LINGUISTICS INSIDE OUT
AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE General Editor E.F. KONRAD K...
42 downloads
1098 Views
19MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
LINGUISTICS INSIDE OUT
AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE General Editor E.F. KONRAD KOERNER (University of Ottawa) Series IV - CURRENT ISSUES IN LINGUISTIC THEORY
Advisory Editorial Board Henning Andersen (Los Angeles); Raimo Anttila (Los Angeles) Thomas V. Gamkrelidze (Tbilisi); John E. Joseph (Edinburgh) Hans-Heinrich Lieb (Berlin); Ernst Pulgram (Ann Arbor, Mich.) E. Wyn Roberts (Vancouver, B.C.); Danny Steinberg (Tokyo)
Volume 148
George Wolf and Nigel Love (eds) Linguistics Inside Out Roy Harris and his critics
LINGUISTICS INSIDE OUT ROY HARRIS AND HIS CRITICS Edited by
GEORGE WOLF University of New Orleans
NIGEL LOVE University of Cape Town
JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM/PHILADELPHIA
The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences — Permanence of Paper for Printed Library Materials, ANSI Z39.48-1984.
Library of Congress Cataloging-in-Publication Data Linguistics inside out : Roy Harris and his critics / edited by George Wolf, Nigel Love. p. cm. — (Amsterdam studies in the theory and history of linguistic science. Series IV, Current issues in linguistic theory, ISSN 0304-0763 ; v. 148) Includes bibliographical references and index. 1. Harris, Roy, 1931- . 2. Linguistics. I. Harris, Roy, 1931- . II. Wolf, George, 1950 - . III. Love, Nigel. IV. Series. P85.H34L56 1997 410-dc21 96-52209 ISBN 90 272 3652 6 (Eur.) / 1-55619-863-9 (US) (alk. paper) CIP © Copyright 1997 - John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O.Box 75577 · 1070 AN Amsterdam · The Netherlands John Benjamins North America · P.O.Box 27519 · Philadelphia PA 19118-0519 · USA
ψ
Roy Harris
[He] has many of the traits of the best English pamphleteers: courage, an individual mind, vehement opinions, an instinct for stirring up trouble, the arts of appealing to that imaginary creature the sensible man and of combining original observation with sweeping generalisation, of seeing enemies everywhere and despising all of them... he writes a lucid conversational style which wakes one up suddenly like cold water dashed in the face. The sting of it is sometimes refreshing; sometimes it makes one very angry... in the name of common sense [he] is capable of exaggerating with the simplicity and innocence of a savage... - V.S. Pritchett, on George Orwell
... the fact that missionary linguists sometimes end up in tribal cooking pots is no more surprising than the fact that even well trained troops sometimes end up as casualties. - Roy Harris
Contents Preface Contributors Roy Harris: Publications 1956-1995 Prologue 1 The "Language Myth" Myth: Roy Harris's Red Herrings JOHN E. JOSEPH
ix xi xv 1 9
2 The Language Muddle: Roy Harris and Generative Grammar ROBERT D. BORSLEY and FREDERICK J. NEWMEYER
42
3 Telementation and Generative Linguistics PHILIP CARR
65
4 Phonography: Setting a Term to the Evolution of Writing JOHN SOREN PETTERSSON
84
5 A New Mentality DAVID R. OLSON
99
6 Science and Significance: Making Sense of Wittgenstein's Ways of Seeing ANTHONY HOLIDAY
106
7 Rules and Algorithms: Wittgenstein on Language ROM HARRÉ
136
8 Contextualizing "Context": From Malinowski to Machine Translation NAOMI S. BARON
151
9 Is Ethnomethodological Conversation Analysis an "Integrational" Account of Language? DAVID FLEMING
182
10 Linguistic Theory and the Multiple-Trace Model of Memory JOHN R. TAYLOR
208
11 Language, Art and Kant TREVOR PATEMAN
226
12 From an Integrational Point of View ROY HARRIS
Epilogue References Index
229
311 319 339
Preface This book started life, or gestation, as a traditional Festschrift. For a number of reasons that project was soon abandoned. In the first place, the traditional Festschrift, compiled as it apparently so often is by inviting a random selection of the schriftee's friends and colleagues to send in whatever they happen to have handy, is all too likely to consist of an assortment of pieces with no better reason to appear bound together in a single volume than their authors' wish to be associated with it. Surely something more useful, and more appropriate, could be contrived. Instead we decided on a collection of critical essays. Contributors were asked to focus on (an aspect of) Roy's work, and either develop or rebut some specific point he has made, or else endorse or reject his whole approach to a given area of linguistic inquiry; Roy himself was invited to write a composite reply. The idea was to advance the debate on the issues he has been concerned with. Books of this kind are perhaps commoner in philosophy than in linguistics, and there are plenty of precedents there for what we had in mind: collections of essays that mark a scholar's achievement by discussing the work itself, and are to that extent worthwhile irrespective of any celebratory intention, as opposed to the sort of unfocusedly encomiastic production that lionises its hero by presenting a miscellany of articles some of which may have no point of contact with what was supposed to make such a book worth publishing in the first place. A collection such as the following is appropriate because, whether he would answer to the name of philosopher or not, Roy is best known for the many controversial things he has said not so much within linguistics as about it. For reasons to do with the way it has come to be institutionalised, academic linguistics is on the whole unfriendly to controversialists as wholehearted and thoroughgoing as he has never flinched from being. Having issued our invitation to potential contributors, we were therefore confident that there would be a great deal more rejecting than endorsing in the eventual book. So confident were we of this that the working title was A Blastschrifl for Roy Harris. We sat back waiting for adverse, even hostile commentary. On one level, a Blastschrift is an antiFestschrift. But a Blastschrift is not as backhanded an honour as it might at first appear: on another level it pays its anti-honorand the compliment of taking his work seriously enough to subject it to criticism.
χ
LINGUISTICS INSIDE OUT
In any case, given the above considerations, Roy Harris would be a hard man to honour with a Festschrift of the traditional, more unctuous kind. Outside the ranks of what one contributor calls his "cadre of talented and devoted students", most linguists are inclined to give short shrift to his ideas. The resulting volume would have been slim. A Shortschrift, in fact. In the event not all the contributions proved to be blasts, so the working title was dropped. But all make contact in one way or another with Roy's work. And Roy in his reply has something to say about all of them. In that sense his reply comprehends them all. Whether it comprehends them all in the other sense is something we originally envisaged taking it upon ourselves to pass judgement on, in a long editorial epilogue. In the end, to avoid trying the reader's patience, not to mention usurping his rightful role, we settled for drawing attention to one or two of what strike us as the more salient points arising from the debate. The sandwich is completed by a brief scene-setting prologue.
Cape Town New Orleans
N.L. G.W.
Contributors Naomi S. Baron is Professor of Linguistics and Director of the Summer TESOL Institute at the American University in Washington D.C. Former Guggenheim Fellow and past president of the Semiotic Society of America, she has taught at Brown University, Emory University and Southwestern University. Her research focuses on language acquisition in children, language in social context, the evolution of writing, and the historical development of English. Author of five books, including Growing Up With Language: How Children Learn to Talk, Computer Languages: A Guide for the Perplexed and Speech, Writing and Sign, Professor Baron is currently writing a book on the effects of technology on spoken and written language. Bob Borsley is Senior Lecturer in the Linguistics Department of the University of Wales, Bangor, U.K. His main interest is syntactic theory, and he has written extensively on English, Welsh, Breton and Polish syntax. He is author of a syntax textbook: Syntactic Theory: A Unified Approach (Edward Arnold). He is also co-editor ofJournal of Linguistics. Philip Carr is Senior Lecturer in Linguistics in the Department of English Literary and Linguistic Studies at Newcastle University. His interests lie in the areas of phonological theory and the philosophy of linguistics. His book Linguistic Realities (Cambridge University Press, 1990) addresses issues in the philosophy of science as they relate to generative linguistics, and issues concerning the ontological status of linguistic objects. His textbook Phonology (Macmillan, 1993) is an introduction to current work in generative phonology. He is currently working on an undergraduate textbook on English phonetics and phonology (to be published by Blackwell) and is co-writing a book with N. Burton-Roberts on I-language and the ontology of phonology. David Fleming is a doctoral student in rhetoric in the English Department at Carnegie Mellon University in Pittsburgh, Pennsylvania. His dissertation examines the role of discourse (narrative, argument, conversation and written text) in the practical and social activity of student designers. His background includes work in literary studies, composition and rhetorical theory.
Xii
LINGUISTICS INSIDE OUT
Rom Harré is Emeritus Fellow of Linacre College in Oxford University. He is Professor in the Department of Psychology at Georgetown University, Washington D.C., and Adjunct Professor of Social and Behavioral Sciences at Binghamton University. Roy Harris, Honorary Fellow of St Edmund Hall, Oxford, was born in 1931 and educated at Queen Elizabeth's Hospital, Bristol. After reading Modern Languages at Oxford, he pursued research in comparative Romance linguistics there, and subsequently in general linguistics at the School of Oriental and African Studies, University of London. He was successively Professor of the Romance Languages (1976-7) and of General Linguistics (1978-88) at Oxford, and Professor of English Language (1988-91) at the University of Hong Kong. He has also taught in the United States and France. He is the founder-editor of the journal Language and Communication. Anthony Holiday teaches philosophy in the School of Government at the University of the Western Cape, South Africa. His book Moral Powers (Routledge, 1988) attempts to derive an ethical theory from a reflection on the relationship between Wittgenstein's philosophy of language and Marxist philosophy of history. He has published papers on Wittgenstein's thought and on philosophical semantics, and has a particular interest in the implications of political theory for moral philosophy. John E. Joseph succeeded Roy Harris as Professor of English Language and General Linguistics and Head of the Department of English at the University of Hong Kong. His publications cover a range of topics in language standardisation, Romance historical syntax and the history of linguistics, and include the books Eloquence and Power (1987), Ideologies of Language (coedited with T.J. Taylor, 1990), and Linguistic Theory and Grammatical Description (co-edited with F.G. Droste, 1991). He has recently completed a new translation of Plato's Cratylus, to be published in conjunction with a book situating Plato's theory of language within the broader framework of Western linguistic thought. Nigel Love read Modern Languages at Magdalen College, Oxford, and subsequently held a junior research fellowship at Wolfson College. He has taught English in France and French in Jamaica, and is currently Associate Professor and Head of the Department of Linguistics at the University of Cape Town. Topics he has written on include Welsh phonology, the work of J.L. Austin, and
CONTRIBUTORS
Xiii
the linguistic theory implied by procedures of statutory interpretation. He is editor of the journal Language Sciences. Frederick J. Newmeyer is Professor of Linguistics at the University of Washington. He is the author of English Aspectual Verbs (Mouton, 1975), Linguistic Theory in America (Academic Press, 1980), Grammatical Theory (University of Chicago Press, 1983), The Politics of Linguistics (University of Chicago Press, 1986) and Generative Linguistics: An Historical Perspective (Routledge, 1995). He edited the four-volume Linguistics: The Cambridge Survey (Cambridge University Press, 1988). From 1989 to 1994 he was Secretary-Treasurer of the Linguistic Society of America. David R. Olson is Professor at the Ontario Institute for Studies in Education. He has held Fellowships at Harvard's Center for Cognitive Studies and Stanford's Center for Advanced Studies in the Behavioral Sciences. He is a Fellow of the Royal Society of Canada and holds an honorary doctorate from the University of Gothenburg, Sweden. His latest book, The World on Paper: The Conceptual and Cognitive Implications of Writing and Reading was published by Cambridge University Press in 1994. Trevor Pateman was born in 1947, studied philosophy, politics and economics at Oxford and since 1979 has taught at the University of Sussex, where he is now Reader in Education. His essay "Communicating with Computer Programs" appeared as the first article in the first issue (1981) of Language and Communication. His defence of Chomsky an linguistics in Language in Mind and Language in Society (Oxford University Press, 1987) includes an engagement with Roy Harris's work. John Soren Pettersson was born in 1961 and holds a BA in mathematics as well as a licentiate in general linguistics from Uppsala University, his thesis for which was a critique of evolutionary accounts of writing. He has also written on the origin of writing, writing and phonetic concepts, learning to read and write, and numerical notations. John R. Taylor has taught in Germany and South Africa, and is currently in the Linguistics Section, University of Otago, New Zealand. He is author of Linguistic Categorization (Oxford University Press, 1989), and has published widely on topics in semantics and cognitive linguistics.
XiV
LINGUISTICS INSIDE OUT
George Wolf took his B.A. in French at Columbia University, and his doctorate in General Linguistics at Oxford. He is currently Associate Professor of French at the University of New Orleans. He has edited and translated a collection of essays by the French linguist Michel Bréal, on whose concept of semantics he is now writing a book, has edited New Departures in Linguistics (Garland, 1992), and is collaborating with Eisuke Komatsu on an edition and translation of student notes to the first and second of Saussure's courses on general linguistics (Pergamon, 1996, and forthcoming).
Roy Harris: Publications 1956-1995 1956 "The White Stag in Chretien's Erec et Enide ". French Studies 10/1.55-61. 1957 "Et Liconaus ot non ses prre". Medium Aevum 26/1.32-5. "A Terminus A Quo for the Roman de Thebes ". French Studies 11/3.201-13. 1958 Review of W. Ziltener, Chrétien und die Aeneis. Medium Aevum 27/3.186-9. 1960 Review of G. Tilander, Mélanges d'étymologie cynégétique. Modern Language Review 55/3.439-40. 1961 Review of F.W. Locke, The Quest for the Holy Grail Medium Aevum 30/3.186-8. Review of H. Adolf, Visio Pacis. Medium Aevum 30/3.188-9. Review of L. Thorpe (ed.), Le Roman de Laurin. French Studies 15/4.360-1. 1962 Review of C. Grassi, Correnti e contrasti di lingua e cultura nelle valli cisalpine di parlata provenzale e franco-provenzale. Modern Language Review 57/1.115-6. Review of C.W. Dunn, The Foundling and the Werewolf. Medium Aevum 31/1.70-2.
XVi
LINGUISTICS INSIDE OUT
1964 Review of R.A. Haadsma and J. Nuchelmans, Précis de latin vulgaire. Modern Language Review 59/3.473. 1965 Review of S. Heinimann, Das Abstraktum in der franzosischen Literatursprache des Mittelalters. Medium Aevum, 34/1.83. Review of K. Urwin (ed.), A Short Old French Dictionary for Students. Medium Aevum 34/1.83-4. 1966 "Criteria in Descriptive Semantics". Theoria 32/3.237-9. Review of J.A. Fodor and J.J. Katz (eds.), The Structure of Language. Mind 75/299.450-1. "Gallo-Romance Third Declension Plurals". Revue de linguistique romane 30/117-18.58-70. "Chansons de geste". Chambers's Encyclopedia, New Revised Ed., 3.270-1. Oxford. "Chrétien de Troyes". Chambers's Encyclopedia, New Revised Ed., 3.525. Oxford. "Roman courtois". Chambers's Encyclopedia, New Revised Ed., 11.757. Oxford. "François Villon". Chambers's Encyclopedia, New Revised Ed., 14.306-7. Oxford. 1967 Review of S. Cigada, La leggenda medievale del Cervo Bianco e le origini della 'matière de Bretagne '. French Studies 21/1.50-1. "The Semantics of Self-Description". Analysis 27/4.142. Review of D'A.S. Avalle, Latino 'circum romançum ' e 'rustica romana lingua'. Medium Aevum 36/1.52-4. "Piedmontese Influence on Valdôtain Syntax". Revue de linguistique romane 31/121-2.180-9. Review of G. Moignet, Le pronom personnel français. French Studies 31/2.182-3.
ROY HARRIS: PUBLICATIONS 1956-1995
XVii
Review of A.M. Colby, The Portrait in Twelfth-Century French Literature. Medium Aevum 36/2.183-5. 1968 Review of R. de Dardel, Recherches sur le genre roman des substantifs de la troisième declinaison. Modem Language Review 63/1.239-40. "La structure des paradigmes en latin vulgaire". Actas del XI Congreso Internacional de Linguistica y Filologia Romanicas, 1.391-8. Madrid. Review of L.C. Porter (éd.), JaquesPeletier du Mans, Dialogue de VOrtografe e Prononciacion Françoese. French Studies 22/2.146-7. Review of P. Porteau, Deux études de sémantique française. Modem Language Review 63/3.699-700. "Translation Propositions". Linguistica Antwerpiensia 2.217-27. Review of R. Posner, The Romance Languages. Modem Language Review 63/3.698-9. "Notes on a Problem of Franco-Provençal Morphology". Zeitschrift fur romanische Philologie 84/5-6.572-81. Review of W. Meyer-Lubke, Historische Grammatik der franzôsischen Sprache. Zweiter Teil: Wortbildungslehre. French Studies 22/3.266-7. "Self-Description and the Theory of Types", Analysis 28/6.207-8. 1969 "Pronominal Postposition in Valdôtain". Revue de linguistique romane 33/12930.133-43. "Type-Meanings and Token-Meanings". Journal of Linguistics 5/1.143-4. Review of H. Rheinfelder, Altfranzosische Grammatik, Zweiter Teil: Formenlehre. French Studies 23/3.273. 1970 "Semantics and Translation". Actes du Xe Congrès International des Linguistes, 2.461-7. Bucarest: Editions de l'Académie de la République Socialiste de Roumanie. "Deviance and Citation". Journal of Linguistics 6/2.253 -6. "The Strasburg Oaths: a Problem of Orthographic Interpretation". Revue de linguistique romane 34/135-6.403-6.
XV111
LINGUISTICS INSIDE OUT
1972 Review of V. Introduzione al latino volgare. Medium Aevum 41/1.81-2. "Translation into Martian". Mwrf 81/322.276. Review of K. Bal dinger (éd.), Dictionnaire étymologique de l'ancien français, Fascicule Gl. Medium Aevum 41/3.294-5. "Performative Paradigms". Transactions of the Philological Society 70.44-58. "Words and Word Criteria in French". History and Structure of French: Essays in Honour of Professor T.B.W. Reid, ed. by F J. Barnett et al., 117-33. Oxford: Blackwell. 1973 Review of W.L. Chafe, Meaning and the Structure of Language. Journal of Linguistics 9/1.115-20. Synonymy and Linguistic Analysis. Oxford: Blackwell. 1974 Review of D D R . Owen (ed.), Arthurian Romance. French Studies 28/1.52-3. (with N. Love) "A Note on French Nasal Vowels". Linguistics 126.63-8. 1976 Review of G. Lavis, L 'expression de l'affectivité dans la poésie lyrique française du moyen âge. Medium Aevum 45/1.114-7. "Early Generative Semantics". Journal of Literary Semantics 5/2.78-90. 1977 "Semantics, Performatives and Truth". Journal of Literary Semantics 6/2.63-75. On the Possibility of Linguistic Change. Oxford: Clarendon Press. 1978 Communication and Language. Oxford, Clarendon Press. Review of B. Comrie, Aspect. Journal of Literary Semantics 7/2.114-7. "Word Wars: On Thought and Language". Encounter 50/3.42-9.
ROY HARRIS: PUBLICATIONS 1956-1995
XIX
"The Descriptive Interpretation of Performative Utterances". Journal of Linguistics 14/2.309-10. Review of G.A. Miller and P.N. Johnson-Laird, Language and Perception. Journal of Linguistics 14/2.342-7. 1979 Review of G. Cannon, An Integrated Transformational Grammar of the English Language. Review of English Studies 30/119.324-5. Review of P. Trudgill (éd.), Sociolinguistic Patterns in British English. Review of English Studies 30/120.447-9. 1980 The Language-Makers. London: Duckworth. "The Englishness of English". Review of S. Greenbaum, G. Leech & J. Svartvik (eds.), Studies in English Linguistics for Randolph Quirk. London Review of Books, 6.11.80, 21-2. 1981 "Communication About Communication". Language & Communication 1/1.1-2. "The Need for Glottotherapy". Review of D. Bolinger, Language - The Loaded Weapon. Times Literary Supplement, 30.1.81, 117. "A Passion for Words". Review of D. Crystal (éd.), Eric Partridge: In His Own Words. Times Literary Supplement, 30.1.81, 302. "Archisemiotics". Review of G. Broadbent, R. Bunt and C. Jencks (eds.), Signs, Symbols and Architecture. Design Studies, April 1981, 122-3. "The Dialect of Fleet Street". Review of K. Waterhouse, Daily Mirror Style. Times Literary Supplement, 22.5.81, 559-60. The Language Myth. London: Duckworth. "Scoring the Language Game". Review of J. Lyons, Language and Linguistics: An Introduction. London Review of Books, 15.10.81, 13-14. "Truth-Conditional Semantics and Natural Languages". Language, Meaning and Style. Essays in Memory of Stephen Ullmann, ed. by T.E. Hope et al., 21-37. Leeds: Leeds University Press. "Performing in Words". Review of E. Goffman, Forms of Talk. Times Literary Supplement, 18.12.81, 1455-6.
XX
LINGUISTICS INSIDE OUT
1982 "Something Every Speaker Has". Review of J.C. Wells, Accents of English. Times Educational Supplement, 14.5.82, 26. "The Exchange of Signs". Review of R.E. Innis, Karl Buhler: Semiotic Foundations of Language Theory. Times Literary Supplement, 18.6.82, 669. "From Adam to Aarsleff '. Review of H. Aarsleff, From Locke to Saussure. London Review of Books, 19.8.82, 11-12. "The History Men". Review of R.W. Burchfield (éd.), A Supplement to the Oxford English Dictionary: Volume 3. Times Literary Supplement, 3.9.82, 935-6. 1983 "Means of Communication". Review of J.J. Gumperz, Discourse Strategies. Times Literary Supplement, 14.1.83, 37. "Semantics". Review of J. Aitchison, Language Change: Progress or Decay?', R. Chapman, The Language of English Literature ; R.K. Fenn, Liturgies and Trials', J. Lyons, Language, Meaning and Context. Times Educational Supplement, 14.1.83, 24. "Beware of Standard(s)". Times Educational Supplement, 25.2.83, 2. "Putting the Users First". Review of D. Leith, A Social History of English. Times Educational Supplement, 8.4.83, 21. "Power and the Power of Speech". Review of P. Bourdieu, Ce que parler veut dire. Times Literary Supplement, 20.5.83, 524. "Language and Speech". Approaches to Language, ed. by R. Harris, 1-15. Oxford: Pergamon. F. de Saussure: Course in General Linguistics. London: Duckworth. "Literary Translating: Theoretical Ideas". Times Literary Supplement, 14.10.83, 1119. "The Speech-Communication Model in 20th-century Linguistics and its Sources". Proceedings of the XIIIth International Congress of Linguists, ed. by S. Hartori and K. Inoue, 864-9. Tokyo. "All My Eye and Betty Martin". Review of L.G. Pine, A Dictionary of Mottoes', J. Green, Newspeak: a Dictionary of Jargon; E.S.C. Weiner, The Oxford Miniguide to English Usage; A.P. Cowrie, R. Mackin and I.R. McCaig, The Oxford Dictionary of Current Idiomatic English, vol.II; Κ. Hudson, A Dictionary of the Teenage Revolution and its Aftermath; E. Partridge, A Dictionary of Catch-Phrases. London Review of Books, 1.12.83, 12-13.
ROY HARRIS: PUBLICATIONS 1956-1995
XXi
1984 "The Misunderstanding of Newspeak". Times Literary Supplement, 6.1.84, 17. Review of J. MacNamara, Names for Things. Language Sciences 6/1.173-7. "Speaking Silently". Review of M. Deuehar, British Sign Language. Times Literary Supplement, 20.7.84, 807. "Soft Sell and Soft Soap". Review of D. Crystal, Who Cares About English Usage? and W.F. Bolton, The Language of 1984. Times Literary Supplement, 3.8.84, 859. "The View from the Blackboard". Review of M. Meek and J. Miller (eds.), Changing English. Times Literary Supplement, 14.9.84, 2031. "The Semiology of Textualization". Language Sciences 6/2.271-86. "What Was Meant By What Was Said". Review of G.P. Baker and P.M.S. Hacker, Language, Sense and Nonsense. London Review of Books, 20.9.84, 18-19. "Must Monkeys Mean?". The Meaning of Primate Signals, ed. by R. Harré and V. Reynolds, 116-37. Cambridge: Cambridge University Press. 1985 "Mizzled". Review οf Longman Dictionary of the English Language, L. Heller, A. Humez and M. Dror, The Private Lives of English Words; B. Bryson, The Penguin Dictionary of Troublesome Words; J. Shipley, The Origins of English Words; E. Partridge, A Dictionary of Slang and Unconventional English, ed. P. Beale. London Review of Books, 21.2.85, 18-19. (with I. Griffiths) "The Semiotics of Mental Representation". Semiotica 53.179-214. (with Ch.-J.N. Bailey) "On Developmentalism". Developmental Mechanisms of Language, ed. by Ch.-J.N. Bailey and R. Harris, vii-xii. Oxford: Pergamon. "Saussure and the Dynamic Paradigm". Developmental Mechanisms of Language, ed. by Ch.-J.N. Bailey and R. Harris, 167-183. Oxford: Pergamon. 1986 The Origin of Writing. London: Duckworth. "Sherlock Holmes Meets the Semioticians". Review of U. Eco and T.A. Sebeok (eds.), The Sign of Three. Language Sciences 8/2.197-201.
XX11
LINGUISTICS INSIDE OUT
"No Arguing with English". Review of The Story of English (BBC 2) and the book by R. McCrum, W. Cran and R. MacNeil. Times Literary Supplement, 26.9.86, 1062-79. Review of M. Toussaint, Contre l'arbitraire du signe. Language in Society 15/3.444-5. "Language Difficulties". Review of R. Kirk, Translation Determined. Times Literary Supplement, 28.11.86, 1338. 1987 "The Grammar In Your Head". Mindwaves, ed. by C. Blakemore and S. Greenfield, 507-16. Oxford: Blackwell. Reading Saussure. London: Duckworth. The Language Machine. London, Duckworth. "The Ephemerality of Translation". Times Literary Supplement, 28.8.87, 924/933. "Mentioning the Unmentionable". International Journal of Moral and Social Studies 2/3.175-88. "In and Out of the Language Lab". Review of L.-J. Calvet, La guerre des langues et les politiques linguistiques and F.J. Newmeyer, The Politics of Linguistics. Times Literary Supplement, 11.12.87, 1373. "Language as Social Interaction: Integrationalism versus Segregationalism". Language Sciences 9/2.131 -43. "The Ideological Implications of Onomatopoeia in the Eighteenth Century". Studies in Eighteenth-Century Culture 17.209-216. 1988 Language, Saussure and Wittgenstein: How to Play Games with Words. London: Routledge. "Across the No Man's Land". Linguistic Thought in England 1914-1945, ed. by R. Harris, ix-xi. London: Duckworth. (French version in Modèles linguistiques 9/2, 1987, 7-10.) "Murray, Moore and the Myth". Linguistic Thought in England 1914-1945, ed. by R. Harris, 1-26. London: Duckworth. (French version in Modèles linguistiques 9/2, 1987, 11-38.) 'Language & Communication. The First Seven Years". Robert Maxwell and Pergamon Press, ed. by E.J.J. Maxwell, 657-9. Oxford: Pergamon. Review of E.J. Reuland and A.G.B. Ter Meulen(eds.),77TheRepresentation of (In)definiteness. History and Philosophy of Logic 9/2.250-1.
ROY HARRIS: PUBLICATIONS 1956-1995
XXiii
"Who Will Translate the Translators?". Encounter 72/2.68-73. "The Semiotic Basis of Applicative Grammar". Review of S. Shaumyan, A Semiotic Theory of Language. Semiotica 74/1-2.121-32. "On a Sticky Wicket". Review of R. Bartsch, Norms of Language and S. Greenbaum, Good English and the Grammarian. Times Literary Supplement, 30.9.88, 1082. 1989 "The Worst English in the World?" An Inaugural Lecture. Supplement to the Gazette, University ofHong Kong 36/Ί, 24 April, 37-46. "The Garden of Eden". Hongkong Standard (Education), 9.6.89, 5. "Theories About the Origin of Language". Hongkong Standard (Education), 16.6.89, 5. "Lessons from Evolution". Hongkong Standard (Education), 23.6.89, 5. (with T.J. Taylor) Landmarks in Linguistic Thought: the Western Tradition from Socrates to Saussure. London: Routledge. "Degree-0 Explanation". Behavioral and Brain Sciences 12/2. 344-5. Review of R. Schleifer, A.J. Greimas and the Nature of Meaning: Linguistics, Semiotics and Discourse Theory. Review of English Studies 40/160. 541-2. "How Does Writing Restructure Thought?". Language & Communication 9/23.99-106. "English Fights for Survival in Hong Kong Schools". Hongkong Standard (Education), 1.11.89, 5. 1990 "Quelques réflexions sur la tyrannie de l'alphabet". L Ecriture: le cerveau, l'œil et la main, ed. by C. Sirat, J. Irigoin and E. Poulie, 195-200. Turnhout: Brepols. "The Integrationist Critique of Orthodox Linguistics". The Sixteenth LAC US Forum 1989, ed. by M.P. Jordan, 63-77. Lake Bluff: L.A.C.U.S.. u Lars Porsena Revisited". The State of the Language, ed. by C. Ricks and L. Michaels, 411-21. Berkeley: University of California Press. "The Scientist as Homo Loquens". Harré and his Critics, ed. by R. Bhaskar, 64-86. Oxford: Blackwell. "Semantics and its Critics". Hong Kong Papers in Linguistics and Language Teaching 13.35-9. "The Dialect Myth". Development and Diversity. Linguistic Variation across Time and Space. A Festschrift for Charles-James N. Bailey, ed. by J. A.
XXiV
LINGUISTICS INSIDE OUT
Edmondson, C. Feagin and P. Miihlhausler, 3-19. Dallas: Summer Institute of Linguistics. The Foundations of Linguistic Theory. Selected Writings of Roy Harris, ed. by N. Love. London: Routledge. "On 'Folk' and 'Scientific' Linguistic Beliefs". Meanings and Prototypes. Studies in Linguistic Categorization, ed. by S.L. Tsohatzidis, 449-64. London: Routledge. "On Redefining Linguistics". Redefining Linguistics, ed. by H.G. Davis and Τ J. Taylor, 18-52. London: Routledge. "On Freedom of Speech". Ideologies of Language, ed. by J.E. Joseph and T.J. Taylor, 153-61. London: Routledge. Review of M. Johnson, Attribute-Value Logic and the Theory of Grammar. History and Philosophy of Logic 11. 252-3. 1991 (with T.J. Taylor) "A Second Decade for Language & Communication". Language & Communication 11/1-2.1. Review of J.R. de J. Jackson, Historical Criticism and the Meaning of Texts. Review of English Studies 42/166.301-2. Review of A. Rai, Orwell and the Politics of Despair. Review of English Studies 42/166.307. "Words, Which Go Together. Are Languages Really Fixed or Error-Free?". Review of Z. Harris, A Theory of Language and Information. Times Literary Supplement, 26.7.91, 24. "Hobgoblins of a Cold Warrior". Review of N. Chomsky, Deterring Democracy. The Independent, 7.9.91, 29. Review of J. DeFrancis, Visible Speech: The Diverse Oneness of Writing Systems. Journal of Language and Social Psychology 10/3.217-220. "Language". The Blackwell Companion to the Enlightenment, ed. by J.W. Yolton, R. Porter and B.M. Stafford, 272-4. Oxford: Blackwell. (with M. Chan) "Introduction". Asian Voices in English, ed. by M. Chan and R. Harris, 1-2. Hong Kong: Hong Kong University Press. "English versus Islam: the Asian Voice of Salman Rushdie". Asian Voices in English, ed. by M. Chan and R. Harris, 87-100. Hong Kong: Hong Kong University Press.
ROY HARRIS: PUBLICATIONS 1956-1995
XXV
1992 "On Scientific Method in Linguistics". New Departures in Linguistics, ed. by G. Wolf, 1-26. New York: Garland. "Writing and Proto-writing: From Sign to Metasign". New Departures in Linguistics, ed. by G. Wolf, 180-192. New York: Garland. 1993 "Ecriture et Notation". Proceedings of the Workshop on Orality versus Literacy: Concepts, Methods and Data, ed. by C. Pontecorvo and C. Blanche-Benveniste, 9-42. Strasbourg: European Science Foundation. "Saussure and Linguistic Geography". Language Sciences 8/1.1-14. "Another Crystal Maze". Review of D. Crystal, An Encyclopedic Dictionary of Language and Languages. Times Literary Supplement, 16.4.93, 10. "Signs and Sentences". Review of Y. Tobin, Semiotics and Linguistics. Semiotica 95/1-2.141-6. "Three Models of Signification". Structures of Signification, ed. by H.S. Gill, 3.665-77. New Delhi: Wiley. "Introduction" to Adam Smith's Considerations Concerning the First Formation of Languages. British Linguistics in the Eighteenth Century. London : Routledge/Thoemmes. "Introduction" to James Beattie's The Theory of Language. British Linguistics in the Eighteenth Century, v-xii. London: Routledge/Thoemmes. "Introduction" to James Harris's Hermes. British Linguistics in the Eighteenth Century, v-xi. London: Routledge/Thoemmes. "Introduction" to John Home Tooke's Diversions of Purley. British Linguistics in the Eighteenth Century, v-xii. London: Routledge/Thoemmes. "Introduction" to Joseph Priestley's Lectures on the Theory of Language and Universal Grammar. British Linguistics in the Eighteenth Century, v-xi. London: Routledge/Thoemmes. "Introduction" to William Jones' Discourses Delivered at the Asiatic Society 1785-1792. British Linguistics in the Eighteenth Century, v-xi. London: Routledge/Thoemmes. "Linguistic Questions That Linguistics Does Not Ask". Actes du XVe Congrès International des Linguistes, 4.243-6. Sainte-Foy: Presses de l'Université Laval. "Integrational Linguistics". Actes du XVe Congrès International des Linguistes, 1.321-3. Sainte-Foy: Presses de l'Université Laval.
XXVi
LINGUISTICS INSIDE OUT
(with E. Komatsu) F. de Saussure, Troisième Cours de Linguistique Générale (1910-1911), ed. and trans. Oxford: Pergamon. (with R. Harré) "Preface". Linguistics and Philosophy. The Controversial Interface, ed. by R. Harré and R. Harris, ix-xi. Oxford: Pergamon. "What Is Philosophy of Linguistics?". Linguistics and Philosophy. The Controversial Interface, ed. by R. Harré and R. Harris, 3-19. Oxford: Pergamon. "Saussure, Wittgenstein and la règle du jeu ". Linguistics and Philosophy. The Controversial Interface, ed. by R. Harré and R. Harris, 219-31. Oxford: Pergamon. "Graphically Speaking". Review of T. Conley, The Graphic Unconscious in Early Modern French Writing. Times Literary Supplement, 24.9.93, 11. 1994 "Criticizing Saussure". Review of D. Holdcroft, Saussure: Signs, System and Arbitrariness. Semiotica 98/1-2.181-6. La Sémiologie de l'écriture. Paris: C.N.R.S. "Introduction" to Alexander Murray's History of the European Languages. British Linguistics in the Nineteenth Century, v-xi. London: Routledge/Thoemmes. "Introduction" to excerpts from works by Dugald Stewart, Nicholas Wiseman, Robert Chambers, Charles Darwin and Henry Sweet's History of Language. British Linguistics in the Nineteenth Century, vii-xiii. London: Routledge/Thoemmes. "Introduction" to Max Millier's Lectures on the Science of Language. British Linguistics in the Nineteenth Century, v-x. London: Routledge/Thoemmes. "Introduction" to Richard Chenevix Trench's On the Study of Words. British Linguistics in the Nineteenth Century, v-xi. London: Routledge/Thoemmes. "Introduction" to William Dwight Whitney's The Life and Growth of Language. British Linguistics in the Nineteenth Century, v-xi. London: Routledge/Thoemmes. "Semiotic Aspects of Writing". Schrift und Schriftlichkeit, ed. by H. Giinther and O. Ludwig, 1.41-8. Berlin: de Gruyter. 1995 "Talking Pictures (No Vision Needed)". Review of W.J.T. Mitchell, Picture Theory. Times Higher Education Supplement, 6.1.95, 22.
ROY HARRIS: PUBLICATIONS 1956-1995
XXVii
"A Well-Defined Corpus". Review of J. Willinsky, Empire of Words: The Reign of the OED. Times Higher Education Supplement, 14.4.95, 29. "Opening the Door to Language". Review of RE. Asher and J.M.Y. Simpson (eds.), The Encyclopedia of Language and Linguistics. Times Literary Supplement, 21.4.95, 10-11. "Translating Mitspeak". Review of S. Pinker, The Language Instinct. How the Mind Creates Language. Essays in Criticism 45/3.21r2-9. "Introduction". The Wellesley Series. Nineteenth Century Sources in the Humanities and Social Sciences. I Language and Linguistics, l.vii-x. 4 vols. London: Routledge/Thoemmes. "Put That in Your Pipe and Smoke It". Review of F. Merrell, Semiosis in the Postmodern Age. Times Higher Education Supplement, 22.9.95, 21. "The Lesson to Learn from Ern". Review of J. Harwood, Eliot to Derrida: the Poverty of Interpretation and N. Royle, After Derrida. Times Higher Education Supplement, 13.10.95, 22. "Saussure, Generative Grammar and Integrational Linguistics". Saussure and Linguistics Today, ed. by T. De Mauro and S. Sugeta, 203-213. Roma: Bulzoni.
Prologue It is more than twenty years since Roy Harris set out on his self-imposed mission to the academic tribe concerned with inquiry into language. The intricately interwoven complex of intellectual sins that the missionary has excoriated over those years, and the redemptive gospel he has more recently begun to preach, may be briefly summarised as follows. Although linguistics tends to present itself as a culture-neutral science of language, the enterprise is based not on ab initio consideration of the phenomena under investigation, but on a projection of certain preconceptions about language derived from a particular tradition of linguistic thought. These preconceptions are so pervasive as to determine the form taken by the inquiry itself: modern linguistic science sees itself primarily as engaged in elucidating the synchronic structure of spoken languages. But the very idea, says Harris, that there are (in any relevant sense) such things as "spoken languages" with a determinable "structure" is a product of a particular culture-specific way of apprehending linguistic phenomena. The most fundamental preconception is a consequence of the word "language" itself: Unfortunately, lip-service paid to the diversity of language does not detract from the lure of looking beyond this diversity in the hope of finding an underlying unity that the single term language can plausibly be held to stand for... Scholars who openly proclaim the variety of human linguistic activities are quite capable of speaking in the same breath of the capacity of human beings for language acquisition, as if nature had provided mankind with a single piece of biological equipment from which there somehow flowed the whole range of diverse behaviour acknowledged as "linguistic". Others find no inconsistency in pointing out how various are the uses to which men put the words at their command, and then proceeding forthwith as if nothing else mattered in the analysis of language apart from constructing an account of truth telling, or an account of syntax. In these and similar ways, academically respectable forms are found to accommodate the urge to find some core of unity which justifies one comprehensive term, language. (Harris 1981a: 8-9)
2
LINGUISTICS INSIDE OUT
The word "language" is a Baconian "idol of the market": a name "confused, badly defined and hastily and irregularly abstracted from things" (Harris 1981a: 8, quoting Bacon). It functions as such because of the deeply-rooted conviction that words "stand for" things: if there is a word, there must be a corresponding thing for which it stands. This notion Harris dubs "surrogationalism", and he identifies it as one of three ideas as to the nature and function of language that have dominated Western thinking since classical antiquity. The others are "contractualism", which envisages language as manifesting a tacit collective understanding among members of a community as to how a certain range of social affairs shall be conducted, and "instrumentalism", according to which words are understood as instruments for accomplishing human communicative objectives. Against the background of these ideas there has developed a Western myth about languages whose entrenchment as the basis for an academic science of language was the achievement that made Saussure the progenitor of modern linguistic theory. The myth is that languages are "fixed codes" - systems of correspondences between forms and meanings whose function is to facilitate "telementation" - the transfer of thoughts from one mind to another. These two doctrines are interdependent: if communication is indeed a matter of transferring thoughts from one mind to another, there must be a device or mechanism that enables this to happen. That device is the "fixed code". Thus the fixity of the code (in the sense that the same forms are paired with the same meanings for all speakers of the language) is a requirement of the telementational model of communication. It follows, therefore, that a language must be thought of as an abstraction from the variability of actual utterances. But given this variability, how is the underlying fixed code to be identified? A viable practice of descriptive linguistics, as Bloomfield recognised (Bloomfield 1933 [1935: 144]), depends on the possibility of counting certain utterances as "the same" in respect of both form and meaning. The required "sameness" is established by appealing, in effect, to the decisions about sameness already embodied in an antecedently given practice of writing, which is taken to provide, grosso modo, a ready-made representation of speech at the level of abstraction with which descriptive linguistics is concerned. But, says Harris, there is not only no plausible sense in which writing "represents" speech, it has no intrinsic or necessary connection with spoken language at all. It is a communicational medium sui generis whose use in connection with language is merely a function of the fact that oral communication is available "as an incomparably rich source of analogies for the graphic expression of all kinds of information..." (Harris 1986: 156), and whose
PROLOGUE
3
invention was profoundly significant in restructuring modes of thought including thought about language. How does the fixed code come to be available to users of a given language? The answer takes the form of (a supporting explanatory myth - the myth of the "language machine". This myth is responsible for the recently developed idea that linguistics is in the business of investigating the psychobiological realia that allow one to acquire one's native language. Acquiring one's native language is seen as a matter of deducing a "grammar" from the evidential data provided by the speech that one hears. Given that this evidence is fragmentary, variable and deficient, the question arises how children can so swiftly achieve error-free mastery of their language. The answer offered is that they do it with the aid of a built-in guidance mechanism: the language machine in the brain. So if language-learning is indeed the acquisition of unconscious knowledge of a fixed code, the child must be equipped with a function-specific mental machine to help him with the task. Thus the generativist theory of the language machine is seen to be a logical requirement imposed by the initial step of taking a language to be some fixed set of invariants underlying the vagaries of speech. Under the title "integrational linguistics", Harris has outlined the principles that would guide an attempt to work out a demythologised account of language and language use. The most radical of these principles is founded on the perception that for anyone who hankers after fulfilling Saussure's ostensible wish to develop a linguistics that looks at language "as it is" for language-users, simply treating "languages", in any sense of definition of the term, as so many "givens" for the linguist to set about analysing without further ado leads at best to an impoverished and at worst to a downright distorted view of the phenomena. Languages, according to Harris, are not the primary objects of the linguist's inquiry at all, but second-order products of communites of individuals whose actual linguistic experience is what first has to be made sense of, both for its own sake and in order to understand what languages are. In so far as the demythologising project, as envisaged by Harris, does not automatically point to a definite programme of work for an institutionalised academic discipline, he may be seen as endorsing a Wittgensteinian refusal to release the fly from the bottle only to entrap it forthwith in another. And this perhaps goes some way towards explaining both Harris's interest in Wittgenstein's philosophy of language and the reception accorded to Harris's work by the academic tribe called "linguists". For the fact is that for every benighted recipient of this complex and iconoclastic message who has either been converted by it or, alternatively, has consigned its importunate bearer to the tribal cooking-pots, there must be ten who have reacted with silence. The silence may be a prudent mask for bewilderment
4
LINGUISTICS INSIDE OUT
or a polite mask for contempt. It may be based on the feeling that Harris is doing philosophy rather than linguistics, and can therefore legitimately be ignored by linguists. Or it may simply reflect the tendency towards intellectual solipsism engendered in an academic world where professional advancement depends above all on transmitting (never mind whether anyone is receiving) a message (no matter how ill-considered) of one's own, and where evaluating the messages of others is consequently so much waste of time and effort. However that may be, this collection was conceived as an attempt to encourage further discussion by engineering a confrontation between Harris and his critics. The theme of John Joseph's essay is that Harris's language myth is "conceived so broadly as to take on its own mythical dimensions" (p. 12): Harris's endeavour to root out Baconian idolafori is flawed by the very kind of intellectual error under discussion. Arguably, Joseph's conception of the language myth is itself unduly broad, in that a large part of his essay deals with surrogationalism, which is identified by Harris as one of several conceptual strands that collectively make up "the concept of a language which survived as part of the intellectual inheritance of modern linguistics" (Harris 1980: 33), but whose precise relevance to or role in the language myth stricto sensu (i.e. the combination of the telementational and fixed-code fallacies) is obscure. Joseph claims that Harris fails to draw a clear distinction between surrogationalism and nomenclaturism, and that, partly as a consequence of this confusion, the attempt "to read surrogationalism into the whole of the Western tradition" (p. 17) is unconvincing. In particular, Joseph objects to the "rhetorical cleverness" that allows Harris to interpret Bloomfield's behaviorism, "the main purpose of which was to do away with...psychocentric surrogationalism..." (ibid.), as plunging linguistics yet more deeply into the surrogationalist mire. If Bloomfield is a surrogationalist, says Joseph, then either Harris doesn't understand what surrogationalism is, or surrogationalism itself is an idol of the market. Turning to the language myth proper, Joseph raises a similar objection to Harris's claim that, since Aristotle, Western thought about language has been in thrall to a telementational theory of communication. In particular, he objects to the ascription of a telementational theory to Saussure, who, according to Joseph, had no theory of communication at all. Like surrogationalism, telementation turns out to encompass too much to be a useful notion with which to map the history of linguistic thought. Joseph's final point is that these historical oversimplifications are in any case irrelevant, in that, even if they were justified, they have no bearing on Harris's positive views about language. The next two chapters discuss Harris in relation to generativism. Borsley and Newmeyer's contribution is a comprehensive rejection of Harris's ideas, on the
PROLOGUE
5
interesting basis that they "represent a regression to a form of empiricism more extreme even than that advocated by Leonard Bloomfield and his followers" (p. 62). But Borsley and Newmeyer's main focus is on the role of those ideas as a critique of generative grammar. The gravamen of their charge is that Harris's work "is remarkable for the ignorance it manifests of the actual claims and results of generativist theorizing, to the point where it attributes to Chomsky positions that he is on record as rejecting" (p. 43). According to Harris, generativism, like all other species of twentieth-century structuralism, is based on the language myth. But Borsley and Newmeyer argue that both components of the myth are irrelevant to generative grammar. As far as telementation is concerned, they claim that, far from being committed to such a view of communication, generativists not only explicitly reject it, but deny that generativism is interested in, or obliged to have truck with, any theory of communication at all: "since generative grammar does not seek to explain communication, it is under no obligation to propose an alternative to the telementational conception" (p. 47). As for fixed codes, Borsley and Newmeyer concede that in the adult speaker's "I-language" representations of phonetic form are linked with representations of logical form, and that to that extent the I-language constitutes a "code", in some sense. Moreover, the I-language is a "relatively stable... state...which then undergoes only peripheral modification" (Chomsky 1986b: 25). Does this mean, then, that generative grammar is concerned with fixed codes, in Harris's sense? No, for at least two reasons. First, because the linkages between phonetic form and logical form envisaged by Chomsky do not add up to the kind of mechanism for effecting telementation that Harris is concerned with, in that they do no more than help "to provide a basis for determining what a sentence means when it is used... Since L[ogical] F[orm] is not equivalent to the actual message conveyed, Chomsky's view of grammar is not consistent with Harris's idea of what a 'code' is". Secondly, it is crucial to Harris's conception of a fixed code that it should be the same for all its users. But an I-language is the property of an individual: a unique cognitive system that is only "partially shared by others in the various communities with which people associate themselves in their normal lives" (Chomsky 1991: 19; quoted by Borsley and Newmeyer). Philip Carr argues, contra Borsley and Newmeyer, that current generativist theory is committed to a version of the telementational thesis, that since that thesis imputes telepathic powers to human beings it cannot be sustained, and that therefore Harris's rejection of generativism is pro tanto justified. But rather than reject generativism himself, Carr suggests that the answer is to be more royalist than the king, and to elaborate a form of generativism free of the telementational taint. Not content to join with Borsley and Newmeyer in merely reiterating the
6
LINGUISTICS INSIDE OUT
familiar assertion that generativism is not bound to espouse any particular view of how communication works, Carr makes the starker and much more radical claim that "Chomsky has consistently denied that language serves communicative ends" and that "this denial is fundamental to the generative enterprise" (p. 82). Accordingly, what is required is that the generativist's I-language be disconnected from the outside world altogether, via a revision of "the current generative conception of linguistic expressions" whereby "they are...oriented towards behaviour, in the sense of being instructions to behave or intentions to behave" (ibid.). Two essays address Harris on writing. Pettersson pinpoints two novelties in Harris's thinking. First, the idea that it is idle to seek answers to such questions as how and when human beings first wrote on the basis of arbitrary terminological and definitional decisions as to what is to count as writing. Rather, there is a prior and more fundamental question: what conceptual advance was prerequisite for the systematically meaningful use of graphic signs? Pettersson offers no criticism of Harris's answer to this question; instead, he wonders how anyone writing the history of writing - and therefore concerned with how and when - is to apply it. The second novelty is the rejection of the idea that writing has an essential connection with speech: this, says Pettersson, has the striking implication that writing has a future - that its evolution did not come to an end with the development of the alphabet. David Olson aims to fill out, or at any rate modify, Harris's account of how the availability of writing "restructures thought". The obvious difference between a written text and a spoken utterance is that a written text is not ineluctably connected with a particular utterer. There thus opens up a conceptual gap between the (written) sentence and the (spoken) utterance: sentences, Harris says, are "authorless" and "unsponsored" (Harris 1989b). The concept of the unsponsored sentence must obviously add a dimension to the language-user's apprehension of the possibilities of language, but it is not immediately obvious how this new dimension affects the language-user's thinking. Olson's conjecture is that writing affects the mind by requiring the development of new concepts required for dealing with writing itself: "a kind of literate metalinguistics" (p. 102). Two contributions deal with Harris's discussions of Wittgenstein. Anthony Holiday is upset by what he sees as Harris's cavalier attitude to Wittgenstein's distinction between science and philosophy. This distinction, says Holiday, is crucial to Wittgenstein's conception of philosophy, and resides in the fact that "philosophy does not enable us to criticise or justify what we do by showing that our doings match or mismatch criteria laid down in the course of our philosophising" (p. 109).
PROLOGUE
7
Rom Harré offers a variety of comments on Harris on Wittgenstein. He suggests that Harris's complaint that Wittgenstein "never tells us where grammar comes from" results from a misunderstanding of the Wittgensteinian notion of "grammar". There are two sorts of Wittgensteinian grammar: "immanent grammar", which comes, says Harré, from "training in a practice", and "transcendent grammar", which comes from "someone sitting down and working out a system of rules". He suggests that Harris's belief that "Wittgenstein did not leave room in his treatment of language for the instability of linguistic practice and the fact of language change" (p. 142) is likewise mistaken. And against Harris's scepticism about linguistic rules he says that we cannot make sense of any human practice without the distinction between correct and incorrect performance, proper and improper actions, well and ill done projects and so on. We all agree that all human practices are normative, from drinking and cooking to calligraphy and doing mathematics, and even to persecuting people. If this is so the notion of "rule" must surely have a central place somehow (p. 145). Two contributions focus on Harris's integrational linguistics. As the word "integrational" implies, at the heart of his theorising is the idea that language must be understood in the light of its imbrication with the non-linguistic. Baron seizes on the consequential emphasis on "context", and argues that the importance of context in using and understanding language is itself a context-dependent variable: to the extent that it is unimportant, the argument for an integrational linguistics is weakened. David Fleming tackles the problem of providing integrational linguistics with a method and a programme. He considers the possibility that ethnomethodological conversation analysis, a mode of linguistic inquiry that appears to have certain assumptions in common with integrationalism, may offer an appropriate and useful exemplar. However, ethnomethodologists hypostatise a "system" of trans-situational formal mechanisms governing conversational interaction "abstract structures external to the particular act drive that act in a way surprisingly compatible with the orthodox structuralism that ethnomethodologists have claimed to reject" (p. 199). Fleming proposes an "integrational rhetoric" that "would produce close, detailed, empirical analyses of situated discursive action.. .without resorting to the limitations of ethnomethodological conversation analysis" (p. 207). Two further contributions revert in different ways to the issue of linguistic rules. John Taylor points out the incompatibility of fixed codes with a "multipletrace" model of memory. According to such a model,
8
LINGUISTICS INSIDE OUT
knowledge of a language... would comprise nothing other than a vast set of memory traces of previous encounters with the language. There would be no place in the model for "rules", whether these be understood as procedures for the manipulation of categories of symbols, or simply as abstractions or generalisations over individual linguistic events. For in its pure form at least, the multiple-trace model makes no provision whatsoever for the mental storage of generalisations over individual events. The archetype of language knowledge is not the generalisation captured in the grammar rule, but the particularities of the corpus, (pp. 214-5) Finally, Trevor Pateman suggests that the dispute between the orthodox linguist and the position taken by Harris vis-à-vis linguistic rules can be characterised in terms of Kant's distinction between determinant and reflective judgements as applied to art. Harris's response to all this, published here as the final chapter, is intended both to advance the various debates in which he is engaged and to provide him with the opportunity to re-state and refine his message in the light of critical reactions to it.
1 The "Language Myth" Myth: Or, Roy Harris's Red Herrings John E. Joseph
1. Introduction:
Idols of the
market
The criticism of Roy Harris most frequently voiced by other linguists is that his work destroys without rebuilding. Nowhere in his writings does he articulate the kind of practical program on which the linguistics industry could continue to survive, let alone expand. As if his relentlessly trenchant judgements on other theoreticians of language from the beginning of history to the present day didn't do damage enough, he casts them in a lucid and elegant prose style that can actually be read by people outside the field. Hence he is a threat to linguistics not only from within - where he can be safely ignored - but from without, where his scepticism over the possibility of any scientific linguistics puts the general reputation of the discipline into peril. The failure to establish a practical programme is indeed a problem that Harris should confront, for no one can reasonably expect people to embrace ideas that threaten their livelihood. It is like asking the tobacco industry to recognize a link between smoking and cancer. Linguistics, after all, really is an industry employing thousands of people and creating significant business for related industries such as abstracting, library cataloguing, publishing, printing, binding, all the way down to logging. It is not just an abstract "field" in Bourdieu's sense. Hundreds of millions of dollars per year are involved. Harris wants linguistics to give up its claim to the status of science which it won in the last century and has fought hard to keep, and upon which its economic position is founded. He seems to have forgotten the wisdom of Proverbs 11.29: "He that troubleth his own house shall inherit the wind". But even if we acknowledge the economic function that academic disciplines serve, they are nevertheless supposed to do more than that. They are supposed to be concerned with the validity of the ideas they pursue. The notion that Harris need not be attended to because he merely criticizes without offering anything positive in return may be fair enough from a pragmatic view, but intellectually, it
10
JOHN E. JOSEPH
is empty. It implies that what matters first and foremost is to keep the economic wheels of linguistics turning - give people work to do. If cogent arguments are put forward for reconfiguring the rationale for this work, they should be heeded only if they allow an overall increase in the amount of work produced or producible, and shunned if they threaten the workflow. Surely academic linguistics should aim higher than this. There is no denying the power of the economic drive; no linguist, including Harris, whose cadre of talented and devoted former students includes the editors this book, wants to put other linguists out of work. But it is a moral duty for linguists and other academics to resist the economic tide when intellectual integrity is at stake. For sheer integrity, Roy Harris has it all over his critics. His refusal to be a cog in the economic wheel is a large part of what makes him uniquely worth reading among modern theoreticians of language. Yet he does not stand aloof from the market whose idols he tries to shatter; he wields his hammer while standing within the marketplace itself For anyone who knows Harris's work, the phrase "idols of the market" will strike a familiar note. It is the title of Chapter One of The Language Myth (1981), the second book of a trilogy that forms the core of the Harrisian canon. The myth of the title is the concept of language as used, not by lay speakers, but by linguists. It is a reification, postulating the existence of a thing where in reality there is a diversity of forms of behavior and knowledge. Harris compares this myth with the second type of Bacon's "idola fori" {Novum Organum, 1.59-60): Idols of the market, in Bacon's view, are of two kinds. The less pernicious sort arise when words function as "names of things which have no existence". These idols are relatively easy to unmask, once it can convincingly be shown that there is nothing existing to which the word could correspond. But the more pernicious idols are "intricately and deeply rooted". This class includes the cases of words which do indeed correspond to actualities and not to fictions; but they are names "confused, badly defined, and hastily and irregularly abstracted from things". They mislead because they conflate under a single designation things which are distinct in reality. The harm which may thus be caused to the understanding is clearly this: that we may be led to believe, by reason of the single designation, in an underlying unity present in what in fact is a collection of merely associated but disparate things. (Harris 1981a: 7-8) The "language myth" thus engendered is responsible for the "segregationalism" of mainstream modern linguistics, which Harris's "integrationalist" program is aimed at resolving. He characterizes segregationalism as follows:
THE "LANGUAGE MYTH" MYTH
11
... segregational analysis treats language and languages as objects of study existing in their own right, independently of other varieties of communication and amenable to description in terms that are quite separate from those used in any other discipline. The alternative approach, the integrational approach, sees language as manifested in a complex of human abilities and activities that are all integrated in social interaction, often intricately so and in such a manner that it makes little sense to segregate the linguistic from the non-linguistic components... ...the segregationalist approach to language typically abstracts from the linguistic community and from the communication situation, and proceeds by setting up decontextualized systems of linguistic units and linguistic relations. The integrationalist, on the other hand, insists that language cannot be studied without distortion except in its normal functional context. (Harris 1987a [Love (ed.) 1990: 199]) Harris asserts that we must pass beyond the myth in order to arrive at a kind of linguistics that might produce real knowledge about human linguistic and communicative functioning, rather than continuing to elaborate the foundational myth itself. He describes the language myth as a "complex of interconnected errors" and as "the product of two interconnected fallacies: the telementational fallacy and the determinacy fallacy. The telementational fallacy is a thesis about the function of language, while the determinacy fallacy is a thesis about the mechanism of language" (Harris 1981a: 9). Telementation holds that "linguistic knowledge is essentially a matter of knowing which words stand for which ideas. For words, according to this view, are symbols devised by man for transferring thoughts from one mind to another" (ibid.). The determinacy fallacy, alternatively called the "fixed-code" fallacy (ibid. 10), explains how telementation works: languages are "fixed public plans", sets of "correlations between ideas and verbal symbols" shared by a community "in order to provide themselves with a viable system for exchanging thoughts (ibid.), i.e., for telementation. It soon becomes clear that the two fallacies are more than interconnected: the fixed-code fallacy is actually a corollary to the telementational one, since "the telementational model imposes what may be called an 'invariance condition' upon the language" (ibid. 88). As Harris would later write, "if speech communication is a telementational process, it demands a fixed code" (Harris 1990c: 30). The "complex of interconnected errors" that make up the language myth include the two fallacies which are the main focus of the first part of the trilogy, The Language-Makers (1980), namely surrogationalism and nomenclaturism, and two others which receive somewhat less attention in that book, instrumentalism and contractualism. The other one which he treats at length is the
12
JOHN E. JOSEPH
mechanistic conception of language which is the subject of The Language Machine (1987). The key fallacies he identifies as constituting the language myth, including surrogationalism, instrumentalsm, telementation, and fixed-code theory, have undeniable power. To what extent that power is logical or conceptual, and to what extent rhetorical, is an open question. Nor is it even certain that conceptual and rhetorical power can be meaningfully separated from one another in linguistics or any other realm of inquiry-cw/w-persuasion. At the very least, they are much more difficult to separate than is generally assumed. From this point of view it is not necessarily a criticism of Harris to assert that the component fallacies of the language myth, whatever their conceptual power, possess definite rhetorical power within his writing. Indeed his writing would not be so compelling otherwise. Yet the perception of a gap between rhetorical power and conceptual justifiablility is precisely what characterizes an idol of the market. Since Harris's "language myth" and its component fallacies are aimed at rooting out idola fori, it seems only fair that their own status in this regard should be examined. The present paper is therefore based upon the premise enunciated above by Harris, that certain terms can lead us "to believe, by reason of the single designation, in an underlying unity present in what in fact is a collection of merely associated but disparate things". The terms I shall investigate in the light of this premise are the following: surrogationalism, nomenclaturism, telementation, and the language myth. For if there turns out to be cause for a sympathetic reader to suspect that the "language myth" is conceived so broadly in Harris's writings as to take on its own mythical dimensions, that the desire to establish its existence prompts historically unconvincing readings, and that the whole idea of "myth" in Harris's work is finally at odds with his own linguistic, historical, and epistemological scepticism - then the opportunity for Harris to disabuse the reader of those suspicions would certainly be welcome. I should say at the outset that, unlike some of the other contributors to this volume, I really do see my disagreements with Harris as slight in comparison with how many of his central conclusions I accept. Yet neither are the divergences insignificant, for although they have mainly to do with the treatment of history, they point to quite another exit route from segregationalism than the one Harris would have linguists take.
THE "LANGUAGE MYTH" MYTH 2. Surrogationalism
and
13
nomenclaturism
Two of the fallacies that Harris discusses at greatest length are never clearly defined by him, and his use of them throughout his work shows no obvious or consistent pattern. Surrogationalism, the idea that "words are essentially surrogates or substitutes for other things" (Harris 1987b: 33), is arguably the oldest of the fallacies which compose the language myth: Languages are thus surrogational systems, which provide the language-user with a set of verbal tokens which stand for, or take the place of, non-verbal items of various kinds. Accordingly, it is the relation between words and what they stand for which is central to understanding how languages work. The concept of languages as surrogational systems has as its principal component a doctrine of names which goes back to the very beginnings of the Western tradition. In the Judaeo-Christian branch of that tradition it is embodied in one of the most influential etymological myths ever told: the account given in ch. 2 of the Book of Genesis of how Adam gave the animals names. In the Graeco-Roman branch, it makes it first appearance in the extended discussion of words presented in Plato's dialogue Cratylus. (Harris 1980: 33) The passage then goes on to discuss the "nomenclaturism" of Cratylus without any remark upon the terminological transition. The reverse occurs in Harris (1988a: 10), where after a discussion of nomenclaturism from the Bible through Locke, he suddenly comments that "Locke and Leibniz, no less than Cratylus and Hermogenes, espouse an essentially surrogationalist view of language". At times Harris appears to use the terms interchangeably (see ibid. 41 and 44, which respectively cite Saussure's condemnation of nomenclaturism and Wittgenstein's of surrogationalism, apparently in reference to the very same thing; Harris 1987c: 56, which states that "Saussure's attack on nomenclaturism in the Cours has been compared, not unreasonably, to Wittgenstein's in the Philosophical Investigations"', also ibid. 213, where Locke's account of communication is called "psychocentric surrogationalism" and "basically a form of nomenclaturism"). Elsewhere he speaks of nomenclaturism as a subtype of surrogationalism ("The difficulties which Locke's theory encounters are typically surrogationalist, and specifically nomenclaturist difficulties", Harris 1980: 69; "nomenclaturism, or, in its more general form, surrogationalism", Harris and Taylor 1989: 38). The closest he comes to an explicit definition of nomenclaturism is this: "the things in question are antecedently given; that is, exist independently either of their being named at all, or of what they are named. .. .what is named is prior, and the name is secondary" (ibid. 35-6).
