they demonstrate that the opposite is true.
3.4. The Meaning of Uncle

3.5. The Meaning of Bird

As I have argued earlier, bats, pace Rowh and EakoK' aan no more birds than cows are, but ostriches and emus-which do not fly-are birds. Does this mean that Wying is not a m essential part of the concept %bird"? In my view, flying & an essential part of this c o m q t , and the h l l definition of bird, which P have proposed in Wierzbicka (1985: 1801, does mention flying (or the ability to move in the air), alongside components referring to feathers, beaks, eggs, and nests. But the definition of bCd (like ail the other definiticms of 'natural kinds') is phrased in such a way that it doesn"t ;imply that a 1 1 the essential features of the concept 'birdy are realized in all creatures categorized as birds. The definition opens with the folllowing frame:

According to Chomsky ((11972:$51, it is obvious that expressions such as the folllowing (Chomsky's numbers] ' b u s t have the same semantic representation''.
(33) b b ' s uncle
(34) the person who is the brother of John's mother or father or the Bpuslbaplld of

imagining creatures of this kind people would say these things about them


the sister of John's mother or father (35) the person who is fie son of one of John's grandparents crs U h e husband of a daughter of one of John's grandparents, but is not his father In my view, the meaning [and the resemanticrepresentation'" of expression 35; is vastly different from that of 34. What is more relevant in the present

Since the c o n q t 'imagine' is no longer included in the set of primitives, and since in the present version of the NSM grammar 'wodd' requires a complex sentence ("if . . wouPdY),I would now rephase this apening f r m e as follows:

peopb &ink things like this about creatures of this kind

B 64 General Jxsues
Accorclingly, properties such as flying, feathers, and so on are presented as essential parts of the prototype, not as necessary features of every bird. In addition, however, h e full explication of bird includes the following proviso: 'some creatures of this kind cannot move in the air, but when people want lto say something about creatures of this kind they say something like this; "they can move in tihe airv'." What applies to birds applies also, murcrri$ mulandis, to frsrit (and of course to countless other concepts]. Thus, Geeraerts (1993) questions so~lne components of my (Wierzbicka 1985) definition of fruit on the grounds that they don't apply to all fruit, even though the definition itself presents these features as part of the prototype, not a necessary kaltwre of aU denotata. This applies, in particular, to the component "wanting to imagine such things, peopb would imagine them as gowing on trees', Geraertsplsdts out, quite rightly, that raspberries are fruiri and yet they don't grow on trees. In my view, however, this indisputable fact doles not disprove the existenoe of a conceptual link bellween Trwil' and Yrees"ust as the fact that ostriches don't fly does not disprove the existence of ar conceptud link bclwccor %in?dshnd 'flying". Gccraerts (1992%: 266) obscrves thar "wc probably wouYd not claim that other people Lend to think of raspbcrries as growing on trees". But neither would we claim thar other people think of ostriches as flying. From the fact that people think of ostriches as birds, m d of birds as flying, it doesn't follow that they think of ostriches as flying. 11 has L o be stressed, however, that llhc two cases (birds aandfiwis) are not exactly parallel, bemuse bird is a taxonomic category ('a kind of creature'), wbereas Jmir-, !ike f~rniE[cdre, is a collective heterogeneous one Cdifferent kinds of things'). The heterogeneity of the conceptual categoryfr~ruifmakes typical (but not necessary) features offrwib such as 'growing on ereesbnavch less salient than typical (but not necessary) features of B i d such as 'flying'.
3.6. The Meaning of ,fernaft?,Cflibb~ge, and Apples

4. Protofype$ and Invarianls


Hut to define either natural kind3 or cultural khds, we do need the concept of prototype. For example, for cups we have to predict both the fact that a prototypical cup has a handle and the h c t that some cups (e.g. Chinese teacups and Turkish couee-cups) don't have handles. Similarly, in the case of tomabes we have to account both for the k t that prototypical tomatoes are red and for the fact t b t there are also yeBPow tamatws, which are also called tomatoes, or at least ye!low Eomolees. For cabbage, we have to predict both the fact that mbbage without modifier is greenish (except in elliptical sentences) and the fact that there is also the so-called red cabbage. For a p p h , we have to predict the fact that they can be red, green, or ye!lion; but also the fact that wanting to imagine (or paint) 'good apples', people are m r e likely to imagine thlem red than either yeillow or green. To account for facts of this kind, it is justified, I think, to have recourse to analytical etcvices similar L o that which has lbecrr used te account for flightless birds. For example, in tlzc dcfirrition of mbbagc P t l a ~ c included the hllowing components: the Peaflike parts are greenish or whitish-greenish in some things of this kind the Ieadlike parts are reddish wanting to imagine things of this kind people would imagine them as greenish In the gresenlt version of NSM, I would rephrase the last component as follows: when people want to say what things of this kind look like they say they are greenish

3,7, The Meaning of O;"lhb

Alongside bachelor, the verb cd&nb has played an important role in semantic theory as a key example of a word which-aUegedly-+annot be defined in terns of auty newessary m d sufficient components and which can only be analysed in terms of a prototype. For example, Vesschueren (1985: 46) wrote:
To show that a similar analysis is feasible for verbs, 1 adopt an example given by Fillmore (1978); the verb TO CLIMB typically describes an mcendhg motion in a cfamberi.,g fashion. ]I quote: 'A monkey climbing up a flagpole satisfies both of

It has often been claimed that the names of biological species and other "natural kinds" cannot be fully defined. [See Putnam 1975; Kripke 1872. For an excellent discussion, see Duprk 1981.) In Wierzbicka (1972, 19&101]1 B advocated Ithis theory myself. Since t h , however, 1have found-through extensive lexicographic research-that this is a fallacy, and that figem or lemons are no more indefinable than other concrete concepts (such as cup3 or mugs] or than abstract conoepts (such as freedom, love, or prerni8e)"see Chapter 55.
For a definihon of hve, see Wierxbicka, (1986bS; offreedom (fortlwoming 9; oF promb /1987a]. For defioltims of cay, mug, and many other similar concepts, see: Wienbicka (1915).

these conditions. The monkey climbing down the flagpale satisfies the clambering Icomponent only, but is nevesthelless engagled in an action thar c a r n u be properly called cilimbing. A snail1 climbing up the flagpole satisfies the asclending condition and cam stiPll be said to be climbing. But the snail is not privileged to c h b down the flagpole, since that activity would involve neither clambering nor ascending.'

However, this analysis fails to explain why a sentence such as "the monkey dimbed the flagpole" cannot be interpreted as meaning that the monkey climbed dawn the flagpole. If the direction upward was part of the prototype but not part of the invariant, how could we be so sure that the monkey who "climbed the flagpole" was climbing upwards? Difficulties of this kind have prompted Jackendoff (1985') to devote L o the verb cEimlG a whole study, and to use it as evidenoe for his o m version of groltotype semantics, developed in Jackendoff (1983). In essence, however, .Ia&endoWs analysis is not very different from EiPlmore's: he, t c m c r , posits for climb components such a3 ~ p w a r dand ' 'clambering fashion', and he, too, claims that either of these components can be "suppressed" though they cannot both be suppressed at the same time. For example, in the sentence "the train climbed the mouniain'Ythe 'clambering manner' eompcmnent is suppressed, and the component 'upwards9s present, whereas in the sentence "'Bill climbed down the ladder" it is the other way round. The semantic formulae proposed for these sentences are as follows (Jackendoff 1985: 288-9): The Lrairm dimbed the mountain. = TO TOP OF WOPINTAILN], [place ON [ ~ h i n MOUNTAIN ~ 1 GO {TRAIN,

d a Eitltle more precisely:

X climbed . . .=
sometimes in some places if people want to rnove upwards they have to move both their Pegs and their a m s X moved like people move at those times in such pjaoes temperature, the similarity in question can hardly be interpreted as rring to anything other than slowness. For trains, it can be interpreted referring to slowness and apparent difficulty. For people, too, it can be p t e d as referring to slowness and apparent dificulty; but it can also r to a quick and apparently enartless movement upwards in pilaces ere normally people wcruld have L o use their arms and Begs to move srds at ail [see "Watching him climb the cliff quickly and effortlessly P filled with pride and admiration"'). bus, a proltoltype is indeed relevant to the concept %limbJ. But this prois not '%suppressed" in less typical uses of the verb. It is part of the tic invariant' itself,


4. Conclusion
ere was a time when almost any problem in linguistic analysis could be appealing to the distinction between "competen~e"and "per(For discussion, see e.g. R.A. Harris 1993.) These days, this olution to linguistic problems is, usually viewed with! suspicion. sire to find simple solutions to a range of linguistic problem has d. "'Grice saves" m d a facile resort to prototypes are two characosner (1986: 5B) wrote: "As impressed as I am with the insights e ta get very excited ned from Wosch's work, It is rather hard fbr m t the great Aristotle versus Rosch debate." R~~osch's work indeed cons interesting insights, but it would be difficult to maintain that they e contributed a great deal to semantic description. Sn too many cases, ideas have been treated as an excuse for intellectual Iminess and , In my view, the nolion of prototype has to prove its usefulness rnantic description, not through semantic theorizing (see 11985). But if it is treated as a magical key to open all doors rt, the chances are that it will cause more h a m than goad. encoded in natural language are, In a sense, vague (see Black The chaIleng consists in portraying the vagueness inherent in natnglnage with precision. I agree entirely with Hemch and Caramma . 272) when they say that "'natural language concepts are inherently

Bill climbed down the ladder. = GO [BILE, DOWN THE LADDER]) Event 1[~annea @LAMBERING1

But this analysis is unsatisfactory, too, because it fails to predict, far example, that if a train w n t quickly up a hill it couldn't be described as cciimbi~g'. There is a difference in meaning between the two variants in the following paiw of sentences:

(I) The train climbed the mountain. (2) The train shot up the mountain. (3) The temperature climbed to 102 degrees. (4) The temperature shot to 102 degrees. Despite his rich anend of descriptive devices, including multiple brackeb and '"referential featwes"",ackendoffs analysis camat account far facts of this kind. In my view, what is really needed to account for such facts is a more careful, and more imaginative, phrasing of the necessary and suf& cient components of the concept Uimrir". I would propose the following (GF. Taylor 1989: 108): X climbed . . .= X moved like people move iin places where they have to use their a m and legs to rnove upwards

they go on to say that vague". ~ u I tcannot agree with them meaning of a tern could be spwified as a f u z z ~ set ofmew% cQImPon Natural langmge concepts are characterized by referential kkternin the sense that while "there are t11inrgs of which the desc~ption" clearly true and things of which the dcscriplion 'bee' is clearly fa1 there: are a host of borderline cases" ((Pwnam 8975: 1331. This mean, however, that the meaning of the word free @anonly be a fuzzy set of meaning components. 1 have tried to ckm~nstrate t by providing precise, main-fuzzy definitions of tree, and c~uinera concepts in Wiealbickra (1985). 1 hiwe allso attempted to showsth "fmxiest" caconcepts of ;all-"hed@~" W C as ~ ll$?$?r~~imPrrnQfe!y~ Q r Q least, or ra&+-can be given precise, non-fuzzy definitions, discrete components (see Wierzbickm of fully IF people argue whelther or not a radio is "fu~-nitur amount for this by saying that r @ &possesses ~ the meanin c f ~ t u r e " o a certain degree, less than table or desk- There Oinpisticl reasons for not including the feature 'furnitwe' i of either radio or tiable at all, as them are sufficient masons ing features such as %kitchenware" 'tableware', of 'cracker ing of ctp. It is not a matter of degree whether concepts o&, or rase contain i n their meaning components such as ' 'flower'; they simply do contain them. Nor is it a matter 0 concepts such as table, radio, re$igerllf0r9 Or CUP contain components such as 'furniture', 'kitchenware', 'tool', 'devise', or ' h ~ l e menty;hey simply don't. (For justification of this claim, andl for detailed semantic analyses, see Wierzbicka 1985.1 vagueness may reside In the semantic WmPQn C o q o n e n t s such as 'like the colow of gars' (in "reen? are ~n~~ "JWW this vagueness is mirrored inr the referential indeteAnai@~ of the cQF responding words. Components such as [r~u&ly slpeaki a man who can rnamykre perhaps not v a ~ but e are su tiye; they refer not to the "reality out thete", but to th conceptualizing reality. But neither vagueness nor ~omlponents should be iconfused with any " ~ e s e n c e to the Aristotelian notion of necessary and sufficient Ife &orubiein semantic analysis; it is the tacit bahaviowrist as~umptlon necessary andl sufficient features should c~rrlespondit tivelly ascertainable aspects of external rediltr. Many psychologists and philosophers Qve embraced the PrototYPe aP?r' on the assnu;nption that most concepts have resisted all dtemPQ define them and that "enon~~ous efforts have gone ink a featural substrate" (Amstrong et ial. 1983: 29 sion. 1n fact, rea;itlfive]ylittle effort by profi?ssional

$0 iKah and Eodor (31963) and llatz (1972). But, with all! due hese writers, they are, essentially, semantic theorists, not practi-

83: 268) point out, 'We only good answer [to the doubt the validity of the dcfinitioruaPview?'] is is difficult KO work out E n the reglllired detaii. ding the supposed simplest categories (the feaemanticists have tried a0 do that? It is true: that hers and psychologists, but also '"enerations g eb a/. 1983: 301) have hiled to pro&ce suem s of everqrday concepts. But lexicography has sis. Theoretical semanttics has flourished in am ework (see Wierzbicka 1985). Given this , it is the Bexicographers"achievements, markable (see Chapters 8 and 9). gun (see e.g. Mel'Eub and iolkovskij ; kpresjan 1991). The s ~ m e s s of this ity to absorb and to develop insights ical inquiry into fie role of protohowever, it will also depend on the fy, the basic stock of human conson and Smith l9$J: 55; also itive conceptual repertoire cananof 9810: 52) is being validated in the based on clear and rigorous thepeg cannot relieve us from the work. Prototypes cannot '"aye n the other hand, they can cerU s mnstruct the best, the most revealing, definitions, a b e d once~tualbatiomp d reality refiected and embodied in Banguage.

semantic Primitives and Semantic Filelds

I. Introductiiora
It has oken k e n said that the meaning of a word "dependsg~n the mean

we couldn't really investigate this possibility in a systematic and methodical way. If we do? h m v e r , have a list of hypothetical indefinables, and if learn how to discover configurations of indefinables encapsulated in vidua! words, we can reveal the hidden structure of these words and turd relations linking diffe~entwords , if we establish that the meaning of one word is abc, other, bcd, and of a third, bcf, we will know that their common Gore nsrbitrary semantic fields, and we can . Thus, semantic primitives offer us a tool for invessemantic groupings or fields. In particular, they can itrary semantic groupings from arbistinguish discrete, self-contained groupings from follows, E illustrate the preceding tends with a number of examning to different areas of the lexicon. Before doing so, however, notion of "mnfiguration",which was just illustrated (perly] with combinations of letters such as abc or cornpiex structures, built not directly from ty, or 'this', but ehom stru~ctured 'this is good' or 'you did someonents of this kind are ordered, and because h e y often e temporal element 'after', or the causal component 'because', of such companemts can often be regarded as "scripts" or ""smn 19811; Schank and Plbelson 1977). This applies, for to words designating emotions or to words designating speech point later.) "Concrete" nouns (is. names of natural 1 usually exhibit a more static semantic structure, but, usually invoived, and these comfeatures of the referents, but also es"-such as habitat, behaviour or typical interaction people in the case of animals, or the typical situatim of use in the case tefacts (see Fillmore 19756, 1977). Here too, the components have to ee Wierzbicka 1985). The general assumptions stated w illustrated with three sections devoted to specific atural kinds'bnd "cultural kinds", n concepts. Since the first "to domains red in considerable detail elsewhere (see Wierzbicka he present discussion of these domains will be brief, and the section on emotions wiH be disproportion-

I do not believe that this is the case: shoe the meaning of a

meanings of any "'neighbouring" w r d s in the lexicon. The meanin ferent words can overlap (as abc overlaps with bc4, but both the has been identified. Furthemore, what applies to semantic synchrony applies also to

if we are able to describe both the original meaning (arbc) and the s qwenlt meaning (4x4.To do this, we must be able to analyse each mea

this, we @an compare them again, this Ithe more precisely, be identify the elements that are &Efferent. Proceeding in this way, w discover remarkable symmetries and regularities in the semanti

least some of hese groupings are non-arbitrary, is intuitively even irresistible. But if we cowldn7 decompose meanings inlto co

5. Semantic Primi1ives, Semantic Fields 173

2. Natural Kinds and Cultural Kinds

Names of animals (in the everyday sense of the word, not in the scientik sense), of birds, fishes, Rowers, or trees embody, 1believe, taxonomic concepts, that is, concepts based on the idea of 'kind". It is reasonable, therefore, that they are usually rdernd to as "natural kind" words. For example, dog or lion can be defined as "a kind of animal"--plus, in each case, a long sequence of components, specifying the habitat, appearan* hehauiour. relation to people, and so on (for illusbations, s a Wierzbicka 1985; see &so Chapters 11 and 12). Words such as dog, lion, tiger, squirrel, and so on can be said to form a they all have definitiom well-defined, discrete semantic field beheaded, so to speak, by the s m e component, 'a kind of animal'. Similarly, words such as swnllos, eagle, penguin, or e m can be said to form a welldefined, discrete semantic field became they all have definilions headed by the same component, 'a Lind of bird'; and words such as ook,willow, b i d , or p?lm can alro be said to form a well-dehned, discrete semantic field because they all have definitions headed by the svne mmponent, 'a kind of tree'. Furthermore, Lhe conceptual supercategories on which the names of "natural kinds" are based have also a taxonomic character. For example, animal, bird, or @h can all be justifiably defined as "a kind of creature" (plus of course a number of additional components), whereas tree orpower can be justifiably defined as, mughly, "a kind of thing growing out of the groun8' (plus, again, a number of additional components). On the 0 t h hand, it i an illusion to think that words such as doll, b& nicycie, rattle, swing, and teddy bear can be similariy defined in terms of one non-arbitrary supercategory such as toy. As 1have uied to show in my Le*icogmplry and Conceptual Analysis (Wierzbicka 1985), words such as toy, vehicle, container, or weapon embody functional concepts, not taxo nomic concepts; and they are not related to c'cuItumlkind" words (such as rricycle, battje, ctq. or kni/e) in the same way in which taxonomic supercategories (animal, bird, fish, flower, or tree) are related lo their subordinates (such as dog, camnary, fruit, rose, or oak). For example, toy doesn't stand for any particular, describable, and recognizable kind of thing; rather, i l svands for things of uny kirtd made by people for children to play with. Therefore "toys", "weapons", and so on are not taxonomic supercalegories, in the sense that "animaUs", "birds", or " t m " are, Consequently, one cannot speak of "semantic Gelds" of "toys", "vehicles", or "weapons" in the same sense in which one cao speak of semantic fields of "aoimals", "birds", "treesy', and so on. If one wishes to, one can of course group words such as doll, ball, bi-

cycle, and mrtie together, and for certain purposes this may be useful Ifbr example, as a list of various kinds of objects that can be bought in a toy department of a department store). But a grouping of this kind would not have a semantic basis. This is not to imply that words such as doll, bag, tricycle, and so l e ~ nare not all "headed"(in their semantic structure) by the same semantic wmponent. They are. But the component in question is not "a kind of toy1'; uaitlker, it is a much more general one, subsuming a vast number of names of human artefacts: roughly "a kind of thing made by people'" There is no reason, of course, why one should not speak of all the words headed by this component as forming one discrete, non-arbitrary semantic field. But it is is not hierarchicallly structured: it is not divided, caking, into "toys", ccvehicles"/'weapons'~ and so on, ausle these are functional categories, not taxonomic ones. Of course if want to, we can classify cultural kinds into toys, weapons, instruments, henware, and so on, but this classification would be arbitrary from the nit of view of semantic strruetum. From the point of view of folk wteization reflected in the semantic structure, cups are not "a kind d cles are not "a kind of vehicley',balls are inot 'k kind of toy'" and knives are not "a kind of weapon". I suggest, therefore, that names of cultural kinds do not form nonarbitrary, discrete fields, whereas names of natural kinds do. In my view, it wdd be misleading to speak, for example, of "the field of containers'' (see khrer 1974) as if there were a non-arbitrary, self-eontarined field of"names of containers". QOf course, words such as cup, mug, botth, jar, j r r g ,bucket, and barrel are nnutmally closely related, and in fact their full explications reveal a degree of symmetry even grater than one might have expected (see the explications of these words in Wierzbicka 1385). But bucket is also felt to be rePatad to bowl or tub, bottle is related to carafe, carafe is related to to pan, and so on; and tubs, vases, and pans wodd not described as 'kontdners'" As argued in Chapter 4, birds do similarly into bab, ffishes into animals (in Ithe everyday sense , or flowcrs inlo trees (for example, magnolias are thought of ereas roses are thonght of as a kind of flower; emus thought of as a kind of bird, whereas bats are not; and so on). Thus, mantically, pcra and pans, buckers or bowl$ are not "a kind of container", ''a kind of birdM,androses, "a kind of BBower"'. of words and meanings cannot be established by e questions, or giving them simple sorting tasks. It stawished by methodical semantic analysis. E n the absence of is, different schemes of "semantic'bnd "conceptual" catecent literature, particularly in psychological Piterect the pre-theoretical ideas of the researchers rather than


S e m t i c Primitives. Semantic Fie!&


results of valid, well-conceived empirical investigations. (For further discussion, see Chapters 18 and 12.) To illustrate: parrot a kind of bird if people wanted to say many things about them, h e y could say these things: . . . SpdTrrQW a kind of bird if people wanted to say many things about them, they could say these things: . . . bird a kind of creature if people wanted to say many things about them, they codd say these things: . . . bucket a kind of thing if people wanted to say many things about them, they could say these thing: . . . barrel a kind of thing if people wanted to say many things about them, they could say these things: . . .

In addition, order indndes the component:

(I think:) you have to do it because of this

whereas ask contains the opposite assumption:
[I think:) you don" have to do it because of this

Forbfd is in some ways symmetricdl with respect to order, and it includes the component:

[I say:) I don't want you to do it

as well as a similarly confident assumption:
(3 think:) you cam? t o it because of this

Compiain includes the components:

(I say:) something bad is happening to me I fw1 sometlning bad because of this

ReproaLTSI, rebke, scold, and reprf~laa~d include the component: (I say:) you did something bad Thank and apologize include, respectively, the components:
@ say:) yyo did something good for me

@ say:) I did something bad to you so on. It is not my purpose to provide here exhaustive explications of speech act verbs. (The interested reader can find such explications in W~erzbicka1987a.) Rather, I am trying to show here how the "field'hf speech act verbs can be delimited in a non-arbitrary way. The class of verbs that I am talking about does not coincide with the class of '"IPemrmafive'v verbs. h r example, whereas ask, order, firbid, oor apobgsie can all be used performatively, reproach, ~hreaben, and boast cannot:

3. Speech Act Verbs

In English, and in other European languages, there are hundreds af wrbs that can be said to form, together, one coherent, self-contained field; these are verbs referring to "'different things that one can do with wordsn",hat is, to different types of speech acts. 1 have invesltigated some 250 such words in my Engllish Speech Act Verbs: A Semantic Dicfionary (Wierzbicka 1987a), and I have found a very high degree of patterning. What gives ccrherence to the field of ''speech act verbs" is the presence of some well-defined semantic components. These components underlie whet is usually called the "ilpQocutionaryforoe" (see Austin 1962a; Searle 1976) of the speech act described by a given speech act verb. This illmutionary form comprises components that spell out the speaker" inten. tlons, assumptions, or emotions, expressed in speech. For example, the verbs ask and order describe an attitude that lnciluldes the falllowing cornponant:
(1 say:) I want you to do it

I asWorderIforbid you to do it. I apologize for what I have done. *I reproach you shouldn't have done it. *I threaten you I will do something bad to you if you do it. *I boast I am the best.
one the less, all these verbs exhibit the same kind of semantic structure. They aittrribute to the speaker a certain attitude that can be portrayed in terns of first-person illoc-utionary components such as

10 say:) you did something bad [repriman4

(I say:) I will do something bad to you [threar] (I say:) I am good (other people are not like me) [bocrst]

176 Geneml Issues

I believe that components of this kind, all framed, explicitly or implicitly, o identify a dass of words i n a non-arbitrary way; by "I say:" do allow US L and that this class does constitute a "real", relatively self-contained part d the English lexicon. It is particularly interesting to note tlut the phrasing of components of this kind can be supported not only with semantic h t also with syntactic evidenlc3e, as different speech act verbs that share certain semantic compcanents (or oambmations of components) om be shown also to share certain syntactic frames or combinations of frames. (See Lehnr 1988; Pinker 1989.) Consider, for exmple, the component
you did something badr'good which is associated with tbe folbwhiarg frame:

order, commaaf isssrarcr, urge, ask, and beg, all of which can be said to imply the semantic component

I want you to do this

and all of which can occur in the syntactic frame Some of these wrbs, however, have an additional frame:
X V-ed Y for Z (6.g. X askedhegged Y for Z)

which the others don't have

* X ordeaedEcommandecllurged Y for Z

X V-ed Y for doing Z.

For example, the following verbs share this mmpouent and this frame: mproach, rebuke, reprove, repimmd, admonish, scold. Utterances such as

X reproacb&reb&edJreprimandedlscoIda&thanked

Y for Z

This additional syntactic frame links ask: (for) and beg [for) with verbs such as piead (for], appiy (f06, or wish (for]--alll verbs ahat imply tbat the speaker cannot expect to have control over the outcome, that is to say, verbs that attribute to the speaker the intention to convey [among others) the fallowing combination off components:

imply that X said to Y something that included the semantic cornponeat "you did something bad (good)" (cf, * X rebukedmproache&scolded Y , 9. The frame "X V-ed Y for Z"" is also uscd with verbs such as criticize or praise, which describe acts that can be performed, so to speak, behind the back of the target person: one cannot reprimand or rebuke people behind their backs, but one can criticize n o r prate them to a third person. However, the two groups of verbs can be distinguished in terms of another syntactic ' : frame: "X V-ed P% 2

I say: I want you to do Z I don't think you lyliJ1 do it because of this

he other hand, verbs such as order, commafid,or urge, which take the e 'A? X-led Y to do Z ' 3 u t not "X X-ed Yfor Z", imply, as mentioned r, a more confident attitude on the part of the speaker:
1 say: I want you to do Z

I think you w i E 1 do it because of this

The interested reader is likely to raise at this point some objections pointg to apparent aspmetries and idiosyncrasies. For example, why can't andI which should be similar to ordm and commandsoeciur in the frame d P to do Z"? ?r why can plead and appily occur with FOR (like ask but (unlike mk or beg) cannot occur in the frame "X F-ed F to do
W demanded Y to do Z X allow&forbade Y to do Z " X pleaded for Z X applied for Z *X pleaded Y to do Z *X pleaded f" for Z *Xapplied Y to do Z *Xapplied Y for Z

X criticlm&praisd P for Z X criticidpraised Y % Z *X rebuke&reprimamded F s Z

Thus, speech act verbs that imply the component person P did something baagood allow both of the syntactic frames in question:
(1) X V-ed Y h r doing Z (2) X V-ed P's Z

whereas speech act verbs that imply the component you did samething bad allow only the first of these two frames. As a second example, compare the syntactic possibilities of verbs such as

5. Semantic Prjmifives, Semantic FieE&


At first sight, diflerenms of this kind may seem idiosyncratic and semantically arbitrary. But if one studies them more closely, one discovers that far from k i n g arbitrary, they, too, point to very red semantic differences and ~ t h confirm s the reliakdity of syntactic clues in semantic analysis. For iexample, onepIea& WITH a person, as one argues or remons WITH a person, bec;ausep!ed, like argue or reman, involves an exchange of agurnents rather than a direct appeal to the addressee's will. One demaurh SOMBTHING, not SOMEONE, because what the person wiho demands something wants is, above all, a certain outcome (which may be brought about by somebody's actiom], not a specific action by a particular addressee. Far the same reason, one appiies for SOMETHING, and one doesn't apply SOMEONE, because what the applyixsg person wants is, above d l , a particular outcome, not a specific action of a particular addressee. At the same time, the attitude of a person who appiex for something is less cclnfident than that of a person who demsmrrd something; and this is why one appiies FOR something, as one hopes or mks FOR something, whereas one demands SOMETHING, not FOR SOMETHING (for evidence and justification, see Wierzbicka 1987a)i. Certainly, this method of verification cannot be applied to a I B areas of the lexicon, (Generally speaking, it is more applicable to verbs than to nouns.) T t can, however, be reliably applied to speech act verbs; and for this reason alone, speah act verbs consdkwte a parlirclularly fruitful domain for semantic experimentation. In particular, they offer a golden opportunity Ita investigate the structure of a large and highly patterned "semantic field"; and to explore, on an empirical basis, the very notion of a "semantic field1*.

sometimes a person thinks something Pike this: B want to do something I can do it after this, this person thinks something like this: I can't do it this pason f ~ l something s bad because of this X feels like this Refief &.g. X feds relieved) X feels something sometimes a penon thinks something like this: something bad will happen I don" want this after this, this person thinks something like this: I know now: this bad thing will not happen because of this, this person feels something good X feels something like this Qimppimtment X feels something sometimes a person thinks something like this: something good will happen I want this after this, this person thinks something like this: P h o w now: this good thing will not happen because of this, this person feels something bad X feels something like this Surprhe X feels something sometimes a person thinks something like this: something is happening now I didn't think before now: this will happen I want to know more about it because of this, this person feels something X feels something like this Amazement

Emotion concept8 encoded in the English lexicon constitute a coherent and reasonably self-contained (though not shauply delimited) cognitive domain, with a characteristic and specifiaibEe type of semantic structure. All the words belonging to this domain can be d e h e d in Items of cognitive s t m awes b a t are typically associated with the emotions in question. In this swtlon, I will try to substantiate this claim by andysing a reasonably Barge group of emotion concepts, drawn from two separate areas: roughly speaking, emotions linked with-"events contrary to expectations", and emotions linked with '"misfortunes". ((For a discussion of a third group of emotion concepts, including 'terrified', 'petrified" and 'horrified" see Chapter 7.) The first group includes fwbratiofz, relie& dkapp~hsrnend~ surprise, and amazement: Frustration I[e.g. X feels frustrated) X feels something

X feels someithjlng
mmetimes a person thinks samething like this: something is happening now I didn't know before now: this can happen I want to know more about it became o f this, this person feek something X feells something like this

180 General I s m

5. Seemanric Primitives:, Semntic Fields 18 1

As these exmples show, the d~efinition of an emotion concept takes the form of a prototypical scenario describing not so much an external situation as a highly abstract cognitive structure: roughly, to feel a certain ennotion means to Feel like a person does who has cerltain (specifiable) thou&& charancteristic of that particular situation (and to undergo some internal proass because of this]. TypicaIly, though not necessarily, these thoughts involve referenas to 'doing' or 'happening', to something " o d ' or 'bad", and to 'wanting' or b o t wanting'. (See e.g. Wierzbicka 1990c, 199&,

someltking bad happened to me I don't want this if I could, 1 would want to do something because of this b e a u s of this, this person feels something h d X feds something like this The maim difference between unhappy and sa$ consists in the personal character of the f o m r : if my neighboor's close friend dies I maq, be sad but not unhappy, bur if my own dose friend dies I may well feel unhappy. Furthermore, unhappy suggests rr less resigned frame of mind than sad. For example, if one says "'H am unhappy about it" o m may wli intend to try to "'do something about it", but one daesn" say "I am sad about it"". This differenoe is accounted for by the unaccepting component '1 don't want this', and by the absence of the resigned component 'I can't do anythingyin the explication of ut~happy. The combination of a past event ('something bad HAPPENED') w i t ha current rejection ('I don't want this1)may seem illogical, but in natural language "ilillogicalities" o o f this kind are very common. (One characteristic example was provided by the sign "119410 Annexation NO!", displayed by Lithuanian demonstrators in Villnius in January 1990.1 Distressed X feels something sometimes a person thinks something like this: something bad is happening to me now I don" want this because of this, I want to do something I don't know what I can do I want someone to do something because of this, this person feels something X feels something like this The common phrase distress signah, used with reference to ships, points in Ule same dirstion. The ship's crew may well wish to signal a message dong the following lines: 'something bad is happening to us" 'we don't want his', "because of this, we want to do something', '"we don't tknw what we can do', 'we want someone (else) to do something'. But there would be no point in any ship sending out "signals of sadness'" or, fear that mtter, "signals of unrhappiness"". T b word now in the explication of distressed proposed above may seem redundant, given the present tense of the verb in 'something bad is hapfor pening to me" none the less it may be justified, as it helps to aoxoua~t the short span of distress. Joy, too, has a present orientation Csomeltking p o d is happening'), and so does worry ('something is happening')^, but Uley

Thew examples show also how by analysing individual concepts we can $sicrfacto show Plow they are mutually related. Thus the definitions of relief and disappointmeat (differing only in the choice of "oodkr 'bad" are symmetrical and in a way support one another. Similarly, the definitions o f s ~ r prise and amazement are almost Identical, and differing only in one point Vwillr versus 4cangJ; and hese two, loo, support one anollner. On the other hand, fimmrsmriicrxj does not have a symmetrical counterpart; and yet its meaning, too, cam be established with precision and clarity. On lthe whole, the rclations between emotion concjcpts can be quite diverse (as the examplas given above and those which I;ollow illlustrake); at the same time, a large nubier of such concepts can be shown to follow the same overall semantic pattern. Turninn - now to the socond group, we will note that in English, as in many other languages, many emotion terms refer to 'bad things' happening to people. They include (among others) sad, unhappy, distressed, upset, and depressed, which E will now define one by one, using the format illustrated above. Snd (e.g. X feels ad) X feels something sometimes a person thinks something like this: something bad happened if I didn? know that it happened I would say: I don't want it to happen I don't say this now because I know: I. can't do anything because of this, this person feels something bad X fmls something like this In a prototypical scenario, the "bad event'Vs in the past G"something happened"; for example, somebody died). Unhappy (e.g. X feels unhappy) X foeis something sometimes a person thinks something like this:

182 General Lsues

ban both refer to the '"resent time"in a broad sense; by contrast, distress always refers to the '"'resent time" in a more narrow, more specific sense. For example, if I know that somebody that I Ilwe '3s coming" next month, this may fill me wilth joy for many days; and if I worry about my child's poor progress at school, I may be tlninking about months rather than days or hours- But distress swms to involve an immediate reaction to what is happening now (""lday" rather than "in the present period").

5. Semantic Primitives, Semantic Fields


5. Ciclnclusion
we want to establish what the meaning of a word is, and if we want to monstrate the validity of our analysis, comparisons with other words are lly necessary. But the rnemings of individual words do not have to be ndent on "whatever other lexical items may be available in the inveruand, ultimately, a definition, too, has to stand on its o m . A definiexpresses a hypothesis about the meaning of a particular word, and it d if it accounts correctly for the range of use of this particular m r d . undaries of this range may be "fuzzy'" but even this fuzxiness.can Id be predicted by a well-phrased and well-researched definition. s can be rilgorowslly described and compared if they are recognized they are: unique and culture-specific cornfigurations of universal nhc primitives. When the cofigmations of primitives conceptualized ividual words are revealed, the relations between different words also 1 themselves. 1 think, therefore, that the semaartic primitives approach mantic analysis also offers a necessary firm ground for the study of



X feels something
sometimes a person thinks something like this: something bad happened to me now if P couild, E would want to do something baause of this 1 don't know if I can do anything 1 can" think now k a u s e of this, thls person feels something bad X feds something Iike this A person is ~gse6by something that has happened to him 'before now', not by something that is happening to him 'now'. But the event in question is very recent; so much so h a t the experiencer ham%had the time to regain, his balance (as he is expected to do shortly]. The combination of the past knse with the word now in the explication Is meant to capture both the pastness of the event and its immediate character (see the same combination in the sentence "lt happened to me just now"]. At the moment, the experienmr is off 'balance and cannot think as uswl. His attitude is not passive, or resigned, as in sadness ('I can't do anything'); rather, he is mnfwed and temporarily cannot cope ['I dooin't know if I can do anything'). But unlike a person who is d&fresse$, someone who is upset is not crying for help or otherwise drawing attention to himself. The semantic structure of most emotion concepts, then, can be represented as follows:
X feels something soanetimes a person thinks something like this:

. . . .
because of this, this person feds something X feels something l i h this This structure can be said to define a large and coherent semantic field. The size of this field difkrs from language to language, but most, if not all, laawages do appear to have a reasonably self-contained "held" of this k i d .

6. Semantics and "Primitive Thought " " 5

Semantics and "Primitive Thought"

a1 question addressed in Hallpike's banned on a priori ideological ds: Are there essential qualitative d i k e a c e s in the cognition of difpseopbs? The view that there are such differences is not fashionable days, and Hallpike deserves some credit for the courage he shows in ing it, as do its other rmenl proponents [see e.g, Hain 1992; Bain and

"t'lllc fuauctlonls of Plie humon ~nind arc coonmorn lo lllrc w v l i o B c 06

humanity. QFrannz Boas 1938~: 135)

I. Intrloductiic~n
The question of universal semantic primitives is closely linked with t the "psychic unity of mankind"(CBoas 1938~). Just over two decades leading American psychologist, George Miller, wrote: Every culture has its myths. One of our most persistent is that nonliterate people i less developed countries possess something we like to call a "prkmitlve mentality" that is both differemkt From and inferior to our owm. . . . No one would ca ithat tiifferenoes exist. Any denial would be tantamount to saying that expriemce that result from Bvimg, in widely dfiermt cultures and tmh no important psychological consequences. Rather, the argument conmnms nature of those differences, and their soums. (1971a, p. vii]

In linguistics and anthropollogy such t e r n as '"riwnitive t h discredited ralther more than two decades ago (although ow still creep into print-witness the title of Hallpike's (19791 Eoumdatioms o f Primitive TAowght, described in a serious recent ( k Pan 1989: 3) as "monumental"). But the precise nature of ferenlcies beltween diflerent societies-in particular, Western s non-Western tribal societies-aemains an open question (see Discussions of this question have d r a y s relied to a consi on language. Rightly so, since language is " 'e best thoughtY'(ILeitbnix 1 7 6 5 S 1 1 9 1 8 U : 334) and evidence from 1 determining the fundamental thought patterns of d but evidence from language can be misinterpreted, and field-workers require serious semantic analysis before they ca a source of information about mnceptud systems. Discussing the alleged absemce of abstract thinking i n societies, Ndlpike writes: "It is . . . necessary to do some prel semantic ground-work before we can usediPlly discuss the extent to primitive thought is or is not abstract" "979: 171).

stabrlish whether a word has one or more meanings. In my diswill focus on several crucial1 concepts which have been alleged to ng in this or that language (notably, 'if ', 'Llwecause', %sameowe', '"all', and 'think'));I will1 start, however, with an exampk from a language usually identsed with '"primitive thought". eakers have n0 concept of that both the English ref and girl are translated into French as pile. Would such a pted? Presumably not. Rather, it would be pointed out that

such as ie garcon el awghter'), and S e jib the girnl). One could , 'daughter', have diffewnt grammatical tte can only mem 'little girl', ,not 'little iEerent syntactic frames (e.8. b $ E k de

ddl. (For further disclnssion of pdy-

18B Ge1;aerall8.~ues

6. S"enaanlit,-8 arfd "'PrBnifive Thought



2. The Universality olif BECAUSE

In English, the idea of causality is expressed in an ahsolute1 way in the simple everyday word becaruse. U n many other l ever, there is no word which means %mausehnd nothing ellse. ample, In Italian the basic word for 'because"as well as perch&-a word which literally means Tor what" In French, the wor 'w$y"s pourg~oi(again, nitewIly Tor what'), and the ' k a u s e ' i s pasce que [literally 'through this that"* Presmably nobody would claim, however, that speakers of French o Italian don" have, or cannot express, the conoepl; of 'be languages have other words which e parfly became unambiguously [e.g. the nouns Ecr muse, 'ause', in F Italian), and partly because nobody has any doubts t garm que or perch&do mew 'because" despite their far~llal analrsab inlo morphemes with other meanings. m e n it comes to 'exotic' languages (such as Austral guages], hawever, doubts conoerning the availability of have sometimes been expressed. As Goddard (11991a)ip commonplace of an oEder generation of ethnogra sometimes encountered, e.g. Sayers and Bain (898 guages a b u t the expression of causality, or even indifferent to it.'' Along similar lines (though more cautiously than Sa (19816) armed in his paper entitbd ""On the Unimportanlca: Kayardild" that the Australian language Kayardild has no st exponent of 'because" and that expression, of ~ausalityis co languaage with a purely temporal notion of sequenm. the concepts 'because' and 'after'are rendered in Kayardil the same "consequential suffhY%gasrba, how can we know that the s ers of this language distinguish lthe two concepts in question? IS it pa to establish that ngarrba is polysemous ( (1) after, r(21) because), rather having only one meaning, with different interpretations being due to ferent context? The most iranportant thing to do in a case like this is to formu hypotheses and to test them. One hypothesis is that the word always means "fter', with any causal overtones Ireing conk text. On this hypothesis, any senltencje with the word in, m k e sense on the 'after' interpretation. If we b d , however, that contexts the "fier'interpretation doesn't make sense whereas a ' interpretation does, then we have to postulate polysemy.

English the word md is often linked with a causal interion, but we don't have to posit a separate, causal meaning lPDs and to s ~ s of c all atad sentences. For inslance, in tfve sente~cje He fell down and cried.
;a causal

interpretation is implied but it is not absolutely neoessary, and the sentence makes sense even if we assume that alrd means here co-occurrence After her husband died, she fell ill. is (contextually) implied, but ithere no need to usal waning for the word afler (because a sequential tion still makes sense). other hand, in the Australian language Yankunytjatjara the i(-ng~~il~], which can Ire interpreted in different contexts as 'from', rYy Or "because" can be used in sentences in which a temporal (sequenintevetati~n wodd make no sense (Goddard 1994l~).For example: -Why are you crying? -I have a toothache. That-ABL I am crying.

sequen~al H("dterY] interpretation wound not allow us to make sense of sentenms, and, as Goddard argues, a separate 'because"e;aning has postulated. [See Goddard P9911a.) usively, Goddard (1991~)devised a test in the form of n: How, if at all, can one say in that language things "swe, ' I happened after X, but not because of p? For test shows conciusively that -ngu~u,which 1s fdly and clause, does indeed have a separate meaning fie meaning "after' can also be expressed by the m~nosemous
pie of the polysemy of an ablative suffix in an

You should go visit your mother [because] she is very sick. e Amrnte counilepart of this scsltence the morptrerne glossed here as can in other contexts m a n 'from' or 'after', but in this context the rpretation is causal. The speaker is clearly urging the son to is mother at the time of her illness, not after ilt or away from it. If we kers themselves sentences of ]thiskind do make sense de that the morpheme in question is polysemous between 'because' (as argued in Harkins and Wilkins 1994).

6. SernanSia md "Prilxsifive Thought" 189

It is important to add that different meanings of a polysemous word or morpheme are often associated with distinct syntactic frames, and that differences of this kind can provide crucial evidence for the polysemy of lexia 1 items. For example, in the Australian language Ngaanyaltjarra the s u f i -tjanu can mean dther %ecause' or kfter' (Amee Glass, personal c o m u nication; see also Glass and Hackett 1470). When one asks, however,

3. The Universality of IF
According: to Bain (1992: 871, 'The hypothetical conditional sentenoe is not found in Bitjlantjatjara. In Pitjantjatjara one cannot put forward a pure1y hypothetical condition, something that is merely possible, or a supposition. In practiuce, when Westerners attempt to do so, the Aboriginal person receives the idea as a fact." The claim is disturbing. Is it true that in some Australian languages "one cannot put forward a purely hypothetical condition, a supposition"? To plat forward a supposition one needs a word for 'if" Bain says that in Pitjantjatjajal-athe concept 'ifVoes have a lexical exponent, but that, none the less, when a Westerner wishes to advance a mere hypothesis "hidher listeners treat the statement as fact. Accordingly, what is intended as, for instance, 'if you were to get the money . . .' is received as either "hen you get the money . . . h r "dsince you are getting the money . . .""1992: 90). E t is easy to believe that sentences such as "if you (I) were to . . ." may lead to miscommunication in enmunters between white people and Aborigines in Australia, but it doesn't follow from this that a purely hypothetical supposition cannot be expressed in Pitjantjatjara or any other Aboriginal language; or that one cannot forestall misunderstandings in this area when addressing Aboriginal persons in English. All one needs to do is that one is not to state explicitly [whether in English or in Pitj~amtjaitjara) asserting the condition:
1 don? know whether X will happen X it happens, then Y

the sentence m m a only 'Why did he pun away?, not 'WWh did he imn away?' On the other hand, -0amrr m mean Wter' when it is used in an answer to a question about time: Wmpjawara Lukurrarnu? when run-PAST When did he run away?' Tur&utjana. corroboree-TJANU 'After the corroboree.' What applies to Yankunytjatj~ara, Arrernte, and Ngaanyatjarra applies also to Kayardild; and it is interesting to note that in a more recent paper on Kayardild, Evans (19941 also reached the view that the so-callled "consequential" suffix -ngarrba is polysemous between 'after' arrd %ecauseY. A word (or molrfieme'j which can bt: glossed as either %ecausekr 'after" cannot have some unitary meaning "more abstract than either 'becauseh~ 'after' ": there is no identifiable meaning mom abstract than 'kalnse' aslid 'after' and contained in them both. If someone claimed that there was some such meaning but that we had no word for it and couldn't articulate it, therr I would say with Wittgenstein that what one can't say one should be dlent about. Semantic hypotheses based on "ghost meanings'' which cannot be articulated are not falsifiable and therefore have no glace in semantic analysis. It would be wrong, therefore, to think that by allowing polysemy we are rendering our hypotheses immune to empirical disconfirmation. Polysemy hais to be established; it can never be posited without justification. For example, as pointed out earlier, the hypothesis that the English word rlllfier is poIysemous between a "sequential" sense and a "causal" s s e n is disctmfirmed by the fact that, in any context, after can be shown to 'be compatibb with a sequential interpretation.

a language doesn't have a word [or morpheme) for 'if', as it n alleged to be the case in some Australian languages? What if does not distinguish between 'ifhand "hen', or ' i f ' a n d
t hcrc as elscwhcre allegations oF this kind oftcnr stem from dare to recognize lexical polysemy. As h4cConvel1 notes:"Yack of a fordistinction betwesa $and when in Aboriginal languages, in contrast to nglish, is supposedly linked to absence of hypothetical conditional n Aboriginal discourse" (199 1: 15). Rejecting such claims vePl argues that Aboriginal languages do haw lexical and gramrces to mark conditionanity, and he points out that even if the and 'whenkre identical, they may appear in diflerent frames.

[IF.. . WOULD) sentences, and it is no@ always clicar which typc I P ~has in mind when he
talks of "hypothetical conditional statements"'.

