This book offers introductory entries on eighty ideas that have shaped the study of language up to the present day. Entries are written by experts in the fields of linguistics and the philosophy of language to reflect the full range of approaches and modes of thought. Each entry includes a brief description of the idea, an account of its development, and its impact on the field of language study. The book is written in an accessible style with clear descriptions of technical terms, guides to further reading, and extensive cross-referencing between entries. A useful additional feature of this book is that it is cross-referenced throughout with Key Thinkers in Linguistics and the Philosophy of Language (Edinburgh, 2005), revealing significant connections and continuities in the two related disciplines. Ideas covered range from Sense Data, Artificial Intelligence, and Logic, through Generative Semantics, Cognitivism, and Conversation Analysis, to Political Correctness, Deconstruction, and Corpora.

Features

• The only single-volume reference book to focus specifically on ideas from both linguistics and the philosophy of language

• Accessibly written for use at all levels, including undergraduate, postgraduate, academic, and other general readers in the fields of linguistics and the philosophy of language

• Extensively cross-referenced both within itself and with Key Thinkers in Linguistics and the Philosophy of Language to provide a unique reference resource

Siobhan Chapman is Senior Lecturer in English Language at the University of Liverpool. http://siobhanchapman.co.ukChristopher Routledge is a freelance writer and editor. http://chrisroutledge.co.uk

Edited by Siobhan Chapman

and Christopher RoutledgeEdinburgh

KEY IDEAS IN LINGUISTICS AND THE PHILOSOPHY OF LANGUAGE

Edited by Siobhan Chapman and Christopher Routledge

KEY IDEAS IN LINGUISTICS AND THE

PHILOSOPHY OF LANGUAGE

KEY IDEAS IN LINGUISTICSAND THEPHILOSOPHY OF LANGUAGEEdited by Siobhan Chapman and Christopher Routledge

ISBN 978 0 7486 2619 9

Edinburgh University Press22 George SquareEdinburgh EH8 9LFwww.eup.ed.ac.uk

Cover illustration: Image © Corbis 2007

Cover design: Michael Chatfield

Pantone 330

Key Ideas in Linguistics and the Philosophy of Language

Key Ideas in Linguistics and thePhilosophy of Language

Edited by

Siobhan Chapman and Christopher Routledge

Edinburgh University Press

© in this edition Edinburgh University Press, 2009© in the individual contributions is retained by the authors

Edinburgh University Press Ltd 22 George Square, Edinburgh

Typeset in 10/12pt Sabonby Servis Filmsetting Ltd, Stockport, Cheshire, andprinted and bound in Great Britain by MPG Biddles Ltd, King’s Lynn, Norfolk

A CIP record for this book is available from the British Library

ISBN 978 0 7486 2618 2 (hardback)ISBN 978 0 7486 2619 9 (paperback)

The right of the contributorsto be identified as authors of this work has been asserted in accordance withthe Copyright, Designs and Patents Act 1988.

Preface viiAcknowledgements ixNotes on Contributors x

Acceptability/Grammaticality 1Adequacy 2Ambiguity/Vagueness 3Analytic Philosophy 4Analytic/Synthetic 11Artificial Intelligence 14Behaviourism 17Cognitivism 21Compositionality 24Connotation/Denotation 25Continuity 26Conventional Meaning 28Conversation Analysis 29Corpora 36Correspondence Theory 42Creativity 43Deconstruction 44Deduction/Induction 47Definite Descriptions 54Descriptivism 57(Critical) Discourse Analysis 58Distinctive Features 65Emic/Etic 66Empiricism/Rationalism 67Feminism 74Generative Phonology 77Generative Semantics 78Glossematics 79Holism 81Ideational Theories 84Implicature 86Indeterminacy 92Innateness 96Integrationism 97Intentionality 100

Intuition 103Language Games 106Language of Thought 110Langue/Parole 113Linguistic Relativity 116Linguistic Variable 119Logic 120Logical Form 127Logical Positivism 128Mentalism 130Metaphor 131Minimalism 134Modality 138Model-theoretic Semantics 139Names 142Nonnatural Meaning 144Optimality Theory 145Ordinary Language Philosophy 149Performative 155Phoneme 156Politeness 157Political Correctness 161Port-Royal Logic 163Possible World Semantics 164Poststructuralism 170Presupposition 173Private Language 179Propositional Attitudes 181Propositions 185Prototype 186Psychoanalysis 189Relevance Theory 193Sense Data 199Sense/Reference 200Signs and Semiotics 205Situational Semantics 209Speech Act Theory 212Structuralism 219Systemic-Functional Grammar 225

Contents

Transformational-GenerativeGrammar 232

Truth Theories 238Truth Value 244Type/Token 246

Universal Grammar 247Use/Mention 251

Index 253

CONTENTS

vi

Preface

The ideas described in this book have been developed in linguistics and the philosophy of language, as well as in some related disciplines such as mathe-matics, logic and psychology. They necessarily represent only a very small pro-portion of the long tradition of the serious study of language; we have chosenthem because of their impact on current work in linguistics and the philosophyof language. These two disciplines are subdivided into many different branches.Linguistics, for example, includes work undertaken in semantics, pragmatics,phonology, syntax, sociolinguistics and many other fields. In general we havenot treated these individual fields as key ideas in their own right. You will not,for instance, find an entry here on ‘Pragmatics’, but you will find topics fromthe field of pragmatics discussed under entries such as ‘Implicature’, ‘RelevanceTheory’ and ‘Speech Act Theory’. The names of different branches of linguis-tics and the philosophy of language do, however, appear in the index. Similarly,we have avoided allocating entries to descriptive categories such as ‘adjective’,‘phrase’ or ‘conjunction’.

The entries are arranged in simple alphabetical order, and aim to elucidateeach key idea, offering a succinct definition followed by a more discursiveaccount of the development of the idea and of its impact and current relevance.The book can therefore be used as a stand-alone reference work. However, it isalso designed to be used in conjunction with our Key Thinkers in Linguisticsand the Philosophy of Language (2005). The coverage of these two volumes issimilar: broadly, the study of language in the Western tradition from antiquityto the present-day, with an emphasis on work that has been influential on lin-guistics and the philosophy of language as they are practised in the early partof the twenty-first century. However, the two are complementary in that theyarrange and present the material in different ways. Key Thinkers considers dif-ferent ways of thinking about language in the context of the work of the particular figures with which they are most closely associated, drawing out con-tinuities and developments of thought in their particular historical and socialcontext. Key Ideas focuses on the development of specific ways of thinking,sometimes across many decades or centuries, considering the influences onthese ways of thinking, the relationships between them, and their overall significance.

Each entry is cross-referenced both internally within this book and externallyto Key Thinkers. The internal cross-references (‘See also’) draw attention toother key ideas that are relevant to the particular entry. These may be ideas thatare concerned with similar or related issues (such as the reference from ‘Deduc-tion/Induction’ to ‘Empiricism/Rationalism); ideas that form a specific topicwithin a more general approach (such as the reference from ‘Speech Act Theory’

vii

to ‘Performative’); or conversely ideas that provide a broader framework forthe discussion of the specific issue in question (such as the reference from (‘Ade-quacy’ to ‘Transformational-Generative Grammar’). Within each entry, anyterm that is itself a key idea with its own entry is marked with an asterisk onfirst use. The external cross-references (‘Key Thinkers’) point the readertowards relevant entries in Key Thinkers.

The entries are all concerned with encouraging the reader to find out more,and therefore with pointing outside these two volumes to other and moredetailed reading. Each entry concludes with a list of suggested readings, dividedinto ‘Primary sources’ and ‘Further reading’. Between them these sections givethe full references of any works mentioned or quoted in the entry, along withdetails of other particularly salient works. The section of ‘Further reading’ mayalso include suggestions of useful introductions to or overviews of the relevantkey idea.

PREFACE

viii

Acknowledgements

A book of this kind depends on the cooperation and goodwill of a large numberof people and it has been a privilege to work with all of them. We would liketo take this opportunity to thank all of our contributors to this book and itscompanion volume Key Thinkers in Linguistics and the Philosophy of Lan-guage (2005) for their diligence and professionalism. Two individuals deservespecial mention. We would like to thank Sarah Edwards and her team at Edin-burgh University Press for their enduring enthusiasm and support over thecourse of two lengthy book projects. Jeremy Kaye, whose father Alan S. Kayewas working on entries for this book when he died in May 2007, generouslygave his time searching for material that Alan was preparing for us and deliv-ered the one entry his father had managed to complete before becoming too illto continue. Alan was passionately committed to the study of language and avaluable contributor to these books. We dedicate this volume to his memory.

ix

Notes on Contributors

Varol Akman, Chair, Department of Philosophy Bilkent University, Turkey.

Salvatore Attardo, Professor, English Department, Youngstown StateUniversity, USA.

Annalisa Baicchi, Professor of English, Department of Theoretical andApplied Linguistics, University of Pavia, Italy.

Jennifer A. Baldwin, freelance translator (http://jenniferbaldwin.com).

Alex Barber, Department of Philosophy, The Open University, UK.

Philip Carr, Professor of Linguistics, Department of English, MontpellierUniversity, France.

Siobhan Chapman, Senior Lecturer in English Language, University ofLiverpool, UK.

Billy Clark, Lecturer in English Language, Middlesex University, UK.

John Collins, Department of Philosophy, University of East Anglia, UK.

Yousif Elhindi, Associate Professor, Department of English, East TennesseeState University, USA.

Iván García Álvarez, Lecturer in Linguistics, School of Languages, Universityof Salford, UK.

Agustinus Gianto, Professor of Semitic Philology and Linguistics, PontificalBiblical Institute, Rome, Italy.

Hans Götzsche, Associate Professor and Director, Centre for Linguistics,Aalborg University, Denmark.

Eva Herrmann-Kaliner, computational linguist, Germany.

Patrick Honeybone, Lecturer in Linguistics and English Language, Universityof Edinburgh, UK.

Asa Kasher, Laura Schwarz-Kipp Professor of Professional Ethics andPhilosophy of Practice, Tel Aviv University, Israel.

Alan S. Kaye (1944–2007), Professor of English, California State UniversityFullerton, USA.

Stavroula-Thaleia Kousta, Research Fellow, Department of Psychology,University College London, UK.

x

Guy Longworth, Assistant Professor, Department of Philosophy, University ofWarwick, UK.

Stephen McLeod, Department of Philosophy, University of Liverpool, UK.

Marie Nilsenová, Assistant Professor, Department of Communication andInformation Sciences, Tilburg University, Netherlands.

Ingrid Piller, Professor of Linguistics, Macquarie University, Australia.

Geoffrey Poole, Lecturer in Linguistics, School of English Literature,Language, and Linguistics, University of Newcastle upon Tyne, UK.

Kanavillil Rajagopalan, Professor of the Semantics and Pragmatics of NaturalLanguages, State University at Campinas (UNICAMP), Brazil.

Christopher Routledge, freelance writer and editor (http://chrisroutledge.co.uk).

Karl Simms, Senior Lecturer, School of English, University of Liverpool, UK.

Jürg Strässler, Universities of Berne and Zurich, Switzerland.

Marina Terkourafi, Department of Linguistics, University of Illinois atUrbana-Champaign.

Geoff Thompson, Senior Lecturer in Applied Linguistics, School of English atthe University of Liverpool, UK

Dominic Watt, Lecturer in Forensic Speech Science, Department of Languageand Linguistic Science, University of York, UK.

David Witkosky, Professor of German and International Studies, AuburnUniversity, Montgomery, USA.

NOTES ON CONTRIBUTORS

xi

1

by a native speaker. A grammaticalstring is not necessarily meaningful, asexemplified by Chomsky’s famous sen-tence ‘Colorless green ideas sleep furi-ously’. However, it is assumed thateven nonsensical grammatical sen-tences can easily be pronounced with anatural intonation and that speakersare able to recall them more easily thanungrammatical sentences. Speakersalso supposedly have intuitions aboutgrammaticality (or grammaticalness;these two terms appear to be inter-changeable), determined by their com-petence (in other words, knowledge ofa language).

Acceptability, on the other hand, isrelated to speaker’s performance,that is the actual use of her languagein concrete situations. As stressed byChomsky, acceptability should notbe confused with grammaticality:while an acceptable sentence mustbe grammatical, not just any gram-matical sentence is necessarily accept-able. For a sentence to be judgedacceptable, it must also appear natu-ral and appropriate in a given con-text, be easily understood and,possibly, be to a certain extent con-ventionalised.

Both grammaticality and accept-ability are considered to be gradientproperties by Chomsky and in the lin-guistic literature they are typicallyexpressed with a combination of ‘?’and ‘*’ for grammaticality – a sen-tence marked with ‘**’ being stronglyungrammatical, a sentence markedwith a ‘?*’ being questionable, and soon – and ‘#’ for acceptability.

It has been observed that althoughspeakers may have intuitions aboutgrammaticality, they cannot beexpected to translate their intuitions

ACCEPTABILITY/GRAMMATICALITY

Acceptability is the extent to which asentence allowed by the rules to begrammatical is considered permissibleby speakers and hearer; grammatical-ity is the extent to which a ‘string’ oflanguage conforms with a set of givenrules. It is assumed that a nativespeaker’s grammar generates gram-matical strings and that the speakerhas the ability to judge a certain stringto be either acceptable or not in herlanguage. In practice, the two notionsare frequently confounded and speak-ers are typically asked to give their‘grammaticality judgements’ insteadof ‘acceptability judgements’.

See also: Innateness; Mentalism;Transformational-GenerativeGrammarKey Thinkers: Chomsky, Noam

In prescriptive grammar booksemployed in language teaching, astring is considered to be grammaticalif it conforms with a set of prescribednorms. These norms are usually basedon conventionalised rules that form apart of a higher/literary register for agiven language. For some languages,these norms are defined and periodi-cally updated by an appointed body ofexperts whose rulings are occasionallyquestioned by the frustrated membersof the public.

In theoretical linguistics it hasbecome customary to utilise the term‘grammaticality’ in line with NoamChomsky’s generative approach. Ingenerative grammar, a string is gram-matical – or well formed – if it could begenerated by a grammar internalised

2

ADEQUACY

mathematics known as formal lan-guage theory. The idea was that onecould formulate a finite set of ruleswhich generated the set of sentenceswhich constituted a specific language,such as English. The set of sentencesconstituting a language was said tobe infinite. That infinitude is guaran-teed by the recursive nature of therules. Recursion is the phenomenonwhereby a given syntactic categorycan be embedded within a category ofthe same type. For instance, relativeclauses in English can be embeddedwithin relative clauses, as in the sen-tence ‘I know the woman who shotthe man who held up the bank’. Therelative clause ‘who shot the manwho held up the bank’ contains therelative clause ‘who held up thebank’. Recursion is widely believedto be a universal feature of humanlanguages.

One of the aims of early generativegrammars was to generate all andonly the sentences which constitute aspecific language. A grammar wassaid to be observationally adequate ifit did this. The aim was to constructgrammars which did not generatesequences which were ill-formed, suchas ‘*The house have might been built’,where the asterisk denotes ungram-maticality. Sequences are judged to begrammatical or ungrammatical on thebasis of the intuitive grammaticalityjudgements of native speakers of thelanguage. In making such judgements,native speakers are said to be access-ing a mind-internal grammar, which istaken to constitute largely uncon-scious linguistic knowledge, known ascompetence.

A given generative grammar willgenerate not only strings of words,

into judgements; hence, a categorisa-tion task can only rely on their abilityto judge the acceptability of a string(Schütze 1996). This assumption isoften violated in the process of datacollecting and speakers are typicallyencouraged to give grammaticalityjudgements of marginally soundingsentences independently of an actualdiscourse situation.

Primary sourcesChomsky, N. (1957). Syntactic Structures.

The Hague: Mouton.Chomsky, N. (1965). Aspects of the Theory

of Syntax. Cambridge, MA: MIT Press.

Further readingSchütze, C. (1996). The Empirical Base of

Linguistics. Chicago and London: TheUniversity of Chicago Press.

Marie Nilsenová

ADEQUACY

In the theory of transformational- generative grammar* generative gram -mars are said to be evaluated withrespect to three levels of adequacy:observational adequacy, descriptiveadequacy and explanatory adequacy.

See also: Acceptability/Grammaticality; Intuition;Transformational-GenerativeGrammar; Universal GrammarKey Thinkers: Chomsky, Noam

In Syntactic Structures (1957) NoamChomsky defined a language as a setof sentences. Generative grammarswere taken by Chomsky to be formalgrammars, based on a branch of

3

AMBIGUITY/VAGUENESS

Ambiguity and vagueness are relatedbut distinct concepts. Ambiguity, theproperty of having more than one lin-guistically encoded meaning, is animportant notion in several areas oflinguistics (including syntax, seman-tics and pragmatics) as well as in phi-losophy of language. Vagueness, theproperty of having no uniquely deter-minable value or interpretation, hasbeen discussed by philosophers sinceancient times and is still a focus ofattention for philosophers, semanti-cists and pragmatists.

See also: Connotation/Denotation;Conventional Meaning;Nonnatural Meaning; Implicature;Indeterminacy; Logic; LogicalForm; Sense/Reference; Signs andSemioticsKey Thinkers: Frege, Gottlob;Grice, H. P.; Peirce, C. S.; Quine,W. V. O.; Russell, Bertrand;Wittgenstein, Ludwig

Ambiguity is the property of havingmore than one linguistically encodedmeaning. Lexical ambiguity is a prop-erty of words (or lexemes) and syntactic ambiguity is a property ofphrases, clauses or sentences.

The English noun ‘seal’ is anambiguous lexical item referring,among other things, to a particularkind of sea creature or a device formaking sure an opening is fullyclosed. The sentence ‘I’ve brought theseal’ is ambiguous because of the pres-ence of this lexical item.

A syntactically (or ‘structurally’)ambiguous expression is one which

but also structural descriptions ofthose strings. For instance, one gram-mar might generate a structuraldescription of the phrase ‘The oldman’ in which there are only threeconstituents: ‘the’, ‘old’ and ‘man’, ina flat linear string. Another grammarcould generate a structural descrip-tion in which there are two interme-diate constituents: ‘the’ and ‘oldman’. The two grammars differ in thestructural descriptions they generate.In Aspects of the Theory of Syntax(1965), Chomsky argued that a gen-erative grammar is justified to theextent that the structural descriptionsit generates correspond to the intu-itive grammaticality judgements ofthe native speaker. He also sug-gested that linguists should seekexplanatory adequacy: they shouldgo further than descriptive adequacy,devising generative grammars whichwere embedded within a generaltheory of human language. Thetheory in question was Chomsky’suniversal grammar*.

Primary sourcesChomsky, N. (1957). Syntactic Structures.

The Hague: Mouton.Chomsky, N. (1965). Aspects of the

Theory of Syntax. Cambridge MA: MITPress.

Further readingSampson, G. (1979). ‘What was trans-

formational grammar?’ Lingua 48:355–78.

Smith, N. (2004). Chomsky: Ideas andIdeals. Cambridge: Cambridge Univer-sity Press.

Philip Carr

4

ANALYTIC PHILOSOPHY

‘heap’ or ‘painful’ are inherently vaguesince there is no way to say for certainhow much of a particular substancewe need to create a heap nor exactlyhow much discomfort is required foran experience to be painful.

Primary sourcesWilliamson, T. (1994). Vagueness.

London: Routledge.

Further readingAtlas, J. (1989). Philosophy Without Ambi-

guity. Oxford: Oxford University Press.Hurford, J. R., B. Heasley, and M. B.

Smith (2007). Semantics: A Course-book. Second edition. Cambridge:Cambridge University Press.

Saeed, J. (2003). Semantics. Second edi-tion. Oxford: Basil Blackwell.

Billy Clark

ANALYTICPHILOSOPHY

Analytic, or analytical, philosophycan be any of many things. The termis used to describe philosophy thatproceeds via analysis, broadly byseeking to understand the composi-tion of its subject matter (or conceptsof that subject matter) out of simple(or simpler) components. In a promi-nent but secondary sense, ‘analyticphilosophy’ applies to most philoso-phy carried out in the mainstream ofAnglo-American university philoso-phy departments together with philos-ophy that bears a suitable familyresemblance to it: work within theso-called analytic tradition.

can be associated with more than onesyntactic structure. For example, onereading of the expression ‘grey sealsand walruses’ refers to seals which aregrey and to walruses which are grey,while on another reading it refers toseals which are grey and to walrusesregardless of colour. Each of theseoptions is associated with a differentstructure which can be represented bybracketing:

(a) [grey] [seals and walruses](b) [grey seals] and [walruses]

A full understanding of human lan-guage requires an account of knowl-edge about the ambiguity of linguisticexpressions, the processes by whichwe understand utterances containingthem and what happens when we failto disambiguate them.

Vagueness is the related but distinctproperty of failing to determine aunique value or interpretation. Vague-ness can be understood as a very gen-eral, and fairly informal, notion or asa more systematically understoodproperty of concepts, propositions orutterances. Informally, all linguisticexpressions are vague since they havemore than one possible interpretation.The expression ‘He has brought theseal’, for example, could have severalinterpretations depending on who isunderstood as the referent of ‘he’,which sense of ‘seal’ is intended andwhat is the referent of ‘the seal’. Inmost contexts, this kind of vaguenessis relatively easily resolved and aunique interpretation can be deter-mined. On the more systematic sense,vagueness is understood as the prop-erty of not having a uniquely deter-minable meaning. Concepts such as

5

ANALYTIC PHILOSOPHY

(knowledge that X is the case) byviewing it as composed of other con-cepts: belief, truth, and justification.Such analysis attempted to explain theapparent inferential behaviour of theconcept, for example the apparentfact that if someone knows that thesun is shining then it may be inferredthat they believe that the sun is shin-ing and that their belief is both trueand justified. The success of themethod is dependent both upon thecorrectness of the analysis and uponits capacity to deepen understandingby explaining the less well under-stood – the whole – on the basis ofwhat is better understood, the simplercomponents and the effects of theirconfiguration in the whole. But theanalysis of the concept of knowledgearguably fails on both counts. It hasbeen argued that the concept ofknowledge is a simple concept and sonot composed of other concepts; andthat the concept of justification (andperhaps also the concept of belief) isno simpler, or easier to understand,than is the concept of knowledge.

The method of analysis played arole in early Greek philosophy (see, forexample, Plato’s Theaetetus) and reap-peared to take a major part in the earlymodern period, in the work of RenéDescartes and his followers. However,the method began gradually to lose itscentrality during the rise of Germanidealism, as philosophers responded tothe work of Immanuel Kant by seekingto develop grand systematic theories.Although it was a central tool of someother important thinkers in the nine-teenth century, most notably FranzBrentano and C. S. Peirce, its rise toprominence as the central methodin twentieth-century Anglo-American

See also: Analytic/Synthetic;Compositionality; DefiniteDescriptions; Holism;Indeterminacy; Logical Positivism;Ordinary Language Philosophy;Sense/Reference; Truth Theories;Truth ValuesKey Thinkers: Arnauld, Antoine;Austin, J. L.; Carnap, Rudolf;Frege, Gottlob; Moore, G. E.;Pierce, C. S.; Quine, W. V. O.;Russell, Bertrand; Ryle, Gilbert;Strawson, P. F.; Wittgenstein,Ludwig

Although the two senses of ‘analyticphilosophy’ described above are quitedifferent, it is not an accident that thesame expression is used for both. Thetype of philosophy that now domi-nates the Anglo-American main-stream began its rise to prominence inthe very late nineteenth century at thesame time as the emergence of analy-sis as the central method of that typeof philosophy. Analysis, broadly con-strued, has continued to play a largerole in much work in that tradition.

The method of analysis is themethod of seeking to understand asubject matter by coming to under-stand its composition. The aim is tounderstand the behaviour of a wholeby tracing its behaviour to the influ-ences of its parts and their organisa-tion. In philosophy, the aim has oftenbeen to understand the inferentialbehaviour of a concept, to understandwhat may be inferred from a judge-ment involving the concept and whata judgement involving the conceptmay be inferred from. For example,analytic philosophy has attemptedto understand the behaviour of theconcept of propositional knowledge

6

ANALYTIC PHILOSOPHY

1905). A very similar logic was devel-oped in the medieval period, but thediscovery of quantificational logic inits modern form is usually attributedto Gottlob Frege (1879) and (inde-pendently) Peirce (1885). This logicenables the systematic treatment ofthe inferential behaviour of a verylarge range of the statements that canbe made in natural language (and sothe thoughts expressible by the use ofthose statements) as well as the sharpstatement of complex positions andarguments. Of special importancewas Russell’s treatment of definitedescriptions* in sentences of the form‘The F is G’. Using the example ‘Thepresent King of France is bald’, Rus-sell represented sentences of that formas conjoining three claims:

(1) There is at least one F (there is atleast one present King of France).

(2) There is at most one F (there is atmost one present King of France).

(3) Whatever is F is G (what/who is thepresent King of France is bald).

In modern logical notation, the analy-sis becomes:

(4) (�x) [Fx & (�y) (Fy � x = y) & Gy].

By systematising a statement’s inferen-tial (or, more broadly, logical) behav-iour, the representation of thestatement in a favoured logical systemshows the (or, perhaps, a) logicalform* of the statement. Russell’streatment provided a model on whicha definite description that fails toapply to exactly one individual maybe meaningful, and so provided apotential solution to old problemsabout the functioning of talk that

philosophy was due mainly to thework of G. E. Moore and BertrandRussell.

In the very late nineteenth century,Moore began a revolt against Germanidealism. There were four main areasof dissent. First, Moore felt that Ideal-ism – according to which mind andworld are interdependent – was anerroneous metaphysical view andthat, where possible, there should be areturn to realism, according to whichthere is an external world that is con-stituted independently of operationsof mind (except, of course, where theexternal world contains individualminds). Second, Moore felt that grandsystem building should be suppressedin favour of careful attention to detailand rigorous argumentation. Third,he objected to what he saw as unnec-essary obfuscation in the writings ofGerman idealists. He felt that philo-sophical theories and arguments forthose theories should be open toobjective assessment and so should bestated as clearly and sharply as possi-ble. In Moore’s view, work in the sci-ences at the time offered better modelsof philosophical methodology andthis belief also motivated the fourthsource of his dissent, his negative reac-tion to the Idealist suppression of themethod of analysis. Calling for areturn to the method, Moore wrotethat ‘a thing becomes intelligible firstwhen it is analysed into its constituentconcepts’ (Moore 1899: 182).

Inspired by Moore and enamouredin particular with the science of math-ematics, Russell began systematicallyto develop the programme of analysis.He was helped by his discovery (orrediscovery) of modern quantifica-tional logic (see, in particular, Russell

7

ANALYTIC PHILOSOPHY

Ludwig Wittgenstein was the mostfamous pupil of Russell and Moore.Wittgenstein’s Tractatus Logico-Philosophicus (1922) further devel-oped the analysis of statements, andof representation more generally, inthe direction of logical atomism.According to logical atomism, themost fundamental level of representa-tion involves point-by-point connec-tions between simple representationalelements – representational atoms –and simple represented elements –represented atoms. This paradigm ofanalysis was driven by the view that‘A proposition [i.e. the content of astatement] has one and only one com-plete analysis’ (1922: 3.25). Thework’s major conclusion was thatthere are two types of statement:those that represent the world, and socan be either true or false dependingupon how the world in fact is – thesynthetic truths and falsehoods – andthose that either cannot be true orcannot be false – the logical or ana-lytic truths and falsehoods – that failto represent the world as being oneway or another. Since genuinelyphilosophical statements, derived byanalysis, were taken to fall on thenon-representational (analytic) sideof this divide, they were taken to bedevoid of real content and to have arole other than that of conveyinginformation. And since a core sense ofmeaningfulness was identified withrepresentational significance, suchstatements were taken to be – in thatcore sense – meaningless. (For anintroduction to Wittgenstein’s earlywork, see Mounce 1989. For a dis-cussion that relates logical atomismwith Plato’s Theaetetus, see Burnyeat1990.)

purports to make reference to partic-ular non-existents. On Russell’saccount, the fact that there is no pres-ent King of France makes the sentence‘The King of France is bald’ false,rather than meaningless, because itmakes false the first clause in hisanalysis. His treatment also madeespecially evident that the logical formof a statement might not be obviousfrom its superficial form. However,the space between logical and superfi-cial form involved in Russell’s treat-ment of ‘The F is G’ via (4) is anartefact of Russell’s favoured logic.An alternative, though slightly lessperspicuous, treatment is given in (5):

(5) ([The x: Fx] (Gx).)

Russell’s treatment of definitedescriptions showed that philosophi-cal progress could be made by dis-cerning the (or a) logical form of aphilosophically problematic range ofstatements and that some philosophi-cal disputes are usefully viewed as atleast in part concerning how best torepresent the logical forms of state-ments involved in those disputes.Together with the new treatment ofquantification more generally, thisbecame a model for a variety ofapproaches to philosophical prob-lems that involved attention to theforms of language used in the state-ment of those problems. It supportedthe view that philosophical problemscan arise due to the misleading super-ficial forms of the language we use,and it provided a model for howproblems that arise in that waymight be solved through uncoveringthe true logical forms of the state-ments involved.

8

ANALYTIC PHILOSOPHY

empirical content with a more expan-sive view of philosophical analysis.According to the more expansiveview, the analysis of statements was toinclude the tracing of their roleswithin larger systems of languagedriven by careful attention to the waythose statements are actually used inordinary contexts. The more expan-sive view therefore involved a partialreturn to the sort of holistic app -roach involved in German idealism.Wittgenstein’s work within the moreexpansive paradigm developed hisearlier view that the role of philo-sophical analysis should be largelytherapeutic: that it should serve theremoval of philosophical perplexityby uncovering and excising thesources of confusion in the misleadingsuperficial forms of language. Butmany other philosophers workingwithin the new paradigm of analysis,including J. L. Austin, Gilbert Ryle,and P. F. Strawson – the so-called‘ordinary language’ philosophers –returned to the view that some philo-sophical questions might be genuineand hoped that the analysis of lan-guage would deliver answers to thosequestions.

W. V. O. Quine took a differentpath away from the Vienna Circle.Quine was strongly influenced by thework of Carnap and retained theCircle’s view that scientific observa-tion is the only source of cognitive sig-nificance. However, he thought thatCarnap had failed fully to draw outthe consequences of that view. Quine(1953, for example) argued that itserved to undermine the distinctionbetween statements that can be veri-fied or falsified on the basis of experi-ence and those that cannot be so

Members of the Vienna Circle –including especially Rudolf Carnap –were inspired by Wittgenstein’s workand sought to embed its centralthemes in an approach to philoso-phy, known as logical positivism*,shaped by epistemological concerns.They replaced Wittgenstein’s distinc-tion between statements whosetruth value* depends upon worldlycontingency and statements whosetruth value is fixed independently ofsuch contingency with a distinctionbetween statements that admit of ver-ification or falsification on the basis ofexperience and statements that cannotbe so verified or falsified. The task ofphilosophy was taken to be the analy-sis of statements into experientiallysignificant components, an analysisthat would either indicate preciselythe course of experience that wouldverify or falsify the statement or showit to be beyond verification or falsifi-cation. In that way, philosophy wouldeither show how a statement can beassessed on the basis of scientificobservation, or show the statement tobe (in the Circle’s proprietary sense)meaningless. Since the only properlycognitive activity was taken to be thecollection of observations, the pro-gramme of the Vienna Circle wasshaped by the view that ‘what is leftover for philosophy . . . is only amethod: the method of logical analy-sis’ (Carnap 1932: 77).

In the early post-1945 period, manyphilosophers retained the ViennaCircle’s animus towards traditionalmetaphysics but viewed its approachto analysis as overly restrictive. Theysought to replace what they saw asan empirically unmotivated fixationupon a very narrow conception of

9

ANALYTIC PHILOSOPHY

under mine a distinction between theanalysis of statements and the empiri-cal assessment of those statements.And the rejection of determinatetranslation appears to undermine theassumption that there is such a thingas the analysis of a statement.

The main effects of Quine’s workhave involved the further broadeningof the analytic horizon. Many con-temporary analytic philosophers whoare influenced by Quine’s rejection ofan analytic-synthetic distinction allowtheir work to be shaped by the find-ings of empirical science. Similarly,many contemporary analytic philoso-phers allow that philosophy – evenwhen approached from the armchair –can be a source of discovery about theworld. Thus the culmination, throughQuine, of the Vienna Circle’s anti-metaphysical empiricism has led tothe reinstatement of metaphysics as alegitimate area of cognitive inquiry.Finally, many contemporary philoso-phers have followed Quine in relegat-ing the study of meaning from itscentral place in the analytic philoso-phers’ armoury, and have sought amore direct approach to answeringphilosophical questions.

The postwar period of intense focusupon language and linguistic meaningcoincided with Frege’s work becomingwidely available in translation and ledto a re-evaluation of his place in thedevelopment of analytic philosophy.In particular, Michael Dummett(1993) has claimed that it is criterialof modern analytic philosophy that itapproaches the study of thoughtthrough a study of the way language isused to express thought and that,from that perspective, Frege should beassigned priority over Russell as its

verified or falsified: that is, the Circle’sversion of the distinction betweenstatements that are analytic and thosethat are synthetic. Quine thereforetook the range of statements that areup for scientific assessment to includestatements in logic and mathematics.He took our total theory of the worldto form an interconnected web ofstatements that can only be assessedas a whole on the basis of the range ofpredictions it makes about the courseof experience. Quine’s work involvedin particular a rejection of the goal ofatomist analysis that formed the impe-tus for the earliest work in modernanalytic philosophy.

Quine’s rejection of the existence ofa category of analytic truths wenthand in hand with a general rejectionof the philosophical utility of appealsto unreconstructed notions of linguis-tic meaning. In particular, Quineargued for the indeterminacy* oftranslation, the claim that for anytranslation from one language intoanother (including translation from alanguage into itself), there will beother translations that have equallygood empirical credentials. Famouslyhe argued that evidence that anexpression is to be translated by theEnglish expression ‘rabbit’ can beequally good evidence that the expres-sion is to be translated by the Englishexpression ‘un-detached part of arabbit’ (Quine 1960: 1–79). Whenconjoined with Quine’s rejection of aprincipled distinction between ana-lytic and synthetic statements, accept-ance of Quine’s views has seemed tomany to undermine the possibility ofa philosophy based upon methods ofanalysis. The rejection of an analytic-synthetic* distinction appears to

10

ANALYTIC PHILOSOPHY

(1959), Logical Positivism. Glencoe, IL:The Free Press.

Dummett, M. A. E. (1993). Origins ofAnalytical Philosophy. Cambridge MA:Harvard University Press.

Frege, G. (1879). Begriffsschrift: eine derarithmetischen nachgebildete Formel-sprache des reinen Denkens. Halle.Reprinted in Frege (1972), ConceptualNotation and Related Articles. Trans.and ed. T. W. Bynum. Oxford: OxfordUniversity Press.

Frege, G. (1884). Die Grundlagen derArithmetik. Eine logisch mathematischeUntersuchung über den Begriff derZahl. Frege (1950) The Foundation ofArithmetic. A Logico-mathematicalEnquiry into the Concept of Number.J. L. Austin. Oxford: Basil Blackwell.

Moore, G. E. (1899). ‘The nature of judge-ment’. Mind 8: 176–93.

Mounce, H. O. (1989). Wittgenstein’sTractatus: An Introduction. Chicago:University of Chicago Press.

Peirce, C. S. (1885). ‘On the algebra oflogic; a contribution to the philosophyof notation’. American Journal ofMathematics 7: 180–202. Reprinted inPeirce (1933), Collected Papers, ed.C. Hartshorne and P. Weiss. CambridgeMA: Harvard University Press.

Quine, W. V. (1953). ‘Two dogmas ofempiricism’. In From a Logical Point ofView. Cambridge, MA: Harvard Uni-versity Press.

Quine, W. V. (1960). Word and Object.Cambridge, MA: MIT Press.

Russell, B. (1903). The Principles of Math-ematics. London: Allen & Unwin.

Russell, B. (1905). ‘On denoting’. Mind14: 479–93. Reprinted in Russell(1956), Logic and Knowledge, ed.R. Marsh. London: Allen & Unwin.

Wittgenstein, L. (1922). Tractatus Logico-Philosophicus. Trans. C. K. Ogden and

progenitor. Dummett’s claim is con-troversial for several reasons, not leastthat it appears to place many contem-porary philosophers who work withinthe analytic tradition outside thebounds of analytic philosophy proper,and it also seems to exclude the earlywork of Russell and Moore. Second, itis controversial that Frege would him-self count as an analytic philosopherby Dummett’s own standard, sincehe distrusted natural language andsought to construct artificial systemsbetter able to capture the nature ofthought. Third, although Frege’s worknow occupies a central place in thecurriculum of analytic philosophy,there is little consensus concerning itsprecise role in shaping the initialdevelopment of the modern analytictradition. Other philosophers, forexample, also played important rolesin the development of Russell,Wittgenstein and Carnap. What is un -controversial, however, is that Frege’swork possesses many of the qualitiesassociated with analytic philosophy,including narrow focus, clarity, rigourand depth, and exhibits all those qual-ities to a very high degree. To thatextent, no education in analytic phi-losophy would be complete that didnot involve a careful study of hiswork.

Primary sourcesBurnyeat, M. F. (1990). The Theaetetus of

Plato. Trans. M. J. Levett. Indianapolis:Hackett.

Carnap, R. (1932). ‘Überwindung derMataphysik durch logische Analyseder Sprache’. Erkenntnis 2: 219–41.Reprinted as ‘The elimination of meta-physics through logical analysis of lan-guage’ trans. A. Pap, in A. J. Ayer (ed.)

11

ANALYTIC/SYNTHETIC

its use by empiricists as a ‘dogma’. Thereverberations of Quine’s criticism arestill felt today.

See also: Generative Semantics;Holism; Logical Positivism; SenseData; Truth ValueKey Thinkers: Carnap, Rudolf;Kant, Immanuel; Quine, W. V. O.

The contrast between analytically trueand synthetically true statements is,on the face of it, an intuitive one,easily conveyed by either definition (asabove) or example:

(1) ‘All bachelors are unmarried’ (ana-lytically true).

(2) ‘Bachelors have a higher averagedisposable income than spinsters’(synthetically true).

The terms ‘analytic’ and ‘synthetic’ asthey apply to true judgements werecoined by Immanuel Kant (1781). Buthis actual definition, that a true judge-ment is analytic if its predicate con-cept is contained within its subjectconcept and is synthetic otherwise, isnow usually treated as having at mosthistorical interest. Some philosophersargue that statements are analyticallytrue if understanding them requiresacknowledgement of their truth.Others define the distinction in twostages: first, as it applies to either abelief, proposition, statement, or sen-tence, and second, as it applies, deriv-ing from this, to some or all othermembers of this group. For example,if an analytically true proposition is aproposition true solely by virtue of thecontent of its component concepts,then an analytically true statementwould be a statement that expressesan analytically true proposition.

F. P. Ramsey. London: Routledge andKegan Paul.

Further readingArnauld, A. and P. Nicole (1683). La

logique ou L’art de penser: contenantoutre les regles communes, plusieursobservations nouvelles, propres àformer le jugement. Second edition.Paris: G. Desprez. Reprinted asArnauld, A. and P. Nicole (1996). Logicor The Art of Thinking. Trans. and ed.J. V. Buroker, K. Ameriks, and D. M.Clarke. Cambridge: Cambridge Univer-sity Press.

Descartes, R. (1701). Rules for the Direc-tion of the Mind. Amsterdam: P & J.Blaeu. Reprinted in Descartes (1985),The Philosophical Writings of Descartes,vol. 1. Trans. and ed. J. Cottingham,R. Stoothof and D. Murdoch. Cam-bridge: Cambridge University Press.

Hylton, P. (2007). Quine. London: Rout-ledge.

Hylton, P. (1993). Russell, Idealism andthe Emergence of Analytic Philosophy.Oxford: Clarendon Press.

Guy Longworth

ANALYTIC/SYNTHETIC

An analytic truth is a statement (orproposition) that is true solely byvirtue of the meaning of its componentwords (or concepts). A synthetic truthon the other hand is a truth deter-mined at least in part by features of theworld outside of the language system.Almost as well known as the distinc-tion is W. V. O. Quine’s description of

12

ANALYTIC/SYNTHETIC

such as ‘photon’. Quine thought thatthe analytic definitions needed tobridge the gap between the more theo-retical, higher-level synthetic claimsand the basic-level observation sen-tences were simply not available.

The problem, Quine said, was alack of adequate guidance on how todistinguish analytic from syntheticclaims in practice. Kant’s account ofthe distinction, with its talk of con-cepts being ‘contained in’ other con-cepts, is too metaphorical to behelpful here. Similarly, to say that ana-lytic truths are truths of meaning, ortruths by definition, leaves us needinga way of distinguishing truths ofmeaning from deeply held convic-tions. Deferring to a dictionary merelypasses the buck up to dictionary writ-ers. To fall back on the thought thattruths are analytic just in case they arelogical truths, as Carnap himself did,is to forget that many supposedly ana-lytic sentences, including (1) above,are non-logically true. Perhaps, Quineconjectures, analytic truths are sup-posed to be stipulations, true by lin-guistic convention. Against this heclaims that we lack criteria for decid-ing whether a particular assertionshould be classified as true by conven-tion rather than as an empirical con-jecture. Is the assertion that photonshave wave-particle duality a stipula-tive truth about the word ‘photon’ ora substantial empirical assertion? Nopre-established glossary for thevocabulary of physics exists that isuncommitted to any particular empir-ical outlook.

For all that has been said so far, log-ical positivists could cling to thethought that the analytic sentences aresimply those that link theoretical

In the twentieth century the analytic/synthetic distinction has beenboth important and controversial,particularly because of the weightplaced on it by logical positivists, andQuine’s insistence that it could notbear that weight. Whether or notQuine is right, there may be uses forthe distinction in linguistics that arealso compatible with his position.

Logical positivists such as RudolfCarnap had an ambitious goal: toconstruct a language suited to theneeds of science. Anyone with anunderstanding of this language wouldbe able to extract empirical conditionsfrom any sentence of it used to expressa claim. Any properly scientific claimcould thus be made to wear its confir-mation conditions on its sleeve, so tospeak, making it straightforwardlysusceptible to empirical confirmationor refutation.

Quine argued that the reason logicalpositivism* struggled to achieve thisgoal of a scientific language was itsreliance on an unworkable distinctionbetween analytical and synthetic sentences. Synthetic sentences weresupposed to include basic-level obser-vation sentences such as ‘There is red-ness to the left of my visual field’ aswell as higher-level theoretical sen-tences such as ‘Photons have wave-particle duality’, with many levels inbetween. Higher-level synthetic sen-tences had to mean the same as (thatis, be analytically equivalent to) com-plex basic-level observation sentences,since the latter expressed the former’sempirical commitments. Logical posi-tivists needed all synthetic sentences toreduce to complex basic-level observa-tion sentences by the repeated applica-tion of definitions of theoretical terms

13

ANALYTIC/SYNTHETIC

‘know’. One motive for this concep-tion of their role is that philosophersrealise they are not scientists. Whatthey provide must therefore comple-ment rather than compete with scien-tific knowledge. Another is immodest:they aim to understand the world at ana priori level, that is, at a level imper-vious to the contingency of experience.

If Quine is right, this model of phi-losophy may be in trouble. Accordingto him, there are no analytic truths inany interesting sense. Some support-ers of the model insist they are notinterested in the analysis of scientificconcepts. But they need to defend aninterest in the analyses they offer.After all, ‘folk’ ways of thinking areoften inferior to scientific ones. Goodanalyses of concepts such as ‘causa-tion’ or ‘knowledge’, if Quine is right,may be susceptible to improvement asscience develops. Advances in the cog -nitive sciences, for example, haveshaped our understanding of whatknowledge is. Whereas ‘Knowledgemust be conscious to the knower’would once have been regarded asanalytic; it is now commonly regardedas not even true, let alone analyticallytrue. Disputes over the status of ana-lytic truths have, then, broadened outinto a split between those who see thephilosophy of x and the science of x asdistinct in both subject matter andmethodology, and those who see onlycontinuity.

A very different approach to ana-lytic truth emerges from within lin-guistics itself. It may be that to call asentence analytically true is merely togive voice to an accidental feature ofour language faculty. Certain appar-ently analytic entailments, such asthat killing requires causing to die,

sentences to their confirmation condi-tions, plus those that enable this linkto be derived. It is up to us to identifythose links. No one ever claimed thiswould be easy. Quine’s final objectionto this faith has, rightly or wrongly,probably been the most influential. Heobserves that evidence for or against aparticular clause in some scientifictheory can, in principle, come fromanywhere. Evidence for a particularclaim in syntax, for example, may pre-suppose a particular morphology; evi-dence for this morphology maydepend in turn on an analogy with atheory of colour vision tied to specificassumptions about photons; and soon. Given this evidential holism, stat-ing the confirmation conditions forhigher-level theoretical claims wouldrequire making impossibly detailedand disjunctive reference to all of sci-ence. Developing and applying a language that would allow these state-ments to be formulated, if possible atall, would be no easier than doing sci-ence as it is done already.

Interest in and scepticism towardsthe analytic/synthetic distinction thusgrew out of a concern for how scien-tific methodology might be optimised.But the distinction is also central to apopular conception of what sets phi-losophy apart from other disciplinesand also to an approach to semanticsconceived as a part of psychology.

Many philosophers follow Socratesin conceiving of their role as being toanalyse concepts. There are more orless sophisticated implementations ofthis thought, but a typical illustrationis the giving of necessary and sufficientconditions that approximately match,or perhaps improve on, those forthe proper application of the word

14

ARTIFICIAL INTELLIGENCE

Further readingBoghossian, P. (1996). ‘Analyticity recon-

sidered’. Noûs 30(3): 360–91.Jackendoff, R. (1990). Semantic Struc-

tures. Cambridge MA: MIT Press.Jackson, F. (2000). From Metaphysics to

Ethics: a Defence of Conceptual Analy-sis. Oxford: Oxford University Press.

Alex Barber

ARTIFICIALINTELLIGENCE

The ability of machines to think forthemselves. More generally the sci-ence dedicated to exploring and devel-oping the ability of technology – mostcommonly in the form of a softwareprogram – to think and/or act ration-ally or similarly to human capability.As it relates to linguistics, the flourish-ing field of artificial intelligence aimsto develop systems that can under-stand, process, and even generate lan-guage as well as a human can.

See also: Generative Semantics;Intentionality; Logic; Speech ActTheoryKey Thinkers: Chomsky, Noam;Montague, Richard; Russell,Bertrand; Searle, John

The study of language often leads toquestions over what defines intelli-gence, as exhibited via language com-petence and performance. Artificialintelligence (AI) provides us with pre-cious opportunities to apply what weknow about language toward usefuland often valuable purposes, whetherto serve as a proof of concept, to

may be grounded in the semantic fea-tures of our concepts and/or lexicalitems (see generative semantics*).This approach to semantics has a che-quered history, in part because of thedifficulty of distinguishing deeply-held beliefs about something fromsemantic features – the same problemQuine drew attention to in a differentguise. That said, appealing to thenotion of analyticity in this contextwould not commit one to logical pos-itivism*. Discovering semantic fea-tures could be interesting for what ittells us about the structure and devel-opment of the human mind, notbecause of any help it offers in inves-tigating the world outside the mind.

The intuitive distinction between theanalytic and the synthetic has beendrawn on by philosophers of science inan attempt to understand how sciencemight be improved, and by otherphilosophers to explain how what theyare doing complements science. Inter-est in analyticity has also emerged outof the mentalist approach favouredby some semanticists. The challengefacing all who invoke the notion ofanalyticity is to ascertain how to dis-tinguish in practice between some-thing’s being analytically true and itsbeing profoundly but syntheticallytrue.

Primary sourcesCarnap, R. (1928). Der logische Aufbau der

Welt. Berlin-Schlachtensee: Weltkreis-Verlag.

Kant, I. (1781/1933). A Critique of PureReason, trans. N. Kemp Smith. Secondedition. London: Macmillan.

Quine, W. V. O. (1951). ‘Two dogmas ofempiricism’. Philosophical Review 60:20–43.

15

ARTIFICIAL INTELLIGENCE

who is interacting with the computer, itwould seem that the computer under-stands Chinese. However, if we replacethe computer with a human who doesnot know Chinese but has been trainedto apply rules from a book against Chinese language input to arrive atChinese language output, we wouldnever say that the human knows Chinese. That is, the simple act ofapplying predetermined rules (as acomputer does during a Turing test)does not constitute thinking, nor doesit represent consciousness. Neverthe-less, variations of this approach arecommonly used to evaluate AI systemsagainst a standard of human ability.

In order to process language as wellas a human, an AI system must havethe ability both to understand and togenerate language. Its means ofaccomplishing this may differ fromhuman mental processes, but theeffect should be equivalent. At a highlevel, natural language understandingin an AI system entails accepting language input (whether throughspeech recognition or optical charac-ter recognition), analysing it (throughparsing and semantic interpretation),and disambiguating the message (atthe lexical, syntactic, semantic andpragmatic levels, calling upon outsideknowledge of the world, as needed).Likewise, natural language generationrequires that the AI system determinea reasonable response, organise itsstructure, form comprehensible andnatural-sounding sentences, and out -put them as written text or synthe-sised speech. AI forms an extension oftraditional computer science, with linguistic AI being commonly cate-gorised as computational linguisticsor natural language processing. These

innovate, or to fulfil a specific need.The roots of AI extend to the originsof computers and even logic*. Fromits foundations in the works of RenéDescartes and Blaise Pascal, throughthe nineteenth-century innovations ofCharles Babbage and Ada Lovelace,AI flourished in the twentieth century,directly paralleling advancements inboth linguistics and computer science,through Bertrand Russell’s formallogic, the Turing test, and contribu-tions by such innovators as JosephWeizenbaum and Hans Kamp, toname a few. Not without its share ofcriticism, AI has resulted in manypractical applications for everydaylife, including speech recognition sys-tems, machine translation software,and information retrieval systems.

Humans are capable of many tasks,with language being among the mostcomplex. Language allows us to com-municate abstract thoughts through aninfinite combination of words andmeanings. With this in mind, it is notsurprising that Alan Turing designedhis now-famous Turing test aroundlanguage. The Turing test is a way tojudge the intelligence of an AI system.It involves a human subject, who car-ries on a conversation with two enti-ties, one being a human and the othera computer. If the subject cannot dis-tinguish between the human and thecomputer, the computer is said to havepassed the test. The best-known coun-terpoint to the Turing test is JohnSearle’s Chinese Room argument. Webegin this scenario with a computerthat has been programmed to return aparticular set of Chinese characters asa result of applying a set of rules toanother set of Chinese charactersreceived as input. To a Chinese speaker

16

ARTIFICIAL INTELLIGENCE

tion as economically and as efficientlyas possible. Statistical AI systems offerpromising alternatives within naturallanguage processing, incorporatingprobability theory, Bayesian networks,Markov models, and neural networks.The most effective examples of lan-guage AI, in fact, are hybrid models,combining linguistic knowledge withstatistical methodologies.

Just as AI draws upon developmentsin the field of linguistics, it presentsvaluable opportunities for linguists totest their theories in a computationalenvironment. An AI system ultimatelydemonstrates linguistic performancewith its software resources comprisingits linguistic competence. Language-based AI has many practical applica-tions in business and leisure. Weencounter AI, for example, in busi-ness, where it may be used to forecastmarket conditions, to analyse largeamounts of data, and to save time oth-erwise spent on tedious tasks. In ourhomes, we see AI in our appliances, inour automobiles, and on our personalcomputers, and we may even playchess and other complex games againstAI systems. Practical uses of AI are allaround us, originating in linguistic andAI circles and ultimately being distrib-uted around the world for an abun-dance of practical applications.

AI represents an application of lin-guistic knowledge, and it serves as adriving force to learn more, not onlyabout language, but also about humanknowledge and the nature of intelli-gence itself. It encourages us to revisitquestions over what defines intelli-gence, what sets apart human capabil-ity, and what we can programmeor train a computer to do. Fictional portrayals of AI systems have created

disciplines call upon generative gram-mar and linguistics in order to applywhat we know about language totechnology that may be capable ofnatural language understanding andgeneration. Linguists involved in AImaintain a keen interest in advance-ments within the general field of lin-guistics, as their goal is to equipcomputers with a comparable sense ofthe linguistic knowledge and abilitywe humans innately have, such asgrammaticality, phrase structure, gov-ernment and binding and so on.

Critics of AI argue that computersare incapable of achieving the level ofcognitive ability that humans exhibit.To this point, Searle made a distinctionbetween strong AI and weak AI. StrongAI describes a system that is able toreason and perform as well as – orbetter than – a human. A strong AIsystem is conscious of itself and oper-ates beyond a strict adherence to pre-programmed rules and algorithms.Most AI systems, however, are betterclassified as weak AI, capable of only alimited range of skills and lacking in theability to ‘think’ as a human does.Hubert Dreyfus, one of the most out-spoken critics of AI, states that com-puters simply cannot accomplish thecomplex tasks typically associated withstrong AI. For example, an AI systemcan never learn to love, no matter howit is programmed. System design israrely flawless and the mere prospect ofdeveloping a fully capable AI programis downright intractable. Difficultiesthat face AI systems encompass thevariation and creativity inherent in lan-guage, such as differences in grammat-icality across dialects and a heavy useof metaphor*, with the prevailing constraint that a system should func-

17

BEHAVIOURISM

Woods, William A. (1970). ‘Transitionnetwork grammars for natural languageanalysis’. Communications of the ACMVol. 13: 591–606.

Further readingCharniak, Eugene (1993). Statistical Lan-

guage Learning. Cambridge, MA: MITPress.

Hutchins, W. John and Harold L. Somers(1992). An Introduction to MachineTranslation. San Diego, CA: AcademicPress.

Jurafsky, Daniel and James H. Martin(2000). Speech and Language Process-ing. Upper Saddle River, NJ: PrenticeHall.

Rabiner, Lawrence R. and Biing-HwangJuang (1993). Fundamentals of SpeechRecognition. Upper Saddle River, NJ:Prentice Hall.

Russell, Stuart and Peter Norvig (1995).Artificial Intelligence: A ModernApproach. Upper Saddle River, NJ:Prentice Hall.

Jennifer A. Baldwin

BEHAVIOURISM

The study of behaviour withoutappeal to mental states. Behaviourismhad a profound effect on linguistics(especially American structuralist linguistics) and philosophy (logicalbehaviourism), but was effectively dis-placed by the cognitive science revolu-tion in the 1960s.

See also: Cognitivism;Empiricism/Rationalism;Innateness; Mentalism;Psychoanalysis; Structuralism;

rather humourous public misconcep-tions, from robot uprisings to worlddomination, but also some ideas forvaluable applications of AI, many ofwhich require linguistic AI in particu-lar. In some cases, such as searchengines and some speech recognitionsystems, language-related AI hasbecome a welcome part of daily life. Itbrings together many of the things weknow about language and tests thesetheories, sometimes even in the form ofannual competitions. Such challengeslead to innovative approaches andservices that would otherwise be unat-tainable through mere human effortand ability. There are endless possibili-ties for practical applications of AI,especially where there is consumerdemand, and it is an important area inwhich businesses, technological inno-vators, and philosophers of languagecan work together.

Primary sourcesDreyfus, H. (1972). What Computers

Can’t Do: The Limits of Artificial Intel-ligence. Cambridge, MA: MIT Press.

Kamp, H. (1981). ‘A theory of truth andsemantic representation’. In Groe-nendijk, J. A. G., T. M. V. Janssen and M.B. J. Stokhof (eds), Formal Methods inthe Study of Language. Vol. 1: 277–322.

Schank, Roger C. (1972). ‘Conceptualdependency: a theory of natural lan-guage understanding’. Cognitive Psy-chology Vol. 3: 532–631.

Turing, Alan (1950). ‘Computing machin-ery and intelligence’. Mind Vol. 59:433–60.

Wilks, Yorick (1973). ‘The StanfordMachine Translation Project’. InR. Rustin (ed.), Natural Language Processing. New York: AlgorithmicsPress. 243–90.

18

BEHAVIOURISM

involuntary behaviour by repeatedlypairing a neutral stimulus (for exam-ple, a bell sound that does not natu-rally elicit a given response) with anunconditioned stimulus (for example,food, which naturally produces agiven response such as salivation indogs). After repeated pairing, the neu-tral (now conditioned) stimulusbecomes associated with the (nowconditioned) response: a dog that hasundergone this type of training willsalivate when hearing a bell sound,even in the absence of food.

Classical behaviourism had a pro-found effect on American structuralistlinguistics, through the work of lin-guists such as Leonard Bloomfield,and his student Charles Hockett.Bloomfield espoused the behaviouristcredo that only behaviour can be theobject of scientific study. He advo-cated for the study of linguistics whatWatson had advocated for the studyof psychology: that it should beapproached as a natural science, witha firm emphasis on empirical data andrigorous description. In his enor-mously influential Language (1933),he provided an account of meaningwhich adapted the classical condition-ing schema to the interpretation ofacts of speech. Language provides asecondary stimulus-response mecha-nism, pairing a stimulus from thespeaker with a response from the lis-tener. He illustrated this in the nowfamous ‘Jack and Jill’ story, in whichJill experiences hunger (speaker stim-ulus) and asks Jack to fetch an applefor her (speaker response). Upon hear-ing the request (hearer stimulus), Jackfetches the apple (hearer response).Although Bloomfield himself admit-ted the difficulties of such an analysis

Transformational-GenerativeGrammarKey Thinkers: Bloomfield,Leonard; Chomsky, Noam;Hockett, Charles; Quine, W. V. O.;Ryle, Gilbert; Skinner, B. F.;Wittgenstein, Ludwig

Behaviourism was the dominant para-digm in psychology for much of thefirst half of the twentieth century anduntil the late 1950s, especially in theUnited States. Despite important dif-ferences between different schools ofbehaviourism, all behaviourists sharedthe conviction that behaviour was theonly legitimate object of investigationin psychology, and that behaviourcannot be explained by appealing tohypothetical internal states, such asconsciousness. As such, behaviourismfollows in the tradition of Britishempiricism*.

J. B. Watson is credited as thefounding father of behaviourism. In anarticle that has come to be known asthe ‘behaviourist manifesto’, Watson(1913) set out the main principles ofclassical behaviourism: psychology isa natural science; the only legitimateobject of psychological study is overtbehaviour; internal mental states andemotions are not behaviour andcannot be the causes of behaviour;behaviour is environmentally deter-mined; the same principles are appliedto the study of both animal and humanbehaviour, which do not differ quali-tatively. In later work, he acknowl-edged the importance of Ivan Pavlov’swork on what came to be known asclassical conditioning and incorpo-rated it into the behaviourist agenda.Classical conditioning is a form ofassociative learning that modifies

19

BEHAVIOURISM

complex behaviours, but languagewas not one of them. B. F. Skinnerundertook to fill this gap in VerbalBehavior (1957). In this work, whichtook more than twenty years to com-plete, Skinner asserted that verbalbehaviour is a complex habit arisingas a result of the interaction betweena speaker’s environment and his/herpast learning history. In his researchwith animals, Skinner had shown howvoluntary behaviour could be modi-fied through operant conditioning: afood-deprived (‘motivating opera-tion’) rat is placed in a cage and a lightis turned on (the ‘discriminative stim-ulus’) to signal that food can beobtained if the rat performs the appro-priate action. The rat may initiallyperform a number of actions, such asrunning around the cage, scratchingthe cage bars, or pressing a lever in thecage. If by performing one of theseactions – for instance, pressing thelever (‘response’) – the rat obtainsfood (‘reinforcer’), this particularbehaviour is reinforced. Skinnerclaimed that the same type of func-tional scheme accounts for verbalbehaviour as well, with only minormodifications. For instance, if you arehungry (‘motivating operation’), andyour mother is at home (‘discriminat-ing stimulus’), you may say ‘Giveme some food’ (‘response’). If yourmother gives you some food (‘rein-forcer’), she rewards your verbalbehaviour and you are more likely tosay ‘Give me some food’ next time youare hungry and she is in the house.

In a review of Verbal Behavior thatbecame more influential than thebook itself, Noam Chomsky (1959)launched an all-out attack on Skinner’sapproach to language as behaviour. He

of meaning, his commitment tobehaviourist principles in the study oflanguage guided research in theoreti-cal and applied linguistics up until thecognitive revolution in the late 1950sand early 1960s.

In philosophy and the philosophyof language, behaviourism becameknown as logical behaviourism. Logi-cal behaviourism, unlike classicalbehaviourism, did not deny the valid-ity of inquiring into mental states.According to logical behaviourism,however, all meaningful – that is veri-fiable – statements about mental phenomena are translatable intostatements which refer to physicalconcepts only. For instance, to beangry includes changes in tone ofvoice, facial expression, gesture andso on. For the logical behaviourist, allthese physical properties are not justmanifestations of ‘anger’, but defini-tions of the concept ‘anger’. In otherwords, logical behaviourists werecommitted to the notion that mentalphenomena can be reduced to behav-ioural dispositions. Logical behav-iourism is mostly linked to the workof Gilbert Ryle, and the later workof Ludwig Wittgenstein, but otherphilosophers with behaviourist affilia-tions include Rudolf Carnap and laterW. V. O. Quine. The main criticismslevied against logical behaviourismare that several mental states are notreducible to behavioural dispositions(for instance, ‘belief’ cannot be under-stood independently of ‘desire toverify the belief or act on the belief’)and that it has nothing to say aboutconceptual relations between mentalconcepts themselves.

By the early 1950s behaviouristpsychology had tackled a number of

20

BEHAVIOURISM

early 1960s, it still survives in someapplied settings (for instance in thetreatment of autism) and in certainstrands of animal learning theory. Thestudy of the mind/brain and their rolein explaining behaviour are now atthe forefront of scientific investiga-tion, but the legacy of behaviourismhas not disappeared completely:behaviourism, for instance, shareswith contemporary connectionistapproaches to psychology and lan-guage the emphasis on associativelearning and on the role of experienceon learning.

Primary sourcesBloomfield, L. (1933). Language. New

York: Henry Holt.Chomsky, N. (1959). ‘Review of Verbal

Behavior’. Language 31: 26–58.MacCorquodale, K. (1970). ‘On Chom-

sky’s Review of Skinner’s Verbal Behav-ior’. Journal of the ExperimentalAnalysis of Behavior 13: 83–99.

Ryle, G. (1949). The Concept of Mind.London: Hutchinson.

Skinner, B. F. (1957). Verbal Behavior.New York: Appleton-Century-Crofts.

Watson, J. B. (1913). ‘Psychology as theBehaviourist Views it’. PsychologicalReview 20: 158–77.

Further readingBaum, W. M. (2003). Understanding

Behaviorism: Science, Behavior, andCulture. New York: Harper Collins.

Skinner, B. F. (1953). Science and HumanBehavior. New York: Macmillan.

Skinner, B. F. (1986). ‘The Evolution ofVerbal Behavior’. Journal of the Experimental Analysis of Behavior 45:115–22.

Stavroula-Thaleia Kousta

claimed that human behaviour andmore specifically language are infinitelymore complex than the types of animalbehaviour Skinner had previously stud-ied, and thus principles of animalbehaviour cannot be applied to verbalbehaviour. He showed that many of theconcepts Skinner used, such as stimu-lus, response, and reinforcement, werehopelessly vague and had no explana-tory value in accounting for linguisticknowledge. He argued that Skinner’saccount is also a poor model of lan-guage acquisition, since it fails toaccount for the speed, efficiency anduniformity with which children acquirelanguage. According to Chomsky, thebasic facts of language acquisitionpoint to the existence of an innate pre-disposition to acquire language. Thisreview, along with work carried outin computer science, was instrumentalin the ‘demise’ of behaviourism and ineffecting the move from studyingbehaviour to studying mental processesusing a computer metaphor (the cogni-tive revolution).

A direct response to Chomsky’s crit-icisms never came from Skinner him-self. Kenneth MacCorquodale, one ofSkinner’s former students, published aresponse (1970) more than ten yearsafter the review appeared, in which heshowed that Chomsky had misunder-stood several of the points Skinner wasmaking and that a lot of Chomsky’scriticisms were not actually criticismsof Skinner’s work but of other behav-iourist work popular at the time. Thisresponse, however, had a very limitedimpact, partly because a paradigm shifthad already taken place and behav-iourism had given way to cognitivism*.

Although behaviourism ceased tobe in the mainstream of science in the

21

significant role to metaphoricity andiconicity.

The centrality of meaning in cogni-tive grammar is one of the mainaspects of the generative semanticsheritage in cognitive linguistics.Meaning in CL is taken to be funda-mental and directly mapped ontophonological structures. This entailsthe rejection of modularity (the beliefthat there exist encapsulated modulesthat process phonological or morpho-logical information without referenceto other levels of language) and theadoption of an encyclopedic model ofthe lexicon, that is a belief that allsorts of encyclopedic informationabout lexical items are part and parcelof their meaning and correspondinglyno systematic distinction can bedrawn between lexical (word-related)and encyclopedic knowledge (knowl-edge about the world). Typically, lexi-cal knowledge is organised in frames,organised complexes of informationrelating to a given concept.

The centrality of meaning is notmerely axiomatic, but informs theentire CL edifice. CL focuses bothon lexical meaning and on grammati-cal meaning. Generative grammarignored the lexicon, seen as a reposi-tory of exceptions, and assumed thatgrammar was a formal structure, thatis one without meaning. Conversely,grammatical structure, according toCL, is the result of a process ofschematisation, that is abstraction ofconstructional meaning. So, in effect,CL claims that all language is mean-ing, including grammar, and thatgrammatical meaning is very abstractschematic meaning. Relatedly, CL isalso very keen on constructionalmean ing, that is the fact that some

COGNITIVISM

Cognitivism can be defined as thebelief that cognition mediates percep-tion, unlike in behaviourism. In thisbroad sense, cognitivism is synony-mous with mentalism*. But whilementalism has found expression ingenerative grammar, cognitive linguis-tics was initiated as an alternative tothe generative-transformational para-digm, with significant influences fromthe generative semantics* approach,Gestalt psychology and the psychol-ogy of prototypes. Its main earliestproponents were Ronald Langacker,George Lakoff and Len Talmy in theearly 1980s, but the approach soongathered a broad following in Europeand the United States.

See also: Generative Semantics;Mentalism; Metaphor; Prototype;Transformational-GenerativeGrammar,Key Thinkers: Sapir, Edward;Whorf, Benjamin Lee

The basic organisations of cognitivelinguistics (CL) are semantic-phono-logical mappings and in this respectCL is like all semiotic approaches tocommunication. The significant differ-ence is that CL is squarely focused onmeaning and rejects any non-semanticcomponent to the grammar (for exam-ple, empty categories). Other charac-teristics that distinguish CL from othernon-cognitive approaches to grammarare prototypicality, schematicity, andperspectivism, as well as more gener-ally, the rejection of modularity andconsequently the adoption of an ency-clopedic semantics, and the belief thatcognition is embodied and affords a

22

COGNITIVISM

Conceptualisation is a central con-cept in CL. As we have seen, meaningis the object of CL. Meaning can bedefined as the ‘content’ of an idea,concept, and so on, and its construal,that is the way in which the speakerlooks at the content, the perspective inwhich it is seen. For example, the sen-tences ‘Mary kicked the ball’ and ‘Theball was kicked by Mary’ clearlydescribe the same factual situation(content) but with two different per-spectives, one that privileges the activerole of Mary and another, the passiveconstruction, that privileges the recip-ient of the action. The same reasoningapplies to ‘downhill/uphill’, whichdepends on where the speaker, so tospeak, puts him/herself.

Conceptualisation is said to beembodied and to play an active role inthe construal of reality. The embodiednature of conceptualisation refers tothe fact that many metaphoricalexpressions are based on commonhuman characteristics. For example, acommon metaphor is that UP isGOOD and DOWN is BAD. This isseen to be rooted in the upright stanceof the human body and the elevatedposition of the human head. Thus, farfrom being based on abstract cate-gories, conceptualisation is based onshared physical characteristics ofhumans.

The role of language and its cate-gorisation in our perception of realityare a significant philosophical tenet ofCL, which rejects the idea that realityis simply mirrored in language and/ormental representations, and insteadsees an active role of linguistic cate-gorisation in shaping the way theworld is represented linguistically. Theembodied nature of conceptualisation

constructions, such as ‘X let alone Y’have a meaning that cannot be pre-dicted from the sum of its parts. Sig-nificantly, parts of this meaning wouldtraditionally be considered pragmatic(hence reiterating CL’s rejection ofmodularity).

The notion of prototypicality, elab-orated by Eleanor Rosch in the 1970s,and originally applied to categories(such as ‘bird’ or ‘furniture’), is alsoapplied to grammatical categories, ina way recalling generative-semanticistJohn Ross’s notion of ‘squish’. Thismeans that some verbs, to take a clearexample, are better examples of verbthan others (for example, actionverbs, such as ‘kick’, which involvethe transfer of force from one entity toanother, are better examples of verbsthan state-verbs, such as ‘own’).

CL shares with generative seman-tics the programmatic rejection of thecore/periphery distinction. In genera-tive grammar, it is common practice tofocus on the regularities of grammar(core), rejecting exceptions to theperiphery, where they are accountedfor by pragmatic (the waste-baskettheory of pragmatics) or other factors.CL denies the validity of this distinc-tion, and seeks to explain all phenom-ena of language using the samecognitive principles and hence showsa particular interest in the seemingirregular patters found, for example,in constructions. Moreover, CL incor-porates pragmatic and functionalexplanations within its conceptualapparatus, hence rendering a core/periphery distinction impractical.This parallels CL’s rejection of modu-larity, which presupposes that thesame principles and mechanismsapply at all levels of language.

23

COGNITIVISM

show is over’). Blending is a morerecent theory of how mental spacescan interact (blend) and result in anew mental space with emergent fea-tures (features which are not presentin either of the original spaces in theblend). Blending has been applied tometaphors, especially novel meta -phors, and to an increasing variety ofphenomena.

It should be stressed that CL shareswith some theories of language suchas natural phonology and morphol-ogy the use of iconicity (the fact that alinguistic form reflects the form of thereferent it describes) as an explana-tory concept. For example, the com-parative and superlative forms of theadjective are often longer than thepositive form: ‘pretty – prettier – pret-tiest’. Relatedly, CL has shown sinceits inception a fondness for visualschematic representations that useiconic means, such as bolded lines toindicate focus, that have considerableweight in the explication of the con-cepts, at least intuitively and pedagog-ically. CL has also always shown avery distinct interest in space and itslinguistic representation: witness theintensive study of prepositions whichhas been one of the great contribu-tions of CL to grammatical analysis,and also the conceptualisation ofmany events as consisting of a figureand a ground, a trajectory or a land-mark, for example.

Finally CL asserts that a good theoryof language is a theory of language inuse. This is mostly a programmaticclaim, but recent developments of CLhave emphasised the use of corpora*of naturally occurring utterances,analyses of conversational data, and anincreased attention toward pragmatic

extends to these factors as well, withthe assumption that knowledge isorganised in ways that are influencedby the needs, goals, and culturalassumptions of the speakers.

Other concepts closely associatedwith CL are mental spaces, which canbe defined as a conceptualisationinformed by the relevant frames by agiven speaker of a given situation,event, and so on. Mental spaces in CLare similar to possible worlds in formalsemantics. Mental spaces have beenwidely used, for example, to explaincounterfactuals, but have also beenextended to the theory of literature,deixis, and discourse phenomena.

The study of metaphor* has alsobeen a central interest of CL. Mostlinguistic approaches see metaphori-cal meaning as a trope – a figure ofspeech – and would relegate it to theperiphery of language. CL converselyargues that most meaning is meta -phorical in nature. Metaphors areseen as mappings between twodomains; however, the significant dif-ference between CL accounts ofmetaphors is that CL sees meta -phors as grounded in metaphoricalschemata of great generality, such as‘argument is war’, which underliesuch metaphors as ‘he shot downmy argument’ and ‘Mary won thedebate’. Furthermore, CL sees mean-ing dynamically expanding from lit-eral meanings to other metaphoricalones. For example, the meaning of thepreposition ‘over’ is literally relatedto spatial orientation (an object isabove another), but that meaning isthen extended metaphorically tomotion over an object, and eventuallyto being past the object, and hence tothe meaning of finished action (‘the

24

COMPOSITIONALITY

The notion that the meaning of a com-plex expression is a function of themeanings of its parts. The ‘Principleof Compositionality’ has played animportant role in the work of a largenumber of philosophers and semanti-cists including Gottlob Frege, DonaldDavidson and Richard Montague.

See also: Conventional Meaning;Language of Thought; Logic;Logical Form; Mentalism; ModelTheoretic Semantics; PossibleWorld Semantics; Propositions;Truth TheoriesKey Thinkers: Davidson, Donald;Frege, Gottlob; Montague,Richard; Quine, W. V. O.

The Principle of Compositionality is aguiding principle for many semanti-cists and philosophers, notably Frege,Davidson and Montague. It is some-times referred to as ‘Frege’s Principle’,although not everyone agrees thatthe idea originated with him. Frege(1892) states it as follows: ‘The mean-ing of a complex expression is a func-tion of the meanings of its parts andtheir syntactic mode of combination’.

Perhaps the main reason that theprinciple is seen as important is therole it plays in explaining the creativ-ity* and systematicity of language.We can explain how we are able toproduce and understand an infinitenumber of utterances if we assumethat we know the meanings of indi-vidual expressions and how to com-bine them into larger units.

It follows from this principle thatthe contribution of an expression,say the lexical item ‘red’, to a more

phenomena and sociolinguistic varia-tion at large.

Primary sourcesLakoff, George and Mark Johnson

(1980). Metaphors We Live By.Chicago: University of Chicago Press.

Langacker, Ronald W. (1987). Founda-tions of Cognitive Grammar, Volume I,Theoretical Prerequisites. Stanford, CA:Stanford University Press.

Langacker, Ronald W. (1991). Founda-tions of Cognitive Grammar, Volume II,Descriptive Application. Stanford, CA:Stanford University Press.

Rosch, Eleanor (1973). ‘Natural Cate-gories’, Cognitive Psychology 4: 328–80.

Talmy, Leonard (2003). Toward a Cogni-tive Semantics – Volume 1: ConceptStructuring Systems. Cambridge, MA:MIT Press.

Talmy, Leonard (2003). Toward a Cogni-tive Semantics – Volume 2: Typologyand Process in Concept Structuring.Cambridge, MA: MIT Press.

Further readingCroft, William and D. Alan Cruse (2004).

Cognitive Linguistics. Cambridge:Cambridge University Press.

Evans, Vyvyan and Melanie Green (2006).Cognitive Linguistics. An Introduction.Edinburgh: Edinburgh University Press

Geeraerts, Dirk (ed.) (2006). CognitiveLinguistics: Basic Readings. Berlin/NewYork: Mouton de Gruyter.

Geeraerts, Dirk and Hubert Cuykens (eds)(2007). The Oxford Handbook of Cog-nitive Linguistics. Oxford: Oxford Uni-versity Press.

Ungerer, Friedrich and Hans-Jörg Schmid(1996). An Introduction to CognitiveLinguistics. London: Longman.

Salvatore Attardo

25

CONNOTATION/DENOTATION

Further readingDavidson, D. and G. Harman (eds)

(1972). Semantics of Natural Language.Dordrecht: Reidel.

Dowty, D. R., R. E. Wall. and S. Peters(1981). Introduction to MontagueSemantics. Dordrecht: Kluwer.

Thomason, R. (ed.) (1974). Formal Phi-losophy: Selected Papers of RichardMontague. New Haven, CT: Yale Uni-versity Press.

Billy Clark

CONNOTATION/DENOTATION

Connotation, from the medieval Latincompound verb con-noto, refers to animplied or accompanying feature, asthe comitative prefix con suggests.Denotation etymologically derivesfrom the postclassical Latin com-pound verb de-noto which conveysthe idea of singling out an entity byway of distinctive features.

See also: Sense/Reference; Signsand SemioticsKey Thinkers: Aristotle; Hjelmslev,Louis; Mill, J. S.; Saussure,Ferdinand de

The notion of denotation can betraced back to Aristotle’s classifica-tion of denominative names (Cate-gories I) where he labelled ‘paronyma’(denominative) the names of entitieswhich are morphologically derivedfrom the names of something else, forexample ‘grammaticus’ (‘grammar-ian’) derives from ‘gramatica’ (‘gram-mar’). This is tantamount to saying

complex expression should be thesame in all cases. The contribution of‘red’ should be the same in the nounphrase ‘red wool’ as in ‘red cotton’and in any other noun phrase. W. V. O.Quine (1960) pointed out some prob-lems with this assumption. For exam-ple, the contribution of ‘red’ in ‘redapple’ is not the same as the contribu-tion of ‘pink’ in ‘pink grapefruit’, sincea red apple is usually understood to bean apple with (mainly) red skin on theoutside while a pink grapefruit is usu-ally understood to be a grapefruit with(mainly) pink flesh on the inside.

Propositional attitude* reports(utterances which contain a statementabout an individual’s attitude to aproposition) provide a further test forthe principle. For example, we canbelieve all three of the following:

(1) Chris thinks his next-door neigh-bour is considerate.

(2) Chris thinks the person who reversedinto his bicycle is inconsiderate.

(3) Chris’s next-door neighbour is theperson who reversed into his bicycle.

Despite these problems, most theo-rists prefer to retain the Principleof Compositionality rather thanattempt ing to develop a new accountof the productivity and systematicityof linguistic knowledge.

Primary sourcesFrege, G. (1892). ‘On sense and meaning’.

In Peter Geach and Max Black (eds)(1980), Translations from the Philo-sophical Writings of Gottlob Frege.Oxford: Blackwell, 56–78. First edition1952.

Quine, W. V. O. (1960). Word and Object.Cambridge, MA: MIT Press.

26

CONTINUITY

directional relationship between a setof entities in the world and the linguis-tic expressions used to refer to them,whereas reference/extension is viewedas a more specific relationship betweenone specific entity in the world and theword in a given utterance used to men-tion it. In a similar vein, connotation isoften paired with sense/intension, butwhile sense encompasses the whole setof properties of an entity, intensionrefers to the qualities of the entitywhich are implied in a given utterance.In modern linguistics Saussure offereda clear distinction between the twolabels. Stemming from his semioticdichotomy of the sign, which is con-ceived of a signifier and a signified,denotation refers to the meaning of aword as is given in a dictionary, theobjective semantic content codified bythe signifier, whereas connotationrefers to any other semantic implica-tions that are part of the speaker’svalue in a given context.

Primary sourcesLyons, J. (1995). Linguistics Semantics.

An Introduction. Cambridge: Cam-bridge University Press.

Further readingRigotti, E. and A. Rocci (2006). Denota-

tion versus Connotation. In Ency-clopaedia of Language and Linguistics.Elsevier. 436–44.

Annalisa Baicchi

CONTINUITY

A view arising from the theory thathuman language is the product of a

that the abstract name ontologicallyprecedes any morphological variant.This view was spread in the earlyMiddle Ages by St Anselm of Canter-bury who introduced the notion ofindirectness. On the one side, denom-inatives point directly to the property(‘grammar’) and indirectly to the sub-stance (‘grammarian’); on the other,denominatives have both meaning(the concept of grammar, ‘significa-tio’) and reference (the person who isa grammarian, ‘appellatio’). Two hun-dred years later, in the fourteenth cen-tury, William of Ockham tackled theissue of the semantics of adjectivesfrom the opposite perspective andconceived of a distinction between‘absolute’ and ‘connotative’ terms.The former refers either to the namesof qualities (albedinem, ‘whiteness’)or of substances (canis, ‘dog’). Thelatter refers to both the names of qual-ities and the names of substances.Along the centuries St Anselm’s theoryof denotation was progressively aban-doned and it was Ockham’s versionthat became prominent in the nine-teenth century within the philosophi-cal approach to semantics of JohnStuart Mill. It was Mill who first men-tioned the dichotomy between conno-tation and denotation in his System ofLogic (1843), two labels that he usedin a similar way as sense and refer-ence*, or extension and intension.Mill limited Ockham’s ‘absolute’terms to proper names and to abstractnames for attributes (‘whiteness’).

The meaning of the two terms isclosely related to the theoreticalframe work taken as a perspective,and it varies broadly across theories.In logico-philosophical semantics,deno ta tion corresponds to the bi-

27

CONTINUITY

animals. The work of Charles Hock-ett in the 1970s and 1980s is central tothis effort to describe the key elementsof human language, including forexample the use of the vocal-auditorychannel, semanticity, arbitrariness,and creativity*. But Hockett andothers continually changed theirminds about what features should beon the list, and the issue of what con-stitutes human language has sincebeen recognised as a diversion.

One problem for proponents ofcontinuity theory is that they oftenpresuppose that evolution takes placein a linear and hierarchical way, fromthe most ‘primitive’ life forms, to themost ‘sophisticated’, namely humans.However, this view of evolution failsto take account of the fact that whilemany animals, including humans andchimpanzees, may share a commonancestor, they are not necessarilydirectly related. In other words,human language may have evolvedafter the human and chimpanzee‘branches’ of an evolutionary tree hadsplit apart. In that case there is noreason why a chimpanzee should evershow human language capabilities.Most scientists now believe that theynever will.

Noam Chomsky has often expressedscepticism about the idea that the lan-guage faculty evolved in the same wayas other biological features, but thelack of evidence for continuity doesnot mean it is false. As Stephen Pinkerargues in The Language Instinct(1994), the trunk of an elephant is afeature unique to that creature, butbiologists do not spend time testingother animals to see if they are able touse their noses in the same way asan elephant uses its trunk. Viewing

unique biological and geneticallyencoded faculty. The theory of conti-nuity suggests that human languagemust have evolved over time in thesame way as other biological andgenetically inherited features. Yetthere is little hard evidence to suggestthat any other animals have a lan-guage faculty with similar capabili-ties. This is sometimes known as the‘continuity paradox’: language musthave evolved in humans, but there isno surviving evidence that it did. Formany thinkers the ‘continuity para-dox’ is simply an unsolved problemthat, in the absence of any explana-tion, needs to be accepted.

See also: Cognitivism; Creativity;Mentalism; Transformational-Generative Grammar; UniversalGrammarKey Thinkers: Chomsky, Noam;Hockett, Charles

The theory of continuity holds thatthe human language faculty must haveevolved gradually over time, ratherthan appearing fully formed as asingle mutation and that it must bepossible to identify the ‘continuity’ inthe evolutionary development of lan-guage in humans. Attempts to provethat language in humans is the resultof a long process of evolution by nat-ural selection have usually involvedanimals. Parrots have been trained to‘talk’, while chimpanzees, lacking thephysical apparatus necessary forhuman-like speech, have been taughtsign language. One of the problemsencountered by such experiments isdeciding what constitutes ‘language’in a human sense and establishingthe criteria against which we shouldmeasure the linguistic abilities of

28

CONVENTIONAL MEANING

the sounds and the meaning therebyexpressed. The stored meaning ofexpressions in the lexicon is an exam-ple of conventional meaning. Conven-tional meaning must be learnt, as itcannot be inferred based on principlesof rationality. The distinction betweenconventional meaning (what an ex -pression means) and speaker’s mean-ing (what a speaker means by using aparticular expression) is sometimesused to draw the line between seman-tics and pragmatics.

See also: Nonnatural Meaning;Signs and SemioticsKey Thinkers: Aristotle; Grice, H.P.; Lewis, David; Plato; Saussure,Ferdinand de; Strawson, P. F.

The question of whether the origin oflinguistic meaning lies with naturalnecessity or human convention israised in Plato’s Cratylus, whereSocrates expresses the view that lin-guistic meaning is a matter of habit, aview echoed in Aristotle’s De Inter-pretatione. In contemporary linguis-tics the conventionality of linguisticmeaning is acknowledged in Ferdi-nand de Saussure’s arbitrary associa-tion of signifier and signified, whichengenders the possibility for this asso-ciation to shift and change across lan-guages and over time.

More recently, H. P. Grice explainedlinguistic meaning as a type of non-natural meaning*. Contrary to othertypes of signs which are symptomaticof particular states of affairs and thusexpress natural meaning, the meaningof a linguistic sign is dependent onhow speakers use it to bring about par-ticular effects in an audience. Such useis in turn more or less constrained bythe customary use of the sign by

human language as a unique adapta-tion, just as an elephant’s trunk isunique, does not preclude evolutionarycontinuity. Advances in genetics may inthe end show how humans came tohave language when other animals donot. But in the mean time it is notunreasonable to assume that, as Pinkersays, ‘There were plenty of organismswith intermediate language abilities,but they are all dead’ (1994: 346).

Primary sourcesAitchison, Jean (1989). The Language

Instinct: An Introduction to Psycholin-guistics. London: Unwin Hyman.

Sebeok, T. (ed.) (1968). Animal Commu-nication: Techniques and Study Resultsof Research. Bloomington: Indiana Uni-versity Press.

Pinker, Steven (1994). The LanguageInstinct: The New Science of Languageand Mind. Harmondsworth: Penguin.

Further readingUriagereka, Juan. ‘The Evolution of Lan-

guage’, in Seed (http://seedmagazine.com), 25 September 2007. http://seed-magazine.com/news/2007/09/the_evolution_of_language.php?page=1(acces sedon 30 January 2008).

Christopher Routledge

CONVENTIONALMEANING

The fact that a sequence of soundsexpresses a certain meaning in virtueof a tacit agreement among speakersat a certain time and place, rather thanbecause of any necessary link between

29

CONVERSATION ANALYSIS

Grice, H. P. (1989), Studies in the Wayof Words. Cambridge, MA: HarvardUniversity Press. 22–40.

Morgan, Jerry (1978). ‘Two types of con-vention in indirect speech acts’. In P.Cole (ed.), Syntax and Semantics, vol. 9:Pragmatics. New York: Academic Press.261–80.

Traugott, Elizabeth (1999). ‘The role ofpragmatics in semantic change’. In J.Verschueren (ed.), Pragmatics in 1998.Antwerp: International PragmaticsAssociation. 93–102.

Marina Terkourafi

CONVERSATIONANALYSIS

The systematic study of all aspects oftalk in social interaction. The findingsof this important discipline, started byHarvey Sacks in the mid-1960s, aresignificant to all social sciences.

See also: (Critical) DiscourseAnalysis; Speech Act TheoryKey Thinkers: Austin, J. L.; Grice,H. P.; Sacks, Harvey; Searle, John

According to Anthony Liddicoat, con-versation analysis (CA) – also knownas ethnomethodology – originated inthe work of Harold Garnfikel, an eth-nomethodologist whose research inthe 1960s centred on how members ofa certain community conduct andunderstand social actions and interac-tions Liddicoat (2007: 2). These ideaswere developed by Sacks who deter-mined that conversation is an organ-ised activity that has a systematicstructure. Ironically, Sack’s claim was

speakers in a community, that is by itsconventional meaning. According toGrice, ‘the conventional meaning ofthe words used’ is part of the input tothe derivation of conversational impli-catures (1967/1989: 31). Nonnaturalmeaning thus encompasses both conventional (language-dependent)and intentional (speaker-dependent)aspects of meaning.

The continuity between conven-tional and intentional aspects ofmeaning is the topic of David Lewis’sbook Convention (1969), where heinvestigates the rational bases for con-ventions, explaining these as solutionsto co-ordination problems. Contextplays an important role in sustainingthe transition from nonce meaning toconventional, or coded, meaning. Therelated notions of conventionalisa-tion and standardisation have beenproposed to account for the grad-ual decontextualisation of recurringmean ings of linguistic expressions,such that these meanings eventuallybecome part of an expression’s con-ventional meaning.