14
JOHN E. JOSEPH
It seems, then, that surrogationalism and nomenclaturism say more or less the same thing in two different ways. Surrogationalism holds that words stand for things or ideas, and therefore emphasizes the function of language. Nomenclaturism says that words come about as names for already-existing meanings, and therefore highlights the historical and ontological primacy of nonlinguistic reality. If this is indeed an accurate account of Harris's view, then it would seem possible to have surrogationalism without nomenclaturism, but not the other way around. That is, one can hold that words stand for things, yet deny that words simply name realities that already exist; but to hold that words name realities that already exist necessarily entails that words stand for things. This is consistent with the statement about Locke (ibid. 69) quoted earlier in this paragraph, and it means that there are three possible positions for a theory of language to take in this regard: (1) surrogationalist and nomenclaturist; (2) surrogationalist but not nomenclaturist; (3) neither surrogationalist nor nomenclaturist. Harris divides nomenclaturism into two types, "natural" and "non-natural" (ibid.: 36). The first holds that "each name not merely stands for but somehow reflects the identity o f what it names (ibid.: 39), while the second does not make this assumption. He likewise divides surrogationalism into two types, "reocentric", which "supposes that the things words stand for are to be located 'out there' in the world external to the individual language-user", and "psychocentric", which "supposes that what words stand for is to be located internally, that is to say in the mind of the language-user" (ibid.: 44). In Harris's view, the Biblical account of Adam naming the animals (Gen. 2.19) is surrogationalist since the words will stand for the things, and nomenclaturist since the animals and the distinctions among them exist before they are brought to Adam for naming. Of course it is enough to say that the account is nomenclaturist, since surrogationalism is entailed thereby - though actually what independent evidence of surrogationalism there might be in this particular Biblical story is hard to see. Indeed, the first Creation account (Gen. 1) is neither surrogationalist nor nomenclaturist, in that it depicts God using words to create the cosmos, which means that some words, at least, exist prior to the things they name and can have a function very different from "standing for" them. Nevertheless, Harris asserts that ... there is a measure of agreement between Genesis and Cratylus, at least in the following respects. Names are treated as vocables standing in a certain relationship to things [= surrogationalism: JEJ], and the things in question are antecedently given; that is, exist independently either of their being named at all, or of what they are named [= nomenclaturism: JEJ]. (Harris 1980: 35-6)
THE "LANGUAGE MYTH" MYTH
15
At the very least, this statement must exclude Gen. 1 to be accurate. Harris goes on to say that the question of "natural" vs. "arbitrary" names does not arise in this story, since the names Adam gives to the animals are not subject to evaluation. Yet the rest of Gen. 2-4 contains numerous accounts of what might be called natural or motivated naming: Adam calls the woman made from his rib ishshah 'woman' because she was taken from ish 'man' (Gen. 2.23), and later names her Eve ('life') because she is the mother of all who live (Gen. 3.20; see notes in Sandmel et al. (eds.) 1970). Eve in turn names her third son Seth 'granted' because God granted him to her in place of the murdered Abel (Gen. 4.25). In addition there are cases of implicitly natural naming: in the second Creation account, God forms adam 'man' from adamah 'the dust of the ground' (Gen. 2.7); Abel ('herdsman') and Cain ('smith, metallurgist') are a shepherd and a tiller of the soil respectively (Gen. 4.2), and so on. Perhaps it is the case that names for animals are left to the arbitrary choice of Adam, while names for human beings must be naturally or at least rationally justified. The name of God is, of course, taboo. In any case, it is far from clear that questions of natural naming do not arise in the pre-Babel portion of Genesis; though the examples given here certainly suffice to temper the statement "Nor does Genesis explain the origin of any words other than the names of the various species of animals and birds" (Harris and Taylor 1989: 39-40). On the other hand, one can assume with assurance that all surrogationalism, including that of the Old Testament, is reocentric prior to the introduction of psychocentrism by Aristotle. In Harris's view, "The Western intellectual tradition has been mainly dominated by reocentric thinking, from Plato's theory of Forms onward" (Harris 1980: 49). That statement assumes particular definitions of "Western intellectual tradition" and "mainly" on which not everyone would agree, for arguably equal or greater domination could be ascribed to the psychocentric surrogationalism that Harris sees extending from Aristotle through the modistae in the thirteenth and fourteenth centuries, to Locke and Hobbes in the seventeenth, to Saussure in the twentieth, and through Saussure to contemporary segregationalist linguistics. But to go back to Plato, the question may be asked: is he a nomenclaturist as well as a surrogationalist? And if so, what kind of nomenclaturist is he? The answer to these questions depends crucially on how one interprets the dialogue Cratylus. In Harris's view, which I share, Plato is rejecting both the natural nomenclaturism of Cratylus and the non-natural nomenclaturism of Hermogenes, denying that either provides an adequate account of language, which for Plato would depend crucially on the kind of reocentric surrogationalism represented by his theory of forms. Plato "deconstructs" nomenclaturism by dividing "things" into two types: the ideal forms of things, which are eternal and unchanging and
16
JOHN E. JOSEPH
therefore "real", and physical things, which are merely copies of these forms and ephemeral. The ideal forms of things precede both physical things and their names. The question of the "correctness of names" hinges upon whether names are direct imitations of the ideal forms, or only of their physical "shadows", which are illusions; and the answer cannot be given a priori, since it is a historical matter having to do with the person who made the names and how well he knew his art. Taking nomenclaturism again to assume that "things... exist independently either of their being named at all, or of what they are named. .. .what is named is prior, and the name is secondary" (Harris 1980: 35-6), this is a fair assessment of Cratylus' and Hermogenes' positions, but it is precisely the kind of oversimplification that Socrates (and through him Plato) sets out to complexify. Thus, to associate Plato with nomenclaturism, and to identify him as the source of reocentric surrogationalism, is a red herring. The "reocentric" part is already a problem - of course Plato believed that the reality of things is "out there" relative to human beings, rather than internal to the mind, but not in the way that most people have always conceived "out there" to be, namely in the physical "reality" of the world of perceptions that surrounds us. But "surrogationalism" is a complete red herring as applied to Plato, who does after all problematize the whole idea that words stand for things and decides that the problem is insoluble. In calling it a red herring, I do not mean to assert that Plato cannot be made to fit the term "surrogationalist" as Harris has defined it, but only that the application of this term to him is trivial, obfuscating, and ultimately meaningless. And the same goes for the "surrogationalism" of Genesis. But not, I should think, for the surrogationalism of Augustine. The famous account of his acquisition of language in Book I of the Confessions, doubly famous as the point of departure for Wittgenstein's Philosophical Investigations, really does assert that words stand for things, and indeed incorporates this assertion into his explanation of how the child learns to speak: ... cum ipsi [maiores homines] appellabant rem aliquam et cum secundum earn vocem corpus ad aliquid movebant: videbam et tenebam hoc ab eis vocari rem illam, quod sonabant, cum earn vellent ostendere... Ita verba in varus sententiis locis suis posita et crebro audita quarum rerum signa essent paulatim colligebam measque iam voluntates edomito in eis signis ore per haec enuntiabam. (Augustinus, Confessiones 1.8) When they [older people] would name something and then, following their voice, would move their body toward it, I would see this and grasp that the thing was called by the sounds they made when they wanted to point it out... Thus, by hearing words frequently used in their proper places in various sentences, I
THE "LANGUAGE MYTH" MYTH
17
gradually pieced together what things they were signs of, and after my mouth was trained in these signs, I would enunciate my own wishes through them. (My trans.: JEJ) It is specifically the use of words to enunciate and satisfy wishes - a wish implying absence of the thing named - that is the proof of surrogationalism, if admittedly not its only manifestation. That is what makes the attribution of surrogationalism to Augustine so convincing. Harris, however, is determined to extend surrogationalism back in history to the founders of the basic worldview that informs Augustine's works, though the same kind of evidence for it that occurs in Augustine is not found there. Similarly with nomenclaturism: it is easy enough to demonstrate in the story of Adam naming the animals, and in the linguistic theories of Cratylus and Hermogenes; but to extend it to Genesis and to Plato, as Harris does, is a leap whose logic is not apparent. It seems to go something like this: look at the surrogationalism of Augustine - striking isn't it? but in fact it isn't really Augustinian. It is symptomatic of the whole intellectual tradition of which Augustine is a part, even if clear cases like Augustine's are rare. And it testifies to the antiquity and longevity of the "language myth". The problem is that, having said this, we are liable to read surrogationalism into the whole of the Western tradition regardless of whether there is any evidence of it or not. And likewise, we are liable to read Plato as a nomenclaturist and reocentric surrogationalist even if he is as opposed to all these notions as it is possible for anyone to be. If the terms nomenclaturist and surrogationalist are broadened to apply to all the people Harris would apply them to, they give up whatever significance they originally had when applied to the clear cases, like Cratylus and Augustine. In some instances Harris's rhetorical cleverness verges on the diabolical, as when he comes to discuss the behaviorism of Bloomfield, the main purpose of which was to do away with the psychocentric surrogationalism or "mentalism" characteristic of most turn-of-the-century continental linguistics, and which seemed hopelessly metaphysical in the modernist lights of 1920s America and Britain. Yet in Harris's view, behaviorism does not save Bloomfield from surrogationalism at all, but plunges him into it even more deeply: ... Bloomfield had recourse to one of the oldest options available in the Western tradition, and reinstated a crude reocentric surrogationalism of the kind to which Saussure had been utterly opposed. In Bloomfieldian theory, languages thus became again what Saussure had denied they were: nomenclatures. Words were merely vocal substitutes for non-vocal things and events in the external world. To the question of how words came to function as substitutes for things,
18
JOHN E. JOSEPH behaviorism supplied an answer in terms of conditioned responses... (Harris 1980: 159)
To make his point, Harris contrasts Bloomfield precisely with Augustine, and downplays Augustine's surrogationalism in order to make that supposedly implicit in behaviorism appear all the stronger: If we compare Augustine's account of language-learning with Bloomfield's, it is clear that Wittgenstein could have picked a better example of pure surrogationalism without going outside the present century. (Harris 1980: 160) Yet by way of illustration of Bloomfield's "reocentric surrogationalism", Harris supplies quotations not from Bloomfield, but from the behaviorist John B. Watson, who said that "words are but substitutes for objects and situations" (Watson 1924: 184) and that a language provides the speaker with "a verbal substitute within himself theoretically for every object in the world" (ibid. : 187). This is indeed reocentric surrogationalism at its strongest, but it is not Bloomfield. Harris does quote Bloomfield's definition of the meaning of a linguistic form as "the situation in which the speaker utters it and the response which it calls forth in the hearer" (Bloomfield 1933: 139), then offers the following comment: By putting forward this pivotal definition, Bloomfield unhesitatingly took surrogationalism to an extreme which the Western linguistic tradition had not previously seen. That is, he proposed that as far as the linguist was concerned everything in the universe should be divided into just two categories: speech events on the one hand, and the corresponding speakers' situations and hearers' responses on the other. (Harris 1980: 162) This comment loses sight of what surrogationalism is supposed to be essentially about: words functioning as surrogates of or substitutes for things (or meanings or ideas, though with Bloomfield ideas are not in question). Although Watson appears to have thought of words in this way, Bloomfield instead sees languages as a kind of action, on a par with other sorts of behavior; he attempts to undo the traditional reification of words that is already the basis of surrogationalism. That he is not consistently successful in the attempt is apparent, but the attempt itself ought to be recognized as a significant effort in the direction Harris wishes linguistics to go. There are certainly aspects of Bloomfield's behaviorism that Harris might legitimately criticize, but the attempt to force it into the surrogationalist framework for the narrative purposes of Harris's "language
THE "LANGUAGE MYTH" MYTH
19
myth" myth - in part by making Watson a "surrogate" for Bloomfield - is a red herring. The problem with surrogationalism, the reason it is so limiting to a theory of language, is that it implies that language actually covers and consists of much less than it actually does, that linguistic meanings can be inventoried in a list of all the things one might wish to replace with a surrogate. Bloomfield stresses over and over that nothing of the sort is true. Meaning is unlimited: In order to give a scientifically accurate definition of meaning for every form of a language, we should have to have a scientifically accurate knowledge of everything in the speakers' world. The actual extent of human knowledge is very small, compared to this. (Bloomfield 1933: 139) Furthermore, surrogationalism implies that a deterministic, one-to-one match-up between word and meaning is a straightforward matter. Bloomfield says that this might be true for some things, but not for most: We can define the meaning of a speech-form accurately when this meaning has to do with some matter of which we possess scientific knowledge. We can define the names of minerals, for example, in terms of chemistry and mineralogy, as when we say that the ordinary meaning of the English word salt is 'sodium chloride (NaCl)', and we can define the names of plants or animals by means of the technical terms of botany or zoology, but we have no precise way of defining words like love or hate, which concern situations that have not been accurately classified - and these latter are in the great majority, (ibid.) From this Bloomfield concludes that "The statement of meanings is therefore the weak point in language-study, and will remain so until human knowledge advances very far beyond its present state" (ibid.: 140). It will be evident by now that what Harris means by "surrogationalism" is a bit slippery, shifting here and there as his narrative purposes require. Another way to try to get hold of it is to inquire what, in Harris's view, its opposite might be. The only discussion to be found of a non-surrogationalist approach to language concerns the concept of language which informs the dictionary: The dictionary, in short, embodies a semantics which is in some respects the complete antithesis of surrogational semantics. The meanings the dictionary gives for words like chair, table and bread are neither physical objects nor ideas; but simply other words. In a civilisation with a thoroughgoing surrogationalist attitude to language, the dictionary as we know it would be inconceivable.
20
JOHN E. JOSEPH Whereas the fact is that in Europe the dictionary came to be seen as the repository of verbal meanings par excellence, regarded with a veneration and respect for authority amounting in certain cases almost to superstition. Such a development would be impossible if a surrogational view of languages prevailed twenty-four hours a day and seven days a week. (Harris 1980: 78)
There are at least two problems with this passage. Its reasoning goes like this: surrogationalism means that words stand for things or ideas; dictionaries define words not in terms of things or ideas but of other words; therefore, surrogationalism and dictionaries are incompatible. But the conclusion does not follow from the premises, any more than this one does: economic theory holds that paper money stands for abstract value; change machines convert paper money into coins; therefore change machines are incompatible with economic theory, and change machines could never have been invented in a society in which economic theory prevailed twenty-four hours a day and seven days a week. Secondly, the only thing which is truly inconceivable in the passage is a civilisation in which "the dictionary as we know it would be inconceivable" - this is another red herring, intended to mask the fact that the "extended" surrogationalism Harris is discussing is itself mythical. Truly to escape from surrogationalism would mean giving up the notion that words stand for physical things or events, mental ideas, or Platonic ideas - in other words, for anything at all, except, perhaps, other words. It is not something that either the behaviorists or Wittgenstein managed to do, in Harris's view. Indeed there is no historical precedent for it: no record of any society that has thought about language as not exchangeable for things in the world, in the mind, or in heaven. Nor does Harris pretend to be able to conceive of what a linguistics that got beyond this conception of language would be like. His achievement lies in recognizing that unless and until a theory of language is developed that does this, we are trapped within this conceptual prison we have inherited, and it takes nothing away from this achievement that no one, including Harris, has yet seen the way forward. But this is really the point of it all: getting beyond the notion that words have meanings. There is no reason to think that this notion is a peculiarly Western product, since it is not the case that other cultural traditions offer any significantly different understanding of language in this regard. In so far as the naming of surrogationalism, nomenclaturism, and other fallacies lends power to Harris's conceptual apparatus, he might be said to exploit an aspect of nomenclaturism itself: the belief that "what is named is prior, and the name is secondary" (Harris and Taylor 1989: 36). A corollary of this is, as Bacon knew, the belief that whatever is named, exists. If Harris had described
THE "LANGUAGE MYTH" MYTH
21
the phenomenon of telementation without attaching a name to it, it would not function so forcefully within his writing. There is of course nothing inherently wrong with naming new concepts, even if it exploits aspects of the very fallacies the concepts are aimed at unmasking. I suspect that many of Harris's best readers appreciate the postmodern irony of this, however much Harris himself might cringe at the postmodern part. Yet at the same time the fallacy is capable of working even on these best readers; perhaps the rhetorical effect of the name is even stronger when one's attention has been momentarily deflected by the irony. Again, there is nothing wrong with such a deployment of rhetoric; it is just that we should then ask whether the concept so persuasively named is appropriately applied in each particular case. That is, whether the power of the name is backed up by actual historical evidence for the fallacy's existence, or the power of the name functions in place of any such evidence. If the former is true, no reader should feel guilt over being swept off on the rhetorical tide. As Aristotle pointed out in his Rhetoric (in answer to the Socratics, who believed that rhetoric was inherently deceptive and evil), to have truth on one's side counts for nothing if one is unable to persuade a jury. Truth does not come automatically equipped with persuasiveness. But if the latter is true, the strong reader will resist being persuaded by the rhetorical effect. I believe that the evidence presented above suggests that surrogationalism and nomenclaturism, while based upon arguably fair descriptions of what certain people have said about language in the Western tradition, are turned into "doctrines", and simultaneously into idola fori, and red herrings. They draw attention away from, and thus needlessly weaken the logical force of, Harris's ultimate concern to challenge the almost universal assumption in contemporary linguistics that language exists in order to serve the needs of the deeper, more important realm of meaning. We should, moreover, expect those twentieth-century theoreticians of language who problematize the ultimacy of meaning to receive more sympathetic treatment (at least on this particular score) in Harris's writings than do those who assume it. Bloomfield in particular should loom high on this list. But instead, as we have seen, the opposite occurs: Bloomfield is depicted as taking surrogationalism to an unprecedented height. Harris seems initially more drawn to the Saussurean and Wittgensteinian problematizations of meaning, but finally neither of these emerges with any points for trying either. Harris instead stays consistent with his narrative framework of a monolithic language myth - it is as though no one can break free from, or even make a dent in, any part of it so long as he remains captive to other parts of it. One is either an integrationalist or a segregationalist, with no middle ground. But it may be that in time we shall look back on the parallel rise of problematizations of meaning in linguistics and philosophy (as well as
22
JOHN E. JOSEPH
psychology and other areas) in the first half of the twentieth century as the starting point of a movement that would lead in time to the realization of something like Harris's integrationalist project. This would then be the ultimate justification of what Saussure, Wittgenstein, and even Bloomfield were trying to do; and the fact that in their wake the traditional approach to meaning not only refused to give way but was actually shored up will appear not so much as a consequence of their inability to escape the language myth, but as the kind of rearguard reaction that almost always occurs in the long process of establishing a wholly new understanding of physical or human phenomena. If the essence of Harris's critique of linguistics is taken to be that we have no hope of moving toward a radically improved understanding of linguistic behavior unless and until we have passed beyond the basic conceptions of language and mind that have been handed down to us since Aristotle and more particularly since Descartes, then I doubt that any but the most shortsighted of linguists or philosophers would find anything in it to dispute.
3. Telementation This brings us to the cardinal fallacy for the language myth as Harris describes it. As noted toward the beginning of this paper, Harris (1981a: 9) has identified telementation as one of two interconnected fallacies of which the language myth is the product; the other, the deterministic fallacy or fixed-code theory, he comes to see as little more than a corollary of telementation (see Harris 1990c: 29-30). The classic description of telementation occurs in Locke's Essay Concerning Human Understanding (1689): To make Words serviceable to the end of Communication, it is necessary, (as has been said) that they excite, in the Hearer, exactly the same Idea, they stand for in the Mind of the Speaker. (Locke, Essay, III.9.6) Locke intends this not as a description of communication, but as a goal to strive toward, and from which actual communication, alas, falls short. In a later passage he writes: ...the ends of Language in our Discourse with others, being chiefly these three: First, To make known one Man's Thoughts or Ideas to another. Secondly, To do it with as much ease and quickness, as is possible; and, Thirdly, Thereby to convey the Knowledge of Things. Language is either abused, or deficient, when it fails of any of these Three. (Locke, Essay, III. 10.23)
THE "LANGUAGE MYTH" MYTH
23
But just as he has read surrogationalism and nomenclaturism into texts where they are not clearly to be found, Harris does the same with telementation. Indeed he takes it as far back as Aristotle: Speech is a form of telementation. This theory can be traced back through the medieval modistic grammarians to Aristotle. Words, says Aristotle at the beginning of De Interpretatione, are "symbols or signs of affections or impressions of the soul". "As writing," he continues, "so also is speech not the same for all races of men. But the mental affections themselves, of which these words are primarily signs, are the same for the whole of mankind, as are also the objects of which those affections are representations or likeness, images, copies". (Harris 1981a: 9, translation of Aristotle by H.P. Cook, On Interpretation, Loeb Classical Library ed., London, 1938, I) ... although agreeing with Plato that the mind stores "likenesses" of the things we perceive, [Aristotle] does not agree that there is any mimetic relation between these likenesses and the words which stand for them. The latter relationship is purely conventional. Thus he maintains (i) that the world is the same for all its inhabitants, (ii) that the "mental representation" of the world is the same for all its inhabitants, but (iii) that language is not the same for all because it is conventional, and different communities have different conventions. This entails that people will understand one another's speech if, and only if, they base their speech on the same conventions... It is clear that Aristotle presupposes that communication is telementational; in short, that words "transfer" thoughts from one person's mind to another person's mind because - and in so far as - the words are associated with the same thoughts in both minds. And that, precisely, is the role of convention; to establish in a person's mind the connections between words and thoughts. (Harris and Taylor 1989: 33) But if we look to the actual texts of Aristotle, we can only conclude that Harris has interpolated the tele- into the telementationalism supposedly found there. Aristotle never says anything about the transference of ideas from one person's mind to another's. Rather, Harris says it is implied by Aristotle's statements that words are "symbols or signs of affections or impressions of the soul" - that is, the mind - and that they signify by agreement, i.e. convention. But it is an unjustified and pointless logical leap to say that because Aristotle believed that the mind was the locus of meaning, and that meaning was conventional, he therefore believed that communication actually functioned in the way that Locke described
24
JOHN E. JOSEPH
as the ideal of how it should function, as the transfer of ideas from one speaker's head to another. He is again committing the same error of anachronism of which he accuses anyone who reads a Hermogenetic non-natural nomenclaturism back on to Adam's naming of the animals. Indeed, since his "telementalization" of Aristotle would proceed equally unfettered if Aristotle held to a naturalist rather than a conventionalist view of names, Harris is simply equating all "mentation" with telementation. What he is really criticizing, in other words, is not the idea of thought transfer, but the whole idea of mind. Here again Harris's problematization of the notion of mind in linguistic theory is of central importance - but all the rest, all the supposed "telementation", is so many red herrings. Where telementation really comes to the fore is in the case of a post-Lockean figure, Saussure. In some of his more recent writings Harris has stressed Saussure's responsibility for embedding the language myth within twentiethcentury linguistics, particularly through his adoption of the telementational model. Why did the adoption of the principles of arbitrariness and linearity produce a segregationalist linguistics? Was that inevitable? No, it was not inevitable; but it became inevitable once the principle of arbitrariness and the principle of linearity were conjointly wedded to one particular theory of human communication. The theory in question is telementation; that is to say, the theory which explains communication as the transference of thoughts from one person's mind to another person's mind. Saussure adopted telementation as his theory of communication, although he does not designate it by that term. Nevertheless, the adoption is explicitly spelled out in Saussure's account what he calls the "speech circuit". (Harris 1990c: 26) Harris locates Saussure's adoption of the telementational model in his description of the speech circuit, and identifies it as the decisive element in the fallacious grounding of linguistic theory from Saussure onward. Harris's treatment of this theme is powerful and memorable, perhaps in part because Saussure's speech circuit diagram is itself so visually striking, and one measure of its impact is its persistence in work by his former students, for whom Saussure's telementationalism is a given (see Love 1990a: 53; Taylor 1990: 124-5). The previous passage continues as follows: Saussure's speech circuit envisages the archetypal speech act, reduced to its bare essentials, stripped of all possible ramifications. There are just two participants, A and B, who in turn take on the roles of speaker and hearer. A says something to B, and Β in return says something to A. That constitutes one completed lap of
THE "LANGUAGE MYTH" MYTH
25
the speech circuit. This simple scenario assumes that A and Β are speaking the same language. If the only language known by A were Catalan and the only language known by Β were Cantonese, this exchange would not constitute a speech circuit in Saussure's sense, regardless of how A and Β eked out their oral utterances and their mutual tolerance with gestures, facial expressions and other varieties of non-verbal communication. Let us suppose, then, for the sake of argument, that A and Β are both speaking Kalaba. What happens in the speech circuit, according to Saussure, is that certain ideas occur to A. A wishes to transmit these ideas to B. Because A is a speaker of Kalaba, these ideas trigger in A's mind the phonetic image of certain Kalaba words, which A then proceeds to utter. Β hears A's utterance and, being also a speaker of Kalaba, is able to interpret the Kalaba sounds heard as having precisely the meanings which correspond to the original ideas in A's mind, which A intended to transmit to B. Speech communication, on this view, is essentially a process of telementation, or thought-transference. (Harris 1990c: 26-7) This appears to be such a straightforward presentation of Saussure's speech circuit that it is surprising to realize how much of it consists not of citations or paraphrases of anything Saussure said or wrote, but of Harris reading between the lines. What the Cours de linguistique générale (henceforth CLG) actually says is this: Le point de départ du circuit est dans le cerveau de l'une, par exemple A, où les faits de conscience, que nous appellerons concepts, se trouvent associés aux représentations des signes linguistiques ou images acoustiques servant à leur expression. Supposons qu'un concept donné déclenche dans le cerveau une image acoustique correspondante: c'est un phénomène entièrement psychique, suivi à son tour d'un procès physiologique: le cerveau transmet aux organes de la phonation une impulsion corrélative à l'image; puis les ondes sonores se propagent de la bouche de A à l'oreille de B: procès purement physique. Ensuite, le circuit se prolonge en B dans un ordre inverse: de l'oreille au cerveau, transmission physiologique de l'image acoustique; dans le cerveau, association psychique de cette image avec le concept correspondant (CLG 28) The starting point of the circuit is in the brain of one individual, for instance A, where facts of consciousness which we shall call concepts are associated with representations of linguistic signs or sound patterns by means of which they may be expressed. Let us suppose that a given concept triggers in the brain a corresponding sound pattern. This is an entirely psychological phenomenon,
26
JOHN E. JOSEPH followed in turn by a physiological process: the brain transmits to the organs of phonation an impulse corresponding to the pattern. Then sound waves are sent from A's mouth to B's ear: a purely physical process. Next, the circuit continues in Β in the opposite order: from ear to brain, the physiological transmission of the sound pattern; in the brain, the psychological association of this pattern with the corresponding concept. (Harris trans.)
Saussure never says explicitly that A and Β must be speaking the same language, that A wishes or intends to transmit ideas to B, that Β is able to interpret A's utterance as having precisely the meanings which correspond to the original ideas in A's mind, or that any of this is being offered as a model of "communication". The characterization of the circuit as "the archetypal speech act, reduced to its bare essentials, stripped of all possible ramifications" is Harris speaking, not Saussure. Saussure never restricts the circuit to just two participants, but says that this is the minimum number required. And when he says that Β associates the sound pattern he hears with le concept correspondant "the corresponding concept", he means the concept corresponding to the sound pattern, not to "the original ideas in A's mind", which is not necessarily the same thing. Of course the mere fact that something does not appear in the CLG or its source materials does not rule it out as a Saussurean idea. We can legitimately reconstruct things Saussure left unsaid based upon the records of what he did say - records which are themselves subject to reconstructive interpretation from the multiplicity of sources. We can with equal legitimacy eliminate other reconstructions as unlikely or impossible within Saussure's overall linguistic worldview. My purpose is actually not to contest Harris's interpretation of Saussure as a telementationalist - the available evidence allows a convincing case to be made either for or against this interpretation, though not for any definitive judgment - but rather the use Harris makes of this interpretation, turning telementation into Saussure's "theory of communication" and giving it a central place in his idea system, indeed suggesting in the citation above that it is the crucial element in the development of segregationalist linguistics in the twentieth century. First, I should explain in what sense the evidence for Saussure 5 s telementationalism is ambiguous. If he had actually said the words Harris put into his mouth, that Β "is able to interpret" the sounds of A's utterance "as having precisely the meanings which correspond to the original ideas in A's mind, which A intended to transmit to B", he would be as clear-cut a telementationalist as they come. In fact Saussure makes hedging statements that suggest a recognition on his part that such an interpretation can never be "precise":
THE "LANGUAGE MYTH" MYTH
27
Entre tous les individus ainsi reliés par le langage, il s'établira une sorte de moyenne: tous reproduiront, - non exactement sans doute, mais approximativement - les mêmes signes unis aux mêmes concepts... C'est par le fonctionnement des facultés réceptive et coordinative que se forment chez les sujets parlants des empreintes qui arrivent à être sensiblement les mêmes chez tous. (CLG 30; italics added: JEJ) All the individuals linguistically linked in this manner will establish among themselves a kind of mean; all of them will reproduce - doubtless not exactly, but approximately - the same signs linked to the same concepts... The individual's receptive and co-ordinating faculties build up a stock of imprints which turn out to be for all practical purposes the same as the next person's. (Italics added: JEJ) But perhaps that is close enough. Even Locke, after all, recognized that the communication of ideas is not as perfect as he would have liked it to be. Similarly with the question of A's "intention" of transmitting ideas to B: this is not asserted in the presentation of the speech circuit in the CLG, which says instead that a concept "triggers" a corresponding sound pattern, implying that this is an automatic process. This seems to contradict the statement on the following page that the "executive side" of the circuit, the conversion of concept into sound pattern, ".. .is always individual, and the individual is always master of it. This is what we shall designate by the term speech" (30) - that is, parole rather than langue. Part of the problem here is that Saussure did not actually use the word déclenche "trigger' in his lecture on the speech circuit; it was supplied by the editors of the CLG. On the other hand, they did not make it up out of the blue: for CLG 99, reproducing one of Saussure's lectures faithfully, says of the signifier and signified that "The two elements are intimately linked and each triggers the other". A full reading of the CLG and its source materials suggests that Saussure ultimately believed that both views were true, each applying in its particular domain. That is, the "executive" production of speech is first something which an individual speaker must choose to do at all. He then chooses the thoughts he wishes to express, and the language in which he will express them. Only after choosing the language can he organize his thoughts into linguistic concepts and their basic linear sequence, which are part of the language, though still subject to the speaker's will up to the point, apparently somewhere within syntax, where the choosing stops and the triggering begins. Saussure is never clear on this. Nor does he ever specify anything about the process by which the individual's thoughts are organized into linguistic concepts, i.e. signifieds, or about what
28
JOHN E. JOSEPH
happens in the hearer's mind following the association of concepts with the sound patterns heard. Are these concepts then delinearized into post-linguistic thought, and is this a matter of individual parole, or somehow dictated by the structure of langue? Neither the word idée 'idea' or pensée 'thought' occurs in the discussion of the speech circuit, which makes Harris's account of it (especially the reference to "thought-transference") slightly misleading. All Saussure speaks of are les faits de conscience, que nous appellerons concepts 'facts of consciousness which we shall call concepts' (CLG 28), and these 'concepts' will later be identified as signifiés 'signifieds'. Although Saussure elsewhere sometimes conflates "idea" with "concept" and "signified", the imposition of the word "idea" on this discussion forces it into a Lockean mold. Again, I am not denying that Saussure imagined the speaker to intend to transfer thoughts or ideas to the mind of the hearer and to carry off that transference more or less successfully; I am saying that the evidence is too scanty and ambiguous either to support or deny such an interpretation definitively. But the more important point is that it simply does not matter. Saussure did not address all the questions outlined above (and many more that could be outlined) because he did not have a "theory of communication" at all. A fact which Harris never points out is that the word communication never occurs in the CLG or Constantin's notebooks for the third course. The plural, communications, and the verb communiquer occur once each, in the same sentence, but nowhere near the discussion of the speech circuit. Instead they arise in the context of language change, in the discussion of "intercourse" (CLG 281 ff.) as the explanation of why languages are essentially cohesive and yet susceptible to change. Although this discussion occurs long after that of the speech circuit in the published CLG, it actually came five months earlier in Saussure's final course of lectures (intercourse on 29 November 1910, the speech circuit on 25 April 1911). Here we find no mention whatever of mind or brain: "la force d'«intercourse»" is simply ...[ce] qui crée les communications entre les hommes. ...l'intercourse les oblige à communiquer entre eux. C'est lui qui amène dans un village les passants d'autres localités, qui déplace une partie de la population à l'occasion d'une fête ou d'une foire, qui réunit sous les drapeaux les hommes de provinces diverses, etc... C'est à l'intercourse qu'est due l'extension et la cohésion d'une langue. Il agit de deux manières: tantôt négativement: il prévient le morcellement dialectal en étouffant une innovation au moment où elle surgit sur un point; tantôt positivement: il favorise l'unité en acceptant et propageant cette innovation. (CLG 281-2)
THE "LANGUAGE MYTH" MYTH
29
...[that] which establishes communication between men. ...intercourse forces them to communicate with others. Intercourse brings to a village visitors from elsewhere, brings together people from all around on the occasion of a celebration or a fair, unites men from different provinces under the same flag. It is intercourse which is responsible for the extension and cohesion of a language. It acts in two ways. Negatively, it prevents dialectal fragmentation by suppressing innovations as they emerge at one point or another. Positively, it favours unity by accepting and propagating innovations. Here, as virtually everywhere in the Cours, the focus is on langue. Saussure is simply not interested in communication, except for its ramifications on langue, which pertain strictly at the level of diachronic inquiry. For synchronic inquiry, communication is nothing more than a backdrop. This is where I believe Harris commits a serious distortion of Saussure, centered upon the "talking heads" drawing of CLG 27. From Constantin's notebook we can see that Bally and Sechehaye are partly to blame for this: Saussure's own drawing showed two unconnected lines, one extending from A's mouth to B's ear, the other from B's mouth to A's ear, with no lines involving their brains or minds. The figures are not even labeled A and B. Directly to the right of this is the "real" drawing, similar to the one on CLG 28, which does not show speakers at all, but rather signs divided into verbal concept and verbal image, and connected by way of phonation and audition. For Saussure, clearly, this was the one that mattered; and I believe that the same is true for the vast majority of readers of the Cours, except those who under Harris's influence have come to see the initial background drawing as instead the core of it all, the proof of Saussure's telementationalism and the source of segregationalism in twentieth-century structuralist linguistics. I have accused Harris of distorting Saussure, a charge serious enough to require attribution of a motive. What might have induced him to promote Saussure's telementationalism (real or imaginary) to a status of a cardinal explanandum in the development of modern linguistics? The answer lies in the power it confers on the "language myth" archemyth. I noted earlier that Harris's use of the word "ideas" in his account of Saussure's speech circuit channels the reader toward a Lockean, hence a telementationalist, reading. By making the speech circuit look as Lockean as possible, a tradition of telementationalism extending from Locke to Saussure and onward through the twentieth century is established as a linchpin of the language myth. Yet when Harris comes to discuss the relationship between Locke's and Saussure's "theories of communication" directly (Harris 1987c: 204-218), interestingly enough it is the basic discrepancies between them that emerge as most salient. The most severe of these discrepancies has to do with what
30
JOHN E. JOSEPH
Parkinson (1977) calls Locke's "translation theory" of understanding, a part of which, but only a part, does Harris believe Saussure shares: The term "translation theory" refers to the fact that, according to the theory in question, when language is the vehicle of communication understanding requires a double process of "translation": a speaker's thoughts are first translated into sounds, and then the sounds uttered are translated back again into thoughts by the hearer. This is clearly the basic idea behind Saussure's account of what happens when A and Β engage in discourse. (Harris 1987c: 205) However, Locke and Saussure do not conceive of "thoughts" in the same way. Locke believes in simple and complex "ideas" that exist as it were independently of both words and speakers, whereas the equivalent for Saussure, concepts, are themselves part of linguistic signs, hence not independent of words. Thus, There are differences between Locke's version of the translation theory of oral communication and Saussure's. Locke evidently supposed that there could be thought without language, and that the mind could engage in it without the aid of any linguistic instrument... Saussure emerges as a sceptic on this score. He does not explicitly discuss the question of whether or to what extent human beings could think without language, but he describes prelinguistic thought as amorphous. "Psychologically, setting aside its expression in words, our thought is simply a vague, shapeless mass" ([155]). More specifically still, "...were it not for signs, we should be incapable of differentiating any two ideas in a clear and consistent way... No ideas are established in advance, and nothing is distinct, before the introduction of la langue " ([155]). This marks a significant shift of emphasis away from Locke... (ibid.: 209) Indeed, comparing Locke's and Saussure's views on the existence of thought without language is an even more difficult undertaking than Harris suggests here, because these views cannot be disentangled from the way other people used thought and pensée in the late seventeenth and early twentieth centuries. In Locke's time, thought was used by philosophers to denote the entire range of mental processes, starting from simple perceptions (e.g. sights, smells, feelings). In the course of the eighteenth century it gradually shifted focus and became more specialized, so that the "archetypal" thought was conscious and linearly organized, and simple perceptions were usually excluded. Thus when Max Muller took as his slogan in the 1860s "No language without thought, no thought without language", he did not mean to include vision, pain, etc., within "thought". Indeed from our perspective it is not clear how much his slogan
THE "LANGUAGE MYTH" MYTH
31
constituted a theory of language and thought, and how much a normative assertion of the new definition of "thought" as restricted to that mental activity which is linguistic in nature. Nevertheless, Harris is certainly right that there are differences between the Lockean and Saussurean conceptions of thought and language, from which he concludes: To this extent it could perhaps be argued that it is unfair to saddle Saussure with a crude, old-fashioned Lockean theory of communication. For Saussure would certainly have denied that the systems of concepts on which communication depends are given independently of the system of vocal signals. In Saussure's theory, therefore, there is strictly no question of "translation" between two systems: on the contrary, both are constitutive of one and the same langue. (ibid.: 210) Yet Harris does not draw the conclusion he himself suggests, that in fact it is pointless to characterize Saussure as a telementationalist. Another discrepancy which Harris notes is that Locke's theory of communication is designed, first, as part of a more general theory of mind, and second, to account for linguistic behavior which Saussure would classify as parole. Saussure however adapts it to become the basis for his theory of langue'. ...Saussure leaves himself room to dissent from Locke's claim that men talk "only that they may be understood; which is then only done when, by use or consent, the sound I make by the organs of speech excites in another man's mind who hears it the idea I apply to it in mine when I speak it" (Locke 1706: III.3.3). For Saussure is apparently concerned only with giving an account of communication in so far as it is mediated by la langue: he has nothing to say about the more general problem of understanding between individuals, (ibid.: 211) Harris expresses puzzlement therefore that Saussure should'have adopted Locke's telementational model: Why did Sàtisstire apparently try to adapt to his own ends a theory of communication Which ostensibly served a quite contrary purpose? To put the point more shârply, why did Saussure take over a (Lockean) theory of communication concerned obviously (in Saussurean terms) with la parole, and apply it instead to la langue? (ibid.: 212)
32
JOHN E. JOSEPH
The answer, Harris says, lies in Saussure's structuralism. Whereas "preSaussurean linguistics treated language as being reducible, in the final analysis, to a constitutive totality of individual linguistic acts" (ibid.: 212, taking a cue from Hjelmslev 1942: 30), "[s]uch a view is completely antithetical to any form of structuralism, and Saussure could not accept it". To be more precise, the "constitutive totality of individual linguistic acts" is precisely what Saussure called parole, distinguishing it from langue, the virtual system that makes such acts possible; and Harris is asserting in effect that any structuralism had to be based on langue, the socially-shared system, and not parole, the totality of individual acts. In rejecting a theory of communication that took language to be the latter, Saussure ... inherited the obligation to offer at least a minimal account of the individual linguistic act, in order to avoid the charge that his theory was an abstraction which simply failed to come to terms with the realities of everyday speech. Looking at the problem from this angle, one can see why Saussure would have found a modified Lockean model of communication attractive. In effect, it offered two welcome guarantees. In the first place, it established the role of linguistics in the human sciences. By interposing la langue between the individual speaker and la parole within a Lockean framework, Saussure ensures that the study of language cannot be reduced either to the psychology of individual speech acts, or to social analyses of the communication events which such acts constitute. For the individual acts depend on the existence of a linguistic system, of which neither the psychology nor the sociology of the nineteenth century offered any detailed description. In the second place, reducing la parole to the individual implementation of resources provided by la langue ensured that within linguistics the study of the latter take priority over the former and be independent of it. Thus built into the speech-circuit model there is what amounts to a double guarantee of autonomy: (i) autonomy of linguistics among the disciplines dealing with human speech behaviour, and (ii) autonomy of the study of la langue within linguistics. (ibid.: 212-213) I agree completely with the idea that Saussure was motivated by a desire to establish linguistics independently of psychology and sociology (a point I elaborate on in Joseph 1994). What I do not see, however, is what difference the "modified Lockean model of communication" makes in this regard. The "double guarantee of autonomy" to which Harris refers falls out precisely from the privileging of langue as the "real" reality of language, the sole one amenable to scientific inquiry. To imply that this privileging falls out from the theory of communication, rather than the other way around, would be to put the cart before
THE "LANGUAGE MYTH" MYTH
33
the horse - except that there is no cart. As I have stated above, the speech-circuit diagram is not offered as a "theory" of communication, any more than the drawing of a tree on CLG 97 is offered as a "theory" of tree. Instead, to take Harris's own words out of context, "Saussure's speech circuit is essentially a schematic summary" (Harris 1987c: 205), but not of what Harris takes it to be a summary of. It is, in effect, a schematic summary of the general sense of the word communication, applicable to whatever "theory" of communication one may hold - including one which states that no "real" communication ever occurs when people speak, only the illusion of communication - and committed to none. The speech circuit is based on no more elaborate assumptions than that "people communicate" and "words have meanings". If these things are so, then whenever someone tells someone else something (which is the commonsense interpretation of a speech event, though clearly not a complete one, as Saussure points out), then the actual utterance (as phonation) is sandwiched between meaning as speaker's intention and meaning as hearer's interpretation. If this is telementation, then anyone who believes that words have meanings, or that communication can occur, is a telementationalist. That seems ultimately to become Harris's position, and it threatens to turn the whole concept of telementation into a red herring. When Saussure says that the individual act of speech "requires at least two individuals", it is not to assert any controversial corollary within a theory of communication, but simply to say that speech exists for the purpose of communication. That is a theory not of communication, but of speech, parole, which is precisely what Saussure claimed his circuit drawing to represent. Presumably a "Lockean" model of communication is one which asserts telementation of ideas that exist separate from language, while a "modified Lockean" model asserts telementation of concepts that are already part of language. Hence the "essence" of a Lockean model is telementation: a model which does not assert telementation would be non-Lockean, regardless of whether it traffics in ideas separate from language, in which case we might consider it Platonic, or in concepts that are already part of language, in which case I could think of nothing more appropriate to call it than Saussurean. For it is only by a distortion of Saussure that one can assert him to have a "theory of communication" that takes precedence over the distinction between langue and parole. The fact that variants of the word communication appear in only one sentence in the published CLG and Constantin's notebooks at first seems almost unbelievable in a text that insists so strongly on the social nature of langue. But it is a fact nonetheless, one that no amount of Harrisian commentary can obliterate. The explanation for it is that Saussure's lectures focused on the linguistics of langue, but communication is always a matter of parole. Hence
34
JOHN E. JOSEPH
Harris's question, "why did Saussure take over a (Lockean) theory of communication concerned obviously (in Saussurean terms) with la parole, and apply it instead to la langue!" (Harris 1987c: 212), contains two false premises. If he had taken over a theory of communication, he would clearly have restricted it to parole. It does not take an intimate familiarity with the Cours or its source materials to realize that langue is a pivotal concept around which much else revolves and from which much else falls out, while the speech-circuit diagram is tossed in, as Harris rightly asserts, as part of a campaign to position langue strategically between psychology and sociology. But whereas investigation of langue consitutes nearly the entire rest of the Cours and the lectures they were based on, the "model of communication" is never revisited. When Harris invokes Saussure's "structuralism" to answer the question re cited in the last paragraph, it is tantamount to admitting that Saussure's conception of language does not in fact fall out from his "telementational theory of communication". His structuralism, under which Harris (correctly) subsumes his view on the nature οf langue and its status as the principal or sole object of linguistic inquiry, must stand independently of and prior to the "telementation" which it is supposed to explain. But this is not what Harris had been claiming up to this point, or what he would claim more strongly than ever in his 1990 article: namely, that Saussure's adoption of the telementational theory of communication, wedded with the principles of arbitrariness and linearity, are the primal causes from which langue and the rest of linguistic structuralism fall out. The fact is that if we need to have structuralism independently to understand Saussure, then telementation does not "buy" us anything. We no longer need it as the foundation stone of anything else, and that is good, in view of what a poorly set foundation it is in the first place. Moreover, if we ask: would a different (i.e. non-telementational) theory of communication have forced Saussure to change his conception of langue in any essential way? - the answer is no. Suppose for a moment that Saussure had been a communicational sceptic, believing that communication never actually occurs, though we sometimes have the illusion that it does. Could this only be explained by asserting that langue is not a social fact, or that signifiers and signifieds are not intimately bound to one another? Not at all. As hinted at above, Saussure could have continued to assert this structuralist conception of langue, while simply extending parole to the "receptive" as well as the "executive" end, and holding that the understanding of language, as well as its production, involves individual, willful construction. Within such a theory, the socially-shared langue would serve to explain how it is that the illusion of communication occurs, which may ultimately be the most difficult thing to explain of all. Consider too that Bloomfield (1927), whose antimentalism would seem to immunize him from any
THE "LANGUAGE MYTH" MYTH
35
charge of telementationalism, found no difficulty in reconciling Saussure's system with his own behaviorist views (see Joseph 1990: 58-63), which might not have been the case were telementationalism really so central to Saussure as Harris suggests. That the reverse is equally true - telementationalism is not compatible solely with a structuralist conception of language, but with others as well - has already been argued by Harris in the case of Locke. But, finally, is even Locke a telementationalist in the sense that Harris makes Saussure out to be? As noted at the start of this section, Locke describes telementation as the ideal of communication, and explicitly denies that it is the reality. This is the essence of what Taylor terms "Locke's puzzle". I imagine many, perhaps most, people in Locke's time as well as our own would agree that the ideal or goal of communication is the transference of ideas, and that this goal is rarely if ever met. It is not clear from Harris's work that he would classify them as telementationalists, if they believe that perfect transference of ideas is not the reality of communication. I imagine there have always also been more hardcore telementationalists, who really do believe such communication is always taking place. Of course, having no direct access to other people's thoughts, we cannot prove them right or wrong; all we can do is to deny that the mind as such exists at all. In other words, "telementation" does not even count as a fallacy, since a fallacy is by definition a failure of logic, and there is no basis for counting telementation as a failure of logic unless it is based on a "mentation" that is itself nonexistent. And again, as before, this makes both telementation as such, and the "telementational fallacy", a pair of red herrings. In the case of Saussure, what they deflect attention from is the fact that his conception of langue independently accounts for the various subfallacies Harris wants to attribute to him - a fact which does not lend the kind of support to the idea of a "language myth" that a continuity of telementationalism from Aristotle to Locke to Saussure does. Harris (1987c: 206-7) gives examples of linguists from the 1960s onward who express overtly the kind of telementational stance that Harris reads into Saussure, noting that "The family resemblance is unmistakable" (ibid.: 205). This might be taken as showing that, regardless of whether a reading such as mine absolves Saussure of literal telementationalism, that is nevertheless how he was read, and that is part of his legacy to twentieth-century structuralist linguistics. But surely this would be absurd. It is in fact Harris's original contribution to have read the speech-circuit diagram as a telementationalist theory of communication, or indeed as anything at all. A look at any treatment of Saussure not written by Harris or one of his students will show that the diagram has never been accorded any significance, theoretical or otherwise, in the transmission of Saussure's leading ideas: namely, the dichotomies of langue and parole, synchrony and diachrony, signifier and signified, and other tenets of
36
JOHN E. JOSEPH
structuralist linguistics. The simple fact is that, as stated in the previous paragraphs, some people subscribe to a telementationalist view of communication and others do not; there has never been any unitary "myth" on this score, any more than on the related question of whether the "real" meaning of texts is that intended by their author, that inferred by their readers, or that which somehow inheres within the text itself. All three views - all three "myths", if you like have always coexisted, even if at various times and in various quarters a particular one has been temporarily dominant. If we were to attribute to Saussure, even indirectly, views that formed within the context of Saussurean structuralism, we would have to include the doctrine of structuralist literary criticism that the real meaning of a text inheres neither within the intention of the author, nor within the minds of readers, but within the text itself. This view fails a crucial test for being telementational: it stresses that texts (a concept which does not in principle exclude any type of utterance) are fundamentally different from the intentions of their authors, but always contain both less and much more than was in their authors' minds. The goal of reading is not telementation between reader and author, but as rich a construction as possible of the meaning that is latent within the text. And individual minds and individual readings are always made secondary to the ultimate reality of the text itself. This position is in fact quite in line with Saussure's view of linguistic meaning. He locates meaning within the sign. Individual speakers, speech acts, and minds are all subordinated to the sign itself. He is not really concerned with thought before it comes into language, or after it goes out of language, because it is amorphous. Only in so far as it is in language is thought of any concern to the linguist; divorced from sound, which is to say from its existence as one half of the linguistic sign, it belongs to "pure psychology" {CLG 157). I am not asserting that Saussure held the structuralist view of textual meaning, but only countering the use of later developments within structuralism as evidence for the telementational heritage of the CLG. Like most intellectual movements, structuralism was far less homogeneous than we are sometimes tempted to make it appear in hindsight. The problem with an approach like Harris's which tends to focus on individual thinkers and not give much consideration to their contemporaries is that it, for want of a better word, mythifies those thinkers into both more and less original creators than they may actually be, mythifies their times into periods when everybody thought alike, and mythifies the most famous thinkers of various periods into having been in a kind of dialogue with one another that has no basis in reality. Harris's discussion of Locke and Saussure, for example, is absurd in the absence of any substantial placing of either in the context of his time or any serious consideration of all the developments which took place between
THE "LANGUAGE MYTH" MYTH
37
them. Harris actually acknowledges much earlier in the book that "Scholars doubt whether Saussure had ever studied Locke at first hand (de Mauro 1972: 381 n. 11), Aarsleff 1982: 27)", but, he declares without explanation, "the question is irrelevant to a reading of the Cours" (Harris 1987c: 28). Well, so it is if in fact both are simply manifestations of a unitary "language myth" that runs through Western thought from antiquity through the present; but not if that is precisely one of the questions to which one is seeking an answer. In addition, Harris's discussion of the contrast between Locke's and Saussure's notions of thought and language would be clarified tremendously by the inclusion of information on what thought meant to people in the late seventeenth century (when it often included perceptual data) and pensée to people in the early twentieth. And consideration of the intervening history might save him from creating the false impression that the development of early structuralist ideas beginning in the late nineteenth century is the source of Saussure's views on how thought is amorphous until it is analyzed and linearized by the introduction of language (ibid.: 209-210), views which in fact are already articulated in Condillac's Grammaire in the eighteenth century. The chapter on Locke and Saussure in Harris (1987c) closes with a brilliant conceit that the lines connecting the "talking heads" in the speech-circuit diagram resemble telephone wires, and that the power of this analogy helps account for the force with which Saussure's telementationalism was taken up in twentiethcentury structuralism. It would be a marvelous explanans, if only its explanandum were true. There is no evidence that such cases of telementationalism as can be detected in modern linguistics (Harris does not identify more than a handful) have any link to Saussure. What then does the characterization of Saussure as a telementationalist "buy" Harris? A lot, actually. First of all, in deploying a term like "telementation" Harris simultaneously fashions a concrete and simple concept and joins it with a visually suggestive name that moreover carries overtones of magician's trickery as well as of- not just the telephone wires that were seen as a marvelous contrivance in Saussure's day, but the electronic communications media that have come to occupy a much more central role in our lives today. What is more, the reader of Harris does not even need to read the CLG, but only to open to page 27, to see in one split second the apparent visual proof that the language myth not only existed but informed the very Bible of modern linguistics.
38 4. Conclusion:
JOHN E. JOSEPH The "Key to All
Mythologies"
By this point you may be wondering why I should have stated in the introduction that my disagreements with Harris were slight, then gone on to write about them at such length. In fact most of the points I have tried to make in the preceding sections are themselves red herrings, since the real point is whether or not the historical exposition is essential to Harris's theory of language. If, as I believe, it is not, then not only is Harris's historical exposition filled with red herrings - it is itself the biggest, reddest herring of all, deflecting attention from his original thinking on language and communication and his program for redefining linguistics. Herring number two is the "language myth", that archemyth encompassing all of Western civilization and its offshoots. Harris's attempts at bringing it to light are reminiscent of some grand theosophical quest, or of The Key to All Mythologies; the tragically unfinishable life's work of Edward Casaubon, the dried-up old scholar who marries the heroine of George Eliot's Middlemarch. To be sure, "dried-up" is the last description anyone would apply to Harris. But the desire to reduce all the conceptions of language in the history of linguistics to a set of "keys", with the language myth as master key, is in the same universalist spirit. This is a crucial aspect of the rhetorical power of Harris's writing. Harris locates errors that are conceptually concrete, discrete, and simple, within texts that are themselves remarkable for their abstractness and complexity. Few of his readers, including his students, are so well-read as he, and for the others their first introduction to many classic texts is through Harris's treatment of them. His account of the elemental fallacies really do function for such readers as a set of keys, offering a clear and unified reading of sprawling and opaque texts that resist global comprehension even after a lifetime of study. From Harris one learns how to go into a text in the history of linguistics and, using the key fallacies, come out with a reading that is not only coherent but that readily assigns the author a place within a historical chain of thinkers whose work has been analyzed using the same set of keys. From a pedagogical and heuristic point of view, this is a real contribution, and earns Harris a place within a great tradition of teaching and textual analysis. If the method poses the danger that students may interpret their key-derived reading as the real and full meaning of the text rather than as one interesting but partial reading, useful for certain purposes but with no pretensions to being globally right - one might point out that the same danger exists in all teaching and writing, and grows proportionally with the power of the method. Or one might say that these keys "mislead because they conflate under a single designation things which are distinct in reality. The harm which may thus be caused to the
THE "LANGUAGE MYTH" MYTH
39
understanding is clearly this: that we may be led to believe, by reason of the single designation, in an underlying unity present in what in fact is a collection of merely associated but disparate things" (Harris 1981a: 7-8). From this (Harrisian) perspective, Harris contributes to the process of linguistic myth-making which he claims to undo. He is driven to this by an archemyth of his own designing, namely the idea that one single language myth, with all its various components, has characterized the whole of Western linguistic thinking. The title of The Language Myth, the second installment of Harris's "language" trilogy, confirms this idea. While serving as a powerful expository and rhetorical device, this myth of a myth has proven to be too powerful a heuristic device in Harris's hands, leading him to discover fallacies in key linguistics texts that in fact can only be read into those texts via arguments from silence, inferences from side comments, or by throwing the texts to the winds and referring instead to what their authors should have said if they were to be consistent with themselves. If such a unitary tradition existed, it would be a simple enough matter to explain, given the status of the Bible, Plato, Aristotle, and other surrogationalist etc. texts in Western culture, together with the role played by Latin and the way it was taught in Europe from the days of the Western Roman Empire through the eighteenth and in some cases the mid-twentieth century. Yet when has there ever been any universal consensus on the meaning of the Bible, or Plato, or Aristotle, on the role of language and rhetoric in society, or on a theory of how language operates? It is true that a myth of such unity is commonplace, for example that everyone in the twelfth and thirteenth centuries was committed to the basic principles of modistic grammar. But one does not have to dig very deeply into the surviving documents of any period to appreciate the splintering of views extant under even the most oppressive majority; the religious schisms within the Christian church are merely the best known examples. This is not to deny the existence of traditions of linguistic thought of the kind Harris is describing. It is rather to affirm them, while at the same time stressing that traditions (plural) and not myth (singular) is historically the operative concept. Whichever term we use, it is important to insist that such traditions are historical products, in the sense that the ideas they contain can be connected to particular developments in the histories of peoples and languages that could conceivably have gone otherwise. If I prefer to refer to "traditions", it is because "myth" is such a loaded term, implying that a particular tradition runs contrary to logic or some other objectively determinable standard of truth. In my own (post-Harrisian, or perhaps pre-post-Harrisian) perspective, it is not so clear that the language myth and its component fallacies, even if they are the red herrings I claim them to be, are to be dismissed as idola fori. When the
40
JOHN E. JOSEPH
gods have died, and taken absolute criteria for truth with them, perhaps idola fori are all that are left to us. This sort of "rhetorical constructionist" view, which has become quite prevalent in both humanistic and scientific disciplines over the last two decades, does not assume a clearcut distinction between myth and reality and a special knowledge of which concepts fall into which category. It throws into question the commonsense conception of what academic work and scientific knowledge are about, the kind of conception that informs Harris's writing. Yet rather than dismissing work based on such a conception, it retains it as one particularly privileged "myth" among a variety of others. For example, the rhetorical constructionist would recognize that Harris's frequent insistence on the impossibility of any "science" of language is actually based upon a particular "myth" of what a science must be and do - in fact a view of science which epistemologists of the "hard sciences" have largely discarded. Harris gives absolute credence to this myth when he insists that language is not fit to be one of its objects. Yet the rhetorical constructionist will prefer to examine why Harris chooses to argue in the way that he does, rather than to dismiss his argument as discordant with the spirit of his own project for "demythifying" linguistics. The strength of the integrationalist program lies in its commitment to a diversity of perspectives and sources of insight into language and communication. That commitment would be strengthened by the recognition that a diversity of perspectives has always characterized linguistic thought in the West as elsewhere, and not by attempting to fit it all into the Procrustean bed of a single "language myth". In other words, the "'language myth' myth" produces a history of linguistics that is as segregationalist - if not more so - than the linguistics it is purporting to describe. Demythifying the myth allows us to integrationalize that history. De-emphasizing the discourse of myth and fallacy altogether may help us see our way clear to what I believe may ultimately prove the most promising route out of segregationalism: the realization that the "science" of language actually consists of the rhetorical construction of a variety of accounts (myths, if you prefer) of linguistic behavior - not all equally valid, but subject to evaluation against criteria that are neither independent nor universal, but are themselves part of the same language game within which the construction of theories is taking place. Harris's belief in the impossibility of any scientific linguistics is in fact based not in scepticism, but in the lack of it - in a too great credulity about the possibility of "science" as a body of absolute knowledge. If we are to believe our colleagues in virtually every other area of inquiry that has ever laid claim to such a possibility, the epistemological problems that beset the study of language are not so unique to it as Harris's views on linguistics would lead us to believe. To drop the denial that it can be a science would also be a move coherent with the general spirit of integrationalism; for it is
THE "LANGUAGE MYTH" MYTH
41
supposed to be segregationalist to suggest that language is "amenable to description in terms that are quite separate from those used in any other discipline", and integrationalist to assume that "it makes little sense to segregate the linguistic from the non-linguistic components..." (Harris 1987a [Love (ed.) 1990: 199]).
2 The Language Muddle: Roy Harris and Generative Grammar Robert D. Borsley Frederick J. Newmeyer
1. The "language myth" If our sweep of the bibliographical surveys of the field is accurate, a major theme in modern linguistics is that Chomsky has got it all wrong. Functionalists, discourse analysts, structuralists, cognitive grammarians, connectionists, stylistic analysts, and even a sizeable percentage of those committed to formal linguistic analysis take as their point of departure the fundamental inadequacy of Chomsky's approach to language. Typically, such work interweaves its critical remarks with concrete proposals for alternative conceptions of the form that an adequate theory of language should take. In one sense, Roy Harris's scholarly output of the past two decades is very much in the above tradition; we doubt that any other author has devoted as much space to attacking (what are taken to be) the historical and theoretical foundations of generative grammar. 1 But Harris's work is unique in several striking respects. * We would like to thank John Joseph and Keith Percival for their helpful comments on an earlier version of this paper. 1 Our critique is based upon a close reading of the following publications by Harris, which we take to be representative of his ideas: Harris (1970; 1972; 1973a; 1973b; 1977; 1978; 1980; 1981a; 1981b; 1983a; 1983b; 1983c; 1984a; 1984b; 1985; 1986; 1987a; 1987b; 1987c; 1988a; 1988b; 1988c; 1989a; 1990a; 1990b; 1990c; 1993a; 1993b); Harris and Taylor (1989). Throughout this paper, the term "generative grammar" will be used to refer to the constellation of approaches whose intellectual roots lie in Chomsky's Syntactic Structures (1957) and Aspects of the Theory of Syntax (1965). That is, the term will be applied to those models that attempt to characterize the human linguistic capability in terms of grammars whose essential properties are universal and which consist of a set of formal, discrete, and interacting rules and
THE LANGUAGE MUDDLE
43
First, it stresses that it is not just Chomsky who is to blame for the mess that the field is in; rather, most twentieth-century linguists from Saussure onward have labored under the most profound misconceptions. Second, it is remarkable for the ignorance it manifests of the actual claims and results of generativist theorizing, to the point where it attributes to Chomsky positions that he is on record as rejecting. And third, even for a genre in which demolition characteristically outpaces construction, it is noteworthy for its failure to erect even the bare foundations of an alternative approach. After presenting in somewhat schematic form the outline of Harris's critique, we address some specific issues that we feel merit attention in virtue of their importance to the foundations of linguistic theory. According to Harris (see especially 1980; 1981a; 1987b; 1990c), modern linguistics is infected by the "language myth", which is composed of two interconnected elements. One is "telementation", the theory that communication is the mechanical transference of ideas from one person's mind to another "in accordance with a prearranged plan determined by [the grammatical] rules" (1981a: 11). The other is the "fixed code" theory of language, the theory of the "prearranged plan" itself This theory posits that pairings of sounds and meanings "remain invariant from speaker to speaker and from occasion to occasion..." (1990c: 29). The dubious achievement of generative grammar is its extreme mechanistic development of these ideas, extending them to encompass syntax (Chomsky 1957) and soon thereafter meaning itself (Katz and Fodor 1963). Harris does devote some space to substantive empirical criticism of the components of the language myth (to which we will return), though he is clearly far more comfortable casting aspersions on its disreputable origins. While it can be found in rudimentary form in Aristotle and Locke, "in its modern form [the myth] is a cultural product of post-Renaissance Europe. It reflects the political psychology of nationalism, and an educational system devoted to standardizing the linguistic behavior of pupils" (Harris 1981a: 9). Rising nation-states naturally saw the advantage of stable national languages. Hence the birth of the monolingual dictionary to "fix" sound-meaning pairings and normative grammars to ensure that only the speech of the privileged be sanctioned as correct. Linguists readily accommodated to such nationalist fervor by sanctioning a fixed-code view of language, which takes the dictionary as a model, and by enshrining grammars as sets of rules, thereby giving "scientific" cover to what amounted to an endorsement of the prescriptivist enterprise. principles. Any linguist committed to this program will be regarded as a "generative linguist", whether his or her research is devoted to grammar per se, or to some other area, such as psycholinguistics, pragmatics, sociolinguistics, or whatever.
44
ROBERT D. BORSLEY & FREDERICK J. NEWMEYER
But immediately a problem of chronology arises: nation-states were in the making from the Middle Ages onwards. Why did it take until the early twentieth century with Saussure,2 then, for the language myth to assume the status of scientific dogma? Harris explains: Not until a time when the European nation-state was itself in crisis as a viable social and political entity did European linguistics at last embrace a theory of languages which explicitly mirrored the ideally integrated, stable community which the nation-state would have liked to be (Harris 1980: 157). ...What had started out as a patriotic aspiration was eventually given the solemn blessing of modern science. (167) Harris could not be more explicit in linking structural approaches in linguistics to our century's more extreme attempts to save the capitalist nation-state in decline. To wit: The "ideal" speech community [which] is "totally homogeneous" [is] a fascist concept of languages if ever there was one (1983c: 119). Conveniently for those in power, then, the degree of abstraction inherent in the machine model "effectively precludes any possibility of raising issues concerning freedom of speech" (1990b: 153). Generative grammar has enjoyed an extra fillip to help propel it to prominence: it was able to latch on to the postwar infatuation with computers and high technology: Postwar over-optimism about building machines that could handle language made mathematical modelling of linguistic structure a very fashionable area of research. That was the context in which generative grammar was born (Gardner 1985). Suddenly, in the mid 1950s, linguistics seemed to have acquired a new engine which would propel it forward into the computer age; and, as a result, the model sold very well in the universities of the world for a quarter of a century. (Harris 1993b: 15) 2 While the present paper will focus on the extent to which the "language myth" is applicable to generative grammar, one might question whether Saussure held it in anything like the strong form in which it is attributed to him by Harris. True, a strong interpretation can be read into the Cours. But that book was a posthumous compilation from his undergraduates' lecture notes. We would hate to have our ideas judged from our undergraduate students' hastily scribbled interpretations of the already oversimplified material in our lectures to them! For more remarks along these lines, see Percival (1984).
THE LANGUAGE MUDDLE
45
A recurrent theme in Harris's writings is that the basic analytical units of modern linguistics owe their existence and characteristics to the fact that their originators used alphabetic writing. Despite linguists' protestations to the contrary, their work reflects an extreme alphabetic bias (see especially Harris 1980; 1986). The ability to segment speech into phones (or phonemes), the idea of the sentence as a linguistic unit (and therefore the possibility of sentential ambiguity), syntactic categories, and even the notion of the "word" reflect the effects of scriptism. Thus, the linguist's attributing "psychological reality" to these constructs is nothing but a vain attempt to accord scientific status to what are no more than cultural artifacts. Harris proposes to replace the linguistics of the self-contained system with an "integrational linguistics", which redefines linguistics as a mode of inquiry into the construction and articulation of our linguistic experience. It inquires not into the hypothetical structure of abstract linguistic systems, nor into their even more hypothetical representations in the human brain, but into the everyday integrational mechanisms by means of which the reality of the linguistic sign as a fact of life is established. For this purpose, in contrast to all previous linguistic programmes, it rejects any a priori attempt to circumscribe the phenomena of language or to draw a distinction between language and non-language which will be valid in each and every case. (1990c: 50; emphasis added) The remainder of the paper is organized as follows. Sections 2 and 3 deal with the two central aspects of the "language myth", telementation and fixed codes, respectively. Section 4 questions his account of the historical origins of current linguistic concepts and § 5 debunks the idea that they are linked intrinsically to alphabetic literacy. In § 6, we challenge Harris's contention that generative linguists are covert prescriptivists. In § 7, we suggest that an essentially empiricist world-view underlies much of Harris's critique of modern linguistics. Section 8 takes a brief critical look at "integrational linguistics", followed by concluding remarks in § 9.
2.
Telementation
Let us turn to the first component of the "language myth", the telementational conception of communication, the idea that communication is the mechanical transference of ideas from one mind to another, a mere process of encoding and decoding. In criticising this conception of communication, Harris is not alone.
46
ROBERT D. BORSLEY & FREDERICK J. NEWMEYER
The inadequacies of what they call the "code model of communication" is one of the starting points for the "relevance theory" of Sperber and Wilson (1986). We agree with them, in fact, that it is an untenable conception of linguistic communication. The question, then, is whether generative grammar is committed to such a conception. Harris's only support for the idea that it is comes from one decades-old passage from Katz (1966) that he cites over and over again. 3 This hardly constitutes an impressive body of evidence. We have found no more recent example of a generative linguist endorsing telementation. Chomsky, for example, has explicitly rejected the idea that communication involves a mechanical implementation of our mental grammars. As he has noted in a recent paper: ...in real world communication, virtually any information and strategy can be used to try to determine what some person is saying in a given situation. Furthermore, little knowledge need be shared by the speaker and the interpreter. (Chomsky 1991a: 18) One could not ask for a more explicit rejection of the telementational conception of communication. There is, in fact, one proposition about communication that is a central assumption of generative grammar. This is that key aspects of language are not reducible to facts about communication. The "autonomy thesis", as this proposition is known, distinguishes generative grammar from most so-called "functional" approaches to linguistics, where the communication-based grounding of grammatical phenomena is taken for granted.4 Chomsky has missed few opportunities to stress that language is only incidentally an instrument of communication; its "design features" do not manifest any signs of communicative ends (see, for example, Chomsky 1975b: 69; 1980: 230; 1991b: 51). The following passage, from Chomsky (1979), is typical: There is no reason to believe - to repeat myself once again - that language "essentially" serves communicative ends, or that the "essential purpose" of language is "communication", as is often said, at least if we mean by 3
Katz's position, here as elsewhere in his writings, is sui generis. As Harris notes, the theory is also expounded in Cairns and Cairns (1976), a popular introduction to psycholinguistics that takes a generative slant on some issues (but which cites no generativist work to support a telementational view). 4 For the idea that the needs of communication are the central force shaping grammatical structure, see Givon (1979; 1984; 1990). For defence of the contrary autonomist view, see Newmeyer (1983).