McConvell doesn't draw a distinction between conditional (IF) and counter-factual

6. Semantics and *'Primitive Thoughf " 191 For example, "In the Ngarinmn language the concept of q i s distinguished from daen by the use of the doubt sufix ngu following the subordinate clause marker nyamss and the pronoun clitic complex" (16). McConvell points out that devices of this kind are frequently utilized in certain genres of spontaneous discourse, and that older people without Western education frequently describe "imaginary and hypothetical smnarios, including mdtiple chaining3 and embeddings of hypothetical statements within otber hypothetical statements" (15). The polysemy of the primary exponent of the concept 7iF can be illustrated with data from the Australian language Arrernrte (Markins and Wilkins 1994: 298). E n a simple clause, the word peke means 'perhaps', but if a dependent clause is present then it means 'if'. Inpenthe peke kwatye umte-me. tomorrow may& water fall-NPP 'Et could rain tomorrow.' ['Perhaps it will rain tomomow.'] Kwatye peke urnte-me ayenge petye-tyekenhe. water maybe f a l l - ~ e ~ 1 ~ ~ come-VERB : s MEG "If it rains 1 won't come.' Data of this kind provide strong evidence for the presence of a tinpistieally encoded concept of IF (even if this encoding involves a polEysemous illustration of this point, consider some data from Geman, it has also been sometimes asserted (incorrectly) that it doesn't he conoepts of IF and WHEN. d all, German does have some quite unequivocal exponents of W E N , warm (used in questions and in relative clauses) and ab (used in past tense temporal clauses), which can never be used in the sense or IF; so clearly, C a m a n dms distinguish WHEN from IF. Ear example: Wann warst du dart? W e n were you th~ere? Plls du dort wmt, war ich him. M e n you were there, 1was here. Second, German does have a word for IF, namely wenn, and although in subordinate sentences referring to future events wenn can stand for either IF or WHEN, this doesn" mean that it is somehow vague and always covers both senses at the same time. It is more justified to conclude that in future tense sentences wenn is lysemous, and means either IF or WHEN. For example (from Die B&cl hew figm iCeeul$ch),

{If the dependent clause is affirmative rather than negative as above then the verb carries a special "subsequent" marker-tyershenge.) E n another Australian language, Yankunytjatjara (Goddard BW4b], the same word tjfmgng, shows an even more complex pattern of polysemy: when it is used in a simple clause, or by itself (as an exclamation}, it means 'maybe" bult In a subordinate sentence it has either a conditional or a counter-factual meaning, depending on the altpsenoe or presence of an ''iissealis" inflexion on the verb of the main clause (see Chapter 2):
(I} Tjitugum. Maybe! (2) Ka nywntirmtjinguxu kjuklarpa irititjatjara and you if story long.ago:~ssw:HAVING nyakwla kulinitjikiltja rnukuringkula,nyiri S ~ : S E R I A L think:a~T~ntT want:saRrar, paper pala palunya nyawa. that Z]BF:ACC s e e : n ~ ~ 'So if you want to read Old Testment stories, look at that book." (3) Tjingu~u ngayulu waringka, puQkapalyanma. if 1 co1d:~oc big make:^^ 'If I was jin cold (weather], I'd make a bigger (amount).' ('The speaker was explaining that shle had not made a very large amount of spinifex gum because the weather was too hot to do thk easily" (Goddard 19946: 248.)

. . . aler Mens~heasohn w i d kommen, wenn ihr es nicht emartet.


(Matt. 24:

'The Son of Man will mnae when [not "whewllP'1 you do not expect itY. Wenn ihr nur Vcrtteuen hiabt, werdet ihs alPe6 kkommen, wotum ihr Gott bib
'if [not "iflwhen'" you have faith, you will receive all that you ask C d for."

The fact that in certain grammaticai frames (e.g. in the frame wenn n k h f

"if not" or in combinations with the past tense) weptn can only mean ""if"',
nd nat either "if" or "when", supports the view that it is polysemous, not ape. This conclusion is also supported By the fact that if one wants to contrast the two concepts I F md WHEN (e.g. "when, you come-if you came . . ."I, this is possible, too: Wenn du komst-WENN du kommst-wirst du es sehen. WWn you come-if you come-you will see itY. Whalt all these facts show is that Geman does distinguish, iexically, ktween the concepts IF and WHEN, even though in one type of sentence (compl~allex sentences referring to the future) the exponents of these concepts overllap.

192 General P m e s

4. The Universality of SOMEONE

In her recent book similar in its general orientation to Hallpike's, Bain (19912) develops the thesis that Australian Aborigines use only "first degree: abstraction and concrete logid', whereas Westerners use "second degree abstraction and formal logic"". In support of this thesis, Bain argues (9431 that, for examp1e1 in Piitjantjatjara "there are no terns for the indefinite pronouns such as *someone', 'anyone', 'whoever' and that to refer )to an unspecified person the speaker would have to use the word ku$q~a,'other', According to Bain, "linguistic features of this kind are antithetical to the formulation d purely pneral sktements1'@bid.). Couldn't one argue, however, ttnnt in Pitjantjatj~ara the word kugupa is i n fact polysemus and has two distinct meanings, 'other' and 'someone3? Bain rejeds this possibility. Cementing on the sentence
I '

Goddard, personal comunimtion]: if a specific antecedent is present (e.g. this man went k s t ; \then another [ku$wpa)r went), then it cleuly means 'other" but in the absenae of a specific antecedent it can only mean 'someone'.

5. The Universality of ALL

Hallpike [19?9> argues that what kne calls "primitive socjletiesl%ave no concept of Wl', adducing linguistic evidence from several languages. The i w l i cations of this d & are so serious that they deserve to be examined ira some detail. Me writes: Y30;me~d 'all' are thus fundamental notions of logic and bask to propositions of inclusion which relate parts to wholes. "All' denotes the hotaJity of a set A, while ) . Im primitive usage, however, it is 'some' denotes 'A - xi{where x is greater than 0 possible that while words are used that ethnographers translate as 'some'and 'atuPY", "a111' does nab denote "all possible members of set A" but $14 those Irm awr axprienoeks simply % I lot'. In so far as primitive thought is not usluarlly concerned with working out the theoretidy ma?&uum number d items im a set, it will tend to use 'akl' in the sense of 'very many" while if d l possible members of a set ahe physicdy presemt, the primitive may indeed say Will", but in the seause of 'full' or 'compIete', which is derived from a spatial canception, as of ar container that has been Elled up. Dr Neil Warren, for example, tells me (private colmmunication) that the Kamano of the New Guinea Highlands use thein w r d for 'mamy' to do duty for what w e would translate as "a'. In the same way, amang the Taude I laund that the word that I was inridally indined to translate as 'all', k~parfmrs, was more a ~ u r a t d y renm y ' " . Kuparitri is the word For 'two' or 'pair', -oi king the dual suffix, dered as " md -ma is one of tihe plural suffixes;thus kuparhrs seems to have the literal meming of "airsp, i.e. 'many', and is certainly sol used in conversation. It should also be not4 that kupcrrima is not a;n adjective, but a noun, and refers to a state of affairs, 'multipPicity', rather than being a property ofa dass. (HaIlpike 1979: 181-2)

Kugupa ngurakutm anu. another camp to went 'Another @erson) went to camp.'P/"Someonewent to camp." she writes: With the last of these translations there is a mom from the adjective 'another' 10 the more abstract pronoun 'solmeonue'. That shift is ameptakrle in English but its appropriateness must be questioned Tor Pitjanltjatjara. While to some extent the translation used may be ;a matter ol pmonal preference, if wa are la stay as dose as possible to the Aboriginal thought, then the link with the real should be retained. (1992:94) But semantic analysis is not a matter of personal preference, and the hypothesis that kugwpa is poUysernous can be tested. The simplest thing to do would be to conj~we up a situation where 'other'would not make sense but %omeomebwodd, and to check if the word k3stjapa could still be used. The fact that one can use Jcus;Ewpa in such situations (Cliff Goddard, personal comrnrnntcaaion) shows that this word cannot always mean 'othery. On the o h e r hand, one cam make sense of a11 uses of kwtjupa in t e r n of two hypothetical meanings: btheryand 'someone'. For example, if one m n say, using kwfjwpa, '7 saw someone Qkucjiupa] there, it was the s a m person"l then it is clear that kntjupa cannoit mean 6another'in this context ( ? ' I saw another pewon there, it was the same person'). In addition to semantic tests of this kind, one can allso examine the enviro m e n t s in which the two senses occur and see if there are any dierenms between them. For example, Harkins and Wllkins (19134) show that in a related Australian lampage, Arremtey the word arxpenhe can also be used in the same two senses cother'and 'someone'), depending on whether or not a speciifc antecedent is present. The same applies to kuGupa (Cliff

I believe Hallpike% cconclusions are fundamentally wrong. It is true that In many languages ithe word glossed by ethnographers as 'all' is in fact a nominal rather than a determiner (that is, grammatically more Pike the expression the b s in EngPish than the determiner every), and that sentences including a word whose basic meaning is kmanykan sometimes better be Wanslatad with the English word a0 than m m y . But does this mean that these languages make no clear distinction between the concepts 'ail' and 'manyy? I think not. To begin with, HalPpikds remarks on the Tauade data are far from convincing. J R f the stem kupari combined with the dual suffk is the m r d for Ywo'or "air" then by itself it is much more likely to mean 'all' than hamy' (cf. French tous lex deux, Pit. "an the two", that is, "0th'). A stem meaning 'manykould hardly be combined with a dual suffix, sinoe the language clearly has a contrast

194 General .&sues between a plural q"manyq)and a dud ('two"; the combination 'manyy+ 'two'would be incoherent, whereas the combination of %a1and "WOYQ~ a pair has parallelis in many other languages. Furthemore, though E haven't been able to check Ha1lpike's assertions about Tauade (or about Karnano), his remarb on h e apparent canflation of the concepts 'all" and hany'also apply, for example, to Australian lam guages, with respect to which they have been studied in reliable linguistic literature. (The index of Hallpike's book shows clearly that next to the peoples of New Guinea it is Australian Aborigines who epitomize For him the notion of a ""primitiw society".)~~ For example, Bittner and Hale's (1995) amalysis of the Australian language Warlpiri shows h a t in this language, too, there is a word, also a nominal @mu), which is sometimes best translated into English as 'many' and some1 1 " and Harkins (1991) shows that the same applies b another times as " Australian language, Luriqa, But this does not mean that Warlpiri a d Luritja do not have a concept of 'all'distinct from the concept of 'manyy. First, while the Warlpiri word panu can be translated into English as either 'alt'or 'many"i/depending on context), there is another word, jiatdumarsarai, which can never be translated as 'many', but only as %H' (or "1 of them" see blow]. Se~osld, as Eittner and Hale" smlysis shows, pamu can be translated as "In" only in those contexts which imply definiteness, that is, where it can be interpreted as 'the many', "the lot', 'the group (composed of many]" by implication "he whole groupy. In a case like this, the word which means, essentially, b a n f may appear to mean 'all'. From this, it is only one step to the conclusion that the Warlpiri people do not distinguish 'manyYrorn 'all". But such a conduiort would be fallacious. Equally well one could argue that English speakers do nat distinguish the two notions in question because a lFor means 'many', whereas she jot means (roughly) 'all', The fact that Warlpiri has a separate word for 'alll', jiatmkumarsarni, just as English does CaIEp, shows illhat in fact the two concepts in question are distinguished, despite differen= in use linked with other differenoes between the two languages, such as the presence versus the absence of articles. These points can be illustrated with the following data (Bittner and Hale 1995, their mnumbers, and their
Austualian Aborigines (along with Papiuans) have often k n used in the literallnnre as an example olc$rhnitive mentality"-by Uy-Bmhl, by Hallpike, and by many others. The title of W&ek (1872) paper, "The Mental Characteristics of Primitive Nan as 1ExempliEed by the Australian Aborigines", is very characteristic in this respect. For discussionu, see Chase and yon Stumer (1973). Bittner and Hale use &e following abbreviations irn their interlinear glosses: PIBSPlbsalutive, Lac--locative, H,2,3,-kt, Znd, 3rd person, p-plural, s--singular, INF-imfini. tius, PRClX-pronimate, PRF-perCectiw, PRS-pueserat, P S T p a s t , NPST--~~onn-past. Ward internal morpheme boundaries are indicated by '-', clitic boundaries by k'.

6. Semantics and 'Frimilrs've Thought " 195

1. Punu means 'many': (117) (Q.) Nyajangu-O-0-ngku


gu-ngu nyuntu-ku?

NhlkJ!%3?1EU-~~~&-2s boomerang give-rs.r you-Dst 'HOW MANY boomeramgs did he give you?'

(A,) Paau a-Qc-ju yu-ngu karli. PANU PRF-3%1s pjve-~s~ boomlerang We gave me MANY boomerangs.'

2. Jinlakumarrarni means "11': (10) Yuml-jarri ka-lu jhtakumaraarni=1ki ripe-lbecome-t?.~~s~ PRS-3p alll=then Then they get rip, all (parts) of them.' (1 11) Slntakumrm~-jiki-jaIa ka-llu wapa kankarlu-mipa dl-A=oFcourse PRS-3p ~ O ~ Y ~ - N Pabove-only ST paarrpardi-mj~a-rlapinkirrpa-kurlu-lrZI fly-INF-anox father-ones.with-A 'AH of them of course live ody up in the air flying [the feathered ones]." 3. Panu can be used in the sense of 'the many', 'the lot', "he groupy,and, by implication, bsl1"Ci.e. 'the whole group'): (193 Panu ka-ma-jana nya-nyi. many PRS- 1s-3p see-NPST ti] T see a laage group (af them).' (ii) '1 see the large goup [of them)^.' (iii) 'I see hem, who are a large group.' (3.51 Yapa ka-1n nyina panu nyampu-rla ngurrj~u? person-ABS PRS-3pbe-WST many this-am well 'Are all the many people here wePI?"
This is not to say that the Warlpiri wordj;inr[j~kurnarrarmi has exacltly the m e range of use as the English word "1' (a point to which I[ will return later), but it does mean that Wadpiri distinguishes between the concepts 'arll\ad hmanys,and has separate lexical exponents for each of them. Furthemme, although Bittner and Wale (1995) glass jinrakrrmarrarna' as 'all of hem' rather &an 'alll', P see no evidence that this ward mems anything other than simply 'all'. In actual speech, it will usually reffer, no doubt, to some previously mentioned group, and thus will be consistent with an interpretation along the lines of 'all of them" But this is not necessary: when needed, the same word can also be used to make open-ended gen~eralizations,of the kind that Hallpike claims are impossible in "primitive languages". Bittner and Hale's sentence 11 ('"all of !them of course Iive only up in the air flyhg [the feathered ones]'"), which comes fram an "oral essrryl'about living kinds recorded by Bale, and which doesn't refer to any particular group of birds but to birds in genera!, provides a good illustration of this.

6. Semantics and "PrhitSw T h a w g h t J 7 9 7

Two further examples oh such open-ended generalizations from andher Australian language, Kayardild, are quoted in Evms (1985) (DTstands for deitpansitivimd):
(16-279)Maarra diya-a-n-kuru. all at-DT-FUT (Speaking, of yams:) 'Q%ey) are a11 edible." (6-28 1) Maarra maku-karran-d. a l l woman-cEm-Mow [On lioe as food: ) 'Qdy women eat M i c e . ' (Lit. "all lice are w\nmenYs")

(12) ma -meri ma-Merranunggu gusringgi -wanggaQ human ns man human ns Merranunggn 3nsSR "RR" finish -@ -a wakay pl Pst finished T h e Meraaounggu people all "finished up" (="didw).' [The Merraaunggu people died campletely.)

No doubt everyday life in a tribal community doesn't generate much need for generalizations of this kind, but if they are not made frequently it is not for lack of conceptual or linguistic resources (cf. Section 7). I conclude that as far as we know Lhcre is no human language which doesn't haw some lexical means for expressing lthe conceglt of 'all'-n~t something roughly comparable to 'all' but exactly Lhe same conoept. I think Hallpike is right in assuming that without a word (or s o m other lexical component) for 'all'a language wouldn't be able to express oertain thoughts---more than that, that in a language without callknewouldn't be able to think certain thoughts-and that thoughts crucially dependent 011 the concept %ll' have fundmental importance in European culture. But I Mieve Hallpike's conviction that such languages exist is not supported by the evidence. This is certainly not to say that the range of use of words or morphemes which embody the concept 'all' is the s m e in all languages. In some languages, the range of use of the word or morpheme meaning ~ I l restricted ~ s to a relatively narrow range of semantic andfor syntactic environments. This is true, in particular, of the Auslirralian language Marrithiyel (Green 1992) and of the Papuan language Yimas (Foiey 1991). But the range of use is one thing and the existence of a lexicallized concept is another. To siee this, consider briefly the Marsithiye1 facts. According to Pan Green (1992) the only word in Marrithiyel which could possibly be regarded as a m exponent of the concept 'all' is the adverbiinterji~~tion wakay. Leaving aside the use of this word as an interjection, it appears that wakay (as an adverb) can only combine with "semantic unadergoers" (that is, it can only apply to someone to whom something happened, not to someone who did something); and that, moreover, it can convey the idea of "completeness"' as web as "'totality"".For example (Green's n ~ m b e r s ) : ~
-ya (wakay). (1 1) Eiyi winjsjeni gani 3sSR "go" "ST finished head bad 'He went [completelly) silly.'
Green uses the oliow wing abbreviations in interlinear glosses: Pst-pat, auxiliary, +singular; SR-Subject RcaLs.

What wouEd it mean for a group of people to die '~ompleteiy"? Smely, in a sentence like 12 w&ay means simply 'alP"ust as Green's gloss says). y twe used with noon-gradable predicates such as "die' The fact that w a k ~ can suggests that this word doesn't encode same hybrid notion of 'cornpBetelyjrtotdlyjraln" but has two distinct meanings: (1) cornpktely, 42) all. O t k r examples cited by Green [personal comnmunication) support this conn. For example, if wcrkay is added to a sentence which means We gave hters in marriage', the sentence can only mean 'He gave away hmers', not 'Me gave away his daughters COMPLETELY'. of this kind, vve have to concludle, I think, that Marsithiyen s have a lexical exponent for the primitive concept 'all" even though this onent's range d use is more restricted than that of the English word all. r an interesting discussion of the concept of %liY iin Australian languages m a dmerent perspective, see also N. Evans [forthcoming); h r a discussion of the concept 'some'see Chapter 2.)

6. The PJniiversalilty of K M W and THINK

ng to Mallpike (8979) there are languages which cannot express the of 'knowhnd 'think" People who speak such languages are, ke children at the "pre-operatony'\tage of development
gnitively incapable of distinguishing clearly between recognizing the operation of his own mental prl~cesses ord like 'think", he does not grasp its cagniliue bplicakconcentrating" '"making a mental eflort', e.g. when try. . . At the first stage . . . (Cat about six) he supposes when we speak, and by assonciatian also identiiies nd smoke, or else equates thinking with hearing, and g we do with our ears. (Hallpike 1979: 38563

According to Hallpike, ""pPimitire"pnseptes, too, confuse thinking with eakhg and hearing, and they, too, have no woncegt of purely cognitive ses and states such as those linked in English with the words think now. He writes (1979: 3934):
ility to ana1yse private experience, as opposed to social ffiehvionr, the OF the knowable, is well illustrated by ethnographic evidence from the


198 General hme3 iOmmura, of the Eastern Highlands Province of Papua Hew Guinea, Like many primitive peoples in New Culmea and elsewhere, the O m u r a use the same verb [sera) for 'understanbgkor %comprehendingy, amd the 'bearhgkfa sound eltc. D ~ p i corresponds fairly closely to 'clear', 'distinct', as opposed to bobscure' ar 'anfud', and thus the expressiom &pi ienrcs means 'that sound which we can hear clmearlybd also, when wed in the sense of understanding, the notion of 'hearing' imptied in. such contexts relates the souwd of the name, nrutu, of the object being spoken. Similarly, Hallpike (1979: 406) quotes with approval Read's statement about the Papuan people Gahuku-Cama: The Gahoku-Gama do not ascribe any importance to the brain, nor have t h y my conception of its function. Cogniltive processes are associated with the organ of hearing. To 'knowko b Wnk' is to %earsEgekn6ir~e)l; Pomw"C know' or '1 don't understand' is Po not bear' or Ti have nolt heard' @eSemu~e].[Read 1955: 265 n.)

6. Semantics and "Primsirive ;TSEoughf" 199

Hallpike overlooks the crucial distinction bemeen polysermqr and vaguess (and so does Bain)~.Ify for example, in the Australian lamwage ankunytjatjara the same word k u h i can mean either 'hear' or ?hinlc'(see oddlard 8992a), it is (as Goddard argues) a case of polysemy, not of aguemess. For example, in the sentenGe ((Goddard 1994b) Ngayulu alatji W i , "tjiingu~_rulia . . ." I t h k (abonlt it) like this, "maybe w e.

. ."

arly means 'think', not some hybrid between "think' and 'hear' (for the "hink' sense can take a quasi-quotationall complement). On the r hand, in a sentence such as Ngayulu wan@-n&u wangkenytjala kullinu I person-ERG talk NOML.LOC near PAST ' I heard peopPe talking' [for only the "ear' sense can take inlkV6lhear' polysemy in Yankunytjatjera (or ra] is exactly paralIel to the %ee'r'bdarstand' polysemy in both cases, the semantic difference c~rrelates with a difference in by . . ." can only mean 'I underand ajatji kuiini can only mean '[I)l even if there were no syntactic difmantic interpretations, rtwo meanings wouild still y semantic grounds: for example, I ~ e what e YOU my eyes what you mean'. semy involving basic concepts sucb as 'think' ith a possibility of different frames. In the case ustralian Western Desert Language (of which Yankunytjatjjawra is a particularly telling evidence is provided by Glass (1983: 40). This elves the use of a sl;lffivr (-kuk~n$a)and an enclitic (=tkmyu) , w as "mt") Clearly, bath these k'sense d the polysemous. verb
( 6 0 1 Tjilku pirni-lu=ya tjiimya kuli-ra palya-palya=lkanyu cmd many-ma m they you know think-PRfun=mt pituE-pa nganri-rranfljakukantja-llw. kapi-kukaatja-lu W~TQI-ABSI ~ ~ - T P F . M T - ~ G water-m(i-~m The children, you know, mistakenly think that petrol is lying about for fun, they mistakenly tMmk it is (as harmless as] water.
ightly the English gbss of this example. Glass uses the following abbrea glosses: abe-abscolutive, erg--ergatiwe, ipf-imperfective, mt-nrist, ps-present, pt-participle.



B n a similar vein, Bain states: There is no way to diEerenitiate the concepts of thiotkbg, listening and heeding in Pitjantjatjaoa. The same verb kulki does duty for aH. . . . This feature suggests that the psychicd and physical, the self and the environment, are not fuYly distiaguism from one another, a characteristicnoted by Piaget in oonnection with pre-operatory thought. (119912: 8161 X t is true that many non-Western societies use the same word for 'think' and 'hear' or for 'knowy and "ear'. But what exactly does that prove? In English one can use the word see to mean %ndewtand'('" see what you mean . . ."I, but tlis does not prove that the speakers of English do not distinguish ths concept of 'understanding' and the concept of 'seeing'. (Similirrlly, in French entendre can mean either 'hearhr 'understand'; but this does not prove that the speakers of French do not disltinguish the concepts of ' h e a r i n g k d ndunderstanding'.) Admittedly, in English there is also a separate word, underssand (and h French, ccomprersdre], which has only a cognitive meaning. This doesdt change the fact, however, that see is polysemous in English, and that in a sentence such as "I see what you mean" it hhas a purely cognitive meaning. But if see can be polysernclus between %ee (with one's eeys)' and 'understand', why can't the O m u r a word iero be polysemous between 'hear (with one's ears)' and 'understand' or 'knowv ~Cj~us' as t the French woad en tendre is)? In Polish (and a number of other Slavic languages) the word For 'knowing" wwiedztet, is cognate with the word for 'seeing', widziet (see Bdllncknlet 1970). Both! derive from the same proto-lndo-Europeam root veid%now/see' [see also video, '1 see', in Latin and vdda, 'I know9, in Sanskrit; Ernout and Mcillet 1963: 734).

200 General hmes

(61) Klruli-mu

tjarrpa-ngu=lktunyap kapi-ngka palunya-kukantja. water-in that-mt '(They) thought that they had gone into the water, that's what K h e y mistakenly thought.'
think-~.m enter-P.PF=~~

Examples such as these make it crystal-clear that even if the concepts 'think' and 'hearbhare the same lexical exponent, this doesn't mean that the concept of 'thinking' in a purely cognitive sense is missing. It is no less present in the Western Desert Language than in languages where Yhinking" has a unique l e x i d exponent, for example the Australian language Ngandi (Heath 1978ib:1147). What applies to 'thinkkpplies also to 'know'. A word which is used for both "ear' and "noow" amd which can be used in a sentence incompatible with a 'hear' interpretation, must be interpreted as pollysemous; and when one looks for syntactic diflerences linked with the diflerenoe in meaning one can us;llallly h d them. For examplle, in the Papuam language Gahuku-Gma [now called Mekanopl, whkh Hallpike uses as one of his prime examples, the word for both "think' and 'knowys indeed the sane as the word for %ear" yet according to one of the best experts on this language, Chris Deibler @ersonall oramunication), the three senses of this word ('think', 'know', and 'hear" can always be distinguished by the frame in which this word is usedJ so that ambiguity does not arise. If the word in question is glossed as "eroeive', as is usually done, then the sense 'heaskan be associated with phrases such as "talk perceive" or "say permive", the sense 'think" with the phrase "one's ear perceive", and the sense 'how', with phrases such as "thing perceive"". For exa~llrpie (Dei'bler, personal letter):
(1) na-gal guluumb

The traditional! gloss 'perwive' has an obvious value for describing facts of this kind, but it cannot be regarded as an accurate representation of the word" meaning, since quite clearly three diflerent meanings are involved, not one. For example, scntencc I can mean only Y think', not '1 hear', and sentence 5 can mean only 'God knows everything', not G o d hears everything'. The fact that "ear-perceive" means in this langunage 'think' and not 'hear' is particularly telling, since in many other languages (e.g. in Kalam, see below) 'cear-perceive" means 'hear" not %ink'; this shows that in Gahuku-Gama "ear-perceive" is in fact lexidized in the sense 'thinkJ. Words for body parts often provide a convenient idiom for talking about inner states. One language which illustrates this particularly well is Hua (another Papuan language, geographically very close to Gahlaku-Gama, and described authoritatively by Haiman 1980b, 19911). Pru Hua one feels, so to speak, with one's guts, one thinks with one's ears, and one knows with one's eyes. What this reaHy means, however, is ahat Hua exhibits a certain pattern af polysemy (Haiman 1991 and personall comwnication):
1. ear, 2, opinion ( beta havi1. hear, 2. understand ( 1geX;ahavi- think @it.hear one's ear) il ) a i P 1. guts, 2. feel

my-ear I-perceived 'I thought4 think" (2) das~mo gak6 mukii geleneive. God talk all he-has-perceived 'God has heard everything.' (3) l&tu lako Ri m6 nene dasim geleake . . . thus saving he-said Gad having-perceived 'God hearing that he said thus . . ." iC4) dasimo net& rnuki gdeneive. God things dl he-has-perceived 'God haws everythiag." ( 5 ) diasimo net& muki-kumu geleneive. God things all-about he-has-perceived God knows about everything."

Isrr English one can refer to one" thoughts as one's "view", and in Hua, as one's "ear"; but it would be absurd to conclude from this that the speakers of either English or Hua lack the concept of 'thoughit'. Lexical evidence is vital for establishing a culture" concepts, but without in-depth semantic analysis lerricall evidence can be easily misinterpreted. For example, when one hears that in the Papuan language Kalam the same word (rap)can translate both know and hear* one might conclude that the language makes no distinction between the two. In fact, hawever, Andrew Pawley's data and comments (Pawlley 19166, 1975, 1986, personal conmuarication) show that Kalam does distinguish between 'know' and 'heary:ng means essentially 'know', whereas tmwd ng (lit. "ear known")eans 'hear'. In sentences referring Ito sounds (such as thunder) m w d nsg can be abheviated to ng* but in this context the bare farm ng can be regarded as elliptical for ~ m w d ng. This analysis is supported by the fact that Kalam has many other lexical wilts including the s k m no, and that several of those other lexical units, too, can be abbreviated to a bare ng. For example:
wdn dl fib gas niy ng ng ng 'see" 'feel [by touching)' 'taste' 'think' ((lit."eye know") (lit. "touch know") (lit. "eat know") (lit. ""thoghthind know")

202 General ISZU~S

6. a T ~ ~ r ~ msd ~ ~ ~ "SrS~~s{tivc f f c . ~ Thor~gh~ 203


pk bwk mepn sb

ng ng n;(~ ng

kudge" 'read, study' Tcel nWTecction for' 'feel sorry for'

(lit. "'hit know") (lit. "book know") (lit. "liver know") Uit. "gut know"]

'knowheadsee" is strengthened by data from related and surrounding lanEtimres. Thus, if among geographically close and genetically ly related languages oFAusLrialia scr~mnc haw separate words for 'think" 'know7e.g. Arrernte; see Wilkins use 1993; and for Willkikins m s others (e.g. Yankunytjatjara) theHarkins same word both 199411, 'Rear' d "think', it would be bizarre to infer that the Aarernte people do have a concept of 'think'whereas the culturally and liingwisrtically closely related Yankunyltgiatjara people don't. The evidence available to date suggests that all languages do in fact have words for 'know' and %hink'. These words or may not be poEysernoms, but this is irrelevant from the point of view language" conceptual resources. allpike (1979: 391) writes: "Even when we encounter among the prima word we are disposed to translaite as Yhink" it commonly has the meaning of kbvious mental effort', as it does among Fiaget's preTmgu word ngek'ngeki, 'to think, ponder, cogitate, dge 18969: 176) ]." HaJPpike is probably right: it is age will have a separate word for 'think' aving a separate word for 'thinking' in the more k that . . .'.The Australian language Kayardild studied by Evans (1985) is a good case in point. In Kayardild the word ha flit. '"head-put") means "Xhinlk of*recall, come up with somenalmarufh~lc~ nithi, cuncPe will rough thinking', for example k a k ~ j c will emP1 the name' (Evans 1994: 21 1); and there is no word which would mean specifically 'think that' and nothing else (although the word rnarralmaru~ka,lit. "ear-p1at", can be used in the swse 'think thatY, as well as in the sense 'recall'). But Hailpike's argument rests on his assumption that if a word means %ear' it cannot also mean "Bhnk'. Inr fact, if a language has two words, one peaking, 'think w i h effortq,and the other is pollqir h n d 'think (that)', the argument fails. Referring to of ~onciepts such as "a', 'some" 'number'%and 'time', (1979: 390) mites: ame wag, when we are considering wards that relate to cognitive prouesses, aernembler that this word, together with those r', 'stupid', and 'understand', can bear simpler interitive, and that It is possiible for primitives and the in relation to behaviowr, facial expressions, bodch, while leaving out of account their distinctively cognispects. We wauEd not expect to find discussions in primitive society about the emce between knowing and believing, for wampl, or appearance and reality. at issue here is molt the use that different societies make of as Yhink' and %now', but the availability of such concepts. ords for 'think'aarud %now' may be present but may "bear

In e sentence such as b byn nq-k men woman know-he-past (punctual) 'The man saw the woman" the bare stem ng can be used in \the senwe 'seeyand in fact only in that sense], but this doesn't mean that in Kalam h e same verb [ng]~ maem some d k g " f w y E 'a r intermediate between 'think', 'taste', ' r e d , Yeel sorry', and so that K a l m distinguishes lexically between b o w ' , 'hear', 'see', and 's as foBows: n g know; ( w h ) ng see; [tmwdl ssg hear; Q ~ ~ nfjr, Y ]smellAlthough the bare stem ng c m associated with different senses Qkno see, hieirr, taste), in actual speech these senses are clearly distinguiskdr the object refers to a sound, ng has to be interpreted as %ear" if it refers an odour, it has to be interpreted as 'smell; and so on; and if the oibj, refers to a concrete entity (e.g. a person or even a bell), Ithen the only P sible interpretation of ng is 'see" and never %earyor 'smell'. It is also important to emphasize that in wrtain frames the only possibb H know', not perceive"~see, hear, or whatever]. For exam reading is ' (Pawley, personal communicaltion):


yad LJlnike fin akag OW-a-k ngb-yn. I iLJllliLe day when come-Tsc-Past know-Pres-lsa 'P know ["lpemlve] when Ulrike cane." yad Ulrike md-pnuglb-yn P UXke stay-Fres-3s~ know-Pres-1.w 'I know (*grerc.eive] wbew Ulsike is.'6
n e case far positing distinct senses For verbs such as 'knowhear' (

that ng has a meaning COP which there is no word in ward perceive as a label for this inexpressible s m e t a bit like saying that ng means 'X' [the hypothesis hypothesis h a ng by itserf means 'know', and ~ o nu s

Descartes (1701/1931), among many others, has pointed out, nothin

be 'ksimpler" than "hink'ar 'know' (see Boguslawski 19719, 8989).

claim that in some languages words for "think' and 'know'are not sense. Referring with approval to Gilbert RyBe's view that "at the c o m a i n level our assessments of mental processes are in fact assessments of Be iour" "979: 31841, Hallpike goes on to say: But Needham also lists trmslatiolns of some other Nuer words and which mem to refer mom up11embigauously and explicitly to inner states

7, General Discussilora

who share this belief, d l distinctions between ''them" and "us" in

Iber', 'forget', 'devec', 'stupid', 'understand",and so on. But the point is that all th aspects of cognition have behavionral memi~estations, too . . . In shark, it is the external manifestadionns d i r r m n e r states n in which primitives interested, and inn these external mranifestaliions the body has a crucial rolIe. The statement that a given word means "think', or ' b n ~ w or ' ~ 'hear" c be tested by the usual semantic methods; if the application of semantic t a ' shows that a word means 'think', or is polysemous b e t w ~ n results of semantic analysis. The semantic relevance of behaviourai manifesta'Y.ionscan be tested. example, in English the words merry and gbamy refer in their very m do not: 'He was mrryilgloorny, but he didn? show it, He was happyisad, but he didn't show it. To my knowledge, no similar evidence far the relcvance of behaviourlrl manifestations ta the meaning of the words for 'think'ar "knowVin Australian, Papuan, or m y other languages has ever been produced (by Hainpike or by anyone else). 1conclude that while Hallpike's claim that in many nlan-Western cultures "the realm of purely private experience receives very little elaboration or analysis at the level af public disc.aurse9' is uundchubtedly correct (see e.g. Howell 11981; Lutz 19881, his assertions concerning the alleged absence of wards for 'think'and 'knolllr'(and the concornnitant absence of the corresponding concepts) are unfounded.

Hallpike" Faundcztiom o f Primitive Thought;Shwedes 19821, pollogists continue to be ~quike pious about the 'principle ail'

If a shared basis of universal concepts did not exist, the M e r e n t conuniverses associated Gith dflerent languages would be mltudly
in the psychic unity af humankind and i n the primiple that whatan say in one language one can also say (more or less easily] in one cannot at the same time reject the hypothesis of a set of


concepts ~EIE ~ o ~ n m a n demphasis " added]. The question of native speakers of different languages bave the same basic concepts

estipted., and a ttheoretiml framework was lacking within which it could seriously and rigorously investigated. Cole, Gay, Glick, and Sharp 01971: 2151 wrote: 'The allmost universal

cancQusionthat "cultraral differences in cognition reside more in the

206 Geneml h u e s

6. Semml[ic~ crmd "Primitive Thaught " 207'

anakysts of the kind that Uvy-Bmhl was to put into eflecb, almosb exactly uries later, and even in tenns that find ready agreement today; but it is not , not the type of reseauch that he recommended, that have since been called o renewed question. Underlying his proposal w a s the conviction that human ure was uniform a d fixed, and it is pr~wiely this idea that more hecent concrepDual analyses have made difficult to accept.

situations to which particular cognitive ppacxesses are applied than in the existence of a proms in one cultural group and in absence in another"". A4wsali.v mcd~wdi~, the same concEusions emerge fmm research into crosscultural semantics. On the one hand, the almost universal oubome of the semantic study of culture and cognition has been the demonstration of large differences among cuPturaB groups with respect to their patterns of IlexIcalimtion, in particular, their key words and key concepts (see Wiembicka 1991b). On the other Inand, it emerges that in addition to the vast m s s of culture-specific wnmpits, there are certain fundamental concepts which appear to be lexicalized in all languages of the wodd; so that cultural differences b e t w ~ n human groups reside in ways in which these basic concepts are utillhd rather t h in the existence of some ooncepts in one cultural group and thejr absence in another. It might be added that there are also considerable differences between cultures in the extent to whi& certain basic concepts are called upon. For example, the concepts d 'bewuse', "if and 'aal may indeed be utilized much less in the cdture o f hstralian Aborigines than in Western culture (see N. Evans 1994 on 'because'; Bittner and Hale 1995 and Harkias 1991 on 'dl'; also Goddard 19189c3). But this doesn't mean that these concepts are absent, or that they are not lexicallly embodied. It should also be mentioned that (as argued by Goddard 1991a) the low frequency of some indefinabte wards in some languages may be comperrmsated for by high frequency of more inclusive "semantic moledesyJ, Goddard argues that this is the case with the notion of cb'because' in Australian languages, which is often included in "semantic molecu~es" encoded in various purposive constrlactions. Goddard (1Wla: 44) writes:
Prime facie, I submit, the meaning oE these constmcllions involves the notion of became, in combination with the complex notion o r m ~ ~ wanting n e somes"hifig $0 happe~. . . The purposive constructions, in other words, provide a compact means for arliicwllating causal connections within a particular broad domain-that having to do with people" motives or reasons for doinug things. In P N society, I wouEd argue, people lend to be more interested in each others' reasons for doing &in@ than in other kinds of causal Ilinks, Most talk about reasons for actions takes plaoe in the idiom of the purposive, which sesvices the main needs in respect of the expression of causality.

Hallpike (19791, who quotes Needham extensively, drew logical wmrclusions m these relativist statements [although one may doubt whether edhaam himself would have endorsed HaIlpike's theories of ''primitive In a sense, however, serdi~~m nom dcrtur (there is no third possibility): either Leibniz was right, and these is, behind the variability of cdtures, a universal, "fixed and uniform" set of underlying human conciepts, or Needham was right, and there is no '%xed and unifam" conceptual basis of different language-and-culture systems. Linguistic evidence suggests that the truth is on Leibniz's side, as do conp h a l analyses more recent than those referred to by Needham. Language-and-culture systems differ enormously from one another, but there are also semantic and lexical universals, which point to a shared conapturil basis underlying all human language, cognition, and culture. But in order to establish what is and what is not universal in human linguistic and conceptual resaurces we need a rigorou~ methodology. In particular, we need a methodology which would allow us to recognize polysemy where it is really present without positing it in cases where it is This point must be emphasized, because many scholars have deep-seated fears of ever positing polysemy for another language and imagine that by doing so one will inevitably fall prey to ethnooentrism. For examplie, one the anonymous reviewers of an ueadier version of this chapter ieszbicka 1994g) cautioned against "the assumption that if a term in lanage X requires more than one distinct translation to cover a range of es that the translatodanalyst devises, then it must be polysemous . . . , if w can imagine a distinction, as for example, between 'hear' and ink', for which a single term is used ia X (but f a r which multiple terns are available in our Ilanguage), this (putatively universal) conceptual distinction must be polysemously fabeled in X",and illustrated this point as follows:
This is as if to say that if a tern, e.g. ainiwa (Sahaptin] is used to refer to a ''bee" at one time and a "wasp'ht another, that it must have these two distinct senses. Why reject out of hand the possilbility that the distinction is simply of no corrsequlenw: to the spneahrs d A ? "IPerps, if it is ru distinction we find dificull to imagim doing without, as between '"1" and '"ome", we would want to argue strenwlously for a polysemous interpretation.

Eeibniz (176UI981: 3261, who Bmdy believed in the psychic unity of humankind, recommended comparative study of different languages of the world as a way to discover the "inner essence of man" and, in particular, the universal basis of human cognition. Nleedham (1972: 2201 commentled on Leibniz's ""grand propasa!" as follows:
This b d d suggestion . . . was based on the tacit premise that the human mind was everywhere the same. . . . Methodologically, LeLeibniz w a s thus proposing a com-

6. Semantics and "Prhiffve nought "2209

But the question is not whether it is easy or difi6icult to imagine doing without a particular distinction, but whether the necessity of a particular distinction can be established by reductive paraphrases with full predictive pwer. For example, ' ' ~ and " "wasp" can no doubt be r4uw.d to mma thing along the following lines: 'k small flying creature (small enough far a person to be able to hold it between a finger and a thumb); it has a shnp thin long part; it can sting peaple with that part; when this happens to someone it hurts"". If this c o r n o n core fits all the contexts in which h e Sahaptin word ~tni'wa can be used (a point which can be tested), then positing polysemy for this word would be totally unjustified. But the relationship between 'think'and 'hear' is quite different f m that h t w m 'bee' and 'wasp' (m from that between "rl" and 'daughta')I. If a loomon core [substitutable in context) can be articulated for Ehe Sahaptin word atnfwa, no c o r n o n core (substitutable in context) can be articulated for, say, the Pitjmtjatjara word M i n i (or for the French word

Certainly, some of the universal concepts discussed here may be used more frequently in some cultures than in others. For example, it may well that in some Australian languages, in which the concepts 'tlninkhanrd arbbane the same lexical exponent, this exponent is u d much mare frenfly with the meaning 'Ineas'tthan with the meaning 'thinEr"; and also, t references to 'thinkjing'are much more common in English discousse n in, say, Pitjantjatjara discourse. But the question of the extent of use certain conncepts must be distinguished from the question of their availMore generally, the availability of cognitive resources should not be confused with the habitual use of these resources in different societies. Differences in the latter are particularly clearly illustrated by Luria's (1976: I l l ) interviews with Uzbek and Kirghiz peasants. For example:
[Q.] In

One could of course suggest that 'thimk' m d 'hear3ave re a m a m care which simply Itaannot be articulatied, but this hypothesis is utestable (because ta hypothetical meaning which hasn't been articllllated cannot Ire tested in context), and therefore them is little point in entertaining it. The status of such an antestable hypothesis must be seen as quite diRerent from the sutttus of the testable hypothesis that, lfor example, the Sahaptin word atniwa has a unitary meaning. If one takes into account that RuJhi occurs in two difTerent grammatical frames, each of them assodated with a different sense Ctlrink" in one frame and %earJ in anotlner), whereas a m h a o~ocwrs[presumably) in exactUy Ehe same grammatical frames whether it is used with referenw to bees or to wasps, one can see that the analogy between the two cases is more apparent t h m real. Similarly unjustified (though understandable) is the fear oil'ethahwentrisrn expressed in the following comments (by the same reviewer):
Perhaps it is better to attack the mswpUion that every distb~ti!on we judge rdewant md nwsessary for coherent abst~action must be basic to 'kdult" ratiocination. For exampie, re. the alleged "fai8ureqQo distinguish 'tarusal re1ations"from reladons of simple spatio-temporal contiguity: owr philosophers argue interminably about the validity of causal inferems ram observations of spatlo-temporal contiguity. Why should ;dm cultures make the same logical errors we are prone to?

the Far North, where there is snow, all beam are white. Novaya Zedya is in the Far North and there is always snow there. m a t wlou are the Bears there? the beaus there are,I never saw them.

a.] Once I saw a hear In a nruwm, but that's all.

Luria comments that "the most typical responses of the subjects . . . were ;a complete denial of the possibility of drawing conclusions from propositions about things they had no personal experience of, and suspicion about any logicail operation of a purely theoretical nature" '1976: 108). At the s m e time, however, Curia" interviews clearly show that his interviewees did have concepts such as 'allymd 'if', and that when pressed they could draw the desired inferences. For example (1 11):
.I] But on the basis raf what I said, what color do you think the ibears there tare? .] Either one-colored or two-cdored . . . Iponders for a long time]. To judge from ~e place, they should be white. You say that there is a Pot of snow there, but we have never been there?

And another example (1093:

I Brut what kind of bears are there in Nolvaya Xedya?

Brut what do my urords imply? . . . Well, it's like this: o w tsar isn't Pike yours, and yours isn't like ours. Your words lcan be answered onEy by someone who was there, and if a person wasn't there he can't say anything on the basis of your words . . . But on the basis of my wards-in the North, where there is always snow, the ears are white, can you gather what kind of bears there are in Novaya ZemPya? If a man was sixty or eighty and had seen a white bear and had told about it, e could be believed, but I ' v e never seen one and hence 1 can't say. That's my last word. Those who saw can tell, and those who didn't t e can't say anflhilag!


] We always speak only d what we see; we don't talk about what we haven't seen.

But concepts such as "because" '"if", 'think" 'knncrw', and 'ail' are not just "ours": they are well attested in numerous languages of Asia, Afsi@a, America, Australia, New Guinea, and Oceania, and their exponents are by no means always polysemous [for evidence on this point, see Goddard and Wierzbicka b 944b).

2 10 General Ismes (At this point a young Wzbek volunteered, "From Jrourwords it mems that beam there are whifie."] [Q.]Weill, which d you is right? [A,] m a t the cock knows how to do, he dots. What I know, E say, and nothing beyond that! The availability of canmpts such as 'all' and W'is crucial to deductive r a sonimg; all the rest can be learnt (as Luria" data show, quickly learnt, given sufficient cultural exposure]^. Different modes of thought do not make human cultures mutually impenetrable if the basic conwptuaI resouroes are the same. As Franz Boas wrote (19380: 1411-2):
I r m primitive culture people speak only about actual exprienoes. They do mat discuss what is virtue, good, evil, beauty; the demands of their daily life, like hose of our uneducated classes, do not extend beprnd the virtues shown on definite msions by definite people, good or evil deeds of their fellow tribesmen, and the k w t y d a man, a woman, or d an object. They do not talk about abstract ideas. The question is rather whether their language makes impossible the expression of abstract ideas . . . Devices to develop generalized ideas em probably dwsuys p ~ mnt and they are u d m soon as the cultural needs campel the natives, to farm them.

Semantic Complexity and the Role of Ostension in the Acquisition of Concepts

1. Introduction
have been impeded, more than anytatabions. Familiar analyses of meaning along se to die' or bachelor equals 'unmarried n regarded as satisfactory, Itrut they have helped to twate the illusion that If the meaning of a word cannot be stated sstmula of three or four words, then what is needed plex one, of five or six words. who tried to go beyond formulae like ts kii! = se to die' found that trying to go just a little further didn't seem to histicakd semantic formulae were as ttne cruder and simpler ones. in meaning of wards the more elusive this rying to catch it in verbal formulae was like tryce, many scholars previously interested earch altogether and turned to other, ing to justify their abandonment of semantic iption in terns of new ideas about the nature of meaning. Meaning, defined-not bemuse we haven" yet cause it is, by its very nature, indefinable (see 1981; Chomsky 1987; Lakoff and Johnson usion was frequently accompanied by attacks on Plato and Leibniz], by almost ritualistic references stein, and by assurances that meaning cannot be described hawse it is "fmya"" Ofteny'Tuzziness" came to be oelebrated as almasd the ultimate truth about h m a n language and cognition. One could h o s t hear a collective sigh of relierf: meaning is "fwy'" so we don't need to try to describe it. But in linguistics there is no escape from meaning. Meaning is what

Discussions of cm~eptualand lewiml resources in non-Western Ianguages such as those contained in HalUpikeYsbook tend to be based on anecdotal information [see Lave 198P) and often lack linguistic sophistication. But if dalmw like Hallpike" are to be suocessfullly refuted, they have to be refuted on the basis of sdid evidence and sound andysis. The task of determining the hll set of universal concepts which underlie the "psychic unity of humankind" is viitall and urgent. Empirical Ilinpistic investigations reported in Semantic and Lexical Univer8ab (Coddard and Wiemlbkka 1994M suggest that in all prolbillib11ity the rnetapredliwtes 'ifax 'because', and 'dli', the mental predicates 'think'and %now7,and the basic "~ubstantives"%omeoneharrd'sometluingkar among their number.

language is all about, and the study of meanhg whet linguistics is, ultimately, all about. Ohiousl to make the study of meaning his or her primary concern: t important tasks in linguistics. But Binguistics as a whole m its responsibility for the study of meaning; and this means avoid the problem of semantic c~mplexity.

r w e d we need a list of semantic simples) was pointed out not by Locke Leibniz also saw clearly the dilemma stemming from the mutual depenur knowledge of simple concepts and our understanding of plex ones: lea understand complex concepts we have to decompose them what we assume are simple concepts; but to discover which concepts can be reasonably regarded as the simple ones we have to experiment with their power to ccgenerateJy complex d error that we can discover the ulltimate simovered them, all our semantic analyses must ional and to a greater or lesser degree incorrect. s, H shall look briefly at a few diSfercnt types of concept, to assess their complexity (relative to the postulated set of "sims mentioned earlier (Chapter 11, data from child language are relevant ve to be handled with care, since to study conceptual herent semantic theory is needed in the first place. ists sometimes argue about such matters in a some. For example, Chierchia and McConnell-Ginet (19901:
on the issue of lexical decomposition.