Primary sourcesLewis, David (1969). Convention: A

Philosophical Study. Cambridge, MA:Harvard University Press.

Strawson, Peter (1964). ‘Intention andconvention in speech acts’. Philosophi-cal Review 73, 439–60.

Further readingBach, Kent (1998). ‘Standardisation revis-

ited’. In A. Kasher (ed.), Pragmatics:Critical Concepts. Vol. IV. London:Routledge. 712–22.

Grice, H. P. (1967). ‘Logic and Conversa-tion’. William James Lectures, HarvardUniversity typescript. Reprinted in

30

CONVERSATION ANALYSIS

connection has been established. Thisopening is followed by making an iden-tification and exchanging greetings.

This pattern recurred in the major-ity of the telephone calls analysedby Schegloff. However, variationsoccurred as a result of a number ofreasons, including the setting of thecall. If the call is made to a workplace,for instance, the answerer would startby giving the name of the organisationand not the personal name. Althoughface-to-face conversations are charac-terised by similar features, there are afew differences. These result from thefact that interlocutors can see eachother, which would influence the iden-tification part of the opening sequence.

Members of a speech communityalso need to be aware of the way inwhich conversations are closed, orconcluded. Since talk in social interac-tions is a collaborative process, it isimportant to ensure that a conversa-tion is not ended in an abrupt mannerthat would deprive any of the partici-pants of the right to contribute. It hasbeen observed that participants usecertain expressions that signal the endof the conversation, the most commonamong which is ‘goodbye’ or one ofits synonyms. However, a closingsequence cannot just be introduced atany juncture of the conversation tosignal its conclusion. Before terminat-ing a verbal interaction, a pre-closingsequence alerting the participants tothe nearness of the conversation endhas to be introduced. This is meant toensure participants are given theopportunity to make whatever contri-butions they deem necessary. Sche-gloff and Sacks define a pre-closingsequence as a turn after the pro- duction of which every participant

proposed in the mid-1960s, during theheyday of the Chomskyan paradigm,which regarded ‘talk’ as being unfit asa source of linguistic data because it isriddled with irregularities and flaws.However, Sacks and his collaborators,Emanuel Schegloff and Gail Jefferson,proved that conversation is an orderly,collaborative activity characterised byconsistent principles that define whatlinguistic forms interlocutors use toopen and close their contributions,how they repair and rephrase theirtalk, and when they take the floorduring the course of a conversation. Inaddition to these, CA also investigatesthe role of proxemics, or inter-per-sonal distance and body language, incommunication. The principles of CAhave significant implications for anumber of disciplines that include lin-guistics, social psychology and com-munication, to name a few. It istherefore useful to summarise some ofthe areas in which CA has proveduseful or enlightening.

Since CA investigates the sequentialnature of conversation, analysts havestudied the systematically recurring lin-guistic units that open these conversa-tions. The majority of these studieshave focused on telephone conversa-tions because of the limitations result-ing from the absence of physical cluessuch as body language, facial expres-sions and eye gaze. Schegloff has inves-tigated opening sequences in telephoneconversations (1979, 1986) and con-cluded that they are characterised by apredictable sequence in which theanswerer of the telephone call speaksfirst, saying ‘hello’. Since making acall is considered to be a summonsthat requires a response, the answereris expected to acknowledge that a

31

CONVERSATION ANALYSIS

which reflects the turn-taking mecha-nism of conversation, this may sometimes be altered for a certain commu-nicative purpose. If an interlocutor,for instance, needs more informationbefore he/she answers a question, thiswould result in the interruption of thetwo elements of the adjacency pair.Cook gives the following example:

A: Did you enjoy the meal? (B: Did you?A: Yes.) B: So did I.

Cook explains that the initial adja-cency pair, which is contributor A’sfirst turn and contributor B’s secondturn, has been interrupted by a secondadjacency pair, the one between paren-theses. He calls this interrupting ele-ment an ‘insertion sequence’ (1989:54). It is evident from the example thatspeaker B needed more informationbefore responding to the question.

Talk in social interaction is a col-laborative endeavour in which turn-taking seems to be a highly structuredprocess. Although interruptions andgaps are not uncommon, conversationparticipants seem to have tacit knowl-edge of when to ‘take the floor’ andoffer their contribution to the conver-sation. When asked about what cluesthey use to start speaking, peopleoften mention pauses and length ofcontribution as the most reliableclues. Citing several examples, how-ever, Liddicoat argues that neither ofthese clues are valid markers of aturn’s end (2007: 54–6). Sacks et al.(1974) speculate that speakers knowwhen to take the floor because theyhave tacit knowledge of turn con-structional units. They define these aslinguistic forms that are not identifi-able in terms of syntactic structure. A

‘declines at least one opportunity tocontinue talking’ (1973: 214). Expres-sions that could function as pre-clos-ing sequences include ‘alright’, ‘right’,and ‘okay’ and they are usuallymarked by falling intonation. If noneof the speakers introduces a newtopic, the pre-sequence will be fol-lowed by a closing sequence thatwould mark the conclusion of the con-versation. However, if a new topic isintroduced, the conversation will con-tinue unless one of the participants isunable to proceed because of timeconstraint or some other reason.

Another phenomenon characteristicof conversations is ‘sequence pairs’.These are conversation contributionsby different speakers that are sequen-tially placed next to each other. Schegloff and Sacks (1973) proposedthe term to refer to sequences suchas greeting-greeting and question-answer. Since conversation is a collab-orative social act to which participantsare expected to contribute, a questionshould be answered and a greetingreciprocated. The failure to do sowould be considered an act of non-compliance and a violation of socialetiquette. Guy Cook (1989) classifiesresponses to adjacency pairs into ‘pre-ferred’ and ‘dispreferred’. Accordingto his framework, agreeing with anassessment is the preferred sequence,while disagreeing with it is dispre-ferred. As for responses to questions,giving the expected answer is the pre-ferred response, while giving an unex-pected answer is the dispreferredresponse. Cook adds that a dispre-ferred response is usually explicitlyjustified or non-verbally marked.Although the two parts of an adja-cency pair occur next to each other,

32

CONVERSATION ANALYSIS

type of silence may also result in anextended lapse in the conversation.Another type of silence, however,occurs at a juncture of the conversa-tion where a specific participant isexpected to contribute, for example,to answer a question. This type ofsilence is considered problematicbecause it results in an act of non-compliance. Strategies used to repairfailure to contribute include repeatingthe question or changing the topic.

Overlapped speech, on the otherhand, may be the result of a partici-pant’s misreading of a pause as theend of the speaker’s turn and a sign offorfeiting the floor. When that hap-pens, the overlap may persist for sometime before one of the participants dis-continues. Another possible reasonfor overlaps in conversation is when,at the end of a speaker’s turn, twointerlocutors take the floor at thesame time. If an overlap occurs at apoint that is not a potential juncturefor speaker change, it is considered aninterruption. O’Grady et al. believethat how overlaps are perceived is cul-ture-specific. Communities that speakthe standard North American Englishvarieties, they argue, adhere to a rigidturn-taking mechanism, which resultsin very few overlaps. The East Euro-pean Jewish community in New York,however, views overlaps favourablybecause they reflect the speakers’interest in the conversation. Hence,they call overlaps in which speakersfinish one another’s sentences ‘cooper-ative overlaps’ (2005: 491). Besidecultural differences, other factors thatmight determine overlaps include thetopic, type of conversational interac-tion and number of participants. In aninterview setting, for instance, where

turn constructional unit can be asingle word, a phrase, a clause, or alonger stretch of discourse. They addthat speakers have knowledge of thepragmatic function of these context-sensitive units, which enables them toidentify the end of turns. A variety ofother proposals have been offered toexplain what clues participants inconversation use to take the floor.These include utilising knowledge ofsyntactic, semantic and suprasegmen-tal features of conversation, in addi-tion to using visual clues to determineend of turns. Among these is the para-tone, which Brown and Yule define asa conversational paragraph whoseend is marked by a summarisingsequence, a significant pause and fall -ing pitch (1983: 101). Another relatedarea of interest is turn allocation,which investigates whether theperson holding the floor selects thenext speaker, or whether the subse-quent speaker self-selects. In their1974 frame work, Sacks et al. statethat a speaker holding the floor may:

(a) directly select a subsequent speakerby using a personal name, a pro-noun, or asking a question;

(b) indirectly select the next speaker byusing gaze and establishing eye contact;

(c) that a speaker holding the floor maynot select any one, which leaves itopen to participants to self-select.

Despite the tacit knowledge speakershave of turn-taking rules in conversa-tion, gaps and overlaps occur. Accord-ing to the framework in Sacks et al.,silence may take place at a transitionrelevance place, where any of the par-ticipants could claim the floor. This

33

CONVERSATION ANALYSIS

have received specific attention fromconversation analysts. First, a storyrequires a fairly extended turn, whichmay disrupt the orderly turn-takingmechanism of conversation; and,second, in order to maintain coher-ence and topic continuity, a story hasto be introduced at the right juncturein the conversational interaction. Jef-ferson (1978) proposes two factorsthat trigger narratives. The first iswhen a conversation topic reminds aspeaker of the events of a certain expe-rience. When that happens, she or hemay elect to relate the story; however,the conversation participants have tobe alerted by such expressions like ‘bythe way’, or ‘incidentally’. Accordingto Jefferson, these serve as disjunctsmarking topic transition and provid-ing some kind of a justification for thenarrative. Researchers have alsoasserted the interactive characteristicsof story-telling. Contrary to commonbelief, a story is not a monologue thatis delivered by a single speaker. Notonly do participants comment on andevaluate the narrative after its deliv-ery, they also frequently ask questionsabout and make remarks on the storywhile it is being told. Upon the com-pletion of the narrative, receivers areexpected to make some kind of aresponse. That response may be anevaluation of the story, a comment onits plot, or an expression of empathy.If silence follows the end of a story,that would be regarded as an act ofnon-compliance.

Deborah Schiffrin has proposeda model for the analysis and inter -pretation of spoken English thatfocuses on what she calls local coher-ence, or coherence between adjacentunits in discourse. This, she believes,

only two people are involved, over-laps are rare. On the other hand, if alarge number of people are involved inthe discussion of a heated topic, over-laps are likely to occur at a high rate.

Unlike written language, in the pro-duction of which a significant amountof time is spent on composition andrevision, spoken language is plannedin the here-and-now. Consequently,conversations are marked by ‘repair’,a process whereby utterances are mod-ified. Liddicoat states that althoughrepair may target a perceived mistake,people often use repair when they failto retrieve a specific item from theirlexicon, or mental dictionary, in atimely manner (2007: 171). Schegloffet al. (1977) have proposed a taxon-omy for the classification of repairsthat occur in conversation. Their categorisation takes into account whoinitiates the repair and who makesit. This yields four types which are:(1) self-initiated self-repair, whichdenotes deficiencies detected andresolved by the speaker; (2) self-initi-ated other-repair, which refers to defi-ciencies detected by the speaker, butrepaired by another; (3) other-initiatedself-repair, denoting irregularitiespointed out by an interlocutor, butrepaired by the speaker; and (4) other-initiated other-repair, referring torepairs detected and repaired by some-one other than the speaker. Typically,Schegloff et al. argue, repairs involvinggrammatical mistakes are initiated bythe speaker, whereas repairs of hearingproblems are other-initiated. Citing avariety of examples, the researchersindicate the different junctures of theconversation at which repair occurs.

Because of their unique characteris-tics as a specific genre of talk, stories

34

CONVERSATION ANALYSIS

Schiffrin also examined the role of‘well’ in response to requests. She sug-gests that in reaction to requests foraction, it is more likely for ‘well’ to beused in responses that indicate non-compliance to requests. In addition tothese contexts, Schiffrin investigatesother roles of ‘well’, such as markingrequests for clarification, self-repairsand reported speech. Schiffrin empha-sises the need for future research thatexamines the distribution of differentmarkers in different conversationalgenres. Such research, she asserts,would enable us to understand howcoherence is achieved in differenttypes of discourse.

Finally, an issue that is of utmostimportance to every conversationanalyst is transcription. Since speechis transient, it cannot be studied asit unfolds in real time. Hence,researchers have to tape-record andtranscribe it using a variety of con-ventions whose purpose is to captureall of the conversational aspects dis-cussed above. However, it is noteasy to come up with a system of transcription that is neutral and objective. Moreover, a transcriptionis not a substitute for the conversa-tion. It has been argued that tran-scription should be modified asresearchers listen again to the record-ing. Although the objective of thetranscription is to capture as manyfeatures of the conversation as possi-ble, a balance has to be struck be tweenthe comprehensibility of the transcrip-tion conventions and the accessibilityof these conventions to the targetedaudience (Liddicoat 2007).

It is widely believed that a tran-scription should include informationabout the conversation participants,

is accomplished by employing dis-course markers that speakers use toachieve not only semantic, but alsopragmatic goals. Although Schiffrindoes not give a concrete definition, heroperational definition is that dis-course markers are ‘sequentiallydependent elements which bracketunits of talk’ (1987: 31). Despite thedefinition’s limitation, which arisesfrom the difficulty of defining speechunits, Schiffrin identifies eleven dis-course markers: ‘oh’, ‘well’, ‘and’,‘so’, ‘now’, ‘then’, ‘I mean’, ‘but’,‘because’, ‘y’know’ and ‘or’. Then sheapplies these markers to the analysisof authentic conversation explaininghow they are used by participants toachieve coherence through accom-plishing a variety of communicativegoals. Schiffrin (1985) argues that‘well’ is the most versatile discoursemarker because it does not have inher-ent semantic or structural attributes.Its meaning, however, derives fromthe context in which it occurs.Schiffrin distinguished between occur-rences of ‘well’ in adjacency pairs(questions/answers, request/compli-ance sequences), and occurrences of‘well’ which cannot be explained interms of adjacency pairs.

Examining the distribution of ‘well’in answers to wh- questions andyes/no questions, Schiffrin found that‘well’ precedes answers to yes/noquestions 10 per cent of the time; onthe other hand ‘well’ precedes answersto wh- questions 21 per cent of thetime. She attributed this to the factthat yes/no questions delimit theupcoming answer to either an affir-mative or a negative response, hence‘well’ is not regarded as a coherentmarker for answers in these contexts.

35

CONVERSATION ANALYSIS

Practices, and Applications. Malden,MA: Blackwell Publishing.

Jefferson, G. (1978). ‘Sequential Aspectsof Storytelling’. In J. Schenkein (ed.),Studies in the Organization of Conver-sational Interaction. New York: Aca-demic. 219–48.

Liddicoat, Anthony (2007). An Introduc-tion to Conversation Analysis. London:Continuum.

O’Grady, William, John Archibald, MarkAronoff and J. Rees-Miller (2005).Contemporary Linguistics: An Intro-duction. Fifth edition. Boston, MA:Bedford/St Martin’s.

Sacks, H., E. Schegloff and G. Jefferson(1974). ‘A simplest systematics forthe organisation of turn-taking for conversation’. Language Vol. 50:696–735.

Schegloff, E. A. (1979). ‘Identification andrecognition in telephone conversationopenings’. In G. Psathas (ed.), EverydayLanguage: Studies in Ethnomethodol-ogy. New York: Irvington. 23–78.

Schegloff, E. A. (1986). ‘The routineas achievement’. Human Studies Vol. 9:2–3, 111–51.

Schegloff, E. and H. Sacks (1973). ‘Opening up closings’. Semiotica Vol. 7:289–327.

Schegloff, E., G. Jefferson and H. Sacks(1977). ‘The preference for self-correc-tion in the organization of repair in con-versation’. Language Vol. 53: 361–82.

Schiffrin, Deborah (1987). DiscourseMarkers. Cambridge: Cambridge Uni-versity Press.

Further readingBrown, Gillian, and George Yule (1983).

Discourse Analysis. Cambridge: Cam-bridge University Press.

Cook, Guy (1989). Discourse. Oxford:Oxford University Press.

time and place. Since the anonymity ofthe participants should be maintained,pseudonyms are usually used. A majorissue regarding transcriptions is wordrepresentation. Using the conven-tional orthography has the advantagethat it is a system known to the wholeliterate society. However, since con-ventional orthography is based on astandard variety of the language, somescholars, for example Duranti (1997),argue that using it would underminethe neutrality and objectivity of thetranscription. In addition to the con-versation, non-linguistic sounds suchas laughter and back channellingforms have to be in cluded. The tran-scription system also needs to indicatepauses within a contributor’s turn.Moreover, a pause between speakers’turns is also significant because itsabsence is an indication of overlappedconversation, which is significant tothe turn-taking mechanism. Supraseg-mental features such as stress andintonation have to be transcribedbecause of their interactional signifi-cance. In addition to these features,the transcription should account foroverlaps, body language and back-ground noise. Liddicoat asserts thattranscripts should be continuouslyupdated, adding that a transcript is‘not an objective account and that itwill always be a selective representa-tion of the data itself’ (2007: 50). Con-versation analysis is an important toolfor understanding language in use, butits usefulness to other disciplines, suchas anthropology and sociology, is alsovery significant.

Primary sourcesHutchby, Ian, and Robin Wooffitt (1998).

Conversation Analysis: Principles,

36

CORPORA

the empiricist J. R. Firth observed,‘You shall know a word by the com-pany it keeps.’ A corpus provides notonly such company (technically knownas concordance), but also many otherlinguistic insights. Linguists havenumerous pre-existing corpora avail-able to them, or they may choose toassemble their own, such as when theyseek to analyse or test criteria or char-acteristics not adequately representedin existing corpora.

An effective corpus is a representa-tive sample of natural language takenfrom the population being studied.Linguists must match their needsagainst the specific criteria of a corpussuch that the results of their researchand analysis may transfer to the realworld what the corpus is meant toreflect. There are a number of charac-teristics by which corpora vary fromone another. For instance, some cor-pora contain textual data in the formof written words, sentences and punc-tuation, while others contain speechdata in form of utterances, fragmentsand filled pauses. Another majorfactor is whether the corpus is raw orannotated. A raw corpus has not beenmodified from its original form, as inthe case of an online book or maga-zine. By contrast, an annotated corpushas been marked up with tags to iden-tify certain elements, like parts ofspeech. A corpus may focus on a par-ticular subject area or population, orit may represent a variety of subjectsor even genres. Corpora also vary bytime period, geographical region andsize, and they may even use differentlanguages or dialects. Some corporaare multilingual, containing text orspeech in multiple languages. Whenthose different languages express the

Duranti, A. (1997). Linguistic Anthropol-ogy. Cambridge: Cambridge UniversityPress.

Psathas, George (1995). ConversationAnalysis: The Study of Talk-in-Interaction. Thousand Oaks, CA: SagePublications.

Schiffrin, D. (1985). ‘ConversationalCoherence: The Role of “Well” ’. Lan-guage Vol. 61: 640–67.

Yousif Elhindi

CORPORA

Collections of written texts or spokenutterances representative of a particu-lar language, dialect, population orstyle. Linguists use corpora to providereal-world context in language analy-sis and language processing.

See also: Behaviourism,Conversation Analysis, (Critical)Discourse Analysis,Empiricism/RationalismKey Thinkers: Firth, J. R.; Sinclair,John

Linguistics calls upon language data toprovide evidence to support the exis-tence of particular linguistic phenom-ena or to challenge theories andalgorithms. Linguists may derive suchevidence from experience or personalknowledge, or they may refer to a writ-ten or recorded set of external data inthe form of a corpus. A corpus – fromthe Latin for ‘body’, as in ‘body oftexts’ – provides perspective and con-text for language data. Althoughmodern corpora typically exist as com-puter databases or files, the idea is farfrom new. In the mid-twentieth century

37

CORPORA

Times. A more modern version of thecorpus is the Associated Press WorldStream English, published in 1998and containing approximately 143million words. It combines the NorthAmerican News Text Corpus withEnglish-language text written in allparts of the world, usually not inAmerican English. This corpus ismarked up in Standard GeneralisedMarkup Language (SGML) to makeanalysis easier.

Newswire data is also used in theWall Street Journal (WSJ) corpus,which exists in several forms. An earlyversion of it, called CSR-I, containsWall Street Journal news articles readaloud in the early 1990s as part ofan effort to support research on large-vocabulary Continuous SpeechRecognition (CSR) systems. This wassoon revised as CSR-II, containing78,000 utterances, or seventy-threehours of speech. There are also writ-ten text variations of Wall Street Jour-nal data, containing thirty millionwords parsed in a treebank-style (cap-turing syntactic structure) with part ofspeech tagging.

Combining both written and spokentext from a wide range of British Eng-lish sources, the British NationalCorpus (BNC) contains 100 millionwords and has been updated severaltimes since its birth in 1991. Its mas-sive size is roughly equivalent to 1,000average paperback books. The writtenportion of the BNC corpus represents90 per cent of the total and containstext from British newspapers, periodi-cals, journals, non-fiction and fictionbooks, letters and essays. The remain-ing 10 per cent consists of spokenutterances by British English speakersin the form of informal conversation,

same content, this is known as a par-allel corpus. Existing corpora, avail-able through a number of sources,including the popular Linguistic DataConsortium (LDC), may be availablefor a fee or free of charge.

Among the first widely-availablecorpora are those in the the Browncorpus, named for Brown University,where it was created in the early1960s. First published as the StandardCorpus of Present-Day AmericanEnglish by Henry Kucera and NelsonFrancis, it contains 500 written textsand totals more than one millionwords of running text, averaging over2,000 words per text. Although it hasseen a number of proofreadings andrevisions since its creation, the Browncorpus remains an admirable repre-sentation of 1961 written AmericanEnglish, spanning a wide range ofstyles and varieties of prose, includingboth informative and imaginativeprose, such as text from newspapers,novels, non-fiction and academic pub-lications. Kucera and Francis con-ducted a thorough analysis of thecorpus, a landmark in modern corpuslinguistics. Another version of theBrown corpus now exists with taggingfor parts of speech.

Another important corpus includestexts published by the AssociatedPress (AP), aptly known as the NorthAmerican News Text Corpus, or moresimply the AP corpus. There are actu-ally multiple versions of this corpus. A1987 version of it contains fifteen mil-lion words, and a 1988 version con-tains thirty-six million. The contentcame from North American authors,writing news stories in American Eng-lish for the Los Angeles Times, theWashington Post and the New York

38

CORPORA

All of the corpora discussed aboveare monolingual, representing onlythe English language. By contrast, theCELEX corpus is multilingual corpus,including not only English, but alsoGerman and Dutch. It was developedas a joint effort by the University ofNijmegen, the Institute for Dutch Lex-icology in Leiden, the Max PlanckInstitute for Psycholinguistics inNijmegen, and the Institute for Per-ception Research in Eindhoven. Thiscorpus is a database of lexical wordforms, with detailed information onorthography (variations in spelling,hyphenation), phonology (phonetictranscriptions, variations in pronunci-ation, syllable structure, primarystress), morphology (derivationaland compositional structure, inflec-tional paradigms), syntax (word class,word class-specific subcategorisa-tions, argument structures) and wordfrequency (summed word and lemmacounts, based on recent and represen-tative text corpora).

A very popular multilingual corpusis the European Parliament Pro- ceedings Parallel (Europarl) corpus,assembled from 1996 to 2003. Ex -tracted from the proceedings of theEuropean Parliament, it includes par-allel text in eleven European lan-guages: Romanic (French, Italian,Spanish and Portuguese), Germanic(English, Dutch, German, Danish andSwedish), Greek and Finnish. It con-tains roughly twenty million words in740,000 sentences per language, witheach language aligned by sentence orby document. This corpus makes itpossible to align 110 language pairs,including languages (particularlyGreek and Finnish) that were poorlyrepresented in previous corpora. For

business or government meetings andradio shows. Its goal is to represent across-section of written and spokenBritish English, produced by a varietyof sources and intended for a mixtureof ages and backgrounds, representa-tive of the British population itself. TheBNC corpus includes part of speechtagging, as well as built-in identifica-tion of its structure, such as headings,paragraphs, and lists.

An important corpus for spokenlanguage data is the SWITCHBOARDcorpus, which consists of about 2,430spontaneous telephone conversationsby over 500 diverse native speakers ofall major dialects of American Englishbetween the ages of twenty and sixty.The conversation topics come from apredetermined list of topics, and thespeakers did not already know oneanother at the time of their recordedconversations. Developed at TexasInstruments in 1991, SWITCH-BOARD contains a total of over 240hours of data. Unlike the corporamentioned above, the spoken lan-guage within SWITCHBOARD is verynatural, not scripted from written text.However, a written version of thecorpus does exist, produced later viatranscription of the speech data. Foreach transcription, a time alignmentfile indicates the beginning time andduration of each word, informationthat is particularly useful for studyingthe phonetic characteristics of thespeech, as is common in developingand evaluating automatic speechrecognition systems. Another advan-tage of the corpus is that it comes withbackground information about eachspeaker along with details surround-ing each specific conversation, includ-ing the nature of the telephones used.

39

CORPORA

corpus to have multiple instances,each formatted according to a differ-ent standard. One such standard is atreebank, where each sentence isannotated as a tree structure basedupon its syntactic structure, whichitself is often annotated through partof speech tagging. A treebank mayrepresent phrase structure (as inthe case of the Penn Treebank) ordependency structure (as in the caseof the Prague Dependency Treebank).Another formatting standard, usedspecifically for bilingual corpora, is abilingual knowledge bank (BKB),which organises a bilingual corpusinto translation units, or aligned‘bitexts’.

Corpora play a supporting role inseveral fields of research. The mostobvious of these is corpus linguistics,which focuses its study of languagespecifically on corpora and real-worldtext. To accurately represent languageas it is actually used, a corpus must bevery large in volume. Corpus linguis-tics thus relies heavily upon auto-mated methods for conducting properanalysis of data. Technology has givenus new analysis techniques that werepreviously not possible for reasons ofsheer quantity of data, particularlywhen the corpus being studied isannotated with information about thetext structure, parts of speech, seman-tics, and so on. Statistical NLP alsouses data from corpora in order tocount words and compute word prob-abilities based on word frequencieswithin a training corpus, even takinginto account such factors as concor-dance and discourse features. Speechcorpora are particularly useful inspeech recognition technology, whichmay involve training software using

this reason, Europarl is frequentlyused within the field of statisticalmachine translation to train and eval-uate systems to an extent that was pre-viously not possible.

The CALLHOME corpora is a col-lection of speech data, originally cre-ated to support research in largevocabulary conversational speechrecognition, which was suffering dueto a shortage of speech data for lan-guages other than English. Its creatorsoffered its speakers free telephonecalls to their native country inexchange for permission to recordtheir conversations. The result is aseries of corpora, arranged by lan-guage (English, Spanish, Arabic,German, Mandarins and Japanese),containing unscripted telephone con-versations, up to thirty minutes inlength and originating in North Amer-ica. All speakers are native speakers oftheir respective languages, and manyinteractions are between family mem-bers or close friends. The CALL-HOME corpora were thus named fortheir collection method.

The corpora discussed here aregood examples of the range of cor-pora available to linguists, who selecta corpus based upon their specificneeds in analysing data or testing the-ories or natural language processing(NLP) systems. Data within a corpusmay be textual data or speech data,and even speech data may be tran-scribed such that it, too, exists astext. Text corpora are commonly for-matted to identify key elements orstructure within the data, and the for-matting conventions must be clear tothose who use the corpus so that theycan properly interpret the data it con-tains. It is even possible for a single

40

CORPORA

of the language used by the popula-tion of interest. When selecting acorpus or collecting data to build anew corpus, a linguist must identifywhat type of text a corpus must reflectas well as whether the results of workwith the corpus will transfer to thedomain of interest. That is, a corpusserves as a powerful tool for linguists,but it cannot wholly substitute for thereal world. In general, the larger acorpus, the more useful it is and themore likely it is to capture actual lin-guistic use. Yet even the largest corpusavailable is only as useful as the extentto which it reflects the target domainand offers a usable format for analy-sis. In addition to size, the breadth ofa corpus also affects its usefulness andapplicability for meaningful research.A valuable corpus for a linguist study-ing general language use, for example,should contain a variety of textsources, such as newspapers, text-books, popular writing, fictions andtechnical material. Another potentialdrawback specific to speech corpora isthat they lack linguistic cues that areotherwise available in a text corpus.Speech corpora include both frag-mented, incomplete words and filledpauses, such as ‘uh’ and ‘um’. Whilethese elements reflect how peopleactually speak, which is crucial forspeech recognition systems, theyunderstandably make it harder towork with the data.

The concept of a corpus presentsan interesting circular relationshipbetween linguistics and society in gen-eral. A corpus captures a subset ofreal-world language data, while lin-guists use this data in a format thatenables them to study it, to analyse itscontents, and to develop and support

speech corpora data as well as testingsystems against an existing corpora ofnatural, unscripted speech. It mayeven play a role in assembling a con-trolled language to be used for aspeech recognition system, identifyingthe words and phrase structures thatregular people may use when interact-ing with the system. Machine transla-tion (MT) technology also usescorpora for example-based systemsworking on the assumption that trans-lation involves finding or recallinganalogous examples of words andphrases. Corpus-based approaches toMT may be trained using a subset oftexts within a particular corpus andthen tested against other texts withinthat same corpus. Finally, corpora arealso useful for exploring and inter-preting historical documents. Forexample, researchers working ondeciphering an ancient text could usea corpus of texts from the same his-torical period and geographical loca-tion to find clues to better understandthe text they are studying.

A key criticism of corpus analysis isthat it supports statistical methodsthat disregard our knowledge of lan-guage. Within NLP, systems fall alonga spectrum, with purely rule-basedapproaches at one end and corpus-based (statistical) approaches at theother. Evaluations of competing NLPsystems show that hybrid systems,incorporating a combination of lan-guage rules and statistical methods,consistently outperform systems thatuse only one of the two methodolo-gies. Even when a corpus-based (orhybrid) approach is undisputed, a par-ticular corpus itself might not fit thetask. A corpus fails its job when it isnot an accurate representative sample

41

CORPORA

Knowing the importance of size, linguists are using technologicaladvancements to make increasinglylarge corpora to better represent thereal world. They also take advantageof statistically-based techniques forautomatically deriving probabilitiesconcerning word counts, frequenciesand concordance. As long as linguistscontinue to propose theories, they willneed corpora to support them, and aslong as they continue to design andbuild NLP systems, they will need cor-pora to test them.

Primary sourcesGodfrey, John, Edward Holliman and

Jane McDaniel (1992). ‘SWITCH-BOARD: Telephone speech corpus forresearch and development’. Proceedingsof the IEEE Conference on Acoustics,Speech, and Signal Processing. SanFrancisco: IEEE.

Koehn, Philipp (2005). ‘Europarl: A Paral-lel Corpus for Statistical Machine Trans-lation.’ MT Summit 2005. Unpublished.

Kucera, Henry and Nelson Francis (1967).Computational Analysis of Present-DayAmerican English. Providence, RI:Brown University Press.

Marcus, Mitchell, Beatrice Santorini andMary Ann Marcinkiewicz (1993).‘Building a large annotated corpus ofEnglish: the Penn Treebank’. Computa-tional Linguistics Vol. 19: 313–30.

Quirk, Randolf (1960). ‘Towards adescription of English usage’. Transac-tions of the Philological Society 40–61.

Further readingBiber, Douglas, Susan Conrad and Randi

Reppen (1998). Corpus Linguistics:Investigating Language Structure andUse. Cambridge: Cambridge UniversityPress.

theories or NLP systems. These NLPsystems, developed and tested againsttheoretically real-world corpora, inturn can help their real-world users tocarry out tasks. For example, a speechcorpus, like the SWITCHBOARDcorpus or CALLHOME corpus dis-cussed above, originates from real-world speech data collected viaunscripted telephone conversations.Linguists may use such a corpus tolearn about how people communicateover the telephone and therebydevelop a speech recognition systemthat can function properly with prac-tical, real-world speech data. Thequality of the final product is a directreflection of the quality of the speechcorpus by which it was trained andevaluated. A poor-quality corpuswould result in a speech recognitionsystem that proves to be an inade-quate fit for its users’ needs. Likewise,a high-quality corpus would result ina speech recognition system thatbetter anticipates the types of utter-ances its users may give it as input. Acorpus aims to reflect real-world lan-guage so as to support theories andsystems that, too, reflect real-worldlanguage.

A single corpus cannot be a perfectfit for every linguistic endeavour, norcan it exhibit all aspects of languageuse. This is why there are so manyexisting corpora available to linguists.Nevertheless, linguists often createnew corpora tailored to a particularlanguage use scenario, geographicalregion, demographic group or format.Corpora will continue to play animportant role in linguistics and thestudy of language, and automatedmethods will improve our ability tocollect, format and analyse their data.

42

CORRESPONDENCE THEORY

phy, and its contemporary version ismostly influenced by G. E. Moore,Bertrand Russell, Ludwig Wittgen-stein and Alfred Tarski. For manypeople outside philosophy, includinglinguists and lay-persons, the cor -respondence solution seems intu-itively correct: in our daily lives wehave to assume that there is a relationbetween what people say and thethings they talk about. But whetherwhat they say is true or false, we needthe concept of truth because withoutit we would not be able to describewhat happens in the world.

One modern way of formulating thecorrespondence theory solution is tosay that a sentence (a proposition) istrue iff (if and only if) it corresponds tosome fact, or iff it corresponds to somestate of affairs. It is not hard to see theproblem: in order to appreciate thenotion of correspondence one has tograsp the notions of ‘sentence’, ‘fact’and ‘state of affairs’ respectively. Intu-itively we would say that the truth of asentence has to do with the meaning ofthe sentence, and this may be differen-tiated from the state of affairs, whichseems to be what is out there, in reality. But can ‘sentence meaning’ beseparated from ‘fact’? Some sentences– for example, ‘Open the window’ –do not express facts, but in generalsentences are interpreted as expressingfacts of some kind. So, if a sentencemeaning conveys true information,what is the difference between themeaning of a sentence and the factsthat it conveys? In other words, if ‘sen-tence meaning’ and ‘fact’ are the samething, how can they not correspond?

Dictionary definitions on the notionof truth usually take up the nature ofthe relation between language and

Edwards, Jane (1993). ‘Survey of elec-tronic corpora and related resources forlanguage researchers’. In J. Edwardsand M. Lampert (eds), Talking Data:Transcription and Coding in DiscourseResearch. Hillside, NJ: Lawrence Erl-baum Associates.

Morley, Barry (2006). ‘WebCorp: a tool foronline linguistic information retrievaland analysis’. In A. Renouf and A. Kehoe(eds), The Changing Face of Corpus Lin-guistics. Amsterdam: Rodopi.

Jennifer A. Baldwin

CORRESPONDENCETHEORY

A theory that tries to solve the prob-lem of what constitutes truth as aproperty of sentences by claiming thatthere is a relation of correspondencebetween the meaning of true sentencesand the way the world is at a certainplace and time.

See also: Language of Thought;Logic; Sense/Reference; TruthTheories; Truth ValueKey Thinkers: Aristotle; Kant,Immanuel; Moore, G. E.; Russell,Bertrand; Tarski, Alfred;Wittgenstein, Ludwig

Correspondence theory is the mostprominent of several theories trying tosolve the problem of what truth is. Itoriginates in Greek philosophy, specif-ically in the work of Aristotle, andappears in the medieval period inthe work of Thomas Aquinas andWilliam of Ockham. It was later aprinciple in Immanuel Kant’s philoso-

43

CREATIVITY

Semantics, Metamathematics. Indi-anapolis: Hackett. 152–278.

Wittgenstein, Ludwig (1922). TractatusLogico-Philosophicus. London: Rout-ledge.

Hans Götzsche

CREATIVITY

The ability of the ordinary use of lan-guage to be innovative and free fromstimulus control. As associated mostrecently with Noam Chomsky(though with ultimate roots in Carte-sian rationalism), it is claimed thatthis creativity points to the centralplace of language in the study ofhuman nature.

See also: Behaviourism; Metaphor;Universal GrammarKey Thinkers: Chomsky, Noam;Descartes, René; Humboldt,Wilhelm von; Skinner, B. F.

In one sense, linguistic creativity has anarrow meaning, referring to the abil-ity of human beings to innovate withinthe lexico-semantic domain. Speakers,writers and poets can use the elementsof their language to draw attentionto surprising and interesting aspectsof the world through metaphor*,jokes and the like. However, there isanother, broader sense of linguisticcreativity, most commonly attributedin the modern era to Chomsky, whichapplies not to the activities of giftedindividuals, but rather to the ordinaryuse of language by everyday speakers.This creativity underlines the funda-mental role that language plays inunderstanding the essential nature of

reality. Some term for the concept ofrelation seems inescapable and so it isin modern epistemology and philoso-phy of science. Thus the standard defi-nition of knowledge is ‘justified truebelief’ (Dancy 1985: 23), meaningthat there must be a relation betweenbeliefs and what the beliefs are about.If scientific knowledge is to count astrustworthy, it must be true (seePopper 1963: 215–50); there must bean agreement between what sciencesays and how things happen out there.

Kant expressed the correspondenceprinciple in the most transparent way:‘Wahrheit ist die Übereinstimmung derErkenntnis mit ihrem Gegenstand’,‘truth is the correspondence of knowl-edge with its object’ (Kant 1781/1787,1924/1927: 992), even though he didnot consider it a problem, while Tarskioffered a technical and rather com -plicated clarification of the corres- pondence relation based on predicatelogic*. Although it is commonlyaccepted that some idea of correspon-dence is a necessary requirement intruth theory*, it is also commonlyrecognised that the nature of this rela-tion is the source of much controversy.

Primary sourcesKant, Immanuel (1781[A]/1787[B]. Pub-

lished by Raymund Schmidt 1924/1927). Kritik der reinen Vernunft. Wies-baden: VAM-Verlag.

Popper, Karl R. (1963/1989). Conjecturesand Refutations. Fifth edition. London:Routledge.

Further readingDancy, Jonathan (1985). Contemporary

Epistemology. Oxford: Blackwell.Tarski, Alfred (1935). ‘The concept of

truth in formalized language’, in Logic,

44

DECONSTRUCTION

explanation. Therefore, according tothe Cartesians, human linguistic capac-ities provided evidence for the exis-tence of a second substance ‘mind’ (asdistinct from mere ‘body’). Althoughmetaphysical dualism no longer formspart of standard scientific assumptions,these observations regarding linguisticcreativity nonetheless do suggest that,in studying language, we are studyingone of the foundations on which ourhumanity rests.

Primary sourcesChomsky, Noam (1959). ‘Review of

Verbal Behavior by B. F. Skinner’. Lan-guage 35: 26–58.

Chomsky, Noam (1966). Cartesian Lin-guistics: A Chapter in the History ofRationalist Thought. New York:Harper & Row. Reprinted (2002)Christchurch, New Zealand: Cyberedi-tions Corporation.

Geoffrey Poole

DECONSTRUCTION

Deconstruction is an approach todoing philosophy that subverts justabout everything mainstream philoso-phy has stood for through the cen-turies. It is associated with the nameof the controversial Algerian-bornFrench philosopher Jacques Derridawho was strongly influenced by,among others, Friedrich Nietzscheand Martin Heidegger.

See also: Analytic Philosophy;Poststructuralism; Structuralism;Truth TheoriesKey Thinkers: Derrida, Jacques

human beings and is often referred toby Chomsky as ‘the creative aspect oflanguage use’.

The most general aspect of creativ-ity in this broad sense, and one whichconnects Chomsky’s work to the ear-lier rationalist traditions of Wilhelmvon Humboldt and René Descartes isthe observation that human beingscan produce (and understand with nosense of novelty) an infinite numberof sentences which may be new intheir experience or even new inthe history of the language. This cre-ativity is thus intimately connectedwith the ‘generative’ nature of gener-ative grammar, and forms part of thekey motivation for universal gram-mar*.

As noted in some detail by Chom-sky (1959), human language use isalso creative in the sense of being freefrom identifiable stimulus control. It isappropriate to a situation, but it isnot caused by it. A traditional behaviourist account of the response‘Dutch’ to a painting on the wallwould be to say that the speaker wasunder the stimulus control of thepainting’s ‘Dutchness’. However, asChomsky observes, a speaker couldhave just as easily responded ‘It’stilted’, ‘I thought you liked abstracts’or an infinite number of other things.

Chomsky (1966) notes that this free-dom from stimulus control was one ofthe Cartesian arguments for the exis-tence of mind. Machines, once theinternal arrangement of the partsand the external conditions are specified, behave in a completely pre-dictable man ner (or randomly). How-ever, human linguistic behaviour, likehuman thought and action more gener-ally, transcends simple ‘mechanical’

45

DECONSTRUCTION

There is a grain of truth in theclaim that deconstruction engages intextual analysis. But it is important toadd that, as Derrida himself onceremarked, there is nothing outside thetext. What Derrida was trying to presshome in his trademark claim was that,contrary to conventional thinking onthe matter, the reading of a text doesnot consist in pitting the text againstsomething that is essentially of someother order – say, ideas or intentionsin the mind of the speaker/writer or aputative reality ‘out there’. Instead, areading of the text is an extension ofthat very text. That is to say, no textcomes out unaffected by its successivereadings. Each new reading of the textadds something to that text and isfully incorporated into the text’s‘meaning’ so that the meaning of thetext – to the extent there is such athing – may be described as that whichis constantly undergoing change,despite the illusory sensation that it isthe self-same object that one is dealingwith on successive occasions.

So what on earth does Derridamean by deconstruction and exactlyhow does it work? It has beenremarked that it is easier to ‘define’deconstruction by saying what it isnot, rather than what it is. In a famoustext called ‘Letter to a JapaneseFriend’, Derrida wrote that ‘decon-struction is neither an analysis, nor acritique’ and, furthermore, it is ‘not amethod and cannot be transformedinto one’. In fact, as it turns out, thevery quest for definitions is antitheti-cal to the spirit of deconstruction.This is so because the idea that mean-ings can be captured and encapsulatedin neat definitions is the very hallmarkof philosophy in its traditional sense.

The very mention of the word ‘decon-struction’ causes jitters among manyacademic philosophers and calls forthderision from others. It is probablytrue to say that no other term in phi-losophy has recently been the target ofsuch tireless tirade and misrepresenta-tion. In common parlance, it is oftenerroneously and dismissively used asthough it were just a fancy word fordestruction – thus proving the oldadage that one can give a dog a badname and hang it. When Jacques Der-rida first employed the word, scarcelycould he have imagined that it wouldsoon be at the epicentre of so muchcontroversy and acrimony. From the1960s on, when Derrida took theworld of philosophy by storm, decon-struction became a catchword amongacademics across a wide spectrum ofdisciplines, and his influence is felt inareas as diverse as literary criticism,linguistics, sociology, and even suchunlikely disciplines as economics andlaw.

Both Derrida and the movement hestarted have understandably been thetarget of scathing criticism, especiallyfrom those scholars who, speakingon behalf of the philosophical ‘estab-lishment’, rightly sensed a potentialthreat to the way they went abouttheir routine business. However, tocall the approach sceptical or down-right nihilistic, or to characterise it asjust plain philosophical dilettantism,as some critics have done, is to dismissit without a fair hearing. To regarddeconstruction as a form of textualanalysis is to give it a left-handed com-pliment and, in effect to relegate it tothe realm of literary criticism –another way of saying it is anythingbut philosophical.

46

DECONSTRUCTION

ness of the reader to tease it out of spe-cific texts.

The process of unravelling a text’sputative meaning begins the momentone notices that the dichotomies thata text’s author posits in order to con-struct his arguments are actually rela-tions of hierarchies, often ingeniouslycamouflaged in order to be presentedas symmetric. In other words, it isinvariably the case that, in a dichoto-mous pair, one side is privileged to thedetriment of the other. In his book OfGrammatology (1967) Derrida illus-trated this by showing how one of thefounding dichotomies of linguistics –namely, the one that opposes meaningand the linguistic object (say, thetext)–is itself the result of a ‘textualprestidigitation’ at the hands of Ferdi-nand de Saussure, the discipline’sfounding father, who postulated it.Initially presented as a symmetric pair,the distinction between signified andsignifier reveals itself, upon closerinspection, to be an hierarchical one.This is because the very stability ofthat sign relation (the sign being, forSaussure, a relation of bond betweena signifier and a signified) is predi-cated upon there being one side of therelation (namely, the signified) notdependent on the other and thus capa-ble of ‘standing alone’ – unlike the sig-nifier whose contribution to the signrelation is always that of ‘pointing to’the other, namely the signified, con-ceived of as ‘auto-sufficient’ in that itis the meaning, period.

Derrida extended his deconstructiveanalysis to other prized dichotomies ofstructural linguistics such as speechversus writing. He argued that the lin-guists’ preference for speech to thedetriment of writing (the ‘wandering

Recall, for instance, Socrates’s irritat-ing habit of asking his interlocutors toprovide him with a definition of, say,‘piety’ (as repeatedly shown in Plato’sDialogues). Therefore, to the extentthat deconstruction seeks to, as itwere, ‘call the bluff’ of philosophy inrespect of its time-honoured preten-sions, it is only to be expected that itshould stubbornly resist every attemptto enclose it in a neat definition.

The idea that one can define one’sterms once and for all rests on theassumption that there are such thingsas meanings that can remain stableover a period of time and which couldbe captured and ‘imprisoned’ in theform of rigorous definitions. Well,among other things, deconstructionis, as we have seen, concerned withreminding us that there are no suchstable meanings to begin with, and sothe quest for definitions is a wild goosechase. Rather, the meaning of a giventext is precisely what is iterable – aneologism coined by Derrida, meaningthat which comes out different everytime one repeats it, as bizarre as thismight indeed appear at first blush. It is,to use another of Derrida’s coinages,determined by différance, an amalgamof deferral and difference.

Derrida maintained throughoutthat deconstruction is a form of closereading, where all that one needs tomobilise is the text’s own internallogic. In other words, texts decon-struct themselves as it were when sub-jected to relentless close reading. Allthat a reader needs to do is to pressahead with its own internal logic.That logic, Derrida insisted, is builtaround the notion of logocentrism,the belief that there are stable mean-ings out there and that it is the busi-

47

DEDUCTION/INDUCTION

relentlessly and with a rigour it isnot used to, will ultimately turnagainst itself. Derridean philosophy ofanti-philosophy has impacted such movements as poststructuralism* andpostmodernism. But it is important toregister that it has, so to speak, been‘hijacked’ by both the political left andthe right to serve their own respectiveagendas.

Primary sourcesCuller, Jonathan (1982). On Deconstruc-

tion: Theory and Criticism after Struc-turalism. Ithaca, NY: Cornell UniversityPress.

Derrida, Jacques (1967). Of Grammatol-ogy. Trans. Gayatri Spivak. Baltimore:The Johns Hopkins University Press.

Derrida, Jacques (1991). A DerridaReader. Between the Blinds. Ed. PeggyKamuf. Hertfordshire: HarvesterWheatsheaf.

Norris, Christopher (1982). Deconstruc-tion: Theory and Practice. London:Methuen.

Further readingWood, David (ed.) (1992). Derrida: A

Critical Reader. Oxford: Blackwell.

Kanavillil Rajagopalan

DEDUCTION/INDUCTION

Deduction is a form of reasoning inwhich one proceeds from general prin-ciples or laws to specific cases. Induc-tion is a form of reasoning in whichone arrives at general principles orlaws by generalising over specific cases.

outcast of linguistics’) is yet anotherdemonstration of discipline’s complic-ity with Western philosophy’s trade-mark logocentrism, which in the caseof linguistics takes the form of ‘phono-centrism’. Using a very subtle line ofreasoning, Derrida maintained thatthe argument frequently used to bol-ster up the case of speech as the privileged object of scientific linguis-tics – its spontaneity and authenticity– crumbles as soon as one realises thatit is based on the implicit idea thatspeech is closer to the speaker’sintended meaning than writing is.Aristotle condemned writing for beingthe imitation of what is already an imi-tation. This means, Derrida insisted,that speech is always already beingconceived of along very much thesame lines as writing (namely, the rep-resentation of something else), so thatthe very distinction between speechand writing may be said to be based onan unacknowledged notion of ‘arch-writing’. Furthermore, the so-calledintended meanings are themselvesbeing treated as though they wereinscribed in the speaker’s mind/brain,ready to be scrutinised any timethanks to what is credited with beingthe distinguishing feature of writing,its permanence over a period of time.

Using analogous reasoning, Derridaargued that such hallowed dichotomiesas nature versus culture, philosophyversus literature, science versus myth,reason and unreason, and so forth – thebedrock of much of Western thought –would not survive deconstructivescrutiny. This has, as he rightly fore-saw, devastating implications for thevery enterprise of philosophy.

Deconstruction is a constant re -minder that reason, when pursued

48

DEDUCTION/INDUCTION

that these latter types of propositionare less certain. He claimed that induc-tive generalisations are not well-founded in logic*: from the fact thatthe sun has come up on all the preced-ing days of our lives, it does not followlogically that it will come up tomor-row, or, more generally, that there is avalid general law which states that itcomes up every day. A clear example ofthe uncertain nature of inductive gen-eralisations is the case of the generali-sation that all swans are white. Fromthe fact that all swans one has previ-ously observed were white, it does notfollow that the next swan one willobserve will be white. The discovery ofblack swans in Australia showed thatthis particular inductive generalisa-tion, once held to be true, was in factfalse.

These features of inductive general-isations make inductive reasoning different in kind from deductive rea-soning, since deduction is founded inlogic: in the syllogism ‘All men aremortal; Peter is a man; therefore Peteris mortal’, the conclusion that Peter ismortal follows logically as a deduc-tion from the premises. Whether thepremises are well-founded is anothermatter: the argument is logicallysound. The deduction in the follow-ing argument is equally well-founded:(a) ‘All rabbits are carnivores’,(b) ‘Rupert is a rabbit’, therefore(c) ‘Rupert is a carnivore’. In statingthat inductive generalisations have nobasis in logic, one need not deny thatsuch generalisations are well-foundedpsychologically, and Hume’s theory ofinduction was indeed a psychologicaltheory: he argued that repeated obser-vation of the ‘same’ event causes us toarrive at inductive generalisations. He

See also: Empiricism/Rationalism;Metaphor; Propositions; Type/Token; Universal GrammarKey Thinkers: Chomsky, Noam;Hume, David.; Locke, John; Mill,J. S.; Peirce, C. S.; Popper, Karl;Russell, Bertrand; Sapir, Edward;Whorf, Benjamin Lee

The status and role of these two formsof reasoning/inference in humanthought and language have been muchdiscussed in the history of both philos-ophy and linguistics. No-one doubtsthat human beings are capable of bothkinds of reasoning, but there has beenmuch debate as to exactly what rolethey play in establishing humanknowledge, including linguistic knowl-edge. Central to these debates has beenthe role of inductive generalisations. Asimple example of an inductive gener-alisation is the generalisation that thesun comes up every day. We arrive atsuch generalisations by observing spe-cific events, such as the sun coming upon a specific day, and then generalisingover other events which are thoughtby the observer to count as instances ofthe same type of event, for example,the sun coming up on subsequent days.

The eighteenth-century philosopherDavid Hume distinguished between (a)propositions concerning states ofaffairs that are certain (such as theproposition that 2 + 2 = 4), and (b)propositions which are unlike suchmathematical propositions in thatthey rely on observed objects andevents. The latter sorts of propositioninclude propositions based on induc-tive generalisations arising from theperception of those objects and events(such as the proposition that the suncomes up every day). Hume argued

49

DEDUCTION/INDUCTION

is functioning on the basis of cate-gories, and these are said by many toplay a central role in human percep-tion, including speech perception.Central to the formation of inductivegeneralisations and the formation ofcategories is the notion of similarity:in order to take two events or objectsto be instances of the same thing, weneed to be able to perceive similaritybetween those instances. That capac-ity to perceive similarities may well beinnate, and is arguably at the heart ofmuch of human cognition. It appearsto underlie our capacity to establishsimiles and metaphors* which lin-guists such as George Lakoff take tolie at the heart of human language.

There has been much discussion ofthe role of deduction and induction inthe field of child language acquisition.Chomsky’s approach to the child’s lin-guistic development is founded onhis linguistic rationalism (otherwiseknown as linguistic nativism): he pos-tulates innate linguistic knowledge,often referred to as an innate languagemodule or faculty, and also referred toby many as universal grammar*. Thechild is said by Chomsky to be bornwith an innate set of universal linguis-tic principles which are not, by defini-tion, acquired by the child in his/herinteraction with the mind-externalworld. The input (otherwise known as‘the stimulus’) to which the child isexposed is said by Chomsky and hisfollowers to be ‘impoverished’ in cer-tain ways: it is said to be full of hesitations, false starts, utterances ofsentences which are never completed,and so on. Chomsky argues that,given the sheer complexity of adult linguistic knowledge, it could nothave been acquired on the basis of this

took that to be a central feature ofhuman psychology.

Hume’s theory of induction waslater re-examined in the twentiethcentury by the philosopher KarlPopper (1963 and elsewhere), whoargued that no two (or more) events(say, the observation of a flash of light-ning) will ever be exactly the same,and that we therefore need to appealto the idea of two (or more) eventscounting as ‘the-same-for-us’ (orindeed, the same for a member ofanother species, which may not bequite the same thing as it is for ourspecies). Popper argued that humanbeings are born with an innatepropensity to seek out regularities,and that it is this propensity thatunderlies our capacity to form induc-tive generalisations. This is similar, ifnot identical, to Bertrand Russell’sclaim that induction is ‘incapable ofbeing inferred from experience orfrom other logical principles’ (1946:647). Popper’s claim is not to be con-fused with the doctrine of innateideas, associated with the work of,among others, René Descartes in theseventeenth century and NoamChomsky in the twentieth century.

The propensity to seek out regular-ities is, arguably, intimately connectedwith our capacity to form categories,and our capacity to take a specificevent on a specific occasion to be atoken of a type, in the sense invokedby the twentieth-century Americanphilosopher Charles Saunders Peirce.If one takes, say, a specific flash oflightning to be a token of the type‘flash of lightning’, or a specific speechsound token (say, a [t] uttered on aspecific occasion) to be a token of thetype ‘[t]’, then one’s perceptual system

50

DEDUCTION/INDUCTION

the course of that development. SinceChomsky adopts a version of natural-ism, in which a postulated biologi-cally-endowed module of mind, onexposure to the input, undergoes bio-logical growth, linguistic developmentis said not to be something that thechild does; rather, it is something thathappens to the child. It is important tonote that this is an entirely passiveconception of the child’s linguisticdevelopment. The importance of thislies in the fact that rationalists oftendepict empiricist approaches to childlinguistic development as a passiveconception of first language acquisi-tion, a depiction that is at odds withmuch present-day empiricism. Notetoo that appeal by rationalists to theact of setting parameters and engagingin deduction runs counter to the claimthat the acquisition process does notamount to the child actively doinganything to acquire a first language.

Those who oppose Chomskyanrationalism typically emphasise therole of inductive and analogical gen-eralisation in child language acquisi-tion. Examples of such generalisationsare child expressions such as ‘Threesheeps comed’ (‘Three sheep came’),in which the irregular plural ‘sheep’and the irregular past tense ‘came’ areregularised. This suggests that thechild has arrived at the regular rulesfor plural and past-tense formationin English via induction based onrepeated exposure to regular forms,followed by analogical extension ofthe inductive rule to irregular nounsand verbs. The emphasis on analogyand induction is central to the work oflinguists such as Joan Bybee, whoseek to re-establish the importance ofusage (performance, in Chomskyan

‘impoverished’ stimulus alone. Putanother way, adult linguistic knowl-edge is underdetermined by the input.This is the poverty of the stimulusargument for an innate linguisticmodule of mind. Chomsky has alsoappealed to what he calls the logicalproblem of language acquisition. Theproblem is said to be based on the dif-ficulty of arriving at fully-fledged adultlinguistic knowledge on the basis ofthe supposedly impoverished input/stimulus. The problem is said to be alogical one since the child is born inpossession of knowledge of the natureof language, in the form of general lin-guistic principles and a set of parame-ters, and is said to deduce theproperties of the ambient language. Anexample of a parameter is the positionof the verb in sentence structure. Thechild is said by Chomskyans to deducethe parameter settings in universalgrammar, thus yielding the specific lan-guage he/she is being exposed to.

Deduction is thus central to Chom-sky’s vision of child linguistic develop-ment. Chomskyans, such as Smith(2004), argue that general learningmechanisms play a very limited role inthat development. The capacity forinductive generalisation forms part ofour general learning mechanisms: wecan form inductive generalisations inany cognitive domain, from knowingabout the seasons to knowing whatthe traffic might be like at a givenpoint in the day. Because Chomskyansdo not take the child’s linguistic development to constitute a kind oflearning, they play down the role ofinductive generalisation in child lin-guistic development. Equally, theyplay down the role of our capacity forforming analogical generalisations in

51

DEDUCTION/INDUCTION

he/she must acquire knowledge of aset of social conventions, includinglinguistic conventions (such as know-ing that the subject precedes the verbin English sentences, or knowing thatregular plurals of nouns consist of thesingular form with a specific kind ofsuffix). The child can acquire theseconventions by observing the conven-tional word orders used by otherhuman beings in his/her linguisticenvironment. Crucial to Tomasello’sview of child language acquisition isthe social, interactive nature of thechild’s world, in particular the child’scoming to attribute intentions to thepeople that he/she is interacting with,along with the child’s ability to haveintentions towards those intentions.

Not all inductive reasoning is con-scious. While adults and children alikecan engage in explicit learning, involv-ing conscious reasoning, there is aform of implicit learning which evenvery young infants (in the first year oflife) can engage in. This kind of learn-ing rests on the human capacity toextract probabilities from speechinput, such as probabilities aboutlikely sequences of words and speechsounds. For instance, children in thefirst year of life have been shown toprefer made-up words which conformto the phonotactic constraints of thelanguage they are being exposed to.By ‘phonotactic constraints’ is meantthe constraints on sequences ofsounds in the syllable structure ofwords. The sequences /pr/, /pl/, /tr/,/kr/, /kl/ are all permissible sequencesat the beginning of words in English,but the sequences /pn/, /ps/, /tl/ andmany others are not. Infants in thefirst year of life prefer made-up wordswhich contain the permissible, rather

terms) in understanding the nature oflinguistic knowledge. Bybee (2001)argues that frequency of occurrence ofspecific words in language usage playsa major role in the synchronic state ofa given language, and in its histori-cal development. She distinguishesbetween token frequency, which is theextent to which specific word-formsare uttered, and type frequency (seebelow). She argues that words withhigh token frequency, such as the past-tense form ‘kept’ in English, are moreresistant than words with lower tokenfrequency (such as the verb ‘weep’) tothe kind of analogical changes thatwould yield regularised past-tenseforms such as ‘keeped’: lower-tokenfrequency ‘weep’ is likely to regu-larise, via analogy, to ‘weeped’,whereas higher-token frequency‘keep’ is much less likely to be regu-larised, via analogy, to ‘keeped’. Typefrequency is defined as the frequencyof occurrence of a specific pattern,such as the English past-tense ‘-ed’pattern, which applies to the vastmajority of English verbs. High typefrequency is said to determine the pro-ductivity of a specific pattern. In thiscase, the ‘-ed’ pattern is highly likelyto be applied to novel word forma-tions (such as the past-tense verb form‘googled’) and to borrowed verbs(such as ‘nuanced’). Type frequency,like token frequency, is based oninductive and analogical inferences.

Some twenty-first-century child lan-guage researchers, such as MichaelTomasello, argue that induction playsa central role in child language acqui-sition. For Tomasello (2001), the childpossesses no innate linguistic knowl-edge at birth. Rather, the child is saidto be born into a world in which

52

DEDUCTION/INDUCTION

philosophy of science form the basisfor Sampson’s empiricism. Interest-ingly, Sampson and Chomsky bothstress the role of deduction in thechild’s linguistic development, but forvery different reasons, and in differentways. Note that earlier claims byChomskyans that the child is a littlescientist, whose hypotheses are con-strained by a postulated innate lan-guage acquisition device, run counterto the present-day view of Chomskythat the child’s linguistic developmentis entirely passive.

Appeal to the hypothetico- deductive method in the child’s con-ceptual and linguistic developmentcan be found in current work byAlison Gopnik. She proposes a‘theory theory’ of child development:the child is said to be actively con-structing theories about the surround-ing world, including the ambientlanguage. Gopnik (2001) reports onexperiments which seem to show thatcertain stages in the child’s conceptualdevelopment, such as the capacity tograsp the idea of a means towards anend, come on stream just ahead of therelevant child’s expressions denotingsuch notions. This work seems to sug-gest that, for such stages in concep-tual development, it is the conceptwhich comes first, followed by anutterance-type for that concept.Gopnik also argues for a version ofthe Sapir-Whorf hypothesis, accord-ing to which aspects of the structureof the ambient language will inducethe child to conceptualise certainkinds of events and objects in specificways. For instance, in a languagesuch as Korean, ellipsis of nouns ismuch more common than in Eng-lish. Because of this, a child of fifteen

than the non-permissible, sequences.This is because they have the capacityto tune in to the high frequency of thepermissible sequences and to extractthe phonotactic patterns. Infants arecapable of this kind of probabilistic(otherwise known as stochastic)learn ing without conscious effort. Theexistence of this kind of unconsciousinductive learning has been used byopponents of Chomskyan nativism toshow that the stimulus which the childis exposed to is rich in information,which the child can access, rather thanimpoverished, as claimed by support-ers of linguistic nativism.

One of Chomsky’s fiercest critics,the British linguist Geoffrey Sampson,adopts an empiricist approach tochild language acquisition. Sampson,like Tomasello, and unlike Chomsky,argues that children learn language,and that they do so in the same way asthey learn anything else: using generallearning mechanisms. The most cen-tral part of child language acquisition,for Sampson, is the hypothetico-deductive method: the child is said tocome up with hypotheses as to thenature of the language he/she is beingexposed to, and then uses the capacityfor deduction to deduce testableclaims about the structure of that lan-guage. These are then said to be testedagainst the data the child is exposedto. The child can then modify or aban-don his/her hypotheses about thestructure of that language, and thuscome to learn the language. This is theview that the child is ‘a little scientist’in his/her attempts at coming to graspthe structure of his/her native lan-guage. This emphasis on the hypo-thetico-deductive method is central tothe work of Popper, whose ideas in the

53

DEDUCTION/INDUCTION

remains a central topic of researchand debate in contemporary languagestudy.

Primary sourcesBybee, Joan (2001). Phonology and Lan-

guage Use. Cambridge: Cambridge Uni-versity Press.

Chomsky, Noam (2000). New Horizons inthe Study of Language and Mind. Cam-bridge: Cambridge University Press.

Gopnik, Alison (2001). ‘Theories, lan-guage and culture: Whorf without winc-ing’. In M. Bowerman and S. Levinson(eds), Language Acquisition and Con-ceptual Development. Cambridge:Cambridge University Press.

Hume, D. (1748). An Enquiry ConcerningHuman Understanding. Reprinted in A.Flew (ed.) (1962), David Hume OnHuman Nature and The Understand-ing. London: Collier Macmillan.

Karmiloff-Smith, Annette (1998). ‘Devel-opment itself is the key to understand-ing developmental disorders’. In Trendsin Cognitive Sciences 2. Reprinted inM. Tomasello and E. Bates (eds)(2001), Language Development: TheEssential Readings. Oxford: Blackwell,331–50.

Popper, Karl R. (1963). Conjectures andRefutation. London. Routledge andKegan Paul.

Russell, Bertrand (1946). The History ofWestern Philosophy. London: Rout-ledge (2004).

Sampson, Geoffrey (1997). EducatingEve: The ‘Language Instinct’ Debate’.London: Cassell.

Sapir, Edward (1921). Language: AnIntroduction to the Study of Speech.New York: Harcourt Brace.

Tomasello, Michael (2001). ‘Perceivingintentions and learning words in thesecond year of life’. In M. Tomasello and

to twenty-one months acquiringKorean monolingually will be delayedin the ‘naming explosion’ stage, andon categorisation tasks, in compari-son with a child of the same ageacquiring English monolingually. Butthese Korean-acquiring infants areadvanced on means-ends abilities incomparison with English-acquiringinfants of the same age. Gopnikclaims that these differences in devel-opment are the result of the childusing the hypothetico-deductiveprocess to arrive at conclusions aboutthe shape of the ambient languageand the nature of the world theyinhabit.

Current work by AnnetteKarmiloff-Smith (1998), who worksin the tradition established by JeanPiaget, adopts a version of construc-tivism with respect to child develop-ment. She argues that there is noinnate language module, but theremay well be innate biases in differentcognitive domains (such as the recog-nition of familiar faces). These biasesare said to develop with training onthe environment: the child’s percep-tual input is said to shape those biasesinto domain-specific cognitive capaci-ties which appear modular in nature(such as the face recognition module).Inductive generalisations play a role inthis approach to child development,since the child’s initial biases are saidto become increasingly richly devel-oped as a result of training on specificsorts of input, where repeated expo-sure to specific kinds of sensory eventresults in the forming of inductive gen-eralisations in that domain.

The issue of the respective roles ofinduction and deduction in the construction of linguistic knowledge

54

DEFINITE DESCRIPTIONS

Key Thinkers: Frege, Gottlob;Russell, Bertrand; Strawson, P. F.

In 1905 Bertrand Russell published anarticle drawing attention to what hesaw as some specific and philosophi-cally interesting properties of definitedescriptions. Phrases introduced by‘the’ appear to indicate not only theexistence but also the uniqueness ofsome entity. In this respect they can becontrasted with indefinite descrip-tions; ‘the leader of the free world’conveys the idea that there is only onesuch person, while ‘a leader of the freeworld’ does not. According to Frege’saccount of meaning, definite descrip-tions should be treated as names*,having both a sense and a reference. Inopposition to this analysis, Russelldraws attention to the problem of def-inite descriptions that do not have anyactual reference in the world; ‘aphrase may be denoting, and yet notdenote anything’ (1905: 471).

Russell discusses this problem inrelation to his now-famous example‘the present king of France is bald’.According to Frege’s account, this sen-tence expresses a simple propositionof subject-predicate form. Since thesubject fails to refer, the proposition asa whole should fail to have a truthvalue*; it should be nonsense. Russellargues against this reading on thegrounds that ‘it is not nonsense, sinceit is plainly false’ (1905: 484). Thiscan be explained, he claims, once werealise that definite descriptions donot operate in the same way as names.In fact, it is not appropriate to dis-cuss their meaning in isolation; the meaning of a definite description canbe considered only in relation to the sentence in which it appears. The

E. Bates (eds), Language Development:The Essential Readings. Oxford: Black-well. 132–58.

Further readingPinker, S. (1994). The Language Instinct.

London: Penguin.Pullum, G. K. and B. Scholz (2002).

‘Empirical assessment of stimuluspoverty arguments’. The LinguisticReview 19: 9–50.

Sampson, G. (2002). ‘Exploring the rich-ness of the stimulus’. The LinguisticReview 19: 73–104.

Scholz, B. and G. K. Pullum (2002).‘Searching for arguments to support lin-guistic nativism’. The Linguistic Review19: 185–223.

Smith, Neil (2004). Chomsky: Ideas andIdeals. Second edition. Cambridge:Cambridge University Press.

Philip Carr

DEFINITEDESCRIPTIONS

Denoting phrases that are introducedby the definite article ‘the’, or by apossessive such as ‘my’ or ‘SherlockHolmes’s’. Bertrand Russell’s logicalanalysis of sentences containing definite descriptions has been highlyinfluential, and highly controversial,throughout the past century.

See also: Analytic Philosophy;Connotation/Denotation; Logic;Logical Form; Names; OrdinaryLanguage Philosophy;Presupposition; Propositions;Sense/Reference

55

DEFINITE DESCRIPTIONS

that where necessary the messy andimperfect constructions of naturallanguage must be ‘translated’ into log-ically correct form before they canprovide suitable subject matter forphilosophical analysis. This has beenthe cause of a number of responses toRussell. Perhaps the most significantof these has been that put forward byPeter Strawson in his 1950 article‘On referring’. Coming from a back-ground in ordinary language philoso-phy* (OLP), Strawson criticisesRussell for being obsessed with logicand mathematics, at the expense ofattending to the realities of naturallanguage. Russell was overlookingimportant facts about the ways inwhich speakers actually use expres-sions containing definite descriptions;as Strawson puts it, ‘“mentioning”, or“referring”, is not something anexpression does; it is something thatsomeone can use an expression to do’(1950: 326).

Strawson claims that someoneencountering an utterance of ‘thepresent king of France is bald’ wouldnot be likely to reply ‘that’s false’, asRussell’s account would seem to pre-dict. Rather they would be stuck for aresponse of any kind, feeling thatthere was something badly wrongwith this statement; the question ofwhether it is true or false would justnot arise. For Strawson, the proposi-tion that there exists a present king ofFrance is not part of the logical formof the statement, but is a presupposi-tion* attached to its use. It is neces-sary for this presupposition to befulfilled for the sentence to be eithertrue or false. Hence both ‘the presentking of France is bald’ and ‘the pres-ent king of France is not bald’ must be

presence of the definite descriptionensures that the logical form of thesentence is a complex set of proposi-tions that concern existence anduniqueness as well as baldness. Thelogical form of ‘the present king ofFrance is bald’ could be paraphrasedas ‘there is an entity x such that x is thepresent king of France, no entity thatis not equivalent to x is the presentking of France, and x is bald’, or moreinformally ‘there is one, and only one,king of France, and he is bald’.

This analysis allowed Russell toexplain his conviction that the sen-tence is false. The simple propositionthat a present king of France exists isfalse. This is sufficient to make thelogical form* of the sentence, which isa coordination of this and two otherpropositions, false also. The analysisalso allowed Russell to maintain aclassical bivalent logic* for language.If ‘the present king of France is bald’is false, then logic dictates that ‘thepresent king of France is not bald’must be true. This is indeed the case ifthe negation is read as having scopeover the whole of the logical form: asRussell paraphrases it, the negativesentence can be read as ‘it is false thatthere is an entity which is now king ofFrance and is bald’. The negative sen-tence is ambiguous, however. Onanother reading it can be interpretedas ‘there is an entity which is now kingof France and is not bald’, and on thisreading it is as false as its positiveequivalent.

Russell’s theory is very much aproduct of his analytic approach tophilosophy. It draws attention to whathe sees as a discrepancy between thegrammatical form of a sentence andits logical form. The implication is

56

DEFINITE DESCRIPTIONS

chairman of the teetotallers’ society isinformed that someone is drinking acocktail at their Christmas party, hemight ask ‘Who is the man drinking amartini?’ in order to establish theidentity of the culprit. In this case theattribute ascribed to the individual isall-important. If it turned out that theindividual in this case was drinkingonly water, then there is no possi-ble right answer to the chairman’s question. Donnellan also discussesexamples of definite descriptionsintroduced not by ‘the’, but by pos-sessives: examples such as ‘Smith’smurderer is insane’. He argues that,for different reasons, neither Russell’snor Strawson’s account can ade-quately handle the distinctionbetween referential and attributiveuses of definite descriptions.

Despite criticisms such as these,Russell’s theory of definite descrip-tions has been widely regarded as anexemplary application of analytic phi-losophy*. It is a measure of the con-tinuing importance of ‘On denoting’in the philosophy of language that aspecial edition of Mind was devotedto commemorating the centenary ofits publication. At the start of his contribution to this edition, StephenSchiffer comments that Russell’s article not only contributed to defin-ing a new philosophy of language, butalso remains the dominant theory ofdefinite descriptions in the presentday.

Primary sourcesDonnellan, Keith (1966). ‘Reference and

definite descriptions’. PhilosophicalReview 75: 281–304.

Frege, Gottlob (1892). ‘On sense andmeaning’. In Peter Geach and Max

categorised as neither true nor false.By allowing that some statements canlack a truth value* in certain contextsof use, Strawson was arguing that nat-ural language did not conform to theusual laws of logic; it should be stud-ied and analysed on its own terms.Russell strongly rejected this conclu-sion, arguing in a response to Straw-son that ‘common speech’ must bemodified before it is fit for philosophy.He points out that ‘my theory ofdescriptions was never intended asan analysis of the state of mind ofthose who utter sentences containingdescriptions’ (1957: 388).

Keith Donnellan (1966) also criti-cises Russell for being unable toaccount for the intentions and pur-poses with which people actually usedefinite descriptions in everyday lan-guage. He draws attention to a dis-tinction between two different ways inwhich definite descriptions are used,the referential and the attributive uses.Used referentially, definite descrip-tions pick out an individual in order tosay something about that individual.A person who uses a definite descrip-tion attributively ‘states somethingabout whoever or whatever is the so-and-so’ (1966: 285). If someone sees adistinguished-looking man drinkingfrom a martini glass at a party andasks ‘Who is the man drinking a mar-tini?’, that person is using the definitedescription referentially. In this casethe particular attribute ascribed to theindividual is not very important; thespeaker would still successfully havereferred to the distinguished-lookingman even if it turned out that he hadonly water in his glass. The same defi-nite description, in the same sentence,could also be used attributively. If the

57

DESCRIPTIVISM

characterising – not judging or criticis-ing – language produced within a par-ticular speech community.

See also: Langue/Parole;Structuralism;Transformational-GenerativeGrammarKey Thinkers: Bloomfield,Leonard; Boas, Franz; Brugmann,Karl; Chomsky, Noam; Halliday,M. A. K.; Hockett, Charles; Labov,William; Pike, Kenneth; Rask,Rasmus; Sapir, Edward; Saussure,Ferdinand de

Descriptivism became an importanttrend in linguistics after 1900. Its the-ories and principles support an openattitude toward language and linguis-tic study. It is opposed to prescrip-tivism, the term most often used torefer to a linguistic school of thoughtin which individuals seek to promoteone particular variety of a language,formulate its rules, and enforce adher-ence to those rules. Descriptive lin-guists strive to present a picture oflanguage as complete as possible as itactually exists at a specific point intime and place; they first describeobservable facts about a particularspoken and written language and notegeneralisations about that informa-tion. They then draw conclusionsabout that language and tie their conclusions to an analysis of humanlanguage in general.

Nineteenth-century European lin-guists such as Karl Brugmann andRasmus Rask moved the study of language away from ‘fanciful’ notionsof philosophy or folklore to a moredisciplined analysis of written lan-guage. Although these scholars con-centrated on diachronic studies of

Black (eds) (1980), Translations fromthe Philosophical Writings of GottlobFrege. Oxford: Blackwell. (First edition,1952.) 56–78.

Russell, Bertrand (1905). ‘On denoting’.Mind 14: 479–99.

Russell, Bertrand (1957). ‘Mr. Strawsonon referring’. Mind 66: 385–9.

Strawson, P.F. (1950). ‘On referring’.Mind 59: 320–44.

Further readingBezuidenhout, A. and M. Reimer (eds)

(2004). Descriptions: Semantic andPragmatic Perspectives. Oxford: OxfordUniversity Press.

Devitt, Michael and Kim Sterelny (1999).Language and Reality. Second edition.Cambridge, MA: MIT Press. Chapter 3.

Evans, Gareth (1982). The Varieties ofReference. Oxford: Oxford UniversityPress.

Ostertag, G. (1998). Definite Descrip-tions: A Reader. Cambridge, MA: MITPress.

Schiffer, Stephen (2005). ‘Russell’s theoryof definite descriptions’. Mind 114:1135–83.

Siobhan Chapman

DESCRIPTIVISM

Primarily a synchronic method of lin-guistic analysis in which the structureand variation of written and spokenlanguage are portrayed dispassion-ately and non-judgementally. Advo-cates of descriptivism, borrowing frommethodology employed by linguistsin the nineteenth century, furtheredthe development of linguistics as a sci-ence and focused on cataloguing and

58

(CRITICAL) DISCOURSE ANALYSIS

Primary sourcesBloomfield, Leonard (1933). Language.

New York: Henry Holt.Hockett, Charles Francis (1958). A

Course in Modern Linguistics. NewYork: Macmillan.

Further readingHarris, Zellig (1951). Methods in Struc-

tural Linguistics. Chicago: University ofChicago Press.

Hudson, R. A. (1980). Sociolinguistics.Cambridge: Cambridge UniversityPress.

Newmeyer, Frederick J. (1996). Genera-tive Linguistics. London: Routledge.

Pedersen, Holger (1931). Linguistic Sci-ence in the Nineteenth Century: Meth-ods and Results. Trans. John WebsterSpargo. Cambridge, MA: Harvard Uni-versity Press.

David V. Witkosky

(CRITICAL)DISCOURSEANALYSIS

A term applied to a broad and hetero-geneous range of approaches to language which share a family resemblance in their focus on the lin-guistic characteristics, organisationalpatterns and communicative func-tions in context of naturally-occurringtext, be it spoken or written (orsigned). It has been argued that thecentre of gravity of linguistic studyhas shifted over the past thirty yearsfrom the sentence to the text, that is,from syntax to discourse; this shiftreflects an increasing acceptance that

European languages, their systematic,scientific approach appealed to earlytwentieth-century linguists such asFranz Boas, Edward Sapir and Ferdi-nand de Saussure, who refined themethodology and used it to presentsynchronic analyses of various lan-guages, including those of NorthAmerican Indian tribes. Descriptivismis especially associated with LeonardBloomfield, Charles Hockett, WilliamLabov and Kenneth Pike, who estab-lished procedures for examiningphonology, morphology, syntax andsemantics, and who insisted that lin-guists eliminate value judgementsfrom their studies and avoid social,cultural and moral grandstanding.They believed that linguists shouldobserve actual speech in linguisticcommunities, form hypotheses aboutcommon trends in language and inter-relationships between various parts oflanguage, and test their hypotheses.

The heyday of descriptivism span -ned the years 1930 to 1960. After-wards, an increasing number oflinguists found the approach too lim-iting. For example, in the 1960sNoam Chomsky and supporters oftransformational-generative linguis-tics expressed the view that scholarsshould investigate speakers’ uncon-scious awareness and knowledge oftheir language’s abstract system ofrules. In the 1970s M. A. K. Hallidayand other sociolinguists asserted thatlanguage could not be studied andexplained without examining thestructure of the society in which it isspoken. Regardless of reservations orcriticisms, the openness, inclusivenessand rigour of descriptivism continueto provoke and inform linguisticdebate into the twenty-first century.

59

(CRITICAL) DISCOURSE ANALYSIS

remains a central focus. One con -sequence of this openness to otherdisciplines, however, is that it issometimes difficult to trace the limitsof what is and is not included in discourse analysis.

This indeterminacy is com-pounded by the inherently multifac-eted nature of discourse, which can beapproached at a micro level, such asexploring how each utterance in turnrelates to the next in an interaction, ata macro level, for example investigat-ing the discourse of medicine from theperspective of its role as a form of reg-ulative control within a society, oranything in between. It is reflected inthe number of terms which can beseen as alternative labels (althoughmany of the practitioners would stressthe differences): discourse studies,text analysis, text linguistics, registeranalysis, genre analysis and so on. It isalso reflected in uncertainty over theboundaries between pragmatics anddiscourse analysis, and over whetherapproaches such as speech acttheory* or conversation analysis*come under the umbrella of discourseanalysis, or are compatible but sepa-rate fields of inquiry (the plurality ofviews can be seen simply by typing‘discourse analysis’ into an Internetsearch engine such as Google andopening the first five or six links in thelist). It is therefore perhaps best tothink of discourse analysis less as aunified approach and more as a pre-disposition on the part of the linguistto aim to relate language forms totheir contexts of use. What is strikingis that this predisposition has come todominate linguistic inquiry in a waythat would have seemed unlikely inthe early days when the structural

linguistics should be concerned withdescribing and explaining language inuse.

See also: Conversation Analysis;Corpora; Implicature;Integrationism; Langue/Parole;Politeness; Relevance Theory;Systemic-Functional Grammar;Speech Act TheoryKey Thinkers: Austin, J. L.;Bakhtin, Mikhail; Cameron,Deborah; Grice, H. P.; Halliday,M. A. K; Labov, William; Sacks,Harvey; Sinclair, John; Tannen,Deborah

Discourse analysis is often defined asthe analysis of language beyond thesentence. This is over-simple, but ithas the merit of making clear how itdiffers from traditional linguisticscentred around syntax, since the sen-tence is accepted as the maximumdomain within which syntactic struc-tures operate. A further crucial dif-ference is that, whereas theoreticallinguistics has, since Ferdinand deSaussure, insisted on the autonomyof linguistics and attempted to isolateit from other fields by strictly cir-cumscribing its domain, discourseanalysis is strongly oriented towardsinterdisciplinarity. It makes contact,and partly overlaps, with a widespectrum of other social sciencesincluding social and cognitive psy-chology, sociology, ethnography,anthropology, pedagogy and com-munication studies. In turn, it hasbeen applied in those other fields(and in some cases it has then lostalmost all characteristics that relateit to linguistics as a discipline). Acore feature of ‘linguistic’ discourseanalysis is that the language itself

60

(CRITICAL) DISCOURSE ANALYSIS

(‘non-structural’ in the sense that theycannot be adequately described interms of clause structures such as subject-verb-object). These includeconjuncts (for example, ‘However’),reference items (for example ‘These’at the start of the present sentencereferring back to ‘links’ in the preced-ing sentence), and repetition (forexample, the phrase ‘language beyondthe sentence’ above is repeated froman earlier paragraph to signal the con-tinuity of topic). Halliday and Hasan’sdefinition of ‘text’ (1976: 2) marked aconscious departure from the con-cerns of structural linguistics: ‘A textis best regarded as a semantic unit: aunit not of form but of meaning.’

Although much work was donefrom the late 1960s on that laid thefoundations for discourse analysis,the first explicit introduction to thefield was Malcolm Coulthard’s 1977book, which was soon followed byother ground-breaking volumes byBeaugrande and Dressler (1981),Brown and Yule (1983) and Stubbs(1983). There are naturally areas ofoverlap, but the differences betweenthese four books reflect some of themost important directions in whichdiscourse analysis developed. ForCoulthard, whose treatment wasinspired by the findings of an earlierstudy that he and John Sinclair hadcarried out into the classroom lan-guage of teachers and pupils, ‘dis-course’ essentially refers to spokeninteraction: he draws on work inspeech act theory* and other areasof pragmatics, and in conversationanalysis*, as well as work by Hymes(1974) and others on the ethnographyof speaking, and by Labov (1972) andothers on sociolinguistics. Coulthard

linguistics of Bloomfield, Chomskyand others was exerting its hegemony.

The first use of the term ‘discourseanalysis’ is generally ascribed to ZelligHarris in a 1952 paper. However, hispaper did not set an agenda that dis-course analysis has followed since.Although he focused on structureabove the sentence, his aim was toextend existing methods for analysingsentence structures, based purely ondistribution with no recourse tomeaning, to the analysis of texts. Apioneer whose contribution is recog-nised as more relevant is T. F. Mitchellwho, in a 1957 article, identified func-tional stages in interactions betweenbuyers and sellers in a particular cul-tural context. This study had certainkey features that came to be charac-teristic of much discourse analysis: itanalysed naturally-occurring text; itworked at text level; it explored howdiscourse is organised to carry out thecommunicative purposes of the inter-actants, particularly by highlightingthe ways in which each stage serveddifferent functions but all contributedto the achievement of the interactants’goals; it demonstrated that such inter-actions are not randomly recreatedeach time but follow conventions thatare accepted as appropriate by, andreflect the norms of behaviour of,members of that culture; and it identi-fied linguistic signals that distin-guished each stage of the interaction.

One of the most significant start-ing points in the study of languagebeyond the sentence was the publica-tion of Michael Halliday and RuqaiyaHasan’s 1976 book on cohesion. Thisexplored the non-structural cohesivelinks which play a central part intying sentences together into a text

61

(CRITICAL) DISCOURSE ANALYSIS

methodological issues of collecting,transcribing and analysing data. Suchissues are generally of little relevanceto theoretical linguists, who rely onintuition; but they are crucial for dis-course analysts, for whom data is thestarting point of inquiry.

Between them, the four books thuscover ‘text’-oriented aspects of speech(how utterances function in relationto each other) and writing (whatmakes a text hang together as a text),‘process’-oriented aspects (how pro-ducers and receivers construct meaning collaboratively) and ‘con-text’- oriented aspects (how discoursereflects and constructs social contextsof use). These four lines of inquiry,sometimes pursued separately butoften combined in the same study,were to provide rich ground for dis-course analysis over the followingdecades.

At around the same time as thesevolumes were appearing, furtherimpetus for the focus on discoursecame from applied linguistics andmore specifically, from the teaching ofEnglish for Specific Purposes (ESP),which expanded dramatically fromthe 1980s on. Teachers trying to helptheir students cope with, for example,writing academic articles in English,soon realised that it was not enough toteach them to produce grammaticallycorrect sentences. They needed to beshown how to write texts in the waysthat were conventionally accepted intheir field. This involved, amongmany other things, knowing how toorganise their text following acceptedpatterns, such as the IMRAD (intro-duction, methods, results and discus-sion) format of scientific researcharticles, and how to deploy language

is particularly interested in the structure of exchanges (multi-partstretches of conversation that formfunctional units, such as question-answer-acknowledgement), and theways in which utterances constrainwhat can follow them, and providethe context in which the followingutterances make sense. Beaugrandeand Dressler, on the other hand, focuson the general characteristics of texts,with more attention to written text.They outline seven ‘standards of tex-tuality’ which distinguish texts fromnon-texts. They argue that, to beaccepted as a text, a stretch of dis-course must have qualities such ascohesion (largely in Halliday andHasan’s sense), coherence (recognis-able unity of content) and situational-ity (recognisable appropriacy to thecommunicative situation). Brown andYule establish a distinction betweentext-as-product and discourse-as-process: that is, they see ‘text’ as therecord of a process of communication,while ‘discourse’ is the process itself.This leads them to place emphasison the psycholinguistic processingaspects of discourse: the ways inwhich producers organise their mes-sages so as to guide receivers to theintended meaning, and receivers usetheir background knowledge andpowers of inference to constructmeaning. Finally, as the subtitle of hisbook (‘the sociolinguistic analysis ofnatural language’) indicates, Stubbs isconcerned with the ways in which dis-course can only be fully analysed if itis situated in its sociocultural context.Like Coulthard, he explores speechacts and exchange structure, buthe also highlights the ethnographicdimension, and devotes much space to

62

(CRITICAL) DISCOURSE ANALYSIS

successfully and in order for thisprocess to be implemented effectively,the genres need to be analysed anddescribed in a way that simultane-ously uncovers the functions that lan-guage serves in a particular genre andshows the place of that genre in rela-tion to other genres in the culture.

The stages (or, in Swales’s terms, themoves) of a genre are identified interms of their function in achievingthe overall communicative aim. Themore delicate description of the lan-guage choices in the stages theninvolves register analysis. A register isseen as a variety of language associ-ated with particular communicativecontexts which members of a culturerecognise as recurring in accepted pat-terns: for example, classroom interac-tions, school textbooks, recipes, newsbroadcasts, research articles and soon. All registers use the same languagesystem (for example, research articlesand recipes in English are recognis-ably in the same language), but theprobabilities of occurrence of particu-lar configurations of grammatical andlexical choices from the system varyfrom register to register. To take asimple example, recipes (in theinstruction stage) have a markedlyhigh number of imperatives of verbsof action (‘mix’, ‘add’, and so on);research articles, on the other hand,have many fewer imperatives, and theverbs in these cases are from a smallconventional set referring to mentalprocesses (‘see’, ‘consider’, and so on).Martin and Rose (2003) is an accessi-ble introduction to this view of genreand register.