THE LANGUAGE MUDDLE
47
"communication" something like transmitting information or inducing belief. (P. 87) Lately, Chomsky has taken to claiming that the organization of grammar makes it ill-designed fox communication and that only a series of "computational tricks" allow it to be used to these ends at all (see for example, Chomsky 1991b: 49). The goal of generative grammar is not now and never has been to explain linguistic communication. For generative linguists, the central question is not How do speakers use language to communicate?, but rather How do speakers come to have the grammatical knowledge that they do? Generative grammar is concerned, for example, with the fact that an English-speaker knows that Which men does she expect to like each other? is acceptable, although Does she expect to like each other? is not; or the fact a speaker knows that John and him cannot refer to the same person in Did John expect to see him?, whereas they can in Who did John expect to see him? It argues that this knowledge is a manifestation of our "I-language", that is, our internalized linguistic knowledge, and that Ilanguage is in turn a manifestation of the principles and parameters that constitute universal grammar. One might or might not accept this explanation, but it is clear that it is not an attempt to explain linguistic communication. Since generative grammar does not seek to explain linguistic communication, it is under no obligation to propose an alternative to the telementational conception. However, it is interesting to note that Sperber and Wilson stress that their alternative inferential account of communication accepts the basic generativist assumptions (see Sperber and Wilson 1986: 94). It seems, then, that generative grammar is quite compatible with the most clearly articulated alternative to telementation. We must conclude, therefore, that the first of Harris's main criticisms is quite misconceived.
3. Fixed
codes
We now consider the second of the main criticisms that Harris advances against generative grammar: that it is committed to the mistaken view that languages are "fixed codes", defined as "some fixed set of correlations between ideas and verbal symbols" (Harris 1981a: 10), which remains invariant from speaker to speaker and from occasion to occasion within the sphere in which it operates. It is fixed in the sense in which the institutionalized rules of a game such as chess are fixed. (Harris 1990c: 29)
48
ROBERT D. BORSLEY & FREDERICK J. NEWMEYER
As noted in the previous section, no generative linguist since the 1960s has advocated the telementational theory of communication. So even if Harris is right that telementation entails a fixed-code view of language, generative linguists are not required to adopt such a view. Do they adopt one anyway? One might suppose that Harris is on firmer ground here. Chomsky sees an adult speaker's Ilanguage as constituting a "relatively stable steady state.. .which then undergoes only peripheral modification" (Chomsky 1986b: 25). In other words, he sees it as something (relatively) fixed.5 But is I-language a "code"? For Harris, a code is invariant from speaker to speaker. But generative linguists have never made such an assumption. Chomsky, for example, has repeatedly stressed that an I-language is an individual cognitive system that is only "partially shared by others in the various communities with which people associate themselves in their normal lives" (Chomsky 1991a: 19; emphasis added). In this sense, then, I-language does not meet Harris's criteria for being a fixed code. A code for Harris links "ideas and verbal symbols", i.e. meaning and form. Since under the mainstream approach to generative grammar, the principles-andparameters theory of Chomsky (1981; 1986b), representations of phonetic form (PF representations) are linked with representations of logical form (LF representations), there is a sense in which I-language is in fact a (relatively) fixed code. However, as we will argue below, it is not one in Harris's sense of the term, and hence the objections that he advances against the idea that languages are fixed codes are irrelevant to generative grammar. Chomsky's position is that at LF, certain aspects of meaning, notably anaphoric relations and quantifier scope, are made explicit. LF is a level that "interfaces...with conceptual-intentional systems" (Chomsky 1993a: 2), and thus 5
The assumption that one's I-language remains invariant from occasion to occasion throughout one's life has never been regarded as a particularly important one. On this topic, Chomsky has remarked: Suppose contrary to apparent fact, that the system never hit a steady state but kept growing (like a carp) until termination of life. I don't see that anything much would change, (cited in Patemanl987: 91, fn.16) In remarks germane to this issue, Emonds (1985: 238; 1986) and Pateman (1987: ch. 4) argue that many prescriptively-mandated prestige constructions acquired via schooling or parental correction do not represent changes to one's I-language at all, but rather only a tenuous overlay on it. One might also call attention to the debate in the field of second-language acquisition research over whether an adult-acquired second language, however fluent, is an I-language. For answers to the affirmative, see Epstein, Flynn and Martohardjono (1994); White (1985); the negative, see Bley-Vroman (1989); Clahsen (1988); Clahsen and Muysken (1986).
THE LANGUAGE MUDDLE
49
it helps to provide a basis for determining what a sentence means when it is used. The mechanisms that effect this determination, however, are proper to pragmatic theory, not grammatical theory (for discussion, see Kempson 1988). That is, generative grammar distinguishes between the grammatical encoding of meaning (sentence meaning) and the actual messages that are conveyed when language is used (utterance meaning). Utterance meaning is a result of the interaction between sentence meaning and the context of the utterance.6 Since LF is not equivalent to the actual message conveyed, Chomsky's view of grammar is not consistent with Harris's idea of a what a "code" is. Harris thinks that the view that sentence meaning and utterance meaning are distinct is untenable (see especially 1987a). But the only argument that we were able to detect in support of his position is that no discovery procedure reliably separates out the two levels of meaning. This is true, but quite irrelevant in any non-empiricist approach to language (for further remarks on this point, see § 7). Harris does attempt a series of empirical arguments against fixed codes (1990c: 32f), and it is worth reviewing them. We will argue that the first four are all in reality arguments against the telementational theory of communication, which (in order to effect a mechanical exchange of ideas) demands that speaker and hearer have precisely the same code. His fifth argument is simply an unsupported objection to the study of I-language. Harris begins by arguing that a a fixed code theory... must attribute exactly the same linguistic knowledge to A and Β if communication is to be successful. On this theory, therefore, it is impossible for anyone to come to know the meaning of a word by asking another person" (ibid.: 33). But this would be true only if communication were a matter of telementation. Since telementation is clearly an untenable conception, there is no reason why communication should demand identical codes. And, indeed, it manifestly does not; communication can take place even when codes differ to some degree. Thus, if the British co-author of the present paper were to say to the American co-author " You read Synonymy and Linguistic Analysis in one sitting? You couldn 't have done! ", the latter would understand him, even though the second sentence is not admitted by his Ilanguage - the auxiliary systems of our two varieties of English differ. It is an interesting question how different codes have to be before communication is 6
All approaches to meaning within generative grammar, including truth-conditional approaches, distinguish between sentence and utterance meaning. Since Chomsky (1979: 142) "doubt[s] that one can separate semantic representation from beliefs and knowledge about the world", he rejects truth-conditional semantics. Harris, of course, rejects it as well. His major publication devoted to the deficiencies of this approach (Harris 1981b), in appealing to the problems posed by "the practical limitations of speakers' knowledge and interests" (p. 110), echoes criticisms that Chomsky (1975a; 1979) had made earlier.
50
ROBERT D. BORSLEY & FREDERICK J. NEWMEYER
impaired; certainly, intelligibility (i.e. the successful conveyance of utterance meaning) is a function of cultural and personal factors, as well as the purely grammatical features of sentence meaning. Harris's second, third, and fourth objections are also based directly on telementation and its concomitant idea that communication between individuals requires that they have precisely the same code. He asks, for example, how the same fixed code could get established for all speakers, given their differences in linguistic experience; how innovation could ever happen, since that would lead to individuals having different codes; and how speakers and hearers could ever be sure that they were using the same code. But, again, since generative linguists reject the telementational theory of communication, these objections are without force as a critique of generative grammar. His "fifth objection is perhaps more powerful than any of these [others]" (ibid.: 35): If linguistics deals with synchronic speech-systems... and these systems are fixed codes, then they do not correspond to "languages" in the everyday sense in which English, French, and German are reckoned to be languages typically spoken by most people brought up in, say, the United Kingdom, France, and Germany. These are not fixed codes, whatever else they may be, because they are manifestly not uniform... Thus it appears prima facie that linguistics cannot deal with languages like English, French, and German; or if it does it cannot be dealing with fixed codes. Quite so! As Chomsky notes, "... the commonsense notion of language has a crucial sociopolitical dimension.... That any coherent account can be given of 'language' in this sense is doubtful; surely none has been offered or even seriously attempted. Rather, all scientific approaches have simply abandoned these elements of what is called 'language' in common usage" (Chomsky 1986b: 15). Grammatical theorists do not study "languages", but rather "grammars" the internalized systems of speakers. That is, they study I-language. But since Harris arbitrarily rejects the study of I-language, his only recourse is to retreat to another attack on telementation and its requirement of identical codes: Smith's English may not be the same English on all occasions... Worse still, if synchronic systems exist only on the idiolectal level, then ex hypo the si if Smith and Brown ever manage to engage in successful communication it will be sheer good luck. The identification of synchronic systems with idiolects is theoretically self-defeating for orthodox linguistics. It is no good for Smith to have a fixed code which is shared with no one else, (ibid.: 36)
THE LANGUAGE MUDDLE
51
Once again, there is no reason to believe, and every reason to doubt, that Smith's and Brown's I-language must be identical in order for them to communicate successfully.
4. The socio-historical
roots of formal
linguistics
Harris's theory makes a clear prediction about the countries in which structural linguistics (of any variety) should have flourished most and enjoyed the highest degree of "official" sanction. They are Germany and Italy. In the nineteenth century, as the patriotic movements in these lands struggled for belated nationstate status, a model incorporating the elements of the language myth would, one assumes, have served the interests that Harris feels such a model is inherently designed to serve. The intellectuals of the Weimar republic of the 1920s, a "European nation-state in crisis as a viable social and political entity", if any ever has been, should have jumped at Saussurean structuralism. And of course contemporary fascist Italy and (after 1933) Nazi Germany would have been expected to warmly embrace Saussure's "fascist" theory. Harris's predictions are completely wrong. The centers of structural linguistics in the inter-war years were all bourgeois democracies: Switzerland, France, Britain, Holland, Czechoslovakia, Denmark, and the United States. Structuralism in fascist Italy and Germany was not only not hegemonic, it was officially condemned. The egalitarian underpinnings of the approach - the idea that all languages are cut from the same mold and amenable to the same analytical methods - were incompatible with the idea of a master folk, destined to rule over the weak and the inherently inferior (for more discussion, see Newmeyer 1986). Another false prediction is that generative grammar (at least in its Chomskyan variety) should be most appealing to those with mathematical, computational or engineering interests as regards language. Hardly; meetings devoted to artificial intelligence and computational linguistics are among the most anti-Chomskyan milieux. Chomsky for years has infuriated the "technolinguists" by minimizing the importance of constructing a formalized theory (see, for example, Chomsky 1990). We also question the extent to which generative grammar has "sold very well in the universities of the world". The only country that comes to mind in which it is possible that a majority of linguists are, or ever have been, generative grammarians, is the Netherlands. It is still the case in some continental European countries that a wise career move for a young linguist is to write an "incisive" critique of generative grammar. One could speculate on why parliamentary democracy and structural approaches to languages have tended to go hand-in-hand. And one could
52
ROBERT D. BORSLEY & FREDERICK J. NEWMEYER
speculate on the reasons for the success of generative grammar in Holland and for its failure, say, in Belgium. But what point would such speculations serve? The merits (or demerits) of Saussurean structuralism, Chomskyan generativism, and, for that matter, Harrisian integrationalism should be evaluated on their own terms without regard for their real or imagined parentage. No doubt Harris's ideas, like those of the rest of us, are to whatever extent a product of their environment. They seem to be just the sort that one would expect to flourish in an institution (Oxford University) with long traditions of philological and philosophical discussion of language, but no real linguistics. Worth the reflection of a moment or two, perhaps, but quite irrelevant to whether they are right or wrong.
5. Alphabetic
literacy and linguistic
theory
That the constructs of linguistic theory owe their origins to alphabetic writing is stressed over and over again by Harris: ...the myth that spoken English includes a unit called "the sentence Mary had a little lamb " which somehow exists independently of any precisely identified sequence of English sounds. What is the basis of this myth? Simply the fact that Mary had a little lamb occurs as a unit in written English... (1980: 8). Without the transition from syllabic to alphabetic writing, the development of phonemic analysis in modern linguistics would be inconceivable (1980: 15). As is so often the case in his writing, Harris intends such remarks simultaneously to represent an analysis of the historical origins of a particular conceptual scheme and a guilt-by-antecedence negative evaluation of their adequacy.7 In the remarks that follow, we will try to separate the historical issue from the question of their theoretical motivation. In and of itself, the observation that phonemic theory arose in the cultural context of alphabetic writing is not remarkable or even interesting: so did physical theory, biological theory, and just about every other theory departmentalized in academia. Is there a causal link between the two - did the possession of 7
Another of Harris's rhetorical ploys is to impugn the motives of those with whom he disagrees: "Saussure was nothing if not an astute academic strategist. Having adroitly fended off the competing claims of experimental psychology and neurophysiology, he had no intention of surrendering language tamely to the teaching profession" (Harris 1987b: 111). The remarks in footnote 2 above seem particularly germane here.
THE LANGUAGE MUDDLE
53
alphabetic script help lead modern grammarians to discover phonemes in language? That is not out of the question: it is a commonplace observation of historians of science that the process of discovery is abetted (or hindered) by the cultural milieu of the investigator. However, as an examination of the ancient Indian linguistic tradition reveals, the possession of an alphabetic script on which great cultural value is placed is not a necessary condition for the development of phonological theory. The Indian grammarians not only grasped the phonemic principle (Allen 1953: 50, 89), but produced what are uncontroversially the most sophisticated phonological analyses before the mid-twentieth century AD. And yet controversy rages over whether Pânini, the greatest of them all, was even literate! Evidence for the use of writing in India before Asoka (third century BC) is vague and inconclusive; Pânini dates at least from the 4th century, and possibly earlier. Furthermore, even if writing did exist at the time of Pânini, no cultural value was placed upon it; it seems to have been used for centuries simply to record commercial transactions, but not for important things like religion, literature and linguistics. And the Asokan scripts, Kharosthi and Brâhmï, were not even fully alphabetic; they treated each CV sequence as an orthographic unit, with obligatory vowel diacritics attached (Bright 1988). In a rare acknowledgment of the existence of the Indian grammatical tradition, Harris (1980: 119-122) recognizes that the Indian interest in phonology arose from a desire to preserve exactness of phonetic detail in the recitation of the Vedic texts. In other words, we see that the same advance in our knowledge can arise in more than one historical context and for different underlying reasons. Just as one might wonder whether phonemic analysis would have ever been developed by the Japanese or the Aztecs, it is (necessarily) an open question whether the insights of other fields would have arisen in a different historical context. For example, would the principles of modern chemistry have been developed if they had not been preceded historically by many centuries of alchemical attempts to transmutate base metals into gold? Who knows? All that matters is the degree to which such principles successfully model reality. In any event, the claim that current linguistic theory (whatever its roots may be) mimics alphabetic writing in some relevant sense strains credulity. The segment is the only phonological construct that evokes alphabetism at all. Yet it is composed of distinctive features and may be mutated by rule into other segments. Neither distinctive features nor sequential processes have any analogue in alphabetic script. Furthermore, the segment has played an increasingly diminished role in phonological theory. Constructs such as the strict-cycle condition of lexical phonology, principles of syllabification, autosegmental linking principles, and the metrical grid have no analogues in alphabetic script,
54
ROBERT D. BORSLEY & FREDERICK J. NEWMEYER
yet form central parts of an overall approach to phonology in which the classical phoneme is, at best, epiphenomenal. In one of his less prescient passages, Harris writes: Any quite radical approach to phonology, such as was suggested by J. R. Firth, had to face an impossible uphill struggle against the solid conservatism of the alphabetic tradition. Among the ranks of phonologists who remained faithful to that tradition, the more die-hard objected to any innovation likely to disturb the reassuring parallelism between phonological analysis and alphabetic representation. (1980: 11) One might argue that this passage exaggerates contemporary resistance to Firth's "nonalphabetic" theory. After all, Zellig Harris, arguably the leading American structuralist theoretician, put forward a proposal for "simultaneous components" in phonology (Z. Harris 1944) that bore considerable resemblance to Firth's. But as far as the "impossible uphill struggle" is concerned, it should be pointed out that Harris's remarks date from several years after the appearance of Goldsmith (1976) and Liberman and Prince (1977). By 1980 these works had already changed the research agenda of theoretical phonologists in a direction that was able to directly incorporate the insights of Firthian prosodic analysis. 8 To our knowledge, none of Harris's subsequent publications acknowledge the enormous success that this change of direction has come to enjoy. 9 If there is one feature common to all levels of grammatical description in current theoretical work, it is hierarchy. If there is one feature absent from alphabetic writing systems, it is hierarchy. Syntacticians have no interest in strings of words per se, as an orthographically-influenced approach would suggest. Rather it is the hierarchical relations of grammatical elements (encoded by such structural constructs as "c-command", "dominance", and so on) which form the primitive vocabulary in which principles of phonology, morphology, and syntax are framed. This has been true to a large extent since the earliest days
8
For interesting discussion of the extent to which current auto segmental and metrical phonology are compatible with prosodie analysis, see Goldsmith (1992). 9 Of the twenty-nine publications of his that we reviewed, not a single one cites papers in generative phonology or morphology. Aside from Harris's two-column commentary on Lightfoot (1989), the only work in generative syntax published after 1970 that he cites are four papers defending fuzzy grammar. And he cites none of Chomsky's major theoretical works published after 1975 (e.g. 1980; 1981; 1986a; 1986b; 1988). Harris, obviously, is free to read and cite whom he chooses. Yet one would expect a scholar whose expressed goal is to overturn the corrupt foundations of the field to exhibit greater erudition as to what precisely those foundations are that need overturning.
THE LANGUAGE MUDDLE
55
of transformational generative grammar, though, to read Harris, one would never know it. There is work that could be taken to suggest that phonemic theory has an alphabetic bias (a conclusion reached by Olson 1993), though, curiously, Harris does not cite it. Work by José Morais and his associates (Morais 1985; Morais, Bertelson, et al. 1986; Morais, Cary, et al. 1979) demonstrates that alphabetic literates are far more adept at segmentalizing words into distinct sounds than are illiterates.10 It is not clear what bearing this fact has on the question at hand. It is well known that with training, normally submerged mental constructs can be brought to the level of consciousness. Arithmetical abilities before and after one has been explicitly taught the principles of addition, multiplication, and so on provide an example. Before the teaching, they are there latently; even innumerates can perform arithmetical operations when they have to. Yet explicit teaching of the principles makes it vastly easier for them to do so. The mastery of alphabetic writing appears to provide analogous "training" in segmenting speech into its component discrete units.11 The literature is full of references to prodigious feats of grammatical analysis by those illiterate in the language under investigation. Consider, for example, Sapir's Southern Paiute consultant Tony Tillohash and his Sarcee consultant John Whitney (Sapir 1963 [1933]), who would correct his phonetic transcriptions in favor of the phonemic ones that they "heard", i.e. were represented mentally. Since neither Southern Paiute nor Sarcee were written languages, they could hardly have been influenced by orthography. Linguistic s a y often yields subtle insights into the internalized grammars of the players. Consider, for example, the following description in Conklin (1959) of a variety of language games played by the pre-literate Hanunoo of the Philippines: The first type [of game] results from the simple rearrangement of the phonemic content of polysyllabic forms so that the initial CV of the first stem syllable is transposed with the first CV of the second syllable. Prefixes, monosyllabic particles, and vowel length position remain unchanged. In trisyllabic stems, the first and last (third) CV are transposed: a form such as balaynun 'domesticated' is pronounced nulayban.
10
Along the same lines, Read et al. (1986) show that Chinese speakers who have mastered the alphabetic pinyin script are better at segmentalization tasks than those who have not. 11 Derwing and Nearey (1986); Derwing, Nearey and Dow (1986) provide experimental evidence that suggests that speakers do have the intrinsic ability to segment their speech into phonological units.
56
ROBERT D. BORSLEY & FREDERICK J. NEWMEYER In the second type... there is a simple substitution of a particular consonant phoneme... for the final consonant phoneme of each polysyllabic form. The third type... involves short prefixation with partials, i.e. with only the first CVC of each word. The fourth type... involves suffixation with partials, i.e. with only the first CV of each polysyllabic form, not counting prefixes. The fifth type... involves reduplicative suffixation with partials (which consist of the initial CV of each stem-syllable of polysyllabic forms), and the addition of the segment tag sa 'goes to', or pause, between the resulting forms. (pp. 631632)
The deft abilities of the Hanunoo at this complex form of language play bear witness, in our opinion, to their having a mental phonemic representation of each word in the language, which they are able to manipulate at will. We challenge Harris to describe these five games in a vocabulary that does not make reference, explicitly or implicitly, to the constructs of grammatical theory. Psycholinguistic research, as well, bears out the grammarians' hypotheses that their constructs are not mere orthographic after-effects. To cite only two examples from a vast literature, Lahiri and Marslen-Wilson (1991) show that speech processing makes use of underlying phonological representations, not surface phonetic ones, and Levelt (1989) demonstrates that syntactic categories such as "sentence", a noun phrase", and so on are called upon by speakers in speech production. In short, we conclude that Harris is wrong to insist that the units of grammatical theory are a by-product of the conventions of orthography. To a very great extent, it is the other way around. Phonological segments, syllables, morphemes, words, sentences, and so on are represented by orthographic units in the languages of the world because they represent real aspects of our mental grammars.
6. Generative grammar as a prescriptive
enterprise
As we noted above, Harris maintains that the grammatical theorists' rules represent a covert form of prescriptivism. He claims that: The normative grammarian is no sooner evicted from paradise than his intellectual stock-in-trade is confiscated by his evictors. The very rules he was condemned for peddling turn up on the linguist's stall, sometimes - but not always - with new labels on them. (1987b: 128-129; emphasis added)
THE LANGUAGE MUDDLE
57
Characteristically, Harris fails to provide any examples of these "very rules", so we can only guess what they might be. Normative grammarians have peddled rules such as: Never end a sentence with a preposition; Say It is f not it's me, Avoid double negatives; Use which as a relative marker, not that; and Do not split infinitives. Which of these have turned up in our stall? Where specifically? What are their new labels? Harris suggests that "any categorization of usage into 'grammatical' and 'ungrammatical' involves a social value judgment of some kind, and it is idle to pretend otherwise" (1987b: 143; emphasis in original). One wonders if he has ever looked at the sort of data that linguists consider. What sort of social value judgment could be involved when a linguist points out that John is believed to be mad and Mary seems to the men to like herself are grammatical, while John is believed is mad and Mary seems to the men to like each other are not? Harris does give a couple of examples of what he considers normative judgments by supposedly descriptive grammarians, though they hardly represent traditionally-peddled proscriptions of linguistic behavior. For example, he responds to the assertion in Lightfoot (1989: 324) that null-subject sentences like ho trovato il libro are commonplace in Italian, while English-speaking children fail to hear corresponding * found the book, as follows: How many non-English-speaking linguists have been taken in by this notorious example it would be difficult to estimate. The fact is that found the book (with no subject pronoun) exemplifies a combinatorial pattern frequently encountered in colloquial English (and hence, presumably, heard by English-learning children), particularly in response to questions. (Harris 1989a: 345). Harris's observation is correct as far as it goes. Some registers of English do allow subjectless finite clauses under some circumstances. Indeed, this "English diary register" has been subject to a careful study by Haegeman (1990). But Lightfoot's point was that no register of English allows subjectless finite clauses under the conditions that a language like Italian does. As Haegeman notes, the diary register does not allow null subjects in subordinate clauses or in clauses in which some constituent is preposed. Thus, the following are impossible: (1) a. * I said that is a genius. b. * When will be able to see him?
58
ROBERT D. BORSLEY & FREDERICK J. NEWMEYER
This is quite different from Italian (and other null-subject languages), in which null subjects are possible under both these circumstances. We have examples like the following in Italian:12 (2) a. Dissi che è un genio. said-lSG that is-3SG a genius Τ said that he/she is a genius.' b. Quando lo potro vedere? when him be-able-lSG see-INF 'When will I be able to see him?' Thus, there is a major difference between English and Italian. The facts are more complex than Lightfoot's unelaborated remarks suggest; one might accuse him of oversimplification, but it is very far fetched to accuse him of covert prescriptivism. Another example that Harris cites of a prescriptive rule masquerading as a description is "the adjective agrees with its noun in gender, number, and case" (1987b: 130). Now what exactly is the problem here? If an analysis of German leads in that direction, why shouldn't we say that German has an agreement rule? Is it the word "rule" that bothers Harris? If so, we could no doubt find an alternative. Indeed, it is worth noting that in recent years, generative linguists have appealed far more to "constraints" and "conditions" than to "rules", a fact that Harris does not remark upon. In The Language Machine, Harris remarks that "True is a word we use primarily to express our agreement with someone else's assessment of the situation or, somewhat differently, to make it clear that we do not think that person is lying" (1987b: 160). One might wonder what makes such a proposition acceptable to Harris, while he will not tolerate the statement about German adjective agreement. Our guess is that the crucial point is that the former, but not the latter, is phrased in terms of what people do, i.e. it directly describes linguistic behavior. We suspect that Harris would be happy if the German adjectival agreement rule were replaced by a statement about what German speakers do. That would be more consistent with his empiricist belief, which we will document in the following section, that the only thing that is "real" is behavior itself, not the principles underlying it. Prescriptivism is, at root, an attempt to stop people from what they are doing and to get them to start doing something else. How the examples that occur in our 12
We are indebted to Ian Roberts for providing the Italian examples.
THE LANGUAGE MUDDLE
59
technical writing might play even a miniscule part in that endeavor is not revealed in any of Harris's publications that we have had the opportunity to read. It would seem to follow that if generative linguists are covert prescriptivists, then they would be careful to eschew any analysis of non-standard forms and dialects.13 This is hardly the case; two of the major debates in generative syntax, for example, have concerned the conditions under which want to might be contracted to wanna in informal speech and those under which prepositions can be "stranded" at the ends of sentences (see Postal and Pullum 1986 and Kayne 1984 respectively for discussion and reviews of the literature). One recent year saw analyses of the following non-standard dialects in Linguistic Inquiry and Natural Language and Linguistic Theory, the two leading journals of generative grammar: Bavarian German (Reis and Rosengren 1992); Haitian Creole (Deprez 1992); Belfast English (Henry 1992); Northern Italian vernacular dialects (Suñer 1992); Ottawa-Hull French (Reed 1992); American Sign Language (Perlmutter 1992); and Northern Palestinian Arabic (Shlonsky 1992). And it should hardly need to be mentioned that dozens of languages were analyzed which have never been "standardized" at all. One might also wonder what sort of prescriptivism is inherent in the work of generative linguists who work on dead languages (see Dresher & Lahiri51991and Mester 1994 for recent examples) or in that of those who attempt to provide explanations of language change (e.g. Roberts 1985; Lightfoot 1991). Are we to believe that Dresher & Lahiri and Mester were "prescribing" usage for speakers of Old English and Latin respectively? And which stage of English was Lightfoot prescribing in his explanation of the rise of accusative subjects: that prior to their rise or that following it? We must therefore conclude that Harris's claim that generative grammar is prescriptive grammar in disguise is utterly without foundation.
7. Harris's
empiricism
As we have noted, Harris's principal objection to the idea of sentence meaning as something distinct from utterance meaning is the fact that no reliable set of tests or procedures allow the former to be abstracted from the latter. Now, the rejection of abstractions on the basis of the impossibility of deriving them by a I 3 To the best of our knowledge, the first publication that argues that an analytical overemphasis on standard dialects by theoretical linguists reveals a prescriptivist bias is Newmeyer (1978 [1973]). This work is not cited by Harris. However, the increased attention to nonstandard dialects in the intervening twenty-one years has rendered the bulk of the critical remaiks of that paper inapplicable.
60
ROBERT D. BORSLEY & FREDERICK J. NEWMEYER
mechanical procedure from a series of observations is uncontroversially the hallmark of an empiricist view of reality. While Harris never explicitly embraces empiricism, he gives himself away, as observed in Percival (1984), when he argues that "modern linguistics has proved to be no science at all". Harris writes as follows: The abstract object of knowledge, "language", is at a double remove from direct observation. It is once removed in that we cannot hope to see the workings of language except insofar as they are manifested through particular languages. But it is twice removed in that particular languages as such are not directly available to observation either. All that we can directly observe (in the sense in which "direct observation" is conducted in the natural sciences generally) are specific events, utterances, inscriptions and reactions to them by members of a linguistic community. (Harris 1981a: 44) Harris appears to assume that legitimate science is limited to classifying the products of direct observation. If that were true, then physics, as Chomsky (1972: 65) noted, would be transformed into a science of meter readings. However, numerous philosophers of science have emphasized that the goal of science is to identify the mechanisms underlying observed phenomena. This is stressed particularly forcefully in the work of Roy Bhaskar, who writes that "the objects of knowledge" are "the structures and mechanisms that generate phenomena" (Bhaskar 1975: 25) and that "...science is concerned with the behaviour of things only in as much as it casts light on their reasons for acting..." (1975: 212). This is very much the view of generative grammar. 14 Harris's empiricism also reveals itself in his reaction to Chomsky's contention that linguistic competence is just one factor involved in linguistic performance. Harris calls this an "ingenious compromise" (1981a: 109), as if it were some kind of trick. But the recognition that observed phenomena typically reflect the interaction of many different mechanisms is a fundamental feature of science. As Bhaskar notes, "The world consists of mechanisms, not events. Such mechanisms combine to generate the flux of phenomena that constitute the actual states and happenings of the world" (1975: 47). Only an empiricist who does not believe in anything underlying observed phenomena would see something dubious here. Since Harris rejects in principle the idea that actual behavior is the product of the interaction of a variety of different underlying mechanisms, he typically interprets any proposal regarding the properties of such mechanisms as being a 14
See Pateman (1987: Ch. 1 and 2) for relevant discussion.
THE LANGUAGE MUDDLE
61
claim about behavior itself. Nowhere is this as evident as in his reaction to Chomsky's reference to the "ideal speaker-listener, in a completely homogeneous speech-community, who knows its language perfectly" (Chomsky 1965: 3). Harris appears to interpret this statement as being literally a claim about the observable attributes of speakers.15 Thus he has Chomsky believing ludicrous ideas such as the following (emphasis has been added in both quotes): [Speakers have] the ability to produce an infinite set of flawless sentences. (1980:18) By definition, [the ideal speaker-listener] speaks "the language" perfectly. (1981a: 33) In some passages Harris does appear to understand that Chomsky's view is that the mental grammar is just one of many factors affecting verbal behavior. But in others, amply illustrated above, his empiricism leads him to forget this view and to equate cognitive principles with the actual behavior that they underlie. As we noted in the last section, Harris's view that generativists are covert prescriptivists seems to be a manifestation of an empiricist outlook. That is, such a view appears to stem from the belief that all that is real is behavior. From this perspective, if one is not describing behavior, then one must be prescribing it. Since generativists are clearly not describing behavior, the conclusion would seem to follow that they must be prescribing it. The logic here is sound, but only an empiricist would subscribe to the crucial basic premiss. We also suggest that empiricist assumptions underlie Harris's curious belief that the generativist position is that "linguistic rules are rules which we have no option but to obey" (1987b: 136-137) and that language is "the product of mysterious inner machinery over which human beings have no control" (ibid.: 171). The real mystery is where Harris came up with such an idea. There is absolutely no suggestion in Chomsky's writings or in those of other generative linguists that we have no option but to obey linguistic rules. Chomsky (1986b: 261) in fact argues that the cognitive system involved in use of language is "cognitively penetrable" in the sense of Pylyshyn (1984) and other current work; that is, our goals, beliefs, expectations, and so forth clearly enter into our decision to use the rules in one way or another, and principles of rational inference and the like may play a role in these decisions. 15 For discussion of the idealizations embodied in the concept of the "ideal speaker-listener", see Newmeyer (1983:73-76).
62
ROBERT D. BORSLEY & FREDERICK J. NEWMEYER
It is worth noting that this idea is not new in Chomsky's work. For example, in Chomsky (1961), he observes that striking effects can be achieved by the use of grammatically deviant sentences. Obviously, one cannot produce grammatically deviant sentences if one has no option but to obey the rules. Thus, for Chomsky, linguistic rules are just one factor involved in the use of language. It seems to us that Harris's inability to understand this is another manifestation of his empiricist outlook. There is hardly a publication of Harris's that we reviewed in which his empiricism does not come to the fore. For example, it is evident in his attempted deconstruction of the "word" (1972), the "morpheme" (1973b: Ch. 3), and the "phoneme" (1981a: 58f.), which are based explicitly on the demonstration of difficulty in providing discovery procedures that reliably identify them. But most linguists have accepted since Chomsky (1957) that it is unreasonable to expect linguistic theory (or any other theory, for that matter) to provide discovery procedures by which theoretical constructs might unfailingly be isolated. In short, Harris's critique of modern linguistic concepts represents a regression to a form of empiricism more extreme even than that advocated by Leonard Bloomfield and his followers. After all, the Bloomfieldians were willing to admit a pretheoretical distinction between vocalizations that belong to language and those that do not. Harris (1981a: 165; 1990c: 50) is not willing to do so.
8. Integrational
linguistics
Harris's notion of an integrational linguistics, in which linguistics is redefined as "a mode of inquiry into the construction and articulation of our linguistic experience", is hardly "in contrast to all previous linguistic programmes" (Harris 1990c: 50). The call for such an approach in linguistics dates at least from early 1970s work in ethnography of speaking and conversation analysis. So, for example, in 1974 Dell Hymes devoted pages to the same criticisms of "rigid dichotomies" in linguistics that Harris's writings would echo years later. Hymes outlined a program for linguistics that, among other things, called for an approach to "performance as accomplishment and responsibility, investiture and emergence" and to "languages as what their users have made of them (not just what human nature has given)" (Hymes 1974: 206). Even earlier, the papers in Gumperz and Hymes (1972) had (in the words of Duranti 1988: 222) stressed "the role of speech in creating context, the need to take the participants' perspective in the analysis of their interaction, the cooperative nature of verbal communication - the latest feature being related, but not identical, to the claim of the emergent nature of (some aspects of) the social order".
THE LANGUAGE MUDDLE
63
This work has spawned an enormous literature, which is all but ignored by Harris. Likewise, we have found no citations to the "Universalpragmatik" of Habermas (1976); to A. L. Becker' work on "text-building" (e.g. Becker 1979); or to Paul Hopper's "emergent grammar" (Hopper 1987; 1988), even though these programs are, in relevant respects, identical to Harris's, and a good deal better worked out. Even in 1990, practically a decade after his first call for an integrationalist approach (1981a), all Harris could talk about was what such an approach "would" do or "will" do, not about what it has done (1990c: 45ff). By way of comparison, imagine Chomsky writing in 1966, nine years after the appearance of Syntactic Structures, about what generative grammar might be able to accomplish in the future. A bibliography published the previous year (Dingwall 1965) listed no fewer than 962 publications written within the framework of the theory or addressing its claims. We, in fact, do find aspects of Harris's program to be of interest. For example, his advocacy of an investigation of the semiology of written texts (Harris 1984b) strikes us as a reasonable one. And we have no quarrel in principle with investigating the extent to which "communication [is] a function of the individual's experience in the context of a given situation" (Harris 1990c: 50), though our interpretation of what that would involve would most likely differ from Harris's interpretation. But the point is that such undertakings do not challenge to one slight degree the fundamental principles of generative grammar. Again, strictly linguistic knowledge (I-language) is but one system of many involved in actual use of language for communicative purposes. This point can be driven home, perhaps, by an examination of Harris's contention that the underlying assumptions of modern linguistics "effectively preclude[s] any possibility of raising issues concerning freedom of speech" (1990b: 153). Now, whether Harris knows it or not (and there is no reason to think that he knows it), Chomsky has devoted dozens, if not hundreds, of pages to exposing the manipulative use of language, in particular by the leaders of the American political establishment and their apologists, and to showing that the term "freedom of speech" is often used as a protective cover by those who would wish to deny it to others. In Chomsky (1986b), for example, he exposes the use of terms like "aggression", "doves", "hawks", "peace process", and "terrorism" in American political discourse. Surely, the fact that Chomsky sees no inconsistency between such work and his grammatical theorizing (to the point where both occur within the covers of the same book) is prima facie evidence that the study of formal grammars and their properties complements, rather than challenges, the study of language in its communicative setting. It is tempting to draw a moral from the fact that after fifteen years of rhetoric, there is no field of integrational linguistics. Successful linguistic programs, just
64
ROBERT D. BORSLEY & FREDERICK J. NEWMEYER
like successful programs for research in other disciplines, have always been prepared to draw distinctions and to concentrate on some phenomena at the expense of others. Integrational linguistics has not been willing to do so, and has as a result doomed itself to sterility.
9. Concluding
remarks
In one way, Harris's writings are more impressive than those of other critics of generative grammar. Their broad historical sweep and their breadth of reference distinguishes them from the typical critique, which tends to focus on one or two perceived theory-internal defects or inconsistencies. In another way, however, what Harris has to say is far less impressive than the run-of-the mill criticism. As we have seen, it is remarkable for its myopic focus on what he takes to be the corrupt pedigree of the theory and its concomitant refusal to attend to what generative linguists actually say. Despite the fact that Harris, unlike many critics, has a significant band of followers, there is no sign that they are developing an alternative approach to linguistics.16 "Integrational linguistics" is no more than a name for a promissory note. There is no likelihood, then, that Harris will have any impact within linguistics. This being so, many of our colleagues would no doubt ask us: Why bother to discuss his ideas at all? One answer, of course, is that doing so is an interesting intellectual exercise. But there is a more important answer. Our impression is that Harris's writings are mainly read by non-linguists with an interest in language. We think it matters how linguistics is perceived by outsiders. It would be unfortunate indeed if the picture he paints of generative grammar were to be seen as an accurate one. We hope to have demonstrated that it is a seriously distorted picture, based on false premises, profound misunderstanding, and defective reasoning.
16
Papers in Davis and Taylor (1990), in particular those by Nigel Love and Talbot J. Taylor, present a Harrisian perspective on some issues. For a critique, see Borsley (1992).
3 Telementation and Generative Linguistics Philip Carr
1. Introduction One aspect of what Roy Harris (1981a, 1987b) calls "the language myth" is the idea that communication between humans may take the form of the mechanical transference or transmission of ideas from one human being to another in accordance with a largely shared and fixed code. Harris refers to this putative process as "telementation" and denies (with considerable justification) that it exists. Harris also takes it that generative linguistics subscribes to this myth, and thus concludes that the generative enterprise is fundamentally misconceived. In this paper, I will seek to show, following Burton-Roberts (1994) and Burton-Roberts & Carr (in preparation), that the current "minimalist" conception of I-language (Chomsky 1992), under which I-language is conceived of as generating linguistic expressions (structural descriptions) ! which are instructions for performance systems, does indeed commit the generative enterprise to a version of the telementation thesis. I argue that Harris is right to deny the existence of such a phenomenon, and that a strict "internalist" conception of language, of the sort that Chomsky has all along been seeking to elaborate, conflicts with the telementational thesis which current work finds itself committed to. I will focus on the notion that linguistic expressions include the level of representation known as Phonetic Form (PF), details of which are spelled out, as Chomsky (1992: 4, footnote 3) indicates, by Bromberger & Halle (1989). I will 1
As will become clear, the term "linguistic expression" is somewhat unfortunate with respect to what follows, in that the verb "express" suggests externalisation of something internal, a telementational notion which, I will be suggesting, is inappropriate to a strictly internalist conception of language. "Structural description" is no better, since it is founded on a two-place predicate under which the S.D. is a description of something, which is also unsatisfactory, for the same reason. I prefer the term "linguistic object", but I continue to use "linguistic expression" in order to connect what is being said here with current generative literature. The reader should think of "expression" as in "mathematical expression".
66
PHILIP CARR
examine those details and seek to show that, in conceiving of PF either in terms of instructions to an artidilatory performance system, or as articulatory intentions, generative theory adopts a version of the telementation thesis. My general conclusion, again following Burton-Roberts (1994) and Burton-Roberts & Carr (in preparation), will be that a strict internalist conception of language, and with it the innateness hypothesis, is sustainable only if linguistic expressions are not conceived of as instructions to performance systems. In brief, generative thinking on the internalist conception of language has been less radical than it might have been, and has thus laid itself open to Harris's valid accusation that generative linguistics subscribes to the myth of telementation. However, a non-telementational generative linguistics is perfectly possible, if the minimalist programme becomes somewhat more minimalist, and if a new, non-instructional conception of linguistic expressions is adopted. The paper is set out as follows. In section 2,1 stress the fundamentally occult nature of the telementational thesis. In section 3, I introduce the current "Minimalist" model of generative grammar. I then seek to show two things. Firstly, that the current generative conception of linguistic expressions as instructions to behave or intentions to behave is inconsistent with the Chomskyan view that language is entirely internal and not designed for use. Secondly, and more specifically, that the level known as PF is inconsistent in this way, is telementational in nature, and is profoundly problematic. I end, in section 4, with some concluding remarks on the generative understanding of realism and its relation to the notion "linguistic behaviour", subscribed to by both Chomsky and Harris, suggesting that such a notion is an empiricist notion, and inconsistent with the "internalist" ("I-language") conception of language as espoused by Chomsky (1991a, 1991b, 1992, 1993b, 1995).
2. The occult nature of the telementational
thesis
I begin by assuming that human beings do not possess powers of telepathy. By "telepathy", I mean the transmission by one human being of propositional content2 to another without engaging in any observable behaviour. Let us take it
2 I appreciate that Harris and his followers almost certainly do not accept that there are such things as propositions. But one has to start somewhere. I will be assuming that human beings are capable of engaging in mental activities which can reasonably be described as "entertaining propositions". This seems uncontroversial: it is hard to see how deductive or inductive reasoning, which surely exist, can be described other than in propositional terms.
TELEMENTATION AND GENERATIVE LINGUISTICS
67
that this assumption is well-founded.3 The "telepathy" idea is clearly an occult idea. Intriguingly, however, the idea that we may transmit propositional content by means of observable behaviour (telementation) is not generally taken to be an occult idea. And yet, on any kind of close inspection, the telementation idea is strikingly occult. If we subscribe to the telementation idea, we are committed to claiming that propositional content can somehow be "encoded" such that objects which are claimed by generativists to be entirely cognitive in nature (i.e. linguistic expressions) can be "transformed" or "transmogrified" into something externally physical and observable, in the sense of being external to the mind/brain. This idea of transmogrification is, I will seek to show, central to the notion of the "externalisation" of the internal (and indeed "internalisation" of the external). Both notions, externalisation and internalisation, are central to the current Minimalist Programme, as I will seek to show below: speakers are said, in current generative work, to have internalised an I-language. This clearly implies that the I-language was in some sense external in the first place, and yet generativists deny that language is external. The point is an important one: if "I-language" means "internalised language" (and almost all generativists, including Borsley and Newmeyer, in their contribution to this volume, say that it does) then language cannot be entirely internal. If, on the other hand, language is entirely internal, then one cannot coherently speak of it as having been internal/serf. Something has to give, and for the generativist, it ought to be the twin ideas of internalisation/externalisation. If the linguistic is entirely internal then that which is external cannot be linguistic. Thus, there can be no aspects of behaviour which are linguistic in any coherent sense: there can be no such thing as linguistic behaviour. The idea that there can be a transmogrification of cognitive objects into external, observable behaviour (externalisation), and a reverse transmogrification in the opposite direction (internalisation), is more occult than the alchemist's claim that base metals may be transmuted into gold, or the Roman Catholic notion of trans-substantiation, since base metals, gold, wine and blood are at least all physical types whose tokens are externally observable, and at least some physical transformations are known to be possible. But to posit an ontological distinction, as generative linguists do, between cognitive states and 3 Even if experiments such as those carried out at Edinburgh University under the guidance of its Professor of Parapsychology did demonstrate possession of some kind of "parapsychological" powers by some individuals (and it is a matter of some controversy whether they do), such experiments never, apparently, involve "transmission" of large sets of arbitrarily chosen propositions, such as, say "The Chancellor will consider lowering interest rates if inflation targets are met within the forthcoming twelve month period". I take it that not even parapsychologists make the claim that this is possible.
68
PHILIP CARR
observable behaviour, and then allow that the one can be transformed into another, is to allow that transmutation can occur across quite distinct ontological categories. It is clear that any generative linguist would wish to deny that current generative theory is committed to this occult transmogrification proposition. It is my purpose, in what follows, to show that current generative thinking is indeed committed to it, and to suggest that it need not be.
3. The Minimalist
Programme
and problems with PF
3.1 The Minimalist Programme For Chomsky (1992, 1995 and elsewhere), the central component of the language faculty is I-language (which Chomsky identifies with language), where "I" is translated variously as Internalised language or Intensional language.4 A speaker's I-language is said to generate linguistic expressions (known as structural descriptions). Each linguistic expression is a sequence of representations, one at each linguistic level, the levels provided by Universal Grammar. The I-language is said to be embedded in performance systems, which "enable its expressions to be used for articulating, interpreting, referring, reflecting and other actions" (Chomsky 1992: 2). Despite the fact that, for Chomsky, language is not "designed for use", he thinks of linguistic expressions as complexes of instructions for these performance systems (Chomsky 1992: 2, 1993b: 184, 1995: 15): he "expects to find connections between properties of the language and the manner of its use" (1992: 3). Here lies Chomsky's fatal compromise: he ought not to expect this. It is in this instructional conception of linguistic expressions that the inconsistency between the internalist conception of language and the role and nature of linguistic expressions is to be found: if linguistic expressions are said to be instructional in nature, then that is tantamount to allowing that they are indeed "designed for use". Two major sorts of performance system assumed by Chomsky concern articulation and perception on the one hand and conceptual/intentional activity on the other. Linguistic expressions are said to contain instructions for each of these. The instructions to the articulation/perception performance system are referred to as Phonetic Form (PF) and those to the conceptual/intentional system are 4
For the reasons given briefly above, which are elaborated at greater length by Burton-Roberts (1994) and Burton-Roberts and Carr (in preparation), the term "Internal language" is preferable, and the terms "Internalised language" and "Intensional language" inappropriate to and inconsistent with a strictly internalist conception of language of the sort Chomsky seeks to develop.
TELEMENTATION AND GENERATIVE LINGUISTICS
69
generally known as Logical Form (LF). Chomsky (1992: 3; 1995: 15) also assumes that an I-language contains two components: a lexicon and a computational system, whereby the lexicon "specifies the items that enter into the computational system" (1992: 3) which uses these items to generate linguistic expressions. 5 Thus, an I-language "generates an infinite set of linguistic expressions, each a pair drawn from the interface levels (PF, LF) respectively." (1992: 6; 1995: 15). I now consider, in some detail, the nature and function of PF as currently conceived, and the problems associated with that conception. 3.2 Type, token and telementation in PF The distinction between the linguistic, characterised as cognitive in nature, and behaviour has been fundamental to generative linguistics since its inception. The Chomskyan notion of the linguistic, conceived of as "knowledge", is a static conception: to "know a language" is to be in a certain cognitive state. In the expression "to know a language", "know" is not to be interpreted as a two-place predicate: it is not that there is some language or grammar, subsisting external to the individual, which that individual gains knowledge of; rather, "know" is to be interpreted as a one-place predicate. It is therefore unfortunate that the linguistic should be referred to by Chomsky as "internalised language", since that implies the making internal of something which exists outside of the individual, and that is not what is intended by Chomsky. The expression "knowledge of language" is also rather unfortunate,6 since it implies that linguistic knowledge is knowledge of something, but that is not what Chomsky intended "know" to convey; thus the well-known Chomskyan expression "to cognise", rather than "to know". Similarly, when Chomsky speaks of linguistic representations, he does not intend this to mean representations of the linguistic: representation is not to be interpreted as a two-place predicate. Rather, linguistic representations, as conceived of by Chomsky, are part of what constitutes linguistic knowledge: a linguistic representation is something which is mentally constituted.
5
I lack the space to pursue the interesting idea that Universal Grammar allows, via "parameter setting" for very limited variation involving "PF options and lexical arbitrariness" (Chomsky 1992: 5). Note that if neither PF nor the lexicon are taken to fall within I-language, then the way is open to conceiving of UG as entirely invariant, rather than "mostly" invariant, in the way that Chomsky (1991a: 23) suggests. 6 It is thus unfortunate that the phrase constitutes the very title of a book on the subject of the linguistic conceived of as a cognitive state: Chomsky (1986). However, it is clear from the claims made in that work, and indeed all of Chomsky's work prior to that, that no "knowledge o f relation is postulated with respect to the linguistic.
70
PHILIP CARR
Under this view, the phonological {qua object of inquiry), since it is linguistic, is constituted as a cognitive state, and is distinct from behaviour. It must be stressed that Chomsky has often been at pains to point out that the linguistic is not orientated towards behaviour. That is, the linguistic is not constituted as either ability or intention to behave. The term "competence" was probably abandoned by Chomsky (in favour of "I-language") precisely because it lent itself to an interpretation of the linguistic in such a behaviourally-orientated way (i.e. as ability or capacity to behave). If we insist upon a conception of the linguistic which is not behaviourally orientated, we are therefore following the spirit of Chomsky's work. But the problem with the current conception of PF, as adopted by Bromberger & Halle (henceforth B&H: 1989, 1992), and subscribed to by Chomsky (1992) himself, is that it conceives of phonological expressions as mental states orientated towards behaviour and is deeply problematic. Let us see how. B&H (1989: 53) adopt the view that phonology, unlike syntax, "is primarily concerned with the connections between forms that can serve as input to our articulatory machinery". Phonology is, for them, "concerned with the relationship between representations that encode the same type of information phonetic information - but do so in ways that serve distinct functions: articulation and audition on the one hand and memory on the other" (1989: 53). Phonological features are defined by them (p. 54) as "linguistic-phonetic behaviours"(emphasis mine: PC). Speakers, they claim (p. 56), "represent words in their memory by means of a code that is directly related to ways of producing linguistic sounds". That is, phonological expressions, constituted as mental representations, are composed, not of purely abstract, purely cognitive, objects, but of objects that encode details of physical articulation: they are geared towards phonetic events. These statements clearly express the widely held view that it is in the very nature of the phonological to be dedicated to the connection between linguistic expressions and articulatory/acoustic events, that the phonetic and the phonological interpenetrate. This view of the phonological is still adopted by Chomsky himself, thus: "As for the latter [the rules of the phonological component: PC], it still seems to me that they have something like the form originally assumed in early generative phonology, now understood in the familiar broad sense" (Chomsky 1991a: 23). It is in terms of such an interpenetration that the notion "linguistic occurrence" is connected with a more general notion of physical occurrence, such that tokens of linguistic expressions are said to be capable of being articulated and heard. B&H clearly subscribe to this view: "utterances are sequences of word tokens produced one after another" (B&H 1989: 55). Thus, "tokens" of linguistic expressions ("tokens of words") may "figure in utterances" (1989: 53), where "figure" clearly means "occur".
TELEMENTATION AND GENERATIVE LINGUISTICS
71
Similarly, surface phonetic representations are "generated (sic) only when a word figures in an actual utterance" (1989: 53). This is the "manifestation" conception (by now traditional in generative linguistics and certainly subscribed to by Chomsky himself) of the relation between, on the one hand, phonological objects qua linguistic expressions and, on the other, observable phonetic events. As Chomsky (1993b: 184) puts it, "particular signals are manifestations of linguistic expressions". That "manifestation" (or "realisation") conception is entirely telementational: it amounts to the claim that linguistic expressions are encoded in the speech signal, which is then decoded by the hearer. The hearer may thus be said to perceive, within the acoustic signal (via performance systems) its linguistic properties. Additionally, Bromberger (1992: 14) conceives of orthographic events such that they "encode" sentences: a clearly telementational conception of those events. For him, different placement of utterance accent "creates different grammatical objects". But this is to imply that utterances may be grammatical, i.e. linguistic, objects. He asks how it can be that non-spatiotemporal types may be instantiated as tokens; his view of "linguistic types" is that they are merely theoretical constructs, and he commits himself to the claim that the evidential basis in linguistics "consists mainly of tokens" (p. 15) such that tokens, but not their types, are objectively real. This conception of "linguistic tokens" is incoherent. Bromberger seeks to allow for the existence of linguistic tokens while denying that there are linguistic types. But the notion "token" has no meaning independently of the notion "type"; the two are defined in relation to one another. One cannot speak of tokens other than as tokens of types. That is what "token" means. Consequently, one must either abandon the type/token relationship in discussing linguistic expressions, or spell out a coherent conception of it as applied to such expressions. One cannot coherently abandon talk of types while retaining talk of tokens. Equally importantly, talk of manifestation, realisation or externalisation of linguistic expressions is telementational talk: if these expressions are manifested, they "become" observable, are transmogrified into, or encoded in, observable physical events. If this were not being claimed by generativists, there would be no sense in the notion, much appealed to by generativists, of "Primary Linguistic Data" (PLD): this data, crucial to the current generative conception of language "acquisition", is claimed to be simultaneously sensory and linguistic. In short, generativists currently claim that acoustic events may have linguistic properties, an idea which is positively medieval in its occult nature: in the late twentieth century, we ought to be asserting that acoustic events have only acoustic properties. Alas, the myth of telementation is alive and well and embedded in the PLD idea.
72
PHILIP CARR
3.3 Articulatory intentions,phonological "events"and
PFasinstructions
B&H (1992) claim that the objects of phonological theory are "neither noises, nor gymnastics, nor Platonic ideas" (the latter expression reflects Bromberger's interpretation of constructs as types), but "a class of occurrent intentions and occurrent mental states" (Bromberger 1992: 16). B&H therefore claim that the object of phonological enquiry is to be characterised in terms of intentions orientated towards behaviour. Let us consider an example which will make this clear, and then consider the problems arising from that conception of the phonological. The example concerns a particular utterance, uttered by Bromberger during a talk at Uppsala, and labelled by B&H as "event (1)". I will represent that event as in(l): (1) The merchant sold shelves.
B&H say that many things happened when the event represented by (1) occurred. Among these are noises, bodily movements and brain and neurological events. Additionally, they claim, (1) illustrates "a phonological event' (emphasis mine: PC), which can be examined in the light of phonological theory and given a phonological analysis. They then show how a phonologist would "represent" that event, and present a representation for the initial stage in a phonological derivation, another for an intermediate stage in that derivation, and, finally, a "systematic phonetic" representation corresponding to the final stage of the derivation. Each of the three lines, they claim, stands for some fact about the "phonological event" in question. The production of event (1) was, B&H say, an action, and like other actions, it was brought about by a distinctive kind of mental state, namely an "intention". By "intention", they mean "a familiar kind of purposive mental stance" (B&H 1992: 213). In order to explicate this "purposive mental stance", they offer an analogy: Think of someone aiming a rifle at a target. That person moved and positioned limbs, head, eyes, etc. in certain ways. But more went on. After all, the movements were not made accidentally, or by way of checking whether the barrel is in line with the butt. The person was set psychologically in a distinct way, i.e. had distinct intentions. More specifically, a person who aims a rifle has certain effects in mind, plans moves in ways calculated to achieve those effects,
TELEMENTATION AND GENERATIVE LINGUISTICS
73
and, crucially from our point of view, has the intellectual capacity to select those effects and to devise the gestures that achieve them. (B&H 1992: 213) For B&H, event (1) was like the aiming of a rifle, it required a distinctive mind set, distinctive intentions on my part, intentions that I could not have formed without certain pre-existing intellectual capacities. Of course, I had many intentions when I produced it: I intended to give you an example, I intended to be understood, I intended to produce a sentence that you have probably never heard before [emphasis mine: PC]. But only some of my intentions account for the fact that I pronounced (1) [emphasis in original: PC], that (1) was an action of pronouncing something in a language I know, in my idiolect of English. Those intentions are the kinds of facts about (1) that we take (2) [the phonological derivation offered: PC] to represent. (B&H 1989: 213) They characterise the initial level of representation in a phonological derivation, which, they claim, eventuates in that utterance, as one of a series of intentions responsible for event (1), namely, the intentions to use certain words, e.g. the noun "merchant", the verb "sell" marked for past tense, etc. in a certain order. (B&H 1992: 216) The mental state which accompanies the firing of a rifle is a behaviourally orientated state: it is constituted in terms of intentions to behave, to perform certain acts, to bring about certain spatiotemporal events. And it is in precisely these terms that B&H conceive of the phonological expressions which figure in the first stage of their derivation: as stored memories which map directly on to intentions to bring about certain spatiotemporal (for them, articulatory) events. But they also conceive of their "systematic phonetic" representation (the final stage in the derivation) as also "standing for a series of intentions that generated [emphasis mine: PC; see § 3.4] those movements" (219), i.e. the articulatory movements in question. 7 Thus, the "[m]" in their systematic phonetic 7
It is not clear that it is quiterightto speak of intentions of this sort. In the case of therifle,it seems that the person has the intention to hit the target; whether the physical movements carried out in aid of this intention are themselves intentional is a moot point, since many, perhaps all, of them may occur at the level of motor controls, and such movements are not very clearly describable in terms of intentions. Similarly, when one walks towards a given point, it is reasonable to say that one has the intention to arrive at that point, but it is not at all clear that one has the intention, in arriving at that point, to put one leg in front of the other, bring
74
PHILIP CARR
transcription "represents an intention". But it does not just represent an intention to lower the velum, adjust the stiffness of the vocal cords and close the lips (since one can have that intention while intending to hum a tune); rather, "the '[m]' in (2c) represents an intention to act so as to produce a specific English speech sound, a token of the phoneme /m/." (215) It is clear from these comments that the initial stage in a phonological derivation is said to represent a set of intentions, and so too does the final stage. Interestingly, they take the final stage in the derivation, their (2c), to be entirely linguistic: "It only contains linguistic information and takes into account only linguistic knowledge." (215). It is clear from this that, for B&H, the phonological is conceived of as linguistic and is simultaneously conceived of in terms of intentions to perform articulatory events. Thus, for B&H, there are such things as "phonological events", a bizarre idea if, as is claimed, an I-language is a cognitive state, if linguistic expressions are abstract objects (rather than events) generated by the I-language, if those expressions are in part phonological, and if phonetic events are distinct from I-language. This conception of the phonological is deeply problematic. Consider the conception of phonological derivation which it provides. The first stage in the derivation of the expression "sold", in their example, is characterised as { [ s " | ] , Vb...} + { Q, Past...}. B&H are committed to the claim that, at that stage in the derivation, the speaker is possessed of an intention to utter [s"l], and perhaps to "utter it in the past tense" (whatever that may mean: it is not clear exactly what precise intention B&H propose to be represented by "{Q, Past..}"). This then converts, at the intermediate stage of the derivation, into an intention to utter [sol], and perhaps also "to utter the past tense" (again, whatever that may mean). Finally, the speaker is possessed of an intention to utter [sold]. B&H conceive of these fluctuating articulatory intentions in terms of "a mental computation" over intentions (215): the final stage in the derivation is "the result of a mental computation" (215). Thus, in uttering "[sold]", the speaker must formulate a specifically articulatory intention to utter the vowel ["]; that intention must somehow transform itself into a quite distinct articulatory intention: the intention
forward the other leg, and so on. In a similar vein, I would wish to claim that the speaker intends to utter and intends his interlocutor to know what propositional content he is entertaining. It is not at all clear that there are, additionally, intentions to close the lips, slacken the vocal cords, etc, in the manner that B&H envisage. Many, perhaps all, of those articulatory movements must surely take place at the motor-control level, and are thus not very clearly accompanied by intentional mental states of the sort B&H postulate. If we conceive of acts of speech without appeal to the idea of such states, no telementational aspect enters into the picture.
TELEMENTATION AND GENERATIVE LINGUISTICS
75
to utter the vowel [o]. But this is surely an entirely implausible account of a speaker's articulatory intentions in uttering the noises "[sold]". 8 While Chomsky (1992) says that he subscribes to B&H's account of PF, he also adopts the view (1995: 19) that PF constitutes instructions which are common to both articulation (as in B&H) and perception, a view which, as he puts it, is "not at all obvious, hence interesting if true". He points out that the elements of these symbolic objects [PF and LF: PC] can be called "phonetic" and "semantic" features, respectively, but we should bear in mind that all of this is pure syntax, completely internalist, the study of mental representations and computations, much like the inquiry into how the image of a cube rotating in space is determined from retinal stimulations, or imagined. (1995: 19) And yet, We may take the semantic features S of an expression Ε to be its meaning and the phonetic features to be its sound; Ε means S in something like the sense of the corresponding English word, and Ε sounds Ρ in a similar sense, S and Ρ providing the relevant information for the performance systems. (1995: 20; emphasis in original: PC) It is not at all clear how a PF can simultaneously be the "sound" of a linguistic expression in any coherent sense while remaining "pure syntax", or, put another way, that linguistic expressions, conceived of in a purely internalist manner, could in any sense have a sound. Nor is it clear in what coherent sense PF can be phonetic if, as Chomsky claims, the language faculty is not tied to specific sensory modalities, contrary to what was assumed not long ago. Thus, the sign language of the deaf is structurally much like spoken language... the analytic mechanisms of the language faculty seem to be triggered in much the same ways whether the input is auditory, 8
If it is reasonable to speak of "articulatory intentions" at all when describing ordinary speech (i.e. not when describing many of the actions undertaken in a phonetics class), which is doubtful. Note that conscious intentions to lower the velum etc. actually inhibit fluent articulation: in order to become fluent in a foreign language, it is crucial that one gets to the point of not formulating conscious articulatory intentions. The English-speaker who is consciously intending to utter, say a nasalised vowel of a certain sort when speaking French is, almost by definition, not fluent in French. Formulating articulatory intentions and speaking fluently appear to be incompatible activities.
76
PHILIP CARR visual, even tactual, and, seem to be localized in the same brain areas. (1995: 16)
There is no indication at all in B&H (1989) that their account of phonetic features is capable of acting as instructional input to perceptual, rather than articulatory, performance systems, and even then their account of PF as input to articulation is profoundly problematic, as we have seen. It is even less clear in what sense B&H's account of PF could possibly act as input to the performance systems of the deaf, or what the "P(honetic)" in PF could possibly amount to in the I-language of a deaf person. Chomsky (1995: 20) allows that a linguistic expression is "accessed by performance systems that interpret it, on the receptive side, and articulate it typically using it for one or another speech act, on the productive side. How is that done? The articulatory-perceptual aspects have been intensively studied, but these matters are still poorly understood." There is a fundamental problem here: on the one hand, linguistic expressions, under a strictly internalist conception of language, must be "pure syntax"; on the other hand, those expressions, when construed as input to performance systems, cannot be pure syntax. The root of the problem lies in the idea that linguistic expressions act as input to performance systems, and what is required is a conception of those expressions such that they are not conceived of in this way. The linguistic endowment shared by deaf and hearing people could not possibly be modality-specific, as Chomsky points out; what neither he nor B&H point out is that it follows from this that there can be no such thing as PF within those expressions.9 3.4 Production, generation, sentence and utterance It is striking that Harris (1987) does not discuss the distinction often made in the generative literature between "generate" and "produce". Where that distinction has been made, it consists in construing the term "generate" in a special, rather technical, and entirely formal, sense: a generative grammar "generates" a set of sentences in the sense that it formally characterises a set of expressions such that one can tell, by examining the grammar, whether a given expression is or is not generated by a particular grammar. In this sense, generative grammars are said 9
It is as well to note, at this stage, that almost all of the points made here can be dismissed as "pointless terminological controversy" (Chomsky 1993b: 184). But the issues are philosophical, and in philosophy one is obliged to choose one's terms carefully and use them as consistently as possible. Linguists typically get around issues such as these by means of copious use of inverted commas. Thus, PF, while said to be phonetic, is not really phonetic at all; thus it may be referred to as "phonetic".
TELEMENTATION AND GENERATIVE LINGUISTICS
77
not to produce anything, and certainly are not said to produce utterances. With this conception of the notion "generate", we distinguish between grammars generating (infinite sets of) sentences and speakers producing (finite sets of) utterances, where sentences are not observable events, but utterances are. The term "generate" did indeed appear to have rather literal overtones in early generative work, and continued to be used in its literal sense of "produce" in much later work. This literal sense may well have arisen partly out of a confusion between finite-state grammars and finite-state automata, and was, no doubt, not helped by the fact that the series of rewrite rules in finite-state grammars were often referred to as "productions". The term "string" helped muddy the waters further: grammars were said to generate strings, and strings were easily thought of in terms of sequences of physical objects, that is, in terms of observable events. The confusion must also have arisen partly because of the everyday meaning of the word "generate", which is often synonymous with "produce". Harris can therefore be partially forgiven for taking the word at face value (not that he would accept that the word has a face value!); but a closer reading of generative work on his part would have revealed the special, technical, use of the word. It would also have revealed much inconsistency in that literature, with "generate" and "produce", "sentence" and "utterance" frequently used interchangeably; I will give some examples shortly. There is a deep-seated problem for generative theory lurking behind the conflation of "generate" and "produce" and the conflation of "sentence" and "utterance", and that problem resides in the notion "linguistic behaviour", which both Harris and Chomsky appear to subscribe to, and which I return to briefly in section 4. Take Harris's interpretation of the notion "rule" as it has been used in the generative literature. Harris's is a behavioural interpretation: he construes grammars as production devices, by means of which human beings engage in articulatory behaviour which results in the production of acoustic events, events which have linguistic properties. On this view, the generativist's rules are rules which guide linguistic behaviour. This conception of the generative notion of linguistic rule is quite distinct from, and indeed inconsistent with, the non-literal, formal sense of "generate". Under that formal notion, the rules of a generative grammar do not produce behaviour. Borsley and Newmeyer are right to point out that the notion "rule" plays little or no role in current generative work, and that notions such as "principle", "parameter" and "constraint" have superseded rule-based work. It may be thought that this conceptual development rather undermines Harris's criticisms, and perhaps shows them to have been misconceived in the first place. But the problem of linguistic behaviour remains. B&N object that Harris takes it that the goal of the linguist is to directly describe "linguistic behaviour", but that is to
78
PHILIP CARR
concede that behaviour may have linguistic properties, and generative theory cannot make that concession without compromising its internalist basis. Both Harris and the generativists subscribe to the idea that observable behaviour may be linguistic. For Harris, verbal behaviour and linguistic behaviour appear to be synonymous. For him, there are such things as words, and they are externally observable in nature. He claims that communication is a matter of making sense of such verbal or linguistic behaviour, but not in the sense of decoding a signal which is an encoding of thoughts. Nonetheless, he does perhaps think that words, conceived of this way, have meaning, but only in the sense that they are imbued with meaning by the context of utterance, not in the sense that they have meaning in and of themselves. Let us assume, with Chomsky (1986b, 1991a and b, 1992, 1993b, 1995), and against Harris (1981a, 1987b), a distinction between I-language on the one hand and observable, Ε-physical behaviour on the other, where "E-physical" means physical in the sense of being external to cognitive states.10 If generative linguists wish to uphold a sentence/utterance distinction, and want to deny that speakers may "produce sentences", then they must maintain a consistent distinction between the linguistic and the Ε-physical, such that linguistic expressions (including sentences and words) are quite distinct in their ontology from utterances (which are physical events) and are not brought into being by acts of uttering: they are not "produced". Generativists seek to uphold a strict distinction between the wholly abstract, cognitive expressions generated by the grammar and utterances produced by speakers; that is, "generate" is not synonymous with "produce". The problem with the conception of PF as elaborated by Bromberger & Halle is that it is quite inconsistent with respect to these distinctions, distinctions which play a crucial role in determining one's conception of the phonetics/phonology relation. This becomes clear if we attempt to establish where B&H stand on these matters. Let us now attempt to do so. On the one hand, the generate/produce and sentence/utterance distinctions are apparently nbt upheld by B&H: they speak of "producing sentences" that one has probably never heard before, as we have seen: I intended to produce a sentence that you have probably never heard before. (213) That which is physically heard is acoustic in nature, i.e. is Ε-physical; thus, for B&H, sentences are conceived of as physical events, and can be heard; they are 10 The advantage of the term "Ε-physical" is that, conceived of this way, it allows us to distinguish between cognitive states and physical behaviour without addressing the mind/brain issue and thus the matter of the sense in which cognitive states might be physical.