2. Complex Concepts as Configurations of Simple Ones

The complexity of a concept can be viewed as t From the level of indefinables. Some meanings guages can be regarded as ""sirtrpleY~in the sense that t decomprased (without circularity) into any otber meaningsmFor exa as argued earlier, one cannot decompose (or define) concepts s u ~ as h Gsomeonel, or 'know', and any attempt to do so must lea and obscurity [as when one tries to define this in terms of dehb, osfiensiol definiteness, refirentiality, and lsiol on; or k m w in terms of in~ursrratian, fa tiuliby, wesa$co~ion, and the Pike). One could of course say that concep such as 'this', "someone', or 'know' are c~omplexw in som senses of the tern "complex"; I maintain, h plex in the sense of being cdwmpo~bubEeinto sirnlpbr than themselves. in this sense (and only be regarded as the "ultimate simples" of any valid @once natural language. By contrast, most concepts e n d e d in any human language ere "IC~CP plexY9nthe sense that they EM be decomposed in terms of simpler col cepts. To stalte the meaning of a word is to reveal the configuration simple concepts encoded in it-just as Locke said nearly a definitions is nothing else but the showing the meaning of one word by other not synonymous bnns. The meaning of words made to stand for by him that uses them, the meaning of any or the ward is defined, when, by other words, the idea it is made tlhe s i ~ n o
annexed to, ins the mind o l the speaker, view of ansother; md thus its sigdicatio of definitionus; and therefore the only m tian. (Locke 11690il959: 33-43

But to 'tae able to present comp1ex meanings @r ooncepts) as tions of simple ones we must know in advance what the simp1 are, just as to analyse chemical compounds into th we must know irm advance what the chemical elements are. Semantic ds requires a list of semantic dements, just as list of chemical elements. This crucial fact (Ithat

ut empirical findings on chilid language demonstrate that the concept of usation (not in the fonn of the verb lo carrse but in the form of clausall nkers such as 'cause and because) is quite common in the speech of 2-year. Whether the words kill and die are learned earw: the authors probably don't know either but simply guessing. Furthermore, they take for granted that when children they link them wikh the same concepts to the relevant literature. Finally, the composition of the adult concept 'kill' should n the verbs E;O cause and to become (as suggested twenty-five years ago s that the authors'view of semantic . Wierzbicka 1980rs.) The same applies to Fodor" (1987: 161) comments on the meaning of

2 14 General Issues
farher: "Children know about fathers long before they know about males and parents. So either they don't have the concept FATHER when they seem to, or you can have the concept MALE PARENT without having aocess b internal structure; viz., by having the concept FATHER. Of these alternatives, the last seems best." 1would certainly agree that the analysis offafher as "male parent7Ysmisguided (see Wierzbicka B972], and that the child's concept of 'father' is not that of 'male parent'. But what does it prove? For adults, the knowledge of, roughly speaking, 'begetting' is a park of tlveir concept of 'father'; it is necessarily true, then, that when children acquire the adult concept of 'father" they m s t also acquire this knowledge. Obviously, when small children use the word DatEdy they don? teed to have that knowledge; so it is not true that small chilldren necessarily "haow about fathers"(just lsecause they use the word Daddy). Words such as M o m y and Daddy are among the very first wrds learned by children (see Anglin 8977), but they are first learned as names far particular people, and the road from there to the adult concept of bother' and 'father' is a subject for serious study. Fodor's comments on the subject suggest limited familiarity with the work done thus far. His conclusion that children have an unanalysablle concept of FATHER, which i8 the same as that of addts, denies the whole idea of conceptual development and overlooks the vast body of empirical research on child language and on the aquisillion of meaning. And yet Fodor's musings on "hthar'hnd other similar passages are ofkn adduced in books on "'dbrmal semantics'" as evidence that the whole "deoompositional approach'" to meaning is untenable (see e.g. Chierchia and McConndl-Ginet 1990: 363).


7. Osrensionamd Concept Acquisition 215

er, for example, the English word happy and the Polish word given tionairjles as its equivalent: szczg?$liwy. As Barahczak (1990: 12-13] out, the range of use of the two words is not the same. t b word '%appyY" ,rPuerps one of the mast frequently used wards in Basic ican. It's easy ta open an English-Polisb or EngllshRussian dictionary and an equivalent adjective. In fact, however, it will not be equivalent. The Polish d for '%splpyW"nd I biebiavc tlluis ;ulso Bralds foe otJ~crSlavic languages) has B uch more restricted meaming; it is generally reserved for rare states of profound , or tala1 xatishction with serious things such as love, Iramily, the meaning of and so on. Accordingly, it is not used as often. ns "happy" is m i u American comIt is not only the Polish word szcsg.t/iwy or its counterparts in the other avic languages which differs from the English word happy in fie ways ibed: the Geman word gMcklich and the French word heuirem diner happy in much the same way (see Wierzbicka 1992~). To amount for ase diflerenms, [I have postulated for these words he following two expli[A) X feels happy. = X feds something sometimes a person thinks something Bike this: something good happened to me I wanted this E don? want anything more now because of this, this person feels something good X feels. like this (B) X feels 8zczgStlry &lUcklfcEEca, heureu, etc.). = X feels something sometimes a person thinks something like this: something very good happened to naa: I wanted this everything Is good now P can't want anything more now because of this, this person feels somethitg very good X feels like this

3, Abstract Corroeplts: Words far Emotions

Generally speaking, abstract concepts appear to be less complex than concrete ones; but even so, they are usually much more complex than simple d i ~ tionary definitions or illustrative semantic formulae offered in scholaaly literature would lead us to believe. But very simple definitions of this k i d @.g,e"o tie-'to say something untrue') do not have any predictive power, md they cannot m u n t for the differences in the range of use of related concepts. For example, as pointed out earlier (Chapter 4), a definition of "ie' which says that "0 lie' is to say something untrue cannot account for the differences in use betwmn lie and its dosest Russian counterparts wrat' and [gar', both of which also mean, roughly speaking, Yo say something untrue'. To my mind, if we want to assess a concept's real complexilty we must seek to reveal its structure in a formula whose validity could be verified against its actual range of use. Otherwise, the formulae we devise willl reflect nothing but our own preconceptions.

two explications differ in three respects: First, B has one additional ponent, "everything is good now' (by implication, "wrything that is pening to me'); second, 'good' in A contrasts with %err good' i n B; and d, 'I don't want anything more nowqn A contrasts with 'I can't want thing more now' in B. These three diflerenms account, I think, for 'caabsollmte" ~oonotationsof szczgs"[iwy and the more limited, more pragmatic character of happy, discussed by Baraniczak and confirmed by

7. CJ'sserrsbn and Concept Acqaisits'on 21 7

numerous linguistic facts such as, for example, that one can say quite h p p y but not *ca&iema szcz~dEiwgr or * g m z gIr;r'ck/ich, or that one can m y Im happy witk this arrmgement but not * ~ ~ SszczgBiiwy S I P ~ r $egosrkfadu or *Lk bin gIf4ncrfrls'chmit dieser Anordnmg, A few further examples from the area of emotion concepts [sex also Chapter 5): Terrd@ed ' A feels something sometimes a person thinks something Pike this: something very bad is happening I because of this, something very bad can happen to me now I don't wmt this because of this I would want to do something if I could I can't do anything bemuse of this, this person feels something very b d X feels something like this If one is ferr$edr what one is tcrr@ed of is seen not simply as 4sometln;i;ng bad'buk as something 'very bad'. What one is tesr$ed of Is very resealsomething that is already there. A m n d yet the target of iterror is also p d y in the future' because the present 'bad evenlt9s seen here as a source of a tuture threat ('. . . can happen NOW). This future threat is necessarily personal ('sometkning very bad can happen TO ME now'). The exp~enmr's attitude is one of an intense non-ancrceptance ('I don't want it']; at the same time, it is one of totd helpkssness ('1 can%do anything').
,Petr$ed X feeis something sometimes a person thinks something like his: something very bad is happening something very bad will happen to me now I don't want this because of this, I would want to do something if I could I can't do anything because of this, this person feels something very bad bemuse of this, this person can't move X feels something like this iPetr@ed appears to be a more spec& version of terrged: it is a ferrEpr which beads to a kind of paralysis: 'this person can't move".mote, however, the difference between 'can h*rppm'in rerrijied m d l 'will happen' in petr@cd,) Horrged W feels something

sometimes a perlaon thinks something like this: something very bad is happening to someone I didn't think that something like this could happen 1 don't w n t this because of this I would want to do something if I could I can" do anything because of this, this person fecls something vcry bad X feels something like this main difference between horror and terror concerns the relationship the exrperiencer and the victim: in the m e of terror, the two are , whereas in the case of horror they have to be different. One is omeone else, as one is appalied to see A second differemnm between horror to the first one] has to do with the prehe farmer: since horror is, essentially, the feeling of a primady what is happening kow'(Cin a b r a d sense), than what can or ~lrsi81happen after now. is kind are very different from so-called sad on a set of necessary and sufficient conditions, to concepts but to denotata (so that one can say of any extralimn-

is a fairly precise and flexible tool of conoeptual analysis, and itae and elusive aspects of meaning far beyond the ation which was aimed at in earlier analyses of ortantly, definitions of the kind proposed here Iy verifiable; they canan, therefore, be distested against native speakers'intuitions, and ended .on the basis of such discussions. Though not perfect,
all concepts encoded in natural language are reducible to le, univensal, and inherently "c1ear"concepts such as NE and SIOMETBIING, SAY and KNOW, or GOOD and ]BAD, 9 always be so reduced, so to speak, in one go, that is in a single complex. Often, they can only be reduced to the level conceptual simples step by step. In saying this, I am relaxing a pinciplle semantic analysis which H have defended for two decades, and which was stulated by Leibniz as a necessary empirical check on semantic analysis, to some words (notably, particles): oper explanation of the particllcs it i s mot sufficiemli to make an abstract . . but we must proceed lo a paraphrase which may be substituted in

its place, as the definition may be put in the place of the thing defined. When wle hnvc elo'ivcn In ~ c c k stlcl tn dclierfiami~ie tlucse ,f:t~itnble pnrnp~nansc~il~ nil lhc pnrtidcs so far as they are susceptibic or L lm e m , we sUaall Y l u s u x e rcgnrlekcd tllclr sigmilicatlons. (Leibniz 1765/11981: 333).

I still believe that this principle is vitally important, bwlt P no longer think that it can be applied as blroadly as 1suggested in a number of earlier publications (see in particular Wierzbicka 1972, 1980). In many cases, I now recognize, it can only be applied step by step. 1 will illustrate this with a number of examples of concrete concepts, beginning with relatively simplie ones, and tihen turning to more complex ones.

often, when a person feels something, snm~ething happens in this p:nrt otlrer peoplc cam see this because af this, when a person feels something, other geaplle can often know something about it

4. Relatively Simple Concrete Concepts: Body Parts and tbe Natural Environment
Among the simplest concrete concepts are some spatial concepts such as ctopyand 'boffomy: a part of something this part is above all1 the other parts of this something bot~om a part of something this part is under a 1 the other parts of this something
A somewhat larger cxtegory of (relatively speaking) simpb concept8 imncludes those referring to body parts.' For example:

head a part of a person's body this part is abow a11 the other parts of the body when a person thinks, something happens in this part

Despite their relatively simple semantics, many body part ooncepits appear to involve a hierarchical (transitive) structure, with, far example, the 'eyes', 'ears', c n ~ ~ and e y , 'mouth' being defined via 'face'* and "ace', via 'head' (see Cruse 11986; WiPkins 8981; Apresjan 1974, 1992; Mel'Cuk 11974):

a part of a person's hcnd

it is on one side of the head ilt has parts
I For an earlier allilemgt at analysilug the meaning of body part lenns, see Wierzhicka (i980: 77-97).

two parts of a person's face these parts are alike one is on one side of the face the other is on the other side of the face because of these two parts, a person can see ears two parts of a person's head these parts are alike one is on one i d e of the head the other is on the other side of the head they are not parts of a person's face because of these two parts, a person can hear nose a part of a permn's faoe one can think of this part like this: there ;are two parts above this part one on one side of the face one on the other side of the face there is one part under this part because of this part, when a person is in a place, this person can feel something because of this, a person can think something l i h this about a place: there is something bad in this place there is something good in this place mouth a part of a person's faoe this part has two parts, one above the other these parts are alike because of this part, a person can say things to other people other people can hear these things often, there: are some things inside this part because this person wants to do something to these things with this part people can think like this about this part: if there is nothing inside this part for a very long dime a person cannot live

7. Qstension and Concept Acquisition 221

What applies to body part concepts applies also to environmental concepts such as "sky'or " u n k r 'cloud" for which 1 would propose explliations rowg01ly along the following lines:
3 b

something very big people can see it people cain think like this about this something: it is a place it is above all other plaoes it is far from people

something people can often see this something in the sky when this something is In the sky people can see other things bmause of this when h i s something is In the sky people often fzel something be~ause of this cio~d something people can often see many things of this kind in the sky sometimes people cannot see the sun bemuse of these things these things can move Explicated in this way, a set such as 'eyes-face-head' or %loud-sw-sky" reminds one of a set of Russian wooden dolls. A 1 1 1 this may seem mot only very complicated but mneeessarily so: wouldn't it be better to admit, quite simply, that eyes are eyes, a face is a face, and a head is a head, and that's the end of the story? Of course it would be simpler to do this. Brat consider what words such as these [e.g. eyes) can mean to a blind person like Helen Keller (1956), who could use them in ways mlnat make pcrfect sense to sighled people. It seems to me that seen from the point o[ view of someone like Hellen Keller the explications proposed here sound psychologicaUy quite pbusihle. The words eyes, Jace, and head could have to Helen Kdler an ostensive interpretation, because she could know their designeta by touching. But surely, she knew not only which parts of her own face crri of other people's faces wem called eyes, but aIso that these parts-which she coulld b w ~ h - c a u l d give people some special, otherwise inaccessible, knowledge albout places. She also knew that people" faces can reveal something about their thoughts and feelings, and that their heads are not only above other parts of the body but also have sometlhinp, to do with thinking. In the case of 'cloud', 'sun', or " k g h n ostensive definition is even less

plausible, bemuse Helen Keller could neither see nor touch their designata; and yet she clearly did understand the concepts. I do not think, therefore, that the explications proposed here are unnecessarily connp8ex. k~t the same time I acknowledge tbat they are complextoo complex for gIobal, d-embracing, one-level paraphrases couched exclusively in terms of semantic primitives to be fully intelligibb. It is desirable, therefore, and perhaps necessary, that our definitions of concrete concepb such as names of body parts or names. of diiTerent parts or aspeds of the natural ersvironment should include semantic "molecules" as w e B E as semantic atom^". It slniould be emphasized, however, that a semantic analysis conducted in k m s of semantic 'koleculesY>ather than directly in t e m s of semantic '%toms'hakes no claim about the order of acquisition. Thus, if we define 'eyes' in tems off 'face', or "slm'in terms of 'sky', this does not mean that we expect children to learn the wordface before the word qes, or the word sky hefare the word mn,because the acquisition of concepts is none thing, m d the acquisition of words, another. Though we m y know that a child has started to use the word eyes, or the word sun, this does not mean that we h o w what concepts the chiid is associating with these words. Since a young child's use of words does not correspond to that of an adult (see e.g. knglin 1970; Clark and Clark 1977; Clark 1983; Carey 191851), we m n o t a s m e that the chill85 meanings carrespond to those of the adult; on the contrary, we m s t assume that they may be different. It seems reasonable to conjecture that children absorb the semantic universe of their native language gradually, moving, on the whole, from s h pler concepts to more complex ones. It goes without saying that the acquisition, of concepts by children requires much further study (across a wide range of langvages) before any fim conclusions about the nature of this process can be: reached. It is important to remember, blowever, that the task is exceedingly difficult, precisely because the acquisition of concepts cannot be equated with the acquisition of words. l a would seem obvious that systematic and methodologically informed analysis of adult concepts is a conditbn sine qua nora for the study of the grdual acquisition of these concelpts by children.

5 . Temperature Terns and the Colrvciept of Tire'

It s m s very likely that 'Ye~mperatureItems" are first learnt ostensivelyfar example, hot in connection with an oven or a heater, and cold in connection with cold water or drinh, or ice, or cold weather. At that stage of language acquisition, hot and coidrnay not even be thought of as opposites.

Buk the full adult meanings of these words are definitely thought of as opposites (although cold is also thought of as an opposite of warm]. We can amounrt for this if we try to reveal the conceptual point of referen~e contained in these concepts. I believe that such a paint of reference for all these three mncepts (hot, cold, and warm) is provided by the concept oE$re, and that slightly different concepts are embodied in the temperature words referring to the ambient air and in those referring <to objiexlts. (It is interesting to note in this connection that in Russian two different words are used for "hotyheather and for "hot" things, e.g. water.) As a first approximation, I: would propose the following explications:

or dothes and as applied to objects can be amounted for by assigning to the latter type explications along (roughly) the following lines? This thing (Xjl is hot. = if something is very near fire something can happen to this thing bemuse of this if someone touches this thing this person can feel something bad because of this X is like this This thing (X, e.g, some soup, mimilk, water] is warn. = if something is near fire something can happen to this thing because of this if someone touches this thing this person can feel something bemuse of this X is like this This thing CX, e.g. some soup, mill&,water) is cold. = if somethitrg is near fire something can happen to this thing because of this if someone touches this thing this person can feel something because of this X is not like this Though relatively simpile, all the above explications rely, crucially, on the eioncept of 'firey-and %reLitself is quite complex, much more so than the other environmental concepts discussed earlier [%sky','sun', or %@lloud')i. In view of this complexity, we might be tempted to say that "fire" is a purely ostensive term ("fire is what people call fire"). But this muld not be any more satisfactory than saying, for example, that "sky" Is "what we call skyyL, or "sun" is '"hat we call1 sun"'. For any native speaker of Englllshjre is more than just a proper name, and it implies knowlledge which can be spelled out. If we do spell it out, we can see, among other things, howjre is semantically related to sun. Speaking Informa8ly, fire is a phenomenon which can be seen (even at night], which can be felt by people nearby, which causes a profound change iE"h,ming"J in some things or substances, and which cam hurt people if they get too close to it, k mere precise explication is given below. For the reader's convenience, some '"raps" (in square bsackctsjr have been inclluded in a number of places. These props shoulld Lare helpful, but not essential, for the understanding of the explication.

It is hot now- = if someone is very near fire this person can feel something bad because of this people can feel something like this now P t is cold now. = sometimes people want to be near fire because they feel something bad people can feel something bad like this now It is warm now. = sometimes people fed something good because they are near fire people can feel something good like this today

As these explications show, both kof and caId as descriptors of ambient temperature imorporate value judgements, and in both cases, these j~udgemealks arc macgative. This suggcsilion is s~upportedby tehrer's (1990) abserwation that one can talk of a '"arm jumper" and perhaps even of a "cool dress", but not of a "hot jumper" ar a "cold dress": '"am jumpers" and "cool dresses" can be seen as functional, but hot jumpers or cold dresses would not because normalliy peop,e don't want to feel hot or cold. The contrast in acceptability between keep (orreseg] warm or keep cool on the one hand and ?keep hot or ?keep coSd on the other points In the same dirm tion. On t k other hand, hot b o d , or a hot bath, are not seen as undesirable (and keep (food) hot is perfectly acceptaMe); nor is a cold drink, or a cold compress an one's forehead, undcsirablc. The idca a l samething warm coming into such contact with our bodies implies a pleasant sensation, and while the idea of something cold coming inlto such contact is more Bikely ta imply something unpleasant this doesn? have to be the case; for example, the coltocatians caM beer or caid meat do not imply anything arr&sirabla (cf. cold ~~efhoscape), and don" sound odd, unlike coidj~mper or hot ~ocksl. These diflerences between temperature words as applied to ambient air


The? concept 'touch' (or 'ann,'contact') has been proposed as a possible semantic phimiby Cliff Goddmd (persona0 mmmunicailion). It has not yet baen tested cross-lingznisticalEyY

7. CEsfeflsimz md Concept Acquisition 225

There is fire in that place. = (a) something is happening in that place (b) people can see it (c) if at a time [at night] people couldn't see anything else in this place, they could see this (4 if someone is near that place, this person can feel something [warn, hot] bemuse this is happening (el something is happening to some things in that place leg. wood, coals] h a u s e this is happening (fl after this, these things will not be the same [they will1 turn to ashes, eitc.] (g) people carr think about it like this: (la) this is something (K) if someone touches this something, this person will feel something very bad Component (a) indimtes that fire is an event, or a process, (6) and [c) that it is highly visible, (4that it generates warmth, [e) and QSJI that i n the process some substances are "bwnt". Component (g) indicates that people think of this event or process in a special way, namely, (h) that it is a "thing"', a tangible thing ( i ) which, however, should not be touched (because one would bum onesello. What is particularly important about this explication Is that it doesn't include words such ;as hot, het~t, or burn (not to mention Pamres, hflmrmalien, carnbus~ioc~nr, Jrclel, ekc.) as conventional dictionaries usuallly do, which all, inevitably, lead to circularity. Long as this explication is, I have indicated that it may not even be complete, because it a n be argued that the full conmpt of %re'iincludes also some r~eferences b its role in human life: cooking, wamth, Bight as well as destructive unwanted fires. E n simple terns, these further aspecb of 'fire" can be articulated along the following lines:

be an argument about terminology (what is "meaning"") not about the substance: equally well, one could argue that the position of the head in the body (above the other parts), or the location of the sky C'above everything"") hlrellongs to "folk knowkdge"".ethm we call such things 'heaning" or "folk knowledgelideasy" they are part of the speakerskcommunicative competenoe, and they are associated, invariably, with the explicated words. (Far further discussion, see Chapter 11 11. 1 The comgrlexity of such meanings-or such folk knowledge-is lquite remarkable, m d it is hard to escape the conclusion that the human mind has an in-built capacity to acquire such horrors-or wrgnder~-~f ccomplexity and to organize them along certain paths, for which the mind is perhaps somehow prepared. This conclusion is strengthened by consideration of words for cultural and natural kinds, which will be discussed next. (See again Chapter 11.1

6. Cultural and Natural Kinds: "read"

and Water'

According to many theorists of language and cognition, there are two kinds of concept: one kind is acquired via direct experience of the world, that is, ostensivdy, whereas the other is acquired via language. For example, Russell (1948: 78) wrote: 'Ostensive definitionymay be defined as 'any process by which a person is taught to understand a word otherwise then by the use of other words'. Suppose that, knowy ing no Fremch, you are shipwrecked on the coast oFNormandy: you make your m into a farmhouse, you see breed on the table, and, being famished, you point at it with an inquiring gesture. BE the farmer thereupon says pain, you will condude, st least provisionally, that this is the French for "read', and you will be oonfimed in this view i l the word is not repeated when you point at other kinds of eatables. You will then have learnt the meaning of the w r d by ostensive definition. Having in this way inkroduccd tlrc concept of ostensive definitian, Russell goes on to suggest that there are two different kinds of words in language: ithose which are normally learnt by ostensive definitions and those which are learnt via other words. He exemplifies: Most children Lam the word dog ostensively; some learn in this way the kinds of dogs, colllies, St. Bemards, spaniels, poodles, etc., while others, who Raw little ta do with dogs, may Arst meet with these words in books. No child learns the word gwdradped csstensiwllgf, still less the word animal in the sense in which it includes oysu a t , bee, and beerk ostensivdy, and perhaps ters and limpets. He probably lems m insect, but if so he will mistakenly include spiders until corrected. Names of substances not obviously oolbctdons of imdividuals, such as miilk, bread, W O ~ ,are apt b be Pamt ostensively when tbey denote things fmiliar in every-day life. (Russell 1948: 833

(fl often people do something in a plaoe

because they want this to happen [making dire]

(k) sometimes people d o it because they want something to happen to

something [to some food] sometimes people do it because they don" want to feel something h d [cold] (m)sometims people do it at a time when they cannot see things [at night] &awe they want to see things PlgErt]


H t could be argued, of course, that the formula above articulates not the meaning of the wardfire, but something else-say, English speakersyalk knowledge and folk ideas about fire. In my view, however, this would really

7. Ostension and Concept Acquisition 227

In a similar vein, Burling (1970: 801) distinguishes between what he calls "referential definitions" and "verbal definitions": The conclusion seems cfear: any theory of meaning must provide for two essenmtially different ways by which we can learn and define the meaning of' words. Mother must almost always be learwd in context, while second cowin once removed mulld probably newer be Beamed without some degree of verbal explanation. Water is learned in context, hydrogen aFicmi&, with an explanation, and so on. I would also agree that some concepts are learnt (partiy) on the basis of ostension, and bred, milk, wafer, dog, or bee may indeed be of this kind. Whether any concepts can be acquired pureiy by ostension is a problem which I will discuss shortly. First, however, it should be pointed out that the alternative set up by Russell or Burlling omits the third-and perhaps the most important-souroe of our concepts, that is our innate conceptual apparatus, which is both '"logically and onbgeneticalliy'"rior to either verbal explanations or ostemsion. It seems obvious that "'verbal explanations" must rely on words which can be lleaunt without such explanations. Consequently, if ostension were thc only alternative to verball explanations then all concepts would have to be acquired, ultimately, by ostension. But how can one acquire by ostension conceptual distinctions such as that between someom and somethhg, or between you and R Astonishingly, Russell (though not Burling] does mention you and I, as well as afler and before, among those words which are learnt by ostension. But how colllld one possibly "show" to a child what you or I means? Or how could one s b w to anyone what someone, as opposed to somebhfng, mans? Of course, one could give examples of persons, and examples of things; but how could one show the grounds for the respective generalizations? Ass~ng that universal human ooncepts such as SOMEONE and SOMETHING, YOU m d I, or BEFORE and AFTER are prior to all experience, and to a P P explanation, and provide-like Kant's space and time-a prariori f o m s of experience, let us return to the question of how concepts which are rooted in people" experience (such as Russell's "read'or 'pain') may be actually acquired, and what these concepts may really stand for. Russell" charming vignette notwithstanding, to acquire the adult meaning of either pain or b s e ~ d [whicl~,incidenitaily, do not mean exact$ the same, and which do not have exactly the same range of useI3 chiidre11 (or
"or eiexmpb, Engkish distinguishes lexically between roSls thread rolls) and bread, wherea in French "rol8s" are calbd r'es pefiss pains ("little breads"). This ract (among many othea)~ reflects digerent culinary traditions and dimerent expectations with respect to bread and pain: bread is expected to be sliced (and, cons;equently, so& are conwived as a different cultural kind), whereas pain, which plays a far greater role in French culture than bread does in Anglo cuItwlre, is expected to be crusty on the outside and light and puffy inside, and is therefore less

shipwrecked mariners from distmt lands) must do more than simply observe how these words are used in one type of situation, with respect to one type of object: they must also figure out how to extend their initial use to new situations, and to new, unfamiliar, types of referents. To da that, they must go beyond a mere observation of material ob~ects (various pieces or kinds of itaread), and come up with some (unconscious) hypotheses about the way people think about those objects; they must make the leap from ostensicrru to conceptualization, from objects to construals. They must learn that "tPread9ssomething that people eat; they must also learn that it is something that people, generally speaking, eat every dayeven if they, or their o m families, happen not to eat it every day; that it is something people normally eat in order not to be huangry [unlike cakes, which people normally eat for pleasure); that it comes in the form of large objects which can be shared and which can last for some time (loaves) and that people normally eat it with something else (e.g. butter). They must also learn thalt-unlike noodles or rim-it is something that people eat with their hands, and not with forks or spoons. (Words and expressions such as bead~win~wr, bread and EPrrrre~,or sa,rsdwich provide lingnistic evidence for the psychological reality of these compomnents.) The fact thar bread is made from flour or from something like ffour, and that it is baked, is probably less essential than the k t s mentioned above, but it probably is, none the less, part of the adult concept of 'bread': to understand this coancept, one must know thar the thing in question is made from 'something that comes from plmts, that is, something that grows out of the ground' (and that people do something to the ground for that purpose); and also that, while one doesn't normally eat bread hot, to prepare it one does need 'heats7 For Europeans, 'bread7s such a familiar everyday phenomenon that it is often hard h r them to imagine that this conoept may need any exlplanatiion. M e n one looks at it from a cross-cultural persgectiv~e,however, it becomes obvious that bread eating is a highly culture-specificphmornenon, and that the word bseud stands for a complex cultural kind, whose esserutial characteristics are by no mems easy to identify to c d t u r d outsiders (see Nida lg475. The fact that in many languages ~(e-g. in many Australian and keanic languages) the word for 'bread' is a recent loan from English highlights the "foreign" character of this concept in many cultures. 11 is also likely that although loan words of this kind may have the same range of GLenotata as the English word breadI the concept is in fact somewhat different. In particular, in Australian languages, which tend to contrast "flesh food" and "piant food"",breadY (or, rather, 'burridi') is thought of above
suitable for slicing. [For example, a French bagwetre counts as pain, hut it is not expected to be sbced.) Fhe fact that in English bread is a mass noun (e.g. *two brea&) whereas in French paah cam be used as a count n o w (e.g. d e ~ poi@ x points h [he same direction.

7. O'steasion asad Concept Acquisition 2213 a11 as a kind of "plant food"'. For example, in Maimithiyei bread is commonly referred to not just as burridi but as mi-b~rsidi, where mi is a general classihr for food coming from plants (Ian Green, personal mm'~lhimtion). A s mentioned earlier, however, from an English 8pIkerE8 paint of view the 'plant' origin of bread may well be a less salient part of the concept, aquired relatively late. As a first approximation we can, then, explicate the concept of %read9 aPong the following limes: bread a kind of thing that people eat many ;people eat it every day wlipcri pcoplc don't cat somathing fur a long time they can feel something biad people often eat this kind of thing because of this people eat it like this: they eat it with their hands, not with anything else when they eat it, they eat something else at the same time they don't eat all d it at the same time people make things of this kind because they want people to be able to eat them ORIGIN in some places things grow out of the pound because peoplle did something in these places because they want people to be able ta ealt this kind of thing people do some things to Urese things something happens to some parts of these things because of this people do something with these parts near fire This explication, incomplete and imperfect as it no doubt is, is, I believe, gsychloPogi~d1y more real than a typical dictionary definition, such as, for example, that provided by the Shorter Oxfard Ep~gii$h Dic~iomry l(SIOED 1964): bread-"an article of food prepared by moistening, kneading, m d baking meal or flour, usually with the addition of yeast or leaven"; or the one provided by the MacqwarB Dictionary o f Ausrm~imEnglish (UPIEI]: bread-"a h o d made of flour or meal, milk or water, etc., made Into a dough or batter, with or without yeast or the like, and baked". Contrary to what definitions of this kind imply, one can know what bread means without knowing the meaning d words such as yefist, iearuefl, or batter; but i f one doesn't know that bread is thought of as something that many people eat every day ("our daily bread"], and as something &at they eat, roughly speaking, in order not to be hungry (rather than h a pleasure, like cake]; or'that it is made from something that, roughly speaking, "comes from plants" (so that, for example, a "meat loafY3s not a kind of bread), then they do not know what b r e d means. The knowledge encapsulated in this concept goes far beyond ostension, and in this sense "read" is far from a purely ostensive concept. It embodies generalizations which are usually not made explicitly, verba1Ey, but which none the less can be made verbally, and which sometimes have to be made verbally. For example, if a small child refers to a cake or b a meat loaf as bread, he or she is likeiy to be corrected: this is not bread, this is cake; or: this is not bread, this is a meat loaf. Somlehow or other the child must build the generalization that things with meat in them, or sweet things that one eats for pleasure and which one doesn't spread with anything, are not bread. I am not saying that generallizations of this kind are given to the child verbally, or that they are made by the child cnnsciausly, but thal i F lhcy arc not madc at all, the concept has not been Ileamt. Similarly, how can one learn the concept of %ater'(CBudEng% example) by mere ostension? Water is something that one can drink, but so is milk or orange juice. Water is 'kolourless", "see-through"'; but so is vodka, and various other liquids that a child is not alllowed to drink. Water is also the stlzff one can go into in some places, for example, in a place called "the sea"; and this stuff one is not allowed to drink either. Water is also the stuff one can wash oneself with; but the idea of 'wash'preswpposes, it seems, the knowledge of the concept 'water'. (One can %ash1 oneself only with water or with 'someat is lnslnally assumed, 'water'is not a universal human iwencept for which every language has a word. For exsample, the closest Japanese word, mim, does not mean the same as water and refers exclusively to cold water (see Sluzuki 1978). One cannot say, for example, *atmi whereas afmi miruku, 'hat milk', i5 fully aoceptable communication). For 'hot waterqapanese has h an honorific prefix, oyu). the Engiish water can be either cold or a from a tap, but if it comes from the sky, it is normally d twin* not water (unlike the Pitjaritjatjara word k~pi, which applies to the stuff falling sometimes Yrom the skyy;Coddard 1992~). None in can be called drops of water. see Clearly, there is much more to this concept than on; and the most important oomponents those that one learns first. fully acquired the concept of 'water' one doesn't rt from all other substances. If scientists tell us like water and tastes like water, but it is not e to believe them. We think: of water as a particular KIND of thing, a kind which we may not always be able to recognize

230 General Issues

7. Osfemsionand Concept Acquisitio~ 23 1

but which one can 'put into one" body, through the mouth' (La. drink it), and which one can "ee in various places not because people did something in those pIaces"rilrers, creeks, lakes, sea, m d so on). These two we perhaps the most cruclal components of the (adult) everyday concept of 'water" They cannot be acquired by ostension alom. There is more to the a~quisition of concepts like 'bread' and %water1 than RusselB's little fantasy about art Englishman "shipwrecked on the coast of Normandy" sallows for. My conclusion is this: the meaning of a "c~ncrete"wword like $read Jslr water is indeed learnt "'in context", but it is built, in the child's mind, gradually, and contains certain tacit knowledge which may not be verbal but which is verbalizable. The Japanese child learns '7n context" that mizu is old, but when she hies to use it with respect to boiling water she will be corrected (presumably, verbally corrected); m d even if such an error, and such a correction, never occurs, tlne child must somehow make a mental note that a hot liquid which otherwise is quite like mizu is called y~ and is never called mizu, and therefore that mizu must be cold. I am not saying thal children make this mental nome consciously or verballly; but they must make it somehow if they are to learn the p r o p r range of use of the word mizu. The same applies to names of animals, such as Russell5 dog. How could a child learn this concept by ostension alone? Surely, sooner or later a moment will come when the child learns to apply the word dog to an animal that looks different from all those that were called dog in her presence. The word dog doesn? mean simplly 'an animal that looks like those that we have heard people refer to as dog3'. d mall child may call a fox, or ewn a cat, dog (Clark and Clark 1977). To develop a full a d d t concept of Vog' one has to make in one's mind certain generdimtions, such as that t h y bark, that they growl, that they can bite, that they live with people or near people, that they can do diflerent things that people want them to do, and so on (see Wierzbicka 1985: 369-70; see also Tyler 1978). Again, the ability to recognize dogs is not a necessary Eonsequence of knowing the concept 'dog" An adult person may mistake a woif, or a dingo, for a dog, and this doesn't show that they don't know what dog means. To know this one has to have aocess to a number of generalhatiom about dogs (such as those mentioned above), not to a mental image ensuring that one @an always tdU dogs apart Biom other animals. How exactly these generalizations are built, and how they are stored in the mind, we don't know; but they are oertailrly verbaiizable. (Cf. Chapter 11.) Russell was convinoed that quadrtrped is learnt via a verbal explanation: whereas dog, beet&, .or ant is learnt via osbnsion. He was less confident, however, about insecf and seemed in Tact to r m o p b e , implicitly, that bsecl cztmnot be learnt without a verbal explanation or at least a verbal correction.

But is beetle redly very different in this respect from insect? Or, for that matter, from am? In many AustraPian Aboriginal languages there is no word for mat, because, for examplle, "edible ants" are conoeptuallized as something quite different from 'red ants' or from poison^^^ anltsl'. In fact, different kinds of what speakers of English would describe as h n t s h r e not even assigned to the same supercaitegory. (See Hale er aL forthcoming.) This suggests that each "species word" of this kind is learnt in tcrms of some absbacm gencralizalions-not unlike tho% involved in q~rsdruped~ ifiscct, or second cowin once removed. Having said this, I hasten to add: I am not saying that there is no difF on the one hand ference between quadruped or second cclusi~rO R C ~ removed and dog, oras, bread, or wafer on the other. But the difference is not what Russdl 01 Bwrjing say it is. Concepts such as 'quadrupedkr 'second cousin once removed' are probably learnt by verbal explanation alone, so that ostension plays no role in their acquisition, whereas ostension clearly does play a role in the acquisition of concepts such as 'dog', 'ant" 'bread', or 'waterJ. However, this doesn? mean that the latter kind of concept can be acquired by osbnsion alone. Nor does it mean that they can be given a purely ostensive definition, along the lines of "an ant is a kind of creature that people call ANT" or ""bead is a kind of stuff that people call BREAD"". Noit every language has a word for ctmn-ead'or for 'ant" and one could never teach the speaker d a language which doesn? have such words what they mean simply by showing them some appropriate specimens. Specimens are of course useful, but the learners have also to grasp how far they can go in applying the words in question to things which look different-r to things which look similar but are differently handled, differently interacted with. Abstract generalizations have to be built, and allthough the learner QesnY necessarily bluiad them, and store them, In the fonn of verbal definitions, this is the only way, I believe, that testable hypotheses about their content can be formulated. How dse? M e n Burling (1970: 80) says that words such as wakr are ''learned in contextY'(raitherthan with the help of other wards) he is no doubt right. h m . saying that the meming of such words can be But this is different B stated without a verbal definition. The quote adduced at h e beginning of this section sserns to suggest that, in Burling's view, wards learned in context do slot require verbal dehitions-and this is the point which I dispute.

232 General Iasres

7. Oxtension and Concepl Acquisitio~ 233

So far, I ham rocused on the semanljc complexity of words of different kinds. But, obviously, people don" sspeak just in nrrords: they speak in sentences; and the semantic complexity of sentences depends not only on the complexity of the words of which they are composed but also on the mechanics of the composition itsclr. Let us consider, for example, a simple sentence such as I ale m appk. En broad outline, its semantic structure can be represented as follows: [at some time before now], 1did something [EAT] to something [APPLE] something happened to this something EAPPLE] because of this To apply this fomula to a sentence, we need to 'plug in', somehow, the semantic infomation encapsulated in h e words eat and appEe. For ear, we could propose (as a first approximation] the f018wing scenario: Someone ate something. = (a) someone did something to something (b) after this, something was inside a part of this person [MOUTW (c) this person did something to this thing with this part ($3 something Inappened to this thing because of this (e) after this, parts of this thing were inside another past of this person [STOMACIfl IC;rP if people don't t o this they cannot live ( g ) if someone doesn't do it for a long time this person feels something bad The crucial components which ddisfimnguish 'eating'from Urinrking' are (rc] ;and (4: even though food doesn't have to be nieeessarily chewed or bitten, when one 'eats' it (as opposed to "rimking') something happens to it ;in the mouth, and it happens Irecause one does something to it with one's mouth. (For a definition of moluth, see Section 4.) As for appJle(sJ, this could be defined, in essence, as 'a kind of thing that people eat', and in simpler syntax, 'a kind of thing-people EAT this kind of thimg', a fomula whose intelrpretdion depends again on the explication of sat. But a full meaning of apple contains of course more than a basic categorization in terns of edibility. Trying to spell out this meaning (that is, the folk conoept) fully, I hund that I had to posit numerous components, some of great semantic complexity (see Wierzbicka 1985: 302-3). But even if we ignored the enormous semantic complexity of a concept Bike "apple', and if we d e b e d this word simply as % kind of thing that p a For an earlier attempt to state the meaning oreat, sea Wierzbicka ((19810: 901.

p U e eat" or even as 'a kind of thing that people call APPLES" we would still have to conclude that the meaning of the simple sentence I ate a m apple cannot bc represented in tcrms of onle global paraphrase. It can be represented in terns of semantic primitives, but only on a step-by-step basis. I conjecture7therefore, that-to some extent at least-this is how the human mind operates when it confronts the stupendous task of semantic interpretation in general; or at least, that it is capable of doing so when it is needed.

The semantic structure of an ordinary human sentence is about as simple and as "shallow" as the structure of a galaxy or the structure of an atom. Looking into the meaning of a single word, let alone a single sentence, can give one the same feeling of dizziness that can come from thinking about the distances between galaxies or about the irnpenetraltPPe empty spaces hidden in a single atom. The experience can be disconcerting, and perhaps it is not surprising that many theorists of language and cognition prefer to take the view that meanings can't be analysed-as W. Lyons (8981:73-41 put it, 'Tor theoretically interesting reasons"'. Brat no reasons, not even '"theoretically interesting'' ones, can absolve us from the effort of trying to explore the meanings of words to find out what unconscious principles determine the boundaries of their use. We have to try to pin d o m the elusive and culture-specific corhgurations of elements encapsulated in everyday concepts, and to face the formidable complexity d meanings which ordinary people appear to juggle effortlessly in every&y discourse. If we don't tackle this copnlolexity we shall also fail to achieve some of our basic professional tasks, such as laying the groundwork for a more efictive lexicography, developing tools to revitalize language teaching, or promoting cross-culltural understanding via a non-ethnocentric description ofcuBturaP variation. We shall also throw away our chance of exploring and contemplating the dazzling beauty of the universe of meaning.


Against "Against Definitions"

1. LioguisXie Meaning
t try to review the m o m o u s literature on definisent m y a m concluiorions on the subject emergjag from definitions, pursued over nearly three decades. A "definitionya-in the sense which is relevant to linguistics-is an expreso tor speak, the meaning of a word by articulating it to its components. [Gf. the quote from Locke 1690r'8959:33-4 in Chapter

Since to define a word is to deC0mplo~e its meaning into its constituent s, only comp1ex concepts are susmptible of definition. Although keys""simple ideas" do not correspond exactly to semantic primitives as aceived here (see Wierrbich lI4lglCD), his argument agpEes equally to both:

H saw that the names cEfsimpiFpidem, md those only, are hcaThe eeiilwn whereof is this, That the several terns of a defiether by no means represent an idea a dehition, which is properly ~ 0 t h by several others not signifying each as have no place. (1690j1959: 3 4

word, then, I mean, essentially, what h c k e meant: ing" the meaning of a definable (i.e, semantically carnplex) w r d in of indefinable I[i.e. semmtically simple) ones.

2. Definitions as a Tool h s Cross-cultural Research

Why do we need definitions at all? One of the possible answers to this questian is that we need them as ;a took lfais understanding other cultures [and for making ourselves understood). Wards are a society's most basic cultural artefacts, and-properly unrderstood-they provide the best key to a culture's values and assumptions. But to avoid misinterpretation, definitions are needed thait are free of ethnocentric bias; that is, definitions couched in terns of universal, culture-free, primitive concepts. The need for such definitions is still nat wideiy recognized. To quote one

8. Against 'Againsf Dejnifiiom" 239

anthropologist: "It is perhaps a matter for some wonder . . that in so many regards . . . social anthmpoilogists should still1 so commonly write as though in their own language, and in the technical terms of their profession, they already possessed an ideal language" meedham 11972: 222-3). I do not claim that the Natural Semantic Metalanguage is an '"deal language" in the sense of being the final answer to the search for lexically embodied concxptual universals. It is an approximation, to be improved by further trial and error. But it is better to have a tentative and imperfect set of indefinables than none at all. Using the proposed set we can clearly define even those conlciepts which are widely regarded as "unique", that is, absolutely culture-spec$c and thoroughly "untransIatable". I will illustrate this claim here with just one example: that of the Japanese concept of 'amae' (for other examples, and more detailed discussion, see Wierzbicka 1991b, 1 9 9 2 ~ and forthcoming el. According to Doi, amae is "a peculiarly Japanese emotion", although it has "universal relevance" (1981: 169). It is "a thread that runs through all the various activities of Japanese society" (1981: 261, represents "the true essence of Japanese psychology", and is "an important key to mderstmding the psychoIogica1 diflerences between Japan and Western countries" (1974 310). Doi explains that '"mae is the noun form of amaeru, an intransitive verb which m a n s 'to depend and presume upan another's benevolence' " (1974: 307). It indicates "helplessness and the desire to be loved" '19811: 221. The adjective amai means 'sweet', both with reference to taste and with reference to human relations: "if f is said to be ramrrs' to B, it means that he alllows JfP to amamu, i.e., to behave self-indulgently, presuming on some special relationship that exists between the two'"18983: 29). Antaeru can also be defined "'by a comlbination of words such as 'wish to be loved' and 'dependency needs' " @oi 1974: 309). The Japanese dictionary Doigenkai defines amae as "to lean on a person's good williY"(qooted in Dai 1981: 72) or "to depend on another's aIfection'"l6~. Other dictionary glosses include "to act lovingly towards (as a much fondled child towards its parents), "to presume upon", "'to take advmtage oEll (Brinkiey'x 19613); "ta behave Pike au spoilt child'" "be coquettish", "trespass on'" ""betrave in ;a caressing manner; towards a man"; "to speak in a coquettish bne", "encroach on bnels kindness, good nature, etc.]" (Takenobu 1918); "'presume on another's lave", "be coquettish"', "coax" (Kenkyyus/~a's11954); and so on. Morsbach and Tyler (1 9861, who analysed fifteen passages from Japanese literature referring to amae, used in their translations of these passages the following English glosses: ""take advantage of", "play baby", "'make up to [someone] and get their sympathy"",koax", "act spoilt"', and so on. On the basis of these and other similar clues, we can explicate the concept of m a e as follows:


X thinks something Pike this:

when Y thinks about me, Y ifeels something good Ywants to do good things for me (4 Ycan do good things for me (e) when I am near Y nothing bad can happen to me I don't have to do anything because of this Cg)i I want to be near Y (A) X feels something good because of this (b)

u C P f 3

Doi emphasizes that amae presupposes conscious awareness. The subcomponent (a) 'X thinks something like this . . .?eflects this. The presumption of a spcial relationship is reflected in the component (b) "hen Y thinks about me, Y feels something good'. The implication of selfindulgence is rrooted in the emotiomral security of someone who knows that he or she is loved: "it is an emotion that takes the other person's love for granted"' (Doi 1981: 168). This is accounted for by the combination of camponents (b) 'when F thinks about me, Y feels something good', (c] < Y wants to do good things for me', ('4'Y can do good things for me', and (e) "hen I am near 16 nothing bad can happen to me" The component (f) '1 don't have to do anything because of this' reflects the passive attitude of an ansuejunior, who does not have to earn the mother-figure's goodwiU1 and protection by any spacial actions. The component Cg) ' Iwant to be near Y' r~flects CPoi's ((1981:74) idea that the baby in an "amae" relationship to the mother 'komes to feel the mother as something indispensable to itsdf" and that "it is the craving for close contact thus developed that constitutes . . . amaeJ'. Thus even "'euniqwe"",thoroughIy "untranslatable" words such as m a e can be awurately and inteliilgibly d e h e d in terns of universal semantic primitives.