By the mid-1980s, discourse analy-sis had already developed much ofits all-encompassing nature. It was

choices appropriately in differentstages of the text – for example, in anacademic paper, using tense in the lit-erature review to show the degree ofcentrality of cited work to their ownresearch, or using modality* to adjustthe strength of claims in the discussionsection. Since traditional grammar didnot cover such topics, many ESPteachers became discourse analysts inorder to work out what their studentsneeded to know: they were enthusias-tic consumers of the growing researchin the field, and they applied it and fedtheir insights back into the discipline,enriching and extending it. Probablythe best-known example of the ESPdiscourse analysis tradition is JohnSwales’s 1990 book Genre Analysis,based on research going back to thelate 1970s, but this is only one repre-sentative of the huge body of workarising from the educational needs oflearners of English.

At the same period, the educationalcontext in Australia inspired a slightlydifferent kind of discourse analysis,also under the label ‘genre analysis’.Jim Martin and his colleagues (forexample, in Martin 1985/1989)applied systemic-functional gram-mar* in investigating the genres thatpupils and students are required tomaster in their mother tongue. Theirwork has a very strong socioculturalorientation: Martin’s definition ofgenre is that it is ‘a staged, goal- oriented, purposeful activity in whichspeakers engage as members of ourculture . . . Culture seen in these termscan be defined as a set of genericallyinterpretable activities’ (1985: 25).Education is in essence the process ofhelping children to extend the rangeof genres in which they can operate

63

(CRITICAL) DISCOURSE ANALYSIS

tity and norms (that is, discourse doesnot simply reflect existing external‘facts’, it plays a central role in con-structing the world as we see it). How-ever, CDA goes further in focusingespecially on the role of discourse inconstructing and reproducing socialinequalities as embodied, for exam-ple, in racial and sexual discrimina-tion. Certain forms of discourse areavailable to those with power, whileother forms of discourse construct theusers as (relatively) powerless. To takea simple example, each time someonefills in a tax return, providing theinformation required in the formatspecified, they are accepting andthereby reaffirming the authority ofthe institution controlling their behav-iour. This seems completely natural,and most people do not take a resist-ant attitude, whatever their feelingsabout paying tax. The discourseembodied in tax forms and the likehas become so naturalised that we nolonger notice the implications, interms of reinforcing the unequal dis-tribution of power, of doing as we aretold – that is, of playing our assignedroles in a discourse event. The set ofnaturalised assumptions about the‘state of things’ with which languageusers work, and which are reflected inthe language they use, are seen asmaking up the users’ ideology. Themore inequality is built into one’s ide-ology – that is, the less obvious it is –the more difficult it is to resist or evento question. CDA practitioners set themselves the task of using variousforms of analysis to bring to light the ideological assumptions underlyingany discourse, particularly thosewhich embody dominant forces in thesociety.

certainly losing overall coherencethrough the proliferation of often-conflicting methodologies andassump tions about language, societyand the main goals of the analysis, butat the same time it was gainingstrength through the sheer range oftopics that could be covered andthrough the increasing sophisticationof the analytical methods on offer. Asense of the diversity can be gained byinspecting the spectrum of areas cov-ered in surveys of discourse analysissuch as those edited by van Dijk(1985) and by Schiffrin et al. (2003).The diversity can also be seen by comparing publications such as (totake just two examples fairly ran-domly) Georgakopoulou and Goutsos(1997), which keeps to a firmly text-linguistic approach and highlightsespecially the linguistic and organisa-tional differences between narrativeand non-narrative discourse, andWetherell et al. (2001), a readerdesign ed to introduce discourse analy-sis to students of the social sciencesand oriented strongly towards ethno-graphic and socio-cultural issues. Thediscipline is also served by an increas-ingly wide selection of journalsincluding Discourse Studies, Dis-course and Society, Text & Talk andothers.

One of the more recent manifesta-tions of discourse analysis that isworth special mention has been in theform of critical discourse analysis(CDA), as developed by scholarssuch as Norman Fairclough (1995),Teun van Dijk and Ruth Wodak. Thisstarts from the basic constructionistassumption in most discourse analysisthat discourse inherently functions toconstruct and reproduce social iden-

64

(CRITICAL) DISCOURSE ANALYSIS

Martin, J. R. (1985). Factual Writing:Exploring and Challenging Social Real-ity. Geelong, Australia: Deakin Univer-sity Press. Republished by OxfordUniversity Press 1989.

Schiffrin, Deborah, Deborah Tannen andHeidi E. Hamilton (2003). The Hand-book of Discourse Analysis. Oxford:Blackwell.

Stubbs, Michael (1983). Discourse Analy-sis: The Sociolinguistic Analysis of Nat-ural Language. Oxford: Basil Blackwell.

Swales, John (1990). Genre Analysis: Eng-lish in Academic and Research Settings.Cambridge: Cambridge UniversityPress.

Van Dijk, Teun A. (ed.) (1985). Handbookof Discourse Analysis (four volumes).London: Academic Press.

Wetherell, Margaret, Stephanie Taylorand Simeon J. Yates (2001). DiscourseTheory and Practice: A Reader.London: Sage.

Further readingAntaki, Charles, Michael Billig, Derek

Edwards and Jonathan Potter (2002).‘Discourse analysis means doing analy-sis: a critique of six analytic shortcom-ings’. Discourse Analysis Online 2,http://www.shu.ac.uk/daol/articles/open/2002/002/antaki2002002-paper.html(accessed 11 October 2007).

Harris, Zellig (1952). ‘Discourse analysis’.Language 28: 1–30.

Hymes, Dell (1974). ‘Ways of speaking’.In R. Baumann and J. Sherzer (eds)Explorations in the Ethnography ofSpeaking. Cambridge: Cambridge Uni-versity Press. 433–52.

Labov, William (1972). SociolinguisticPatterns. Philadelphia: University ofPennsylvania Press.

Martin, J. R. and David Rose (2003).Working with Discourse: Meaning

As the above outline makes clear, itis difficult in the scope of this article toprovide a straightforward overview ofwhat discourse analysis is. The lack ofa clearly-defined domain has been thefocus of much criticism over the years,together with the complaint (howeverelegantly expressed) that discourseanalysis is not ‘real linguistics’ butsimply impressionistic commentary. Itis certainly true that the centrality oftheoretically-grounded analysis oflanguage is sometimes overlooked:Antaki et al. (2002) is both a usefulsummary of the problems and areminder that they need to be avoided.However, the achievements of dis-course analysis have proved itsrobustness: they have shown that it isas rigorous as other social sciences inits study of language in its contexts ofuse, and that this can provideuniquely valuable insights into howwe conduct our lives through lan-guage.

Primary sourcesBeaugrande, Robert de and Wolfgang

Dressler (1981). Introduction to TextLinguistics. London: Longman.

Brown, Gillian and George Yule (1983).Discourse Analysis. Cambridge: Cam-bridge University Press.

Coulthard, Malcolm (1977). An Introduc-tion to Discourse Analysis. London:Longman. Second edition 1985.

Fairclough, Norman (1995). Critical Dis-course Analysis. Harlow: Longman.

Georgakopoulou, Alexandra and Diony-sis Goutsos (1997). Discourse Analysis:An introduction. Edinburgh: EdinburghUniversity Press.

Halliday, Michael A. K. and RuqaiyaHasan (1976). Cohesion in English.London: Longman.

65

DISTINCTIVE FEATURES

developed these ideas to propose whatbecame, through further reworkingby Noam Chomsky and Morris Halle,the standard model.

Although the notions were implicit inearlier phonetic description, NikolaiTrubetzkoy (1939) focused attentionon phonology’s subsegmental level.Trubetzkoy described ‘oppositions’ be -tween phonemes, invoking the idea thatphonemes are characterised by the contrasts they exhibit in languages.American structuralists also spoke ofpho nemes’ features, principally to op -pose the distinctive and non- distinctiveproperties of languages’ in divisiblephonemes. Jakobson revised theselargely language-specific notions andsought a small language-universal set offeatures, which exist independently ofthe segments that they compose.

Trubetzkoy focused on distinctiveoppositions – those which signalphonological contrast. These could beprivative (a marked property is eitherpresent or absent), equipollent (bothmembers are of equal status) or grad-ual (with several gradations of oneproperty). All Jakobson’s features werearguably equipollent, with two values,each characterising a definite property(for example, tense/lax, nasal/oral).Jakobson’s work (some collaborative,clearly expressed in Jakobson, Fantand Halle 1952) based features prima-rily on segments’ acoustic properties.Chomsky and Halle (1968) redefinedfeatures using principally articulatorydefinitions, and used them in phono-logical rules, setting the scene for standard generative phonology. Theyreinforced features’ binary nature,using plus and minus values (for exam-ple, [+back], [–nasal]) in underlyingrepresentations and virtually all rules.

Beyond the Clause. London and NewYork: Continuum.

Mitchell, T. F. (1957). ‘The language ofbuying and selling in Cyrenaica’. Hes-peris 44: 31–71.

Geoff Thompson

DISTINCTIVEFEATURES

The smallest units of linguistic struc-ture, from which larger units are built,sometimes seen as the attributes bywhich phonemes* can differ. The ideais fundamental in phonology, wheremany generalisations are standardlystated in terms of features.

See also: Generative Phonology;Phoneme; Optimality TheoryKey Thinkers: Chomsky, Noam;Jakobson, Roman; Trubetzkoy,N. S.

One of the few areas of phonologicalconsensus is that segments are com-posed of features. Features play cru-cial phonological roles, being used (1)to express how segments contrastwith each other and (2) what groupsof segments (‘natural classes’) for-mally have in common, (3) to modelwhat changes in phonological pro -cesses in generative phonology*, and(4) in the formulation of constraints inoptimality theory*. The Prague Lin-guistic Circle provided the firstdetailed expression of the ideasbehind feature theory, and relevantideas occur in American structuralistwork. Roman Jakobson, seen as thefather of distinctive feature theory,

66

EMIC/ETIC

people producing that behaviour,while etic refers to categorising behav-iour from the perspective of an out-sider in ways that are applicable todifferent systems and can be used tocompare them. Since the 1950s theterms have gained currency in thehumanities and social sciences, espe-cially in the fields of anthropology andcross-cultural psychology.

See also: PhonemeKey Thinkers: Pike, Kenneth;Sapir, Edward

The emic/etic distinction is part ofPike’s theory of tagmemics, accordingto which verbal and nonverbal aspectsof behaviour are inseparable and mustbe studied in tandem, for the signifi-cance of the one frequently can onlybe discovered through reference to theother. Pike saw the compartmentalisa-tion of levels of analysis prevailing inlinguistics as producing only eticanalyses, or what is worse, disguisingemic categories as etic – for example,as when asking the natives for judge-ments of sameness/difference necessi-tating recourse to meaning, whileexplicitly rejecting meaning as a cate-gory external to the linguistic system.By proposing the emic/etic distinction,inspired by Edward Sapir, Pike aimedto increase awareness of the pitfalls ofprojecting one’s own emic categoriesonto an alien system, and to empha-sise the need to keep apart the twoperspectives, of the native speaker andof the analyst.

Pike did not, however, perceive arigid dichotomy between these twoperspectives. Rather he recognised aprogression from emic to etic, onceemic units discovered via analysis ofone language are applied to analysis

Phonologists have since refined theset of features, or sought structure intheir organisation (to account forgroup behaviour in processes). Fea-tures are now linkable to multiple seg-ments, and sometimes reinterpreted asexclusively privative (and given differ-ent names, such as ‘autosegments’,‘components’ or ‘elements’). Theory isfundamental in phonology: certaincurrents now push for a less categori-cal notion of feature, but these small-est linguistic units are in no danger ofbeing split further.

Primary sourcesChomsky, Noam and Morris Halle

(1968). The Sound Pattern of English.New York: Harper and Row.

Jakobson, Roman, Gunnar Fant andMorris Halle (1952). Preliminaries toSpeech Analysis. Cambridge, MA: MIT.

Trubetzkoy, N. S. (1939). Grundzüge derPhonologie. Baltaxe. Trans. Principlesof Phonology, Berkeley: University ofCalifornia Press, 1969.

Further readingAnderson, Stephen (1985). Phonology in

the Twentieth Century. Chicago: Uni-versity of Chicago Press.

Patrick Honeybone

EMIC/ETIC

Terms coined by Kenneth Pike, from(phon)emic and (phon)etic respec-tively, to refer to two complementaryways of analysing behavioural data.Emic refers to categorising behaviourfrom the perspective of the insider inways that are meaningful to the

67

EMPIRICISM/RATIONALISM

Psychology. Amsterdam: Swets &Zeitlinger. 55–63.

Marina Terkourafi

EMPIRICISM/RATIONALISM

Put simply, empiricism is the view thatall knowledge derives from experi-ence; rationalism is a contrasting viewin which knowledge comes from rea-soning. In epistemology, the branch ofphilosophy concerning theories ofknowledge, empiricism and rational-ism are types of position that havebeen taken about the sources ofknowledge, in particular in discus-sions about what is required for astate to count as knowledge. In psy-chology and its philosophy, empiri-cism and rationalism concern thesources of psychological states andcapacities that may include, but arenot confined to, states of knowledge.

See also: Analytic/Synthetic;Behaviourism;Deduction/Induction; Holism;Innateness; Linguistic Relativity;Logical Positivism; Mentalism;Universal GrammarKey Thinkers: Berkeley, George;Chomsky, Noam; Descartes, René;Frege, Gottlob; Hume, David;Locke, John

In a general sense, a position is aform of empiricism insofar as it holdsthat knowledge (or other psychologi-cal states and capacities, indicatedhenceforth: (etc.)) about some partic-ular subject matter, S, derives from

of another. To the extent that they arerelevant to the latter, these units con-stitute part of an etic inventory that isno longer internal to any single lan-guage. The International PhoneticAlphabet is an example of such aninventory, while the phonologies ofindividual languages correspond tothe emic systems from which theformer is abstracted.

An inventory of etic units is createdby the analyst prior to the analysis ofthe particular language to which it isapplied. Etic units correspond to rawobservational data, often measurableby instrumental means. Since theydo not combine into a system ofmeaningful contrasts, they are pre-structural. Conversely, emic units resultwhen the raw data are interpreted bynatives, though they are not necessarilyconsciously known (named) by them.They are thus structural and valid foronly one language. Combining emicand etic perspectives yields ‘a kind of“tri-dimensional understanding” ofhuman behaviour instead of a “flat”etic one’ (Pike 1954: 12).

Primary sourcesPike, Kenneth (1954). Language in Rela-

tion to a Unified Theory of the Structureof Human Behaviour. Glendale, CA:Summer Institute of Linguistics.

Further readingHeadland, Thomas, Kenneth Pike and

Marvin Harris (eds) (1990). Emics andEtics: The Insider/Outsider Debate.Frontiers of Anthropology. Vol. 7.London: Sage.

Jahoda, Gustav (1977). ‘In pursuit of theemic-etic distinction: can we ever capture it?’. In Y. H. Poortinga (ed.),Basic Problems in Cross-Cultural

68

EMPIRICISM/RATIONALISM

meaning (although issues surroundingthe structure of language and the pos-sibility of a perfect language – a lan-guage designed perfectly to reflect thestructure of reason – have also been ofsome importance). Thus, the empiri-cist David Hume argued that ideas –that is concepts, including the mean-ings of words – must derive ultimatelyfrom impressions – experiential input.On the basis of this constraint, Humeargued that we can have no idea cor-responding with genuine causation –that is, no idea whose content out-strips our experience of one type ofoccurrence regular following anothertype. Similarly, in the twentieth cen-tury, logical positivists like Ayerargued that, in order for a sentence tobe meaningful, it must be possible toverify (or falsify) it – to determinewhether it is true or false on the basisof experience. Opponents of the logi-cal positivists argued in turn that thisVerification Principle was not itselfverifiable by appeal to experience.Some of these opponents took them-selves to be pushing for an empiricismthat was even more extreme than thatof the positivists: they would arguethat this shows that the VerificationPrinciple is itself meaningless and thatwe need to construct a different formof empiricism. Other opponentsargued that knowledge of some verygeneral claims must derive from theuse of reason; the latter form of oppo-sition may reasonably be viewed as aform of rationalism.

Three major traditional points ofdispute between empiricists and ratio-nalists centre on the following threecharacteristic rationalist theses: (1)knowledge of a particular subjectmatter is underwritten by intuition (or

experience of that subject matter; anda position is a form of rationalisminsofar as it holds that knowledge(etc.) about some particular subjectmatter, S, derives from the use ofreason or, more generally, from ourrational nature(s). Traditionally, thesepositions, and disputes among theirproponents, have concerned knowl-edge (etc.) about mind-independentsubject matters, about how the worldis independent of particular views thatwe might take about it.

In the twentieth century a centraldebate between empiricists and rationalists concerned our knowl-edge of mathematics. Gottlob Fregeattempted to show, first, that ourknowledge of mathematics is substan-tive knowledge – knowledge of anindependent subject-matter and,second, that mathematics derivesfrom logic* so that our knowledge ofmathematics can be seen to derivefrom the use of (pure) reason. Frege’sempiricist predecessors, like J. S. Mill,had attempted to treat mathematicalknowledge as deriving from experi-ence. In the face of Frege’s critique ofthat move, his empiricist successors,like A. J. Ayer, attempted to treatmathematical knowledge as derivativefrom knowledge of meaning and so(they argued) not really a substantiveform of knowledge. This debate inturn raised questions about the statusof our knowledge of language: is thisknowledge derived from reason (moregenerally, an aspect of our natures asrational beings), or is it rather a formof experiential knowledge?

With respect to the study of lan-guage, major disputes betweenforms of empiricism and rationalismhave focused upon our knowledge of

69

EMPIRICISM/RATIONALISM

its lack of specificity. First, articulatingthe precise content of a form ofempiricism is dependent upon furtherspecification of the notion of experi-ence employed in the approximateaccount. For instance, one form ofempiricism holds that knowledge(etc.) about a particular subject matterderives solely from sense experience ofthat subject matter – visual, auditoryor tactile experience – and not fromother forms of experience – introspec-tive experience or religious revelation.Further specification of that form ofempiricism would be dependent uponfurther specification of the boundariesof sensory experience. Correlatively,articulating the precise content of aform of rationalism is dependent uponfurther specification of the extent ofreason or our rational nature(s).Whether a particular form of ration-alism is in dispute with a particularform of empiricism depends upon thedetails of such further specification. Inparticular, it depends upon whetherthe type(s) of experience to whichappeal is made in characterisingthe particular form of empiricismincludes, or excludes, the outputs ofthe type(s) of reason, or our rationalnature(s) to which appeal is made incharacterising the particular form ofrationalism.

Second, someone who holds thatthere are sources of knowledge inaddition to experience and reason, orour rational nature(s), might rejectempiricism (or rationalism) about asubject matter without endorsingrationalism (or empiricism) about thatsubject matter. For instance, thatperson might hold that knowledgeabout some subject matters dependsupon aspects of our non-rational

rational insight) and deductive rea-soning, rather than by experience ofthat subject matter; (2) knowledge ofa particular subject matter is innate(very roughly, determined by naturerather than, for example, by the par-ticular course of experience); and (3)the concepts or ideas that constituteour abilities to think about a particu-lar subject matter are innate. Ratio-nalists about knowledge (etc.) about aparticular subject matter characteristi-cally endorse at least one of (1)–(3)with respect to that subject matter.Empiricists about knowledge (etc.) ofa particular subject matter character-istically reject (1)–(3) with respect tothat subject matter. Since epistemolog-ical forms of empiricism and rational-ism concern the justification orwarrant required for a state to countas knowledge, and not the sources ofpsychological states and capacities ingeneral, it is possible to adopt psycho-logical forms of each type of positionwithout also adopting epistemologicalforms. For instance, one might holdthat a particular belief is innate – andso be a psychological rationalist withrespect to the belief – and also holdthat in order to be justified or war-ranted the innate belief must be sup-plied with experiential support – andso be an epistemological empiricistwith respect to the belief. Alterna-tively, one might hold that a particularbelief is only acquired on the basis ofexperience, but that the justificationor warrant for the belief derives fromreason. In that case, one would be apsychological empiricist and an epis-temological rationalist.

Two points are worth noting aboutthe approximate account of the twotypes of position, both pertaining to

70

EMPIRICISM/RATIONALISM

sometimes known as a ‘poverty of thestimulus’ or ‘poverty of evidence’ con-sideration – is that our knowledge(etc.) about the target subject mattercould not have been acquiredthrough, for example, sense experi-ence, so that empiricism is ruled out.The second type of consideration is anaccount of how the rationalists’favoured source – reason or ourrational nature(s) – could have under-written our acquisition of knowl-edge (etc.) of the subject matter. Forinstance, a rationalist about ourknowledge of ethics might attempt toaccount for our knowledge aboutwhat we ought to do in a particularcase by appeal to our possession ofinnate knowledge of general ethicalprinciples from which our knowledgeabout particular cases is derived.

In response to the first type of con-sideration, empiricists will attemptto argue that experiential resourcessuffice to explain the knowledge thatwe in fact possess. Empiricists willdevelop their response either byattempting to provide an account ofhow the knowledge (etc.) could havebeen acquired on the basis of experi-ence, or by arguing that we do not infact possess the knowledge that ratio-nalists claim we do. For instance, anempiricist about our knowledge ofmathematics might attempt to arguethat this knowledge is supported byinduction from our experiences ofgroups of objects. And an empiricistabout our putative knowledge ofethics might claim that we do notreally have such knowledge but onlyvarious feelings about particularcourses of action. Since such disputestypically concern knowledge (etc.)about a particular mind-independent

nature(s) that are not sense- perceptual; or that it depends uponnon-sensory experience.

There are other dimensions alongwhich particular forms of empiricismor rationalism might vary. But we cansee already that classifying a positionas a form of empiricism or rationalismis quite unrevealing. And we can alsosee that there is no such thing as thedispute between empiricism andrationalism; rather, there are – orcould be – various disputes eachtaking place between particular formsof each broad type of position.

Finally, we can see that care isrequired in classifying individualthinkers as empiricists or rationalists,for careless adherence to such a broadscheme of classification can serve todisguise differences between particu-lar members of one of the groups andsimilarities between members of thedifferent groups. René Descartes,Benedictus de Spinoza and GottfriedLeibniz are often aligned as paradig-matic rationalists, the so-called Conti-nental Rationalists. They are oftenseen in opposition to John Locke,George Berkeley and David Hume,the so-called British Empiricists, whoare often treated as paradigmaticempiricists. However, it is importantto look beyond that preliminary clas-sification in coming to a proper appre-ciation of the works of thoseimportant thinkers.

Where a dispute arises between aparticular form of empiricism and aparticular form of rationalism, thedispute characteristically takes the fol-lowing general form. The rationalistscharacteristically offers two types ofconsideration in favour of their posi-tion. The first type of consideration –

71

EMPIRICISM/RATIONALISM

The most prominent contemporarydefender of a form of rationalism isNoam Chomsky. Chomsky, togetherwith numerous co-workers in linguis-tics, psychology and philosophy, hasused ‘poverty of stimulus’ considera-tions in support of the thesis thathuman knowledge of natural lan-guage has a significant innate compo-nent. And Chomsky, again togetherwith co-workers in a variety of disci-plines, has developed an increasinglydetailed account of the developmentof our knowledge of particular natu-ral languages – for example, particu-lar dialects of English or particularBantu languages – that makes appealto innate structures and capacities.

Of course, Chomsky accepts thatexperience plays some role in theacquisition of knowledge of language,since it would otherwise be a mysterythat children typically acquire knowl-edge of language that enables themto communicate with those whosespeech they experienced during acqui-sition. But he holds that the role ofexperience is primarily to select fromamongst the child’s innate repertoirethe bits that will be operative in thecompetence that they come to employ.

Chomsky’s form of rationalism hasbeen subjected to both sides of thestandard empiricist critique. First, ver-sions of Locke’s objection have beenpressed, according to which Chom-sky’s appeal to innate psychologicalstates or capacities either fails to dis-tinguish his position from empiricismor is easily falsified. Because Chomskyaims to provide a detailed, predictiveaccount of the course of acquisition ofparticular languages, his positionappears to avoid the second horn ofLocke’s dilemma by going beyond the

subject matter, the empiricist mightattempt to argue that we have theknowledge that the rationalist claimswe do, but that it is knowledge aboutthe operations of our own minds, orrelations among our concepts orideas, rather than about mind- independent reality.

In response to the second type ofconsideration, the empiricist willattack the rationalist account of howwe come to have the knowledge (etc.)that rationalists claim we possess. Forinstance, the empiricist might followLocke in attempting to argue that theclaim that a piece of knowledge (etc.) isinnate is either false or consistent withempiricism. If the rationalist claimsthat a piece of knowledge (etc.) isinnate only if it is possessed by every-one at birth, then the empiricist willpoint to the absence of that piece ofknowledge (etc.) in the very young orthe dysfunctional. Alternatively, if therationalist opts for looser requirementson innateness by counting a piece ofknowledge (etc.) as innate if we areborn with a capacity to acquire it, thenthe empiricist will also be willing toaccept that all knowledge is innate inthat very thin sense. As Locke puts it:

If the capacity of knowing, be the natu-ral impression contended for, all thetruths a man ever comes to know, will,by this account, be every one of them,innate; and this great point will amountto no more, but only an improper wayof speaking; which whilst it pretends toassert the contrary, says nothing differ-ent from those, who deny innate princi-ples. For nobody, I think, ever denied,that the mind was capable of knowingseveral truths. (Locke, 1690: book I,chapter II, section 5, p. 61)

72

EMPIRICISM/RATIONALISM

rationalism. That should not be sur-prising now that we have recognisedthe variety of possible rationalist posi-tions. But some of the apparent differ-ences are especially striking and havebeen taken by some thinkers to under-mine Chomsky’s classification as arationalist, or the bearing of his workon the standing of more traditionalforms of rationalism.

First, although Chomsky talks ofknowledge of language, it is not clearthat he thinks of this knowledge as thesort of propositional knowledge –knowledge that such-and-such – thatis of concern to epistemologists. Somephilosophers have thought that theknowledge is really knowledge-how –that is, practical knowledge – likeknowledge how to ride a bicycle.Although Chomsky rejects that inter-pretation, it remains an open questionwhether he is right to do so. AndChomsky admits other reasons forthinking that the sort of knowledgein question differs from the sortthat concerns epistemologists. Forinstance, he does not think thatknowledge of language is justified orwarranted and he does not think thatwe are typically conscious of possess-ing it; he thinks of it as tacit knowl-edge. Moreover, some of these reasonsmight also be grounds for thinkingthat knowledge of language is notreally a psychological state, so under-mining Chomsky’s classification evenas a psychological rationalist.

Second, Chomsky does not thinkthat knowledge of language is knowl-edge about a mind-independent sub-ject matter. Rather, on Chomsky’sview, facts about an individual’s language are constituted by factsabout the individual’s psychology, in

bland claim that we are predisposedto acquire language. And because heprovides an account according towhich our initial state of knowledge isshaped by experience in the course ofnormal development, his accountavoids falsification by the fact thatsmall children and subjects of abnor-mal development lack ordinaryknowledge of language.

However, important questionsremain concerning the precise contentof the claim that a basic component ofhuman linguistic capacity or state isinnate. And the fact that many theo-rists who align themselves with eitherempiricism or rationalism – includingChomsky – agree that both innateand experiential factors play a role inshaping knowledge of languagetends to undermine the utility of their classification as empiricists or ratio-nalists. Second, empiricists haveattempted to provide accounts of lan-guage acquisition that make more lim-ited appeals to innate psychologicalstates and capacities. Some empiricistapproaches agree with the rationalistassessment of what the child acquiresand attempt to provide accounts ofhow the child might acquire it more orless solely on the basis of experience.Other empiricist approaches involvean attempt to show that the childacquires less than the rationalist hasclaimed, so that the task of accountingfor their acquisition is made easier forthe empiricist. Thus far, no empiricistaccount has been provided that hasanything approaching the depth,detail and coverage of rationalistaccounts.

Chomsky’s form of rationalismappears to differ in certain respectsfrom some more traditional forms of

73

EMPIRICISM/RATIONALISM

Primary sourcesBerkeley, G. (1975). Philosophical Works,

Including the Works on Vision. Ed.M. R. Ayers, Everyman edition.London: J. M. Dent.

Chomsky, N. (1988). Language and Prob-lems of Knowledge. Cambridge, MA:MIT Press.

Descartes, R. (1641/1984). Meditationson First Philosophy. In The Philosophi-cal Writings of Descartes, vol. 2. Trans.J. Cottingham, R. Stoothoff and D.Murdoch. Cambridge: Cambridge Uni-versity Press.

Hume, D. (1748/1999). An Enquiry Con-cerning Human Understanding. Ed. T.L. Beauchamp. Oxford: Oxford Univer-sity Press.

Leibniz, G. (c. 1704). New Essays onHuman Understanding. In G. H. R.Parkinson (ed.) (1973), Leibniz: Philo-sophical Writings. Trans. M. Morrisand G. H. R. Parkinson London: J. M.Dent and Sons.

Locke, J. (1690/1997). An Essay onHuman Understanding. Ed R. Wool-house. London: Penguin Books.

Further readingCarruthers, P. (1992). Human Knowledge

and Human Nature. Oxford: OxfordUniversity Press.

Cottingham, J. (1984). Rationalism.London: Paladin Books.

Kenny, A. (ed.) (1986). Rationalism,Empiricism and Idealism. Oxford:Oxford University Press.

Loeb, L. (1981). From Descartes to Hume:Continental Metaphysics and the Devel-opment of Modern Philosophy. Ithaca,NY: Cornell University Press.

Pinker, S. (1994). The Language Instinct.London: Penguin Books.

Spinoza, B. de (1677/1985). The Ethics. InThe Collected Works of Spinoza, vol. 1.

particular facts about their knowledgeof the language. If he were right aboutthis, then it would form another dif-ference between knowledge of lan-guage and ordinary propositionalknowledge. Third, although Chomskythinks that knowledge of language isdetermined as a part of human nature,he does not appear to think that it isdetermined as a part of our rationalnature(s). See also the entry on uni-versal grammar*. For a readableintroduction to some considerationsthat support Chomsky’s position, seePinker (1994).

To sum up, let’s return to the earliertwentieth-century debate concerningthe tenability of empiricist restrictionson the possible meanings of sentences.Recall that the logical positivistsargued that meaningfulness coincideswith the possibility of verification (orfalsification) on the basis of experi-ence. The empiricist position here wasthat there could be nothing in ourknowledge of language, including ourknowledge of meaning, that was notput there by experience. Chomsky’swork casts doubt on that view. For ifChomsky is right, then our knowledgeof language is shaped as much by ourbiology as by our experience. More-over, Chomsky’s work makes plainthat views about possible limits to ourknowledge must engage with (experi-ence-based) work in the sciences concerned with human nature or con-stitution. Hence, even those who seeka more empiricist view about ouracquisition of knowledge of languagetypically seek to support that view byappeal to work in those sciences.Needless to say, the various disputesin this area are far from beingresolved.

74

FEMINISM

were silenced and interrupted by men.Another influential study of the area isPamela’s Fishman’s 1978 analysis ofthe home conversations of Americancouples: she found that women did alot of the interactional ‘shitwork’ – asshe termed it – that is mundane butsustains relationships, in the sameway that they do other mundane workthat sustains the family, such as house-work.

Some of the researchers of the timedid not only argue that male domi-nance led to gender differences inspeech but to female deficiencies. The‘deficit approach’ as an extension ofthe ‘dominance approach’ was carriedover from an earlier period where menwere seen as ‘default human beings’and everything that deviates from themale norm was seen as deficient. Thusspeaking like a woman was not onlyseen as feminine but, at the same time,as deviating from ‘normal’ – that ismale – speech. It is one of the lastingachievements of this early work toidentify the ‘double bind’ in whichwomen as speakers found themselves:if they talked ‘like a lady’, they wereconsidered less than a full humanbeing; if they talked like a man, theywere considered insufficiently femi-nine.

By the 1980s the political climatehad changed and while the students ofthese pioneers were still interested inresearching the way men and womenspoke differently, they were no longerready to see those differences as anexpression of female subjugation andmale dominance. Feminism hadturned into a social movement thatwas more focused on celebratingfemale difference. Consequently the‘dominance’ and ‘deficit’ approaches

Ed. and trans. E. Curley. Princeton, NJ:Princeton University Press.

Stich, S. (ed.) (1975). Innate Ideas.Berkeley, CA: California UniversityPress.

Guy Longworth

FEMINISM

Feminism is a hotly contested termbut it can broadly be described as arange of social movements and theo-ries that have discrimination on thebasis of gender as their key concern.Feminist linguistics explores the inter-relationship between language andgender.

See also: Conversation Analysis;Deconstruction; (Critical)Discourse Analysis; PoliticalCorrectness; PoststructuralismKey Thinkers: Cameron, Deborah;Tannen, Deborah

Feminist linguistics, or ‘language andgender’ as the field is more widelyknown, emerged in the context ofthe second feminist movement of the1960s and 1970s. Linguists of thetime who were also feminists startedto ask what part language played inthe widespread discrimination againstwomen. This early work found a focalpoint in Robin Lakoff’s 1975 bookLanguage and Woman’s Place. Theauthor argued that men and womenspoke differently and that those dif-ferences were evidence of male domi-nation: women were described as lessconfident speakers, for instance, who

75

FEMINISM

evidence that women talked more thanmen and equally solid evidence thatmen talked more than women. At thesame time, the universalist underpin-nings of the feminist movement startedto be questioned by women of colour,the queer movement and others whodid not see themselves representedin the stereotypical straight whitemiddle-class US–American womanwho seemed to have been taken to rep-resent the prototypical female experi-ence up until then. Both the‘dominance approach’ and the ‘differ-ence approach’ came in for substantialcriticism on a number of grounds, par-ticularly for treating ‘men’ and‘women’ as homogeneous categories;for a lack of attention to context; foran unsophisticated understanding ofpower relations; and for insensitivityto ethnic, racial, social, cultural andlinguistic diversity.

As a result, poststructuralistapproaches to language and genderstarted to emerge. DeborahCameron’s book Feminism and Lin-guistic Theory (1985) is often consid-ered a foundational text for this newapproach. Poststructuralism* doesnot start from the assumption thatmen and women speak differently – itdoes not even start from the assump-tion that men and women naturallyexist as meaningful categories. Ratherit is language that calls the (gendered)identity of speakers into existence.Gender is thus no longer treated as agiven but the linguistic concern is nowwith the way in which gendered sub-jectivities are constituted in language.Once the question ‘how do men andwomen talk differently?’ had becomeobsolete, feminist linguistics was in aposition to focus on new questions.

of the 1960s and 1970s started to bereplaced by the ‘difference approach’.The key exponent of this approachis best-selling linguist DeborahTannen. In books such as That’s NotWhat I Mean (1986) and You JustDon’t Understand! (1990), she arguesthat communication between men andwomen could best be understoodthrough applying the framework ofintercultural communication – forexample, the way people who grew upin different parts of the world maysometimes create communication dif-ficulties – but men and women, sheargues, have communication difficul-ties because they operate with differ-ent values. Men are apparently drivenby competition and women by coop-eration. These different values areexpressed in different communicativestyles, which lead to misunderstand-ings and fights between the sexes.Representatives of the differenceapproach were keen to stress thatwhile men and women had differentcommunicative styles, these wereequally valid.

By the late 1980s language andgender had changed from an almostesoteric interest of some female lin-guists to become a thriving researchfield which was institutionally based inuniversity departments and curricula.However, this high level of researchactivity also had a surprising side-effect that led to a complete conceptualrethinking of the field: as more andmore researchers tested the variousclaims about female and male speechempirically, differences no longerappeared as clear-cut as they hadseemed to exponents of both the dom-inance and the difference approach.For instance, there was empirical

76

FEMINISM

such as applied linguistics, feminismand language and gender are onlyslowly finding their way into languageeducation (Pavlenko et al. 2001). Irre-spective of the approach, it is a keycharacteristic of most feminist linguis-tics that it is committed to a politicalcause, namely emancipation in vari-ous forms. Since the inception of thefield, feminist linguists have not onlyaimed at describing linguistic prac-tices. They have also been committedin various degrees to engagement with‘the real world’ and to challengingpractices that disadvantage groups ofspeakers.

Primary sourcesCameron, Deborah (1992). Feminism and

Linguistic Theory. London: Macmillan.First edition 1985.

Fishman, Pamela (1978). ‘What do cou-ples talk about when they’re alone?’. InDouglas Butturff and Edmund L.Epstein (eds), Women’s Language andStyle. Akron, OH: L&S Books. 11–22.

Lakoff, Robin (1975). Language andWoman’s Place. New York: Harper andRow. (A revised and expanded editionwith commentaries by contemporaryfeminist linguists was edited by MaryBucholtz and published by Oxford Uni-versity Press in 2004.)

Tannen, Deborah (1986). That’s NotWhat I Meant! How ConversationalStyle Makes or Breaks Relationships.New York: Ballantine Books.

Tannen, Deborah (1990). You Just Don’tUnderstand: Women and Men in Con-versation. New York: Ballantine Books.

Further readingPavlenko, Aneta, Adrian Blackledge,

Ingrid Piller and Marya Teutsch-Dwyer (eds) (2001). Multilingualism,

One such new question is related tothe construction of gendered identitiesin interaction. Such studies haveshown that gender is not some form ofstatic identity but changes from con-text to context: in some contextsgender may not matter at all, in otherswe may highlight or downplay certainaspects of our gender identities.

Another new question that has beenparticularly taken up by scholarsworking in (critical) discourse analy-sis* centres around the representationof particular types of femininity ormasculinity in the media. Media dis-courses have a central influence onmany facets of identity in contempo-rary societies from gendered risk-taking to gendered parenting. Therehave also been inquiries into the waysin which gender structures access tolinguistic resources. For instance,immigrant women often find it moredifficult to learn the majority languageas they may have less access to thepublic spaces where it is taught andused. In other contexts, girls may findit easier to access foreign languagelearning because it may be discur-sively constructed as a ‘girlie’ thing todo while boys are discouraged fromlanguage learning because of thesevery associations.

What all these questions andapproaches have in common is thatthey do not treat gender as given andstatic but as emergent in context andas being comprised of a range of per-formances, some of which may bemore hegemonic than others. Genderthus turned from a noun into a verb.Language and gender today has aplace in probably every sociolinguis-tics undergraduate class in the world.Despite being highly relevant in areas

77

GENERATIVE PHONOLOGY

and only the surface forms of a natu-ral language, focusing on its speechsounds. As with any such set of rules,the measure of its success was the ade-quacy* with which it was consistentwith underlying linguistic knowledge,known as universal grammar*.

Chomsky and Halle’s book TheSound Pattern of English (1968),known as ‘SPE’, is widely regarded asthe defining text of generative phono -logy. In it they attempted to specify thephonological rules under lying thespeech sounds of native Englishspeakers. SPE established the standardframework for this type of theory.That framework is as follows: thatthere are abstract rules deter miningthe actual acoustic output of speech;that the rules apply sequentially toproduce a series of derivations result -ing in an abstract repre sentation of thephonetic representation; that thephonetics consists of a series ofsegments that could be exhaustivelydefined in terms of sets of binaryfeatures; that the rules are strictlyordered.

By the 1970s SPE had become abenchmark against which most otherwork in phonology was measured.But critics began to find problemswith some of its basic assumptions.For instance, SPE’s focus on abstractrules, rather than detailed phoneticanalysis, and the formal complexity ofmany of its proposals, became stum-bling blocks for many linguists. Thesecriticisms led to developments suchas ‘Natural Generative Phonology’,which attempted to establish rulesthat were more psychologically plau-sible than the abstractions of SPE.Later work on optimality theory* didaway with rigid sequential derivation

Second Language Learning, andGender. Berlin and New York: Moutonde Gruyter.

Ingrid Piller

GENERATIVEPHONOLOGY

A branch of generative grammar thataims to establish a set of rules, princi-ples or constraints capable of produc-ing the surface phonetic forms of alanguage and of modelling the inter-nalised linguistic knowledge of thenative speaker. Generative phonologywas a central idea in linguisticresearch throughout the 1960s andalthough it has undergone reformsand changes in subsequent decades, itcontinues to be the dominant frame-work for many developments inphonological theory.

See also: Adequacy; DistinctiveFeatures; Optimality Theory;Transformational-GenerativeGrammar; Universal GrammarKey Thinkers: Bloomfield, Leonard;Chomsky, Noam; Jakobson,Roman; Trubetzkoy, N. S.

Generative phonology originated withthe work of Noam Chomsky andMorris Halle at the MassachusettsInstitute of Technology (MIT) in thelate 1950s. It built on N. S. Trubet-zkoy’s idea of phonemic oppositionsand Roman Jakobson’s later work ondistinctive features*. More specifi-cally it drew on the general aspirationof generative grammar to stipulate aset of rules capable of producing all

78

GENERATIVE SEMANTICS

half of the 1960s and the 1970s. Itopposed the approach dubbed ‘inter-pretive semantics’ favoured by NoamChomsky and others. Among its majorproponents were George and RobinLakoff, Paul Postal, John (Haj) Ross,and James McCawley. Generativesemantics is credited for bringing atten-tion to meaning in linguistics and forushering in interest in pragmatics, cog-nitive linguistics and some aspects ofsociolinguistics.

See also: Cognitivism; Prototype;Transformational-GenerativeGrammarKey Thinkers: Austin, J. L.;Chomsky, Noam; Grice, H. P.;Searle, John

Chomsky’s standard theory (1965)postulates the presence of formal ruleswhich generate syntactic sentenceskeletons which are then filled by lex-ical insertion rules to create deepstructures. These are then turned bytransformations into questions, pas-sives and so on. According to Chom-sky, transformations do not affectmeaning. Chomsky and many of hiscollaborators took the position thatlexical semantics had to take placeafter deep structures had been generated, relegating the role of lexi-cal semantics to the interpretation ofstructures that had already been gen-erated, just like phonology (hence thename given to this position: interpre-tive semantics). The generativesemanticists took a different position,that lexical insertion had to happenboth in the deep structure and aftersome transformations, but most significantly that transformationsaffected meaning, hence the name‘generative semantics’.

in favour of an algorithm for selectingsurface forms from a set of possiblealternatives.

In the twenty-first century, phonol-ogy might appear to have moved someway from the rigid model imposed inSPE. Nevertheless the formal andprincipled approach of generativephonology and the wider generativeproject remains an important founda-tion for much work in the field.

Primary sourcesChomsky, Noam (1964). Current Issues in

Linguistic Theory. The Hague: Mouton.Chomsky, Noam and Morris Halle

(1968). The Sound Pattern of English.New York: Harper and Row.

Halle, Morris (1962). ‘Phonology in Gen-erative Grammar’. Word 18, 54–72.

Prince, Alan and Paul Smolensky(1993/2004). Optimality Theory: Con-straint Interaction in Grammar. RutgersUniversity and University of Coloradoat Boulder. Oxford: Blackwell.

Further readingCarr, Philip (1993). Phonology. Hound-

mills: Macmillan Press.Kenstowicz, M. and C. Kisseberth (1986).

Generative Phonology: Description andTheory. New York: Academic Press.

Christopher Routledge and SiobhanChapman

GENERATIVESEMANTICS

An approach to the treatment of seman-tics within transformational-generativegrammar* popular between the second

79

GLOSSEMATICS

the study of meaning and its relationsto social interaction.

Primary sourcesHarris, Randy Allen (1993). The Linguis-

tics Wars. New York and Oxford:Oxford University Press.

Further readingChomsky, Noam (1965). Aspects of the

Theory of Syntax. Cambridge, MA:MIT Press.

Rosch, Eleanor (1975). ‘Cognitive repre-sentations of semantic categories’. Jour-nal of Experimental Psychology: General104: 192–233.

Ross, John Robert (Haj) (1973). ‘Nouni-ness’. In Paul Kiparsky, John RobertRoss, James D. McCawley and OsamuFujimura (eds), Three Dimensions ofLinguistic Theory. Tokyo: TEC Corpo-ration. 137–257.

Zadeh L. A. (1965). ‘Fuzzy sets’. Informa-tion and Control 8: 338–53.

Salvatore Attardo

GLOSSEMATICS

A structuralist approach to the studyof language that attempts to establisha formal and abstract theory of lan-guage equivalent to the exactness oftheories in the natural sciences by set-ting up a formal system of descriptionbased on a elementary unit called a‘glosseme’. The theory was developedin the 1930s by the Danish scholarLouis Hjelmslev in collaboration withHans Jørgen Uldall and is the mostprestigious outcome of the works ofthe ‘Copenhagen Linguistic Circle’.Any list of linguistic theories mentions

The debate between the two campswas fierce and highly technical,revolving, for example, on the correctdecomposition of the verb ‘kill’ in‘cause to become not alive’. Thedetails are beyond an introductoryaccount (but see Harris 1993) but theresult of the controversy was that thegenerative semanticists, while essen-tially correct, did not present a cohe-sive research programme and movedon to other fields, including genderresearch (Robin Lakoff), pragmatics(Georgia Green), cognitive linguistics(George Lakoff, Ronald Langacker),or developed other theories of syntax(Paul Postal, James McCawley).

In general, generative semantics isto be credited with the attention tomeaning that characterised the fieldsof pragmatics and cognitive linguisticsin the last quarter of the twentiethcentury and the first decade of thetwenty-first. In particular, generativesemantics’ interest in broad data,often resulting in odd, whimsical orarcane examples, led to an interest innotions that defied the rigid categori-sation of transformational grammar,such as Ross’s ‘squish’, Lotfi Zadeh’s‘fuzzy logic’, and Eleanor Rosch’s‘prototypicality’, which are at the coreof cognitive linguistics. Generativesemantics was also instrumental inbringing J. L. Austin’s, H. P. Grice’sand John Searle’s ideas to the fore-front of linguistic theory and con-tributing significantly to the creationof the discipline of pragmatics.

Generative semantics can thenbe seen as the precursor of many ofthe most significant contemporaryapproaches to the study of linguisticmeaning and as having brought abouta significant shift in paradigm toward

80

GLOSSEMATICS

stance’, which may best be understoodas unformed matter.

The most fundamental work inglossematics was written in Danish(Omkring sprogteoriens grundlæggelse(English translation: Prolegomena to aTheory of Language (1953)) and it issignificant that in Paul L. Garvin’sreview of the translation by Francis J.Whitfield, he says that ‘The Prolegom-ena are probably among the mostunreadable books in linguistics’. Partlydue to this and partly due to the factthat, apart from writings in Danish, thebulk of Hjelmslev’s works was writtenin French, the details of the theory are achallenge for the non-initiated, andalthough the intellectual achievementsof glossematics are widely acknowl-edged, the approach has had littleimpact on empirical studies. Its mostprominent status is as an inspiration forlinguists and as a beacon of scholarlyrigour.

Primary sourcesGarvin, Paul L. (1954). ‘Review of Prole-

gomena to a Theory of Language’.American Anthropologist, vol. 56, no.5, part 1: 925–6.

Hjelmslev, Louis (1943). Omkring sprogte-oriens grundlæggelse. Copenhagen: Uni-versity of Copenhagen. ReprintedAkademisk Forlag 1966. Trans. FrancisJ. Whitfield (1953), Prolegomena to aTheory of Language. Baltimore: WaverlyPress.

Hjelmslev, Louis (1975). Résumé of aTheory of Language. Copenhagen:Nordisk Sprog-og Kulturforlag.

Further readingHjelmslev, Louis (1928). Principes de

grammaire générale. Copenhagen: Høst

glossematics, but apart from accountsof the phonematic systems of somedialects, the theory has not beenadopted by many linguists.

See also: Logical Positivism;Phoneme; Signs and Semiotics;StructuralismKey Thinkers: Hjelmslev, Louis;Greimas, Algirdas; Jakobson,Roman; Saussure, Ferdinand de

The point of departure in glossematicsconsists in a few axiomatic claims: lan-guage is one of the semiotic systemsemployed by humans to think andcommunicate, the linguistic system isan immanent, self-contained, structurethat should be described withoutany ‘metaphysical’ or ‘psychological’claims, and the formal system ofglossematics is the adequate tool forthat. The theory is a hierarchically-ordered set of terms conceiving ‘func-tion’ (in a non-mathematical sense) asthe key-concept on which the otherterms are elaborated. Basically itdenotes dependence relations betweenother entities, called ‘functives’.According to the intrinsic logic ofglossematics, it presents a complicatednomenclature of ‘functions’ and ‘func-tives’, the most important of which arethe terms ‘constant’, a ‘functive whosepresence is a necessary condition forthe presence’ of another ‘functive’, and‘variable’ which is a ‘not necessary’condition. The apparatus is utilised onboth sides of the linguistics sign,‘expression’ and ‘content’, implyingthat minimal units of the same nature,‘glossemes’, can be found both in the‘expression form’ and the ‘contentform’. Form as a concept shouldbe conceived of as a synonym for structure, and it is opposed to ‘sub-

81

HOLISM

and ‘long’, the second of which itselfdepends for its meaning on ‘space’,and so on until the meanings of allwords are seen to belong to an inter-connected web. The mentalist formholds that the content of concepts(the components of propositions*)depends on that of all other conceptsin the thinker’s repertoire. The prosand cons of semantic holism are usu-ally unaffected by which form it takes.A word’s meaning is, after all, givenby the concept it expresses. Early dis-cussions of holism tended to beframed linguistically. Latterly it hasbeen framed in terms of the content ofthe language of thought*.

Most arguments for semanticholism have a two-premise pattern.The first premise is that some propertyother than meaning is holistic. Thesecond is that semantic facts aregrounded in this other property.Meaning therefore inherits the holismof the other property. The threeinstances of this form of argumentpresented below centre on evidence,on translation, and on psychologicalexplanation respectively.

Holism about evidence (‘confirma-tion holism’) is the thesis that whethersome empirical discovery confirms agiven proposition depends on the con-firmation status of a good number ofother propositions, and ultimately ofevery other proposition. Science standsor falls as a whole, since it consists ofmutually supporting theses, not eviden-tially isolated ones. If we accept this,and also hold (more contentiously) thatthe meaning of a sentence is consti-tuted by its confirmation conditions,semantic holism results. Tying meaningto confirmation conditions can have different motives. Logical positivists,

and Søn. Reprinted Copenhagen:Munksgaard, 1968.

Hjelmslev, Louis (1932). Etudes baltiques.Copenhagen: Levin & Munksgaard.

Hjelmslev, Louis and Hans Jørgen Uldall(1957). Outline of Glossematics. Copen-hagen: Nordisk Sprog-og Kulturforlag.

Hans Götzsche

HOLISM

Holism about meaning, or semanticholism, is the idea that there are noindependent units of meaning smallerthan the entire representationalsystem, that is, the language. Whilesome have happily embraced semanticholism, others have claimed that itsallegedly unpalatable consequencesundermine otherwise seeminglyattractive theories in the philosophyof language or psychology.

See also: Analytic/Synthetic;Descriptivism; IndeterminacyKey Thinkers: Dummett, Michael;Fodor, Jerry; Quine, W. V. O.

‘Holism’ has several different mean-ings, according to what is being saidto be holistic (as opposed to ‘atom-istic’). Some other forms of holismwill be mentioned here, but the focusis on semantic holism, which is a pivotal idea in the philosophy of language.

Semantic holism itself has both alinguistic and a mentalistic guise. Thelinguistic form holds that what oneword in a language means depends onwhat all the other words mean. Forexample, perhaps ‘giraffe’ depends forits meaning on the meaning of ‘neck’

82

HOLISM

waiting in the cupboard, and so on.Ultimately, it seems, an agent’s actsemerge out of a whole sea of rationallycoherent propositional attitudes* andrelated mental states. According tocognitive holists, this deeper truth isdisguised by the way we highlight justone propositional attitude in our ordi-nary explanatory practices. To getfrom this alleged truth to semanticholism (of the mentalistic kind in thiscase) we must make a further assump-tion: that a mental state’s content isgrounded in the state’s explanatorypotential. Since this potential is holis-tic, mental content is holistic.

Having sketched three populararguments for semantic holism, let usturn to objections, not to the argu-ments, but to semantic holism itself. Ifsuccessful, these objections can tell usmore about meaning than that it isnon-holistic. They also show that, foreach of the three arguments just con-sidered, either the underlying prop-erty is not holistic after all, or thatmeaning is not constituted out of con-firmation conditions, translatabilityor other such aspects.

One common objection to semanticholism is that to understand any wordor concept one would have to instan-taneously understand every word inthe language, or every concept in anentire repertoire. Even if this is notimpossible, it certainly seems not to behow we do achieve semantic compe-tence. One response is to insist thatchildren acquire concepts one by one,but these morph, becoming graduallydistinct and more sophisticated con-cepts, reflecting the beliefs theyacquire on the way to adulthood. Amore radical strategy is to accept thatwe do indeed acquire concepts en

for example, wished to streamline sci-ence by making this link. Others thinkthat if a sentence is independent in prin-ciple to confirmation or disconfirma-tion, then it would never be rationallyassertible and its meaning could neverbe learnt.

Translation holism is the thesis thathow we translate one word candepend on how we translate any otherword in the same language. Considerhow we might complete the transla-tion from Old English of the followingSaxon law:

(1) No man shall kill hwyđer except inthe presence of two or three wit-nesses; and then he shall keep hisskin for four days.

Clearly we need to know somethingabout what skinned creature it mighthave been impermissible (in the cultureof the utterer) to kill save in the pres-ence of witnesses. But we know thatthis is what we need to figure out onlybecause we have already translated theother words in the sentence. Semanticholism follows if we add the furtherassumption that the meaning of a wordis constituted by its translatability. Thisfurther claim is not entirely implausi-ble. After all, communicability is a fun-damental aspect of word meaning, anddepends on our translating (or inter-preting) one another’s words.

Holism about psychological expla-nation (‘cognitive holism’) is bestapproached by example. Suppose weexplain why someone walked towardstheir kitchen by saying that theydesired food. This only explains theirbehaviour if we assume they think thatthere is food in their kitchen, that itsdoor is not locked, that no assassin lies

83

HOLISM

istic alternatives. These latter usuallyidentify the meaning of a word ormental representation with a causalrelation of some kind. It has provenimmensely difficult to get straight onwhat kind of causal relation couldwork. The debate here often collapsesinto that between descriptivist theo-ries of reference (popular with holists)and direct theories of reference (popu-lar with atomists).

Even so, it is a simplification to treatthe debate as a choice between seman-tic holism and semantic atomism.There is an intermediate position –semantic molecularism – and a hybridposition, either of which may evadethe difficulties faced by the purerviews. In semantic molecularism aword or concept’s meaning is from itsconnection to some proper subset ofthe other words or concepts in the lan-guage or conceptual system, ratherthan being either independent from allthese connections or dependent on allof them. The difficulty here, as criticshave pointed out, is how to distin-guish between ancillary connectionsand constitutive connections.

The hybrid position involves adopt-ing an atomistic theory for some pur-poses and a holistic theory for others.For example, ‘two factor’ theories ofmental content say that, for purposesof cognitive explanation, a holisticaccount is better, while an atomistictheory is better suited to the purposeof understanding the conditions underwhich a belief is true. Each hybridtheory needs to be evaluated on itsmerits, but a general problem attachesto the strategy: how to fuse the twoelements of the theory.

Holism about meaning is distinctfrom other forms of holism but it is

masse, but to express this more plau-sibly as the view that we becomeproper subjects for the attribution ofpropositional attitudes in a kind ofGestalt event.

A different objection centres on thecommitment of holists to a fine-grained conception of meaning. Thenature of a person’s concept is read offfrom the sum of beliefs that personhas involving that concept, since abelief is just a linking of one conceptto other concepts. A change in beliefcould change the concept’s identity.Ultimately, a single discrepancy inbeliefs held would mean that twopeople had no concepts in common.This is, to say the least, counterintu-itive: we could neither understand,learn from, nor disagree with oneanother, or with our previous selvesafter we have changed our minds, forthat matter. Semantic holists try toreply by saying that while, strictlyspeaking, we have no concepts incommon, our concepts are similarenough to allow understanding, learn-ing, disagreement and so on.

Finally, some commentators thinkthat holism is incompatible with com-positionality. The meaning of a phraselike ‘pet fish’ ought to be a function ofthe meanings of ‘pet’ and ‘fish’. But ifsemantic holism is correct, the meaningof ‘pet fish’ ought to be associated withthat of ‘golden’, even though there isnothing in the meaning of either ‘pet’ or‘fish’ to predict this association. A pos-sible response, here, is to insist thatsemantic holism applies only to simplephrases or concepts; the meaning ofcomplex phrases or concepts is deter-mined by compositionality*.

Semantic holists respond in kind bypointing out problems with the atom-

84

IDEATIONALTHEORIES

Ideationalists hold that the meaningof words is inherited from the mean-ing of mental entities (ideas or con-cepts) rather than the other wayaround. Two classic statements ofideationalism have had an enduringinfluence, the first by John Locke andthe second by H. P. Grice.

See also: Nonnatural Meaning;Private LanguageKey Thinkers: Grice, H. P.; Locke,John; Putnam, Hilary;Wittgenstein, Ludwig

Locke sought to answer the question:‘[How can] the thoughts of men’sminds be conveyed from one toanother?’ (1706/1997: III.1.2). Havinga thought, according to him, was amatter of having ‘ideas’ – roughly, per-ceptual and abstract concepts – in one’smind. The existence and content of anidea was supposed to be private, thatis, directly accessible only to its posses-sor. Language allows this privacy to beovercome so that knowledge can bepooled. Words ‘stand as marks for theideas within [the speaker’s] own mind,whereby they might be made known toothers’. That is to say, spoken wordshave a proxy content that is inheritedfrom the speaker’s ideas and, withluck, appropriately decoded by thehearer.

Locke treated the principle ofideationalism as obvious, and focusedinstead on confusions that arise whenwe ignore how imperfect a vehicle forideas language can be. The first influ-ential criticism of ideationalism didnot emerge until two centuries later

common to argue for semantic holismfrom holism of these other kinds bymaking a constitutive assumptionabout semantic facts. Equally, thealleged incoherence of semanticholism has led others to go in the otherdirection and reject these constitutiveassumptions. Nevertheless, argumentsabout holism touch on many otherareas of the study of meaning.

Primary sourcesBlock, N. (1986). ‘Advertisement for a

semantics for psychology’. In P. A.French (ed.), Midwest Studies in Philos-ophy, vol. X. Minneapolis: University ofMinnesota Press. 615–78.

Davidson, D. (1975). ‘Thought and talk’.In S. Guttenplan (ed.), Mind and Lan-guage. Oxford: Oxford University Press.

Devitt, M. (1996). Coming to Our Senses.Cambridge: Cambridge University Press.

Dummett, M. (1976). ‘What is a Theoryof Meaning?’ In G. Evans and J.Mcdowell (eds), Truth and Meaning:Essays in Semantics. Oxford: Claren-don Press. 67–137.

Dummett, M. (1991). The Logical Basis ofMetaphysics. Cambridge, MA: HarvardUniversity Press.

Fodor, J. (1981). ‘The present status of theinnateness controversy’. In J. Fodor,Representations. Brighton: Harvester.

Fodor, J. and E. Lepore (1991). Holism: AShopper’s Guide. Cambridge, MA:Blackwell.

Quine, W. V. (1960). Word and Object.Cambridge, MA: MIT Press.

Further readingFodor, J. A. and E. Lepore (1993). Holism:

A Consumer Update. Amsterdam:Rodopi.

Alex Barber

85

IDEATIONAL THEORIES

itself subject to extra-cranial influ-ence, which is hence passed on to thecontent of utterances. To make thismove would, however, undermine theclassification of Grice as an ideation-alist. Ideationalists insist that wordsinherit their meaning from mentalstates. To allow that external factors,including characteristics of the lin-guistic community, can influence thecontent of our mental states is toallow that the inheritance relationmay also run in the other direction.

These reflections suggest thatideationalism is a tendency ratherthan a specific thesis. It is difficult todeny that the meaning of linguisticexpressions and our utterances ofthem depends in some way on lan-guage users having mental states. Buthow that dependency is spelled out,and whether the dependency ever runsin the other direction, are questionswhose answer is as yet unsettled.

Primary sourcesGrice, H. P. (1957). ‘Meaning’. Philosoph-

ical Review 66: 377–88.Putnam, H. (1975). ‘The Meaning of

“Meaning”’. In Gunderson, K. (ed.),Minnesota Studies in the Philosophy ofScience VII: Language, Mind, andKnowledge. Minnesota: University ofMinnesota Press.

Locke, J. (1706/1975). An Essay Concern-ing Human Understanding. Ed. P. Nid-ditch. Oxford: Clarendon Press.

Further readingDavis, Wayne A. (2006). Nondescriptive

Meaning and Reference: An IdeationalSemantics. New York: Oxford Univer-sity Press.

Alex Barber

when Ludwig Wittgenstein criticisedthe notion of a private bearer of con-tent – a Lockean idea. On one readingof his private language* argument,Wittgenstein is suggesting that Lock-ean ideas could never be held mistak-enly, and hence that it is alsomeaningless to talk of their being heldcorrectly. If content is a matter of cor-rectness conditions, Lockean thoughtslack content and so cannot be thesource of our words’ content.

Ideationalists need not be weddedto Locke’s specific model of humanpsychology. Grice, for example,argued in 1957 that utterances inherittheir meaning from the peculiar inten-tions with which they are performed.If an utterance ‘Snow is cold’ meansthat snow is cold, it does so because itwas intended to cause the hearer tobelieve that snow is cold, and to do sothrough the hearer’s recognising thisvery intention. Nothing here commitsGrice to Lockean ideas. In a secondmajor potential problem for ideation-alism, Hilary Putnam in 1975 devel-oped a famous thought experimentinvolving ‘Twin Earth’ to argueagainst the view, then prevalent, that aword’s content depends entirely onthe internal mental states of theperson uttering it. Our intuitionsappear to show that the content of anutterance of ‘water’ or ‘gold’ candepend in part on factors outside theutterer’s skull, such as the physicalcomposition of the colourless liquid inlocal lakes or the opinions of experts.

Grice could embrace Putnam’s con-clusion by allowing that the intentionsfrom which our utterances inherittheir content are not internal mentalstates but external ones. That is, per-haps the content of our intentions is

86

weakest of the Achaeans was it thathad smitten [Hector]’ has been usuallytaken to convey, when uttered by theauthor of The Iliad (15.1) about Ajas,that Ajas was rather the strongest ofthe Achaeans. The usage of suchexpressions of understatement wasmarked in early rhetorical studies as‘litotes’ (Hoffmann 1987), but thetheoretical problem of delineation andproblem of derivation of what is con-veyed but not said remained openuntil the second half of the twentiethcentury.

It was the philosopher H. P. Gricewho made the first crucial contribu-tions to a theory of implicature thatpurports to delineate whatever is con-veyed beyond what is said when a cer-tain speech act is sincerely performedin a certain context of utterance andto show how to derive what is con-veyed but not said from the speech actand its context of utterance. Grice’smajor ideas were made public duringthe 1960s, first in a 1961 paper on thecausal theory of perception, and thenin the 1967 Harvard UniversityWilliam James lectures, parts of whichwere distributed and published in various forms, and then in a revisedversion. They appeared in full inGrice (1989), which includes alsorelated ‘Prolegomena’ and ‘Retrospec-tive epilogue’.

Grice’s theory is about contexts ofutterance that involve a speaker, aspeech act, which includes a sentenceor other expressions, and a hearer.Such a context of utterance is onein which the speaker performs aspeech act intending to have somecommunicative effect on the hearer.Grice’s theory introduces two majorideas about communicative, ‘conver-

IMPLICATURE

When a sincere performance of aspeech act takes place in a certain con-text of utterance, what is conveyed bythe performance, under the circum-stances, beyond what is then being lit-erally said by it, is an implicature of it.

See also: Nonnatural Meaning;Relevance TheoryKey Thinkers: Grice, H. P.

‘I meant what I said, and I said what Imeant’, said an elephant calledHorton in Horton Hatches the Egg, abook for children by an author andillustrator called Theodore SeussGeisel, otherwise known as Dr Seuss.What Horton said about himselfmakes sense, because what Hortonmeant is not necessarily what Hortonsaid. Generally speaking, when aperson performs a speech act in somecontext of utterance, a full report ofwhat the person did under the cir-cumstances will have to include ananswer to the question, ‘What wasconveyed by the utterance made bythe speaker under the circumstances?’It is a major insight into the nature oflanguage use that the required reportis dividable into two separate parts:one, reporting what was said, and anadditional one, reporting what wasconveyed beyond what was said.

Interesting examples abound ofspeech acts that convey more thanwhat they say, but much less so sys-tematic depictions of what is con-veyed but not said by a certain speechact in a certain context of utteranceand systematic understanding of howa speech act conveys what it does notsay. The Homeric expression ‘not the

87

IMPLICATURE

(Quality supermaxim)Try to make your contribution one thatis true.

(Quality maxims)(1) Do not say what you believe to be

false.(2) Do not say that for which you lack

adequate evidence.

(Relation maxim)Be relevant.

(Manner supermaxim)Be perspicuous.

(Manner maxims)(1) Avoid obscurity of expression.(2) Avoid ambiguity.(3) Be brief (avoid unnecessary prolix-

ity).(4) Be orderly.

‘and one might need others’. (Grice1989: 27)

Grice’s Cooperative Principle andsupermaxims and maxims apply toconversations in different ways. Thefollowing distinctions are crucial fora proper understanding of Grice’stheory. For the sake of brevity, we willuse the term ‘the norms’ for the Coop-erative Principle, supermaxims andmaxims.

Conversational implicatures bysimple observation of the normsversus. conversational implicaturesby dramatic observation of thenormsGiven an ordinary context of utter-ance, one often, or even usually,assumes that the speakers who partici-pate in the conversation observe the

sational’ contexts of utterance.Roughly speaking, the first idea is thatspeech acts performed in conversa-tional contexts of utterance are gov-erned by certain principles andmaxims of conversation; the secondidea is that what is conveyed but notsaid is what follows from the speaker’sobserving those maxims and what fol-lows from the speaker’s seeminglyflouting those maxims. It is clear thatsuch ideas can serve in solving thedelineation problem of what is con-veyed but not said. It will becomeclear in the sequel that the same ideasare of much significance in an attemptto solve the derivation problem aswell. We turn now to a brief presenta-tion of these two ideas.

Speech acts that are performedin conversational contexts of utter-ance are governed, according toGrice’s theory, first and foremost, bywhat Grice labelled the CooperativePrinciple:

(CP) Make your conversational contri-bution such as is required, at the stageat which it occurs, by the accepted pur-pose or direction of the talk exchange inwhich you are engaged.

The fundamental (CP) gives rise tosupermaxims and maxims of conver-sation, classified by Grice into fourgroups, ‘echoing Kant’, entitled interms of the categories of Quality,Quantity, Relation and Manner:

(Quantity maxims)(1) Make your contribution as inform-

ative as is required (for the currentpurposes of the exchange).

(2) Do not make your contributionmore informative than is required.

88

IMPLICATURE

However, as every fluent speaker of thelanguage knows, the speech act ofsaying ‘I saw the movie’ under such cir-cumstances counts as an appropriatereaction to the posed question, beingcooperative and informative, thoughin an indirect and implicit way. First,assuming the speaker does observe thenorms, we understand the speaker asholding that the movie is sufficientlysimilar to the book, to the extent thata person who is acquainted with themovie can be regarded as a personacquainted with the book. Notice thatsuch an understanding rests on theassumption that in reacting the waythe speaker did, the speaker observedthe Relation maxim and reacted in arelevant way. Secondly, assuming thespeaker observed Quantity maxim 1and Quality maxim 1, we have a pos-sible explanation for the speaker notanswering the posed question in theaffirmative. Hence, by assuming thespeaker observed the norms, we reachthe conversational implicature that thespeaker had not read the book. Herethe question arises: If the speaker hadnot read the book, why didn’t thespeaker admit it and simply answer‘No’? The response would be in termsof two insights we gain by trying toapply the norms to the conversationunder discussion. Notice, first, that theanswer ‘No’ would have been lesscooperative and less informative thanthe reaction ‘I saw the movie’. Thelatter helps the person who posed thequestion to make the next step in theconversation, pursuing the goal hetried to reach, had the answer been‘Yes’. If acquaintance with the movie isenough, the conversation will go on asplanned for that stage of it. Moreover,by assuming the speaker observed

norms. Conversational implicature bysimple observation of the norms isimplicature that follows from theassumption that the norms have beenobserved, under the given circum-stances, in which there is no indication– real or apparent – to the contrary. If aspeaker says, in an ordinary context ofutterance, ‘It is raining here’, weassume the person is observing thenorms, in particular Quality super-maxim and maxim 1, given no indica-tion to the contrary. A conversationalimplicature by simple observationwould be that the speaker believes, atthat context of utterance, that it is raining there. This explains the so-called Moore’s Paradox, whichemerges when a persons says: ‘It is rain-ing here, but I don’t believe it is’ or asimilar expression. The conversationalimplicature that follows from the firstpart of the assertion contradicts what issaid in the second part of the sameassertion.

Conversational implicatures bysimple observation abound and are ofmuch significance in ordinary interac-tion. More interesting however, at leastfrom a theoretical point of view, areconversational implicatures that aredrawn in contexts of utterance inwhich the norms are actually observedthough there are seeming indicationsto the contrary. When a speakeranswers the question, ‘Have youread the present top best-selling book’,by saying, ‘I saw the movie’, the speak-ers apparently flouts the CooperativePrinciple, because strictly speaking thereaction is not an answer. The questionis of the form called ‘yes’/‘no’-questionand the speaker’s reaction does notseem to be the required one, because itincludes neither a ‘yes’ nor a ‘no’.

89

IMPLICATURE

not all of my colleagues read fiction.Such a conversational implicature iscalled ‘generalised’.

Conversational implicatures versusconventional implicaturesAn implicature in general is what isconveyed by a performance of a speechact beyond what is said when thespeech act is performed. According toGrice’s theory, it is possible to conveyby a performance of a speech act whatis beyond what is said independentlyof the norms of conversation which pro duce conversational implicature.There is a class of implicatures Gricecalled ‘conventional implicatures’,which do not rest on the norms of con-versation but rather on conventionsthat govern the use of certain expres-sions. Consider an ordinary use of‘but’, such as ‘What they did waslegally permissible, but ethicallywrong’. On the level of what is said,one may replace ‘but’ with ‘and’ with-out changing the propositional signifi-cance of the assertion and its truth-value. However, the use of ‘but’ con-veys that there is some significant con-trast between the two parts of theassertion. The significance of the con-trast varies, but not its conveyedappearance.

Grice himself thought the nature ofconventional implicatures should bebetter explained before ‘any free useof it, for explanatory purposes’. Somescholars have tried to reduce all con-ventional implicatures to other phe-nomena of language use, such asconversational implicature, presuppo-sition* or what is said. Related theo-retical debates are still ongoing.

Grice’s theory lends itself to threetypes of study, related to the following

Quantity maxim 2, we draw the con-versational implicature that thespeaker believes that, for the sake ofthe present conversation, it is sufficientto point out the fact that he had seenthe movie, conveying but not admit-ting that the speaker had not read thebook.

The distinction between conversa-tional implicature by simple derivationfrom the norms and conversationalimplicature by what Grice called ‘dra-matic’ derivation from the norms willhave to be manifest in each attemptedimprovements on Grice’s theory.

Particularised conversationalimplicatures versus generalisedconversational implicaturesOur discussion of the conversationalimplicatures that can be drawn from aspeech act of saying ‘I saw the movie’applied to the particular context ofutterance, in which that speech actserved as a response to a particularquestion about the speaker, ‘Have youread the present top best-sellingbook?’. We can easily imagine con-texts of utterance in which the samespeech act appears, even in referenceto the same movie, made by the samespeaker, where none of the above-mentioned conversational implica-tures can be drawn. Hence, thoseconversational implicatures are called‘particularised’ ones.

However, not all conversationalimplicatures are particularised. If aspeaker uses the expression ‘most’ inan ordinary speech act, a conversa-tional implicature is going to be usu-ally drawn in terms of ‘not all’. Thus,for example, an ordinary usage of‘Most of my colleagues read fiction’would conversationally implicate that

90

IMPLICATURE

able from (R) without any resort to(CP).

To move to our second question,several types of empirical studies ofGrice’s theory should be mentioned.One type includes surveys of linguisticbehaviour that result in apparentcounter-examples, such as people regularly making vague commentsor references in conversation. Suchexamples can, however, be shown tobe compatible with and evenexplained by Grice’s theory, where the‘at least cost’ part of (R) involves yetanother parameter, such as commit-ment or guilt on grounds of beingshown to be mistaken, which shouldbe minimised.

Empirical studies of other types arerelated to different theoretical ways ofexplaining data about what is con-veyed but not said (Bezuidenhout andMorris, in Noveck and Sperber 2006),to the psychological processes of com-puting conversational implicatures(Chierchia et al., in Noveck and Sper-ber 2006), to the operation of impli-cature production devices in the brain(Kasher et al. 1999), and to childacquisition of such devices (Noveck,in Noveck and Sperber 2006). Suchstudies are related to the issue of mod-ularity (in the sense of Fodor 1983),where much remains to be studied.

Finally, let us consider three theo-ries of implicature that have beendeveloped in the footsteps of Grice’stheory. The first two have been labeled‘neo-Gricean’. All of them have fun-damental principles of the above- mentioned form ‘most effectively andat least cost’.

Laurence Horn suggested (see Horn1984 and 2004) a replacement ofGrice’s maxims by two principles:

theoretical problems. First, what is thenature of the norms and what justifieseach of the maxims? Second, dospeakers usually observe the norms ofconversation? Third, what alternativetheories of language use would serveas improvements upon Grice’s theory,explaining the facts better and morebroadly?

The first attempt to show the foun-dation of Grice’s system of norms, dis-regarding debatable details, wasKasher (1976), further developed inwhat appears as an appendix inKasher (ed.) (1998). The major idea ofthe explanation is that speech activityis ideally rational, that is, it actsaccording to the Rationality Principle:

(R) Given a desired end, one is tochoose that action which most effec-tively, and at least cost, attains that end,everything else being equal.

When the best means at one’s disposalare the verbal ones, then one follows(R) by performing a certain speechact. It is not difficult to explain the‘most effectively’ requirement of (R)when a desired end is given, but it ismuch more difficult to elucidate the‘at least cost’ requirement. A simplereading of Grice’s norms shows whathe took to be some of the costs, suchas verbal effort, related to number ofwords uttered and the time it takes tovoice them and similar attributes ofspeech. Other ‘at least cost’ parame-ters include inaccuracies (avoided byexpressions of the Henry James style)and hurt feelings (avoided by usingpoliteness manifestations). Kasher(1976) includes arguments against(CP) and arguments that show thatsupermaxims and maxims are deriv-

91

IMPLICATURE

dominant ideas in pragmatics in the1980s and early 1990s.

Primary sourcesFodor, Jerry (1983). The Modularity of

Mind. Cambridge, MA: MIT Press.Grice, H. P. (1989). Studies in the Way of

Words. Cambridge, MA: Harvard Uni-versity Press.

Hoffmann, Maria E. (1987). NegatioContrarii: A Study of Latin Litotes.Assen: Van Gorcum.

Horn, Laurence R. (1984). ‘Toward a newtaxonomy for pragmatic inference:Q-based and R-based implicature’. InD. Schiffrin (ed.), Meaning, Form andUse in Context: Linguistic Applications.Washington, DC: Georgetown Univer-sity Press. 11–42.

Horn, Laurence R. (2004). ‘Implicature’.In Laurence R. Horn and Gregory Ward(eds) (2004). 3–28.

Horn, Laurence R. and Gregory Ward(eds) (2004). The Handbook of Prag-matics. Oxford: Blackwell.

Kasher, Asa (1976). ‘Conversationalmaxims and rationality’. In Asa Kasher(ed.), Language in Focus: Foundations,Methods and Systems. Dordrecht:Reidel. 197–216. Republished with anappendix, ‘Gricean inference revisited’,in Kasher (ed.) (1998), vol. IV: 181–214.

Kasher, Asa, Gila Batori, Nahum Soroker,David Graves and Eran Zaidel (1999).‘Effects of right- and left hemispheredamage on understanding conversa-tional implicatures’. Brain and Lan-guage 68: 566–90.

Levinson, Stephen C. (2000). PresumptiveMeanings: The Theory of GeneralizedConversational Implicature. Cambridge,MA: MIT Press.

Noveck, Ira A. and Dan Sperber (eds)(2006). Experimental Pragmatics. Bas-ingstoke: Palgrave Macmillan.

(1) The Q-Principle (where Q standsfor quantity): make your contribu-tion sufficient; say as much as youcan, given the R-principle.

(2) The R-principle (where R stands forrelation): make your contributionnecessary; say no more than youmust, given the Q-principle.

The Q-principle gives rise to Hornscales, such as <good, excellent>. Asincere speech act that portrays athesis as a good one has a Q-implica-ture that the thesis is not an excellentone.

Stephen Levinson (2000) suggestedan additional fundamental principleof a different nature. The M-principlefor the speaker (where M stands formanner): do not use a marked ex -pression without reason. A markedexpression indicates an unusual situa-tion, while an unmarked expressionindicates a usual one. A sincere speechact that describes a certain action as‘not unreasonable’ has an M-implica-ture that the action is not perfectlyreasonable.

Relevance theory*, as developed bySperber and Wilson (see Sperber andWilson 1987 and Wilson and Sperber2004) and their followers, rests ontheir principle of relevance, accordingto which human cognition is ideallygeared to maximising relevance. Herea major shift from Grice’s theory ismanifest. Whereas Grice was inter-ested in rational speech activity, relevance theory is about cognitiveeffects and processes. According torelevance theory, relevance is a func-tion of cognitive effects, which shouldbe maximised, and of processingefforts, which should be minimised.Relevance theory was one of the

92

INDETERMINACY

ings of expressions and utterancesprecisely because there are no mean-ings about which to be definite. Mean-ing does not exist as an autonomousentity but only in observable behav-iours within a language community.Writers starting from quite diversesets of initial assumptions have foundthemselves advocating this scepticaland, to many thinkers, counterintu-itive thesis.

See also: Behaviourism; HolismKey Thinkers: Kripke, Saul;Quine, W. V. O.; Wittgenstein,Ludwig

Indeterminacy in linguistics is mostcommonly associated with W. V. O.Quine’s 1960 discussion of translation.Quine alleged that translation is inde-terminate because words themselvesdo not have meaning; all that linguistscan do is observe patterns of behaviourin relation to particular language use.Quine’s ‘meaning scepticism’ finds anecho in a puzzle raised by Saul Kripke.Kripke’s puzzle purports to show thatwe do not follow any determinatemeaning-rule in applying a word innovel circumstances.

In his 1960 discussion Quine is notoffering practical advice but scrutinis-ing the notion of meaning. For thisreason he asks us to imagine we arewhat he calls ‘radical translators’,developing a ‘translation manual’ – inother words, a recipe for going fromsentences in some entirely unfamiliarlanguage to sentences of our own lan-guage. To develop this manual wemay use only a narrow range of clues.In particular we may not rely on thetwo languages having a commonorigin and we certainly may not use adictionary. The only evidence we may

Sperber, Dan and Deirdre Wilson (1987).Relevance: Communication and Cogni-tion. Oxford: Blackwell.

Wilson, Deirdre and Dan Sperber (2004).‘Relevance theory’. In Laurence R.Horn and Gregory Ward (eds), TheHandbook of Pragmatics. Oxford:Blackwell. 607–32.

Further readingCarston, Robyn (2004). ‘Relevance theory

and the saying/implicating distinction’.In Laurence R. Horn and Gregory Ward(eds), The Handbook of PragmaticsOxford: Blackwell. 633–56.

Davis, Wayne A. (1998). Implicature:Intention, Convention, and Principle inthe Failure of Gricean Theory. Cam-bridge: Cambridge University Press.

Gazdar, Gerald (1979). Pragmatics: Impli-cature, Presupposition and LogicalForm. London: Academic Press.

Grandy, Richard E. and Richard Warner(eds) (1986). Philosophical Grounds ofRationality. Oxford: Clarendon Press.(See, in particular, ‘Paul Grice: a view ofhis work’ by the editors and the reply byGrice.)

Kasher, Asa (ed.) (1998). Pragmatics: Crit-ical Concepts. Six volumes. London:Routledge. (See in particular vol. IVwhich includes seventeen classicalpapers on implicature.)

Asa Kasher

INDETERMINACY

‘Indeterminacy’ in a linguistic contexthas various interpretations, but inmost senses it is taken to refer to theidea that linguists cannot make defi-nite pronouncements about the mean-

93

INDETERMINACY

part of the same whole as’ rather than‘is the same as’; the alternative manualwould then correctly predict assent.Quine is confident that, after comple-mentary juggling of different clauses,alternative but equivalent translationmanuals will always be available.Translation, he insists, is a matter offitting a complete manual to all thesentences of the other language (seeholism*); Quine treats as myth theview that successful translating is amatter of revealing the one true mean-ing of individual words and sentences.

Some critics have argued that it willalways be possible to isolate at mostone translation scheme compatiblewith assent/dissent patterns. There iscertainly more room for manoeuvrewith Quine’s ‘Gavagai’ example thanhe initially realised. In 1970 Quineexpressed regret at the focus on thisparticular example. He tried to show,quite generally, how his controversialindeterminacy thesis follows from awidely accepted underdeterminationthesis. According to this, empiricalevidence rarely forces any particularscientific theory upon us. His infer-ence from underdetermination oftheory to indeterminacy of translationhas failed to convince the majority ofcommentators. Other critics havequestioned whether indeterminacy oftranslation, even if correct, generatestrouble for the notion of meaning.Most target the behaviourism* builtinto his statement of the evidenceavailable to a radical translator.

Kripke’s 1982 indeterminacy puzzleis inspired by his reading of Wittgen-stein’s Philosophical Investigations(1953). To set up the puzzle, Kripkeuses, for illustrative purposes, theexpression ‘+’ ; or, in spoken English,

use is behavioural: dispositions on thepart of native speakers to assent ordissent in varying circumstances toutterances of sentences of their lan-guage. Quine limits himself to this evi-dence on the grounds that sentencemeanings, in so far as there are suchthings (and he ends up doubting thereare), must be constituted by trans -latability on these terms. Allowingother evidence would be presupposingrather than coming to understand thenature of a sentence’s meaning. Inde-terminacy of translation is the thesisthat, in the context of radical transla-tion, there will always be at least twoacceptable but incompatible transla-tion manuals available.

Quine gives an example. Supposethat, as a rabbit scurries by, a nativespeaker utters the one-word sentence,‘Gavagai’. You guess it means ‘Lo, arabbit’. Tests on future occasionsusing assent and dissent to this samesentence bear out your hypothesis.Before adopting this as ‘the’ correcttranslation, however, you must ruleout alternatives. But this cannot bedone. For example, why not take‘Gavagai’ to mean ‘Lo, undetachedpart of a rabbit’? In those few contextswhere the difference between this andyour translation shows up, Quineholds that it will be possible to tweakother parts of the manual to accom-modate the discrepancy. Imagine youpoint to one part of a rabbit, thenanother part of the same rabbit, andsolicit assent or dissent from ‘This gav-agai is the same as that gavagai’ – orrather, ‘Shiz gavagai sumo shaz gava-gai’, as it may be. Native-speakerassent here will not automatically ruleout the alternative translation manual.We could take ‘sumo’ to translate as ‘is

‘added to’. Suppose you are asked‘What is 68 + 57?’ Suppose further thatthis is a question you have never beforehad occasion to address (there willalways be some such question). Youanswer with confidence: ‘125’. You areassuming that the expression ‘+’ refershere to what it has always referredto, namely, the plus-function. But asceptic challenges you: ‘How do youknow that ‘+’ has always referred tothe plus function?’ Perhaps it hasalways referred to the quus (pro-nounced ‘qwus’) function, shown inFigure 1.

The only way you could ascertainwhich of the two functions ‘+’ referredto earlier is to apply the expression inthe context of figuring out the sum offifty-seven and sixty-eight. But byhypothesis you could not have donethis.

Kripke’s imagined sceptic has thusfar claimed to establish only that youdo not at present know whether ear-lier uses of ‘+’ referred to plus ratherthan quus; and hence that you do notknow if the correct answer to the

present question is ‘125’ rather than‘five’. This is bad enough. But thesceptic also denies the possibility ofknowing whether you meant plus by‘+’in the past, and so mean plus now.It is not a matter of ignorance as towhich you mean: there is simply nofact of the matter as to which youmeant in the past, or which you meannow.

The puzzle can be generalised invarious ways. First, it extends toexpressions other than ‘+’. Supposeyou ask yourself, right now, ‘Is this abook I am reading?’ Your answerdepends on what you mean by‘book’. Perhaps ‘book’ refers now, asin the past, to quooks, defined inFigure 2.

The conclusion we are invited toaccept is that whenever we come toapply a familiar word in a novel con-text, there is nothing in virtue ofwhich our application is correct orincorrect, since nothing in previoususes determines that the wordmeans one thing rather than another. Moreover, since every context of aword being used was novel once, thepuzzle applies to every use of anyword. Finally, the puzzle is not justabout communication. It applies evenwhen we silently ask ourselves a question, so it is a puzzle aboutthought itself, not just about theexchange of thoughts through spokenlanguage.

INDETERMINACY

94

true if x is a book not called Key Ideas in Linguistics andthe Philosophy of Language;

false otherwise.

x is a quook

Figure 2

Figure 1

5 if x = 68 and y = 57

x plus y otherwise

x quus y =

The quus and quook readingscannot be dismissed on the groundsthat ‘quus’ and ‘quook’ are defined interms of the more familiar ‘plus’ and‘book’, and are hence less simple.After all, it is equally possible to definethe latter in terms of the former. Nordoes it help to observe that, since weare disposed to reply ‘125’ rather than‘five’, we must mean plus rather thanquus. If what we mean by a word is aproduct of how we are disposed toapply it, it would be impossible ever tomake an erroneous application andthat would be absurd. Kripke himselftentatively offers what he calls a sceptical solution. This seeks to understand the role of meaning-state-ments within discursive practice with-out allowing that they assert anythingtrue or false. Rather, they show some-thing about the statement-maker’swillingness to use a term like ‘+’ or‘book’ in a particular way. Few haveembraced Kripke’s solution, but thereis no consensus over where his scepti-cal argument breaks down.

Quine’s and Kripke’s discussionsboth threaten the widespread assump-tion that words and sentences have aparticular meaning. A more recent lit-erature (albeit on a very old topic) alsocalls semantic determinacy into ques-tion. Suppose I were to ask you ‘Haveyou had breakfast?’. Am I asking youwhether you have had breakfast ever,or just whether you have had breakfastthis morning? It is tempting to appealto the context. But after setting asideformal indexicals (‘I’, ‘here’, ‘tomor-row’, and so on), context’s contribu-tion to utterance meaning is difficult tosystematise, even for ‘Have you hadbreakfast?’. To say that its contribu-tion is impossible to systematise would

be to say that there is no determinatefunction from sentences and contextsof utterance onto propositions*. Thisapparent indeterminacy lies at theheart of the question of how we oughtto distinguish between words’ contri-bution to an utterance’s meaning(broadly, semantics) and the utterancecontext’s contribution (broadly, prag-matics). See Recanati (2004) and Cap-pelen and Lepore and (2005) foropposing views.

We have seen that challenges todeterminate meaning can take differ-ent forms. The interest attaching tothese challenges often lies not inthe scepticism they generate aboutmeaning but in the resources devel-oped to meet them, and the claritythat emerges from reflecting on theassumptions that give rise to them.