TELEMENTATION AND GENERATIVE LINGUISTICS
79
thus indistinguishable from utterances. B&H also allow, as we have seen, that articulatory movements (which are undeniably physical events) are said to be generated: they speak of a series of intentions that generated those movements (219) If sentences are said to be produced on particular occasions, they are being equated with utterances. So sentences may be said to be produced and utterances may be said to be generated. On the face of it, then, B&H are not concerned to distinguish generation from production, nor sentences from utterances. On the other hand, the final point in a phonological derivation, which constitutes the output of operation of the phonological component of the grammar, is said by them to be entirely linguistic, and is not, they urge, to be equated with actual articulatory events. That output is, they say, like a musical score, something that can be executed through motions, but need not be executed, and often is not. (B&H 1992: 215) It seems clear that B&H, in this latter statement, wish to deny that the grammar generates actual physical events. And yet that is precisely what they have said the grammar does; clearly, B&H have committed themselves to a contradiction. B&H do not offer a consistent conception of the relation between the phonological and Ε-physical, observable behaviour, or a consistent conception of the phonological. B&H are faced with two options: they could claim that, in the course of a derivation, the phonological, not conceived of in terms of intentions, is mapped on to the phonological. They would then need to map the final stage in such a derivation on to articulatory intentions. Or they can choose to incorporate articulatory intentions into the phonological, which they do, but that option results in the implausible claim that one set of articulatory intentions is transformed into a quite distinct set of articulatory intentions in the course of a derivation. Another problem with B&H's conception is that it is not clear what the status might be of computations performed over intentions; that is, it is not clear that those computations need not themselves be intentional. Finally, it is clear that B&H conceive of a phonological derivation as a series of intentional events in real time, and that too is problematic, since it entails a conception of the phonological as event-like, which is inconsistent with the claim that to "know a language" is to be in a certain cognitive state, rather than to bring about certain events. That is, to impute eventhood to linguistic expressions is to attribute properties of the Ε-physical to the cognitive.
80
PHILIP CARR
It is fundamental to the generative enterprise that process terminology used in discussing linguistic expressions be interpreted merely as a façon de parler". if generative grammars are to be conceived of as generating sets of expressions, then that set does not constitute a set of utterances or a set of behaviours; a generative grammar is not a theory of utterance production; it is not a theory of behaviour. It is true that there are those (e.g. Wilks 1987) who justifiably regard such process terminology as a remarkably unhelpful and misleading façon de parler. But generative grammars, even as conceived of in recent work by Chomsky, are not interpreted as accounts of on-line production, are not accounts of behaviour. The current view, interpreting features as "behaviours" or as intentions to behave (articulatory intentions), conflicts with this conception of grammars. If grammars generate linguistic expressions and those expressions have a phonology, then the phonology of an expression is fully a part of that expression. If the phonology of an expression is constituted as a kind of behaviour, or as a set of intentions to behave, then the grammar generates behaviour or intentions to behave. But this view contradicts the definition of a grammar such that grammars generate expressions but do not produce behaviour. 3.5 Phonological derivations, "externalisation"and "manifestation" It is because B&H conceive of phonological expressions as behaviourally orientated that the notion "derivation" is so central to their position: a phonological derivation represents, for them, the "making external" (the "externalising" or "manifesting") of the linguistic: in "uttering an expression", the speaker is exploiting stored representations in memory which are conceived of in terms of articulatory targets; the speaker then harnesses these and intends to utter in accordance with them. The phonological component of a generative grammar is, on this view, conceived of primarily as a connection between the starting point in a phonological derivation (the abstract underlying forms of words) and the end point of that derivation, namely phonological "surface forms", conceived of as "the input to our articulatory machinery", thus (to repeat): Phonology, on the other hand, is primarily concerned with the connections between surface forms that can serve as input to our articulatory machinery (and to our auditory system) and the abstract underlying forms in which words are stored in memory. (B&H 1989: 53) Since B&H do not offer a consistent position on the distinction and relation between the linguistic and the Ε-physical, it is unfortunate that they conceive of
TELEMENTATION AND GENERATIVE LINGUISTICS
81
these surface forms simultaneously as the input to our articulatory machinery and as the result of the harnessing of that machinery, thus: ...underlying phonological representations of words are stored in speakers' permanent memory, whereas phonetic surface representations are generated only when a word figures in an actual utterance. (B&H 1989: 53) "Surface phonetic representations" (the end point of a phonological derivation) are thus, for B&H, both linguistic and non-linguistic and constitute both the input to our articulatory machinery and the result of the harnessing of that machinery. In addition to this inconsistency, and the conception of words as objects which may "figure in an actual utterance", it is worth emphasising that it is this conception of the phonological component of a grammar as a "connecting" device between the I-linguistic and the Ε-physical which lends the notion "phonological derivation" its significance for B&H: the notion "derivation" counts, for B&H, as a characterisation of the "externalisation" or "manifestation" (Chomsky 1993b: 184) of the I-linguistic. Because of the way in which the phonological component is conceived by B&H, there is a clear and theoretically significant sense in which underlying representations are prior to surface representations, a sense that justifies thinking of the surface form as "derived" from the underlying form. (B&H 1989: 53) On the view adopted by B&H, a phonological derivation is a progression from a cognitive state of affairs to an articulatory/acoustic state of affairs; as Harris & Lindsay (1993) put it, on this view, a phonological derivation moves from the cognitive to the "ever-more physical". But "physical" does not admit of degree. On B&H's view, Ε-physical events are derived, via phonological derivations, from I-linguistic states: the internal is "made external", it is "externalised" or "manifested". However, as we have seen, this claim undermines the very distinction being made between the I-linguistic and the Ε-physical: it entails sortal incorrectness, i.e. the mistaken attributing, to the Ε-physical, of linguistic properties, and the attribution, to the I-linguistic, of properties orientated towards the Ε-physical. That attribution is pervasive throughout the generative literature and entails an appeal to the idea of a continuum from the cognitive to the Ε-physical, which is incoherent.
82
PHILIP CARR
4. Realism and "linguistic" behaviour in generative
linguistics
The version of realism adopted by many generativists incorporates a notion of "linguistic" behaviour. Generative linguists often complain that the object of linguistic inquiry is not observable behaviour per se, but that which "underlies" behaviour (see, for example, Borsley & Newmeyer, this volume). A common analogy (owed to Chomsky himself) is often cited between physical inquiry and generative linguistics, whereby taking observable behaviour to constitute the object of inquiry is like taking meter readings to be the object of physical (specifically, say, electro-magnetic) inquiry. A parallel analogy would be to point out that patterns of iron filings on a sheet of paper covering a magnet do not constitute the object of electro-magnetic inquiry; rather, it is the "underlying", non-observable, electro-magnetic forces which are the object of inquiry, at least under a realist interpretation of physical inquiry. These analogies lead us to believe that the generativist's postulated I-language stands in a causal relation to observable behaviour: just as the electro-magnetic forces cause or induce the iron filing patterns or the meter readings, so does the I-language cause or induce observable ("linguistic") behaviour. Under a rule-based conception of generative grammar, the rules of the grammar are conceived of as underlying observable behaviour in the sense of inducing that behaviour. But that, as we have seen, is precisely the interpretation of rule (and, for that matter, "principle", "parameter" or "constraint") that the generativist must seek to avoid if a consistently internalist conception of I-language is to be sustained; it is precisely the interpretation that Harris assumes, and can be forgiven for assuming. Borsley and Newmeyer, in their contribution to this volume (Ch. 2), claim that, in attacking the telementation idea, Harris is attacking a position not adopted by generative grammarians. Their point is that Chomsky has consistently denied that language serves communicative ends. In this they are undoubtedly correct; they are also correct in observing that this denial is fundamental to the generative enterprise. What they fail to point out, and what I have sought to show, is that the current generative conception of linguistic expressions allows that they are indeed orientated towards behaviour, in the sense of being instructions to behave or intentions to behave. In particular, the postulated PF level of representation is both geared towards behaviour and conceived of in telementational terms, as an encoding of linguistic expressions, such that Ε-physical events are said to possess linguistic properties and thus that there are such things as observable events which possess linguistic properties. The notion "linguistic behaviour" is an empiricist one, and clashes with the non-empiricist basis of generative linguistics. The problems with the current generative conception of linguistic expressions are therefore at least twofold: (i) the telementational conception of PF
TELEMENTATION AND GENERATIVE LINGUISTICS
83
is profoundly occult, and (ii) is anyway inconsistent with the strictly internalist conception of the linguistic which Chomsky has always sought to uphold. Harris is right on two counts: that telementation is a myth, and that generativists have, if inconsistently, subscribed to it. He is wrong, however, to believe that generative linguistics must be founded on this myth; it needn't be.11
11 My aim here has been negative in the sense that I have sought to show how a consistently internalist generative theory cannot proceed while conceiving of the linguistic in telementational terms, i.e. in terms of instructions to performance systems or intentions geared towards physical performance. I have not sought to show how such a consistently internalist approach can be sustained; in particular, I have given no alternative to the "instructional" view of the relation between linguistic expressions and observable behaviour. Such an alternative is given in Burton-Roberts (1994) and is developed, with respect to the phonetics/phonology relation, by Burton-Roberts & Carr (in preparation).
4 Phonography: Setting a Term to the Evolution of Writing John Sören Pettersson
How, when, what and where human beings first wrote are ultimately questions for the archaeologist and the philologist... But the origin of writing is a different question, and it does not fall either to the archaeologist or to the philologist to answer it. (Harris 1986: viii) During the present century, many people have presented the history of writing and, not least, presented general schemes for how writing evolves. But such presentations have met with much criticism, at least during the last few decades. Not that anyone questions that the different scripts of the world have had less advanced forerunners, but the history of writing has often been presented as if it has been aiming all along toward the writing systems of today. The alphabet especially has been presented as the ultimate goal for the history of writing, while Chinese script has been viewed as a form of writing which got stuck on the way. When one is dealing with the history of writing one can hardly avoid mentioning Gelb's A Study of Writing (1963), where a very strict order for the evolution of writing was presented. Moreover, in Gelb's book non-phonographic systems, that is systems of notation not indicating pronunciation, are seen as "feeble attempts" to make real writing and are consequently classified as "forerunners of writing" (ibid.: 194). Gelb's view of the alphabet as the crowning achievement of the evolution of writing has been criticized by many, indeed also by Roy Harris (1986, 1989b). Still, however, it seems as if the majority of works in linguistics reveal the same basic assumptions as Gelb (for an extensive review of the littérature, see Rohr 1994, esp. p. 67, although Röhr himself does not pursue the criticism as far as Harris). In this paper I will comment on how the history of writing has been surveyed, but I will not focus on teleology, at least not its more obvious manifestations. Instead, I will address another issue, namely the terminology used when dealing with the origin and evolution of writing. To be sure, Harris has had a few things
PHONOGRAPHY
85
to say here too, but I will start with Gelb's work to see how the philologist approaches the problem. The aim of Gelb's work "is to lay a foundation for a full science of writing" (1963: 23), and one of its means is a comprehensive "Terminology of Writing" (assembled in Ch. 11). This terminology does not include names of specific scripts like Chinese writing, but does include categorizing terms like ideography, logography, syllabic writing, consonantal writing etc., and Gelb devotes his pages to discussing the merits of each; for instance, the term ideogram has been used by some authors to designate any kind of symbol and by some to designate word-signs of phonographic systems; he therefore urges his readers to avoid that term altogether. Further, he does not believe in "consonantal writing". Scripts that only indicate consonants are in his view syllabic systems: the so-called "consonantal" signs indicate whole syllables even if they leave the vowel unspecified. Now I will make some points on the use of a comprehensive terminology. An elaborate "Terminology of Writing" is not necessarily the right way to found a (the) science of writing, as Gelb tries to do. The abundance of terms gives the student an impression of precision and exactness, but when discussing a particular matter, what counts is of course how that particular discussion is carried out. In view of this, the relevance of a general terminology of writing will be the topic of what follows. Concomitant with any terminology, or any single definition, two kinds of error may occur: the definitions are not sensible and/or the use of the defined terms is not in accordance with the definitions. Interestingly, Gelb fails in both respects even when discussing the forerunners of writing. Not that he does not take precautions, as the following passage clearly reveals when he coins the term "forerunners of writing": "The word which I should have liked to coin for Vorstufe is 'forestage', but unfortunately this term cannot be used for our purpose because it already has the meaning 'forecastle' or 'a ship with a forecastle', even though only in obsolete English" (1963: 24). But when he proceeds to subdivide the forerunners of writing, he makes the two errors mentioned above, as will be explained immediately below. Gelb, who states that only phonography should be regarded as full-grown writing, says that within the forerunners of writing it is possible to discern two kinds of techniques that have been utilized: the descriptive-representational device and the identifying-mnemonic device. When he summarizes his theories in a chapter called "Evolution of Writing" (Ch. 6: 190 ff) he says about the "forerunners of writing": "In this stage visible drawn forms - j u s t like gesture language - can express meaning directly without an intervening linguistic form" (ibid.: 191). "Forerunners of writing" are then distinguished according to the two
86
JOHN SÔREN PETTERSSON
ways in which one communicates: the descriptive-representational device and the identifying-mnemonic device respectively. In the first case, drawings "contain only those elements which are important for the transmission of the communication and lack the aesthetic embellishments which form an important part of an artistic picture" (ibid.: 192). In the latter case, "a symbol is used to help to record or to identify a person or an object" (ibid.). Further, Gelb writes: "The binding traditions of art, established hundreds and thousands of years before man first attempted communication by means of conventional marks, were too strong to allow for the development of the descriptive-representational device in the right direction" (ibid.). He holds that the identifying-mnemonic device, which concentrates on objects or beings instead of events, was the device which gradually succeeded in attaching words of the spoken language to its signs. Now I want to focus on "the development of the descriptive-representational device", as Gelb put it in the last quotation above. If a descriptive-representational writing gradually acquired signs for words of the language, Gelb could say that these signs were used to identify those "things" (perhaps non-concrete) which the words "stood for". That is, use was being made of the identifying-mnemonic device. And to be sure, Gelb says on the preceding page that "[a]s the two devices are frequently interlocking it is difficult to assign rigorously some primitive writings to definite categories." The point to be made here is, then, that it is meaningless to state that the descriptive-representational device could not develop in "the right direction", because according to Gelb's own classification it would not be called descriptiverepresentational but identifying-mnemonic if it started to develop in the right direction. It is actually not the device that develops in any direction - since it is fixed by Gelb's definition - but the writing (the "script"). This should be compared with what was said above about defining a term. The definition should be sensible, and the use of the defined term should be in accordance with the definition. Gelb then does not make proper use of the term "descriptiverepresentational device". Of course, one could furthermore ask whether it is a good thing to discuss forerunners of writing in terms of the descriptive-representational device and the identifying-mnemonic device. Is this dichotomy sound? This is not the same question as the one just treated above, where the aim was not to discuss any specific classifications of writings. This latter question should instead be answered in the way that Harris deals with the matter: "contrasting 'mnemonic-identificational' with 'descriptive-representational' devices seems to set up a quite artificial opposition between criteria relating to purpose and criteria relating to design" (1986: 65). To go back to the question of terminology in general, this example of how Gelb handles his "forerunners of writing" merely shows one specific scholar to
PHONOGRAPHY
87
be not quite aware of the implications of his definitions. To be sure, there may be many more such scholars, and one should thus not be blinded by a laborious terminology used by a whole group of researchers. But there is another reason to be cautious when a terminology takes a prominent place in theoretical discussion on writing: indeed, one can actually say that to pre-engage in using a general terminology is a risky matter. Gelb's terminology can serve as an example. There, "full" writing systems are classified as either word-syllabic, syllabic, or alphabetic. Now, with this classification the function of writing will often be discussed in terms of individual signs. It will then seem self-evident that an alphabetic script functions because of the matching between speech sounds and letters. Facing the fact that people born deaf may learn to read and write, however, one may feel the need to rethink the whole matter. In this light, a phonographic description of an alphabetic writing system seems to be wrong, or at least only partly true: one may ask if fluent readers in actual fact read by recognizing word images, or if an alphabetic script functions differently for the deaf than for hearing people. Before plunging into such queries one should take note of the role of the researcher in any analysis. For there is nothing wrong in doing phonetic analyses, but one should realize that the terms (the concepts) reveal a certain point of view, namely the point of view taken in the analysis, in the description. An analysis may of course provide a true picture of what has been analyzed. But this picture (true or not) "exists", or is meaningful, only from the point of view taken in the analysis. In this way, the designation "alphabetic script" of a working writing system reveals intrinsic properties to be sure, although not necessarily of the actual written communication but surely of the description technique used by the researcher (or, indeed, the teacher). Thus, setting the terminology for "a full science of writing" in advance is to put a straitjacket on all future research. Let me again take an example from Gelb's book. He evoked a debate whether Egyptian hieroglyphic writing and its West Semitic derivatives, the forerunners of the Greek alphabet, are syllabic or not. Earlier, these scripts had been called consonantal. Gelb rejected both these designations on the ground that the general development of real writing (i.e. phonography) must be taken to be from wordsyllabic through syllabic to alphabetic, which indicates that hieroglyphic writing is really word-syllabic and not word-consonantal, and West Semitic scripts are syllabic rather than consonantal. In itself, there is nothing wrong with a syllabic interpretation of these scripts since the so-called consonantal signs, as was noted above, can be seen as indicating whole syllables even if they leave the vowel unspecified. However, Gelb forcefully argues against any other interpretation; yet his rigidness concerning different conceptions of the West Semitic scripts clearly stems from his theory of writing evolution. He holds the theory to be an
88
JOHN SÔREN PETTERSSON
explication of intrinsic structures of reality instead of seeing it as a point of view, i.e. as an interpretation of data. He cannot allow for other interpretations, since these appear to be misconceptions of reality instead of mere interpretations. Again, this first and foremost exemplifies how Gelb mistreats his outlook on writing. But it gives a hint of something which I find characteristic of many general works on writing. Note how important the matching between individual signs and the structure of spoken language is. Gelb has merely reinterpreted consonants as syllables; for the rest he follows the path laid out already by the decipherers of Egyptian writing. Or rather, since the decipherment has long been complete, he reiterates their (as well as, for other scripts, other decipherers') approach(es) and tries to generalize them to "a foundation for a full science of writing". This is nothing specifically Gelbian. In my view, the evolutionary schemes set up by philologists when they, for one reason or another, incline to general theoretical outlooks on the history of writing are quite faithful reproductions of some key features in the methods used to "crack the codes". This is of course to fall into the trap described earlier: the question of how writing works will be discussed only in terms of individual signs. (In general, many approaches to writing whether historical or not fall into this trap; most conspicuous is probably the theorizing on how readers read texts and how children should be taught writing. In these fields it is of course the way of teaching writing which has prompted the research. As for theorizing on the evolution of writing, probably not only appreciation of the decipherments plays a role but also the fact that the theorists once were small children who were brought into literacy with the aid of one or another set of teachers' spelling principles.) To be sure, the "consonantal" insight of ChampoUion led him to the right reading of the hieroglyphic texts. To relate Egyptian texts to Coptic, ChampoUion interpreted the signs as consonants. Gelb's syllabic reanalysis, on the other hand, brings no new understanding of the texts handled in either Egyptology or Semitic philology; in actual fact, it demonstrates only his own evolutionary preferences. So there is no particular use of his analysis in the factual disciplines. Or rather, his analysis shows that a rigid consonantal analysis is also out of place since it is possible to maintain both types of analysis when doing philology, which implies that neither analysis, when presented as excluding the other analysis, is essential for the work in the disciplines. This may surprise readers of books like Coe (1992), where the "phonetic" method as compared to fanciful iconographic readings is hailed as the real breakthrough in the short history of decipherment of ancient scripts. Such readers will undoubtedly wonder how ChampoUion could relate Egyptian signs to phonetic values if the phonetic aspects can be blurred to the extent that the analysis makes no distinction between consonants and syllables. The thing is that ChampoUion related old Egyptian signs to signs of
PHONOGRAPHY
89
other scripts, known to indicate pronunciation, and using his own Latin script as a base. Any such transliteration is called "a phonetic reading". Exactly which phonetic motivations are used for each transliteration is not crucial so long as consistent interpretations of whole words and texts are provided. Thus, my criticism of evolutionary theorizing on writing is that it consists of a simple generalization of principles of decipherment to a standard for the historical development of writing. This criticism holds good not only for the analysis of full-fledged phonographic writing but also for the device called rebus writing, which is what writing theorists call word-signs (notably pictograms) in those instances where these signs are thought to have been employed solely for their pronunciation, that is for the sound value only, in order to make up (the pronunciation of) words or word forms lacking their own signs. The belief that ancient "proto-scribes" had recourse to a rebus principle plays a pivotal role in many theories of writing (see for instance Harris 1986: Ch. 3). In my view (explicated in Pettersson 1991: 26-29; 1992) the rebus principle is a concept of the present-day interpreter: it is a concept he applies when the employment of a sign seems to be interpretable through homonymy or at least some rough homophony. Repeated application of this concept is simply his grouping together of some (possible) extensions of the use of the signs in a writing system. Note that extensions could be made in other ways; for instance, Gelb speaks of extension by association of meaning (1963: 99). It is not clear that the early writers viewed a phonetic motivation for an extension as essentially different from other motivations that might have been employed before, such as etymological, "meaning-associative" or sportive motivations. My view could be compared with Bloomfield's notion of asterisked forms of Primitive Germanic and Primitive Indo-European: a constructed form is not a phonetic record but a formula describing systematic correspondences in a set of related languages (Bloomfield 1933, sec. 18.5). In the same vein, rebus hypotheses should be seen as formulas the value of which is determined by the degree of sensible semantic interpretations they provide in a larger context than the mere individual signs. When the theorist takes the decipherers' rebus reading as indicating ancient rebus writing, no new insight is gained about old scripts. The statement "employment of rebus writing" cannot mean anything but "interpretation by the rebus principle" unless very specific evidence is at hand about the ancient conception and utilization of phonetic similarities. In sum, the simplistic generalization of methods of decipherment becomes awkward when "the general principles governing the use and evolution of writing" are to be inferred (quoted from the first page of the 2nd ed. of Gelb's book). Adherence to concepts that have worked well in a certain descriptive
90
JOHN SÔREN PETTERSSON
context will simply lead theorists to restate old descriptions over and over again.1 This suggests of course the notion that really novel theories simply cannot be based on old concepts to any great extent. Harris, when reviewing earlier accounts of the origin of writing, says about the candidates put forward by Gelb and some others: None of these evolutionary accounts can satisfactorily identify the transition from pre-writing to writing, because they do not squarely face the issue of what basic conceptual advance underlay the first attempts to produce written records. Consequently, they equate writing with logography, or else with phonography, or else with the introduction of phonographic elements into pictography, or else with the change from tokens to marks. But these are all quite arbitrary terminological decisions, and not solutions to the problem. Instead of treating the problem as needing elucidation before any line of retrospective inquiry makes sense, they reverse the rationale of investigation and operate on the tacit assumption that if and when we know more about the facts of prehistory the nature of the problem will become clear, concomitantly with its correct answer. (Harris 1986: 74-75) What then would a theory of writing in general or a theory of the origin and evolution of writing look like if one were not enslaved by old philological concepts? First of all, if one does not adhere simply to the systems of principles used by the decipherers, one will be capable of seeing alleged supplemental and secondary systems, like numerical notations and layout conventions, as integral parts of the writing systems actually employed. Even the utilization of the writing material may be seen as an integral part of the actual systems. Some theorists admit the fact that a certain type of writing system is often linked to a certain type of writing material and to a certain technique of making the signs, but this is often noted only in passing as ifjiothing in principle forbids the use of any type of writing with any type of materialization. This should lead one to the following conclusion: if the principles with which the various writing systems have been analyzed admit that, for example, cuneiform wedges can be painted on papyrus, then they cannot possibly be taken as the principles which have governed the use and evolution of writing. (To Gelb's credit, it should be mentioned that he actually considered some other types of principles, like the Principle of Aesthetic Convention (1963: 251); and he stressed the importance of the Principle of Inner Development for establishing legacies among writing systems, and rightly so, 1 My eyes were originally opened to this fact by Professor Sven Ôhman, Uppsala, whose seminars have pointed out the uninventive conceptualization of language in generative linguistics and other linguistic theories as compared to traditional grammars.
PHONOGRAPHY
91
though this approach, which was based on the linking of each script to the linguistic habits of a people, is far from the exhaustive foundation needed for "a full science of writing".) Secondly, one will be obliged to consider social factors, the history of war, etc. These are of course old concepts but at least they bring in new areas within the domain of a full theory of writing. It is true that Gelb considers the hampering effects of "habits and elite classes" on any writing system (1963: 202), but, besides the questionable validity of his sweeping arguments in this case, which will not be brought up here, the conquest of the world by the alphabet is too narrowly seen (by Gelb and many others) as simply the result of the (supposed supreme) internal structure of alphabetic scripts. It is superfluous to speculate upon the exact position of Harris in these matters; suffice it to note that he has expressed discontent with the representational view of writing, as he sees writing as an integral part of "the complex of human abilities and activities we now call 'language'" (1986: 52). More interesting in a general evaluation of terminology is the last chapter of his The Origin of Writing, called "The Great Invention", where he attempts to explain what the real advance is that brought forth writing. I think it is fair to say that he develops concepts unheard of in ordinary accounts of writing. I attempt to outline his ideas below, but would like to throw in the following note of caution. When describing theories à la Gelb, one can, at least within the alphabetic culture sphere, rely on the firm basis of a common conception of language as something manifested in utterances which are divisible into words, syllables, and "single sounds". Then it is quite an easy matter to sketch the evolution of writing as going through the stages of word-syllabic and syllabic before it finally reached the alphabetic stage. Outlining Harris's theory is not that simple at all. Since he brings in new concepts, a great deal of explanation is needed. Probably, a true rendering of his idea would need as many pages as he himself has spent on it (if not more). Thus, here I will only be giving a simplified version which no doubt will not do justice to all the intricacies of the original presentation. The idea is based on the combinatorial effects made possible when qualifier (like "red") and qualified (like "block") are separated into distinct signs and then juxtaposed to make up the same meaning ("red block") but not to make up a single sign; this is in contrast with direct pictorial representation. One may observe that this technique is also used within musical notation: the length of the note is not represented by the length of its sign but by "arbitrary" variations of its sign. Scriptorial signs thus achieve a standard of their own, a standard which separates them from the visual-imitative strategy of pictures, a strategy actually found also in the simple notches on a tally. Writing in the sense of Harris's work may thus include musical notation and mathematical as well as phonographic scripts.
92
JOHN SÔREN PETTERSSON
The combinatorial effects presuppose some kind of slotting system within which to make the combinations. Almost certainly homo sapiens mastered the use of numbers before mastering the use of letters. What is being suggested here is something else: that the human race had to become numerate in order to become literate. No society which could not count beyond three ever achieved writing; at least, not by its own efforts. (Harris 1986: 133) However, according to Harris, it was not enough just to make sixty notches on a counting-stick to become literate, neither was it enough to make a mark for the commodity counted, followed by sixty strokes, since this is more or less equivalent to sixty appearances of the commodity sign. Instead, he views the slotting system to be "the great invention", where one slot is reserved for indicating the commodity type and another slot for indicating the total. This second slot should be filled with just one number sign in the sense that it is not enough to simply count the number of marks it contains (ibid.: 145). (No graphic slot frame is supposed, as the signs themselves can be seen as slots: the item sign is a slot for the item concept and the number sign for the number concept.) Noteworthily, it is explained that "the postulated principle of scriptorial development itself has nothing to do with counting as such" (ibid.: 146), and there is also a tentative example not based on numbers which exemplifies signalling without and with slotting technique. In short, that example goes as follows. There is a situation where one has to signal the need for various red and blue things (blocks, slabs etc.) with the help of cards with symbols on their faces. One could have one card for red blocks and another card for blue blocks etc., so that if one wants to have a red block, one holds up the red-block card. On this card, a red block could be depicted. But there is another solution where slotting technique is implemented. In this solution, one has one blue and one red card and then another category of cards consisting of only one card per item irrespective of colour. Interestingly, even if we allow for drawings of the items on this second category of cards, one can no longer show a picture of a red block, because the red card and the card depicting a block will not together make up a "picture of a red block" (ibid.: 143). One has lost one feature of pictures but gained a smaller set of cards. "Slotting imposes an analytic structure of its own on the way signs are used for purposes of communication, irrespective of the specific message involved" (ibid.: 146). On the whole, this seems to be what Harris's work is aiming at. It is further remarked that "[i]t is typical of linguistic structure, as opposed to pictorial representation, to 'separate' properties and quantities from objects, and express
PHONOGRAPHY
93
them by means of independent signs" (ibid.). In general, the "adaptation to more complex communicational requirements will lead to the intrusion of structurally superimposed features which inevitably open up a gap of just the kind which characteristically separates pictures from writing" (ibid.: 144; cf. also 149). And finally: "The idea of genius behind the great invention of writing was an intuitive grasp of the principle that graphic signs have no limitations for purposes of human communication other than those which derive causally from their primary parameters as visible marks. They have no other semiological constraints." (ibid. : 155) (Note that the "grasp" in the last quotation corresponds to the "gap" mentioned in the preceding quote.) There is more to the "gap" which Harris speaks of. For even if it is allowed for that the structure red + block is criterial for writing, it is not clear how any primitive man may hit upon the idea of seeing red + block as "red block" which, after all, is necessary to make the whole thing function. Thus, to a certain extent he must equate the two signs with the single picture of a red block. However, it does not seem improbable that superimposing structures could arise due to man's other practices; for instance, when painting one may use one pot for blue paint and one for red irrespective of where the painter is to put the paints. Conversely, when it comes to counting one would like to see some explanation of how units bigger than ones arise; how was it possible in proto-literate Mesopotamia to utilize one sign to stand for sixty cattle, for instance. Again, it is not hard to sketch a likely scenario: if larger numbers of livestock are noted by paddocks instead of individual animals, it could be realized that it would be convenient to fix the number of animals in each paddock so that a sign for paddock always stands for a certain number of one-animal tokens. Such a technique could be used when the number of paddocks became larger: a certain sign could be agreed upon to stand for, say, three paddocks even if the paddocks themselves were not kept in some kind of superpaddock containing three paddocks each. Gradually, then, the system of numerals breaks away from imitative one-to-one structures. Cases like these could perhaps provide a cognitive starting point for a more general gapwidening, but, admittedly, the particularity of the solutions in the two sketches makes one feel uncomfortable. As already cited, Harris's 1986 book offers "more complex communicational requirements" as the driving force for the restructuring of graphic communication. Such requirements seem likely to secure slotting efforts that have already been made: the set of cards in Harris's tentative example will be smaller and the system in some sense more efficient. But they will not explain how people came to think of two signs as standing for one sign or one sign as standing for two others. To come to grips with this intriguing cognitive question, Harris, in a later work (1992; repeated as Ch. 21 in Harris 1994a), launches the
94
JOHN SÔREN PETTERSSON
concept of"synoptic equivalences between visual signs. Such equivalences allow a single sign to function as a metasign for a series of signs, thus systematically enhancing the utility of signs as an aid to memory" (1992: 181). According to Harris, such metasigns not only appear in acknowledged writing systems like cuneiform writing or hieroglyphic, but could already have shown up in the early stone age: some researchers (not read by the present author) presenting Paleolithic materials have pointed out that numerous "slate plaquettes" have been marked with scratches, often repeatedly so; it seems as if surfaces had been cleared to make possible repeated use, although earlier traces are discernible. Could such signs be considered proto-writing? To begin with, one can note that Harris writes that these signs were obviously treated as having an existence independent of the object which was repeatedly reused (1992: 190 ff.). "Autrement dit, il s'agit d'une conception très proche de la distinction moderne entre un texte et son support - distinction tout à fait essentielle dans le développement de l'écriture" (1994a: 353). This interesting fact would be easily overlooked by many authors who would instead "immediately" rush to the scratches and try to evaluate these according to this or that standard for writing (cf. the extensive account of Upper Paleolithic in Part Π, "Stages in the evolution of writing", of Röhr 1994: 95-157). Moreover, since the plaquettes were re-used they will bear witness to earlier use. So quite by happenstance the superposition of scratches will represent "a synoptic reduction of a whole sequence of 'messages' which had been previously inscribed and which, with the aid of memory, could be reconstructed." (Harris 1992: 191) "The metasemiotic equation 'X = A + B', where X incorporates A and B, bypasses the need for any prior graphic convention relating the forms A and Β to the form X, but at the same time provides the paradigm for such a convention" (ibid.). He views this equation as a starting point for an analysis of spoken language which makes no reference to the phonetic form of speech but only to semantic relations, as in the Sumerian case where e.g. the sign for "eat" was composed of the signs for "mouth" and "food". One may further note that although a single compound sign is not built up by a linear structure in the sense employed in linguistics where linear means "sequential" (the term linear otherwise means "non-pictographic" when dealing with archaic scripts, as in Linear B), it is nevertheless true that such a graphic constellation "becomes directly iconic of a temporal sequence of signs: exactly the pre-condition required for the development of writing" (ibid.). After this exposé of the ideas in two of Harris's works on the origin of writing we may ask what theorists of the types criticized earlier can learn from these ideas. This, I think, we can ask regardless of how well-founded my earlier suspicions are concerning the impossibility of making a quick exposé of novel
PHONOGRAPHY
95
and unorthodox theories (i.e. regardless of how poorly Harris's ideas were represented above). Exactly how we ask is no trivial matter, though. Let us try this first: how is anyone who is writing the history of writing to apply Harris's idea about the conceptual advance lying behind writing? For instance, where did fortuitous synoptic equivalences lead to the separation of various sorts of qualifying signs from the signs for the qualified objects? In Upper Paleolithic graffiti, in Neolithic token systems, in Sumerian numerical tablets, or in some other graphic device? Obviously, it will be very hard to judge when the "intuitive grasp" governs the use and development of any particular writing system. Similarily, we cannot rely on a premiss that the ancient users of a system had a conscious notion of the principle of slotting (recall the rebus principle dealt with above). In spite of this fact, however, it cannot be denied that insights of precisely this sort have eventually enabled man to reduce writing as well as sound records, pictures, and whole industrial processes to computer "bits", to zeroes and ones, where imitations of visual, acoustic or whatever structures have vanished altogether, leaving only the slot to be filled or not filled. We have accepted the slotting possibility "once and for all". Surely, in specific cases it may be undecidable what is "slotted" and what is not. Harris's theory will not help the evolutionary-minded philologist to decide whether this or that notational system marks the birth of true writing. His theory will not help except as a reminder that the philologist's approach to the origin of writing is not the only conceivable one. (And I might confess, in the shadow of a pair of parentheses, that much of my treatment of Harris 1986 in Pettersson 1991 and 1994, especially the former, is precisely a philological misreading of him, even though I did at the same time stress his novelty.) As to the reception of Harris 1986, one probably dares say that hardly any writer has picked up Harris's account of the origin of writing. When The Origin of Writing is referred to, the references are to its first four chapters reviewing and criticizing earlier attempts to deal with the question. The reason is simple: many people feel ill at ease with the phonocentrism of the linguistics which has dominated the present century. However, this does not mean that people who refer to Harris (no one named, no one blamed) are not on the threshold of remaking the same mistake less some of its more conspicuous errors, such as crowning the evolution of writing with the alphabet or rendering Chinese writing as primitive. Thus, while Harris's book (together with Ch. 1 of Harris 1980) offers a well-wrought criticism which quite a few scholars nowadays incline to, little attention has been paid to Harris's own contribution. Naturally, philologists have their aim of interpreting old texts, but it is a bit surprising that more "general" and theorizing authors on writing have not struggled more with Harris's contribution.
96
JOHN SÔREN PETTERSSON
As already mentioned, Harris's ideas are not applicable to the form philological disputes ordinarily take. To this I want to add that the kind of theorizing criticized here, namely the uncritical extension of methods of interpretation to theories of the origin and evolution of writing, is not saved by turning to philology for an application of it. It is true that Harris's ideas, relying on non-philological concepts as they do, do not fit into the ordinary working of philology or archaeology, and neither do they fit into disputes raised by extrapolating these disciplines into general theories. However, while this is true, it does not permit the extrapolated theories, like Gelb's, to claim any power of judgement in philological questions. This is so because this theorizing has been a disguising of primarily philological methods and conceptions, as pointed out above (for an elaboration, see also Pettersson 1991: 66-75 & passim), and will consequently not yield any other answers than those which are built into it by means of the adopted concepts. After this discussion of the utilization of Harris's and Gelb's theories in the "factual" disciplines, it is fitting to take a look at how the evolution of writing is treated by Harris. Harris does not deal very much with how writing developed after the first intuitive grasp. Perhaps this is just as well, since many people even today, not the least many writing theorists, still have to revive this grasp (I am thinking especially of the common view in which writing appears as first and foremost a matching of graphemes to segments of speech; this view surely does not embrace the nearly unlimited possibilities of writing). At any rate, at the end of his book he calls attention to "the inevitability of the fact that every change in perspective from which the independence of writing is viewed brings with it an automatic re-evaluation of the boundary between the pictorial and the nonpictorial, together with a re-evaluation of the relationship between speech and language" (1986: 157). But most of the time he emphasizes the fact that the mechanism behind the evolution of scriptorial signs is independent of spoken language: "the mechanism itself does not lead automatically or ineluctably in the direction of the kind of writing system on which Western education came to be based. It is just as capable of leading in the opposite direction to the development of graphic systems which are independent of oral communication altogether" (ibid.: 150). Here we should of course note that Harris does not say "in the direction of the kind of writing system which represents speech". After all, writing as representation is one of the main targets in this and several other works by Harris. Nevertheless it is true that Western education has been based on the alphabet and certain principles for its teaching, which doubtless has led our culture to regard graphic communication from a very special point of view. To continue, he points out that "graphic communication, as a mode of communication sui generis, is free to draw upon other modes of communication
PHONOGRAPHY
97
as structural models. In particular, it could draw upon oral communication as an incomparably rich source of analogies for the graphic expression of all kinds of information that human beings might need for purposes of civilisation" (1986: 156). Still, he assures his readers in the Epilogue that when "in the fulness of time a history of writing as writing... comes to be written... speech will be seen as the historical crutch on which writing was obliged to lean in the earliest phases, a prop to be thrown aside when no longer needed" (ibid.: 158). One of the most striking implications of this assertion, at least for anyone who has read the ordinary histories of writing, is that writing has a future. For Gelb, the evolution of writing stopped in the beginning of the first millennium B.C., because then it was completed: "The development of a full Greek alphabet, expressing single sounds of language by means of consonant and vowel signs, is the last important step in the history of writing. From the Greek period up to the present, nothing new has happened in the inner structural development of writing" (1963: 184). Many others have expressed similar views. In conclusion, without the conceptualization of writing as "phonography" the history of writing will appear in quite a different light than it would if we allowed this term to set a term to the evolution of writing. It is not strange at all that the standard method of writing the evolution of writing does not really reckon with a future of writing. One only has to remember that the method is an (over-) generalization of methods of decipherment, that is, methods for the decipherment of long-dead scripts. Such methods cannot conceivably be claimed to lay heavy stress on types of writing systems not yet invented. The delimiting force of phonography is twofold: it does not allow for any writing "proper" until about 5000 years ago, and it ends the evolution of writing about 3000 years ago; at least if the traditional classification in decipherment is taken as a foundation for the concept of phonography. If we include the development of phonetic notations, the evolution would still be on-going. However, the evolution would be circumscribed to include only representational advances, or, more precisely, only advances concerning phonetic representation (this is well illustrated in Ch. 10, "Future of writing", in Gelb's book; the only development he can see concerns refinements, really reinventions, of the alphabet). This is about as bad as the usual termination of the evolution of writing because the concept of phonography still sets a term to the discussions. To this one can add the note that the equation "phonography = real writing" not only affects discussions of the evolution of writing, but naturally also delimits any other kind of discussion of writing. The topic of the present paper is perspectives on the development of writing, but let me elaborate a little on this last point by noting a book which I have had time to only superficially scan but which should be at least mentioned in this paper, La sémiologie de l'écriture
98
JOHN SÔREN PETTERSSON
(1994a). In this book Harris develops the idea of an "integrationalistic" approach to the study of writing. "Le concept d'«intégration» dont il est question dans le domaine de l'écriture est plus proche de celui des biologistes et des neurologistes, disciplines où il est toujours question de rendre compte de l'action réciproque de systèmes dynamiques, et où la dsintegration n'est rien moins qu'un état pathologique" (1994a: 10). The simple correspondences between single signs and alleged independent slices of speech will be left outside this integrationalism. No doubt, the study of graphic signs will be very different if they are studied as dynamic systems, and furthermore as dynamic systems integrated in and integrating other dynamic systems. And Harris is probably right when in an article in HSK 10 he refers to his own La Sémiologie de l'écriture in the following way: "Although it is commonly taken for granted that writing systems are systems of signs, surprisingly little has been published which attempts to apply sign theory in a principled way to the analysis of written communication. Apart from Harris 1994[a], no comprehensive study of this type has so far appeared" (1994b: 41).
5 A New Mentality David R. Olson
The essential innovation which writing brings is not a new mode of exchanging and storing information but a new mentality. (Harris 1989b: 99) Mentalities are mysterious things. They purport to explain by appeal to psychological properties of knowers rather than by appeal to social, political or material conditions. They postulate distinctive ways of seeing, interpreting and talking about what, these days, we can only refer to in quotation marks as "the world". Furthermore, the appeal to "mentalities" goes beyond other more local psychological explanations such as those addressed in terms of theory change, paradigms, conceptual frameworks in that these latter allow alternative modes of representation but modes restricted to particular topics or particular domains. Thus the adoption of a Copernican view of planetary motion involves a mental, conceptual change without bringing with it, necessarily, a new mentality. Even more worrisome about the appeal to the notion of "mentality" is the long tradition going back at least to Wundt's (1916) Elements of Folk Psychology in which he traced the idea to Kant, Hegel and Herder, namely, that mankind has evolved through a series of stages of mentality. History was a matter of the "education" of the human species, beginning with a primitive mentality, through the totemic, the heroic, to the development of an inclusive humanity. Lévy-Bruhl (1926), too, offered an evolutionary account of thinking, focussing on "primitive" mentality, which he characterized as "participatory" rather than "rational". The claim of the Huichol Indians of Mexico that "Corn is deer" was taken as paradigmatic of such participatory thinking. Ordinary rules of logic appear to fail in the analysis of such utterances. Advances in mentality occurred, Levy-Bruhl argued, with advanced social organization and advanced technologies, writing among them. It goes without saying that traditional concepts of "primitive" have been abandoned. "There are no primitive languages" is a basic claim of anthropologists, and advanced modes of thought have been detailed in many anthropological accounts e.g. of cognition in traditional societies, including
100
DAVID R. OLSON
navigation in Micronesia (Gladwin 1970) and biological knowledge of the Maya (Atran 1990). It now seems relatively clear that the human mind, human mentality, has not progressed through a series of stages up to our present one. The way out of our dilemma is to acknowledge the possibility of distinctive mentalities without harnessing them to the notion of progress. Mentalities may be specialized forms of cognition developed for particular purposes, embodied in particular institutions, and passed on to the young through deliberate educational programs. Their claims to superiority rise, not from making persons better people or even better thinkers, but from making them better thinkers in a certain domain. Again, this is not to identify a mentality with the adoption of a contentful paradigm such as atomic theory or planetary theory but with thinking about many things in a new way. So what is this new way? One new way is associated with writing and literacy. Harris's conjecture is the following: "The restructuring of thought which writing introduces depends on prising open a conceptual gap between sentence and utterance" (1989b: 104). He continues: "Writing is crucial here because [this kind of] inquiry presupposes the validity of unsponsored language. Utterances are automatically sponsored by those who utter them, even if they merely repeat what has been said before. Sentences, by contrast, have no sponsors: they are autoglottic abstractions" (ibid.). Utterances are, of course, human actions. We do things with words as Austin (1962) put it. In saying something we are giving our warrant to the truth of what we express. Sentences are linguistic abstractions in two ways. First, it is not clear that speakers even utter proper sentences; utterances are normally composed of "idea units" expressed by a burst of six or seven words with one intonational contour (Chafe 1985). The idea that sentences express a complete thought is the concern of the grammar teacher not the speaker. But more importantly, sentences, as Harris puts it, are "authorless". Two young boys both insisting that "My dad is bigger than your dad" mean or intend different things - that is why they can argue - but their sentences have the same meaning. This is the insight that writing exploits, in Harris's view. It makes possible dictionaries, grammars and works of logic. With this view I am fully in accord (Olson 1977, 1994). But how does the cultural possession of dictionaries, grammars and logics actually alter the conceptual repertory of thinkers? On this point Harris has had much less to say. It is not entirely clear how writing "prises" a conceptual gap between utterance and sentence nor how such prising actually alters one's mentality. It is on those issues that the remainder of the paper is focused. First, it is important to point out that neither Harris nor I assume that just because writing is available, the harvest of treasures opened up by literacy is immediately available. As Harris notes, the uses made of any technology depend
A NEW MENTALITY
101
upon cultural and historical conditions. If writing is used as it was by the followers of Emperor Lothar III "whose great deeds were inscribed on lead tablets and buried in his tomb so they would never be forgotten" (Morrison 1990: 215) the conceptual implications are necessarily different than if writing is used as it was by Robert Boyle who described his experiments with the vacuum pump in such detail as to transform his readers into "virtual witnesses" (Shapin 1984). Admittedly, theorists of writing have been slow to acknowledge these cultural considerations. Only recently has a new brand of anthropology developed to look into just such cultural ways of reading and writing (Boyarin 1993). But that in some way contradicts the claim attributed to Harris and developed here that it is the writing and not merely the cultural needs that have been critical in the formation of the particular cognitive style or mentality that is so prized in the West, namely, a literate mentality, a literate mode of thought. How, then, could writing prise apart utterance and sentence? And secondly, how could that prising alter one's mode of thought? The first question is relatively straightforward. Writing allows the preservation of utterance independently of the speaker. Traditional modes of interpretation required that authority be devolved by passing on the mantle of authority, empowering the inheritor with the right to say what a preserved utterance meant. This mode of interpretation is most clearly seen in the pre-Reformation Church but it is part of all modern institutions as well. The Supreme Court of Canada has the right to determine just what the statements of the Constitution and of the laws mean. Writing in such contexts may be useful for the accumulation of information but does not, in itself, call for a new way of thinking. Decisive for the so-called "Modern" tradition is a new way of reading and creating what I have called "autonomous" texts, texts which aspire to mean neither more nor less than what they say. This, I believe, is close to what Harris calls "authorless texts" - texts whose meaning depends upon linguistic properties of expressions rather than the intentional properties of speakers. These two notions are not identical and it would be worth sorting out the details of their difference, but for my purposes I will use them interchangeably, the important point being that one can look at the words, sentences, paragraphs as textual objects while setting aside, even if only temporarily, the intentions of the author in writing them. Intention, of course, comes back into the picture in that one uses the textual information as the basis for inferring the author's intention. It is, at least, an anachronism for a reader to see in a text a meaning which the author could not possibly have entertained. The important point is that the author's intention is now seen as a product of inference from textual information rather than as being given directly and manifestly in the text.
102
DAVID R. OLSON
The distinction is between what a text "says" and what a text "means", between wording and intention. Again this is not to say that no such distinction is possible without writing - puns, which require some awareness of the distinction between form and meaning, appear in traditional, non-writing cultures. Rather it is to say that writing provides a fixed text, independent of the speaker, which can become the object of such extended analysis, but again, only if someone has the leisure for such analysis or sees the use of it, and teaches the practice to others, thereby institutionalizing it. Why cannot sound and meaning similarly be prised apart in speech? To some extent they can. But the product is distinctive. Oral reflection yields new genres, new ways of meaning - it turns song into poem or utterance into memorable phrases for the use of the orator. Writing allows reflection on linguistic form words, grammar and logic. The claim then is that writing allows a different set of considerations to emerge. The second question - how does prising apart sentence from utterance alter mentality - is much more complex and has invited much more skepticism (Halverson 1992; Smith 1989; Street 1984). My conjecture is that writing affects the mind through the construction of a particular set of concepts designed for dealing with writing. These concepts constitute a kind of literate meta-linguistics. Leech (1983) has pointed out that no language exists without a metalanguage. Indeed, it could be argued that before children learn any words they have to learn the concept of a word. To know a thing one has to know what kind of thing it is, as Keil (1989) has argued in another context. Writing provides new objects that serve as the referent of new concepts. Among them are concepts like words, sentence, verbatim, paraphrase and literal meaning. All of them derive, I have suggested, from refinement of the concepts of saying and meaning. Harris (1986) suggested that word-signs - and hence consciousness of words - may have been instrumental in clarifying the distinction between things, names and words. He suggested that the failure to distinguish words from names produces a form of emblematic symbolism which may extend to various gods and spirits and "is often bound up in various ways with word magic and practices of name-giving. It reflects, fundamentally, a mentality for which reality is still not clearly divisible into language and non-language, any more than it is divisible into the physical and the metaphysical, or into the moral and the practical" (1986: 131132). That is, writing may be an important factor in "disenchanting" nature. We now recognize that disenchanting the universe may bring with it a lack of respect for the world and for other creatures. But we must also acknowledge that disenchanting the world was what made it possible to think that nature follows
A NEW MENTALITY
103
rules or laws which in turn could be formulated in written language of a "mathematical plainness of style" as the Royal Society of London urged in the seventeenth century. The research which my colleagues and I have conducted over the past decade has focused on children's understanding of "the very words" and their literal meanings (Lee, Torrance & Olson 1994; Olson & Torrance 1987). The laboratory for our research has been ordinary Western pre-school and schoolaged children whom we studied as they learned to read and write. Children in these schooled cultures begin to distinguish what was said and what a person meant by it when they are six or seven years of age. The acquisition brings with it a new form of competence. Children who are able to distinguish between what a person says and what they meant by it are also able to recognize who is at fault when communication failure results. One study (Robinson, Goelman & Olson 1983) showed that when playing a communication game in which a speaker was to refer unambiguously to an object in an array to enable a listener to pick out the identical object, children under six or seven years tended to "blame the listener" if the wrong object was selected rather than "blame the speaker" for his or her ambiguous message. This changed by the end of first or second grade. We infer that a new set of concepts was in play for dealing with, and thinking about, expression and intention. Note, too, that this new understanding is extremely important in writing in which one must see to it that, so far as possible, the expression can be read by another only in the way intended. So at the same time that writing is bringing expressions into consciousness, it is also bringing the intentions of persons into focus as well. It appears that the contribution of writing to the consciousness of language is, at the same time, contributing to a new sense of consciousness of the self. This idea is, of course, only a small start on an extremely interesting and important question of the rise of self-consciousness in the West (see Olson 1994, Ch. 11). A similar indication of the new understanding of the distinction between sentence and utterance, between words and meanings, comes from our studies of children's understanding of paraphrase. The concept of paraphrase contrasts with the concept of verbatim repetition. Children use paraphrase along with direct and indirect quotation from an early age, but it is the concepts and the ability to distinguish them that is at issue here. Our studies showed (Torrance, Lee & Olson 1994) that children can recall what was actually said (verbatim) and they can accept paraphrases. What they cannot do is distinguish them. Thus to cite an example from Hedelin and Hjelmquist (1988), if preschool children are told "The hippopotamus is hungry" they are likely to accept "The hippopotamus wants food" as what was actually said. Conversely, our study mentioned above showed
104
DAVID R. OLSON
that if they remembered the wording of a nursery rhyme, they were unlikely to accept a paraphrase as having the same meaning. The appropriate cross-cultural studies need to show that these distinctions are tied to literacy rather than to development more generally. These studies have not been conducted, but some research on metalinguistic concepts in non-literate cultures has begun to appear. Karanth and Suchitra (1993) found that a variety of metalinguistic judgments on topics ranging from words to sentences changed primarily, not as a function of age, but as a function of learning to read and write. Scholes and Willis (1991) made similar observations of non-literate adults in rural America. Finally, as to styles of reasoning, a much analyzed case is that of the studies reported by Luria (1976), which he had conducted in cooperation with Vygotsky, on the reasoning abilities of rural, illiterate peasants in a remote part of Central Asia in the 1920s. We must establish two points, first, that the style of reasoning is indicative, not of the absence of particular knowledge, but of an orientation to language. And second, that literacy is responsible for that orientation to language. A typical example of pre-literate reasoning reported by Luria was the following: In the far North, where there is snow, all bears are white. Novaya Zemlya is in the Far North and there is always snow there. What color are the bears there? To which a non-literate subject not untypically responded: "There are different sorts cf bears". [The syllogism is repeated.] "I don't know; I've seen a black bear, I've never seen any others... Each locality has its own animals: if it's white, they will be white; if it's yellow, they will be yellow". But what kinds of bears are there in Novaya Zemlya? "We always speak only of what we see; we don't talk about what we haven't seen". But what do my words imply? [The syllogism is repeated.] "Well, it's like this; our tsar isn't like yours, and yours isn't like ours. Your words can be answered only by someone who was there, and if a person wasn't there he can't say anything on the basis of your words" (Luria 1976: 108-109). Luria took the subject's failure as a failure of inference, specifically, an inability to reason syllogistically. The subject was repeatedly invited to derive what followed from the phrase "All the bears" and yet he refused to draw the necessary inference. That inferential abilities are at stake has been denied by Scribner and Cole (1981) who showed that if the premises are made into fanciful hypotheticals,
A NEW MENTALITY
105
non-literate subjects have no difficulty with them. In addition, we may note that the subject himself later in the interview used deductive reasoning - "if they are white they will be white". Yet all the while he refused to draw what seems to us the obvious conclusion. I have argued that what Luria's findings indicate is not so much a matter of logical reasoning as a matter of ways of taking texts. Luria intended his text to be taken as premises for an inference; the subject took them as an expression of hearsay. In Harris's terms we could say that Luria intended them as "authorless"; the subject took them as "hearsay" expressions of some unknown, but presumably unreliable, speaker. This was the conceptual gap that Luria and his subject could not negotiate. We may specify this relation even more precisely. Strict implication, the kind of reasoning favored in the sciences, as well as in law and other practical domains, is reasoning by entailment - what follows from a statement. Only certain kinds of statement permit such reasoning. What we call metaphor for example, does not. Thus, to return to the Huichol claim, "Corn is deer" does not permit strict deduction. Aristotle, recognizing that logic applied to only certain kinds of texts, ruled metaphor out of the strict sciences (Lloyd 1990). Our concept of literal meaning, I suggest, is that meaning for which strict deduction holds. Thus the concepts of logic and of meaning are closely related and both a product of writing. This orientation to text which literacy makes possible is important not only to school-like tasks; school-like tasks involve the kind of use of language and the kind of thinking we, in a modern literate society, take as paradigmatic. It is not just that we have come to use language in a new way, it is that we have come to treat language as an object, itself worthy of analysis and reflection. It is that attitude to language which is at the basis of a new mentality.
6 Science and Significance: Making Sense of Wittgenstein's Ways of Seeing Anthony Holiday
To be sure, we can if we wish stipulate that science and philosophy do not overlap, that there is an absolute distinction between conceptual and empirical questions and so on. But this is merely to adopt a certain academic stance in the context of twentieth-century Western culture. It is far from self-evident that such a position can claim to be underwritten by any eternal verities about "the nature of language" or its "limits". Science, philosophy and language, after all, are presumably just words like any others (at least, according to the author of the Philosophische Untersuchungen).
1. This observation, taken from Roy Harris's comparative study of Saussure and Wittgenstein, is disturbing in ways which need to be investigated. 1 The alarm it excites strikes me as being somewhat akin to what we should feel if we encountered a set of people whose sense of humour was not merely deficient by our standards but was altogether absent, people who not only could not see the point of any joke we told them, but could see no point in telling jokes at all, and who protested that our practice of telling them did not seem to be founded on any 1 R. Harris, Language, Saussure and Wittgenstein: How to Play Games with Words, London and New York: Routledge, paperback edition, 1988, p. 125. My defence of Wittgenstein's philosophical method against attitudes typified by the above quotation from Harris has benefited immensely from conversations with Professor David Pears of Christ Church, Oxford, who was so kind as to read an earlier paper of mine on Wittgenstein (see my "Teaching to Rule: Wittgenstein and Educational Practice" in Perspectives in Education, vol. 14. No. 2, Winter, 1993, pp. 157-179) and to discuss it with me during his visit to the University of Cape Town. I have gained much from written exchanges with my friend Dr P.M.S. Hacker of St. John's College, Oxford, who has, since my days as a graduate research student at Oxford, so often placed his immense scholarship and sharp but kindly criticism at my disposal. I should also like to thank members of the Philosophy staff seminar at the University of the Western Cape for their comments and criticism.
SCIENCE AND SIGNIFICANCE
107
"eternal verities" about the nature or limits of the languages in which our delight in the paradoxical, the incongruous and the ridiculous found expression. We should, I suggest, view such people as we regard victims of certain degrees of autism or "idiot savants". We should think of them as tragically deprived of something which made our lives valuable, fuller and, indeed, worth examining in the Socratic sense. The analogy between doing philosophy and joking is likely to strike many professional philosophers working in the contemporary mainstream of AngloAmerican analytical philosophy as overdrawn, and, indeed, given the theoryladen ambience in which they work and to which their own theories contribute, this is not to be wondered at. Nonetheless, the comparison is not far fetched. You cannot (except in some stretched use of the word "teach"), teach someone the point of a joke, just as you cannot teach a student to experience the peculiar fascination exerted by a philosophical enigma. In neither case will any amount of instruction do the trick, although, in both instances, carefully selected ranges of examples may prove elucidatory. There can self-evidently be no "jokeology", no science or general theory of jokes - which does not, of course, preclude the possibility of someone trying to draw up an exhaustive classification of the sorts of jokes there are - riddles, puns, "dirty" jokes, practical jokes and so on. That there can be no such science is not an hypothesis about the anatomy of humour; it is a grammatical remark about a salient feature of human life and language. As such, it does not claim to rest on some further and more unshakeable hypothesis as a truth of science might. To put the matter another way, this remark about joking is not, strictly speaking, a proposition, expressive of a judgement or opinion that things are thus-and-so as far as what touches humour is concerned; it does not record some fact of the matter about jokes and so makes no appeal to other more wide-ranging propositions to support its claim, for there is no claim that it makes. The remark that there can be no "ology" of jokes, or (what comes to the same thing) that jokeology would be a funny kind of science is itself a (rather weak) sort of grammatical joke and, so, is on a level with the species of conversational art on which it comments. Wittgenstein, as it happens, is especially consistent about the conceptual gulf which separates philosophical activity from scientific theory and is also at pains to be clear about the connected issue of the affinity between joking and philosophising. The Tractatus Logico-Philosophicus is driven by its author's conviction that the necessitarian claims made for scientific hypotheses usurp what belongs to logic alone - viz. an ineffable form which is the sole origin of such certainties as are vouchsafed to us, the only possessor of ineluctability and a form, moreover, which it is the philosopher's business to make visible, not by recourse to propositions, but by elucidation. Thus the Tractatus says plainly that
108
ANTHONY HOLIDAY
philosophy is not one of the natural sciences, is not a body of doctrine but an activity and results, not in "philosophical propositions" but in the clarification of propositions (TLP 4.111-4.112).2 The Philosophical Investigations puts the matter even more firmly: ".. .we may not advance any kind of theory. There must not be anything hypothetical in our considerations. We must do away with all explanation and description alone must take its place" (PI § 109, emphasis Wittgenstein's). And then, in just the context of these stipulations, and of the pronouncement that the "superstition" that language is something unique is the product of grammatical illusions (PI § 110) comes the comparison of the depth of philosophy with the depth of a "grammatical joke" (PI § 111). Wittgenstein's point here is that jokes and philosophical enigmas have common sources of significance in the forms of our language and in our liability to misunderstand, to impose misinterpretations on those forms as when, for instance, someone believes they have a problem about whether or not it will be (logically) possible for them to attend their own funeral because, if the answer is no, then it will not be their funeral, whereas if it is yes, they will not be dead. The matter has only to be put like this for it to become evident that the reason why science is helpless to alleviate difficulties of this kind is not merely that philosophy deals in concepts and science trades in empirical wares, although this may be true as far as it goes. It is, rather, that someone in the grip of philosophical perplexity will not be liberated by the provision of more information, does not need to be provisioned with a further list of propositions, but does need to be helped to look at the way language functions, to try alternative arrangements of what is already known, what is there on the surface and open to our view. This is why, for Wittgenstein, philosophy "leaves everything as it is" (PI § 124). 2
I have followed the standard practice of referring to Wittgenstein's writings in the body of my own text, giving the initial letters of the work cited, followed by the appropriate page, proposition or section number. Works referred to in this way are: TLP RLF BB PI Ζ CV
Tractatus Logico-Philosophicus, transi. D.F. Pears and B.F. McGuinness, London: Routledge & Kegan Paul, 1961. First German edition in Annalen der Naturphilosophie, 1921. First English edition, transi. C.K. Ogden, 1922. "Some Remarks on Logical Form", Proceedings of the Aristotelian Society, supp. vol. 9, 1929, pp. 162-171. The Blue and Brown Books: Preliminary Studies for the Philosophical Investigations, Oxford: Basil Blackwell, 1958. 2nd éd., 1969. Philosophical Investigations, transi. G.E.M. Anscombe, Oxford: Basil Blackwell, 1953. 2nd éd., 1958. Zettel, ed. G.E.M. Anscombe and G.H. von Wright, transi. G.E.M. Anscombe, Oxford: Basil Blackwell, 1967. 2nd éd., 1981. Culture and Value, ed. G.H. von Wright and H. Nyman, transi. P. Winch, Oxford: Basil Blackwell, 1980.
SCIENCE AND SIGNIFICANCE
109
What he intends by this dictum is not that, since "philosophy" is a "word like any other", philosophical language-games should be left alone and exempted from criticism because "what we do" is where the "spade is turned" and is perfectly in order and wholly, as it were, self-justificatory. What he does mean is that philosophy does not enable us to criticise or justify what we do by showing that our doings match or mismatch criteria laid down in the course of our philosophising. Wittgenstein, after all, went in for abundant criticism of the language games other philosophers and he himself had played or thought there was some point in playing. Indeed, insofar as it is helpful to situate him within a period or tendency within the history of Western philosophy, he belongs firmly within the tradition of critical philosophy inaugurated by Kant. But what his criticism results in is not demonstrations that some or other set of philosophical requirements have or have not been met. Rather it issues (for those his arguments convince) in the insight that certain sorts of philosophical language-game might no longer be worth playing because the point of playing them depended on fantastical mythologies - about the need to establish a logical foundation for arithmetic, say, or to find a definitive proof for the existence of God - and because the playing of them held out only an illusory promise of satisfying the urge which drew us to philosophy in the first place. Harris's dismissive posture in respect of Wittgenstein's methodology seems to me to show that he misses all or most of this. It might even be argued - although I have neither the wish nor the competence to argue the point here - that the very project of trying to compare Saussure's structural linguistics with the descriptive explorations of Wittgenstein's later philosophy of language is evidence of a deep failure to understand the latter, the critical objective of which is precisely to expose as chimerical the notion that our capacities to speak and be understood are made perspicuous by reference to some or other structural arrangement. Aside from that, however, before Harris would be remotely entitled to insinuate as he does that Wittgenstein's prohibition on philosophical theorising was groundless and arbitrary, he would need to offer a critical account of the crucial distinction, made in the Tractatus, between what a proposition says and the way in which its sense is shown. This distinction is the ultimate source of the methodological distinction between science and philosophy, since sciences consist of what is sayable - i.e. of the data propositions convey - and not of what propositions show, since this cannot, according to Wittgenstein, be said. Yet Harris makes no mention of the saying/showing distinction, as though he thought Wittgenstein's claim that a philosophical work consists essentially of "elucidations" (TLP 4.112) could be understood without it! But I have, at best, only a marginal interest in proving that Harris's exegesis of Wittgenstein is wrong. The mistakes philosophers make are, as a rule, far less
110
ANTHONY HOLIDAY
interesting than the points of view and chains of argument that led to the making of them. So I shall rest content with what I have so far had to say about the misapprehension Harris's comment on Wittgenstein's methodology betrays and devote the rest of this paper to an attempt to discover how the misunderstanding might arise.