3. The Concept of a S e m a n t i ~ Invariant

A linguistic definition is a scientific hypothesis about the concept encoded

iTY a given word [see Robinson 91950: 41). Like other scientific hypotheses,
it cannot be proved to be right, but it can be tested and proved wrong-in which case it is discarded, or revised and tested again. While the concept is not ac~essibleto direct observation, it is manifested in a word's use. Amdingly, a definition can be tested against a word's range of use; this range of use may be very broad, but it has its boundaries, which are detcrh e d l by the different components of the concept. The components of the conciept determine which aspects of a word's use

are variable and which are invariable. It is the purpose of a definition capture the invariable aspects off a word's use, that is, its semantic kvari ant. Om might ask at this point: How do we know that a word real a concept with some invariable components, that is, that it really a semantic invariant? The answer is that we don" know it, but proved to be a fruitful working hypothesis. The alternative hypo the meaning of words is claangeable and "fuzzy", and cannot in rigorous definitions] is sterile and can hardly provide an eEe large-scde iexicographi~ research. The goal of capturing a word's semantic invariant by investigating its , O m years ago, by Socrates, in Platoys range of use was set out more than 2 dialogue Laches, devoted to a search for the definition of the Greek cancepa 'andreia"usuaQly rendered i n English as clsrwrage):

definition doesn't say what danger, fear, and d@ccrFby have in com(nor what is common to mental ssrengrh and rnaral strmgtlC3). phic devices such as 'br", "often", 'hsudEy",oor 'ktc." are unmiss af the Iaxiwgrapher's failure to find a semantic invariant, as iferation of different senses and subsenses in a dictionary entry. words don't h v e any meaning in isolation, but only in sentemces, g for t k semantic invariant of a word we have to s t a t (contrary lexicographic practice) with some syntactic frame, and try to he sentence serving as our point of departure in terms of words slier as indefinables. For example:
X is courageous. = X can do very good things when other people can3 because when other people think something Bike this: I don't want bad tbings to happen to me ' A thinks something l i h this: it is good if I do this it is bad if I don't do it I want GQ do it h a u r n of this this is good

meant to ask you not only about the wurag of the heavily-amdl so[diers, but about the courage of cavalry and every other style of soldier; md not only who are couciagleous in war but who are courageous in perils by sea, md who in disease, or in poverty, or again in politics, are courageous; and mot only who are couragzous against pain or fear, but mighty to conbnd against desirm and pleasures, either h e d in their rank or turning upon the enemy. There is this sort of courage-is there not, Laches? LACHES.Certainly, Socrates. . . . s m w m What i that o o m o n quality, which is the same in all these cases, and which is called courage? ffPlallo 19710: 116-17) Dictiomary descriptions usually do not succeed in capturing a word's semantic invariant. Often, they don't even try to provide a definition, but instead offer more or less random lists of quasi-synonyms, as in Web~ter's New School and Ojfice Dictionary (1965) entry for tlne ward courageow: "courageous-brave, bold"". Entries of this kind inevitably lead to vicious circles. For example, to cantinue with the same dictionary: corurageoz~s brave bold intrepid fearless brave, bold bold, courageous, intrepid courageous, venturesome bold, fearless intrepid

@or justification and discussion, see Wieezbicka 19192a.)1

4. Determinacy of Meaning
search B;or a word's semantic invariant presupposes a belief' in the detering. E n recent decades, this belief has often been questioned, and philosophers alike. Anyone who has engaged in Iexicoarch can: readily sympathize with such scepticism: constructing which matches a word's entire range of use is a huge task m r e work ahan most are prepared to put in; and the temper the first two minutes or the &st two hours may well BUG the fact that one Baas tried, and failed, to come up kctory definition in two minutes, or even two hours, es the conclusion that meaning is indeterminate. (See e.g. out the meaning of pafrtt, discussed in Section 9 below.) unrealistic expectations abaut the amount of time and y for success, several other factors have dearly contributed e widespread soepticism about the feasibility of definitions, notably the wing: (1) the lack of a set of imdehables; (23 ithe Pack of a coherent envy; (3) confusion of lexical meaning with illocutionary and such as metaphor, irony, sarcasm, hyperbole, and so on; seductive power of the theory of 'cfarnilly resemblances". The first of

If a rare dictionary does manage to define courage or coiurageasrs without a vicious circle, it is likely to pay for it by n d even attempting G o capture the invariant. For example, the h n g m n Dictionary of the Ersglkh Lmguage H(LDOT8L 1984) defines courage as follows: courage - mentaP or moral strength to confront and withstand danger, fear, or difficulty

242 Lexical Senamtics these four points has already been discussed. Of the other three, only fa@ tor(3) hasn't yet been discussed at allp but while the question of polysemy has already been discussed in Chapter 6, and that of 'Tanwhly resemblances" in Chapter 4, both these topics need some elaboration.

8. Against "Against Definzlions"


5. Problems of Polysemy
As Zgusta justly points out, "the lexicographer will1 do well to reckon always with polysemy: this will cause him to undertake very deep analyses of the wordsheanings" "971: 6163. Polysemy can be obvious and unmistakable. For example, the Gnglish noun sprig has at least four different meanings, defined by the Qxfod Paperback Dictionary (OPD 1979) as follows:
1, the act of springing, a jump 2. a device (usually of bent or coiled metal) that reverts to the original position

guide-lines ar generid principles, and this applies to ambitious and innovative dictionaries as well as to commercial hack jobs. (For rare exceptions, see Neltuk $c Zolkovskij 1984; Mel%uk eit a[. 1984, 1988, 1992; or Apesjan and Rozenman 1979; for a critique oh one ambitious and innovative dictionary, the Songman Dictionary of Cmfemporary English (LDQGE 19781, see Wierzbicka 1987a: 4-7.) For example, the American Herirarge D k ~ i o m r yo f the English Language (AHDOTEL 1973) postulates the following three meanings for $
1, in the event that: If I were to ga, I would be late. 2, granting that: Even if that's true, what should we do? 3. m condition that: She will sing only ifshe is paid.

after being compressed or tightened or strelchad . .

3. e place where water or oil comes up natlurally from the ground 4. the season in which vegetation begins to appear . . .

. ..

agree that the meaning of v i n sentence 1 is indeed different from the eaning of if in sentences 2 and 3 [see Chapter 21. But the difference etwelen the supposed senses 2 and 3 am be attributed to different conexts, not to the word $ itself (note, in particular, the contrast between 'even i f " h 2 and "only ik'yin 3). The following rephrasings of the same ententes (without m'e'ir'en"and " ~ n l y ' ~ show ) better L t n e similarity in their eanings:

It is obvious that if we tried to cover these four senses with one definition we woufd get nowbere; the notion of "semantic invariant" makes sense oniy with respect to specific senses of a word, not with respect to a mechanical sum of a11 its diiFferent uses, and to be able to construct adequate d e G tions, we have to .separate the diflerent senses of a polysemous word. Here again we must proceed by trial and error, assuming atways, to begin with, that there is only one meaning, mnslructing a tentative defirrition, checking it against a word's possible range of use, then, if newssary, powiting a second meaning, and so on. A great deal of hard work m y be required, but this fact by itself has no bearing on the issue of discreteness of meaning. Often the meanings to sort aut will be much closer to one another than the four senses of sprhg. In sorting out ciosely related senses it is important that the differences beltween what we posit as two separate meanings be linked with the ward itself and not simply with different contexts in which this word is used. For example, it would not be justified to posit polysemy for the noun itwe along the lines of (1) ramantic love, (2) parental love, (3) brotherly love, and SO on, be~auseall these supposedly different kinds of "love" cam be seen as sharing a semantic invariant, and as differing only in ways implid by the modifying adjective. Generally speaking, dictionaries Itend to posit polysemy on a truly massive scale; but it is usuatlly posited on an ad Roc basis, without any clear

(I> If [we suppose thaq that's true [we have to decide: ] What should we
do? (23 If she is paid she will sing. [If she is not paid she will not.] ent contexts may hlelp clarify which meaning the
t they are not the source of the four different inter-

tions, which are inherent in the word spring itself. consider the three distinct meanings postullaked by the same diction1. to olffer advice to; to counsel

2. to reloommend; suggest 3. to inform; notify

In fact, there is no evidence of any semantic difference between the alleged meanings 1 and 2 (while 3 is indeed different). It is true that advfse can m u r in two different syntactic frames, with either a person or an abstract noun as its object Qe.g. The doctor advised B S E E $0have compkte rwr, The doctor a$lvised comph?te resf), and different syntactic frames can be associated with differences in the semantic structure. But the verb itself does have an invariant meaning, evident in both these frames (associated with the alleged meanings I and 2). (For justification and detailed discussion, see Wierzbicka 1987a.) Finally, consider the verb fee!, for which the same dictionary, AHDOTEL (19731, posits several different meanings, including the following ones:

244 Lexical Semantfc8

lb. to peuoeive as a l o c a l d physical sensationr: feel a s h q pain. lc. to perceive as a nonEocalized physical sensation: feel the cold. 3a. to experience (an emotion): I felt my interest rising.

8. Agsrfnst "Against Dejni'tiom" 2 4 2 4 5

However, the distindion between Pb and l c implies that a senten= Sheyeif pain is ambiguous, and that it means two diFFemt things, ing on whether the pain is nocallized or non-localized. F u r t h e m o ~ the , tinctiom 'taetwmn the fed of sensation and the feel of emotion implies sentences such as How me youfeeling?, Ifeelg~lod,or IjreeI better are all ambiguous. But in f x t , the two supposedly different meanings can easily conjoined, as genuinely different meanings cannot:
I felt cold and miserable. +He ordercd several books and an inquiry inlo the matter.

refuge",but OPD (rightly) refrains h m mentioning this use as a seprate sense of the word rejuge. If metaphorical uses and other similar rhetorical devices me not distini s k d from lexical polysemy, then meanings may indeed appear to be determinate and not amenable to precise definitions. But this is not the dtY'of the meanings themselves.

7. Family Resemblances
key lfactor militating against large-%ale study of the lexicon until rwiently st the absence of a suitable methodology but also the widespread t such a methodology could not be devised. ParticwlarEy d m a g s respect was Wittgenstein" doctrine of "family resemblances", gained extraordinary popularity (see Wierzbicka 1990b; also T f it is a s s u d that meanings have no clear boundaries and that they are utually related by vague and elusive family resemblances, then it is natcondude that no precise definitions are possible. But without such ons meanings can be neither stated nor earnpared. As a result, all investigation of the lexicon comes to a halt. e doctrine off family resemblances never had any empirjcal basis. It the speculative idea of a philosopher-to be sure, a philosopher of ast unparalelIed genius and charisma, but none the less one who was immune to error. I believe that lexicographic research of reant years rrs proved Wittgenstein wrong on this particular point. As pointed out by van Brakel(199l: 61, in reicient literature on the nature if meaning and categorization, the notion of family resemblances is usually ked with the notion of prototypes ("it is almost canonical to refer to genstein and Rosch in the same breath when prototype theory is intnsdm).Van Buakel argues against the prevailing current use of the cona p t of prototype, and against the whole '"rototype approach'"t~~language and cognition, and tries to separate prototypes ;thorn family resemblances, of which he does approve. But while I agree with much d what be says about the prototype oach to meaning (see Wierzbika 1990b and Chapter 41, I believe that doctrine of family rwmbllances is equalPy at fault. Van Brake1 writes: With the notion of "family resenrbllane9Tittgenskb wanted to illlustrate that t h e is no uniform sea of attributes which constitutes the meaning of a linguistic expression. Them are omlly similarities between diflerent uses oF an expression. Whatever rules are proposed, they only conventiondlly define some asplecb of the Language use. It i not possible to grasp, or theoretically pin down, the meaning of axpressions or the competence of a speaker-hearer. (1991: 6)

But while dictionaries often posit a great deal of unjust% they also Frequently fail to recognize polysemy which is really striking illustration see, for example, the discwssion of the Chapter 9.1 Zgusta is off course right when he states that 'Yaken on the whole, pollysemy will always prove a hard riddle for the lexicographer" "971: 73). We do not have to assume, however, that this riddle cannot be salved. I a system of semantic analysis based on a finite set of indefinables the principle of reductive paraphrase, meanings re-emerge as d determinate entities, and the '"riddle of palysemy'keases to seem ansca uble. 6. Lexical Meaning and Pllocutlonary RhctoricaE Devices A word can be adequately defined only if its literal meaning is distiaguishedl from its metaphorical use, ironic use, playful use, euphemistic use, m d other similar uses. Dktionaries frequently fail in this respect, and, far example, treat a word's metaphorical use as a separate lexical meanhg. For example, OPD lists (among others) the falowing two meanings the English wordfr{en$: (1) a person with whom one is on terms ofmut affection independently of sexual or family love; (2) a helpful thing or ity, e.g. darkness was our friend. In fact, this second use can be satisfac rily a s o m t e d for i n terns oF the first on the basis of a general princi which allows us to use words of different kinds in the sense "ike x' (where kx' stands for the word's literal meaning). Thus, i n the example above, darkness was not a TriendVn s o m separate Beximl sense of the word friend, ibu it was 'like a friend'. Since the 'like x'device is productive in English [as i is in many other languages), them is af course no need to mention it i n individual dictionary entries. One can also say, for example, that 'Varlmess was

246 Lexical Semmtics

But the proposition that ""I is not possible to grasp, or theoretically pin down, the meaning of expressions" is refuted by hundreds of definitions which have been worked out by semanticists in the last two decades (e.g. Apresjan 1974S1992; Md'Cok and kolkovskij 1984; MdTuk el a/. 1984, 1988, 1992; Wierzbicka 1985, 89187~~ 1992;0,e; Goddard 1991a, 1992a,E995; or papers included in Wierzbicka E99Dc].L Once the different senses o f 8 polysemous word (or expression) have been sorted out, ewry meaning can bc piillncnll down, alld Ihe scn~al~mhc lbrunwla rcprcsenting it cat1 be tested against a wide range of examples and hypothetical counter-examples. Van Brakel" convicltion that every anallysis of the meaning of a word that one may come up with "gives no more than one way of looking at it" ii not borne out by the experience of methodical lexicographic work based on a systematic semantic theory. Van Blrakel states: "As I hope my exampla have shown, the meanings of words have a vagueness and flexibility that resists ultimate rational reconstructionr' (1991: 16). But there is no necessary conflict between precision and vagueness. As I have tried to show in my study of English '%approximative expressions", even a vague meaning such as kataud"e.g. around twenty), 'about' (8.g. abf~~tjifteen] or 'raugihly3 [e.g. mugFEEy half) can be captured in a precise semantic formula: [see Wierzbicka 1 9 8 6 ~ l991aS. ~ Van Brakel's examples, intended to show that the meanings of words "resist rationail reconstnuction", are worth considering because they illustrate some of the fallacies (as I believe) on which the belief in the indeterminacy of meaning is based. Thus, according to Karl Heider (19701 the Dani people of the Mew Guinea Highlands have more than seventy words referring to sweet pa&toes, which is not surprising, given that "sweet potatoes are extremely important in the Dani culture" and that " m r e than half of all conversations are directly or indirectly about sweet potatoes" (van Brakel 1991: 81, Nome the less, "After having spent some time trying to find out which sweet potatoes were called lay what name, Heider gave up, concludimg that Whough there are mare than seventy terns for sweet potatoes, they are
Important work on the structure of the lexicon has or comse been done 'by many other semanticisls, e.g, by David Cruse (19861, I t . M. W. %%on [IPSZ), Charles Fillmore (11971, 1977), Dirk Geeraerts (1993), Jeffrey Gmber (196%, Way JackendotT (1983, 1990I), Adrienrne h h r e r (1974, 19831, John Lyons (1977), Eugene Nida n(19751, Leonard T a h y (19&5),IeI Verschueeen (19851, and Uriel Weinmich (19803. To my knowledge, however, none .of t h e authors bas attempted to test their ideas, original and fruitful as they may be, in large-scale lexicographic studies, involving hundreds of lexical items and hundreds oh definuitions. (Qlue outstanding exception is the work of Russian semanticists, in particullar Jwrij Apresjm and lgoa Mel'Euk; me Wierzbicka 1976d319&6%,i,c.)i This relative lack of intevest r im Bexicmgrapluy as a testing-ground For semantic theories is in, nuarbd contrast to the attitude of the two great semandicisns of the seventeenth mmbury, Leibniz and Wilkias (see Leibnss Table de dbfinitions' 1170411903,and John Wikins 1668; For discussion, see Wiembicka E975, 1994JT;D o l d 19923.


970: 33). He also noted that the different types of sweet potatoes have different 1) there are many diflerent names for not seem to be linked with different , these names are used ""with consid-

us, bcmuse the dirferent names may s other than purely referential ones. example, in Russian the words oseI (Yonkey" and izok H('donkey7are linked with different types of donkey, and yet the semantic dilfferemce ords is stable, and can be captured with full1 precision see Apresjan 1974k1!$%)I. (Roughly stion as a symbol of stupidity and willingness to work hard without clusion, therefore, that the Russian words osel and are used with considerable imprecision would be false; and in all probty what applies to these two words applies aiso to the Dani words for t it is on Ygllse conclusions of this kind that the docy of meaning rests. sider also van Brakd's second example, that is, the area of English used for describing wine. He writes: r way to look at the incoosistenq of the way m y off the Dani words far o h t o e s are used-not the way to look at it-is to cantemplate that also in Pa such incoosistemies may aocur. For example, vocabullary for the descripn of the taste and odour o f winles does not seem to have any fixed meaning, even experts, but seems to function primarily in constructing a vague tone off dgeableness and concern about wine among a group of speakers [ k h r e r Such use o f language is what Nalinawski called pharic r o m u n i ~ n "'an: oes not fmctiron here as, a means of transmission off thought. . Each uttering the direct aim off binding hearer t o speaker by a tie of some senliment or other." Hence, in such a case it doesn't make sense t o look for icular cognitive model that underlies the speaker's specific thoughts aod utter-

Wan Brakel appears to assume that an expression whose function is ainly "phatic" cannot have a specific meaning at the same time. But in ct, although sentences such as Lovely day, isn't it? Nice day, isn't it? Beautifid day, isn? tit? a n he said to have the same phatic fkanctlon, this doesn't mean that in this cclmltext there are no semantic differences between the words bvely, nice,

248 Lexical Semanfics and beaub~$l, or that these words cannot be assigned co The same applies to adjectives used to describe wine. Consider, for example, the following dinner-party conversation a wine:
-This l[Yuim]i s soft m d semsuous-quite an impravewent over the 67s, which unstylish and flabby. -Yes, it's sari, but 1 would say lhal it's grucefu'ul rall~erlhan scrrsuous.

sentence in which it is used, but overall meaming of the sentence s also on other elements and aspects of the scnlence, including varr a n g of semantic stratagems used in talking about wine is quite

It may well be that the main purpose of remarks of this kin or impress other people, allowing the speaker ta show off his kn experience, and expertise" (ibid.), but from this it doesn't word5 themselves have '"ittle or no meaning"', as Lehmr sa assume" "bid.). What Is presupposed in such a view is that words have meaning when they are used to inform, not when they are used, impress or amuse. But one can &o try to impress people with lies, m d ~ o u l d n "be fies if they didn't mefin samething. Words such as soft, sensuow, ar graceful may or may n with some objective differences between xarious .wines (as the plasams, and b ~ e l y may or may not be associated with objective between the people or things to which they axe applied), but t mean that in certain contexts these wards have no meaming. The meaning of the word so$ as applied to wine a n be explicated as foilows: This wine is soft. = when this wine is in a person's mouth this person f e k something good this persan m say then something like this: if I was touching something soft I could feet something like this It is not difficult to see bow this analysis could be applied to other "wine wards" such as S S X E O O veEvesy, ~~, or silky: This wine is smooth (velvety, silky). = when this wine is in a person's mouth this penon feels something good this person can say then something like this: if I was tauchhg something s m o t h (or: some velvet, some silk) H could feel something like this M e n applied to wine, words such as soit, smooth, velvety, or silky ar kind of simile, and they make sense in the way similes or metaphors ma

This wine is graceful. = when this wine is in a person's mouth a person feels something good a person can say then something like this: if I saw someone moving gracefully I could feel something like this This wine is sensuous. = when this wine is in a person" s o u t h this person feels something good this person can say then something like this: if I felt something in my whole body because of one sense (or: because of one part of the body] I could fed something like this Explicatians of this kind may be very different from traditional definitions put forward by either philosophers or lexicographers, but they do ofler an alrxurate representation of meaning.

8. Dictionary Definitions
There is a widespread view that no rnatltes what reservations a smanticist, or a philosopher, or a cognitive psychologist may have with respect to dictianasy definitions, these definitions are basically all right-because dictionaries are sold in minillions of copies and herefore they must be useful. For emmple, speaking of the ubiquitous circularity of dictionary definitians, Feixer writes: The dietiornary for an ordinary language, such as English-Websrer5 Mew WorSd DidEomary (1988), fat example-appears to succeed in providing useful definitions

8. Against '%gainsf Defi~ri~ions"2251

for the terns that it contains in spite of resorting to definitional circularity. IS the is n problem here. therefore, then it needs to be made apparent. haazose them see b bc no prollbrlcrns in practice with dictionary delirultjorms. (1998: 51) This faith in dictionary definitions is based on an illlusion. As pol by Sledd (11972)1, the users of a dictionary cannot know what is them, and how much information they could extract From a diction dictionary definitions were as good as they can be (rather Itlaan as b d they frequently are; see Wierzbicka 11992fi. "The average man and the aver age reviewer m n a t demand the. best in a big di:tianary, because they have no idea what the best might be: and even if they did demand the &st, the businessmen who run commercial publishing houses would not give it to them unless they saw a direct relation between qudilty and pofits1"I(Sedd 1972: 1136; quoted in Landau 1984: 112). There are no reasons to thinnlk, therefore, that commemial success can provide an adequate measure of a dictionary's usefulness. Among many people who have never practised lexicography tbemelves and who have (partly for that reason, no doubt) formed the strange view that it doesn" matter how good or bad dictionary definitions might be because their purpose is 'herely practical", one finds, with some surpris, Noan Chomsky. According to Chomsky, dictionary definitions don" offer anything like a faithful representation of a word's meaning, and yet they am all right as they are, because there is no need for dictionary definitions to be even approxirnakly correct. Anyone who has attempted to define a word precisely knows that this is an extremely difficult matter, involving intricate and complex properties. Ordinary dktionmy definitions do not come close to characterizing the meaning of words. Tb speed and precisicpn (DEvocabulary acquisition leaves no real alternative t c p the consomehow has the concepts available prior to experiewe with clusion that the ~ h i l d language, and is basically learning labels for cancepts that are already part cif hls or her conoeptuall apparatus. This is why dictionary defimuitionrs can be sufficient F o r their purpose though they are so imprecise: the rough approximation su%oe:s because the basic principles of word meaning (whatever they are] are k dictionary user, as they are to the language learner, independently of tion or experilenoe. The point is dramatically illustrated in the case of even the deaF-blind, who can acquire knowledge of the visual voc remarkable precision thotlgh extremely limited evidence; the meanings of such words as watch, gaze, glare, scrrrfinize, etc. (1987: 21) What is most striking In Chomsky's remarks is the absenoe o cultural perspective, and the complete disregard for the fact that fer in meaning across language and culture boundaries. For an and culture learner, a g o d dictionary is a tool of prime importance, an is an odd view for a linguist to take that there is no need to try to improve n existing dictionaries, however bad they may be, lxcause one can allways rely on one% innate conceptual apparatus. To take Chomsky's own: examples, how can students trying to learn English acquire the necessary knowledge of words such as watch, gaze, b e , or scrlxfinize (short of moving for many years to an EngBish-speaking if dictionaries do not offer them any reliable guidance in this ct, dictio~aries do offer some guidance, but their help is far and could easily be improved on (ir practical lexicography lexical semantics w r e to work hand in hand). For example, BPB d e h e s thle relevant meaning of watch as fo11ows1"to ak at, to keep one's eyes fixed on, to keep under observation". This defition, which includes three different would-be paraphrases, does not show what the invariant meaning o f watch is, or how it differs from the meaning of the closely related but more basic verb look of. In fact, the difference between the two lies partly in the temporal qualification (with watch implying the components 'for some time' and 'all the time') and partly in the nature of the object (to which I[ will return in a moment):

X was watching Y. for some time, X was looking at Y all the time
The validity of the component Tor some time'is supported by the fact that iook a5, but not watch, can be used with respect to a momentary evenk2
At that moment, he looked at her. *At that moment, he watched her.

The component 'all the time"% supported by the somewhat incongruous effect of the following sentence: ?While he was watching her, he was frequently looking at other girls. As for the object of uvatchiag, it must be capable of change, For example, one can watch a film but not a painting in a museum (unless one thinks that something may happen in, or to, the painting, for example that it may get sto1en). Hence:

X was watching Y* = for same time, X was looking at Y all the time because X thought something like this: something can happen i d t o Y now I[ want to see it
The expressionf i r some time i s Rat constrwcted out of semantic primitives and even the word some used in it is not used here in date sense of the primitive SOME [see Chapter 2). A mom articulated version of the semantic component in question could read (perhaps):'at at1 times. after one time, before another time'.

Glare is defined by the same dictionary as "to stare angrily or fier cially in astonishmentW,and gaze as "'Lo look long and steadilyq" initions, too, fail to show the exact relationship between ghse thing bad towards you, T want you to know itY.

X scrutinized 3E: = for some time, X was looking at E." during that time, X looked at all parts of Y if someone wanted to know everything about all parts of something this person could look at this thing like this

4. Fodolr on Definitions

X glared at K = X looked at P if someone wanted to say something like this to someone else: when I think about you I fed something bad I want you to know this this person could Book at this other person like this people can think: X wanted to say this to 3Y
and gaze) is incorrect, since one can glare at (though not sfare a t or lvsll someone briefly or even mmentarily. As for gaze, it implies, in addition to 'Ilooking', the followin ncnts: (1) 'for some time' (duration), (2) 'all the time' (steadily), (3) a of a knowledge-seeking purpose, and (4) 'feehngs"wonder, delight, fa ness, disbelief, etc.). The validity of these components is supported by following negative material: *He gave her a gaze. W e gazed at her for a moment. *Wile he was gazing at her, he was throwing glances at the other ?He was gazing at her intently, trying to read her thoughts. *He was gazing at her wilth hatred. This leads us to the following definition:
X was gazing at . ' l = for s o m time X was looking at Y sail the time if the person was looking at something for some time because this person felt something molt because this person wanted to know something this person couEd look at this thing like this people could think this about X

present proposal. (1 981: 3 15)

Fodor's reall challenge lies in his daim that there is a '"triking y of working examples [of definitions] in the standard iiteratwre" 2841, that "there seem allways to be counter-examples to the prodefinitions" (285)' and that ' h o s t of the morphemically simple sions of English are undefinable"(285). lieve that, for example, the definitions of the verbs watch,glareL., gaze,

Finallly, scrwtinize adds !tohokiag srf dements of, roughly speaking, r;ho
I am very graleIu1 Lo lgor Mel'Euk, who discussed villlr me my definitions point and Lr~sir"and who owered a number OF valuable insights.

oF the verbs

inle in detail"..)

Eodor writes pahi):

with refereme to George Miller's 11978: 285 disciussi

rson could $ay what calour Z was

What we have here is a proposal For defining "re transitive verb >plainty in Ile the noun 'paint', together with some further conceptual apparatus [COVER, FACE, and WITH]; and what I claim is that the definition doesn't work; X Y with pafnt mag be a necessary condition for Xpainis Y but it is certadnly

sufficient condition. He concludes his lengthy discussion of paint as follows:

brief c o w n b on this definition are in order. First, whib it is w n ter 71, and also, some non-primitive sptactic constnnctions. Second,

tion base that indudes 'ainosaur' and khlorodent'. Either way, the that Milkrk example doesay work. That's not surprising; when ilt lions, the exampbs almost always don't work. (1978: 288)

sary condition for Xpakts K For example, when a woman is painting: finger-nails she is not putting paint on them, bulk nail polish. This does mean that one can "paint" an object with just about anything; one wo for the activity in question. A sentence such as Xpainted Y with Z (where the verb pafm' is used in tlrc rcEcvant scnsc pcdi~ai?~) iiu~pPi~s 11381 a person OX) did soancthing to some object [ f l , putting some stuff (QZE on all the parts c~lrY that could be: seen (tlzat is, on the surface of l ' ) ; it furlher implies that the sluff Z was Illquid or semi-liquid at the time when X was putting it on , ' l and that Z had a defnitc colaur [so that one could say what its colour was); and that X wanted that stuff Z ta remain on the surface of Y and become, as it were, part of this surSace. Finally, the verb purist implies a function: if one puts some liquid or sehi-liquid stuff Z on the surface of object Y it is for a reason, and normally (though not always) this reason is to make this object look g a d (1 will return b this point shortly). As a first approximation, then, the following definition can be proposed:

language-specific. Far example, the closest Palish equivalent

X painted Y with Z. = (a) X did something to E" Qib] like peaph d o Qc) when they want something to look good (4 when X did it

X pokrasil Y Z-om ( X LrasitLed I ' with = (a) Xdid samething to Y (b;P like people do when they want something to be a certain coQour (ED because they think that this thing will look good bemuse of this


(4 when X did it (el X did something to some stuff Z (f)i i f someone looked at Z at that time
iCg) this person could say what colow Z was (h) at the m e time tlnis person could think that part of Z was water ( s ] X wanted Z to be like part of Y 0) after Xdid it Z was like part of Y The difllelrenm lie in h e : wmpanents Qbl), expllication of paint, and (el, which is less spe oomponenf of pa&. Thus, conwpts such as those encoded in the: words p a i ~ ~mabw~E, t, krmit'are language-specific. Consequently9 they cannot be innate unleamt; and since Iexicological research shows tha universal, the "p"prmitiveconceptual basis" of h a t e , pie"', indehablle concepts mwt be quite: dimerent fro [if I may say so] by Fador and his associates.
10. ConcacPusian
To quote Chomsky again, '%ordinary dictionary definitions do not close to characterizing the meaning of wordsyp. b harn a n ~ l k language r and understand another that words -not be defined can hardly be anyt Fortunately, Ithis propasition is not true. As Armstrong, Glei Cleiltman (1983: 2681 say, "the only good answer so many doubt the validity of the definitional vie theory is difficult to work out in the required detail". But "diRcuYt" not mean "impossible"". As I wrote in my Lexicography and Conc Analysis:
st obvious, and in a sense the most important, to the ordinary Ianguage the lexicon. This remarkable state: of atudYairs reflects the wide gap which any linguists' declarations and actual elffarts continues to separalte acatde~ngu~slijcs from "real life" as manifested in the needs and concerns oF ordinary language users. It reflects also the failuire of linguistic science to develop adequate ethodofogical tools for dealing with the lexicon-and a widespread M a c k of Faith the possibility of a purposerid, methodical and revealing scientific study of this

o the study of the lexicon is beginning

ry may bencome in linguistics the era of the

d approach to linguistic description"see

r, and Parkes (1980) put forward the destructive ns", which led to tlne conclusion that semantic general is impossible and should be abandoned. It is time to ge ithe self-defeating nature of this ssogan. Meanings mn be rigared if they are recognized for what they are: ue and culture-specific configurations of universal semantic primitives. ecogniae the: role of these primitives as a foundation on which all x meanings are b a d we c a n use them as an: instrument for impmvich will be developed in the fallowing chapter.

habitual devices. But the theoretical imssumptians implicit in these routines h dom been the subject of serious analysis. (Wieszbicka 1985: 11)

And in English Speech Act V;?rbs:

IT modern linguistics were b be judged by the contribution it made to lexica

it would be bardl to understand why linguistics is said to have made d r m

9. Semantics and lexixklograplrhy 259

9 Semantics and Lexicography aa

For years, I b e argued that semantics as a schokdy discipline must prove itself in lexicography. "'Lexicography needs linguistics, and Ihngukxics needs lexicography, As Zgusta (1971: 1P Y) points out, for the treatment of meaning in dictionaries to be radically improved, preparatory work has to be done by linguists" (Wierzbiclca 1987a: 1-2). I believe that during the two decades which have elapsed since Zgusta made this c o m e n t , much of h i s preparatory work has i n fact been done. In this chapter I will U y to show that, as a result ah this work, the treatrncnt of meaning in dictionaries can indeed 'be radically improved.

2. Scope versus Adequacy and Truth

Dictionaries are books about words. Unlike, however, various more or less selective "studies in wordsY"e.g. Lewis 1960), dictionaries are meant to be relatively complete-a~t least with respect to one thematic domain, or om aspect of language. Since they are also meant to be prwtically useful and commercially viable, one of the first dllcrnmas for a dictionary-maker is how to combine completeness with a reasonable six. It is at this point, I bdeve, that a practical lexicographer often Laecom impatient with theoretical lexicography. Theoretical lexicographen tend to maintain that to describe one word adequately one needs a grealt deal of space (many pages, if not many dozens, scores, or even hundreds of pages). As one leading lexicographer and semanticist, Igor Mel'Euk (1981: 571, pot it: "Not only every language, but every lexeme of a language, is an entire world itir itself." In a sense this is tme-but if so, then of course a praclt4ml lexicographer does not have the room to do justice to even a single word, let alone to the thousands of words with which he or she usually has G o deal.
An earlier version of this chapter was published as one oE two "lead papers" in a special issue of the journal Dfcrfonories (14. 1992-3: 6781, devoted to the )theoryand practia: of Meximp raphy. In the same issue, a oumks of commentaries on the t w lead papers sere published, along with the authors' replies [Dictionaries. 14. 1992-3: 139-59). Several OF these com~~nmtaries are referred to in this chapter.

One possible response to this situation on the part of practical Uexicographers is to turn their back on, theoretical lexicography and to continue doing what they have always done: to rely on experience and common sense. I believe that in doing that practical lexicographers have frequently produced valuable and useful works, and can still do so. But I also believe that if they try, instead, to look theosetical lexicography in the eye and to take from it what it has to offer, they can do a lot better. Landau C1984: 5) writes: "A dictionary is a book that lists words in alphabetical order and describes their meaning." It is only as an afterthought that he adds: "Modern dictionaries often include information about spelling, syllabification, pronunciation, etymology (word derivation), ma@, synonyms, and grammar, and sometimes illustrations as well." I agree with 'Landau's emphasis: although a good dictionary has to include, as Apresjan (forthcoming)i points out, morphological, syntactic, prosodic, pragmatic, and phraseological information, as well as infomation about meaning, it is the latter which normally constitutes the core of a dictionary. In what follows, I will not try to c o m e n t on all0 aspects of the relationship between Lhcorc.elical and practical lexicography, but rather will Bbcus, primarily, on the one feature which is truly essential: Ithe description of the meaning d words.' My main thesis with respect to this central problem i& this: The description of a word's meaning may vary, Iegitimateiy, in completeness from one work to another, but it should not differ in its basic content. A "definition" is meant to represent the truth about a word" meaning, and there is only one such tmith, whether it is to be presented in a research paper devoited to one particular word or in a dictionary intended for a general audience, including various dictionaries addressed specif cally to 'khildren", "learners'', 4'sbudents's, and so on. It is a curious but widespread illusion that by saying things which are untrue, meaningless, obscure, or theoretically untenable, the dictionarymaker can gain in either insight or spaw, and that the dictionary user is better served. If space is of paramount importance in a "comercialty e'victionary, then all the space available, however limited, should be for saying things which without being complete are none the less true, eaningful, illuminating, and clear. It might be thought churlish to deny that reputable c o m e r c i d dictionaries do say, by and large, things that are '"me, meaningful, illuminating, and clear". But unfortunately they often don't.

a Que partimlarly important area which has not been discussed at all in this chapter is that of relationships Between the meanings OF words and their syntactic properties. For both general discussion and ample exempiiffication, see Wierzbicka [M9&Te]. See also Ch. 5.

260 Lexical Semntic3

3. Saying Something that is not True

Sometimes dictionary definitions say things which are simply false. Far example, the Oxford AwbsaJ"hy3Junior Dictionary (OAJD 1980) offers the foillowing definition, of sure: sure-knowing something is true or right But of course '%knwingg' and "being sure" are two very different things, and even in a dictionary intended for children they should never he equated. An "adult" dictionary, the Oxford .$@perbackDictionary r(0P.D 1979) offers a more complex and 'kophistimte8" but in fact equally false, definition:
sure-having or seeming to have sufioient reasons For one's beliefs, fma Rom doubts

CobuiPd Engli$h Languuge Dictioy3ary [Cobui/d s, Cobuiiild defirnes empathy as '"the ability to share another personas s and emotions as if they were one% own""'.But in fact, as shown by (19921, empathy does not imply that one shares another person's (but rather, that one understands them, as if they were your own); athetic counsellor cannot be expected to share his efinition off forgive implies incorrectly that to be one has to be first angry with them and want to give someone who has done something wrong or one has done, you stop being angry with them, unish them'" But stories of saints and martyrs lences in which someone is said to have forgiven thalwt any implicallioru tlrat at first he or she was ted to punish them. Likewise, in the Gospel story the prodigal son [which for many people epitomizes forgiveness) there is er was at first angry with his son m d wanted

The apparent afterthought ""Tee from do~brs'Ysbasically right, but doubt itself is defined by the OPjD via certaintry, and cerrahfy via dosxbt (doubt"feeling of' uncertainty about something'" ceastaia--"having no doubts"). Leaving aside this circular detour (see Secltion 101, we will note that for ~ u b jmtive certainty (being sure of something), having sufficient seasons far one" belief is neither necessary nor suficienf. Similarly, amounce is defined R a y the OAJD as "to say something in front of a lot of people"'. But in fact, one can also announce something [for example, an important decision) to one's parents, and the presence of a lot of people is not necessary at all. BoMis defined by the OAJD as "brave and not afraid". But this is wrong, too: one can be bold without being brave, and be brave without being bold. In particular, bolldness is shown in relation to other people, whereas none can be brave even in solitary confinement (see Wiembicka 1992a: 208-91. Srondard is defined by the same dictionary as "how good something is1'* But in fact, it is rather "how good you think something has to be". Threat is defined as "a promise that you will do something bad if what you want does not happen". However, a threat is mot a kind of promise, although it can be called that ironically; and one can say, for example, threa~ mdpromises, whereas one cannot say *spmie;eba d dogs (because a spaniei is indeed a kind of dog]^. AbiUty is defined by the 8?$JD as "the power to do something"". But although the notions of 'ability'and 'powerkre related, the fomer cannot be reduoed to the latter: "ower' implies that one can do things ]that someone doesn" want, and so it implies actual or potential conflict of wills; 'abillity', however, does not h p l y this. To show that errors of this kind occur also in ambitions, prestigious, and innovative modern dictionaries, I will condude this section with two ex-

4. Saying Something that is Superfluous

Given space constraints under which practical dictionaries usually operate it is surprising to see how often they waste precious space by saying things which are entirely superfluous. For example, the Longman Dictionary o f Conrlemporary EngIish (LDBTEL 1984) defines the word weapm as follows:
weapon-an instrument o f okTenstve or defensive cornbat; something to fight with

The simple phrase "something to Bght with'qs perhaps not a perfect definition of weopon but it is a pretty good approximation; the definition is qoiln, however, by the completely unnecessary addition of "an instrument off offensive or defemnsive combat"".One can almost sense the nervousness of the lexiooguapher who, having produoed an excellent short definition, realizes that he or she has nothing to add to ilt-and panics at what appears to be m unfamiliar, unconventional level of simplicity, and tries desperately to add something to make it longer, more complex, more "respectru.ble". Theoretical lexicography can be very useful at this point if it can reassure the practical lexicographer: "'There is no need to add anything; the simple short definition is okay; on the contrary, it is the longer one which is faullty, hecause, as Aristotlle pointed out twenty five centuries ago, in a definition every superfluous word is a serious transgression."

5. Confusing Mleaning with Knowledge

Another way to waste space in a dictionary is to include in it technical or scientific knowledge. For example, SDBTEL defines the word dentist as follows:
dentist-a person who is skilled in amd licensed to practise the prevention, diagnosis, a11cl tre:uil~nient of disciusce, irrju~rics,: P I I ~ nialra~rmuiutio~rs nf thc teeth,

h between sugar and saccSiariue we could say that while both sugar ccharke are "added" to some things that people eat a r drink, only sugar can it be said that people can "eat it as part of some things they onsider also the following definitions of horse from three different dica solid-hoofed perissodactyl qluadrupd ~(Equrt~r ccrr*~!fu~fi (SOED 3964) a large sollid-hoofed herbivorous mammal [EQMMS cabaJ/u.r)domesticated by man since a prehistoric period (Web~tes k 19881 arge solid-hoofed pllamt-eating Clegged mammal (Eqcttrs eabal[us,family Equidae, e horse Family), domesticated by humans since prehistoric times and used as beast of burdem, a draught animal, or For ridimp; esp. one over 14.2 hands in height

jaws, and mouth and wha makes and inserls false teeth

It may be instructive for a reader to learn that a dentist does all the things enumerated in this definition, but information of this kind, however useful, is out of place in a dictionary. The short definition oflered by IOAJD (though not perfect) is much more satisfactory:
dentist-someone whose j~obis to look after teeth

The line between knowledge and meaning is not always easy to draw, but in principle it can be drawn (see Wierzbicka 1985 and Cbapkr 1]I), and in any case crBicfionaries are often full of infomation which quite clearly belongs in an encyclopaedia, not in a dicctbnalry. Consider, for example, the following definition of sugar (LDOTEL; my emphasis):
sugar-+ sweet substance that consists wholly of s~crose, is mlaurless or white when pure, boding to brown when less refined, is usually obtained commercianly fiom sugarcane or sugar beet, and is nutritionally important as a source of carbohydrate as sr sweetener amd preservative of other foods

ardly needs to be pointed out that dehitions of this kind do not repwhat ordinary speakers of English have in mind when they talk of ses. The information imncluded in such "definitions" isi for the most part, perfluous in a dictionary of English. It would be much better to ply that a horse is 'k kind of animal called horse"'. It would be betto try to explicate, in an abbreviated f o m , the folk concept encoded ngllish word horse; but if a dictionary cannot afiord the space to do waste space on infomation which is given in all encyclopaedias h has nothing ta do with ordinary speakers' knowledge of their guage anyway? (For further discussion', see Chapbr I 1.)

Clearly most of what the LDBTEL definition offers is not g a d of the everyday concept at all (not to mention the fact that sugar is defined here via sugar-cane and mgar-beet; see Section 10). As usual, lthe OAJD ddinition, though not perfect, is much m r e piausibte: 'ki sweet food ithat Is put in drinks and other foods to m&e them taste sweet'". (It would probably be better still to say something llike this: 'csomething that peopb add to things they drink or eat when they want to make them taste sweet; it ~jcsmes from some things growing out of the ground w[i.e. plants); it is normally white".) McCawlley (1492-3: 823) suggests that, in one respat, the OAJD's definition of sugar "is more accurate than Wierzbicka", since QAJD's initially guzzling use of 'food' in the definition neatly distinguishes sugar from such sugar substitutes as saocharine and Nutrasweet". But while sExgilEr should indeed be distinguished from saccharine, do we have to call1 sugar a 'Tood'~something that McCawley himelf Gnds counter-intuitive) to achieve this god? Iffiod stands, roughly, for things that people eat, then it is understandable why people would normally not call sugar "a food": one noranally doesn? eat sugar (on its own). To dis-

6. Dehitilons which are too Broad

0 1 1 s which are too broad do not contain any falsehood (because ng they include is true), but their implalcations are false (because ey leave out certain necessary components). For example, taiemr is defined by the QAJD as '"he ability to do someh g very well"'. But this implies that an acquired skill could be called tab t, which is not true. The definition misses the crucial component 'inborn' 'if someone can do things of a certain kind very well not because hei'she $I something to be ab11e to do them well"). 7%succeed is defined as "to do or get what you wanted to do orget". By is definiition if one gets a present that one wanted to get, this could be esleribed as suweding; once again this is not true. (To sulciceed one has to ca something; the disjunction ""dl or get" is therefore wrong.) 7% defy is defined as "to say or show that you will not obey". But if a ild says to his or her brother or sister, 'You are not my mother or father, will not obey you"', this would not be described as defying. [One can only

264 Lexical Semantics defy orders given by someone who actually does have authority over m and can be expected to be obeyed.)~ The definition of steal says "to take something that does not could rcfer to robbery as well as to st you and keep it". But tll~is stealing, it is essential that the actor does not want people to or she is doing, and expects that they will not know it. Secret is defined as "something that must be kept hidden from 0th plc"; bull this could rchr LO physical objects, whcrcas in fact secret only for something that one knows (and must mot tell other peo T h i r s t is presented as '"he need to drink"; but in fact it r sation, to w h d one feels Ywhen you feel you need to drink"). A ribbon is, according to lthe OAJD, "a strip of nylon, silk, or some a material"". But if this were true, any strip of any material co ribbon, which of course is not lrue. In Fdct, the word ribbon [in the re1 sense) refers only to a kind of thing made (by people) in order to something look good. [See Section 9.) IFcl shed is defined as "to let something falYy[and it is illustaate sentences "trees shed leaves, people shed tears, and caterpillars skins"). Hut this implies that if 1 let a book fall 1 am shedding cia1 concept missed by this definition is that of 'part" A can only shed B before the event B can be thought of as part of A (and after the event, can not]. Finally, a wswm is according to the O"AJD '% fully g r o m ffemak which turns a bitch or a mare into a woman. Needless to say, definitions can also be too narrow, but this fault seld oacurs on its own, and P will discuss it in the context of which it is most commonly combined. Here, just one exa The QAJD defines appahtment as "a time when you have ar and see some one'^ s. is too restrictive, because the l a y e r who r w i v clients or the professor who sees students can also have an without having to "go" anywhere outside their ofice.

mplain-l. to say that ame is dissatisfied, ta pmtest that something is wrong; 2.

to state that one i s sufTe'feritl.g from a pain, etc.

first meaning is stated in two different and non-equivalent ways. I1 needs to be pointed out that one can complain without feeling (or retending that one feels] pain; that one can protest that something is

d with oneself without complaining about anyone or anything. in addition to being wrong in almost everything it says, the whole exudes lexicographic despair and apathy: "It is impossible to czupture variant, or even to dmide how many different meanings are invohed; t is, we do not know how to go about it." the lexicographers responsible for this entry had a reliable Ilexicotheory at their disposall, they would have need of neither their halfspeaking (for a detailed and more predse discussion see a 8987a3, the complaining person has to convey the following 'something bad happened to me-1 feel something bad bemuse

7. Capturing the Invariant

Althaiugh this may sound too grand far (what tends to be seen as) humble task of a lexicographer, the process of constructing a definritioln is-or s h o l d be-a search for truth. To find the truth abou meaning of a word means to find the invariant concept which is part o native speakersQacit knowledge about their language and them in their use of that word. Yet lexicagraphers often lack the confidence, the resolve, reach for the invariant, and thus bccome unfaithfull to their t

nd the failure b even aim at capturing the semantic invariant. The exlcogmpher realizes that the phrase "'to state that one is suffering from a " is too narrow as a definition of comptain, but instead of looking for s restrictive f o m d a he or she simply adds am ebc. (and, for good mea-

9. Semanfics and Lexicography 267

are others which are inherently negative (that is, which reflect a negative: evaluation), for example 'reckless', Yoolhardy', or Yinapudent', "Bold' belongs to neither of tl~esetwo calcgorics, bcing cormpalibPc with either a positive or a negative evaluation. By splitting it into two supposedly different meanings, one positive and one negative, the dictionary is misrepresenting the truth about this concept and blurring the difference between the neutral concept 'bold', the positive concept "courageous', and the negative concept 'impudent'. (P4wdless to say, 'boldy,'courageoms', and 'impudentJ differ also in other respects; for detailed discussion see Wierzbicka 19920: 203- B S .) Another characteristic example is the OPD definition of boast: '"0 speak with great pride and try to impress people, esp, about oneself"".This time no polysemy is postulated, but the little esp. (especially) is no less of a sigh of resignation (or a moan of despair?) than the etc. of the previous defiuolition. Clearly, the authors of the entry could not make up their c01iective mind as to what the essential features of boast are. If they had a reliable lexicogtaphic theory to lean on, they would have k n o w that an esp. is a sign of defeat and they would have fdt obliged to think a little longer. I f they had done this, they would have realized that the concept of "oast' always involves oneself (whether directly or indirectly), and that it is not diflerent in this mspect from 'pride" Again, for detailed discussion the reader is referred to Wiembicka 1987a; here, it wiU suffiw to say that hasting always involves the following attitude: saying something very good about someone or something, thinking something very good about myself, comparing myself with other people ('other people are not like me'), and wanting other people to think something very good about me. Thus, if a father says something very g m d about his children, evidentlly thinking something very good about himself, comparing himsePf with other people ('other people are not like me'), and wanting other people to think some thing good about him bemuse of this, then this can indeed be described as boasting. It may be noted in passing that the OPD definition of impress, via which the OPD defines boast, is equally inadequate: "'to make (a person) farm a strong (usually favourable) opinion of something"".suadJ"y, like esplechlly, suggests that the component of "favourable opinion" is not n a e s s q , whereas in fact it is absoUutely necessary both for beast and for the relevant meaning of imprw (as in "He wanted to impress h1erJ'). The Failure to capture the invariant is manifested in a particularly spectacular manner in dhc use of the conjunction or, with which dictionary definitions are usualBy peppered. For example, the QPD defines tempt as follow: "to persuade or try to persuade (especialEy into doing something wrong or unwise) by the prospect of pleasurc or advantage". The first or can be dispensed with immediately:

ting can of collrrse be suwssful, but so can trying; it is enough, therey "try to persuadem,thereis no need for "to persuade or try to '. The disjunction c'~omething wrong or unwise" caw be reduced hing bad" (not necessarily " " e ~ " a r "morally very badW",ul that is thought of as a bad thing to do]; and the "prospect of asure or advantage" cam be reduced to the prospect of "something good temptee)'". The qualifier espech#y cam be dispensed with altogether: ply cannot tempt somebody to do something good; it has to be Lng that is seen as "something bad"' (although the speaker can of e be using the word tempt in jest). conclusion, devices such as w ,espercial/ysu~aaS!y, the use of mulltiple es [whether words or phrases) to portray the same meaning, or the stting of arbitrary polysemles are all different manifestations of the same failure of practical lexicography. This failure mars most entries in of the existing dictionaries, and makes them m c h less usefid to the r than they otherwise could be [under the same limitations of space d other practical constraints). A rigorous and consistent lexicographic om,with a firmly established principle of determinacy of meaning, can illy remedy this weakness. n particular, a sound lexicographic theory can prevent the common phenomemoar of unfounded proliferation of meanings, as well as the (less common, but even more harmful) conflation of meanings which are related but not the same. It can prevent the confusion of ironic, sarcasltlc, jocular, or metaphorical usage with the literal meaning of words (as in the case of threat, discussed earlier). It can oRer lexicographic criteria on the basis of which meanings can be firmly separated from one another, clearly identified, and intelligibly stated.