Primary sourcesCappelen, H. and E. Lepore (2005). Insen-

sitive Semantics. Oxford: Blackwell.Kripke, S. (1982). Wittgenstein on Rules

and Private Language. Oxford: Black-well.

Quine, W. V. O. (1960). Word and Object.Cambridge, MA: MIT Press. Chapter 2.

Quine, W. V. O. (1970). ‘On the reasonsfor indeterminacy of translation’. Jour-nal of Philosophy, 67: 178–83.

Recanati, F. (2004). Literal Meaning. Cam-bridge: Cambridge University Press.

Wittgenstein, L. (1953). PhilosophicalInvestigations. Oxford: Blackwell.

Further readingMiller, A. (1998). Philosophy of Language.

London: UCL Press. Chapters 4–6.

Alex Barber

INDETERMINACY

95

96

(information about grammatical sen-tences), yet they successfully acquirelanguage.

From this basis one has to concludethat humans are genetically hard-wired for language: children are bornequipped with a universal grammar*(Chomsky 1981), containing informa-tion about linguistic universalsenabling them to form hypothesesabout the structure of the languagethey are learning. The proposed exis-tence of a critical period for language(typical language acquisition is notpossible after a certain age) and stud-ies of deaf children who sponta-neously develop sign language havebeen used as further evidence in sup-port of linguistic nativism.

Opponents of the Chomskyan viewclaim that the richness of the dataavailable to children is vastly underes-timated; the stimulus argument failsonly given the generative definition ofwhat language is. One of the strongestchallenges to nativism comes fromConnectionist psychology (Elman etal. 1996) where artificial neural net-works have been trained to reproducelinguistic behaviours such as recursion,previously thought to be possible onlythrough innate pre-programming.Research on innateness has spurredresearch with animals and in genetics.In 2001 Cecilia Lai and colleagues sug-gested that a mutation in the FOXP2gene is causally involved in languagedisorders. Despite widespread enthusi-asm in the popular press about the ‘lan-guage gene’, the exact role of FOXP2in relation to language development isfar from clear.

Although the innateness questionis still unresolved, interdisciplinaryresearch that straddles the gap

INNATENESS

The claim that some aspects of lin-guistic competence are geneticallyspecified rather than learnt throughexperience. This claim has been driv-ing research in generative linguisticsand language acquisition since the late1950s.

See also:Acceptability/Grammaticality;Continuity; Mentalism;Transformational-GenerativeGrammar; Universal GrammarKey Thinkers: Chomsky, Noam;Descartes, René; Plato

Noam Chomsky has proposed thathumans possess domain- and species-specific knowledge of the structure ofpossible languages, which enableshuman young to acquire languagewith speed, efficiency and uniformity.This view can be traced back to Carte-sian cognitivism and Platonic philoso-phy. Opponents claim that languageacquisition is innately constrained butonly by mechanisms that underlie gen-eral cognitive ability. In other words,it is currently uncontroversial thatlanguage acquisition is innately con-strained; what is the object of heateddebate is exactly what is innate.

The argument that has most force-fully been used in support of the lin-guistic innateness position is that thestimulus argument, in which childrenlearn language by experience, is seri-ously flawed (Chomsky 1980: 34).For example, since language is a com-plex system, it could only be acquiredthrough experience if negative evi-dence was available. Children onlyever have access to positive evidence

97

INTEGRATIONISM

challenge what they characterise as‘segregational’ linguistics, which treatslanguage as an autonomous object ofstudy separate from non-linguisticaspects of communication, and argueforcefully that the attempt to decon-textualise language leads to a pro-found distortion.

See also: Acceptability/Grammaticality; ConversationAnalysis; Creativity; (Critical)Discourse Analysis; Indeterminacy;Systemic-Functional Grammar;Structuralism;Transformational-GenerativeGrammarKey Thinkers: Chomsky, Noam;Firth, J. R.; Halliday, M. A. K.;Malinowski, Bronislaw; Saussure,Ferdinand de

The dominant tradition in Westernlinguistics over the twentieth centuryand into the twenty-first century hasinsisted on the autonomy of linguisticsas a science. This has meant definingand delimiting the subject in such away that language can be investigatedin isolation from other phenomenawhich might be thought to impinge onit. It is accepted that areas of contactwith the non-linguistic world may beworth exploring, through sociolin-guistics, speech act theory*, seman-tics, discourse analysis* and so on;but these are all seen as in some wayless ‘pure’ extensions from the investi-gation of the essential linguistic core,which is autonomous syntax. Integra-tionism, which grew from a seriesof books published by Roy Harrisfrom 1980 on, rejects this stance completely. Following Harris’s lead,integrationists have set out to demol-ish what they see as the suspect

between linguistics, anthropology,artificial intelligence, genetics and neu-roscience has opened up novel, excit-ing ways in which the question can beaddressed.

Primary sourcesChomsky, N. (1965). Aspects of the

Theory of Syntax. Cambridge, MA:MIT Press.

Chomsky N. (1980). Rules and Represen-tations. Oxford: Blackwell.

Chomsky, N. (1981). Lectures on Govern-ment and Binding: The Pisa Lectures.Holland: Foris Publications.

Further readingElman, J. L., E. A. Bates, M. H. Johnson,

A. Karmiloff-Smith, D. Parisi and K.Plunkett (1996). Rethinking Innateness:A Connectionist Perspective on Devel-opment. Cambridge, MA: MIT Press.

Hauser, M. D., N. Chomsky and W. T.Fitch (2002). ‘The faculty of language:What is it, who has it, and how did itevolve?’ Science 298: 1569–79.

Lai, C. S. L., S. E. Fisher, J. A. Hurst,F. Vargha-Khadem and A. P. Monaco(2001). ‘A forkhead-domain gene ismutated in severe speech and language’.Nature 413: 519–23.

Stavroula-Thaleia Kousta

INTEGRATIONISM

An approach to linguistics which isradically opposed to the assumptionson which much of mainstream linguis-tics has been based since the time ofFerdinand de Saussure, and insists thatlanguage must be seen as actionembedded in context. Integrationists

98

INTEGRATIONISM

obscurity and ambiguity* in order tomaintain the smooth flow of commu-nication. Speakers are often uncon-cerned with avoiding these apparenthindrances, secure in the knowledgethat the hearer will, all things beingequal, collaborate in making enoughsense of the utterances for communi-cation to take place satisfactorily. Avalid integrationist model of linguisticbehaviour cannot relegate this to ‘per-formance factors’ which do not affectthe underlying ‘competence’: it is akey distinguishing feature of com -munication which must be taken intoaccount.

The second point on which integra-tionists take issue with the mainstreamis the determinacy of the linguisticsign. If, as in the telementation model,ideas were transferred undistorted, thiswould have to rely on signs (words)having determinate meanings whichare known to, and accepted by, bothinteractants. But, as Harris and othershave stressed, if this were the case,lawyers would be completely unneces-sary, since much of their work involvesnegotiating over the precise applica-tion of laws to individual cases – thatis, exactly what the law ‘means’ in aparticular context. This seems odd,since legal language is designed to be asexact and determinate as possible. Ifindeterminacy* occurs here, it is evenmore likely to be inherent in everydayuses of language. For integrationists,interactants are engaged in a constantprocess of making sense of communi-cation in particular contexts. They candraw on their experience of previoussituations which they perceive as simi-lar, but these perceptions of similarityare not dependent on some kind ofcommunally held ‘dictionary’ which is

foundations on which the linguisticedifice has been constructed. Therehave been integrational studies ofareas such as fictional dialogue, irony,legal arguments and so on, but the pri-mary aim has been to force main-stream linguists to at least question, ifnot revise, many assumptions thatthey have taken for granted.

The integrationist critique high-lights three key interrelated problemswith the linguistic tradition foundedon Saussure’s work. The first is whatHarris in his 1981 book terms themyth of ‘telementation’. This is theimage (introduced in a famous dia-gram in Saussure 1916/1922 of twoheads with arrows passing betweenthem representing ideas being trans-ferred) of language as serving toconvey the thoughts of the speakerinto the mind of the hearer. The ideasare assumed to be transferred undis-torted, fully comprehensible and fullycomprehended. Integrationists arguethat, even when mainstream linguistsclaim publicly not to work with thismodel of communication, their prac-tice indicates that they do. A funda-mental flaw in the model is that thereis in fact no way of knowing to whatextent, or even whether, the meaningsunderstood by the hearer are thoseintended by the speaker. It alsorestricts ‘meaning’ to ‘ideation’, thusignoring interpersonal or other aspectsof meaning (compare systemic- functional grammar*, where theseother types of meaning are fully inte-grated in the model). Experienceshows that the telementation modeldoes not correspond to reality: veryoften hearers do not understand every-thing that they hear, and are contentto cope with degrees of vagueness,

99

INTEGRATIONISM

ers (as represented by the linguisthim/herself), and around this is aperiphery of variation in forms,including the usages of individuals orgroups (such as dialects), which havean uncertain relationship to the codebut which can be left to sociolinguistsand other scholars with ‘fringe’ inter-ests to deal with. Integrationistswould see this as merely ducking theissue (and, in the process, taking answeepingly arrogant view of society).For them, language is in a state of con-stant flux, geographically and histori-cally but also moment-by-moment asspeakers negotiate and create mean-ing in context; and generalisationsabout the characteristics of a particu-lar ‘language’ will necessarily excludethe variability that is fundamental tothe way that language operates.

A charge that has been levelledagainst integrationism is that itsrobust attacks on mainstream linguis-tics have not been balanced by anyclear programme or methodology toreplace current practices. MichaelToolan counters this by arguing thatsuch a complaint ‘reflects a way ofthinking deeply committed to the pre-vailing post-Saussurean cognitivist-mechanistic paradigm’ and that theterm integrational linguistics ‘names aprinciple rather than a method’ (1996:23). He notes that much of the workin ethnography, conversation analy-sis* and discourse analysis is in prac-tice congruent with integrationism,and the way forward will probablyinvolve adapting methods from thoseapproaches. A more theoreticalcounter-attack from a ‘segregationist’position (though they would reject thelabel) can be found in Borsley andNewmeyer (1997) who insist that

the final arbiter – they are the resultof general problem-solving abilitiesapplied (unconsciously and instanta-neously, because of years of practice)to seeing what function utterancesneed to serve in context in order to bean appropriate contribution to thecommunicative event. Signs are inher-ently indeterminate, and are only rendered determinate (to whateverdegree is judged necessary by the inter-actants) in particular contexts. Takingthis to the next level, segregationistapproaches are underpinned by theview that a language is a fixed code, sothat sentences can be deemed unequiv-ocally to be ‘grammatical’ (part of thecode) or ‘ungrammatical’ (not part ofthe code). However, for the integra-tionist this is an equally untenable dis-tortion: this could only be true of agrammar held in the mind of the ‘idealspeaker-hearer in a homogeneousspeech community’ – an imaginaryentity that does not, and cannot, exist.

The third point of contentioninvolves a claim which seems counter-intuitive, but which follows logicallyfrom the points above: ‘language’exists but ‘languages’ do not. Speakersdo, of course, find it useful to referto ‘languages’ such as English andFrench, but the integrationist sees thatas a convenient fiction which is unnec-essary and misleading as a basis forlinguistics. ‘A language’ implies deter-minate signs and a fixed code whichare, at least potentially, shared by allspeakers of that language. But whodecides which parts of ‘English’belong to the code and which do not?Some linguists would attempt to getround this question by arguing thatthere is a fixed core that is recognisedand accepted by all (educated) speak-

100

INTENTIONALITY

Muddle: Roy Harris and GenerativeGrammar’. In G. Wolf and N. Love (eds),Linguistics Inside Out. Amsterdam.Philadelphia: John Benjamins. 42–64.

Saussure, Ferdinand de (1922/1983).Course in General Linguistics. Secondedition. Trans. Roy Harris. London:Duckworth.

Toolan, Michael (1998). ‘A few words ontelementation’. In R. Harris and G.Wolf (eds), Integrational Linguistics.Oxford: Pergamon. 68–82.

Geoff Thompson

INTENTIONALITY

Directedness or aboutness; the prop-erty of mental states whereby they areabout, or directed towards, states ofaffairs in the world, typicallyexpressed in language through such‘intentional verbs’ as ‘believe’,‘desire’, ‘know’, and ‘intend’. Inten-tionality is an important concept inboth philosophy of mind and ethicaltheory as a way of recognising respon-sibility towards other beings, and indistinguishing humans and otherhigher animals from computers.

See also: Artificial Intelligence;Nonnatural Meaning;Phenomenology; Speech ActTheoryKey Thinkers: Husserl, Edmund;Searle, John

Intentionality is an important conceptin both continental and analytic phi-losophy. It was introduced by theGerman philosopher Franz Brentano,and developed by his pupil EdmundHusserl, for whom it is one of the

mainstream linguistics does not in factwork with a telementation model.This paper, and the response in Toolan(1998), give a flavour of the debate.It is hard to judge how far the in -tegrationist critique has altered thecourse of linguistics as its proponentswould wish. An increasing numberof approaches have independentlyarrived, by different routes, at rathersimilar conclusions, and key aspects ofthe integrationist position (such as theindeterminacy of the sign) have them-selves become mainstream, thoughgenerally under different labels. Inte-grationism can claim to have con-tributed to weakening the hegemonyof a particular view of linguistics, butthere is as yet little sign that it hasbrought about its demise.

Primary sourcesDavis, Hayley G. and Talbot. J. Taylor

(eds) (1990). Redefining Linguistics.London: Routledge.

Harris, Roy (1981). The Language Myth.London: Duckworth.

Harris, Roy (1998). Introduction to Inte-grational Linguistics. Oxford: Perga-mon.

Harris, Roy and George Wolf (1998).Integrational Linguistics: A FirstReader. Oxford: Pergamon.

Toolan, Michael (1996). Total Speech: AnIntegrational Linguistic Approach toLanguage. Durham/London: Duke Uni-versity Press.

Wolf, George and Nigel Love (eds) (1997).Linguistics Inside Out: Roy Harris andHis Critics. Amsterdam/Philadelphia:John Benjamins.

Further readingBorsley, Robert and Frederick, J.

Newmeyer (1997). ‘The Language

101

INTENTIONALITY

around his room, Husserl describesthe experience of perceiving a sheet ofwhite paper in a dim light, to draw adistinction between perceptual experi-ence and something perceived. Theconcrete experience of the paperappearing from a particular angle,with lack of clarity because of thelight, is a conscious experience, anddescription of that experience is phenomenological description, asopposed to a scientific description ofsomething perceived. A descriptionof consciousness and a descriptionof phenomenal experience are forHusserl one and the same thing: ‘thebasic character of intentionality [is]the property of being a “conscious-ness of something”’ (Husserl 1931:120). Moreover, ‘the pregnant mean-ing of the expression cogito’ (as inDescartes’ cogito ergo sum, ‘I thinktherefore I am’), is ‘I have conscious-ness of something’, or ‘I perform anact of consciousness’ (Husserl 1931:118). Thus for Husserl the Cartesiancogito is also an intentional act, andDescartes is unwittingly the truefather of intentionality.

More recently, intentionality hasbecome of interest to philosophersworking in the analytic tradition. Inhis ‘Minds, Brains and Programs’(1980), John Searle demonstrates thata programmed digital computercannot have cognitive states such asunderstanding. He does this through afamously elegant thought experimentcalled the ‘Chinese Room’. Searleimagines himself locked in a room andgiven a batch of Chinese writing. He isthen given a second batch of Chinesewriting, with a set of rules (in English)correlating the second batch with thefirst. Finally, he is given a third batch

foundational and essential concepts ofphenomenology. Intentionality is alsorecognised as Husserl’s most impor-tant influence on the analytic traditionof philosophy, in which it is especiallyassociated with the speech act theory*of John Searle.

The term ‘intentionality’ was firstused in its modern sense by Brentanoin his Psychology from an EmpiricalStandpoint (1973, first published1874), although the idea is ultimatelyderived from Aristotle’s concept of‘mental inexistence’, the notion thatwhen one thinks of an object, one hasan object ‘in mind’, but the object inmind does not ‘exist’ in the same waythat the object in the world exists.According to Brentano (1973: 88),‘every mental phenomenon is charac-terised by . . . the intentional (ormental) inexistence of an object, andwhat we might call . . . reference to acontent, direction toward anobject, . . . or immanent objectivity’.Brentano’s purpose here is to distin-guish the mental from the physical,and in turn to define the mental and tocharacterise mental states. Accordingto Brentano, all mental phenomenaare intentional; indeed, that some-thing is intentional defines it as amental phenomenon. This is what hasbecome known as an ‘irreducibilitythesis’: the mental cannot be reducedto the physical.

Brentano’s biggest influence was onHusserl, who in his Ideas (1913)extends the concept of intentionalityto all experiences of thought, feel-ing or will: intentionality and phe-nomenal experience are for Husserlinextricably linked. In a celebratedpastiche of René Descartes’ descrip-tion of the experience of looking

102

INTENTIONALITY

conversely’ (ibid.). Nevertheless, thespeech-acts distinction betweenpropositional content and illocution-ary force has a parallel in intentionalstates, where it is expressed as a dis-tinction between representative con-tent and psychological mode. Forexample, a sentence containing aspeech-act verb, such as ‘I predict thatyou will leave the room’, can beanalysed into its illocutionary force ofpredicting, and its propositional con-tent that you will leave the room. Asentence containing an intentionalverb, meanwhile, such as ‘I believethat you will leave the room’, can beanalysed into its psychological modeof belief, and its representative con-tent that you will leave the room. Fur-ther, just as there is a range of ‘fits’ ofspeech acts – they may be true or false,obeyed or disobeyed, kept or broken,for example, depending on whetherthe acts are assertive, directive, com-missive and so on – so there is a rangeof ‘fits’ of intentional sentences.Beliefs are true or false, while inten-tions are complied with or not com-plied with, and desires are fulfilled orunfulfilled. Moreover, an intentionalstate is a sincerity condition of its typeof speech act; ‘the performance of thespeech act is eo ipso an expression ofthe corresponding intentional state’(Searle 1982: 263), so that it soundsodd to assert, for example, ‘Congrat-ulations on winning the prize, but Iam not glad you won the prize’. Andfinally, speech acts and intentionalstates are linked by ‘conditions of sat-isfaction’ or ‘conditions of success’.Just as a statement is satisfied if andonly if it is true, a promise is satisfiedif and only if it is kept, and so on, so abelief is satisfied if and only if it is true,

of Chinese writing, with instructions,again in English, for correlating ele-ments of the third batch with the firsttwo batches. Unknown to Searle, whounderstands no Chinese, the peoplegiving him the batches of Chinesewriting call the first a ‘script’, thesecond a ‘story’, and the third ‘ques-tions’. Eventually Searle becomes sogood at following the instructions formanipulating the symbols that, fromthe point of view of someone outsidethe room, his answers to questions areindistinguishable from those of anative Chinese speaker. However, weremember that Searle still understandsno Chinese. Unlike the native Chinesespeaker, Searle in his room is merelymanipulating symbols. This showsthat although a computer might intheory be programmed with the sumtotal of all information there is to behad about the world, this would notmean that the computer understoodthe world. Understanding is an intentional phenomenon, along withperceiving, acting and learning. Com-puters merely manipulate symbols,and hence are incapable of beingattributed with these intentional phe-nomena.

Searle elaborates this position intwo subsequent essays, ‘What is anIntentional State?’ (1982) and Inten-tionality: An Essay in the Philosophyof Mind (1983), which connect inten-tional states with speech acts. Accord-ing to Searle, ‘intentional statesrepresent objects and states of affairsin exactly the same sense that speechacts represent objects and states ofaffairs’ (1982: 260). This is not to saythat intentionality is essentially lin-guistic: on the contrary, ‘language isderived from intentionality, and not

103

INTUITION

Searle, John R. (1983). Intentionality: AnEssay in the Philosophy of Mind. Cam-bridge: Cambridge University Press.

Further readingDreyfus, Hubert L. (1992). What Com-

puters Can’t Do: A Critique of ArtificialReason. Cambridge, MA and London:MIT Press.

Harney, Maurita J. (1984). Intentionality,Sense and the Mind. The Hague: Nijhoff.

Karl Simms

INTUITION

Speakers’ introspective judgementsabout aspects of their language. Asource of evidence for theories of lan-guage and cognition which has beenused extensively in recent decades,particularly under the influence ofNoam Chomsky.

See also: Acceptability/Grammaticality; Adequacy;Corpora; Creativity;Empiricism/Rationalism;Innateness; Mentalism;Transformational-GenerativeGrammar; Universal GrammarKey Thinkers: Austin, J. L.;Chomsky, Noam; Greenberg,Joseph; Katz, J. J.; Sinclair, John

Informant intuitions about languagehave been a major source of data forlinguists since the work of Chomskywas first published in the late 1950s.Chomsky argued that there was clearevidence that speakers share system-atic intuitions about their languagesand that these should be used asdata in studying language. Chomsky’s

a desire is satisfied if and only if it isfulfilled, and so on. In short, inten-tional states, like linguistic entities,are not things, but representations –they have logical properties, not onto-logical ones.

The concept of intentionality con-tinues to be of relevance today, since itis recognised as essential to the ascrip-tion of consciousness to beings. Thequestion of how intentionality is pro-duced, and what the criteria are forhaving a mental state, is thereforeimportant to computer scientists andphilosophers of cognitive science.Daniel Dennett (1987, 1993), forexample, takes a contrary view toSearle, being prepared to ascribeintentionality to computers and tovarious animals if they manifest whatis perceived to be intentional behav-iour.

Primary sourcesBrentano, Franz (1973). Psychology from

an Empirical Standpoint. Trans. A. C.Rancurello, D. B. Terrell and L. L.McAlister. London: Routledge.

Dennett, Daniel C. (1987). The Inten-tional Stance. Cambridge, MA andLondon: MIT Press.

Dennett, Daniel C. (1993). ConsciousnessExplained. London: Penguin.

Husserl, Edmund (1931). Ideas: GeneralIntroduction to Pure Phenomenology.Trans. W. R. Boyce Gibson. London:Allen & Unwin.

Searle, John R. (1980). ‘Minds, brains,and programs’. The Behavioral andBrain Sciences 3: 417–57.

Searle, John R. (1982). ‘What is an inten-tional state?’. In H. L. Dreyfus with H.Hall (ed.), Husserl, Intentionality, andCognitive Science. Cambridge, MA andLondon: MIT Press.

104

INTUITION

consistent intuitions about this and sothis constitutes good data about theirunderlying system of linguistic knowl-edge. A large amount of work in lin-guistics and psychology has beenbased on evidence from intuitions likethis.

From the outset, there was consid-erable discussion of issues associatedwith the use of intuitions as data.Chomsky himself pointed out that notall intuitions are as clear, or as clearlyshared, as the intuitions about exam-ples like (1)–(2), and he suggested thattheories should be based as much aspossible on the clearer cases. Otherlinguists have questioned the useful-ness of intuitions at all, arguing that itis better to look at naturally-occurringdata, such as data gathered in acorpus. Chomsky has always rejectedthis view, pointing out what he sees asserious problems with such naturally-occurring data. One problem is thatthere is an element of luck in whethera corpus will provide relevant exam-ples. Related to this is the fact that acorpus cannot provide negative evi-dence; it cannot show us what is notpossible in a language. We cannot, forexample, use corpus data to find outthat sequences like ‘It is eager toplease John’ (with the same reading as(2a)) or ‘Which student did Sylvia tellthe teacher who taught to go home’are not possible in English.

But the reasons why these are notpossible are important if we aim tounderstand the nature of our languagesystem. Another reason is that Chom-sky sees the object of study as an internalised system, which he origi-nally termed ‘competence’ (Chomsky1965) and distinguished from language in actual usage, which he

work led to major changes in linguis-tics and in psychology (often referredto as ‘the Chomskyan revolution’)and a large amount of work in these disciplines is now based on data fromintuitions.

Chomsky’s work changed linguis-tics in several ways. He introduced anew way of thinking about what wasthe object of study (mental grammars)and also about the methods to be usedin studying it. Before Chomsky’swork, the most common methodsused by linguists were ‘discovery pro-cedures’, which were techniques for‘discovering’ facts about the languagebeing studied. Chomsky argued thatthese techniques must be based on anumber of factors, including intuition,and also that speakers’ intuitions pro-vide evidence for the existence of aninternalised language system whichshould be the object of study for lin-guistics. He suggested that the bestmethod for studying the system was toexplore the intuitions. One famousdata set discussed by Chomsky con-cerns the examples in (1) and (2):

(1) (a) John is easy to please.(b) It is easy to please John.

(2) (a) John is eager to please.(b) It is eager to please John.

Chomsky pointed out that (2b) is nota possible way of expressing the samething as is expressed by (2a) (it is apossible utterance, of course, but onlyif we take ‘it’ to refer to a particularentity capable of being eager to pleasesomeone). There is no logical reasonwhy (2b) could not be a way of sayingthe same thing as (2a) by analogy withthe pattern in (1a) and (1b), but this isnot possible in English. Speakers have

105

INTUITION

sun and observe directly what is hap-pening there. But we can observe lightand heat coming from the sun andmake inferences about what must becausing what we observe. Similarly,we cannot look directly at the compe-tence of speakers but we can makeinferences based on their intuitionsabout particular examples.

Birdsong (1989) and Schütze(1996) considered methodologicalissues with the use of intuitions, andused the term ‘metalinguistic perform-ance’ to describe what speakers aredoing when they make these judge-ments. This term makes clearer whatthe judgements are and highlights thefact that this approach is using per-formance data to explore competence.

One major issue which all re -searchers need to address is that theintuitions of speakers vary. Whilemost speakers will agree on (1)–(2)above, a large number of examples arejudged differently by different speak-ers. One source for the variation injudgements is that all speakers haveslightly different language back-grounds and experiences resulting notjust in ‘dialectal’ variation (variationbased on where the speaker lives orhas lived) but also ‘idiolectal’ varia-tion reflecting the speaker’s individuallinguistic history. While there is sig-nificant agreement about exampleslike (1)–(2), there would be variationin responses to (3)–(4):

(3) He and Sylvia will go.(4) All of the staff are pleased to work

here.

In fact, Chomsky has suggested thatlanguage is ‘an individual phenome-non’ and that ‘no two individuals

termed ‘performance’. A large num -ber of irrelevant factors affect per-formance, for example interruptions,reformulations, the physical andmental state of the speaker. Anycorpus shows many examples of‘ungrammatical’ utterances involvingoverlap, false starts, hesitation, repeti-tion, reformulation, unfinished utter-ances and so on. Chomsky arguedthat what we should do is look forsophisticated ways of finding datawhich reveal facts about the underly-ing system of competence. The bestdata, he argued, come from the intu-itions of speakers.

What are the intuitions exactly?They are often referred to as ‘gram-maticality judgements’ but this isslightly misleading, particularly if wetake seriously the distinction between‘grammaticality’ and ‘acceptability’.Chomsky argues that mental gram-mars are systems of ‘tacit’ knowledge:speakers know that (2b) does notwork like (1b) but they cannot explainwhy. There are also a number of well-known examples which speakersreject but which are nevertheless con-sidered to be grammatical, and viceversa. Strictly speaking, then, speak-ers cannot be relied on to make judge-ments about ‘grammaticality’ but onlyabout ‘acceptability’, that is aboutwhether particular examples ‘soundOK’ or whether they can imaginehearing or saying them.

This means that intuitions are quitefar removed from the system of gram-mar which is the object of study.Chomsky (1980: 189–92) acknowl-edged this remoteness, making ananalogy with the study of thermonu-clear reactions inside the sun. Wecannot set up a laboratory inside the

106

LANGUAGE GAMES

data should be used instead. Others,like Schütze, envisage research carry-ing on using a range of kinds ofdata. For methodologically eclecticlinguists, the range of possiblesources of data includes elicitation(engaging speakers in conversationswhich are likely to elicit utterances ofthe forms being considered), inter-views, questionnaires, experimentsand corpora.

Primary sourcesChomsky, N. (1965). Aspects of the

Theory of Syntax. Cambridge, MA:MIT Press.

Schütze, C. T. (1996). The Empirical Baseof Linguistics: Grammaticality Judge-ments and Linguistic Methodology.Chicago: Chicago University Press.

Further readingBirdsong, D. (1989). Metalinguistic Per-

formance and Interlinguistic Compe-tence. New York: Springer-Verlag.

Chomsky, N. (1957). Syntactic Structures.The Hague: Mouton.

Chomsky, N. (1980). Rules and Represen-tations. Oxford: Basil Blackwell.

Chomsky, N. (1988). Language and Prob-lems of Knowledge: The Managua Lec-tures. Cambridge, MA: MIT Press.

Billy Clark

LANGUAGE GAMES

The idea that language use can becompared to a game, where conver-sational participants are the playersand the goal of their conversationcan be reached if they perform certain types of moves within the

share exactly the same language’(1988: 36).

Given this variation, linguists needto make idealisations about the lan-guage they are studying. They needeither to abstract away from the vari-ation and imagine a group of speakerswho do share a language, or to focuson describing the competence of justone speaker. In practice, for many lin-guists, the object of study is their ownmental grammar. But even individualspeakers do not always make the samejudgements about particular exam-ples. So linguists need to be careful notto accept too readily the judgementsof informants. A number of studieshave compared actual recorded usageof speakers with what they reportabout themselves and found signifi-cant differences.

Schütze (1996) makes a number ofsuggestions to help address these diffi-culties. He suggests that it is legitimateto continue to use intuitions as databut suggests that certain ‘precautions’should be followed and that linguistsshould aim to develop other tech-niques, in particular experimentalwork. As well as a number of specificsuggestions about materials and theprocedure of gathering intuitions, oneof the main precautions urged bySchütze is that subjects who provideintuitions should never themselves belinguists since there is a risk thattheir training will interfere with theirjudgements.

Chomsky has been resolute inrejecting alternative kinds of data,suggesting for example that gatheringcorpus data is unlikely to lead to significant insights. Other linguistssuggest that intuitions are unreliableand that ‘naturally-occurring’ corpus

107

LANGUAGE GAMES

work of game-theoretical semantics,an account closely related to PaulLorenzen’s dialogue games (Lorenzen1955, Lorenzen and Lorenz 1978). Inthis approach, semantic gamesbetween a verifier (the player who istrying to prove that the statement istrue) and a falsifier (proving that thestatement is false) are used to evaluatelogical formulas in a negation normalform. For example, a formula inwhich the outmost component is anexistential quantifier would be inter-preted as a game in which the verifierhas the first move, selecting a witnessto the truth of the formula. The veri-fier thus substitutes a variable with aproper name referring to an entitywhich substantiates the property orrelation contained in the formula.Similarly, the presence of a disjunctionprompts a verifier’s move in that shechooses which disjunct she will sup-port. Universal quantifier and con-junction, on the other hand, aremoves appointed to the falsifier. Aplayer is said to have a winning strat-egy (a proof) if she can win disregard-ing the choices of the opposing player.The semantic game can become morecomplex if we impose imperfectknowledge on the players, for exam-ple by making them ‘forget’ some oftheir earlier moves (imperfect recall).Hintikka’s game-theoretical semanticshas received important applications inlogic and on fragments of natural language.

Other recent developments in thefield of language and games have pri-marily been based on research in gametheory. Classical game theory describesthe behaviour of interacting rationalagents who take into account theirknowledge of each other’s preferences;

context of publicly-known rules.The philosophical outline was mostfamously provided by LudwigWittgenstein and further developedby a number of researchers in logic*,linguistics, artificial intelligence* andcomputation.

See also: Artificial Intelligence;Implicature; Speech Act TheoryKey Thinkers: Grice, H. P.; Lewis,David; Wittgenstein, Ludwig

The use of games as paradigms fordialectical situations involving a ques-tioner and an answerer goes back toAristotle. The idea that all conversa-tion can be understood in terms of alanguage game, however, is com-monly ascribed to Ludwig Wittgen-stein who utilises the term to refer tothe complex of a particular languageand the actions in which it isanchored. In Philosophical Investiga-tions, Wittgenstein strives to under-stand how we acquire the meanings ofwords. He notes that simply knowinghow to name a concept is not suffi-cient for us to be able to use the wordin a conversation; we must also knowabout the conventions or rules gov-erning its use. In this sense, languagecan be compared to a game of chess; ifwe were to teach someone how to playchess, it would not be sufficient to givethem the names of the chess figures,we would also have to explain to themhow the figures can move on the chessboard. Depending on the context, theuse of a word or a sentence that existsin the language – a move – can beviewed as more or less appropriate;non-existing language elements arenot moves at all.

Jakko Hintikka further extendsWittgenstein’s notion in his frame-

108

LANGUAGE GAMES

For example, in Prashant Parikh’sgame-theoretic model of com -munication (an extensive cooperativegame with partial information), thehearer disambiguates a speaker’sutterance by reasoning about itsalternatives and by taking intoaccount the cost associated with pro-ducing (and processing) complexexpressions.

According to other accounts, a con-versation can more appropriately bedescribed as a game of opposing inter-ests (a zero-sum game) or a bargaining(mixed-motive) game, in which theinterests of the players are only par-tially aligned (Merin 1994). Accordingto Arthur Merin’s Algebra of Elemen-tary Social Acts, the general purposeof the game is to establish the con-tent of the conversational commonground. The players must have strictlyopposing preferences regarding theadoption of the proposition under dis-cussion, since otherwise there wouldbe no point debating it. The commu-nicating agents also have different bar-gaining powers in the game, expressedby the dominance parameter. Differentstates of the game – actually, differenttypes of dialogue acts (claim, conces-sion, denial and retraction) – are char-acterised as sets of values for thedominance parameter, as well as theactor role, initiator role and preferencewith respect to the proposition under discussion.

As observed in a number of studies,game-theoretical approaches to com-munication offer the possibility toaccount for the construction of form-meaning pairs in terms of solutionconcepts. A well-known solution con-cept is the Nash Equilibrium – a pro-file with the players’ best response to

it makes use of tools employed by utility and decision theory. Evolution-ary game theory includes a descriptionof the changes in players’ strategies asthey dynamically adapt to their envi-ronment.

Various lines of research have beenbased on the idea that language canbe compared to a coordinationgame (Lewis 1969; Parikh 2001), inaccordance with H. P. Grice’s viewof a conversation as an interactionbetween rational agents with ashared goal. In a coordination game,players’ ‘payoffs’ are aligned – that is,unlike in other sorts of games exam-ined in the context of game theory,the players do not have opposinginterests. Based on their prior beliefsand preferences, the agents wouldinteract in a way that leads them to agood decision as to which actions toperform (problem-solving games) or,in a special case of problem solving,which propositions to consider to betrue (inquiries). The shared knowl-edge of the purpose of the game helpsthem to code and interpret meaningefficiently and successfully. Forexample, in David Lewis’s signallinggames, a signalling system (in fact, alanguage) evolves between a senderand a receiver if the receiver reactsappropriately to the sender’s signalby choosing an action that matchesthe state of the world. The speakerand the receiver can be compared totwo people travelling in a boat; therole of the sender is to signal possi-ble danger ahead; and the role ofthe receiver is to steer the boat accordingly. A signalling system willevolve if the interaction results in suc-cess above chance level, dependingon the strategy the agents employ.

109

LANGUAGE GAMES

metaphor rather than as a vehicle forprecise formalisation.

Primary sourcesBenz, A., G. Jaeger and R. Van Rooij

(2005). Game Theory and Pragmatics.Basingstoke: Palgrave Macmillan.

Hintikka, J. (1973). Logic, LanguageGames and Information: KantianThemes in the Philosophy of Logic.Oxford: Clarendon Press.

Lewis, D. (1969). Convention. Cam-bridge, MA: Harvard University Press.

Merin, A. (1994). Algebra of ElementarySocial Acts. In Tshatzidis, S. (ed), Foun-dations of Speech Act Theory. London:Routledge.

Parikh, P. (2001). The Use of Language.Stanford, CA: CSLI Publications.

Wittgenstein, L. (1958). PhilosophicalInvestigations. Trans. G. E. M.Anscombe. Oxford: Blackwell.

Further readingDekker, P. and R. Van Rooy (2000). ‘Bi-

directional optimality theory: an appli-cation of game theory’. Journal ofSemantics 17: 217–42.

Horn, L. R. (1984). ‘Towards a new tax-onomy for pragmatic inference: Q-based and R-based implicatures. In D.Schiffrin (ed.), Meaning, Form, and Usein Context. Washington: GeorgetownUniversity Press. 11–42.

Lorenzen P. (1955). Einführung in dieoperative Logik und Mathematik.Berlin: Springer.

Lorenzen, P. and K. Lorenz (1978). Dialo-gische Logik. Darmstadt: Wis-senschaftliche Buchgesellschaft.

Osborne, M. J. and A. Rubinstein (1994).A Course in Game Theory. Cambridge,MA: MIT Press.

Marie Nilsenová

the choices of other players, all ofwhom are trying to maximisetheir payoff. Dekker and Van Rooy(2000) propose to use Nash Equilib-ria to characterise winning candi-dates in two-dimensional optimalitytheory, an extension of optimalitytheory* employed on the semantics-pragmatics interface. Under thetwo-dimensional optimality theoreticinterpretation, a form-meaning pair isoptimal if it satisfies both the speaker’sand the hearer’s communicative goals,originally expressed by the Griceanmaxims and later subsumed under theQ- and R-principle (Horn 1984). Theprinciples capture the intuition thatthe speaker’s and the hearer’s attemptto minimise communicative effortleads to opposing preferences (for thespeaker, to say as little as possible; forthe hearer, to maximise the informa-tion value of the message). If thespeaker acted only on the basis of herown preferences, our language wouldconsist only of expressions involvinglow articulatory effort with a highdegree of homonymy. A division of thepragmatic labour results in the use ofunmarked forms in unmarked situa-tions and marked forms in marked sit-uations.

Other recent accounts of pragmaticphenomena in which the authorsmake use of game-theoretical resultsconcern indirect speech acts andunderspecification, credibility, ques-tion-answer pairs and grounding(Benz, Jaeger and van Rooij 2005).The game jargon has also been utilisedin computational analyses of dia-logue, where dialogue acts are charac-terised as ‘moves’ appropriate fordifferent game types. Here, the idea ofa language game serves as a loose

110

fore, in which LOTH is an alternativeto the theory of intentionality*.

LOTH is predicated on a number oftheses. The first is so-called ‘represen-tational realism’, or the holding of aRepresentational Theory of Mind(RTM). According to this theory,there is a unique, distinct, dedicatedpsychological relation for each propo-sitional attitude, and each thoughtincorporating an attitudinal verb is atoken of this mental type. Thinking isthoughts joined up – in other words,causal sequences of tokenings of thismental representation of the attitude,be it belief, desire, or whatever. It fol-lows that there is a strong rationalistbias in LOTH: thinking is defined asrational thought, the ratiocinativeprocess consisting of thoughts incausal sequence.

The second LOTH thesis is thatthese mental representations belongto a representational or symbolicsystem that is itself language-like. Thisis the part of the hypothesis that ismost closely influenced by NoamChomsky: roughly speaking, this sym-bolic system (often called ‘mentalese’)corresponds to what Chomsky calls‘competence’ or ‘deep structure’.Mentalese has its own grammar, andone which is, moreover, universal.LOTH, despite being concerned withpropositional attitudes rather thanpropositional content, is thus situatedin the tradition of linguistic thoughtstretching back to Port-Royal logic*,which sees the purpose of language asbeing predicative, that is to makepropositional statements about theworld which are demonstrably true orfalse. LOTH is incompatible with the‘meaning is use’ tradition of linguisticphilosophy promulgated by Ludwig

LANGUAGE OFTHOUGHT

The Language of Thought Hypothesisis the hypothesis that thinking andthought are conducted in a mentallanguage (‘mentalese’) that is innate,distinct from all natural languages,universal among all thinking beings,and physically realised in the brain.The language of thought is an im -portant concept in the attempt tofind a cognitive and neurologicalexplanation for language and con-sciousness.

See also: Innateness; Intentionality;Mentalism; Private Language;Propositional Attitudes;Propositions; Universal GrammarKey Thinkers: Chomsky, Noam;Fodor, Jerry

The Language of Thought Hypothesis(LOTH) was first postulated by JerryFodor in 1975. According to thishypothesis, thought and thinking aredone in a mental language, ‘men-talese’. It is a bold hypothesis insofaras this language is held to be innate,distinct from all natural languages,and physically realised in the brain.LOTH is derived from work on‘propositional attitudes’*, which aredescribed by sentences of the form‘Mary believes that pigs might fly’.The general form of such sentences is‘S As that P’, where S is the subjectwho holds the attitude, A is an ‘atti-tudinal’ verb such as ‘believe’, ‘desire’,‘hope’, ‘intend’ and P is any sentence.It may be seen that ‘attitudinal’ verbscoincide with what elsewhere in thephilosophy of mind are called ‘inten-tional’ verbs: there is a sense, there-

111

LANGUAGE OF THOUGHT

capacity. Innateness is necessary to thehypothesis in order to guard againstan infinite regress: if mentalese isinnate, then there is no need to postu-late a language that a being wouldneed in order to learn a language inorder to learn a language, and so on,since mentalese is not learned.

As well as being a nativist thesis,LOTH is also a naturalist one. Itattempts to answer the ‘hard question’of consciousness, which in Fodor’sown words is, ‘How could any-thing material have conscious states?’(1991: 285). LOTH answers thisquestion by collapsing mind intobrain: contrary, for example, to psy-choanalysis*, it has no need of psychicprocesses.

According to LOTH, propositionalthought cannot be accounted forexclusively through mental images, butrequires a syntax that combines sen-tences. It follows that, finally, LOTH ispredicated on a theory of ‘semanticcompleteness’; in other words, anypredicate that has meaning in anyspoken language is also expressible inthe language of thought.

It is the second LOTH thesis – thatmental representations belong to alanguage-like representational orsym bolic system, mentalese – which isthe most original feature of thehypothesis, and also the most con-tentious. LOTH holds that eachpropositional attitude (belief, desireand so on) is a function in the mathe-matical or computational sense.LOTH is often characterised as aspecies of ‘functionalist materialism’.According to this view, that a beingshould have a propositional attitude isenabled (some would say, caused) byphysical properties of that being,

Wittgenstein and developed by suchphilosophers of language as J. L.Austin, H. P. Grice and John Searle.Indeed, Fodor (1975: 55–97) explic-itly argues, contra Wittgenstein, that aprivate language* is not impossible,but necessary, and that mentalese isjust that private language.

A third LOTH thesis is that the lan-guage of thought is distinct – that is, itis not identical to any spoken lan-guage. This is again a strengthening ofa Chomskian position: in LOTH, anattempt is made not only at grammat-ical (syntactic) transformations fromdeep structure to surface structure,but also at semantic (representational)transformations from mentalese tospoken language.

Fourth, LOTH is a nativist orinnatist hypothesis: the language ofthought is held to be universal inhumans, and to be genetically deter-mined. LOTH is therefore also anempirical hypothesis: it is believedthat the language of thought can bediscovered as a material reality, as canthe biological mechanism that gener-ates it. While this is a development ofChomsky’s theory, again it is moreradical: for Chomsky, language isuniquely human, whereas for LOTH,it is at least partly possessed by allspecies which have cognitive pro -cesses. And while Chomsky holds thata being needs a language acquisitiondevice in order to learn language,LOTH holds that a being needs a lan-guage in order to learn a language –and that language is mentalese. Thelanguage of thought is the hypothesisused to explain how it is that certainanimals, and pre-linguistic infant chil-dren, can be seen to be thinking, eventhough they have no spoken language

112

LANGUAGE OF THOUGHT

despite the controversy it continues togenerate, is an important hypothesisin cognitive psychology and neuro-science. It is especially important as aconceptual tool in attempts to makethe findings of cognitive scienceaccord with folk psychology (people’scommonsensical beliefs about otherpeople’s propositional attitudes), asseen, for example, in the work ofDaniel Dennett.

Primary sourcesCarruthers, Peter (1996). Language,

Thought and Consciousness: An Essayin Philosophical Psychology. Cam-bridge: Cambridge University Press.

Dennett, Daniel C. (1987). ‘Reflections:the language of thought reconsid-ered’. In The Intentional Stance. Cam-bridge, MA and London: MIT Press,227–35.

Fodor, Jerry A. (1975). The Language ofThought. Hassocks: Harvester.

Fodor, Jerry A. (1987). Psychosemantics:The Problem of Meaning in the Philos-ophy of Mind. Cambridge, MA andLondon: MIT Press.

Fodor, Jerry A. (1990). A Theory of Con-tent and Other Essays. Cambridge, MAand London: MIT Press.

Fodor, Jerry A. (1991). ‘Replies’. In B.Loewer and G. Rey (eds), Meaning inMind: Fodor and his Critics. Oxford:Blackwell. 255–319.

Further readingCain, M. J. (2002). Fodor: Language,

Mind and Philosophy. Cambridge:Polity.

Loewer, Barry, and Georges Rey (eds)(1991). Meaning in Mind: Fodor and hisCritics. Oxford: Blackwell.

Karl Simms

especially neurological properties.These physical properties realisemental representations through per-forming computational operations,expressed as functions, on them.Moreover, each function is specificto a certain attitude: belief is ex -pressed through a belief-function,desire through a desire-function andso on. Thus the general form of thelanguage of thought, ‘S As that P’ maybe read as ‘S performs a computa-tional operation on P’. According tothis view, subjects do not directlybelieve that, for example, it is raining,but rather believe the proposition ‘it israining’ to be true.

This computational theory of lan-guage and mind implies that a com-puter could be constructed withpropositional attitudes, and that con-versely, a certain kind of computercould serve as a model of the humanmind-brain. This once again pitsLOTH against speech act theoristssuch as Searle, who in his ChineseRoom thought experiment sought todemonstrate that a digital computercould not have intentionality*.Searle’s answer to the question ofwhat intentionality ‘is’ was that it isnot a thing, but a logical property.LOTH’s answer to the same ques-tion, on the other hand, is that inten-tionality is reducible to a specificcomputational operation. In this view,‘intentionality’ becomes a misnomer:propositional attitudes are proposi-tional functions of mental processes.Interactions between thoughts andconcepts are merely computations,and ‘meaning’ is merely the internalinteraction of thought with concept.

As an empirical, naturalistic theoryof how the mind works, LOTH,

113

habits of any one member of a givenspeech community, Saussure sought toexamine language in general and toidentify the systems or rules and con-ventions according to which languagefunctions. Saussure’s views on lan-guage influenced linguistics during thetwentieth century, and his imprint canbe found in theoretical works dis-cussing phonetics, phonology, mor-phology, syntax, pragmatics andespecially semantics. Indeed, the dis-tinction between langue and paroleforms an important part of the theo-retical basis of structuralism*.

A popular lecturer at the Universityof Geneva, Saussure suggested ideasand concepts that fascinated his stu-dents, yet he did not personally writean authoritative guide to his views.Two colleagues of his, Charles Ballyand Albert Sechehaye, collected andedited student notes from three occa-sions during 1906–11 when he deliv-ered his lectures, publishing theassembled remarks under the titleCours de Linguistique Générale in1916. In the 1990s newly-edited ver-sions of student notes based on Saus-sure’s lectures, along with translationsinto English, appeared. At the begin-ning of the twenty-first century, thereis still disagreement about a numberof Saussure’s statements, and prob-lems surrounding the fragmentednature of some of the student noteshave not been fully resolved.

Through Cours de LinguistiqueGénérale, Saussure’s views concerninglanguage and the study of languagewere introduced to scholars through-out the world. Saussure rejectedthe nineteenth-century notion that linguistics should be primarily histor-ical and comparative, and disagreed

LANGUE/PAROLE

Referring to two aspects of languageexamined by Ferdinand de Saussureat the beginning of the twentieth cen-tury, langue denotes a system of inter-nalised, shared rules governing anational language’s vocabulary, grammar, and sound system; paroledesignates actual oral and writtencommunication by a member or mem-bers of a particular speech commu-nity. Saussure’s understanding of thenature of language and his belief thatscholarship should focus on investi-gating the abstract systematic princi-ples of language instead of researchingetymologies and language philosophyled to a revolution in the field of linguistics.

See also: Phoneme; Signs andSemiotics; Structuralism;Transformational-GenerativeGrammarKey Thinkers: Bakhtin, Mikhail;Barthes, Roland; Bloomfield,Leonard; Boas, Franz; Chomsky,Noam; Jakobson, Roman; Pike,Kenneth; Sapir, Edward; Saussure,Ferdinand de

The discussion concerning langue andparole was first suggested by Ferdi-nand de Saussure and popularised inhis Cours de Linguistique Générale(Course in General Linguistics), aseries of Saussure’s university lecturescollected by his students and publishedposthumously in 1916. Abandoningthe mindset, goals and objectives ofhistorical linguistics, Saussure advo-cated a synchronic examination of lan-guage. Not interested in studying aparticular language or the linguistic

114

LANGUE/PAROLE

speaker’s ingenuity or experimenta-tion. Moreover, he felt that an indi-vidual’s potential influence onlanguage is minimal. An individualmight create a memorable turn ofphrase, but that person is unable toaffect the overall structure or soundsystem of a given language. Finally,speakers can manipulate language inminor ways, but language imposes itsrules, order and possibilities on allspeakers without exception.

As part of their intuitive knowledgeof langue, members of a speech com-munity share possession and compre-hension of a body of signs (signes).According to Saussure, a sign consistsof two components: a signifier (signifi-ant) and a signified (signifié). Linguis-tic signs can encompass words, units ofgrammar, and expressions. The signi-fier is a sound or series of sounds, andthe signified is the meaning that thesounds represent. Saussure was carefulto note that signs are actually linked toclusters of meanings or associationsand not to specific things. For example,the word ‘house’ does not refer to aspecific object in the world but ratherto a concept involving images andassociations that speakers have inmind when they say or write the word.Furthermore, the connection betweenthe series of sounds and the cluster ofimages and emotions is arbitrary. Thewords ‘girl’, ‘Mädchen’, and ‘niña’might all refer to a female child, butthere is no direct connection betweenthe sounds of each word and the mean-ing. Even so, speakers form a strongconnection in their minds betweensounds and meaning.

Saussure stated that langage,the psychological and physiologicalfaculty to produce meaningful lan-

vigorously with the idea that substan-tial effort should be made to identify,codify and promote the standard formof any national language; he felt it wasmore worthwhile to focus attentionon describing language as it exists at agiven point in time, and believed thatthis activity could be conducted in animpartial manner.

For Saussure, three aspects of lan-guage could be potential objects ofconsideration in linguistic study, andhe used the French words langage,langue and parole to designate theseaspects. Langage refers to the anatom-ical ability and psychological need orurge of humans to create a system oflinguistic signs for expressing ideas.Langue represents a system of rules,usages, meanings and structures thatare products of the human ability tocreate language and are shared bymembers of a specific speech commu-nity. Parole is often equated withspeech. It is the concrete realisation ofa collectively-internalised system andalso reflects the personality, creativityand physiological capabilities of anindividual speaker.

Overall Saussure paid little atten-tion to langage, considering it the sub-ject matter of other fields of inquiry,and he regarded parole as too idio-syncratic. Instead, he believed that lin-guistics should study langue in orderto gain a picture of the comprehen-sive, complex, ordered assemblage ofsounds, words and syntactical units.Making use of a concept suggested inthe writings of the French sociologistand philosopher Émile Durkheim,Saussure viewed language as a socialfact. According to Saussure, languageis acquired through the socialisationprocess; it is not created through a

115

LANGUE/PAROLE

phonology. On occasion, agreement ordisagreement with Saussure’s beliefscan be traced back to an individual’spolitical and philosophical leanings.The Marxist linguist Mikhail Bakhtindisapproved of Saussure’s efforts todistinguish individual production oflanguage (parole) from collectiveknowledge and linguistic awareness(langue), a division that, to Bakhtin’sway of thinking, isolates an individualfrom society; he was much more infavour of a theory of language thatportrays speech as dependent on, and aproduct of, a specific social context.Stimulated by Saussure’s discussion ofthe sign and its two components – thesignified and the signifier – RolandBarthes investigated the contrastbetween the message of our speech andits form and articulation, and KennethPike advanced his system of tag-memics, a type of grammatical analysisdeveloped in the 1950s. Noam Chom-sky, too, responded to Saussure’s ideaswhen he transformed Saussure’s con-cepts of langage, langue and paroleinto ‘language capacity’, ‘competence’and ‘performance’, and achieved a newunderstanding of the Saussurean con-cepts. Twenty-first-century linguistsremain attracted to Saussure’s conceptof the dual nature of language and tohis theory of meaning.

Primary sourcesSaussure, Ferdinand de (1996). Premier

Cours de Linguistique Générale (1907):d’après les cahiers d’Albert Riedlinger.French ed. Eisuke Komatsu. Englished. and trans. George Wolf. Oxford:Pergamon.

Saussure, Ferdinand de (1997). DeuxièmeCours de Linguistique Générale (1908–1909): d’après les cahiers d’Albert

guage, does not manifest itself solelyin the creation of individual sounds,words or units of meaning, and hestressed that parole, individual com-munication within a speech commu-nity, does not take on the form of astring of unrelated utterances. Lan-gage becomes a reality in langue – andultimately in parole – through therules governing the use and organisa-tion of signs. These linguistic conven-tions are expressed in the form ofsyntagmatic and paradigmatic rules,two types of systems that enable lan-guage to convey messages by organis-ing and sequencing the buildingblocks of sound and meaning. Syntag-matic relationships refer to the limita-tions governing sequences of sounds,parts of words, and complete wordsoffered by a given national languageto create meaning. Paradigmatic rela-tionships concern the existence ofwords of similar meaning or gram-matical form that can substitute foreach other in a given context.

Saussure’s views concerning langueand parole, as well as his understand-ing of the purpose and goals of linguis-tics, have exerted immense influenceon linguists in Europe and NorthAmerica. Leonard Bloomfield, FranzBoas and Edward Sapir adopted Saus-sure’s method of objective, synchroniclanguage study as the basis for theirdescriptive analyses of various NorthAmerican Indian languages. Bloom-field also incorporated elements ofSaussure’s innovative teachings intohis writings, most notably Language(1933). Roman Jakobson and othermembers of the Prague School of Lin-guistics were inspired by Saussure asthey investigated sound systems anddeveloped theories of phonetics and

116

LINGUISTIC RELATIVITY

limits of the native speaker’s languageare the limits of his/her world. In the-ories of linguistic relativity, the vocab-ulary and linguistic structure of one’snative language limits or influencesone’s Weltanschauung or world view.

See also: Feminism; PoliticalCorrectness; StructuralismKey Thinkers: Humboldt, Wilhelmvon; Peirce, C. S.; Sapir, Edward;Saussure, Ferdinand de; Whorf,Benjamin Lee

The idea that the native languagecolours the speaker’s world view hasbeen in the forefront of linguistic sci-ence since the time of Wilhelm vonHumboldt, and has found advocatesand critics from various disciplines.The American philosopher CharlesSanders Peirce postulated that thesymbolic universe could only makesense through language, which hedefined as semiotic, a system of signs.Ferdinand de Saussure, in the Coursde Linguistique Générale, stated that:‘No ideas are established in advance,and nothing is distinct, before theintroduction of linguistic structure’(1916: 155). However, the notion oflinguistic relativity has largely becomeassociated with Benjamin Lee Whorf,who along with Edward Sapir, his lin-guistic mentor at Yale University, usedmodern linguistic concepts to advo-cate the position that language limits,or at least influences, the way a speechcommunity conceives of its worldview and reality.

Part of the groundwork for thishypothesis was laid by Whorf’s workas a fire insurance investigator. Duringhis career, he had the opportunity toanalyse many reports as to why firesbroke out in factories. He found that

Riedlinger et Charles Patois. French ed.Eisuke Komatsu. English ed. and trans.George Wolf. Oxford: Pergamon.

Saussure, Ferdinand de (1993). TroisièmeCours de Linguistique Générale (1910–1911): d’après les cahiers d’Émile Con-stantin. French ed. Eisuke Komatsu.English ed. and trans. Roy Harris.Oxford: Pergamon.

Saussure, Ferdinand de (1966). Course inGeneral Linguistics. Cours de Linguis-tique Générale. Trans. Wade Baskin.New York: McGraw-Hill. First Frenchedition 1916.

Further readingChomsky, Noam (1964). Current Issues in

Linguistic Theory. The Hague:Mouton.

Harris, Roy (1987). Reading Saussure: ACritical Commentary on the ‘Cours deLinguistique Générale’. London: Duck-worth.

Harris, Roy (2004). Saussure and HisInterpreters. Edinburgh: EdinburghUniversity Press.

Koerner, E. F. K. (1973). Ferdinand deSaussure: The Origin and Developmentof His Linguistic Thought in WesternStudies of Language. Amsterdam: Ben-jamins.

Sanders, Carol (ed.) (2004). The Cam-bridge Companion to Saussure. Cam-bridge: Cambridge University Press.

David V. Witkosky

LINGUISTICRELATIVITY

The idea central to the Sapir-Whorfhypothesis, which states that the

117

LINGUISTIC RELATIVITY

esis’. When viewed in terms of output,one could counter that a more appro-priate label would be the ‘Whorf-Sapir hypothesis’.

A rather interesting development inthis debate over giving credit wherecredit is due has been the attempt todisassociate Sapir from the hypothesisentirely. Desirous of preventing theimage of the great maestro Sapir frombeing tarnished by the taint of contro-versy, some, most notably Alfred L.Kroeber, have claimed that EdwardSapir’s views were not really pro-Whorfian. This viewpoint is not borneout by an examination of Sapir’s ownwritings. For example, as one canplainly see in the following passage,there can be no doubt that Sapir’sposition was fundamentally one thatequated language with culture andthinking. In Sapir’s words:

Language is a guide to ‘social real-ity’ . . . it powerfully conditions all ourthinking about social problems andprocesses. Human beings do not live inthe objective world alone, nor alone inthe world of social activity as ordinar-ily understood, but are very much atthe mercy of the particular languagewhich has become the medium ofexpression for their society. It is quitean illusion to imagine that one adjuststo reality essentially without the use oflanguage and specific problems ofcommunication or reflection . . . Notwo languages are ever sufficiently sim-ilar to be considered as representingthe same social reality . . . We see andhear and otherwise experience verylargely as we do because the languagehabits of our community predisposecertain choices of interpretation.(1929: 209)

workers would use extreme cautionwhen around ‘full’ drums of gasoline.Just as one would expect, workerswere careful not to smoke around‘full’ drums. Yet, these same workerswhen around ‘empty’ drums of gaso-line would often toss lit cigarettesnearby. This caused a violent explo-sion because an ‘empty’ drum(unknown to the smoker) still con-tained volatile gasoline vapour; an‘empty’ drum was really much moreof a threat than a ‘full’ one. Usingthese data, Whorf concluded that themeanings of certain words had aneffect on a person’s behaviour.

It was the research of both Sapirand Whorf into the grammatical sys-tems of many American Indian lan-guages, however, that proved to havethe greatest impact on this hypothesis.By predicating their insights into theinterrelationships of language andculture on what they had learned fromthe structures of these so-called‘exotic’ languages, the basic idea oflanguage shaping the perceptions ofits speakers and providing for them avehicle so that their experiences andemotions can be placed in significantcategories, was given scientific under-pinnings. Generally, Sapir is creditedwith giving the problem of establish-ing the link between language and cul-ture its initial formulation, continuingin the tradition of Johann GottfriedHerder and Humboldt. Whorf is hon-oured as the one who took this ideaand developed it into a bona fidehypothesis. Hence, the resultant sup-position is commonly given the desig-nation the ‘Whorfian hypothesis’.Pointing to Sapir’s pre-eminent statureas a linguist, some writers prefer theappellation the ‘Sapir-Whorf hypoth-

118

LINGUISTIC RELATIVITY

what I have called the ‘linguistic relativ-ity principle’, which means, in informalterms, that users of markedly differentgrammars are pointed by their gram-mars toward different types of observa-tions and different evaluations ofexternally similar acts of observation,and hence are not equivalent asobservers but must arrive at somewhatdifferent views of the world. (In Carroll1956: 221)

Perhaps the most incontrovertiblepiece of evidence in favour of linguis-tic relativity comes from the realm ofnumbers and numerals. There are lan-guages, such as Hottentot, also knownas Nama, which only have words forthe numerals for ‘one’ and ‘two’ and aword roughly translatable as ‘many’for three or more. A few languageshave no numerals whatsoever andtheir speakers are consequently unableto undertake even basic arithmetic.

In Whorf’s undated manuscript ‘ALinguistic Consideration of Thinkingin Primitive Communities’, it isshown that the problem of ‘thinking’by so-called ‘primitive’ peoples is‘approachable through linguistics’ (inCarroll 1956: 65). Further, as linguistshave come to fully appreciate onlyfairly recently, Whorf maintained that‘linguistics is essentially the quest ofMEANING’ (in Carroll 1956: 73).Example after example is given ofthings which are relatively easy to sayin Hopi but awkward or clumsy to sayin such Standard-Average-European(SAE) languages as English, Spanishand German. The term SAE was ofWhorf’s own invention.

In the classic ‘An American IndianModel of the Universe’, Whorf arguesthat since there is neither an explicit

There are really two different yetrelated versions of the Whorfianhypothesis, which is understandablewhen one considers that Whorf did allof his professional writing in therather short period between 1925 and1941. His ideas, quite naturally, werecontinuously developing. The strongversion of the hypothesis, which iscalled linguistic determinism, holdsthat language determines thinking, oras Stuart Chase writes in the forewordto Whorf’s collected works: ‘Allhigher levels of thinking are depend-ent on language’ (in Carroll 1956: vi).This position is most difficult todefend primarily because translationbetween one language and another ispossible, and ‘thinking’ can take placewithout language at all, as evidencedby fine art.

Mirroring Sapir’s thoughts as men-tioned above, Whorf notes in his 1940article ‘Science and Linguistic’:

We dissect nature along lines laid downby our native languages . . . We cutnature up, organize it into concepts,and ascribe significances as we do,largely because we are parties to anagreement to organize it in this way – anagreement that holds throughout ourspeech community and is codified in thepatterns of our language. (In Carroll1956: 213)

The milder version of the Whorfianhypothesis is labelled ‘linguistic rela-tivity’. This states that our native language influences our thoughts orperceptions. In fact, it was Whorfwho coined the phrase ‘linguistic relativity’. In the article ‘Linguisticsas an Exact Science’, Whorf com-mented:

119

LINGUISTICVARIABLE

A descriptive unit defined as a cate-gory of two or more linguistic alter-natives co-varying with one another inone of three ways: in a categorical way(the variation always occurs given cer-tain circumstances); in a quasi- predictable or probabilistic way (inline, for example, with another lin-guistic variable or a social variable);or in an apparently unpredictable,random way (‘free variation’). Thecentral idea is that there are ‘multipleways of saying the same thing’,though debate continues as towhether two different linguistic struc-tures are ever exactly equivalent infunction and/or meaning, and aboutthe extent to which individual lan-guage users are able to exercise con-scious choice over which alternativeto select from the range available.

See also: Acceptability/Grammaticality; ConversationAnalysis; Corpora; (Critical)Discourse Analysis; Descriptivism;Emic/Etic;Empiricism/Rationalism; Intuition;Speech Act Theory; Type/TokenKey Thinkers: Cameron, Deborah;Halliday, M. A. K.; Pike, Kenneth;Labov, William; Milroy, Lesley;Sapir, Edward

Variables operate at all levels of lin-guistic structure. Grammatical vari-ables capture some aspect ofoptionality in the occurrence of mor-phological or syntactic forms. Forinstance, the use of double modal constructions like ‘might could’ (+main verb) is not obligatory in those

nor an implicit reference to time in theHopi language and thus no tenses forits verbs, according to the Hopi viewof the world ‘time disappears andspace is altered’ (in Carroll 1956: 58).Whorf’s basic contention is that Hopimetaphysics, which underlies its cog-nition, is different from our own. Inother words, the Hopi calibrate theworld differently because their lan-guage defines experience differentlyfor them.

As more information has surfacedabout Hopi, some of Whorf’s specificgrammatical points have not with-stood the test of time. Although mostlinguists today dismiss many ofWhorf’s claims, it should be pointedout that Whorf’s basic idea of linguis-tic relativity – that the structure andvocabulary of one’s mother tongueinfluences one’s world view –although not proven to be correct,also has not been proven to be wrong.

Primary sourcesCarroll, John B. (ed.) (1956). Language,

Thought and Reality: Selected Writingsof Benjamin Lee Whorf. Cambridge,MA: MIT Press.

Mandelbaum, David G. (ed.) (1949).Selected Writings of Edward Sapir inLanguage, Culture, and Personality.Berkeley: University of California Press.

Sapir, Edward (1929). ‘The status of linguistics as a science’. Language 5:207–14.

Further readingSaussure, Ferdinand de (1916/1966).

Course in General Linguistics (Cours deLinguistique Générale). Trans. WadeBaskin. New York: McGraw-Hill.

Alan S. Kaye

120

LOGIC

machinery of variationist sociolin-guistics, enabling quantitative analy-sis of language data.

Primary sourcesWatt, Dominic (2007). ‘Variation and the

variable’. In C. Llamas, P. Stockwelland L. Mullany (eds), The RoutledgeCompanion to Sociolinguistics.London: Routledge. 3–11.

Wolfram, Walt (1991). ‘The linguisticvariable: fact and fantasy’. AmericanSpeech. Vol. 66:1, 22–32.

Further readingCornips, Leonie and Corrigan, Karen

(eds) (2005). Syntax and Variation: Rec-onciling the Biological and the Social.Amsterdam: John Benjamins.

Guy, Gregory (1997). ‘Competence, per-formance and the generative grammarof variation’. In F. Hinskens, R. vanHout and L. Wetzels (eds), Variation,Change and Phonological Theory. Ams-terdam: John Benjamins. 125–43.

Dominic Watt

LOGIC

Logic is the study of argument. Logicis central to philosophy, linguisticsand many other fields.

See also: Logical Form; Port-RoyalLogic; Truth ValueKey Thinkers: Aristotle; Frege,Gottlob; Russell, Bertrand

An argument is an attempt to per-suade using reasoning. Arguments arecomposed of statements. As standard,each statement has a truth value*: it iseither true or false, but not both. Nor-

varieties of English which permitit. The study of lexical variables, traditionally the province of nine-teenth-and twentieth-century dialectgeographers, involves examination ofalternative words or phrases for thesame object or concept in a language’sdifferent dialects or sociolects. Dis-course variables, such as tag questionsand the focus marker ‘like’, are moredifficult to handle given that it is notalways clear that utterances in whichthey occur can necessarily be viewedas functionally equivalent to those inwhich they could occur but do not.Problems involving incomplete syn-onymy of syntactic structures, dis-course variables and lexical variablestend not to arise with phonologicalvariables, as the equivalence of twoalternative pronunciations is generallyunambiguous. For example the use of[�] rather than [w] in the word‘which’ indicates a difference intwo individuals’ geographical and/orsocial origins, rather than a differencein the word’s meaning.

Knowing that there are differentways of expressing the ‘same thing’ isalmost certainly an intrinsic compo-nent of untrained native speakers’knowledge of their language(s).Awareness of the variable’s utility as ameans of formally accounting for lin-guistic alternations is apparent in theSanskrit grammar of Pānini (seventhto fourth centuries bc) which containsvariable rules allowing for differingoutputs. The notion is implicit in thehistorical linguistics and dialectologyof more recent centuries, for examplein the work of William Jones ongenetic affiliations between Indo-European languages, and it forms acentral element of the theoretical

being valid.) To appreciate that it ispossible, rather than actual, truthvalues that matter to validity, considerthe following arguments embodying 2and 4:

Argument 2*

P1 All US presidents are communists.P2 George W. Bush is a US president.C George W. Bush is a communist.

Argument 4*

P1 All US presidents before 2007 weremale.

P2 Eugene McCarthy was a US presi-dent before 2007.

C So, Eugene McCarthy was male.

In 2*, P1 is false and C is false. How-ever, if we assume, for the purposes oftesting for validity, that P1 and P2 areboth true, then we cannot deny Cwithout being committed to a contra-diction. So, the argument is valid: if itspremises were all true, then its con-clusion would be true. In 4*, P2 isfalse. If the premises of 4* were alltrue, then the conclusion wouldhave to be true as well, on pain of con-tradiction. So again we have a valid argument.

mally, an argument has at least onepremise and exactly one main conclu-sion. The main conclusion of an argu-ment is the statement it ultimatelyaims to support. The premises are thestatements that are meant to supportthe conclusion. An inference is a stepin reasoning.

Logic is primarily concerned withdeductive arguments. These aim to bevalid. An argument is valid if and onlyif the truth of the premises would guar-antee the truth of the conclusion. If itis consistent to assert an argument’spremises and deny its conclusion, theargument is invalid. The validity orinvalidity of an argument does notdepend on the actual truth values of itsparts: it depends on their possibletruth values. As Figure 3 shows, fourpermutations of truth-values are pos-sible.

The only permutation of truthvalues that debars an argument frombeing valid is 3. This is because validinferences are truth-preserving: ifthe inference from a given set of premises to a given conclusion isvalid, then if the premises are alltrue, the conclusion must also betrue. (Embodying 1, 2 or 4 does notmake an argument valid, but nordoes it disqualify the argument from

1 3

All the premises are true. All the premises are true.The conclusion is true. The conclusion is false.

2 4

At least one premise is false. At least one premise is false.The conclusion is false. The conclusion is true.

Figure 3

121

LOGIC

122

LOGIC

Argument FEvery human is mortal.Every philosopher is human.So, every philosopher is mortal.

By using term variables, Aristotle isable to discuss the logical forms of var-ious arguments in abstraction from thegeneral terms (for example, ‘human’,‘mortal’, ‘philosopher’) those argu-ments employ in natural language.Formal logic goes beyond the informalmethod of testing for validity. Theformal logician codifies deductive rea-soning to help distinguish valid argu-ment forms from invalid ones.Aristotle is a formal logician in oursense: the term variables he usesenable him to discern various patternsof inference valid by virtue of theirstructures. As well as being a termlogic, Aristotle’s logic is syllogistic. Asyllogism is an argument with twopremises and a conclusion, in whichboth premises and the conclusion aregeneral sentences. The sorts of generalsentence Aristotle is concerned withare what he calls ‘universal’ and ‘par-ticular’ sentences (De Interpretatione,17a–17b): sentences that use expres-sions equivalent to ‘all’, ‘some’ and‘no’. The following types of generalsentence can feature in a syllogism:

Every A is B. (Universal affirmative)No A is B. (Universal negative)Some A is B. (Particular affirmative)Some A is not B. (Particular negative).

The Stoics were also ancient pioneersof logic. Chrysippus (c. 280–c. 207BC) was the most important Stoic logi-cian. He regarded as valid such formsas ‘If the first, then the second; but thefirst; therefore the second’ and ‘If the

To assess an argument for validityby informal means, first suppose thatall its premises are true. Then askwhether, under this supposition, theconclusion must also, on pain of contradiction, be considered true aswell. If so, the argument is valid. Ifnot, it is invalid. Consider the follow-ing argument:

Argument T

P1 If the war in Iraq was a war for oil,then it was an unjust war.

P2 The war in Iraq was not a war foroil.

C The war in Iraq was not an unjustwar.

Suppose, for the test, that P1 and P2of Argument T are both true. Wouldthis then guarantee that C is true? No.P1 does not assert that the only way inwhich the war could have been unjustis by being a war for oil. Nothing inthe content of the premises rules outthat the war was unjust by virtue ofsome other factor, so the conclusiondoes not follow.

Aristotle founded the discipline oflogic. He defines a proposition as ‘astatement denying or affirming some-thing of something’ (Prior Analytics,24a). Propositions* are of subject-predicate form, where a subject termpicks something out and a predicateterm says something about it. Aris-totle uses letters to serve as term vari-ables, that is, to stand proxy for anyterms whatever. For example, ‘If Apredicated of every B, B every C, A bepredicated of every C’ (Prior Analyt-ics, 26a) depicts a valid argumentform. Here is an instance of thatform:

123

LOGIC

‘logical’ and the ‘non-logical’ vocabu-lary of the lexicon. Let us consider alanguage of propositional logic, called‘PL’ (after Tomassi 1999). The lexiconof PL consists of:

i. The sentence letters: ‘P’, ‘Q’, ‘R’,and so on, which symbolise atomicsentences – that is sentences with noother sentences as parts.

ii. The propositional operators: ‘~’,‘&’, ‘v’, ‘→’, ‘↔’, at least roughlyequivalent to ‘it is not the casethat . . .’, ‘. . . and . . .’, ‘. . . or . . .’,‘If . . . then . . .’, and ‘. . . if andonly if . . .’.

iii. The parentheses: ‘(’and ‘)’, used asa type of punctuation.

A PL formula is any string of symbolsfrom the lexicon of PL. A well-formedformula (WFF) of PL is a formula con-structed in observance of the follow-ing formation rules:

1. Every sentence letter is a WFF.2. Prefixing a WFF with ‘~’ gives a

WFF.3. For any WFFs containing A B, put-

ting them either side of ‘&’ or ‘v’ or‘→’ or ‘↔’ and putting parenthesesround the resulting formula gives aWFF.

4. Nothing else is a WFF.

The propositional operators are the log-ical vocabulary of PL. In propositionallogic, the quantifiers ‘some’, ‘all’ and‘no’ (and equivalent expressions) arenot treated as logical vocabulary. Thus,not every valid argument is proposi-tionally valid. For example, the validityof a valid syllogism, like Argument Fabove, depends on expressions otherthan the propositional operators.

first, then the second; but not thesecond; therefore not the first’ (seeKneale and Kneale 1962: 162–3).‘The first’ and ‘the second’ here areplace-markers for complete ‘assert-ibles’. Assertibles are the meanings ofdeclarative sentences (see Bobzien2003). Stoic logic is a type of proposi-tional (or ‘sentential’) logic. TheStoics put what are now called the‘logical operators’ of propositionallogic at the centre of their investiga-tions. Their English equivalents are‘If . . . then . . .’, ‘ . . . or . . .’, ‘ . . . and. . . ‘ and ‘It is not the case that. . .’.These operators are used to formcomplex statements: that is, state-ments with other statements as parts.

In symbolic logic, artificial, ‘formal’languages are constructed and em -ployed, in contrast with the mixture ofartificial symbols and natural lan-guage used by earlier formal logicians.Symbolic logic in this sense developedwith the work of Gottlob Frege and ofBertrand Russell and A. N. White-head, following earlier nineteenth-century innovations. Frege’s work isof unparalleled influence and impor-tance.

Among Frege’s most valuable con-tributions to logic were the inven-tions of modern predicate logic (alsoknown as ‘quantificational logic’ and‘predicate calculus’) and the firstformal system, both originallyincluded in his Begriffsschrift (1879).While Frege’s notation is obsolete,contemporary logic is built upon hispredicate logic and his systematisa-tion of the notion of proof.

A formal language consists of a lex-icon (a set of symbols) and a syntax(a set of rules for using the symbols).Logicians distinguish between the

124

LOGIC

Argument A

Socrates is a philosopher.Therefore, someone is a philosopher.

To translate ‘Socrates is a philosopher’into propositional logic, we assign asentence letter, such as ‘P’ to the sentence. Since ‘Someone is a philoso-pher’ is a different sentence, we assignit a different sentence letter, such as‘Q’. With these assignments, Argu-ment A comes out as:

Argument A*

Premise PConclusion Q

The problem with this is that no logi-cal form is revealed, since there are nological operators in play. So, A*cannot be an instance of a valid logicalform of PL and cannot capture theintuitive validity we assign to A. Pred-icate logic solves this problem. Logicalform within atomic sentences isrevealed. The way A is dealt with inQL is as follows.

First, we specify a domain of quan-tification, D. This is the set of entitieswe are quantifying over. Then wespecify the meanings of the non-logi-cal vocabulary we are going to use intranslating from English to QL. So, wecan begin to translate Argument A byspecifying the following.

D{human beings}

a SocratesF . . . is a philosopher

We can now translate the premise ofArgument A as:

Fa

After Aristotle and Frege, predicatelogic treats these quantifiers as logicalvocabulary. A language of predicatelogic both retains and supplements thelogical and non-logical vocabulary ofa language of propositional logic. Letus specify such a language of predicatelogic called ‘QL’ (for ‘QuantificationalLogic’) broadly after Tomassi (1999).The lexicon of QL consists of the lex-icon of PL (that is, i–iii above) plus:

iv. Individual constants: ‘a’, ‘b’, ‘c’, andso on. These are lower-case lettersfrom the beginning of the alphabet,used to symbolise proper names(such as ‘Socrates’, ‘London’).

v. Predicate letters: ‘F’, ‘R’, and soon, used to symbolise predicativeexpressions (such as ‘. . . is a man’,‘. . . loves . . .’).

vi. The quantifiers ‘�’ (the existentialquantifier, for ‘some’) and ‘�’ (theuniversal quantifier, for ‘all’).

vii. Individual variables: ‘x’, ‘y’, ‘z’,and so on.

viii. The identity sign: ‘=’.ix. The brackets: ‘[’, ‘]’, which are

another form of punctuation.

Though formation rules are specifi-able for QL, let us instead note twofeatures of QL. First, its logical vocab-ulary includes the logical operators ofPL plus the quantifiers, the individualvariables and the identity sign.Second, the non-logical vocabulary ofQL, specifically the individual con-stants and the predicate letters, allowsus to formalise, at a sub-sententiallevel, sentences that do not containany propositional operators. Bothpoints can be illustrated using a validargument that is neither proposition-ally nor syllogistically valid:

125

LOGIC

tional logic is first-order when it quantifies over individuals but notover properties. ‘Higher-order’ sys-tems quantify over properties. Suchsystems include predicate variablesamong their vocabulary. The bearingof the first-order/higher-order distinc-tion cannot be explored here.

The term ‘classical’ in ‘classicallogic’ does not refer to the ancientworld, but to some basic featuresshared by all classical systems. Theseinclude, among others, adoption ofthe principle of bivalence (see theentry on truth value*) and an accountof logical consequence according towhich a given conclusion followsfrom a given set of premises if andonly if the negation of the conclusionis inconsistent with the premises.Non-classical logic is obtainable byextending or revising classical logic.

What is it to extend classical logic?A system of logic S* is an extension ofa system of logic S if S* supplementsthe language and derivational appara-tus of S. By supplementing the lan-guage of S, we mean keeping all thesymbols and formation rules of S andadding further symbols and formationrules. The inference rules of S* willalso supplement those of S. Thus,every S-argument will be S*-valid, butnot every S*-valid argument will beS-valid. But why extend classicallogic? The usual motive is that classi-cal logic does not capture enoughvalidity. For example, classical logicprovides us with no formal means ofcounting as valid such inferences as:

Argument H

It is necessary that Socrates is human.Therefore, it is not impossible thatSocrates is human.

We read this as ‘a is F’. By convention,we put a predicate letter before aname to which it applies.

The existential quantifier ‘�’ andthe universal quantifier ‘�’ are used,with the individual variables, toexpress ‘some’ and ‘all’. Thus, we cantranslate the conclusion of ArgumentA as follows:

�x[Fx]

We read this as ‘There is an x that x isF’, or ‘For some x, x is F’. (‘x’ is anindividual variable.) So, the transla-tion of Argument A is:

Argument A**

Premise FaConclusion �x[Fx]

The validity of A** (and thus of A) isformally demonstrable in predicatelogic. (For example, the argument canbe proven by using an inference rulecalled ‘Existential Introduction’.)Modern predicate logic has greatlysuperior expressive power to its pre -decessors: it captures all the logical formthey could capture and logical form thateluded them. This is a legacy from Frege(see Noonan 2001, ch. 2).

Frege also presented the first formalsystems of propositional and predi-cate logic. Minimally, a formal systemconsists of a formal language plus amethod of proof. The latter is a pro-cedure for demonstrating the validityof valid arguments in a formal lan-guage. In working with this idea of aformal system, contemporary logi-cians follow Frege.

The sort of logic normally studiedon an introductory course is first-orderclassical logic. A system of quantifica-

126

LOGIC

Ernest is brave and Ernest is not brave.So Ernest is a mountaineer.

The premise here is a contradiction, soit cannot be true. So, we cannot be ina situation in which the premise is trueand the conclusion false. Thus we havea classically valid argument. Logiciansin the field hold that arguments likeRead’s example show that the classicalaccount of validity over-generates.

Linguists and philosophers of lan-guage investigate the relationshipbetween logic and natural language,asking, for example, about the rela-tionship between logical and gram-matical form and the extent to whichsemantic theory for natural languagecan be modelled on the formal seman-tics for logic. Logic is also interestingfor linguists because the syntax of aformal language is finitely specifiableby reference to a set of recursive for-mation rules (for example, the PL formation rules set out above). Ananalogous set of rules for a naturallanguage would be explanatory withrespect to a speaker’s capacity to formsentences never previously encoun-tered.

Analytic philosophy* of language,inaugurated by Frege and Russell, haslogic at its heart, and some of thefoundational works of modern symbolic logic were its impetus. In lin-guistics, formal semanticists and con-temporary syntactic theorists use logicin their analyses of natural language.A solid grounding in logic goes a longway for the student of linguistics orthe philosophy of language.

Primary sourcesAristotle. De Interpretatione. Trans. J. L.

Ackrill. In Jonathan Barnes (ed.)

Supplementing the vocabulary andrules of inference of a classical systemwith logical operators standing for‘It is necessary that’ and/or ‘It is pos-sible that’, enables the developmentof systems of modal logic. A formalsystem S is said to ‘under-generate’relative to natural language if thereare valid arguments of natural lan-guage that are not S-valid. Modallogicians hold that, in failing for-mally to account for the validity ofsuch arguments as H, classical logicunder-generates.

What is it to revise classical logicand why do so? A system of logic S*is a revision of a classical system oflogic S if and only if S* discards oneor more of the inference rules of S. Aformal system is said to ‘over- generate’ if there are arguments thatought to be considered invalid butwhich come out as valid within thesystem. Some logicians object to theclassical account of logical conse-quence on the basis that it results inover-generation. This objection some-times stems, as in the case of intu-itionist logic, from rejection of aprinciple of classical semantics (seethe entry on truth value). If there areclassically valid arguments whichought not to be considered valid at all,then at least one of the inference rulesof classical logic must be rejected orrestricted.

Revisions to classical logic includeintuitionist, relevant, fuzzy anddialethic logics. The relevant logician,for example, notes a seemingly oddfeature of classical validity: that fromcontradictory premises any conclu-sion whatever can validly be inferred.Read (1995: 55) gives the followingexample:

127

LOGICAL FORM

why it logically entails other sentencesthat it entails and why it is entailed byother sentences that logically entail it.Questions about logical form areimportant to the characterisation oflogic and to its relationship with nat-ural language.

See also: Definite Descriptions;Logic; Logical PositivismKey Thinkers: Aristotle; Chomsky,Noam; Russell, Bertrand

Bertrand Russell aimed to put a con-ception of logical form at the centre ofthe practice of philosophy. Logicalform in this sense is distinct from thenotion called ‘LF’ by linguists afterNoam Chomsky. The investigation oflogical form begins with the inquiryinto deductive reasoning initiated byancient logicians. For example, Aris-totle noted that any argument that isan instance of the pattern below isdeductively valid:

Some As are Bs.All Bs are Cs.So, some As are Cs.

Statements themselves have logicalforms: ‘Some philosophers are logi-cians’ is an instance of the pattern‘Some As are Bs’.

With the development of modernsymbolic logic*, a view arose thatgrammatical form can be misleadingas to logical form. For example, onRussell’s account, sentences that usedefinite descriptions* have logical fea-tures not evident at the level of surfacesyntax. In ‘The present King of Franceis bald’, none of the words is a logicaloperator. Russell claims, however,that the sentence is to be analysed asemploying covert logical operators.

(1984), The Complete Works of Aris-totle, vol. 1. Princeton: Princeton Uni-versity Press. 25–38.

Aristotle. Prior Analytics. Trans. A. J.Jenkinson. In Jonathan Barnes (ed.)(1984), The Complete Works of Aris-totle, vol. 1. Princeton: Princeton Uni-versity Press. 39–113.

Frege, Gottlob. Begriffsschrift (1879/1967). Trans. Stefan Bauer- Mengelberg.In Jean van Heijenoort (ed.) (1967),From Frege to Gödel: A Sourcebook inModern Mathematical Logic, 1879–1931. Cambridge, MA: Harvard Uni-versity Press. 5–82.

Further readingAllwood, Jens, Lars-Gunnar Andersson

and Östen Dahl (1977). Logic in Lin-guistics. Cambridge: Cambridge Uni-versity Press.

Bobzien, Susanne (2003). ‘Stoic logic’. InBrad Inwood (ed.), Cambridge Com-panion to the Stoics. New York: Cam-bridge University Press. 85–123.

Noonan, Harold W. (2001). Frege: A Crit-ical Introduction. Cambridge: Polity.

Kneale, William and Martha Kneale(1962). The Development of Logic.Oxford: Clarendon.

Read, Stephen (1995). Thinking aboutLogic. Oxford: Oxford UniversityPress.

Tomassi, Paul (1999). Logic. London:Routledge.

Stephen McLeod

LOGICAL FORM

On one central conception, the logicalform of a sentence consists in those ofits structural features which clarify

128

LOGICAL POSITIVISM

lenges’. Philosophy Compass 1/3,303–316, 10.1111/j.1747–9991.2006.00017. x.

Stephen McLeod

LOGICALPOSITIVISM

A movement in twentieth-century phi-losophy whose adherents argued thatphilosophically legitimate discussionmust be limited to statements thatcould be assigned a determinate truthvalue*. Logical positivism triggeredsome important debates about thenature of meaning and the appropri-ate methods and terminologies for theanalysis of language.

See also: Analytic Philosophy;Analytic/Synthetic;Deduction/Induction;Empiricism/Rationalism; Logic;Truth ValueKey Thinkers: Ayer, A. J.; Carnap,Rudolf; Russell, Bertrand

Logical positivism was practised invarious forms in Europe and the USAin the early and mid-twentieth cen-tury, but it is most closely associatedwith the work of the Vienna Circle.This was a group of philosophers ledby Moritz Schlick and includingRudolf Carnap, Otto Neurath andFriedrich Waismann who held meet-ings at the University of Vienna duringthe 1920s and 1930s. They were com-mitted to establishing a rigorously sci-entific mode of philosophical inquirywhere positive, or empirically justi-fied, knowledge was expressed in

On one view, logical form is con-strued as relative to a formal language.A sentence in a natural language thenhas a certain logical form in virtue ofits being translatable into an instanceof a given sort of formula in the formallanguage. Alternatively, logical formcan be conceived as already present innatural language itself, independentlyof a given formal language’s attemptsto represent it. On this second concep-tion, a sentence’s logical form is not amatter of it having a certain relation tosome construction in a formal lan-guage. Rather, the aim of the formallanguage is to capture intrinsic struc-tural features of sentences of naturallanguage. A related question concernswhether logical form is distinct from,or instead a level of, grammatical form(see Bach 2002).

A central criticism concerns howany division of the vocabulary of anatural language into logical and non-logical terms is to be justified (seeJackson 2006). That is, which wordsin a natural language are logical oper-ators and which are not? Scepticismabout logical form may start from theclaim that there is no entirely non-arbitrary answer to this question.

Primary sourcesRussell, Bertrand (1914). ‘Logic as the

essence of philosophy’. In Our Knowl-edge of the External World. Chicago:Open Court. 33–59.

Further readingBach, Kent (2002). ‘Language, logic, and

form’. In Dale Jacquette (ed.), A Com-panion to Philosophical Logic. Oxford:Blackwell. 51–72.