2. Cora Diamond, in her penetrating study of Wittgenstein, insists that his view of philosophy as not consisting of doctrines or theses is not only essential to a grasp of almost all that is valuable in his thought but is itself something that has first to be seen in the Tractates if the later forms it assumes are to be understood, "and in the Tractatus it is inseparable from what is central there, the distinction between what can be said and what can only be shown".3 That judgement says a great deal for her perspicacity. But Harris, as I say, ignores the saying/showing distinction and his treatments of the Tractarian thoughts are cursory at best. Why is this? I suggest that his carelessness is due, in the first place, to a failure to take seriously that it is true of Wittgenstein, in a sense and to a degree that it is true of few other philosophers of this century, that the development of his philosophy needs to be understood as a whole and that the unique way in which that whole is constructed, the special configuration of the relationship between its parts and periods of construction needs to be taken fully into account as well. The pioneering studies produced by Diamond herself but also by scholars like Baker,4 Hacker5 (whom Harris relies on to some extent) and Pears6 3
C. Diamond, The Realistic Spirit: Wittgenstein, Philosophy and the Mind., Cambridge, Mass./London: MIT Press, 1991, p. 179. 4 For some best fruits of a partnership in collaborative scholarship which ultimately broke down for reasons which included disagreements over interpretation and nuance, see G.P. Baker and P.M.S. Hacker, Wittgenstein: Understanding and Meaning, Oxford: Basil Blackwell, 1980, and Wittgenstein: Rules, Grammar and Necessity, Oxford: Basil Blackwell, 1985. These works constitute Volumes 1 and 2 respectively of a projected four-volume analytical commentary on the Philosophical Investigations. See also G.P. Baker, Wittgenstein, Frege and the Vienna Circle, Oxford: Basil Blackwell, 1988, which marks Baker's break with Hacker and indicates one source of the disagreement through its reappraisal of Frege's philosophical logic in the light of Wittgenstein's later writings, more especially the Investigations and the Remarks on the Foundations ofMathematics. 5 P.M.S. Hacker, Wittgenstein: Meaning and Mind, Oxford: Basil Blackwell, 1990. This is the third volume of the proposed four-volume commentary and deals with §§ 243-427 of the Investigations. At the time of writing, Hacker was engaged in preparation of the final volume in the series. For an instructive study in the difficulties of presenting a clear overview of
SCIENCE AND SIGNIFICANCE
111
have brought us some distance from the eccentric picture of Wittgenstein's philosophy as an essentially fissured production in which the later writings, notably the Investigations, have a solely critical and corrective function in relation to the early texts, notably the Tractatus. But - and this may be at the bottom of Harris's apparent insouciance - if we are now more disposed to pursue continuities in Wittgenstein's thought than were a previous generation of exegetes, we are by no means certain of how to go about this, of how, that is, to disentangle from the fine-spun tapestry of his meditations threads which are indeed continuous and to distinguish them from strands which are broken or frayed. We are, connectedly, uncertain of what to make of his unrealised intention of having the Tractatus and the Investigations published between the covers of the same volume, because he believed that his new thoughts "could be seen in the right light only by contrast with and against the background of my old way of thinking" (PI : viii). This plan was approved by the Syndics of Cambridge University Press when Wittgenstein reopened publishing negotiations with them in 1943 but was never put into effect. But, even had the project been implemented, this would not of itself have told us what its architect intended by "the right light" or what kind of "contrast" between old and new thoughts he had in mind. The suggestion I once made 7 that Wittgenstein hoped that the two works, if presented in this format, would be perceived as an ironical dialectic in which old and new thoughts mocked one another, now seems to me, although not misleading or entirely unhelpful, too far from the mark to stand alone as an exegetical guide. For all that, enough work has been done to sharpen our sensitivities to the subtlety and many-faceted character of Wittgenstein's approaches to all the major objects of his concern. Whether he is discussing logical form, the status of ethical values, rule-following or (in the case towards which I am making my way) the referentialist proto-theory of how language works, he seldom permits himself to draw sharply defined boundaries between "correct" and "incorrect" accounts, trying instead to show how imperceptibly the former shade into the Wittgenstein's philosophy see P.M.S. Hacker, Insight and Illusion, revised edition, Oxford: Basil Blackwell, 1985, in which Hacker breaks decisively with positions adopted in the first (1972) edition, where he had tried to treat the Tractatus as a paradigm of truth-conditional semantics and the Investigations as a work informed by anti-realism or assertion-condition semantics. 6 D.F. Pears, The False Prison: A Study of the Development of Wittgenstein's Philosophy, 2 vols., Oxford: Oxford University Press, 1987, 1988. In my view, this work constitutes an unrivalled treatment of the continuities and disjunctures in Wittgenstein's thought. 7 A. Holiday, Moral Powers: Normative Necessity in Language and History, London and New York: Routledge, 1988, pp. 2-9.
112
ANTHONY HOLIDAY
latter and how what begins as a natural expression of our unobjectionable sense of the way things are may migrate towards metaphysics. This is certainly true of the Investigations\ the work on which Harris chiefly relies for his treatment of the doctrine he calls "nomenclaturism". But it is also true of the Tractatus which, notwithstanding its very different style and mode of argumentation, demands the closest possible attention if there is to be any hope of sorting out, for instance, where Wittgenstein agrees with and where he differs from Frege on such issues as the categorial distinction between functions and objects and the nature and extent of his agreements and disagreements with Russell on such matters as the theories of logical types and definite descriptions. It follows that a superficial reading of the Tractatus is liable to result in a simplistic view of the criticisms Wittgenstein makes of it in the Investigations. A crass understanding of the Tractarian view of how names relate to their bearers will issue in a crude grasp of the way that view is modified and demythologised in the later philosophy and this, in its turn, will result in a failure to see how and why both the notion of "showing" and the anti-doctrinal provisos survived the demolition of the metaphysics of symbolism which characterised Wittgenstein's first magnum opus. The interpretation Harris offers in his Language, Saussure and Wittgenstein is enfeebled in precisely this way. It is flawed in respect of its omissions and weak perceptions rather than by his having set down what is straightforwardly false. He is not incorrect, for instance, when he follows Baker and Hacker in asserting that the famous quotation from Augustine's Confessions with which Wittgenstein begins the Investigations is not intended to represent (otherwise it would misrepresent) what Augustine qua philosopher believed about language and language-learning, but is being used instead as a convenient device for the identification of real and less naive philosophical theses. Neither is Harris wrong when he says that in attacking the "Augustinian picture", according to which the main function of language is to affix names to extra-linguistic objects, Wittgenstein "is really attacking his own earlier views, those put forward in the Tractatus, and at the same time closely similar views held by other philosophers, notably Russell and Frege".8 But there is really very little to be got from being told this unless we are favoured with a more enriched account of what these views of Frege and Russell were, so that we can see what drew Wittgenstein (who after all never shared either Frege's ambition of reducing arithmetic to the fundamental laws of logic or Russell's even grander plan for deriving all mathematics from logic) to them and unless we have an inkling of his own reservations about and modifications of the positions he embraced. Why, for 8
Harris, op. cit., p. 13.
SCIENCE AND SIGNIFICANCE
113
example (to take a matter Diamond deals with), 9 should Wittgenstein attach himself to Russell's theory of denotation, which rejects as artificial Frege's SinnlBedeutung distinction in trying to deal with the problem of apparently denotationless denoting expressions, when we know that Wittgenstein had a better opinion of Frege's work than of Russell's? What role is the theory of definite descriptions playing in the Tractatus! How, given that it is a doctrine, a theory, did it get into the Tractatus in the first place? Harris's rendering of Russell's theory is far too shallow to address questions of this kind. He tells us little more than that "Russell had developed the idea that words are names by positing that they stand not only for concrete things but for abstract things as well" 10 and that Russell was "a foremost champion" of the idea that philosophy was the most general of the sciences.11 Before we can hope to understand Wittgenstein's early or his later ideas about naming and logical necessity we need a far more fine-grained account of Russell's theory than this, one which will help to explain how Wittgenstein might have thought to reconcile his total disagreement, noted by Harris,12 with Russell's view of philosophy as a super-science with adherence to Russellian referentialism and which will, additionally, connect that adherence to the saying/showing distinction and the socalled "mystical" part of the Tractatus. But as a preliminary to giving this more detailed version of Russell's theory we need to remind ourselves of some home truths about the Tractatus. The first of these is that a sober and non-simplistic reading of the book demands a firm grasp of the slogan that absolute determinacy of sense requires absolute defmiteness of denotation. The demand that the sense (Sinn) of a sentence be determinate is both something Wittgenstein owes to Frege and part of the former's thrust against philosophical scientism. It was, Wittgenstein held, a scientistic superstition that observation-based hypotheses could yield predictions which were certainly true. "It is an hypothesis that the sun will rise tomorrow: and this means that we do not know whether it will rise" (TLP 6.36311). This claim flows from the circumstance that genuinely propositional well-formed sentences are truth-functionally bipolar - i.e. they are always possibly true or possibly false - unlike tautological sentences which will always yield a line of Ts on any truth-table drawn up according to the rules of the propositional calculus or contradictions which will always disclose a line of Fs. The sense of a proposition, therefore, shows itself quite independently of whether what the 9 Diamond, op. cit., pp. 186-187. 10 Harris, op. cit., p. 133. n Ibid., p. 122. 12 Ibid., p. 127.
114
ANTHONY HOLIDAY
proposition says is true or false, and the Sinn is nothing other than the logical form of the proposition which, of course, is what it is entirely non-contingently. Thus "the only necessity that exists is logical necessity" (TLP 6.37) for it is only, so to speak, within the transcendental limits set by logical space that the boundaries of real possibility and impossibility are defined. The so-called "propositions" of philosophy can elucidate, not explain, these boundaries, for the latter are the preconditions of all sense-making sentences and, hence, of all explanation by means of sentences. But that which philosophical sentences do properly aim to illuminate lies outside the scope of all propositional utterance and so is the exclusive preserve of philosophy or of the family of gestural practices which are akin to philosophy simply because "what can be shown cannot be said" (TLP 4.1212). The saying/showing distinction, then, is the basis for Wittgenstein's attack on what he regarded as a kind of sorcery, inflicted on the modern intelligence by the epistemic paradigms of science and the ground for his saying that The whole modern conception of the world is founded on the illusion that the so-called laws of nature are the explanations of natural phenomena. (TLP 6.371) His point is that the explanatory pretensions of science rest on opinion, not on certainty - a theme which recurs powerfully in the later writings - or, to put the matter as Plato might have done, they are rooted in doxa, not in epistemê. It follows from all this that Wittgenstein would have to have vehemently opposed the idea that logic was a science or could be the object of scientific study. For logic was not a body of doctrine but a mirror-image of the world; it was "transcendental" (TLP 6.13). The fundamental idea underlying Tractarian philosophical logic was that there were no logical constants and that Frege was mistaken in thinking that the logical connectives were names for "logical objects", instead of recognising that what logical operators signified was not a set of objects but a set of operations such as conjunction, disjunction, implication and negation. There were, therefore, no objects for aspirant logical "scientists" to study. There was, moreover, in Wittgenstein's view, no foundation for the Fregean-Russellian aspiration (one which harks back to Leibniz) to invent a logically perfect language, because "all the propositions of our ordinary language, just as they stand, are in perfect logical order" (TLP 5.5563). Frege and Russell had seen in the former's new "concept-script" the making of an ideal tool for a special sort of scientific inquiry. But Wittgenstein, as we have seen, thought the project of conducting such an inquiry to be entirely misconceived. He saw the value of the new function-theoretic logic as lying, not in substituting a more perfect medium for the rough infelicities of natural language, but in
SCIENCE AND SIGNIFICANCE
115
providing, in the form of the propositional and predicate calculi, a notation which would perspicuously reveal the hidden logical structure which, through the workings of its iron-clad necessity, made all meaningful utterance possible. It is at this point in the chronicle that a tension emerges between Wittgenstein's resistance to what he took to be the illicit intrusion of science into philosophy, advocated by Russell, and his own urge to communicate his vision of how a transcendental logical essence was made manifest in the mystical union of language and world. The world was the totality, not of things, but of facts {TLP 1.1), and language was the totality of propositions {TLP 4.001). Yet, at the level of natural language, the translucent perfection of this order was quite unapparent. He had written that The limits of my language mean the limits of my world. {TLP 5.6) The task he had set himself was to trace those limits from within. Yet now it seemed that this could only be done if he followed Russell in trying to delve beneath the semiotic surface in order to uncover the fine detail of how those limits were constituted. The alternative was to rest content with a camouflage. Thus he says: Language disguises thought. So much so, that from the outward form of the clothing it is impossible to infer the form of the thought beneath it, because the outward form of the clothing is not designed to reveal the form of the body, but for entirely different purposes. {TLP 4.002) It is to this decision of Wittgenstein's to yield, as he was later to warn others not to yield, to the intellect's hunger for generality {BB: 17) and to follow Russell in treating natural languages as mere epiphenomenal concealments of a uniform logical structure - quite as much as to Frege's function-and-argument analysis that we owe the Tractarian inventions of the logically independent "elementary" propositions, the simple objects which are supposed to make up the invisible metaphysical substance of the world and the atomic names which denote those objects. The latter combine immediately to constitute the elementary propositions {TLP 4.22-4.221), and never occur outside the nexus of such propositions, which stand to them as function stands to argument {TLP 4.23-4.24). The logical independence of the elementary propositions is a function of the requirement that sense be absolutely determinate. For the proposition pictures states of affairs, given by the various configurations of the simple objects constitutive of the world's substance. Were there no such substance, "then whether a proposition had sense would depend on whether another proposition was true" {TLP 2.0211)
116
ANTHONY HOLIDAY
- i.e. on something contingent - in which case no picture of the world (true or false) could be sketched. The atomic names and their corresponding metaphysical atoms are functions of the matching requirement of absolute definiteness of denotation. The atomic names denote the sempiternal atoms. If we credit the existence of these atoms then we will be driven to conclude with Wittgenstein that A name means an object. The object is its meaning. (TLP 3.203) Thus the requirements of determinacy of sense and definiteness of denotation in fatal combination with Russellian logical essentialism lead on inevitably to acceptance of the dogma, which Russell adhered to for the rest of his life, that the meaning of a name is that name's bearer. It is natural to wonder whether Wittgenstein could not have found an alternative way to satisfy the twin requirements - one which would not have compelled allegiance to the dogma which was to be the target of so much of his later criticism. But, leaving that issue aside for the moment, it is safe to say that Russell's motives for devising his theory are far less mysterious. For his scientistic project, which included in its aim the reduction of mathematics to logic, was bound to run foul of features of natural language which, from the standpoint of mathematical logic, appear to be inexactitudes, anomalies and paradoxes. From the point of view of the philosophical scientist, the most vexing of these is the circumstance that everyday language contains such denotationless denoting expressions as "the present King of France". In his early work, The Principles of Mathematics, which introduced Wittgenstein to Frege's philosophy of mathematics, Russell made his first stab at resolving problems of this sort.13 He set forth a very pliant version of the doctrine that name-meanings are identical to named objects, allowing in principle for a name to count as a genuine referring expression even in cases of logically impossible "objects", like round squares, and mythical beasts, like unicorns. In his 1905 paper, "On Denoting", 14 however, Russell adopted a far more rigid approach. He rejected Meinong's solution which allowed a kind of existence to pseudo-entities of the round-square variety and, in place of Frege's SinnlBedeutung distinction, introduced the stipulation that in the case of such puzzling phrases as "the present King of France", the Bedeutung should be a null class. Russell's strategy in this matter was dictated by his logicism regarding mathematics. He had to ensure that he produced a theory of logical syntax which 13 B. Russell, The Principles of Mathematics, London: Allen and Unwin, 1903, pp. 53-65, 73 and 502. 14 B. Russell, "On Denoting", in R.C. Marsh (ed.) Logic and Knowledge, London: Allen and Unwin, 1956.
SCIENCE AND SIGNIFICANCE
117
did not violate the Law of Excluded Middle as denotationless denoting phrases seemed to do. For that law stipulates that either "the present King of France is bald" or "the present King of France is not bald" must be true. But since France presently has no king, he cannot appear in any list of bald or hirsute beings. The solution Russell proposed to this dilemma was first expressed in the clumsy notation of "On Denoting", and later given a more polished form in Principia Mathematica by the introduction of quantifiers. It consisted of a refusal to recognise denoting phrases like "the present King of France" as being names at all! Instead, they were held to be incomplete symbols, which were to be treated as significant solely on the basis of their contribution to what was to be understood by the sentences in which they occurred. Logical analysis could, Russell held, demonstrate that such phrases as "the author of Waverley9 did not, even on the assumption that one and only one person (Sir Walter Scott) wrote Waverley, signify that person. The belief that such locutions were logically proper names was an illusion, generated by surface grammar. Accordingly, all sentences of the form "the such-and-such is thus-and-so" were to be analysed according to the schema "one and only one thing Xs and it also Ys". Ambiguous cases like "the present King of France is not bald" could be settled by determining the scope of the existential quantifier- i.e. the logical symbol which roughly translates as "there is (exists) at least one X such that..." Where the troublesome expression was taken to mean that there is one and only one present King of France and he is not bald, the descriptive phrase was said to have "primary" occurrence. Where, on the other hand, what was meant was that it is not the case that there is one and only one present King of France, bald or not, the descriptive phrase was said to have "secondary" occurrence. Russell was undismayed that his theory rescued his credo about the way names were tied to bearers at the price of dramatically reducing the number of symbols which could qualify as logically proper names. He agreed with philosophers such as Strawson,15that everyday language had no exact logic, but was unable to sympathise with the then fashionable "ordinary language" movement at Oxford and elsewhere which sought to preserve philosophical discourse from encroachments by the artificial exactitudes of the language and symbolism of the special sciences. His obduracy was a function of his empiricism. Language had to relate directly to phenomenal facts, to the data, or it was meaningless. If ordinary language in its vagueness and ambiguity was unequal to the task of preserving that relation, then so much the worse for ordinary language. But Wittgenstein, contrary to the early anticipations of the Vienna Circle positivists and despite interpretations of his philosophy offered by 15
B. Russell, "Mr Strawson on Referring", in B. Russell, My Philosophical Development, London: Allen and Unwin, 1959, pp. 238-245.
118
ANTHONY HOLIDAY
contemporary anti-realists, had no empiricist axe to grind. So why (to return to a question which came up earlier) did Russell's Theory, and not some other method of analysing sentential structure, attract him? To answer this question we need a fresh perspective on the theory so that we see it not as we have so far done, only from the point of view of its architect, but from the standpoint of someone who felt impelled to employ it for the very different purpose for which Wittgenstein put it to use. The challenge here is to elaborate on the account just rehearsed in a way which still leaves it recognisable as Russell's Theory, so that we do not breach the exegetical canon of charging Wittgenstein with distorting Russell's thought to suit his own ends without bringing adequate evidence to support such a charge. In order to do this, we must first clear up an area of potential misunderstanding: we must recognise that, whatever its faults and however trenchant his later criticisms of it, it was not the doctrine that the meaning of a name is its bearer which was responsible for the collapse of the Tractarian metaphysical system. The responsibility for that catastrophe is ultimately traceable to the impossibly stringent requirement that the elementary propositions exhibit complete logical independence one from another - a stipulation which, as we saw, was introduced in conformity with the demand for absolute determinacy of sense, the demand which alone guaranteed that the sense of a proposition stemmed from what was wholly non-contingent and purely formal. The stipulation that elementary propositions be logically independent was subverted by the notorious intractability of the colour-exclusion problem - viz. the threat posed to propositional independence by the logical truth that it is impossible for something to be at once, say, red and blue all over at the same time. Wittgenstein had thought that the entailment "If Rx then not-Bx" could be shown by reductive analysis to be no genuine entailment at all but a mere feature of distortive surface grammar. Such analysis, he thought, would transform the degrees of colourproperties into a logical product of single statements of quantity, adding to each the qualification, "and nothing else". The thought powering this procedure is that the colour red contains all degrees of its own coloration and no degree of the colour blue and vice versa, so that the presence of R at Tl, T2... and at PI, P2... excludes, but does not contradict B's presence at those points in time and space. That contention, however, is empty. It serves only to delay, not to resolve, the problematic crisis which colour-exclusion poses at the level of natural language and thought. For if colour-predications contain statements of degree, then numbers inevitably enter into them. If numbers occur in the way colourpredications are formulated, then such formulations must fail as analyses of statements of degree, aspiring to shore up the proposal of the logical independence of elementary propositions. They must fail because analysis of a
SCIENCE AND SIGNIFICANCE
119
statement like "s is red" can only show either that the degrees of vividity into which the statement is analysed are identical, so that the logical product will be no more than that specific degree - since the logical operation of conjunction does not function like the arithmetical plus-sign - or the analysis will show the degrees as being different, so that their co-presence is contradictory and we are back to the problem we started out with. Moreover it is an insoluble problem, because, as Wittgenstein conceded in the paper which marks his return to professional philosophy in 1929 (RLF: 166), the entrance of numbers into the statements of degree resulting from the analysis of colour predications is not just a feature of a special symbolism but an essential part of the representation. When he returned to the issue in the Philosophical Remarks, he was forced to admit that if different degrees of colour were mutually exclusive, elementary propositions could be contradictory, and once this admission was made the demise of these "elements" themselves along with such other lynch-pins of the Tractarian edifice as the atomic names and the atoms they were supposed to denote was inevitable. Hopefully this recital of the reasons for the downfall of the metaphysics of the Tractatus does more than exonerate Russell's theory from the accusation of being villain-of-the-piece in the catastrophe. It ought also to serve to remind us of what was of cardinal importance to Wittgenstein at this juncture, to give an indication of why he risked his entire system for the sake of an impossible requirement and thus to shed light on how the interests of that attachment might have seemed to him to have been best served by adopting Russell's account of how namemeanings are tied to meant objects. What Wittgenstein thought of first importance, we saw, was that the absolute determinacy of a proposition's sense be preserved or, to put the matter accurately, he set his chief priority in erecting his elucidatory scaffolding on showing that determinacy of sense had been preserved absolutely. The contribution Russell's theory of denotation made to his overall enterprise in the Tractatus was precisely that it allowed him to do this by affording him a notation and method of analysis which clarified, without the need for recourse to any further explanation, just what kind of function a sentence was by distinguishing its real from its apparent logical form. How did Russell's theory accomplish this for his former pupil? It did so by dint of a technique which seems to lay bare the truth that a "sentence", if it is anything, is a judgement, an opinion which bisects logical space by virtue of its possible truth or falsity. Only propositions do this, because only propositions are possibly true or possibly false. On Russell's finding, if a sentence says something true or false, it does so quite independently of whether any definite description it contains is empty or satisfied; its sense does not rely on such possible satisfactions or disappointments concerning the emptiness or plenitude of the name-places in its everyday formulation as such a formulation may excite.
120
ANTHONY HOLIDAY
Put differently, the Russellian analysis appears to display the significance of genuine sentences by letting us see that their truth-functional bipolarity is unaffected by variations in truth-value in a given range of sentences. Thus it will not emerge that only if a sentence is true will it have any truth-value at all; that only if the denoting phrase in "the present King of France if bald" is satisfied will the sentence be truth-valued; and it will be clear that in uttering it someone is expressing a judgement, an opinion. Moreover, and most importantly, if the assumption that Russell's account is the right one is granted, then that account affords a way of making the crucial distinction between genuinely sentential structures which can, and bogus sentential structures which cannot, be laid against reality like a measuring tape. The latter comprise two sorts of pseudosentence - viz. contradictions, which have the form "p and not-p" and tautologies which have the form "p or not-p". Tautologies and contradictions are nonsensical in the sense that their logico-syntactic make-up is degenerate and the transformation rules of modern mathematical logic will show that the operation of negation turns one of these forms of nonsense into the other. But these quasisentential forms are the sole modality in which the "truths", the "laws" of logic can find expression; they mark the boundary between the world which genuine fact-stating propositions are supposed to be about, and the transcendental realm which gives the world its sense. That is why contemplating a tautology evokes the feeling of the world as a limited whole, the feeling Wittgenstein equates with what is mystical {TLP 6.45). Russell's theory of descriptions, insofar as it enables the unsayable difference between propositions and tautologies simply to evince itself, therefore provided the Tractarian Wittgenstein with a seemingly indispensable device for marking the limits between the world of clearly statable facts and the domain of that respecting which nothing can be stated at all.
3.
We are now in a far better position than Harris's meagre treatment makes available to approach the mature Wittgenstein's self-critical response to his Tractarian efforts to cope with the questions posed by referring expressions and ostensive definitions. A first thought might be that the special emphasis on practice over theory which is a feature of the later writings simply dissolves the problematic. But the matter is less straightforward than that, as this key passage from the Investigations shows:
SCIENCE AND SIGNIFICANCE
121
For a large class of cases - though not for all - in which we employ the word "meaning" it can be defined thus: the meaning of a word is its use in the language. And the meaning of a name is sometimes explained by pointing to its bearer. (PI § 43) The caveat this remark carries is instructive. It signals that the baby of our realistic, pre-philosophical intuitions is not to be thrown out with the Russellian bath-water; that, however wide of the mark it may have been, there is something about the analysis of naming and named objects, given in the Tractatus, which it is imperative to salvage and that what this "something" is can be brought out by paying the right kind of attention to the practice of explaining the meaning of a name by pointing to that which it names. With this in mind, let us look at the passage where Wittgenstein launches his most direct assault on Russell's doctrine concerning denotation: Let us first discuss this point of the argument: that a word has no meaning if nothing corresponds to it. - It is important to note that the word "meaning" is being used illicitly if it is used to signify the thing that 'corresponds' to the word. That is to confound the meaning of a name with the bearer of the name. When Mr N.N. dies one says that the bearer of the name dies, not that the meaning dies. And it would be nonsensical to say that, for if the name ceased to have meaning it would make no sense to say "Mr N.N. is dead." (PI § 40). Notice that Wittgenstein's point here is not that there is no relationship whatsoever between the meaning of a name and its bearer, but that such relationship as there may be is not one of direct one-to-one correspondence. Notice further that, in order to make that carefully qualified point and in both the above-cited remarks, Wittgenstein appeals to well-established practices, firmly founded institutions embedded in our lives and the ways in which we unquestioningly live them. In the first case the practice invoked is our custom of using indexical gestures to explain name-meanings, while in the second attention is drawn to our practice of using the same name for a person after that person has died. Just because these practices, however ancient and deeply ingrained in our lives they may be, are normative and not basic natural regularities, it is possible to wonder whether they might have been other than they are without thereby fatally subverting the rationality of the order they serve. We can, if we like, conduct thought-experiments in which we try to imagine a community whose members never explained the meaning of a name by pointing to its bearer with their index
122
ANTHONY HOLIDAY
fingers because they held that to do so destroyed the essence of whatever had borne that name, or we might attempt to imagine a tribe who never used people's names to speak about them after they had died but only referred to them in terms of kinship relations such as "your departed elder sister", and so on. We can even test the cogency of such scenarios by expanding the examples and their settings in order to see how the "alternative" customs would ramify out into other reaches of the lives of our imagined communities and so try to test whether what emerged as possible in those lives really makes sense to us. The essential point, however, about all such actual or imaginary examples is that it is what human beings do that establishes the semantic link between their signifying activities and that which they intend by those acts to signify or pick out. It is human practice and only human practice that, in the rich variety of its manifestations, establishes, over long periods of time, that cluster of rulegoverned activities whereby we play our naming games when we play such games together, and it is this practice which displays the living form which is inseparable from our capacity to make meanings as we do and to respond with understanding to what we have made. We do not, of course, establish these norms just as we please. Normative must co-operate with natural necessities if babble is to be distinguishable from speech. In a world where objects appeared and disappeared randomly, changed shape and colour for no apparent reason, the institution of naming could not be sustained. Nonetheless, the descriptive overview of language in general, and of the assignment of names in particular, which is offered in the early sections of the Investigations, does assign to speaking beings a special kind of responsibility for establishing and maintaining the limits segregating sense from nonsense. Does this way of putting matters, in Wittgenstein's phrase, "abolish logic" (PI § 242)? Not at all. What it does do is to redescribe the logic of our language in a way which avoids the metaphysical eccentricities of both the Russellian and the Tractarian theories of reference. In so doing, of course, it abandons certain metaphors and concomitant habits of thought, discouraging us from thinking of the canons of logic as a kind of supernaturally sharp invisible razor-wire or as being preternaturally "hard". But, not only does this revisionary procedure leave the legislative force of the rules of logical syntax unimpaired, the focal point of the inquiry remains exactly what it was in the Tractatus - viz. logic and the quest for the most cogent description of the way logical necessity underpins semantic possibilities. What has changed is the perspective from which the investigation is to be conducted. For we are now being invited to look at logic from a wholly different and more holistic point of view. As Wittgenstein puts it,
SCIENCE AND SIGNIFICANCE
123
. . .the axis of reference of our examination must be rotated, but about the fixed point of our real need (PI § 108). Now it might well seem that, even if what we are dealing with in the Investigations is not more than a gestalt-switch in our view of logic, it has at least deprived that discipline of the transcendental status it enjoyed in the Tractatus. What speaks in favour of this reading is the virtual absence in the Investigations of the neo-Kantian aroma which is so unmistakable a feature of its illustrious predecessor. The special employment Wittgenstein gave in his first book to Kant's version of the two-worlds concept by making the sense of that slice of reality which can be captured by propositional attitudes depend on an ineffable transcendental reality, an unworldly "Beyond" (the idea that logical form determines language from the outside, as it were), seems to have been made redundant by the proclamation in the Investigations that "to imagine a language means to imagine a form of life" (PI § 19). Two considerations, however, strike me as telling decisively against this position. Firstly, there is nothing in Wittgenstein's later self-critical writings which constitutes an unambiguous retraction of the Tractarian standpoint regarding the status of logic. The critique of metaphors just mentioned does not constitute such a retraction. To suggest that logical form should not be imagined as a "halo" surrounding thought, a crystalline essence (PI § 97) and so on is not at all to deny what the Tractatus asserts in stating that logic shares with ethical and aesthetic values that numinous absolute character which is reminiscent of Plato's forms. In the Investigations, however, the form has, so to speak, been rendered incarnate by dispensing with the term "logical form" and adopting in its stead the Spenglerian phrase "form of life". The shift in terminology reflects - as Wittgenstein acknowledges in a remark penned in 1931 (CV 19) - the influence of Spengler's pessimistic historiography in assisting the mature phases of his work of clarification, an influence which (as the remark evidences) was comparable in its potency to that exerted by such figures as Russell, Frege, Kraus, Loos, Weininger and Sraffa. Some students of Wittgenstein's development are bound to have difficulty in accepting that the switch in terminology is as I describe it, because they are unable to see how the idea of a form of life connects up with the notion of logical form.16 These critics will say that, if the conception of logical form has been replaced at all, it has been replaced by the notion of the description of the "depth grammar" of an expression - i.e. by descriptions of the rules which provided the framework for our representations of the world by means of language. But Wittgenstein cannot have intended the term 16 I am much indebted to P.M.S. Hacker for setting out his objections to my construal of Wittgenstein's development on this point (personal communication, January 3, 1994).
124
ANTHONY HOLIDAY
"depth grammar" to be the sole agent of the transition in his thinking about logic which he tried to express in the Investigations and elsewhere in his later writings, because if this term alone substitutes for the notion of logical form then it will misleadingly suggest precisely what is denied in the post-Tractarian philosophy viz. that logical form constitutes an occult substratum hidden in the invisible depths of language and requiring analysis to expose its nature. The terminological innovation by recourse to Spengler's notion has the advantages of offsetting the dangers inherent in talk of "depth-grammar" while injecting a culturo-historical dimension into the philosophy of the Investigations (since Spengler uses "forms of life" to refer to cultural phenomena) which had been absent from the Tractatus. The introduction of this dimension does indeed radically alter the approaches one will be disposed to adopt to questions in the philosophy of language and of logic, but it does nothing to reduce the status of logic to that of a profane set of opinions which might be caught within the net of some or other scientific or pseudoscientific theory. In fact such a reduction would be quite out of keeping with Spengler's attitude towards the topic, which is on all fours with the Tractarian view in holding that "Both logic and ethics alike are systems of absolute and eternal truths for the intellect..." 17 Secondly, Wittgenstein's refusal in the Investigations to treat the topic of names and naming in the theoretical mode characteristic of Russell's semantics and (arguably to some extent) of his own procedures in the Tractatus, coupled with his scrupulous effort to draw attention to the existence of language-games in which the meanings of names really are quite legitimately explained by pointing to their bearers, has the effect of reminding us of something highly significant about ourselves which is displayed in our capacity to use signs in this way. It is this: human persons alone are able to use gestures, sounds and symbols to fix references in a normative, and hence regulated and repeatable fashion. Animals, even trained hunting dogs, cannot do this. For what a pointer hound, say, does when his stance indicates the presence of prey in the thicket is to "point" to the animal, not to point to it. In order to describe the stance it adopts as fixing a reference we would need to remove the quotation marks from the expression for what the dog does and, in order to do that, we would need to fill in special features forming the backdrop to his behaviour, features which do not attach to the lives of real dogs, but only to those anthropomorphised imitations of them we encounter in animated cartoons and comic strips. While Wittgenstein is clear that what humans do in their referring activities is not to be thought of as an "occult process" in which a queer kind of connection is established between words and objects (PI § 38), he is equally adamant that the 17 O. Spengler, The Decline of the West, abridged ed. by Herbert Werner and Arthur Helps, transi. C.F. Atkinson, London: George Allen and Unwin, 1961, p. 260.
SCIENCE AND SIGNIFICANCE
125
linguistic practices of humankind - including, of course, the whole range of techniques we deploy in identifying persons and things - although they are certainly natural to persons - are not to be assimilated without remainder to the natural world in the sense in which that world includes the animal kingdom. Thus he insists that "If a lion could talk, we could not understand him" (PI II.xi: 223); that a dog cannot simulate pain (PI § 250) and that there are no more grounds for claiming that it means something by wagging its tail than for saying that by drooping its leaves a plant means that it needs water (Z § 521). The import of these observations manifests itself when one considers how human ostensive behaviours are expected to be intrinsically responsive to challenges requiring self-reflection in a way in which even neurologically highlydeveloped animals or very sophisticated machines are not. Priscilla may be asked whether she really meant to point out Peter rather than Paul, who stands adjacent to him at the identity parade. The wooden arrow by the roadside points in a southerly direction. But should it be read as indicating the path leading to the farmhouse or the track to the church? Might it not have been deflected by last night's gale so that it no longer expresses the intention of whoever erected it? The ostensive gesture which explains a meaning is pregnant with the possibility of a further explanation. Thus the philosophy of the Investigations and some of later texts proximate to it make explicit something about language and its limits which is already implicit in the Tractatus. The overall achievement of Wittgenstein's revolution within a revolution consists in redrawing those limits in a fashion which overturns the entire tradition of modern philosophy from Descartes to Hume and Kant and returns us to a pre-modern conception which informed the thought of Plato, Aristotle and Augustine. This it does by showing, not only that language is inseparable from the human agency which alone is capable of it, but also that it is the rationality of such agents - as evidenced inter alia by their verbal behaviour in assigning names to themselves and things together with the living cultural forms which accompany these activities - and not merely the sensate consciousness they share with animals, which is the test of what does and what does not constitute personal being. Since scientific procedures depend on these volitional and rational capacities it is obvious that science, no matter how general its scope or how sharp its analytical tools, is helpless to provide a foundation in the form of hypotheses and explanations for these defining capabilities. His heightened sensitivity to the importance of this constraint led Wittgenstein to give his anti-scientistic credo a more stringent formulation in the Investigations, forbidding himself to set forth "any kind of theory", to shun all hypotheses and to do away with all explanation, allowing description alone to take its place (PI § 109). Thus although, as Harris says, "philosophy" and "language" remain
126
ANTHONY HOLIDAY
"words like any others", the attempts to respect the limits of the latter issue in an entirely new way of doing the former and give a new meaning to that activity. While it may not, in Harris's phrase, be "self-evident" (whatever that means) that language and its limits underpin Wittgenstein's dichotomy between science and philosophy, I hope the preceding examination has already gone far enough to show that vigilant attention to those limits is an absolute prerequisite for anything like a proper appreciation of the Wittgensteinian achievement. As is often the case with would-be critics of great philosophy, Harris's dismissive rhetorical flourish conceals a blind spot which prevents him from treating the writer and texts he comments on with an appropriate degree of seriousness.
4.
Our exploration thus far has taught us something of what will be lost to sight or distorted when a commentator adopts a cavalier attitude to the way in which the phases of Wittgenstein's philosophical development relate to one another, ignores the saying/showing distinction and underplays the importance of the idea that there are absolute limits circumscribing language and the world, segregating sense from nonsense and illuminating that which they circumscribe in the light of a transcendental eternally present source of the world's meaning. The labour which remains in large part unaccomplished is that of showing why these weighty matters are shoved aside; we must go a further distance towards achieving this aim if we do not want to give the impression (which we certainly don't) that commentaries like Harris's are the result of culpable ignorance or mere stupidity. We can, I think, best deepen our insights into the sources of Harris's failure to give due weight to Wittgenstein's defence of his philosophical techniques by attending to the way that failure impoverishes the treatment Harris attempts of a theme in Part I of the Investigations which he rightly judges to be an important one -viz. Wittgenstein's exemplification and discussion of "primitive" languagegames, elementary systems of communication which fall short of being natural human languages. Wittgenstein proposes the following famous thoughtexperiment with a view to testing the viability of the Augustinian picture of how language functions: Let us imagine a language for which the description given by Augustine is right. The language is meant to serve for communication between a builder A and an assistant B. A is building with building stones: there are blocks, pillars, slabs and beams. Β has to pass the stones and that in the order in which A needs them. For
SCIENCE AND SIGNIFICANCE
127
this purpose they use a language consisting of the words "block", "pillar", "slab", "beam". A calls them out; - Β brings the stone which he has learnt to bring at such-and-such a call. - Conceive this as a complete primitive language. (PI § 2) Now, although this example is certainly part of a set of arguments designed to assist our understanding of how the Augustinian proto-theory gets its grip on the intellect, it is, fairly obviously, not intended as an element in some rival theory of the genesis of language. Yet Harris insists on situating it in a chain of putatively explanatory reasoning, designed not only to expose the fallacies of the Augustinian picture, but also to demonstrate how a communicative system might be devised and sustained. He begins, unobjectionably enough, by remarking that the scenario presses home an attack on the nomenclaturist assumption that every word must stand for something and that this "something" is the word's meaning. Harris rightly takes Wittgenstein to be making the point that what the builders say and do only fits the nomenclaturist case because their utterances and activities relate "to a communication situation which does not place any more complex demands on the language".18 But, having made this promising start, Harris then proceeds to problematise the example in ways which might have seemed licit had the builders' languagegame been intended as illustrative of a pattern of natural speech, instead of being, as it is, an exemplification of "primitive" semantic transactions (as I have called them elsewhere)19 which fall outside the scope of human conversation and which are not, therefore, enigmatic in the way that fully-fledged human language-games sometimes are when they become the focus of philosophical interest. He alleges that the language-game trope is a mechanism for soothing rather than resolving doubts about the workability of the system the builders employ. With a breathtaking disregard for either its context in the Investigations or the topics to which it contributes, Harris drags in the notion of "agreement" to which Wittgenstein draws attention - at a point well beyond his discussion of the Augustinian doctrine and exactly midway between the end of his discussion of rule-following and the start of the series of arguments often miscalled the privatelanguage argument. There Wittgenstein points out that:
18
Harris, op. cit., p. 102. A. Holiday, "Conversations in a Colony: Natural Language and Primitive Interchange", in Pretexts, vol. 4., no. 2, Summer 1993, pp. 3-19. 19
128
ANTHONY HOLIDAY
If language is to be a means of communication there must be agreement not only in definitions but also (queer as this may sound) in judgements. (PI § 242). According to Harris, the games analogy beguiles us into assuming that some such agreement must pre-exist the builders' success in communicating as they do, while we fail to notice that we have nowhere explicitly been told that the builder has actually taught his apprentice how to respond to the words the builder calls out to him. This supposed sleight of hand provokes Harris to object that "Nevertheless, there is a world of difference between producing the 'agreement' story as a description of what the builder and his assistant do, and producing the 'agreement' story as an explanation of what the builder and his assistant do".20 But this is to charge Wittgenstein with neglecting to do something he has explicitly set out not to do. As Harris knows very well, Wittgenstein is determined to eschew explanations of how the builders' game (or any other language-game, primitive or not, for that matter) is generated, invented or kept in play. Indeed, the builders' game forms part of an argumentational strategy, directed precisely at exposing the absurdity of genetic explanations in philosophical semantics and the futility of supposing that languages were erected systematically from elementary "building blocks" like the one under discussion. Explanations of a mechanistic, means-ends sort can certainly be given in respect of the builder and his apprentice's verbal antics but, since such explanations deal in external causal relationships only, they will be void of philosophical interest. What would serve the interests of a philosophical inquiry would be a piece of conceptual work which helped us to see what a human language-game was really like by drawing our attention to a game which failed to achieve that status, and that is what the builders' game is brought in to do. For their "game" is no more a game than the pointer-hound's posture was an instance of pointing. Their signals are not part of human language and neither their signs nor what is done in response to them count as actions. The builders are not agents, if by "agency" we mean a continuum linking desires, acts and reasons-for-acting. They are not men but manikins. Their system of communication could easily be duplicated by a pair of marionettes, driven by a mechanism, in which case what they "said" or "did" could be explained by explaining how the machine worked. But, as I have said, such an account would address none of the philosopher's concerns. It is noteworthy that Harris's misapprehensions concerning the builders are long-standing ones. In a book which appeared eight years before the publication of Language, Saussure and Wittgenstein he had this to say on the matter: 20
Harris, op. cit., p. 107.
SCIENCE AND SIGNIFICANCE
129
The moral of Wittgenstein's fable is the wrong moral, because his fable attempts an impossibility. The impossibility is to construct a model of what language would be like if it consisted solely of names of things; that is to say, of words which functioned intrinsically as names of things, as distinct from word which might be described, in a chosen metalanguage, as names of things.21 That Harris should have clung for so long to a misleading picture of what Wittgenstein is doing in this early section of the Investigations says something important about how powerful an influence such pictures may exert. But enough, I hope, has already been said to make it plain that Wittgenstein is not attempting the impossible feat Harris takes him to be attempting; that despite Wittgenstein's subsequent remark that one could even imagine that the language of A and Β was "the whole language of a tribe" (PI § 6), his use of the adjective "primitive" to describe A's and B's interactions is intended to convey his conviction that this "language" is not a natural language among other natural languages in the sense that English is one natural language and Zulu another; that the builders' "language" is an artifice, like Esperanto, not a live language, like Afrikaans, nor yet a dead one like Latin or Sanskrit, and that it is this artificiality and not merely the paucity of its verbal furniture which makes it unthinkable that the languagegame "played" by the builders could be internally connected to the modalities of genuine conversation. It is precisely in respect of this matter of internality that the well-springs of Harris's misreading of Wittgenstein's writings and method begin to emerge. Harris expects to be told the causes of the builders' behaviour and he also expects that this causal story can cogently be amplified to include the whole range of behaviours constituting natural language. His expectation is unreasonable because it demands that an account in terms of external relations should do what only a description of a network of internal relations could achieve, so the anticipation is doomed to disappointment from the start. His impatience with the Wittgensteinian dictum that explanations cannot do the work of conceptual descriptions is partner to his failure to take seriously the all-important distinction between giving reasons and stating causes. The failure is not uncommon. Some philosophers are prone to treat the former as a special case of the latter - a move which would be harmless if it did not so often conceal a mistake about the grammatical differences between reporting causal states or the action of causal powers and making statements of a reason-giving kind, an error which sees reasons as the mere "innards" of causes. 21
R. Harris , The Language-Makers, London: Duckworth, 1980, p. 87.
130
ANTHONY HOLIDAY
But to conflate reasons and causes in this way is to ignore the very different ways in which they are established. Typically, we try to ascertain the reason a statement is made or an action performed by asking the author of the statement or agent of the action why they said what they said or did as they did. We do not trace the workings of mechanisms (physical or psychical), do experiments or call for statistics, as we would do in trying to establish a relationship of cause and effect. Moreover, an internal or conceptual relation exhibits certainty in a way which a causal relation does not. If an agent gives a cogent reason for an action, or an argument is judged to be valid, we do not say that the reason is "very probably" linked to the action or that the argument "might" be valid. Our interest in reasons, as distinct from the kind of interest we have in causes, depends on this being so. In trying to discover a set of internal connections we do not try to meet such obstacles as may arise by trying to think up better experimental methods or by wondering whether our procedures might not be confounded. Rather we look more closely at the relata, scrutinise them more carefully until the relation we seek is seen certainly to be there. For, as far as internal relations go, all evidence is self-evidence which will show up if the bond we are looking for really holds. Moreover, as Paul Johnston notices,22 causal explanation conflicts with the notion of rationality, because rational action is characterised by its being determined by the merits of the situation rather than by the prior state of the agent. Notions of decision and choice which are central to our talk of human action cannot be captured by accounts in terms of causation. Johnston is correct also, it seems to me, to argue that our reasons, intentions, beliefs and so on are not related to each other and to our actions in terms of their causal efficacy, but in terms of their content, so that if someone offers a reason for doing one thing which could equally well have applied to another available course of action, they will usually be called on to point to a difference in the alternatives which led to one being favoured over the other. This reflection on the nature of rational agency as distinct from causally determined behaviour brings us to another layer of sources of the misreadings of Wittgenstein perpetrated by Harris and commentators who share his orientation. As should by now be apparent, Wittgenstein in his later work is trying to present a descriptive overview (Ubersicht) of language, a sketch-map, or series of sketch-maps, of our concepts and their mutual affinities and disaffinities. But his purpose in so doing remains, not analytical, but elucidatory. In the instance before us he essays to shed light on the religio-ethical significance of such things as the distinction between artificial semantic systems, like the system used by the 22
P. Johnston, Wittgenstein and Moral Philosophy, London and New York: Routledge, 1989, paperback ed., 1991, p. 42.
SCIENCE AND SIGNIFICANCE
131
builders, and genuinely human speech. In the case of the builders, the distinction in question indicates an important difference in the notions of subjectivity which accompany artificial signing systems on the one hand, and spontaneous human dialogue on the other. In the former instance, the interlocutors will be thought of as having minds, where the word "mind" probably displays affinities with the Cartesian res cogitans and is, in any case, a term of art in the epistemologist's amply stocked repository of such terms, and where the attribution of mental capacities takes the form of an hypothesis as to what sort of subjective structure best explains the behaviours under scrutiny. In the latter case, by contrast, nothing hypothetical can enter into the participants' attitudes towards one another in respect of the kind of subjectivity which makes their dialogue possible. For this is constitutive of their interaction and of their sense of what they have in common as speaking agents, beings ensouled by a will which is the ground of their activity and the source of the speech act they originate and share. For a participant in such a dialogue, the ensouled state of fellow participants cannot be a matter of opinion, since the assumption that it could become so is logically inseparable from the supposition that what is taking place is not a conversation but something entirely and unthinkably different and thus not something in which the participants could be taking part unless they were beings unthinkably different from the beings they know themselves to be. But, in order to see language and its users in the light of their ensoulment and to recognise, in the phenomena of understanding and meaning, something which is the unique and irreducible expression of human personhood, Harris and those who do philosophy as he does it would need, not to adopt some or other set of religious doctrines, but to survey this matter of language and agency from a religious point of view. Further, if I have not wholly misread his purposes, it is precisely in the interests of assisting his readers to adopt such a viewpoint that Wittgenstein devised and deployed the philosophical techniques he did in his later writings. But these techniques were inspired by his anti-theoretical precepts, his determination to preserve the dichotomy of philosophy and science, and it is just this stance which Harris and those philosophers he keeps company with find opaque. So it looks as though we are caught up in the toils of a circular argument. It is, I think, impossible to deny that a measure of circularity has crept into our proceedings. However, the consequences of this are less dire than might appear, at first sight, to be the case. We do not need to think of the stances Wittgenstein asks us to adopt or the conceptual survey-tours he wants us to undertake as demands that we enter unreservedly into a system of thought in order to reap the putative advantages of such a surrender, as a certain stratum of the psychoanalytic fraternity seem to require their clients to do. Rather, Wittgenstein invites
132
ANTHONY HOLIDAY
us to undertake a variety of thought experiments, to allow ourselves to be guided through multiple mazes of explorations such that the results are known, albeit very imperfectly, in advance, and the destinations dimly recognised from the start as places of revisitation. Thus the benefits of his therapies are not discoveries but rediscoveries, not new revelations but illuminations and assisted recollections. If, then, the exercise Wittgenstein suggests we undertake is a relatively innocuous one, why should so important a body of thinkers in the mainstream of the modern intellectual enterprise so fiercely resist it? The answer (or partial answer) I would advance is this: not only is it true that our addiction to general theories distracts us from that vision of the speaking being as an ensouled being, a vision which, as I have argued, is both the end and the impetus of Wittgenstein's method, but the addiction itself is fed by a tradition in occidental thought which runs directly counter to and is at least as powerful as the neoPlatonic tradition which inspired Wittgenstein. The former tradition stretches back at least as far as Aristotle. It is wedded to a pagan ethic of ennoblement through flourishing, to a metaphysic which assimilates reasons to causes and to an anthropology which makes no radical distinction between the personal and the biological. The way in which these various strands in Aristotelian naturalism combine to obliterate the insight into the nature of personhood which informs Wittgensteinian methodology comes out in this very matter of the conflation of reasons and causes we have been discussing and which I said was implicit in Harris's treatment of the lessons he seeks to draw from the language-game played by the builder and his assistant. Notice how one of the most eminent contemporary proponents of the thesis that reasons are a kind of cause, Donald Davidson, expresses himself when arguing for the view that what he calls the "primary reason" for an action is the cause of that action: In the light of a primary reason, an action is revealed as coherent with certain traits, long-or-short-termed, characteristic or not, of the agent, and the agent is shown in his role of Rational Animal.23 (emphasis added) Here one feels inclined to interject at once that of course if what we are talking about is a rational animal, if agents and their actions (including their actions in speaking, listening, reading and writing) are to be viewed solely under their aspect as the agency and behaviour of homo sapiens, then it will seem much more plausible to suppose that whatever it is that moves them, no matter how different it may be in some respects from the potentialities and powers which 23
D. Davidson, "Actions, Reasons and Causes", in D. Davidson, Essays on Actions and Events, Oxford: Oxford University Press, 1980, p. 8.
SCIENCE AND SIGNIFICANCE
133
move inanimate things, is really a special sort of cause. And because nobody wants to deny that what animals do and their modes of communicative interaction are explicable, and that the explanations we give of them will be in terms of causes, nobody wants to deny that, given the element of animality which attaches to human actions and utterances, these too are best explained causally. These concessions, once made, will set us on the high road towards a scientific approach to the philosophies of action and of language, bravely unaware that too much has been conceded; that an aspect of the human condition has been mistaken for the totality, a petitio principii perpetrated, and a spiritual reality altogether lost to sight. Insofar as he is consciously or unwittingly an ally of this dominant naturalism and scientism, Harris must have his share in its blindness to a spiritualised, and in the end holistic and unashamedly idealist, view of people and their speech - and it is this blindness, which is at the bottom of his inability to sympathise with Wittgenstein's ideal of philosophy as a descriptive activity.
5. We began by talking about joking and the way in which what strikes us as profound about a philosophical dilemma is very like what suggests depth in a grammatical joke. We noticed how, in both cases, there was a sense in which the idea of teaching people to see the point of these activities and enigmas - and hence the idea of there being a science of either of them - was incongruous. There was a certain obviousness about this which led us to wonder how anyone could fail to see it and, in failing to see it, neglect the seriousness of Wittgenstein's strictures against conflating the scientific and philosophical enterprises. Our quest for the sources of this insouciance led our discussion to meander through various aspects of the development of Wittgenstein's philosophy of language in general and his views on names and naming in particular. Thus we explored the Tractarian origins of the saying/showing distinction and its relation to Wittgenstein's adoption of Russell's theory of definite descriptions, went on to examine the critique of the doctrine that the meaning of a name is its bearer, set forth in the Investigations, and finally scrutinised the language-game played by the builder and his assistant with a view to eliciting the full force of the distinction between natural human language as it is spoken by persons and artificial systems of signs typical of machines or animals. This last consideration has brought us round full circle and back to the themes of humour, joking and seriousness. For it must surely be evident that, as Gaita
134
ANTHONY HOLIDAY
argues24 contra Malcolm,25 we cannot, without importing into the stage-setting of the builders' scenario features which would qualitatively transform the signals which make up their language-"game" into symbols forming part of authentic human speech-patterns, imagine them as responsive to calls to sobriety, injunctions to reflect on whether they mean to say what they say and to do as they do. If the builders are incapable of responding to reminders of this sort, then they are incapable of telling sense from nonsense, cogency from incongruity and, thus, incapable of humour, of being struck by the absurdity of, say, calling for a slab when a beam is what is needed or even by the oddness of a working life which is as linguistically impoverished as their life is. They are a little like the clowns in Samuel Becket's Waitingfor Godot, with the important difference that their activities can never seem clownish to them. If we want to imagine the builders as capable of humour, then we have to be able to think of what they are as a human, personalised way of living in the world, and in order to do that we have to be able to think of their humanity as internally related to their speech. There is intellectual nutriment for us in considering what we have in the form of a capacity for humour by contrast with what the builders lack in this regard, something good for us in being made to remember what the uneven terrain of our natural modes of self-expression offers us in the form of an openness to selfirony and an ability to find happiness in our propensity to mismatch what we say and do to what we mean to do and say. It is our gift for humour which allows us to discriminate between degrees of seriousness attaching to the language-games we engage in and create with and for ourselves and with one another, and it is this same ability which enables us to set standards of appropriateness and inappropriateness regarding how seriously we ought to take ourselves, other people and the games people play. Because we can see the point of a joke, we ought to be able to see the pointlessness of treating scientific procedures and paradigms with undue seriousness. To say this is not to say we should not treat them seriously at all, only that it is ludicrous to try to import them into all significant regions of our lives, to indulge ourselves in the idolatries of a Comtean religion and to anticipate that the fruits of science can substitute for what is sacred and mysterious to us and will always remain so. It was Wittgenstein's aim in setting forth his philosophical method and in trying against so many odds to remain faithful to it, to restore us to an awareness of these things, to bring 24
R. Gaita, "Language and Conversation: Wittgenstein's Builders", in A. Phillips Griffiths (éd.), Wittgenstein: Centenary Essays, Cambridge: Cambridge University Press, 1991, pp. 101115. 25 N. Malcolm, "Language Game (2)", in D. Phillips and^P. Winch (eds.), Wittgenstein: Attention to Particulars, Essays in Honour of Rush Rhees, London: Macmillan, 1989, pp. 4243.
SCIENCE AND SIGNIFICANCE
135
through the intellectual fog which infected his scientistic age as it infects ours some particles of the spiritual light which comes with self-knowledge.
7 Rules and Algorithms: Wittgenstein on Language Rom Harré
A glimpse of
biography
Nearly forty years ago Roy and I used to give a class on reasoning, or something like that, at the old Edmund Street site of Birmingham University. My job was to trot out the elements of deductive logic, and then to display their limitations as a model for the way scientists thought. Roy was doing much the same for grammar as a formal model of the normative constraints on how people made sense to one another. I was just beginning to get a fondness for the richly multivocal concept of "rule", having picked up, I thought, a little late Wittgenstein. I had much to learn. In this paper I want to set out what I take to be some leading Wittgensteinian insights, and here and there defend them against some critical remarks of Roy's, some of which derive, dare I say it, from a too hasty reading of some pretty dense paragraphs. Older and perhaps a little wiser, we have both kept faith with at least the spirit of our youthful intuitions. Roy has persistently rubbed all our noses in a cluster of problems that anyone must cope with who wants to hold on to the notion of "rule" as an expression of the way that standards of correctness are maintained in human affairs. His intuitions about these problems were already immanent in his contributions to our class. If language is continuously created and recreated in its daily uses, and if in the course of these re-creations the uses of words change along various dimensions, isn't "rule" just the wrong notion to capture whatever is consensual and normative (and these are not the same thing) in our symbolusing practices? (I had originally written "communicative" here, but Roy has been severe, and rightly so, on simplistic uses of this concept.) He has asked whether conversation is really all that similar to chess or tennis. Or to take a more Harrisian image, cricket. As the joint promoter of the eighteenth-century cricket matches we used to play in Leicester in the late fifties, Roy knows that one can use rules to create and recreate a practice. Of course cricket has changed, but its changes have been quantum leaps, constrained by discontinuous sets of rules. Innovations such as body-line bowling are ruled out! But innovations like
RULES AND ALGORITHMS
137
hugging a successful bowler, as if he had just scored a goal for United, just come about. Patently the Académie Française is not as potent a force in the realm of language use as the Imperial Cricket Conference is in the realm of sport. But is this a difference in kind? This is the question I want to explore. To do so I shall make some corrections to Roy's presentations of Wittgenstein's views, as he expounded them in his book on Saussure and Wittgenstein, and then explore the notion of "rule" as I see it in use in linguistic and social-psychological discourse. But first I shall add some further support to Roy's trenchant criticisms of formalist linguistics, of grammar as algorithm.
Critique of the formalist
tendency
Descriptive and auxiliary formalisms In the above phrase I mean "critique" in the sense of Immanuel Kant and not of American undergraduates, who persistently confuse a critique with a criticism. A critique is an examination of a practice, for example, to display its limits, that is, to show where it is proper and where improper. Who would deny that formal linguistics, expressing grammar as algebra, can be fun! My former pupil Justin Leiber and I "cooked" up a transformational algorithm for working out proper menus for the Burgundian cuisine, as defined by Brillat-Savarin (Harré 1993: 88). We would certainly not wish to claim that this little exercise displayed the structure of the knowledge shared by Brillat-Savarin and M. Raymond Blanc. The mere fact that one can produce a formal system that mimics the structures produced by some "natural" process does not show that the formal system is interpretable as a model of that process. There are two distinctions needed to keep our feet on this slippery slope. One is between the means by which we express what we have found out about something, and what that something is; and the other is between a convenient device for making inferences and predictions about some process and a genuine explanation of how what one has pre- or retrodicted comes about. Rather than subject the Chomsky ans and their algebraic algorithms to yet another hammering I shall make these points in an adjacent field, the philosophy of physics. There we shall see the distinctions I wish to make very clearly and the problems that arise by ignoring them. Both the above points can be illustrated with the distinction between descriptive and auxiliary mathematics. I owe this terminology to John Roche (personal communication). Let us compare the formula for calculating how far a body in uniform motion in a straight line will have travelled so many seconds after it has been released, with the formula for
138
ROM HARRE
calculating where a body in simple harmonic motion will be so many seconds after the motion has begun. In the first case the formula is s = vt
I
Each term in this expression has an interpretation in the actual process being studied. "s" is the distance travelled, "v" the linear velocity and "t" the time taken. We could be slightly more sophisticated and link these variables to the actual process via the correspondence between idealised sets of numbers each variable expresses and the numbers produced by measurement. Each variable in the formula corresponds to a measuring process. We can call the formula (I) descriptive mathematics. The formula for the velocity "v" of a body in simple harmonic motion of amplitude "a" at some point "y" in its oscillation is ν = +/- w a2 - y2
Π
where "w" is a constant. Only some of the terms in this formula correspond to real aspects of the reciprocal process. There is no physical process that corresponds to "w". It can be pictured as the uniform tangential velocity of a point in an imaginary circle the motion of which is projected on to the diameter of that circle, as the amplitude of the reciprocal motion. Devices like "w" and the imaginary circle Roche calls auxiliary mathematics. When working with classical mechanics physicists seem to know very well which variables are descriptive and which auxiliary. For example, though the expression for kinetic energy(1/2mv2) is a descriptive expression, the expression for potential energy in a gravitational field, "mgh", where "h" is the vertical displacement of a body above an arbitrary bench mark, is auxiliary. But in some branches of physics, though the distinction is robust it is not always uncontroversial where it is to be drawn. The Hilbert space technique for expressing the formalism of quantum mechanics is based on the rotation of a vector in an infinite dimensional "space". Though some have struggled to conceive of this formulation as descriptive, that is, that the vector should have a physical meaning, others, and I count myself among them, see this mathematical system as ingenious, useful but auxiliary. What has been found out about the behaviour of subatomic particles and the radiation associated with them can be expressed very nicely in the formalism of Hilbert "space". Predictions of probability distributions of different classes of experimental results can be predicted with astonishing precision. But none of that gives us good grounds for
RULES AND ALGORITHMS
139
believing that Hilbert "space" is a real something behind the space and time within which the experimental apparatus of physics is set up. The role of ontologies But these comments merely describe the situation. They do not diagnose it. What is it that gives physicists such confidence in some cases and permits such deep seated divisions of opinion in others? The answer is the implicit working ontology with which interpretations are controlled. Working ontologies are nicely expressed as type hierarchies, with generic ontological prescriptions of what there is in a domain occupying the role of supertype under which a branching layout of subtypes and subsubtypes can be arranged. There are various "logics" for construing type hierarchies, for example there is the genus/species relation that orders Linnaean phylogenies (cats are feline) and the determinable/ determinate relation that orders hierarchies of properties (reds are colours). A third hierarchical ordering relation that is important in physics is the whole/part relation that orders hierarchies of substances (subatomic particles are parts of atoms are parts of molecules are parts of observable things). Whether a formal system which has been shown to be adequate to the expression of at least some of the discourse about the beings in the same domain is descriptive or merely auxiliary will depend upon the ontology which is brought to bear upon the task of interpreting it. If we are to take Chomskyan linguistics seriously as a theory of language we should be obliged to take seriously an ontology of word classes. It is to this that I think Roy (and Wittgenstein) are deeply opposed. The notion of a "word class" is a notion appropriate to the task of formal linguistics as an auxiliary "mathematics". But since there are no word classes in Roy's (and Wittgenstein's) linguistic ontologies, formal linguistics is not a descriptive mathematics, that is, it is not a theory of language in the sense, say, that genetics is a theory of heredity. Mendel's algebra, coupled with the molecular ontology of Watson and Crick, is interpretable as a descriptive mathematics of the recombination of genetic material and its role in the genesis of adult characteristics, because there are genes. Tools and rules But now we come to the hard part. If it is not word classes that are ontologically fundamental, what is? Another answer might be: the word as a lexical item. And we would then use a type/token distinction to help us arrive at an account of language as a Bloomfieldian domain of sentential objects. This would tie back on to Saussure's langue/parole distinction were we to hypostatise the types into an
140
ROM HARRE
ontology. Langue is the collection of linguistic supertypes, and the huge variety of paroles consists at various possible levels of abstraction from concrete speakings upward, of subtypes, and hierarchies of subtypes. At this point we run up against the Philosophical Investigations, and in particular the tool-box image. If that were the only illustrative image we might think that Wittgenstein's tool-kits were Saussure's langues. But this would be a mistake. There is another Wittgenstein image of equal potency, the handles in the locomotive cab, that look alike but work in different ways. This can be construed as an attack on the utility of grammatical word classes in understanding how language works. It is just exactly the assumption that nouns all work the same way - an assumption not exactly entailed by but implicit in transformational grammar and barely escaped by formal semantics - that in Wittgenstein's view is the source of some of the illusions of the intellect that constitute most of philosophy. But tools have uses, and it is their uses that make them the tools they are. If we take the two images together we reach the famous advice that in trying to express meaning we should describe use. ("Meanings" disappear along with other non-entities.) Tools can be used correctly (skilfully) or incorrectly (unskilfully). To express criteria of correctness we could use the concept of a rule, and as a metaphor for "A knowing how to use X properly", we could say "A knows the rules for using X". Does this move fill the mind of the competent carpenter, speaker of Inuit, jumbo pilot, accountant and so on with rules? Boundaries and "agreements" In discussing Wittgenstein's use of the (generalised) concept of "grammar" (Harris 1988a: 70-90) Roy seems to misunderstand the thrust of Wittgenstein's arguments and illustrations in a way that leads him to miss some crucial things about the way the notion of "rule" is invoked by Wittgenstein. There are both immanent and transcendent rules. There are normative regularities discernible in a practice, which could be expressed by some onlooker as a system of rules. And there are rules which exist prior to their use in applications in guiding a practice. Wittgenstein even imagines a set of rules which, had they ever been implemented, would have been realised in a game, but they never are. Even so, they are rules. On the other hand there are ways of acting that are subject to standards of correctness and incorrectness for which no propositional formulation of their norms had been attempted until they had been in play for a longtime as Wittgenstein says. But when the game analogy is worked out for language uses, and we record both transcendent and immanent rule systems, the special role of rules in
RULES AND ALGORITHMS
141
expressing the bounds of sense is brought out. Most of this is best read in On Certainty, but it is present in Wittgenstein's thought from the beginning. Grammatical propositions, those which express the bounds of sense, are such that while their affirmations are true, their negations are not false but senseless, that is, have no application. So "Nothing can be red and green all over at once" is true and looks like a report on how the world is. When we consider its negation "Something is red and green all over at once" we cannot think how to apply it. If that is the case then the former proposition was not a report of how things are. In discussing this and other related points, summed up in the thesis of the autonomy of grammar, Roy seems to have missed the key step. In PI 242 Wittgenstein says that we must agree in a form of life, a grammar, but ("queer as it may sound") we must also agree in judgements. Roy finds problems with both "agreements". I can reassure him on the former. In Harris (1988a: 107) he suggests that agreement in grammar might be serving as an explanation of the possibility of a practice. But that would be a very unWittgensteinian thought, tantamount to theorising about linguistic practice. "Agreement" just means "having been subjected to a common regime of training". Roy's gloss is too intellectualistic. Similarly, in his remarks about Wittgenstein's seeming commitment to a "communication" account of the role of language in life, he seems to miss the anti-mentalist thrust of the Investigations. "Communication" is not "transferring thoughts or wishes", which when moved across into the mind of the shopkeeper in language Game No. 1 bring it about that five red apples are handed over. Rather it is getting something done that we want done and getting it done by the use of language. As Wittgenstein says it could be just getting the other people to behave in a suitable way. Now we come to the other context in which agreements are needed for a language to be viable. We must agree in judgements, that is that the grammar of "eggs", "baking powder", "sugar", "flour" and "shortening" should eventuate, when used in the kitchen, in a cake we agree is just the thing for tea. A "grammar" of baking that led to radical disagreements in the kitchen would be in need of renovation. We can now rebut the critical thrust of Roy's remark (Harris 1988a: 83) that "Wittgenstein never tells us where grammar comes from." But he does. It comes from training in a practice (immanent grammar, capable of being expressed propositionally though not serving a normative role as propositional edicts). And there are games which have just grown up and no one has formulated the rules, yet or ever, though most people know how to play. But "grammar" may also come from someone sitting down and working out a set or system of rules, say for Monopoly, or the differential calculus, and then people go on to carry out these practices by attending to the rules and obeying them. This sort of grammar
142
ROM HARRE
is "transcendent". Neither would be possible if there were no natural regularities in the human form of life.