8. Standing Firmly on the Ground of Dis~reteness

(Dineof the major reasons that most dictionary definitions are much less useful than they could be is the widespread lack of faith in the discreteness of meaning. As Aristotle realized better than many contemporary linguists do, there are few things harder than constructing a good definition. How can a lexioagrapher be expected to undertake the necessary effort if he or she does not believe that lthe task is feasible at all? Theoreticians who undermine the lexicographer% faith in nhe possibility of stating the meaning oE words truthfully and accurately are doing both the lexicographer and the dictionary user grievous disservice. ln fact, most of the problems which plague practid lexicography are linked with the issue of discreteness. For example, how can lexicographers search with all their might, and patience, for an invariant if they do not know whether they can expect to

9. Setnantic..vand Lexicography 269

find a definite number of meanings? To recdl the OPD definition of cornplczim, cited earlier: "I, to say that one is dissatisfied, to protest that something is wrong; 2. to state that one is suffering from a pain etc.". On the face of it, two meanings are postulated here, but in f a t the Arst alleged meaning is stated twice, in two different ways, and the relation between these two different attempts at a definition is left unclear. The use of the numbers I and 2 implies that this particukr lexicographer does believe in the discreteness of meanings, but the constant practice of throwing together different fomulations of what is counted as "one meaning" indicates the shakiness of this belief, If the lexicographer felt obliged to state just one definition for each @ypothetid) meaning, this would encourage him or her to look for the true invariant; as a result, the multiple meanings postulated In present-day dictionaries w u l d o f t ~ n be reduced to a smaller number, invariants would be captured, superfluous phrases wodd be omitted, space would be gained, semantic rdations between diflerent meanings (and dit= ferent lexical item) would be made much clearer, and, an tog of all thiss, the language used wodd be much simpler and clearer, as well as more =anomical. For example, as pointed out ember, far camphin one could propose just one unitary formula, and there would be no need for positing polysemy, no need for an etc., and no need for agonizing between two nonequivalent phrases "to say that one is dissatisfied'hand "to protest that something is wrong" (not to mention other advantages linked with n stmdarbarian and reduction of the metalanguage used). Or consider the way LDOTEL defines the English verb rltl pray:
la, b a entreat earnestly; esp. to call devoutly on (God or a god] b, b wish or hope fervemdy 2. arch~icor JomaE to request oourtesy---often used ta introduoe a qlueslti~m,

have been alleviated, and the efforts rewarded with more satisfactory ts, if theoretical lexicography had sent a clew md unequivocal mesthat meaning is determinate, that a definite (and minimal) number of eanings must be looked for, that there are no "shades" of meaning, no and (bls, and that no hedges [no ~specSd~Sly, +em, etc.1 are necessary or ptable; and if, in addition to this message, clear criteria for establishd distinguishing different meanings had been provided. I believe that h a dear message had been sent, and if the nemssary guide-lines had ided, the entry for pray would have ended up with just two meanarchaic and one contemporary (without any submeanings), and e two meanings would have been stated clearly and a m a t e l y hedges, OM, ees., or other visible signs of i n k i s i o n and ana;see WieraBPicka 19934. sider the lists of quasi-equivalents offered in dehitions such as the following ones from Webster" (1959): l y (noun) - an answer; response; ccounnter-attack - to ybld to aaather; surrender Formally; withdraw from; submit calmly neport (verb) - to give an amaunt af; relatq telU from one b another; circubate publicly; take down (spoken words) iequest [noun)- desire expressed; petition; prayer; & m a & entreaty er (nourn) - method of regular arrangement; settled mode of procedure; rule; regulation; mmmand; class; rank; degree; a religious fraternity; an associatiom of persons possessing a common honorary distinction . . .

request or plea 3. archaic to get or bring by praying: (ia address God o n . a gad wiflh adoralion, confession, supplication, or thanksgiving; engage in prayer Although three figures are used (I, 2, and 31, the actual number of meanings postulated is far from clear: Ps (a) a separate micafling? Or @]? and what about all those especiaI1ys, or$, eftens, semicolons? The relationship between the different alleged meanings is even less clear than the number of meanings postulated. For example, why should the meaning "to address God or a god" be given under heading 3, and "to call devoutly on (God or a god]'hPuner I? Are these two alleged meanings more different f m one another than those listed under IH[sr)and 1(b] (only one of which mentions Gad)? Why is the alleged meaning "to engage in prayer" given lander 3 and "to call devoutly on (God or a god)", under 12 I am not saying that it is easy to define pray in a satisfactory way, and E sympalhlize with the lexicographer" painful erlirorls. But 1 bclieve this pain

Walt is most striking about such lists is the fact that the Pexicographer making no attempt to indicate how many different meanings are involved ch case. Should the first four entries above leave anybody in doubt t the lexicographer% mmotivatim for this failure, the hotchpotch of I-equivalents thrown into the entry for order makes it quite clear that e only possible motivation is despair. This despair is understandt is not jzrslified. As 1 have tried to show in my Dicfionary o f eecb Act Verbs (which includes, in particular, the verbs reply, igm, seporr, and repest), meanings, can be sorted out from one another, ace Wittgenstein and kllowers) boundaries between meanings can be .The doctrine of family resemblances must not be used as arm excuse eximgraphic laziness or as justification for lexicographic despair. (For ther discussion, see Chapter 4.) one final point related to the question of discreteness, B e t us consider 's (1982: 1012) claim that there is no reason why "dictionary definitions are to be mad as mutually exclusive'" and that ' " i n practice, the warding of definition 1 normally colors the interpretation of definition 2'" 1agree that,

9. Setnass tics and LexicograpIry 27 1

fmm a reader's point of view, the wording of one definition may colour the interpretation of the other definitions of a polysemous word. The main problenn, Inowever, is to establish whether the word in question is reallty polysemous, and to ensure that its meaning or meanings be correctly identified. If this is achieved, theu I think there is no need for such a "cross-fwtilization of definitions"". For example, Hanks quotes a sentence describing two girls as "shultaneously bold m d innocent"",sking: "Does boM here mean forward or impudent or daring?"; and he answers: "'A bit of both, really"'. But why should we assume that $aEd reaily does have two meanings which can be stated as (I) forward or impudent and (2) daring? If none of threse supposed "definitions" " s the sentence well, it is, I think, not because there are two meanings which colour one another, but because neither of the proposed "definitions" of boId is correct. (For an alternative definition, and justification, see Wierzbicka 1992a: 208-9.) A definition should always be able to stand on its own. If a word is %enuincPy polysemous, h e n each of its meanings should be stated separately, and each definition should be able to defend itself. This is not incompatible with Hanks" statement that "secondary meanings have a tendency Ito contain traces of primary meanings'". I, too, believe that different meanings of a word are usually interrelated, and that adequate definitions should reveal those links. (For many illlusitrations, see Wierzbicka 1987; also Meltuk's concept of '"emantic bridges", Me16Euk et a l 1984.) But this does mot change the basic requirement that each definition shodd be able to stmd on its o w . 91. Distinguishing Polysemy from Vagueness One of the main reasons why lexicographers often find it di@cult, indeed impossible, to a p t w the semantic invariant is that they do not know hoy to distinguish polysemy from vagueness. It is not that lexicographers do not believe in poiysemy: frequently, polysemy is posited in dictionaries on a truly massive sale; but it is posited on an od k c basis, without any clear guide-lines or general principles. Consider, for example, the definition of ribbon mentioned earlier: "a strip of n)llon, silk, or some other material" (COAJD). As pointed out, this definition implies that any strip of material could be called ribbon, whereas in fact many different "strips of material" (e.g. a piece of sewing-taw would not be so ailed [because their hnction is clearly different from that of ribbon: sewing-tape is clearly not made for tying things and, equafly clearly, it is not made h r a decorative purpose). But the generalization poposed here may seem to apply to many [even

mrpst) cases rather than to a E I cases. For example, what aboult typewriter ribbon? Is it meant for tying things? Or is it decorative? And yet it is called ribbon, too, isn't it? Confronted with an apparent exception of this kind, lexicographers often tend to lose faith i n the existence of a semantic invariant, take recourse to hedges, qualifiers, and various other 4d hoc devices, and Pose the generalization. But in fact the counter-example is apparent rather than real: the so-called Jypew~iterribbon is not called ribbon but bypetvriler ribbon (even if it can sometimes be referred to, elliptically, as ribbon]. The common belief that a modger-head constructiolv must indicate a taxonomic ('kind of") relationship is based on a falEacy, which feeds on the fact that compounds with such a structure can often be abbreviated, in an appropriate context, to the head alone. For example, ill is often assumed that an artificial leg is a kind of leg, that a plastic flower is a kind of flower, that an electric chair is a kind of chair, or that a house of cards is a kind of house. Since people cannot live in a house of cards, m d since "a house of cards is a kind of house", the generalization that houses are m d e for people to live in appears to be easily refuted. Similarly, since plastic flowers do not grow out of the ground, and since "a plastic icaawer is a kind of flowr", the generalization that flowers grow out of the ground may also seem to be easily refuted. Reasoning of this kind is fallacious because it confuses semantic relationships based on the notion of VikehlNlth those based on the notion of 'kind' [i.e. "horizontalY'and'%ertical" rerelionships; see Bright m d Bright 1969; Berlin 1992). For example, a rose is a kind of flower, but a plastic Bower is Me a flower, not a kind of flower. Similarly, a deck-chair is a kind of chair, but an electric chair is not a kind of chair ("somhhing for people to sit on . . .'"; rather, it is an object which is like a chair, bwlt whose function is quite different from that of a chair. Finally, typewriter ribbon is not a kind of ribbon; rather it is something which is like a ribbon, but whose function is quite different from that of a ribbon. There is, however, one important difference between, the case of typewriter ribbon and that of plosticflowers: the fact that one can rala a plastic imitation of a flower aflower is language-independent, whereas the Fact that the "typewriter strip"is called in English rfbbosl! (typewriter ribbon) is language-specific ((e.g. in Polish the corresponding rcompound is r a h a do maszyny, lit. 'typewriter tape". Consequently, sypewriier ribbon has to be Iisted in a dictionary as a separate item, with its own definition, whereas pI~$tlflowesdoes mat. To prove that a typewriter ribbon is not 'a kind of ribbon' and that the definition of ribbon does not have to cover typewriter ribbon, we proceed as fol~ows: we first assume, far the sake of argument, that typewriter ribbon

272 Lexical Semwtfcs

9. 8emts~n~ics and lexicography 273

does have to be covemd in the definition of ribbon and we ask what the two categories have in common; when we establish the common denominator [roughly, "a strip d material"] we ask whether any object which fits this common denominator can be called by the word in question (rfbbon]; we find that the answer has to be negative; from this we conclude that the supposed common denominator cannot account for the wohd's range of use; and fmm this we infer the exisltenm of polysemy. I[f we proceed in this way yve can arrive at dehitions with full predictive power, not in a diachronic sense, explaining why certain objects came to be called by certain names I(e.g, why the "'typewriter strip" came to be called in English by a compound including the ward ribon, whereas, for example, in Polish it came to be called by a compound including a noun which on its o m means but in a synchronic sense, which m e a n s that the definition matches a word's actual range of use. This procedure will enable us to be precise and to dispense with hedges and qualifying expressions such ojrsen, typica!Ilp,or etc., which are meant to make up for the inaoas ~rswally, curacy of the definition itself. Another m y of making basically the same point is this: a hafr ribbon differs from other kinds of ribbons in only one respect, spmified by the m d irfier (being used for tying hair with), so there are no grounds for positinrg polysemy in this case. But typewriter ribbon differs from other ribbons in more than one respect (it is used in typewriters, it is not suitable for tying things with, it is not decorative); in this case, therefom, polysemy has to be postulated (see Chapter 83. Thus, although with respect to nouns polysemy often has to be established on purely semantic grounds, this does not mean that there are no guide-lines for establishing whether a word has one meaning or two (or more). With respect to verbs, the task of establishing the number of meanings Is often facilitated by differences in syntactic frames. This point can be illastrated with the English verb to order. The Concise Oxford Dictionary (GOD 19M) offers for lthis verb the following glosses: "'put in order, array, regulate . . . ; . . . ordain . . . ; command, bid, prescribe . . . ; command or direct . . . ; direct tradesman, servant, etc., to supply r( dinner, settle what it shall consist of)". These glosses do m t make it clear how many meanings are being posited (and perhaps the underlying assumption is that the different meanings of this verb cannot be sorted out from onre another). But in fact, if the s p tactic frames are sorted out in orderiy fashion, clear semantic dis.Xiwctians emerge, too. Leaving aside (for reasons of space) the meaning illustrated with the sentences " 'e ordered his affairs'kr ""He orde~ed his troops", 1will focus here on the distinction between what 11 lnrilil call orderl and olrser2, which can be

tences "'She ordered him to leaveY'and "She ordered with only minor s in m y Dictiomtzr'y in a restaurant or a book in a boohhop to have that thing and expects him to do llar tto order]. But there are numerous order2 but absent from order,. The pers to have something; an action by the r ithis desired state of affairs to eventuate, but it is oal itselt it does not matter who carries s long as it is carried out. This semanin a syntactic difference: one ordersl a person (to rders2 a thing (from a person). One cannot order2 r2 a steak, and one cannot order! a steak (as one rs 110 a steak), even though concrete nouns can be reported door', he ordered"). ect represents the focus of the speaker" interest. For kiject has to refer either to the addressee or to the action to IPZ~IV~", 'We ordered1 an inquiry"") For 0rder2, it has to an object ('We ordered2 a steak""), with the addressee being cony and syntactically demoted to a prepositional phrase CX ordered2 eakes who orders2 something assumes that the addressee ny people may want to have, and that he is willing: to people with some d these things, on certain conditions. The cerns not only the product which the speaker wants to sesxioes: the addressee has to do something; to the object desired I(i.e. get it, prepare it, wrap it, serve it, and so on). For his prt,the speaker undertakes to do something, too: whatever is required (for example, to pay). He also has to wait, because his wish cannot be complied with immediately [as in the case of buying), but onlly after some delay, allowing the addressee to perform the necessary actions. Finally, the person who orders2 something does not assume that the ddressee has L o do what Ere wants him to do; he does assume, however, that the addressee is willing to do SO. A person who orders1 someone to do k that 9 . a ~ have to thing is a little less confident, because out ithe arderz. But, in this case, too, nfident fcl think that if you can do it, you will that order2 has developed out of orderl by a shift of phasis from the addressee tla the object provided by the addressee, and a concomitant disappearance of certain assumptions conoeming the

9. Siemanrfsrtics and Lexicography 2275

relationship between the speak~er and the addressee: order1 implies that the addressee is subordinated to the speaker and has to do what the speaker wants him to do. In o r $ e ~ this , assumption is absent; instead, there are new assumptions which emphasize the object ('something has to be done to itY] and which de-individualize the addressee: he is not seen as an individual subordinated to the speaker but as a person or group of persons who i d bgly provide certain kinds of objects to unspeEified customers. Thus the last gloss in the COD entry for order ("'direct tradesman* servant, etc., to suppIy9')confuses two diflFerent meanings, associated with tiifferent types of social relationship and with different sets of assumptions: it is one thing to order someone I(e.g,a servant) to do something, and another to order something from someone (e.g, from a trades man)^. In an adequate 3iexiloographic description, these two senses whould have to be clearly distinguished, and the readers would have to k informed of the different grammatical frames associated with the different meanings. rst : "a jumping movement". Thus r o j m p Is defined jumping, and jmping is not defined at ail, being a form of the verb ro jump. Substituting the da, we obtain the falltowing: ta j u m p C ? omove g movement"; lihat is, "to move suddenly with a of moving suddenly with a movement characterisa movement". And so it goes on-like a record ould take a real jump of faith to believe that that ition which d l serve the dictionary user best. another example, the QPD defines fate as "a persomn's destiny", as "that which happens. to a person or thing thought of as deterplacing the wordjate with its definition we get the folthat which hahagens to a person or thing thought of as t which happens to a person or thing thought of as that a person or thingY'(and so on, ad kjinitum!). Quite apart oF this kind is an insult to the readers' inteielwhat use could it possibly be to them? B f somebody knows what ans, what destiny means, and how they differ from one another, any definitions at all; but if they do not know, or are ictionary will not teach them anything. BoCh the proney is wasted on definitions of this kind. be added that fate is also given another definition in the same thought to control all events and impossible to resist", Iso defined as '"ate considered as poweryy. This time, definition we get: destiny Is "a power . . . considered mare illuminating result. that this last example is unfair because the concepts iculady difficult to define. But in fact circwat the easiest concepts are often defined in the same For example, LDQTEL defines besr as "excePling a11 others". Excel is efined, in turn, via superior and surpass, superior via surpass, and surpass ia betser, as well as via exceed, with exceed being defined in turn via merior. What baffles the reader most is why besf ~ouldnV haw been defined ia better in the %st place rbetter than all the others'"), instead of going und the circles involving superior, surpass, and exceed Similarly, OPD defines question as "a sentence requesting information or an answeryys and answer as ""smthing said or written or needed or done to deal with a question, accusation, or problem". Omitting, to save space, the numerous ors, we get something like this: an answer is "something said to deal with a sentence requesting something said to deal with a sentence requesting . . ."-and so on, ad infis~ffum. And yet the essence of a guestion, or an smmwer, is not difficult to state: tbe gueslianr refers to a situation

1 1 0 . Avoiding Circullirrity
There was one thing that Edith could anat beet, and that was the d i e liionary. ""The Larausse i s a big cheat. You look for a word, you find It, they send you back to another word and you haven" got anywhere." (Blerteaot 1973: 1103) Conventional dictionaries are, generailly speaking, vitiated by all-pervasive circularity in their definitions. Some didionaries are better in this respwt than others (for example, the OMD is much better than all the Oxford dictionaries addressed at adults), but while there are differences of degree, there are hardly any exceptions+ircularity is a malady (in a more or less advanced form) to wbicln virtually no conventional di~tionary is immune. Practical lexicographers are often well aware of the circwtrity of their definitions, but not knowing how to avoid it they try to make a virtue of "mecessity"', and attempt to justify this circularity as something that may bother theoretical semanticists but that is quite amptable in a practical die tionaq and which will never bother the ordinairy user. In fact, however, they are dwxiving themselves (and unwittingly insulting the intelligence of their intended audience). A few examples. The verb tbohmp is defined by the OPD as "to move off the ground ete. [try bending and then extending the legs or (of fish) by a movement of the: tail"' (Ifitst meaning] and "to move suddenly witla a jump or bound, to rise suddeniy from a seat etc." (second meaning). As for the f i s t definition, one might query the unexpected attention it gives to fish (as we18 as its use of etc.), but it is the second definition which is relevant in the present context. For what is "a jump'? ?e OPD offers several definitions, but the relevant

when, rough,y speaking, someone says '7 want to know something, I want someone to tell me" and nnswex means, roughily, '%elling someone soma thing that Ithey said they wanted to kmwY'(for more precise definitions, see Wierzbicka 1987~). The examples given above were relatively simple, with A being defined via B, and B via A. TypicaBly, however, vicious circles are like huge webs enveloping whole extended families of wards, or Bike gigantic tmtades extending throughout the pages of a dictionary. Far example, A is defined via B, B ria C, and iC via A; or A, B, C3D, E, and F are defined via one another-in circles, criss-crosses, and all imaginable sorts of combinations rand patterns (for example, A via B and D;B via D, E, and E T; via A, B, and Cn; C via A and l$ and so on)-with repercusions throughout the entire dictionary, which becomes an entangled web of overlapping circles. For example, in the OPD, one ~ l a a find a little circle (Fig, 9.1) and a larger one, with the little one within it (Fig. 9.2).


offer accept






In a system of this kind, every answer generates new questions, and these new questions either lead us to further questions or bring us back ltc~the starting-point (or both). It is just as Descartes said in his "Search after Truth" (speaking through the mouth of Eudoxus and mfersing to the socalled metaphysical steps or the tree of Porphyry]:
~uoaxw. You pay n o attention to my question, and the reply that you make to me, simple as it may appear to you, wiill bring us into a labyrinth d &ifficulties, iF 1 try ever so little to press you. Were 1 for example to ask E p i s t a o n himself what a man is, and were he to reply, as is done in the Schools, that a man is a rationall animal; and if, in addition, in order t o explains these two terms which are not hss obscure than the first, he were to conduct n u s by a31 the steps which are termed metaphysical, we should ltwe dragged into a maze from which it would be impossible k r us to l m e r g e . As a matter of Faiet, from this question two others anise, the Brse is what is an animal? The second, what is reasonabie? And further, if, ltcr explain what sun analmall is he were to reply that it is a living thing possessed of sensations,

that ia living thing is an animate body, that a body is a corporeal substance, you see that the question, like the bramhes of a geneaiagical tree, would go an increasing and multiplying; and finarly all these wonderfd questions wouId finish in pure tautology, which would dear up nothing, and would leave us in our original ignorance. (Descartes 1701/E93t: 31181

Circularity involves the same kind of regressus ad injinitwsr. Pascal denounced circularity with his mock-definition of fight, He wrote:
There are people who go as Far in absurdity as to explicate a word by itself. I know some who have defined f w m i a r e ("light') Bike this: ""La lumiboe est un rnouvement luminaire des corps lumineux" [Light is a Itminary movement of luminous bodies]; as if one could understand the words l w m i n a i r e and ~ ~ ~ E Y Ewithout w understand8 1 0 ) ing l~lmi8re. (Pascal 1667/1954: 5

Three centuries, and hundreds of dictionaries later, the problem of circularity has not only not been solved (in practical lexicography) but, on the contrary, has by and large ceased to be seen as a problem! The hydra of circularity is rearing its ugly heads with more and more self-assurance. There mn be no doubt that-although dictionary makers are, naturally,

9. S e m t f c s m d Lexicography 279

reluctant to admit it-what est of the reader.

is sacrificed to this hydra is, above all, the inter-

at all, and thus is free to define both evil and wicked via bad, imperfectly, no doubt, but at least without circularity:

- very wicked wicked - very bad


11. Relying on Indefinables

One cannot define everything. For any sound hexicographic undertaking it is crucial to decide which words are going to be defined and which can be raken as indefinable. The point bas been made so many times, so clearly and lioroefully, that one feels embarrassed having to repeat it again and again (see, for example, the quotes given in Chapbr !)I. Yet repeat it one must, until this basic point is generally understood and finally universally amepted. But how should a lexicographer decide on the set oF indefinables on which the dictionary /s lo be based? For obvious reasons, the set of indefinables must 'be reasonably small. For example, if half the words in ru, dictionary were defined and the other half not, the reader w d d have the right to complain, and perhaps even to demand a 50 per cent rehnd. Sccond, the indefinables must be cliosen horn among words which are intuitively dear; otherwise, they are useless (or worse than useless] as building-blocks out of which the definitions of all the other (definable) words are constructed. For example, if the words good and brrd are defined, directly or indirectly, via mom! and immoral this is useless to the reader because the former pair is by far clearer and mire intelligible to everybody (including small children] than the latter. Thus, explaining good via moral or bad via immoral is a parody of an explanation. And yet this is how dictionaries often proceed-not 'because their makers rue foolish but because they do not have a f i m and clear semantic theory at their disposal. To illustrate, the QPD offers the following: bad - wicked, evil wicked - morally bad, oflending against what is right - morally bad, wicked evil If once pursues the leads offered a little further, the web of vicious circh thickens: to oflend - b do wrong wrong - morally bad, conltrary to justice or to what is right moral - of or mnoerned with the goodness or badness of human character or with the principles of what is right and wrong in conduct This, means bad -+ wicked -+ bad; wicked -+ oJJen!d --> wso~g + had wicke& bad --> wicked + mord -+ bad; on and on. The QAJD shows more wisdom, in that it docs not attempt to define bad

Unfortunately, the same wisdom was not sffiom in the case of good, which b defined, causing, predictably, a vicious circle (and, incidentally, committing the dictionary to a dubious and dangerous doctrine that 'goody Is the same as 'socially acoeptablle'): good - of the kind that people like and praise praise - to say that someone or something is very good solution to all this is very simple: to accept that both good and bad are mong the most basic human concepts and that they neither can nor need o be defined-and then to define everything else clearlly and accwraki8y. Critics are often sceptical of the defining power of simple and general m s such as good and bad. For example, Landau (19921-3: B 151 asks how can be sufficient "to distinguish between, say, mistake, blunder, lapse, ng, and sinyy; but 1 believe that simple and general words sllpch as bad an achieve this goal much better than the unrestricted set of words used conventional dictionary definitions. Consider the set of circular definitions offered by AHDOTEL (1973) [abbreviated here for reasons of space): ht&e

- an, error or fault - . . . 4. a mistake - . . . 2. a mistake; error - 1, a transgression of religious or moral iaw, espclallly when delib- 1. the act of violating . . . 2. an instance or violation; a transgreserate; 2, amy offense, violatian, fault, or error


sion sgression - I. the violation of a law, command, or duty Pictoridly this is represented in Fig. 9.3.




L transgression
FIG. 9.3

280 Lexical Semantics

CobuiFd takes a diflerent approach, but, in my view, it is also far fram sumssFul. For example: mistake I. 1. an action ar opinion that is jmcamect or foolish, or that is not what you intended to do, or whose result is undesirable; 1.2. something or part of something which is incorrect or not right The comment on the margin says, in ddition, that "mistake = errorWIf the definitions offered by AHDOTEL are striking in their blatant circularity, those oflered by CobtddJ'd are striking in their failure to capture an invadant. ]Inmy view, however, misrukr (noun) does have a unitary meaning, which can be stated as follows:

If. S ~ ~ ~ L Tand F I Lexicography IS'CS 28 1

X knew that God wants people not to do things like this X did it because X wanted to do it
e shplicity of all the elements used in this definition [except the concept d" which is not very simple, but which underlies thc concept of 'sin'] us to avoid circularity, portray stru~turat relations, and avoid blind uch as the one in which, for example, both AHDOTEL and Cobuild mselves in their attempt to d e h e sin via law. To illustrate from a sin - is an action a type of khaviour which is 'believed to break the laws

nsissde [Xmade a mistake] something bad happened k a u s e X did sometiling X didn't want it L o happen X wanted something elm to happen X thought that something else would happen
Hy analysing the concept of 'mistake' into its components, we can not only avoid drcularity and capture a unitary meaning, but also show the diflerences as well as similarities between related concepts such as 'mistake', "Iunder', and 'sin'. For example, GanGtriSd defines blunder as 'b big mistake, especially owe which seem to be the result of carelessness or stupidity". But not every "big mistake", or even "terrible mistake", is a blunder, not even if it 2 s due to carelessness. Something like "stupidity'" on the other hand, is a necessary part of this concept (so it shouldn't be introduoed in an '%specia1111ym frame]. I would propose the following:

- 1. is a system d rules that a smiety or government develops over

time in order 'to deal with business agreements, social relationships, and cldmes such as theft, murder, or violence.

dy, the definition of law quoted above does not allow for any "laws of " ~ the 0 attempted dehition of s i n via h w does not allow for any
interpretation. ,a few words about wrong (adjective), for whkh AHDOTEL finds ged to posit no less tban seven different meanings, and C~buiM, many as eleven, without being able b show what all these supposed ngs have in common. Of murse, one could write a whole study about noepts of 'right' and 'wrong', but basically, the meanings of these two adjectives are quite simple. In essence, they could be stated as follows: It is wrong to do this (like this). = it is bad to do this Oike this) if one thinks about lit one can know it It is right to do this (like this]. = it is good to do this (like this) i f one thinks about it one can b o w it Unlike the concepts of 'goods and 'bad', wwhh are universal, the concepts of ' r i g h t b d 'wrong'are culture-specific, and in fact they are very revealing in the links which they postulate between values ('good' and 'bad') on the one hand and 'thinking' and 'knowing' on the other (see Wierzbicka 19X9a; 19920, ch. 1). A list of indefinables that has proved itself valuable in lexical senrantics should be of great potential benefit to practical lexicography. Since the same indefinabSes, and the same simple syntactic patterns, which appear to be mast useful for ananysing the English lexicon appear to be allso wry usefull in anallysing the Peximl resources of other languages (including ones as diverse as Japanese, Pitjantjatjara, and Ewe), the ~onclusiondoes not seem

blmdes ( X made a blunder) something bad happened because X did something X didn" want it to happen if X thought about it For a shorr time, X wouldn't have done it people can think something bad about X because of this
Landau doean" believe that a general! and simple word like bad could Ise used in defining both a word like mistake (with no moral or religious h p l i cations) m d a m r a l and religious concept like sin" But if E am not R Y j l s taken, and if I am not sinfully over-confident, it can be done. Here is my sin (in the serious, non-jocular, use of the word):

sin (Xcommitted a sin) X did something bad X knew that it was bad to do it

282 Lexice~l &man tics

premature that the same set of indefinables can be used as a core ofa "natural lexicograplnic mctalanguagc" swilablc ror both mo~l~olinngual and bilingual dictionaries in most, if not all, Irmguages of the world. The practical advantages of such an outcome would seem to me to be huge, as any Ianguage learner would thus have relatively easy access to dictioruaries of amy language via their common coree2

9. Semantics and lexicography 283

obligatory - everyone has to do it optional - one can do it if one wants to, one doesn't h a w

tio do it

12. Using Simple Language

The issue of indefinables is linked closely with that of simple language. Thle use of excessivelly complex m d obscure language is one of the greatest obstacles to effective cornmication in m y area of human endeavour; brut in a dictionary, which s&s to expkria the meaning of words, it is particularly out of place. k n a u l d refers, in this mnnection, to the writings of tbe philosopher and mathematician Gassendi, m d he wribs: "Gassendiasexposition makes clear that there is scarcely a more reprehensible t m of mind than is exhibited by these enigmatical writers who believe that the most groundless thoughts-not to say ttne most false and impious ones-will pass for grand mysteries if reclothed in f o m s unintelligible to the c o m o n m n ' " [Arnauld 166219164: $8). Despite repeated pleas from thinkers like Gassendi, the use of complex and obscure language is a great plague of Western civilization, which mars, in particular, most encyclopaedias, textbooks, manuals, printed instnnctiom Ebr the use of machines and devices of dinerent kinds, and so om. M m often than not, it also mars, and diminishes the usefulness of, dicieionarim of diflerent binds. Consider, for example, the definitions of the words obFigatory and crpltbnol, given by AHDOTEL: obligatory - 1. legally or morally constraining; binding; 2. imposhmg or rseordhg a m obligation. 3. of the nature of an obligation; compulsory. optional - left to choice; not compulsory or automatic wouldn? it be better to explain what these words mean in very simple words, along the following lines:

@'m not proposing these as fully accurate definitions, but only as an improvement on those offered by the dictionary.) My respect for the work of practical Iexicogqhers is so high that I would not wish to be seen as accusing them of passing (as Gassendi put it, according to Arnaud] "groundless and the most false thoughts" for "grand mysteriesm&ut I think it is fair to demand of them that their definitions avoid k i n g "clothed in f o m unintelligible to the common man".

13. Exploring New Models of Definition

xicographic definitions can be improved immensely while maintaining a or less traditional f o m . This can be done, above all, by simplifying egularking tlne language of definitions, by using a discrete model of tionrs, by trying to capture invariants (and thus banishing all ors and eras), and by getting rid of circularity. It is possible, however, to ove definitions still further, if one is prepared to give up the traditional! s of definition and to explore new formats and new models (drawing on the discoveries olF,contemporary semantics). Consider, for exampje, the LDOTEL entry for the verb so p~rms'sh: la. to impose a ~w:nalty on for a hult, offence or violation lb. to inflict a penalty for [an offence) 2. )totreat roughly, harshly, or damagingly Many weaknesses of this entry are quite apparent: the positing of a completely unjustified polysemy (1 versus 2) and semi-polysemy (la versus lb]; the failure to capture the invariant (what do "fault, ofince or violation" have in common?); the latent circularity. The curious distinction between "impose" [la) and "in63ictY"r(lb) collapses in the entry for inpict, which is defined as 'Yo hrce or impose (sornethimg damaging or painful) on someone". Penalty is defined, predictably, via pam&krnent: "punishment imposed for, or incurred by, committing a crime or public offence1'- And if that much circularity was not enough, more is introduced via crime, which is defined as "an act or omission punishable by law'"unth ; . pemcrify + pmbhment; punish + penc~ity+ crime + pmi~h). Here as in many other cases, the simple and unitary definition offered by OAJD is incomparably better:
to punish-to
to d o

I a n not suggesting lrhat the Natural Senantic Metalang~age devised by the author c a r n d colIeagues should be wed, unaltered, as a lexicographic metalanguage, but onUy that It a n ke used as the core of a iexicographic metalanguage. Nor am I suggesting that all bxical item in a dictionary should b defined directly in terns OF the indefinables. In particular, the n m oC natural kinds (e.g. cgt, ps~oort~c, burlerfly) or af culliural kinds (e.g. iGortC, bicycle, chaw require a different approach, with a much larger defining lexioon than abstract vocalbula~ does (see Wierzbicka 11985; dso Ch.7).

make someone who has done wrong suiFfer, so that he will not wanti wrong again

284 Lexical Semantics

In fact, this simple short dediniiion can bc made even h t k r by shortening it further: since one can speak of "capital punishment" (by death) or of "eternal punishment" [by hell), the corrective purpose cannot be a necessary part of the concept. We are left, therefore, with the short formula "to make someone who has done wrong sufler"', and this is probably almost a$ good as one can get within the traditional lapodell. The definition is no more, however, than an approximation: it does not capture correctly all the camponenits of the concept, and in some respects it manages to be over-speciik. In particular, it posits suffering as a nemssary part of punishment, whereas in fact an intention to cause suffering is sufficient, even if no suffering actually occuw. What is missing from the definition is some indication of the: relationship between the punisher and the punished, and of the punisher's view of the action as morally justified. For example, if a little boy (Johnny] hits his younger sister (Smie) on the head, and Swzie retaliates by biting Johnny's finger, the OAJD definition would fit the situation (since Johnny did something wrong and Swie made lnim suffer because of it), but the word pmish would not. To portray the concept of 'pwnishnnent' accurately we need, I think, a scenario, not a definition of the traditional kind:

9. Semantics and SexicegrqAy 285

As a second example of Lhc need for a scenario, consider the concept of 'revenge" which the OPD defines as follows:

or injury inflicted i n return For what one has suffered

The definition is unsuccessful for many reasons, two of which can be linked with the use of the words punisFfimemt and rerrrm. Contrary to what U h e definition implies, revenge is not a kind of punishment, because it does not imply the assumptions which as we have seen are part of the latter concept. The expression in return is not defined at all, and the definitions assigned to the noun retzdrm are useless and irrelevant from the point of view d defining revenge (e.g. "coming or going back"). Again, the definition offered by the OAJD is much more satisfactory ("a wish to hurt someone because he has hurt you or one of your friendspy)il but is; not quite correct: rewnge refers to an action, not merely to a wish, the mendon of "friends'2s swpefluous, and the crucial idea of 'Voing the same" is missing. To portxay the concept of %evengay accurately we need a scenario:

X punished IY [for 4. = I ( . I Y did Zll (b) X thought something like this: (c) P did something bad [a (4 I want Y to feel something bad because of this (el it will be good if Y feels something bad bemuse of this it will be good if P do something to Y because of this (g) X did something to Y because of this


Y took revenge on X [for Z]. = (a) someone ( X ) did something bad [Z'j to someone ( 0 (b] because of this, Y felt something bad (c) after thh, Y thought something like this: (4 this person (A) did something bad to me (el because of this, I want to do the same to this person (X) (f) P thought about it for a long time (g] after this, Y did something lbad to X because of this
Component (a) refers to the action of the offender, and lCg) to that of the evenger; (b) shows what the revenger felt, and components ( 4 and Q e ) hew his or her thoughts [with their focus on "paying in kind"). Finally, consider the concept of ?temptingy,which Web$~ster$ Dictionary 9) "defines" as follows: "temps-to put to trial; test; persuade to evil; ; allure; entice". It is hardly niecessay to point out that this entry does t tell the reader whether the verb tempt is supposed to have one meanor more, and if more, then how many; that no attempt is made to cape the semantic invariant or invariants; and that the entry offers no clues to the differences in meaning between all the different verbs which it s as supposed equivalents of fempt. It does nat require much imaginaion to guess that the same dictionary will "define" entice and a h r e via As pointed out earlier, a much more illuminating alternative is provided by the BAJD, which offers the simple, short definition "to fempt-to try to make someone do wrong". But of course this is only an approximation; for example, one can "'tryto make someone do wrong'9y threats, and this

Component (a) refers to the culprit's actian, (b) to ilfr) describe the punisher's attitude, and (83 refers to the punisher" action. The punisher's attitude incl.udes, roughly speaking, a desire to inflict pain ( d ) , and three assumptims: that the target person did something bad (c], that it will be right and julst if Ine or she "suffers" [feels something bad) because of \this (e), and that the punisher is callled upon, to inflict the necessary pain wlprcsumably, as the person in charge). Cruse (1992-3: 89) questions my analysis or thc verb psmi.rh on ithe grounds that "the punisher may actually hate having to cause sufferinggL, 1 entirely agree that the punisher may hate having to cause suflerlng; but this is not incompatible with an intention to cause suffering. For example, the father or mother imposing the punishment on the child may s u f i r intensely themselves; but if they didn't intend to cause some pain for the child they wouldn't be "punishing" him or her. On the other hand, if the child doesn't really feel any pain this doesn't stop the parents' action from being d e m i b able as punidrrrmesrt.

286 Lexica/ Semantics

could not be called tempting. To portray Rhis concept adequately we need, 1 think, a scenario along the following lines:

X tempted Y to do 2. = (a) X wanted Y to do Z (i5) ' l thought somethilrg like this: (c) if I do Z it will be bad [d) because of this, I don" t w i t to do it (e) X knew this [J)because of this, X said something like this to E 6g3 if you do it, something very good will happen to you [h] you will feel something very good because of this (fi X thought something like this: maybe 3 ' will do ii t u s e of this (j') (k) X wanted this

The Meaning of Colour Terms and the Universals of Seeing

Phemomenological analysis . . . is analysis of concepts and can neither i t h mar contradict physics. agree w
(Wittganstcin 1937: 16)

I am not suggesting that new models of definition such as those illustrated here with rewage, psilnisfisrrent, and tempt should be nemssarily accepted in dl practical lexicography, although I think it wodd be useful to adopt---or to adapt-them for some types of dictionaries. But [I believe it is ~ s e f b ~r a the practical lexicogaaphers to know that new models o f definition are available-and to let them draw on whatever is availablle in ways which they would judge most appropriate in any given case.

"dour' is not a universal human concept. It can of course be created in all human societies, j~ustas the concepts Ltelevi~iona, 'computer', or honey" can, but despite the rapidly increasing contact between human societies illl many which have neither borrowed nor developed the canuryand of course there have been many more such societies in are ""colourterns" a universal phenomnonr. It is certainly not true as often been daimed, all languages have words for black and s point will be documented and discussed in detail later, but in a sense it is quite obvious: if a word is used to describe not only black, but also brown, grey, or dark-blue objects, then it cannat possibly mean %blacky. sh, and in many other languages of the world, 'kolour" can be arded as a reasonably self-contained semantic domain. But in the mise of hiurnan discourse it is not. To try to carve out, for all languages, a "coolaur semantics" means to impose on the study of all cultures arising from only some of them (in particular, from modern, y complex, Western culture). ltures, people are interested in "seeingY'andin describing what they see, but they don" necessarily isolate "colour" as a separate aspect of their visual experience. All languages haw a word for SEE (see Chapter 21, but not all languages haw a word for "colour'? From what we know about the vocabulary of "eeingyn different languages, we can conclude that in ures the discourse of 'seeing' is contextualized, and the experience is described as a complex and integrated one, with colour, texture, lion, and many other attributes being treated as an indivisible
This chapter owes a great deal to many long discussions with CUiff Goddard.

14. Conclusion
There is more to practical lexicography than getting the meanings right, but trying to get the meanings right is vitally important-more importmlt, I think, than anything else. If theoreticall lexicography could not help i n this respect-by providing ideas, principles, criteria, models, and guide-line+ onc could really doubt its mliscrn dlt;rre. I have tried to show, however, that theoretical lexicography indeed offers a31 these thing^.^ Most imporbntb (from the present writer's point of view), lit oflers a too1 which can by itself remedy a large proportion of the ills of traditional lexicography: a natural Ilexicographphic metalanguage, derived from the Natural Semantic Metalanguage, and based on universal semantic primitives.
For briEliant di5cwi*mr of various aspects of theoreticall lexicography and of the principles of Pexicographic definition, w e in particular Apmsjan (1974, 1992, foraskhcaming]; Mel'Euk (1974b); and BogwsEawski [198%]. For discussion or the lexicography (PS the conclete Iexicotu, see Wienbicka (19851 and Ch. 8. For some recent dictionaries which orer~onale the indeterminacy and the circularity of trardiiiona! Ucxicography, and wlriclh aim a1 enjpirical adequacy without departing from traditional models, see e.g. Apresjan and Rozenrnan (1975'9; Boguslawski S1983); Gddard (U992a]. For a new model of a monolingual dictionary, see Me16Cukand Zonkovskij (198411; Mel'Euk e l or! (1984, 1988, 11992); Rwdzka ef el! (E481).


whole. As van Brake1 (1993: 1 13) writes, '"n western languages, the domain d cdour is clearly separated from other categories and there is a bias towards hple at the expense of brightness anrd saturation. In other cultures, the hue aspect of colour may, as it were, be subsumed under dflerenlt categories, so that it isn%really present as a separate domain." 1 do not doubt that there are some "universals of seeing"', which mn be discovered and validated through the study of the world's languages. But to establish what tlnese universals are, the focus of research must shift from the search for "colour universals" to the search for '%niversaEs of seeing", There can't be any colour univ~enals,if colowr itself is not a human miversal. But 'sedng'is indeed a universal human concept. To say that the search for colour universals has been largely misguided is not to say that it has not been fruitful. The massive research into the description of colours initiated by Berlin and Kay's (19641) classic study has generated a great body of knowledge about the discourse of SEIEing, and has contributed a great deal to any future theory d the universals of seeing. In particular, this research has made it clear that the notion off colous is not only far from universal but that its hale in human discourse is, relatively speaking, quite limited. What does seem universal, or near-universal, in the domain of seeing is, first d all, the distinction between times when people can see ('Uay") and times when people cannot see ("night'". This universal or near-universal distinction between, roughly speaking, nights (dark times] and days (times of light), appears to be linked, universally or near-universally, with same distinction or distinctions in the description of what one sees. Roughly speaking, people tend to distinguish, universally, between seeing things which look "light" and '%shinym and seeing things which look "'dark" and ""duaa (that is, light-less, shine-less). Cleady, the first kind brings to mind the experience of ""sun-timevision", and the second, that of "night-time vision". (It is worth recallling in this connection Birren" (14178: 3) observation: "'A111 civilizations since the beginning of man's existence worshipped the sun, and from the sun came light and co1o~."3 The distinction between "dark" colours and ''light''colows appears to play an important role in most languages of the world. For example, Hargrave (11982: 208) makes the following comment about the Australian language KKnnku: Yalanji (a "stage I1 language" in Berlin and Kay" cotourencoding sequence):
According to the investigators, bingaji and ngtnmbw mean 'light' and 'dark' as well as 'white' and 'blmk" Several participants appeared to name chips Blight or dark ia comparison with the frame around the chip or in comparison with the &.hip, just shown them previously.

Hargrave adds:
Terns denoting light and dark have beem recorded by other researchers in Australia. Jones and Meehan, carrying aut an investigahiom of Anibarra (north-ceoemutral Punhem Land) colous concepts, concluded that there were only two real coIour terms, those ror light and dark. Four additional h3colouur terms' were names for mimeral pigmemts and could only be used to describe a limited range of objects (Jones and Meehan 1978: 2630). Davis faund that children at Mdinghbi, also in Punhem Land, first classified all colowrs as vvatharr 'light' or mol "ark'. As they got older, they added further tems which classified ~olours by hue and saturation as well as brightness [Davis 119821.

Another universal or near-universal has to do with the impartanoe of the environment as a fundarnenltal frame of refemce for any human descripdon of 'seeing'. English words such as vfew, scenery3or landscape provide useful hints in this respect, since they link the idea of %seeingy with the idea of "lace'. For what do human beings normally '%ee"? Presumably, objects, m h a l s , or people positioned or moving against a background (cf. the "figure" and '"round" distinction in psychology). Of the two, backgrounds are m a doubt more stable and more predictable than "figures'? the sky (often blue], the ground (often brown), the grass (typically green), the sun (often yellow and brilliant), the sea (often dark blue), the broad expanse of snow (normally wlnite)]. Of course, the landscape doesn't look the same everywhere. Not all human beings are familiar with the sea nar with snow, the ground is not evebywhere brown (and in some places it may be seen as predominantly red, yeifow, or black), and even the greenness of the grass depends on the availability of water and on the exposure to sun (e.g. in Australia the grassy landscape is, typically, yellowish or brownish rather than green). I am proposing, however, that the principle of using common features of landscape as a friarme of reference for visual categories in general, and for 'colour3n particular, is an important human universal, and also that this principle is responsible for many recurring features of human discourse on 'seeing', Yet another universal or near-universal feature of human discourse on seeing is the important role of comparison, or, more precisely, of the universal concept LIKE, in the experience of visual experienmw. The English adjectives goEd and gotden Illustrate this mode of description very well, and so do numerous other "non-basic colour terms" such as silver, navy blue, khaki3mh blond, and so on. Another example is provided by some of the main colour t e r m in the Australlian language Warllpiri: y~11ys~-yabu, 'red' (lit. 'bbnod-blood'), karntawarclr, 'yellow' (lit. 'pllow ochre'), which, together with two "environmentall" terms, waEya-waEyos, 'brown' (lit. "earthearth'] and ytrlk~ri-yukusi,'peen-blue' (lit. 'plants-plants"$ and with something like UdarVlolack' and 'lightfwhite', form the core of the Warlpiri ''co1our''vocabulary [Hasgrave 1982: 23 (13).