Jackson, Brendan (2006). ‘Logical form:classical conception and recent chal -

129

LOGICAL POSITIVISM

physics, of aesthetics and, most controversially of all, of ethics andreligion.

Logical positivism was beset withvarious problems, for instance to dowith the reliability of empirical evi-dence and the method of induction*on which it depended, and was grad-ually modified or abandoned. How-ever, its influence has continued to befelt in the philosophy of language andsubsequently in linguistics, largelybecause of the alternative ideas aboutmeaning that its critics suggested. Forinstance, W. V. O. Quine advocatedholism* in an account of meaning,rather than expecting each statementin a language to be justified by anindependent process of verification.J. L. Austin argued that ‘true’ and‘false’ were not the only philosophi-cally interesting labels that could beapplied to uses of language, a posi-tion that led to his development ofspeech act theory*. This in turn hashad a significant impact on howmeaning is discussed in present-daylinguistics.

Primary sourcesAyer, A. J. (1946). Language Truth and

Logic. Second edition. Harmonds -worth: Pelican (1971). First editionLondon 1936.

Ayer, A. J. (ed.) (1959). Logical Posi-tivism. Glencoe, IL: The Free Press.

Carnap, Rudolf (1932). ‘The eliminationof metaphysics through logical analysisof language’. In A. J. Ayer (1959), Log-ical Positivism. Glencoe, IL: The FreePress. 60–81.

Schlick, Moritz (1930). ‘The turning pointin philosophy’. In A. J. Ayer (ed.) (1959)Logical Positiviism. Glencoe, IL: TheFree Press. 53–9.

logically coherent language. Membersof the Vienna Circle published anumber of monographs and journalarticles in German, but their ideaswere disseminated more widely, andin particular became available to anEnglish- speaking audience, when A. J.Ayer published Language, Truth andLogic in 1936, after a attending anumber of meetings of the Circle onan extended visit to Vienna.

Logical positivism was not prima-rily concerned with the analysis ofnatural language; in fact Carnap andothers were openly dismissive ofeveryday usage. They saw it as impre-cise and illogical; its statements werein need of translation into a logicallyacceptable form before they could bethe subject of serious discussion.Rather, the impact of logical posi-tivism on language study was duelargely to its criterion of meaningful-ness. To be counted as meaningful andtherefore admitted into philosophicaldiscussion, a statement must be capa-ble of being classified as either true orfalse. Analytic statements, a classtaken by the logical positivists toinclude the statements of mathematicsand logic*, were meaningful becausethey were true by virtue of their owninternal properties. Synthetic state-ments could be admitted as meaning-ful if they were capable of beingsubjected to an identifiable processof verification: that is, if it was possi-ble to establish what sort of empiricalevidence a statement could be evalu-ated against in order to establishwhether it was true or false. All otherstatements – synthetic statements thatwere not amenable to verification –were simply meaningless. This classincluded the statements of meta-

130

MENTALISM

investigations on objectively observ-able data. Behaviourists see languageacquisition as a conditioning process,while mentalists propagate an inborndevice according to which language isacquired systematically.

The first mentalists were psycholo-gists such as Edward Thorndike andEdward Bradford Titchener, whoseaim was to study the mind scientifi-cally using methods of association andintrospection. Their objects of inquirywere ‘mental facts’, and they wereconvinced that by replacing specula-tion with experimentation, the mindcould be analysed into the compo-nents from which complex thoughtsare constructed.

Another kind of mentalist linguistictheory is rooted in the study of NativeAmerican languages. Edward Sapir(1921) and Benjamin Lee Whorf(1956) claimed that every languagehas its own descriptive categoriesthrough which the world is seen. Lan-guage shapes thought and is not justan expression of thought; thus it is evi-dence of how people think.

Mentalism was revived and devel-oped by Noam Chomsky (1965,1968), based on René Descartes’rationalism, as a reaction to LeonardBloomfield’s structuralistic approachand B. F. Skinner’s radical behav-iourism. In contrast to these theories,which concentrated on observablesurface structures, Chomsky postu-lated an underlying deep structure aswell as an innate language acquisitiondevice that enables human beings todevelop their linguistic competence.

Cognitive linguistics investigatesthe mental processes involved in theacquisition and use of language and ofknowledge in general. It is closely

Further readingGower, Barry (ed.) (1987). Logical Posi-

tivism in Perspective. London: CroomHelm.

Stadler, Friedrich (ed.) (1993). ScientificPhilosophy: Origins and Developments.Dordrecht: Kluwer Academic Publishers.

Siobhan Chapman

MENTALISM

A notion that can be applied to anylinguistic theory that deals with therelationship between language andthe mind. It explores the relationshipbetween language, thought and real-ity, and describes the internal lan-guage devices which explain thecreativity in language acquisition andthe processes involved in thinking,speaking and understanding.

See also: Artificial Intelligence;Behaviourism; Cognitivism;Innateness; Language of Thought;Linguistic RelativityKey Thinkers: Bloomfield,Leonard; Chomsky, Noam;Descartes, René; Sapir, Edward;Skinner, B. F.; Whorf, BenjaminLee

Mentalism dates back to the emer-gence of scientific psychology in theearly twentieth century. It has been inconstant rivalry with behaviourism*ever since John B. Watson’s (1919)reaction to the mentalist methods ofintrospection and descriptions of feel-ings and thoughts. Whereas for men-talists the main concern is the questionof whether the mind is accessible forintrospection, behaviourists base their

131

METAPHOR

Lakoff and Mark Johnson (1980)affirm that metaphors are deeplyingrained in our thoughts, actions andeveryday language.

See also: Ideational Theory;Linguistic Relativity; PossibleWorld Semantics; Signs andSemioticsKey Thinkers: Derrida, Jacques;Saussure, Ferdinand de; Whorf,Benjamin; Wittgenstein, Ludwig

Lakoff and Johnson (1980) state that,contrary to common belief, a meta -phor is not a rhetorical deviceemployed to embellish literary lan-guage. They argue that metaphorsconstitute the foundation of our con-ceptual system and influence ourthoughts, actions and communica-tion. In addition to their descriptionand classification of metaphors,Lakoff and Johnson discuss their sig-nificance in both philosophy and lin-guistics. Zoltán Kövecses (2002)proposes a detailed framework of cog-nitive metaphors and their role in lin-guistics, literary analysis, ethics andpolitics. In his (2005) publication,Kövecses focuses on the diversifica-tion of metaphors as a result of cul-tural differences. He explores thecorrelations between metaphoric andcultural variations. Murray Knowlesand Rosamund Moon (2006) give acomprehensive account of how meta -phors pervade a diverse number ofdisciplines, including semantics, liter-ature, religion, cinema and music.

Criticising philosophers and lin-guists for their failure to appreciatethe significant role metaphors play inour perception and communication,Lakoff and Johnson (1980) point outnumerous everyday expressions that

linked to cognitive psychology and toartificial intelligence*, which tries tomake machines carry out tasks requir-ing intelligence and in doing so exam-ines how humans perform such tasks.

The theory that mental states andprocesses are independent of but canexplain behaviour has had its impacton linguistics and related fieldsthroughout the twentieth century.Although criticised heavily by thebehaviourists, it remains a concept tobe considered in any approach to lan-guage and its applications.

Primary sourcesChomsky, Noam (1965). Aspects of the

Theory of Syntax. Cambridge, MA: MITPress.

Whorf, Benjamin L. (1956). Language,Thought, Reality: Selected Papers. Ed.John B. Carroll. New York: Wiley.

Further readingChomsky, Noam (1968). Language and

Mind. New York: Harcourt, Brace andWorld.

Sapir, Edward (1921). Language: AnIntroduction to the Study of Speech.New York: Harcourt, Brace and World.

Watson, John B. (1919). Psychology fromthe Standpoint of a Behaviorist. Philadel-phia: Lippincott.

Jürg Strässler

METAPHOR

A figure of speech in which a term thatis usually associated with a certainentity is used to describe another, as in‘the dawn of history’. In their seminalwork Metaphors We Live By, George

132

METAPHOR

nonphysical entities as physical ones.Examples include the quantificationof abstract entities, like speakingabout someone as having ‘a lot ofpatience’, or describing the world asbeing ‘filled with intolerance’. Themost prominent type of ontologicalmetaphors, according to Lakoff andJohnson, is personification, which isgiving a nonhuman entity a humanquality or attribute. This is evident inexpressions like talking about infla-tion ‘eating up’ someone’s savings, ora disease ‘catching up’ with some-body.

Another figure of speech that ismetaphoric in nature is metonymy,which is employing an entity or fea-ture to make reference to another.Examples of this include making ref-erence to restaurant clients by themeals they order, for example ‘Thechicken lasagna left a big tip’, or ‘Thebeef burrito spilled his drink’. A majorcategory of metonyms, according toLakoff and Johnson’s framework,involves using the part as a whole.This phenomenon, which has tradi-tionally been called synecdoche, isexemplified by expressions such as‘wheels’ to make reference to ‘cars’, or‘fresh blood’ as a reference to ‘newpeople’ in a work-place setting.Lakoff and Johnson refute the claimthat metonyms are purely referentialin nature when compared withmetaphors. They argue that the ‘part’used to make the reference would usu-ally have a vital role in determiningthe significance of the utterance.

In Philosophy in the Flesh, Lakoffand Johnson argue that ‘the banish-ment of metaphor from the realm oftruth explains why metaphor has tra-ditionally been left to rhetoric and

are metaphoric in nature. Theseexpressions include such conceptualmetaphors like argument as war(Lakoff and Johnson use small capi-tals to denote conceptual metaphorsand differentiate them from meta -phoric expressions). To support theirhypothesis they provide a multitude ofexpressions associated with this andother metaphorical concepts. Theseinclude our speaking about ‘argu-ment’ as if it were ‘war’ when we usesuch expressions as ‘attacking ordefending a position’, ‘winning orlosing an argument’, or talking aboutclaims as being ‘indefensible’. Anotherconceptual metaphor is time asmoney, which generates a number ofexpressions that include ‘saving’,‘wasting’ or ‘investing’ time.

In addition to this type, which theyregard as complex and call ‘structuralmetaphors’ because ‘one concept ismetaphorically structured in terms ofanother’ (1980: 14), Lakoff and John-son identify a number of simple typesof metaphors, including orientationaland ontological metaphors. Themajority of the morphemes belongingto the former type are related to spa-tial orientation. They speculate thatthese metaphors may have resultedfrom our cultural and physical experi-ence. The most commonly used ofthese is the up-down metaphor. InWestern culture ‘up’ is associated withpositive concepts like ‘happiness’ and‘health’, whereas ‘down’ is related tonegative ones such as ‘sadness’ and‘illness’. This gives rise to suchmetaphoric expressions as ‘high spir-its’, ‘feeling up’, ‘falling ill’ and ‘beingdepressed’. Ontological metaphors,on the other hand, include diverseexpressions most of which refer to

133

METAPHOR

nutrient, has been that the targetdomain ‘food’, a physical entity, isunderstood in terms of the sourcedomain ‘love’, a nonphysical entity.Such a simplistic view of the relation-ship between the source and targetdomains, Kövecses argues, is not sufficient to explain the variousmetaphorical expressions based onthe complex relationship between thetwo domains. He argues that the intri-cate mappings between ‘love’ and‘nutrient’, and our ability to highlightdifferent aspects of them enable us touse metaphors like ‘love’ as ‘food’,‘the desire for love’ as ‘hunger’, and‘consequences of love’ as ‘effects ofnourishment’. These complex map-pings, Kövecses argues, enable speak-ers to highlight specific source-targetrelationships that would make it pos-sible for them to express subtle mean-ings (2002: 79–92).

Since its inception as an independ-ent, vital discipline, metaphor theoryhas seen a number of developments.One of these is the universality ofmetaphors and their variation in dif-ferent cultures. Although Lakoff andJohnson (1999) alluded to this aspectof metaphors, Kövecses (2005) iscredited with its development. Kövec-ses predicates the potential universal-ity of many conceptual metaphors onthe similarity of human physiologicaland conceptual experiences. Anger inhumans, for instance, results in manyphysiological changes, including anincrease in body temperature andblood pressure. Kövecses believes thisaccounts for diverse cultures utilis-ing figures of speech based on thepressurised container conceptual metaphor. English metaphoric expres-sions like ‘boiling blood’, ‘simmering

literary analysis’ (1999: 120). Chal-lenging this – among other – Westernphilosophical principles, they assertmetaphors form the basis of concep-tualisation. Without them, therefore,the discussion of science, morality orphilosophy would not be possible.Even a universal concept like ‘time’ ismetaphorically anchored because itis perceived of, and spoken about,metaphorically, not temporally. Citingexamples from English, they discuss anumber of expressions related to thevarious aspects of ‘time’. Some ofthese are metaphoric expressions thatindicate the passage of ‘time’ as‘approaching’, ‘arriving’, ‘running’and ‘flying’. This metaphorisation oftime, they believe, is culturally specificas reflected in different languages. InEnglish, for instance, we ‘look for-ward’ to future events and regard pastones as being ‘behind us’ whereas inAymara, a language spoken in theChilean Andes, the future is ‘behind’.In this culture, the metaphorisation offuture events as being behind indicatesthe unforeseeable nature of suchevents (1999: 141).

Lakoff and Johnson’s pioneeringresearch on metaphor sparked interestin the field and paved the way for amultitude of subsequent publications.Reiterating the basic principles oftheir framework, Kövecses (2002)argues that the formula proposed toexplain conceptual metaphors, whichstates that a target domain is under-stood in terms of a source domain, isinsufficient. Using expressions such as‘someone is starved for love’ or‘hungry for affection’, he argues thatthe traditional explanation of such fig-ures of speech, which are based onthe conceptual metaphor love as a

134

MINIMALISM

Cambridge: Cambridge UniversityPress.

Lakoff, George and Mark Johnson(1980). Metaphors We Live By.Chicago: University of Chicago Press.

Lakoff, George and Mark Johnson(1999). Philosophy in the Flesh: TheEmbodied Mind and its Challenge toWestern Thought. New York: BasicBooks.

Further readingAitchison, J. (2002). Words in the Mind:

An Introduction to the Mental Lexicon.Oxford: Blackwell.

Cameron, L. and G. Low (eds) (1999).Researching and Applying Metaphor.Cambridge: Cambridge UniversityPress.

Kövecses, Z. (2000). Metaphor and Emo-tion: Language, Culture and Body inHuman Feeling. Cambridge: Cam-bridge University Press.

Lakoff, George and M. Turner (1989).More Than Cool Reason: A Field Guideto Poetic Metaphor. Chicago: Univer-sity of Chicago Press.

Whorf, B. L. (1956). Language, Thoughtand Reality. Cambridge, MA: MITPress.

Yousif Elhindi

MINIMALISM

A cover term for ideas related to theMinimalist Program, an approach tothe study of the human language fac-ulty chiefly associated with NoamChomsky. It is driven by a radical con-ceptual and technical parsimony andexplores the possibility that the designof the human computational system

down’ and ‘blowing one’s top’, forexample, have parallels in languagesas diverse as Chinese, Japanese, Hungarian, Wolof, Zulu and Polish(2005: 39). However, Kövecses adds,there are a number of variations inmetaphoric expressions within thesame culture and between differentones. He attributes this variation toregional, social and experiential dif-ferences.

Other developments in metaphortheory include work on metaphor andthe brain, acquisition of metaphors bychildren, and nonverbal metaphors.Despite the inconclusiveness of theevidence, research suggests thatmetaphors are processed by the righthemisphere of the brain. This hypoth-esis is consistent with lateralisationstudies that characterise the left hemi-sphere as being specialised in holistictypes of processing that are creativeand nonliteral. Knowles and Mooncite research showing that patientswho had right-hemisphere aphasiafound it difficult to process metaphors(2006: 62). Since figurative compe-tence is acquired relatively late, it hasbeen determined that children acquiremetaphoric processing between theages of ten and twelve. Finally, non-verbal metaphors in cinema, musicand pictorial representation have alsobeen areas of interest.

Primary sourcesKnowles, Murray and Rosamund Moon

(2006). Introducing Metaphor. NewYork: Routledge.

Kövecses, Zoltán (2002). Metaphor: APractical Introduction. Oxford: OxfordUniversity Press.

Kövecses, Zoltán (2005). Metaphor inCulture: Universality and Variation.

135

MINIMALISM

approaching the same order of magni-tude of complexity as the phenomenathemselves. Similar concerns moti-vated the transition from the RevisedExtended Standard Theory to Govern-ment-Binding Theory in the late1970s. Additionally, various principlesbegan to emerge which had a ‘leasteffort’ flavour, or were computation-ally more simple: namely local metricsfor determining the domain of certainsyntactic effects. As a result of consid-erations such as these, the strong min-imalist thesis began to emerge in themid-1990s.

Not surprisingly, this thesis has farreaching implications for every aspectof the computational system, andmuch of the organisation and technol-ogy of Government-Binding Theoryhas undergone revision. If PF and LFare the only linguistic levels with anysignificance (being the only two seem-ingly mandated by virtual conceptualnecessity, though see below), thenother linguistic levels, such as D-struc-ture and S-structure, must be elimi-nated. Linguistic principles stated interms of these levels must be reformu-lated.

Structure-building has also under-gone radical revision, returning to aversion of generalised transforma-tions from Chomsky’s early work inthe 1950s. Rather than building upsyntactic objects in a top-down fashion through recursive phrase- structure rules, the so-called Mergeoperation builds structure from thebottom up, combining two syntacticobjects of arbitrary complexity: thatis, individual lexical items or larger,previously created syntactic objects.Consistent with the goal of appealingonly to virtual conceptually necessary

for language optimally satisfies con-straints imposed only by the need toservice a sensori-motor interface (PF)and a conceptual-intentional interface(LF).

See also: Adequacy; Logical Form;Transformational-GenerativeGrammar; Universal GrammarKey Thinkers: Chomsky, Noam

The Minimalist Program, the succes-sor framework to Noam Chomsky’sGovernment-Binding Theory, was ini-tially articulated in the book of thesame name (Chomsky 1995) whichbrought together several earlierpapers with some new material. It wasexpanded upon principally in Chom-sky (2000, 2001, 2004, 2005). Theleading idea, often called by Chomskythe ‘strong minimalist thesis’, is thatlanguage is a ‘perfect solution’ to theproblems imposed by virtual concep-tual necessity. In other words, thereare certain inescapable constraints onthe computational system of thehuman language faculty. If it is torelate sound and meaning, the com-putational system must interface withat least a sensory-motor system (Pho-netic Form/PF) and a semantic/ conceptual-intentional system (Logi-cal Form/LF). The strong minimalistthesis hypothesises that the only con-straints are those imposed by theseinterfaces, and that computationalsystem satisfies these constraints in anoptimal fashion.

As with previous radical changes inframework developed by Chomsky,questions of simplicity, and ultimatelythe logical problem of language acqui-sition, were central. By the early 1990sthere was within Government-BindingTheory a sense that explanations were

136

MINIMALISM

Merge. As Chomsky notes, this viewof structure-changing entails thecopy theory of movement. The twoelements of the chain have to be identical, rather than one being aninclusiveness-violating ‘trace’, be -cause Merge cannot change what itoperates on. What would tradition-ally be referred to as the head and thetail of the chain can still be distin-guished, however, by examining thecontext in which the copies appear.

Insofar as it is features, rather thanlexical items themselves, that are rele-vant to the PF and LF interfaces, thesebecome the focus of the syntactic derivation within the Minimalist Pro-gram. As alluded to above, a distinc-tion is made between interpretableand uninterpretable features. Certainfeatures are plausibly interpretable atLF, though only on certain elements.Consider ‘agreement features’, such asperson, number and gender (usuallyreferred to as phi-features). Whether agiven Noun Phrase (NP) is singular orplural, second-person or third-person,affects the way in which the itemis interpreted. However, althoughperson and number features canappear on functional categories suchas Tense (reflected ultimately in sub-ject-verb agreement), they seem not tohave the same interpretive implica-tions for Tense at LF. Thus the phi- features on NPs are an instance of‘interpretable’ features at LF, whilephi-features on Tense are uninter-pretable at LF and must be eliminated.

In the earliest minimalist literatureof the early-mid 1990s, movement orInternal Merge was the mechanism bywhich uninterpretable syntactic fea-tures were eliminated. However, inthe early 2000s these operations

mechanisms, it is claimed that Merge‘comes for free’, insofar as language isundeniably hierarchically recursive.

As language is hypothesised to bean ‘optimal’ solution to the conditionsimposed by the two interfaces, ques-tions of ‘economy’, both of represen-tation and derivation, have taken on anew prominence. Economy of repre-sentation prohibits ‘superfluous’ sym-bols. This entails that the interfacelevels of PF and LF may only be com-posed of symbols which are inter-pretable at that interface. This is theprinciple of Full Interpretation. Fea-tures uninterpretable at a given inter-face must be eliminated prior to thatinterface. The syntactic computationtherefore becomes principally drivenby the need to eliminate these uninter-pretable features. Economy of repre-sentation also entails the principle ofInclusiveness, which states that thecomputational system may not itselfintroduce any symbols into the deri-vation. As a result many mechanismsin previous Chomskyian frameworks,such as traces, binding-theoreticindices and the like, do not exist. Syn-tactic principles which make referenceto them must be eliminated or refor-mulated. Economy of derivation stip-ulates that the syntactic derivationitself proceed in an optimal fashion.

In another radical change from pre-vious Chomskyan approaches, theMinimalist Program claims that struc-ture-building and structure-changingare essentially the same operation.Movement is simply an instance ofMerge where the object merged at theroot comes from inside the existingsyntactic object rather than from out-side; Chomsky usually distinguishesInternal Merge (Move) from External

137

MINIMALISM

object over to the sensory-motor inter-face and the conceptual-intentionalinterface, preserving a strong versionof the strict cycle. On this view, thedifference between overt and covertmovement lies in whether an elementis displaced before or after deciding tospell out the phase.

Cyclic spell-out and interpretationentails a Phase Impenetrability Condi-tion. Since material within a phase hasalready been spelled out and inter-preted, it is no longer accessible to thesyntactic computation. However, soas not to completely rule out succes-sive-cyclic movement in any form, it ishypothesised that material at the edgeof a phase (a specifier or adjunct of aphase head) and the phase head itselfare accessible to further operations.

Potentially one of the most interest-ing things about the minimalist Pro-gram, as noted by Chomsky (2004), isits contribution to the longstandingquestion of adequacy*. The highestlevel of adequacy, explanatory ade-quacy, is attained when there is a gen-eral theory of grammars that providesthe basis for selecting the mostdescriptively adequate grammar froma selection of possible grammars.However, explanatory adequacy maybe just description at a higher level, tothe extent that the general theory ofgrammars merely incorporates what-ever features the most descriptivelyadequate grammars happen to pos-sess. Minimalism allows one to gobeyond explanatory adequacy, intheory at least. Not only can the gen-eral theory of grammars select themost descriptively adequate grammar,but if language truly is a ‘perfect’ solu-tion given the constraints imposed,then we can explain why a minimalist

became separated. The feature-check-ing operation is Agree. It consists of aprobe, a syntactic element which hasan uninterpretable feature (say, phi-features on Tense), and a goal, an ele-ment which has an interpretableversion of this feature (say, an NP).The Agree-relation deletes the unin-terpretable feature of the probe andprovides a ‘value’ for morphologicalpurposes based on the appropriatefeature of the goal. This recreates thetraditional claim that it is tensed verbsthat agree with their subjects, not viceversa. Displacement itself is triggeredby an edge feature (often also referredto as an EPP feature) which particularfunctional categories possess andwhich can act as a probe. Edge fea-tures further underline the essentialidentity of Merge and Move withinthe Minimalist Program, as it is alsothe edge feature of a lexical itemwhich permits it to be merged withanother syntactic object (its comple-ment).

One of the most important mini-malist developments relating to struc-ture-building and changing concernsthe concept of the phase, first intro-duced in Chomsky (2001). In a bid toreduce computational complexity,Chomsky suggested that derivationswere inspected for legitimacy at par-ticular key points; specifically, afterthe Merge of a phase head: vP (thelight verb sister to VP which is impli-cated in transitivity) or CP.

Chomsky (2004) notes that eventhe levels of PF and LF can be elimi-nated under the assumption that spell-out and interpretation take placecyclically, at the level of the phase.When a phase is completed, the oper-ation Transfer hands the syntactic

138

MODALITY

former is usually called epistemicmodality and the latter, deonticmodality. In formal semantics andmodal logic, modality concerns thepossibility or necessity of the predica-tion and its negation.

See also: Speech Act Theory;Politeness; PropositionsKey Thinkers: Aristotle; Austin,J. L.; Searle, John

A proposition such as ‘They go homeat this hour’ can be communicatedwith different attitudes marking itsmodality. There are two basic types:(1) epistemic modality, which is aboutthe truth of the proposition, such asguessing (‘They may have gone homeat this hour’) and (2) deontic modal-ity, which is about the realisability ofthe proposition by some agent, suchas granting permission (‘They may gohome at this hour’). Closely related todeontic modality is dynamic modalitywhich expresses ability (or inability)to put the assertion into practice inde-pendently of the judgement or will ofthe speaker, such as ‘They can gohome; they still have the return tick-ets’. Often held as part of epistemicmodality is evidentiality – that is thewarrant the speaker has for the asser-tion, such as in ‘The show seems to bea success – whereby the speaker relieson some evidence, for example,applause heard, rather than subjectivejudgement.

Epistemic modality is also calledproposition-modality, or speaker-ori-ented modality. Deontic modality isalso known as event-modality since itis about the realisability of the state ofaffairs expressed in the proposition. Insome recent literature, deontic modal-ity is also described as agent-oriented

grammar ends up being the mostdescriptively adequate.

Primary sourcesChomsky, Noam (1995). The Minimalist

Program. London: MIT Press.Chomsky, Noam (2000). ‘Minimalist

inquiries: the framework’. In R. Martin,D. Michales and J. Uriagereka (eds),Step by Step: Essays on Minimalism inHonor of Howard Lasnik. Cambridge,MA: MIT Press. 89–155.

Chomsky, Noam (2001). ‘Derivation byphase’. In M. Kenstowicz (ed.), KenHale: A Life in Language. Cambridge,MA: MIT Press. 1–52.

Chomsky, Noam (2004). ‘Beyond explana-tory adequacy’. In A. Belletti (ed.), Struc-tures and Beyond – The Cartography ofSyntactic Structure, vol. 3. Oxford:Oxford University Press. 104–131.

Chomsky, Noam (2005). ‘Three factors inlanguage design’. Linguistic Inquiry36.1: 1–22.

Further readingHornstein, Norbert, Jairo Nunes and

Kleanthes K. Grohmann (2005). Under-standing Minimalism. Cambridge:Cambridge University Press.

Boškovic, Željiko and Howard Lasnik(2007). Minimalist Syntax: The Essen-tial Readings. Oxford: Blackwell Pub-lishing.

Geoffrey Poole

MODALITY

The expression of the speaker’s attitude concerning the truth of aproposition or the realisability of aproposition by some agent. The

139

MODEL-THEORETIC SEMANTICS

Discourse. Amsterdam and Philadel-phia: Benjamins.

Lyons, J. (1977). Semantics. Two vols.Cambridge: Cambridge UniversityPress.

Agustinus Gianto

MODEL-THEORETICSEMANTICS

One of the mainstream approachesto the study of meaning in natural languages. Model theory is a branchof mathematical logic concerningthe description of the semantics ofartificial languages. A model- theoretic approach to the semanticsof natural languages was first put for-ward by the Polish logician AlfredTarski in 1935. Following theTarskian tradition, current model-theoretic approaches in linguisticsusually embrace a truth-conditionaltheory of semantics. Model-theoreticsemantics provides a mathematicallyrigorous and elegant way of describ-ing and explaining extremely intricatesemantic phenomena.

See also: Analytic Philosophy;Compositionality; Logic;Logical FormKey Thinkers: Frege, Gottlob;Montague, Richard; Tarski, Alfred

The most influential application ofmodel theory to natural languagesemantics can be found in the work ofthe American logician and philoso-pher of language Richard Montague.Analytic philosophers advocated thesystematic study of formal, idealised

modality because it characterises howsome agent will carry out the proposi-tion.

Modality is expressed lexically orgrammatically or a combination ofboth. Lexical expressions includeadverbs such as ‘possibly’, ‘maybe’,‘probably’, ‘presumably’, ‘suppos-edly’, for example. Grammaticalexpressions make use of the modalverbs ‘can’, ‘may’, ‘must’, ‘shall’,‘will’. In a number of languages, themorphology of the verbal predicatealso indicates modality, hence the tra-ditional term mood, such as state-ment, question, imperative. Note thatthe illocutionary force of a statementmay generate modality, such as ‘Thisroom is a bit dark’ can be interpretedas a request to switch on the light, ora question like ‘Have we finished?’may mean an invitation to leave.

Primary sourcesPalmer, F. R. (2001). Mood and Modality.

Second edition. Cambridge: CambridgeUniversity Press.

Von Wright, G. H. (1951). An Essayin Modal Logic. Amsterdam: North- Holland.

Further readingHoye, Leo (1997). Adverbs and Modality

in English. London and New York:Longman.

Palmer, F. R. (1990). Modality and theEnglish Modals. Second edition.London and New York: Longman.

Bybee, J. L.,R. D. Perkins and W. Pagliuca(1994). The Evolution of Grammar:Tense, Aspect and Modality in the Lan-guages of the World. Chicago: Univer-sity of Chicago Press.

Bybee, J. L. and S. Fleischman (eds)(1995). Modality in Grammar and

140

MODEL-THEORETIC SEMANTICS

Montagovian model-theoretic seman-tics, and its influence in linguisticsemantics.)

Central to the model-theoreticapproach to semantics is the notion of‘truth with respect to a model’. Inother words, the specification of thetruth conditions of sentences in a lan-guage is not provided in a vacuum,but rather relative to an abstractmathematical model of some state ofaffairs in the world, which we couldthink of informally as a ‘snapshot’ ofreality. The construction of idealisedmodels of certain aspects of reality isa common practice in many scientificcontexts, as these can provide thebasis for successful explanations ofdifficult phenomena in the natural sci-ences.

Model-theoretic semantics providesinterpretations for sentences in a lan-guage by first specifying the entitiesthat exist in the state of affairs in theworld being modelled. Thus modelscan be very complex and they cancomprise abstract as well as real enti-ties. For example, models usually con-tain two special abstract entitiesreferred to as True (or 1) and False (or0), the truth values, which are oftenassumed to be the semantic values oftrue and false sentences, respectively.Models are, in effect, idealised repre-sentations of what sentences in a lan-guage can be about. It is important topoint out, however, that logicalmodels of natural language are notmeant to capture our knowledge oflanguage, and therefore should not beconstrued epistemologically.

Having specified the ontology of themodel, the task of a model-theoreticsemantics is to assign interpretationsto the basic expressions of the object

languages as a means of better under-standing slippery notions like entail-ment (logical consequence) andcontradiction. They also believed thatby approaching such semantic phe-nomena through the analysis of artifi-cial languages, one could avoid theambiguities, vagueness and paradoxesthat are frequently found in naturallanguages.

Montague was of the same opinion.He famously held the view that natu-ral languages are not fundamentallydifferent from formal languages andthat, consequently, the same approachcan be employed fruitfully in theirstudy (Montague 1970). Thus Mon-tague extended the mathematicaltechniques that were traditionallyused in the semantics of formal lan-guages to the study of meaning in nat-ural languages.

One such technique was modeltheory. Model theory was used bylogicians to provide a precise specifi-cation of the truthconditions of sen-tences in artificial languages. Startingwith Tarski’s seminal essay ‘The Con-cept of Truth in Formalized Lan-guages’ (1935), the concept of truthhas played a fundamental role in thesemantics of formal languages. Tarskisuggested that describing the interpre-tation of a given sentence in a formallanguage involved giving, in a meta-language previously understood, aprecise specification of the conditionsunder which that sentence is true.Montague’s model-theoretic seman-tics for natural language is truth-con-ditional, as it provides a theory oftruth for a fragment of English.(Partee (1975, 1996, 1997) provides acomprehensive overview of the mainideas and historical development of

141

MODEL-THEORETIC SEMANTICS

tions that must be obtained for sentences in the object language to betrue relative to the model under consideration. Bach (1989), Cann(1993), Dowty, Wall and Peters(1981) and Gamut (1991) provideexcellent introductions to model- theoretic semantics as it is currentlypractised in linguistics.

Perhaps the most important advan-tage of the notion of ‘truth withrespect to a model’ is that it allows usto define, in a rigorous and formallyprecise way, key semantic propertiesof natural language sentences likevalidity and contradictoriness, as wellas important meaning relations thatexist between them such as entailmentor logical equivalence. Equipped withthis notion, we can quantify over theclass of possible models and say, forexample, that a sentence S1 entails asentence S2 just in case every model inwhich S1 is true is a model in which S2is true as well; or that any two sen-tences S1 and S2 of our object lan-guage are logically equivalent just incase S1 is true in exactly the samemodels in which S2 is true and in noothers; or that a sentence S1 is contra-dictory just in case it is false withrespect to every possible model. A cor-rect understanding of these propertiesand relations is at the heart of any sci-entific theory of natural languagesemantics.

An issue that causes disagreementbetween different theories of model-theoretic semantics is the syntax-semantics interface. For example,some theories postulate a separate levelof linguistic representation mediatingbetween natural language syntax andmodel-theoretic semantics. This addi-tional level of representation, which is

language by associating them withentities in the model. Of course, dif-ferent expressions may denote, orrefer to, different kinds of things in themodel. For example, a proper namesuch as ‘Mary’ denotes a member ofthe set of individuals in the model,whereas the adjective ‘clever’ is takento denote a subset of that set, namelythe set of individuals in the model whoare clever. A sentence like ‘Mary isclever’ is assigned a truth value*: trueif the sentence is actually true – that is,if Mary is in fact a member of the setof clever individuals in the model – orfalse otherwise. As we can see, modelshave internal structure. In order tohave a systematic way of describingthese distinctions, models are oftenstructured into different domainsaccording to the types of things thateach domain comprises. For example,the three expressions mentionedabove denote members of differentdomains in the model: respectively thedomain of individuals, the domain ofsets of individuals and the domain oftruth values.

A set of recursive syntactic rulesdefining the class of well formed sentences of the language must be pro-vided, together with a precise specifi-cation of how the various basicexpressions can be combined syntacti-cally to yield complex expressions.For each of these syntactic rules, thereis a corresponding semantic rulewhich determines the interpretation ofthe complex expressions on the basisof the interpretation of their parts, inaccordance with the principle of com-positionality* of meaning (sometimesalso referred to as ‘Frege’s Principle’).Ultimately, a model-theoretic seman-tics must explicitly define the condi-

142

NAMES

(eds), Handbook of Logic and Lan-guage. Amsterdam and Cambridge, MA:Elsevier and MIT Press. 5–91.

Tarski, A. (1935). ‘The concept of truth informalized languages’. In J. Corcoran(ed.) (1983), Logic, Semantics, Meta-mathematics. Indianapolis, IN: HackettPublishing. 152–278.

Further readingBach, Emmon (1989). Informal Lectures

on Formal Semantics. Albany: SUNYPress.

Cann, Ronnie (1993). Formal Semantics:An Introduction. Cambridge: Cam-bridge University Press.

Dowty, David R., R. E. Wall and S. Peters(1981). Introduction to MontagueSemantics. Dordrecht: D. Reidel Pub-lishing.

Gamut, L. T. F. (1991). Logic, Languageand Meaning (Volume 2): IntensionalLogic and Logical Grammar. Chicago:University of Chicago Press.

Iván García Álvarez

NAMES

In philosophy the term has been usedsometimes to describe any word orphrase that can refer, and sometimesin the more restricted sense of a wordthat uniquely identifies an individual.In linguistics, names or proper nounsare recognised as a separate class oflinguistic signs, distinct from commonnouns. The field of name studies, oronomastics, is increasingly movingfrom language-internal criteria tocontextual ones.

See also: Definite Descriptions;Sense/Reference; Use/Mention

derived from other linguistic levels bytransformational operations in thesyntax, is called logical form* (May1985). On the other hand, the strictcompositionality constraint betweenform and meaning built into Mon-tague’s original theory precludes theneed for such an additional level ofsyntactic representation.

Model-theoretic semantics can pro-vide interpretations for (semantic rep-resentations of) linguistic expressionsin a way that accurately captures ourpre-theoretical semantic judgementsof such expressions. It is currently thefocus of a great deal of research in lin-guistics, but much of this research alsounifies and expands various other dis-ciplines, such as philosophy, cognitivescience, artificial intelligence*, andtheoretical computer science. Work inmodel-theoretic semantics has madesignificant contributions to our under-standing of truth and meaning in nat-ural languages, for example in areasas diverse as tense and aspect, gener-ics, negation, and plurality.

Primary sourcesMay, Robert (1985). Logical Form: Its

Structure and Derivation. Cambridge,MA: MIT Press.

Montague, Richard (1970). ‘Universalgrammar’. Theoria 36: 373–98.

Partee, Barbara H. (1975). ‘Montaguegrammar and transformational gram-mar’. Linguistic Inquiry 6: 203–300.

Partee, Barbara H. (1996). ‘The develop-ment of formal semantics in linguistictheory’. In S. Lappin (ed.), The Hand-book of Contemporary Semantic Theory.Oxford: Blackwell Publishing. 11–38.

Partee, Barbara H. (with Herman L. W.Hendriks) (1997). ‘Montague grammar.’In J. van Benthem and A. ter Meulen

143

NAMES

because of a chain of communicationestablished within a community ofspeakers.

In linguistics, unique reference isusually taken to be the key distinctionbetween names and common nouns.However, this distinction is not asclear-cut as it may seem. First, somecommon nouns such as ‘sun’ or ‘hell’have unique reference but they arenot considered names. Second, thereferents of many names have cer-tain things in common: all referentsof the name ‘Mary’, for instance, arefemale, born to English-speaking parents (or they are learners of Eng-lish as a second language who areusing an English name in addition totheir native one), and, given thatmost personal names are subject tofashion, one may even make an edu-cated guess about the age of manyMarys.

The syntactic criterion that distin-guishes proper nouns from commonnouns in English is the absence of adeterminer. However, not all commonnouns can take the full range of deter-miners: ‘a music’, for example, is nota possible expression in English. Thenon-translatability of names has beenput forward as another distinction criterion. However, counter-examplesare easy to find such as German‘Schwarzwald’, which becomes ‘BlackForest’ in English.

Thus, it is obvious that proper andcommon nouns form a gradient, withprototypical cases at either end andmany fuzzy ones in between. Ulti-mately, the intention and perceptionof language users is the only way todistinguish between common nounsand proper nouns. However, this dis-tinction is no longer systematic and

Key Thinkers: Frege, Gottlob;Kripke, Saul; Mill, J. S.; Russell,Bertrand; Searle, John

The nature of names has exercisedphilosophers of language for a verylong time. The best known early treatise is probably Plato’s CratylusDialogue which centres around thequestion of whether names havean intrinsic relation to their referentor are as arbitrary as other linguisticsigns. The discussion of names inmodern philosophy was initiatedby John Stuart Mill (1867), who distinguished between ‘connotative’names, such as ‘The father ofSocrates’, which refer to an individ-ual by means of describing that individual, and ‘non-connotative’names, such as ‘Sophroniscus’, whichsimply refer. In contrast, GottlobFrege (1892) argued that all names,including personal names, haveboth a reference and a sense. Accord-ing to Bertrand Russell (1919), per-sonal names can be used to referdirectly to an individual who isimmediately present, but in mostinstances are used as ‘abbreviateddescriptions’. There are various problems for Russell’s account,including the fact that, since differentspeakers may identify an individualby means of different descriptions,the meaning of a name would seem tovary depending on who is using it. Inone attempt to address this problem,John Searle (1958) developed his‘cluster theory’ account. In perhapsthe most influential philosophical discussion of names since Russell,Saul Kripke (1972) argued thatnames are ‘rigid designators’ thatsucceed in referring to individuals

144

NONNATURALMEANING

A type of meaning, which includes lin-guistic meaning, in which there is nonecessary link between a sign andwhat it represents. In his influentialaccount of nonnatural meaning, H. P.Grice emphasised the importance of aspeaker’s intentions and a hearer’srecognition of these intentions.

See also: Conventional Meaning;Implicature; Intentionality; Signsand Semiotics; Use/MentionKey Thinkers: Grice, H. P.; Peirce,C. S.; Saussure, Ferdinand de;Searle, John; Strawson, P. F.

The idea that linguistic meaning canbe distinguished from natural mean-ing is long standing. It can be tracedback to classical philosophy, and inmore recent times can be found inPeirce’s account of words as ‘sym-bols’ and in Saussure’s discussion ofthe arbitrary nature of the sign. How-ever, Grice’s short article ‘meaning’attempted a more precise characteri-sation.

Grice draws attention to two differ-ent ways in which the verb ‘mean’ isused, exemplified by ‘Those spotsmean measles’ and ‘Those three ringson the bell (of the bus) mean that thebus is full’. The first case involves nat-ural meaning; the spots simply are asymptom of the disease. In the secondcase, there is no necessary connectionbetween the three rings and the mean-ing, but rather someone (the bus con-ductor) meant something by the rings;we can say that the rings mean ‘Thebus is full’. Unlike in the case of naturalmeaning, the rings are not a guarantee

inherent in language but a matter ofnaming practices in context.

There can be no doubt that the rela-tionship between a name and its refer-ent has more psychological reality forlanguage users than that pertainingbetween any other linguistic signs andtheir referents. Modern linguistics andonomastics, however, have tended todismiss this widely held assumptionas ‘primitive’ or ‘superstitious’. Thisstance has led to an ever-wideningchasm between limited academicinterest in names and naming andwidespread popular interest. In thephilosophy of language, the relation-ship between names and their refer-ents remains an enduring focus ofinterest.

Primary sourcesFrege, Gottlob (1892). ‘On sense and

meaning’. In Peter Geach and MaxBlack (eds) (1980), Translations fromthe Philosophical Writings of GottlobFrege. Oxford: Blackwell. 56–78. FirstEdition 1952.

Kripke, Saul (1972). ‘Naming and neces-sity’. In D. Davidson and G. Hartman(eds), Semantics of Natural Language.Dordrecht: Reidel.

Mill, J. S. (1867). A System of Logic.London: Longman. Chapter 2.

Russell, Bertrand (1919). Introduction toMathematical Philosophy. London:George Allen and Unwin.

Searle, John (1958). ‘Proper names’. Mind67: 166–73.

Further readingEvans, Gareth (1982). The Varieties of

Reference. Oxford: Oxford UniversityPress.

Ingrid Piller and Siobhan Chapman

145

OPTIMALITY THEORY

Searle, John (1969). Speech Acts. Cam-bridge: Cambridge University Press.

Strawson, P. F. (1964). ‘Intention and con-vention in speech acts’. The Philosophi-cal Review 73: 439–60.

Further readingAvramides, Anita (1989). Meaning and

Mind. Cambridge, MA: MIT Press.Chapman, Siobhan (2005). Paul Grice:

Philosopher and Linguist. Basingstoke:Palgrave.

Siobhan Chapman

OPTIMALITYTHEORY

A framework in theoretical linguisticsused to formalise analyses in phonol-ogy, and less frequently other areas oflinguistics. Its core is the assumptionthat linguistic generalisations shouldbe described using a set of violableconstraints on surface representationswhich are ranked in terms of theirimportance.

See also: Generative Phonology;Universal GrammarKey Thinkers: Chomsky, Noam

Optimality theory (OT) took centrestage in theoretical linguistics duringthe 1990s. Its basic tenets were bothfamiliar and revolutionary and thishas doubtless contributed to its suc-cess. It developed ideas long presentin linguistics, but gave them new characteristics, considerably chang-ing the understanding of the gram-mar. Despite some influence fromneural networks, OT was essentially

of the truth of this statement; the busconductor may be mistaken or deliber-ately trying to deceive. Grice labelledthis second type of meaning ‘nonnat-ural meaning’, for which he coined theabbreviation ‘meaningnn’. Linguisticmeaning is a type of meaningnn.

For Grice, meaningnn is determinedby a speaker’s intention to communi-cate something, and by a hearer’srecognition of that intention. If A is aspeaker and x an utterance, then ‘“Ameantnn something by x” is roughlyequivalent to “A intended the utter-ance of x to produce some effect in anaudience by means of the recognitionof this intention”’ (1957: 385). Griceraises but does not fully develop theidea that linguistic meaning (‘what xmeansnn’) may itself be determined byspeakers’ intentions, and hence thatconventional meaning is to be definedin terms of psychology. He also hintsthat what speakers intend to commu-nicate in specific contexts may some-times go beyond what their wordsactually meannn.

‘Meaning’ has largely beenfavourably received by philosophersand linguists alike, although somehave put forward criticisms that sug-gest flaws in Grice’s theory (Strawson1964; Searle 1969; Schiffer 1972).‘Meaning’ also introduced Grice’sinterest in the distinction betweenwhat our words mean and what wemean by using our words, which heexplored further in his work on con-versational implicature*.

Primary sourcesGrice, Paul (1957). ‘Meaning’. The Philo-

sophical Review 66: 377–88.Schiffer, Stephen (1972). Meaning.

Oxford: Clarendon Press.

146

OPTIMALITY THEORY

tial, setting out the OT approach(along with work by McCarthy andPrince).

OT analyses are formulated in a‘tableau’, exemplified below (seeTable 1) for German Final ObstruentDevoicing (FOD) in its standarddescription: syllable-final voicedobstruents are devoiced, as in Bund[bʊnd] ‘federation’ (a contentious butstandard assumption here is that [d]in German is specified for [voice],while [t] is unspecified). The tableau’stop row shows the input (the under-lying representation), and then theconstraints, in ranked order from leftto right. The first column shows theset of candidates, one of which ischosen as the output (indicated by apointing finger). The other columnsshow constraint violations, eachreceiving one asterisk. The ‘fatal’ vio-lation (which rules candidates out) isindicated by an exclamation mark.There are two basic types of con-straint: markedness and faithfulnessconstraints. Markedness constraintspenalise candidates which containmarked structures: the analysis ofFOD uses *FinalObstruent/Voice(*FOV), which dictates that finalobstruents may not be specified for[voice]. Faithfulness is a crucial inno-vation of OT: markedness constraintscan exert pressure for an output todiffer from its input, but faithfulnessconstraints do the opposite, requiringidentity between input and output.Individual constraints regulate par-ticular aspects of faithfulness:Ident() that the value of [voice] bethe same in input and output, andMax requires everything in the inputto have some correspondent in theoutput, ruling out deletion.

conceived as a development of gener-ative phonology*, so it maintains thebasic competence/performance dis-tinction, and that between underlyingand surface levels of representation,although these are reinterpreted. Onecentral OT assumption is that onlyconstraints should be used to charac-terise linguistic generalisations. Thereare no substantive rules or transfor-mations. A set of potential surfaceforms (‘candidates’) are evaluated bythe grammar and one is chosen asthe optimal candidate (the ‘output’)because it violates a language’s con-straints in the least bad way. This ispossible thanks to another key char-acteristic: all constraints are violable,and languages rank them in order ofimportance, so it is worse to violate ahigh-ranked constraint than a low-ranked one. As the standard assump-tion is that the set of constraints(known as Con) is universal, lan-guage learners must work out howconstraints are ranked in the lan-guages they acquire.

OT provided linguists with newways to work and new theoreticalproblems to crack, although somehave rejected it as fundamentally mis-guided. It was introduced by AlanPrince and Paul Smolensky and hasbeen developed by many others, mostnotably John McCarthy. OT was cre-ated with phonological problems inmind, and is still most popularamong phonologists, but is also usedin syntax and other areas. After earlypresentations in 1991, Prince andSmolensky distributed the manu-script Optimality Theory: ConstraintInteraction in Generative Grammarin 1993. Although not publisheduntil 2004, this was highly influen-

/bʊnd/ *FOV Max Ident(voice)

bʊnd !*

☞bʊnt☞ *

bʊn !*

bʊ !**

Table 1

147

OPTIMALITY THEORY

indistinct: thus it does not matterwhether or not stops are aspiratedunderlyingly in English, as these seg-ments do not contrast and, assumingthat aspiration is enforced by a high-ranked Aspiration constraint (‘initialstops must be aspirated’), then [pɪk]could be derived from either [pɪk] or[phɪk]. Which of these is the under -lying form is simply not an important(or relevant) question. This character-istic is known as the Richness ofthe Base, and has some theoreticalbenefits. In the German case, how-ever, alternations (such as Bund~Bundes [bʊnd]~[bʊndəs] ‘federation’~‘federation-genitive’) show that theunderlying segment must be voiced.

OT analyses can often be linked topre- or non-OT work by implement-ing as a violable constraint what suchwork sees as a universal principle.This has likely contributed to OT’ssuccess because it expresses what pho-nologists have wanted to be able tosay: certain principles often play a rolein the phonology of languages, butnot always. For example, the Obliga-tory Contour Principle (OCP) wasintroduced in the 1970s to forbidadjacent identical aspects of structure,such as adjacent high tones or specifi-cations for place. By the late 1980s, ithad been shown that the OCP is not

Fully faithful [bʊnd] is rejectedbecause it violates high-ranked *FOV.An unlimited number of candidatesare in fact produced (by the functionGen,which simultaneously performsevery possible process on the input toderive the candidates). In practice,only a few ‘reasonable’ candidates areconsidered in analyses (outlandishcandidates are assumed to be ruledout by high-ranked faithfulness con-straints). The candidates include[bund], which satisfies *FOV, with novoicing in its final obstruent. [bund]violates Ident(), but is still the outputbecause other candidates violatehigher-ranked constraints. The othertwo candidates do not violate *FOV,as they have no final obstruent(having undergone deletion), but theydo violate Max, also ranked aboveIdent(). This ranking is shown by theunbroken line between Max andIdent(). The broken line between*FOV and Max shows that theirmutual ranking is irrelevant here. Theabove ranking is specific to languageswith FOD. In English, with no FOD,the constraint *FOV must be rankedbelow Ident().

It is worth noting that the right can-didate would still be selected if theinput were /bʊnd/. In fact, in somecases, the precise nature of the input is

148

OPTIMALITY THEORY

analyse simple FOD are subject todebate: there is not necessarily onestraightforward way to analyse anyphenomenon, because there is no cleartheory of Con. Some see this as anopportunity for debate, others as aproblem: how can we know whichconstraints exist? A more fundamen-tal criticism questions whether wereally should formalise all tendenciesas cognitive constraints on surfaceforms. Should ‘ease of articulation’(formalised in OT as Lazy) be seen asthe same kind of thing as constraintson the faithfulness of features?

Despite such criticism, OT has afirm place in theoretical linguistics, asthe framework for most analysis anddebate in phonology, and consider-able work in other areas of linguistics.It provides a novel set of principles toguide analysis and has allowed newanswers to old problems. For thosewho follow its direction, it hasfocused work on the status of con-straints and moved attention awayfrom representations and underlyinglevels. The framework for constraintinteraction has also been adoptedby linguists who reject generativeassumptions. Its influence is feltwidely in linguistic theory.

Primary sourcesPrince, Alan and Paul Smolensky

(1993/2004). Optimality Theory: Con-straint Interaction in Grammar. Rut-gers University and University ofColorado at Boulder. Oxford: Black-well.

McCarthy, John and Alan Prince (1993).Prosodic Morphology: ConstraintInteraction and Satisfaction. Universityof Massachusetts, Amherst and RutgersUniversity.

absolute, but is best understood as atendency in languages, elevated to thelevel of an inviolable principle insome. This is exactly what OT pre-dicts, and OT thus allows phonolo-gists to formalise tendencies.

Theoretical development in OT haslargely involved either work on thetypes of constraints allowed, drivenby attempts to find new, better ways toexpress linguistic generalisations, orwork on additions to the basic theo-retical machinery, often to remedyperceived shortcomings in the model,such as its inability to account foropacity. The former includes work on‘positional faithfulness’, which prefer-entially preserves input specificationsin strong positions, such as the onset.FOD can also fit into a positional faithfulness model: a positional IdentOnset(voice) can require thatthe value of [voice] be the same in inputand output in onsets, and conflict witha general *Obstruent/ Voice, whichrequires that no obstruent be specifiedfor [voice]. This achieves the sameresult, but also opens up new ‘posi-tional’ analytical options.

Opponents of OT often argue thatits central tenets are mistaken. Criticshave claimed that the status of inputsis problematic. The Richness of theBase has been criticised on both psy-cholinguistic and theoretical grounds(much previous theoretical successwas due to the ability to specify thecharacteristics of underlying forms).The fact that OT has largely led itspractitioners to abandon work onphonological representation (espe-cially at the segmental level) has beencondemned. The contents of Conhave also proved controversial. As wehave seen, the constraints needed to

149

ORDINARY LANGUAGE PHILOSOPHY

philosophy or linguistic philosophy,but these terms should not taken asaccurately descriptive. It is true thatOxford University served as its princi-pal locus, with key Oxford philoso-phers such as Gilbert Ryle, P. F.Strawson and J. L. Austin contribut-ing to its development. But it is oftenpointed out that many of the ideas ofthe later Wittgenstein also had animpact on the movement, and JohnWisdom, associated like Wittgensteinwith the University of Cambridge, isgenerally considered a sympathiser ofOLP. Furthermore it is not true thatthere was anything approaching aconsensus among the philosophers ofOxford regarding the main ideas ofthe movement; it is arguable that thegroup did not even constitute a move-ment or a school in any real sense. Atbest they shared some common atti-tudes concerning the nature of lan-guage and ways of doing philosophy.To use a term attributable to Wittgen-stein, the farthest one may go in char-acterising the group is to say that therewas a ‘family resemblance’ among thepositions assumed by many of itsmembers.

The term ‘linguistic philosophy’ ismore to the point. These philosophersput language at the centre of theirinquiry. This was in sharp contrast withthe customary practice of regarding itas merely (and, in the views of some,lamentably) a tertium quid (third ele-ment) between René Descartes’ resextensa (extended matter or materialreality) and res cogitans (thinking sub-stance or, simply, the mind). In thissense, linguistic philosophy marked amajor departure from the long tradi-tion of analytic philosophy, of which itis nevertheless an offshoot. However,

McCarthy, John and Alan Prince (1995).‘Faithfulness and reduplicative iden-tity’. In Jill Beckman, Laura WalshDickey and Suzanne Urbanczyk (eds),University of Massachusetts OccasionalPapers in Linguistics 18: 249–384.

Further readingMcCarthy, John (2002). A Thematic

Guide to Optimality Theory. Cam-bridge: Cambridge University Press.

Patrick Honeybone

ORDINARYLANGUAGEPHILOSOPHY

Ordinary language philosophy (OLP)is the name given to a philosophicalmovement that developed in Englandduring the years between the twoWorld Wars and enjoyed its heyday inthe late 1940s through the early1960s. Its adherents saw ordinary,everyday language as the startingpoint for their philosophical inquiries.Through the central ideas of its lead-ing figures such as J. L. Austin andLudwig Wittgenstein, OLP continuesto echo in a wide variety of disciplines.

See also: Analytic Philosophy;Implicature; Speech Act Theory;Truth TheoriesKey Thinkers: Austin, J. L.;Derrida, Jacques; Grice, H. P.;Russell, Bertrand; Ryle, Gilbert;Searle, John; Strawson, P. F.;Wittgenstein, Ludwig

Ordinary language philosophy isalso variously referred to as Oxford

150

ORDINARY LANGUAGE PHILOSOPHY

arguing that such an approach wouldclarify matters by blocking misuses oflanguage. In other words, most of phi-losophy, on careful inspection, wouldboil down to straightforward linguisticanalysis.

J. L. Austin, one of the principalexponents of ordinary language phi-losophy, used to quip that if one wantsto embark on a career in philosophyone had better equip oneself with agood dictionary. His point was thatour ordinary, everyday language,unlike its ideal or formal counterpart,was fashioned into its present form bygenerations of speakers. In that longprocess, the language has beeninvested with most, if not all, of thedistinctions that its speakers felt nec-essary. Not that the ordinary languagecannot be improved or further dis-tinctions introduced into it to suitfresh needs. Ordinary language maynot provide us with the last word onphilosophical problems, but it shouldcertainly be the starting point of allphilosophical inquiry.

The philosophical importance ofconferring pride of place upon ordi-nary language can hardly be overesti-mated. In a way it debunked the veryenterprise of philosophy or at least atraditional way of going about doingphilosophy. The image of the proto-typical philosopher is best captured bythe famous statue called ‘The Thinker’by the French sculptor Auguste Rodin.Lost in meditation, the philosopher iscompletely out of touch with thework-a-day world. Introspection is hispreferred modus operandi and soli-tude his self-imposed ambience. This isperfectly in tune with philosophy’sproverbial disdain for language in theeveryday sense of the word. Gottfried

even here, one cannot point to any con-sensus. John Searle, of the University ofCalifornia at Berkeley, probably themost famous of Austin’s followers, hascategorically denied being a linguisticphilosopher, claiming to be only aphilosopher of language.

The term ‘ordinary language phi-losophy’ captures an important ele-ment that distinguishes the movementfrom the work of earlier philosophers,particularly those inspired by logicalpositivism* or what may contrastivelybe referred to as ‘ideal language phi-losophy’. Bertrand Russell, for one,was openly scornful of ordinary lan-guage which he thought was full ofambiguity* and vagueness and henceinadequate for the philosophers’ pur-poses. Unlike many of their con -temporaries and predecessors whobelieved in first positing a logicallyperfect language and then lamentinghow our ordinary, everyday languagepales in comparison with it, ordinarylanguage philosophers insisted onexamining ordinary, everyday lan-guage at its face value.

In stark contrast with the apologistsof ideal language, Wittgenstein arguedin his Philosophical Investigations(1953) that ordinary language was per-fectly in order as it stood, and thatmany of the puzzles that professionalphilosophers and linguists encounteredand sought to resolve were actually theresult of an inadequate understandingof the subtleties of ordinary, everydaylanguage. In his essay ‘SystematicallyMisleading Expressions’, Ryle (1932)made a strong case for a careful analy-sis of ordinary language expressions asa way of doing philosophy, or rather,as a way of ‘dissolving’ many of the problems that crop up in philosophy,

151

ORDINARY LANGUAGE PHILOSOPHY

the hilt. True to the austere tradition ofsober empiricism, the hallmark ofBritish thought, he preferred to starthis philosophical inquiry with the tan-gibles – to start from the concrete(words) and work his way towards thegeneral or the abstract, rather than theother way around. And he was in nohurry to finish that job. He believedthat philosophy demanded a great dealof painstaking spadework. He oncesaid that the sentence ‘Neither a be-allnor an end-all be’ could make an excel-lent motto for philosophy.

Another essay by Austin, ‘A Plea forExcuses’, is widely regarded as aprime example of OLP at work. Bycarefully teasing out the different usesof the word ‘excuse’ (which excul-pates the doer of an action), Austindistinguishes it from ‘justification’(which, by contrast, ‘absolves’, as itwere, the action itself of any imputa-tion of wrong-doing). This essay alsohighlights what may be seen as yetanother hallmark of OLP – deflectingthe focus from the naming of an act(along with the hypostatisation that itinvariably involves) to the more mun-dane doing of the act or, simply put,the action itself. This idea found itsfull expression in How to Do Thingswith Words (1962a), undoubtedlyAustin’s most famous work. Beforeproceeding any further, it is importantto comment on the title of the book,originally presented as a series oftwelve lectures presented at the Uni-versity of Harvard in 1955 under thetitle ‘Words and Deeds’ and publishedposthumously. It has been suggestedthat Austin chose that title somewhatlightheartedly after Dale Carnegie’sbestselling How to Win Friends andInfluence People. In fact, many an

Wilhelm Leibniz, the German philo -sopher, is famously said to haveexclaimed that if God Almighty wereone day to descend upon the earth, Hewould certainly address us ordinarymortals in the language of mathemat-ics – it was inconceivable for thephilosopher that an all-perfect Godwould have recourse to any languageother than the all-perfect language ofmathematics.

Traditionally, philosophy is con-cerned with conceptual analysis. It isnot primarily concerned with words.Words are believed to embody con-cepts at best. For the most part, tradi-tional philosophers distrusted words,judging them misleading representa-tions of the concepts behind them.Recall the Socratic practice, illus-trated in several of Plato’s Dialogues,of insisting that his interlocutor pro-vide a definition of, say, piety, insteadof pointing to examples of pious per-sons. The moral is that true under-standing comes from pure conceptualanalysis which is what a definition isall about. In Socrates’ view, examplesgive the false impression that one hasgot to grips with the concepts behindthem.

Austin’s attitude to this time- honoured practice among philoso-phers was eloquently expressed in apaper entitled ‘Are there any a prioriconcepts?’, originally presented in1939 before the Aristotelian Society. Init Austin surprised his audience byarguing that he had no idea of whatthat question meant because he did notknow what concepts were to beginwith. So the question of concepts beinga priori or a posteriori simply did notarise. On this question and on manyothers, Austin was an Aristotelian to

152

ORDINARY LANGUAGE PHILOSOPHY

expected answer would have it, ‘abarn’. Austin thus challenges whatmay be described as a mainstay of Pla-tonic realism, namely, the idea thatepistemology can only play secondfiddle to ontology or, to put it moresimply, what there is can by no meansbe affected by what one comes toknow about it.

The performative view of languagethus has important things to say inrespect of what has, from time imme-morial, been the philosopher’s HolyGrail: the concept of truth. P. F. Straw-son advanced what is referred to asthe ‘performative theory of truth’according to which truth is not some-thing to be approached solely with thetools of the trade available in philoso-phy in its conventional sense, butsomething invariably mediated bylanguage. Truth claims, in otherwords, take precedence over truthipsis. To call something true is to per-form the speech act of endorsing it orgiving it one’s stamp of approval.

Historians of philosophy often treatOLP as a chapter in the unfinishedbook of analytic philosophy but thepoint is debatable. Both the laterWittgenstein and Austin producedimmense challenges to what was theestablished dogma in philosophy. Per-haps the best proof of the radicalnature of their philosophical positionsis the publication in 1959 of a book byErnest Gellner entitled Words andThings: An Examination of, and anAttack on, Linguistic Philosophy.The book even carried a laudatorypreface by Bertrand Russell in whichhe lamented that his former pupilWittgenstein seemed to have taken aholiday from serious intellectual activ-ity, meaning thereby that he considered

Austin reader has a rather difficulttime coming to terms with what is bestdescribed as his blithe irreverence forthe sober, lofty style that one typicallyassociates with philosophy, and hisuse of a thoroughly colloquial style.But it is important to realise thatAustin is making an important philo-sophical point here (Rajagopalan2000a).

In How to Do Things with WordsAustin argued that a sentence such as‘The cat is on the mat’, which has longbeen seen as a declarative (or, ‘consta-tive’ as he calls it) and hence capableof being judged true or false, is actu-ally a ‘performative’* which cannotbe judged true or false but only happyor otherwise, depending on the cir-cumstances in which it is uttered or‘performed’. Thus was launched theidea that speaking a language is per-forming a series of acts. This seminalidea has inspired a number of scholarsfrom a number of diverse disciplinesincluding linguistics, psychology,anthropology, sociology, and evensuch unlikely fields as economics.

A discussion of Austin’s legacywould be incomplete without a men-tion of his book Sense and Sensibilia(1962b) (the echo of Jane Austen’s cel-ebrated novel is unmistakable here).Austin takes on the fashionableapproach to the analysis of perceptionin terms of sense data* and proceedsto deconstruct the so-called ‘argumentfrom illusion’. For instance, he says,the right answer to the question as towhat it is that you actually see, posedapropos of a church camouflaged soas to look like a barn, is precisely whatthat description says, namely, ‘achurch camouflaged so as to look likea barn’ and not, as the traditionally

153

ORDINARY LANGUAGE PHILOSOPHY

he was widely acclaimed as Austin’sintellectual legatee. The book waswelcomed by many as genuinely Aus-tinian in spirit and an improvementupon Austin’s own How to Do Thingswith Words, which had left manyreaders perplexed and unsure of justwhat Austin had in mind. What manyreaders of Searle’s book failed to takeserious notice of was something Searlehad declared right in the subtitle of hisbook: An Essay in the Philosophy ofLanguage. The philosophy of lan-guage casts a much wider net than lin-guistic philosophy, which is one wayof philosophically looking at languagebut not the only one. In fact, in worksince then, Searle has evinced littlesympathy for many of the trademarkpositions assumed by linguistic philo -sophers, notably Austin himself(Rajagopalan 2000b).

It has been argued that Searle rein-terpreted Austin and put his ideasback on the beaten track of traditionalanalytic philosophy. In the process,many of the potentially subversive ele-ments in Austin’s thought were eitherdownplayed or simply ignored. With-out doubt this made Austin a house-hold name in disciplines such aslinguistics, but many critics have com-plained that the Austin appropriatedby Searle is a far cry from the Austinwho so vigorously and implacablydefended linguistic philosophy.

Be that as it may, the later Wittgen-stein and Austin have had a consider-able influence elsewhere in academia,and continue to do so. Wittgenstein’sinfluence on contemporary linguisticthought is undeniable. His insistenceon the normative character of lan-guage has profound implications forcontemporary emphasis on ethics and

OLP at best a diversion from philoso-phy proper and at worst an aberration.But the very fact that the book becamea major sensation immediately after itwas published speaks volumes for thepopularity of OLP at that time and thethreat it represented to mainstreamphilosophy. No doubt Gellner’s bookdid quite some damage to OLP, thoughonce the heat of the moment had dieddown, many scholars were of the opin-ion that the book threw very little lighton substantive issues, instead content-ing itself with mud-slinging and vitu-perative rhetoric.

Perhaps the real damage to OLP wasdone not by die-hard detractors likeGellner, but by those responsiblefor clever appropriations of the her-itage of some of its principal figures,notably Wittgenstein and Austin. Wehave already seen how Bertrand Russell, a towering figure in Britishphilosophy who personally supervisedWittgenstein’s earlier work TratactatusLogico-Philosophicus (Wittgenstein1922/1961), simply brushed aside hislater work, Philosophical Investiga-tions (Wittgenstein 1953/1968), dis-missing it as at best frivolous. Thisfurnished the vital clue to many suc-ceeding generations to stay with theearlier Wittgenstein and to summarilydismiss the later Wittgenstein.

With Austin, the philosophicalestablishment adopted a differenttactic. After his untimely death in1960, his mantle was widely consid-ered to have fallen on John Searle whohad studied at Oxford in the early1950s and who therefore knew himpersonally. In 1969 Searle publishedhis book Speech Acts: An Essay in thePhilosophy of Language which wasan instant success in linguistics, where

154

ORDINARY LANGUAGE PHILOSOPHY

Judith Butler’s notions of ‘performa-tivity’ and ‘performance’ (Butler1997), both inspired by Austin’s ‘per-formative utterance’. Building onAustin’s insights, Butler argues thatgender is not a matter of what one is(constative), but rather what one does(performative). While OLP is oftenassociated with a specific time andplace, its influence goes much widerthan linguistics and the philosophy oflanguage thanks to the ongoing workof scholars in other disciplines whofind inspiration in its central tenets.

Primary sourcesAustin, J. L. (1939). ‘Are there a priori

concepts?’. In J. L. Austin (1961), Philo-sophical Papers. London: Oxford Uni-versity Press. 32–54.

Austin, J. L. (1956). ‘A plea for excuses’.In C. Lyas (ed.) (1971), Philosophy andLinguistics. London: Macmillan Press.79–101. Originally published in theProceedings of the Aristotelian Society.1956–7. Reprinted in J. L. Austin(1961), Philosophical Papers. London:Oxford University Press. 175–204.

Austin, J. L . (1961). Philosophical Papers.London: Oxford University Press.

Austin, J. L. (1962a). How to Do Thingswith Words. Oxford: Clarendon Press.

Austin, J. L. (1962b). Sense and Sensibilia.Oxford: Clarendon Press.

Gellner, Ernest (1959). Words and Things:An Examination of, and an Attack on,Linguistic Philosophy. London: VictorGollancz.

Ryle, Gilbert (1932). ‘Systematically mis-leading expressions’. In Richard Rorty(ed.) (1967), The Linguistic Turn.Chicago: University of Chicago Press.

Searle, J. R. (1969). Speech Acts: An Essayin the Philosophy of Language. Cam-bridge: Cambridge University Press.

politics in relation to language. Notethat Wittgenstein’s position is in starkcontrast with the position assumed inmuch of linguistics where the beliefhas been that a grammar must ideallybe descriptive rather than prescrip-tive. Another area that has been influ-enced by Wittgenstein’s thought iscognitive science. The theory of pro-totypes* – a model of graded cate-gorisation that eschews the classic,Aristotelian distinction betweenessences and accidents – is a directspin-off from Wittgenstein’s notion of‘family resemblance’. The idea thatclass membership does not have to bea matter of all or nothing, but can bea matter of more or less, has impor-tant consequences for how differentsocieties categorise entities in thework-a-day world (Rosch and Lloyd1978; Lakoff 1987).

Austin’s influence on contemporarythought is less often acknowledged,thanks to the widespread tendency tocredit it to Searle instead. JacquesDerrida has made no secret of theenormous influence of Austin’s think-ing on his own. He even becameinvolved in a protracted exchangewith Searle over the legacy of Austin.An unlikely area of inquiry taken bystorm in the 1980s, thanks to theinfluence of OLP, was economics. In1985 Deirdre (then, Donald) Mac-Closkey published a book in whichshe argued that economists are closerto poets and story-tellers than theythink, and that their claims are mistakenly taken to be constative,whereas they are just as performativeas any other. Perhaps a most signifi-cant testimony to the vitality of lin-guistic philosophy in general andAustin’s heritage in particular is

155

PERFORMATIVE

A description of utterances, sentencesand speech acts the use of which ismeant to create facts in addition to thefact of one of them having been used.Often, the appearance of ‘hereby’marks the performative nature of theexpression, as in ‘Master Little John ishereby created Sheriff of Nottingham’(Robin Hood).

See also: Speech Act TheoryKey Thinkers: Austin, J. L.; Searle,John

For a long period, philosophers of lan-guage and logicians of natural lan-guage were interested in sentences theutterance of which in appropriatecontexts expresses propositions thatare either true or false, for example,‘Shakespeare is the author of Romeoand Juliet’ or ‘All ravens are white’. Itwas always clear that not all sentencesare of that type, for example Englishsentences in the interrogative or theimperative. It was J. L. Austin (1961)who drew attention to sentences of aseemingly different type that are quitecommonly used, the ‘performatives’.

The leading intuition underlyingthe distinction between propositionalutterances and performative ones isthat in the former case the story ofthe utterance, as commonly intended,conveyed and understood, is that ofpresenting a given fact about the stateof affairs, while in the latter case thestory involves some additional ele-ment that creates a new fact. Thus sin-cerely saying ‘The door is open’ isdescribing what the speaker takesto be a fact about the door, underthe circumstances of utterance, while

Wittgenstein, Ludwig (1922/1961). Trac-tatus Logico-Philosophicus. London:Routledge and Kegan Paul.

Wittgenstein, Ludwig (1953/1968). Philo-sophical Investigations. Oxford: BasilBlackwell.

Further readingButler, Judith (1997). Excitable Speech: A

Politics of the Performative. New York:Routledge.

Hacking, Ian (1975). Why Does LanguageMatter to Philosophy? Cambridge:Cambridge University Press.

Lakoff, George (1987). Women, Fire andDangerous Things: What CategoriesReveal about the Mind. Chicago: Uni-versity of Chicago Press.

McCloskey, Deirdre (1985). The AppliedTheory of Price. Second edition. NewYork: Macmillan.

Rajagopalan, Kanavillil (2000a). ‘Austin’shumorous style of philosophical dis-course in light of Schrempp’s interpreta-tion of Oring’s “incongruity theory” ofhumor’. Humor: An International Jour-nal of Humor Research 13 (3): 287–311.

Rajagopalan, Kanavillil (2000b). ‘OnSearle [on Austin] on language’.Language and Communication 20 (4):347–91.

Rajagopalan, Kanavillil (2004a). ‘JohnLangshaw Austin’. In Philipp Strazny(ed.) (2004), Encyclopedia of Linguistics.New York: Fitzroy Dearborn. 98–100.

Rajagopalan, Kanavillil (2004b). ‘JohnSearle’. In Philipp Strazny (ed.) (2004),Encyclopedia of Linguistics. New York:Fitzroy Dearborn. 936–8.

Rosch, Eleanor and Barbara B. Lloyd (eds)(1978). Cognition and Categorization.Hillsdale, NJ: Lawrence Erlbaum Associates.

Kanavillil Rajagopalan

156

PHONEME

Bach K. and R. M. Harnish (1992). ‘Howperformatives really work’. Linguisticsand Philosophy 15: 93–110.

Searle, John (1989). ‘How performativeswork’. Linguistics and Philosophy 12:535–58.

Asa Kasher

PHONEME

A term generally understood in theBritish and American structuralist tra -ditions to refer to a fundamentalabstract linguistic unit physically re -alised by multiple context- dependent,phonetically similar speech sounds,and possessing the capacity to changemeaning but bearing no semanticproperties of its own. Several defini-tions of the term exist, and disagree-ment persists over when and by whomit was first used.

See also: Distinctive Features;Emic/Etic; Generative Phonology;Optimality Theory;Poststructuralism; Type/TokenKey Thinkers: Bloomfield,Leonard; Firth, J. R.; Hockett,Charles; Jakobson, Roman; Jones,Daniel; Martinet, André; Pike,Kenneth; Saussure, Ferdinand de;Trubetzkoy, N. S.; Whorf,Benjamin Lee

Daniel Jones (1967) characterisesphonemes as ‘small families ofsounds, each family consisting of animportant sound of the languagetogether with other related soundswhich, so to speak, “represent” it inparticular sequences or under particu-lar conditions of length or stress or

sincerely saying ‘I promise to returnthe book’ is creating a commitment onthe part of the speaker to return thebook. The commitment is created bythe linguistic institution of promising.According to the rules that constituteand regulate that institution, utteringa sentence of the form ‘I promise to dothis and that’ counts as undertaking acommitment to do this and that.

Promising is a linguistic institution.Appointing is a non-linguistic institu-tion, according to the rules of whichan utterance of ‘I hereby appoint youdeputy sheriff’, under appropriate cir-cumstances, counts as creating you adeputy sheriff. What is common toboth linguistic and non-linguistic insti-tutions is that they determine felicityconditions for creating new facts, suchas a certain person having a certaincommitment or serving in a certaincapacity, which are institutional facts.

The question whether a performa-tive utterance is an utterance of anassertion has been debated. Searle(1989) answers it in the negative,while Bach and Harnish (1992) arguethat such an utterance is primarilyassertoric.

According to the debatable Perfor-mative Hypothesis developed withingenerative semantics*, the deep struc-ture of every sentence, includingindicative ones, includes a prefix ofthe form ‘I + (performative) verb’.Hence, every utterance is made withinthe framework of some institution.

Primary sourcesAustin, J. L. (1961) ‘Performative Utter-

ances’. In J. O. Urmson and G. J.Warnock (eds), J. L. Austin, Philosoph-ical Papers. Oxford: Clarendon Press.233–52.

157

POLITENESS

taken as self-evident by researchers indomains outside theoretical phonol-ogy, for instance in the study of firstand second language acquisition,speech and hearing disorders andremediation, speech technology, psy-cholinguistics and psychology moregenerally. The phoneme is also arguedto be a structural element of signedlanguages.

Primary worksJones, Daniel (1957). The History and

Meaning of the Term ‘Phoneme’.London: International Phonetic Associ-ation.

Jones, Daniel (1967). The Phoneme: ItsNature and Use. Third edition. London:W. Heffer & Sons.

Swadesh, Morris (1934). ‘The phonemicprinciple’, Language Vol. 10: 117–29.

Further readingAnderson, Stephen R. (1985). Phonology

in the Twentieth Century: Theories ofRules and Theories of Representations.Chicago: University of Chicago Press.

Goldsmith, John (ed.) (1996). The Hand-book of Phonological Theory. Oxford:Blackwell.

Dominic Watt

POLITENESS

The formulation of linguistic utter-ances in such a way that they con-tribute to good social relationshipsbetween participants in an exchange.Since their inception in the 1970s,politeness studies have developed intoone of the most active research areasin pragmatics and sociolinguistics,

intonation’ (p. 7). These family mem-bers are known as allophones (derivedfrom the Greek for ‘other sound’) fol-lowing Benjamin Lee Whorf’s first useof the term in the early 1930s.

A language’s phonemes stand inrelationships of contrast with oneanother in that commutation (switch-ing) of one phoneme for another in agiven phonological frame through thecreation of a ‘minimal pair’ (for exam-ple, substituting /p/ for /b/ in ‘big’ toproduce ‘pig’) brings about a changein the words meaning. Phonemes aretherefore said to be in ‘parallel distri-bution’. Their allophones, which arecontextually determined, are said tobe in ‘complementary distribution’.

In The History and Meaning of theTerm ‘Phoneme’ (1957), Jones attrib-utes the simultaneous ‘discovery’ ofthe phoneme to the late nineteenthand early twentieth-century linguistsBaudouin de Courtenay and HenrySweet, although elsewhere (Jones1967) he cites J. R. Firth’s contentionthat the term ‘phoneme’ as distinctfrom ‘phone’ (that is, an actual speechsound) was coined in 1879 by NikolajKruszewski, one of Baudouin deCourtenay’s students. Jones points tovarious kinds of evidence indicatingthat all spoken language users haveintuitions about phonemic units, andcites the development of alphabeticwriting systems as a good example ofevidence of the ‘natural sense’ inwhich native speakers considersounds to ‘have a kind of sameness’,even when their phonetic forms differfrom each other.

Although the phoneme was super-seded in the 1980s by non-linearmodels of phonological representa-tion, its existence still appears to be

158

POLITENESS

endearment terms and in-group lan-guage), and negative face satisfiedthrough avoidance of contact (as sig-nalled through ‘for example, ques-tioning and hedging). These twoaspects of face allow for a four-wayclassification of speech acts into actsthat inherently threaten the speaker’sor the hearer’s positive or negativeface. Prior to performing an act x thatinherently threatens face, known as aface threatening act or FTA, thespeaker assesses the weightiness ofthe threat using the formulaWx=D(S,H)+P(H,S)+Rx. In this for-mula, D(S,H) stands for the distancebetween the speaker and the hearer, asymmetric measure of familiarity orsimilarity between them; P(H,S) is thepower of the hearer over the speaker,an asymmetric measure of the amountof control the hearer can exercise overthe speaker; and Rx is the ranking, orseriousness, of the imposition entailedby x in the culture in question. Thesum of these three sociological valuesguides the speaker’s choice among fivestrategies for the realisation of FTAs,ranging from ‘Bald on record’ (Wx isnegligible), to ‘Don’t do the FTA’ (Wxis very high) (see Figure 4). The bulkof Brown and Levinson’s essay is ded-icated to cataloguing the linguisticexpressions that may realise the firstfour strategies (since the last oneresults in silence), which they illus-trate with examples from English,Tamil (a Dravidian language) andTzeltal (a Mayan language).

It is hard to overestimate the impactof Brown and Levinson’s model tofield of politeness studies. Integrat-ing politeness with anthropologicalnotions such as face, as well as the the-oretical frameworks of implicature*

with several studies focusing on therealisation of politeness in differentcultures.

See also: Implicature; Speech ActTheory; Conversation AnalysisKey Thinkers: Grice, H. P.; Searle,John

The importance of politeness as afactor motivating particular turns ofphrase was acknowledged early on bylinguists such as Charles Bally andOtto Jespersen. Later, it was alsoacknowledged by H. P. Grice (1967)and John Searle (1975), who associ-ated it with departures from rationalefficiency and with indirectness,respectively. However, it was not untilthe 1970s with Robin Lakoff’s workon the rules of politeness (1973) and,most notably, the publication of Pene-lope Brown and Stephen Levinson’sessay ‘Politeness: Universals in Lan-guage Usage’ (1978) that the inte -gration of politeness into existingtheoretical frameworks became a seri-ous theoretical concern. A major con-tribution of Brown and Levinson layin proposing the notion of face,inspired by Ervine Goffman’s work onritual aspects of everyday exchanges,as the unifying principle underlyingthe expression of politeness throughlanguage.

Brown and Levinson distinguishtwo aspects of face: positive facerefers to the desire to be liked andapproved of, while negative face cor-responds to the wish for privacy andfreedom from imposition. The twoaspects of face may thus be viewed aspulling in opposite directions, withpositive face promoting sociabilityand contact with others (as signalled,for example, through, the use of

Est

imat

ion

of r

isk

of f

ace

loss

Lesser

Greater

5. Don’t do the FTA

4. Off record

with redressive action

on record

1. without redressive action, baldly

2. Positive politeness

Do the FTA

3. Negative politeness

Figure 4 Strategies for performing FTAs (Brown and Levinson 1987: 60)

159

POLITENESS

and surrounding social conventions(an aspect of ‘discernment’) with con-comitant over-emphasising of individ-ual rationality (also termed ‘volition’);and the universal applicability of theirclaims.

More recently, and in tandem withthe challenging of Gricean and Sear-lean accounts of meaning as not flexible enough to account for theongoing co-construction of meaningby participants in interaction, a newset of concerns focusing on social theoretic aspects of politeness andfavouring a more holistic approach topoliteness phenomena have emerged.Responding to a paradigm shiftwithin politeness studies, recent stud-ies distinguish between first-orderpoliteness (Politeness1), correspon-ding to participants’ own definitionsand perceptions of politeness in inter-action, and second-order politeness(Politeness2), corresponding to thetechnical definition of politenessby researchers, building on, for exam-ple, anthropological notions suchas face. Although researchers do not

and Speech Act Theory*, their workhas been instrumental in shaping ourthinking about how human dyadicrelationships are reflected in, and con-stituted through, language. Moreover,their comprehensive analysis of field-work data has inspired a multitude ofarticles and books exploring politenessphenomena in a variety of languagesand cultures, often using the data col-lection method of Discourse Comple-tion Tests (DCTs). Nevertheless, theseworks have also produced criticismsand refinements of Brown and Levin-son’s model. Some of the most impor-tant ones concern the definition of faceand prioritising of negative over posi-tive aspects; their focus on face threat-ening acts to the exclusion of faceenhancing/boosting acts; the defini-tions and adequacy of the three socio-logical variables and their assumedindependence; the association ofpoliteness with degree of indirectness;the nature and content of politenessimplicatures; the (inscrutable) role ofsilence in their model; the little atten-tion paid to the role of the audience

160

POLITENESS

this debate. In other words, the fre-quent, contextually-conditioned asso-ciation of a particular perlocutionaryeffect with a particular expression incorpus data, if such association can beestablished, provides an empiricallyverifiable baseline as to what consti-tutes the ‘unmarked’ case, and an analyst-independent vantage pointfrom which to locate and analyse var-ious cases as ‘marked’.

Politeness is neither inherent in lin-guistic forms in isolation from theircontext of utterance (it is not a matterof structure), nor does it reside(wholly) in the speaker’s intention,independently of its recognition by thehearer (it is not a matter of agency).Rather, politeness may be viewed asthe mutual constitution of face by par-ticipants in an exchange, which isgreatly facilitated by their partakingof similar societal norms – what maybe described in terms drawing onBourdieu as their having developed‘homologous habitus’. Approaches topoliteness have thus increasinglyturned to social-theoretic notions,such as habitus and communities ofpractice, to analyse the contributionof language in constructing, maintain-ing and endangering good social rela-tionships.

Primary sourcesBrown, Penelope and Stephen Levinson

(1978). ‘Politeness: universals in lan-guage usage’. In Goody, E. (ed.), Ques-tions and Politeness: Strategies in SocialInteraction. Cambridge: CambridgeUniversity Press. 56–324. Reprinted asBrown, Penelope and Stephen Levinson,(1987) Politeness: Some Universals inLanguage Usage. Cambridge: Cam-bridge University Press.

necessarily agree on which of the twoshould be the focus of scholarlyanalysis, the distinction betweenPoliteness1 and Politeness2 has sev-eral important consequences, includ-ing the possibility of finer gradationbetween behaviour that is merely ade-quate (now termed ‘politic’), andbehaviour that goes beyond that (nowtermed ‘polite’), with behaviour thatis inappropriate falling at either endof a continuum that ranges from‘over-polite’ to ‘impolite.’ In this way,the question of the scope of a theoryof politeness is also raised, with impo-liteness/rudeness increasingly attract-ing scholarly attention.

At the level of methodology, thefocus has shifted from isolated utter-ances to longer chunks of discourse,which are often analysed using con-versation-analytic tools. Rather thanaiming at providing an inventory ofdevices, either lexical or structural, bywhich politeness is expressed, asin earlier studies, emphasis is nowplaced on the utterance situation as awhole, including the addressee’sreception of the speaker’s utterance,prosodic aspects, and any paralinguis-tic cues available. The use of recordedconversational data is paramount inthis respect. Further to the qualitativeanalysis of these data, the value oftheir quantitative analysis is alsoincreasingly acknowledged. Suchquantitative analysis presupposes theavailability of large conversationalcorpora*, in which regularities ofusage (or norms) may be investigated.While the existence of such norms,and hence their value as analytic tools,have been brought into question,quantitative analysis of conversa-tional corpora may hold the key to

161

POLITICALCORRECTNESS

In its most general sense the term‘political correctness’ refers to an indi-vidual’s or group’s conscious avoid-ance of linguistic terms associatedprincipally with race, ethnicity,nationality, gender, religious beliefsystem and sexual orientation that areperceived to be pejorative or at anyrate to have the potential to causeoffence to others. A more specialisedsense refers to an uncritical adherenceto and advocacy of a particular polit-ical credo, such as Marxism.

See also: Conversation Analysis;(Critical) Discourse Analysis;Feminism; Metaphor; Names;PolitenessKey Thinkers: Cameron, Deborah;Milroy, Lesley; Tannen, Deborah;Whorf, Benjamin Lee

The term ‘political correctness’ origi-nated on American campuses amongNew Left activists in the 1960s and1970s, when it was used primarily inconnection with feminist criticisms ofperceived sexism in language. Nowa-days the term is used mainly by criticsof proposed linguistic and socialreforms and is therefore almost exclu-sively pejorative. Opponents of politi-cal correctness claim that is is overlyconcerned with trivial linguistic pre-scriptivism. Those on the other side ofthe debate draw on a form of ‘weakWhorfianism’ to claim that languageinfluences perception, therefore thatlinguistic change is a necessary pre-cursor to changing social attitudes.

While practically all language usersobserve the conventions prohibiting

Lakoff, Robin (1973). ‘The logic of polite-ness; or minding your p’s and qs’. InPapers from the Ninth Regional Meeting of the Chicago Linguistic Soci-ety. Chicago: Chicago Linguistic Soci-ety. 292–305.

Leech, Geoffrey (1983). Principles ofPragmatics. London: Longman.

Further readingBousfield, Derek and Miriam Locher (eds)

(2007). Impoliteness in Language.Berlin: Mouton de Gruyter.

Eelen, Gino (2001). A Critique of Polite-ness Theories. Manchester: St Jerome.

Grice, Herbert Paul (1967). ‘Logic andconversation’. William James Lectures,Harvard University typescript. In P. Coleand J. Morgan (eds) (1975), Syntax andSemantics. Vol. III: Speech Acts. NewYork: Academic Press. 41–58. Reprintedin Grice, H. P. (1989), Studies in the Wayof Words. Cambridge, MA: HarvardUniversity Press. 22–40.

Kerbrat-Orecchioni, Catherine (2005).‘La Politesse dans le discours en interac-tion’. In C. Kerbrat-Orecchioni. Le Dis-cours en Interaction. Paris: Colin.187–284.