Wittgenstein and language
change
Breaking with the Tractatus Roy has sometimes been known to declare that Wittgenstein did not leave room in his treatment of language for the instability of linguistic practice and the fact of language change. Of course if our only document from the Wittgenstein Nachlass had been the Tractatus Logico-Philosophicus that conclusion would be inescapable. In that work the members of a repertoire of elementary names stand in one-to-one correspondence with elementary objects, the latter being (fixing) the meaning of the former. In so far as the world is stable so the language which depends on it is unchanging. Concatenations of names, atomic propositions, mirror the structure of atomic facts. No one could have been harder on this doctrine than Ludwig himself, and particularly on the various corollaries of it. Let us begin by looking at his later remarks on criteria and symptoms. Surprisingly Roy (Harris 1988a) dismisses this distinction and its subtle applications. Criteria are those conditions which are used to fix the meaning of a word, in some privileged language game, and symptoms are the consequential occasional conditions under which we use the word. A criterion for "red" would be a sample, and a symptom for "red" would be the colour of the Labour Party's rose. Someone who overlooked the lability of language would surely keep these strictly segregated. But in PI he says "It may be practical to define a word by taking one phenomenon as the defining criterion but we shall easily be persuaded to define the word by means of what, according to our first use, was a symptom". "The fluctuation in grammar between criteria and symptoms makes it look as if there were nothing at all but symptoms." This might tempt us into forgetting that there are norms. I think that in some moods Roy does indeed suppose that there is nothing at all but symptoms. Now we have to be careful about the "rule" metaphor - "For not only do we not think of the rules of usage - of definitions etc. - while using language, but when we are asked to give such rules, in most cases we aren't able to do so." Of course what this shows is not that language is normless, but that it is a practice, not a carrying out of calculations according to algorithms (PI 354-61). The key notion of "language game" also opens up other dimensions of linguistic change. As Wittgenstein remarks, language games are dropped, "cease
RULES AND ALGORITHMS
143
to be played", and others are invented. In the well known image of the City of Language, he contrasts the crooked streets of the old city with the rectangular grid of the new suburbs. But it is not just adding new language games to a conserved core, some of the core will cease to be played. Though our repertoire of language games must, so to say, respect the human form of life, what is needful for people to be able to manage as a natural species, everything is in flux, to different degrees. In another of his telling images Wittgenstein suggests we look on the framing of our language games as a river bed which is partly sand that changes easily, and partly rock that persists over long stretches of time (Conway 1989). The notion of "form of life" as used sporadically by Wittgenstein to try to express the enormously variable lability of language games must yet cohere in practices that make life possible for us, human beings - a sort of anthropic principle. One of the sources of misunderstanding of Wittgenstein's conception of the normativity of language, though I am sure it is not one that Roy is unaware of, is the misleading translation of Sprachspiel as "language game". The point comes out well when Wittgenstein, in discussing how it is possible to think about the future, remarks (in the standard translation) that "it is in language that expectation and its fulfilment make contact". But the original passage uses "Sprache", 'speech' and that is important. It is in the occasioned use of a form of words that the type of event that would fulfil the expectation is laid down. So an expression of an expectation is like a rule in that it specifies a kind and not a particular, while it is unlike a rule in that it is fulfilled only once. Nothing coerces us Wittgenstein emphasises again and again the difference between someone following a rule and someone behaving as an automaton. But the difference is subtle. In PI 218 he says "Whence comes the idea that the beginning of a series is a visible section of rails invisibly laid to infinity? Well, we might imagine rails instead of a rule. And infinitely long rails correspond to the unlimited application of the rule." But when we do what the rule says how do we do it? By interpreting the rule for this case, or by choosing to do this rather than that? Neither. If there were always an interpretation between the rule and its being applied then we could always escape an accusation of doing something wrongly by adjusting the interpretation. But that would mean that the notions of correct and incorrect would lose their point. Again if the rule simply advised us which alternative to choose, it might look as if we had a choice. But if we want to do what is correct we do not have a choice. But are we then coerced, made to so proceed? Is the rule a kind of persistent cause? Like rails? Again no, since the mathematics pupil
144
ROMHARRE
does get the series progression wrong. That is the clue. When we adopt a rule we take on a certain set of standards of correctness, by which we can then judge whatever it is we did. The rule does not belong among the causes of what we do. Once I take some edict as my rule of action then there is no room for further justifications. This is what Wittgenstein means by "I obey the rule blindly". Not because it is an edict, but because it sets the standards of correctness. Because much of human life, and especially the way people use words and other signs, is normative not causal, a matter of convention and not a matter of coercion, we can set the conventions freely, within certain constraints set by the material character of human organisms and their material environment. In PI 86 the image of alternative mappings between words and colour patches illustrates the point about conventions, while the story of the woodyard which tries to do business by selling wood by the base area of the stack rather than by its volume illustrates the point about the constraints that the material environment exercises on viable language games. In these and other ways the Wittgenstein of the Philosophical Investigations explores the possibilities of language change in several dimensions. But language is a normative practice: it is subject to local standards of correctness and incorrectness, and though these mark only fuzzy boundaries, language use must lie somewhere with bounds. Tropes, metaphors and all sorts of varieties of language use, some of which Wittgenstein lists early in the Investigations, need a background of implicit norms to have the force they do. Waismann added the notion of "open texture" to our armoury of Wittgensteinesque terminology, to express the idea that Wittgenstein himself expresses with the rejection of the image of the tramlines laid into the future. Nothing decided now, no contemporary practice however well entrenched, is able to fix the use of words in the future. The most obvious means for expressing the labile norms of a practice is the word "rule". It is easy to misunderstand a remark of Wittgenstein's to the effect that "if we look at the actual use of words what we see is something constantly fluctuating. In our investigations (my emphasis) we set over against this fluctuation something more fixed [as methodological device], just as one paints a stationary picture of the constantly fluctuating landscape". Roy seems to suggest (Harris 1988a: 90) that Wittgenstein means that there is "something more fixed" behind the fluctuations. But that is not the point of the remark; rather that to discuss a language-game we must take a snapshot. But that snapshot shot is not an X-ray of a bony skeleton.
RULES AND ALGORITHMS Rules and the interpretation
of
145
Wittgenstein
Rides and practices Is Roy hard or soft on rules? Scouring his works for an answer will also help us adjudicate his "debate" with Wittgenstein. Roy seems to reject the notion of "rule" as a main device for use by linguists to catch the normativity of language use. Sometimes in both The Language-Maker s and Language, Saussure and Wittgenstein he seems so intent on pointing out disanalogies between gameplaying and speaking and the use of the "game-derived" concept of "rule" for expressing the normativity of a practice, as almost to be inclined to drop the "rule-following" tropes altogether. But we cannot make sense of any human practice without the distinction between correct and incorrect performance, proper and improper actions, well and ill done projects and so on. We all agree that all human practices are normative, from drinking and cooking to calligraphy and doing mathematics, and even to persecuting people. If this is so the notion of "rule" surely must have a central place, somehow. What is the general problem that some people seem to have with the notion of rule? As I have suggested it seems to be that the notion of "rule" is used for two very different explanatory roles. There are rules "transcendent to a practice". These come in several subspecies. There are explicitly formulated maxims or edicts, rules as cognitive guides to correct performance, for example a manual or set of instructions, e.g. for assembling a model plane, or performing a ceremony, etc. There are various necessary psychological conditions for these to work, such as attending to the rule as written or otherwise displayed, understanding it, making the correct assumptions about how it is to be applied, and so on. Then there are the rules which are used to teach someone some way of performing correctly, which will not be accessed explicitly once the practice has become habitual, such as playing the musical scales. The notion of "rule" is also a device used by students of human performances to express the normative constraints that seem to be limiting the actions proper to some practice. None of the above psychological conditions are necessary for skilled or competent performers to manage what they are doing. The point of using the notion of "rule" to write down what we think expresses the normativity of the practice is only that we are sure that we have been investigating a practice and not a causal process, that is, the subject of our enquiry is an activity to which criteria of correctness and incorrectness etc. can be and usually are routinely applied. How can we write down the implicit norms? By writing a set of rules. These have the status of hypotheses as to the norms immanent in the practice, unlike transcendent rules, and can be corrected a posteriori. Typical cases for the
146
ROM HARRÉ
use of "immanent rules" are the skilled performances, by highly trained performers, such as cricket players, stage actors etc., or exercises of "native" wit by idiots savants. Naive calculators could not be using rules in the sense in which the Deans of Degrees do when performing a degree-giving at Oxford, and unlike the skilled cricket players could never have used transcendent rules in their apprenticeship. I believe that we should assimilate language use to (b), so that grammars are sets of rules of the second type. But where do linguistic norms fit into Roy's way of thinking? For the ephemeralist of The Language-Makers the very idea seems problematic. There are stabilities and there is moment-by-moment change. The original Chomsky an gloss on the notion of grammatical rule was correct, namely that it is not a psychological concept but a way of expressing what a competent speaker must know. "Such a syntactical theory is concerned precisely with the rules that specify deep structures and relate them to surface structures and with the rules of semantics and phonological interpretation that apply to deep and surface structure respectively" (Chomsky 1966: 38). This may sound like a psychological theory, and indeed many have taken it as such. Suppose we stick to the "auxiliary formalism" interpretation. This interpretation raises the question of how we are to understand the possession of skills as knowledge. Grounding skills "Knowledge" is something someone has. A skill is a grounded disposition: to say that someone has a certain skill, say to saw a straight line, is to say that if set the task he or she will perform it well or correctly. It is a commonplace that to account for the fact that we can say that someone has a skill when they are not displaying it, there must be some persisting state of the person which grounds the skill. For manual skills the persisting state is some structural property of the nervous system (Luria argued that this was also true of cognitive skills). I claim that language-using and other cognitive skills are also grounded in states of the nervous system, not in hidden cognitive conditions. The investigation of mechanisms of "natural language processing" is not a task for AI enthusiasts or cognitive scientists but for neurophysiologists. All that the former can do is to invent models of the normative systems which people use and conform to. These are not the foundations for explanations either of the power of the norms in everyday life, nor of their modes of implementation in skilled action. This raises the question of what the sets of rules, through which the normativity of such a practice as speaking or writing one's mother tongue is expressed, should be taken to represent. It can only be the causal processes in a
RULES AND ALGORITHMS
147
neurophysiological mechanism. But what has happened to the norms? They persist in the fact that it is a person who is using the mechanism of the brain (the mechanical properties of their tennis racquet) for some cognitive or communicative or entertaining task. The norms reside in the reactions of oneself and others to one's performances. That is how norms can change. If the norms were identical with their individual representations as concrete realisations of rule systems, then there would be rule conformity or chaos and never norm change. Even in practices with transcendent norms, such as tennis, there is room for creative variation in the implementation of the practice, and for rule change, though not as the explicit recognition of a local shift in what is accepted as proper practice, the case with immanent norms.
The conduit
metaphor
Saying as expressing We can only applaud Roy's determination to root out the picture of linguistic activity as communication, that is as a kind of conduit along which meaningless "marks" pass to and fro, encoded by the speaker and decoded by the interlocutor. Psychology, alas, always the last branch of the sciences to pick up the latest insight at the cutting edge, is still infested by the conduit picture. But it is worth remarking that the latter part of Wittgenstein's Private Language Argument is also fatal to the conduit image. In the latter part of the argument Wittgenstein tests out the widespread way of thinking of "the subjectivity" of each person as a swarm of mental objects by comparing the way criteria of identity and individuation work for those beings which are paradigmatically objects, namely material things, and putative mental objects, such as thoughts and feelings. In PI 293 the image of each person having ("being") a box in which each one has a beetle (a private experiential object) is tested out by showing that the word "beetle" would be usable in the language of this group even if each had a different object or none at all. The object drops out. Again the interlocutor says to Ludwig, "so you are saying that sensations are nothing?" To which Ludwig replies "They are not nothings nor are they somethings either!" If there are no mental objects whose existence is known only by inference from the effect they have on outward behaviour, how could subjective and private experience become a part of the public domain? The answer is the holistic expression account of how, for example, laughing and feeling happy, groaning and suffering pain, are related. To be happy is, inter alia, to be disposed to laugh and smile (Winnie-the-Pooh for instance). Someone who was disposed to frown and weep, and make gloomy
148
ROM HARRÉ
prognostications (Eeyore, for instance) could not be said to be happy. This would not be because we would inductively infer that Eyore's inner realm was occupied by sad thoughts and feelings (and paranoid ones as well) but because how it is with someone is not defined by their possession of certain kinds of inner object, but holistically; what it feels like and what we are disposed to do by way of expressing those feelings is one way of being.
The last remnants of
surrogationalism
A surrogationist, such as John Locke, is one who thinks of words as surrogates for and in close correspondence with states of affairs in a non-linguistic world. Roy has always taken a strong stand against this account of language (Harris 1980: 84-5). But in his discussion of Wittgenstein's block-slab game he seems to me to miss the point of Wittgenstein's example. It is not I think an example of a surrogational language, "block" meaning block, "slab" meaning slab and so on, which Wittgenstein is using to explore the limits of Augustine's apparent surrogationalism. Its point I take to be this: it is possible, unless we take "use-ina-context" account of meaning an Austinian or instrumental language use can look very like an Augustinian or surrogationalist one. Though the four words in the builder's language look like nouns (the handles in the locomotive cab all look alike) this language (and this is all there is to it) is not a language of things, but of commands and orders. The difference in "block" as a noun and "block!" as a command is not expressible in words, since in the builder's language there are no more words. The example also highlights the role of training in the uses of a language in which there is no set of words which could be used for formulating rules. So if the only words are "block", "slab", "pillar" and "beam" then the use of these words is to command the assistant to bring a certain piece of bricolage. Is there an ur-language? This question has been raised in a notoriously fallacious argument first promulgated by Fodor (1979). According to Fodor to learn a word is to test a hypothesis as to what it means. Hypotheses are propositional and so must be couched in some language or other. To learn the first words of one's mother tongue one must be able to formulate hypotheses about their meaning, and this must be in a language other than one's mother tongue, an inherited ur-language. The falsity of the major premise is evident. Wittgenstein's remarks on how language is acquired and what it is to have command of something linguistic, available long before Fodor's "reasoning", are supported by masses of empirical
RULES AND ALGORITHMS
149
studies, from the work of Vygotsky (1963) to the present day. Learning a language is acquiring a certain range of skills, according to Wittgenstein. It is a kind of training, in just the way that learning to play an instrument or to ride a bike is a kind of training. Certain biological prerequisites must be in place for the training regime to "take" but there is nothing propositional in the business of such learning. It may occur wholly as a manual apprenticeship. Furthermore, and again according to Wittgenstein, the ability to use a public vocabulary for feelings and the like, experienced only by the speaker, depends on the inborn ethology of a person, qua human being. There are natural expressions of feeling, which, with the feeling, form a whole. By substitution of verbal expression for natural expression a linguistic way of expressing one's bodily feelings, one's sense of self and so on is acquired. Again we need no assumptions about the testing of hypotheses, just the idea of training in a skill. But that is not the only fallacious move in Fodor's development of his account of language. There is the complementary step to "mentalese". This too comes from a fairly simple mistake. A computer operator inputs a word by pressing the keys that correspond to its written form. This flows into the machine as a bit string, a series of pulses that set a register (which is a set of switches) in a certain pattern of "offs" and "ons". But this may not be the appropriate form in which the causal processes in the machine transform and shift about these patterns of switches in the registers of the machine. So there is a compiler which transforms the input pattern into a pattern appropriate to a certain machine. The "rules" for operating the new patterns, which are of course no more than causal processes, are sometimes and very misleadingly called a "machine language". Now draw an analogy between machine computing and a brain thinking and we have just the opening for mentalese as the brain's machine language. Once one runs through the above description of what computing machines do, the grossly metaphorical character of the use of the word "language" to describe the inner workings of the machine is apparent. Of course if all that Fodor had meant by "mentalese" had been the electrical processes in the machine (electrochemical processes in networks of brain cells) one might have it a bit poetic, but not a serious mistake. But when combined with the other argument, that there must be an ur-language, we get the seriously misleading claim that the ur-language is mentalese. It would be inborn in the way that the machine "language" is inbuilt into a computing machine. By the same token there would be a gearbox language in my car since there is an inbuilt mechanical system by which the gear ratios are matched to the power output and speed of the car! What is the cure for this kind of extravagance? Surely the Wittgensteinian insight about fields of family resemblances in the use of a great many important words. Just because the same vocable is used in a great variety of ways, ways which have different patterns of
150
ROM HARRE
similarities and differences between their uses, there is no ground in that fact alone for supposing that there is some common essence in all the things for which the word is used, be it "game" (Wittgenstein), "plate" (Rundle 1990) or "language" (Fodor).
8 Contextualizing "Context": From Malinowski to Machine Translation Naomi S. Baron
1.
Introduction
In March of 1992, a shell diver named Kuzuto Harada was harvesting abalone on the bottom of Japan's Inland Sea. Suddenly, crew members in the boat waiting above heard an anguished cry from the microphone in Harada's diving helmet: "Somebody help me! Shark! Shark!" While shark attacks are rare in these waters, they can be deadly. When the crew pulled up Harada's lifeline, all they found was his orange and gray diving suit, with an enormous hole on one side. Harada's fate was clear. But the imminent threat for other divers was not. Were they looking for one shark? For more? They didn't know. For when Harada shouted "Shark!" {same in Japanese), he was not specifying whether he meant one shark or several. In Japanese, nouns are not marked as singular or plural. When a specific count is needed, speakers of Japanese follow the noun with a number plus classifier (e.g., jisho san-satsu = "dictionary three boundobjects"), but much of the time the Japanese rely, instead, upon context to disambiguate possible confusions. Unfortunately, Harada and his support crew had no time. How do language users figure out what words and sentences mean? We use four basic strategies, either individually or in combination. First, we rely on the shared meanings that linguistic elements are generally taken to have by members of the language community. When I go to a bakery and ask for whole-wheat bread, the person waiting on me understands what I am expecting because we both speak the English language. When Harada shouted "Shark!", his crew knew the sort of creature that was threatening him. Second, we rely on grammar to tell us what kind of meaning relationships are being expressed in a word, phrase, or sentence. In Indo-European languages,
152
NAOMI S. BARON
these relationships are typically marked by grammatical inflection, through word order conventions, or both. The -s on books indicates pluralization, and the "subject-verb-object" order in English tells us who initiated the communication in "Michelle phoned Max." A third source of meaning interpretation comes from looking at language that precedes the utterance in question. If I say "The blue one." you can hardly be expected to figure out what I am talking about - unless you know that my comment is responding to a question such as "Which ball do you want: the red one or the blue one?" The fourth avenue for decoding meaning is to look at the extrahnguistic context in which the linguistic form in question is embedded. If I say to the bakery clerk "I'll take that kind." the referent of "that kind" is only clear from the non-linguistic context. In Harada's case, the extrahnguistic context was the physical presence of one - or more - sharks. Within this paper, whenever we speak of "context", we will specifically mean "extrahnguistic context". Interest in the role of context in determining linguistic meaning has been growing in a number of quarters over the past two decades. Linguists (e.g. Germain 1979), anthropologists of various ilk (including ethnographers of speaking, ethnomethodologists, and students of conversational analysis - see, for example, Duranti and Goodwin 1992; Auer and di Luzio 1992), philosophers of language (e.g. Barwise and Perry 1983), and computer scientists interested in machine translation of natural language (see Kay et al. 1994) have all been exploring the issue. The foundational framework upon which many of these contemporary investigations rely was developed in the 1920s and 1930s by the social anthropologist Bronislaw Malinowski. Malinowski's analysis of the role of extrahnguistic context in determining linguistic meaning is grounded in two premises about human language. The first is that the purpose of language is not to reflect thought, but to act. As Malinowski writes in Coral Gardens and Their Magic. All our considerations have led us to the conclusion that words in their primary sense do, act, produce, and achieve. (1935: II, 52)
CONTEXTUALIZING "CONTEXT"
153
And again, The meaning of a single utterance... can be defined as the change produced by this sound in the behavior of people. (1935: II, 59) The second claim is that meanings of words can never be derived from texts themselves. Rather, to understand meaning, one must necessarily consult the extralinguistic "context of situation" in which words are used. In "The Problem of Meaning in Primitive Languages", Malinowski writes: A statement, spoken in real life, is never detached from the situation in which it has been uttered.... [I]n the reality of a spoken living language, the utterance has no meaning except in the context of situation. (1923: 307) Earlier in the same article, Malinowski hints how broad that "context" must be. In discussing the difficulty of translating "native" words (in this case, of Melanesian tribes of Eastern New Guinea) into English, he argues that Such words can only be translated into English, not by giving their imaginary equivalent - a real one obviously cannot be found - but by explaining the meaning of each of them through an exact Ethnographic account of the sociology, culture, and tradition of that native community. (1923: 299-300) That is, meaning is defined modulo the entire culture. While coming at the problem of meaning from his own presuppositions and analytic agenda, Roy Harris endorses Malinowski's two central premises about language. With regard to the purpose of language, Harris rejects what he calls the "telementation" model (Saussure's "thought transfer", wherein the purpose of language is taken to be transferring the speaker's thoughts to the hearer) in favor of a model resembling that of both Malinowski and Wittgenstein, in which the meaning of language lies in its use (see Harris 1981a; Harris 1988a). With regard to determining what that meaning is, Harris uses Mahnowski's notion of "context of situation" as a gateway to his (Harris's) model of "integrational linguistics", wherein (1) human beings are recognized as inhabiting "a communicational space which is not neatly compartmentalised into language and non-language" (Harris 1981a: 165) and
154
NAOMI S. BARON
(2) the point of departure for linguistics should be not 'the language' but rather 'the individual linguistic act in its communicational setting' (1981a: 166). In light of his "integrational" pursuit, Harris acknowledges Malinowski's focus on the importance of context in defining meaning, but criticizes Malinowski for not going far enough: Malinowski's most famous dictum, that language is 'a mode of action, rather than a countersign of thought', when watered down into such statements as 'the context of situation is indispensable for the understanding of the words' or 'the utterance has no meaning except in the context of situation' (Malinowski 1923: 307), appears to reduce to truisms with which nobody would disagree. (Harris 1987a [Love (ed.) 1990: 202]) For Harris, Malinowski's error is in using non-linguistic information to interpret linguistic data, whereas in Harris's model, there should not be two data sources but one: as human beings, whose humanity depends on social interaction, we do not inhabit a communicational space that Nature has already divided for us between language and the non-linguistic. (Harris 1987a [Love (ed.) 1990: 203]) Given the centrality of Malinowski's work to contemporary discussions of the role of extralinguistic context in determining meaning, it behooves us to take a closer look at his work. As we shall see, Malinowski's preoccupation with "context of situation" appears to derive from some questionable assumptions about the languages he studied and the people who spoke them. Moreover, following Malinowski, contemporary discussions of context often confuse or ignore a range of important variables that, in fact, help determine when knowledge of extralinguistic context plays - and does not play - a critical role in determining meaning. The goal of this essay is to "contextualize" discussions of the role of extralinguistic context in analyzing linguistic meaning. That is, to look at the sociohistorical "context" in which Malinowski's work emerged and then to broaden the discussion to other "contexts" for doing theoretical analyses of meaning. We begin (Section 2) by exploring Malinowski's insistence that an understanding of "context of situation" underlies all linguistic meaning. Section 3 dissects the notion of extralinguistic context, distinguishing between language users and language analysts, while Section 4 lays out a spectrum of variables in
CONTEXTUALIZING "CONTEXT
155
terms of which to predict the significance of extralinguistic context in determining linguistic meaning. Drawing upon these analyses, Section 5 looks at ways in which culture and context need to be factored into contemporary attempts to use computers to translate between natural languages. Section 6 reflects upon what we have learned.
2. Malinowsk's
"context of situation":
new insight or bad
science?
Bronislaw Malinowski has always been an enigmatic figure in anthropology. Like Franz Boas, Malinowski took his initial training in physics and, like Boas, left his native land (for Malinowski, Poland; for Boas, Germany) to embrace the study of non-Indo-European, non-literate cultures - Malinowski studying in England and doing field research in the Western Pacific, and Boas living and doing research on Amerindian languages in the United States. But here the similarity ends. While Boas devoted his professional life to establishing the basic human equality of all peoples and their languages (see, for example, Herskovits 1953; Boas 1911), Malinowski was at once a founder of modern British social anthropology and a believer in the late nineteenth century notion that the world is divisible into "civilized" and "primitive" peoples, and that primitive peoples speak structurally primitive languages. Belief in primitive peoples and primitive languages pervaded turn-of-thecentury writing. For example, we find in an article on the languages of the Philippines, appearing in the Journal of the Royal Anthropological Institute, the assertion that "Compared with an Aryan language, Tagalofg] is deficient in many qualities which have made European tongues the vehicle of civilisation" (Mackinlay 1901: 214). In the same vein, Malinowski distinguishes between "modern" languages and "native" languages, the latter representing earlier stages in the development of human speech. According to Malinowski, these native languages are "the living monuments of very primitive languages as they still exist in the native-speaking communities" (Malinowski 1920: 71). Throughout his writings, Malinowski described the languages he was studying as "native" or "primitive" rather than as "indigenous", or simply as "languages". Yet unlike his predecessor armchair anthropologists who speculated about faraway peoples and languages from European drawing rooms, Malinowski insisted upon the importance of getting first-hand knowledge of native languages and peoples: "only a good working knowledge of a native language on the one hand, and a familiarity with their social organization and tribal life on the other, would make it possible to read all the full significance into these texts" (Malinowski 1922: 463). A noble thought, but how did it work in practice? It would appear (at
156
NAOMI S. BARON
least according to his own reports - e.g., 1915: 501) that Malinowski was fairly good at picking up new languages. However, to this observation we must add a crucial caveat: despite his acquaintance with British anthropological linguistics of the time, Malinowski remained, by his own admission (Malinowski 1920: 72 n.l), largely ignorant of contemporary work on formal language analysis, particularly of developments in the United States by, for example, Boas, Bloomfield, and Sapir, who insisted upon analyzing non-Indo-European languages in their own terms. In the words of Hilary Henson, "when Malinowski first went into the field, he had no linguistic theory to support him. But even when he had read some linguistics, he felt its lessons were irrelevant to the situation which he experienced in Kiriwina" (Henson 1974: 44). Malinowski reveals his linguistic naivete in Coral Gardens and Their Magic : "The real difficulty of this language [Kiriwinian] consists not in the complexity of the grammatical apparatus, but rather in its extreme simplicity. Its structure is on the whole what might be described as telegraphic" (1935: II, 36). Mirroring the sentiments (and linguistic prejudices) of armchair anthropologists, Malinowski makes the mistake of assuming that "primitive" peoples must speak "primitive" (that is, not very grammatically complex) languages. This erroneous conclusion on the part of both Malinowski and some of his well-meaning predecessors in the field (mostly missionaries) grew out of an understandable tendency to look for European grammatical categories in the "native" languages being explored. Since as often as not these "native" languages lacked familiar categories such as ablatives or past perfects, they were judged deficient. Given the paucity of linguistic skills among the observers, including Malinowski, indigenous grammatical categories, which were often actually more complex than their IndoEuropean counterparts, were routinely ignored. (For further discussion, see, for example, Hanzeli 1969.) Unlike his confreres who had judged "native" languages to be structurally primitive, Malinowski took a further fateful step. If, he argued, a language such as Kiriwinian was grammatically impoverished, then one needed to resort to some means beyond the language itself to understand what meaning the language was attempting to convey. Malinowski's solution was context: "the relation of the words, as well as the relation of the sentences, has mainly to be derived from the context" (1935: II, 36). Thus, context becomes a critical component in understanding "primitive" languages not because Malinowski had established that context is intrinsically a sine qua non in analyzing human language but because, in Malinowski's mind, the languages he was dealing with were too grammatically impoverished to express meaning clearly:
CONTEXTUALIZING "CONTEXT
157
Malinowski's theories of context-dependency are to an important extent the result of his ethnocentric and linguistically naive assessment of the language as being basically inadequate and limited. [Following Malinowski's model,] [t]he translator is forced to depend on contextualization as the language itself is incapable of making sufficient distinctions. (Henson 1974: 52) None of us, to paraphrase Harris (1987a [Love (ed.) 1990: 202]), would doubt that context has an important role in determining meaning. But is that role as ubiquitous as Malinowski (or Harris) would have us believe? As Henson aptly puts it, Malinowski thus tries to say that meaning is completely context-dependent. If this were so, utterances could share no common structures, no common features, and the language of fishing [one of Malinowski's examples] would be of a completely different species from the language of cooking, for example. (Henson 1974: 54) Or to put the issue in more general terms, if Malinowski is right, no communication is possible between people who don't share one another's context of situation. Since we know that even total strangers manage to share a good bit of information with one another, obviously the claim that meaning is "completely context-dependent" is too strong. But how much too strong? And under what circumstances? If Malinowski's relentless invocation of "context of situation" reflects poor linguistics as much as it does enlightened anthropological field work, the time is ripe for a more fine-grained analysis of when extralinguistic context is - and is not - critical for understanding meaning.
3. Meaning for whom?
Linguists9
"context" / users'
"context"
Some years ago, I was on my way to Boston University's Department of Philosophy, which was located a number of floors up in an academic building. As I waited on the ground floor level for the elevator, a woman approached a cluster of undergraduates standing nearby. "Do you know where the Philosophy office is?", she inquired. "Yes", responded the self-appointed leader of the group. He said no more, but smugly basked in his sophomoric cleverness. For the woman had not explicitly asked where the office was located - only if he knew its location.
158
NAOMI S. BARON
At the literal level of the conversation, the undergraduate was responding appropriately to his interlocutor's inquiry. Yet in practice, all language is situated in a context. Sometimes an awareness of that context is critical for understanding what an utterance means; other times, not. If a judge asks a juvenile defendant, "Do you know the difference between right and wrong?", the judge is asking for a yes/no response, not a definition. However, if a stranger to an area asks a yes/no question as a polite alternative to a more direct request such as "Please tell me where the Philosophy office is", an appropriate response is not "Yes" or "No" but a contentive explanation (e.g. "On the third floor"). The undergraduate, perhaps flush from a lecture on how speakers usually do things with words (e.g. Austin 1962; Searle 1969), chose to ignore the normal conventions. One critically important linguistic convention for language users is knowing when to request (or provide) additional information in order to interpret an utterance or text, and when both parties can make do with the information at hand. To illustrate what we mean: suppose I pick up the phone and call Western Union to send a telegram to my grown son. The message I dictate is: "Dad died. Come home now." The operator who relays the message does not know who "Dad" is, that the intended recipient of the telegram is my son, or even where "home" is (the address from which I am sending the message isn't necessarily "home"). Yet the operator clearly understands that someone died and that a relative or friend is being summoned. Even lacking further situational context, the message is highly meaningful. Interlocutors are aware that language is always situated, and that meaning is ultimately defined modulo linguistic and extralinguistic context, including the larger cultural milieu and the participants' individual life histories. Yet in practice, speakers and writers, listeners and readers agree to ignore much or all of that information as they work to interpret language meaning. Why? Because language is, at base, a tool for assisting human beings in going about their daily business (be it finding a particular office or summoning a person home). How do interlocutors use this tool in real-life circumstances? Through a process I have elsewhere called language orienteering (Baron 1992). Language Orienteering In the outdoor sport of orienteering, adventurers make their way across uncharted territory through the use of minimal tools (e.g. a compass, a few matches) and strategies of their own design: No path is inherently right or wrong. Nonetheless, participants may reach dead ends, need to double back, go hungry for stretches of time, and become
CONTEXTUALIZING "CONTEXT"
159
frustrated. Individual paths share many similarities: the same magnetic North guides each compass, the same forest provides blackberries for all to pick. But in the end, each forager charts a unique path to the goal. (Baron 1992: 24) In much the same way, I have argued, children journey into mastery of their native language: Language acquisition is fundamentally language orienteering. Each participant sets out with a basic set of tools: our genetic endowment that permits us to acquire language. The natural environment through which learners blaze a trail is the social community into which the child is born and the language modeled by that community. The goal is to join the group as a full-fledged member. (Baron 1992: 94) Young children make do with whatever linguistic equipment they have on hand to get their messages across, saying things such as "Why you don't like my chooses?" (uttered by a child whose mother had gently complained about her daughter's selection of an inappropriate birthday present for an older child). "I can ran so faster than you!" (the boast of a 4-year-old boy to his older sister) "Do you know what something?" (a 3-year-old's melange of "Do you know what?" and "Do you know something?") "Are you still have any of these candies before?" (a 4-year-old's attempt to inquire of a friend whether he had ever sampled a piece of candy like this one before) Similarly, adults typically not only ignore grammatical infelicities in their interlocutors' speech, but willingly truncate their own rich contextual arsenal in favor of homing in on what their interlocutor may be trying to express. It matters less whether their path to understanding is elegant or even complete than that they reach whatever level of comprehension is necessary for that particular linguistic event.
160
NAOMI S. BARON
Language Users and Language Analysis What role does context play in determining meaning? It depends upon the task at hand. One task is that of language analysis, carried out not only by linguists and philosophers, sociologists and anthropologists, but also by dictionary makers, professional translators, and computer scientists interested in natural language processing (i.e. by computers) or machine translation of natural languages. It is the responsibility of such professionals to worry the details. To look for holes in the dyke. To find places where a translation system fails or where ambiguity can only be resolved by resorting to knowledge of context. (See Baron 1994 for a discussion of dictionary-based versus contextual notions of meaning.) But language users have a very different task: to get their meaning across. As language orienteerers, most of our energies are focussed on making the system work, not on probing its limits; on getting the essential communicative job done, not on style and nuance. Needless to say, style and nuance are critical parts of the human linguistic landscape. The translator of Dante and the expert witness called in a legal suit to distinguish between an implied and an explicit threat must concern themselves with precision in meaning. Yet for every translation of poetry and for every legal testimony, there are thousands more cases of simultaneous translation and everyday linguistic exchange. In talking about the extent to which language users rely on context to determine meaning, it is useful to distinguish between language producers and language receivers. Language producers (speakers, writers) who are intent upon maximizing the success of their linguistic messages take into consideration a range of factors that may affect how their interlocutor will interpret their message. Adults addressing young children frequently speak more slowly than usual, reduce their range of vocabulary and grammatical complexity, and tend to repeat - both their own speech and that of the child. Good teachers take into consideration the background of their students, trying neither to bore their audience nor to confuse them. Rhetoricians and politicians pay as much attention to the nuances of word meaning (like that of red during the McCarthy era in the US) as they do to dictionary definitions. But these linguistic sensitivities are always modulo the audience. Language receivers (hearers, readers) are typically orienteerers in meaning. The first step is trying to figure out what meaning was most likely intended. If you say to me "Please draw a picture of a dog", I can respond perfectly well without requesting further clarification. This is not because I have read your mind as to what kind of dog you have in mind, or because some ideal type of "dog" actually exists. Rather, it is because as language pragmatists, we primarily use language in order to get things done rather than as an exercise in metalinguistic
CONTEXTUALIZING "CONTEXT"
161
reflection. As part of our socialization, we learn the sort of thing that is meant when we use the word dog without further specification, and as participants in social communities, we are generally cooperative in linguistic exchanges. The philosophy undergraduate was the exception, not the rule.
4. Users' meaning: the varying role of extralinguistic
context
The importance of extralinguistic context, then, in determining meaning must be viewed in terms of the language participant: analyst or user, producer or perceiver. What other variables are significant in assessing the role of context in determining meaning? There are at least four: Mode of Representation In-Group/Out-Group Domain of Communication Lexico-Grammatical Profile Mode of Representation The role of context in determining or clarifying meaning tends to vary depending upon whether the message is spoken or written. Because spoken language is typically addressed to an identifiable audience, while written language is available now (and in the future) to readers who may be unknown to the author, speakers can draw upon contextual assumptions that writers cannot. In face-to-face communication, interlocutors can use visual cues such as facial expressions both to convey meaning and to signal lack of comprehension. Writers must rely on making themselves clear through the written word, the first time around (see Baron 1984). A second important consideration is the options available to the language user (or language community). Does the community have a writing system? If so, is it available to the entire populace, or only to a limited subgroup (such as scribes in ancient Egypt)? Is wide-spread literacy valued in the society (as became the case, for example, in early modern Europe with the introduction of the printing press, the Protestant Reformation, and the rise of a sizeable middle class)? The jury is still out on whether the use of literacy demonstrably affects the way the readers and writers think and reason, compared with their non-literate counterparts (see e.g. Goody and Watt 1968, Scribner and Cole 1978, Olson 1994). However, there does seem to be good evidence that the schooling process, through which most of us learn to read and write, affects the way we organize our thoughts
162
NAOMI S. BARON
(Greenfield 1972). One of the things we learn in school is to make our written language more self-contained, that is, to make deciphering the meaning of written language less dependent upon context than is typical in face-to-face spoken language. When Malinowski spoke of "context of situation", the mode of representation he had in mind was speech. He was, after all, working exclusively with "primitive" languages that had no written expression. In fact, in his seminal article on "The Problem of Meaning in Primitive Languages" (1923), Malinowski introduces the notion "context of situation" for spoken language in contradistinction to written language, whose meaning, he suggests, is selfcontained. In the article, Malinowski begins by contrasting the study of living languages with the nineteenth century preoccupation with using written texts to study evolution of Indo-European languages: context of situation... makes clear the difference in scope and method between the linguistics of dead and of living languages. The material on which almost all our linguistic study has been done so far belongs to dead languages. It is present in the form of written documents, naturally isolated, torn out of any context of situation. (1923: 306) But then he goes on to generalize about differences between any written and spoken texts: In fact, written statements are set down with the purpose of being self-contained and self-explanatory... To take the clearest case, that of a modern scientific book, the writer of it sets out to address every individual reader who will peruse the book and has the necessary scientific training... The book by itself is sufficient to direct the reader's mind to its meaning; and we might be tempted to say metaphorically that the meaning is wholly contained in or carried by the book. (1923: 306-307) "Primitive", unwritten languages are different: when we pass from a modern civilized language, of which we think mostly in terms of written records, or from a dead one which survives only in inscription, to a primitive tongue, never used in writing, where all the material lives only in winged words, passing from man to man - there it should be clear at once that the conception of meaning as contained in an utterance is false and futile. A
CONTEXTUALIZING "CONTEXT
163
statement, spoken in real life, is never detached from the situation in which it has been uttered. (1923: 307) Or writing more than a decade later in Coral Gardens and Their Magic, Whereas the language of literature in more highly developed communities is handed down to us on marble, brass, parchment or pulp, that of a savage tribe is never framed to be taken outside its context of situation. The speech of a preliterate community brings home to us in an unavoidably cogent manner that language exists only in actual use within the context of real utterance. (1935: II, vii) In sum, Malinowski's "context of situation" was developed to deal with the "primitive" languages of "savage tribes", not with "modern civilized language[s]" that can be and are written. What of Roy Harris's position on possible differences between linguistic modalities in the role of context for understanding meaning? Unlike Malinowski, for whom an understanding of extralinguistic context was only necessary in determining the meaning of spoken messages, Harris sees context as central in interpreting meaning in both speech and writing. As Nigel Love writes, for Harris written texts supply convincing and readily assimilable illustrations of a principle that in fact applies in respect of language-use in any medium: all utterances, whether written, spoken, or manifest in some other way, have to appear in a context; and that context must be assumed to be, not merely potentially relevant to the message, nor even always and necessarily relevant to the message, but, quite simply, an integral part of the message. There is no such thing as a contextless utterance. (Love 1990b: 20) In-Group / Out-Group The role of context in determining meaning also varies with the type of social relationship between the interlocutors. We can call the extremes "in-group" and "out-group", meaning language users who share presuppositions and experiences in common versus those who don't. Since they share so much extralinguistic context, in-group members can afford to be less linguistically explicit than outgroup members as they proceed in their orienteering quest for meaning. In the linguistic literature, the terminology that perhaps comes closest to what we mean by in-group versus out-group is Basil Bernstein's "restricted" and
164
NAOMI S. BARON
"elaborated" code (e.g. Bernstein 1971). The distinction was initially presented in the context of differential socialization of classes in England. Bernstein argues that lower-class speakers are essentially socialized to speak in restricted code, while middle-class speakers have access to both restricted and elaborated styles: forms of socialization orient the child towards speech codes which control access to relatively context-tied or relatively context-independent meanings. Thus...elaborated codes orient their users towards universalistic meanings, whereas restricted codes orient, sensitize, their users to particularistic meanings.... Where codes are elaborated, the socialized has more access to the grounds of his own socialization, and so can enter into a reflexive relationship to the social order he has taken over. Where codes are restricted, the socialized has less access to the grounds of his own socialization, and thus reflexiveness may be limited in range. One of the effects of the class system is to limit access to elaborated codes. (Bernstein 1971: 176) In later work (and perhaps in response to critics), Bernstein indicates that speakers of all classes have access to both codes. The real difference between classes comes in how often they make use of each code. Whether or not one agrees with Bernstein's arguments concerning causal relationships between class, language socialization, and cognition, the linguistic profiles he constructs for the two codes clearly reveal differential roles for context in determining meaning. Where is in-group language (restricted code) typically used? Between family members, between friends, between co-workers, between members of groups who share common experiences. Such groups tend to be relatively small (be they social or special interest groups), though sometimes larger groups are also homogeneous (Malinowski's Kiriwinians or the inhabitants of a proverbial small town who know everyone else's business). What matters is that their interlocutors share enough contextual presuppositions to make more contextbound language correctly interpretable. When is out-group (elaborated code) more typical? Among interlocutors who have different interests, different experiences, different life histories. Much as writing historically grew out of urban conditions, out-group code is more cosmopolitan language, usable between speakers who don't know one another and don't share contextual presuppositions. Cities not only bring together diverse populations; they also bring alternatives that need to be differentiated linguistically. A hundred years ago in rural Vermont, when you went "to the store", there was only one store in town; in today's cities and suburbs, there are scores of options.
CONTEXTUALIZING "CONTEXT"
165
Like most initial definitions, the dichotomy between in-group and out-group language needs to be further refined. Cities, for example, are often divisible into smaller neighborhoods, which in turn are more tribal in character (New York's Chinatown, Boston's Italian North End). At the same time, mass communication and global transportation bring together information and visitors to people who were once geographically and communicatively isolated. Grandparents from the old country can live in New York or Boston and never learn English, and rural dwellers can live in cyberspace. Moreover, as we will see later on in this paper, the same individual can be both an in-group member and an out-group member, depending upon the larger social group against which he or she is being measured at the moment. Formal schooling and experience with "out-group" speakers provide opportunities for learning and using out-group language, that is, language whose meaning is less dependent upon an understanding of "context of situation". But what about the populations that Malinowski studied? They had no formal schooling. Moreover, how often did they have occasion to use elaborated code? Given the social structure of tribes such as the Kiriwinians, would the distinction between restricted and elaborated code, developed to deal with contemporary urban speech in a class-stratified society, make sense? Even if the answer were "yes", would Malinowski have had the sociolinguistic sophistication necessary to elicit and identify such distinctions? In short, Malinowski's insistence upon the centrality of "context of situation" to understanding language appears to be a product of the populations he studied (and, of course, the tools of linguistic analysis he had available). Had Malinowski been observing the speech of Berliners or New Yorkers, it seems unlikely that he would have concluded that "in the reality of a spoken living language, the utterance has no meaning except in the context of situation". Contemporary students of language who cite Malinowski as the source of their belief that knowing the context of situation is always critical for interpreting linguistic meaning will need to make their own case. Malinowski himself only made the case for a very limited linguistic community. Domain of Communication Thus far, we have looked at the role of modality (spoken, written) and of social grouping (in-group, out-group) in predicting the extent to which meaning is context-dependent. Does the subject matter about which people are talking (or writing) - that is, the domain of communication - also influence the extent to which extralinguistic context plays a role in determining meaning?
166
NAOMI S. BARON
We can think about domain in two related ways. The first is a simple dichotomy between direct and indirect reference: is the topic of conversation something physically at hand, or is it distant in time or space? Direct reference communication about the here and now - can be less linguistically explicit, since extralinguistic context is there for the asking. A mother inquires of her child "Does it hurt?" and the child doesn't need to ask, "Does what hurt?" - he is holding his toe that both of them saw him stub. Indirect reference - communication about what can't now (or perhaps ever) be seen (last year's Christmas tree, the first moon landing, the consumer price index, the causes of World War I) - requires more explicit language. Not only is the physical context absent but the likelihood that interlocutors share the experience (or idea) under discussion is far lower than for direct reference. Context of situation plays a more critical role in determining meaning with direct reference than with indirect. The second component of domain is the actual content of what is being talked about or written about. Is the linguistic interchange about things, actions, or ideas - friends swapping recipes or college roommates debating the pros and cons of the death penalty? Is the subject narrow or broad in scope - instructions for dialing up the campus computer or a general lecture on computer networks? Is the exchange part of everyday discourse between ordinary language users, or is it limited to a special semantic domain (be it physics, steel manufacture, or soccer)? Is the language at hand a conversation? a narrative? a legal document? a shopping list? Not surprisingly, the role of domain interacts with the profile of the participants involved. Are they, for example, members of an in-group or outgroup? In an operating theater, the surgeon and surgical nurse are members of an in-group, while medical students are members of an out-group. All the nurse needs to hear is "scalpel" to know what instrument to hand and how, while the medical students need to have the contextual cues made explicit before they can usefully begin their surgery rotation. All of these variables tend to influence the significance of extralinguistic context in determining meaning. The more that language is used for indirect reference, the broader the scope of the domain of communication, and the more the communication is between out-group members, the less meaning interpretation is dependent upon extralinguistic context. And, of course, vice versa. Given the vast range of possible domains of communication, we can reasonably ask of a general theory of meaning interpretation that it take into account meaning in different domains of communication. Yet when we examine
CONTEXTUALIZING "CONTEXT"
167
Malinowski's framework for analyzing meaning, we find its domain of communication particularly narrow. At the beginning of this essay, we noted that Malinowski's analysis of language is based upon two premises: that the purpose of language is not to reflect thought, but to act; and that the meanings of words can never be defined in isolation. These premises were formulated on the basis of a very restricted domain of language use. As Firth writes, Malinowski essentially limited the scope of discourse he analyzed to everyday conversation between people sharing activities in common: Ranging himself with the primitive man's pragmatic outlook and regarding language as a mode of action rather than as a counter-sign of thought..., Malinowski selected for notice only such features of his languages as were essentially bound up with his contexts of situation in trading, fishing..., gardening and similar pursuits. (Firth 1957: 105) In terms of our own framework of analysis, Malinowski's subjects were members of an in-group making significant use of direct reference to talk about limited topics - circumstances under which extralinguistic context is most likely to be valuable in determining meaning. Had Malinowski been analyzing abstract ideas rather than shared actions, global agriculture rather than local gardening, conversations between members of out-groups rather than in-groups, it is unlikely that context of situation would have appeared as central in determining meaning. Lexico-Grammatical Profile The final variable we will look at is what we will call a language's lexicogrammatical profile. As anthropologists and linguists have long observed, languages differ from one another in the meanings they encode through individual words or grammatical markers. To the extent that meaning is overtly expressed lexically or grammatically, it is made explicit and does not need to be deduced from extralinguistic context or clarified through circumlocution. Leaving aside the Whorfian question of whether or how such differences mold speakers' Weltanschauung (see Whorf 1956; Lucy 1992), it is empirically clear that differences in lexical and grammatical expression pervade the linguistic landscape. Examples of differential encoding of the physical color spectrum are well known (see e.g. Berlin and Kay 1969). Two other obvious cases are body parts and verbs of motion.
168
NAOMI S. BARON
Body Parts Like the color spectrum, the human body is not divided, a priori, into individually named sections. While most languages make the same basic distinctions (head, upper limbs, lower limbs, trunk), languages sometimes differ in the degree of labeling given to contiguous body parts. In Old Norse, for example, the single word hond refers to the entire upper appendage from shoulder to fingers. Thus, when we read that a character in one of the Old Norse sagas has hidden his sword "undir ['under'] hendi ['hand'] ser ['his']", we deduce that he has hidden the sword under his armpit, not his hand. Verbs of Motion All languages have words of motion for talking about getting from here to there. While languages commonly use a single verb of going, e.g. English "I am going to the store." Russian uses different verbs, depending upon the means of locomotion, e.g. "go by foot": idit "go by vehicle": exat (The Russian verb system for movement is actually more complex, distinguishing, for example, between going one way and making a round trip.) In English, if we wish to specify the means of locomotion, we need to add more (or different) language (e.g. "I am walking to the store", "I am going to the store and taking the car") or rely on extralinguistic context (e.g. if I live only a few blocks away from the store, you assume I will walk; if the store is a mile off and I typically drive, you assume I will drive). In classifying languages, we typically use such characteristics as genealogy (i.e. historical origins), areal criteria (similarities between languages spoken by geographically contiguous peoples), and typology (e.g. grouping languages on the basis of how they tend to order subjects, objects, and verbs in sentences). Such typologies have been developed by historical linguists or linguists interested in language structure. For the sociologist or anthropologist of language interested in how language functions in a social context, it would be fascinating to try classifying languages by the degree to which they make meaning explicit in lexicon or grammar (e.g. use of an overt morphological marker of plurality, as in English -s) as opposed to the extent they rely on extralinguistic context for interpreting meaning (as in the expression of plurals in Japanese). (For discussion of such classification issues, see Comrie 1988, 1989.)
CONTEXTUALIZING "CONTEXT"
169
While a full-scale investigation of a language's lexico-grammatical profile would require extensive data collection and analysis, we can illustrate the kinds of issues that would need to be explored by looking at selected aspects of one language's profile and comparing them implicitly (i.e. through translation) with English. The language we will look at is Japanese. This comparison helps set the stage for discussion of machine (i.e. computer) translation in Section 5 below. Japanese and Context: A Brief Introduction The Japanese have long viewed their language (not to mention other aspects of their culture) as unique. In the words of the linguist Haruhiko Kindaichi, author of the classic book Nippongo (The Japanese Language), The Japanese language has a unique position among the languages of civilized countries. That is, there is absolutely no other language of a similar nature. (Kindaichi 1978: 30) Moreover, many Japanese believe that only someone born Japanese is capable of fully learning the language (Miller 1982). While both of these sentiments are obvious exaggerations, they convey an attitude long shared by the Japanese, who, despite periods of invasion and even conquest, have, until quite recently, maintained cultural isolation. Undergirded by a strong system of family and community, strict behavioral norms, and, at least over the past century, by state-imposed patterns of socialization through formal education, Japanese culture is well suited to a language system that relies heavily on extralinguistic context for conveying meaning. For despite their size, the Japanese people constitute a linguistic in-group. (Note: the Japanese people as a whole constitute a linguistic in-group as compared with gaijin, i.e. foreigners. However, as we will see, within Japanese society, the same individual may be a member of an in-group or an out-group, depending upon the make-up of the subgroups being compared.) From even a handful of examples, we can see how context-dependent much of meaning interpretation is in Japanese. Number As Kuzuto Harada so starkly illustrated with his last word - "Shark!", Japanese does not have a regular grammatical means of distinguishing between singular and plural. If you say to me
170
NAOMI S. BARON
Kono pen wa? [this/these] [pen/pens] you may either mean "How about this pen?" or "How about these pens?" To know which one, I need to see what you are pointing at. While Japanese does not grammatically mark singular and plural, it does have a fairly complex way of enumerating items by number, when specification of number is called for. The system concatenates the name of the object with a number and a classifier. Japanese has several dozen classifiers, e.g. -satsu is used with bound objects such as books; -mai denotes thin, flat objects like pieces of paper; -dai specifies vehicles and machines such as cars, e.g. hon go-satsu [book] [five] [bound object] kami ni-mai [paper] [two] [thin, flat object] kuruma roku-dai [car] [five] [vehicle or machine] As we might expect, Japanese has words indicating amount, such as takusan ('many,' 'much') and sukoshi (ca few'), but the use of such quantifiers is at the discretion of the speaker, not required by the grammar. Gender Japane has a plethora of grammatical distinctions between the language of women and men: distinctions in the pronouns used to refer to oneself in the first person, distinctions in some of the verb forms, distinctions in lexical appropriateness. Yet Japanese does not grammatically mark distinctions between male ar-1 female proper names. Hence, Yamamoto-san des ka? [Proper name] [politeness suffix] [is] [question particle] could be asking "Is it Mr Yamamoto?" "Is it Mrs Yamamoto?" "Is it Miss Yamamoto?"
CONTEXTUALIZING "CONTEXT"
171
"Is it Ms Yamamoto?" The suffix -san marks polite reference to people who are not members of one's immediate social in-group (e.g. such as an acquaintance who is not a close friend - see Section 5 below). Given the close-knit structure of Japanese society, the speaker can typically presume the listener knows who Yamamoto-san is, and hence grammatical or lexical specification of gender becomes redundant. Grammatical Subjects When Westerners begin learning Japanese, they are often confused as to who is being talked about in a sentence. First and second pronouns are rarely expressed; third person pronouns are typically reserved for peers or subordinates whose position does not require a demonstration of respect. Hence, when we encounter a sentence such as Yoku ikimasu ne. [often] [go] [sentence particle of confirmation or agreement] it could mean any of the following: "You go often, right?" "He/she go/goes often, right?" "They go often, right?" Similarly, because pronominal subjects are typically omitted, a sentence like Sensei de su ka? [teacher] [is] [question particle] might mean "Is he/she the teacher?" "Are you a teacher?" "Is it you, Teacher?" The meaning, presumably, comes clear in context. But extralinguistic context is hardly the only avenue for conveying meaning in Japanese. We must not make Malinowski's mistake of assuming that a non-IndoEuropean language lacks grammatical complexity simply because that complexity is not revealed in typical Indo-European conjugations or declensions. Anyone
172
NAOMI S. BARON
who has studied the Japanese verb system knows how grammatically intricate it is. In addition to being divided into several category-types for conjugation, verbs in Japanese must reflect keigo, the system of Japanese honorifics: how you address in-group members, out-group members, people before whom you must humble yourself, and so forth. There are, for example, three basic levels of politeness for all verbs. But by adding further nuances, you can distinguish up to fourteen levels of politeness - all linguistically marked (see Niyekawa 1991). Kindaichi notwithstanding, Japanese does not stand wholly apart from other languages of the world. Other languages share the grammatical characteristics of Japanese, including its reliance on extralinguistic context to determine number, gender, or person, and its multiple levels of grammatical distinctions for verbs. The question of interest is how many meanings are typically derived by users from context, and how many and which through lexico-grammatical specification.
5. Culture,
Context,
and Machine
Translation
[Translation is a matter of compromise among competing criteria (Kay et al. 1994: 9) How can we understand the lexico-grammatical profile of a language other than our own? The answer, as we have just illustrated with examples from Japanese, is through translation into a language we already know (in this case, English). Yet translation introduces a whole new set of challenges. The Problem of Translation When we speak of "translation", we generally mean the process of representing the meanings (and sometimes the grammatical techniques and stylistic cadences) of one language in another. In principle, the notion of translation can be applied to any communication between interlocutors (including those using different dialects or even individual idiolects - see Baron 1986), though here we will limit our scope to translation between what are commonly recognized as distinct languages. Translation is about much more than word equivalences. It is about converting messages originally expressed by speakers (or writers) with unique sets of individual and cultural experiences and expressed through one lexico-grammatical formulation into a second formulation for listeners (or readers) with very different personal and social histories. It is roundly acknowledged - by
CONTEXTUALIZING "CONTEXT
173
professional translator and vacationing traveler alike - that translation is hard. Try as we may, we have difficulty in fully grasping (at least in the time typically available) the "situation" that makes the linguistic message correctly interpretable. This "situation" may be linguistic (e.g. the text preceding or following; the literary character of the original) or non-linguistic (e.g. the relative social positions of the interlocutors). And so we make compromises. We may, for example, agree when translating poetry to sacrifice the rhyme scheme of the original to make sure that the meaning is conveyed as accurately as possible. In courtroom translation, we might be willing to ignore the finer subtleties of a system of honorifics in order to further the proceedings. The problem of translation is knowing what compromises to make. Computers and Translation For over thirty years now, a new variable has been introduced into the challenge of translation: the computer. Work on "machine" translation got its initial boost during the early years of the Cold War in America, where the impetus to translate Russian scientific research into English brought forth considerable federal funding. However, in the mid 1960s, a government panel judged progress to date to be questionable (see National Research Council 1965), and funding was abruptly curtailed. During the next fifteen to twenty years, a number of projects, largely housed or funded outside of the United States, continued to work on machine translation. Some of the best known include a weather-forecasting translation system developed at the University of Montreal (see below) and work at the University of Texas at Austin, supported by the Siemens Corporation in Germany, for translating technical manuals. By the 1980s, the prospects for machine translation were once more seen as promising. Developments in computer hardware, along with progress in natural language processing (a necessary preliminary step for any translation system) rendered machine translation systems increasingly successful. Most work on machine translation is economically driven. Japan, for example, committed itself in the 1980s to making machine translation viable because of the high costs of human translation: [Japanese] businesses normally operate with the world market in mind, and every time a new product is introduced or an old product is updated, a huge volume of technical material related to its use and maintenance must be translated into various foreign languages. For commercial reasons, such
174
NAOMI S. BARON
translation must be completed at virtually the same time as the original Japanese documents.... Currently [ca. 1986] all such technical translations are done by human translators, and it has been estimated that sums in excess of 100 billion yen [one billion US dollars in 1994] per year are spent on this alone. If even half of that sum can be saved by machine translation, a considerable saving will have been made. (Nagao 1989: 2) More recently, Germany has undertaken an even more challenging venture, known as Verbmobil, that aims to provide simultaneous translation of spoken language, thus compounding the already hefty problems of natural language processing and translation with those of speech recognition (see Kay et al. 1994). The payoff for such a system would be enormous. Commerce could function with relative linguistic ease worldwide, and the vagaries of international fortunes would not force businessmen back for linguistic retooling every few years. Spinoffs from Verbmobil would, of course, benefit others as well, from the scientific community to travelers to intelligence offices. Is a project such as Verbmobil realistic? The German government's Federal Ministry of Research and Technology commissioned Stanford University's Center for the Study of Language and Information to find out. In their report, the Stanford researchers offered a frank assessment of the difficulties the project would face, and recommended the development of two research prototypes, to be built according to a number of delimiting parameters. One of these parameters was cooperation between interlocutors. Prototypes should be designed to encourage dialogue on subject matter of mutual interest rather than creating "adversarial interaction." (Kay et al. 1994: 7) I am reminded, with some embarrassment, of the "adversarial interaction" I once encouraged when, at a computer show demonstrating a new commercial natural language processing program, I intentionally used my linguistic training to stump the system. The company's glossy brochures were boasting that the program could "understand a sentence just the way a person does", and I was out to establish the claim's hyperbole. The demonstrators at the booth were not amused. For a translation system to work, both parties have to agree to cooperate. Generally this cooperation entails making compromises. In return for users agreeing to limit their messages (lexically, grammatically, stylistically), the machine "agrees" to translate these messages, without promising to replicate native-speaker accuracy. The second parameter the Stanford report identified was the domain of meaning over which the translation prototype was to function. By limiting the domain of discourse, it becomes possible to make both the dictionary size and the
CONTEXTUALIZING "CONTEXT"
175
grammatical parser more tractable, hence increasing the chances of translation success. To date, machine translation programs that have produced the most robust results are those that restrict the language domain. A good example is the TAUM ( = Traduction Automatique de l'Université de Montréal) METEO system, developed in Canada to translate weather forecasts from English into French. The system, introduced in 1978, receives weather information from meteorological stations across Canada several times a day, translates them directly into French, and sends them out immediately to Francophone newspapers, radio, and television stations. TAUM METEO is successful because the task (and language) at hand are tractable. The vocabulary used for talking about weather forecasting is limited, as are the syntactic constructions necessary for formulating forecasts. Moreover, French-speaking Canadians are more interested in learning whether it will rain tomorrow than in playing "stump the system" as I had once done. Even those who are aware that the grammar used for translations is a truncated version of French are willing to compromise, to trade off possible linguistic nuances for the efficiency gained through machine translation. The need for compromise in contemporary machine translation is eased somewhat by two of the factors we identified in Section 4 as influencing the relative importance of context in determining meaning, namely Mode of Representation and In-Group/Out-Group. While projects such as Verbmobil suggest that in the future, real-time machine translation of ordinary spoken language may become possible, successful machine translation projects to date have used writing as their mode of linguistic representation. We have seen that, paradigmatically, the use of writing reduces dependence upon context for interpreting linguistic meaning. Similarly, machine translation is used for communicating with members of an out-group, who are not expected to share personal and social histories with the language producer. The only domain that is assumed to be shared is the domain of discourse (e.g. what sort of information goes into a weather forecast, and what basic terms such as "low pressure system" or "cold front" mean). As a result, when messages are formulated in the source language, they are constructed with an out-group interlocutor in mind, and thus tend to minimize reliance upon context to convey meaning. Cultural Challenges to Machine Translation In principle, out-group status reduces dependence upon context for interpreting linguistic meaning. However, when the interlocutors belong to distinct cultures
176
NAOMI S. BARON
(the ultimate out-groups), a new layer of problems arises in attempting to do translation. Eugene Nida, linguist and Bible translator, noted years ago how critical an understanding of cultural patterns is to rendering usable translations. Even when word-for-word equivalences exist between languages, literal translation may be wide of the mark. For example, if the Bible translator attempts to translate literally the expression "He beat his breast" (speaking of the repentant publican, Luke 18:13), he may discover that, as in the Chokwe language of Central Africa, this phrase actually means "to congratulate oneself (the equivalent of our "pat himself on the back"). In such instances it is necessary to say "to club one's head." (Nida 1959: 11-12) Malinowski himself was driven to develop a contextual based theory of meaning because of cultural disparities between Oceanic languages and English: in working out [my linguistic material from research among the Melanesian tribes of Eastern New Guinea], I tried to translate my texts into English, and ...was faced by fundamental difficulties. These difficulties were not removed, but rather increased, when I consulted the extant grammars and vocabularies of Oceanic languages [which]... would give the next best approximation in English to a native word. (Malinowski 1923: 299) As Malinowski correctly notes, such a casual approach to translation will not do: the object of a scientific translation of a word is not to give its rough equivalent...but to state exactly whether a native word corresponds to an idea at least partially existing for English speakers, or whether it covers an entirely foreign conception. (Malinowski 1923: 299) How does one arrive at a "scientific" translation? By calling upon extralinguistic data: Such words can only be translated into English, not by giving their imaginary equivalent - a real one obviously cannot be found - but by explaining the meaning of each of them through an exact Ethnographic account of the sociology, culture and tradition of that native community. (Malinowski 1923: 229-300)
CONTEXTUALIZING "CONTEXT"
177
While designers of contemporary machine translation systems are hardly concerned with translation between languages of such disparate cultures as America and New Guinea, they are nonetheless faced with the challenge of incorporating an understanding of cultural variables into their translation schemes. In some instances (e.g. translation between English and French), the cultural differences are small in comparison with distinctions between, for example, English and Japanese. Consider differences in the social importance - and the linguistic markings of notions of politeness in Japanese versus most European languages: although they are infrequently used in European languages, in Japanese there are many words of respect and politeness which reflect the social positions of the speakers, as well as distinctly male or female expressions which lie at the heart of Japanese culture. These are factors which must be considered when translating between Japanese and European languages, but they are far less important when translating solely within a European context. (Nagao 1989: 6-7) In Japan, social relationships are defined by the group to which you belong: your family, your circle of friends, your company. Each individual belongs to a number of different groups. Linguistically, when referring to members of your group, you use one kind of language; when speaking with (or about) people outside of your group, you use another. To complicate matters, depending upon the situation, the same individual might be seen as either an in-group or an out-group member. Suppose you are a secretary in a Japanese company. When talking with a co-worker about the president of the company, you must refer to the president using an out-group honorific form, since the president is not part of the in-group made up of you and your co-workers. However, when talking with a visitor to the company about the president, you must refer to the president with in-group humble forms, since regardless of his or her status in the hierarchy, the president and all employees in the company form an in-group in comparison with the visitor, who, for purposes of conversation, is a member of the out-group. (See Brown and Levinson 1978 for a more detailed discussion of linguistic markers of politeness.) To get a sense of how politeness is marked lexico-grammatically in Japanese, let us look at some examples involving first nouns and then verbs. Politeness Indicators on Japanese Nouns: -san, o-, noun pairs As we have already seen, the suffix -san is generally added to the name of the person with whom (or about whom) you are speaking as a marker of respect, e.g.
178
NAOMI S. BARON
Suzuki-san imasu ka? [Mr/Mrs/Miss/Ms] [be located] [question particle] "Is Mr/Mrs/Miss/Ms Suzuki here?" However, -san can only be added to the name of a person who is in the outgroup vis-à-vis the speaker. When the secretary described above talks about the boss with her co-worker, the boss is called, e.g. Suzuki-san. However, when the secretary discusses the boss with a visitor to the company, the boss must be called only Suzuki, since with respect to the visitor, the boss is now a member of the same in-group as the secretary. Japanese has several lexico-grammatical means of marking politeness levels on common nouns as well. One regular morphological mechanism used with many words is to add the prefix o- For example, sushi osushi nee san oneesan
'sushi' (not marked for politeness) 'sushi' (marked for politeness) 'older sister' (not marked for politeness) 'older sister' (marked for politeness)
In other cases, distinct lexical forms exist for talking about substantives vis-à-vis in-group versus out-group members. For instance, among the words for 'home' are uchi and otaku. While uchi is used for talking about the homes of in-group members, including oneself, e.g. Tomodachi no uchi wa, tiisai desu. [friend] [connective] [home] [particle] [small] [is] "My friend's house is small." otaku is used for identifying homes of members of the out-group, e.g. Takana-sensei no otaku wa, dochira desu ka? [Tanaka] [Professor/Doctor] [connective] [home] [particle] [where] [is] [question particle] "Where is Professor/Doctor Tanaka's house?" Politeness Indicators on Japanese Verbs: Plain, Humble, Polite As we noted earlier, Japanese also has a grammatically elaborate way of marking politeness levels through its verb system. At its most basic level, Japanese distinguishes among
CONTEXTUALIZING "CONTEXT
179
(1) plain verbs (not marked with respect to politeness) (2) verbs that humble the speaker (generally used to refer to oneself or members of one's in-group when addressing a listener from the out-group) and (3) verbs that show respect for the listener (generally used when referring to a listener or a third party who is a member of the speaker's out-group). And so we find the following kinds of contrasts: Doitsu e ikimasu. [Germany] [to] [am going - not marked for politeness] "I am going to Germany." Sensei, doitsu e mairimasu. [teacher] [Germany] [to] [am going - marked for humility] "Teacher, I am going to Germany." Sensei wa, doitsu e irassyaimasu ka? [teacher] [particle] [Germany] [to] [are going - marked as honorific] [question particle] "Teacher, are you going to Germany?" Japanese offers an interesting blend of lexico-grammatical explicitness (as we have just seen with markers of politeness on nouns and verbs) and contextdependency (as we saw in Section 4 above). Every language, of course, makes use both of lexico-grammatical markers and of extralinguistic context in defining meaning. These unique balances reflect not only linguistic differences but also cultural distinctions. In sum, the cultural challenge for machine translation comes in figuring out how to handle disparities in the lexico-grammatical profiles of the source and target languages. Not only must the program designer be aware of what these differences are, but he or she must decide which of these differences to incorporate into the translation program.