10. The M e ~ n i ~ o fgCoIour Terms 291

But miversais or near-universals such as those mentioned above ccsulld not be stated within the Berlin and Kay (1969) framework, with its emphasis on ""bsic"colour terns. Shweder and Bourne (1984: 160) described Berlin and Kay" (1969) theory of "colour universals" as an example of the "data attenuation rule"". They wrate:l Not infrequenlly, the discc~weryof a universat is the product of a sophisticatedl pPQ6lesSof data restriction and data attenuation. Berlin and Kay (1196911, far e m ple, discover unIuversal prototypes for the definition of color categories, and a miversa1 sequence For the emergence of a color lexicon. Their study begins w i t h twsr app1icrutions ofthe data attenuation rule. First, 'ec~lor" ccllassifimtion is equated with the task aF partitioning a perceptual space, predefined in terms at" hue, saturation, and intensity (thus, attenuating the referenda1 range of the ""co1or"wncept as understood by, at least, some cultures [Conklin 19551). Smorrmdly, all1 collar categries whose linguistic expression fails t0 meet oertain formal criteria (e.g., snperordination, monolexemic unity) are eliminatledl born oonsideratian. The consequence of the applicdion of these two data attenuation rules is that 95% of the world's expressions for color and most of the world's mlar categoties are dropped from the investigation. In the intervening time so much counter-evidenm to Berlin and b y ' s theory has been presented that one could no longer say that they discovered "universal prolatypes for the definition of colour categoriesy' or "a uniwersal sequence for the emergenoe of a color Bexican" (see e.g. Kay eS 01. 1991; MacLaury 19187, 1992; Hewes 11992; Kinnear m d Deregowski 1992; Saunders 1992; Toren 1992; van Brakel 1992, 1993). 1 suggest that if we wish to discover, and to explain, the universals of human discourse on seeing, we must, SO to speak, look in a direction different from that chosen in Berlin and Kay's (1969) classic and further explowd in the huge body of research b ~ l on t the foundations laid out in that wok.

2. Meaning and Scientific Kncawledge

The hardest things to observe are those which one sees every day. (""H faut beaucoup de philosophie pour savoir observer une lois ce q u b n voit taus les jours"; Jean-Jaqws Rousseau.") The question "'What do words like red and blue mean?" may sound, to many, offensively foolish. But in fact, it is
"f. also van Brakehi (1993: 1112) comment: 'kll subsequent work in )the Berlin and Kay trdition has 'been carried out with Mwnsdl Ccalom Chips and standardized procedures to elicit B a s [basic crolour t e r n ] . lt has been estimlled that in doing ahis 95% of the nodd's mlour terns are eliminated." I, borrow LMS quote from Moore and Caslling 1019821, who used ill as one of the mottoes of their hook.

very difficult Ito answer. Although the psychological, anthropological, and linguistic literature on colour t e m s is very extensive, it usually addresses itself to other questions. The simple, "naive" question raised here tends to be largely owedooked-as simple and ""laaTveY'questions concerning our everyday experience often me. It is, of course, true that the meaning of colour terms has often been discussed by philosophers, and linguists and psycholagists can profitably draw on the writings of thinkers such as Loclce, Hume, Carnap, RusseUl, or Wittgenstein. But the philosopher's perspective is necessarily different from that of a psychologist and, even more, from that of a linguist. The crucial difference is that philosophers are interested in language, whereas Linguists [qua linguists] are Interested in languages. To a linguist, the problem is not only to discover what the (English) words red or blue mean, but also what the Hungarian words vdkfi3 and piros (roughly, types of red) mean, what the Russian words goluboj and (roughly, types of blue] mean, what the Polish words niebieski and granatowy (roughly, allso types of blue, but different from the Russian ones] m a n , or what the Japanese word aoi (roughly, blue, but much broader in range than the English blue) means. Glosses such as niebieski = blue or iaoi = blue or sing = blue will clearly nolt do, since the range of each word is language-specific and cannot be correctly established on the basis of interlingual matching procedures of this kind. ]But if miebieski, singy or aoi do not m a n the same as bIue, what do they mean? And what does blue mean, in the first place? To some scholars, questions of this kind may seem foolish, because they believe that the meaning of every colour tern can be identified in tems of physical properities of Bight such as wavelength or relative energy. For example: ""When the wave lengths are between 400 and 470 nm [nanametres, lP9d,the field is reported to look violet for an average light levell; around 475 nm it is seen as mostly bllue" (Hurvich 1981: 391). In fact, however, scientific knowledge of this kind is entirely beside the point, if we are interested in meaning, and if by meaning we understand, essentially, what people mean when they use the words in question. Clearly, when someone says a blue dress, niebieska ('FEM) sukienka (Polish)l, or sinee ( M E E P T .pSat8e (Russian), they may have no idea what wavelengths, or what relative energy, are associated with the words &he,yriebieski, and sin& and yet, surely, it would be foolish to conclude from this that they ddn't tnaw what these words mean. Scientific knowledge of wavelength associated with different colour terns is valuable in a textbook of physics, but when it is repeated in linguistic books and articles and presented as if it were an answer to questions about meaning, it only clouds the issue and stands in the way of our search for a real understanding of what people mean when they use these wards. As

294 Lexical Semantics workings of our minds may, indirectly, reflect this; but the conceptualhatlons in our minds must be Linked to something that at constitute the canLent of our thoughts. The faith which some scholars have in the relevanae of nemrophysiology to the study of meaning can only be equalled by their faith in the retevar~~ce of formalisms. For example, Kay and McDaniel write: We have found furUher that tlhe facts of colour semantics are modelled Felicitousky in fumy set theory, and are not readily modelled in the traditional theory ol discretely contrasting semantic features. This finding casts doubt on the general usefwllness of the Feature model, and suggests that mare powerful formalism, employing a range o o S strwtures much broader than the restricted Boolean algebra implicit in the discrete semantlic-feature approach, are probably neoessary to provide realistic accounts of the semantics of words. (1978: 6441 The full title of their artic6e reads: "'The Linguistic Significance of the Meaning of Hasic Colour Terms". This isianplies that the authors know what the meaning of basic colour terms is; and that they me going to b d d on that knowledge (which, one is to understand, has emerged from the neurophysiological research reported in the arlicle). But all1 the reader is told, al the end, is that ithe authors believe that the facts of colour semantics can ibe modelled felicitously in fwzy set theory, and perhaps in some other "powerful mathematical formalisms"'. In my view, if some scholars are interested in translating linguistic ffacts ~ I Y L O "powcr~~U mathematical formalisms'"sucl~ as, for example, fuzzy set Iheory) they have every right to do so, but I do not think that I b l y doing wa they are bringing us any doser to discovering what words mean.

l i O l The Meaning o f CoIour Terms 295

w e d as follows: What do words such as bhe, niebieskd, or gohboj mean? Simple: we can show this by circling appropriate areas in a universal colour chart. Far wordis such as Mue, niebfeski, and gsrlubaj these areas may overlap, but since they will not coincide with one another, the language-specific range of each word's use will be correctly amounted for. Bwt Berlin and Kay achieved the apparent suluiess they did because they were investigating not the meaning of colour terns but the interlanguage stability of colour foci-and the method they chose seemed initially appropriate for the task which they had set themselves. They saw clearly, however, that tihieis method was totally inappropriate for the investigation of colour bowndairiex. Thus, they wrote: Repeated mapping trials with tbe same informant and also across informants showed that category foci plamrrranls are Rigluly reliable. . . . Category boundaries, however, are not reliable, even for repeated trials with the same informant. (Badin and Kay l969: 13) They concluded:

ia is possible that the brain's primary storage procedure for the plsysical reference of coEour categories is concerned with points (or very small volumes] of the colour solid rather thm extended volumes. Sacondary processes, of lower salience and intersubjective homogeneityswould then mount for the extensions of reference to points of the croloue solid not equivalelvK to (or included in) the focus. Currenl formal theories of lexical definitions are not able to deal naturailly with such phenomena. (itaid.)

3. Meaning and C~plepurCharts

Another popular approach to the semanltics of colour t e m s is based on the identification of meanings with denotata. One recalls in this connection the scholars fmm Swift's G~IIiver'sTravcb (1728), who believed that verbal explanations could be replaced with the demonstration of denotata, and who carried everything they wanted to talk aboult on their backs. In the same vein, it is proposed that instead of defining colour terns in different languages we can simply produce samples of colours themselves. In particulm, great faith is placed in commercially produced colour cbips such as those which were used by Berlin and Kay (1969) in their investigation of universals of colour naming. To some linguists it seems self-evident that the method which initially at least seemed to produce so much insight in Berlin and Kay" investigsltiolll of colour unniversals provides also an obvious solution to the problem o f the meaning of colour terns. The attitude of these linguists can be par-

I believe that in 11969 this conclusion was correct; and that the concomitant decision not to pursue, at that time, speculations about the meaning of colour t e m s was prudent and justified. I think, however, that in the course of the intervening twenty years, enough progress has been made in the areas of both the theory and practice of lexicall definition to enable us to tacklie the protPaem which in 1969 may have-rightly-appeared inltaactable; Berlin and Eay had every right to limit their attention to colour foci, and to choose not to explore colour boundaries. But if we wish to reveal the concepts encoded in the colour ]lexiconsof different Eanguages of the world we have to take into account both the foci and the boundaries. But to return for the moment to the possibility af "showimlg'We meaning ofcolour t e r n in colour charts, consider also Frumkinaysobservations:
any colour model is characterhued by some degree of indeterminacy, as far as its possible naming is cmcermed. Often, people who are not professional colour experts . . . i.e. who have nothing to do with the science of callour or with other areas of howledge where precision in the naming of colour perceptions is important, will not be able to find any intuitively satisfying conour terms to designalte a given 'colonrr madel'. In other cases, they will propose several tems for one colour model. Since in practice there are situatioms where denotlaliive indeterminacy of colour

296 Lexical Semantics

designations and 'naming' indeterminacy of colour models may be very inconw nient, specid nomative charts are produceded, which show what lcolanr designations should be given to a given colour model. For example, the charts of the Bditig;b Colour Coundl, 1939-11942, have such a nomative character. Charts of this kind have purely pragmatic gads; for example, to achieve mutual understanding in the description of different genres and species of plants it is necessary to ensure, in an artificial manner, one-to-one correspondence between colour designation and colour modds, despite the fact that in natural languages the correspondence is commonly of the many-to-many type. The charts of the British Colonar Council, just like other normative charts, are a terminological guidebook, whose validity is strictly limited to that do~nain for which it was prepared, so that, for example, the nomenclalt~reol colouc designations for aolou~photo-reproduction (i.e. the system d pairs: coiour designation-colclur model) requires already a separate guidebook. (Frumkina 1984: 26) Frumkina concludes: The problem of colour naming, that is OF assigning callour designations to s p G c coPour models, deserves separate investigation as one aspect of the problem oFnaming in general. As for the possibility of describing colour designations by means d pictures, the fact that the relations between cogour designations and colour mdels are o E the many-to-meny kind makes it rather unrealistic. (1984: 27) This echoes Conklin's f1973: 940) remark: "There is obviously more to the study of colour categorisation than the matching of spectrographic readings with human verbal responses." h o u l d add lo this that the use of pictures and CIO~QPR~I charts can be useful in tbe investigation of meaning if one makes proper use of them, without placing unreasonable demands on them. They cannot automatically show the meaning of a colom tern, but they may help establish what the meaning of a colour term is. For example, Jones a d Meehan Q1978), who investigated the use d the two basic colour terms (-gungar&@ and -gungmt$a)i in the Australian Aboriginal language Gu-jingarliya with the help of the Munsell charts, obtained results which are indeed highly instructive b i n a semantic point of view. But they are instmdive because they raise fascinating questions, not because they field any ready-made answers. Equally instructive is the fact (which the chart itself cannot show) of how the data were obtained:

10. The Meansing o f G'oI~wrTerms 297

The question of w h d the words -gefngaI&ihand -gmgwndja mean is a fascinating one, and 9: believe it is the kind of question that is more pertinent to linguistic research than any questions concernling the neurophysialogical bases of mlour perception, important and interesting as the Matter may be their own right. I will return to this question later, after I have: discussed e meaning of the English coilour terns white, biarck, bhe, green, red, and law, and their closest counterparts in a Few other languages with comx colour lexicons. First, however, I should like to clarify what 1mean by eaaingy",and how "'meaning" is related to "psychological reality".

4. Meaning and Psycholagicd Reality

leaning of a word can, roughly speaking, be defined as what people or have in mind when they use it.' Since what they mean or have in may differ somewhat depending an Icrantext and situation, we should that 'beaning" has to d o with the constant, not with the variable, of ;a word's use. These constant aspects can Ibe ascertained in a variays, indwding a methodical introspective study, a study of common ology, w m m m metaphors, questianing of informants, psyi?noQintic experiments of different kinds, and so on. All these methods reveal in the speakers%nds, words are mutually related in different ways, they allow us to establish, how tiley are related (see Wierzbicka 1985: For example, Frumkina (1984: 30) reports that she asked a number of infommts to explain to her "what pink ( r ~ z o v yis" ~ ) and that she obtained m them the foPlowing E n d of answer: '"ink is a very3 very light red cdour, quite light, but sumciently specik for people to be atale to it is similar to red or has a shade like red." In Frurnkina's s e w , swen of this kind help us to reveal ' U e organization of memings in the linguistic consciousness". 9: believe this is correct, and I think that ate semantic description of the word pink (or rolzovyj] should intuitive link with the concept encoded in the word red for kmsnyjl and with that encoded i n the ward light (or sve~J'yj]. It is important to keep in mind, however, that "linguistic consciousness" has many diflerent levels and that while there are facts which lie on its very surface there are others which are buried deeply, ellien very deeply, under the 5urface. As was stressed forwfully by Boas (1911/1966: 63-41 and Sapir (11949: 416-71, and more recently by Halliday 1(1987),native speakers'
Of COIPCSR one could also define the tern "meaning" in many other wags [see e.g. Ogden and Richards 1923), and P have no desire ta argue about terminology. But the question o f "what people mean'"when they use a particular word) is clearly a very important and worthwhile question t~p ask. Obviously, nemophgsiology cannot answer questions like that.

first, Gummmnana (the in8;orrnanlt)said these were no -gamgaIoa colours them at U d and pointed from the chart to a piece of reflective foil used for coding, Igbg on a bench in the bnt. 'That one here, properly number one gum-grtmgrelfjia,no mare this mob.', . . Having made his proibeslt, Gnrmanamana then proceeded to outline colours as shown in Fig. 2 [not reprothe approximate boundary af the -gu~gaIsjra dumd here]. It can be seen that only about 1 0 1 % of colour chips are included in this category, the main bulk of the chart belonging to the -gangundjaclass. (Jones and Meehan 1978: 27)

10. The Meaning oifcCaIo~r Terms 299

knowledge of their language is, by and large, subconscious. It is important to distinguish tacit knowledge, which is hidden ""ithe depths" of a, person" mind but which can be dragged to the surface (see Sapir 1949: 3111, from scientific knowledge, which native speakers simply do not have rand which no amount of searching could ever reveal. The latter, in contrast ta the fcrrm~er, is not reflected in lanjg~ageafi~d plays no role in ttre li~iguisajc piatlcrning. (See Cllapters E l and 12.) Semantics is a search Ibr meaning, not a search for scientific or encyclopaedic knowledge; but this does not mean that it is concerned onlly with facts whicli lie on thc surfacc, or vcry near L l n r swrracz, or spcakers' cansciousncss. O f we confuse '~syclrologicalscality" (sce Buding 1469) with "consciousness" w shall never find out what goes on in people" minds a d what conceptualizations are reflected in human languages. Let us consider, for example, the following facts, reported by Erumkima (1984: 30): ""W~lilcfor Russian infornuants, rlozovyj 'pink' and krasnyj 'red" arc sirtrilal- in colour . ., feltyj 'yellow\afild kuriZnrvyj "brwnhre [far tlicm) simply different colours, j~usltas different as krasrayj 'red' aandfi~iesovyj "urple'."Shoonzld we then con~lude from this that an adequate semantic descriptionr of Russian colour terms should present krastryj and$loSessrvyj as totally z~nrelatcd-as unrclaled as, say, ZeEfyj and ,;fr'oS~fovyj, or zelenyj, "recn" annd jiu/e~o9yj? I tlzink thal we should not. It is important, B believe, not to draw hasty conclusions from anything that informants may tell us. Rather, we should take tl~eirinitial responses as one kind of evidence, to be used jointly with other kinds of evidence. Informants' responses should never be taken at face value; they should be interpreted and made sense of (see Wierzbicka 1985: 89-90; see also Chapter 12 Section 21." -and we have learnt, on this basis, what people mean by bhe. Bhe means 'what people call blue'." 1 believe that an answer dong these lines is probably correct, as far as it goes (although, as I shall argue shortly, it does not go very far). In particular, it is important to note that an answer along these lines is not circular, as it Is not circular lo say that, for cxannplc, llhc word JoSfjlSe~r i u ~RPIC senlencc referring to some: particular person called John, means, roughly, "'the person: whom 1 call JOHN" (where JOHN, in capital fletters, refers not to a person but a sound). Assuming, them, that colour terms are Ileaart, essentially, by ostension, and that their mcaning reflects this, we could propose, as a starting-point For further discussion, the following explicaltion: X is blue. = people say of things like: X . "this is BLUE""

A formula of this kind represents the colollnr term as a kind of proper name:
it implies that just as the word h h n means, essentially, ""Le person called

JOHN'', the word !due mleans, essentially, "the colour called BLUE"".
S h m to understand a word like blue one must know that this word has something to do with seeing (rather than, say, with hearing or tasting), we could expand our first Formula as follows: Xis blue. = when people see things like A' they say of them: this is BLUE It seems reasmuable to suppose that a formula of this kind may reflect the child%first meaning of the w n ; c Pbhe; and it is important to note that to sketch such a formula we do not even need the word clobur, which, one must surmise, would be normally acquired much later than blue or red. (As pointed out by LeikPniz (19616], the concept of cdour is not indefinable: it can be defined via seeing, since colour is the only property which we can perceive only by xeeingn5) Since, however, the formula sketched above refers to some specific models (such as a particular object "A"" which would be different in each person's individual experience) without attempting to draw any boundaries, it cannot have full predictive power with respeclt to an adult's use of the word bhe. N t e r all, ithirugs that people call greeFa or purple can also be seen as similar to those which they ucaill biue-and yet mature speakers of English do nolt extend their use of the word biue to objects which they call green. As it stands, the formula sketched above dms not account for this. In learning a second language, we often acquire the boundaries with the help of negative feedback. For exampie, my daughters, who are bilingual
The concept of 'colour' is actually quite complex, and I will not attempt a dehition here. It is clear, howelier, that a definition of 'collour' would have ni@r be based on the concept of

5. Coloulr T e r n s as Quotations
Once again, then: what do people mean when they say, for example, "I bought a bkue dress" or '" saw a blue car"? One "ciommon-sense" answer to this question takes the following form. 'Tolour terms are learnt ostensively, and their meaning is also based an ostension. We have all heard the word blue applied to a variety of objects
It should be emphasized, however, that "folk comments'" as well as folk definitions, can provide precious insight into the meaning oF colour terms. To see this, consider, for example, an informant's comments on the word hyi-hyi, 'grey', in the h t d h language Warlpirk " M e n a gum tree is first in good condition, in Its Foliage, it is first green. But if it should then (die and] dry up, its leaves would then become grey 1i.e. layayCFoj?iI . . . hyf-Eayi that is dry grass and dry foliage, old dry leaves. And ludlayi-Sayi is while hair d ;people, that white . . . We also refer to old people as /oyi-Eayi"*(Simpson 1989: 2). Clearly, the concept encoded in the Warlpiri word SnjC/ayi is not the same as that encoded in the English word grey.

3 1 0 0 Lexical Semantics
but who iive in an English-speaking environment and for whom English is their primary language, as childntem tended to extend the range 0 6 the PoEish word niebiexki (%lue" rrom nieba 'sky') to dark shades which in English are still called bhe, but which in Polish would have to be described as granatowy, not as niebie8ki. When they did this, I corrected them: 'hot nlbieski, grsrrrtsfowyr"".(See nlsa Diiritlctt 1978.) 1 don't know what role corrections of this kind may play in the acquisition of the first language; presumably, a more limited one. It is known, however, that in the first language, too, a child3 lexicon of basic colour words is more limited than that of an adult, and that-from the adult's point of view-children "over-extend" even words such as yellow and blue, let alone b o w n , pink, p;urp3"e, orange, or grey. (See Warkness 1973: 182; E. R. Heider 1972b.J This suggests that in the speakers' consciousness or subconsciousness, neighbouring tems may detimit one another" range, to some extent ([although the boundaries are, of course, hzzy). In his discussion of the logic of colour terms, Bertrand Russell wrote: We oertainEy know-though it is difficult to say how we know-that tw different collours cannot coexist at the same place in one visual field. . . . More simply 'this is red%and 'this is blwekre incompatible. The incompatibility is not logical. Red and blue are no more Sogicajly inlcomplalliblle than red and round. Nor is the incompatibility a generalisation from experience. I do not think I can prowe that it is ncat a generalisallion from experience, but I think this is so olbvious that n0 one, nowais grammatical. I do not days, would deny it. Some people say the inu@ompakibility deny this, but I am mot sure what it means. [RzosselE 1965: 781 Quoting this passage in my Lexicogrtrphy and CTsrncepruaJ' Analysb (Wierzbicka 1985: 79-80; see also Wierzbicka 1940a) I suggested that the incompatibility between red and blue is in fact semantic, and I proposed semantic components to that effect, along the following lines:

10. The Me~~nSmg of CoIour Terms 301

and rnidori 'green'appear to be "basic", and yet they dcfinc ihc notion of a a1 exclusiveness mandan follows, anyway, from identification (as argued p@rsonrel;crred L o srs JOHN is r~ormally not reFerrcd oes mean that when I refer to someone as "John" I om I call JOHN and whom E don't call anything ula: "he person whom I call0 JOHN'is sufficient. if an explicit "exclusionary" component Is not necessary for proper it is probably not necessary for colour terms either and a rudiy formula such as 3 is blue = when people see things Ilike X they hem: Ithis is B L U E h a y be essentially correct (not for all colour e caabstract"ones, such as red, blue, green, yellow,

. . . what all green objects that we've learnt to call them 'green'aand what all kwaals 'yellow-green" a SShswap tern en the N.S. Pacific coast, which d Ma~hury (1987) 3 objects have in common is that Shuswap t ltcl call them kwaai"sand can teach us which things are as we can bach them which obj~ercts are green." is is probably true; but, as I shall argue below, this is not the
3: 132) writes: '30perhaps

6. ""Black" and ccWhilte"7 "Dark"' and "Light"

If we say only that white means, essentially, '"hat people call WHITEy'$ bhck "what people call BLACK'" we will1 fail to account for the fact these words are felt to be opposites, and also for the fact rhat they are to be closely related to dark and Sight. For example, one can form in English compounds such as light blue or dark blue, but one cannot form mmpowds such as *lighr kvhite or *dark black. Nor, for that m t t e r , can one call something *dark white or *Sight black. The expressions dark white and light black sound self-contradictory, whereas lfght white and dark bIack sound foolishly tautollogous. To account for these facts we have to analyse both pairs of adjectives into components, and to see what they have in c a i m a n . H believe that (as hinted earlier) the clue to the semantics of dark and light lies in the concept oE seeing, and that the prototypical use of these words has to do not with any objects but with the ambience. We say, above all: It was (allready) dark. It was (still) light.

X is blue. = when people see some things they say of them: this is BLUE Xis like this when people see other things they say some other things of them they don't say these other things of X

I have come to doubt, however, whether ""leclusiocrary'" components of this kind are really necessary, for a number of reasons. First, not all pairs of "basic colour terms" are felt to be incompatible in the same way and to the s m e degrlee. For example, wkik and black are felt to be opposites, whereas red and bhe are not. Furtihermore, re$ and pink, though incompatible, are felt to be closely related, whereas, again, red and blue are not. Even more importantly, some "basic colour t e m s ' b a y not be mutually exclusive at all. For example, in Japanese both the t e m s aoi 'blue, blue-

Sentences including expressions such as '% dark ball" or "a light flower" seem neither as common or as onar~iral as sentences wilh PIC words dark or referring to the ambience. One can also speculate that in children's speech the words light and h r k refer, plrredominantly, if not exclt~sirsely, to the ambient, not to calours. Pllthozlgh h have no data to support these speculations, it is hard to irnagine small children talking about a "dark dress" or a "light dress", whereas sentences such as the one quoted earlier (by a two-year-old, Ellloom 1991), "I was crying because I dicriln" want to wake up, because it was dark, so dark" are of course well attested. Bult what dlo we mean whew we say tbat "it was (already) darY or tbat "it was still light"? I would suggest that we mean something along the fdlowing lines:
it was dark {at that time). = at some times people can't see much it was like this at that time It was light (at that time). = at some times people can see many things it was like this at that time

X is dark. = at some timcs pcopEe can" scc mlucll when one sees things like X one can think of this X is light (in colour)." at some times people can see many things when one sees things like X one can think of this I do not think Ebalt the words durk and J'igSr~ (as colour desigmnatiows) are art by ostension, with reference to some objects wbich provide models "a dark oolour" or ''a light colour"'. E f here is a model of "darkness"", a model of "li&tnessm>t is to be found in the darkness of the night, or the light of the day. To put it diffierently, seeing dark objects reminds us the experience of seeing things at a time when it is dark; and seeing ligbt ects reminds us of the experience of seeing things at a time when it is ligbt. It is interesting to note in this connection that in some languages, for example in the Australian Aboriginal language Luritja, one of the two basic wolour terns ["light' and "dak'"~ Is in fact identical with the word for night-time ([Ian Green, personal comunication)~;and d s o that in Alice Springs Aboriginal English, night is often called "dark t h e " (Jean Harkins, personal communication). Turning now to the English wards 6Iack and white, I would suggest that their semantic sltaucture would reflect both theh status of "basic colour terms learnt by ostmsion" and their association with the concepts 'dark' and "light'. (Cf. Leonardo da Vhci's comment made in his "Treatise on Painting'" "We shall set d o m white for the representative of light, without which no color can be s m ; . . and black for total darkness"; quoted in Birren 1978: 4)~.As a first approximation [to be refined later) I would propose the following:

It could 'rae argued, quite plausibly, that these definitions are too broad, and that they could be linked explicitly with, roughly speaking, d a y h e and night-time, along the following lines: It wm dark (at that time). = at some times people can't see much because the sun is not in the sky it was like this at that time It was light (at that time]. = at some times people can see many things because the sun is in the sky it was like this at that time The matter requires further consideration. For sentences referring to "darlr"' and "Eight" objects, we could then propose explications along the folllowing lines:"
The word one in the sense used here has no place in the Natural Semantic Metalanguage, and should, strictly speaking, be replaced with sameone, along the following lines: when someone sees Lhiogs like X, this someone (this person) can thiak of tintes o r this kind

X is black. Ipartial explication] when people see some things they say of hem: this is BLACK X is like this at some times people can" see anything because the sun is not in the sky when one sees things like X one can think of this

For white (of which more vdl be said later) we could initially consider a symmetrical exgllicalion:

X is white. [partial explication]

when people see some things they say of them: this is WHITE

X is like this
One could also can~siderexplicating 'light' via 'dark', so ba speak @ecause even in broad daylight, dark ahiogs stiil look datk, whereas when it is dark, even light Lhings look dark).

I haw dccidcd L o use one, rrulher than someone, however. lo make tlne explicatiaaus easier to

at some times people can see very many things because the sun is in the sky when one sees tbiangs like X one can tllrio~kof this Explications of this kind amount both for the intuitively felt antonymous relation between bhck and white, and for the intuitiveiy felt links betwen bbck and dark, and between white and light. They do not imply that peaple think of the day as "something white" and of the night as "somethlrrg b1ack"ds they might think of snow as something white and of charmal as something black. Nor do they imply that white objects necessarily make us think of daylight, and black objects, of the darkness of the night. Bwt they do imply a potential conceptual link: "when one sees things like this one can think o f . . .". There are good reasons to think, however, that the meanings af bhck and while should not be presented as fully symmdriml. The associaition between black and night is no doubt more straightforward and more tramsparent than that between while a d day. The link kbween white and high visibility seems intuitively indubitable, but it is not white itself which is highly visible. Red and orange are no doubt more visible, or more obvious, than white. On the other hand, white provides the best background far all other colours: dl other colours are better visible in "broad daylight" (in Polish, w bhfy dzieh, E ' n white day" than at dusk, and also they are better visible i n places which provide a white visual background, such as, for example, a snowy landscape or the white paper on which we write or type. To account for this property of white as the best possible background far objects of all kinds (other than white) we could consider adding to its explication some further components aQongthe fotlowing lines: X is white. [parltisl explication] in some places, people can see very many things when one sees things Bike X, one can think of this It should be added, however, that 'white' is a much more complex, and more problematic, concept than %lack" no doubt because 'black' has a universal prototype in a '"pitch-black" (very dark) night, whereas 'white" doesn't have a similarly uniform universal prototype in a very bright day i(becausewhen it is very light people see Pots of different collours), and may in fact embody in its meaning two very differenit points of referenoe: a ternporal one (day versus night) and a spatial one (a white wintry landscape, covered with snow). AS mentioned earlier, snow (unlike day) cannot be a universal referencepoint in the semantics of vision, but of course 'white' is not a universal conloept either. For English, and for o t h r languages which do have a semantic equivalent of h e English word white, a snowy Iandscape seems a plausibb mferenoe-point, of course not as a neoessary feature of every individual's

personal experience but as a feature of the collective memory of speakers of English, reflected in their shared semantic universe (see expressions such ;as snrow-while, snowy whIfc, or Samw Whil'e, and white CIsris~irwas)~. A number of observations about the concept kwhitehade in the Iiterature appear to support this suggestion. Thus k h i t e 9 s often described as a '%whce colour",not a "voluanl: co1our'~seee.g. Westphal 1987: 14; Katz 1935: 7); a colour which "more than any other colour, offers resistance to the eye" Westphal 1987: 14). It is also described as a quintessentially "opaque" colour, incompatible with transparency. "White is an opaque colow", noted Wittgenstein (1977: 4), and he puzzled: "Why is it that something can be transparent green but not transparent white?" (1977: 5). It seems to me that the prototype of the ground covered with snow explains, to some extent at least, intuitive observations of this kind. For if "white" k " 'e lightest of colows" (Wittgenskin 1977: 23, and is that "which dms away with darkness" (11977: 151, this is explained by the fundamental contrast between day and night (roughlly speaking, the night is "black", the day is the opposite of the night, and 'khite" is the opposite of "black"). But if "white" is also "opaque", a barrier to the eye, surely this is consistent with the image of snow covering and '%hiding" the ground? The "bjue" of the sky, or the ' ~ e l l o w ' k fthe sun, can hardly be thought of as a "barrier" between the eye and somethhg dse; the green of the vegetation is also something that one can normally see through (except in the thickest jungle); and of course deep waters of the sea, or of a Iake, are anything but opaque. But the white expanse of snow is indeed an exceedingly light and yet opaque '"barrier" to the e p , a covering of the ground, which cannot be seen though, although it sets off, and lightens, the visibility of all "figures'" visible against this light and opaque background. Finally, it is worth noting that "black'hnd "white" are by no means symmetrical in the world's languages, and that '%kicky' is a more common term than 'khite"'. For example, Hargrave (11982: 211) writes this of the Australian Ismguage Martu Wangka:
The primary question raised by thc data in comparison with IBerPin and Kay's] mllour-encoding sequence is the absence of a basic tern for white---ar even for 'macro-white" Aawrdlng to the sequence, a Pamguage with two basic colour terms has the: categories 'macro-blackkand 'macro-white', and foci for the latter can be expected to vary between white and red, while the former may have foci in black, green or blue [Kay and McDanieJ 1978: 63%. The Martu Wangka data, however, show clear Gategorim focused in btack and red. Twemty-two participants d;$ focus a colour tern iru the pure whine area, but a variety of terms were used, and five participants used two or three terns. In all, twelve different terms were used which were focused in white.

Hargsave notes further that "This same lack of agreement on a tern For white is found in the Warlpiri area" (1982: 2121, and she concludes:

3016 Lexical Semontta

The above data suggest that Aboriginal groups who were @aditiolnaIly desert m w d s did not abstract the collaur white as a separate property of a variety of mahral phenomena, and fierefare it canmot be oonsidered a basic colom term i n tlieir language. when Wittgenstein wrote &at "white is an opaque collou~" 41977: 41.31,a ""tmsparent white is impossible'"l977: 191, and that " ' white water is inconceivable" (1977: 51, he was clearly thinking of the q'white". But the words from other languages glossed in ~ n g l i may have a different image. For example, AQexandraAikhen~al co~m~icatjlo reports n that in the Tariana language of Bra "whi&" fi~~sajire) means also 'transparent"as w Birren (1978: 3) qudes (in English] the following (so fmm the Upanishads: "The red color of burning fire is white color of fire is the color of water, the black color of fire is of earth". Then, while the experience of snow is of course far fro gal, so is the idea of an opaque, "surface co1ourq', 'white'.

PO. The Meaning o f Cobur Ter~ns 307

hings growing out of the ground" '8 no doubt ot ody for the Engllish word green and far its semantic equivalents r example, the Polish word zielhlny), but also nearest counterparts of green in languages in which this word does have exact semantic equivalents. counterpart of greea is gwyrdd, whose er than that 01geen: and some English reens" are lexically identified with "b1ues"in Welsh [see Hjelmslev 1953: n on Welsh to interpret these facts in uld appear that Welsh restricts gwyrdd to er, brighter, fresher greens. Trying to account for this in lnitively plausible prototype, I would propose (as a startingssian) the following (partial) exp1icatian:

in some places many things grow out of the grounnd at some times there is walter in these places (after rain) when one sees things like X one can think of this
The reference to "wetmess after rain" brings to mind, 1 think, fresh, glisore consisBnt with the range of g w y r d than a mere owing out of the ground'" But it is not so much wet is a point of reference; rather, it is a growing out o f the ground, rain comes, thing glistens with wetness". The whole vivid, natural greenness evoked also by the inages ture lore ('"ow green was my vdley owdaries of gwyrdd are as '"fuzzy" as ifferent phrasing of the conceptual reference-points somewhat difleremnt range of reference. tation in the case d green, and to vegetation and dd, may seem rather speculative, but they are case of the Hanurrdo word laruy, desc~bed by 64: 191) as 'Yight green and mixtures of green, yellow, and light c ~ l o u "which r most tangibly visible in their [the Hanundos'] e surro~~ndings'~, and whose focal point is '"ear light- or yellow-green". onklin shows that htq is clearly associated both with plants and with 3s- Hanundo has four basic colour terns, which can be PooseIy glossed e m s of wetness versus dryness. siccation and wrewess or freshness @u@environment which are reflected in the tively. This distimction is of particular sig~ficanw lterms of plant life. Almost all living plant types possess some fresh, succulent,

7. Greea, gwyrdd (Welsh), h t u ~ !yHsunlsnd~S

In mamy languages of the world, the nearest equivalent o f t green is eiher morphologically or etymdogically related to words herbs, or vegetation in general. For example, in Polish the ward etymoQ~gkally derived from zi&, %erb3.In 'Wartphi (as m far "green" or "green-blue" is a reduplication of itbe plants. B n fact, even tkne English word green is believed to be e rejated to grow (see Swadesh 1972; Mein 19661. Native speakers of English, when asked to give some ex usual[y mention grass, leaves, or fresh vegetation (most cOmmon This does not mean that their range ol greens is restricted to the grass, or even mare generally, to the colours of vegetation; sonably clear that they do associate the concept e gees with vegetation ("things gowing out of the grou this, I would propose h e falllowing (partial) explanation

X Is green. =
in some places many things p o w out of the ground when one sees things like X one can think of this

I haye deliberately refrained firom using here the phrase '" is li bemuse there are shades which native speakers of English w0ulc.I pared to call green although they would be reluctant to colour of vegetation. None the less, a vaguer association by the explication sketched above seems to be valid.

308 Lexical Semom tics

and often 'greenish' parts. To eat amy kind of raw, uncooked food, garliiGlularly fresh knits or vegetables, is knowm as;pag-hry-un (Ssr~uy]. A shiny, wet, baown-colowed section of mewly-cut bamboo is rnakatuy (not marrrraq. Drkd-out or matured phnt material such as certain kinds of yellowed bamboo or hardened kernels of' mature n o r parched corn are marara7. (Conklli~u1964: 191)

10. The Meaning of (Cobm Terms 309

On the basis of Gonkclin" comments I propose the following tentative explication: X i s latuy. = in some places many things grow out of the ground there is something llike water in these things when one sees ~thikingslike X one can think of this 1 1 1 1 this case, unlike the case of the Welsh gwyrdd, "uiciness" rather tha external wetness appears to be part of tihe conceptual model. (I doubt if" shiny, wet, brown-coloured section of newly-cut bamboo" co gmryrdd.) It is. also inkeresting to mote that although Hanunbo d a separate word for "blueW>ts word for ""geen'Uoes not exten blue at all, dark blues being categorized with "black", and light "wfite". The association between Iarrduy and "juicy plants", which as a part of the meaning of this word, is consistent with this fact, Conklin" iiluminatimg description or Hanunbo makes it partic clear that the use of colour charts are not a suitable melttrod the meaning of cdour terns. To understand words such as Satuy an we have to understand the conceptual prototypes to which these refers to fresh, juicy plants, whereas th refer. The prototype off laarls~~y ciation between redness and dryness can be expiained if we assum word rara? refem, in its semantic structure, to fire and to burning. colour chips cannot account for facts of this kind, but verbal can. t . r r yis a sep Of course one could say that the "wetness" implied by b semantic feature, which can be added to a description in terns of brightness, and saturation. But the evidence presented by Con that in the speakrers' mind this ""wetness" or '~uiciness"is not an dent semantic feature: rather, it is an integral part of Ithe same p which determines the kind of greenness associated with this warn, succulent, ripe, closer to yellows and to light brown th What applies to Hanunfio applies also, rnoratt mulandis, to the S term Rwaalt, Yellow-green', which was menltioned earlier, and
g in ahis mnnectim Toren's [I992 1693 remark: "co abstract& From other domains of classification as if it formed a flnn Hanunbo warm and cojid categories, with their connatations . . . wggest a connection with plant classification, far which IE than with ~ 0 3 as 0 such". ~

""Much to the amazement of MacLaury (1987) reporting this fact . . . 'contradicts present physidogical knowledge? " (van Brakern 1993: B IS). To van Brakel's highly pertinent question "Wound it really be possible that the meaning of a word in a far-away-culture mntradicts our ph!ysi~r'ogical howledge"?"'lwould add my own: Isn't it likely that in the natural surrowdings of the Shuswap people these is something that is visually salient and that is "ellow-green' (perhaps sun-burnt grasslands, like those which define the yellow-green Australian landscape)?

8. Bhe, niebie~ki (Polishi), gohboj md shij (hssian], aoi [Japanese), armd Jiia [Thai)
[mr numerous languages of the world, the nearest equivalent of the English
word blue is morphologically, or etymologically, related to the word for ple, the Polish word niebieski is derived from niebo, %ky; and Latin ccaeruillew is derived from the word caelm, 'sky" Words of this association between the colonr concept in question and "sky"". The English word jbhe and the Japanese word aoi are related to the words for sky, but (judging by informants' these languages, too, there is a strong association between the nd the concept of 'sky" h :e m asked to give shme examples [we, or of something a&, informants invariably mention the o account for these facts, I would propose not only for niebieski and lso for ErSW and mi, the following stmamatic camponent: " X i s bluelaolJniebieskir'caeruIeu~~ = at some times people can see the sun above them in the sky when one sees things like X one can think of the sky at these times However, although words such as B S w , aai or niebhski are all semanticoncept 'sky', they are not identical in meaning, since of these words is unique. For example, as mentioned earrs only to light and medium dark blues, to veq dark glish would still be callled blue). To account for this, I he, but not to niebieski, an additional reference-point: "big water places", such as the sea or lakes (especially is allows us to contrast the meaning of niebieski and blue Xis niebieski. =

(4 at some times people can see the sun above them in the sky
when one sees things like X one can think of the sky at these times

3 10 Lexfc'cal Sem~asa'cs

10. The Me'caning ox CoIaur Terms 3 1 1

at some times people can see )thesun above them m i p the sky when one sees things like X one can think of the sky at these times (b) in some places there is a lot of (very much) water when people are far from these places they can see this water when one sees things like X one can think of this

I have refrained from using in the explication d b/ue the phrase '" is like this" "cause pople can distinguish the cdour "blue" from the spec5c shade '"sky-blue'" None the less, I believe that a vaguer, more general m a ciation between blue and sky is valid, a m p d my informal cjuestioning of a number of informants confirms this. I recognize Ithat the '%st", focal bIue is darker than sky-bhe, and IIIOR "vivid'9harv the fcRbl~e'y of the sea. Bts exact shade may indeed depnd some properties of the human perceptual apparatus rather than on the shades prevalent in the human envimnment, such as the sky or "big water places visible from afarY'".ut the mnge of blue is II it cannot be explained in purely biological terms. can account for it if we assume that the concept b of swuctnare which is characteristic of the compoun or pea (as collour terms): it does not provide an ex some points of referen~e.~ Polislr Itas no pllrrase Pike '"sky-blue", h i t it has a more specific adjective B!gkitrty, which is associated primarily with the sky. Accordingly, 1have no used the phrase "Xis like this" for the Polish word nfebieski either, and have used instead the vaguer formuia: "one can think of". Let us turn now to the two Russian ccruntcrparts of b 'dark blue', and galcrboj, Vight blue'. Berlin and Kay (11969: 3671 have r some doubts about the "basic" character of gnl~Eubaj, referring to some dence showing t b t among Russian children goSuboj is less salient than (Istomina 1963). However, there is also some evidence s
low", which in turn is less salient than the word 11973); and yet all of these words are regarded as Furthermore, Frurnkina (1984: 31) reports that Russia prised when they learn that English has only one sing (i.e, blwe). This suggests, it seems to me, th and grohbojas "basic". She also notes: "Some inform spakers of Russian--don't want to regard the wo m d seryj, "ref, aas basic "because: they are absent a m n g the mB0iurs of
It is interesting to note, none the less, that aocosdixug to van Erakd (1993:1141, "English speakers often volunteer two foci For 'blue"[one dark and one liglnt].'"

'-."' llvey are all convinced, however, that golvboj and sinfjare there. o Corbett and Morgan 1988; Moss 1989.) ieve that the semantic relationship between gohboj and sinfj (and these words and the English word blue) can be satisfacr if w show, in the explications, that d l three words Y, but that gohboj'is directly likened to the sky, whereas sing be "like the skyY'althoughit can make people tthink of the specified in this respect (so that it can include both the -sky-blue shades of the sky-sea range]. In addition, the proposed below link golluboj directly with broad daylight and sing nce of full daylight.
people can see the sun above them in the sky when one sees things like X one can think of the sky at these times X is Pike this kind of sky at some times pe~ple can see many things when one sees things like X, one can think of this at some times people mu see the sun above t Y r m in the sky when one sees things like X one can think of the sky at these times [b) Xis nor like tlve sky at these mimes CC) at some limes people can't suee very nqu& when one sees things like X one rn think of this ed out that the explication assigned here to sffig does, as a ' " a r k mlour'", although it does present it as darker than see this, compare the following two variants: eople can't see very much eople cm't see much t i(l), which I have assigned lio the English word assign to other words which stand for colours thought and it differs iflg not thought of as "a dark colourYy; Polish word granralowy, also "ark blue" wwhh defhas been assigned a d a d colom. The h c t that accounts for these differences. Panese word dlr~iI(@@), we note that its rmge covs which in English are called bhe, but also some e calked green.1 Thus, it is not only the sky which calbd anlFf, but also wet grass and the crGo!" traffic lights. In fact,
Y"' information on the use of aoi comes chiefly from discussions ~ & h Takako To&, md m her reports of infomaxutsSresponses. I am very grateful to her for her help in mat-

3 12 Lexical Semantics

1 1 0 . The Meaning o f CoIour Terms 3 13


Japanese has set phrases referring to both grass and traffic lights as (Takako Toda, personal communication). For example, when teacher Japan teach children tramc rules they say: WWn the lights turn aoi look to the right, look to the left>and cross." When there is a need to contrast the colotnr of the sky with the wlo grass, a different colour adjective is used for grass: midorr'. But when

to note that the sea, umi, is normally described in

, next to the sky, what could be visually more changethe ocean? In fact, according to my informants it is the ocean

n grass or plants in general which provides the second best exafter the phrase aoi SOM,'blue sky', it is the phrase aoi umi, comes to mind most naturasly in oonnection with sea can be seen, at different times, as either blue or triple model, based on the sky (primary point of reference), the sea ference), and vegetation after rain {a tertiary point of to accord better with the way aisli is used, and with to it. Fosiowing this line of thought, one could prowing (partial) explication: at some times people can see the sun in the sky when one sees things like X one can think of the sky at these times [b] in some places there is a lot of water when people are far from these places they can see this water when one sees things like X one can think of this (c) in some places things grow out of the ground at some times there is water in those places when one sees things like X one can think of this


also restrictions on the use of a d as an attribute: o mry state of trees or grass, for example: Arne no ato, ki ga ao ao to shite im. After the rain, the trees are (look) very m i .
a e underscores the vivid, fresh look of the troes a The reduplication ao a the rain. Hnterestingly, midori wauld not be used like this:

? h e no ato, ki ga rnidori midori to shite 2m.

to a temporary visual impression. It is also intereslting to note that while troi Is readily applied to cha a book with a green cover w u l d be described as midmi, not as aoi. these facts point to a link between aai-mess and a possibility of change.

ally, k t us consider briefly the situation in Thai, described authoritaby Diller and Juntanamalaga [forthcoming). In ''lower", rural. Thai, are only four basic colous terns, with foci in the areas of white, black, nd green. In the '%igher"+uutban Thai, however, there are also two of the "blues' type: f i a (lit. 'sky') and nam-gars (lit. "silver-tarnish". efers only to very light blues ("sky-bluebr even lighter], tter designates, in particular, the dark blue of the Thai flag. ered in English the focal blue is too dark to be callledfda and called nam-gan, and informants regard it as a "difficult to e, allthough if pressed they may call it fhtr-k& (lit. 'sky-dark' '1. There is, therefore, a "no man's sand" between the two remarkably, it is this no man's land which corresponds to the pposedy determined by universal human neurophysiology. hlights the irreducible gap between neurophysiology anrd g. Surely, what the Thais Pack is not a perceptual category but a conone. It is likely that sooner or later they will develop o n e - o n the lish bhe, of the Polish rslbieski, or of the Japanese aoi, or ems mast likely, in fact, that they will follow a path ne, with two "basicL'words far "blue".) In any case, X the exact shape of this future category on the basis of

3 14 Lexical Semansics

JO. The Memring O ~ C O I OTM er Pm 315

ordinary visual interest, things that dominate people" view (like the sky, the sun, the sea, or a white snowy landscape do). But if we do not define red via blood, or not just via blood, what else can we do to ducidate this ooncept? Trying to approach the problem from a different angle, I shall take as my point of departure the suggestion made by Manning (1989) that red is 'b rich, warm colour". The words "rich"amd ""warn" are used here metaphorically, but I think that these metaphors provide useful clues to the meaning of red. Of the b u r "'basic colour categories" encoded in English as red$yellow, greem, and blue, two, red and yeS!okv, are thought of cornmonly as "'warn caIours". Why is that? What does the notion of "a warn mlour" mean and why is it that people associate 'karmth'kwith red and y e h w rather than with green and h h e l The answer seems to me rather obvious: ye/tmv is thought of as "warn" bcause it is associated wit11 the sun, whereas red is thought of as "warm" because it is associated with fire. It seems plausible, therefore, that although people do not necessarily think of the colour of fire as red, they do associate red colour with fire. Similarly, they do not necessarily think of the colour of the sun as yelllow, and yet they do think of yellow, on some lewek of comnsciousness or subconsciousness, as a "sunny collour". It seems likely that the association between red and fire, and between yeifow and the sun, is a little hrther removed from the surface of speakerskcansciousness than that between bhe and the sky, or grem and things growing out of the ground. None the less, it is not difficult to bring these links to the surface. I have asked a number of informants what colour they think fire is, and several of them replied: orange. However, when F ask informants which colour fire makes them ehhk of, many of them reply: red. I think the reason may be that when asked about the colour of fire people think of the flame; but when asked about "'what fire makes them think or', they think of the whole situation involving fire, and this includes glowing red coals. The association between fire and red is supplorted by the existence of set phrases such as red-hot, red coals, or fiery red (cf. also the name of the mast popular Australian brand of matches: Redheflds). Other European languages have similar phraseological reflexes of this association. Far example, in Polish the expression czerwony kksrr, literally 'red rooster" is a synonym of fire. It is also worth noting that fire-engines and other paraphernalia used by fire brigades are often painted red; that fire extinguishers are also painted red; that red is generally used as a symbol of danger or warning (for exampk, in traffic-light systems). It seems reasonable to suppose that all of these facts reflect a common association between fire and red. The common association of the colour red with fire is well attested from a wide variety d times and places. For example, Birren (393833 notes that

any past or future findings in either ~chromat~lagy ar neur~physiology~ Furthermore, the situation in Thai seems to be at variance with the @laera1 amount ofcolour category formation proposed by Rosch (1975b: 1M): "There are perceptually salient colours which more readily attract attention and are more easily remembered than other colours. When category names are learned, they tend to become attached first to the salient stimuli, only later generalizing to other, physically similar, instances. By this means these natural prototype colowrs become the foci of organisation for categories." But in Thai [like many other languages) it is the sky which is twated as the "nalural prototype" of a bluelike category; and this "natural prototype" is different from the kind of "blue" which is said to be perceptually most salient and which is (perhaps) most likely to become the focus of the: not-yet-born basic "blue'btegory. For this category to be born, the focal, pcrccptually 111osd salient ""bwc" must 'becronne conceptually linked with srvllc ~ii~l~iccnblc ~~cl'crelnec-hroilrl it1 Llsc speakers' cxpcric~~cc-suclr as, for example, the idea of the sky on a sunny day.