Searle, John (1975). ‘Indirect speech acts’.In P. Cole and J. Morgan (eds.), syntaxand Semantics. Vol. III: Speech Acts.New York: Academic Press. 59–82.

Terkourafi, Marina (2005). ‘Beyond themicro-level in politeness research’, Jour-nal of Politeness Research Vol. 1:2,237–62.

Watts, Richard (2003). Politeness. Cam-bridge: Cambridge University Press.

Watts, Richard, Sachiko Ide and KonradEhlich (eds) (2005). Politeness in Lan-guage. Second edition. Berlin: Moutonde Gruyter.

Marina Terkourafi

162

POLITICAL CORRECTNESS

forms of legislation seeking to pro-scribe the public use of ‘offensive’ lan-guage. Since 2000 extensions toexisting linguistic legislation havebeen made in the United Kingdom, forexample, ostensibly in the interests ofprotecting religious minorities frompersecution in the aftermath of theevents of 11 September 2001. TheRacial and Religious Hatred Act 2006specifies that it will be an offence toincite (or ‘stir up’) hatred against aperson on the grounds of his or herreligion, by prohibiting ‘threateningwords’ and ‘the display of any writtenmaterial which is threatening’ if theintention by so doing is to stir up reli-gious hatred.

Resistance to the introduction ofsuch measures takes several forms.Key among these are the holding up ofPC language to public ridicule, forexample James Finn Garner’s series of‘politically correct stories’, and theappropriation (and thereby subver-sion) of ‘non-PC’ terms such as‘queen’ and ‘nigger’ by the gay and theblack communities, respectively.

Primary worksCameron, Deborah (1995). Verbal

Hygiene. London: Routledge.Dunant, Sarah (1995). The War of the

Words: The Political CorrectnessDebate. London: Virago.

Further readingFairclough, Norman (2003). ‘Political cor-

rectness: the politics of culture and lan-guage’. Discourse and Society Vol. 14:1,17–28.

Holborow, Marnie (1999). The Politics ofEnglish. London: Sage.

Dominic Watt

the public use of taboo forms, thedegree of sanction attached to individ-ual words or phrases in the socialdomains mentioned above is highlyfluid, such that terms consideredacceptable at a particular place or timemay quickly become stigmatised ifthey are considered insufficientlypolitically correct or ‘PC’ by one ormore politically or economically influ-ential groups. Awareness of the oftenserious consequences of deliberatelyor accidentally flouting the conven-tions of politically correct languagehave strongly influenced linguistichabits at almost all levels of society inEnglish-speaking countries, perhapsmost particularly in the United States,where a high degree of sensitivity tocertain terms perceived to betray racistattitudes has resulted in the sacking ofseveral high-profile public figures. Forexample, the use of the word ‘coon’ byAmerican radio talk-show host DaveLenihan in an item about US Secretaryof State Condoleezza Rice broadcastin 2006 was apparently accidental (heclaims to have intended to say ‘coup’),while the previous year Las Vegas TVweatherman Rob Blair’s alleged use ofthe same form, supposedly substitutedfor ‘King’ in the phrase ‘Martin LutherKing Junior Day’, could plausiblyhave resulted from an anticipatoryspeech error. Both Lenihan and Blairwere nonetheless dismissed from theirposts almost immediately. In 1999vociferous objections to the contextu-ally valid use in a private meeting ofthe word ‘niggardly’ by DavidHoward, an aide to the mayorof Washington DC, led to Howard’sresignation.

Political correctness is tightlybound to censorship, and to other

163

modifications from the scholasticstudies. Propositions* are understoodto be complexes formed by acts of will(judgement) that combine a subjectidea with a predicate idea, with theverb carrying affirmative force; a neg-ative judgement is the converse, wherethe predicate is separated from thesubject. The Logic falls prey here tonot being able to distinguish a propo-sition from an attitude towards it. Thereasoning, however, brings into reliefthe ‘unity problem’ that was to bedevilBertrand Russell.

The Logic also distinguishesbetween explicating and determiningsubordinate propositions (relativeclauses). For example, ‘The invisibleGod created the visible universe’ canbe analysed as ‘The God who is invis-ible created the universe which is visi-ble’, where the first relative clauseexplicates its subject, and the secondfurther determines its subject. Thisdevelopment was perhaps the firstformal account of the analytic/syn-thetic* distinction.

The Logic fell into neglect after therise of modern function-argumentlogic and the prevailing behav-iourism* empiricism* in the first partof the twentieth century. In the mid-1960s, however, Chomsky arguedthat the Port-Royal approach was thehigh point of a Cartesian first ‘cogni-tive revolution’, in its search forunderlying mental structures behindthe surface of language. AlthoughChomsky’s historical remarks havebeen broadly misconstrued, they didstimulate fresh interest in Port-Royaland both the Logic and Grammar arenow recognised as the forerunners ofmuch that is characteristic in modernlinguistic analysis.

PORT-ROYAL LOGIC

Colloquial name for La Logique oul’Art de Penser (Logic or the Art ofThinking) written by Antoine Arnauldand Pierre Nicole but publishedanonymously in 1662 (the fifth edi-tion was published in 1683). TheLogic elaborated a traditional syllo-gistic logic wedded to a Cartesianepistemology and metaphysics. Thework also contains many insights innatural language syntax that were notsurpassed until the work of NoamChomsky in the mid-twentieth cen-tury.

See also: Analytic/Synthetic;Empiricism/Rationalism; Logic;Propositional Attitudes;Transformational-GenerativeGrammarKey Thinkers: Aristotle; Arnauld,Antoine; Chomsky, Noam;Descartes, René; Frege, Gottlob;Russell, Bertrand

With its companion, GrammaireGénérale et Raisonnée de Port-Royal(published by Arnauld and ClaudeLancelot in 1660), the Logic seeksto show that the new Cartesian philosophy of ideas provides a gen-eral account of judgement and rea-soning and a sound basis for ademarcation of ‘good’ and ‘bad’arguments. It further provides manydetailed analyses of how linguistic‘surface forms’ systematically mis-match the complex combination ofideas they express.

Following Aristotle, the Logic categorises judgements in terms ofquantity and quality, and adoptsthe standard syllogistic forms with

164

POSSIBLE WORD SEMANTICS

Key Thinkers: Carnap, Rudolf;Kripke, Saul; Leibniz, GottfriedWilhelm; Lewis, David; Montague,Richard

From a common-sense point of viewthe idea of possible worlds may seemat odds with normal intuitions sinceone may think the world is what it is;what it might have been instead is justa matter of speculation. But Leibnizsaw the world as a manifestation ofGod’s will and in God’s mind there arean infinite number of possible worlds,from which God has chosen this one.Since God is both benevolent andomnipotent He has chosen the best ofall the possible worlds. When takingup the idea in a modern context, thefirst kind of concern is what is under-stood by the concept ‘possible world’if it is not the notion proposed byLeibniz. It may most easily be cap-tured in the perspective of the choiceswe make as human beings. Considerthe sentence:

(1) John bought a red car yesterday

Whoever John is, he chose to buy a carand he chose the colour red. Recon-structing the situation in retrospect weknow that he (i) might have chosen notto buy a car and, if anyhow he did buya car, he (ii) might have chosen anothercolour. Had he chosen otherwise, theworld would also have been otherwisein the respect that John might have nocar or a car with another colour. Wecan also, in retrospect, construe thesituation in which John made hischoices and imagine the scenarios thatwere going through his mind at the cardealer: (i) car versus . no car, and (ii)red versus other colours. Each sce-nario represents a minor difference in

Primary sourcesArnauld, A. and P. Nicole (1996). Logic or

the Art of Thinking. Trans. and ed. J. V.Buroker. Cambridge: Cambridge Uni-versity Press.

Arnauld, A. and C. Lancelot (1975). Gen-eral and Rational Grammar The Port-Royal Grammar. Trans. and ed. J.Rieux and B. E. Rollin. The Hague:Mouton.

Further readingChomsky, N. (1966). Cartesian Linguis-

tics. New York: Harper and Row.

John Collins

POSSIBLE WORLDSEMANTICS

A theory of formal and natural lan-guage semantics that assumes truth-conditions to be the meaning oflinguistic expressions and sentences,and claims that the truth value* ofparticular sentences relative either tothe actual world or to one or more ofthe other possible worlds is whatmakes any unique sentence true orfalse, thereby establishing its mean-ing. The idea goes back to GottfriedWilhelm Leibniz and has been elabo-rated by Rudolf Carnap, but the mainfigures of the modern version, espe-cially associated with modal logic,were Saul Kripke and David Lewis.Possible world semantics was, to someextent, utilised by Richard Montague.

See also: Connotation/Denotation;Correspondence Theory; DefiniteDescriptions; Logic; Modality;Names; Sense/Reference

165

POSSIBLE WORLD SEMANTICS

world – what then is the meaning of afalse sentence? Can a lack of corre-spondence be the meaning of a sen-tence? Possible world semantics maynot have a clear answer to this.

Another problem is the questionabout the relation between words andthe world, in that we may ask: whatworld? It seems evident that we cantalk directly about things in our envi-ronment because we can observe themwhile talking. In general we also feelfairly sure that our environment is rel-atively stable: for instance, that whenwe get up in the morning, our homewill look the way it looked when wewent to bed. We may realise that thebuilding may have burned down overnight, but only as a hypothetical pos-sibility. Philosophers have spent muchtime debating the fact that we thinkwe know, and can talk about, thingswe do not observe while talking, andbasically we cannot be totally surethat we do know. The fact that wenevertheless maintain that we actuallydo know is based on our experienceand the belief that our experience isreliable. Therefore one of our linguis-tic skills – affiliated with our mentalfaculty and our memory – is the abil-ity of displacement, namely that wecan talk about things that are notpresent, or even things we have neverexperienced anything about, likeNapoleon, the far side of the moon orJack the Ripper. This, too, is one of theissues for possible world semantics. Inthe special and technical frameworkof formal logic, possible world seman-tics will also comprise a kind of modallogic, the modality* of which is so-called alethic modality, which dealswith propositions expressing what ispossible versus what is necessary.

the way the world may look depend-ing on the choices made by John; in thehistory of the universe it may seeminsignificant what John chose to do,but in the context of what linguisticexpressions – and especially sen-tences – mean, it may be crucial.Taking (1) as an ordinary life utter-ance, it is expected to have an inherentmeaning more or less evident to theinterlocutors engaged in a conversa-tion, and in modern philosophicalsemantics this is supposed to be estab-lished by its truth value, that is,whether it is true or false according toits truth conditions. This in turn is theway the world looks like in the cir-cumstances surrounding John’s buyinga car or not.

So far one might accept the claimthat the meaning of sentences has todo with the relation between thewords uttered and the world, butthere are two problems here. One isthat it may not be evident how themeaning of sentences is identical withthe question of whether they are trueor false. Intuitively one may think thatin order to decide the truth value (rel-ative to a specific state of affairs) of asentence, one has to understand it,and to understand it is to know itsmeaning. But, it is claimed, to knowits meaning is to know its truth valuebased on its truth conditions, which itfollows is the same thing as its mean-ing. There seems to be a kind of circu-larity predicament as an immanentpart of truth-value semantics, and thisis one of the problems possible worldsemantics tries to solve. A furthercomplication is the following ques-tion: if meaning is truth conditions –that is, there is a kind of correspon-dence between a sentence and the

166

POSSIBLE WORLD SEMANTICS

and, accordingly, contingent proposi-tions are true at some possible worldsand false at some other possibleworlds. Obviously things might havegone otherwise, and (4)

(4) The present queen of Denmark wascrowned queen in 1971.

might have been true if only her fatherthe king had not lived the last fewdays of his life. Therefore (4) is aproposition that is possibly true – thatis, in possible world semantics it istrue in some (at least one) possibleworlds. In contrast, (5) is an impossi-ble proposition:

(5) The present queen of Denmark is aman.

This is true in no possible world and istherefore false by necessity, while (6)

(6) The present queen of Denmark is awoman.

is a proposition that is true by neces-sity – that is, it is true at all possibleworlds. When using the word ‘possi-bly’, one is able to set up an explicithypothetical scenario:

(7) The queen of Denmark is possiblymeeting the American president thismoment.

and the same job can be done with anepistemic modal:

(8) The queen of Denmark might bemeeting the American president thismoment.

Both (7) and (8) may be true in somepossible worlds and false in others,

The modern notion of possibleworld semantics has its roots inCarnap’s Meaning and Necessity(1947) in which, as opposed to his ear-lier work, he engaged in semanticquestions, adjusting Gottlob Frege’sbasic notions of Sinn (‘sense’) andBedeutung (‘meaning’) in order tomake them work in his system ofmodal logic. In the general modernversion of possible world semantics –in which the term ‘possible world’ hadits technical meaning coined byKripke in a paper in 1959 – it worksas a way of clarifying the meaning ofeither normal declarative sentences orsentences containing the words ‘possi-ble’ and ‘necessary’ (or some of themorphological derivations of them) orcontaining epistemic modals (verbs inthe past tense expressing counterfac-tuals or unsubstantiated information).Consider the sentence below:

(2) The present queen of Denmark wascrowned queen in 1972.

This proposition is true in the actualworld because her father the king diedthat year. This is a normal declarativesentence and its truth value is a matterof contingency, that is, it may be eithertrue or false, and one has to check theevidence in order to decide what thefacts are. One also has to do so whenlistening to (3) in order to concludethat it is a false contingent proposition:

(3) The present queen of Denmark hasan older brother.

A possible world semantics will tacklethese details by saying that (2) is trueat the actual world, as mentioned,while (3) is false at the actual world,

167

POSSIBLE WORLD SEMANTICS

(12) ‘if the “Big Bang” initiated theuniverse [antecedent] then it willexpand forever [consequent]’

and adding the alethic expression ‘it ispossible that’ as a prefix. This can beformalised, first, as

(13) p → q

where p is the antecedent and q theconsequent in (12) and, second, as

(14) ◊( p → q) = ‘it is possible that (12)’.

In possible world semantics thiswould be equivalent to ‘ “if the ‘BigBang’ initiated the universe then it willexpand forever” is true at some (atleast one) possible worlds’. On thisbasis one is able to combine morecomplex formulae of deduction.Whether modal alethic logic is theideal (for) modal logic or just somesubtype of it is a debated issue amongphilosophers, but some commonground seems to be that there are certain relations between possibleworlds. Sometimes this is referred toas accessibility or alternativenessbetween worlds (see McCawley 1981:276), figuratively how you can ‘get’from, say, w1 to w2, irrespective ofthe stipulated ontological status of theworlds. That means that the tradi-tional mathematical principles ofreflexivity, symmetry and transivity(McCawley 1982: 276) apply to theformulae produced by systems offormal modal (alethic) logic. A furtherextension of such systems is the pro-duction of predicate logic systems formodal logic (see McCawley 1982:285–96). Lewis (1973) takes up thequestion of relations between possible

but the trouble with such sentences isthe fact that their truth value is notgiven by the meanings (reference ofthe subject and extension of the predicate) of their parts, as is thecase with the previously mentionedsubject-predicate sentences. Hence(7) and (8) cannot be used as truth-functional conjuncts in the same way.This other kind of logic may beexpressed in a formal and symbolisedlanguage (so-called alethic formallogic) using special logical operators.Thus in (9)

(9) �p = ‘it is necessary that p’

the symbol � is an operator signifyingthe necessity of the content of theproposition ‘p’. If p is supposed to be(6) ‘The present queen of Denmark isa woman’ then p is true – that is, (6) istrue at all possible worlds. However, ifp is, for instance, (5) ‘The presentqueen of Denmark is a man’ andsomebody claims that

(10) �p = ‘it is possible that p’

using � as a possibility operator, then(10) is false by necessity because thesentence p is true in no possible world.By combinations of these operatorsand connectors from traditional logic,one is able to construct larger formalsystems (languages) with compoundpropositions. Using the letters w, v, uas variables for possible worlds, onemay state the propositions:

(11) ‘the ‘Big Bang’ initiated the uni-verse’ and ‘the universe willexpand forever’

putting them together in

168

POSSIBLE WORLD SEMANTICS

designators. He developed an ideanow called the causal theory of refer-ence, and he has profoundly changedthe common view on the basic logicalnotions of a priori and a posteriori rel-ative to analytical and synthetictruths. But in the context of modallogic his major contribution is thesolutions he offers to the problemsconcerning modal logic, possibleworld semantics and the metaphysicalconsequences of a number of princi-ples appealed to by logicians. This hasto do with, for instance, the above-mentioned realism as one kind ofassertion about whether possiblethings exist or not; another view isactualism, which says that everythingthat exists is what is actual – that is,what we can say there is a fairly clearconsensus about in the real world.Now, the problem is that if one is anactualist and accepts the formal con-sequences of modal logic in what iscalled a simplest quantified version,one also has to accept the unattractiveoutcome that possible things do exist.Kripke’s solution to the problemrequires a rather technical accountbut it has to do with the way he han-dles the semantics (specifically thedomains) of individual terms in modalpropositions, a theoretical suggestioncalled ‘Kripke Models’. An unpreten-tious interpretation might say that heclaimed that possible things can besaid to exist but only, so to speak, intheir own worlds, from which it fol-lows that propositions about themput forward in the actual world endup being false. In this way the prob-lem with their truth value, that theyseem to be true also in the actualworld (which has some unpleasantmetaphysical consequences), has been

world in the context of counterfactu-als, that is what is not true at theactual world, and he offers a clarify-ing account of these relations in thathe proposes not an abstract distance –illustrated by closeness in numericalmeasure – but a three-place predicatenotation expressing the relative close-ness between three possible worlds(McCawley 1982: 312).

Lewis also defends the idea thatpossible worlds do in fact exist asentities in some way or other. This iscalled absolute realism (Schurz 2006:443) and is a subtype of realism, ametaphysical claim that holds thatuniversals (what can be used as pred-icates in propositions about morethan one entity) are real. An alterna-tive view is reductive realism whichsays that possible worlds do exist butcan be reduced to entities that aremore familiar, and this view is themost common among scholars in thefield. A third position is anti-realism,and from the meaning of this labelone may infer that it denies boththe realistic stance (that universalsexist) and, accordingly, that possi-ble worlds exist and that possible-world-sentences are meaningful. Inthis respect Kripke conceived of apossible world as ‘a counterfactualcourse of history’ (Kripke 1981: 6),and, in a broader perspective, onemay set up a quite impressive numberof combinations of different kinds ofexistence of possible worlds, maybeending with the mundane view thatthey are just useful fictions used inlogical reasoning.

Kripke made a number of contribu-tions to philosophical logic, notablyto the theory of the semantics ofnames* – which he called rigid

169

POSSIBLE WORLD SEMANTICS

Kripke, S. (1981). Naming and Necessity.Oxford: Blackwell.

Lewis, D. K. (1973). Counterfactuals.Oxford: Blackwell.

McCawley, J. (1981). Everything ThatLinguists Have Always Wanted toKnow About Logic, But Were Ashamedto Ask. Oxford: Blackwell. 273–359.

Menzel, C. (1990). ‘Actualism, ontologi-cal commitment, and possible worldsemantics’. Synthese 85: 355–89.

Schurz, G. (2006). ‘Alethic modal logicsand semantics’. In D. Jaquette (ed.), ACompanion to Philosophical Logic.Oxford: Blackwell. 442–77.

Further readingBenthem, J. van et al. (2006). ‘Part IX.

Modal logic and semantics’. In D.Jaquette (ed.), A Companion to Philo-sophical Logic. Oxford: Blackwell.389–509.

Dancy, J. (1985). Introduction to Contem-poraryEpistemology.Oxford:Blackwell.

Groenendijk J. and M. Stokhof (2002).‘Type-shifting rules and the semantics ofinterrogatives’. In P. Portner and B. H.Partee (eds), Formal Semantics. Oxford:Blackwell.

Kratzer, A. (2002). ‘The notational cate-gory of modality’. In P. Portner and B.H. Partee (eds), Formal Semantics.Oxford: Blackwell.

Martin, R. M. (1987). The Meaning ofLanguage. Cambridge, MA: MIT Press.

Robertson, T. (1998). ‘Possibilities and thearguments for origin essentialism’.Mind 107. 428: 729–49.

Saeed, J. I. (2003). Semantics. Second edi-tion. Oxford: Blackwell.

Stalnaker, R. C. (2002). ‘Assertion’. In P.Portner and B. H. Partee (eds), FormalSemantics. Oxford: Blackwell.

Hans Götzsche

dissolved. Kripke’s solution is notundisputed but his intellectualachievements in the field are widelyacknowledged.

Possible world semantics presentsthe linguist and the philosopher witha number of suggestions that maysolve some traditional problems. Forinstance, the apparent problem of cir-cularity in a truth-conditional seman-tics for linguistic expressions may besolved by rephrasing the questionfrom asking about relations betweenmeanings of sentences in this worldto asking about relations betweendifferent worlds, called possibleworlds, as meanings of sentences.This may also solve the questionabout the meaning of a false sentencein that a false sentence may be true insome possible world, which therebypresents an intuitively acceptablepositive correspondence between thesentence and that particular world.However, this leaves open the ques-tion of whether this will make usunderstand what a false sentencemeans in our world. It also does notdissolve the relation between sen-tence, meaning and possible world; itdoes not solve the correspondenceproblem. It would not be unfair tosay that possible world semanticschooses to ignore such problems con-centrating on constructing formalsystems that can handle traditionaltechnical problems in other formalsystems.

Primary sourcesCarnap, R. (1947). Meaning and Neces-

sity. Chicago: Chicago University Press.Kripke, S. (1959). ‘A completeness theo-

rem in modal logic’. Journal of Sym-bolic Logic 24: 1–14.

170

entities themselves are not positivelydefined, but identified at best as mereplace-holders. This has the conse-quence that all structures are by defi-nition hermetically closed untothemselves. It is only on this conditionthat a structure can be regarded ascomposed of pure negativities. To putit differently, the ‘structurality’ of agiven structure is conditional upon theentire space having been taken up bybinary relations: the forceful removalof any entity will automatically triggera rearrangement of the remainingentities. This in turn means that allstructures are fully integrated, eachwith respect to itself, and autonomouswith respect to other structures.

Structuralism was a huge success asan idea in the early decades of thetwentieth century and it held swayover several academic disciplines wellinto the second half of that century. Itrevolutionised linguistics and set itapart from philology and historicallinguistics, which dominated the disci-pline in the nineteenth century. In factthe neat separation between syn-chrony and diachrony that Ferdinandde Saussure advocated was but astraightforward consequence of therequirement that a structure, in orderto function the way it was required to,had to be closed unto itself. But it soonbecame clear that this was at oddswith the further requirement that astructure also had to be resilient, so asto permit ‘the rearrangement of itsinternal units’ in order to accommo-date eventual structural changes.From a structuralist point of view, allchanges were, so to speak, sudden andcataclysmic and, while keeping theoverall structure intact, would resultin an internal rearrangement of the

POSTSTRUCTURALISM

Poststructuralism is an outgrowth ofstructuralism* It is a method of analy-sis which had a tremendous impact inthe last decades of the twentieth cen-tury on work done in the fields of lin-guistics, anthropology, psychology,literary criticism and elsewhere.Arguably, it remains a potent force toreckon with in the early twenty-firstcentury, though it has been at thecentre of intense and often acrimo-nious debates.

See also: Signs and Semiotics;Structuralism; DeconstructionKey Thinkers: Bourdieu, Pierre;Derrida, Jacques; Lacan, Jacques;Saussure, Ferdinand de

Like so many other terms that take theprefix ‘post-’, poststructuralism is acomplex notion. Part of the difficultyin getting to grips with the term has todo with just how one is supposed toview its relation to structuralism. Inone sense, the term does imply that itregards structuralism as a thing of thepast (as in ‘post-World War Twoperiod’). But equally it also capturesthe idea that it is an offshoot of struc-turalism or, alternatively, it is a move-ment that draws its strength from thelegacy of structuralism (as in ‘post-graduate student’).

In other words, to understand whatpoststructuralism is about, one needsto have some idea of what constitutesstructuralism. Structuralism is a modeof inquiry which looks at the phe-nomena under its scrutiny as made upsolely of the relations among the enti-ties in question, rather than those enti-ties themselves. In other words, the

171

POSTSTRUCTURALISM

to incorporate into structuralisminsights from Marxism and Lacanianpsychoanalysis. The speaking subjectwas unceremoniously dislodged fromthe high pedestal where he/she hadremained safely and majesticallyensconced. The subject was decentredand demoted to the status of a mari-onette. Marxist structuralists likeLouis Althusser advocated this secondalternative, on the strength of the con-viction that the ‘mature Marx’ hadmade a crucial break with his own ear-lier Romantic humanism, wherebywhat would be left at the centre is anillusory subject, seeming to be in con-trol of him/herself and his/her mean-ings, while actually being subjected orsubjugated by the structure.

Derrida contended that both of theoptions signalled by structuralismresult from a misunderstanding of thefunction of the absent centre. Thestructure’s absent centre functionsrather like the black hole at the centreof a galaxy (a structure like anyother). It is here that Derrida recog-nises the effects of history within thestructure, traditionally thought tooffer no room for origin or history(whereof the separation of synchronyfrom diachrony). What this means isthat the subject can be neither Carte-sian (all-knowing, intending) nor abeing at the mercy of forces beyondhis/her control. The structuralist sub-ject, in other words, must give way toa poststructuralist agent who must actin history (not independently of it)while, no doubt, constrained by thestructure around him/her.

This then can be pointed out as thehallmark of poststructuralism: theemergence of the historical subjectwithin the very entrails of a structure.

network of relations within the struc-ture. History, in this world-view, pro-gressed by fits and starts, rather thanin a smooth and gradual continuum.In the philosophy of science, ThomasKuhn famously advocated a view ofthe progress of science which had allthe trappings of structuralism in thissense.

As already noted, the twin require-ments of the closure and the resilienceof a structure are mutually incompat-ible. In a ground-breaking essay called‘Structure, Sign, and Play in the Dis-course of the Human Sciences’,Jacques Derrida (1966) drew atten-tion to this irreconcilable incompati-bility and argued that the only waystructuralism could be salvaged is byadmitting that a structure can haveonly an imaginary centre. ‘The centrecannot be the centre’ concluded theFrench philosopher in what is oftencited as one of his most enigmaticstatements. Now, this imaginarycentre does indeed appear to be eithercontrolling or, contrariwise, beingcontrolled by the structure.

The first alternative results in thepositing of an all-powerful subject,Cartesian in its lineage, whose inten-tions underpin meanings and preventthem from going astray or floatingaround freely. Humanism, alongsideits offshoots existentialism and phe-nomenology, provides the necessaryphilosophical backdrop against whichsuch a subject could be envisaged.Intentionality* (‘aboutness’) would bethe hallmark of the subject of lan-guage. He/she would constitute thefulcrum around which language turns.

In the late 1960s and early 1970s,however, there was a concertedmove among French intellectuals

172

POSTSTRUCTURALISM

notion of ‘habitus’. In Bourdieu’sview, social agents develop, over aperiod of time, a ‘feel for the game’which is a kind of eminently practicaland bodily knowledge.

Poststructuralism is closely tied toand often confused with postmod-ernism. While the two indeed sharemany common concerns, it is impor-tant to bear in mind that the latter isconcerned to critique, in a way theformer is not, the Enlightenment proj-ect and the sheer arrogance of thosewho claimed to be fighting the battleof Reason against Irrationality.

Poststructuralism, alongside post-modernism, opens up an entirelynew set of possibilities for thinkingabout ethical issues. With the meta- narratives of foundationalist, essen-tialist and totalising discoursecom pletely discredited, poststruc-turalism leaves us no option but thatof regarding the path of ethics as anexercise in tight-rope walking, withno safety net in case a false step istaken. In other words, it foregroundsthe question of personal responsibil-ity. In the United Kingdom, a special-ist group called ‘Post-Structuralismand Radical Politics’ has been activelyengaged in promoting discussion onthe political implications of poststruc-turalism in all walks of life.

Primary sourcesBelsey, Catherine (2002). Poststructural-

ism: A Very Short Introduction.Oxford: Oxford University Press.

Bourdieu, Pierre (1977). Outline of aTheory of Practice. London: CambridgeUniversity Press.

Derrida, Jacques (1966). ‘Structure, sign,and play in the discourse of the humansciences’. In Jacques Derrida (2001),

This new subject is one who has beenempowered to act on his/her own andis endowed with agency. Once thepresence of an agent who is in a posi-tion to subvert the order of things,thwarting it from within, is recog-nised, it is but a short step to reject theexistence of all pre-ordained, founda-tionalist, essentialist and totalisingconceptual schemes.

Michel Foucault played an impor-tant role in the development of post-structuralism. Having been a die-hardstructuralist himself, Foucault grewincreasingly discontented with two ofstructuralism’s key assumptions. Onthe one hand, while recognising thepervasiveness of structures in many ofhuman activities, Foucault becamesceptical of the idea that structuresinvariably control and regulate thehuman condition. On the otherhand, he also came to recognise theinevitable ‘situatedness’ of our gazeand the impossibility of ever attaininga transcendental standpoint fromwhich to contemplate things.

Freeing the subject of languagefrom the shackles of structurallyimposed subjugation is key to a poli-tics of identity. But then the escapefrom the prison-house of structureshould not herald a return to liberalindividualism whereby the individ-ual determines his/her own destinythrough the exercise of a series ofrational choices. This is the centralthrust of and moving impulse behindthe work of Pierre Bourdieu. In hisbook Outline of a Theory of Practice(1977) Bourdieu made a proposal foravoiding both the Scylla of the totalsubjugation of the subject of languageand the Charybdis of navel-gazingindividualism, by putting forward the

173

PRESUPPOSITION

See also: Connotation/Denotation;Definite Descriptions; Implicature;Names; Possible World Semantics;Sense/Reference; Speech ActTheory; Truth ValueKey Thinkers: Frege, Gottlob;Grice, H. P.; Russell, Bertrand;Strawson, P. F.; Montague,Richard

The notion of presupposition as a the-oretical concept in linguistics and thephilosophy of language tries to solvethe problem that in discourse utter-ances may be articulated and fullyunderstood by the interlocutors, whilea detached and uninitiated observermay find it hard to understand themeaning of what is being said. Imag-ine I go to the supermarket togetherwith my wife and, seeing among theamount of goods a certain kind ofsoup in small plastic bags, I say ‘Look,there is this special Italian soup insmall plastic bags’, and my wiferesponds ‘My Italian language coursehas no more classes’. The backgroundknowledge assumed by both of us isthat I do not cook and therefore I justwarm up some prefabricated meal forus when she comes home from herItalian classes. The line of reasoning isthat when she has no more Italianclasses, I will not have to do this anymore and therefore we will not haveto buy more plastic bag soup. Accord-ing to H. P. Grice, this instance ofbackground knowledge and tacit lineof reasoning may be called a kindof (conversational) implicature*, or,in later developments, presumptivemeaning or preferred interpretation(Levinson 2000), but it might also beconsidered a kind of presupposition.Which option is chosen depends on

Writing and Difference. London: Rout-ledge.

Sarup, Madan (1993). An IntroductoryGuide to Poststructuralism and Post-modernism. Second edition. Athens,GA: University of Georgia Press.

Sturrock, John (1979). Structuralism andSince. London: Oxford University Press.

West, David. (1996). An Introduction toContinental Philosophy. Cambridge:Polity Press.

Williams, James (2006). UnderstandingPoststructuralism. Canada: McGill-Queen’s University Press.

Further readingValentine, Jeremy and Alan Finlayson

(eds) (2002). Poststructuralism and Pol-itics: An Introduction. Edinburgh: Edin-burgh University Press.

Kanavillil Rajagopalan

PRESUPPOSITION

A type of extra and assumed meaningattached to the basic meaning of utter-ances or sentences. In a broad sensepresupposition can be defined as whatmust by necessity be assumed to be thecase in order to interpret a shortordered sequence of linguistic expres-sions in a meaningful way. The ideawas introduced by Gotlob Frege whosaid that presuppositions (in German:Voraussetzung ‘(pre)condition’) areparticular conditions that have to besatisfied for single linguistic expres-sions to have a denotation (Beaver1996), but twentieth-century linguis-tics and philosophy has focused moreon the presuppositional meaning asso-ciated with sentences and utterances.

174

PRESUPPOSITION

heated discussion and controversyamong philosophers and linguists,and while later on Strawson modifiedhis views, many twentieth-and twenty-first-century scholars prefer not tocommit themselves to any of the standpoints.

However, the philosophical impli-cations of this issue are by no meanstrivial since, on the one hand, theydeal with some aspects of the basicproblem of how linguistic expressionscan be used to refer to things in theworld and, on the other hand, how wecan express knowledge of the world inso-called existential sentences like‘there is a [one and only one] queen ofDenmark’, as opposed to so-calledidentity sentences like ‘a queen is afemale monarch’ and predications like‘the queen of Denmark is intelligent’.One further concern is how we referto fictional and abstract entities and,in the present context, how we usesuch kinds of reference in presupposi-tions. These are, in the end, epistemo-logical questions, and some of theformal properties of presuppositionshave some bearing on deeper philo-sophical contemplations. One suchcommon property shared by presum-ably all theories of presuppositions isthe way presuppositions work undernegation. Consider the following sentence:

(1) The queen of Denmark is intelligent.

Accepting the convention that (1) canbe represented by the symbol q, thenit follows that the truth value of q aswell as its negation ~q depends onthe truth value of the presupposition‘there is a queen of Denmark’, whichcan be represented by the symbol p.

whether one sees presupposition as asemantic or as a pragmatic functionand furthermore on whether one’sconcern is natural languages or formallanguages. Frege’s main objective wasto nurture the technical languages ofscience in the form of formal logic*and mathematics, and the specific lin-guistic functions of these languagesthat now are regarded as presupposi-tional functions were the main interestof the central figures Bertrand Russelland P. F. Strawson.

Working with his famous, and tosome scholars notorious, example ‘Thepresent king of France is bald’ (Russell1905: 483), Russell claims that theproposition does not fail to have atruth value* (a point of view held byFrege) but has to be broken down intoa number of component propositionseach of which has to be true for thecompound proposition to be true.Since one of the propositions (‘thereexists a king of France’) is false, thenthe compound proposition is false, andtherefore meaningful according to theview that truth conditions are themeaning of sentences. This view waschallenged by Strawson (1950) whosaid that the proposition singled out byRussell – together with the secondcomponent proposition, ‘there is nomore than one king of France’ – is notasserted but presupposed since the def-inite description* ‘the [present] king ofFrance’ fails to refer (it did in 1950 andhas done since then). In this respectStrawson agrees with Frege. Thereforethe proposition containing the definitedescription cannot be used to make anassertion and accordingly it is mean-ingful but has no truth value. The posi-tions held by Russell and Strawsonrespectively have given rise to much

175

PRESUPPOSITION

The denial of (4b) automaticallymakes (4a) false while the denial of(4a), as in (4c), has no influence on thetruth value of (4b) since the womanmay have died from other causes. Thisis opposed to (3a) and (3c) in that if(3b) is denied: ‘John has not beensmoking’, then both (3a) and (3c)become false by necessity, because onecannot stop doing something one hasnot been doing. This is the basis ofwhat is sometimes called presupposi-tion failure. Imagine somebody isbeing asked the question:

(5) Have you stopped beating yourdog?

Questions like these are much debatedin informal logic because the respon-dent is being trapped: he can neithersay ‘yes’ or ‘no’, because in both caseshe will confess to the fact that he hasactually beaten his dog in the past,even though he may not have done so,and then the only relevant way torespond is to question the questionitself, viz. its presupposition(s). In thiscase many would feel that the presup-position is disguised by the implicitmeaning of the words. It may seemmore evident that somebody is dead ifone has been killed (4a and 4b), and itmay seem even trivial to say of thesentence ‘the king of France is bald’that the existential sentence ‘there is a(one and only one) king of France’ is apresupposition of that sentence,because if somebody mentions a ‘kingof France’ then there must be a king ofFrance.

Nevertheless this was the subjectmatter of the controversy betweenRussell and Strawson and of the fol-lowing debates. Russell’s interest had

Thus, it seems to be a matter of factthat p is the necessary condition ofboth q and ~q. In this respect a pre-supposition differs from an entailmentin that the same does not seem tohold for entailments. Consider thesentence:

(2) Jack the Ripper killed the woman.

It follows from the meaning of thesentence that the woman must bedead, and this has to do with themeaning of the word ‘kill’. The basicmeaning of ‘kill’ is that somebodybeing ‘killed’ must be dead. This is notthe case with, for instance, the word‘shoot’. The sentence ‘the policemanshot the murderer’ may refer to anincident where the murderer survived.Thus entailments are closely con-nected with the lexical meanings ofthe words used (verbs like ‘kill’ and‘shoot’) while presupposition is a rela-tion between sentences and betweensentences and the world. This can beillustrated by sets of sentences such asthe following:

(3a) John has stopped smoking(3b) John has been smoking(3c) John has not stopped smoking

If John has stopped smoking (3a),then he must have been smoking in thepast, but the same goes for the case inwhich he has not stopped smoking(3c), and therefore (3b) is a presuppo-sition of both (3a) and (3c). Not sowith entailments:

(4a) Jack the Ripper killed the woman.(4b) The woman is dead.(4c) Jack the Ripper did not kill the

woman.

176

PRESUPPOSITION

above, held that propositions pickingthem up as subjects were false. Straw-son, then, said that Russell had got itall wrong in that he ignored the dis-tinction between ‘a sentence’, ‘the useof a sentence’ and ‘an utterance of asentence’. In Strawson’s view, sen-tences have no truth value, only theuse of a sentence can produce aproposition which is either true orfalse. As for the definite description‘the king of France’ Strawson arguedthat ‘To say “The king of France iswise” is, in some sense of “imply”, toimply that there is a king of France’(1950 (1971: 12)), and this is the kindof implication that gave rise to thenotion of presupposition; the termwas coined by Strawson in his 1952publication. The outcome was theposition that certain sentences, like‘the king of France is bald’, are seen ashaving unfulfilled presuppositionsand therefore they have no truthvalue, or they have the truth valuezero. One way to go around the prob-lem, if one does not feel at ease withsentences that have the form of pred-icate propositions but neverthelessseem to lack truth value and seman-tics, is to adopt a ‘possible world’semantics*. In this kind of semantics,propositions can be seen as functionsfrom the set of all possible worlds tothe truth values ‘true’ or ‘false’ and inthis kind of logic, presuppositionscan be conceived of as fulfilled inonly a restricted set of worlds. Hencethe propositional functions can bedefined as having value only in the setof these worlds, that is where the pre-suppositions of a proposition are ful-filled, a set which is sometimes calledthe domain of that proposition (All-wood et al. 1977: 150).

to do with the semantics of what hecalled ‘definite descriptions’ like ‘the(red) car’ which he conceived of aswhat he called ‘names*’ (that is, notonly proper names in our understand-ing), a stance based on the argumentmentioned above, viz. that a phraselike ‘the (red) car’ formally should bedissolved into the proposition ‘thereexists a red car’. Accordingly Russellclaimed that ‘names’, which hedefined as linguistic expressions beingable to be used as logical subjectsin propositions, included definitedescriptions, and therefore thesecould be used – as referring expres-sions in propositions – to refer tothings in the world, for instance a redcar. This solved the problem that inordinary predicate logic there are nodefinite descriptions, only individualor quantitative terms like ‘Aristotle’,‘he’ or ‘all dogs’, and accordingly Rus-sell was able to justify predicate sen-tences like ‘the (red) car is rusty’.

This has further implications inmodern logic but it also presentedRussell with the problem that definitedescriptions may be empty, at vari-ance with our common-sense viewthat the use of the definite articleimplies that there is some well-knownentity out there. But as is also wellknown, definite descriptions maysometimes have no referents – forinstance in fictional and abstract con-texts – that is, no entities that satisfytheir description, such as is the casewith the definite noun phrase ‘theking of France’. From this it followsthat propositions containing themcannot specify truth conditions andaccordingly they have no semantics.According to Frege, they were mean-ingless, while Russell, as mentioned

177

PRESUPPOSITION

‘assume’, and it may seem a little oddto say that sentences are able to assumesomething, because a common-senseunderstanding might claim that onlypeople can assume. The same questionmay arise when considering the theo-retical idea that a sentence like ‘theking of France is bald’ presupposes thesentence ‘there exists a king of France’,because it would be counter-intuitiveto think otherwise. This has led to thesuggestion that there are, at least, twotypes of presupposition: one is asemantic notion dealt with in modernlogic where scholars try to solve thetechnical problems of formal lan-guages, and another is a pragmaticnotion dealing with what people pre-suppose in their discourse interactions.To contemplate this question is likeopening a Pandora’s box in linguisticsand philosophy. As Levinson puts it,‘there is more literature on presupposi-tion than on almost any other topic inpragmatics (excepting perhaps speechacts)’ (1983: 167).

In order to approach the subject wemay take the Italian-soup-in-small-plastic-bags scenario described aboveas an example. A main topic here willbe what are in general called presup-position triggers. A presuppositiontrigger is one or more words, or‘aspects of surface structure in gen-eral’ (Levinson 1983: 179) that gen -erate a relation of presuppositionbetween what is actually expressedand what must be assumed. At facevalue only the word ‘Italian’ seems tobe common to the two utterancesthereby suggesting there is a connec-tion between them. But this similaritymay elicit a rather large number ofpossible scenarios and it is in factdeceptive. My wife and I may have

Apart from these technical and epis-temological questions, some issuesconcerning presuppositions are on theborderline between such basicallyphilosophical problems and the dailyuse of language. Some of them dealwith the formulaic or grammati-calised expressions associated withspecific speech acts:

(6) What did Jack the Ripper do?

In (6) the wh-expression connectedwith the fixed word order (and a cer-tain prosodic contour) signifies thespeech act of a question, and it pre-supposes that Jack the Ripper didsomething (cf. Lyons 1977: 597). Thiskind of presupposition is not indiffer-ent to negation because the sentence:

(7) What did Jack the Ripper not do?

presupposes that there was somethingthat Jack the Ripper did not do.Another kind of presupposition (cf.Lyons 1977: 599) is found in sen-tences expressing so-called proposi-tional attitudes*, for example,sentences containing expressions like‘think’, ‘believe’, ‘realise’:

(8) The police realised that Jack theRipper had killed the woman

In (8) it is a presupposition of the com-pound sentence that Jack the Ripperactually killed the woman and as forthe ‘historical’ Jack the Ripper, ofwhom little is known, a sentence like(8) does not seem justified.

As pointed out by Lyons (1977:600), a pre-theoretical notion of pre-supposition as a verb ‘presuppose’means almost the same as the word

178

PRESUPPOSITION

be interpreted as a ‘No, we should buyno more soup’. But this is, in a narrowsense, not something presupposed. Itmay belong to the ‘normal’ state ofaffairs in certain cultures or subcul-tures, but none of the words or othersurface aspects can, directly, triggersuch presuppositions. Only knowledgeabout the family and a number of cir-cumstantial facts may yield the infor-mation that this has nothing to do witha scenario where, for instance, the stu-dents of my wife’s language class eatItalian soup every time they meet.However, in a broad sense theseinferred assumptions may be calledpresuppositions. Some scholars havedefended radical theories of pragmaticpresupposition in which the key con-cepts are ‘appropriateness’ or ‘felicity’(Levinson 1983: 204) and ‘mutualknowledge’ or ‘common ground’,mean ing that a presupposition can be‘appropriately used if it is assumed inthe context that the propositions indi-cated by the presupposition-triggersare true’ (Levinson 193: 205). The con-textual prerequisite is furthermore thatthis is known by the participants, andconsequently what is not mutuallyknown by them will not come up aspresuppositions because they are inap-propriate. In this framework the chainof inferences assumed in my conversa-tional transactions with my wife on‘Italian soup’ may well be called pre-suppositions, but one problem is justthe criterion about mutual knowledge.When examining actual discourse it isfairly hard to make justified claimsabout what people know – as opposedto claims about what they say.

At the core of theories on presup-positions there are other difficulties,among them the problem of whether

any kind of fast food after her Italianclasses. If, instead of fantasising, welook at the details of what is beingsaid, my utterance was ‘Look, there isthis special Italian soup in small plas-tic bags’, and this may be interpretedas a man-and-wife convention for thespeech act of a question: ‘Should webuy some?’ Taken as a question, thepresupposition is ‘either we buy someor we don’t buy some’. In this case itis the unspoken but intended questionthat triggers the presuppositional rela-tion in that the specific clause surfacestructure of questions – the syntacticinversion and the prosody of ques-tions produced by native speakers ofEnglish – will come to my wife’s mindwhen I paraphrase my intended ques-tion as the declarative sentence I actu-ally utter. If this reading seems fairenough, the logical response from mywife should be an answer implyingeither that we should or we should notbuy the soup-in-small-plastic-bags,and it actually may be so. The sen-tence ‘My Italian language course hasno more classes’ may be paraphrasedas ‘My Italian course has stopped’,and the verb ‘stop’ is a so-calledchange-of-state verb triggering thenecessary presupposition(s) ‘I haveattended an Italian language course inthe past and now I don’t attend it anymore because there are no moreclasses’.

In order to establish a connectionbetween my question ‘Do we buy ornot buy?’ and the ‘Italian classes’, onemay speculate what my wife’s absenceor late home-coming has to do withfast food and one only has to infer thatin this family fast food is on the menuwhen the housewife is away or late,and that my wife’s response may boldly

179

PRIVATE LANGUAGE

Strawson, P. F. (1950). ‘On referring’.Mind 59: 320–44.

Strawson, P. F. (1952). Introduction toLogical Theory. London: Methuen.

Further readingAllwood, J, L.-G. Andersson and Ö. Dahl

(1977). Logic in Linguistics. Cam-bridge: Cambridge University Press.

Levinson, S. C. (1983). Pragmatics. Cam-bridge: Cambridge University Press.

Levinson, S. C. (2000). PresumptiveMeanings. The Theory of GeneralizedConversational Implicature. Cam-bridge, MA: MIT Press.

Lyons, J. (1977). Semantics. Cambridge:Cambridge University Press.

Saeed, J. I. (2003) Semantics. Second edi-tion. Oxford: Blackwell Publishing.

Hans Götzsche

PRIVATELANGUAGE

A private language is a putative lan-guage the meaning of whose expres-sions are essentially grounded in thesubjective or private states of the usersof the language. The notion was firstemployed by Ludwig Wittgenstein(Philosophical Investigations) in orderto refute both classical and contempo-rary conceptions of linguistic mean-ing. The notion remains at the centreof Wittgenstein scholarship and has ahighly controversial use in philosophyof language and mind more generally.

See also: Deconstruction;Empiricism/Rationalism; LanguageGames; Logical Positivism;Presupposition

presuppositions are compositionalrelative to the complex clauses theyare presuppositions of (Levinson1983: 191). There is not a consensusamong scholars in the field aboutwhat constitutes a standard notion ofpresupposition in linguistics and thephilosophy of language.

Primary sourcesAsher, N. and A. Lascarides (1998). ‘The

semantics and pragmatics of presuppo-sitions’. Journal of Semantics 15, 3:239–300.

Beaver, D. I. (1996). ‘Presupposition’. In J.van Benthem and A. ter Meulen (eds),Handbook of Logic and Language.Amsterdam: Elsevier Science B.V.

Frege, G. (1892). ‘Sinn und Bedeutung’. InZeitschrift für Philosophie undphilosophische Kritik, NF 100:25–50.English translation, ‘On sense andmeaning’. In Peter Geach and MaxBlach (eds) (1980), Translations fromthe Philosophical Writings of GottlobFrege. Oxford: Blackwell. 56–78. Firstedition 1952.

Journal of Semantics Vol. 9 no. 3 1992.Special issue on presupposition.

Journal of Semantics Vol. 9 no. 4 1992.Special issue on presupposition.

McCawley, J. (1981). Everything ThatLinguists Have Always Wanted toKnow About Logic, But Were Ashamedto Ask. Oxford: Basil Blackwell.

Portner, P. and B. H. Partee (eds) (2002).Formal Semantics. Oxford: BlackwellPublishing.

Reyle, U. (1993). ‘Existence presupposi-tions and background knowledge’.Journal of Semantics 10, 2: 113–22.

Russell, B. (1905). ‘On denoting’. Mind14: 479–99.

Russell, B. (1957). ‘Mr Strawson on refer-ring’. Mind 66: 385–89.

180

PRIVATE LANGUAGE

in approach and associated with JohnMcDowell, takes the argument to be aprincipled declining of a justificationof our normative activity, for thereason that any non-normative basis,such as consensus, would never sufficeto establish the kind of correctnessinvolved in meaning.

Wittgenstein’s argument has alsosurfaced in cognitive science. JerryFodor and Noam Chomsky haveargued that meaning, linguistic andnon-linguistic, has an empirical basisin an unconscious system of rules thatdetermine interpretation as opposedto normatively governing it. Theydefend this approach against the pri-vate language argument by pointingout that the rules at issue are notessentially private, but simply notavailable to first person consciousaccess; ultimately, whether there aresuch rules or not is an empirical issuethat cannot be a priori determined.

Primary sourcesWittgenstein, Ludwig (1953). Philosophi-

cal Investigations. Oxford: Blackwell.Kripke, Saul (1984). Wittgenstein on

Rules and Private Language. Oxford:Blackwell.

McDowell, John (1984). ‘Wittgensteinon following a rule’. Synthese 58 (3):325–64.

Fodor, Jerry (1975). The Language ofThought. Cambridge, MA: HarvardUniversity Press.

Chomsky, Noam (1986). Knowledge ofLanguage. Westport: Praeger.

Further readingMiller, Alexander, and Crispin Wright

(eds) (2002). Rule-Following andMeaning. London: Acumen.

John Collins

Key Thinkers: Ayer, A. J.; Carnap,Rudolf; Chomsky, Noam; Fodor,Jerry; Frege, Gottlob; Kant,Immanuel; Kripke, Saul; Russell,Bertrand; Wittgenstein, Ludwig

In the first half of the twentieth cen-tury, the received understanding of aclassical empiricist theory of languagewas that words are outward signs ofprivate states (ideas or images). Such aposition found its contemporaryadvocates in Bertrand Russell, RudolfCarnap and A. J. Ayer. Arguably fol-lowing Immanuel Kant and GottlobFrege, Ludwig Wittgenstein arguedthat such an essentially private lan-guage was impossible and, therefore,the empiricist conception of meaningis confused. Although the argumenthas been interpreted in many differentways, its basic structure is as follows.(1) Rule following is constitutive ofcompetent language use. (2) Rule fol-lowing is a normative activity – that is,there are essential conditions onwhether one is following the rules cor-rectly or not. (3) If linguistic meaningis private, then the agent must be fol-lowing private rules. (4) If the rulesare private, then there is no differencebetween following the rules correctlyor not; whatever seems right will beright. (5) Therefore, there can be nosuch private language.

In general terms there are twoschools of thought on the argument.One school, most prominently repre-sented by Saul Kripke, views the argument as an attempt to showthat linguistic meaning is essentiallypublic, in that only a public checkby other speakers could establishwhether one is following the rules ornot. Another school, more diagnostic

181

(2)a. Jane believes that Fido is barking.b. Fido exists.c. There is something that Jane

believes is barking.

Like other ordinary relational claims,(1a) has existential import that (2a)and other attitude reports seem tolack. In particular, (1a) entails both(1b) and (1c) while (2a) apparentlyentails neither (2b) nor (2c). After all,Jane could have the belief reported in(2a) even if Fido is a figment of herimagination, contrary to (2b); and inthat case, contrary to (2c), nothingwould have the property of beingbelieved by Jane to be barking, for ifthere were, that thing would be Fido,and there is no Fido.

Another difference relates to thelaw of identity. This law, compellingin the context of ordinary relations, isless plausible in the context of attitudereports. Thus, suppose Fido is Rex.This identity and (1a) could not bothbe true without (1d) also being true. Incontrast, the identity and (2a) do notentail (2d). After all, Jane may notrealise that Fido and Rex are one andthe same dog.

(1)a. Jane strokes Fido.d. Jane strokes Rex.(2)a. Jane believes that Fido is barking.d. Jane believes that Rex is barking.

These two distinctive features ofpropositional attitude reports –apparent lack of existential importand apparent substitution failure –give rise to a range of semantic puz-zles. Any solution to these depends on

PROPOSITIONALATTITUDES

A propositional attitude is what some-one is described as having by a propo-sitional attitude report: for example,by an utterance of ‘Jane believes thatFido is barking’ or ‘Henry wishes hiscat would have kittens’. The subtlesemantics and pragmatics of proposi-tional attitude reports, as well as therole of propositional attitudes them-selves in language use, have madethem one of the more controversialtopics in linguistics.

See also: Compositionality;Definite Descriptions;Descriptivism; Intentionality;Propositions; Sense/ReferenceKey Thinkers: Frege, Gottlob;Kripke, Saul; Quine, W. V. O.;Russell, Bertrand

A standard route into the topic ofpropositional attitudes begins withthe observation that propositionalattitude reports differ from ordinaryrelational claims. On the face of it,for Jane to believe that Fido is barking is for her to bear a two-placerelation to the proposition thatFido is barking – just as for her tostroke Fido is for her to bear a two-place stroking relation to the dogFido. This surface similarity is deceptive.

One difference can be illustrated bycomparing (1) with (2):

(1)a. Jane strokes Fido.b. Fido exists.c. There is something that Jane

stroked.

182

PROPOSITIONAL ATTITUDES

conditional properties of their simplercomponents. For example, the truthcondition stated in (4a) is a derivableconsequence of (4b), (4c), and somebackground semantic principles.

(4)a. ‘Fido is a dog’ is true iff Fido is a

dog.b. ‘Fido’ refers to x iff x = Fido.c. ‘is a dog’ is true of x iff x is a dog.

But what about (1a)? Its truth condi-tion, stated in (5a), ought to be deriv-able from (5b) plus backgroundsemantic axioms for the other parts ofthe sentence.

(5)a. ‘Jane believes that Fido is barking’

is true iff Jane believes that Fido isbarking.

b. ‘Fido’ refers to x iff x is Fido.

The problem is that any compositionalderivation of (5a) using (5b) seems tolicense a parallel derivation of the ille-gitimate (6a) from (6b). Given thatFido is Rex, (6b) follows from (5b), andthe derivation of (6a) from (6b) couldthen mimic that of (5a) from (5b). Yet(6a) seems wrong: Jane’s belief thatFido is barking is not enough to makethe sentence ‘Jane believes that Rex isbarking’ true, even if Rex is Fido.

(6)a. ‘Jane believes that Rex is barking’ is

true iff Jane believes that Fido isbarking.

b. ‘Rex’ refers to x iff x is Fido.

Notice that this problem does notarise for ‘Rex is a dog’, which is trueif Fido is a dog and Rex is Fido.

a proper appreciation of the nature ofa fundamental notion in the philoso-phy of language and mind: proposi-tions*, that is the things that attitudereports report us as believing, desir-ing, intending, and so on. After pre-senting two such puzzles, one relatingto each distinctive feature we willthink about how different theories ofpropositions require differing treat-ments of the second of these puzzles.

The lack of existential import threat-ens to sabotage solutions to the alreadynotorious problem of negative existen-tials. This is the problem of under-standing how we can ever correctlydeny something’s existence when, inorder to do so, we would need to referto it, and hence need it to exist. Variouspurported solutions are available, but aplausible condition on their acceptabil-ity is that they make sense of the legiti-macy of inferences with the form in (3):

(3) a does not exist, therefore not . . . a. . .

For example, if Saladin never existed,it follows that Saladin could not havefought Richard the Lionheart. But theinference appears to fail when (3) is‘Fido does not exist, therefore Janedoes not believe that Fido is barking’.Any theory of negative existentialstherefore has an additional burden: toexplain exceptions to (3) thrown upby propositional attitude reports.

The second feature, apparent sub-stitution failure, threatens the compo-sitionality thesis, a cornerstone ofmost approaches to formal semantics.In one form, this thesis holds that thereferential or truth conditional prop-erties of complex expressions are afunction of the referential or truth

183

PROPOSITIONAL ATTITUDES

Since the elements of propositions areactual entities, the referents of thewords used, substituting co-referringexpressions will alter neither the iden-tity of the proposition expressed nor,therefore, the truth value of the sen-tence. Russellians must explain theappearance of substitution failure inpropositional attitude reports with-out granting that this appearance isgenuine.

Before 1972 the standard way wasRussell’s own. Ordinary names* like‘Fido’ and ‘Rex’ are not in fact refer-ring expressions at all, he said. Theyare really disguised definite descrip-tions*, which have an internal struc-ture of their own but no referents. Thecontrast between (2a) and (2d) doesnot, therefore, constitute a case ofchanging truth value by substitutingco-referring expressions.

In more detail, suppose ‘Fido’ inJane’s idiolect is a disguised version of‘The dog I bought last year’ and ‘Rex’is a disguised version of ‘The dog I ranover two years ago’. In that case, (2a)and (2d) would express the distinctpropositions (2g) and (2h) respec-tively (where � can be read as roughlyequivalent to ‘for every’, � as ‘exists’and → as ‘entails’; see the entry on‘Logic’ for a fuller account of theselogical constants).

(2)g. <believes, <Jane, �x(x is a dog I

bought & �y(y is a dog I bought →x=y) & x is barking) >>

h. <believes, <Jane, �x(x is a dog I ranover & �y(y is a dog I ran over →x=y) & x is barking) >>

No actual dog is a component ofeither proposition, so ‘Fido’ or ‘Rex’

These puzzles may seem like meretechnical challenges, but theirendurance suggests otherwise. Whatfollows is a summary of two promi-nent theories of propositions, stressinghow each confronts the puzzle ofapparent substitution failure. Both the-ories treat propositions as structuredcomplexes of elements, but they dis-agree on the nature of these elements.

Followers of Bertrand Russellassume that propositions are struc-tured complexes of real-world entities. Sentences map onto thepropositions they express via a map-ping that takes individual words ontothe real-world entities they refer to.The proposition expressed by a typi-cal utterance of (1a), for example,would be (1e), the first element ofwhich is a two-place relation, and thesecond element of which is an orderedpair consisting of Jane then Fido.

(1)e. <strokes, <Jane, Fido>>

Thus (1e), and hence (1a) itself, will betrue just so long as the ordered pairsatisfies the two-place relation. Moregenerally, the truth value* of proposi-tions is determined compositionallyfrom its components.

So (2a) expresses the proposition(2e), which breaks down further into(2f). The simple proposition Jane pur-portedly believes is a component ofthe more complex propositionexpressed by (2a) itself and is repre-sented in (2f) as ‘<barks, <Fido> ‘.

(2)e. < believes, <Jane, that Fido is bark-

ing>>f. < believes, <Jane, <barks, <Fido>>>

184

PROPOSITIONAL ATTITUDES

But Fregeans have a ready explana-tion for exceptions such as (2a) and(2d): when a word occurs within thescope of a propositional attitude verb,Fregeans say it refers to its customarysense, not its customary referent.Since ‘Fido’ and ‘Rex’ have distinctcustomary senses, the propositionsexpressed by (2a) and (2d) have dif-ferent components even after con -version into Russellian propositions,and hence potentially different truthvalues.

The most pressing questionFregeans must answer is: what is asense? If a sense is just a disguised definite description, anti-descriptivistobjections to Russell carry over. Sometake the sense of a word to be the concept it expresses. But concepts areoften defined as components of pro -positions, and if ‘the sense of “Fido” ’is unpacked merely as ‘that compo-nent of a proposition which isexpressed by “Fido” ’, Frege’s solutionseems hollow. More substantial andcontroversial theories of concepts areavailable, but at the very leastFregeans must address profoundquestions in the philosophy of mindbefore they can claim to have solvedthe substitutivity puzzle.

To sum up, propositional attitudereports appear to behave differentlyfrom sentences devoid of psychologi-cal attitude verbs. These differencespose a challenge to semanticists. Thecompositionality thesis, in particular,is difficult to defend in this contextwithout taking potentially controver-sial stands on the nature of proposi-tions, a nature whose investigationbelongs as much to the philosophy ofmind as it does to the philosophy oflanguage.

lack a shared referent, or indeed anyreferent.

Russell’s defence of Russellianismfell out of favour after Saul Kripke’s1972 criticisms of descriptivism aboutnames. More recent Russellians (forexample, Salmon 1986) have insteadtried to explain the appearance of sub-stitution failure as a pragmatic effect.While (1a) and (1d) express the sameRussellian proposition at the semanticlevel, the small difference betweenthem carries pragmatic significanceeasily mistaken for a semantic differ-ence.

The second view of propositions,Fregeanism, is similar to Russellian-ism save that the elements in a propo-sition are not objects but perspectiveson, or ways of thinking about, or (touse Gottlob Frege’s own terminology)‘modes of presentation’ of, objects.The mapping from sentences onto thepropositions they express is thereforea function from a word onto, not anobject, but a specific mode of presen-tation of an object. Frege calls thisfunction the word’s sense*. Twowords, such as ‘Fido’ and ‘Rex’ (or‘George Orwell’ and ‘Eric Blair’), canhave a common referent but distinctsenses. For this reason, substitutingone of the words for the other withina sentence will change the propositionexpressed. It will not (normally)change the truth value, however. Thetruth value of Fregean propositions isdetermined in the same way as Rus-sellian ones, by referents rather thanmodes of presentation; and Fregeanpropositions can be converted intoRussellian propositions by replacingeach perspective on an object with theobject itself. Substitution will there-fore not (normally) affect truth value.

185

PROPOSITIONS

In the analysis of sentence meaningthe application of the concept ofpropositions has proved to be veryfruitful. In formal semantics, truth-conditional semantics and speech acttheory sentences are analysed withreference to the underlying proposi-tions that they contain. It is central tothese theories that in natural lan-guages propositions are not sentences.‘I am hungry’ uttered by differentspeakers is obviously an instance ofthe same sentence but as the referencechanges with every speaker we have asmany different propositions as wehave speakers.

In formal semantics mathematicaland logical techniques are used todescribe the semantic structure of nat-ural languages. Applying proposi-tional calculus, we can deduce howlogical operators such as negation,conjunction, disjunction and implica-tions build up complex sentences outof other sentences. Propositions arecentral to truth-conditional seman-tics based on Alfred Tarski (1956)and to Montague Grammar, whereRichard Montague (Thomason 1974)applies semantic principles developedfor artificial languages to natural languages.

In his development of J. L. Austin’s(1962) speech act theory, John Searle(1969) divided locutionary acts intoutterance acts (uttering noises andwords) and propositional acts, bywhich the speaker is referring andpredicating. In uttering the sentences(1) ‘Sam smokes habitually’. (2) ‘DoesSam smoke habitually?’ (3) ‘Sam,smoke habitually!’ (4) ‘Would thatSam smoked habitually’ (Searle 1969:22), the speaker performs an asser-tion, asks a question, gives an order

Primary sourcesFrege, G. (1892/1997). ‘On sinn and

Bedeutung’. In M. Beaney (ed.), TheFrege Reader. Oxford: Blackwell.

Kripke, S. (1972/1980). Naming andNecessity. Cambridge, MA: HarvardUniversity Press.

Russell, B. (1919). Introduction to Math-ematical Philosophy. London: GeorgeAllen and Unwin.

Salmon, N. (1986). Frege’s Puzzle. Cam-bridge, MA: MIT Press.

Further readingKripke, S. (1979). ‘A puzzle about belief’.

In A. Margalit (ed.), Meaning and Use.Dordrecht: Reidel. 239–83.

Alex Barber

PROPOSITIONS

A term adopted from philosophy andformal logic*, referring to the lan-guage-independent common denomi-nator of the meaning of all thesentences that express the truth of acertain state of affairs, independent oftheir illocutionary form. Propositionsare the central elements used in theinvestigation of sentence meaning informal semantics and in speech acttheory*.

See also: Artificial Intelligence;Logic; Logical Form;Model-Theoretic Semantics;Possible World Semantics; SpeechAct Theory; Truth Theories; TruthValueKey Thinkers: Austin, J. L.;Montague, Richard; Searle, John;Tarski, Alfred

186

PROTOTYPE

Prototype theory explains how peopleunderstand the meaning of a word byreference to the best example of theobject indicated by the word. Thismeans that, for example, in identifyingan object as belonging to the categoryof furniture, the human mind is gearedto thinking in terms of ‘chair’, that is,the prototype member of that cate-gory, rather than with the more mar-ginal ones such as ‘lamp’ or ‘stove’.

See also: Ambiguity/Vagueness;Cognitivism; Intuition; LanguageGames; Language of ThoughtKey Thinkers: Aristotle; Fodor,Jerry; Whorf, Benjamin Lee;Wittgenstein, Ludwig

Every conceivable thing is in principledistinct from every other. In order forthem to make sense, people grouptogether exemplars of the same kindinto categories. But how precisely themind works in this respect is notalways clear. In the Aristotelian tradi-tion, categorisation is explained as theconjunction of necessary and suffi-cient features. Hence a bird has to bea winged creature as its necessary fea-ture and possess a beak as its sufficientfeature. All members of a category arethought to have an equal status. Onceestablished, a category will separateentities that belong to it from thosethat do not. No ambiguity or varia-tion is allowed.

In the 1970s Eleanor Rosch con-ducted a number of cognitive psycho-logical experiments on how peopleoperate with categories and how theylearn about things through them. Notall members of a category turned out

or expresses a wish. But at the sametime she/he refers to a certain personcalled Sam and expresses the predica-tion ‘smokes habitually’ with respectto that person. In other words, thereference and the predication are thesame in all four sentences, and thusthe same proposition is made, regard-less of their respective illocutionaryacts. This is in opposition to mathe-matical languages, where everyproposition is expressed in the formof a statement.

Logical analysis based on proposi-tions has had a major influence onsemantic studies of natural languages.Although the theories are controver-sial, they can lead to more advancedstudies in semantic linguistics, partic-ularly with respect to sentence mean-ing, and in artificial intelligence.

Primary sourcesSearle, John R. (1969). Speech Acts: An

Essay in the Philosophy of Language.Cambridge: Cambridge UniversityPress.

Thomason, Richmond H. (ed.) (1974).Formal Philosophy: Selected Papers ofRichard Montague. New Haven: YaleUniversity Press.

Further readingAustin, John L. (1962). How to Do Things

With Words. Oxford: Oxford Univer-sity Press.

Kneale, William (1972). ‘Propositions andtruth in natural languages’ Mind 81:225–43.

Tarski, Alfred (1956). Logics, SemanticsMetamathematics: Papers from 1923 to1938. Trans. J. H. Woodger. Oxford:Clarendon Press.

Jürg Strässler

187

PROTOTYPE

a higher, superordinate level in contrast to lower, subordinate levelterms such as ‘robin’, ‘canary’,‘parrot’, and so on. Between these twolevels stands the basic level term, inthis case, ‘bird’. In creating categories,people tend to opt for the basic levelterms. Thus when referring to anavian creature, one normally says,‘Look at this bird!’ rather than men-tioning a specific bird (robin, and soon) or using the superordinate term‘animal’. Basic level terms are themost natural form of categorisation,and for that reason they are said tohave the highest degree of cue-validity.

After being shown a word that rep-resents a higher level category, peopletend to find it easier to agree that twowords representing central membersof a category are the same. Thus whenprimed with the word ‘furniture’,people generally take less time indeciding that ‘chair-chair’ are thesame words than ‘stove-stove’. Peopleare fairly consistent in identifying thebest examples of a category. Again,the speed with which people judgewhether an object belongs to suchand such category also determinesthe place the object will occupy on thegraded scale within the category. Theshorter the response time, the morecentral an exemplar will be. Peoplealso seem to rely on the properties ofthe central members in dealing withother exemplars. This means thatattention is normally given first to thecentral members. It is therefore easierto elicit them than the marginal mem-bers.

All these are called prototypeeffects. They play an important role inthe way people mentally organise theoutside world. However, these effects

to be equally representative. Some aremore central to the category; othersare marginal and even dubious. Forexample, the people in Rosch’s studycategorised different kinds of birdsaccording to a graded scale of recog-nisability ranging from the best exem-plar, such as ‘robin’, to the least likelymembers like ‘ostrich’. Between theseopposite ends there is a range ofwhat people will consider as a bird.Very close to ‘robin’, but already lesstypical, were birds like ‘dove’, ‘spar-row’, ‘canary’. Then slightly awayfrom the centre was another groupwith ‘parrot’, ‘owl’, ‘pheasant’, and‘toucan’ in it. Somewhat more distantare ‘peacock’ and ‘duck’, then ‘pen-guin’, until finally, with some hesita-tion, the ‘ostrich’ is considered asbelonging to the category. Of all thebirds in Rosch’s study, ‘robin’ excelledas the most representative member. Itwas the prototype of the category.

Similar experiments with other cate-gories such as vegetables and furnituregive the same result. These categoriesexhibit a graded scale of membership,each with a prototype of its own, in thiscase, ‘pea’ and ‘chair’. This suggeststhat people group things together andexclude others according to their simi-larity to a prototype. Hence the notionof prototype gives a better account ofhow the mind sets up categories anduses them than the Aristotelian frame-work. The finding has a strong impactnot only in cognitive psychology butalso in linguistic research, especially inthe works of R.W. Langacker in cogni-tive grammar and in George Lakoff’ssemantic studies.

An important notion that links pro-totype with linguistic meaning is the‘basic level term’. The term ‘animal’ is

188

PROTOTYPE

about domestic cats and dogs as typical pets, allowing for snakes andcaimans as less common pets. But ‘petfish’ creates some problems. It doesnot matter whether the fish in ques-tion, usually a goldfish, is prototypicalfish or not, and it is not relevant eitherwhere in the scale of pets the goldfishstands. In fact the combinationbetween ‘pet’ and ‘fish’ here is nolonger understood in terms of the proto types of each category.

Prototypes are a mental construc-tion that helps us to understand theworld. This construction or model isbuilt upon an immense storage ofconcepts that can be retrieved tocombine with one another in order toform a model to represent reality inthe mind. One such powerful modelis family resemblance, a notion thatgoes back to Wittgenstein. This isexemplified by the use of the word‘game’. Although every game hassome similarity with other games,there is no single feature that linksthem all. Instead, there is a compli-cated network of similarities overlap-ping and criss-crossing. There issomething that makes all members ofa category resemble one another evenif it cannot be said that they share acommon set of properties. What hap-pens is that each member is linkedto the others by an intermediatemember with which each is in closecontact and thus shares some proper-ties. Thus, schematically, it is possiblethat in a category ABC the membersA and C have nothing in commonand yet both are part of a familysimply because each is in contact withmember B.

As synthesised by John R. Taylor,the notion of prototype is very influen -

are not the same as the cognitive representation itself. The difficultycomes from the role of prototypeThough intuitively a robin is the pro-totypical bird, it cannot be used as aunique criterion for being a bird.Properties of some other birds are alsotaken into consideration. The ques-tion is which properties are to beincluded and which ones are to beignored – in other words, how far onecan extend the category. While anostrich may be quite far removed froma typical bird, it is still a bird, not abirdlike creature the way a bat is. Acolour is no longer considered to bered as it moves towards being orangeor purple. But where are the bound-aries of these categories? Such casesshow that there are no clear-cutboundaries – in other words, cate-gories are inherently fuzzy.

Furthermore, many words have dif-ferent meanings, evoking differentsets of membership. The word ‘jail-bird’, for example, refers to a criminalwho is locked up behind bars like acanary and who may be persuaded to‘sing’ (confess) in order to get bettertreatment. This shows that the word‘bird’ is associated not with a singlecategory, namely the avian category,but also with other categories, in thiscase prisoner. Polysemy is the basisfor metaphors*. The expression ‘he isa fox’ to describe a cunning and slyperson is an example where the meta -phorical use of the word ‘fox’ gives ita new sense. But obviously slyness isnot part of the ordinary sense of theword ‘fox’.

In addition to the problems above, acombination of categories can pro-duce a more complex concept withoutreference to prototypes. People talk

189

PSYCHOANALYSIS

Löbner, Sebastian (2002). UnderstandingSemantics. London: Arnold. Chapter 9.

Agustinus Gianto

PSYCHOANALYSIS

A method of analysing discourse inorder to discover its hidden meanings,originally developed as a cure for various mental disorders, but laterextended to become a method ofanalysing literature, film and othercultural phenomena. Psychoanalysiswas developed as a technique by Sig-mund Freud from the 1890s to the1930s; his disciple Jacques Lacanimported structural linguistics into thetheory, and extended it into a generalphilosophy of everyday life that hasbeen influential in the continental tra-dition.

See also: Signs and Semiotics;Structuralism; PoststructuralismKey Thinkers: Hegel, G. W. F.;Jakobson, Roman; Kristeva, Julia;Lacan, Jacques; Plato; Saussure,Ferdinand de

Psychoanalysis was founded as a dis-cipline by Sigmund Freud in the1890s, but it was through the work ofLacan that it took a ‘linguistic’ turn.Freud, who was contemporary withFrans Boaz, Émile Durkheim and Ferdinand de Saussure, conceived ofpsychoanalysis as a scientific disci-pline and his successors typicallyviewed it as one of the ‘human sci-ences’ in the manner of anthropology,sociology or linguistics. The range ofconditions which psychoanalysis is

tial in forming grammatical cate-gories. For example, what is proto-typical in transitive verbs is thepresence of two different participants,namely the agent and the goal. If theagent and the goal are the same, thenthe verb is transitive-reflexive and issaid to be less prototypical of the classof transitive verbs. The notion of pro-totypes also operates in other areas oflinguistics but it is in this cognitivearena that the idea has had its mostsignificant influence.

Primary sourcesLakoff, George (1987). Women, Fire, and

Dangerous Things. What CategoriesReveal About the Mind. Chicago:Chicago University Press.

Langacker, R.W. (1987, 1991). Founda-tions of Cognitive Grammar. Vols 1–II.Stanford: Stanford University Press.

Rosch, Eleanor (1975). ‘Cognitive repre-sentations of semantic categories’. Jour-nal of Experimental Psychology:General Vol. 104: 192–233.

Rosch, Eleanor (1978). ‘Principles of Cat-egorization’. In E. Rosch and B. B.Lloyd (eds), Cognition and Categoriza-tion. Hillsdale; NJ: Lawrence Erlbaum.27–48.

Rosch, Eleanor and C. B. Mervis (1975).‘Family resemblances: studies in theinternal nature of categories’. CognitivePsychology Vol. 7: 573–605.

Taylor, John R. (1995). Linguistic Catego-rization. Prototypes in LinguisticTheory. Second edition. Oxford:Oxford University Press.

Further readingAitchison, Jean (2003). Words in the

Mind. An Introduction to the MentalLexicon. Third edtion. London: Black-well. Chapters 4–6.

190

PSYCHOANALYSIS

and a baseball. The interpreter putsthe two images together to arrive atthe single word ‘screwball’, which as asignifier of Daffy’s daffiness partakesof the meaning of neither ‘screw’ nor‘ball’. In order to arrive at this inter-pretation, the interpreter must per-form a two-stage ‘decoding’: first heor she must translate each picture intoa word – a movement from symbolto signification – and then translatethe individual words into a combina-tion that is homophonically, butnot semantically, related to them – amovement from signification to meaning. The dreamer, meanwhile,has already performed the samesequence of operations as an encodingprocess, this constituting the dream-work.

In later works Freud extends theprovince of psychoanalysis as an inter-pretive technique from dreams toother, more explicitly linguistic phe-nomena, such as jokes and flashes ofwit, and ultimately to the discourse ofthe patient more generally. It isthis fact, that psychoanalysis takesas its working material the discourseof the patient, that inspires Lacan’s‘linguisticisation’ of Freud, famouslyclaiming that ‘the unconscious is structured as a language’ (Lacan 1972:188). According to Lacan, Freud’s‘condensation’ corresponds to RomanJakobson’s concept of metaphor*,‘displacement’ corresponds to Jakob-son’s metonymy, while ‘figurability’(or ‘transposition’) corresponds to ‘thesliding of the signified under the signi-fier’ (Lacan 2006: 425).

To take the third of these phenom-ena first, Lacan develops Saussure’stheory of the sign, as represented inthe Figure 5 (Saussure 1992/1983).

used to treat has shrunk since Freud’stime, from encompassing psychosis,aphasia, schizophrenia and whatFreud and his contemporaries called‘hysteria’, to being confined to the rel-atively minor condition of neurosisand associated personality disorders.Yet psychoanalysis continues to havecurrency as a tool of cultural critique.

Freud’s claim to importance in lin-guistics comes primarily throughChapter VI of The Interpretation ofDreams (1899), ‘The Dream-Work’.According to Freud, the ‘dream-work’ (the work in which the uncon-scious mind of the dreamer engages inorder to create a dream) consists offour stages: condensation, displace-ment, figurability and secondary revi-sion. Condensation is the processwhereby a detailed and seeminglylong dream can take place within avery short amount of real time. Displacement is the process of replac-ing material the dreamer wishes torepress by other elements whichapparently have a different meaning.Figurability is the process of turningdream material into pictures – mostdreams are pictorial, or are a combi-nation of pictures and language. Andfinally secondary revision is theprocess of turning the dream into a‘day-dream’, of organising the mate-rial so that it may be remembered asa coherent sequence.

The first three of these processes,claims Freud, organise the dream intoa series of rebuses. A rebus is a kind ofpicture-puzzle, whereby pictures sym-bolise words in an indirect manner.For example, in a 1940s Warner BrosMerrie Melodies cartoon, Daffy Duckis shown with a flag attached to histail, on which are depicted a screw

Figure 5

191

PSYCHOANALYSIS

which determines how psychic phe-nomena are organised, and hence howsubjects behave: according to thisexample, men and women line upbefore the doors according to the sig-nifiers that label them. In other words,sex is determined socially, rather thanbiologically. Moreover, the bar sepa-rating signifier from signified, whichwas a relatively minor part of Saus-sure’s theory, assumes a major role inLacan’s: it represents nothing otherthan the bar of repression. Thus evenin everyday life, signifiers do not sig-nify signifieds directly, but rather, theysignify, according to Lacan, other sig-nifiers in the signifying chain that con-stitutes the subject’s discourse. It is thetask of analysis to uncover the truesignifieds of these signifiers, whichhave been displaced.

This theory of linguistic displace-ment leads to Lacan’s appropriation ofJakobson’s theory of metaphor andmetonymy. In a series of papers onaphasia, Jakobson defines metaphor asthe selection or substitution of termsone for another, while metonymy is

Here, the picture of the tree representsthe concept (‘signified’) of a tree in thespeaker’s mind, while the word arborrepresents the ‘sound image’ (‘signi-fier’) of that concept. The horizontalline shows that signifier and signifiedare discrete manifestations, while theoval shows that nevertheless togetherthey form a psychological unity. Thechoice of a tree (arbre in French) forillustrative purposes is a little joke onSaussure’s part, since the relationbetween signifier and signified is heldto be arbitrary (arbitraire) – and like-wise the signified is separated from thesignifier by a barre (barre being ananagram of arbre).

Lacan (2006: 416), meanwhile,replaces Saussure’s diagram with theanother (see Figure 6).

This both inverts Saussure’s dia-gram and reduplicates it. The signifieris shown to have priority for the sub-ject over the signified, by being placedabove it. Meanwhile, the two doorsare identical, so that only the signifiersdistinguish them. Accordingly, the sig-nifier for Lacan has a material reality,

Figure 6

192

PSYCHOANALYSIS

the desire to be desired’. This latterversion introduces the Other into theequation, so that an alternative for-mulation is ‘desire is the desire of theOther’, in which the word ‘of’ canlikewise be read as either ‘for’ or ‘by’.

Lacan’s theory has always beencontroversial: as a linguist, Lacan isindebted to his predecessors Jakobsonand Saussure, and hence his theory isopen to the same criticisms as is struc-turalist linguistics more generally – forexample, that it fails to account forcreativity* in language. However, psy-choanalysis still has currency as amethod of cultural analysis, particu-larly within literary and film studies,and as exemplified through the workof Slavoj Žižek.

Primary sourcesFreud, Sigmund (1976). ‘The Dream-

Work’. In The Interpretation ofDreams. Trans. James Strachey. Har-mondsworth: Penguin. 381–651.

Hegel, G. W. F. (1967). The Phenomenol-ogy of Mind. Trans. J. B. Baillie. NewYork and Evanston: Harper.

defined as combination of one termwith another. Lacan, meanwhile, conceives of Jakobson’s distinctionbetween metaphor and metonymy interms of Hegelian dialectics: his exam-ple, ‘Your freedom or your life!’, is bor-rowed from Hegel’s Phenomenology ofMind. Superficially it looks like aJakobsonian metaphorical equation,freedom=life, where one term canbe substituted for another. However,if the slave ‘chooses freedom, he losesboth [freedom and life] immediately –if he chooses life, he has life deprivedof freedom’ (Lacan 1979: 212).According to Lacan, insofar as the sub-ject exists as a speaking being, he issubjected to this demand by being assuch.

While metaphor is thus equatedwith demand, metonymy is equatedwith desire. The formula for themetonymy of desire is, says Lacan,‘desire is the desire of a desire’. Thiscan be read either as ‘desire is thedesire to desire’ (which, followingPlato, means that by definition desirecan never be fulfilled), or as ‘desire is

193

RELEVANCE THEORY

the study of human communicationwhich is firmly grounded in a generalview of human cognitive design. WithH. P. Grice, relevance theorists assumethat human communication is charac-teristically intention-based, and sothey see verbal comprehension asinvolving not just the decoding ofspeech signals, but also the recogni-tion of the speaker’s communicativeintentions.

See also: Implicature; LogicalForm; PresuppositionKey Thinkers: Chomsky, Noam;Fodor, Jerry; Grice, H. P.

Relevance is usually defined as a poten-tial property of inputs (such asassumptions, thoughts, utterances) tocognitive processes. Sperber andWilson (1995) advance two principlesof relevance embodying two centralclaims about human cognition andcommunication: a first or ‘cognitive’principle of relevance, and a second or‘communicative’ principle of rele-vance. The first of this principles statesthat: ‘Human cognition tends to begeared to the maximisation of rele-vance’ (1995: 260). The second is thestatement that ‘[e]very act of ostensivecommunication communicates a pre-sumption of its own optimal relevance’(1995: 260). The communicative prin-ciple of relevance does not have thesame status as Grice’s CooperativePrinciple and conversational maxims,for it is simply a generalisation aboutostensive-inferential communication.This generalisation about communica-tion applies without exception: it is notsomething that speakers ‘follow’ orcan ‘opt out’ of, for example.

Relevance theory claims that lin-guistic communication (and, in fact,

Jakobson, Roman (1971). Selected Writ-ings II: Word and Language. The Hagueand Paris: Mouton.

Lacan, Jacques (1972). ‘Of structure as aninmixing of an otherness prerequisite toany subject whatever’. In RichardMacksey and Eugenio Donato (eds),The Structuralist Controversy: TheLanguages of Criticism and the Sciencesof Man. Baltimore and London: JohnsHopkins University Press. 186–200.

Lacan, Jacques (1979). The Four Funda-mental Concepts of Psychoanalysis.Trans. Alan Sheridan. Harmondsworth:Penguin.

Lacan, Jacques (2006). Écrits: The FirstComplete Edition in English. Trans.Bruce Fink with Héloïse Fink and Rus-sell Grigg. New York and London:Routledge.

Saussure, Ferdinand de (1922/1983).Course in General Linguistics. Secondedition. Trans. Roy Harris. London:Duckworth.

Further readingForrester, John (1980). Language and the

Origins of Psychoanalysis. Basingstoke:Macmillan.

MacCabe, Colin (ed.) (1981). The TalkingCure: Essays in Psychoanalysis andLanguage. Basingstoke: Macmillan.

Karl Simms

RELEVANCETHEORY

A cognitive theory of pragmatics orig-inally developed in the 1980s by DanSperber and Deirdre Wilson. Rele-vance theory offers a new approach to

194

RELEVANCE THEORY

On the basis of these two compet-ing factors, the relevance of an inputto an individual may be comparativelyassessed as follows: (1) other thingsbeing equal, the greater the positivecognitive effects achieved by process-ing an input, the greater the relevanceof that input; (2) other things beingequal, the greater the effort requiredin processing an input, the lower therelevance of that input. The followingexample, adapted from Wilson andSperber (2004: 609), can be used toillustrate the comparative relevance ofalternative inputs to an individual.Suppose that Peter, a friend of ours,asks us who we phoned last night. Letus assume, furthermore, that wephoned Kim and Sandy last night.Each of the following three alternativeutterances would constitute a true andrelevant answer to Peter’s question:‘We phoned Kim and Sandy’, ‘Wephoned Kim’, and ‘We phoned Kimand Sandy or 2 � 2 � 5’. However,these answers would not be relevantto the same degree: the first would becomparatively more relevant than theother two. Notice that ‘We phonedKim and Sandy’ entails ‘We phonedKim’, and so it is a more relevantanswer because it yields the positivecognitive effects of the second utter-ance and more. The first utterancewould be a more relevant answer than‘We phoned Kim and Sandy or 2 � 2� 5’ because, although these twoutterances are semantically equiva-lent, the latter is obviously more costlyto process. On the whole, when a sim-ilar amount of effort is expended inprocessing alternative inputs, themore relevant of these inputs is theone that yields more positive cogni-tive effects. Conversely, when similar

all human communication) is rele-vance driven. Relevance is definedwithin this framework as a trade-offof two competing factors: cognitive orcontextual effects and processingeffort. A positive cognitive effect is ‘aworthwhile difference to the individ-ual’s representation of the world’(Wilson and Sperber 2004: 608), suchas a true belief or conclusion. Accord-ing to relevance theory, an input is rel-evant to an individual just in casepositive cognitive effects result fromthe processing of that input.

Processing new information in acontext may yield three main types ofpositive cognitive effect. First, itmay yield a contextual implicationdeducible from the combination ofnew and existing assumptions, butfrom neither of these alone. Second, itmay provide evidence that strengthensan already existing assumption.Third, it may contradict and eliminateinformation already held. For exam-ple, given the contextual assumption‘If the lights are on, then Mary ishome’, an utterance of the sentence‘The lights are on’ as we approach ourhouse may yield the contextual impli-cation ‘Mary is home’. Contextualimplications are the central type ofpositive cognitive effect.

Processing effort, on the otherhand, is the effort of perception,memory and inference that must beexpended in computing cognitiveeffects. For instance, an utterance of awordy and syntactically complex sen-tence would take more effort toprocess than an utterance of a lesswordy and simpler version of that sen-tence. Likewise, an indirect answer toa question would require more pro-cessing effort than a direct one.

195

RELEVANCE THEORY

one that satisfies their expectations ofoptimal relevance. Notice that theterm ‘interpretive hypotheses’ as usedin the definition above includes notjust the proposition the speakerintended to communicate, but also thecontextual assumptions, implicaturesand attitudes intended by the speaker.

The relevance-theoretic compre-hension procedure does not guaranteethat communication will always besuccessful, of course. Misunderstand-ings do occur from time to time.Rather, this heuristic offers an accountof how addressees select the interpre-tive hypothesis they are entitled toassume is the one overtly intended bytheir interlocutors.

Relevance theory rejects the tradi-tional assumption that every pragmat-ically determined aspect of utteranceinterpretation other than referenceassignment and disambiguation mustbe an implicature*. Central to thisframework is the claim that theexplicit side of communication shouldalso fall under the scope of a theory ofpragmatics (Bach 1994; Carston2002, 2004b; Sperber and Wilson1993, 1995). Thus, according to relevance theory, there are twotypes of communicated ‘assumptions’(conceptual representations of theactual world): explicitly communi-cated assumptions, or ‘explicatures’,and implicitly communicated ones(implicatures). When is an assump-tion communicated by an utterance‘explicit’? Sperber and Wilson (1995:182) suggest that an explicature is aninferential development of the propo-sitional template or ‘logical form’*encoded by an utterance. In otherwords, an explicature involves a com-bination of linguistically decoded

positive cognitive effects are derivablefrom the processing of alternativeinputs, the one which is less costly toprocess is the more relevant.