NAOMI S. BARON
180
6.
Conclusions [While] we may not agree entirely with [Malinowski's] reasons...nothing that has happened in the field of language studies since Malinowski's day suggests that in the end he got the answer wrong. (Harris 1987a [Love (ed.) 1990: 2089])
What have we learned about context and meaning? We began by tracing the "context" of Bronislaw Malinowski's notion that an understanding of the "context of situation" for any linguistic utterance is necessary to interpret the meaning of that utterance. We found that Malinowski's thesis was itself contextbound by poor linguistic methodology, presuppositions about the "primitive" nature of his informants' language, and a single mode of representation (speech) used among members of an in-group. When some of these variables are altered (e.g. looking at written language or language directed to an out-group), we saw that the role of context in determining meaning diminishes. We also found that languages structurally differ from one another in their lexico-grammatical profiles, that is, in the extent to which a language formally expresses meaning as opposed to relying on extralinguistic context for making meaning clear. This profile, in turn, often reflects cultural presuppositions of the language community. Growing interest in machine translation has helped sharpen our awareness of differences in language structures, cultural patterns, and language modalities. Moreover, the notion of translation as compromise suggests useful ways for us to rethink the larger question of the role of context in determining meaning. In Section 3, we differentiated linguists' context from users' context. All language is, of course, language in context, and all meaning is derived modulo that context. Yet language analysts sometimes conflate this theoretical reality with questions of actual language use. Language users, as orienteerers, are typically more interested in making whatever compromises are necessary to get on with the work of language than in ensuring that each nuance of meaning is clearly expressed and understood. Generally speaking, interlocutors are cooperative. As needed, they limit the domain of discourse, ignore ungrammaticality, and extrapolate context rather than requiring it be made explicit. Most of the time, this pragmatic approach will suffice. Occasionally, it leads to confusion, as when a potentially ambiguous situation is interpreted differently from the way the speaker intended. If I say that The conference luncheon was awful.
CONTEXTUALIZING "CONTEXT"
181
I might be commenting on the quality of the food, but my listener may think I am talking about the caliber of the meeting. Such mistaken interpretation may be clarified ("I was talking about the food, not the speaker!"), left uncorrected, or even go unnoticed. In fact, much of the time, interlocutors are unaware that an utterance is ambiguous, even when speaker and hearer derive alternative interpretations. The case of Kuzuto Harada's ambiguous cry of "Shark!" is the exception. Roy Harris's model of integrational linguistics takes context to be an essential component in determining all linguistic meaning. He finds Malinowski to be right, even if for the wrong reasons. My own conclusion is more qualified. Context matters - and to the language analyst, may matter ultimately. But for language users - from New Guinea tribesmen to families with young children to members of a Japanese company to users of machine translation systems, what matters most is using language to get on with social interaction. To the extent knowledge (or expression) of context is critical, interlocutors depend upon it. When it is expendable (or so perceived), it can be and is ignored.
9 Is Ethnomethodological Conversation Analysis an "Integrational" Account of Language?* David Fleming
The contemporary study of human communication may be notable for, among other things, a growing tendency to treat language as activity rather than thing. This perspective is discernible in at least two trends in recent research on language and discourse: an increased attention to particularity and context and a privileging of procedural over substantive phenomena. The methodological implications of such inquiry may be somewhat obvious: if language is inextricably situated in place and time, then the analysis of language should concentrate on local practices and particular acts. An example of this kind of research can be found in the emerging field of literacy studies. In Kintgen, Kroll, and Rose (1988), Gee (1989), Beach, Green, Kamil, and Shanahan (1992), Cook-Gumperz (1986), and DeCastell, Luke, and Egan (1986), "literacy" has been re-configured as a constellation of situated acts and local practices, analyzed through simultaneously social, cognitive, cultural, historical, and material investigation. Traditional disciplinary approaches have dealt with such acts and practices by abstracting "language" out of them and hypostasizing it in mental or social structures. In psychology, for example, literacy has often been reduced to an abstract decoding process. In the research cited above, on the other hand, literacy is approached through particular literate acts and the social, historical, and practical contexts in which they are situated. Such work has been useful in its re-conceptualization of one aspect of human language. Missing, however, is a theory of how language in general might be approached this way. The integrational linguistics of Roy Harris may provide just such a theory. In his work, Harris has criticized accounts of communication in which language is treated independently of the occasions of its use; and he has proposed in their place an account in which linguistic and non-linguistic phenomena are integrated in situated communicative acts. Harris's theory provides language researchers * I wish to thank Paul Hopper, Roy Harris, and Talbot Taylor for their comments on earlier versions of this paper.
ETHNOMETHODOLOGY AND INTEGRATIONAL LINGUISTICS
183
with a powerful critique of orthodox linguistics and a forceful re-conceptualization of the relationship between language and practical action. It has been less helpful, however, in motivating a productive program of research that moves theory beyond critique. One contemporary school of social science may offer a model of "integrational" analysis. In ethnomethodological conversation analysis, researchers investigate with admirable precision and rigor the intersections of language and social life. As such, conversation analysis, or "CA" as it is often called, appears to be a useful tool for the integrationism proposed by Harris. And CA is a flourishing area of research - in the sociology of science, the analysis of institutional communication, and the study of everyday talk, it has claimed notable advances in knowledge. Silverman (1993: 120) refers to it as the leading approach in the analysis of naturally-occurring linguistic interaction. And among rhetoric and composition scholars, ethnomethodological analyses of language behavior, like CA, appear to be on the rise (see, for example, Brandt 1992; Geisler 1994). More to the point of this paper, conversation analysis and Harrisian linguistics share the belief that language is inescapably contextualized, time-bound, communicative activity. But can CA provide a thoroughly integrational account of language in the sense proposed by Harris? In three specific respects, my answer here is no. First, conversation analysis unnecessarily restricts the study of language to the analysis of transcripts of informal conversation. Second, it formalizes conversational structure in ways that work against a radical contextualization of language. Third, it equates temporality with sequentiality, thereby reducing the temporal dimensions of discourse. On these counts, ethnomethodological conversation analysis fails to meet the challenge of Harris's goals for an integrational account of language.
1. Different
approaches,
a shared
orientation
Harrisian integrationism For Roy Harris,1 the fundamental error of modern linguistic theory is the belief that human communication can be explained, even in part, by appealing to 1
Harris is not the only scholar associated with integrational linguistics. A handout from a panel discussion on integrational linguistics held at the XVth International Congress of Linguists in Quebec in 1992 lists more than fifty articles and fifteen books from eighteen different scholars, including Naomi Baron, Deborah Cameron, Tony Crowley, Paul Hopper, Nigel Love, Peter Muhlhausler, and Talbot Taylor.
184
DAVID FLEMING
structured systems of contextless signs and the rules of their combination. In such works as The Language-Makers (1980), The Language Myth (1981a), and The Language Machine (1987b), Harris argues that orthodox accounts of communication, in their postulation of linguistic invariants underlying particular linguistic acts, abstract language from the occasions of its use and segregate it from the situated activity of which it is always a part. The philosophical underpinnings of this mistake, he claims, are two widely-held Western assumptions: the "telementational fallacy", in which communication is viewed as the transference of thoughts from one mind to another, and the "fixed-code fallacy", in which a shared set of correlations between thoughts and signs is the vehicle of that transference. In subscribing to these assumptions, Harris argues, linguistic science commits itself to the belief that there are, in fact, languages; that different languages are distinct, autonomous structures existing outside the activity of situated human beings; that linguistic signs are intrasystemically related, fixed dualities of form and meaning; and that languages are capable of being contained in dictionaries and visualized in scripts. Harris (1988a: 113) labels this view of language "segregational", in the sense that it assumes the possibility of a strict segregation between linguistic and nonlinguistic phenomena within the universe of human activity. More prosaically, it assumes that human linguistic behaviour can be separated out from accompanying non-linguistic behavior, and treated independently. In segregationism, then, languages are self-enclosed, rule-governed, synchronic systems of contextless form/meaning dualities. In Saussure's version, for example, segregationism has five essential features: 1) centrality of the sign as a unit, 2) strict bi-planarity, 3) self-containment, 4) finite, unique itemisation, and 5) supra-individuality (Harris 1980: 155). But segregationism shows up in other places, too; for example, in the role accorded the dictionary in Western culture. For Harris (1980: 133), the dictionary is a key artifact of the segregational belief that words stand determinately for ideas and things external to the language itself: [B]y exhibiting each word as an established item with its own identity, the dictionary effectively discouraged its users from seeing a language as consisting in a form of continuous activity. It gave visible embodiment to a distinction between the word and its use. Words were units somehow having their own static and separate existence from the ongoing course of human affairs... A language thus came to be seen as constituting, in principle, a finite system of elements at any given time, and the psychological foundation was laid for all modern forms of structuralism.
ETHNOMETHODOLOGY AND INTEGRATIONAL LINGUISTICS
185
Wittgenstein's notion that languages are self-enclosed, autonomous games and Chomsky's belief that linguistic performance presupposes abstract linguistic competence would also be segregational by Harris's standards. In contrast to these approaches, Harris (1987a [Love (ed.) 1990: 199]) proposes that we think of language as "manifested in a complex of human abilities and activities that are all integrated in social interaction, often intricately so and in such a manner that it makes little sense to segregate the linguistic from the non-linguistic components". An integrational linguistics, then, recognizes that "human beings inhabit a communication space which is not neatly compartmentalized into language and non-language" and which takes as its point of departure "the individual linguistic act in its communicational setting" (Harris 1981a: 166). The ontological primacy accorded the sign in orthodox linguistics is thus transferred to communication in integrational linguistics (Harris 1993d: 1). In other words, the sign is no longer given in advance of the communicative situation "but is itself constituted in the context of that situation by virtue of the integrational role it fulfils" (Harris 1990c: 45). The alternative to segregationism is the attempt to understand, as Austin (1975: 148) urged, "the total speech act in the total speech situation". Language is seen to be "continuously created by the interaction of individuals in specific communication situations" (Harris 1981a: 167). So, rather than explain communication through the postulation of context-free linguistic invariants, Harris (1987a [Love (ed.): 206-7]) proposes that we try to understand how particular utterances are "communicationally integrated in a particular continuum of activity". As Love (1990b: 20) claims, all utterances for Harris appear in a context, and that context "must be assumed to be, not merely potentially relevant to the message, nor even always and necessarily relevant to the message, but, quite simply, an integral part of the message". Unfortunately, from an integrational perspective, most models of linguistic analysis suffer from segregationism. Even sociolinguistics involves the implicit admission that the study of linguistic communities is only one subdivision of linguistic theory, whose "disciplinary heart" is still the study of language structure (Harris, 1987a [Love (ed.) 1990: 199]). If we share, then, Harris's (ibid.: 202) desire to "integrate linguistics into the general study of communicative behaviour", an important question to consider will be: on what methodological basis would such studies be carried out? Ethnomethodological conversation analysis One answer has been offered by ethnomethodology. Originating in the United States in the 1960s as "alternate sociology", ethnomethodology survives today
186
DAVID FLEMING
in, among other things, the theoretical and methodological roots of "conversation analysis".2 The term "ethnomethodology" was first used by Harold Garfinkel to index the study of everyday practical reasoning in human activity. In his earliest investigations, Garfinkel (1967) had been interested in the ways people used instructions, laws, principles, rules, theories, and the like to reason their way through the contingencies of mundane social situations. He studied, for example, the methods that jurors used to make sense of legal cases, the procedures employed by coroners to determine cause in equivocal deaths, and the ways psychiatry students used coding instructions to categorize patients. What Garfinkel found was that, rather than simply applying pre-given abstractions to the particular situations in which they found themselves, participants managed specific cases locally and internationally, through what he called "indexical" and "reflexive" means. The reason, order, and coherence accomplished on such occasions was "accountably" (i.e., publicly and verifiably) displayed within the occasions themselves, through the ad hoc, temporally-constituted, practical actions of the participants. In other words, the social structure that Durkheim and Parsons had posited as a thing Garfinkel saw as a, process, achieved within the socially organized occasions of its use. Once sociology is thus reconstituted, informal conversation - seen in most structuralist accounts as merely epiphenomenal - becomes a key object of inquiry. As Zimmerman and Boden (1991: 3) argue, "Talk - more precisely, talk-in-interaction - provides the fundamental framework of social interaction and social institutions". Talk also conforms to the twin ethnomethodological principles of indexicality and reflexivity; that is, conversation is not merely about actions, events, and situations, it is "a potent and constitutive part of those actions, events and situations" (Potter and Wetherell 1987: 21, emphasis in original). Individuals interpret what their conversation partners say in light of the context in which it is said; at the same time, they formulate through talk the very social order that they use to interpret further interactions. Conversation is thus a locally-managed, joint enterprise which both reflects and produces social structure. The ethnomethodological conversation analyst, then, shares with the integrationist the belief that language is first and foremost activity, inextricably contextualized and time-bound. Drew and Heritage (1992: 17) argue that it is through this "activity focus" that conversation analysis distinguishes its treatment 2
Taylor and Cameron (1987) write that the field of ethnomethodology is now often referred to simply as "CA" (99). It is important to remember, however, as Moerman (1988) argues, that no field has a monopoly on either the phrase or the activity of "conversation analysis". For my understanding of CA, I have relied on the work of Emmanuel Schegloff, John Heritage, Paul Drew, Stephen Levinson, Michael Lynch, and others.
ETHNOMETHODOLOGY AND INTEGRATIONAL LINGUISTICS
187
of interaction and language from other theoretical orientations. "Conversation analysis", they write begins from a consideration of the interactional accomplishment of particular social activities. These activities are embodied in specific social actions and sequences of social actions. Thus, the initial and overriding conversation analytic focus is on the particular actions that occur in some context, their underlying social organization, and the alternative means by which these actions and the activities they compose can be realized. (emphasis in original) The "underlying social organization" that Drew and Heritage refer to is thus no longer seen as existing prior to particular interactions and activities. Instead, as Garfinkel (1991: 11) wrote, order is "every society's locally, endogenously produced, naturally organised, reflexively accountable, ongoing, practical achievement". Such an orientation is at the heart of ethnomethodology's foundational critique of structural sociology. In the work of Durkheim and Parsons, social structure had been something that caused social interaction. It was a collective phenomenon which provided a pre-existing social order to human activities. But in conversation analysis, as Zimmerman and Boden (1991 : 19) argue, social structure is not something "out there" independent of members' activities, nor are the structures of social action located at the unobservable level of Durkheim's collective. Rather, they are the practical accomplishment of members of society. Members can and must make their actions available and reasonable to each other and, in so doing, the everyday organization of experiences produces and reproduces the patterned and patterning qualities we have come to call social structure. The organization of talk displays the essential reflexivity of action and structure and, in so doing, makes available what we are calling structure -in-action. The ethnomethodological refusal to posit pre-existing, abstract structures to explain human action also implicates cognitive theories of language because, just as social structures are no longer "out there", they can no longer be "in here" either. Ethnomethodological researchers like Coulter (1991) and Lee (1991) see this as a neo-Wittgensteinian attempt to deny underlying mental structures or grammars as the basis of logic and understanding. As Coulter (1991: 30) writes, "it is in the domain of ordinary, practical, social affairs where rules for using linguistic symbols, signs, and concepts are to be revealed in operation".
188
DAVID FLEMING
So, just as integrationists refuse to see language as meaningful apart from the occasions of its use, conversation analysts hold that there can be no determination of social behavior independent of local interaction. Lucy Suchman's (1987: 179) ethnomethodological claim that "the contingence of action on a complex world of objects, artifacts, and other actors, located in space and time, is...the essential resource that makes knowledge possible and gives action its sense" would therefore appear to mirror Harris's (1988a: 113) Wittgensteinian view that "language has no segregated existence; words are always embedded in a 'form of life'.. .inextricably integrated into purposeful human activities". Language, Harris (1987a [Love (ed.) 1990: 208]) writes elsewhere, involves not just vocal behaviour but many kinds of behaviour, and to engage in face-to-face linguistic communication is, in the simplest type of case, to comonitor with one other person a behavioural continuum along which a succession of integrated events can be expected to occur. Harris (1980: 46-7) himself has written that ethnomethodology, although a powerful critique of certain conventional views of language, falters by assuming that it can, by attending to the local management of meaning, "repair" the indeterminacy of language. Ethnomethodology is thus heir to traditional scientific and philosophical prejudices against language itself (Harris, 1981a: 22-23). Unfortunately, Harris nowhere else engages the ethnomethodological project; and ethnomethodologists have in turn ignored his "integrational" challenge. This lack of mutual attention is especially unfortunate given the (at least superficial) similarities between the two approaches.
2. Language from the user's point of view In addition to a shared orientation to language, conversation analysis and integrational linguistics are also united in certain key methodological assumptions. For example, both claim as their analytical territory language use in naturally occurring social interaction. Both attempt, in other words, to study the situated behavior of real individuals in their everyday activities, a goal which, for Lee (1991: 217), amounts to a rejection of decontextualized linguistic analysis in favor of a discourse analysis which attempts systematically "to integrate 'language' and 'the social'". This is closely related, I believe, to Harris's (1987a [Love (ed.) 1990: 199]) argument that linguistic intercourse never is and never can be a form of human behavior which is sui generis.
ETHNOMETHODOLOGY AND INTEGRATIONAL LINGUISTICS
189
A shared methodological orientation towards language in social use may also explain a shared analytical focus on the interactants themselves in investigations of social structure, a clear move away from the social institutions of Durkheim and Saussure and the ideal speaker-hearer of Chomsky. Just as Harris (1980: 3) has written about language that "the language-user already has the only concept of a language worth having", ethnomethodologists claim that, in studying social interaction, the interactants' own understanding - through their sequential reaction to the moves of the conversation - is the only empirical basis for inferring social structure in the events analyzed. Ethnomethodology becomes, then, the study of lay methodologies, "the body of common-sense knowledge and the range of procedures and considerations by means of which the ordinary members of society make sense of, find their way about in, and act on the circumstances in which they find themselves" (Heritage 1984: 4). For Garfinkel (1991: 17), order, meaning, and reason are the accomplishments of members themselves, through their joint, everyday activities. Interest in such lay rationality can be traced, in part, to Alfred Schutz's Verstehen sociology and its concern for how people interpret their own social lives (Garfinkel 1967; Heritage 1984; T.J. Taylor 1992). In this view, social actors are not "dupes" of deterministic social structures but are instead knowing and acting agents who co-construct meaningful social structure in an ongoing and practical way. In Lynch's (1985: 6) ethnomethodology of laboratory work, this principle manifests itself in studies which attempt "to situate themselves within the settings of work they describe", taking advantage of what Garfinkel had called the publicly "accountable" display by the members themselves of their own rational procedures for accomplishing social life. Likewise, Levinson (1983: 295) claims that "the categories of analysis should be those that participants themselves can be shown to utilize in making sense of interaction". And, in their classic study of conversational turn-taking, Sacks, Schegloff, and Jefferson (1974: 729) write that the ways that interactants themselves respond to one another's turns reveals the very social structure that they are managing: Since it is the parties' understandings of prior turns' talk that is relevant to their construction of next turns, it is THEIR understandings that are wanted for analysis. The display of those understandings in the talk of subsequent turns affords both a resource for the analysis of prior rums and a proof procedure for professional analyses of prior turns - resources intrinsic to the data themselves, (emphasis in original) So, just as Harris (1980: preface) proposes that we think of human beings as language-waters' rather than language-users, ethnomethodologists reject the view
190
DAVID FLEMING
of human agency as simply the playing-out of pre-determined abstractions, language-users being nothing more than "servants bearing tureens of linguistic signs into the great dining halls of life" (Billig 1991: 15). The commitment to the language user's own perspective on language, meanwhile, also corrects, according to Harris, the scientific denigration of ordinary language. In orthodox Western philosophy, words are misleading, imprecise, and hopelessly indexical - that is, inextricably and unfortunately tied to the accidental circumstances of their use. Philosophers and scientists have been concerned, therefore, with finding a "more soundly based descriptive system" with which to acquire knowledge (Harris 1981a: 20-1). For the integrational linguist, however, the flight from ordinary language is a flight from social reality. Harris (1990c: 51) writes [T]he strategies and assumptions people bring to bear on the communicational tasks of daily activity, tasks they are obliged to deal with by whatever means they can, are all an integrational linguistics needs to study in order to advance our understanding of what language is and the part it plays in our lives. Orthodox linguistics joined the philosophical fight against ordinary language, Harris argues, when Saussure abandoned the historian's perspective on language but then failed to adopt the language-user's perspective in its place. A genuine language-user's perspective, Harris (1981a: 163) argues, leads us to ask (with the conversation analyst as well), "[W]hat do the speaker and hearer have to do in order to integrate speech relevantly into a temporal flow of episodes which they are jointly co-monitoring?"
3. Data and theory in the analysis of language Belief in the local, interactional, and temporal constitution of structure is methodologically manifest for both the integrational linguist and the conversation analyst in a commitment to unremittingly empirical descriptions of what people actually say and do in social interaction. For Harris (1981a: 164), the goal of integrational linguistics is the yielding of statements which both "correspond to the language-user's experience and are open to the verification and disproof characteristic of the empirical sciences". Likewise, ethnomethodologists often refer to their work as a much needed "respecification" of social behavior whose goal is the description of "detailed, embodied, situated action and interaction" (Button 1991: xii). On the matter of social data, then, Harris and the ethnomethodologists would appear to be in agreement: proper social-scientific
ETHNOMETHODOLOGY AND INTEGRATIONAL LINGUISTICS
191
investigation begins with actual social behavior, viewed from the point of view of the participants themselves, and treated in an insistently empirical manner by the analyst. On closer inspection, however, the two accounts diverge in fundamental ways. For the ethnomethodologist, order is found in the recurrent accomplishment of ongoing, practical social life, and this entails study of the smallest and most mundane interactions. In such research, a phenomenon like conversational turn-taking is not an abstract structural order imposed from outside on our actions and interactions; instead, it is the very site - local, sequential, and practical - of social structure. But the insistently empirical stance accompanying this view is not just an attribute of the analyst. According to Garfinkel, the participants in social interaction are also empirically motivated - they publicly display to one another, in verifiable fashion, their own ongoing interpretations of the interaction. That is, interactants themselves "prove", by the sequentiality of their contributions, that the interaction is socially ordered. As Sacks, Schegloff, and Jefferson (1974: 728-9) put it: Herein lies a central methodological resource for the investigation of conversation...a resource provided by the thoroughly interactional character of conversation. It is a systematic consequence of the turn-taking organization of conversation that it obliges its participants to display to each other, in a turn's talk, their understanding of other turns' talk. What ethnomethodologists believe they are analyzing, then, is a social reality that is "visible" to participants and analysts alike. Enter the portable tape-recorder. Audio-recorded mundane conversations, "neither idealized nor recollected retrospectively" (Heritage 1984: 235), can be listened to repeatedly, examined in all their detail, and shared with interested others. With the portable tape-recorder, interaction data become available "in raw form, neither idealized nor theory-constrained" (Heritage 1984: 238; see also Moerman 1988: ix). For Drew and Heritage (1992: 5), the emphasis on "recorded conduct" is a corrective to conventional sociology's indirect access to social action through questionnaires, interviews, and self-reports. Indirect accounts, they argue, can never escape the problematic gap between what people say they do and what they actually do. By contrast, conversation analysts use audio and video records to reveal how particular interactions are "enacted and lived through as accountable patterns of meaning, inference, and action". The audio- or video-recording and the transcript thereof thus play a central role in CA inquiry. But for what is by origin a lay methodology, conversation analysis is curiously beholden to the technological mediations of recording equipment and
192
DAVID FLEMING
the cumbersome transcription notations that translate speech events into print. As Edwards (1993), Ochs (1979), and others have reminded us, the transcription of speech is always a theory-laden enterprise, despite the presumption by conversation analysts that they are looking at "raw", "visible" social life. The distance travelled, therefore, between the lay experience of conversation and the analyst's interpretation of conversation is hardly inconsequential. An exclusive concern for informal conversation is another point of principle for the ethnomethodological conversation analyst. Heritage (1984: 238) writes that "the central domain of data with which conversation analysis is concerned is everyday, mundane conversation". Talk which is highly intellectual in content or "specifically fateful" for one or another of the participants is purposefully avoided. The primacy of mundane conversation is supported by the pervasiveness of conversation in the social world at large and the role of such conversation in introducing children to that world. Heritage (1984: 239) argues further that "institutional interactions" (for example, the decidedly non-trivial exchanges of the courtroom, doctor's office, or classroom) are simply reductions or concentrations of "the base environment of ordinary talk". Levinson (1983: 284) agrees: informal conversation is "clearly the prototypical kind of language usage". Sacks, Schegloff, and Jefferson (1974: 722) describe conversation as "a major, if not THE major locus of a language's use" (emphasis in original). And for Drew and Heritage (1992: 19), ordinary conversation or "mundane talk" explicitly associated here with casual conversation between peers - is the predominant medium of social interaction, constituting a kind of benchmark against which other more formal or institutional types of interaction are experienced. It might be argued, then, that comparisons between Harris and the conversation analysts are invalid because Harris's theory concerns language in general while CA is concerned with only one part of human language - mundane conversation. But, for the conversation analyst, the focus on talk is not simply a choice made by the researcher. In ethnomethodology, talk is the key fact of social life and is at the center of any theory of situated action. The assumption of much conversation analysis, in other words, is that the study of mundane conversation is the study of social life. In fact, conversation analysts argue, the more mundane the conversation, the more obvious will be the local and practical accomplishment of everyday social order. One well-known example of CA research, for example, is a fifty-page account of telephone conversation openings (Schegloff 1979). Billig (1991: 16), for one, sees the tendency to concentrate on such seemingly insignificant interactions of social life to be both asset and liability. The ethnomethodological impatience with abstraction, he writes, leads to highly detailed examinations of the ways language is actually used in real social occasions. But the restriction of
ETHNOMETHODOLOGY AND INTEGRATIONAL LINGUISTICS
193
such study to trivial conversations can also be "puritanical". Billig himself does ethnomethodologically influenced discourse analysis on non-conversational data in which subjects are interviewed about highly significant and controversial issues. He admits that many recent conversation analyses are also breaking through the earlier restriction to mundane conversation; but, he writes, they are still more interested in the "forms of talk" than in their content. Lynch's (1985: 8-9) ethnomethodology of science, for example, distances itself from the lack of substance in traditional conversation analysis but then embraces the "general relevance" of abstract conversational architecture (the sequential design of turn-taking, for example). As for Harris, he nowhere restricts the analysis of human communication to conversational data. In fact, Harris has shown a sustained concern for the role of writing in human communication, a subject largely ignored by the ethnomethodologist. Having collected the mundane data on which he or she depends, what does the conversation analyst do next? One thing he or she doesn 't do is turn to theoretical abstractions. Button (1991: 4-5) writes of ethnomethodology that it never bought into the business of theorising, it was iconoclastic, it would not theorise foundational matters... [I]t wished to make them investigable, available for enquiry. In holding them up for scrutiny, ethnomethdology came to respecify foundational matters. According to Heritage (1984: 235), the CA respecification of social data can be traced to Sacks's impatience with the averaging and idealization of interaction specifics in conventional sociology; he deliberately sought therefore "a method of analysis which would keep a grip on the primary data of the social world, the raw material of specific, singular events of human conduct". For Lee (1991: 224), such "respecification involves...suspending general questions, such as the question of the relationship between 'culture' and 'language' until these have been described with respect to the question of how they translate into the witnessable understandings and activities of social interactants". Conversation analysts thus put aside questions concerning the connection between conversational interaction and the larger contexts of social action as "premature and unprofitable diversions for the central task of discovering and describing the organization of conversational interaction as such" (Zimmerman and Boden, 1991: 3). Heritage (1984: 242) describes this as the conversation analyst's "retreat from premature theory construction... an avoidance of abstract theoretical constructs". Similarly, Levinson (1983: 295) praises CA's "healthy suspicion of premature theorizing and ad hoc analytical categories".
194
DAVID FLEMING
But the ethnomethodological re-invigoration of empirical research against the excesses of late twentieth-century theorizing may harbor an over-abundant faith in the social reality captured by CA transcripts. When Sacks, Schegloff, and Jefferson (1974: 699) write that their conversational material, "accessible to rather unmotivated inquiry, exposes the presence of turn-taking and the major facets of its organization" (emphases mine), they seem to imply that the meaning of such data is self-evident. This assumption may undermine the ethnomethodologist's own goal of contextualizing social behavior. If all the analyst needs is some record of the conversation itself, the transcript of a mundane telephone conversation, for example, then, ironically, the context doesn't matter that much after all. It is as if the decontextualized behavioral record makes accessible to participant and analyst alike the visible, cross-situational organization of human conversation. And when Levinson (1983: 295) writes that "the emphasis is on the data and the patterns recurrently displayed therein" (emphasis mine), he seems to imply that, just as interactants themselves "display" their rationality to one another through their various public actions and reactions, so the rationality of conversation is unproblematically available to analysts. In other words, the ethnomethodological project might be said to begin with a rejection of independent, abstract, theoretical constructs but end ultimately with their re-affirmation in the form of interactional abstractions like "the turn-taking system" and the "adjacency pair". The actual content of particular conversations, certainly relevant to any account of context, is curiously expendable. Harris's students Talbot Taylor and Deborah Cameron (1987) have demonstrated in their deconstruction of an analysis by Levinson (1983) that the data of conversations may not be as self-evident as ethnomethodologists would like to believe. What Levinson sees as "visible" in the conversation, as empirically evident in the interactants' own behavior - in this case, a conversational pause interpreted as a "dispreferred" response to a request - may require more presuming on his part than he has imagined. Talbot Taylor (1992) has written elsewhere of the ways ethnomethodologists avoid the intellectual problems of human communication by positing entirely practical solutions. Rigorous theoretical speculation about intersubjectivity - the central plank in the ethnomethodological platform - is by this device effectively sidestepped. Mutual understanding is so obvious to the ethnomethodologist as to require neither abstract theorising nor negative scepticism; it is a voluntary, practical, moral accomplishment of the interaction itself, necessitating no "in-principle" explanation, and discernible as a matter of course. The "normative turn-by-turn implementation of adjacency-structuring of conversational action" is thus both proof of mutual understanding and an explanation of how such understanding occurs (T.J. Taylor 1992: 224). As Taylor and Cameron (1987: 107) write, there
ETHNOMETHODOLOGY AND INTEGRATIONAL LINGUISTICS
195
is no need to impose an analysis on the data because "the conversation itself wears its (or the participants') own inherent ('emic') analysis on its sleeve". For Harris, highly sensitive to the culturally conditioned assumptions behind all our theories, data can never speak for themselves. In his writings, historical and theoretical issues often crowd out the particularities of local data. But this is not merely a matter of preference, as if one could either theorize or specify. From the perspective of Harris, we are always language theorists whether we like it or not. There is no such thing as self-evidence in the analysis of talk, and it is a delusion to believe that, as analysts, we can postpone larger questions while we work out the particularities of smaller ones: To have a concept of a language necessarily involves going beyond mere understanding of what the acceptable verbal moves are, even in the case where what is claimed is that to know a language is just to know the acceptable verbal moves. A concept of a language involves, and is most often clearly manifest in, acceptance or rejection of what requires explanation about the ways in which languages work. This means that a concept of a language cannot stand isolated in an intellectual no-man's-land. It is inevitably part of some more intricate complex of views about how certain verbal activities stand in relation to other human activities, and hence, ultimately, about man's place in society and in nature. (Harris 1980: 54) Integrational linguistics, then, is everywhere concerned with the inescapably cultured nature of our thinking about language. Just as language itself cannot be abstracted away from temporally-constituted practical, social activity, neither can thinking about language be segregated from the culture in which such thinking is always an inheritance.
4. Structure
and occasion in social
action
In both integrational linguistics and ethnomethodological CA, I have argued, language is seen as contextualized, time-bound activity. I would now like to look more closely at the conversation-analytic notion of "structure-in-action" and assess whether such an account squares with Harris's integrationism. Zimmerman and Boden (1991: 5) hold that, although the conventional sociological view of structure as social abstraction (gender, for example) is rejected by conversation analysts, the concept of structure itself is retained. They define structure as "regular, repetitive, nonrandom events that stand in a systematic relationship to one another"; and, drawing on the methodological
196
DAVID FLEMING
principles introduced earlier - that analysis of social interaction take place as much as possible through insistently empirical investigations from the point of view of the participants themselves - they characterize structure as accomplished in and through the moment-to-moment turn-taking procedures of everyday talk in both mundane and momentous settings of human intercourse. The instantiation of structure is, moreover, a local and contingent matter, one that is endogenous to interaction and shaped by it. The enabling mechanisms of everyday talk thus practically and accountably accomplish structure-in-action. (Zimmerman and Boden 1991: 17) The order of social life is thus reproduced in each interaction. In other words, for the ethnomethodologist, structure is a procedural and not a substantive phenomenon, a notion seemingly compatible with Harris's rejection of hypostasized linguistic entities. Several empirical claims follow from the view that structure is always structure-in-action. For one, individuals would be seen to make their way through social life (that is, to make society rational) in an insistently ad hoc fashion. Garfinkel (1967), for example, argued for the necessarily ad hoc nature of actualizing any set of abstractions, rules, or theories. Rather than treat codes as autonomous, deterministic instructions, applied to diverse particulars, he held that consideration of particularity is what gives sense to abstraction. As with Wittgenstein, rules are an insufficient explanation or determinant of human action; to account for usage we must see, in every instance of use, separate, specific, local, contingent determinants (Heritage 1984: 120-1). For the ethnomethodologist, then, talk is understood not by reference to a pre-existing semantic or syntactic calculus but through the very social situations which give particular utterances their meaning and which are in turn constructed by those utterances. In other words, the order we achieve in our social lives is an order constituted in time, in social interaction, and in particular contexts. Conversational structures are therefore best seen not as invariant abstractions but as patterns of situated activity. According to Levinson (1983), for example, conversations are organized 1) by a turn-taking system that results in the orderly transition of speakers in natural social interaction; 2) through adjacency pairs in which the first part (a question, for example) sets up certain expectations for the second (an answer); and 3) by an overall structure that would include such things as openings, topic slots, and closings. And from more recent work on "institutional" talk, conversations in doctor's offices and courtrooms are seen to be structured by such mechanisms as the pre-allocation of turns (for example, of questions to
ETHNOMETHODOLOGY AND INTEGRATIONAL LINGUISTICS
197
institutional participants and answers to lay participants) and a pervasive professional cautiousness (Drew and Heritage 1992). But note that, for the conversation analyst, these patterns and typifications are meaningful only as accomplished, that is, they make sense only as the participants themselves enact them in their everyday activities. So, rather than assume that the meaning of a particular utterance can be fixed through correspondence to a pre-existing structure (a dictionary definition, grammatical rule, social convention, etc.), conversation analysts hold that the meaning of an utterance makes sense only as actual speakers and hearers accomplish the utterance in a sequence of interactions unfolding in a particular situation. One might say, then, that, for the conversation analyst, context is all; that context is a temporal, interactional, and local construct; and that conversational acts and their embedding context are reciprocally related. As Heritage (1984) puts it, any particular communicative act is doubly contextual: it is at once context-shaped and context-renewing. In other words, an interactional event cannot be understood except by reference to its context, including the immediately preceding events of the interaction. At the same time, every action contributes to the terms by which the next action is understood. Thus, an utterance in a conversation is "shaped" by the context in which it occurs; simultaneously, it "renews" the context by which future utterances will be interpreted. The work on conversational turn-taking by Sacks, Schegloff and Jefferson (1974) might serve as an example here. In that paper, the authors postulate a form of communicational structure that is enacted through time by the joint effort of situated, social interactants. Their research suggests that informal conversation involves an orderly allocation of turns such that only one person speaks at a time and there is a finely coordinated transition between speakers with little overlap and only barely discernible gaps. Orthodox accounts of communication have tended to treat such interaction, when they have treated it at all, as the product of structured systems of units and rules, existing either inside each participant's head, as abstract linguistic competence, or in a shared cultural system. In other words, analysts have tended to treat conversational structure, when they treat it at all, as simply another abstract variable ontologically prior to the communicative event itself. Drew and Heritage (1992) call this the "bucket theory of context", in which the social framework "contains" local action. The ethnomethodologist, on the other hand, is interested in what Schegloff (1991) calls the "procedurally consequential" aspects of context. So, as Sacks, Schegloff, and Jefferson argue, analysts cannot afford to hypostasize communicational structures. As perhaps the single most important feature of informal conversational behavior, turn-taking only makes sense as accomplished over time through the joint action of
198
DAVID FLEMING
coordinated individuals, that is, as a sequential, interactional, and thoroughly local affair. Conversational structure is thus not substantive but procedural, not synchronic but temporal, not static but ongoing, not an attribute of individuals but the joint activity of interactants. Language here is firmly integrated into social action, and this would seem to be perfectly suited to Harris's proposals for re-defining what it means to study language. Or is it? Although conversation analysts like Sacks, Schegloff, and Jefferson claim that structure (the organization of turn-taking in mundane conversation, for example) is always structure-in-action (a sequential, interactional, and local affair), there are still hypostasized entities at work. The postulated system may be different from the abstractions of Saussure, Durkheim, Parsons, and Chomsky, but the system is still there; and every time we want to say that this is a thoroughly situated account of human communication, fully sensitive to the necessarily sequential, interactional, and local nature of language, we come up against cross-situational abstractions. In the Sacks, Schegloff, and Jefferson piece, for example, this paradox is particularly noticeable. Early in the article, the authors claim that the turn-taking organization of conversation is both "context-free and capable of extraordinary context-sensitivity" (1974: 699). Because this distinction is important in assessing the CA view of the structure/occasion binary, we should look closely at their discussion: [A] problem for research on actual conversation is that it is always 'situated' always comes out of, and is part of, some real sets of circumstances of its participants. But there are various reasons why it is undesirable to have to know or characterize such situations for particular conversations in order to investigate them. And the question then becomes: What might be extracted as ordered phenomena from our conversational materials which would not turn out to require reference to one or another aspect of situatedness, identities, particularities of content or context? One reason for expecting the existence of some such type of organization is as follows. Conversation can accommodate a wide range of situations, interactions in which persons in varieties (or varieties in groups) of identities are operating; it can be sensitive to the various combinations; and it can be capable of dealing with a change of situation within a situation. Hence there must be some formal apparatus which is itself context-free, in such ways that it can, in local instances of its operation, be sensitive to and exhibit its sensitivity to various parameters of social reality in a local context.... ... Depiction of an organization for turn-taking should fit the facts of variability by virtue of a design allowing it to be context-sensitive; but it should be cast in a manner that, requiring no reference to any particular context, still captures the
ETHNOMETHODOLOGY AND INTEGRATIONAL LINGUISTICS
199
most important general properties of conversation, (ibid.: 699-700, emphasis in original) Thus, although Sacks, Schegloff, and Jefferson offer impressive evidence for the local, interactional, and temporal features of mundane conversational action, the situated aspects of communication are still "managed" (to use their word for it) by a "system" (again, their word) that can be unproblematically abstracted from the particular events it governs. In other words, communication is situated only on the surface; abstract structures external to the particular act drive that act in a way surprisingly compatible with the orthodox structuralism that ethnomethodologists have claimed to reject. What, for example, is the ontological status of "the system" (also referred to as "it") in the passage below? (1) The system deals with single transitions at a time, and thereby with only the two turns which a single transition links; i.e., it allocates but a single turn at a time. (2) The single turn it allocates on each occasion of its operation is "next turn". (3) While the system deals with but a single transition at a time, it deals with transitions: (a) comprehensively... (b) exclusively... and (c) serially... (ibid.: 725) It is this "system", I would argue, that makes any association of ethnomethodological conversation analysis with "integrational" accounts of language problematic. Rather than begin, as Harris (1981a: 166) proposes, with "the individual linguistic act in its communicational setting", Sacks, Schegloff, and Jefferson (1974: 699) impose on that act hypostasized invariants that exist independently of it. Their ultimate concern is not, as they themselves put it, "particular outcomes in particular settings", but "the organization of turn-taking per se" (emphasis mine). The decontextualized abstractions of structuralism are thus resurrected in full force. To this objection, conversation analysts would likely claim that they make no analytic distinction between structure and the occasions of its manifestation. But how then does one account for their reification of structure, of context-free "systems" like turn-taking, which, they argue, play such a powerful role in determining the shape of everyday social interaction? Despite, then, the ethnomethodological predilection for dealing with talk in action, there is a contradictory feature of such analyses that, at times, resembles the structuralist search for the rules behind social interaction, the normative structures which operate in social interactions to orient and guide interactants in their behavior. Goffman had written of Garfinkel's early work that he had attempted to "uncover the informing, constitutive rules of everyday life" (quoted in Taylor and Cameron 1987: 4) and that this attempt had followed the success of
200
DAVID FLEMING
Chomsky's search for the generative rules of language. For later ethnomethodological conversation analysts, the adjacency-pair structure seemed to provide just such a "normative framework" for verbal interaction (Taylor and Cameron 1987: 109). Cicourel's (1974) account of ethnomethodology ties it closely to generative linguistics, although for him the relevant rules are manifest not in syntax but in practical action. Taylor and Cameron (1987: 116-17) point out the "analytical circularity" of this feature of ethnomethodology - the notion that although conversation is an entity that is meaningful only in practical, local, situated, ongoing activity, the machinery that ultimately gives it structure is "relatively stable and context-independent". Zimmerman and Boden (1991: 8) are untroubled by this apparent paradox: [Talk] is organized by use of machinery deployed in and adapted to local contingencies of interaction across an immense variety of social setting and participants... The organization of talk provides the formal resources to accomplish these interactional tasks, and deploys these resources in a manner that is sensitive to just what circumstances and participants happen to be at hand - which is to say, locally. The shape of talk found in a specific site thus reflects the context-sensitive (and thus particularized) application of a more general, context-free (and thus anonymous) interactional mechanism. From an integrational perspective, then, ethnomethodology has simply substituted an objective abstraction called "conversation" for the previous one called "language". Though this may be an improvement over structuralist linguistics, it can never be more than a partial response to the call for a thoroughly integrational approach to communication. Talk cannot be separated from the other mental, physical, social, and cultural activities in which it is embedded and by which it is constituted, including the activity of analysis itself. Questions about setting, social roles, rhetorical occasion, technology, content, etc., cannot be suspended. What Zimmerman and Boden (1991: 8) call the ethnomethodological "treatment of conversation as a virtually autonomous domain" may then actually impede efforts to integrate the linguistic and non-linguistic.
5. Language
and
temporality
In a crucial sense, everything truly iconoclastic about both integrational linguistics and ethnomethodological conversation analysis comes down to how they account for the role of time in human action. In both, temporality is accorded
ETHNOMETHODOLOGY AND INTEGRATIONAL LINGUISTICS
201
a central position, in contrast to the predominantly detemporalized structuralism of conventional sociology and linguistics (see Hopper 1992, for a useful discussion of temporality in post-structuralism). For Harris, succession in time is the key to understanding the human experience of language. Because language is situated communicative action, Harris (1981a: 154-55) argues, language is always embued with a temporal stamp, and this provides "a unique contextualization" for "everything that is said, heard, written, or read". Progression through time therefore makes of language a continuously creative process. "Every linguistic act", Harris (1981a: 155) writes, "is integrated into the individual's experience as a new event, which has never occurred before and cannot occur again". In other words, once language is seen as temporalized action rather than detemporalized state, there can no longer be an unproblematic abstraction of language from the rest of human activity, and there can no longer be a strict segregation of linguistic from non-linguistic events. The chronological "parity" of linguistic and non-linguistic events Harris (1981a: 158) terms the principle of cotemporaiity. What Harris (1990c: 47) proposes, then, is neither the historically-oriented diachronic analysis that Saussure had rejected nor the synchronic linguistics that was proposed in its place, but instead a "panchronic" analysis which "considers as pertinent to linguistic communication both the integration of simultaneously occurring events and also the integration of present events with past events and anticipated future events". The crucial role played by temporality in Harris's theory cannot be over-emphasized: [T]he integrational perspective sees us as making linguistic signs as we go; and as having no alternative but to do this, because language is time-bound. For the integrationist, we are time-bound agents, in language as in all other activities. Whereas segregational semantics tacitly...presupposes an a-temporal linguistic arena, usually disguised as a static present (Saussure's synchronique), the integrationist accepts that there is no way we can step outside the time-track of communication. Once this is conceded, it follows that there is no such thing as a contextless sign. A sign cannot exist except in some temporally circumscribed context. That contextualization is a foundational condition of the very existence of the sign, and whatever falls outside it also falls outside the province of an integrational semiology... Nothing about language or languages - and, more generally, nothing about communication - is timeless. (Harris, 1993d: 9) In short, it is Harris's profound appreciation for the temporal dimensions of the human experience of language that allows for his devastating critique of linguistic structuralism.
202
DAVID FLEMING
Ethnomethodological conversation analysis takes a strikingly similar stance towards temporality. For the ethnomethodologist, the "architecture of intersubjectivity" which orders social life is a thoroughly sequential phenomenon. That is, human beings structure their rational lives together by publicly displaying, over time, their mutual understanding of the situation-underconstruction. In ethnomethodology, an action is always seen as a response to the action immediately prior to it, and any future action is necessarily construed in accordance with the context already established. Drew and Heritage (1992: 18), for example, write that "the sense of an utterance as an action is an interactive product of what was projected by a previous turn or turns at talk and what the speaker actually does". It is for this reason that, as Levinson (1983: 321) remarks, interactants display the order of social life both for one another and for the ethnomethodologist: "As each turn is responded to by a second, we find displayed in that second an analysis of the first by its recipient. Such an analysis is thus provided by participants not only for each other but for analysts, too". The structure of social life is thus an ongoing, publicly accountable, joint accomplishment. Garfinkel's emphasis on the temporal concerting of action, Heritage (1984: 108) claims, was a direct challenge to those theories of social action in which abstract social norms determined local conduct. In such a view, the scene of action is essentially pre-defined, and time is nothing more than a "fat moment". In the ethnomethodological alternative, norms become "reflexively constitutive of the activities and unfolding circumstances to which they are applied"; and social structure is only available through "temporally extended sequences of action" (Heritage 1984: 108). Schegloff (1992: 125) describes this as a kind of serialization of context: There is, to my mind, no escaping the observation that context, which is most proximately and consequentially temporal and sequential, is not like some penthouse to be added after the structure of action has been built out of constitutive intentional, logical, syntactic, semantic, and pragmatic/speechact-theoretic bricks. The temporal/sequential context rather supplies the ground on which the whole edifice of action is built (by the participants) in the first instance, and to which it is adapted "from the ground up". Zimmerman and Boden (1991: 9) agree: For conversation analysis, the varieties of sequential organization - the turn-taking system for managing the construction, allocation of turns at talk, sequences for entry into and exit from conversation, and for the repair of
ETHNOMETHODOLOGY AND INTEGRATIONAL LINGUISTICS
203
trouble or for doing invitations, requests, assessments, and the like - provide the structure for conversational encounters, and talk-in-interaction more generally. As a consequence, the sequential environment of talk provides the primary context for participants' understanding, appreciation, and use of what is being said, meant, and most importantly done in and through the talk. Thus, in Sacks, Schegloff, and Jefferson (1974: 722), the turn-taking system is a system for allocating turns in a series. Each turn is seen as having a three-part structure: "one which addresses the relation of a turn to a prior, one involved with what is occupying the turn, and one which addresses the relation of the turn to a succeeding one". Similarly, for Lynch (1985: 53), ethnomethodology of work is inquiry into "the serial ordering of tasks in the immediacy of an organizational setting". At first glance, then, it would seem that we have a powerful convergence of integrational linguistics and conversation analysis regarding the fundamentally temporal constitution of social life. But there is an important distinction here. In conversation analysis, time has been reduced to a sequencing mechanism. It orders social interaction the way a clock governs life in a factory: it is dominating and relentless. The ethnomethodological position seems to be that not only are conversational turns inescapably connected to the turns which precede and succeed them, but that this is the only way that social action is temporalized. Time therefore becomes another objectified abstraction. In fact, for the ethnomethodologist, time is the one structural element that is denied a reciprocal relationship with occasion. Time constructs local action but is itself unconstructed; it is beyond agency and beyond language. For Harris, temporality is irreducible to sequentiality. Harris's principle of co-temporality ensures that the integrational position is not that language activity is sequential but that linguistic and non-linguistic time are chronologically integrated. Such parity implies both sequentiality and simultaneity. Linguistic acts progress through time, but they are also integrated into the temporal progression of other, simultaneously occurring, acts. For Harris, then, the focus on temporality becomes a way to stress that language is situated, social action, inextricably contextualized and irreducible to static objectivities. An interesting temporal fact about language, in such a view, is creativity, in the sense that we "make" language as we go. And as we "make" language, non-linguistic things happen as well; ideology, for example, might be said to progress simultaneously with linguistic time but still be largely hidden if one looks only at the transcription of a few conversational moves. Narrative devices are another temporal "fact" of communication irreducible to strict sequentiality. Frentz (1985: 7), for example, distinguishes "conversations which structure time as an emergent historical unity
204
DAVID FLEMING
from conversations in which time is experienced as a linear sequence of temporal units". In the latter, participants experience time "on the encounter level", as the sequencing of chronological units. In the former, however, they experience time "on the form of life level", talk serving to fuse past and future in the present. In such conversations, individuals place themselves self-consciously in narrative contexts, acting out stories in which past and potential conversations emerge. In other words, time here is not so much a constructor of social reality as it is itself constructed as part of social reality. Schegloff s (1991) notion that context is only interesting when "procedurally consequential" might be a way, then, to deny the impingement of such non-sequential temporalizations of human communication. There is another way that temporality for Harris defies reduction to sequentiality. Language analysts are also time-bound, and their analysis of language is itself another linguistic act. In other words, temporality contributes to the constitution of language, but language is used to constitute temporality as well. We construct, for example, the "time" of past events; they are not simply given to us in the raw records of human behavior. This reflexivity is lost upon the ethnomethodologist, for whom the sequential unfolding of conversation, as represented in a written transcript, is a visible fact, unproblematically available to all.
6. Conclusion:
Towards an integrational
rhetoric?
My purpose in this paper has been to contribute to the methodological re-configuration of language study. Setting up as a worthy goal Roy Harris's integration of linguistic phenomena into the practical continuum of human activity, I have sought to judge how one approach to the study of language, ethnomethodological conversation analysis, measures up to that goal. I hope to have shown that CA and integrational linguistics share many goals and some methodological tactics. The two accounts converge in their insistence that communication analysts begin with data from naturally-occurring social behavior and that such data be viewed as much as possible from the perspective of the interactants themselves. And they share the belief that language is not a finite, static thing-, it is instead an activity, situated in time, and understandable only as contextualized in the particularities of actual social action. These are important commonalities that set the two accounts apart from more orthodox studies of social life. But there are important differences between the CA and integrational approaches to language. One finds in most ethnomethodological conversation analysis, for example, an exclusive concern for informal, mundane conversation,
ETHNOMETHODOLOGY AND INTEGRATIONAL LINGUISTICS
205
an often suspicious stance towards theoretical reflection, a recurring appeal to formal mechanisms organizing social action, and a reduction of temporality to sequentiality. These are tendencies, I would argue, that preclude the kind of radical contextualization needed in language studies. The integrational linguistics of Roy Harris, I believe, avoids these problems. Unfortunately, Harris's work, while devastating in its critique of conventional linguistics, has failed to produce the kind of "positive" methodology employed by conversation analysts. If a useful goal for the integrational study of language is empirical (albeit non-structuralist) analyses of situated discourse, and if ethnomethodological conversation analysis is found lacking as an integrational approach, then we are left pretty much where we started. Let me close, then, by briefly outlining what an integrational methodology might look like, what I call here an integrational rhetoric to further distance the non-structuralist study of language from modern linguistic science and take advantage of the centuries-old rhetorical tradition, where language has never been meaningful except as actually used in practical, situated, social interaction. The turn to rhetoric here cashes in on the more general turn to rhetoric in much late-twentieth-century philosophy and language theory. From such a perspective, language is approached not as an abstract structured, rational, or formal system. It is neither supra-individual nor primarily mental. Rather, language is situated, social, discursive action. Rhetorically speaking, individuals in social situations construct (through discourse) realities designed to influence the action of others (see, for example, Bender and Wellbery 1990; Billig 1987; Maillioux 1989; Nelson, Megill, and McCloskey 1987; Perelman 1982; Roberts and Good 1993; Shotter 1993; and White 1985.) What would such a perspective look like, given both Harris's integrational principles and the desire for close analyses of actual discursive practice? First, any integrational study of human communication must share with both Harris and the conversation analysts a commitment to firmly basing social inquiry on observations of naturally-occurring social behavior. Analysts would concentrate, in other words, on actual linguistic performances collected with minimal intrusion by the analysts themselves. Second, that data must be approached with a thoroughgoing sensitivity to the context (including the temporal context) in which such action exists. Following Harris's principle of co-temporality, an integrational rhetoric would reference particular linguistic acts both to simultaneously occurring non-linguistic acts and to the sequential working-out of the discourse itself. That is, contextualized analysis of any particular language act must attend to both its situation in time and its distribution over time.
206
DAVID FLEMING
Third, the integrationist must be cognizant of the positionality of analysis itself. An integrational rhetoric would attempt close, detailed descriptions of actual communicative events but do so without assuming that such events will be in any sense transparent. Linguistic acts are inescapably polysemous, and any theory about social life using such acts as evidence bears a heavy argumentative burden. Fourth, integrational accounts of language can afford to use neither speech nor writing as the basis for understanding language in toto; in most studies, it must be assumed that "events" and "texts" co-exist in complex ways. Thus, a theoretical or methodological preference for informal conversation, on the one hand, or formal monologues and texts, on the other, would be counter-productive. The integrational analysis of language should consider both to be species of situated communicative action. Fifth, an integrational rhetoric must be able to connect language to functionality, thus integrating form and content. No inquiry, in other words, can ignore the very real work that language does in and with the world. A radical contextualization can never be content with purely formal phenomena; it must situate linguistic practices in materially constructive actions. A conversation, from such a perspective, is more than a series of turns. It is also a site for the advancement of various constructions of reality. The analyst cannot afford to ignore those realities in favor of content-free (and contextless) abstractions. Sixth, integrational analysis must be able to connect communicative contexts to human invention, that is, to combine the traditional rhetorical interest in the planning of purposeful monologues and texts with the give-and-take of situated, dialogic action. An exclusive concern, on the one hand, with conversing or, on the other, with planning are both to be avoided. A theory of situated action must be able to account for both. Seventh, and finally, an integrational rhetoric would be oriented to instability and conflict and therefore to argument. The terrain of rhetoric is not only social and situated; it is also contested. In this view, we use language less to coordinate our thoughts with external objects or calibrate them with the thoughts of others than to advance through social discourse various partisan constructions of the world. Such an orientation would deny the presumption of intersubjectivity held by many ethnomethodologists, a presumption that allows the researcher a quietistic attitude towards ideology, power, and conflict. By focusing on the informal interactions of peers, and assuming that such interaction is managed by shared communication norms, the conversation analyst becomes blind to what the rhetorician sees as the inescapably argumentative nature of human discourse. And informal conversation, even among peers, need not be excluded from a theory of argument. Farrell (1993), for example, distinguishes between what he calls
ETHNOMETHODOLOGY AND INTEGRATIONAL LINGUISTICS
207
"rhetorical" and "conversational" discourse but sees conversations as becoming rhetorical to the extent that they involve premeditation, disturbance, or disputation. Our social lives, in this sense, are characterized less by intersubjective agreement than by what Burke (1969: 23) called "the state of Babel after the Fall". In sum, an integrational rhetoric may be one answer to the methodological problem raised at the start of this paper. In line with Harris's theory, such an approach would eschew the hypostasization of language endemic to orthodox linguistic analysis. But an integrational rhetoric would motivate more than a theoretical critique of structuralism. It would produce close, detailed, empirical analyses of situated discursive action; and it would do so without resorting to the limitations of ethnomethodological conversation analysis.
10 Linguistic Theory and the Multiple-Trace Model of Memory John R. Taylor
...every linguistic act is integrated into the individual's experience as a new event... (Harris 1981a: 155)
1. Introduction In this essay I address, in a very preliminary fashion, some consequences for linguistic theory of a psychologist's theory of memory, ι The theory in question is the "multiple-trace model" of Douglas Hintzman. The central assumptions of the model are: (i) every experience to which one attends is encoded in memory; (ii) memory consists solely of such episodic memory traces; (iii) abstractions emerge in the act of retrieving episodic memory traces; (iv) there are no abstractions, or generalizations, over experiences in memory, except insofar as these have been the object of conscious attention, and thus have become candidates for encoding in episodic memory. Hintzman summarizes the mechanics of the model as follows: The model assumes that each experience produces a separate memory trace and that knowledge of abstract concepts is derived from the pool of episodic traces 1 My reflections on the multiple-trace model were triggered by some remarks made to me in correspondence by John Williams, of the Research Centre for English and Applied Linguistics, Cambridge, England.
THE MULTIPLE-TRACE MODEL OF MEMORY
209
at the time of retrieval. A retrieval cue contacts all traces simultaneously, activating each according to its similarity to the cue, and the information retrieved from memory reflects the summed content of all activated traces responding in parallel. (Hintzman 1986: 411) Although Hintzman (1986: 423) does briefly touch on issues in language processing, he does not explicitly link this model with any specific theory of language. However, the topic of this essay should need little justification, to the extent that any viable model of human memory ought to be of inherent interest to linguists. Infants are not born with a knowledge of this or that language; languages have to be learned, slowly and gradually; the product of the learning activity has to be stored in the brain, i.e. in memory; and using language, whether in production or reception, involves retrieving some aspects of stored knowledge from memory. No matter how we may ultimately wish to characterize a speaker's knowledge of his language, this knowledge, uncontroversially, resides in memory, and must, therefore, be consistent with some plausible model of human memory. Especially linguists who would wish to claim that their work possesses the quality of psychological realism - i.e. that their linguistic descriptions capture some aspects of what a speaker of a language actually knows, and does - ought, therefore, to have some interest in what the implications of memory theories might be for the representation of language knowledge, and for the retrieval of that knowledge in the production and reception of linguistic utterances.
2. The multiple-trace
model
Hintzman (1986: 425-6) cites some historical antecedents of the multiple-trace model, including ideas of Thomas Huxley, Francis Galton, and F. C. Bartlett, as well as the memory theory of a now forgotten French psychologist, Richard Semon, whose work was first published in 1909. As we shall see below, Saussure and Bloomfield used turns of phrase which might suggest that these linguists, too, seriously entertained the notion that speakers have access to memory traces of specific linguistic events. Hintzman's model is a radical example of what Smith & Medin (1981) call exemplar theories of concept representation. As Smith & Medin (1981: 143) point out, exemplar theories might seem, at first sight, to conflict both with accepted wisdom and with common-sense intuitions. Nevertheless, exemplar theories are able to account naturally for a range of empirical phenomena that are problematic for abstractionist theories. Smith & Medin (1981: 144) cite the
210
JOHN R. TAYLOR
finding that in problem-solving tasks, people typically justify their argumentation by recalling specific instances, rather than by appeal to general principles. Hintzman (1986: 411) draws attention to the ability of experimental subjects to remember the presentation frequency of an item, its list membership, presentation modality, exposure duration, serial position, and so forth. These, he writes, are "all episodic memory tasks". Concerning the multiple-trace model itself, we need to distinguish between a "pure" version, and a "weakened" version. Let us consider, first, the pure version. Hintzman recognizes an active, working memory ("primary memory"), which contains representations of current experience, and a dormant, long-term memory ("secondary memory"), which contains traces of past experiences. Secondary memory consists solely of traces of specific episodes that have been entertained in primary memory. Importantly, an experience encountered for a second time does not lead to a strengthening of the memory trace of the first encounter, it simply produces a second memory trace, alongside that produced by the first encounter. Likewise, a set of similar experiences does not generate a memory trace which abstracts what is common to the various experiences, neither can a new experience in any way modify an existing memory trace. A current experience in primary memory can serve as a "probe" to secondary memory. Each trace in secondary memory is then activated according to its similarity to the probe. The sum of the activated memory traces is returned to primary memory as an "echo". The echo may be characterized, firstly, in terms of its intensity. The greater the similarity between probe and individual memory traces, and the greater the number of such similarities, the greater the intensity of the echo. The intensity of the echo determines the "familiarity" of the probe, i.e. of the current experience. The echo may also be characterized in terms of its content. The content of the echo is the summed contribution of all activated traces. The echo need not, indeed in many cases will not, be identical to the probe, nor indeed with any individual trace in secondary memory. For example, if the probe strongly activates a cluster of similar traces, the echo will reflect mainly their common properties, with unique properties of individual traces making only an insignificant contribution. In this way, the content of the echo may constitute an abstraction over a set of individual memory traces, in terms of which the probe may be categorized. On the other hand, if the probe weakly activates a large number of rather dissimilar traces, the echo will lack clarity, and the probe will be judged ambiguous. Further refinements might be incorporated into the model, thus allowing for the possibility (Hintzman 1988: 529) that memory traces might be degraded with
THE MULTIPLE-TRACE MODEL OF MEMORY
211
respect to the original experience, or that memory traces might slowly decay over time. Hintzman tested the pure version of the model in a series of computer simulations. The aim was to find out to what extent the multiple-trace model was able to replicate the behaviour of human subjects in a variety of experimental tasks. In the series of simulations reported in Hintzman (1986), the procedure was to setup a secondary memory consisting of exemplars, of varying degrees of representativity, of a number of pre-defined prototypes, whilst the probes were generated by distortion of the prototypes. A principal finding was that a predefined prototype could be reliably retrieved from secondary memory when cued with a distorted exemplar - even though the prototype itself was not actually represented in secondary memory. Hintzman was also able to replicate a range of experimental effects that had been reported for human subjects, concerning, for example, the effects on concept formation of category size and category overlap. In the 1988 paper, he reported on the ability of the model to replicate experimental findings on recognition memory, and on judgments of frequency and similarity. We can imagine a weakened version of the multiple-trace model, which foresees the possibility that the echo may itself come to have the status of an experience in primary memory, and may thus be laid down as a trace in secondary memory. By way of this looping mechanism, abstractions generated via the echo can enter secondary memory as episodic traces. Hintzman himself observes (1986: 422) that it would be highly unlikely for human memory not to contain such abstractions. His simulations suggest, however, that acts of categorization and concept formation do not require the existence in memory of abstract representations. And even allowing for the possibility that some abstractions are directly encoded in secondary memory, he speculates that this may occur "only rarely", and only in situations in which the abstraction has been the object of conscious attention, as when a person "attempts to define a term or to describe or imagine a typical member of a category" (ibid.). Furthermore, the abstraction would have the status of "just another episodic memory trace" (ibid.), not qualitatively different from any other memory trace. A major theoretical advantage of the model, therefore, is its parsimony. To the extent that abstractions are indeed encoded in memory, it is not necessary to draw an epistemological distinction between generic knowledge and specific knowledge, nor to propose a special generic memory system for the storing of abstractions, distinct from episodic memory, which stores specific experiences. Hintzman adduces some further arguments in favour of his model. He notes that there exists considerable disagreement in the psychological literature on the question of how abstract knowledge can be learned. The multiple-trace model
212
JOHN R. TAYLOR
provides a natural and explicit account of this process. Furthermore, the model predicts that abstractions are flexible, and context-dependent. An abstraction that is generated in the retrieval process will be crucially dependent on properties of the probe, as well as on properties of the stored memory traces. This prediction, incidentally, accords well with findings of e.g. Barsalou (1983), who showed the context-dependence of categories and concepts. Linguistic evidence for the fluidity of categories has been cited by Tsohatzidis (1995). As Tsohatzidis puts it, there is no "Great Book" in which entities are assigned to types once and for all. How an entity is to be categorized depends crucially on a speaker's current concerns.