The nearest equivalent of the English word red is in many languages etymologically related to the word for "blood" (recall, for example, the Warlpiri word for "red" niccntioned in Section 1); i r can, bowcvcr, be related to many other putative models, such as, For example, various minerals (for example, red ochre) or other sources of pigments and dyes. The Polish ward czertvmay is synchronically no longer analysable, but it is believed to have come from the name of a red worm, czerw (Brwckner 1930). The English word red is not synchl;onically analysable either. It is possible, none the less, that here, boo, we can discover a common association which might unite the speakers of English in their conceptualization of the category in question. In an earlier work (Wierzbicka 1980: 43) H proposed that red may be conceptualized via the concept of %ladd', and I proposed the following rough formula: red--co!our thought of as the colour of blood Further work with informants, as well as an introspective exploration af my own concept of 'czerwony' Iprompted by objections expressed by other linguists), have led me to question the adequacy of this kind of explication. The relevance of blood to the concept of 'red' can be compared to the relevance of milk to the concept of 'white', or to the relevance of chasmal to the concept of 'bback'; while objects of this kind may provide good exemplars of certain commonly recognized colours, they are not things of extra-

'"he Jewish historian Josephus in the first century AD associated . . . r d with fire" (p. 3). Be mentions the same association in Chinese culture; he addu~esLeonlardo da 'urinci" statement "We shall set down . . . red For fire" (4); and he quotes from the Upanishads: 'The red color of burning fire is the color of fire . . The red color of the sun is the color of fire . . . The red color of the moorz is tlre color of fire . . . The red color caE the lightning is the color of fire . ."'. (pp. 2-3). It rnay also be worth recalling here Swadesh's (1972: 204) speculaticans on the possible etymological links 'between red and the Latin nrdere, "tcp born"~:is well as bcllwccu~/igh and tlic Latin ceSb~c'.s, whit^').^' Tlve Pdct that pcoplc Lcnd to picraive the coPour of fire as orange or yellow rather then red does not undermine the conceptual link between fire and redness, lf i t is true that '"be Four hue classes, red, yellow, green, md bluc . . are ncurophysiologI~aIly ' w i r e d b r 'programmed' in human bcirigs" (Witkowski and U;rowna 1978: 4421, for the purposes of conceptualizatioo~hand communication tlvesc ncurophysiologhcal categories have b be proj~ected on to analogues given in shared human experience. For "bluen and "green", the choice is obvious: the sky [and perhaps the sea) end vegelatioo~.For "yelllow"" the sun ofSers-perhaps-one natural point csT reference. (The k t that in children's drawings and paintings the sun is represented as yellow reflects this association).'* For "red'" however, there is no invariable environmental model. Although there Is the invariable experientiat model of blood, most lzuman beings do not see blood nearly as often as they do the sky, the sun, or grass, and in any case, blood Is not nearly as salient, visually, as is fire. It is natural, therefore, that in manyw perhaps most, human cultures, in addition to "local'hssociaitiorus such ar that with red ochre, a deeper association should have been established between "red" and its nearest analogue in human environment which is both visually salient and culturally, or existentially, o v e r w h e l m i n important: Bre. The fact that we can find traces off such a con~eptuall associakiau even in English provides, it seems to me, a striking confirmation of this hct. These considerations lead us to propose the following (partial) explieations:

PI further diflerence between red and yeIIow is that yellow is thought of as

a light ccrlour, whereas red is not thought of as either light or dark. Since we haw already explicated the notion of a light colour, we can ow use it in a fuller explication of the concept 'yellow':
X is yellow. = when one sees things like X one can think of the sun at some times people can see many things when one sees things like X one can think of this
RRh colours"'aace ' U c c p ' ~ w tnot dark; they look as iS "there was much lour in them", as if the dye, or the paint, was thick. They cannot be light, ecause light collours look as if '"here was not much colour in them", as if he dye, or the paint, had been used thinly. On the other hand, red is cerainly a " ~ v i ~ i d ' k ~ l that o~r, is, one that is very easy to see. If we think, hen, that red is "a rich colour" as well as a "warm colour'" and a "vivid colour", we might consider, initially, an explication along the following

X is red. = when one sees things like X one can think of fire when one sees things like X one can think of blood one can see things like X at times when one cannot see other things
On the other hand, additional components would presumably have to be ostulated for the two Hungarian words for "red", vcs'rliis Q"dark red') and kos ('light red". As a starting-point for discussion, I would propose the Plowing:

X i s red. = when one sees things like X one can think of fire when one sees thing like X one can think of blood X is yellow. = when one sees things llike X one can think of the sun
IN Swadesh's speculations rnay seem to us rather Fantastic in some 016 the details, but thls does not invalidate his basic insight into the importance af Might and fire to tbe human conceptualization of calour. As pointed out 'fry Xu (19941, however, this is not universal. Far example, in Chinese ml. tnre the sun Is usually represented as md, no! as yellow.

X is piros. = when one sees things like X one can think of fire when one sees things like X one can think of blood at some tines people can see many things when one sees things like X one can think of this X is vbras. = when one sees things like X one can think ;of fire when one sees things like X one c m think of blood at some Xhes people can see very little when one sees things like X one can think of this
These (tentative) expllications present the two Hungarian words for "red" as manogous to the two Russian words for 'cblue"".


10. Malcrlo-white and Macro-black Following the findings of Berlin and Kay (119691, it is widely believed that the concepts "lack'and "white' are, in some sense, lexical universals. Berlin alurl K a y tIicirusePvcs (0969: 2) plrrascd tluc clairuns In questio~u as FoIPows: "(I) All languages contain terns for white and black. (2) If a language contains three terms, then it contains a tern for red.'"^ infomall, abbreviatory ways of referring to generalizations precisely formulated elsewhere, statea~icnls of Llliis kind can bc scci~ as quite legitimate. It is unforrwnate, however, that these informal abbreviations have led many scholars to conclude that if a language has only two basic colour words, we know what them wards mean: ""they must mean 'black' and "white"'. But, as mentioned earlier, this cannot be true. T f a language has only two colouo. woods, wluich divide betweetr ~lliemselves all the colours perceived by tllc speakers, tliesc words cannot possibly mean the same as what the words bkck and mean. What do they mean, then? For example, what do the Eu-jingarliya (Buram) words -guingaIsja and -gu~gun4a,discussed in Section 3, mean? Onc suggestion which has sometimes been made Is that words of this kind mean 'lighthand 'dark" respectively. But this cannot be the whole truth either if the word supposedly meaning '~ightY~ndulcfes red in its range. For example, the fact that the Gu-jingarliya counterpart of white includes also highly saturated, medium light red suggests that, in this case, tl~e contrast between light and dark, that is, ultimately, between day and night, cannot constitute the only underlying model. Since the combining of light colours and red in one class groved to be a rule rather than an exception (see e.g. E. R.Heider 1972a; Turner 1966; Conkiin 19731, the simplistic model opposing 'Yight" to "dark" h d to be abandoned and an alternative had to be sought. At that point, a number olpscholars suggested that the putative categories "light" and "dark" should be replaced with composite ones: "light warm'" versus "dark cold" (e.g. E. R. Heider 19724, and this is how "Stage I languages'kre now often represented. But this reinterpretation, which may seem quite reasonable, presents serious difficulties, too. For what evidence do we have that speakers of languages such as, for example, Dani or Gu-jingarliya possess a notion of a "'warm colour"? In English, there is at least h e expression "warm colows"; but in Gu-jingarliya and in, Dani the anly evidence we have is the very fact which we am trying to explain: that is, that the speakers of the languages in question include in one category colours which we [Ithat is, speakers of English) think of as "warm colours"'. From the point of view of the Dad and the Gu-j~kngarliya, the idea of a "warm colour" or of a "cooI ccplour"

may be just as alien as the undulatory theory of light. Consequently, the expressions "wmm co1ours"and "cool colours" may help WJ to identify the ranges distinguished by these speakers, but they tell us nothing about the meaning of the relevant terns-what the speakers mean when they use hem. To my mind, the Fact Lhal languages commonly (tho~ughnot invariably) EinL red with light rather than with dark colours suggests that this type of categorization may have some explanation in common human experience. My hypothesis is that the explanation lies in the natural association between fire and sun, both of which are associated for human beings with warmth and light: even if the sun is seen as, primarily, a source of light, it must also be perceived as a source of warmth; and even if fire is seen as, primarily, a souroe of warmth, it must also be perceived as a souroe of light. The natural association between fire and the sun (reflected, indirectly, in the notion of a "warm calourm~applying to both yellow and red] may also account for the variation in the way diflerent languages which have only two basic lcolour terms treat reds. If a language distinguishes light colours from dark and medium ones, one would expect red to go with the latter, and sometimes this indeed happens (for example, in the Papuan language lalc; see Berlin and Kay 1969: 23). In other languages, Itowever, such as Gu-jingarliya, red "unaccountably"' goes with very light colours. It seems to me that the association between sun and fire would explain this. The fact that reflective foil may be seen by Gu-jlngarliya informants as the '%best example" of the category in question ("properly number one gun-gungaSQaM') suggests that in this particular language the idea of sunlight may be especially important for the conceptualization embodied in this category: shining, glistening, bright objects bring to mind things lying in the sun [and, possibly, reflecting sun). The fact that some languages, for example the Papruan language Dani, put even deep dark reds with their light colours, and that they alllow their speakers to think of dark red as the ""bst example" of the category in questiom, is also consistenlt with the idea that both the concept of sun and the concept of fire may play a role in the conceptualization of "macro-whites". Presumably, in Dani it is neither daylight nor sunlight which plays the central role in the conceptualization af "macro-white", but fire-and possibly not even fire but glowing embers. if we assume that universal human experienrce suggests a number of potential foci, interrelated but distinguishable (daylight-sunlight-fire-glowing embers)~,and that each of these potential foci can be given priority in conceptualization by a particular culture, then cross-linguistic variation in the behaviour of "macro-whites" begins to make sense. In any case, it is quite clear that the differences between the diflerent concepts of "macro-whites" call for carefully differentiated explications, and

320 Lexical Semm tics

that the two colour t e m s of languages such as Jale, Gu-jingarliya, and D;)n~i C;IIUIII(D hc ~ ~cruaau11lic:11!y itlcaiPi6nctf wit11 cllclr oPkucr, rus dhcy canrlall be scmaoilicalily identified with the English words bbck and wwhire, dark and /igk, or warn? and cool. References to the neurophysiology of vision will not help, since this is, presumably, the same for all human beings. Ilaving said Ilris, I will now try lo construct sonvc explications, hoping that they may become a starting-point for a constructive discussion. To start with the two Dani collour terms inlrestigabd by Rosch (E.W. Heider 197201, we note that "the focal points (best exanaplles) of miti and mob were not 'black' and 'white'. . . . Examples of mili w r e reliably placed among the darkest greens and blucs. Mob, however, appeared to have two focal points: Ithe most common a dark red, the less common a pale pink" (1972~: 4511). The poinlt is importanit and one must be grateful to the author for skting it in quite unequivocal Iterms: "After each informant had pointed to an exemplar for mili and mob,I asked if he were sure that was a better example than the pure 'black' and pure 'white' chips which were available; innfornants reliably insisted that it was" (ibid.]. On the basis of Rosch's discussion, we can propose the following explication for !no/@:

when one sees things like X one can think of this a1 sorurc Li~lmcspcuple carr'l scc llie sun when one sees things like X one can think of this when one sees some things one can think of fire X is not like this Turning now to the Gu-jingarliya terns -guaga/tjb and -gwssgundja, we recall that lthey oppose light and brilliant shining colours to dark and dull ones, and that bright red is included in the fanner group. To account for these facts, I would propose the following explication for -gungaltja:

X is -gunngaltja. = (a] at some times people can see very much (marry ithings)~ when o m sees things like X one can think of this (a) at some t i m s in some places some things are in the sun when one sees things like X one can think of this (c) when one sees things like X one can think of fire
The first component (a) of this explication is the same which has been posited for the English word light; component (b) accounts for the link between -gungaltja and "brilliance"; and (c) accounts for the link between -gunga!ga and redness. As for the "darkfdulll" tern -gmgm&a, we have even less basis for speo ulation about its possible meaning because we are not told what its ""bst examples" might be. It would appear, however, that at least the dbllowing components should be included:

X is mola. = when one sees things like X one can think of fire at some times people can see many things when one sees things like X one can think of this at some times one can see the sun when one sees things like X one can think of this
This explication accounts for the fact that mola includes, as Rosch puts it, light and "warm" colours. Unlike the explication of red, it does not include a reference ta blood; it does, however, include a reference to fire, and so it also accounts for, or at least is consistent with, the fact that most infarmants see red as the "best example" of mafa It does not aocount for the fact that some informants see pale pink, rather than red, as the "best example" of this category; this, however, can probably be explained in tems of a change in progress (those informants who chose pale pink appear to have moved to a system with three, rather than two, basic colour tern). As for the opposite of moia (is. milt), w e note that it inclludes both dark and "co1d''colours; and that its focus is "among the darkest greens and blues". This description suggests that the concept in question has a largely negative character and centres around absence of light and absence of sun. This can be portrayed as follows:

'A is -gungundja. =
(a) at some times people can't see very much when one sees things like X one a n think of this (b) at some times in some places things are not in the sun when one sees things like X one can think of this

X is nilili. = at some times people can't see very much

The existence of "'macro-whites" and "macro-blacks"' at the beginning of the alleged "evolutionary sequence" cannot in my view be explained in tems of either physics or neurophysiology of vision. "'Black" and "white" are indeed opposites in t e m s of physical properties of light and "psychophysicat" properties of vision, but the fundamental contrast between "jight or wam'kolours and "dark or cool" colours does not seem to be similarly explainable. It could, however, be explained if we gave credence to Swadesh" (1972: 205) speculations which place h e and light at the root of human conceptualization of colour. Given the importance of fire in human life, and given its percepitual salience, derived not only from its collours but also from its movement, and from its brillianoe and lwninosity, these speculations seem to me to be intuitively plausible.

This link of "'macro-whites" with light, sun, and fire (in a31 their aspccms, i~tcluding brillDiarucc and Iu~nit~ersity) lvigllligluns the fact that, contrary to what is commonly assumed, "mlour" is not a universal human concept-not only because there are marry languages which do not have a word b r 'kolour" but also because in languages with only two "basie colour berms", like Gu-jingarliya, the allleged "colour terns" are not really "ccrlour terms" tub" general descriptors o f f appearance, or of visual impression. Witkowski and Brown (1978: 4411) argue that if in primary macro-classes red is usually combined with yellow, and green with blue, this '"provides evidence that a dimension based on wavelength order . . . is important in human colaur categorization. Only conjunctive primary colours or, in other words, those adjacent to each other in wavelength arder are combined in composite classes." But this does not explain why yellow-green is a very rare [though not unattested) category. Nor does it explain why dark colours should be associated with green and bllue, and tight ones, with yellow and red. Witkowski and Brown try to explain this fact, too, in terms of "wiringy'(1978: 442): "'Wring also underlies the pairings of warm hues w i t h white and cool! hues with black in the categories macro-white and macro-black resptivdy. (The converse associations, warm-dark and cool-light, are not attested.)" But this begs the question. I believe that a hypothesis which links light, sun, and fire provides a better explanation of the recurring regularities thaa a mere reference to the supposed ""wring"'.

"light" colours and 'Vark'kolours), it is bright (luminous], and it is focused in red but includes also yellows and oranges? l [ ~ lmy view the answer is clear: the concept in question must take fire as its point of reference. This leads us to the following type of explication:
X is 'macro-redy. = things Bike X are "easy to see" (i.e. people can see things like X at times when they cannot see other things) when one sees things like X one can think rkof ffire at some times when one sees things Pike X one can think of the sun

11. Macro-red and Grue

Those languages of th~eworld which have only three basic colaur t e r n appear to oppose the concept of a "coloured'~chomatic)visual experience to a "non-coloured" (achromatic) one. AS a rule, the ''co1oured''colour is fo~used in "red", which is the most salient hue far human beings (Bornstein er SF!. 1976). At the same time, however, it is a "warm" colour, that is to say one which is opposed not simply to light and dark colours, but to light colrours on the one hand, and to dark-cool or dull-cool ones on the other (see Kay and McDaniel 1978: 6 4 0 ) . This means, in effect, that "macro-r&, wlrile focused in1 red, includcs not cli~rly red, but also yellow and orange; and also, that it is associated with "brightnessyy. What could be the conceptual counterpart of a colour category which people intuitively call "warmq'aandwhich has the following properties: it is vivid H("coUourful"), it is highly noticeable both during the day and duriag the night (and therefore is perceived as maximally distinct from bath

Moving now from ccmacro-red~" to the next stage of the alleged "evoluionary sequence'" we note with Kay and McDaniel(ll9178: 630)thaa "'many e world's languages have a basic collour term that means grue". But does "grue'kean? or many writers on the subject, the first (and often, also, the last) er which comes to mind is that '"rue" means "cool"'. But vague phors like "cool" are not satisfactory explanations of meaning, ough they may provide helpful hints. We must, therefore, ask further: nd what does "cool" meam? Once this question has been asked, however, e answer is not difficult to h d : ""cool" (when applied to colours) means, essentially, "not-warm", and since "warn" makes sense only as an indirect reference to fire and/or sun, "cool" must mean a colour which-while vivid and highly visible w('kolowsed"')-does mot bring to mind fire or sun. But this is not aiL The most striking feature of "grue" is that, while it streltches over both bl!ues and green, "focal grue selections have often proved to be bimodal, being chosen from both the focal blue and focal green regions. But grue has never been found to be focussed in the intermediate blue-green region" '(Kay and McDanieI 1978: 6310). This is an extremely intriguing finding, which requires an explanatiomn. Kay and McDaniel imply that they have one: "The absence of focal chdces from this intermediate region is strong evidence that these colours have lower grue membership values, and that guue has the 'membership structure stipulated by the fumy union analysis" [ibid.). But how can "fuzzy union analysis" explain the faact that the "bestqy example o r "gme'"that is, of a "cool" colour] is chosen either From focal blues or from focal greens, whereas the '%bestw example of a "macro-red" (that is, of a "'warn'' colaur) is not similarly bifocal, and is always focused b "red"? Of course one can nzodel the bifocal structure of "grue" in a 'Tuzzy union analysis", but I don't see haw one could exptuia it this way. It seems to me that both the biabcal character of "'grw'band the "manofm1" character of "macro-reds" can 1Pe explained on the basis of the in a sense, defined negatively, as a "non-warm" hypothesis that "'grue" isis,

10. The Meaning of CoJaur Terms 325

colour, whereas 'hacro-red" k ddicnned positively, as a "warm" coIowr. The notion of a "warm" ccolour refers us to a positive experiential model: fire. The concept of a "non-warm"colowr is dafined above all in opposition to that model. It is only in addition to this contrastive oore that two positive models are involved; and these are, clearly, the sky and vegetation.13 Admittedly, one could suggest that "grue" has its positive point of ueferenm in natural '%ahr places", that is, in lakes, rivers, or seas, which cm be seen as blue, green, or blm-green. This, however, would not account for the bifocal character of "grues"; whereas the hypothesis that they are cmaeptualized primarily with referenoe to the sky and to vegetation would explain this. These considerations lead us towards the following definition of '"me'" (as it is llndersitaod by speakers for whom '%he" is more focal than "geen""):

X is 'grueBs.=


(a) when one sees some things one m think of fire

s something noru-yellow to saamething yellow. Both in d brilliance fire can be seen as a unitary mode18 of all ugh its lrocllns is identified as red rather than orange or e sky nor vegetation can be seen as a unitary model is why gmes are bifocal, whereas "macro-reds" are s can also be bifocal because the category in question ifferent way: by its "cooll", 'cnon-wann" character, that is, of, roughly speaking, " h e associations". On the other hand, [which emerge as a category before ""grille") are not defined s", that Is, they are not conceptualized as "non-cool" are defined with reference to a unitary positive model: nd positive model here (the sun), it plays a secondary case it can be seen as similar to the first one in terms of lities: visibility and warmth. By contrast, the two models of "gmgsess" can only be united on a negative basis, as k i n g different from "warn ~olcaurs",that is from "ffiery'hnd "sumny'"'macro-reds".

X is not like this (b] when one sees things like X one can think of the sky (c) in some places many things grow out of the ground sometimes when one sees things like X one can think of this

12. Names of Mixed CoOours

According to physicists, there are three primary colours in light: red, green, and blue. "White light can be made by mixing red, green and blue Bight" I( WarEd of Science n.d.: 1163). But, of course, this is not how ordinary people think about colours. Acleordhg to psychologists, the smallest number of colour t e m s by means of which we can systematize our colour experience is not three but six. "'If pressed to the greatest possible economy of colour terms we find that we can describe all the colours we discriminate by using only six tems and their various combinations. These are red, yellow, green, and blue, the four unitary hues, and black and white, the two extremes of the series of hueless colours. All other colour narncs . . . can bc described by referring to these six t e m s and combinations of them" [Hurvich 1981: 3). I believe that the meanings of colour terms in languages with an elaborated colour lexicon (such as English) accord reasonably well with the above statement: beyond the list of the first six colours in Berlin and Kay's sequence all the other ones [with the exception of brown, to which [ I will return later) are conceptualized, on some level, as "mixtures". Very roughly: orange = yellow + red pink = red + white purple = blue + red grey = Mack + white

For speakers for whom "green" rather than "'bue'\epreserrts the lkst example of "grue'" we would plaoe the components referring to the sky after, not befare, those referring to vegetation, and we would include the word "sometimes" in the component referring to the sky:
X is '$nuea9. = (4 when one sees some things one can think of fire Xis not like this (GI in some places many things $sow out of the ground when one sees things like X one can think of this ( c ) saametimes when one sees things like X one can tlvink o h " the sky

It will be noticed that for "macro-red", too, two positive points of reference have been posited: fire and the sun; but the relationship between these two models is quite dimerent than that between )the sky and vegetation. First, one can presume that fire is visually much more salient than the sun, whereas the sky and vegetation are on the same level of salience. Second, fire itself can be seen as yeliow, orange, and red, and therefore it is not
Van Brake1 [IW3: 117) notes (halt Zulu speakem, who haw one term Far '%he" and use at times expressions "grue like the sky" or " m e like grass" to difkrenrtiak between the twa. It is also inltenesllilng to note that although the Tarilana word for "red" is derived from Ithe ward far blood, ills rcrerenlirul range (red. orange, dark yellow) points t~ tine, rather Itham blood, as a wnceptual prototype (especially given that dhe lanauage does have r separate word far r~el!ows'].


AS mentioned earlier, in Warlpiri, where colour terms are forme reduplimtion, the nearest counterpart of the English word brmvn nze erally, "arthearth" just as the nearest counterpart of the Engli green means, literally, "grass-grass"l4. TTh manifest association something " b r o m ' k n d the colour of the ground is instru~tiw. Of course, the colour of the ground can vary, and it varies more than colour of the sky or the colour of the sun. This is consistent witb the that "brown" was further down in Berlin and Kay's sequence than "re 'cyellow", "green", and c'blue"'. At the same time, the hypothesis "brown" does have a positive model (albeit a notoriously heter one) would explain why it came in that sequence before '"grey' "orange", and "purple"'. These considerations lead us to the following [partial) explicatio English word hrow~l!:

X is brown. = When one sees things like X one can think of the ground (earth) at some times people can't see much when one sees things like X one can think of this
In support of this 'knvironmentaE prototype" approach to the 'brown" I would add that "brown colour" is often regarded by s a puzzle. For example, Westphal (1987: 53) notes that if red, green, and blue darkemzed, "llze resulting maroons, navies and dark greens seem to re their parent hue in a way in which b r o w doe Boyruton" (1975: 315) view that "brown is certainly the dark colours created by experiments of this kind] because it almost entirely to rescmblle the original bright co~ourY'. (See also G 1977: 127.1 WestphaP (1987: 4 4 1 maintains, none the less, that " a kind of darkened'yetiow", but this is counter-intuitive uuwnviucing, "What dws it meam to say, 'Brawn i Wittgenstein (1977: 25) asked, incredulously, and he noted: "'Bro above all, a swrface colour, i.e. there is no such thing as a clear brow only a muddy one." I would like to suggest that "braiwn" and "yelllo colours, and not as different versions of the same colour, because I % associated, ~mconsciously, with diRerent prototypes: if "yellow" isis, p ily, the colour of tihe sun CiigQlb, "warn'" and I
l4 Similarly, Alexandra AikhenrvaUd (pexsslnal cornmicadon) reports that in L r notes ~ p l rthe Tarjana language oF Brazil [from the Arana Family), the nearest eoun bsam is glassed as 'muddy, dirty, brovmislm', and that this ward is clearly associated colour of the earth.

primarily, the conour of earth. Wittgenstein's observation that "'brown'" Oike "whiteJ') is a '%surface colour", talllies well with the idea that the concept %row' (like 'white') has its prototype in the surfaoe of the earth. From t h i point of view of chramataiogy, it might seem strange that human beings should treat "brown" as an important concept and honour it with a separate ""itpasic calour brm". 8ut from the point of view of people's W e on earth, "the naked earth9Vsan important visual (and existential) reference-point (like the sky above our h e d , or the vegetation all around US]. It is this visual and existential saliena of the earth which explains, I suggegt, the scienltistk "puzzle of brown"".ewes (1992: 863) writes: "Fixation am t!he speGtrum colors and on physical m d neurophyslological explanations for color proefion obscures the fact that many colors of cultural interest to human being, such as the variety of browns and tans, while now understandable as compli~ted mixtures of light of different wavelengths, etci@., are not present as distinct components of the sojar spectrum." I agree with this, but I would add that the cultural interest of browns and ns (which, presumably, has to do with the value of soil and cattle in man life) must be seen in the corntext of the visual salience of big expanses such as the sky (often light blue), the sea (often dark Hue], the grass-covered ground (typically green), the snow-covered ground (white), the naked earth (often brown). It should be added that while I haye explicated a number of colour concepts via "mnvironmental" concepts such as those encapsulated in the English wards $re, sun, sky, graax, sea, and ground, these 'cenvironmentai" concepts are not p~stul;rfedhere as indefinable conccptuai primitives in s of which people conceptualize their experience. On the contrary, ,too, are regarded as constructs built by human beings on the basis of ejr experience of life on earth. (See Chapter 7.)

14. Names of Specific (Locally Salient) Referents

It seems to be a universal feature of language that callour perceptions are dewxibed, at some stage, not only with reference to wisuallly saPient features of the ""macro-environment" (such as the night, the sky, the sea, or the sun), but also in items of locally salient or particularly important referents, such as certain minerals, animals, or plants of characteristic appearance. This applies, for example, to the English words gotd arrd silver,and presumably, it used to apply to the English word oraJrge.Words of this kind provide evidence for the psychological reality of a comparative semantic component in colour semantics in general. But words off this kind, like any words, are subject to semantic change. For example, the fact that the Russian w o d pIuboj, "light blue', is

330 Lexical SemlcJ.mtics

etymologically related to the word for pigeon, or that the Pollish word czerwomy, 'red', is etpologically related to the name of a particular red worn, does not mean that the associations in questions are synchronically alive. They are definitely not: in present-day Russian, godtrboj' is clearly associated with the coljour of the sky, not with the colour of pigeons. Similarly, in the Dani language of New Guinea (E. R. Heider 19'72a), three other colour terns are widely (though not universally] used in addition to the two basic words, mi&('dark-cool') and m o h Clight-warm'), discussed earlier: pimtdl, the name of a kind of red clay, is also used for ""rrle', bod& the name of the root of the turmeric plant, is used for "ye11owW, and io'uaiegen, the name of the bud of a particular flower, is used for '"Jue" (whereas no special word is used for "green"]. It seems to me that facts of this kind do not demonstrate that the Dani associate (what is called i n English) the colonr red with red clay and not with fire. For those Dani speakers who have moved, or are moving, towards a threeaolour system, and who are beginning to differentiate the old concept of 'rnolaq [focused in fire but extending to sunlight and daylight) into two concepts, the name of red clay may constitute a usefuim8 point of reference, but it dws not have to dominate one of the emerging new cicrnceptualizations. At some stage the word ppisrtur may be Oinked with both n?ed clay and fire [and perhaps also with blood), an~d at sofinc point it may dissocialc itself from its etymon altogether and attach itself excOusiwely, in the speakers' linguistic consciousnuess, to a ddiflerent, more salient perceptual model. I conjecture that this is precisely what has happened in the case of the Russian word gotuboj and the Polish word czerrvomy. It has also happened in the case of the English word ommge and is probably happening i n the case of the English word siFver (if not yet got&.

10. The Meaning ofih:olour Terms 331

of ""dable coloring agents'"Hewes calour words'9in tlre Tariana language a Aikhenvald (forthcoming): kadite, orange, dark yellowy,mite, 'yellow', h@oJ.Hse, "reen, blue" halite, 'white, light, transparent', kesoiire, "uddy, dirty, brownish' (from rlcesole, %mud". In some respects, this set meets Berlin and Kay" expectations since bers it inciudes five which could be said to match with 'kvolutionary sequence" [bkack, white, red, yellow, and iana set contradicts Berlin and Kay's 'brown", even though it doesn't have blue" and '"reen". If the word for "brown"is rejected the set as non-basic (because it is derived from the word ifar "earth"), ord for "reed would have to be rejacted too (because it is derived from word for blood)-and this would contradict Bedin and Kay's expectatween the terms for "black" and "white", e curious overlap of the terms for ""rdY'and"'yelllow'" remainn totally ained from this perspective. the other hand, the hypothesis that ""colour terms'hre oriented ntal prototypes makes perfect sense of all ana data, and it allows us to interpret the set with reference to night daylight (white, light, transparent-unlike: a snow-covered ground, the inhabitants of the Amazonian tropics), fire (red, orange, vegetation (blue, green), and earth ur brains, not in the world outside, and by our human biology (which links us, measure, with other primates); but to be able to communicate em on to something in our shared enviAs pointed out by Witkowski and Brown (1978: 42): "'Several authors classes, red, yellow, green, and blue, and white, are neuropffiysiologimlly beings.'Vn addition, however, it has
lS As for I l k "evolubiona~y sequence" proposed by Berlin and Kay,it is no longer seen es fiienable and it is not clear which, if any, parts or aspects of it w i 1 9 sunrive the current onslaught of criticisms. P believe, however, that an alternative inlterpretation of this "evolutionary aeg~enw"",hich I proposed in Wierzbicka [lg90crfi, i s worth keeping on record as a dimerent way d thinking about the issues involved, which can eccommodale all the new insights emerging from the ongoing research h t o the history of human conceptualization of vision.

15. Camclusionr Chrolmatollagy, Cognition, and Culture:

The main conclusion which emerges from the analysis proposed here is that the language of 'seeing' is rooted in human experience, and that its basic frame of reference is provided by the universal rhythm of "light"days and ""drk'hnights and by the fundamental and visually salient kaltures of human environment: the sky, the sun, vegetation, fire, the sea, the naked earth, the earth covered with snow. Since some of these fundamental and visually salient features of buman experience are universal, it is only to be expected that they will be reflected, in some way, in recurring features of the vocabulary d seeing. Since, however, they are aPso variable, with different kinds of smnery prevailing inr different parks of the globe, it is also to be expected that the vocabulary of seeing will be far from uniformquite apart from such obvious and often discussed differences as the avail-

been claimed that these 'knerophysiolcsgicaaly wired" categories are reflected in tanguage, For example:
ar particular structure is inherent in the human perception of which is mot deducible from the physical praprties of light a1

se o n our common human experience.

process analysis identifies and describes four specific ~alegories the R (red], G (green), Y (yePIow], andl B (blue) response states. . . . of basic co1our terns in all languages d'brecsfy reflects the existence o human neural response categories. (Kay and McDaniel 1978: 621; added) But how can language be "directly" linked to neural response

lea, their linguistic cadability across languages, and their superior retention in art- and larigterm memory-it would seem most economical to suppose that aktaibutes are derived fsom the same underlying fadors, most likely having to ith the physiology of primate coPous vision. In sho~t, far from being a domain suited to the study of the effects of language o m thought, the colour space

human neurai responses), whereas concepts can be shared. To be a talk with others about one% private sense data one must be a b b b late them first into communicable concepts.

perception to certain basic aspects o r human existence we can lend ing to what otherwise would be nro more than a mysterious play and mnes in the human retina, and of the cells in the neural p

llinks colour naming with common-but variablee anvirotrment and human wisud experience but not

anchors"', It is the shared concepts of fire, sun, sky, vegetation, and s which function as mgnitiwe anchors far colour naming. The visual

communicable to others. not mean the same as goluboj, a d green does not mean the same as g

te from Machaury will show just how far it has been necessary to tailor the and Kay (1969) silhouette to explain the endless 'kcdour-naming" anomalies: Ibrightness, similarity, and distimctiweness are not the only coordinates by

334 Lexical Semntics

Similarly, Hewes (1992: 163) remarks on the harmful effects of the "fixation on h e spectrum colors and on physical and neurophysiological explanation for colour perception"; and he comments that "The criteria employed by the color-name evolutionists for rejecting most commonPy used color terms as 'not basic* are unrealislic'"ibid.). Finally, van Brakcl (1992: 1169) comments as dbhllows on M a ~ h u r y ' s (31992) attempt to "save" Bedin and Kay's theory:
I applaud MacLaury's recognition that if one "goes out into the fidd'bith 320 Munsell coloulr chips, one dmsnt always come back with pure hue words. He r s . ognizes "'the myriad compllexities, subtbtks, and difrerences" which may turn up in naming collour chips, and he allows many other "Ldimensions" to play a role. I m also very sympathetic to his suggestion that the first ofieial "preliminary analysis of data" ((Kay, Berlin, and Merrifield 1991) from the World Color Survey shows

The Slemantics of Natural Kinds

1. Introduction
How is knowledge stored and organized in the human mind? In particular, does the mind draw a distinction between "linguistic knowledge" and "nonlinguistic kncrwledge", or between a '"mental dictionary" and a 'hental encyclopaedia"? For example, what do ordinary people know and how do they thinlt about mice, crocodiles, or moths? Can the knowledge encapsulated in the everyday meaning of words such as morrse, crocodiEe, or mot11 be separated from the knowledge that people may have about mice, crocodiles, and moths? There w hardly be a better way of approaching these questions than malysing language. Language can allow us-better than anything e-to discover how knowledge is represented and organized in the uman mind. IF by analysing language we find evidence suggesting that nguistic knowledge" differs somehow from "non-linguistic knowledge"', that a distinction between the two a n be drawn in a non-arbitrary support the view that the mind itself draws a distinction tal dictionary" and a ccmental encyclopaedia". In this e that this is indeed the case, and that by examining e can learn how to draw the lime between "meaning" between '7linguistic knaw1edge"and "encyclopaedic y, until reoently the structure of the lexicon was not subjected d methodical large-scale study of a kind which might throw t on the organization of knowledge in the human mind. The main reafor this was the absence of a suitable methodology ;and also the wideF faith in the very possibility of developing such a articularly harmful in this respect was the attractive but not doctrine of family resemblances, which was put forward in his PhiIos~phic~l hve$1fga6fons and which has gained xtraordinary popularity in contemporary philosophy, psychology, anthropology, and also linguistics {see Chapters 4 and $1. As mentioned earlier, I believe that lexicographic research of recent years has proved Wittgenstein wrong on this particular point. Meanings do have boundaries, words can be rigorouslly defined, lexical fields with analogous semantic structures can

"'the inevitable collapse" of the "empirical approach" to "calar-category evo!utionEq. The choice between linguistic arbltrtxrhess and neurophysiological dieterminism in colour caltegorization is a false one. Human conceptualization of colour, which is reflected In language, may be constrained by the neurophysiology of vision, but it can be neither described nor expilain& in t e r n of neurophyiologr. To describe it, we need to take recourse to human conceptual unlvemds (such as SEE, TIME, PLACE, and LIKE). To explain itin both its variable and its lmnilul~rsal or near-universal fcaturcs-we need to pay attention to the way people actually talk about what they see, without limiting our data in an artificial manner, and without, as Saunders (1992: i65)aptlly put if, "reifying the neurophysiological into the phenomenal",

336 Lexica? Semantics

be uncovered, and, on this basis, more or less reliable and accurate cognitive maps can be drawn. Linguistic theories which deny the possibility of drawing the line between one meaning and another have the tendency to become seEEffuPf2ling prophecies: they make the study of the lexicon an unpromising and urnattractive enterprise, and so prevent the discovery of evidence which would lest their validity. The view that it is impossible to draw a line between "meaning9' and '%nowledge" or between 'Uictionaries" and "encyclopaedias" (sw e.g. Haiman 1980a, P982; Langacker 1987, 1990) has had, I believe, a similarly unfortunate effect on the study of the lexicon. For knowledge is open. ended. The belief that a dictionary definition represents nothing other than a selection from a (real or imaginary) encyclopaedia entry, with the choiwe: being determined by practical considerations and having no theoretical jusCificailtion, lcads to stagnation in lexical scmanlics. On the othcr hand, the hypoehcsis that the meaning of a word is dderminate, that it can be established in a nom-arbitrary way and clearly delineated, has the opposite effect. It encourages the semanticist to turn full attention to the lexicon, to examine carefully all available evidence, to come up with specific hypotheses, to look for possible counter-examples, and p n erally to engage In serious and thorough study of a language's l & d resources. But to be able to disoaver the boundaries between meaning and know!edge we need to have a clear idea of what meaning is and how one can establish the meaning of a word or linguistic entity of any kind; in other words, we need a mherent semantic theory and a rigorous semantic methodollogy. AS was shown in Chapter 7 , different types of words show dfierent Jerels of semantic complexity. In particular, there are words whose mearnimg can be portrayed directly in terms of primitives. [for example, emotion items), and words whose meaning is so complex that it can only be reduced to the level of conceptual primitives step by step. The bulk of concrete vocabulary, and in particular the names of ""living kinds" and c'culturd kinds", is of the latter type. None the less, allthough the tacit kmowJedge implicit in words such as m o w or crocodile is quite extensive, it can be separated, in a mom-arbitrary way, from encyclogaedic howledge about mice or crocodiles; and there are some types of information about denotata which can never become part of the folk commpt (e.g. information about the average weight, in grams or kilograms, of a particular kind of animal).

2. Abstract Conmptts m d Ccmcrate Concepts

apter 7, both concrete conoepts and abstract concepts usually much more complex than one might expect before trying to ts are usually sufficiently of the primitives, while an b concept written purely in terms of the primitives rehension (as a whole). I believe the correct concluat concepts associated with natural and cully complex, One can, of course, try to deny and chim that concepts of this kind are learnt largely by nsltitute "gestalts'" not conceptual configurations, ng in an encycllopaedia, not lact remains that explicat a kind of knowledge which native speakers of a age have and which is part of their communicative competence. (For psycholinguistic evidence of the reality of this knowledge, see e.g. mmkina and Mirkin 11986; Fmmkina and Mostovaja lsi8$.)1 Given that it is one of the tasks of a (monolingual) dictionary to portray, faithfully and accurately as possible, the howledge encapsulated in ge, it follows that iexplications of this kind belong, ry. Of course, practical considerations may impose breviations on the dictionary representations of ds, but this is an entirely diflerent matter, tical! line separating a 'cdictionasy" from an "encyds "dictionasyY'aand'kencyclopaedia" in a language-related "folk knowledge" (everyguage-independent scientific knowledge [and cerm y concrete reference works such as the Oxford E~ag&h guage-related folk know,edge is different, and shed in a principled way, from other kinds of knowledge at it can be articulated in a coherent and rigorous manner, and I will o do it with reference to a domain regarded as particularly dificullt and Ily the domain of "living kind^".^

Whether or not 'Tolk-knowledge"' in general can allways be distinguished, in a principled y, from "scie~tific knowledge'" is mother matter, which need not ooncem us here rut the omen!. (See Section 6 . ) 1 am lalking here about the kaowicdge which has become entrenched u n the words of a particular hguage.

Pt is widely believed that the names of animals acquire their meaning from science. The BEoomfielldian claim that am words acquire their meaning fro111 SC~CIICCIrias bccn generally rcjectcd in corltcanporary linguistics as selif-evidently unacceptable, and (as far as 1know) no dictionary has ever sought to define words such as hale,jear, or ~enderness on the basis of the latest findings of neurophysiolagy, as recommended by Bloodidd (1933/1935) [see Chapter 111. None the less, in the area of the names of animals and plants, the "scientific" approach has always been strandy represented, even in dictionasies, and is still widespread. For example, the Shorter Oxford Englid~Dictionary (SOiElD 19164) defmes hama as "a sol& hoofed perissodactyl quadruped (Equus cahahs]'" Wehater (1976) informs us that horse is "a large sdid-hoofed herbivorous mammal (Xquxnrs cabaISus) domesticated by man since a prehistoric period"', and the: Llomgmass. Dictiamasy [LDwOTEL 1984) finds it necessary to mention that the animal designated by the word horse belongs to the family Equidae, and that the word is applied specifically to animals "over 14.2 hands in height". It seems hardly n~ecessary to argue that scientific definitions of this kind do nor represend: the native speaker's concept. Science is, or tries to be, oniversa1 and to reflect the know1edge a ~ c m u l a t e d by mankind as a whole (and, more specfidly, by the professiional experts in different fields of knowledge); languages w a r e not universal, and each of them reflects the experience of a particular part of mankind, united by a common culture and ar c o m o n existential framework (and not the experience of any llocah m p r t s but that of the "people-in-the-street"). This point was argued by the Russian linguist Jurij Apresjan, in his masterly work Lexical Semantics (1992: 32-3, 35):

In the area of names for animals, the language-specific character of the onccpls c~~capsularcd inr words ~lma~lilksls ilslclf oa a number of difrercnl s. To begin with, the basic categoriratioa of the "animal kingdom" diUer considerably from language to language. For example, in lpiri the basic categories lexicalized in language are these: (1) kuyu -meat, creature with edible meat (2) pama -edible, not meat, doesn? grow out of the ground (3) jwrllpu- flying creatures with feathers data are from Hale ee: aI forthcoming; the phrasing of the glosses is
.) K q u includes not only edible "meaty" animals such as kangaroo,

The folk picture of the world that developed in the course of centuries and includes h l k geometry, physics, psychology, elic., reflects the material and spiritual expienoe of a people (native speakers of a certain language) and therefore is languagespecific in the fojlowing two respects. First, a folk picture of a certain portion of the world may be ~:rucidly dtsFenenr from a purely logical scientific picture of the same portion of the world that is shared by speakers of a variety of languages. The task of a lexicographer (unless he wants to g0 beyond his discipline and turn into an encyclopedist] consists of disoovering the naive picture of the world bidden in lexical meanings and presemting it in a system of definitions. Seaand, folk pictures of the wodd, obtained through analysis of meanin& of words in various languages, may dif"Eerin details, whereas a scientific picture aT the world does not depend on the language used to describe it.

sects such as honey ants; it also includes to such creatures, such as honey or nectar, but it (the d a s s i h t i o n is not exhausr example, doesn't tnclude emu. ,the categorization embodied in the lcxlanguage. Far example, Japanese and n mice: and rats, and for native uite difiwlt to learn the difference in iwh words rat and mousese. lOrr the other ally between the moths that eat t fly around lamps at night, which native speakperceive as two totally different kinds of insect (moQ and &my). types and the 'ToPk knowledge" embodied in the names of anilanguage to language. For example (as men, in Russian iJak ("donke;y"")pitombes hard work, whereas "works (and eats) like a horse" rather than "like a donkey". hard w r k is epitambd by a! buffalo, in Malay, by a bullmk, and m. In English, cats are thought of as c o r n o n pets (as well of mice), but in Warlpiri they are thought of as kcryu, 'keaeat"'.

pulsive creatures spreading diseases among people. But in words glossed as "hopping mouse, rat" [and regarded as quite different connotatiomns for the native speakers: burrows. It digs a big burrow. . , . Wc kill them L o eat. We cat them. The meat of that small animal is good. They dig down in the burrows to find that mima1 and they kill it in its burrow. They take it to cook itY"IHale er a[. forthcoming).


An 18Ius1trration:Folk Elilia versus Scientific Micle

To show more dearly the diflermce between scientific knowledge lrnd kind of knowledge which is encapsulated in a folk concept of a natural k I will adduce here one detailed illusltration. This will take sonably full explication of the folk concept mice (a revised version proposed in my Lexicography and Conceptual Anai!ysi8, 1985), co with the full entry for mice i~ the Encyclopaedia B r i t m To show that the folk concept of 4mmwse>arallels in folk conoepts encoded in other animal names lnre would have to here, and discuss, many other such explications. Since this is i reasons of space, ~tlne reader is again referred to Coancepfual AmJ'y~is.~
M I C k n expIica~iono f the folk concept a kind of creature people call them MICE people "Chink that they are ail1 of the same kind because they come from other creatures of the same kind people t h i k these things a b u t them: they live in or near places where people live k a m s e they want to eat things that people keep for people to eat people don" want them to live there (some creatures of a similar kind live in fields) a person could hold one easily in one hand (most people wouldn't want to hold them) they are greyish or brownish one cannot notice them easily (some creatures of this kind are white some people use them when they want to find ou creatures when peoplle do various things to them some people keep them in or near their houses bccausc [Eucy like to watch thcm and to look efics thcruu'p they have short legs
Both the lexiwn and the syntax oF this explication are quite complex. simgify the language, we could do so fairly readily, but at the: oost of lengt cations considerably, and making them much harder lo read. For example, 'because:they come from creatures of I t b same kind' we could say: bemuse before they were things of this kind they were parts of other creatures of this kind

becausa: of this W E I C ~ dhcy move onc can't sce their legs moving it seems as if their whole body touches the ground because of this they czlm get quickly into small openings in the ground they are soft they can squeeze into very narrow openings their head looks as if it was not a separate part of the body the whole body looks like one small thing with a long thin, hairless tail the front part of the head is pointed it has a few stiff hairs sticking out sideways here are two round ears sticking up one on eilthea side of the bop of the head they have small sharp teeth that they bite things with they don't want to be near people or other animals BEHAVIOUR when people or other animals are near they make no noise they hide from people and animals in places where people and animals can't reach them animals of another kind living in places where people live [cats) want to catch and kill creatures of this kind people put special things in or near their houses to catch creatures of this kind and to kill them when they are caught they make little sounds it sounds as if they wanted to say that something bad was happening to them they move in places where people live looking for something to eat they can move very quickly they can move without making noise sometimes when they move one can hear little sounds it sounds as if something light and rigid was moving quickly on something hard sometimes one can see very ssmalE, dark roundish bits of something (dung) in places where they haw been eople think of t h like this: RELATION they are small creatures TO PEOPLE they are quiet they don" want people or other animals to come near them one cannot notice them easily they can do bad things in places where people live i m yeilowish stuf3 of a certain kind (cheese) they like to eat h that people eat Before we proceed to an encyclopaedia entry for mice, a few brief comments are in odes.