How much relevance are individu-als entitled to expect? According torelevance theory, maximal relevance isan unreasonably high expectation incommunication because, for example,our interlocutors might be unwillingor unable to produce information thatwould yield the most positive cogni-tive effects for the least processingeffort (Higashimori and Wilson1996). In light of this, Sperber andWilson (1995) have argued that whilecognition tends to be geared to themaximisation of relevance, acts ofostensive communication simplycreate an expectation of optimal rele-vance. In other words, for any osten-sive stimulus (for example, a verbalutterance) addressees are only entitledto expect a degree of relevance that issufficient to warrant their effort inprocessing it, and which is also thehighest degree of relevance that theirinterlocutors are able to achieve giventheir abilities, goals and preferences.

The presumption of optimal rele-vance suggests the following generalcomprehension procedure: ‘Checkinterpretive hypotheses in order ofaccessibility, that is, follow a path ofleast effort, until an interpretationwhich satisfies the expectation of rele-vance is found; then stop’ (Carston2002: 45). Every utterance gives riseto a number of possible interpretivehypotheses that are compatible withthe linguistic meaning of the sentenceuttered. According to this general criterion, addressees follow a pathof least effort in considering suchhypotheses, stopping once they reach

196

RELEVANCE THEORY

intention that a specific implicationshould be recovered. As Sperber andWilson (1995: 197) put it: ‘Some impli-catures are made so strongly manifestthat the hearer can scarcely avoidrecovering them. Others are made lessstrongly manifest’. Consider, for exam-ple, the utterance ‘I don’t like actionthrillers’ as an answer to the question‘Have you seen the latest HarrisonFord film?’. It is not difficult to see thatsuch a response can be used to implic-itly communicate the assumptions that‘The latest Harrison Ford film is anaction thriller’ and, in addition, that ‘Ihaven’t seen the latest Harrison Fordfilm’. According to relevance theory,these assumptions are strong implica-tures because their recovery is essentialto understand the speaker’s intendedmeaning. In other words, the expecta-tion of optimal relevance that an utter-ance of ‘I don’t like action thrillers’gives rise to requires that the addresseetake these highly salient assumptions(or very similar ones) as being implic-itly communicated. The responseabove may also implicitly communi-cate the assumption that the speakerhas a general dislike for commercialblockbusters, for example. However,this assumption is a weak implicaturebecause it need not be supplied by theaddressee in order to satisfy his expec-tation of optimal relevance. Indeed,there are many other similar implica-tures which may have been derived onthe basis of the speaker’s response, forexample that the speaker is not really afilm enthusiast, or the speaker frownsupon the use of violence. In general,the more obvious the speaker’s com-municative intentions, the stronger thecommunication. Conversely, the widerthe range of interpretive possibilities

material and pragmatic enrichment.By contrast, implicatures are commu-nicated assumptions resulting frompragmatic inference alone. Let us consider an example of a relevance-theoretic explicature involving therestriction of the domain of a quan-tificational expression. An utteranceof the sentence ‘Every cat has whitepaws’ does not make a patently falseuniversal claim about cats, but willtypically be used to express a morerestricted true proposition – for exam-ple, a proposition about cats in ourlocal shelter. Thus, an explicature ofthis utterance would be ‘Every cat inour local shelter has white paws’. Aspointed out above, this conceptualrepresentation is a pragmatic develop-ment of the propositional schema corresponding to the conventionalmeaning of the sentence uttered.

The class of explicatures of a givenutterance within the relevance- theoretic framework includes not justthe proposition expressed by thatutterance, but also a range of so-called‘higher-level’ explicatures, which areobtained by embedding that proposi-tion under an appropriate pro -positional attitude* or speech actdescription (Sperber and Wilson1993). For example, an utterance ofthe sentence ‘The boss is coming’ maybe developed inferentially into thehigher-level explicature ‘The speakerbelieves that the boss is coming’, oreven ‘The speaker is warning that theboss is coming’.

An interesting distinction is made inrelevance theory between strong andweak implicatures (and communica-tion, more generally). The strength ofan implicature crucially depends onthe manifest strength of the speaker’s

197

RELEVANCE THEORY

spective, using expressions whichencode procedures for the identifica-tion of intended cognitive effectswould obviously reduce the processingcost involved in achieving thoseeffects, a result that is in consonancewith the communicative principle ofrelevance. For example, Blakemore(2002) links the use of the sententialconnective ‘but’ with the cognitiveeffect of contradiction and elimina-tion. Hence, according to Blakemore,the use of ‘but’ activates an inferentialprocess resulting in the contradictionand elimination of an assumptionwhich the speaker has reason tobelieve is accessible to the hearer. Forexample, an utterance of the sentence‘Kim is rich, but unhappy’ activates aninferential process whereby the hearercontradicts and eliminates the accessi-ble assumption that wealth leads tohappiness.

Research in relevance theory hasalso made an important contributionto the study of figurative language.Verbal irony, for example, is analysedin this framework as an ‘echoic use’ oflanguage in which the speaker dissoci-ates himself tacitly from an attributedutterance or thought (Wilson 2006;Wilson and Sperber 1992). For exam-ple, uttering ‘You really are good atthis!’ after a friend has failed to scorean easy goal in a quick counter-attackcan be construed as an instance ofverbal irony because we are tacitlydissociating ourselves from a thoughtor utterance with a similar content(such as a reassurance that our friendis a skilful footballer) which mayhave been attributed to us had the cir-cumstances of the game been differ-ent. This analysis is a departure frommore traditional Gricean accounts,

allowed by the speaker, the weaker thecommunication. The relevance-theo-retic approach to the implicit-explicitdistinction has led to a significantreassessment of the interface betweensemantics and pragmatics: the exis-tence of pragmatic aspects of pro -positional content which do notcorrespond to items present in thesyntactic representation, as argued byrelevance theory, strongly suggeststhat context-sensitivity at this level iswidespread.

An important development withinrelevance theory was the recognition,due to the seminal work of DianeBlakemore (1987, 2002), that linguis-tic meaning can encode constraints onthe inferential phase of utterance com-prehension. Thus, linguistic meaningmay affect the inferential processesthat characterise utterance compre -hension in two different ways: while amajority of linguistic expressionsencode constituents of conceptual rep-resentations, there are also expressionswhich encode inferential procedures,which we could think of as instructionsto increase the salience of a particulartype of inferential process.

Blakemore justified the distinctionbetween conceptual and proceduralencoding in both cognitive and com-municative terms. Since, as it isassumed in relevance theory, the inter-pretation of utterances involves carry-ing out computations over conceptualrepresentations, it is reasonable toexpect from a cognitive point of viewthat languages encode informationabout the inferential procedures inwhich such conceptual representationsenter (and not just the constituentsof conceptual representations them-selves). From a communicative per-

198

RELEVANCE THEORY

versus implicating) in the study ofmeaning in natural language. More-over, it has helped to enhance ourunderstanding of the semantics- pragmatics interface by arguing contro-versially that the contribution ofpragmatics to the propositional contentof utterances goes far beyond disam-biguation and reference assignment.

Primary sourcesBlakemore, Diane (1987). Semantic Con-

straints on Relevance. Oxford: Black-well.

Blakemore, Diane (2002). Relevance andLinguistic Meaning: The Semantics andPragmatics of Discourse Markers. Cam-bridge: Cambridge University Press.

Carston, Robyn (1988). ‘Implicature,explicature and truth-theoretic seman-tics’. In Ruth Kempson (ed.), MentalRepresentations: The Interface betweenLanguage and Reality. Cambridge:Cambridge University Press. 155–81.

Carston, Robyn (2002). Thoughts andUtterances: The Pragmatics of ExplicitCommunication. Oxford: BlackwellPublishing.

Carston, Robyn (2004a). ‘Relevancetheory and the saying/implicating dis-tinction’. In G. Ward and L. Horn (eds),The Handbook of Pragmatics. Oxford:Blackwell. 633–56.

Carston, Robyn (2004b). ‘Explicature andsemantics’. In S. Davis and B. S. Gillon(eds), Semantics: A Reader. Oxford:Oxford University Press. 817–45.

Carston, Robyn and George Powell(2006). ‘Relevance theory – new direc-tions and developments’. In E. LePoreand B. Smith (eds), Oxford Handbookof Philosophy of Language. Oxford:Oxford University Press. 341–60.

Higashimori, Isao and Deirdre Wilson(1996). ‘Questions on relevance’. UCL

according to which verbal irony con-stitutes an overt violation of the Qual-ity maxim, thus giving rise to a relatedtrue implicature which contradictsthe literal meaning of the sentenceuttered. The relevance-theoreticapproach to verbal irony is in har-mony with the framework’s assump-tion that an expectation of relevance,rather than one of truthfulness, is astandard of verbal communication(Wilson and Sperber 2002).

A recent line of research in relevancetheory explores in detail the idea thatthe interpretation of words is alsohighly context-sensitive, and that lexi-cal context-sensitivity is not restrictedto indexical expressions. This line ofinvestigation puts forward the hypoth-esis that lexical-pragmatic processessuch as narrowings, broadenings,approximations and metaphoricalextensions are the result of a singlepragmatic procedure which fine-tunesthe conventional meaning of words incommunication (Carston and Powell2006; Wilson 2003).

In sum, research carried out withinthe framework of relevance theory hasyielded interesting and insightful results(see, for example, Carston 1988, 2002,2004a, 2004b; Carston and Powell2006; Sperber and Wilson 1987, 1993,1995). First, it has provided a cognitivealternative to Gricean and neo-Griceantheories of pragmatics. It has also intro-duced (or thrown new light into) anumber of important theoretical con-cepts (explicature, echoic uses of lan-guage, strength of communicatedassumptions) and distinctions (decod-ing versus inference, explicature versusimplicature, conceptual versus proce-dural meaning, interpretive versusdescriptive uses of language, saying

199

SENSE DATA

F. Keil (eds), MIT Encyclopaedia of theCognitive Sciences. Cambridge, MA:MIT Press. 719–22.

Iván García Álvarez

SENSE DATA

Mental images of physical objects thatwe perceive directly, with observedproperties representing those of thephysical object itself. Popular amongphilosophers through the early twen-tieth century, the sense data theory hasbeen criticised for its inability tosolidly define and recognise sense dataas such.

See also: Empiricism/Rationalism;Ideational Theories;Indeterminacy; Mentalism;Sense/ReferenceKey Thinkers: Austin, J. L.; Ayer,A. J.; Berkeley, George; Descartes,René; Hume, David; Kant,Immanuel; Locke, John; Moore,G. E.; Russell, Bertrand

Although philosophers as far back asJohn Locke, George Berkeley andDavid Hume wrote about human per-ception and understanding, the term‘sense data’ (or its singular form ‘sensedatum’) is more modern, having firstappeared in the early twentieth cen-tury in the works of such thinkers asG. E. Moore and Bertrand Russell. Inthe philosophy of perception, thesense data theory argues that proper-ties of mental images directly correlateto those of the physical objects theyrepresent. Growing criticism to thecontrary – most notably by J. L.Austin, Frank Jackson and Wilfrid

Working Papers in Linguistics 8:111–24.

Sperber, Dan and Deirdre Wilson (1987).‘Précis of relevance: communicationand cognition’. Behavioral and BrainSciences 10: 697–754.

Sperber, Dan and Deirdre Wilson (1993).‘Linguistic form and relevance’. Lingua90: 1–25.

Sperber, Dan and Deirdre Wilson (1995).Relevance: Communication and Cogni-tion. Oxford: Blackwell. First edition1986.

Wilson, Deirdre (2003). ‘Relevance theoryand lexical pragmatics’. Italian Journalof Linguistics/Rivista di Linguistica 15:273–91.

Wilson, Deirdre (2006). ‘The pragmaticsof verbal irony: echo or pretence?’.Lingua 116: 1722–43.

Wilson, Deirdre and Dan Sperber (1992).‘On verbal irony’. Lingua 87: 53–76.

Wilson, Deirdre and Dan Sperber (2002).‘Truthfulness and relevance’. Mind 111:583–632.

Wilson, Deirdre and Dan Sperber (2004).‘Relevance theory’. In G. Ward and L.Horn (eds), The Handbook of Pragmat-ics. Oxford: Blackwell. 607–32.

Further readingBach, Kent (1994). ‘Conversational implic-

iture’. Mind and Language 9: 124–62.Blakemore, Diane (1995). ‘Relevance

theory’. In J. Verschueren, J. Östmanand J. Blommaert (eds), Handbook ofPragmatics. Amsterdam: Philadelphia.443–52.

Blakemore, Diane (2002). UnderstandingUtterances: An Introduction to Prag-matics. Oxford: Blackwell.

Huang, Yan (2007). Pragmatics. Oxford:Oxford University Press.

Wilson, Deirdre (1999). ‘Relevance andrelevance theory’. In R. Wilson and

200

SENSE/REFERENCE

data receive increasing support in lin-guistic circles, where words in lan-guage trigger direct awareness ofinternal interpretations, not property-bearing sense data.

Primary sourcesAustin, J. L. (1962). Sense and Sensibilia.

Oxford: Clarendon Press.Berkeley, George (1998). Three Dialogues

between Hylas and Philonous. Ed. J.Dancy. New York: Oxford UniversityPress.

Hume, David (2000). An Enquiry Con-cerning Human Understanding: A Crit-ical Edition. Ed. T. Beauchamp.Oxford: Clarendon Press.

Russell, Bertrand (1964). The Problems ofPhilosophy. New York: Oxford Univer-sity Press.

Further readingFirth, Roderick (1965). ‘Sense data and

the percept theory’. In R. Swartz (ed.),Perceiving, Sensing, and Knowing. NewYork: Anchor Books. 204–70.

Chrucky, Andrew (1992). ‘The alleged fallacy of the sense datum inference’.Eastern Pennsylvania PhilosophicalAssociation. Bloomsburg University.

Huemer, Michael (2004). ‘Sense data’.Stanford Encyclopedia of Philosophy(http://plato.stanford.edu/).

Jennifer A. Baldwin

SENSE/REFERENCE

The distinction between sense and ref-erence (German ‘Sinn’ and ‘Bedeu-tung’) goes back to the Germanlogician Gottlob Frege. According tohis understanding, sense and reference

Sellars – maintains that we form inter-pretations, not mere mental images,that lack the same properties of thosephysical objects.

In support of the sense data theory,Hume and Russell use an example ofhow a table appears differently to usas we physically move alongside it oraway from it in order to describe howour sensory experience adjusts to per-spectival variation, or changes in ourphysical relationship to externalobjects. Because the size and shape ofthe physical table do not change, ourperception is our awareness of thetable’s corresponding mental image.Other arguments supporting sensedata have to do with perception ofmental images not directly related tophysical objects. This includes opticalillusions, hallucinations, doublevision and even time delays betweenthe existence of a physical object andour perception of it.

More recently, criticisms of the sensedata theory have directly refuted itssupporting arguments. For example,sense data theorists would argue that ifthing X were made to look exactly likething Y, what we are directly aware ofin our perception is thing Y. Thismental image of thing Y is evidence ofsense data. Opposing arguments, how-ever, state that this illusion does notchange the fact that we are still lookingat and perceiving properties of a real(or material) thing X.

Originating from discussions onhuman knowledge and understanding,the sense data theory is an importantcomponent of the philosophy of per-ception. As the debate continues overwhether mental phenomena behindperception exist as mental images or aspure interpretation, critics of sense

201

SENSE/REFERENCE

contrast, identified meaning withthe reference of the expression. In con-nection with this denotational concep-tion of meaning, he worked out atheory around two main puzzles about language.

The first puzzle is in dealing withidentity statements such as ‘the Morn-ing Star is the Evening Star’. Of coursethis sentence conveys the informationthat the objects called the MorningStar and the Evening Star (this pair isused by Frege) are physically identical(both are the planet Venus). If we sub-stitute the expression ‘Evening Star’with the co-referring expression‘Morning Star’ in the example above,it results in ‘the Morning Star is theMorning Star’. In contrast to the firststatement, this is a tautology. It givesonly trivial information and does nottell us anything new because the state-ment holds a priori. It is not possibleto comprehend and above all toexplain this difference in the meaningof the two sentences solely by the ref-erence of a sign, because this is thesame for both expressions.

This observation leads Frege to con-clude that there is a second aspect tothe meaning of a sign besides the ref-erence, namely the sense (‘Sinn’). Thesense of a sign corresponds to themode of presentation (‘Art desGegebenseins’). In the example theexpressions ‘Morning Star’ and‘Evening Star’ present two differentways of characterising the planetVenus, the brightest shining visiblecelestial body in either the evening orthe morning. The difference betweenthem in the example given above liesin a difference in the mode of presen-tation of the signs since their refer-ence, the planet Venus, is the same.

are two different aspects of meaning oflinguistic expressions, reference beingthe entity to which an expressionrefers and sense being the way inwhich that entity is presented. Forexample, the expressions ‘Tom’ and‘the prisoner’ may well refer to thesame person, but they differ in sense.This highly controversial distinctionhad, and still has, a strong influence onsemantics and philosophy of language.

See also: Analytic Philosophy;Compositionality;Connotation/Denotation; DefiniteDescriptions; Names;Propositional Attitudes; TruthValueKey Thinkers: Carnap, Rudolf;Dummett, Michael; Frege,Gottlob; Kripke, Saul; Leibniz,Gottfried Wilhelm; Mill, J. S.;Russell, Bertrand; Searle, John;Strawson, P. F.

Gottlob Frege, the father of modernmathematical logic, and also consid-ered one of the founders of analyticphilosophy, developed the distinctionbetween sense and reference mainly inhis essay ‘On Sense and Reference’(‘Über Sinn und Bedeutung’, 1892).This paper is considered one of themost important historical sources ofmodern semantics. In it Frege sug-gested that in order to grasp the mean-ing of an expression, one mustcomprehend not only the object itrefers to but also the way the object ispresented, that is to say the sense ofthe referring expression.

Frege’s theory was opposed to thethen common understanding, datingback to John Locke, that identifiedthe meaning of a word with the inter-nal idea connected with it. Frege, in

202

SENSE/REFERENCE

But a referent can correspond to sev-eral senses (not necessarily a uniquesense), for example in different lan-guages. With ‘sign’ Frege refers toproper names, under which hewidely subsumes all expressionswhich denote or could denote objects,because a grammatically correctmeaningful expression having a sensemight lack a reference. For instance,the expression ‘the least rapidly con-vergent series’ has no referencebecause for each convergent series aless rapidly convergent series can befound. Other examples of emptynames are figures from mythology like‘Odysseus’, as already seen above.This was another reason for Frege todistinguish a second aspect of mean-ing. Frege holds the view that thisinconsistency marks a deficiency ofnatural language and even of symboliclanguage of mathematical analysis.For a logically perfect language(‘Begriffsschrift’) he demands thatevery correctly formed expression hasa reference.

Sense and reference need to be dis-tinguished from the associated idea(‘Vorstellung’), the internal image ofan object, which is subjective. It variesfrom person to person and even forone person it is not always the same.In Frege’s theory of meaning the senselies between the reference of a propername (the object itself) and the sub-jective idea. It is neither subjective noris it the object itself. Frege developedthe theory of sense and reference intoa philosophy of language. Thus,departing from proper names he gen-eralises the twofold semantic functionto entire declarative sentences. Themeaning of a sentence (or complexexpression) is determined by the

The direct reference approach –assuming that the meaning of anexpression is its referent – was alreadyclaimed by Mill but the two problemsdescribed above (identity statementsand empty names) led Frege and laterP. F. Strawson and John Searle to claimthat proper names* have sense, andadditionally but not necessarily, refer-ence. This view is also referred to asmediated reference theory. Philoso-phers like Saul Kripke, John McDow-ell and Gareth Evans have arguedagainst this position, convinced thatFrege’s treatment of so-called emptynames was inconsistent. Kripkeattacked Frege’s and Bertrand Rus-sell’s position in his lectures ‘Namingand Necessity’ asserting that propernames do not have senses at all.Kripke said that the reference of aproper name is determined by theassociated description and heexplained that contrary to the des -criptive properties a proper namerefers to the same object across allpossible worlds (counterfactualworlds included).

The terms used by Frege are ‘Sinn’and ‘Bedeutung’. This terminology isconfusing. ‘Bedeutung’ is generallytranslated as ‘meaning’. But what iscommonly understood as meaning, language internal and independentfrom facts, rather corresponds toFrege’s other term ‘Sinn’ or sense. Otherauthors adapt the connotation/denota-tion or intension/extension distinction.The essay is translated as ‘Sense andReference’ by Max Black (1948), but as‘Sense and Meaning’ in Frege’s Posthu-mous Writings (1979). They are hereinafter called ‘sense’ and ‘reference’.

According to Frege’s theory, a signhas a sense, and this has a reference.

203

SENSE/REFERENCE

This theory of sense and reference canonly be maintained if we assume thatin these contexts the expressions donot have their customary reference. Inhis essay ‘Sense and Reference’, Fregeanalyses various types of subordinateclauses in detail.

If we substitute the proper nameCharlotte Brontë with her pseudonymCurrer Bell in the following true sen-tence ‘Dorothy believes that CharlotteBrontë is the author of Jane Eyre’, themeaning of the entire sentence ischanged because the resulting sen-tence is not necessarily correct:‘Dorothy believes that Currer Bell isthe author of Jane Eyre’. Dorothymight not be aware of the fact thatCharlotte Brontë at that time pub-lished under the assumed masculinename Currer Bell; therefore she mightbelieve that ‘Charlotte Brontë is theauthor of Jane Eyre’ and at the sametime believe that it can not be the casethat ‘Currer Bell is the author of JaneEyre’, despite the fact that they areone and the same person.

The issue described above is alsoknown as Frege’s second puzzle aboutlanguage. How can it be that the prin-ciple of substitution fails in the con-text of propositional attitude* reports(or of indirect quotation)? The answeris that in these cases we talk about thewords themselves. Expressions corre-spondingly do not have their custom-ary reference, but they have theirindirect reference, coinciding withwhat is customarily their sense.Frege introduces the term ‘indirect’(German ‘ungerade’) reference/sensecontrary to the ‘customary’ (‘gewöhn-lich’) reference/sense. This leads to theposition that under the given circum-stances a clause’s reference is not a

meanings of its constituents. This isthe principle of compositionality*controversially attributed to Frege.Hence it should be possible to substi-tute constituents of the sentence withco-referring constituents withoutaffecting the reference of the entiresentence.

Leibniz had formulated this regu-larity as his law of substitution and inFrege’s theory the objective content ofa declarative sentence is called athought (‘Gedanke’) and is under-stood as the sense of a sentence. Sub-stitution might change the sense orthought as in the example givenbelow.

‘The morning star is a body illuminatedby the sun.’‘The evening star is a body illuminatedby the sun.’

In contrast what remains the same isthe truth value of the sentence, whichFrege considers the reference. Theidea of truth values comes from thediscipline of logic or logical semantics.It refers to a function which maps sen-tences on to the truth values true orfalse. Corresponding to what is saidabout proper names, Frege claims thatsentences need to have sense but notnecessarily reference. This can beshown in the sentence ‘Miss Marpleinvestigated the assassination in theOrient Express’. This sentence ismeaningful but, as it contains thename ‘Miss Marple’, a fictional char-acter whose reference is doubtful, itcannot be said to have a reference,that is a truth value*.

The following paragraph highlightsseveral subordinate clauses as exemp-tions from the substitution principle.

204

SENSE/REFERENCE

the antecedent clause nor the conse-quent clause qualifies to express anentire thought. It is their combinationwhich expresses one single completethought. While several types of subor-dinate clauses are in a way incom-plete, others – for example, concessiveclauses – are complete in this respect.

The remaining cases of failure ofthe substitution test are explainedwith additional subsidiary thoughtsnot explicitly expressed. One examplegiven by Frege himself is a causal sub-ordinate clause: ‘Because ice is lessdense than water it floats on water’.Frege argues that the compoundexpresses more than one thought perclause namely three thoughts alto-gether:

1. Ice is less dense that water.2. If anything is less dense that water,

it floats on water.3. Ice floats on water.

The clause ‘because ice is less densethan water’ does not only express thefirst but also part of the secondthought. This overlapping is thereason why it is not possible toexchange the subordinate clause givenabove by another one with the sametruth value without doing harm to thetruth of the entire sentence. It thusdoes not serve to disqualify the theoryof sense and reference of sentences inthis view.

Frege’s ideas had a significantimpact on the development of modernsemantic theories. The distinctionbetween sense and reference continuesto be the subject of research in philos-ophy. Besides the topics already men-tioned, another issue still discussed inthe literature is the questions raised by

truth value but a thought. Thus wecan only substitute the subordinateclause with another with the samethought (that is, reference) in order tokeep the truth value of the complexsentence but not necessarily substituteanother with same truth value.

The situation is comparable in sen-tences with ‘it seems that’ and ‘com-mand’, ‘ask’, ‘forbid’, or ‘doubtwhether’, to name a few. In all thesecases the words in the subordinateclause have an indirect reference andthis determines that the reference ofthe subordinate clause is not a truthvalue but a thought, a command, arequest, a question and so on. Accord-ing to Frege, the subordinate clausefollowing these expressions may alsobe understood as proper name of thethought it represents in the compoundsentence.

The discussion about sense and ref-erence of propositional attitudereports, initialised by Frege, is stillalive. One topic is the question, forexample, of how nested quotations orpropositional attitudes – leading to ahierarchy of senses – fit into Frege’stheory. Apart from cases of indirectreference of words, there are othercases where sense and reference donot correspond to thought and truthvalue respectively. In the sentence‘whoever invented the ingenious MissMarple was a fanciful person’, thegrammatical subject ‘whoever’ has noindependent sense, but only in con-nection with the main clause. Thisaccounts for the fact that the sense ofthe subordinate clause is not a com-plete thought. Besides, the reference isnot a truth value but the personAgatha Christie. A further example isconditional sentences. Usually neither

205

SIGNS AND SEMIOTICS

Harman (eds), Semantics of NaturalLanguage. Dordrecht: Reidel.

Russell, Bertrand (1905). ‘On denoting’.Mind: 479–99.

Quine, W. V. O. (1980). ‘Reference andmodality’. In From a Logical Point ofView: Nine Logico-PhilosophicalEssays. Cambridge, MA: Harvard Uni-versity Press. 139–59. First edition1953.

Further readingBeaney, Michael and Erich H. Reck (eds)

(2005). Gottlob Frege: Critical Assess-ments of Leading Philosophers. Vol. IV:Frege’s Philosophy of Thought and Lan-guage. London: Routledge.

Burge, Tyler (2005). Truth, Thought,Reason: Essays on Frege. Oxford:Oxford University Press.

Carl, Wolfgang (1994). Frege’s Theory ofSense and Reference. Its Origin andScope. Cambridge: Cambridge Univer-sity Press.

Dummett, Michael (1973). Frege: Philos-ophy of Language. London: Duck-worth.

Ricketts, Thomas (ed.) (forthcoming). TheCambridge Companion to Frege. Cam-bridge: Cambridge University Press.

Sluga, Hans D. (1993). The Philosophy ofFrege. Four vols. New York: Garland.

Eva Herrmann-Kaliner

SIGNS ANDSEMIOTICS

The term Semiotics (Semiologie) in thesense that it is understood in thetwenty-first century was first used byFerdinand de Saussure in his 1908

sentences which contain expressionsdepending on context, that is deicticalterms such as ‘today’. As Frege con-siders a thought to be complete, thequestion of what is the sense of theseexpressions arises among Frege’sinterpreters and is broadened to a gen-eral quest for an adequate theory ofdeixis.

Primary sourcesBlack, Max and Peter Geach (eds) (1980).

Translations from the PhilosophicalWritings of Gottlob Frege. Oxford:Blackwell. First edition 1952.

Church, Alonzo (1956). Introduction toMathematical Logic I. Princeton:Princeton University Press. Revised andenlarged version.

Frege, Gottlob (1879). ‘Begriffsschrift,eine der arithmetischen nachgebildeteFormelsprache des reinen Denkens’.Halle: Louis Nebert. Trans. J. Van Hei-jenoort as ‘Concept script, a formal lan-guage of pure thought modelled uponthat of arithmetic’. In J. van Heijenoort(ed.) (1967), From Frege to Gödel: ASource Book in Mathematical Logic,1879–1931. Cambridge, MA: HarvardUniversity Press.

Frege, Gottlob (1892). Über Sinn undBedeutung’. Zeitschrift für Philosophieund philosophische Kritik Vol. 100: 25–50.

Frege, Gottlob (1892/1948). ‘Sense andreference’. Trans. Max Black. ThePhilosophical Review Vol. 57, no. 3(May 1948), 209–30.

Frege, Gottlob (1979). Posthumous Writ-ings. Ed. Hans Hermes, Friedrich Kam-bartel, Friedrich Kaulbach. Trans. PeterLong, Roger White. Chicago: Universityof Chicago Press.

Kripke, Saul (1972). ‘Naming and neces-sity’. In Donald Davidson, Gilbert

206

SIGNS AND SEMIOTICS

Ferdinand de Saussure defined whathe called Semiologie as the science ofsigns. He conceived of semiotics aspart of social psychology and there-fore of general psychology. Semioticsis part of psychology because the signis a twofold psychic entity: such anentity is composed of two sides, theconcept and the acoustic image, bothof which reside in the ‘same psychicsite’.

Saussure was introduced to a sci-ence of signs by his professor MichelJules Alfred Bréal, whose course insemantics he attended in 1881. In hislessons, later collected in his famousEssai de Sémantique (1897), Bréalclaimed that ‘les mots sont des signes:ils n’ont pas plus d’existence que lesgestes du telegraphe aérien ou que lespoints et les traits du télégrapheMorse’ (‘Words are signs: they haveno further existence than the move-ment of the railway semaphore signalor the dots and dashes of Morse code’)(Bréal 1897: 835), and departed fromthe eighteenth-century semiologic tra-dition in order to show the inconsis-tency of the organicistic conception ofthe language. Bréal was influenced bythe studies of Franz Bopp and WilliamDwight Whitney and the notion ofsign they utilised in their languageinvestigation, and by the philosophi-cal perspective of Etienne Bonnot deCondillac and the importance he gaveto psychology in his philosophicalstudies. Bréal’s semantics is thusbased on the psychological laws ofthe human mind, and traces of JohnLocke’s and Condillac’s ideas can befound in his work as well as insightsfrom the studies on psychology andmemory sciences like those offered byHippolite Taine, Théodule-Armand

lessons on general linguistics. A sign isany entity representing another entity:smoke as a sign of fire, or a stop signalalerting drivers to come to a halt at acrossroad. To the whole set of signs inhuman and non-human communica-tion belong different subsets depend-ing on their quality. They can thus beencompassed as visual signs, auditorysigns, verbal signs, cloth signs, and thelike. As far as linguistics is concerned,semiotics is the science that studies lin-guistic signs, which is tantamount tosaying that linguistics is part of a moregeneral science of semiotics which inturn is a branch of general psychology.

See also: Langue/Parole;StructuralismKey Thinkers: Bopp, Franz;Hjelmslev, Louis; Morris, Charles;Peirce, C. S.; Saussure, Ferdinand de

The very first mention of the label‘semiotics’, spelled ‘semeiotics’, datesback to 1670 when the English physi-cian Henry Stubbes used it to denotethe branch of medicine relating to theinterpretation of symptoms. Twentyyears later John Locke used the labelto discuss his tripartite subdivision ofscience:

the third branch may be called Semei-otike, or the doctrine of signs; the mostusual whereof being words, it is aptlyenough termed also Logike, logic: thebusiness whereof is to consider thenature of signs, the mind makes use offor the understanding of things, or con-veying its knowledge to others. (Locke1963: 174)

But we have to wait until the end ofthe nineteenth century to see the labelapplied to language.

207

SIGNS AND SEMIOTICS

Whitney’s claim of the arbitrarynature of sign, Saussure contendedthat the meaning of a sign is not con-tained within it, but arises in its inter-pretation. This means that meaningsdo not exist per se but are establishedby the language users in relation tothe context of use. In other words,the role of the interpreter must beaccounted for, either implicitly as inSaussure, or explicitly as in Peirce.

Saussure’s apparatus was followedby Hjelmslev (1943) who substitutedthe terms expression and content torefer to the signifier and signifiedrespectively. He also referred toplanes of expression and content, eachhaving substance and form. Thusthere are four categories which mayfacilitate analytical distinctions: sub-stance of expression, form of expres-sion, substance of content, and formof content.

In the United States semiotics devel-oped in the second half of the nine-teenth century within the field ofphilosophy where Charles Peirceclaimed that semiosis is ‘an action, orinfluence, which is, or involves, adynamical operation of three subjects,such as sign, its object, and its interpretant, this tri-relative influencenot being in any way resolvable intoaction between pairs’ (Hartshorne etal. 1958: 5.484). Whereas in the Saus-surean theoretical apparatus, the signis a bipartite entity, Peirce conceivedof it as a dynamic triadic relation:

a sign or representamen is somethingwhich stands to somebody for some-thing in some respect or capacity. Itaddresses somebody, that is, creates inthe mind of that person an equivalentsign, or perhaps a more developed sign.

Ribot and Paul Pierre Broca. To thethirty-year shaping of his theory ofsign, Bréal included a strong psycho-logical value that he drew from thestudies on the human mind whichwere so pervasively spreading in thesecond half of the nineteenth century.He thus attempted to search the cog-nitive and intellectual features of thehuman mind which regulate thenature of language.

This aspect of Bréal’s semantics isevident in the work of Saussure, whoaimed to find a bridge between thepsychic essence of the concept and theconcrete reality of the word. He cameup with the idea of ‘sign’ as an indi-visible pair consisting of abstract con-cept and concrete realisation. Untilone of his 1911 lessons, when he firstused the two terms signifiant and sig-nifié, Saussure generally spoke of signin the same way as was done by hiscontemporaries, that is as the phono-logical counterpart of an entity. Laterhe developed a more refined theorywhere the sign is conceived of as adyadic entity formed by the indivisiblecombination of a signifier (signifiant,the acoustic image) and a signified(signifié, the mental representation ofreality). The signifier and the signifiedare intimately linked by an associativelink whereby each triggers the other.Hence they stand in a static dyadicrelationship which goes under thelabel of signification. With referenceto language, a linguistic sign is not alink between a thing and a name, butbetween a concept and a sound pat-tern. This means that language ismainly symbolic, since the relationsbetween the sound sequences andtheir meanings are conventional, orarbitrary, and have to be learnt. On

208

SIGNS AND SEMIOTICS

iourism* and investigated the under-standing of the unitary process ofsemiosis. He proposed to focus on therelation that the sign can establishwith the other entities in the semioticprocess. Thus the sign-object relationpoints to the dimension of semantics;the sign-sign relation refers to syntax;and the sign-interpreter relation refersto pragmatics. Such a tripartite divi-sion became normalised in linguistics.

Semiotics is a broad discipline whichdeals with any type of significationand communication. It encompassesbranches like social semiotics, visualsemiotics, zoosemiotics, music semiol-ogy, computational semiotics, and lit-erary semiotics, to mention but a few.

Primary sourcesBréal, Michel (1897). Essai de séman-

tique. Science des significations. Paris:Hachette.

Eco, Umberto (1976). A Theory of Semi-otics. Bloomington: Indiana UniversityPress/London: Macmillan.

Hartshorne, Charles, Paul Weiss andArthur W. Burks (eds) (1958). CollectedPapers of Charles Sanders Peirce. Eightvolumes. Cambridge, MA: HarvardUniversity Press.

Hjelmslev, Louis ([1943]1953). Prolegom-ena to a Theory of Language. Balti-more: Indiana University Publicationsin Anthropology and Linguistics.

Morris, Charles W. (1938/1970). Founda-tions of the Theory of Signs. Chicago:Chicago University Press.

Saussure, Ferdinand de (1916/1983).Course in General Linguistics. Trans.Roy Harris. London: Duckworth.

Further readingChandler, Daniel (2001). Semiotics: The

Basics. London: Routledge.

The sign which it creates I call the inter-pretant of the first sign. The sign standsfor something, its object. (Hartshorneet al. 1958: 2.228)

Semiosis is therefore an endless pro -cess, something that Eco (1976)defined as ‘unlimited semiosis’.

In his investigation of the triadicnature of sign (Triadism), Peirceclaimed that ‘all thought whatsoeveris a sign, and is mostly of the natureof language’ (Hartshorne et al. 1958:5.421). In his triadism, Peirceanalysed the sign in itself, in relationwith the object, and in relation withthe interpretant, thus developing ahuge number of categories. As far aslanguage is concerned, the most rele-vant triad is the one which originatesfrom the relation that a sign estab-lishes with itself. The sign-to-sign rela-tion produces three modes, an icon,an index, and a symbol. The icon is amode in which the signifier is per-ceived as imitating the signified, as isthe case with onomatopoeia. An indexis a mode in which the signifier is arbi-trary but connected in some way,physically or causally, to the signified,for example in the case of demonstra-tive pronouns. A symbol is a mode inwhich the signifier is fundamentallyarbitrary. These three modes give riseto three basic principles which arefundamental to investigate linguisticphenomena: the principle of indexi-cality, of iconicity, and of symbolicity.

Peirce’s scientific contribution wentalmost unknown during his life-time, and his semiotic approach wasspread by Charles Morris who, draw-ing largely on the Peircean theoreticalframe work, approached semioticsthrough the lenses of Mead’s behav-

209

SITUATIONAL SEMANTICS

called ‘situations’ and can be indi -viduated by cognitive agents. Thus,people perceive situations, cause themto be brought about, and have all sortsof attitudes toward them. One factremains: we are at all times in situa-tions (cf. Norbert Hornstein: ‘Situa-tions people the world. They are datedand located.’).

While the Barwise-Perry volume(1983) is exceptional in its pro -grammatic employment of situations(applied, among others, to naked-infinitive perception and beliefreports), historically there was alwayssome interest in situations. Two note-worthy – albeit cryptic – passages inZettel (Wittgenstein 1981: 2, 13)show that Wittgenstein thought thatsituations a person is embedded in areof key value in making their behav-iour intelligible. Authorities of prag-matics like J. L. Austin, H. P. Griceand Peter Strawson could be regardedas friendly to a situational approach,for they try to come to terms withthe notion of ‘context’. And for some,situations are generalised versionsof ‘events’ as conceived by DonaldDavidson and others.

A situation is a rich object consist-ing of individuals enjoying variousproperties and standing in a variety ofrelations. It is a ‘small’ world. Inci-dentally, there is a crucial differencebetween situation-theoretic and mathematical relations. The latter areset-theoretic constructs whereas theformer are relations of the kind recog-nisable by cognitive agents. A situa-tion may extend quite far in space andtime. An agent can watch a film abouta past assassination, scrutinise thelatest videos from the Jupiter mission,or chat with someone who relates

Eco, Umberto (1984). Semiotics and thePhilosophy of Language. Bloomington:Indiana University Press.

Locke, John (1689/1963). ‘Of the divisionof the sciences’. Book IV, chapter XXI.An Essay Concerning Human Under-standing. Ed. Peter N. Nidditch.Oxford: Clarendon.

Annalisa Baicchi

SITUATIONALSEMANTICS

An information-based approach tonatural language semantics. Formu-lated by Jon Barwise and John Perry intheir influential book Situations andAttitudes (1983), it is built upon thenotion of a ‘situation’ – a limited partof the real world that a cognitive agentcan individuate and has access to. Asituation represents a lump of infor-mation in terms of a collection offacts. It is through the actualist ontol-ogy of situations that the meaning ofnatural language utterances can beelucidated.

See also: Logic; Possible WorldSemanticsKey Thinkers: Austin, J. L.;Davidson, Donald; Frege, Gottlob;Grice, H. P.; Lewis, David;Montague, Richard; Strawson, P.F.; Tarski, Alfred; Wittgenstein,Ludwig

Situational semantics (‘situationsemantics’ in the sequel) starts withthe hypothesis that what is called ‘theworld’ is an inconceivably large total-ity. Limited parts of the world are

210

SITUATIONAL SEMANTICS

intricacy was cited by Gadamer(1975: 268–9) who saw that the veryidea of a situation necessitates that anagent is not located outside of it andhence may be unable to have objectiveepistemic access to it.

Human beings and lower organ-isms display a fundamental ability todiscern similarities between situa-tions. This is accomplished via regu-larities, that is individuals, relations,or locations that endure from one sit-uation to another. Thus, I believe thatsnow makes driving difficult, thatdoctors are available for medicalassistance, that parents care abouttheir offspring, that I will receive apresent on Father’s Day.

Barwise and Perry note that agents‘must constantly adapt to the course ofevents in which they find themselves’(1983: 10). This adaptation takesplace as an upshot of attunement tosimilarities between situations (‘uni-formities’). Thus, a useful uniformityin my life has to do with the milkman.Every morning (a different situation),he brings the milk at about 8 o’clockand leaves it on our doorsteps. By justbeing attuned to this uniformity, I con-tribute to my well-being. Violation ofa uniformity is possible; there is nomilk service on holidays.

Representation of uniformitiesyields ‘types’. Suppose Bob was eatingcookies yesterday and is eating cook-ies now. Both of these situations sharethe same constituent sequence <eats,Bob, cookies>. These events, occur-ring at different times, have the sametype. In the same vein, consider two‘parametric’ infons <embraces, ĝ,Carol, yes> and <embraces, ĝ, ĥ, yes>,where ĝ and ĥ are placeholders forindividuals. Their meaning can be

their adventures in the Pampas ofArgentina.

One of the features of situationsemantics is its information-based dis-position. Let us define something’sbeing P (a property) or something’shaving R (a relation) to something elseas a ‘state of affairs’ (Armstrong1997). In situation semantics, ‘infons’are posited as discrete items supplyingsuch bits of information. An infon isshown as an (n + 2)-tuple <R, a1, . . .,an, p>, where R is an n-place relation(properties being 1-place relations);a1, . . ., an are objects appropriate forthe respective argument places of R;and p is polarity. If p=yes (respectively,no) then a1, . . ., an stand (respectively,do not stand) in the relation R.

Abstract situations are proposed tobe counterparts of real situations inorder to make the latter amenable toformal manipulation. Given a situa-tion s, the set {i | s |= i}, where i standsfor an infon, is the correspondingabstract situation. Notice that this setcollects all facts (infons that are madetrue by s). Alternatively, s is said to‘support’ (make it the case that) i –denoted as s |= i above – just in case iis true of s.

Devlin (1991: 31) has studied whatsituations might amount to and howwe can ‘individuate’ them. A schemeof individuation – a way of carving theworld into uniformities – is an essen-tial facet of the situational approach.This way we can single out – say, viadirect perception or thinking – andtreat situations as entities that canlater be referred to. When agents indi-viduate a situation, they cannot beexpected to give clear-cut descriptionsof all that the situation comprises: sit-uations are vague objects. Another

211

SITUATIONAL SEMANTICS

For instance, an utterance of ‘I amsmiling’ defines a meaning relation.Given d, c, and e, this relation holdsjust in case there is a location l and aspeaker s such that s is speaking at l,and, in e, s is smiling at l. In interpret-ing the utterance of an expression f incontext, there is a flow of information,partly from the linguistic formencoded in f and partly from contex-tual factors provided by the utterancesituation u. These are combined toform a set of constraints on thedescribed situation e.

Ideas from situation semanticshave been applied to a number ofissues in logic*, language, cognitionand information. To take three comprehensive projects, Barwise andEtchemendy (1987) analyse self- reference and paradox, Gawronand Peters (1990) deal with pronom-inal anaphora, and Cooper (1996)focuses on generalised quantifiers.Unlike the classical approaches tomeaning (including Fregean senses,Tarskian truth, Montague grammar),there is an ordinary feel to situationsemantics; it does not impose human-made assumptions in our conceptualscheme (in contra-distinction toLewisian possible worlds, for exam-ple). It is an archetype of what a nat-uralised theory of semantics shouldlook like.

Primary sourcesBarwise, Jon (1989). The Situation in

Logic. Stanford, CA: CSLI Publications.Barwise, Jon and John Etchemendy

(1987). The Liar. New York: OxfordUniversity Press.

Barwise, Jon and John Perry (1983). Situ-ations and Attitudes. Cambridge, MA:MIT Press.

rendered as ‘Someone embraces Carol’and ‘Someone embraces someone’,respectively. Anchoring parametersof an infon yields (parameter-free)infons. For example, given <embraces,ĝ, Carol, yes>, if F(ĝ) = David (F is ananchoring) then we obtain <embraces,David, Carol, yes>.

Networks of abstract links betweensituation types provide informationflow (Dretske 1981). Thus, the state-ment ‘smoke means fire’ expresses thelaw-like relation that links situationswhere there is smoke to situationswhere there is a blaze. If a is the typeof smoky situations and b is the typeof fire situations, then having beenattuned to the constraint a » b (read ‘ainvolves b’) an agent can pick up theinformation that there is a fire in aparticular site by observing that thereis smoke.

According to situation semantics,meanings of expressions reside in sys-tematic relations between differenttypes of situations. They can be iden-tified with relations on discourse situ-ations d, connections c, the utterancesituation u itself, and the described sit-uation e. Some public facts about u –such as its speaker and time of utter-ance – are determined by d. The ties ofthe mental states of the speaker andthe hearer with the world constitute c.A discourse situation d involves theexpression uttered, its speaker, spa-tiotemporal location of the utterance,and the addressee. Each of thesedefines a linguistic role (role of thespeaker, of the addressee, and so on).The utterance situation u constrainsthe world in a certain way, dependingon how the roles for discourse situa-tions, connections and described situ-ation are to be filled.

212

SPEECH ACT THEORY

Gadamer, Hans-Georg (1975). Truth andMethod. Trans. and ed. Garrett Bardenand John Cumming. New York:Seabury Press.

Gawron, Jean Mark and Stanley Peters(1990). Anaphora and Quantificationin Situation Semantics. Stanford, CA:CSLI Publications.

Grice, H. P. (1989). Studies in the Way ofWords. Cambridge, MA: Harvard Uni-versity Press.

Strawson, P. F. (1997). Entity and Identity.Oxford: Oxford University Press.

Wittgenstein, Ludwig (1981). Zettel. Ed.G. E. M. Anscombe and G. H. vonWright. Trans. G. E. M. Anscombe.Oxford: Basil Blackwell.

Varol Akman

SPEECH ACTTHEORY

Speech act theory accounts for an actthat a speaker performs when pro-nouncing an utterance, which thusserves a function in communication.Since speech acts are the tools thatallow us to interact in real-life situa-tions, uttering a speech act requiresknowledge not only of the languagebut also of its appropriate use withina given culture.

See also: Logical Positivism;Ordinary Language Philosophy;PerformativeKey Thinkers: Aristotle; Austin, J.L; Ayer, A. J.; Grice, H. P.; Husserl,Edmund; Kant, Immanuel; Ryle,Gilbert; Searle, John; Wittgenstein,Ludwig

Devlin, Keith (1991). Logic and Informa-tion. New York: Cambridge UniversityPress.

Devlin, Keith (2006). ‘Situation theoryand situation semantics’. In DovGabbay and John Woods (eds), Hand-book of the History of Logic, vol. 7.Amsterdam: Elsevier. 601–64.

Seligman, Jerry and Larry Moss (1997).‘Situation theory’. In van Benthem,Johan and Alice ter Meulen (eds),Handbook of Logic and Language.Amsterdam: Elsevier. 239–309.

Further readingAczel, Peter, David Israel, Yasuhiro Kata-

giri, and Stanley Peters (eds) (1993). Sit-uation Theory and Its Applications, vol.3. Stanford, CA: CSLI Publications.

Armstrong, D. M. (1997). A World ofStates of Affairs. Cambridge: Cam-bridge University Press.

Austin, J. L. (1979). Philosophical Papers.Ed. J. O. Urmson and G. J. Warnock.Oxford: Oxford University Press.

Barwise, Jon, Jean Mark Gawron, GordonPlotkin, and Syun Tutiya (eds) (1991).Situation Theory and Its Applications,vol. 2. Stanford, CA: CSLI Publications.

Cooper, Robin (1996). ‘The role of situa-tions in generalized quantifiers’. InShalom Lappin (ed.), The Handbookof Contemporary Semantic Theory. Cambridge, MA: Blackwell Publishers.65–86.

Cooper, Robin, Kuniaki Mukai, and JohnPerry (eds) (1990). Situation Theoryand Its Applications, vol. 1. Stanford,CA: CSLI Publications.

Davidson, Donald (1980). Essays onActions and Events. Oxford: ClarendonPress.

Dretske, Fred (1981). Knowledge and theFlow of Information. Cambridge, MA:MIT Press.

213

SPEECH ACT THEORY

Although he explicitly deems thenature of sentences to be uninterestingin his inquiry on apophantic logos,Aristotle represents the first accountof language as action.

Aristotle’s standpoint influencedthe study of language for centuriesand paved the way for a tradition ofresearch on verifiability, but severalGerman and British philosophersanticipated a view of language as atool to change a state of affairs. Theissues of language and conversationwere addressed by Immanuel Kantwho anticipated some concepts like‘context’ and ‘subjective idealisation’,the rules that articulate conversation,and the para-linguistic gestures usedin the accomplishment of speech acts.But it was only at the end of the nine-teenth century that a more elaboratetreatment of language as action wasinitiated.

The first, although non-systematic,study of the action-like character oflanguage was conducted by ThomasReid, who described different actsthat can be performed through lan-guage, and grouped them into twocategories: ‘solitary acts’ like judge-ments, intentions, deliberations anddesiring, which can go unexpressed;and ‘social operations’ like command-ing, promising or warning, which, bytheir very social nature, must beexpressed. Reid’s contribution to theinception of a speech act theory can befully understood if viewed from thewider perspective of the philosophicaldevelopments of his time.

Franz Brentano’s distinctionbetween physical and psychologicalphenomena is particularly relevant inthis respect because it reintroduced tophilosophy the scholastic concept of

Speech act theory was first developedby J. L. Austin whose seminal OxfordLectures in 1952–4 marked an impor-tant development in the philosophy oflanguage and linguistics. Austin’s pro-posal can be viewed as a reaction tothe extreme claims of logical posi-tivists, who argued that the meaningof a sentence is reducible to its verifiability, that is to an analysiswhich verifies if utterances are true orfalse. Austin contended that most ofour utterances do more than simplymaking statements: questions andorders are not used to state something,and many declarative sentences do notlend themselves to being analysed interms of their falsifiability. Instead,they are instruments that allow speak-ers to change the state of affairs. Thisis tantamount to saying that we uselanguage mainly as a tool to do things,and we do so by means of performinghundreds of ordinary verbal actions ofdifferent types in daily life, such asmake telephone calls, baptise children,or fire an employee.

The fact that not all sentences are amatter of truth verifiability was firstadvanced by Aristotle who, in his DeInterpretatione, argued that:

as there are in the mind thoughts whichdo not involve truth or falsity, and alsothose which must be either true or false,so it is in speech. [. . .] A sentence is a sig-nificant portion of speech [. . .] Yet everysentence is not a proposition; only suchare propositions as have in them eithertruth or falsity. [. . .] Let us therefore dis-miss all other types of sentence but theproposition, for this last concerns ourpresent inquiry, whereas the investiga-tion of the others belongs rather to thestudy of rhetoric or of poetry. (1–4)

214

SPEECH ACT THEORY

Marty offered the first account of usesof language meant to direct others’behaviour, like giving an order,requesting, or giving encouragement.Marty stated that sentences may hintat the speaker’s psychic processes andargued that ‘deliberate speaking is aspecial kind of acting, whose propergoal is to call forth certain psychicphenomena in other beings’ (1908:284). Stemming from Brentano’s tri-partite subdivision of mental phenom-ena into presentation, judgements,and phenomena of love and hate,Marty discriminated linguistic formsinto names, statements and emotives(utterances arousing an interest),which is a model that closely resemblesKarl Bühler’s Sprachtheorie. It is pre-cisely to Bühler that we owe thecoinage of the label ‘speech act theory’.He offered the first thorough study ofthe functions of language – Darstel-lung (representation), Kindgabe (inti-mation or expression), and Auslösung(arousal or appeal) – thus endowingnon-representational sentences withtheir own status.

A more complete treatment we findin the work of Adolf Reinach, whooffered the first systematic theory ofspeech acts. Reinach received a doc-torate in philosophy from the Univer-sity of Munich; his dissertation wason the concept of cause in penal law.It was within the context of legal lan-guage that Reinach argued in favourof the relevance of speech acts whichhe referred to, presumably independ-ently of Reid’s work, as ‘socialacts, that is acts of the mind that are performed in the very act of speak-ing’. Reinach (1913) provided adetailed taxonomy of social acts asperformative* utterances and their

‘intentionality’, which allows for adistinction between mental acts andthe external world. As far as speechact theory is concerned, suffice it hereto say that Brentano argued that everymental, psychological act has a con-tent and is directed at an object (theintentional object), which meansthat mental phenomena contain anobject intentionally within themselvesand are thus definable as objectifyingacts. The Brentanian approach tointentionality* allows for a distinc-tion between linguistic expressionsdescribing psychological phenomenaand linguistic expressions describingnon-psychological phenomena. Fur-thermore, Brentano claimed thatspeaking is itself an activity throughwhich we can initiate psychic phe-nomena. Edmund Husserl picked upthe importance of what Brentano’spsychological investigation couldbring to logic*, in particular the con-trast between emotional acts andobjectifying acts. Husserl tackled theissue of human mental activities(‘acts’) and how they constitute the‘object’ of knowledge through experi-ence. In his Logical Investigations(1900/1) he developed a theory ofmeaning based on ‘intentionality’which, for him, meant that conscious-ness entails ‘directedness’ towards anobject. It is on the notion of ‘objecti-fying acts’, that is acts of representa-tion, that Husserl shaped his theory oflinguistic meaning, thus emphasisingthe referential use of language. Collat-erally he treated the non-representa-tional uses of language, that is actslike asking questions, commanding orrequesting.

Following Brentano and movingwithin the field of psychology, Anton

215

SPEECH ACT THEORY

Oxford linguist philosophers likeGilbert Ryle and J. L. Austin, whoinstead were greatly influenced byLudwig Wittgenstein. He claimed thata language consists of a wide multi-plicity of structures and usages thatlogical positivists had neglected toanalyse but which encompass themajority of what human beings say intheir construction of meaning.

Following Wittgenstein’s insightsinto language and putting himselfagainst the positivist background,Gilbert Ryle rejected the Cartesianmind-body dualism in The Concept ofMind (1949), and revived the central-ity of the standard uses of language,thus contributing to the developmentof ‘ordinary language philosophy’* inOxford.

Taking the same veil and influencedby Husserl, Austin rejected theaccount that only sentences that aremeant to describe a state of affairs areworth studying, and he observed thatverifiable sentences are only a smallpart of the large amount of utterancesproduced by language users. Not allutterances express propositions:many perform actions as, for exam-ple, greetings or orders, which resista truth-conditional analysis. Indeed,most of the sentences uttered byspeakers are used in such a way as toperform more fundamental things inverbal interactions, such as naming aship, marrying a couple, or making arequest. In daily life we perform manyordinary verbal actions, and utter-ances are used in speech events toaccomplish all that is achievedthrough language. Austin’s speech acttheory was first delineated in the noteshe prepared for some lectures interest-ingly entitled Words and Deeds which

modification, and stated very clearlythat the utterance (Äusserung) of asocial act is different from the innerexperience of emotions like anger orshame and from statements (Konsta-tierungen) about experiences. It is pre-cisely the recourse to the physicalmedium, the Äusserung, that trans-forms the philosophical category ofaction into a social act. Drawing onprevious literature, Reinach separatedactions from internal experiences.Then he discriminated between exter-nal actions like kissing or killing andlinguistic actions, and within this classhe distinguished between social acts,which are performed in every act ofspeaking, and actions, where signs areused but no speech act is performedsuch as in ‘solitary asserting’ and emo-tive uses of language. The final dis-tinction refers to the linguistic actionsperformed in uttering performativeformulae and the linguistic and non-linguistic actions whose performancehas an effect on the state of affairs andeven changes it.

While Reinach’s ideas were spread-ing through the Munich scholars, atOxford A. J. Ayer, considered thephilosophical successor of BertrandRussell, deemed philosophically inter-esting only those sentences that can besubject to the truth-condition analy-sis. In line with the logical positivism*of the Vienna Circle, Ayer developedthe verification principle in Language,Truth and Logic (1936) where hestated that a sentence is meaningfulonly if it has verifiable import. Sen-tences expressing judgements, evalua-tion and the like were not to be objectsof scientific inquiry. This stance,which is now known as the ‘descrip-tive fallacy’, led him into conflict with

216

SPEECH ACT THEORY

act to be successful, it must fulfil someappropriateness conditions, or ‘felic-ity’ conditions: locution is successful ifwords and sounds are correctly pro-duced; illocution is appropriate if itmeets the conditions for its realisation;perlocution may be effective when itproduces consequences desired bythe producer. The notion of illocution-ary force embodies the philosophicalnotion of intentionality, which canbe expressed by performing a speechact through three modalities: (1)directly or indirectly through the per-formance of another speech act (‘Passme the salt’ versus ‘Can you pass methe salt?’); literally or non-literallydepending on the way words are used(‘Stick it in your head’); (3) explicitlyor inexplicitly when meaning is spelledout fully or incompletely (‘I’ll be backlater, Mary’s ready’). Indirectness andnonliterality are disambiguated byway of a conversational implicature*,whereas explicitation is achievedthrough expansion or completion ofwhat one says.

John Searle, one of Austin’s stu-dents, contributed widely to develop-ing speech act theory, which headdressed from the viewpoint ofintentionality. Specifically he con-ceived of linguistic intentionality asderived from mental intentionality. Inhis Speech Acts (1969) Searle claimedthat Austin’s ‘felicity conditions’ areconstitutive rules of speech acts to theextent that to perform a speech actmeans to meet the conventional ruleswhich constitute a specific speech act.Moving from this approach andanalysing the act of promising, Searleproposed a classification of speechacts into four categories: (1) proposi-tional content (what the speech act is

he delivered at Oxford Universityfrom 1952 to 1954. Such notes con-stituted the basis on which he devel-oped his Harvard lectures in 1955,posthumously published in 1962. Inthe first phase of development of histheory, Austin retained the Aris-totelian distinction between apophan-tic and non-apophantic logos, andintroduced the terms of constativeutterances and performative utter-ances, where the former describe orconstate a state of affairs and thelatter perform actions. Austin laterrealised that a clear distinctionbetween the two types of utterances isunsustainable. If, for example, we say‘There is a rat under your chair’, wedo more than assert a state of affairs:we warn someone about a possibledanger. Assertions can thus be used toperform such acts as to warn, to apol-ogise, and many more. Austin thenabandoned the dichotomy and con-tended that to say something equals toperform something.

According to Austin, when we saysomething, we perform three actssimultaneously: a locutionary act, anillocutionary act, and a perlocutionaryact. At the locutionary level, a speakerproduces sounds (phonetic act) whichare well ordered with respect to thephonological system and grammar ofa particular language (phatic act), andcarry some sense with respect to thesemantic and pragmatic rules of thatlanguage (rhetic act). At the illocution-ary level, he is expressing his intentionby virtue of conventions shared in hisspeech community. At the perlocution-ary level, he performs a third act whichincludes the consequences of hisspeaking, and he has only limited con-trol over them. In order for the speech

217

SPEECH ACT THEORY

what they call ‘illocutionary scenar-ios’. They are formed by a before, acore, and an after component. If aperson wants someone to bring himhis pen, he can utter a direct speechact like ‘Bring me my pen’, whichexploits the core component, or hecan make his request indirectlyexploiting either the before compo-nent (‘Can you bring me my pen?’)where the modal verb ‘can’ points tothe hearer’s ability to perform theaction, or the after component (‘Youwill bring me my pen, won’t you?’)where the auxiliary ‘will’ instantiatesthe after component of the requestscenario. Panther (2005) makes thepoint that metonymies provide natu-ral ‘inference schemas’ which areconstantly used by speakers in meaning construction and interpreta-tion. Scenarios may be accessedmetonymically by invoking relevantparts of them. Indirect requests like‘Can you open the door?’, ‘Will youclose the window?’, ‘Do you have hotchocolate?’ exploit all pre-conditionsfor the performance of a request, thatis, the ability and willingness of thehearer, and his possession of therequired object. Such pre-conditionsare used to stand for the wholespeech act category. By means of theexplicit mention of one of the com-ponents of the scenario, it is possiblefor the speaker to afford access to thehearer to the whole illocutionary cat-egory of ‘requesting’ in such a waythat the utterance is effortlessly inter-preted as a request. With a view toimproving Panther’s proposal, Fran-cisco Ruiz de Mendoza (2007) con-tends that illocutionary meaning isdirectly tied to the notion of IdealisedCognitive Models (ICMs), which are

about); (2) preparatory condition,which states the prerequisites for thespeech act; (3) sincerity condition (thespeaker has to sincerely intend to keepa promise); and (4) essential condition(the speaker’s intention that the utter-ance counts as an act and as such is tobe recognised by the hearer). One ofSearle’s major contributions to thetheory refers to indirectness, that isthe mismatch between an utteranceand an illocutionary force.

The interpretation of indirectspeech acts has drawn a great deal ofattention. Drawing on H. P. Grice’spragmatics, most scholars assumethat some inferential work on the partof the hearer is required in order toidentify the speaker’s communicativeintention and the core question is howsuch inference can be computed.Searle (1975) assumes that the hearerrecognises both a direct-literal force,which he understands as the second-ary force, and an indirect-nonliteralforce, which is the primary force. Sim-ilarly Dan Gordon and George Lakoff(1975) argue that inference rules thatthey label ‘conversational postulates’reduce the amount of inferential com-puting necessary to disambiguate anindirect speech act. Jerrold Sadock(1974) departs from the inferentialhypothesis and proposes ‘the idiommodel’ by claiming that a speech actlike ‘Can you pass me the salt?’ ispromptly interpreted as a request andneeds no inference.

Speech act theory is now receivinggreat attention and valid theoreticalproposals from cognitive linguists.Klaus Panther and Linda Thornburg(1998) claim that our knowledge ofillocutionary meaning may be sys-tematically organised in the form of

218

SPEECH ACT THEORY

Panther, K. U. and L. Thornburg (1998).‘A cognitive approach to inferencing inconversation’. Journal of Pragmatics30: 755–69.

Panther K. U. (2005). ‘The role of concep-tual metonymy in meaning construction’.In F. Ruiz de Mendoza and S.Peña (eds),Cognitive Linguistics. Internal Dynam-ics and Interdisciplinary Interaction.Berlin: Mouton de Gruyter. 353–86.

Reinach, A. (1913). ‘Die apriorischenGrundlagen des bürgerlichen Rechtes’.In Jahrbuch für Philosophie undphänomenologische Forschung 1:685–847.

Ruiz de Mendoza, F. (2007). ‘High levelcognitive models: in search of a unifiedframework for inferential and gram-matical behavior’. In Krzysztof Kosecki(ed.), Perspectives on Metonymy.Frankfurt: Peter Lang. 1130.

Ryle G. (1949). The Concept of Mind.London: Hutchinson.

Sadock J. (1974). Toward a LinguisticTheory of Speech Acts. New York: Aca-demic Press.

Searle J. R. (1969). Speech Acts. Cam-bridge: Cambridge University Press.

Searle J. R. (1975). ‘Indirect speech acts’.In P. Cole and J. L. Morgan (eds),Syntax and Semantics 3: Speech Acts.New York: Academic Press. 59–82.

Wittgenstein L. (1953). PhilosophicalInvestigations. Oxford: Blackwell.

Further readingAyer, A. J. (1936). Language, Truth and

Logic. London: Gollancz.Brentano, F. (1874). Psychologie vom

empirischen Standpunkt. Leipzig:Duncke and Humbolt.

Marty, A. (1908). Untersuchungen zurGrundlegung der allgemeinen Gram-matik und Sprachphilosophie. Halle:Nyemeier.

principle-governed cognitive struc-tures. Illocutionary scenarios repre-sent the way in which language usersconstruct interactional meaning rep-resentations abstracted away from anumber of stereotypical illocutionarysituations. In an indirect request like‘I fancy going out for dinner’ thehearer understands the implicatedmeaning by relying on high-level sit-uational ICMs – that is, on thegeneric knowledge that expressing awish indirectly corresponds to askingfor its fulfillment. Thus, it is exactlythe quick and easy retrieval from ourlong-term memory of a stored illocu-tionary scenario that allows us toidentify the nature of indirectness.

Speech act theory is a thought- provoking issue which has attractedthe interest of philosophers of lan-guage and linguists from diverse theo-retical persuasions. Manifold aspectsof the theory are being debated suchas the classification of speech acts, therelationship between speech acts andculture, and the acquisition of speechacts by children, which proves howthis area of language research stillprovides room for developments andnew insights.

Primary sourcesAristotle (1941). De Interpretatione. New

York: Random House. 38–61.Austin, J. L. (1962). How to Do Things

with Words. Oxford: Oxford UniversityPress.

Gordon D. and G. Lakoff (1975). ‘Con-versational postulates’. In P. Cole and J.L. Morgan (eds), Syntax and Semantics,Speech Acts. New York: AcademicPress. 83–106.

Husserl, E. (1900/1). Logische Unter-suchungen. Halle: Nyemeier.

219

STRUCTURALISM

Jakobson, Roman; Sapir, Edward;Saussure, Ferdinand de; Whorf,Benjamin Lee

The term structure is derived fromLatin structura (from struere, ‘tobuild’) and just as human beings buildhouses, so structuralism contends thathuman existence – the physiologicaland mental set-up of individuals andtheir social life – is also built fromstructures in a way that more or lessgoverns what people are able to thinkand do. In addition to the limitationsof the laws of physics and rules ofsocial behaviour, structuralism alsomaintains that less overt structuresrestrict psychological and behaviouralalternatives by controlling individu-als’ preferences. In linguistics struc-turalism is affiliated with the so-calledSaussurean ideas about language andother sign-systems (dealt with in‘semiology’, later ‘semiotics’) and itmay be most easily understood in theconceptual framework that is attrib-uted to Saussure.

As opposed to the use of language(French parole), there is a system oflanguages and language (langue) thatis a set of inherent relations that builda structure. In order to arrive at anexhaustive and consistent descriptionof this system, one has to assert thatthe description is historically specific:characteristic of an abstraction fromlanguage use at a certain time andplace. In other words, it produceswhat was labelled a synchronicdescription of the system. This isopposed to the view prevalent in thenineteenth century that the historyand the genealogy of languages wasthe only (legitimate) theme in the lan-guage sciences; the results yielded in

Reid, T. (1894). The Works of ThomasReid. Edinburgh: Maclachlan and Stewart.

Annalisa Baicchi

STRUCTURALISM

A theoretical and methodologicalapproach in linguistics and otherhuman (including social) sciences thatattempts to gain insights into its sub-ject matter by assuming that every-thing to do with human beings is builtof more or less autonomous systems asrelations of oppositions. These oppo-sitions may be of different types but ingeneral are binary relations. The Swisslinguist Ferdinand de Saussure is oftensaid to have initiated a structuralistmovement, school or intellectualworld view, rather than developinga coherent theory, and in linguisticsa distinction is traditionally madebetween (Saussurean) European struc-turalism and American structuralism,the main figure of which is the linguistLeonard Bloomfield. Structuralism ina broad sense has mainly been appliedin anthropology, especially by theFrench anthropologist Claude Lévi-Strauss and other French thinkers, andin literary studies.

See also: Distinctive Features;Glossematics; Langue/Parole;Poststructuralism; Sense/Reference;Signs and Semiotics;Transformational-GenerativeGrammarKey Thinkers: Barthes, Roland;Boas, Franz; Bloomfield, Leonard;Derrida, Jacques; Hjelmslev, Louis;

220

STRUCTURALISM

system. Phonology is the study ofphonemes*: that is, abstract entitiesidentified as generalisations of speechsounds and building a mental systemof oppositions that makes the individ-ual language user able to decode thesound chains as words in a language.The enterprise of setting up suchphonological systems is often a fairlystraightforward project, but theanalysis of word meanings is a morechallenging task. Some approaches insemantics can be regarded as con-ceived within the framework of struc-turalism, for instance the notions ofsemantic fields and semantic compo-nents describing the features creatingthe basis of semantic oppositions.

The methods of structuralism makeno consistent conceptual framework.Phonology investigates the sound systems of a language or a dialect inorder to find the distinctive sound fea-tures that separate words. But somelinguists take different methodologi-cal approaches. For example, LouisHjelmslev, the inventor of glossemat-ics*, started out by dividing a text intotwo and he continued like this until heended up with the phonemes. One ofthe basic problems of structuralism iswhat answers are given to the questionabout the nature of meaning. There isobviously a difference between wordmeaning and the meaning of life, eventhough appeal to word meanings asoppositional relations may not beunequivocally convincing and eventhough the question of the meaning ofgeneral, abstract and fictional expres-sions may not be straightforwardlyanswered. This vagueness in whatmeaning is may be the background ofthe diffusion of the concept of struc-turalism into other academic fields

this tradition was labelled by struc-turalists as diachronic description. Itis a general experience that words areput together in chains, and a basicnotion in structural linguistics is thatof syntagm. But since this is a matterof language use, the correspondingnotion of paradigm is often moreinteresting for structuralists. A para-digm is not a discernible and evidententity like a syntagm because we canonly identify a paradigm by abstrac-tion and experiment: by playing withthe words of a sentence by substitut-ing them. In that way we learn thatparadigms can be said to be sets ofwords that can replace each other oncertain positions in chains of words,words that accordingly must be dif-ferent in one respect and similar insome other respect. This potential ofwords is said by structuralists to be amatter of how their properties arestructural and can be described assuch.

The basic opposition, then, is thatbetween the form of a word, the signi-fier (signifiant), and its meaning, thesignified (signifié). According to Saus-sure, this relation is arbitrary, a claimthat has caused some controversy. Butone might say that the relation is, inprinciple, arbitrary while it is, evi-dently, not historically arbitrary sincewe use words in – almost – the sameway as we have experienced themwhen acquiring our mother tongue.From this point of departure, bothform and meaning can be subject tostructural scrutiny. The extremes ofsuch scrutiny are, on the one hand,phonology and, on the other, semanticorderings of the sense and meaningsof words and their mutual relationswithin the whole of the language

221

STRUCTURALISM

It may seem ironical that Jakobson,a linguist and passionate reader ofRussian poetry, with a backgroundaffiliation with Russian formalism,was to become the main inspiringforce of European structuralism, thePrague School (which was, withthe Geneva School, the centre of European structuralism), and ofAmerican structuralism. Passingthrough Europe and ending up in theUnited States, he and other Russianscontributed significantly to what hasbecome known as the Prague Schoolof structuralism. Fleeing from WorldWar Two, he visited Denmark andinfluenced the glossematic version ofstructuralism, and in the United Stateshe inspired anthropology and othersciences, at the same time more or lessdirectly influencing Chomsky’s ideasabout linguistic universals.

European structuralism was notonly a countermovement with respectto positivism but after World War Oneit was also an opportunity to find analternative to the dominance of thenineteenth-century German neogram-marians and their ‘sound laws’. Mostof the characteristics of structuralismmentioned above apply to Europeanstructuralism, and a number of otherfeatures can be mentioned. Structural-ism deals with systems (a word thatcan, in this context, be taken as a syn-onym for structure; structuralists whotalk about ‘structured systems’ use apleonastic term), and knowledgeabout these systems (the entities ofwhich are considered at least as real asobservable entities) are arrived atthrough abstraction and analysis. Thesystems are also regarded as social.Even though Saussure’s basic assump-tion about the linguistic sign was that

such as anthropology and literature,and its popularity in semiotics.

Although it is often stated as fact,the attribution of the idea of struc-turalism in linguistics to Saussure maynot be totally justified. The source ofwhat we know about his thoughtsis Cours de Linguistique Générale(Course in General Linguistics) from1916, a work that is based on the lec-ture notes of some of his students. Fur-thermore, Saussure may be consideredonly part of an emerging movement inlinguistics that could not reconcileitself to nineteenth-century positivism.Finally, the term structuralism hadbeen used by psychologists in the nine-teenth century but it was not used bylinguists before the end of the 1920s;Roman Jakobson probably was thefirst to offer a definition of the theoret-ical concept. The expressions ‘struc-ture’ and ‘structural’ in discussinglinguistic phenomena were also used inthe first half of twentieth century.

Nevertheless Saussure is usuallyconsidered the founding father of lin-guistic structuralism, and his thoughtsas they are presented in Cours havehad a considerable impact on generalstructuralist thinking; structuralismreached its peak as an -ism between1930 and 1960. The broad movementof structuralism was not a unifiedendeavour but rather a patchwork ofdifferent groupings with differentgoals, different basic assumptions anddifferent kinds of subject matter. Inretrospect, European structuralismcan be said to encompass some moreimportant groups of people and somemore peripheral groupings and indi-viduals, while one person in particularis a travelling herald of the structural-ist message: Roman Jakobson.

222

STRUCTURALISM

conceiving of them as ‘grammars’intrinsic to the human mind. WhereasLévi-Strauss studied the built-in mean-ing of tribal behaviour, other Frenchstructuralists like Roland Barthes andJacques Derrida have applied struc-turalist thought to literature; theFrench intellectual Jacques Lacan didthe same to psychoanalysis and theFrench philosopher Louis Althusserfounded a ‘Marxist structuralism’. InEurope, therefore, structuralism as abroad notion is nowadays mostly asso-ciated with French structuralism, andwhile the structural mind in twenty-first century Europe is most likely nota member of a club named ‘structural-ism’, structuralist beliefs are almostcommon sense in many theoretical,often implicit, basic assumptions.

In contrast to the European struc-turalistic vein – which was linguisticin its outset and all the time concernedat some level with linguistic meaning –American structuralism tended toignore meaning and focus on linguis-tic form, while also in general main-taining so-called linguistic relativity*.Both attitudes can be explained interms of the background of Americanlinguistics. The United States wasbuilt on a mixture of immigrants andtheir descendants whose cultures hadlong philological traditions, whilealso comprising the ethnic NativeAmerican cultures, the anthropologi-cal and linguistic documentation ofwhich was an immense task. It maytherefore seem natural that describingeach language on its own terms wouldbe a reasonable goal, and this view, incombination with the low prioritygiven to semantics, logically gave riseto the fairly extreme idea that the waypeople think depends on the structure

it is a entité psychique (Saussure 1916:99), or psychological entity, in generalstructuralists maintain that they talkabout social phenomena, albeit stud-ied through analysis of what peoplesay. Indeed this may be a necessarytheoretical prerequisite if one wishesto avoid philosophical inquiry into theproblem of what is private, what ispublic, and what role language playsin making the private public.

European structuralism has hadwide-ranging consequences for thesometimes not particularly explicitbasic assumptions of all disciplines ofmodern European linguistics. This isespecially true in phonology, whereany textbook on pronunciation andspeech sounds takes up the phonemesystem of the language in question.This involves ascertaining the distinc-tive, contrastive features of minimalpairs (of words) in order to identifysingle phonemes, and, following thisprocedure, the objective of the processis to find all the phonemes of a singlelanguage. This modus operandi isnow almost a standard method inphonology, whether phonologists per-ceive themselves as structuralists ornot.

Phonology is a fairly technical disci-pline in linguistics, but structuralistideas have not confined themselves tointricate linguistic phenomena: theyhave also been adopted by scholars infields outside linguistics. In anthropol-ogy, one of the main figures is ClaudeLévi-Strauss, who met Jakobson inNew York, and who attempted toshow that the myths and rituals oftribal cultures work as regulating kin-ship systems and other social institu-tions. He also analysed them in linewith structural linguistic analysis

223

STRUCTURALISM

that goes back to Boas and the (basi-cally pragmatic) idea that humanbeings communicate in sentences, notby using words in isolation. The otherthing is the fact that, contrary to gen-eral opinion, Bloomfield did nottotally abandon linguistic semanticsfrom his linguistic approach. Forinstance, he appeals to a special kindof meaning:

Since our study ordinarily concernsonly the distinctive features of form andmeaning, I shall henceforth usuallyomit the qualification linguistic or dis-tinctive, and speak simply of forms andmeanings, ignoring the existence ofnon-distinctive features. A form is oftensaid to express its meaning. (Bloomfield1935: 141)

This is also by necessity consistentwith his hierarchy of categories whichwould otherwise have been impracti-cal because the notions ‘meaningful’and ‘meaningless’ are necessary theo-retical concepts for the distinctionsbetween the classes of linguistic units,for instance in the expression the‘smallest meaningful unit . . .’.

But some of Bloomfield’s followersin American structuralism took it fur-ther away from semantics and linguis-tic meaning, one of them being ZelligHarris, whose principal work actuallyhas the word ‘structural’ in its title.And these two themes, the meaningsof the words ‘structuralism, structuraland structure’, and the question ofwhether semantics can be disposed of,are being transferred to the modernworld’s most famous linguist, NoamChomsky. It may seem ironical thatChomsky’s programmatic work, Syn-tactic Structures (1957), dealt with

of their language, the so-called Sapir-Whorf hypothesis.

Franz Boas is regarded as thefounder of American structuralism.An anthropologist untrained in lin-guistics, he contributed notably tothe description of the phonologicaland grammatical structures ofNative American languages. Boasincremented the prestige of languagestudies as a part of anthropologicalstudies, which traditionally includedarchaeology and cultural and physicalanthropology. But he was also anadvocate of relativism, a view that hepassed on to his students and follow-ers, and, in line with this, one of hismain concerns was to promote thebasic methods of fieldwork. If lan-guages are more or less self-containedentities, it sounds reasonable to callfor careful and detailed investigationinto each particular language in orderto offer exhaustive descriptions. Andthis may be one of the only reasons forthe predicate structuralism to thisstage of American structuralism: whatwas accounted for through linguisticfieldwork were the internal structuresof particular languages.

The same characteristic may beattributed to the most prominentAmerican structuralist, LeonardBloomfield. In his principal theoreticalwork, Language (1935), the words‘structure’ and ‘structural’ are not fre-quently found. The book offers moreof a methodological account of cate-gories and their hierarchies than anall-encompassing and consistenttheory, but there are two features thatare worth mentioning. In oppositionto European structuralism, Bloomfieldand American structuralism empha-sise sentences as linguistic units, a bias

224

STRUCTURALISM

phenomena they study. Saussure andhis followers see language as words,primarily their phoneme structure andtheir semantics, while Chomsky seeslanguage as phrase structures.

One of the distinctions set up bysome structuralists in linguistics isthat between form and substance,meaning that, for instance, the physi-cal nature of speech sounds studied inphonetics is interesting but secondaryto the important effort of finding andmodelling the phonemes – as mentaland material entities – as the form oflanguage systems. This illustrates afundamental problem with structural-ism, that some of its theoreticalnotions are extremely general, almostuniversal. Likewise, there is a certainkind of vagueness in the terms ‘form’and ‘structure’. From their classicalorigin they are the heirs of words thatdid not have different meanings.‘Form’ and ‘structure’ meant almostthe same, and in some contexts theystill do. In formal logic, for instance,the raison d’être of arguments is thatthey have a certain structure, while inother contexts, for instance in archi-tecture, it may be reasonable to distin-guish between the form (shape andsurface) of a building and its innerstructure. So basically, the proposal ofstructuralism seems to be no morethan the idea that things in the worldare ordered in ways that make themmore than silhouettes in a nebulouslandscape. As such, structuralism haspervaded modern Western civilisa-tion.

Primary sourcesBenveniste, Emile (1939). ‘Nature du

signe linguistique’. Acta LinguisticaVol. I, fasc. 1: 23–9.

formal (mathematical in a broadsense) descriptions of phrase struc-tures, accordingly using the term‘structure’ while explicitly abandon-ing the idea of meaning having anyrole to play in this formal approach tolinguistics. The position is repeated inNew Horizons in the Study of Lan-guage and Mind:

As soon as the first attempts were madeto provide actual descriptions of lan-guages forty years ago, it was discov-ered that the intricacy of structure is farbeyond anything that had been imag-ined, that traditional descriptions ofform and meaning merely skimmed thesurface while structuralist ones werealmost irrelevant. (2000: 122)

The outcome of the structuralist proj-ect will find its place (cf. Chomsky2000: 5) in this the latest theoreticalparadigm, labelled generative gram-mar, in which the term ‘structure’ is abasic and axiomatic one, and in whichthe unearthed structures are assumedto be mental, but have nothing to dowith meaning. On the one hand,Chomsky seems to carry further someof tenets of American structuralismand, on the other, he dissociates him-self from its basic theoretical ideasand methods. In a way he is back atthe starting point of Saussure (maybenot as a historical figure but as theicon of European structuralism), andin another way he is not. He says, likeSaussure, that he studies mental (whatSaussure calls psychological) phenom-ena, and, like Saussure, he does so asabstractions from mental products:words. In this respect they put forward comparable ideas. Wherethey part is the question of which

225

SYSTEMIC-FUNCTIONAL GRAMMAR

Newmeyer, Frederick J. (1998). LanguageForm and Language Function. Cam-bridge, MA: MIT Press.

Hans Götzsche

SYSTEMIC-FUNCTIONALGRAMMAR

An approach to linguistic descriptionwhich aims to provide a comprehensiveaccount of how language is used in con-text for communication. The approachviews language as a resource that is fun-damentally shaped by the uses thatpeople make of it; it therefore aims toexplain the forms of language in termsof the meanings that they express, andto develop a grammar which isdesigned to ‘make it possible to say sen-sible and useful things about any text,spoken or written’ (Halliday 1994: xv).

See also: Corpora; (Critical)Discourse Analysis;Integrationism; Langue/Parole;Metaphor; Modality;Transformational-GenerativeGrammarKey Thinkers: Halliday, M. A. K.;Bernstein, Basil; Firth, J. R.;Hjelmslev, Louis; Malinowsky,Bronislaw; Sinclair, John; Whorf,Benjamin Lee

Systemic-functional grammar (SFG)originated with M. A. K. Halliday,building especially on the ideas of histutor J. R. Firth, in publications fromthe 1960s on, with major contributionsby other scholars such as RuqaiyaHasan and, in more recent years, Jim

Bloomfield, Leonard (US 1933, UK 1935).Language. London: George Allen andUnwin.

Boas, Franz (ed.) (1911, 1922, 1934).Handbook of American Indian Lan-guages, Parts 1, 2, 3. Washington, DC:Smithsonian Institution, Bureau ofAmerican Ethnology.

Bröndal, Viggo (1939). ‘Linguistiquestructurale’. Acta Linguistica Vol. I,fasc. 1: 2–10.

Chomsky, Noam (1957). Syntactic Struc-tures. The Hague. Paris: Mouton.

Chomsky, Noam (2000). New Horizonsin the Study of Language and Mind.Cambridge: Cambridge UniversityPress.

Harris, Zellig (1951). Methods in Struc-tural Linguistics. Chicago: University ofChicago Press.

Jakobson, Roman (2002). Selected Writ-ings Vol. I–VIII. Third edition.Berlin/New York: Mouton de Gruyter,1962–88.

Lévi-Strauss, Claude (1958). Anthropolo-gie structurale. Paris: Plon.

Saussure, Ferdinand de (1916). Cours deLinguistique Generale. Paris: Payot.

Zwirner, Eberhard (1939). ‘Phonologieund Phonetik’. Acta Linguistica Vol. I,fasc. 1: 29–47.

Further readingde Beaugrande, Robert (1991). Linguistic

Theory: The Discourse of FundamentalWorks. New York: Longman.

Huck, Geoffrey J. (1995). Ideology andLinguistic Theory. Noam Chomsky andthe Deep Structure Debates. Londonand New York: Routledge.

Koerner, E. F. K. and R. E. Asher (eds)(1955). Concise Histrory of the Lan-guage Sciences. From the Sumerians tothe Cognitivists. Oxford: Elsevier Sci-ence, Pergamon.

226

SYSTEMIC-FUNCTIONAL GRAMMAR

be represented as a system, labelleddetermination (names of systems aretraditionally written in small capitalsin SFG). The entry condition (the lin-guistic context in which the choicesapply) is ‘nominal group’; the firstchoice (at least in English) is between‘specific’ (‘the [cat]’) and ‘non-specific’(‘a [cat]’). Each option taken opens upa further set of choices until a formalrealisation is reached: for example,selecting ‘specific: personal: interac-tant: addressee’ leads to the deictic(determiner) form ‘your [cat]’, whereasselecting ‘specific: demonstrative:selective: near � plural’ leads to ‘these[cats]’. As this last instance shows,some sets of choices in the system maybe simultaneous: that is, rather thanonly choosing one of two or moremutually exclusive options, thespeaker chooses from two sub-systemsat the same level. Thus, taking the‘selective’ option means choosing bothbetween ‘near’ and ‘far’ and between‘plural’ and ‘non-plural’. Part of thesystem is shown in Figure 7 (three dots indicates where more delicate choiceshave been omitted). Simultaneouschoices are enclosed by a curly bracket;and the formal realisations (in this casespecific words rather than generalstructures) are signalled by downward-slanting arrows. A complete version ofthe system can be found as Figure 6-2in Halliday and Matthiessen (2006:313).

Systems do not operate in isolation:they interact with each other. Forexample, the system of polarity (positive/negative) interacts with anumber of other systems, includingdetermination: here, a combination of‘negative’ with ‘non-specific: total’gives the deictic ‘no’ as in ‘no [cats]’.

Martin and Christian Matthiessenamong many others. From early in itsdevelopment, SFG has had two maindistinguishing features, which arereflected in the name. First, whereasmany approaches focus on the syntag-matic, ‘horizontal’ dimension of howconstituents may be combined withother constituents in a well-formedstructure, SFG prioritises the choicesthat are open to the speaker at any par-ticular point in an utterance – the para-digmatic, ‘vertical’ dimension. Thegrammatical structures are then seen asthe outcome of choices from thoseavailable (the technical term in SFG isthat structures ‘realise’ choices). Sets ofchoices between options can most eco-nomically be shown in the form of sys-tems: for example, ‘if A is the case,there is a choice between B and C; if Bis chosen, there is then a choicebetween D, E and F; but if C is chosen,there is then a choice between G andH’. Systems embody the Saussureanconcept of valeur: a linguistic form hasmeaning by virtue of the other possibleforms that could have been choseninstead. Hence this is a ‘systemic’grammar. Second, the model is ori-ented primarily towards meaningrather than form: that is, its aim is todescribe how wordings are used inexpressing meanings. What a linguisticform consists of is seen as less impor-tant than the function that it performsin the clause: hence, this is a ‘func-tional’ grammar. The following para-graphs expand on these distinguishingfeatures in turn.

The systemic nature of the grammarcan be illustrated with a relativelysimple example. The choices at nomi-nal group level between different kindsof determiner (such as ‘the’ and ‘a’) can

speaker Deictic: ‘my’

interactant speaker-plus Deictic: ‘our’

personal addressee(possession) Deictic: ‘your’

non-interactant …specific

non-selective Deictic: ‘the’

nominaldemonstrative non-plural

Deictic:group

(proximity)‘this’pluralDeictic:‘these’selective

nearDeictic:‘that’farDeictic:‘those’

non-specific …

Figure 7 Part of the system of DETERMINATION in English

227

SYSTEMIC-FUNCTIONAL GRAMMAR

inal group is being determined, orspecified, in terms of possession or oflocation in relation to the speaker.Even traditional labels such as subjectare reinterpreted in SFG as not beingpurely syntactic (‘controlling’ agree-ment with the verb) but as havingmeaning: the subject expresses theentity which is represented as respon-sible for the validity of the propositionexpressed in the clause (that is, theproposition is represented as true for,or applicable to, this entity).