The explication of the h l k concept proposed here starts with four components, labelled here (for the sake of convenience] as "caLegory", "name", "essence", and "'origin". Jointly, these components present mice as what Bedin (19%) calls a "folk genus": a category of living things which is thought of as having a biologically transmitted inherent nature linked with a name (see Chapter 112). What follows is a series of camponenits spelling out what Putnam (119751 callls the stereotype: what people think about mice. A s with most other stereotypes of animals, the stereotype is organized around the following signposts: habitat, size, appearance, belhaviour, relation to people. The sequence in which the components sure given is not a b i trary but seeks to elucidate the internal logic of the folk mncept (see Wiembicka 19851. All this is very different from a typical encyclopaedia entry, such as the one which Follows.

haadled adequately and most safely by trapping. Mice, especially the together with their near relatives, the brown and house rats, are r emormous econmic damage annually. Even those living in natural so numerous in limited areas as to become serious, although usopests. Except In abnormal circumstances, however, mice

MOUSE (an encyclopaedic dewiption] An imprecise term designating any small rodent but often meant to apply b the common house mouse (Mus musculw), the type of the genus MSCS and the fanmily Muridae. In North America most species of the widespread and varied family Cricetidae also are called mice. Specific kinds of mice are usually designated by a compound term such as harvest mouse ~MlcrramysIof Europe; Reishrado'oJstomysd America), wood mouse (Apodemus of Eurasia], whitefooted mouse (SJeromysccrsef America) or pocket mouse (Perog'oJsarh~ of North America). Mim are indigenous to almost every land area and in a given area are likely b be the commonest of mammals. Some species are of narrowly restricted ocxummce and habitat; others are Yrride-spread and versatile. The genus Mws, for example, occurs naturally on all major land masses; the typical s p i e s has been distributed by man to a11 inhabited areas of tbe earth and has became naturalized. One species of Peromyscw occurs over most of North and Central Amerka from the subarctic to the tropics, in swamps, deserts, forests, mowtains and prairies. Mice eat a variety of foods, some oonsurnirrmg almost anything edible-seeds, vegetatiorrm, arthapods and Besh when availabb. Tbeg are in tum preyed upon by a 8 1 1 manner of larger carnivorous mammals, rapacious birds and reptiles. Mice constitute the most important prey group of any of the mamamais. Mice mature relatively young; the house mouse typicallly is ready to mate two to three months after birth. Gestation periods, averaging about three weeks, is less than two in some species. From l to 1 1 8 young comprise a litter, the size of the Bitter depending upan the species, the number oT3itlers already produced by the female (the second or third is usually the largest) and the season. Breeding may take place a m any season in some species and be seasonallly restricted in others; in a widespmlead species the climate of a given region is usually lthe determinant. Some species are social and live in common burrows or in colonies. Others are solitary. Even in the social house mouse, however, excessive: crowding produces metabolic disorders and abnormal behaviour patterns that may result in the decimation of the population, a phenotnenon thought to be rebatedl to the well-knom periodic '%uicidaI" migrations of lemmings to the sea @ee LEMMING]. Many mice, notably the house mouse, seem to prefer dwellling in man-made s t w

M a n y other variant strains of known genetic andlor nutritional history have eveloped for experimental purposes. (Encyclopaedtb Britlmnica, 1969, xv.
According to the Encyclopaedia Britamim entry r e p r o d u ~ e d above, o w e is an "imprecise term" which doesn? correspond exacdy to any taiogical tsuxon. I n natural language, however (that is, in ordinary English), is not felt to be any less precise than folk genera such as horse, rabbit, o r tific category M m mlrslcuhs does not correspond to the folk c a t reflected in the concept ?nouse? in everyday English "the commouse" [Mus muscuh) Is not thought of as a specific variety ''Mus", and the expression "lkouse mouse" is nat used it is precisely the "house mouseY%hiclln is normallly callled "field nuice" or "'white mice" being seen as horizankd extentegory (see Bright and Bright 1969; Hunn 1476; Serllin t as other varieties, on a par with '%ouse mice"". , the folk concept 'mice'stands for what ordinary people see as a ture" different horn scientific categories and identifia m ('people call them MIECE'), and of its presumed

paedia entry For rmostse, one is struck by the mass of

maximum size of' a litter, on metabolic disorders overcrowding, and s o on. Oln the other hand, a great deal of n which is essential to the everyday concept is missing. This

ser"') documents the psychocats, which is further supported r sayings such as "when the s that a reference to cats should s a reference to mice should be English).3 But there should also be another kind o fcompdim turaI dictionaryE')which would explicate the everyday concept of ' some componemt such as 'anipeople live want to catch and

5 . The Evidence for the Folk Conapt

How do we know that th The methodology of given folk macept L a ~ornplex topic which cannot kw discussed hare length. (For more detailed discussion, see Wierzbicka (1985) and Chap 1I; see also Apresjan (197411992). I will, however, adduce here examples difierent kinds of evidence that can be used). Ta begin with, there words and phrases s to many dictionaries, there is also a verb to mouse, which can ct to mouse-hunting by cats (and perhaps ow1s), activities: "00 go or move about softly in search something, to prowl"'(OED), 'Yo search furtively for somethimg, to use attests to the pqchological reality of the seless, furtive movements i n search of foad. mome-ear, noted in m n y dictionaries, stands for "army of t have soA hairy leaves" (LDOTEL), Obviously, this is y with mouse ears, and thus it constitutes evifor the psychological reality of ears in the folk concept of m o w . rly, the cornpound moasa-hit ('b plant of the genus Myosrrsw, esp. nssnimw, Born the shape of its seed teceptade"",ED) attests to the concept of mouse. (See d s o the reference to e about "the thtee blind mice")). e m t use of the word mouse to refer to a iemce of certain aspects s characteristic of mioe. In a that '"he whole body looks

compact, light little '%ody"with a long ck, light, noiseless movements, fib both case d many other names of animals, lexical evidence comes also of wards semantically though mot necessarily morplrolagically]~ names (in Cruse" (1 986) terminology, c c e n d o n ~ s " ) . e is a whole Family OF "horse wards" reins, stirrups, star, stud, steed, ey, m ~ ~ t a ncavalry, g, and so on (see Wierzbicka B985: 201). In ch endchnyms (other than those 3 mouser), but there are words a; these include scurry, squeak,

344 Lexical Semansfcs nibbk, and perhaps p caw per. These words give evidence for various as of the folk conoept, such as character ties. omo on phrases such as quiet as a also a source of evidence. The first of these t perceived quietness of mice, and the second, th human houses as a place which (unlike a chu nent source of f a d . Another source of evidence is to be found in conventional metaphor person [nomaBly a woman) who is called a! mowe is perceived not very noticeable, not drawing attention to herself. As ment metaphors of this kind are partly language important aspects of the folk concept. Often, valuable evidence comes fuam proverbs and c o r n o n sayin such as the one about carts and i c e cited earlier. Literature, in particular poetry, is also although one which has to be used ations can be separated fram shared mixture of fantasy and stereotyped every useful in this respect. For example, the w mim, see how they run" highlight the importance of quick and ap enartless movement in the h l k concept of kouse" and so does lowing rhyme: Hickory dickory dock, The mouse ran up the c k k , The clock struck one, The mouse ran down, Hickory dickory dwk. Finally, we can mention ethnographic such as h e appearance of mice in loommn c a r conspicuous round ears); the schematic drawi Pang )thin tail at the end of a compact, vi little body; common children" games such as the farmer's wife, a mouse, and cheese; and so on. It is important to emphasize that t established simply by putting direct qu knowledge which we are trying to establish is tacik knowledge, which den below the surface of conscious the surface through painstaking, rneth longed dialogue with a wumber of inr more as a method of verification than as a direct "'dscoveuy For example, one reviewer of the earlier explication of mfce (in

I I . The Semandcs o f Natural Kind8 347

485) has objected to "the statement that mice make no noise when movaround'hnd pointed out that "many people think mice make scratchy, '(Malt 1987: 266). Other informants with whom I have since this point have iagwed with the reviewer, and the explication has ended in the relevant mspect. I presume, however, that the reviewer ea of the "scratchy, mstling noises" by exploring her own tions rather than by asking large numbers of inifomants. Obviously, and warking with informants. But er, not the latter, which needs to so needs to be stressed that methodical introspection can widely used procedures hvolvparticular with linguistic evidence of various kinds. (For

6. General Discussion
methodical exploration of folk concepts, using all available evidence, allows us to delineate their contours with a precision pearances to the contrary, is simply inawessible to an encyentry. The editor of an encyclopaedia has to decide, in an include from the mass of to arrange the information d so on. In investigating folk psiition is quite different, the task of choosing what to includq the full concept as it really is, using trying to use exclusively slimich is being explicated. These ncept fully, and to do it as far as posleft for individual choices pt accurately we need that attention Zolkovskij (U964b1,who callled for dividual words. From the present view, however, the metaphor of a portrait is not fully apposite artistic freedom. By cmtuast, an houlld leave no freedom of choice; er compared to that of an ine of a statue or an artefact hid-

(1982: 3.54) has argued against ""the separation of dictionaries opaedias" on the grounds that "there are no hard facts, and all

I d . The Semwdics o f NQEUMI Kind8 3491 scienm is ethnoscience" (3371, and that, for example, "our present knowledge of cats and elephants is as provisional, and specific to our culture, as the definitions of words like care and taboo in the languages where they omur" (337; see allso the exchange between Frawley 1981 and Haiman 1982). ]But the conclusion doesn't follow from the prerniss. The encyclopaedia entry for mouse quokd earlier may indeed be ""provisional'" '%specific to our culture", and, E would add, based on a number of somewhat subjective and arbitrary choioes. Yet the purpose of an encyclopaedia would not be served by replacing entries of this kind with the explication of the folk concept d the kind outnined in this chapter. PwadmicaIly, of the two, it is the dictionary entry$ not the encyclopaedia entry, which can be said to be "objective'bd non-arbitrary, and to represent a '"ard fact", Psychocultur;al fact, of course, not biological fact. An encyclopaedia entry for mouse may be provisional, biased, and subjective in its choices and in its emqdmses, but it doesn" aim a t establishing psychocultural facts; it does not aim at dimovering conceptual sltrwtures. Enqclopaedic knowledge is cumulative and inexhaustible. By contrast, the meanings of words are discrete and finite. They embody a special kind chF howledge [and pseudoknowledge, such as that about mice's fondness F lc F r cheese), and they consdtute a vital point of rekrencc far both communication and cognition, B t is true that the meanings of names of living kinds-unlike, for example, those of emotion t e r n 4 0 have a masure of indeterminacy, since the amount of "folk knowledge" encapsulated in a naive conoept may vary somewhat from speaker to speaker (see Gal 1973; Gardner 1476; Mays 1976; and also Locke 169W1959: 82). But there is a limit to this variation. For example, information such as that mioe are ready to mate two or thee months after birth, or that gestation period averages about hthree weeks, cannot be part of anybody's "najve concept". It can, of course, be part of their individual knowledge, but nol of that mnceplt which they themelves see as a shared stereotype, on the basis of their life-experiences in the community (see Wierzbicka 1985: 212-18; Tyler 1978: 2331-48). This is not to say that living kind concepts such as 'mice"ave, after all, the same kind of semantic structure as, Tor example, emotion concepts. The two domains are fundamentally different, and the greater variability of living kind concepts constitutes one of the important diflerences between them. But it doesn't follow from this that living kind concepts cannot be defined (or explicated); and from an explication the differences separatirrmg them from scientific concepts [and from scientitic knowledge) can be clearly seen. In rejecting the (theoretical) distinction between dictionaries and encyclopaedias Harman (1982) was in fact defending the view that natural kind words such as mice or J F F o ~ s ~ s embody a great deal of "cuitural knowledge* and that they could therefore be defined. In saying this, he was arguing against the earlier claims of Kripke (14'72-31, P u t n m (19751, and others, including myself, that natural kind words are like proper names and carinot be defined. I bdieve now that Haiman is partly right, and that natural kind words (such as mice] a n indeed be defined, explicating the cultural knowledge encapsdded in them. (The reasons why P have abandoned my I972 position in this regard are set out in Wiembicka (19851.1 But a line can be drawn between, culltural howledge which has become deposited in language itself and other knowledge-whether scient'fi a c or non-scientific. Hairnan argues that "the distinction between linguistic and cultural knowledge" is misconceived, and I would partly agree with this, in so far that, for example, the "ling~~istic" knowledge about mice, spelled out in the explication proposed in this chapter, represents cultural knowledge. But there is also knowledge about mice which is not part of the folk concept reflected in language-and a line can be drawn between that knowledge and the knowledge [and ideas] encapsulated in the word mouse itself. The fact that different languages draw such boundaries in different ways demonstrates that these boundaries can indeed be drawn. For example, if Japanese doesn't distinguish lexically between mice and rats, or English between clothes moths (in Polish, nroCe1 and other moths (in Polish, tmy]~, this shows that semantic boundaries between differear, living kind concepts do exist, and that they are different from those drawn by biologists. Hajiman's claims that "dl science is ethnoscience" (1982: 337) alnd that "the difference between everyday experience and scientific experience is a digerence: in degree of precision and generality" do not affect the present argument: the question is not how to draw the line betwen science and etkuoscience, or between cultural knowledge and linguistic knowledge, but how to draw h e line between knowledge and ideas which are encoded in language and knowledge and ideas which are not.


Trying to discover how knowledge (or at least basic, 'Youndationd" lknowlledge) is stored and organized in the human mind, we can rely in considerable measure on language. There may be concepts which are not lexicxlized in natural language, but these are probably less common, less basic, less salient in a given speecln community than those which have achieved lexicalization; they are also less accessible to study. Words provide evidence for the existence of concepts. Lexical sets, sharing a similar semantic structure, provide evidence for llhe existence of cohesive conceptual wholes (or fields). Ef it is hypothesized t b t knowledge is organized in the mind in the form of "cognitive domains", then conoeptual fields detectable through semantic analysis of the lexicon can be regarded as a guide to those domains.

The organization of cognitive domains is reflected in language, aad above all in the structure of the lexicon. The lexicon of a ianguag is the speakers' fundamental icagnitive resource; it is a treasury where the shared knowledge of the world, and the shared models off biollogical, mental, and social aspects af life, are held. Exploring ths: lexicon in a systematic and methodical way we can dkcover how "ordinary people" (in contrast to experts and scientists) conoeptuallize the world; and we can learn to discern the line which separates tanguage-related everyday knowledge from the specialist's knowledge, which is--or should be-largely language-independent-

Semantics and Ethnobiology

1. Introduction
It is widely agreed today that 'cculture Voes nolt consist of things, people, behavior, mi emotions', but the Eoms ar organization of these things in the o;pEey"(Frake 1942: 85, with reference to Goodenough 19573. ause disagreement is this: How can the orgaruihe minds of people k di%orvered? d yet much neglected-path to discovery lies the area of language, and that there is a whole battery of linguistic tests ~ a different l aspmts of the organization af ople. In this chapter, E will try to show the ach with respect to some of the basic issues and detailed study of botanical and zoolagical nomenclaworld (especially that undertaken by Brent ealed thak different societies differ corusidceptualization of the biological universe; but it has also of strikingly regular structural principles hssifhcation which are quite generaIY"i(Berlin e e ba i . 1973: nt lines, univenals have also been suggested by k i E of eithnobiologjcal universals and the ensuing ated interest in the conceptuaPization of plants s, md they are largely responsible for the key position of this domain in current anthropology. In pasticuiar, the recent interwhich is rapidly becoming one of the most topical issues in mgrmitive science, is focusing very much on this particular domain (see e.g. Gelman 1992; Keil 1989; Hirshfeld and Gelman 1994). In the asrent debate on the question of ""whether there are domain-specific cognitive universals that amount for the peculiar kinds of regularities apparent in folk systems of knowledge and belief. . .or whether those regularities are the product of general prooessing mechanisms that cross . . . domains" (Atran 1990: 471, the domain of '"living kinds" is accorded a spedal place and is often treated as a natural testing-ground. One issue which attracts particular attention in this context is that of the

e D


" e w A

E- . :

2 L E*


f g E.Z I f E : ga3 Y" a@ 8. F 2 sg

; Ee ! =


2- ?-

E g g 8

I EB. t;- w
P ,



' = - v l

a; i%8


5 % El &

* g5.
2 2-


11987a) shows that direct questioning of infomants, and even expelrimenlhs involving informants, can only be a subsidiary methodological tooll. Information coming directly from informants is often superficial, unreliable, and mislleading. Linguistic tests based on methodical! examination of usruyc (such as, h r cxampllc, bhosc crlvisuged in the elassic smudics by Conklin (1962) and Frake (3962)) and on systematic examination of lexical and grammatical evidence (such as those applied in Apresjan (1374r'1992], Berlin (1992), or Dixon (19821 3 provide more reliable results. To illustrate, in McCloskey and Glucksberg's (1978) sessions with college to t h e students ilt emerged that 3 per cent of their informants said '"No"" question "Are geese birds?'" whereas as rnany as 1 1 7 per cent said "Yes" ' 0 the question "Are bats birds?" McCloskey and Giucksberg draw from such data the condusion that the category "bird" in English is fuzzy. In fact, however, closer exwinadon of linguistic evidence shows that this category is not fuzzy at all, and t h t while geese may be thought of as atypical birds, they are none the bss conceptualized as 'b kind of bird", whereas b t s are not conceptualized as birds at ail (see Chapter 4 and Wierzbicka 1985). The crucial point is that, in certain circumstances, an individual goom can be referre$ to as bird (as a turkey a t a Christmas dinner may well be referred to as "the bir8'1, whereas a bat cannot. Similarly, all the infomants that McCloskey and Glucksberg worked with agreed that butterflies" are iasecbs, while 3 @ e r a n t stated that bee3 were not insects, and 7 per cent that bars were insects; for Psh, 60 pier oen2 of infomants classifiedjeIlyfisSz!as fish, 1 1 0 per cent daasikd alKgators asfssh, and 2 per cent stated that tuna and saImon are notfish. This, too, was imterprreted as evidence for the alleged fuzziness of human thinking, and for the lack of clear boundaries between different categories. But in fact such results am artefacts of the particular techniques of investigation, Generally speaking, directly dassiffimtory questions ("'Is an X a I"?' or ""What is an X?"') are unhelpful in the malysis of folk taxonomies. What is more illuminating and reliable is the acceptability of sentences referring to individual creatures, such as "Look at that fish over there?" said with respect to jellyfish, or "How many birds can you see?"with respect to groups including geese, buts, or bees; as d l as other kinds d linguistic evidence (such as those discussed in Apresjan 1974/899.2). Similarly, Dougherty ((1978: 78-9) reports that all students in one of her samples placed butterflies in, the insect category (as does, incidentally, Cruse (3986: 136) in his discussion of English folk taxonomles]~.But presumably, the same informants wouSd not say "Sea that insect over there?" or "What a beautiful insec&l3%th reference to a bufterfly. This suggests that their responses were confusing scientific categorization with everyday language, and misrepresented the subconscious folk taxornomy reflected in their actual use of language.

As pointed out by Boas (1911) and Sapir (19271, the categorization reflected in language is unconscious, and it is for this very reason that language is such a revealing and valuable guide to culture and to social psychology. Direct qliwestioning of informants appeals to their ccmsciows kmowlcdgc, not lo lhdr tacit k~aowlcdgc,and is unlikely to reveal much about their unccmscious canceptuallization of the world, whicE1 is revealed more clearly and reliably in linguistic structures and in linguistic usage (see Randall and Hunn 1984: 3331."

3. Types 0f Linguistic Evidenoe

There am rnany types of linguistic evidence which provide clues to the conceptualization af the biological universe. I will not try to undertake here a wmprehensive survey or detailed discussion (see, however, Wierzbicka 19851, but the five Eypes listed below [Sections 3.11-3.5) are particularly important. The last two of these have already been introduced, with respect to mice, in Chapter I 1. 3.11. Ways of Referring Qne @an refer to a rose as ""tat flower" or to a magpie as "that bird'" but in ordinary (non-scientific) usage one cannot refer to a particular tree as "that plant" or to a particular pmpkin, as "that vine". This shows that, despite many daims to the contrary, a tree (singular) is not conceptualized ins folk English as "a kind d plant", or a punrpkin, as "a kind of vine":
t that plant over there! It's the oldest oak in this country. 'Look a

Similar tests show, incidentally, that the recent wave of '"anti-Ros%:hianY' thinking sometimes goes too far, as when, for example, it is claimed that in the English folk taxonomy dissimilar creatures such as caterpillars and buttedies, or frogs end tadpoles, are classified together (Atran 1 9 1 9 0 : 153;
The difficulty in identifying lire fonns in diKerent rangnagas of the world i s well illustrated By Randall and Hum (1984: 334). I suggest that the operational critecia proposed in this chapter WOUM lead 10 much more reliable results, although the resulting repertoires of 3ife forms would not always coincide with those proposed by Berlin et aL (1973) or by C. Brown (1977, 1979). In particular, the diflerence between anonolelnemicallly and polglexernial~labelled m a , whose va'aedity n u importance Randall and Hunn 1(1984:34131 question, sbodd wnbinue to be regarded as important evidence pointing to direrences in the underlying conceptualizait should be kept in mind that "~unctional" and 'ktilitilean" mncepts such tion. Eurthennl~re, as vegetaa/eo r j ~ ~often i t behaw differently (in Einguistic usage) from "'moaphotypes" such as free or 16irdJzfienbicka 1984, 1985, 198h). On the other hand, linguistic evidence shows that some categories, for examplejffower, which have been judged as "functional"and disquali6ed on this basis as a potential life form (Atran 1987~: 32; C. B r o w 1977: 3201, may in Folk aaxonoang play a role malogous to that of genuine, taxonomic life forms such as bird, fish. or

356 Lexical Semanrrrf ics Gclon:~n auld lCoBcy 2991). For cxnlnl~lc, uole ~ a t ~ r say ~ o ti l l Emlgiish while pointing to a tadpole or a c~terpikr: *Look at that frog! *Look at that butterflyl Linguistic tests of this kind sllrow that while an AEsalim and a dachshwd, despite their dissimilarity, are put into the same folk category (dog), a mdpole and afrog, or a bburser$y and a caterpillar, are not. 3.2. Grammatical Congruity k a c b (1964: 41) has drawn a taxonomic tree of what he calls "the English language discriminations of living creatures", which implies that in English gigs are "a kind of farm animals", farm csnimnraEs are "a kind of l i v e s t o ~ k ~ ~ ~ iilvexrock is "a kind of tame beast", tome beasr is '% kind of beast", and beast is a kind of "laad creature", Many different tests could be used to show that this is an arbitrary scheme invented by the researcher, not a failthful representation of the clitsshficatiom embodied in the English language, One of these tests Is tlnat of "grmmatical congruity": pig is a countable noun, whereas livmt~ck is a mass noun ((cf,rhree pigs versus 'fhree Hves~ocks); on these grounds alone we can establish that a pig is not conceptualized in English as "a kind of livestoek", just as an apple (countable noun] is not conceptuali~edas "a kind of fruit" (mass noun), or a chicken (countable noun) as "a kind of poultry" (see Wienzbicka 1984, 1985, 1988~; see also Section 10 below). 3.3;. Morphological Structure The morphological structure of a linguistic expression provides an important clue to its meaning. For example, as pointed out by Mel%uk es a l 41984: 411, the two Russian expressions leJEa ('wife's mother', one word) and mar' feny Cwife's mother', lit. ' m o t h of wife', two words] do not haw quite the same meaning. Both SeSZa and mat' feny specify a certain relationship, but seFEa in addition identifies a certain (recognizable) kind of relationship and it is the latter, not the former, which is always used in the innumrablle Russian jokes about mothers-in-law. Similarly, as pointed out by Langacker 419931, the two English expressions wentan and deer m ~ a t suggest different conceptuallizations. Likewise in Polish, wolowim, 'beef', is not used in quite the same way as rnip.rn wolowe, literally 'ox meat' (referentially, the same thing as 'beefy). For example, in the context of meat exporls, one would be more likely to use the phrase msipso ~valsrwe,whereas in the context of a dinner menu one would be more likely to use the noun wofewfna. M i p o wolo~vedifferentiates this particular kind of meat from

12. Semaslrics and E

o 357

other possiblc kinds, wlucrcas walorrvi!~a idcnkifies a, recognizable standard kind of food. Considerations of this kind svnpporlt the significance of the distinction between "secondary Ilexemes" (such as blue spruce or scrub oak) and "primary Eexemes" (whether analysable, such as fulfp-tree, or unanalysablc onirs, sucln as J J ~ ~ J Sor ~ Jc/!,r), ~ drawn by Hcsbin, Brcedlove, and Raven ( 1973).

3.4. Phraseological Evidence

English concept of, say, 'butterfly" we can draw on the To ducidate tl~e stock of common English phrases and secondary lexemes, such as, for example, ''a sot5al butterfly", "butterfly kiss", ''bbuttefiy clip", or "butterfly stroke"; for 'rabbit' we can, and should, draw on phrases such as "breed like rabbits'" "a rabbit warren", 'Yabbit teeth", "rabbit mouth", "run away like a scared rabbit", and so on (see Jauncey 1990). The fact that the common English collocations involving mlosrse difler considerably from the Japanese collocations involving nezumi, houselrat', suggests that the two folk concepts in question are very diflerent. Analysis of the two sets of collocations shows in what respats these two folk concepts differ (see Miyokawa 1989).

3.5. Lexical Evidence

English has many different "endonyms" of the word dog, that is to say, words which are, so to speak, semantically derived from it (see Chapter 11 1)-firsf of all, names of kinds of dogs, such as ~pmfel, poodk,fox-terrier, jbuSE$09, boxer, dachshund, and so on, but also other kinds of 'Yog-words", such as bark, growl, muzzle, feash, and kernel. All these words provide evidence for some aspects of the cot3ceptwl~zatlonlinked with the English wad dog. The existenoe of special nouns for kinds of dogs reflects aspects of the folk classification, and shows that the conceptual hierarchy is in this case more complex than it is for any other part of the English folkbiological system (see Wierzbicka 1985 and Section 7 below). The h c t that English has numerous nouns for kinds of dogs (e.g. poodle, xpmiel, boxer, and so on] but no nouns for kinds of cats or kinds of mice suggests that the two domains are conceptua6xed differently. Moreover, it suggests that the domain of dogs, but not cats or mice, involves a special Bevel of taxonomic categorizationn (("'subgenus'" see Cecil Brown 1987; Wieszbicka 11985: 232-6). Roughly spcalking, a word such as poodk OF spaniel identifies a oertain kind of dog, whereas an expression such as bhxe whak, white nnorsse, silver fox, or bush turkey identifies a kind of animal [namelly, whale, mouse, fix, or turkey) and differentiates some subset or quasi-subset of the class of animals corresponding to that kind from other

12. Semantics and EtEnobiolog


possible subsets. This distinction between positive identification and differentiation, whose importan~e was first pointed out by Berlin, Breedlove, and Raven (1973), is. reflected in a number of ways in linguistic usage [see Wiembicka 8985). It is also often reflected in diachrony, since a cuiturd change may lead to a change in conventional conceptuallization, and henw to a linguistic change such as that from AIsarian l(adj.3 dog to ASsabjEQn (noun]. I am not saying that linguistic evidence is the only reliable guide b human conceptualization in general and t o human categorization in particdar. There are, of course, many other types of ethnographic evidence that anthropologists have traditionally relied an. But linguistic evidence is particuiart y revealing.

ernin, Breedlove, and Raven (19731 offer tflzammul as an example of a form, and mention animal only as a ""unique beginner", not as a life ut mmml-like q ~ d r u p e dcnmiwre, , or amprbsibiom-is a scientific , and doe~n'tbelong to the English folk taxonomy at all. This can ed lby linguistic tests such as the following one: Look at that animallbirdlfishl "Look at that mmaikrquadrupe&amphibian! cientific concepts such as mammal stand for dasses, not for individuals, d it is remarkable that although educated speakers of English are, so ta speak, bilingual (in "scientific English" and in "folk Engli~h'"~ and can mix elements from both in their speech, none the less they unconsciozlsly apply different rules to them and in particular do not use scientific concepts such as m a m a 1 with reference to individual creatures: What a beautihl animalhirdIfialh! *What a beautiful mammalEguadrupedlamphibian! Using the same linguistic tests, we have to oonclude that aniinul is not a "uarique beginner" in folk English, as it cannot be used with reference to individual spiders or ants. For exampk, seeing an insect on sowbody%solr one cannot say (except in jest): *There is an animal on your collar. Similarly, one cannot (seriously] say of a spider or a butterfly: *What a beautiful animal? "Look at that animal over there! If &ere is ;a zoological "unique beginner" in ordinary English it is creamre, not animal, animal being rather a life f o m , on a par with bird, fish,snake, or &sect Qcf. phrases such as "animals and birds", or series of children's books such as "Auimals of Australia", ""Birdsof Australia", and "'Fish of AustmQia"'). The ward creature is perhaps not used very often in colloqluial English, but it is certainly there (especiallly in the plural) as shown by phrases such as 'kfl creatures great and srnall"flr by sentences such as the following one: Diving in the Coral Sea, one can see all sorts of strange and fascinating creatures. If the word creature did not exist in the English folk taxonomy reflected in ordinary English we could still speak of a covert category: "omething that can move and feel' (mavens senriew], opposed to the covert category oF 'things that grow out of tfne ground'. Returning to anha!, in the non-scientific sense of the word, it is

4. "Life Forms" i n English Folk Zoology

How many "life forms" a n be found in the zoological folk taxonomy embedded in the English Ilanguage? And by what criteria can we reccogniiue them? The first criterion proposed by Berlin, Breedlave, and Raven (1973) and Berlin (89811) is that of number: life form categories are few in number. "Ibis is useful, but it doesn't really help us to decide on individual cases (abr example, on shg, ssrake, spider, or busterfly). A second, and operationally much more useful, criterion is that of " p l y typicity": a life f o m is a category which is thought of as comprising mamy different (named) kinds of entities. Pipplying this criterion to the folk zoology embedded in English, we could identify the fdliowing as life forms: srnC ma/, bird, j k h , snake, and perhaps insect. All these words naturally invite the question "'What kind of (animal, bird, fish, snake)?" and all are readily used in reply to questions of the form "WWt is a Cplatypus, emu, salmon, python)?" Interestingly, the words bug and worm are normally not used like that; rather, they suggest a lack of interest in the identity of the creatures in question: There arc bugs on my windscreen. ?What kind of bugs? There is a worm in that apple! ?What kind of worm? In mast contexts, questions inquiring about the genus of b;ulg.s or worm sound either jocular or scientific, which is not the case with unfamiliar birds, fishes, or snakes, or with wnfmiliar animals in a zoo: is that? What kind of bir&fishr'snake~animan/?wor~bug

360 Lexical Semmfic-v

12. S"erir~antics a d EfRnobia!ogy


interesting to note that. ir is not just a col'loqnuial eq has sometimes been suggested in the literature, For ex easily be referred to as an animal, whereas a spider neither can a slug, a snail, or a butterfly, On the other hand, human b are mammals, but they are not amimoh in the mryday sense of the Atran (1987a: 55) writes: T o r the most part, vertebrate life f o m spond to modern [scientiBc] classes: mammals, birds, fish, eL. . . . often, folk views on the extensions of these life forms diffe callly construed extensions of the corresponding dasses in from the folk viewpoint, are rattnea marginal cases . . . bat, ostrich, etc.'. But in fact mammal is not a part of the folk-English tax foEk English conrcept animd, whose extension is reallly quite that of rnamrnoJ', does inc1ude whale and bas (as well Furthermore, in the present-day folk-zoological system, though a highly atypical one. In some ways, then, Englis is closer to the scientific one than is usually assumed. In olther ways ever, it is much further from it-but to sea this it must be recognize words such as mammal or qundradped belang only to the language of ence, and are not part of the English faRk taxonomy at ail. It is worth redling in this connectjam Russell's (1948: 83) re words such as dog or ant are probably learnt by oslnsi such as quadruped or aniimajl ("in the sense in which it includes o limpets") are definitely mot, Russell doesnY spell out the intende iizatiann, and it might be mnje~tusred that the distinction t was similar to &at between ""Talk genera" and higher-I presumably life form terns such as bird or tree are also learnt ostensiom"; on the other hand, scientific concepts such hn, or animal (in the sense including spiders] are cle way. (CK Chapter 7.)

nds of spiders, ants, or snails. as redback spider, or funnel web spitain circumstances9 '" saw a redback is kind are elliptical, and redbnck is used here as fier in the secondary lexeme ~ d b a c k spider, not as a primary Bexeme o w right. To see this, it is enough to compare the aaeptabiiity of e phrase a redback spider with that of *a swallow bird or *a troutjish. But if we allowed ~t~ails, spiders, ants, or bats to count as "'manogeneric life forms" "wan 11987b), we would be undermining the main operational criterion we have for distinguishing folk genera from life forms: the linof kinds) of t h latter. It s m s B i n el a t 1973) that categories of generics" rather t h n "monogeneric life forms", that not all1 living kinds are thought of in however shaU1~4w".Atran has argued expljcesemt an exhaustive partitioning of the local flora" erics'hre, essentially, generics, not 'hmanoim has to be rejected (see Berlin 1992: 21 1). be rejected anyway, given that (in English, at least] bcrsk, vine, rb are not taxonomic life f o m terns (and that, semanticallly, r pumpkin ''a kind of vine"). ce of the categories under discussion in their semantics), it might be best eir special shtus (for example, by callling e this). In any case, it is misleading m d "manogeneric life fomsm",&e@;ause the notion "life nk is particularly usefull and particularly revealing ind includihng many kinds")-indeed pera kind of living thing there are many kinds of things of this kind all these kinds '%we their names" (i.e. there is a word for every onc of Lhesc kinds)

5. Are there Manogeneric "Life Forms"?

But if creatures such as spiders or snails are not thought o (folk-English] life form do they come under? Clearly, slrakes,@shes, or bids. Could we assume that they aw om right? Amesag the many different criteria for life Form whic gested in the lirteraltpure the taxonomic level (the first one "unique beginner") has often been mentioned (e.g. Atra this criterion, spiders, ants, or snaih could count as life Bb ever, they are not (lexicallly) polytypic: ordinary Engl

6. "Life Forms" i n English Folk Botany

ow more specifically to English folk botany, I will argue-in dist with most other writers on the subject-that only one of all the life forms usually mentioned in the literature is reallly thought of kers of English as something that comes in many different kinds an therefore be recognized as a tavronomic life farm, namely tree. Even

362 Lexica/ Semanrics urban dwellers in America who cannot tell an oak fro whom tree is psychorogically moue salieot than oak, Domgherty 1978) are aware that there are many diffe trees, as there are many different named kinds of bir& or $shes, and question ""What kind of tree is this?'"s perfectly natural to them, unlike question "'What kind of maple is this?" This brings ws to another putative csiterion for distinguishing life fo from folk genera: thalt of psychollogical salienoe. Berlin, Breedlove, Raven (1973: 216) described generic taxa as the basic, and psychologi most salient building-blocks of a11 folk taxonom Dougharty (1978) and others have showtrr that for m bird may be psychologically more salient than oak or 1 have argued in detail elsewhere wiembicka 19851, we do not have to clude from this that for these speaken tree and bS Psychological salience may be a characteristic feature of fol many s p h communities (especially those living largely in a na human-made, environment) but it need not be, and should not be, for defining and distinguishing these categories [see Berlin 19 was, it would lose most of its value as a theoretical consltruct for cross-linguistic cornpadsons. The criterion of polytypicity is rably more useful and more illuminating. Assuming, then, that tree definitely is a life form in ordinary En what of the other putative folk-botanical life forms-that is, what of shrub, vbze, moss, grms, kerb, fern, or muskroom!? In same sense, all1 words can indeed be regarded as botanical "1i however, in which tree or biirdare life forms; B h a taxonomic supercategory. Take, for example, the word gross. On sometimes refer to "difirent grassesW",n the plural; but if ol speaker of English "What kinds of grasm are there?', unable to comc up with difrerent names. Some might mention b but this is a secondary lexeme, not a primary one For example, referring to grass, that someone "was sitting For most ordinary speakers, grass is simply grass. Not only do th know of any named kinds of grass, f hey are unaware that there are names, a situation quite ~dlifferentfrom that pertaining to free. Sinnil wvl~en one asks nativc spcakcrs or English ""What kind there?" or "What kinds of ferns are there?", they look baffled and are ga erallly unable to come up with any names. With barshes, the situation is somewhat different. A know that bushes can be of different kinds (e.g. lilac b biackberry brrskes, thorn bcrslges, gsrae Gwkes, nt but they do not think of roses, tihc, blackberries, of bush". Tyler (1978: 190) has actually defined lilac as

12. Semonrics and EtJsaobioJ'ogy 363

asked "What is lilac?' y "a kind of bush". though it can also

""Look at that tree!" (as mentioned eariier) but it is difto imagim anybody who would say "Look at that vine!", referring to

b,and so on. Conversely, when asked 'What is

so on)?", they do not hesitate to reply "A kind

'. Furthemore, inlfcrmants do not hesitate to accept sentences in

s to an individual tulip, daisy, or daffodil, as in ing growing out of the ground, but in real as the "part" sense. Intuitively, above all, a kind of thing which grows out of the ground, and mes In many different kinds, just as a free is a kind of thing which omes in many diffemnt kinds. (1987a: 32) argues that "when living kinds enter the space of with human function anldl use, such as eating, gardening (weeds n), entertainment (pets, circus and omic importance". E believe that,

364 Lexical Semmbfcs

generaly speaking, this is quite true, and as I have argued in detail (Wierzbicka 1984, 1985, 198&;crE, linguistic tests show that, conoeptua1111y, appEes are not "a kind of fruit", nor are CarroEs "a kind of vegetable'" But the same linguistic tests show that just as oaks, birches, or maples are, conceptually, "a kind of tree"~td@s, hJfodiQ, or roses are, conceptually, "a kind of flower'" aand that both tree andflower are, conoegrtudly, comparable "kinds of things that grow out of the ground". The circumlocutjion '"thing which grows out of the ground"'may seem clumsy and unnecessary: why not say, simply, plant? But the point is that plan6 does not redly function as a botanical '%aique beginner'" ordinary English. In particular, trees are not thought of as 'Vi kind of plant", anad neither are mushroom. When asked whether a tree, or a nzuskroom, is a kind of pbnt, educated informants may of cowwe reply in the affirmative, but they do not aocept that plant wuld be used with reference to an indivldual oak or birch tree (cf. Section 3.1 above): Look at that plant over there! Which one? *That oak treeithat mushroom.

12. Sewramsics and Efhnobiologg! 365

further construct: a "quasi-life form" as distinguished from a taxonomic, '" We also need some additional criterion distinguishboth "life forms" and "'quasi-life forms". I will try to such a criterion in Section 9.

7. Plollytypic Genera
b m s (such as bird fish, tree, orflower) differ from in being thought of as polytypic and in including many named el. This characterization, however, needs to be supnal criterion, referring to the position of a given o other named categories. To see the need for such itional criterion, it is useful to consider the concept of dog. existence in English of nouns such as spaniel or poodle is a particuthough not the only) piece of linguistic evidence showing that ught of in English as coming in many diflerent kinds. ;But rly, it would be absurd to conclude from this that dog is a life form in par with ranixscl.!, bir4 and Jfs'sh, and that spaniel and poodie , on a par with cot and COW. hought of as "a kind of animalm,and the phrase unds absurd, whereas "animals and birds" is pero the following sentences: "It's not a bird, it" an dog, it's an animal",] g felt, ~ ts involving co-ordination show that d ~ are e on the same level as cars or sheep, not on a higher level, are felt to be on a level lower than that d cat o m . or p ~ o d i e es cats, but Mary prefers dogs7', versus "*John likes cats, ary prefers spaniels"'. ic evidence shows that primary lexemes such as spaniel or belong to a s p i a l level lower than that of folk genera, though than that of specific taxa, normally represented by secondary lexh as Siamese cat or blue spruce. I have suggested that this level, y available only for taxa of particular cultural importanoe, can be subgeneric level (Wierzbicka 1985: 232-61; and a similar suggestion forward, independently, by Cecil Brown (1987). (Cf. Section 3.5 hen, that not only life forms but d s o some-rather excepbe thought of as coming in many different kinds. re, a proviso: not a E B biological categories thought of ;as m e d kinds are life forms. If a biological category comes narmed kinds but is treated linguistically [e.g. in conjoined phrases] on the same level as folk genera (that is, as categories which do

A pJ'ans (in the everyday sense of the word, and in the shingula small (much smaller than a person], and it has to be green. A pro p/t,trr is a snnall grecn IcaTy plant such as 2u poltcd plant; but a f
mw~kirsrom, is not just an atypical piant-it is not a pion$ at all (in English sense of the word planr). In English folk taxonomy, then, planit. is not a "unique begin it is a category comparable to bush, darub, gross, or moss. As cepts gross, mtuss, fern, or mushrclom it is quite obvious that they are thought of as included in the concepts of some folk genera subordinate them, bemuse in ordinary English there are no such [known, named) genera subordinated to grass, moss, jFern, or mushsoom. What are these categories, then? If they are not (taxon.ornic, lik forms, could rhey be unamlialed folk genera, like cacesrs, to shg? It seems clear that such a conclusion would be counter [partly for reasons discussed below, in Section 7). It is lntu more satisfying to conclude that categories off this kind are in more like life forms than like folk genera, but that they are not (polytypic) life forms.3 To sustain Ithis point of view, h o w m

366 Lexica! Semantics not come in many named kinds] h e n it is not a life form but a (rather exceptional and cdturallly salient) follk genus subdivided into named subgenera.

12. Semianlies c s m d Erhnobiology 367

9. "Hidden Matures" and "Proper Names"

One further possible difference between folk generic concepts and life h r m concepts is that the former-but perhaps not the latter-imply a "hidden nabure"ar an "underlying essence" which cannot be reduced to any observable attributes (see Atran E9XTa,b,c; Keill 1489; Kdpke 1972; Putnam 1975, 1977; Schwartz 8978). For example, we may think that some particular bird is a spurow, but if other people, whose judgement we trust, assure us that this bird only fooks like a sparrow but in Fact is not a sparrow [but another kind of bird), we are quite likely to accept their judgement, assuming that the "underlying essence" or the "hidden nature" of tfis particullar bird is not that of a sparrow. But could we simdarly accept the judgement that a creature which looks and behaves like a bird in f a d is not a bird? codd probably accept the judgement that a tree which e an oak in fact is not an oak (but another kind of tree), or a flower which looks like a tulip in fact is not a tulip but another kind ower. But would we similarly alcvoept that what Books to us like a tree stance but from dose by, in good visibility conditions] is in a tree but some other kind of thing growing out of the ground it doesn't have the '"idden naturey' of a free)? Or that what looks e a flower in fact is not a flower (but, say, a mushroom), because have the "hidden nature" of aflonarefl Similar questions mn be h respect to n~n-taxonomicquasi-life forms such as brrshes or in each w e the answer appears to be in the negative: we assume ow how to racognize a Sree, or ajower, or how to recognize a and we would not defer in such matters to experts; but with might well defer to an expert. ests that the meaning offfolk generic words like sparrow or opllk have some component absent fliom the meaning of life form wards. as bird or tree, a component which would account for the assumption idden nature". This component can be linked with the idea that a IFOR enus-in contrast to life f o m s and to various other groupings-provides " b r living things which belong to it. For example, the explition of the wad tiger could start as follows: a kind of animal there is a word for animals of this khnd h i s word is tiger assumes that the dividing line runs between living things and 3, has argued, i n apparent support of a presumed '%idden nature'" , that "a given tree may not be as large as a person, but atornab are

8. "Gestatants" a n d 'Distinctive Features"

It has sometimes been claimed in the fiterature that folk genera are conceptlaali~das holistic indefinable '"estaltsv" whereas life f o m s can be defined by m a n s of it few abstract features. For example, racoobssn is conm p t d i m d (it has been claimed) in terms of its unanalysed, global "ramonness", whereas bird can be represented i n the speaker's mind in terns of a few abstract features swh as feathers, beak, and eggs Nunn 1976: 588). But in fact, it is intuitively f a from clear that gestalt of a free, a bird, or a j s h , just as we have a glo ire/. Indeed, elsewhere Hunn (1977: 47) himself men wccoon, dog, and bmbr'e-bee-as an example of not a "deductiveve"' one. And indeed, if we can have a despite the wide variety of types of dogs, why sho global image of a bird? In fact, Bruner, Goodnow, and Austin (195% offer bird as their prime example of a holistic c and bill and feathers and characteristic legs. But . feathers, illhe bill and legs are highilly predictable. ID coding environment, one builds up an expectancy of all of these features sent together. It is this unitary conception that has the conffigur Gestalt property of 'birdness'." What can be said of bird of tree,$& or snake, as well as of dog, squir tions of life forms such as bird, free, or@sh a md C O I I ~ C ~ Ana/ysia ~ I U O ~ are at least as long as (in fact lon sqlrkrel (see Wierzbicka 1985). 1t is doubtful, therefore, t be distinguished from folk genera along the lines of "a tures" versus complex unitary gestallts. Furthemore, as pointed out by Cecil Brown (1990: have unitary gestalts for "cultural kinds" too, wh ones s u ~ h as bottle or jug or relatively complex ones such as The impression that artefactual concepts of this kind c terns of "a few abstract features" is just as illusory as forms such as free or bfrd could be so defined (for detail and other similar concepts, see Wierzbicka 1985).

still \trees %y nature"'((39815: 302).4 But linguistic tests show that bonsais are not thought of in English as "trees". Just as a crob-ap$e would not k referred to, in ordinary English, as "an apple"', a bonsai free would not be referred to as "a t r e " " 5 a sapling could not be so referrled to): *Look at that apple! "Look at that tree'! It is interesting to note that "unaffiliated generics" appew to behave in this respect as the names of other generics, not like the names of life fern. Far exmpPe, if we are told that a creature which looks like a bat (or a mail) in fact is not a hat (or a snaio, we might well accept this claim--certainly more readily than the claim that what looks to us like a bird, or a tree, is in fact not a bird, or a tree. In talking about a possiblc "lllddcn nanurc" of living kinds, 1 do nor[ lapcsvn no abandon EIIY lclarlier claim (Wierzbicke 1985) thal folk genera are also definable in terms of specific features, such as stripes, claws, and teeth in the case of a tiger. On the contrary, 1 hope to have shown that folk genera can be fully defined, and that in this respect they do not differ fmm the names of cdtwral kinds (or any other words). The definitions which I have provided list the newssarry and sufficient campnents of a concept such as "gerkr 'squitreB" not necessary and sufficient features of all the referents d the words in question. I cllaimd in earlier work (1972) that biological kind terms, unlike the namnes of arlefacis, cannot be defined and are a kind of generic proper name. In subsequent work (19&5)1, on the basis of trial and error, I reached the conclusion th