This connects with a second impli-cation: whereas many approaches areparsimonious in that they place greatvalue on achieving maximally eco-nomical descriptions of grammar, SFGis ‘extravagant’ in Halliday’s term. Inorder to capture the multivariate rela-tions between meaning and structure,the grammatical model has to allowfor different perspectives on theclause, providing a much richer and

As relatively simple systems build upinto system networks in this way, thecomplexity increases, but that reflectsfaithfully the complexity of the mean-ing choices that are realised in anyutterance.

The other key feature of the gram-mar is its functional nature. In con-trast to form-oriented approacheswhich concentrate more or less exclu-sively on how clauses are constructed,with meaning excluded and dealt withseparately in a semantics component,SFG is designed to map the relation-ships between forms and meanings ina consistent way. This has a number ofimportant implications. First, the mostimportant labels are functional, orsemantic (telling us what the elementsmean), rather than formal (telling uswhat they consist of). This can be seenin the terms used in Figure 7: they indi-cate what meaning choice is beingmade: for example, whether the nom-

more informative set of descriptions.This then leads on to a third importantimplication: since the communicativefunction of utterances can only befully understood in relation to theirmeaning in context, the grammar hasto be designed in such a way that theanalyst can ‘shunt’ between specificchoices at clause level or below and thecontext within which the utterance isused. For example, an interrogativesuch as ‘Why aren’t you leaving?’ mayfunction as a question, but in certaincontexts it may instead function as acommand. The ‘context’ here wouldinclude who the interactants are, whattheir relationship is, how power is dis-tributed in the culture (who has theright to give commands to whom inparticular circumstances), and so on.In many approaches, the use of theinterrogative as a command would betreated as in some way an extension ofits ‘literal’ decontextualised use as aquestion and dealt with separately as amatter of pragmatics; but, in SFG, theanalyst would take all these factorsinto account in arriving at a full gram-matical characterisation of the utter-ance, exploring the reasons why aparticular meaning expressed in a par-ticular way at a particular point in aninteraction is likely to have appearedappropriate to the speaker in a partic-ular situation.

The two key bases of the model,system and function, come together inthe concept of metafunctions. Halli-day argues that the system networksthat can be identified fall into threemain groupings, with interactionbetween systems in any one group, butlittle or no interaction between the sys-tems across groupings (Matthiessen2006 provides corpus-based evidence

to support this claim). Thus, mood(declarative, interrogative, imperative)interacts with polarity in that, forexample, the position of the negativeparticle ‘n’t’ varies in relation to thesubject with different mood choices;but transitivity roles (see below) areunaffected by whether the clause isdeclarative or interrogative. The threegroupings of systems make up whatHalliday calls three metafunctions,each of which realise different broadtypes of meaning.

The interpersonal metafunctioncom prises those systems which func-tion to enact social relations betweenaddressers and addressees, to expressthe speaker’s viewpoint on events andthings in the world, and to influencethe addressee’s behaviour or views.Apart from mood and polarity, someof the main systems are modality*,mood tag, and, at the semantic level,speech function (whether the clausefunctions as statement, question, com-mand or offer). This is the area of thegrammar in which differences in mean-ing such as the following are captured:

The flight is confirmed. (declarative,statement)

Is the flight confirmed? (interroga-tive, question)

Confirm the flight. (imperative, com-mand)

Would you confirm the flight?(modalised interrogative, command)

The flight mustn’t be confirmed. (neg-ative modalised declarative, command)

As the examples show, SFG highlightsthe crucial distinction between moodand speech function: for example, notall interrogatives function as ques-tions. The choice of forms may seem

SYSTEMIC-FUNCTIONAL GRAMMAR

228

to be imposed naturally by the con-text. For example, if you want infor-mation you use an interrogativequestion; but in fact speakers in dif-ferent contexts consistently exploitchoices. There is, for instance, animportant difference between express-ing a command through the impera-tive or the modalised interrogativewordings shown above; and there isalso a choice between the interroga-tive question ‘Is the flight confirmed’and the declarative question ‘Theflight is confirmed?’. These choicesdepend on, and simultaneously reflectthe speaker’s view of, the context ofutterance. A basic tenet of SFG is thatany difference in wordings is the resultof meaningful choice.

The second metafunction is theexperiential. This is language seenfrom the perspective of how it is usedto talk about events, states and entitiesin the world, to construe the speaker’sview of the world. This kind of mean-ing is traditionally taken – even inmany linguistic approaches – as ‘real’meaning. In SFG, however, it is onlyone kind of meaning: in simple terms,why something is said (the interper-sonal angle) and how it is said (the tex-tual – see below) is as important aswhat is said (the experiential). Anexperiential analysis of a clausefocuses on the process (realised by theverb), the participants in that process(typically, but not always, realised bynominal groups), and the circum-stance(s) in which the process happens(typically realised by adverbial groupsor prepositional phrases). Processescan be divided into six main types,which reflect the cognitive categoriesthat we use to make sense of the eventsaround us (Halliday and Matthiessen

2006 expand on this relation be -tween grammar and cognition): mate-rial (processes of action), mental(processes of sensing, which con- strue our interior worlds), relational(processes of being and having), verbal(processes of conveying messages, bysaying, and so on.), behavioural (char-acteristically human physiologicalprocesses), and existential (processesof existing). Each of these types can befurther subdivided into more delicateoptions: for example, one subdivisionof material processes, the largestgroup with the greatest number ofcross-cutting options, is between ‘cre-ative’ (a process which results in anew entity, such as ‘build’) and trans-formative (a process which involves apre-existing entity, such as ‘cut’). Theparticipants in each process type aregiven different labels which reflecttheir relation to the process. There isno space to give a full account, butTable 2 gives a flavour of transitivityanalysis.

It is worth highlighting that transi-tivity may be ‘blind’ to certain partsof the clause. The clause in Table 3has the same transitivity configurationas the first example in Table 2.

The unlabelled constituents are dealtwith in another part of the grammar (inthe example above, they are interper-sonal in nature). Transitivity labels aresemantic, but it is important to notethat each corresponds to differentgrammatical possibilities. Halliday(1994: xx) stresses that ‘all the cate-gories employed must be clearly“there” in the grammar of the lan-guage’. There is only space to mentiontwo of these grammatical reflexes asexamples: material processes cannotproject (that is have a ‘that’ clause as

SYSTEMIC-FUNCTIONAL GRAMMAR

229

their complement), whereas mental andverbal processes can; mental and behav-ioural processes must normally havehuman sensers/behavers, whereas mate-rial processes have no such restriction.

The third metafunction is the tex-tual. This is the part of the grammarwhich is concerned with how themeanings in the clause are organisedto fit in with the co-text of the sur-rounding messages and with thewider context of the utterance. Thetextual system that has been mostfully explored is that of theme: thetheme of a clause is the initial experi-ential constituent, which has a specialrole in signalling how the currentclause relates to clauses around it.Very broadly, there is a choicebetween unmarked themes (wheretheme and subject are the same),which typically signal continuity ofsome kind in the topic, and markedthemes (where something other thansubject, such as an adverbial adjunct,

is theme), which signal that there issome kind of specific contextual pres-sure at work, often associated with achange of textual frame. For example,in the following extract from aguide to a historical monument, thechoice of themes (underlined) isdesigned to help the reader to followthe text organisation. The markedthemes (in italics) signal that the textis moving on to deal with a differentpart of the church, whereas theunmarked themes signal that the textis, for the moment, continuing tofocus on the same part of the churchintroduced in the preceding markedtheme.

On the northern side of the churchwas a porch . . . Further east, on thewest side of the north transept was aroom . . . The east wall of the room wasdivided into three parts . . . Its use is notknown for certain, but it may have beena vestry and sacristy . . .

SYSTEMIC-FUNCTIONAL GRAMMAR

230

The man drove the car fastActor Process: material, Goal Circumstance

transformative

I recognised the driverSenser Process: mental, cognition PhenomenonThe man was a fast driverCarrier Process: relational, Attribute

attributive

Table 2

Might the man have driven the car too fast perhaps ?- Actor - Process:

GoalCircumstance

material

Table 3

One factor which complicates thealready complex three-dimensionalpicture above is grammatical meta -phor, which has been identified inSFG as a crucial mechanism by whichlanguage can expand the potentialmeanings that can be expressed.Grammatical metaphor* involvesthe use of wordings which do notmap in a straightforward way ontothe meanings that they express. Thiscan be seen in the following news -paper headline: ‘Rise in radiationexposure leads to warning’. A transi-tivity analysis of this clause is in Table4.

This captures part of the meaning asit is expressed; but it is clearly possibleto unpack the clause into more ‘con-gruent’ wordings which reflect morenaturally the events in the world thatare being represented. One possi-ble wording (some of the ‘missing’information is supplied from the news-paper article itself) would be ‘Because[= ‘leads to’] doctors are exposing [=‘exposure’] patients to radiation [=‘radiation’] more often [= ‘rise’],researchers are warning [= ‘warning’]that . . .’. This is a much less econom-ical way of representing the meaning,but it is also more natural, in that it iseasier to process in speech and easierfor a non-specialist to understand.Grammatical metaphor of this kind isparticularly associated with formaltechnical writing. It is not just animpressive-sounding way of sayingsimple things (though it is certainly

used in that way in some texts); itallows new kinds of meanings to bemade. In the headline, ‘radiation’ is aninstance of metaphor, but it isextremely difficult to unpack it: it hasbecome an accepted term for referringin a condensed form to a complex phe-nomenon. Grammatical metaphor isoften ignored in other grammaticalapproaches; but, since SFG is a gram-mar of discourse, it has a central placein the model.

SFG is the most fully-developedalternative to what was, for much ofthe twentieth century, the main lin-guistic paradigm, which aimed to splitthe problem of describing languageinto separate areas such as syntax,semantics, sociolinguistics, and so on.SFG rejects that kind of ‘divide-and-conquer’ solution, on the grounds thatlanguage is a resource for social com-munication and can only be properlyunderstood if that whole picture istaken into account at all stages ofinvestigation. Its orientation to lan-guage in use means that it hasbeen widely adopted in discourseanalysis* and corpus linguistics andin a range of other areas, mostnotably education, natural languagegeneration and language acquisitionstudies.

Primary sourcesHalliday, M. A. K. (1978). Language as

Social Semiotic: The Social Interpreta-tion of Language and Meaning.London: Arnold.

SYSTEMIC-FUNCTIONAL GRAMMAR

231

Rise in radiation exposure leads to warningActor Process: material Circumstance

Table 4

232

TRANSFORMATIONAL-GENERATIVEGRAMMAR

Any formal description of languageconsisting of an algorithm which gen-erates sentence structures and of a setof transformations which modifythem systematically. In particular thetheory developed by Noam Chomsky,whose goal it is to account for theimplicit knowledge of language inher-ent in the human mind by means of aformalised system of rules. Transfor-mational-generative grammar claimsto produce all and only the possiblesentences of a language determined byintuition* and the evaluation by com-petent native speakers and to accountfor language acquisition and speakers’knowledge of language.

See also: Acceptability/Grammaticality; Behaviourism;Empiricism/Rationalism;Innateness; Language of Thought;Logical Form; Minimalism;Universal GrammarKey Thinkers: Chomsky, Noam;Fodor, Jerry; Katz, J. J.; Montague,Richard

Transformational-generative gram-mar has been one of the most influen-tial linguistic theories since thepublication of Chomsky’s (1957) Syn-tactic Structures. It is a systematic,objective, scientific formalisation ofgrammar, based on the belief that thestructure of language is determinedby the structure of the human mind,that all languages share somecommon, universal characteristicsand that the species-specific creativ-

Halliday, M. A. K. (1985, second edition1994, third edition, with ChristianMatthiessen, 2004). An Introductionto Functional Grammar. London:Arnold.

Halliday, M. A. K and Ruqaiya Hasan(1976). Cohesion in English. (EnglishLanguage Series 9). London: Longman.

Halliday, M. A. K. and Christian M. I.M. Matthiessen (2006). Construing Experience Through Meaning: A Lan-guage-based Approach to Cognition.Second edition. London and New York:Continuum.

Halliday, M. A. K. (2002–). CollectedWorks of M. A. K. Halliday. Ed.Jonathan Webster. Ten vols. Londonand New York: Continuum.

Martin, J. R. (1992). English Text: Systemand Structure. Amsterdam and Philadel-phia: John Benjamins.

Matthiessen, Christian M. I. M. (1995).Lexicogrammatical Cartography: Eng-lish Systems. Tokyo: International Lan-guage Sciences Publishers.

Further readingButler, Christopher S. (2003). Structure

and Function: A Guide to Three MajorStructural-Functional Theories. Twovols. Amsterdam: John Benjamins.

Eggins, Suzanne (2004). An Introductionto Systemic Functional Linguistics.Second edition. London: Pinter.

Matthiessen, Christian M. I. M. (2006).‘Frequency profiles of some basic gram-matical systems: an interim report’. InGeoff Thompson and Susan Hunston(eds), System and Corpus: ExploringConnections. London: Equinox. 103–42

Thompson, Geoff (2004). IntroducingFunctional Grammar. Second edition.London: Arnold.

Geoff Thompson

233

TRANSFORMATIONAL-GENERATIVE GRAMMAR

sentences of the language under analy-sis’ (1957: 11). Furthermore such atheory ‘is to provide a general methodfor selecting a grammar for each lan-guage’. In order to achieve this aim,Chomsky relied heavily on featuresthat are important for the success oftheories in physics, and thus arguablyestablished linguistics as a science.

After rejecting linear finite stategrammars and after showing the limitations of phrase-structure (PS)descriptions, a formalised versionof immediate constituent analysis,Chomsky developed a grammar witha tripartite structure. It consisted ofa set of phrase-structure rules, a setof morphophonemic rules that con-vert sequences of morphemes intosequences of phonemes, and an inter-mediate level of transformations mod-ifying the output strings of elementsgenerated by the PS-rules into stringsthat can serve as the input to the mor-phophonemic rules. With a generativecomponent generating the underlyingstructures and a transformationalcomponent modifying them into sur-face structures, this approach was thebirth of what was to be known astransformational-generative gram-mar.

The PS-rules given in Chomsky(1957: 111), which generate theunderlying structures, are the follow-ing (slightly modified): (1) � → NP �VP, (2) VP → Verb � NP, (3) NP →{NPsing / NPpl}, (4) NPsing → T � N, (5)NPpl → T � N + S, (6) T → the, (7) N→ {man, ball, etc.}, (8) Verb → Aux +V, (9) V → {hit, take, walk, read, etc.},(10) Aux → C(M) (have + en) (be +ing), (11) → {will, can, may, shall,must}. � stands for sentence, NP fornoun phrase, VP for verb phrase, Ɵ

ity* of human language – that is, thecapacity of all native speakers to produce and understand an infinitenumber of sentences that they havenever heard before – must beaccounted for. Furthermore, such atheory should also reflect processes oflanguage acquisition and languageuse. Transformational-generativegrammar has been constantly devel-oped into more powerful theories,mainly by Chomsky himself, based onsuggestions and criticism from manyscholars, perhaps the best known ofwhom are Jerry Fodor, J. J. Katz, andRichard Montague. Chomsky’s viewshave always been controversial, butthey cannot be ignored.

In Syntactic Structures, the first ofChomsky’s continually developingmodels, Chomsky declared thatsyntax is a completely autonomouspart of language, independent ofsemantics and the phonologicalsystem. Furthermore he proclaimedthat the syntax of a language can beformalised in a mathematically pre-cise way by means of different rulesoperating on different levels. His pro-gramme was revolutionary in twoways. First, it was an attempt to for-malise at least some of the features oflanguage; and, second, Chomsky dis-sociated himself from the prevailingidea that a grammatical description ofa language can only be derived fromthe observation and analysis of actu-ally occurring data. For Chomsky‘[s]yntax is the study of the principlesand processes by which sentences areconstructed in particular languages.Syntactic investigation of a given lan-guage has as its goal the constructionof a grammar that can be viewed as adevice of some sort for producing the

234

TRANSFORMATIONAL-GENERATIVE GRAMMAR

face structure. In contrast to PS-rules,transformations are not applied tosingle elements but to phrase markers,the structural description of theunderlying structure generated by thePS-rules. In other words, a transfor-mational rule takes one structure as itsinput and modifies it into anotherstructure.

Formally, transformations consistof two components: the structuraldescription, which serves as the con -dition for the transformation to beapplicable, and the structural change,which describes the output. As anexample of an optional transforma-tion, consider the structural analysisNP – Aux – V – NP. In a formalapproach these four elements arenumbered linearly from left to right asX1 – X2 – X3 – X4. The structuralchange now gives the precise instruc-tion X1 – X2 – X3 – X4 ⇒ X4 – X2 +be + en – X3 – by + X1 , that is, in aless formal way, to re-write the inputstring NP1 – Aux – V – NP2 as NP2 –Aux + be + en – V – by + NP1. It caneasily be seen that this transformationturns the underlying structure of anactive sentence such as ‘The man hitthe ball’ into the passive ‘The ball washit by the man’.

It must be noted that in the theoryof Syntactic Structures, the PS-rules,together with the obligatory transfor-mations, only generate simple, active,affirmative, declarative sentences, so-called kernel sentences, and that allother sentences are generated byapplying optional transformations.Though it would not be impossible togenerate all the interrogative, nega-tive, passive sentences, and so on by aset of PS-rules alone, the obvious rela-tionship between, for example, pas-

for zero, T for the definite article, Nfor noun, V for verb, S for the pluralmorpheme, Aux for auxiliaries, M formodals, and C is the element that willlater be interpreted by a transforma-tional rule as concord; en stands forthe past participle morpheme and ingaccounts for the progressive form.Wavy brackets indicate a set of possi-bilities from which one must bechosen, and round brackets stand foroptionality. Thus rule (10) states thatC must be chosen and that we maychoose zero or more of the parenthe-sised elements in the given order,resulting in eight different possibilitiesto rewrite Aux.

Phrase-structure rules are re-writerules which formalise immediate con-stituent structures of sentences on anabstract level. All the elements to theright of the arrow are the proper con-stituents of the single element to theleft. In other words, rule (1) formalisesthe fact that in the sentence “The manhit the ball” ‘the man’ and ‘hit theball’ are proper constituents of thesentence, whereas ‘the man hit’ is not.The output of the PS-rules are kernelstrings, that is the underlying struc-tures of kernel sentences. In order toarrive at the surface structure of thekernel sentences we have to apply allthe obligatory transformations whichregulate the morphological processes,for example, concord.

Transformations were first intro-duced in linguistics by Zellig Harris(1952) as a means to account forthe relationship between linguisticexpressions at surface structure. InChomsky’s terminology, transforma-tions are re-write rules applied on theoutput of the PS-rules, the underlyingstructure, in order to arrive at the sur-

235

TRANSFORMATIONAL-GENERATIVE GRAMMAR

were responsible for the recursivenessof language, another solution to han-dling this vital feature of a comprehen-sive theory had to be found. So, insteadof taking two sentence structures andconjoining them, it was proposed thatby integrating a rule of the type NP +N + (S) the task could be shifted intothe PS-rules and the generalised trans-formations could be discarded. Simi-larly, instead of having a negative orinterrogative transformation, insertingthe necessary element into the respec-tive affirmative or declarative kernelstring, it was suggested that an abstractmarker (neg) or (Qu), or both of them,is inserted in the underlying structureas an optional element, yielding therule S (Qu) (neg) + NP + VP. If thesemarkers are chosen in the generativeprocess, the respective obligatorytransformations are triggered off bythese markers to produce negative,interrogative or negative-interrogativesentences.

J. J. Katz and Paul M. Postal sug-gested in their seminal book An Integrated Theory of LinguisticDescriptions (1964) that the underly-ing structure already contained all thenecessary elements for the semanticinterpretation of sentences, and thattransformations only account for thenecessary adjustments in order toarrive at the surface structure and thusare obligatory. Furthermore as themeaning of the sentences is stated bymeans of PS-rules, these transforma-tions must also be meaning-preserving.

Based on these suggestions, Chom-sky proposed in Aspects of the Theoryof Syntax (1965) a much strongertransformational-generative modelthan his original one. The base com -ponent of the Aspects Theory still

sive and active sentences would belost. All transformations are based ondeletion or insertion of elementsresulting in substitutions or permuta-tions and are either obligatory oroptional. Whereas the obligatorytransformations account for the necessary morphological processesneeded to arrive at the actual sen-tences, optional transformations suchas negation, passivisation, questionformation and so on change the mean-ing by introducing new semanticinformation. Note that sentences suchas (a) ‘The man didn’t hit the ball’ (b)‘The ball was hit by the man’ (c) ‘Didthe man hit the ball?’ and so on, arenot derived from (d) ‘The man hit theball’ but that the kernel sentence (d) isgenerated by means of obligatorytransformations and sentences (a) to(c) by optional transformations fromthe same underlying kernel string.

Furthermore Chomsky distin-guishes between singular and gener-alised transformations. Singulartrans formations operate on individualstrings of elements, whereas gener-alised transformations combine twodifferent strings into one, resulting inembedding and conjunction and thusaccounting for the possibility of recur-sive constructions and the infinitecapacity of the grammar.

The model proposed in SyntacticStructures had some major draw-backs: the growing complexity of thetransformational component, particu-larly with respect to the generalisedtransformations, the covert relation-ship between, for instance, affirmativeand negative sentences, and above allthe fact that it was a purely syntactictheory.

As the generalised transformations

236

TRANSFORMATIONAL-GENERATIVE GRAMMAR

lar to Chomsky’s phrase markers assemantic representations from whichthe surface structures are generated bytransformations. However, by tryingto include more and more phenomenainto their theory, they had to widenthe range of rules so drastically thatthe model lost its explanatory power.Furthermore, such a theory cannotaccount for universal features of language and is a move away fromexplanation back to descriptivism*.

Apart from generative semanticsthere were a very large number ofpost-Chomskyan models, the mostimportant of which were Charles Fill-more’s (1968) Case Grammar, operat-ing with logical cases such as agent,instrument, experiencer, victim, placeand so on, as constituents in the basecomponent; Valency Grammars,which considered the verb as the onlygoverning element from which allother expressions are dependent; andMontague Grammar, as an attempt toformalise the semantic structure ofnatural languages.

Insights from these approaches ledto a continuous development of theStandard Theory to what were to beknown as the Extended StandardTheory and the Revised ExtendedStandard Theory. The major change inthe Extended Standard Theory wasthat semantic interpretation could notbe based on the deep structure alone,but that it is determined by the deepstructure as well as by the surface struc-ture. However, the deep structurekeeps its important syntactic role. Inthe Revised Extended Standard Theorywe have a strict delimitation of the dif-ferent grammatical components, thatis syntax, semantics, as well as phonol-ogy, stylistics and pragmatics, the

consisted of a set of PS-rules and a lex-icon to create the deep structure. But incontrast to the 1957 model, the basenot only contained syntactic informa-tion but also all the semantic elements,and could thus serve as the input forthe semantic component, on the onehand, and the transformational andphonological component, on the other,to account for semantic interpretation,that is meaning and phonological formrespectively. The syntactic part stillconsisted of a generative and a trans-formational part, but their functionswere different. It was now the basecomponent that accounted for recur-siveness and for all the semanticallyrelevant options, and therefore therewas no need for optional transfor -mations anymore. In the 1960s and1970s this theory was so importantthat it became known as the StandardTheory, and there was hardly any syn-tactic research done that was not basedon this model by either accepting anddeveloping it or by rejecting andreplacing it by other theories.

Major criticism of the StandardTheory came from within generativegrammar itself. Some of Chomsky’sstudents felt that the scope of gram-mar was too narrow and should beextended into other areas of language,particularly into semantics. Theycalled their approach generativesemantics*. Observing that the gram-maticality of a sentence is not inde-pendent of the beliefs of the speaker orthat it has a lot to do with the lexemesactually chosen, they suggested thatsyntactic and semantic processescould not be separated and thus dis-carded the notion of deep structure.They claimed that all aspects of mean-ing could be captured in a form simi-

237

TRANSFORMATIONAL-GENERATIVE GRAMMAR

assignment of abstract Case and its mor-phological realization. Control theorydetermines the potential for reference ofthe abstract pronominal element PRO.

It is the aim of GB-theory to find theprinciples and parameters commonto all languages so that the syntax of aparticular language can be explainedalong these lines. Evidence andcounter-evidence from specific lan-guages has led to continual refinementof the theory, so that as much varia-tion in human language as possiblecan be accounted for.

In The Minimalist Program (1995),the latest step in the continuous development of transformational- generative grammar, Chomsky pro-vided a radically new approach to theimplementation of his underlyingideas. The well-established conceptsof D-structure and S-structure havebeen discarded as well as government,the central element in GB-theory. Eventhe ubiquitous phrase-structure ruleshave been eliminated from the theoryto a large degree. The only conceptu-ally necessary categories left are thelexicon and the two levels of phoneticform and logical form* and it is therole of a grammar to map them ontoeach other. It is suggested that all theother categories be dealt with in exter-nally specified systems outside theactual grammar. Using conceptions ofeconomy, assuming that humans useas economic a system as possible, it issuggested that an optimal solution ofrelating phonological form and logi-cal form can be found.

As suggested by the title of Chom-sky’s book, minimalism* has not yetbeen fully developed into a compre-hensive theory, but is a program,

introduction of marked elements andabove all the reduction of the numberof transformations to a single, heavilyconstrained move-� rule. This devel-opment led to the Government andBinding Theory (GB-Theory) advo-cated in Chomsky (1981).

In GB-Theory, sentence descriptionsare simultaneously created on the fourinterdependent levels of syntax (or D-structure), S-structure, phonetic formand logical form, each of which is con-cerned with specific aspects of thedescription of the sentence under con-sideration. It is based on the principlesand parameters theory, which statesthat there is a finite set of fundamentalprinciples common to all natural lan-guages and a finite set of binaryparameters that determine the rangeof permissible variability in language,language acquisition and languageunderstanding. A major differencefrom the Aspects model is the fact thatlogical form is now derived from theS-structure and not from the D- structure.

The principles that constrain therange of possible sentence structuresare contained in closely related sub-systems as formulated in Chomsky(1981: 5–6):

Binding theory poses locality conditionson certain processes and related items.The central notion of governmenttheory is the relation between the headof a construction and categoriesdependent on it. -theory is concernedwith the assignment of thematicroles such as agent-of-action, etc.(henceforth: -roles). Binding theory isconcerned with relations of anaphors,pronouns, names and variables to possi-ble antecedents. Case theory deals with

238

TRUTH THEORIES

Fodor, Jerry A. (1964). The Structure ofLanguage. Readings in the Philosophyof Language. Englewood Cliffs: Pren-tice Hall.

Jackendoff, Ray S. (2007). Language,Consciousness, Culture: Essays onMental Structure (Jean Nicod Lectures).Cambridge, MA: MIT Press.

Lyons, John (1970). Noam Chomsky.Glasgow: Fontana/Collins.

Otero, Carlos (1994). Noam Chomsky:Critical Assessments. Four vols.London: Routledge.

Smith, Neil (1999). Chomsky Ideas andIdeals. Cambridge: Cambridge Univer-sity Press.

Further readingFillmore, Charles J. (1968). ‘The case for

case’. In E. Bach and R. T. Harms, Uni-versals in Linguistic Theory. New York:Holt, Rinehart and Winston.

Harris, Zellig S. (1952). ‘Discourse analy-sis’. Language 28: 1–30.

Katz, Jerrold J. and Paul M. Postal (1964).An Integrated Theory of LinguisticDescriptions. (Research Monographs,26). Cambridge, MA: MIT Press.

Lakoff, George; and John R. Ross (1976).‘Is deep structure necessary?’. In JamesD. McCawley (ed.), Syntax and Seman-tics 7: 159–64.

Salkie, Raphael (1990). The ChomskyUpdate. Linguistics and Politics.London: Routledge.

Jürg Strässler

TRUTH THEORIES

Truth theories are philosophicalattempts to pin down the concept oftruth. Truth is arguably one of the

according to which a theory of thelanguage faculty could be developed.Although the minimalist program iswell established in linguistics, it isseverely criticised by developmentalpsychologists, cognitive scientists andneuroscientists, particularly becauseof its claims with respect to languageacquisition and innateness*.

Transformational-generative gram-mar has influenced the study of syn -tax more than any other theory, but italso had an impact on other fieldswithin linguistics, such as phoneticsand phonology, sociolinguistics, prag-matics, discourse analysis and polite-ness theory. Furthermore Chomsky’sinnovative approach to the study oflanguage and his methodology haveinfluenced anthropology, philosophy,psychology and sociology, as well ascomputer science, neuroscience, liter-ary criticism, music theory, and otherfields. They have provided newinsights for second-language teachingand learning and speech pathology,and above all, they have changed ourconception of the mind.

Primary sourcesBach, Emmon W. (1974). Syntactic

Theory. New York: University Press ofAmerica.

Chomsky, Noam (1957). Syntactic Struc-tures. The Hague: Mouton.

Chomsky, Noam (1965). Aspects of theTheory of Syntax. Cambridge, MA:MIT Press.

Chomsky, Noam (1975). Reflections onLanguage. New York: Pantheon Books.

Chomsky, Noam (1981). Lectures onGovernment and Binding. Dordrecht:Foris Publications.

Chomsky, Noam (1995). The MinimalistProgram. Cambridge, MA: MIT Press.

239

TRUTH THEORIES

the name should avoid. Consider thesentence, ‘I am lying’. If one askswhether this sentence is true or false,one is soon confronted with a para-dox. If it is true, then the speaker isadmitting that he/she is lying or thatwhat he/she is saying is false. On theother hand, if it is false, then we havethe speaker’s word that he/she is notlying or that what he/she is saying istrue. So, either way, the question as tothe truth or falsity of the sentence ‘Iam lying’ leads one to an intellectualquagmire. There are many variants ofthe liar paradox that produce similarresults. Among these are ‘Epimenidesthe Cretan says all Cretans are liars’(is Epimenides lying or telling thetruth?) or ‘The barber of Seville is onewho, while continuing to live inSeville, decides to shave every man inSeville who does not shave himself’(does the barber shave himself?). It isclear that, irrespective of the defini-tion of truth one works with, one hasto make sure that the paradox pro-duced in each of these cases mustsomehow be skirted.

The solutions proposed by phil -osophers for the liar paradox and itsvariants differ considerably. Some rec-ommend that we regard such sen-tences as meaningless on the groundsthat one part of a sentence cannot sen-sibly refer to the sentence as a wholeand therefore predicate of it that it istrue or false. Others concede that suchsentences do make sense (how else dowe know that they are odd?) but areneither true nor false. Others decreethat such sentences are ill-formedbecause they violate a ban on self- reference. Still others decree that allparadoxical sentences, being outrightcontradictions, are false and argue

most central issues in philosophyand truth theories vary a great dealfrom one another in terms of whatthey claim as well as what they takefor granted, that is, their metaphysi-cal baggage. Philosophers havealways been at pains to define truthand suggest ways of ascertainingwhen something is true and when it isnot.

See also: Analytic/Synthetic;Correspondence Theory; DefiniteDescriptions; Logic;Presupposition; Propositions;Truth ValueKey Thinkers: Aristotle; Ayer, A.J.; Davidson, Donald; Frege,Gottlob; Kant, Immanuel; Peirce,C. S.; Quine, W. V. O.; Ramsey, F.P.; Russell, Bertrand; Strawson, P.F.; Tarski Alfred; Wittgenstein,Ludwig

Ever since Pontius Pilate asked thequestion ‘What is truth?’ and, as thesaying goes, preferred not to wait foran answer, the issue has been posedand intensely probed time and timeagain by philosophers. But truth hasproved to be evasive. Over the cen-turies, philosophers have proposedcompeting theories of truth, but this isan area of investigation where newproblems keep cropping up faster thanthey are able to tackle them. In philos-ophy, the predicate ‘true’ is always incontrast with ‘false’, unlike popularusage where it is often opposed to ‘fic-titious’, ‘fake’ or ‘spurious’.

Philosophers have long contendedwith a mind-boggler referred to gener-ically as the ‘liar paradox’. The liarparadox illustrates an early conun-drum which any theory of truth worth

240

TRUTH THEORIES

cal term proposition*, which they dis-tinguish from a sentence (a syntacticentity) and a statement (a unit of prag-matics). Another property of truth isthat a sentence, once true, will alwaysbe true no matter what the specific cir-cumstances of its utterance, makingdue allowances for such elements as‘tomorrow’ that, as it were, date thesentence. In other words, truth is notonly a universal but also an eternalproperty. Furthermore, it is an inalien-able property. With reference to thespecific example above, it must bepointed out that what is being consid-ered truth-apt is neither the sentenceper se nor the proposition expressedby it, but a dated instantiation of it,that is to say, a token of that sentenceas produced on a certain date and at aprecise time. This caveat is absolutelynecessary because, unless we are fur-ther told when/on which day the sen-tence was uttered, it would beimpossible to verify its truth, for thesimple reason that this can only bedone retrospectively – a day after it isuttered.

The philosophers who broadly sub-scribe to the view of truth adumbratedin the foregoing paragraphs are work-ing within what is referred to as thecorrespondence theory* of truth. Thisis a way of conceiving of truth thatdates back to at least as far back asAristotle, who famously claimed in hisMetaphysics (1011b25): ‘to say ofwhat is that it is not, or of what is notthat it is, is false, while to say of whatis that it is, and of what is not that itis not, is true’. Underlying the corre-spondence theory is the assumptionthat truth is a matter of correspon-dence between what one says on theone hand and what there is out there

further that any attempt to derive thetruth of part of a paradoxical sentenceby logical means is doomed to fail.Bertrand Russell identified a versionof the ‘liar paradox’ in set theory byasking us to imagine ‘a set of sets thatare not members of themselves’ andwondering, apropos of this larger set,if it is a member of itself. What Russellsuccessfully showed by his discussionof this antinomy was the importanceof weeding out the threat of contra-diction from a mathematically soundtheory of truth, or for that matter, anytheory whatsoever.

Anyone setting out to propose atheory of truth has to sort out anumber of initial questions. To beginwith, it is important to know what itis of which we may sensibly predicatetruth or, to put it technically, whatsorts of entities can be said to be truth-apt. For instance, the sentence ‘It willrain tomorrow’ is bound to be eithertrue or false – we will soon discoverwhether it is one or the other –whereas ‘Will it rain tomorrow?’ or ‘Ifonly it rained tomorrow/Would that itrained tomorrow’ is not considered tobe a candidate for truth ascription.On the other hand, ‘Il pleuvrademain’, ‘Vai chover amanhã’, ‘kalpani barsega’, which are the French,Portuguese and Hindi translation-equivalents of ‘It will rain tomorrow’respectively, are just as true or falseunder the same set of circumstances asthe English sentence. This means thatwhat is true or false is not the sentence‘It will rain tomorrow’ (which, ofcourse, belongs only to the Englishlanguage). In other words, truth hasto do with what a sentence says orwhat we may call its meaning.

Many philosophers use the techni-

241

TRUTH THEORIES

Frege claimed in a paper publishedtoward the end of the nineteenth cen-tury that the truth of the sentence ‘thepresent king of France is bald’ wasconditional upon there being a king ofFrance. In other words, the sentence(or proposition, if you will; note thatFrege had in mind an undated version)would be neither true nor false if putforward at a time when France was nolonger a monarchy. To control thedamage such a claim represented tobinary logic, Bertrand Russell pre-sented his famous Theory of Descrip-tions, under which what appears to bea reference to the putative king ofFrance turns out to be a claim to theeffect there is a king of France at thetime the sentence is produced andwhich, thanks to its declarative mood,can be adjudged true or false, depend-ing on the facts of the matter.

The Frege-Russell dispute over theneed or otherwise to posit presupposi-tion* as a condition for a sentence tobe either true or false impacted onwork done in linguistic semantics inthe 1970s and the first half of the1980s. The initial enthusiasm infavour of the Fregean approach soongave way to the Russellian alternativeof reinterpreting the putative presup-positions as straightforward entail-ments, thanks to the work of scholarssuch as Deirdre Wilson and RuthKempson.

There are exceptions to the corre-spondence theory’s claim that truth isa matter of correspondence between alinguistic object and a language exter-nal reality. These are so-called ‘ana-lytic’ sentences, first identified byImmanuel Kant, following in the foot-steps of John Locke and GottfriedWilhelm Leibniz. For Kant, analytic

in the world (say, a fact or a state ofaffairs) on the other. It is perhaps themost satisfactory theory from a purelyintuitive point of view and is certainlythe oldest. It is implicit in such routinepractices as a police investigation intoa possible crime in which a witnessaccount is checked and double-checked by comparing it to everysingle detail of the crime scene, or thepolice coming up with an artist’ssketch of the physiognomy of the sus-pect on the basis of details furnishedby persons who claim they witnessedthe crime.

The correspondence theory isunderwritten by metaphysical real-ism, which is a claim for the existenceof things independently of what any-body might say or believe about them.It also posits, as in Ludwig Wittgen-stein’s celebrated analysis in TractatusLogico-Philosophicus, a world con-sisting entirely of facts which can betruthfully and exhaustively mirroredby language. Furthermore, everyproposition is either true or false andnone can be both simultaneously orneither. Now, there can be no doubtthat correspondence theory and itsunderlying metaphysical realism,whether or not explicitly spelt out orthought through, informs most ofwhat laypersons believe about theseissues. Incidentally, when it comes tothe layperson, philosophers are wontto distinguish epistemic truth (whatone knows) and doxastic truth (whatone believes) from aletheutic oralethic truth (truth ipsis).

But the time-honoured claim ofclassical, two-valued logic* to theeffect that between truth and falsitythere could be no third value wasthrown into disarray when Gottlob

242

TRUTH THEORIES

herself, may be seen as putting topractical use the central principle ofcoherence theory–the assumptionbehind the practice being that, if therespondent is telling all the truth, onlythe truth, and nothing but the truth,he/she will not give an answer thatcontradicts any other during the samecross-examination.

In the nineteenth century, CharlesSanders Peirce, William James andJohn Dewey defended a theory oftruth that, despite differences amongthese thinkers on details, claimed thattruth is that which effectively works.This is called the pragmatic (or prag-matist) theory of truth. Peirce addedthe further requirement that truth is‘the opinion which is fated to be ulti-mately agreed by all who investigate’,thus emphasising the public characterof truth. William James replaced thatwith an emphasis on the process ofbelief-formation by the individualrather than the product and pointingto the utility of holding something astrue as the criterion for judging it to betrue. John Dewey, agreeing withPeirce that truth is the conclusion tobe arrived down the end of the road,went on to define it in terms of hisfavourite concept of warranted assert-ibility. One might say that pragma-tism, with its steadfast opposition tothe Cartesian dogma, realism, Conti-nental rationalism and its emphasis oninstrumentalism, is opposed to corre-spondence theory. But it has also beenargued that pragmatism incorporates– or at the very least is not averse to –certain elements of both correspon-dence and coherence theories.

There are other versions of coher-ence theory that concede that,although individual propositions have

statements are those in the declarativemood, where the predicate is con-tained in the concept of the subject, asin ‘All bachelors are unmarried’. Bycontrast, synthetic statements such as‘The dog barked at the stranger’ canbe adjudged true or false, only on thebasis of whether or not they faith-fully report something that actuallyoccurred in the real world. All syn-thetic statements are thus claimed tofall between the limiting cases of tau-tology and contradiction (true andfalse, respectively, regardless of whathappens to be the case in the world).

There are many other theories oftruth that, in one way or another,challenge the correspondence theory.Coherence theory is one such. Coher-ence theorists hold that truth is amatter of internal consistency within awhole system of propositions. In otherwords, truth is to be predicated pri-marily of the system as a whole andonly secondarily and derivatively ofindividual propositions, based onwhether or not they cohere with oneanother. Under such a generous defi-nition, a system comprised entirely offalse propositions (that is, false underthe correspondence theory) may nev-ertheless receive the certificate oftruth, thanks to the fact they coherewith (do not contradict) one anotherand thanks also to the fact that thecoherence theory does not require val-idation by any factor external to thesystem. The popular saying ‘Truthknows no contradiction’ may be seenas drawing on the key insight of coher-ence theory. So too the practice ofcross-examination of a suspect by thepolice or of a defendant by a prosecu-tor, where the attempt is to catch therespondent contradicting himself/

243

TRUTH THEORIES

alinguistic observation concerning thetruth of that sentence in the object lan-guage. In other words, the truth of ‘p’is claimed to be guaranteed by the truthof p. This means that one can predicatetruth only in a language other than theone in which the sentence for truthascription is presented or, alternatively,‘p’ and p belong, as it were, to two dif-ferent languages. Evidently, as Tarskihimself was the first to admit, his defi-nition of truth does not work for natu-ral languages, where self-referentialsentences of the form ‘This sentence isnot true’ are permitted, in spite of thedifficulty in assigning to it a meaning.In a formal language, such trouble-some cases can be blocked by fiat.Tarski’s theory of truth is a version –albeit a very sophisticated one – of thecorrespondence theory.

In 1927, Frank Ramsey published apaper in which he defended the idea,originally entertained by Frege, that tosay of a sentence that it is true is notto say anything else about it. That is tosay, someone who says ‘p’ is therebyalready saying, as it were, ‘p is true’.In other words, the only differencebetween the statements ‘p’ and ‘p istrue’ is that the latter makes explicitwhat the former says implicitly. Thuswas born so-called ‘redundancy’ or‘deflationary’ or ‘disquotational’theory of truth. A. J. Ayer tookRamsey’s thesis to its obvious conclu-sion, namely that truth is an emptypredicate. This has been called the‘disappearance theory’ or the ‘notruth theory of truth’. P. F. Strawsonused Ramsey’s thesis to come to anequally interesting conclusion, byarguing that to say of a sentence thatit is true is to endorse it or give it astamp of approval. Strawson’s thesis

nothing do with a language-externalreality or the way we experience it, thesystem as a whole does, at least at itsfringes. For instance, W. V. O. Quineheld that, in talking about the world,we typically move from talking in cer-tain terms to talking about them, thatis to say, from the material mode tothe formal mode, in what he famouslyreferred to as a ‘semantic ascent’. Atthe other extreme of the spectrum isthe American neopragmatist philoso-pher Richard Rorty who flatly deniesthat there is any relation whatsoeverbetween language on the one handand the external world on the other, sothat all talk of correspondence isphilosophical balderdash.

That truth itself might be definableonly by means of a sort of semanticascent was what led the Polish logi-cian Alfred Tarski to develop hissemantic theory of truth – a theory oftruth in formal or artificial (not natu-ral) languages. He captured theinsight in the formula: ‘ “p” is true ifand only if p’ (Tarski 1933). The keyto interpreting the formula lies inadmitting a neat separation betweenobject language (the language underinvestigation) and metalanguage (thelanguage employed to say thingsabout the object language). Tarski’scentral idea was that truth in a given(object) language can only be explicated by means of a sentence inthe corresponding metalanguage or,equivalently, truth is invariably a met-alinguistic predicate.

Thus, in the formula above, ‘p’ is asentence in the object language in thesense that it is about this sentence thatone predicates truth or whatever. Theoccurrence of p (the one without theinverted commas) is part of the met-

244

TRUTH VALUE

Russell, Bertrand (1905). ‘On denoting’.Mind 14: 479–44.

Strawson, Peter. F. (1949). ‘Truth’. Analy-sis: 83–97.

Tarski, Alfred (1933). ‘The concept of truthin formalized languages’. Reprinted inAlfred Tarski (1956), Logic, Semantics,Metamathematius. Trans. J. H. Woodger.Oxfurd: Clarendon Press. 152–78.

Further readingAlston, William P. (1996). A Realistic

Conception of Truth. Ithaca, NY: Cor-nell University Press.

Davidson, Donald (1984). Inquiries intoTruth and Interpretation. Oxford:Oxford University Press.

Dummett, Michael (1978). Truth andOther Enigmas. Cambridge, MA: Har-vard University Press.

Horwich, Paul (1990). Truth. London:Basil Blackwell.

Kirkham, Richard (1992). Theories ofTruth: A Critical Introduction. Cam-bridge, MA: MIT Press.

Rescher, Nicholas (1973). The CoherenceTheory of Truth. Oxford: Oxford Uni-versity Press.

Kanavillil Rajagopalan

TRUTH VALUE

At their simplest, truth values are oftenthought to be governed by the princi-ple of bivalence, in which any declara-tive sentence’s truth value can be eithertrue (when what the sentence states isthe case) or false (when what the sen-tence states is not the case). There isdebate in logic and the philosophy oflanguage about the appropriateness ofthe principle of bivalence and hence,about the nature of truth value.

has received the name of the ‘perfor-mative* theory of truth’ because itdraws on J. L. Austin’s speech acttheory* which initially posited animportant distinction between sayingand doing. Quine, to whom the dis-quotational theory of truth is gener-ally attributed, interpreted Tarski’ssemantic theory in the light ofRamsey’s insight and argued thatTarski’s was, at bottom, a deflationarytheory to the extent that, in Tarski’sapproach, the predicate ‘is true’ onlymade sense in a metalanguage, not inthe object language.

Truth theories, dedicated to answer-ing the deceptively simple-lookingquestion of what it is to say that a sentence, statement or proposition is‘true’, have dominated the work ofsome of the most eminent thinkers, andhave touched on many of the majorissues in the philosophy of language.

Primary sourcesAristotle (350 BCE). Metaphysics. Trans.

W. D. Ross. Available at http://clas-sics.mit.edu/Aristotle/metaphysics.html(accessed 4 October 2007).

Ayer, Alfred J. (1952). Language, Truth andLogic. New York: Dover Publications.

Frege, Gottlob (1892). ‘On sense and mean-ing’. In Peter Geach and Max Black (eds)(1980). Translations from the Philo soph-ical Writings of Gottlob Frege. Oxford:Blackwell. First edition, 1952. 56–78.

Quine, Willard V. O. (1970). Philosophyof Logic. Englewood Cliffs, NJ: PrenticeHall.

Ramsey, Frank P. (1927). ‘Facts andpropositions’. Aristotelian Society Sup-plementary Volume 7: 153–70.

Rorty, Richard (1979). Philosophy andthe Mirror of Nature. Princeton: Prince-ton University Press.

245

TRUTH VALUE

sition that a given statement is nottrue, we can conclude merely that it iseither false or indeterminate. Usingbivalent semantics it could only befalse.

Non-bivalent semantics has beenmotivated by ‘empty terms’, lackingany truth value at all (Strawson1950); ‘semantic presupposition’* inwhich a logically resolvable sentencepresupposes something impossible(Strawson 1952; see Allwood, Ander-sson and Dahl 1977: 149–55);‘semantic paradoxes’, such as the sen-tence ‘This statement is untrue’ (seeRead 1995: chapter 6; Haack 1978:chapters 8, 11); ‘future contingents’,in which a statement refers to eventsthat have not yet occurred (see Haack1978: chapter 11); and mathematicalstatements, which in some approachesare always false because they referto non-existent entities (see KeyThinkers: Dummett, Michael; Read1995: chapter 8). While truth valuein bivalent semantics may appearstraightforward, wider investigationby philosophers of language and bylogicians suggests that truth value isboth a complex and a contestableidea.

Primary sourcesFrege, Gottlob (1892). ‘On sense and ref-

erence’. In Max Black and Peter Geach(eds) (1960), Translations from thePhilosophical Writings of GottlobFrege. Second edition. Oxford: Black-well. 56–78.

Further readingAllwood, Jens, Lars-Gunnar Andersson

and Östen Dahl (1977). Logic in Lin-guistics. Cambridge: Cambridge Uni-versity Press.

See also: Definite Descriptions;Logic; Logical Form; LogicalPositivism; PropositionsKey Thinkers: Aristotle; Dummett,Michael; Frege, Gottlob; Strawson,P. F.

Though problems about truth and fal-sity are discussed in ancient logic, forexample by Aristotle, the phrase‘truth value’ as we use it in modernphilosophy comes from GottlobFrege. He defined the truth value of asentence as its being true or false, andclaimed that there are no other truthvalues in declarative sentences, whichnormally refer to ‘the True’ or ‘theFalse’. Though Frege’s further viewthat truth values are objects is dis-putable, the main controversies abouttruth value concern what it is that hastruth value (see Kneale and Kneale1962; Haack 1978: chapter 6) andwhether every declarative sentence iseither true or false.

A sentence must be declarative,rather than, for example, interroga-tive or imperative, if it is to be a can-didate for having a truth value. Thus,‘London is in England’ and ‘Torture iswrong’ are candidates; ‘Where isLondon?’ and ‘Don’t go there’ are not.Declarative sentences are usuallycalled ‘statements’. By the principle ofbivalence, each statement is either trueor false, but not both. Bivalent seman-tics, then, recognises only two truthvalues; the semantics of classicallogic* is bivalent.

In contrast, in a non-bivalentapproach to semantics a statement’sfalsehood does not follow from itsbeing not true. For example, supposeour truth values are: true, false, andindeterminate. Then from the suppo-

246

TYPE/TOKEN

mental apparatus. The contrast be -tween type and token maps quiteclosely onto the contrast between uni-versals and particulars or onto thatbetween properties and their instances.Most uses of the words ‘type’ or‘token’ by linguists or philosophersoperate broadly along these lines.

Most appeals to the type/token dis-tinction in a linguistic context belongto one or other of two kinds. For thefirst, consider how many letters thereare in the word ‘London’. Countingby letter type, there are four: ‘L’, ‘O’,‘N’, and ‘D’. Counting by letter token,there are six: one ‘L’, two ‘O’s, two‘N’s, and a ‘D’. This contrast alsoapplies at the level of whole expres-sions. There are, for example, nineword tokens in (1) but only sevenword types:

(1) ‘The blue cat ran circles around thegreen cat.’

This manifestation of the type-tokendistinction clearly generalises to manyother linguistic categories. Think, forexample, of how many phonemes*there are in ‘Mississippi’.

The second manifestation of thedistinction crops up when authorsmark the contrast between sentencesand utterances of sentences bydescribing the latter as tokens of theformer. As with the first use, this alsohelps to eliminate ambiguity*. Sup-pose Jane and John both utter the sen-tence ‘London is in England’. Havethey performed the same utterance?The answer is affirmative if we meanthe same utterance type, where thetype is defined as the sentence pro-duced, but negative if we mean thesame utterance token.

Haack, Susan (1978). Philosophy ofLogics. Cambridge: Cambridge Univer-sity Press.

Kneale, William and Martha Kneale(1962). The Development of Logic.Oxford: Clarendon. Chapters 8 and 10,especially 576–93.

Read, Stephen (1995). Thinking aboutLogic. Oxford: Oxford UniversityPress.

Strawson, P. F. (1950). ‘On referring’.Mind 59: 320–44.

Strawson, P. F. (1952). An Introduction toLogical Theory. London: Methuen.

Stephen McLeod

TYPE/TOKEN

At its most intuitive, the distinctionbetween types and tokens is a distinc-tion between two different ways ofcounting linguistic entities. Onereason for paying attention to the dis-tinction is the ease with which failureto do so causes confusion in discus-sions of language.

See also: Language of ThoughtKey Thinkers: Peirce, C. S.

Suppose you are asked how many carsthere are in a car park. You couldcount by type or by token. For exam-ple, if there are five Fords, one Fiat andthree Toyotas, that gives three types ofcar but nine tokens of car. The termi-nology of ‘type’ and ‘token’ was coinedby C. S. Peirce (1966: 4.537), but thedistinction it marks has been knownunder different labels for far longer.Indeed, the distinction is arguably aprimitive component of everyone’s

247

UNIVERSAL GRAMMAR

thought* as having been uttered, and‘tokened’ is a useful alternative. Totoken a sentence of the language ofthought is for that sentence to occurexplicitly in one’s brain. Philosophersof mind also make a related distinctionbetween types and tokens of actions(where only the latter are specificevents) and types and tokens of mentalstate (such as the belief that London isin England, which anyone might have,versus Jane’s belief that London is inEngland).

The type/token distinction admits,then, of several different manifesta-tions in linguistics. Using the label tomark the contrast between a sentenceand a particular utterance of it may beslightly out of kilter with other usesbut it has become dominant and is atleast comprehensible in context.

Primary sourcesPeirce, Charles S. (1966). Collected Papers

of Charles Sanders Peirce. Cambridge,MA: Harvard University Press.

Further readingWetzel, Linda (2006). Types and Tokens:

An Essay on Universals. Cambridge,MA: MIT Press.

Alex Barber

UNIVERSALGRAMMAR

In Noam Chomsky’s theory of humannatural language, the geneticallydetermined initial state of the lin -guistic competence in the individ-ual human mind/brain, which is the

This second usage may seem atodds with the first. To obtain twotokens of the word ‘Alfred’ in thesecond usage of the type/token dis-tinction would require two utterancesof the word. But in the first usage thereare two tokens of this word in the sen-tence ‘Alfred Brendel was named afterAlfred the Great’, even if that sentenceis not uttered even once.

The relation between the first andsecond usages can be understood if wereturn to our non-linguistic example.The vehicles in a car park can be typedaccording to manufacturer, to colour,to engine size, or to age, and so on.Similarly, the utterances produced at aparticular party can be typed in differ-ent ways: by the identity of the utterer,for example, or by loudness, pitch,time of occurrence,and so on. But themost useful way of typing utterancesin linguistics is usually by the expres-sion produced. So dominant is thisway of comparing and contrastingutterances that it has become normalto talk as though sentences themselvesare the type, while utterances aretokens of them. Though this useis, arguably, erroneous, it is simpleenough to understand what an authorhas in mind in writing this way. Thepractice has even given rise to talk of asentence’s being ‘tokened’, by which ismeant its being uttered. A relatedpractice is to describe the indexicalityof expressions such as ‘I’, ‘now’, andso on as ‘token reflexivity’, sinceindexicals depend for their referentialproperties on features of the context oftheir utterance.

A further factor driving this seconduse of ‘token’ derives from the philos-ophy of mind. It makes little sense totalk of sentences of the language of

248

UNIVERSAL GRAMMAR

the linguistic competence of a child,beyond the lexical dimension.

Amazement is often a plea for expla-nation. What enables the human mind/brain to move from a seemingly poorinitial state to an obviously complexmature state? This is the fundamentalquestion of language acquisition,dubbed by Chomsky ‘Plato’s problem’.Within the spectrum of answers givento that question, there are two extremeand commonly rejected positions andthen two intermediate ones, both ofwhich are related to rich philosophicaltraditions that have been in rivalry formany years. One of these intermedi-ate answers is the idea of universalgrammar.

All the different answers are posi-tions with respect to the role played inlanguage acquisition by the humanmind/brain and by human experience.One extreme position portrays thewhole process of language acquisitionas resting solely on human experience;in the initial state of the process thebaby’s mind/brain is a blank tablet.The opposite extreme position por-trays the whole process as restingsolely on innate human knowledge.Both of these positions are obviouslyuntenable. The former cannot explainthe fact that different children, whoseexperiences, in terms of the utterancesmade in their vicinity, are quite differ-ent from each other, turn out to speakthe same natural language. The latercannot explain the fact that there aremany different natural languages.Thus, conceptions of language acqui-sition that have the form ‘either natureor nurture’ are wrong. Accordingly,the intermediate positions see bothnature and nurture as offering essen-tial contributions to the language

starting point of the linguistic devel-opment of the individual into amature user of a natural language.

See also: Acceptability/Grammaticality; Continuity;Empiricism/Rationalism;Generative Phonology; GenerativeSemantics; Innateness; Mentalism;Transformational-GenerativeGrammarKey Thinkers: Chomsky, Noam;Kant, Immanuel.

The linguistic development of everyhuman being is amazing. When a babyis born, it does not speak, nor does itunderstand the language of its par-ents. But some years later, every childunderstands much of what is beingsaid in the natural language used athome and uses parts of it in its spon-taneous speech. After some additionalyears, the boy or girl is a fluent user ofthe same natural language and seemsto share his or her linguistic compe-tence with adult, mature users of it.

Such a development is amazing onseveral accounts. First, it apparentlyrequires no learning, in the sense usedwhen we describe the child as havinglearnt to use a spoon or a pencil prop-erly and regularly. Second, it appar-ently requires no effort–neither on thepart of the baby or the child, nor onthe part of the parents. What isrequired is that the latter use the lan-guage in the presence of the former.Third, it takes place independently ofother facets of human development,such as spatial orientation or facerecognition. Fourth, and most impor-tantly, the product of development isvery complex, as is easily realised byanyone who tries to describe in full

249

UNIVERSAL GRAMMAR

inductive learning functions from theutterances babies and children hear.An example of such an abstract property would be the very idea ofstructure. A mature speaker’s linguis-tic competence induces a distinctionbetween what counts as a proper sen-tence in one’s natural language andwhat does not count as such. A gen-eral learning device has to capture thisdistinction, simply by processingutterances of sentences or other lin-guistic expressions. Such a generaldevice would have to reach the con-clusion that the distinction does notinvolve numerical calculation butrather structural processing. There isno reason to assume that generallearning devices will have the abilityto reach such a conclusion and also doit quickly. This is especially so in thecontext of the limited linguistic data-base of a small child.

The major alternative to the idea ofa ‘general learning device’ in lan-guage acquisition is that of universalgrammar. Here the position is thatthe mind/brain of a new-born babyincludes an innate concept of natu-ral language. That is a system ofproperties shared by all human lan-guages, historically actual, or psycho-logically possible, which constitutesthe essence of human language. Thus,for example, the fact that the syntaxof a human natural language is struc-tural rather than numerical is a factabout language that is included inthe universal grammar. It is notlearnt by a general learning devicefrom experience, but innately givento human beings as a feature of themind/brain.

It should be noted that universalgrammar is not a grammar in the

acquisition process. However, theyinvolve different views of the contri-bution of the human mind/brain to theprocess, and consequently, also of thecontribution made by human experi-ence to the process.

A position that reflects the philo-sophical tradition of empiricism takesthe mind/brain of the baby when bornto include a general learning device,one that processes what is observedduring experience and gradually devel-ops all human cognitive competencies,including language, but also facerecognition and calculation, for exam-ple. The mind/brain of the baby knowsnothing about language but acquires itby applying the general learning deviceto the linguistic experience.

It is important to understand thatsuch a position is actually a researchprogramme. To hold such a positionwithin the framework of scientificresearch is to be committed to aninvestigative effort to propose atheory that specifies the structure andoperation of the general learningdevice and to put it to empirical test ina variety of areas of human cognitivecompetency, including language.

The major argument that has beenraised against that position is called‘the poverty-of-the-stimulus argu-ment’. It has been argued by Chomskyand many of his followers that thenatural language acquired by the childhas highly abstract properties thatcannot be the product of any generallearning device, especially since theinput received consists only of utter-ances made in its presence. For exam-ple, each sentence of a naturallanguage has a syntactic structure.Those structures have abstract prop-erties that cannot be formulated by

250

UNIVERSAL GRAMMAR

though ancient, is particularly relatedto seventeenth-century philosophy oflanguage and cognition. Its genericnature is related to Kant’s conceptionof the mind. Much of generative linguistics should be viewed as takingplace within the research programmeof universal grammar.

Primary sourcesChomsky, Noam (1966). Cartesian Lin-

guistics: A Chapter in the History ofRationalist Thought. New York andLondon: Harper and Row.

Piattelli-Palmarini, Massimo (ed.) (1980).Language and Learning: The Debatebetween Jean Piaget and Noam Chom-sky. Cambridge, MA: Harvard Univer-sity Press.

Chomsky, Noam (1986). Knowledge ofLanguage: Its Nature, Origin, and Use.New York, Westport, CT and London:Praeger.

Chomsky, Noam (1982). The GenerativeEnterprise. A discussion with Riny Huybregts and Henk van Riemsdijk.Dordrecht: Foris; The Generative Enter-prise Revisited (2004). Discussions withRiny Huybregts, Henk van Riemsdijk,Naoki Fukui and Mihoko Zushi, with anew Foreword by Noam Chomsky.Berlin and New York: Mouton deGruyter.

Chomsky, Noam (1988). Language andProblems of Knowledge: The ManaguaLectures. Cambridge, MA and London:MIT Press.

Chomsky, Noam (2002). On Nature andLanguage. Ed. Adriana Belletti andLuigi Rizzi. Cambridge: CambridgeUniversity Press.

Further readingCook, Vivian and Mark Newson (1996).

Chomsky’s Universal Grammar: An

ordinary sense. It is a system of con-ditions that every natural languagefulfills precisely because every natural language is an end-product of an acquisition process that begins withthe universal grammar. Universalgrammar is uniquely linguistic. It iswhat the human mind/brain knows,by genetic endowment, about naturallanguage. This is one facet of themodularity of language: its cognitiveindependence of the rest of themind/brain. Universal grammar isgeneric, in the sense of serving as theinitial state of different acquisitionprocesses that lead to different natu-ral languages.

The unique linguistic nature of theuniversal grammar and its genericproperty are manifest in a theory ofprinciples and parameters that hasbeen developed within the researchprogramme of universal grammar.Universal grammar includes a class ofparameters that have to be set, ongrounds of linguistic experience.When the parameters are set, much ofthe nature of the natural language tobe acquired has been determined. Anexample is the head parameter: thehead of a noun phrase is a noun, thehead of a verb phrase is a verb, and soon. In a language, all heads appear onthe same side of the phrase. Some nat-ural languages are head-first ones:English is an example. Some are head-last ones: Japanese is an example. Inorder to set the head parameter, thechild does not have to hear more thana few examples of appropriate expres-sions or sentences.

The universal grammar researchprogramme emerged from the philo-sophical tradition of rationalism. Therelated conception of innateness,

251

USE/MENTION

(1) Paris is a bustling metropolis.(2) ‘Paris’ begins with the sixteenth

letter of the English alphabet.

Here are two implausible claims:

(3) ‘Paris’ is a bustling metropolis.(4) Paris begins with the sixteenth

letter of the English alphabet.

(1) and (3), while differing in accu-racy, sound the same when spokenaloud. The same goes for (2) and (4).This is why failure to be clear onwhether one is talking about what aword refers to or about the word itselfcan cause confusion.

In writing various conventions allowfor easy disambiguation between useand mention. Mentionings are oftenenclosed within single invertedcommas; under another convention,they are italicised. In speech, contextusually suffices to make clear whichinterpretation is intended. If that is notpossible, speakers sometimes make adeliberate effort to highlight the dis-tinction by explaining that they mean‘the word “Paris”’ or ‘Paris the city’.

For an example of where confusionmight occur, consider the contrastbetween (5a) and (5b):

(5) (a) What does ‘citizenship’ mean?(b) What does citizenship mean?

An utterance of (5a) would typicallybe inviting a simple definition of theword, such as might be given in a dic-tionary. (5b), though it sounds thesame, is more plausibly asking forsomeone’s view on the broader impli-cations of citizenship as a status.

Several factors make the use/mentioncontrast difficult to define precisely.

Introduction. Second edition. Oxford:Blackwell.

McGilvray, James (ed.) (2005). The Cam-bridge Companion to Chomsky. Cam-bridge: Cambridge University Press.

Asa Kasher

USE/MENTION

To utter the word ‘Paris’, for example,while referring to the French city, is touse that word. To utter the word whilereferring to the word itself is to men-tion the word. The distinction mattersto those investigating languagebecause they are required to talksometimes about a word and some-times about what that word refers to,giving rise to potential confusion overwhich they are doing.

See also: Propositional Attitudes;Sense/ReferenceKey Thinkers: Frege, Gottlob;Leibniz, Gottfried Wilhelm; Quine,W. V. O.

In spoken language the distinctionbetween the use of a word and themention of a word is often difficult towork out. In most languages there isno convention for indicating throughpronunciation which is beingemployed. Speakers and hearers relyon the context in which a word is usedin order to convey and understandwhat is meant. Yet the distinction hasa significant effect on meaning to theextent that it can make the differencebetween sense and nonsense. Forexample, here are two plausibleclaims:

252

USE/MENTION

(7) The final letter in ‘Paris’ is leftunvoiced by the people who livethere.

(8) The people of Paris do not pro-nounce it that way.

(9) ‘Giorgione’ was so-called becauseof his size.

The use/mention distinction can bedifficult for language users to negoti-ate, yet it is important in discussionsof language precisely because it existsat the interface between expressionand meaning. The fact that for speak-ers and hearers the distinction isoften imperceptible without addi-tional communicative actions, includ-ing gestures and intonation, has madeuse/mention a central problem inpragmatics as well as in the philoso-phy of language more generally.

Primary sourcesFrege, Gottlob (1892). ‘On sense and ref-

erence’. In P. Geach and M. Black (eds)(1980), Translations for the Philosoph-ical Writings of Gottlob Frege. Thirdedition. Oxford: Blackwell.

Quine, W. V. (1956). ‘Quantifiers andpropositional attitudes’. Journal of Phi-losophy 53, 177–87.

Further readingDavidson, Donald (1979). ‘Quotation’. In

Donald Davidson (2001), Inquiries intoTruth and Interpretation. Oxford:Oxford University Press.

Devitt, Michael and Kim Sterelny (1987).Language and Reality. Oxford: Black-well. 34–5.

Alex Barber

One is that the very label for the con-trast is misleading: to mention a wordoften is to use it, albeit to refer to itself.Another is that one can mention a wordwithout actually using it, as with (6):

(6) The five letter word that refers tothe capital of the largest French-speaking European country has twosyllables.

Since such cases do not give rise toconfusion, they are not what peoplewould have in mind as typical men-tionings, but mentionings they never-theless are. Yet another complicationis that non-mentioning uses of a wordhave little in common beyond beingnon-mentionings. They are certainlynot all cases of using the word to referto a standard referent. Think of ‘the’or ‘Santa Claus’, for example, orthe use of words to refer to a non- standard referent in malapropisms.

Anaphora – in which one word orexpression is referentially dependenton another – introduces yet morecomplexity. In (7), the final word isanaphoric to the fifth word, eventhough the fifth word refers to itselfwhile the final word refers to a city.With the ‘it’ in (8) the situation isreversed. Following W.V.O. Quine’swork on propositional attitudes*,example (9) is still more complex.The ‘so’ in ‘so-called’ refers to thename ‘Giorgione’ (which means ‘tallGeorge’) while ‘his’ refers to the Ren-aissance artist himself, making itunclear whether the name to whichboth are tied anaphorically should bein quotation marks or not.

Entries are indicated in bold.

acceptability/grammaticality, 1–2,105

acquisition (of language), 20, 49, 50,51–3, 71–3, 96, 111, 130, 157,231, 232–3, 237, 238, 249–50

adequacy, 2–3, 137adjacency pairs, 31, 34alethic, 165, 167, 241Althusser, Louis, 171, 222ambiguity/vagueness, 3–4, 98, 150analytic philosophy, 4–11, 126, 152,

153analytic/synthetic, 9, 11–14, 129,

163, 168, 241–2animals (communication of), 19,

27–8, 96anthropology, 35, 59, 66, 97, 152,

170, 189, 219, 221, 223, 238Aquinas, Thomas, 42arbitrariness, 27, 28, 114, 143, 144,

191, 207, 208, 220Aristotle, 25, 28, 42, 47, 101, 107,

122, 163, 213, 240, 245Arnauld, Antoine, 163Artificial Intelligence (AI), 14–17, 97,

107, 131, 142, 186Austin, J. L., 8, 79, 111, 129, 149,

150–6, 185, 199, 209, 213–16,244

Ayer, A. J., 68, 129, 180, 215, 243

Babbage, Charles, 15Bach, K., 128, 141, 156, 195Bakhtin, Mikhail, 115Bally, Charles, 113, 158Barthes, Roland, 115, 222Barwise, Jon, 209, 210, 211Beaugrande, Robert de, 60–1behaviourism, 17–20, 21, 93, 130Berkeley, George, 70, 199

Birdsong, D., 105Blakemore, Diane, 197blending, 23Bloomfield, Leonard, 18, 58, 60,

115, 130, 219, 223Boas, Franz, 58, 115, 189, 223Bourdieu, Pierre, 160, 172Bréal, Michel Jules Alfred, 206Brentano, Franz, 5, 100–1, 213–14Broca, Paul Pierre, 207Brown, Gillian, 32, 60–1Brown, Penelope, 158–60Brugmann, Karl, 57Bühler, Karl, 214Butler, Judith, 154Bybee, Joan, 50–1

Cameron, Deborah, 75–6Carnap, Rudolf, 8, 10, 12, 19,

128–9, 164, 166, 180Carston, Robyn, 195, 198Chase, Stuart, 118Chomsky, Noam, 1, 2–3, 19–20, 27,

30, 43–4, 49–50, 52, 58, 60, 65,71–3, 77, 78, 96–7, 103–6,110–11, 127, 130–1, 134–8, 163,180, 221, 223–4, 232–8, 247–51

cognitivism, 20, 21–4, 96cohesion, 60, 61compositionality, 24–5, 83, 142,

184, 203computational linguistics, 15concordance, 36, 41Condillac, Etienne Bonnot de, 206conditioning, 18–19, 130connotation/denotation, 25–6, 202continuity, 26–8conventional meaning, 28–9, 145,

196, 198conversation analysis, 29–36, 59

253

Index

Cook, Guy, 31Copenhagen Linguistic Circle, 79corpora, 23, 36–42, 106, 160corpus linguistics, 39, 231correspondence theory, 42–3, 240–3Coulthard, Malcolm, 60, 61creativity, 16, 27, 43–4, 130, 192(critical) discourse analysis, 58–65,

97, 99, 238

Davidson, Donald, 24, 209deconstruction, 44–7deduction/induction, 47–53definite descriptions, 6, 54–7, 127,

176, 184deixis, 205, 226, 227Dekker, P., 109Dennett, Daniel, 103, 112denotation see

connotation/denotationDerrida, Jacques, 44–7, 154, 171–2,

222Descartes, René, 5, 15, 44, 49, 70,

101, 130, 149descriptivism, 57–8, 83, 184, 236Devlin, Keith, 210Dewey, John, 242discourse completion tests, 159discourse markers, 34dispreferred response see

preferred/dispreferred responsesdistinctive features, 25, 65–6, 77Donnellan, Keith, 56Dressler, Wolfgang, 60–1Dreyfus, Hubert, 16Dummett, Michael, 9–10Durkheim, Émile, 114, 189

Eco, Umberto, 208edge feature, 137emic/etic, 66–7empiricism/rationalism, 9, 18, 43,

49–50, 52, 67–74, 130, 151, 242,249, 250

English for specific purposes (ESP),61–2

epistemology, 43, 67, 152, 163ethnography, 59, 60, 99etic see emic/etic

existentialism, 171extension, 26, 167, 202

Fairclough, Norman, 63feminism, 74–7, 161Firth, J. R., 36, 157, 225Fishman, Pamela, 74Fodor, Jerry, 90, 110–12, 180, 233Foucault, Michel, 172Francis, Nelson, 37Frege, Gottlob, 6, 9, 24, 54, 68,

123–6, 141, 143, 166, 173–4, 176,180, 184, 201–5, 211, 241, 243,245

Full Interpretation (principle of), 136

Gadamer, Hans-Georg, 210game theory, 107–8Garnfikel, Harold, 29generative grammar see

transformational generativegrammar

generative phonology, 65, 77–8generative semantics, 14, 21, 78–9,

156, 236genetics, 28, 96–7genre, 33, 34, 36, 59, 62glossematics, 79–81Goffman, Ervine, 158Gopnik, Alison, 52Gordon, Dan, 217government and binding theory, 135,

237–8grammaticality see

acceptability/grammaticalityGreen, Georgia, 79Grice, H. P., 28–9, 79, 84, 85,

86–92, 108–9, 111, 144–5, 158,159, 173, 193, 197–8, 209, 217

Halle, Morris, 65, 77Halliday, M. A. K., 58, 60–1,

225–31Harnish, R. M., 156Harris, Roy, 97–8Harris, Zellig, 60, 223, 234Hasan, Ruquaiya, 60–1, 225Hegel, G. W. F., 192

INDEX

254

Heidegger, Martin, 44Herder, Johann Gottfried, 117Hintikka, Jakko, 107Hjelmslev, Louis, 79–80, 207, 220Hockett, Charles, 18, 27, 58holism, 13, 81–4, 93, 129Horn, Laurence, 90–1Hornstein, Norbert, 209humanism, 171Humbolt, Wilhelm von, 44, 116–17Hume, David, 48–9, 68, 70,

199–200Husserl, Edmund, 100–1, 214, 215

iconicity, 21, 23, 208idealism, 5–6, 8ideational theories, 84–5, 98implicature, 86–92, 145, 158–9, 173,

195–6, 198, 216indeterminacy, 9, 92–5, 98–9, 100induction see deduction/inductioninnateness, 71, 96–7, 111, 238, 250insertion sequence, 31integrationism, 97–100intentionality, 100–3, 110, 112, 171,

214, 216international phonetic alphabet

(IPA), 67intuition, 1, 61, 68, 85, 103–6, 157,

232

Jackson, Frank, 199Jakobson, Roman, 65, 77, 115,

190–2, 221–2James, William, 242Jefferson, Gail, 30Jesperson, Otto, 158Johnson, Mark, 131–4Jones, Daniel, 156–7

Kamp, Hans, 15Kant, Immanuel, 5, 11–12, 42–3, 87,

180, 213, 241, 250Karmiloff-Smith, Annette, 53Kasher, Asa, 90Katz, J. J., 233, 235Kempson, Ruth, 241Knowles, Murray, 131, 134Kövecses, Zoltán, 131, 133–4

Kripke, Saul, 92–5, 143, 164, 166,168–9, 180, 184, 202

Kroeber, Alfred, 117Kruszewski, Nikolaj, 157Kucera, Henry, 37Kuhn, Thomas, 171

Labov, William, 58, 60Lacan, Jacques, 171, 189–92, 222Lakoff, George B., 21, 49, 79,

132–3, 154, 187, 217Lakoff, Robin, 74, 79, 158Langacker, Ronald, 21, 79, 187language games, 106–9language of thought (LOT), 81,

110–12langue/parole, 113–16, 219Leibniz, Gottfried Wilhelm, 70, 151,

164, 203, 241Lévi-Strauss, Claude, 219, 222Levinson, Stephen, 91, 158–60, 173,

177–8, 179Lewis, David, 29, 108, 164, 167–8,

211lexicon, 21, 28, 33, 123, 124, 237linguistic relativity, 116–19, 222linguistic variable, 24, 119–20literary criticism, 45, 170literature, 47, 189, 221, 222Locke, John, 70–1, 84–5, 199, 201,

206, 241logic, 6–10, 55, 107, 120–7, 139,

163, 167–9, 175–7, 203, 244–5logical atomism, 7logical form, 7, 55, 122, 127–8, 142,

195, 237logical positivism, 8, 12, 68, 73,

81–2, 128–30, 215Lorenzen, Paul, 107Lovelace, Ada, 15

McCarthy, John, 146McCawley, James, 78, 79, 167–8MacCorquodale, Kenneth, 20Martin, James, 62, 225–6Marx, Karl, 171Marxism, 115, 161, 171, 222mathematics, 2, 9, 55, 68, 70, 129,

151, 174

INDEX

255

Matthiessen, Christian, 226, 228, 229Mead, Margaret, 208Mendoza, San Francisco Ruiz de, 217mentalese, 110–11mentalism, 21, 130–1mention see use/mentionmerge operation, 135Merin, Arthur, 108metaphor, 16, 21–3, 43, 49, 109,

131–4, 188, 190, 191–2, 198, 231metonymy, 132, 190–2, 217Mill, J. S., 26, 68, 143, 202minimalism, 134–8, 237–8Mitchell, T. F., 60modality, 62, 138–9, 165, 228model theoretic semantics, 139–42Montague, Richard, 24, 139–40,

164, 185–6, 211, 233, 236Moon, Rosamund, 131, 134Moore, G. E., 6, 7, 10, 42, 88, 199morphology, 13, 38, 58, 139Morris, Charles, 208

names, 25–6, 54, 142–4, 168, 176,183–4, 202–3, 214

narrative, 33, 63natural language processing (NLP),

15, 39–41negative evidence, 96, 104Neurath, Otto, 128Nicole, Pierre, 163Nietzsche, Friedrich, 44nonnatural meaning, 29, 144–5noun phrase, 25, 136–7, 176, 233–5,

250

Ockham, William of, 26, 42openings and closings, 30optimality theory, 65, 77–8, 109,

145–9ordinary language philosophy (OLP),

8, 55, 149–55, 215orthography, 35, 38

Pânini, 120Panther, Klause, 217parameters, 50, 108, 211, 237, 250parole see langue/parolePascal, Blaise, 15

Peirce, Charles Sanders, 5–6, 49,116, 144, 207–8, 242, 246

performatives, 152, 154, 155–6,214–16

Perry, John, 209–10phase, 137phenomenology, 101phoneme, 65, 156–7, 220, 222, 224,

233, 246phonology, 23, 38, 58, 65–6, 77–8,

113, 115, 147, 148, 220, 222,238

Piaget, Jean, 53Pike, Kenneth, 58, 66–7, 115Plato, 5, 7, 28, 46, 96, 143, 151, 192politeness, 90, 157–61political correctness, 161–2Popper, Karl, 43, 49, 52Port Royal Logic, 110, 163–4possible world semantics, 23, 164–9,

176, 202, 211Postal, Paul, 78–9, 235postmodernism, 47poststructuralism, 47, 75, 170–3pragmatics, 3, 22, 28–9, 59–60, 78,

79, 91, 109, 113, 157, 177, 181,193, 195, 197–9, 208, 217, 228,236, 252

Prague Linguistics Circle (PragueSchool), 65, 115, 221

preferred/dispreferred responses, 31prescriptivism, 1, 57, 154presupposition, 55, 89, 173–9, 241,

245Prince, Alan, 146private language, 85, 111, 179–80,

222propositional attitudes, 82–3, 110,

112, 177, 181–5, 204, 252propositions, 4, 48, 55, 81, 95, 108,

122, 155, 163, 165, 167–8, 174,176, 182–3, 184, 185–6, 215, 242

prototype, 21, 79, 186–9psychoanalysis, 171, 189–93, 222psychology, 13, 18, 19, 20, 21, 30,

48–9, 66, 71, 72, 81, 85, 96, 104,112, 130–1, 145, 152, 157, 170,187, 206, 214, 238

Putnam, Hilary, 85

INDEX

256

Quine, W. V. O., 8–9, 11–12, 13–14,19, 25, 92–3, 95, 129, 243–4, 252

Ramsey, Frank, 243–4Rask, Rasmus, 57rationalism see

empiricism/rationalismrealism, 6, 152, 168, 241, 242recursion, 2, 96reference see sense/referenceregister, 1, 59, 62Reid, Thomas, 213, 214Reinach, Adolf, 214–15reinforcement, 19, 20, 63relative clause, 2, 163relevance theory (RT), 91, 193–9repair, 30, 32, 33, 34response see stimulus/responserevised extended standard theory,

135, 236Ribot, Théodule-Armand, 206–7Rorty, Richard, 243Rosch, Eleanor, 22, 79, 154, 186–7Ross, John (Haj), 22, 78, 79Russell, Bertrand, 6–7, 9, 10, 15, 42,

49, 54–6, 123, 126, 127, 143,150, 152–3, 163, 174, 175–6,180, 183–4, 199–200, 215, 240–1

Ryle, Gilbert, 8, 19, 149, 150, 215

Sacks, Harvey, 29–32Sadock, Jerrold, 217St Anselm, 26Sampson, Geoffrey, 52Sapir-Whorf hypothesis, 52, 116,

117, 223Sapir, Edward, 58, 66, 115, 116,

117, 118, 130Saussure, Ferdinand de, 26, 28, 46,

58, 59, 97, 98, 113–16, 144, 170,189–92, 205–7, 219, 221–2, 224,226

Schegloff, Emanuel, 30–1, 33Schiffer, Stephen, 56, 145Schiffrin, Deborah, 33–4, 63Schlick, Moritz, 128Schütze, C. T., 2, 105–6

science, 9, 12–14, 18, 20, 43, 47, 52,73, 79, 81, 97, 133, 171, 174,206, 221, 233

Searle, John, 15–16, 79, 101–3,111–12, 143–5, 150, 153–4,155–6, 158, 185, 202, 216–17

Sechehaye, Albert, 113Sellars, Wilfred, 199–200semantics, 13–14, 21, 23, 26, 58, 78,

95, 97, 107, 109, 113, 126, 131,138, 139–42, 164–9, 176, 181,182, 185, 197–8, 201, 203, 206–8,209–12, 220, 222, 224, 227, 231,233, 236, 241, 245

sense data, 199–200, 152sense/reference, 200–5sequence pairs, 31signs and semiotics, 28, 98–9,

114–15, 116, 142–4, 201, 205–9,215

Sinclair, John, 60situational semantics, 209–12Skinner, B. F., 19–20, 130Smolensky, Paul, 146sociolinguistics, 24, 58, 60–1, 76, 78,

97, 99, 120, 157, 231, 238sociology, 35, 45, 59, 114, 152, 158,

189, 238Socrates, 13, 28, 46, 151speech act theory, 59, 60, 86–9, 91,

97, 101–2, 109, 112, 129, 152–3,155, 158–9, 177–8, 185, 196,212–19, 244

Sperber, Dan, 90, 91, 193–9Spinoza, Benedictus de, 70stimulus/response, 18–20, 43–4,

49–50, 52, 70–1, 96, 195, 249Strawson, P. F., 8, 55–6, 145, 149,

152, 174–6, 202, 209, 243, 245structuralism, 113, 170–1, 219–25Stubbs, Michael, 60–1Swales, John, 62synecdoche, 132syntax, 2, 13, 38, 58–9, 79, 97, 111,

113, 123, 126, 127, 141–2, 146,163, 208, 231, 233, 236–7, 249

synthetic see analytic/syntheticsystemic-functional grammar (SFG),

62, 98, 225–32

INDEX

257

tagging, 37–9tagmemics, 66Tain, Hippolite, 206Talmy, Len, 21Tannen, Deborah, 75Tarski, Alfred, 42–3, 139–40, 185,

211, 243–4tense, 50–1, 62, 119, 136–7, 142,

166Thornburg, Linda, 217Thorndike, Edward, 130Titchener, Edward Bradford, 130token see type/tokenTomasello, Michael, 51–2Toolan, Michael, 99–100transcription, 34–5, 38, 61transformational-generative

grammar (TGG), 1, 2–3, 21,58, 77, 78–9, 111, 135, 232–8

translation, 9, 15, 39–40, 81–2,92–3, 118, 143

Trubetzkoy, N. S., 65, 77truth theories, 43, 238–44truth values, 56, 120–1, 125–6,

140–1, 164–8, 174–6, 183–4,203–4, 244–6

Turing, Alan, 15turn taking, 31–3type/token, 49, 51, 110, 240,

246–7

Uldall, Hans Jørgen, 79universal grammar (UG), 3, 44,

49–50, 77, 96, 247–51use/mention, 251–2

vagueness see ambiguity/vaguenessvan Dijk, Teun, 63Van Rooy, R., 109verification, 8, 68, 73, 129, 215Vienna Circle, 8, 9, 128–9, 215

Waismann, Friedrich, 128Watson, John B., 18, 130Weizenbaum, Joseph, 15Whitney, William Dwight, 206–7Whorf, Benjamin Lee, 116–19, 130,

157, 161Wilson, Deirdre, 91, 193–8, 241Wisdom, John, 149Wittgenstein, Ludwig, 7–8, 10, 19,

42, 85, 93, 107, 111, 149–50,152–4, 179–80, 188, 209, 215,241

Wodak, Ruth, 63World War I, 149, 221World War II, 149, 170

Yule, George, 32, 60–1

Zadeh, Lotfi, 79Žižek, Slavoj, 192

INDEX

258

ERRATUM The publisher and editors of Key Ideas in Linguistics and the Philosophy of Language wish to apologise for the errors of fact which they mistakenly introduced into the entry on Innateness in this volume, errors which were neither originated nor seen by the author. The corrected and author-approved entry is printed here. Innateness The claim that some aspects of linguistic competence are genetically specified rather than learnt through experience. This claim has been driving research in generative linguistics and language acquisition since the late 1950s. See also: Acceptability/Grammaticality; Continuity; Mentalism; Transformational-Generative Grammar; Universal Grammar Key Thinkers: Chomsky, Noam; Descartes, René; Plato Noam Chomsky proposed that humans possess domain- and species-specific knowledge of the structure of possible languages, which enables children to acquire language with the speed, efficiency and uniformity that they do. This view can be traced back to Platonic philosophy and Cartesian cognitivism. Opponents of this view claim that language acquisition is innately constrained but only by the same mechanisms that underlie general cognitive ability. In other words, although it is uncontroversial that linguistic development is innately constrained, exactly what is innate is still a matter of debate. The argument that has most forcefully been used in support of the position that we are born with innate knowledge of linguistic constraints and principles is the poverty of the stimulus argument (Chomsky 1980: 34):

1. Language is a complex system that could only be acquired through experience if negative evidence was available (that is, information about what sequences are grammatically illicit).

2. Children only ever have access to positive evidence (information about grammatical sentences).

3. Despite this, children successfully acquire language. If one accepts these premises, one has to conclude that humans are genetically hard-wired for language: children are born equipped with Universal Grammar (Chomsky 1981), containing information about universal linguistic principles which enable children to form specific hypotheses about the structure of the language they are learning. The proposed existence of a critical period for language (i.e. typical language acquisition is not possible after a certain age) and studies of deaf children who spontaneously develop sign language have been used as further evidence in support of linguistic nativism. Opponents of linguistic nativism claim that the richness of empirical data actually available to children is vastly underestimated; that general cognitive ability can explain how language is learnt from experience; and that the poverty of the stimulus argument holds only given the generative definition of what language is. The most articulated alternative to the nativist perspective has been provided within connectionist psychology (Elman et al. 1996), where artificial neural networks are trained to reproduce complex linguistic behaviour based solely on past experience with relevant data.

Research on innateness has spurred comparative research with animals in an effort to identify exactly what aspect of linguistic knowledge could be species- and domain-specific. Parallel research in genetics also attempts to specify the genetic basis of language. In 2001, Lai and colleagues suggested that a mutation in the FOXP2 gene is causally involved in language disorders. Despite widespread enthusiasm in the popular press about the ‘language gene’, the exact role of this gene in relation to language development is far from clear. Although the innateness question is still unresolved, interdisciplinary research that straddles the gap between linguistics, anthropology, artificial intelligence, genetics and neuroscience has opened up novel, exciting ways in which the question can be empirically addressed. Primary sources Chomsky, N. (1980). Rules and Representations. Oxford: Blackwell. Chomsky, N. (1981). Lectures on Government and Binding: The Pisa Lectures.

Holland: Foris Publications. Further Reading Elman, J. L., E. A. Bates, M. H. Johnson, A. Karmiloff-Smith, D. Parisi and K.

Plunkett (1996). Rethinking Innateness: A connectionist perspective on development. Cambridge, MA: MIT Press.

Hauser, M. D., N. Chomsky and W. T. Fitch (2002). ‘The faculty of language: What is it, who has it, and how did it evolve?’ Science 298, 1569–79.

Lai, C. S. L., S. E. Fisher, J. A. Hurst, F. Vargha-Khadem and A. P. Monaco (2001). ‘A forkhead-domain gene is mutated in severe speech and language’. Nature 413: 519–23.

Stavroula-Thaleia Kousta