3. Some
problems
The architectural simplicity of the multiple-trace model by-passes many of the complexities of real-world experiences, and real-world language processing. I would like to briefly mention some of these issues, before passing on to consider the more general implications of the model for linguistic theories. It should be mentioned that the problems raised in this section are by no means peculiar to the multiple-trace model. On the contrary, these issues would have to be addressed in any explicit account of language processing and language memory. A major problem, as Hintzman (1988: 545) observes, concerns the format in which experiences, and their memory traces, are represented. A basic assumption of the model is that each memory trace is activated according to its similarity to an experience in primary memory. Crucial to the model, therefore, is a mechanism for evaluating degree of similarity between representations. Hintzman handles this matter by assuming that experiences and memory traces are represented in terms of a "configuration of primitive properties" (1986: 412). Degree of similarity between representations can then be computed by a comparison of feature values. For the purpose of his computer simulations, Hintzman (1986) represented each "experience" by a string of twenty-three features, each of which could take the values +, -, or 0 (i.e. indeterminate). As mentioned, this format makes possible a straightforward computation of similarity. But even assuming a very large set of binary features, we need to ask whether a "raw" linguistic experience, and its encoding in memory, could be reasonably represented in such a way. Hintzman himself (1988: 545) notes that a feature configuration is "unlikely to be able to capture the complex structure that characterizes the memory code"; and while feature-based computations of similarity were able to model several aspects of human behaviour, they "should not be construed as a serious account of the
THE MULTIPLE-TRACE MODEL OF MEMORY
213
basis of that similarity or of the structure of the representations themselves" (1988: 545). Consider, for example, the acoustic-phonetic form of a linguistic experience. Many traditional phonological theories certainly do pretend that utterances canbe represented as strings of discrete segments, each of which can be exhaustively decomposed into a set of binary features. Any phonological representation, though, is bound to entail a massive abstraction over acoustic events. There are a thousand and one ways to say the word tree, each of them compatible with the phonological representation /tri/. Yet the essence of the multiple-trace model is that abstractions - in this case, phoneme-sized segments - are not represented as such in memory. On the pure version of the model, the generalized concept "voiceless alveolar stop", or "high front vowel", would only emerge in the retrieval from secondary memory of a cohort of memory traces which are "similar" to the acoustic event currently represented in primary memory. Yet these memory traces, to the extent that they record the raw acoustic experience, will eo ipso not be in the form of a string of discrete feature specifications. A further problem is that an acoustic image in primary memory is known to decay very rapidly, within about a quarter of a second, in fact (Garman 1990: 183 ff). If an acoustic event is to be committed to secondary memory, the memory trace will be very much degraded vis-à-vis the original experience. A reasonable assumption, therefore, might be that a good deal of phonological analysis takes place already in working memory, or as part of the perception process. Linguistic events would therefore be committed to secondary memory in a fairly "broad" phonemic form, rather than in a "narrow" phonetic form. Even so, it would probably be wrong to claim that memory traces are always stripped of all phonetic information. For example, we can often recognize a person by his distinctive voice quality, and can "place" a person geographically or socially on the basis of sometimes minute details of vowel quality. That these feats are possible at all suggests that representations in secondary memory might have to be rather rich in phonetic-acoustic detail. Presumably, also, attention to phoneticacoustic detail will play a crucial role in language learning, whether by children learning their mother tongue, or by adults learning a foreign language (or even a new dialect of an already familiar language). A purely featural representation becomes even more problematic when we turn to the semantic aspects of a linguistic experience. In the first place, the context of the speech event, and the conceptualization symbolized by a linguistic expression, will display a degree of variation at least equivalent to the amount of variation in the acoustic-phonetic event. Again, it seems reasonable to suppose that a good deal of contextual information will be stripped away in the very act of perceiving a scene, and processing the signal in primary memory.
214
JOHN R. TAYLOR
Furthermore, feature theories of meaning have turned out to be much less viable than feature theories of phonology. Only with respect to some very limited semantic domains, such as kinship relations (e.g. Goodenough 1956), or dimensional adjectives (e.g. Bierwisch 1967), has an exhaustive decompositional analysis ever been seriously pursued. Even these componential approaches have been criticized; on kinship terms, see Wierzbicka (1992, Ch. 9), on dimensional adjectives, see Dirven & Taylor (1988). Indeed, some scholars, most notably Fodor (1981), have rejected in principle the very notion of semantic decomposition. The majority view seems to be that while semantic decomposition is possible, and even desirable, there is no pretence that the semantic components thus isolated exhaustively represent the meaning of a linguistic expression. Once we identify [+animate] and [-human], and perhaps even [+bird], as components of duck and goose, we are still left with the task of distinguishing ducks from geese. As Jackendoff (1990: 33) points out, it would be "patently ridiculous" to handle the distinction by proposing all sorts of ad hoc features, of the kind [±hasa-long-neck]. "To put a ± sign and a pair of brackets around any old expression doesn't make it into a legitimate conceptual feature". These considerations show that any reasonably sophisticated account of language processing and language memory may well be inconsistent with the feature-based format assumed by Hintzman in his implementation of the multipletrace model. It must be emphasized once again, however, that the kinds of issues raised in this section will need to be addressed by any theory of linguistic representation; the above remarks should not therefore be construed as being directed specifically at the multiple-trace model. We may still be entitled to suppose, with Hintzman (1988: 545), that "the complex processes underlying everyday memory reduce to a simpler, multiple-trace mechanism", while leaving open the difficult, and unresolved question of the format of linguistic representations in primary and secondary memory.
4. Language knowledge
and the multiple-trace
model
Let us consider, then, in general terms, what a multiple-trace mechanism might mean for an account of language knowledge. As the reader will no doubt have already appreciated, the implications of the model could be very far-reaching indeed. The central assumption of the model is that secondary memory consists solely of traces of previously entertained experiences. Knowledge of a language, therefore, would comprise nothing other than a vast set of memory traces of previous encounters with the language. There would be no place in the model for "rules", whether these be understood as procedures for the manipulation of
THE MULTIPLE-TRACE MODEL OF MEMORY
215
categories of symbols, or simply as abstractions or generalizations over individual linguistic events. For in its pure form at least, the multiple-trace model makes no provision whatsoever for the mental storage of generalizations over individual events. The archetype of language knowledge is not the generalization captured in the grammar rule, but the particularities of the corpus. Many linguists, if presented with such a prospect, would probably be inclined to dismiss it as absurd, and inconsistent with just about everything that the modern science of language has discovered about the nature of linguistic competence. After all, is not the most basic insight of linguistic enquiry over the last half century or so that speakers of a language have internalized a set of rules for the generation of sentences? Abstraction, and generalization over data, play an absolutely central role in just about all mainstream theories, whether in the domain of syntax, phonology, or semantics. For many linguists, "doing linguistics" is tantamount to formulating rules and spotting generalizations. Linguistics without rules and without generalizations is almost a contradiction in terms! Before we succumb to the initial impulse to dismiss the multiple-trace model as unworthy of a linguist's attention, we should bear in mind one major result of Hintzman's work. This was to show that the multiple-trace model was able to simulate a range of behavioural effects commonly assumed to rely on abstraction and generalization, even though abstractions as such were not represented in the system. Neither is the multiple-trace model unique in this respect. Some other recent attempts to simulate cognitive (including linguistic) behaviour have also sought to eliminate explicit procedures for the categorization and manipulation of symbols. I refer to experiments in parallel distributed processing (PDP), first reported extensively in Rumelhart & McClelland (1986) and McClelland & Rumelhart (1986). Rules and generalizations, in PDP systems, are emergent properties of self-organizing systems, they are not explicitly represented in the system. Reactions to PDP as a model of language processing have ranged from the highly critical (e.g. Fodor & Pylyshin 1988) to the sympathetic (e.g. Sampson 1987; Langacker 1991: 525-36). Indeed, these linguists appear to see in PDP systems the possibility of a model of grammar that is at once descriptively adequate, and psychologically (and perhaps even neurologically) real. The multiple-trace model was developed independently of PDP, and its architecture differs significantly from that of PDP systems. (The matter is briefly touched in Hintzman 1986: 425.) Especially interesting for our purpose is the fact that the multiple-trace model makes explicit the mechanism whereby a current representation can be evaluated against episodic traces in secondary memory. The mechanism corresponds rather closely to the linguist's notion of analogy. On a multiple-trace model of linguistic knowledge, utterances would be generated and
216
JOHN R. TAYLOR
understood, not by the application of general rules, but on the basis of their similarity with previously encountered utterances. As is well known, the modern linguist's predilection for rules has been accompanied by a downgrading (demonisation, even) of the role of analogy. Chomsky, especially, has been particularly scathing. He associated analogy with behaviourism, with the "fear of 'mentalism'", with the notion that language knowledge is a matter of "habits and dispositions" (Chomsky 1988: 9). Likewise, Langacker, in his earlier days (i.e. before his work on Cognitive Grammar), dismissed analogy as "simply false" (Langacker 1967: 22). He added, as if explicitly rejecting a multiple-trace mechanism, that "we do not go around collecting sentences to hold in memory for future use in speaking and understanding". Nor, when we create and understand sentences, "do we have to search through our personal linguistic archives" in order to find models which might sanction our current linguistic activity. Yet, before the advent of generative grammar, analogy held pride of place in the linguist's theoretical arsenal. It was, in fact, the very mechanism that was claimed to be responsible for phenomena which nowadays we should want to capture in terms of rules. For Bloomfield, the essence of syntax lay in analogy; furthermore, it was analogy that sanctioned creativity, i.e. the ability of speakers to produce and understand utterances they had never previously encountered: A grammatical pattern (sentence-type, construction, or substitution) is often called an analogy. A regular analogy permits a speaker to utter speech-forms which he has not heard; we say that he utters them on the analogy of similar forms which he has heard. (Bloomfield 1933: 275) Observe Bloomfield's wording in this passage: his talk of forms which the speaker "has heard" strongly suggests that Bloomfield took seriously the notion that a speaker recalls from memory the traces of specific linguistic events. Saussure was even more explicit: ...il faut attribuer à la langue, non à la parole, tous les types de syntagmes construits sur des formes régulières. En effet, comme il n'y a rien d'abstrait dans la langue, ces types n'existent que si elle en a enregistré des spécimens suffisamment nombreux. Quand un mot comme indécorable surgit dans la parole...il suppose un type déterminé, et celui-ci à son tour n'est possible que par le souvenir d'un nombre suffisant des mots semblables appartenant à la langue {impardonnable, intolérable, infatigable, etc.). Il en est exactement de même des phrases et des groupes de mots établis sur des patrons réguliers; des combinaisons comme la terre tourne, que vous dit-il? etc., répondent à des types
THE MULTIPLE-TRACE MODEL OF MEMORY
217
généraux, qui ont à leur tour leur support dans la langue sous forme de souvenirs concrets. (Saussure 1922: 173) This passage shows that Saussure, also, seems to have seriously entertained the notion that langue, as represented in a person's brain, could be construed as a depository for a vast number of individual past encounters with a language. Note, in particular, the references to "memories" (even: "concrete memories") and "specimens". Analogical creation, whether in word formation or in syntax, presupposes the recognition, by the speaker, of "regular patterns", which are "supported" by memories of concrete instances. In terms of the multiple-trace model, the regular patterns emerge as properties of the echo, which in turn sums the properties of individual usage events, as these have been recorded ("enregistré") in secondary memory. A first implication of the multiple-trace model, therefore, could be a reinstatement and re-evaluation of analogy as the mechanism underlying syntactic and morphological regularities, and as the mainspring of linguistic creativity. Contemporaneously, rules - especially procedural rules - would be downgraded to the point of elimination. A further point is that the elimination of rules understood as procedures for the manipulation of symbols - would entail strict monostratalism. The term is due to Hudson (1990: 7), who used it to refer to the thesis (quite widely accepted, nowadays) that there is only one level - namely, the "surface" level - at which linguistic structure is represented. Excluded in principle, therefore, would be all kinds of "deep", or "underlying" structures, which get mapped, through rule application, on to a surface structure. Some further implications are as follows: (i) A major issue concerns the specificity vs. generality of language knowledge. Mainstream linguistic theories favour abstract, generic knowledge, whilst the multiple-trace model predicts specific knowledge, in the limiting case, knowledge only of previously encountered linguistic events. (ii) The multiple-trace model entails that language learning will be a life-long process. Every linguistic encounter will have some effect on the contents of secondary memory. In the case of routine utterances - these, precisely because they are routine, are unlikely to impose themselves on a person's conscious awareness - the effect may be minimal. Innovative utterances, however, are more likely to become the object of conscious attention, and may have a more substantial impact. Mainstream theories, in contrast, assume that language knowledge achieves a "steady state" at a rather early age in a person's development. (iii) A further issue is the very construct of "a language". A guiding assumption of mainstream theories is that a language - barring obvious dialectal
218
JOHN R. TAYLOR
variation - is a homogeneous entity. People who "speak the same language" have in their brains identical representations of this homogeneous entity, the point of linguistic enquiry being, precisely, to explicate this unitary mental construct. (The myth of the homogeneous language community, by the way, is still alive and well; see Chomsky 1988: 9-10.) The multiple-trace model, in contrast, predicts no homogeneity of an entity called "a language". Each person's linguistic experiences are unique to that person. There can be therefore no inter-speaker identity of mental representations of language. (iv) A multiple-trace model will tend to highlight the uniqueness of each linguistic event. The use of any linguistic form will involve assessments of similarity, not only of the linguistic utterance to previously encountered linguistic utterances, but also of the situation in which the expression is used to previously encountered situations. Each of the above entailments of the multiple-trace model, it seems to me, is not only very plausible, but probably correct. Consider, for example, the question of language-learning as a life-long process. It is, I would have thought, uncontroversial that even adult native speakers are confronted, almost daily, with new uses of familiar words, with new turns of phrase, with locutions which might strike them initially as syntactically slightly deviant. These, if encountered often enough, and if they serve a communicative need, are apt to become entrenched in the person's linguistic repertoire, just as the multiple-trace model predicts. By allowing for the possibility that traces that were laid down long ago might decay over time, the multiple-trace model also accounts for the atrophy, or progressive forgetting, of linguistic devices. It is again a common observation that linguistic uses that were familiar in one's youth gradually become dated, and may eventually fall out of one's linguistic repertoire. Consider, also, the question of the generality vs. specificity of a person's linguistic knowledge. Linguists are trained to make generalizations over linguistic data, and linguistic descriptions tend to be valued to the extent that they make generalizations over a wide range of apparently disparate data. It by no means follows, however, that a linguist's generalizations necessarily correspond to anything in the mind of a speaker. The matter receives some discussion in Hudson (1984: 27-29) and Starosta (1988: 41-42). Hudson (1984: 28) points out that a person does not need to access a general rule in order to produce the forms generated by the rule. It could be the case that the person has simply "learned all the right facts" about the instances covered by the rule. New instances would then be sanctioned by analogy with learned instances, rather than by application of a general rule.
THE MULTIPLE-TRACE MODEL OF MEMORY
219
There is, in fact, some psycholinguistic evidence that complex morphological forms may be stored and accessed as wholes, rather than assembled by rule from their constituent parts (Bertinetto 1994). It is, furthermore, a common observation of any working linguist that speakers are often sublimely unaware of the kinds of abstractions that linguists are wont to make. Starosta's comments on Palauan demonstratives are symptomatic: ...the elaborate demonstrative system of Palauan [is] a real morphologist's delight: a detailed morphological analysis can discover a consistent meaning for almost every phoneme in every (non-suppletive) form, but it has so far proven impossible to get a native speaker to show any awareness of the existence of this charming system. (Starosta 1988: 42) To claim that the linguistic competence of a speaker comprises generalizations which are inaccessible to that same speaker, could be to commit a serious metaphysical error. A more reasonable conclusion is that speakers of, in this case, Palauan, have simply learned each of the demonstratives as unanalysable wholes. They have not proceeded to make any generalizations over these forms, neither do these generalizations govern their everyday use of the forms. A number of writers have suggested that even syntactically complex expressions may be stored and accessed as wholes. This goes not only for true idioms, but also for expressions that are compositionally transparent, and which therefore in principle are subject to assembly by rule. Langacker (1987: 35-6) draws attention to the "huge set of stock phrases, familiar collocations, formulaic expressions, and standard usages" that permeate fluent use of a language. These expressions, through repeated use, have become thoroughly automated, and are employed by a speaker as integrated wholes, without attention to their internal structure. There are, Langacker suggests, "literally thousands" of these conventional expressions in any given language, and learning them "is probably by far the largest task involved in mastering" a language. A similar point was made by Pawley & Syder (1983). They maintain that fluent and idiomatic control of a language rests to a considerable extent on knowledge of a body of "sentence stems" which are "institutionalized" or "lexicalized". A lexicalized sentence stem is a unit of clause length or longer whose grammatical form and lexical content is wholly or largely fixed; its fixed elements form a standard label for a culturally recognized concept, a term in the language. Although lexicalized in this sense, most such units are not true idioms but rather are regular form-meaning pairings. (Pawley & Syder 1983: 191-92).
220
JOHN R. TAYLOR
Pawley & Syder speculate that the number of such units, each of which needs to be individually memorized, may run into "hundreds of thousands". The thesis that a person's linguistic knowledge might be highly specific is especially compelling with respect to word meanings. It is worth mentioning in this context that a number of rather diverse linguistic theories appear to be converging on the thesis of "lexicalism", i.e. the notion that a very great deal of a person's knowledge of a language consists in a knowledge of words, and the properties of words (Hudson 1990: 3-4). Observations on word meanings, therefore, could have rather broad significance for linguistic theory, far beyond the domain of what Thomason (1974: 49) disparagingly referred to as "lexicography". It is a remarkable fact, as Johnson-Laird (1987: 191) has noted, that although speakers of a language readily concur that words have meanings, people are generally hard put to say what the meaning of a given word is, especially if the word is presented devoid of context. One reason for this could be that words, in themselves, do not actually have meanings. It is only collocations, which have been individually learned by speakers, that have meanings, and it is only these meanings that speakers of a language are able to retrieve and formulate. Take the verb cut (the example is suggested by Searle 1980). If you cut the grass you perform a very different kind of activity than if you cut a cake (you do not, for example, cut a cake by running a lawn-mower over it); cutting your hair is again very different from cutting your finger (cutting your finger does not entail lopping off a one-inch portion with a pair of scissors). In spite of these differences, we might still, perhaps, want to claim that cut, as well as the other lexemes in these expressions, such as grass, cake, hair and finger, has a unitary meaning. The meaning of a complex expression would then be calculated by combining the unitary meanings of the component parts. Subsequently, the compositionally derived meaning would be subject to a pragmatically driven process of interpretation. This would fill in information concerning, e.g., the typical kind of cutting instrument, the typical shape and consistency of the substance cut, the purpose and expected end-result of the activity, and so on. This, I take it, is the standard view, as defended by Johnson-Laird (1987), Katz (1981), Bierwisch & Schreuder (1992), and many others. Alternatively, we could say that a person understands cut the grass in virtue of his having observed that this is the appropriate expression to use in referring to a certain kind of activity. On this view, a speaker's knowledge of the verb cut would consist, not in knowledge of a unitary meaning, but in knowledge of a large set of accepted collocations, and of the kinds of activities of which these collocations can be used. There are, I think, several reasons why this latter approach is to be preferred (see also Taylor 1992, 1994). Note, first, that it is
THE MULTIPLE-TRACE MODEL OF MEMORY
221
infinitely easier to say what cut the grass means, than it is to give a definition of cut that abstracts away from all reference to the material cut, to the cutting instrument, to the motor movements involved in the activity, and so on. And even supposing that one could come up with some kind of general definition, along the lines, say, of "separate parts of a solid object with a sharp instrument", the trouble, now, is that there are many activities to which this definition might apply, but which are not conventionally denoted by cut. The barber cuts my hair, but I do not, each morning, cut my beard; for this latter kind of activity we conventionally use the verb shave. If I splice a plank of wood lengthwise with an axe, I would say that I chopped (or spliced) the wood, not that I cut it. Likewise, if I divide a plank of wood cross-wise with a saw, I would prefer to say that I sawed the plank in two, not that I cut it in two. Knowing the verb cut, then, entails not only knowing the accepted collocations of the word, but also knowing that certain kinds of activities are not conventionally described as cutting. There is a further point. Although cut the wood with an axe is not the standard way of describing a chopping activity, the expression (if produced, say, by a child, or a foreigner) is still comprehensible to a speaker of English, on the basis of analogies with the conventional uses of the word. Even if it is accepted that knowing a word involves knowing a vast number of accepted collocations, one might still feel unhappy about the claim that linguistic memory contains records of individual linguistic events. As we have noted, the multiple-trace model does make provision for the encoding of generalizations in memory; arguably, these generalizations might well be at a rather low level of abstraction (the level exemplified by collocations such as cut the grass, cut your fingernails, cut the cake, etc.), rather than at high levels of abstraction, such as would be required in definitions of context-free lexical items. Even so, there is some circumstantial evidence for the plausibility of the multiple-trace model even in its pure form. Speakers have quite clear intuitions on the relative frequency of elements in their language, and frequency has been shown to be a crucial variable in many experimental tasks, e.g. in word recognition and lexical access (Garman 1990: 140-1, 257-9). Furthermore, frequency effects are not restricted to individual lexical items, they show up also with respect to collocations (Potter & Faulconer 1979). A reasonable conclusion, therefore, is that speakers somehow keep count of the number of times linguistic units (including collocations) have been encountered. This conclusion is entirely in keeping with a pure version of the multiple-trace model.
JOHN R. TAYLOR
222
5. Cognitive
grammar
I have looked at some of the implications for linguistic theory of a multiple-trace model of memory, and contrasted these with the practices of mainstream linguistic theories. I would like to briefly mention one linguistic theory - not, perhaps, a mainstream theory, but one, nevertheless, which appears to be gaining increasing recognition - which is rather compatible with the multiple-trace theory. I refer to Cognitive Grammar, as presented by Ronald Langacker in various publications (see, especially, Langacker 1987, 1991). Langacker (1990) has characterized cognitive grammar as a "usage-based model". It is taken as self-evident that a speaker acquires his language through encounters with specific usage events, also that each usage event is unique, with respect both to the acoustic properties of the linguistic utterance, and the conceptualization that it is meant to symbolize. In spite of the uniqueness of two usage events, a speaker may nevertheless perceive similarities between them, regarding their phonological properties and/or their conceptual (i.e. semantic) properties. On the basis of these perceived similarities, usage event [B] may be construed as an "extension" of usage event [A]. At the same time, the speaker may be able to extract the commonality of [A] and [B], and thus construct a representation [C] that is "schematic" for [A] and [B]. As the speaker is exposed to an ever wider range of usage events, the processes of extension and abstraction may apply recursively. On this model, knowledge of a linguistic unit - whether a phonological unit, a semantic unit, or a symbolic unit relating a phonological unit to a semantic unit - will take the form of a possibly highly complex network of nodes, linked by relations of extension (one node is perceived to be similar to another) and schematicity (one node extracts the commonality of a range of more specific representations). Whilst the higher nodes of the network summarize the commonality of lower elements, at the very bottom of the network stand individual usage events. Langacker (1987) endorses the various entailments of the multiple-trace model that I have mentioned earlier: monostratalism (pp. 46-7), the dynamic nature of a person's linguistic knowledge and the inevitability of inter-speaker differences (p. 376), the role of analogy (pp. 445 ff), and, above all, the specificity of language knowledge. A crucial aspect of Langacker's network model is the claim that abstraction of a higher-level schema need not erase from memory knowledge of the more specific instances, whose commonality the schema is meant to represent. On the contrary, Langacker suggests that the production and understanding of new utterances may well proceed by the activation of rather low-level schémas, rather than on the basis of maximally general abstractions.
THE MULTIPLE-TRACE MODEL OF MEMORY
223
Though regularities are obviously noted and employed in the computation of novel expressions, it is quite conceivable that low-level schémas are more important for this purpose than highly abstract schemas representing the broadest generalizations possible. If high-level schemas are extracted, they may be of only secondary significance, serving more of an organizing function than an active computational one. (Langacker 1990: 265) The general emphasis of cognitive grammar, therefore, is on the mass of specific knowledge accumulated by speakers over their lifetimes. Even though the linguist may be able to summarize a plethora of linguistic data in a few, highly general statements, these general statements may play only a marginal role in a speaker's production and comprehension of new utterances.
6. The
fixed-code
fallacy
I conclude this essay by pointing to the relevance of its subject matter to an important theme in the writings of Roy Harris. Over the years, Harris has led a vigorous critique of what he calls the fixed-code fallacy. The fallacy lies in the assumption that a language consists of a fixed set of signs, each associating a unitary meaning with a unitary phonological representation, in association with a fixed set of rules for the combination of the signs. A language, on this view, can be exhaustively described as a code book accompanied by a rule book. A grammatical sentence results when signs from the code book are combined in accordance with the rules from the rule book. According to Harris, this view of language has informed much twentiethcentury theorizing about language. Formal theories of syntax, whether structuralist, transformationalist, relational, categorial, or whatever, all presuppose that smaller entities (words, phrases) get combined in accordance with fixed rules. Much semantic theorizing, whether of the structuralist, truthconditional, or model-theoretic varieties, takes word meanings and sentence meanings to be fixed entities, with sentence meanings being compositionally computed from word meanings. Likewise, much of phonology, past and present, presupposes that for each sign in the code book there exists a fixed, invariant representation, that is itself composed from elements (phonemes, features) selected from a small, finite set. The sub-disciplines of pragmatics, psycholinguistics, and sociolinguistics (not to mention stylistics and phonetics) owe their legitimacy precisely to the fact that these sub-disciplines propose to handle those aspects of language use that are not accounted for by the code book
224
JOHN R. TAYLOR
and the rule book. Thus, for various reasons which are extrinsic to the language per se, a speaker, in using a sentence, may care to intimate certain intentions that are not encoded by the sentence; he may produce a sentence which deviates from the prescriptions of the rule book; or, with increased speech tempo, the actual pronunciation of a word may diverge quite radically from the prescription of the code book. And so on. Harris presents a number of arguments which aim to show why the notion of a language as a fixed code must be wrong. The fixed-code thesis fails to take proper account of the facts of variability, creativity, and context-dependence. The stock of words ("signs") at a person's disposal is not a fixed quantity. Words can be used in all manner of new and innovative ways, and hearers usually have no difficulty knowing what was meant. Words can be invented (and invented words can be understood), rules can be extended to new instances (with no impression of deviance), and meanings can be negotiated. And the way in which an utterance is evaluated may be highly dependent on the total context in which the utterance was made. No doubt, a defender of the thesis of language as a fixed code could still attempt to account for variability, innovation, and context-dependence, by appeal to a battery of pragmatic principles. Still, it may be countered that variability etc. are so all-pervading in language use that any proposal to shunt them off onto peripheral sub-disciplines brings with it an impoverishment, not to say distortion of the object of investigation. A descriptively adequate theory of language must treat variability, innovation, and context-dependence as normal and expected, not as phenomena which require special explication in terms of principles that are extrinsic to the language as such. More far-reaching are the epistemological questions raised by the fixed-code model. On the fixed-code model, communication via language will only be possible if each party to the act of communication has access to the same code book, and to the same rule book. But what guarantee can a person have that the code and the rules inside that person's head are commensurate with, let alone identical to, the code and the rules inside the interlocutor's head? One does not have to be a trained phonetician to appreciate that speakers of the "same" language pronounce the words of the language in different ways. If interpersonal variation exists with respect to the "public" side of the linguistic sign (where, after all, it would in principle be possible for a person to accommodate his pronunciation to that of other people), what right have we to suppose that the "private" side of linguistic signs (i.e. "meanings") will exhibit a lesser degree of variability? The irony, as Harris points out, is that not only does the fixed-code model presuppose identity across speakers of "meanings in the head", but the private
THE MULTIPLE-TRACE MODEL OF MEMORY
225
meanings of different speakers are matched up precisely in virtue of the fixed code which is the language they share. The fixed-code model construes linguistic communication as a process of "telementation", whereby meanings that I entertain in my head get encoded into signals, which are then decoded by the hearer, who ends up with, in his head, a replica of the meanings that I have in my head. Comparing "meanings" thus becomes as straightforward as comparing the monetary values of two bank notes. A view of linguistic knowledge based on a multiple-trace model is attractive precisely because it can offer a viable alternative to the fixed-code view. It gives us, not the image of a code book supplemented by a rule book, but the image of a language as a vast mental corpus. New expressions are sanctioned, and interpreted, with reference to prior experience with the language, not on the basis of algorithmic rules operating on symbols of predetermined value. Linguistic knowledge, and its application to new situations, will be inseparable from conceptual knowledge, with which it is associated in secondary memory. Furthermore, each linguistic event will be truly creative, in that it will involve active assessment by speakers and hearers of the relevance of previously learned expressions to the current situation. Importantly, also, the model carries with it no expectation that two speakers will share exactly the same mental representation of their language, nor that language knowledge will remain static throughout a person's adult life. It is, after all, an everyday observation that communication is perfectly possible even between people (e.g. adult and child, native speaker and foreign learner) who manifestly do not share the same code, or the same rules.
11 Language, Art and Kant Trevor Pateman
Roy Harris's love of art is, perhaps, less well known than his love of cricket and, though it may be less fun for him to read (or for me to write), I'm going to offer some thoughts on language and art, rather than language and cricket. There is at least one tradition which is defined by its commitment to thinking about language as an art - thinking about language in the same categories as are used to think about art. In this century, that tradition is represented in the work of Benedetto Croce and at least two major figures influenced by him: on the left, Antonio Gramsci, the good Communist; on the right, Karl Vossler, to whom we would probably pay more attention if he hadn't got mixed up with the kind of people who put people like Gramsci in prison or worse. When this tradition is tracked back into the nineteenth century, we quickly turn up Hegelian idealism and organicism as an inspiration, and nowadays, I'm sure, no straight up and down linguist really wants to get involved with that, do they? However, if we go back one more step, we get to Kant's philosophy of art as expressed in the third Critique of Judgement, a dense and difficult work to which very interesting access is provided by Mary Mothersill's book Beauty Restored, which is the immediate source of inspiration for what follows. As part of what we might now think of as a romantic reaction to classicism, Kant wants to have beauty in art without a rule book for beauty. He tries to show that in the ordinary and (Kuhnian) normal scientific understanding, we fit instances of things, events, perceptions to pre-existing rules, concepts, schemata or categories. Such fitting takes the form of what he calls determinant judgement - we determine into which pre-existing pigeon-hole something ought to go. In contrast, says Kant, when it comes to aesthetic judgement (the judgement of beauty) we (in some sense) frame a (new) rule, concept or category to fit the (unique) instance presented by the work of art before us. We do this in an act of imagination, or what Kant calls reflective judgement - we have to find (reflexively) the rules which will allow us to demonstrate post hoc the beauty we intuited without a rule to guide us.
LANGUAGE, ART AND KANT
227
I explicate this a bit more in my little essay on "Beauty". The extension of the argument and distinction drawn to language may well be obvious. On the one hand, we have linguists (Saubloomsky, in particular) who tell us that judgements of grammaticality are determinant judgements which involve categorising candidate sentences as grammatical/ungrammatical, meaningful/meaningless, and so on, against rules of grammar which can and ought to be specified in advance and, in principle, exhaustively specified. On the other hand, we have philosophers and linguists (let's just name Roy Harris on this occasion), who say that it ain't like that at all, and that the reason we ain't yet found no Rule Book that works is because there ain't no Rule Book to be found. Or, rather, the Rule Books cover only the fossilised bits of language, the bits which have sedimented, and come out now as determinately grammatical only because there has been a history of reflective judgements out of which the lineaments of grammar have emerged. The important thing is that it is this way round and not the other. In other words, the essential linguistic judgement is a reflective one, rather than a determinant one, and does the job of allowing a candidate bit of language to be OK - "acceptable", in the jargon, but not now in the sense of being OK despite falling short of Full Grammaticality, but in the sense of being acceptable for the subsequent job of finding a rule which explicates its OKness as part of living language. With a work of art, one might want to say and have every right to, "This is beautiful, but I don't yet know how to show you/state/define its beauty". Likewise, with a piece of language one might want to say, "This undoubtedly belongs, but I can't yet give you a rule of grammar which will show it belongs". In the romantic and idealist traditions, poets are given the job of showing us new ways with language, to present something which as it were exploits the vagueness, incompleteness and contradictions of our Rule Books and comes out as something we hear (or look at) and respond to as appropriate, Just Right, Neat, or even, Beautiful. In other words, in these traditions, the creativity of the poet is located not in conforming well to old rules, nor in breaking old rules, nor in inventing new ones, but in using language where there are no (clear) rules, yet provoking us into finding a rule for the language so used. (For, undoubtedly, human beings though enjoying their reflective judgements do like their determinant ones, too). There is an exact analogy with the notion of judge-made law. Of course, judges make law, because the Rule Book is never complete, never contains all the rules necessary for its own interpretation. But when judges make law, the law they make has to go somewhere - and it goes into the Rule Books. But why poets? Well, of course, we are all poets now - that's the democratic version of the tradition. But that isn't to say that poetry is easy. For we will only
228
TREVOR PATEMAN
get an OK for our poetry if it is sufficiently motivated - linguistic innovation doesn't come about through arbitrary word-play. For example, linguists like P.H. Matthews have pointed out the indeterminacy in the Rule Book of conventionalised English regarding the colour words which may add -ness to give something nominal and substantive: so redness is definitely OK, and magentaness is definitely not OK and in between it gets pretty hazy. But to get something out of the hazy category and into the OK category takes talent, the talent of finding a context where what is done can be seen as exactly the right thing to have done by someone who was not at all expecting it to be done. Linguistic innovation ought to be a pleasant surprise, a breath - if not a blast - of fresh air. I hope this blastschrift is an appropriate response to the frische Lufl Roy Harris has brought to linguistics.
12 From An Integrational Point of View Roy Harris
A demythologised linguistics (or, to give it a less negative designation) an "integrational linguistics" would need to recognise that language allows and requires us to do both far more and far less than the telementational model claims. Language is a process of making communicational sense of verbal behaviour. Our training in language is a training to use words in such a way that, in the context of a particular situation, our total behaviour will make the kind of sense to others that we intend that it should, and effectively implement our interactional objectives. The Language Myth, p. 165
Introduction In this essay I would like to do something more than offer replies to specific points raised by the contributors to this volume. I shall try to situate the replies within the framework of what I hope will be a more comprehensive and concise statement of my theoretical position than has appeared hitherto. I shall also take up some criticisms which have been raised in other quarters. Let me say straight away that I am grateful to all the contributors for taking the time to write about my work, even when they clearly regard it as worthless. It would be impossible to deal in detail with all the points they raise, and I trust they will grant me the respondent's privilege of selecting those which seem to me of most relevance. 1 1
Although some points caused great amusement, I did not always think them strong enough to be worth taking up. Into this category fall, for instance, the arguments (i) that because Chomsky spends a lot of his spare time ranting against the U.S. government, this shows that generative linguistics is capable of dealing with the problem of free speech, and (ii) that because structuralism was not approved of by fascist governments in Germany and Italy, we can conclude that structuralism does not reflect totalitarian ideology. I am not sure how seriously these were intended, but I thank Borsley and Newmeyer for the entertainment anyway. I would also like to take this opportunity of thanking Nigel Love and George Wolf for pertinent comments on a number of issues addressed in my reply.
230
ROY HARRIS
Any criticisms which they - or other readers - think I have culpably evaded I would be more than willing to take up in private correspondence. I live at 2 Paddox Close, Oxford 0X2 7LR, U.K. Although some of my academic colleagues evidently see both my work on Saussure and my work on writing as being quite separate from my critique of orthodox linguistics, I have always considered all three to be intimately - and, I trust, coherently - interconnected. Not merely adventitiously, but intrinsically so. For all three, in my eyes, follow from the adoption of an integrational view of human communication. I do not think I overstate the case when I say that the integrational approach offers a radically different view of language from that which became institutionalized in mainstream linguistics in the twentieth century. It proposes a quite different foundation for linguistic theory, as well as for the interpretation of our everyday linguistic experience. But at the same time it offers an explanation ofand thus transcends - the orthodoxy it rejects. The orthodoxy, on the other hand, offers no comparable transcendance of the integrational view. (It is therefore obliged to dismiss integrationism as "irrelevant": see Borsley and Newmeyer, this volume, Ch. 2.) This assertion of asymmetry should not be interpreted as a hubristic claim to originality. I have always held that recognition of the validity of an integrational point of view is to be found, both implicitly and explicitly, in the work of a number of theorists.2 And I have repeatedly directed the attention of students to these sources. But, at least in the field of linguistics, none of the pioneers in question managed to avoid compromise with the dominant orthodoxy. In this, I feel, they failed to follow through the logic of their own insights. I have tried to spell out this logic of integrationism in a number of publications, beginning tentatively in 1973 with Synonymy and Linguistic Analysis and elaborated most fully in Signs, Language and Communication, which was written in 1994 and has yet to be published. Rather than attempt to give a chronological summary of my own thinking during the intervening two decades, I shall begin by offering some exegetical comments on the above paragraph from The Language Myth, which I wrote in 1981 and which seems to me still valid. It conveniently brings into focus a number of key issues. But it may be useful to preface these comments with some autobiographical remarks, without which the comments may not make much sense to readers unfamiliar with the history of linguistic studies at Oxford. 2 See "Language as Social Interaction: Integrationalism Versus Segregationalism", Language Sciences, Vol.9 No. 2, 1987, pp. 131-143. Reprinted as Ch. 12 of N. Love (éd.), The Foundations of Linguistic Theory. Selected Writings of Roy Harris, London: Routledge, 1990, pp. 197-209.
FROM AN INTEGRATIONAL POINT OF VIEW 1. The study of
231
language
Mastery of the techniques of linguistic communication, and the problems involved in understanding what that mastery involves, were the topics that preoccupied me over the period of ten years during which I held the Chair of General Linguistics at Oxford. All that I wrote during that period (1978-88) may be read as an expansion of my inaugural lecture3 on appointment to that Chair, and represents the point of view from which I taught throughout my tenure of it. The Chair itself had been established belatedly and, in the eyes of some, rather reluctantly by a University not overwhelmingly convinced by the academic credentials of general linguistics. For many at Oxford, it was still difficult in 1978 to distinguish between general linguistics and comparative philology. The latter was a subject long recognized by the University, as witnessed by its appointment of Max Muller to a professorship of comparative philology in 1868. And it was, after all, Max Muller who had proclaimed the advent of what he called "the science of language" and affirmed that "we do not want to know languages, we want to know language".4 The University still had a Chair of Comparative Philology (in spite of the recommendation by one eminent holder that it should be abolished). Furthermore, my immediately previous post had been as Professor of the Romance Languages at the same institution. So there was plenty of room for doubt and confusion about what Oxford thought "general linguistics" was. All this provides the background for two comments that were reported to me at the time of my election. One, from within the University was, "Why do we need two professors of the same subject?" The other, from a waggish colleague at another university, was: "How typical of Oxford to appoint someone who doesn't really believe in the subject." In retrospect this latter comment might seem to have been justified, inasmuch as The Language-Makers, The Language Myth and The Language Machine were widely interpreted as a series of deliberate attacks on the sacred cows of twentieth-century linguistic theory. A similar motivation was also attributed to two other books of mine dating from the same period: The Origin of Writing and Reading Saussure. Given this interpretation, it was hardly surprising that among the ranks of academic linguists much of the reaction was indignant and hostile. Less predictable were the reactions from other quarters. Philosophers in particular evinced a remarkable capacity for outlandish misreadings. 3
Communication and Language. An Inaugural Lecture delivered before the University ofOxford on 24 February 1978, Oxford: Clarendon Press, 1978. Reprinted in Love (ed.) (1990), pp. 136-50. 4 F.M. Muller, Lectures on the Science of Language, London, 1861, p. 24.
232
ROY HARRIS
The trouble began with The Language-Makers, which seems to have been especially vulnerable in this respect, doubtless because the line it took was hardly what might then have been expected from a newly appointed professor of general linguistics. The review in Ethics5 managed to see it as arguing that a general science of language is impossible because each language is unique. But this interpretive tour de force was capped in Canadian Philosophical Reviews, whose readers were told that the book offers inter alia a "defense of structural semantics".6 The review in Choice7 praised its "brilliant discussion of scriptism" but described it as presenting a "counterculture view". According to The Times Educational Supplement 8 (John Weightman) it was unclear exactly where the author of The Language-Maker s stood; whereas according to The Times9 (Randolph Quirk) nothing could be clearer - he stood "with Darwin, high on theology's unwanted list". Potentially higher still, it might have been added, on another unwanted list: the professional linguist's. On the basis of reading the review in The Times, one reader wrote a letter to the editor which saw the book as heralding the demise of the discipline. 10 This prognostication elicited a prompt Establishment response. It came in the form of an official reply from the Chairman of the Linguistics Association of Great Britain, assuring the general public that linguistics was "flourishing in the universities" and the Association's meetings were "excellently attended".11 A related brouhaha in the correspondence columns of The Times Literary Supplement was also initiated by officers of the Linguistics Association of Great Britain,12 who objected to the review in that journal, which described linguistics as being "in a state of extreme crisis verging on collapse". 13 Evidently, what worried the spokesmen of the professional linguists was the public image of their profession, rather than whether what The Language-Makers had to say about it was correctly interpreted or not. Indeed, their letter of protest showed no evidence of their even having read the book. It was subsequently put to me that I was at fault at that point in not taking steps to prevent this dialogue des sourds or attempting to vindicate the study of 5 January 1983, p. 420. Canadian Philosophical Reviews, Vol. 1 No. 5, p. 211. 7 November 1980. 8 8.8.80. 9 27.5.80. 10 Mrs M.J. Wilding, 9.6.80. 11 Dr N.V. Smith, 14.6.80. 12 G. Gazdar and N. Smith, 22.8.80. 13 T.P. Waldron, "For the Want of a Theory", TLS 11.7.80. 6
FROM AN INTEGRATIONAL POINT OF VIEW
233
general linguistics at Oxford. In fact I had tried to do so by writing to The Times to clarify the issue; but to my surprise The Times declined to publish the letter. In order to set the record straight, I here cite the relevant passages: Attitudes to language, and to the study of language, are an important barometer of a society's thinking. Upheavals in linguistic theory are not unconnected with such wider issues as the freedom of the individual, nationalism, the concept of a social collectivity, and Man's place in Nature. Such connexions provide one of the main themes of The Language-Makers... Quite a different matter is whether one finds oneself in agreement with the tenets of some particular brand of linguistic theorizing which may become academically entrenched at a particular time or place. At Oxford, students of linguistics will find a subject which is not disintegrating, but being reintegrated within a broader study of human communicative abilities. They will also, I hope, find a study free from some of the intellectual prejudices which have been built into academic linguistic orthodoxy in recent years. If that counts as hastening the demise of linguistics, I am happy to plead guilty. And I still am. These two paragraphs sum up what was then, and has remained, my position as both a critic and a teacher of linguistics. Whether at Oxford I did succeed in providing a study of the kind described above is a question I am happy to leave to the judgment of my graduate students. By the time this abortive letter had been written, most of The Language Myth was already in draft. I now realized the risk that the reception of its predecessor would colour the reading of The Language Myth; but in the end there is a limit to what one can do as an author to ensure that what one writes will be read intelligently. When the reviews began to appear, it was soon apparent that the alleged sins of the first book were to be visited upon the second. The distinguished philosopher who reviewed The Language Myth for The London Review of Books began by describing it as "a wholesale onslaught" on orthodox modern linguistics.14 His complaint was that the book failed to show that the alleged myth was in fact a myth (a complaint to which I shall return later in this essay). Linguists, however, having been alerted by The LanguageMakers, adopted a different defence against The Language Myth, trying to cry "Foul!" and "Déjà vu!" at the same time. According to the review in The Listener, at least seven other contemporary linguists had made similar criticisms of mainstream linguistic orthodoxy and had thus "done a great deal to attack the very
14 London Review of Books, 19.8.82.
234
ROY HARRIS
myth Harris is concerned about".15 Had this been true, the seven samurai ought to have been welcoming the recruitment of an eighth; but clearly they weren't. No shouts of support went up from their quarter, possibly because the seven samurai failed to recognize themselves in that role at all. This is hardly surprising, since none of them had ever diagnosed "the myth" as such, nor traced its historical provenance. A variant of the foul-cum-déjà-vu defence was adopted in The Times Higher Education Supplement,16 whose reviewer (G. Sampson) both denied the originality of attacking the thesis "that a language has something called 'grammar' which can usefully be discussed in isolation from the communicative purposes to which language is put" and also, for good measure, denied this to be a myth in any case; thus scoring two "own goals" in quick succession. For the thesis identified by this reviewer is nowhere attacked in The Language Myth; and a reader who had not made that initial mistake would see that to insist on discussing "grammar" in isolation from anything else already presupposes a particular kind of communicational purpose. The relevant issue, which the review failed to mention in spite of its being discussed at length in The Language Myth, concerns the nineteenth- and twentieth-century reification of "grammar" as a linguistic component which has a real existence independently of grammarians' communicational purposes. Reactions of this order doubtless reflected the difficulty which most academic linguists during the 1980s experienced in realizing the extent to which their discipline and their own thinking about language were still dominated by the orthodoxy which Saussure and his successors had built up over a period of two or three generations. There was - and still is - great resistance in universities to the notion that the study of language does not need to treat the description of languages as its central task; and even more resistance to the notion that linguistic theory can dispense with the concept of "a language" altogether as one of its basic conceptual tools. As long as that resistance persists, The Language-Makers, The Language Myth and The Language Machine will have a message for students of linguistics to ponder, regardless of whether readers come to agree or to disagree with it in the end. Sceptical reflection upon an orthodoxy must always be of value, for the reasons of which John Stuart Mill gave a classic exposition in his essay On Liberty. If the orthodoxy is fallible, scepticism will tend to weaken it. If the orthodoxy is well founded, scepticism will tend to prompt better defences of it. Neither outcome can reasonably be objected to. 15
The Listener, 17.9.81. The seven linguists named were: Hymes, Labov, Quirk, Trudgill, Halliday, Matthews and Bolinger. 16 16.10.81.
FROM AN INTEGRATIONAL POINT OF VIEW
235
That having been said, however, it must also be said that scepticism was in no way the major impetus for writing all or any of these books. They were intended primarily to focus students' attention on two questions: what purpose or purposes linguistic inquiry serves, and why linguistic inquiry, at any given time or place, should take the particular form it does. True, these are in one sense preliminary or metatheoretical questions. But to ignore them nowadays is to risk substituting a series of rather narrow and contrived scholastic exercises for what should be a great humane study of the broadest possible scope. Over the years since my election to the Chair of General Linguistics at Oxford, I have noticed, even among those not unsympathetic to the approach I was advocating, two types of general misapprehension concerning it. I think it worth while drawing attention to them here before I proceed to examine more specific issues. One misapprehension is the notion that what an integrationist basically objects to is the narrowness of the perspective adopted in orthodox linguistics, and the refusal by linguistic theorists to pay more than lip-service to the existence of other disciplines also concerned with the study of language and languages. Howard Gardner, for example, in The Mind's New Science, describes The Language Myth as "calling for a closer integration of fields",17 which suggests that the term integration alludes simply to the desirability of interdisciplinary co-operation in language studies. On this view, presumably, an ideal university language department would integrate not only linguists but psychologists, sociologists, philosophers, etc., all contributing their bit on language. But from my point of view that could be just as big a disaster as the present situation of academic apartheid. For if the representatives of these various disciplines all subscribed to some version or other of the language myth too, their co-operation would merely serve to strengthen its promulgation. In short, this is a misapprehension about what needs integrating. By "integrational linguistics" I do not mean - and never have meant - some kind of ecumenical movement in language studies. The other type of misapprehension is more subtle. Its point of departure is the fact that orthodox linguistics, throughout the twentieth century, has insisted on certain dichotomies. These dichotomies are presented as intrinsic to the study of linguistic phenomena: they are bifurcations which allegedly offer an ineluctable choice to any investigator. And for each such dichotomy, orthodox linguistics ordains a priority between the alternatives. These dichotomies include langue vs. parole, competence vs. performance, system vs. use, synchronic vs. diachronic, and linguistic community vs. individual speaker. On this basis, any challenge to 17
H. Gardner, The Mind's New Science. A History of the Cognitive Revolution, New York: Basic Books, 1985, p. 221.
236
ROY HARRIS
orthodoxy is easily interpreted as a simple reversal of priorities, or else an insistence that both elements of the dichotomy are equally important. I think it is worth quoting in this connexion some observations of Professor P.H. Matthews, discussing (in 1985) the future development of linguistics. Commenting on the structuralist inheritance apparent in generative linguistics, he wrote: The most rooted structuralist assumption is that the basic unit of study is "a language". But languages as the layman knows them are neither clearly delimited nor homogeneous; on what level, therefore, do the linguist's objects of description exist? There have been various answers, of which the most recent, argued by Katz in particular, is that they are wholly abstract entities, on the plane of Platonic ideas. But it is notorious that the more the concept of a language is idealised, the more difficult it is to relate it either to the practical knowledge of speakers or to the constant processes of linguistic change. To explain these we are forced to abandon idealisations and consider observable individuals. A new doubt must then arise. If the real objects of study are groups of speakers, is the notion of "a language" as fundamental, or even as necessary, as we have thought? There is no consensus on this issue... But one can detect at least two straws in the wind. One is Chomsky's remark, at the beginning of the eighties, that "knowing a grammar" is a fundamental concept, and "knowing a language" merely derivative. In context this means less than it might seem to mean in isolation. But the attention which it has received suggests that it is the kind of pivotal statement on which people may in the end be capable of building much more than was originally intended. The other is to be found in two books by Roy Harris, which argue that our reification of "languages" has its basis in a cultural myth inherited from western antiquity. If the myth is abandoned, the problem of their existence vanishes. Harris's work has attracted less attention, and his history of ideas is perhaps designedly simplistic. But a linguistic theory which is a theory of speakers and not of languages would be a logical conclusion to the critique of generativism that began towards the end of the sixties. Generativism was a development of structuralism, and with it a much older orthodoxy may be dragged down. 18 What I think Matthews has in mind here is an incipient priority-reversal, of which he sees both Chomsky's and my own doubts about "languages" as symptomatic. I describe it as a priority-reversal because clearly, for Matthews, 18
P. Matthews, "Whither Linguistic Theory?", Linguistic Abstracts, Vol. 1 No. 1, 1985, pp. 1-7.
FROM AN INTEGRATIONAL POINT OF VIEW
237
the alternative is "a theory of speakers". But this alternative is one set up in advance by the orthodox dichotomies: it is simply the other side of the orthodox coin. All that the integrationist does, on this view, is insist that we get the priorities right and redress the imbalance resulting from too much concentration on one side of a dichotomy. I have encountered similar views held by other colleagues, who seem to interpret integrational linguistics merely as calling for a more detailed study of parole as opposed to a study of langue, or of speech acts as distinct from linguistic codes. Esther Figueroa, for example, sees integrational linguistics as contrasted essentially with "autonomous" linguistics, and assimilates that contrast in turn to others which have been proposed (including speaker-centred vs. speaker-free, microlinguistics vs. macrolinguistics, and formalism vs. functionalism). 19 Those who would plot the place occupied by integrational linguistics on this chart of oppositions see nothing particularly revolutionary in it: on the contrary, it confirms their view that the basic framework of orthodox linguistics was right all along. All that has to be conceded is that perhaps orthodox theorists were dogmatically over-insistent on certain questions of priority and that the time may have come to revise those priorities. These construals of integrational linguistics, it seems to me, are missing an essential point. What an integrational approach calls in question is not simply the priorities but the basis of the dichotomies themselves.
2. On linguistic
theory
The passage I have chosen as an epigraph above clearly has a number of implications about the function of linguistic theorizing. Let me try to clarify some of these. A linguistic theorist speaks with no greater authority and insight about language than a baker or a bus-conductor; and I doubt whether it is possible to become a linguistic theorist of any stature without reminding oneself constantly of that fact. At best, the theorist speaks with greater circumspection. For it is difficult to do any serious work in linguistic theory over the long term without engaging in a constant process of self-criticism. The reason is all too obvious. A theorist's own linguistic formulations will be the first to come under fire from critics who disagree with the theoretical position those formulations present and defend.
19 E. Figueroa, Sociolinguistic Metatheory, Oxford: Pergamon, 1994, p. 21.
238
ROY HARRIS
A linguistic theorist is also more likely than the baker or the bus-conductor to speak about language from the basis of some historical knowledge about possible theoretical positions.20 If I were asked to put my own views about language into the historical context of twentieth-century linguistic theory and to explain why they take the form they do, the best short answer I could give would run as follows. My own linguistic theorizing would undoubtedly have been very different (although exactly how is another question) had I not belonged to a generation which had inherited a particularly paradoxical situation. We were at school during the second world war and our university education took place in its immediate aftermath. As students of language and languages, we were taught a linguistic orthodoxy which manifestly conflicted with our own linguistic experience. The war was a time when innovations of all kinds abounded in spoken usage and could not be ignored. The conflict between what we were taught and what we could observe for ourselves was blatant and pervasive, even if it went apparently unnoticed by our teachers. It affected the most elementary doctrines concerning grammaticality and vocabulary. Wartime English was clearly not the English described and exemplified in our school books. There's a war on was itself a sentence that defied orthodox parsing. It was a time when familiar expressions abruptly and inexplicably acquired new and sometime contradictory meanings {You've had it). Ubiquitous slogans suddenly appeared which it would have been hard to make sense of at all in peacetime. {Dig for victory is one I remember. It reversed all the usual linguistic associations involving swords and ploughshares. Similarly Is your journey really necessary? brusquely dispelled the aura of adventure surrounding the word journey) Old lexical patterns were thus disrupted. The language seemed suddenly to have been jerked out of its pre-war rut. Dictionaries were outdated virtually overnight. Americanisms flooded in and German words became English words from one day to the next. The realities of wartime English also undermined in advance some of the more sophisticated concepts which lay in wait for us at the university. We already knew that dialects were not the geographically circumscribed phenomena of the linguistic atlas. The upheavals of evacuation had long since taught us that speech went around with people: it was not mysteriously anchored to places. And sound laws were hard to reconcile with the inconsistencies of pronunciation that could be heard within the walls of a single wartime classroom. At no level did one ever feel convinced that the orthodox story actually made explanatory sense of one's own linguistic environment or the linguistic activities in which one was daily engaged. 20
Having said this, I immediately enter the caveat that some linguistic theorists display an ignorance of historical matters which is as deplorable as it is inexcusable.
FROM AN INTEGRATIONAL POINT OF VIEW
239
Language as lived and language as taught were thus, to students of my generation, patently as different as chalk from cheese. We had the sense of being taught something about language that might, perhaps, have been true once upon a time. But when measured against what was going on around us, the orthodoxy clearly did not achieve credibility. It was an educational con job, like being expected to believe that in Britain (a "democratic" country, we were told) all political and religious views were tolerated. Whereas it was obvious to the merest nincompoop of my age and social background that this alleged condition of tolerance in British society simply did not exist. Given the manifest disparity between our first-hand linguistic experience and what we were taught about language in school and at university, we were faced with an awkward choice. Some of us, not unreasonably, decided that the linguistic orthodoxy was antiquated rubbish (like much else the education system had to offer) and promptly lost interest in it. Others, more cautiously, decided that the experts must have got it right after all (because that was what expertise consisted in) and consequently wrote off the perceived disparity between linguistic orthodoxy and linguistic experience as being somehow on a par with the phenomenon of "experimental error", often invoked in the school physics laboratory when one's most careful measurements failed to produce the right results. Only a very small minority, of which I was one, opted for neither of the aforementioned solutions, but treated the disparity itself as a major source of linguistic interest. We never ceased to probe this disparity, much to the exasperation of those in charge of our language studies. That scepticism concerning what we were taught about linguistic matters went hand in hand, unless I am very much mistaken, with scepticism about much else. We were a difficult generation to teach. Those who "ought" to have been teaching us were for the most part away fighting a war which itself generated linguistic and other puzzles every day. When the teachers who survived came back to the schools after the war, they must have found their books out of date, their classes inclined to challenge the answers in the back, and classroom communication a major problem. We, the pupils, perhaps did not realize it at the time. I offer these comments merely as background to the main point I want to make here. Why isn 't language as they say it is? This, in retrospect, seems to me the most obvious, the most basic, the most inevitable question that was bound to emerge from the educational experience of my generation. And that formulation is already more polite than another which can readily be envisaged as an alternative: i.e. Why aren 't they telling us the truth about language? Anyone who takes either formulation of the question as seriously as it deserves to be taken is already halfway to becoming a linguistic theorist.
240
ROY HARRIS
But to go further than half way got progressively more difficult in the post-war years. Examinations, either at school or at university level, required orthodox answers to orthodox questions. The linguistic research one was required to do as a graduate student usually involved the overt or tacit acceptance of some current version of the orthodoxy itself. Junior teaching posts in higher education were open only to those who embraced the orthodoxy with enthusiasm and showed a precocious willingness to pass it on unquestioned to the next generation of students. In short, expanding rapidly in the universities of post-war Britain, linguistics displayed all the classic symptoms of what Ivan Illich calls "the disabling professions". In my own case, it took years before I was in any position to appreciate exactly what was going on in the academic politics of the subject, and even longer to understand what historical circumstances had brought about the bizarre lack of fit between linguistic theory and linguistic practice that had marked my own introduction to the study of language. When eventually I came to a conclusion on these matters, I found it to be shared by few colleagues. It was, indeed, an unwelcome conclusion in more ways than one. For it implied that the twentieth century's most important lesson in linguistic theory was among the least well understood. The lesson in question was epitomized for me in two sentences of the Cours de linguistique générale - a work with which I made my first acquaintance as an undergraduate at Oxford in 1953, although little realizing at that time its importance, or imagining that one day I might feel I ought to translate it into English. The passage I refer to reads: D'autres sciences opèrent sur des objets donnés d'avance et qu'on peut considérer ensuite à différents points de vue; dans notre domaine, rien de semblable... Bien loin que l'objet précède le point de vue, on dirait que c'est le point de vue qui crée l'objet, et d'ailleurs rien ne nous dit d'avance que l'une de ces manières de considérer le fait en question soit antérieure ou supérieure aux autres.21 Those who should have learnt most from this lesson - the linguists of the interwar period - either ignored it or else failed to grasp how radical its implications were. It had turned out, in short, to be a lesson that was far too difficult for them. For in effect it called into question what most linguists still wanted to take as the basis of their teaching; namely, the objective existence of linguistic facts, the objective existence of standard'languages, and the objective 21 Cours, p. 23.
FROM AN INTEGRATIONAL POINT OF VIEW
241
existence of distinctions between correct and incorrect pronunciations, constructions, meanings, etc. Thus instead of marking the end of the naive positivism that had pervaded the linguistic thinking of the nineteenth century, Saussure's distinctions between langue and parole, synchronie and diachronie, had in effect - it seemed to me - been interpreted à contresens; i.e. in spite of what Saussure said, as mere classifications of linguistic facts already given a priori. Nor has that situation greatly altered. The frequent use of the term linguistic data nowadays bears eloquent witness to the survival of that simplistic belief among contemporary linguists. (On this topic, see further remarks in Section 9 below.) It still seems to me that, despite the almost total incomprehension of later theorists (e.g. Bloomfield22 and his successors), what I take to be the original Saussurean view of the function of linguistic theorizing is in all essentials the right one. That is to say, because language does not in any particular instance consist of some determinate set of facts or events lying open to inspection, it falls to the linguistic theorist to develop perspectives from which a certain vision of linguistic phenomena becomes possible. And thus an order of some kind is introduced into what would otherwise be a hopeless confusion of irreconcilable observations. Where I now think Saussure to have been mistaken was in believing that there are just two basic perspectives (synchronic and diachronic) which somehow underlie all other possible linguistic perspectives, and in thinking that the synchronic perspective could be identified with that of the language-user. (He vigorously denied the possibility of a panchronic perspective.) Nor do I think that Saussure sufficiently emphasized the problems (both philosophical and practical) that arise from construing linguistic theory in this way. A perspective is a corridor with invisible walls. Or rather, walls that give the illusion of transparency. All that looking "through" them yields is a reflection of the wall opposite. In order to escape from this illusion, it is necessary to make the effort to discover or develop other perspectives. That effort is the proper domain of linguistic theory.
3. Far more and far less Let me now try to elucidate the sense in which an integrational linguistics needs to be concerned with "far more and far less" than orthodox linguistics proposes.
22
See, for instance, Bloomfield's review of the second edition of the Cours.
242
ROY HARRIS
The trouble with words, as Samuel Butler once observed, is that we expect them "to do more than they can"; 23 and no one has greater expectations of words than the linguist has. The very highest expectation that the linguist entertains in this regard is that words will make it possible to explain what language is. And this will indeed be more than words alone can do. That incapacity is the permanent paradox of linguistics. The explanatory power of words is limited by their conditions of existence. How do words exist? What are the requirements for their existence? These are fundamental questions. But they are questions ignored by the majority of language teachers. To recognize that words have no existence at all except by virtue of their integration into many and varied forms of human activity is the theoretical starting point of integrational linguistics. Integrational linguistics does not resolve the paradox stated above, but at least renders it fruitful as a basis for the study of language. On the other hand, although language is so intricately involved in human affairs, linguistic theorists have never (with the possible exception of Humboldt) had any great reputation for practical wisdom. Governments rarely consult them on matters of linguistic policy. Business executives do not solicit their views on the wording of contracts or advertisements. Scientists do not ask their advice about terminology. Parents do not call on their services for the naming of children. Nor is all this surprising; for the pronouncements of linguistic theorists make little if any contact with the communicational world of daily life. And although the learning of foreign languages and other branches of so-called "applied linguistics" flourish in universities, what they "apply" is the method of trial-and-error rather than any coherent set of principles drawn from general linguistic theory. Herein resides a second paradox: although linguistics claims to be a general "science" of language, and although words are intimately involved in so many human activities, what linguistics has hitherto had to say concerning that all-embracing involvement is minimal. These two paradoxes are related. Whereas linguists often paid lip-service to the integration of verbal and non-verbal activities, the methodology of linguistic analysis, on the contrary, required their segregation. Twentieth-century linguistics thus became segregational linguistics, treating words and their combinations as constituting an independent domain for investigation. 24 By means of this segregationist strategy, a situation was created in which the relevance of the integrational character of language could be denied or ignored, 23
The Notebooks of Samuel Butler, ed. H.F. Jones, London: Fifield, 1912. Repr. London: Hogarth Press, 1985, p. 94. 24 See the article cited in footnote 2.
FROM AN INTEGRATIONAL POINT OF VIEW
243
and the restricted technical terminology of linguistics could be made to appear perfectly adequate to describe and explain the workings of language. In effect, linguistic theorists set up a privileged sub-language of their own, which had the appearance of a powerful, high-precision linguistic tool, but actually made it impossible to address integrational questions other than via the adoption of segregational presuppositions. The attempt to say far more about far less resulted in saying far less about far more. We are dealing here with the consequences of failure to recognize the limitations of a perspective for what they are. The "far more" that an integrational linguistics must deal with accrues from recognition of the fact that although language is "a process of making communicational sense of verbal behaviour", language is by no means restricted to the verbal. (If it were, there would be little communicational sense that could be made of it.) The "far less" derives from being able to dispense with the cumbersome apparatus of rule-systems which that restriction brings in its train.
4. The telementational
model
It has sometimes been put to me that the telementational model, as described in The Language Myth, is a gross parody of a theoretical position, which cannot be taken seriously as something Western linguistic thought was ever committed to. I have occasionally replied to this by quoting the following passage from one of the leading figures in the Neogrammarian school of linguistics towards the end of the nineteenth century: In order to evoke in one mind a train of ideas corresponding to one which has taken its rise in another mind, the latter can do nothing but create by the action of the motor nerves a physical product, which in its turn calls forth the corresponding ideas, correspondingly associated in the mind of the other individual by exciting his sensory nerves. The most important of the physical products which serve this purpose are precisely the sounds of language.25 Here, it seems to me, is a quite sober and explicit statement of the view I am discussing, and it cannot be dismissed as "metaphorical" or otherwise discounted. My account of this view neither distorts nor embroiders it. In The Language Myth, this view of communication is traced from Aristotle down to Saussure via Locke. The account given is, to be sure, the barest outline. It could have been supplemented by documenting in far greater detail the different 25 H. Paul, Principles of the History of Language, trans. H.A. Strong, London: Longmans & Green, 1891, p.xxxviii.
244
ROY HARRIS
versions of the telementational process that theorists, and linguists in particular, have endorsed. For some readers, the outline given in The Language Myth was evidently too sparse. According to Michael Dummett, it was "excessively vague". 26 For it failed to allow anyone to determine whether or not Frege had "committed this crime" (sc. of being a telementationist). This objection raises two issues, one particular and one more general. The particular issue concerns one's reading of Frege. I find Dummett's doubts here somewhat puzzling. Frege makes his position perfectly clear. In his essay Der Gedanke, he writes: People communicate thoughts. How do they do this? They bring about changes in the common external world, and these are meant to be perceived by someone else, and so give him a chance to grasp a thought and take it to be true. Could the great events of world history have come about without the communication of thoughts?27 Again, in Gedankengefuige, he observes: It is astonishing what language can do. With a few syllables it can express an incalculable number of thoughts, so that even a thought grasped by a terrestrial being for the very first time can be put into a form of words which will be understood by someone to whom the thought is entirely new.28 These passages do not seem to leave much room for doubt (i) that for Frege, successful communication by means of language enables A and Β to grasp "the same" thought, and (ii) that Frege's account of the communicational mechanism is different from Locke's. The essential difference between them is that for Frege the thought has an independent, timeless existence and its "grasping" by A and Β is merely psychological and contingent. As he puts it in Uber Sinn und Bedeutung, "one man's idea is not that of another".29 Thus there is, strictly, no question of direct "thought transference", but nonetheless A can - by uttering the right words - prompt Β to grasp the same thought that A grasps. The more general issue Dummett's objection raises is simply this. Why does Frege, of all people, have to be wheeled on as the test case for a theory of 26
M. Dummett, "Linguistics Demythologised", London Review of Books, 19.8.82. G. Frege, Logical Investigations, ed. P.T. Geach, Oxford: Blackwell, 1977, p. 29. ™ Ibid, p. 55. 29 P. Geach and M. Black (eds.), Translations from the Philosophical Writings ofGottlob Frege, Oxford: Blackwell, 1966, p. 59.
27
FROM AN INTEGRATIONAL POINT OF VIEW
245
language? The choice is perverse, given the various specific versions of the telementational thesis discussed in The Language Myth. Even if it were an open question to what extent Frege would have subscribed to any of them, that is hardly a demonstration that the thesis in question has not been given a sufficiently clear historical identification. To suggest that reasonable uncertainty about the intellectual stance of particular individuals casts doubt upon the precision with which the telementational thesis is defined, or even on its existence as a position at all, is a curious line of attack. It probably tells us more about Dummett's own preoccupations than it does about the history of linguistic thought. Another complaint was that The Language Myth failed to spell out clearly enough what is fallacious about the "telementational fallacy" in the first place. Is it a "fallacy" at all? Is not Hermann Paul's description cited above (more or less) right? Do we not recognize ourselves as engaging in the very activity Paul describes? Dummett again: Words surely do have meanings: in listening to an Italian weather report, may I not be stumped by not knowing what nebbia means, and enlightened by looking it up or asking a friend? In some sense, speech serves to convey thoughts...30 So Dummett himself is evidently a telementationist "in some sense". His proclaimed puzzlement over what can possibly be fallacious about a thesis which so obviously accords with common sense might perhaps be taken as a parody of "ordinary language" philosophy. If we can ask for the meaning of the word nebbia and get a sensible answer, he implies, then this by itself shows that nebbia has a meaning; and, furthermore, since finding out the meaning fills a gap in our comprehension of the Italian weather report, is it not obvious that the function of the word is to convey a thought? So what can be wrong with the account of linguistic communication as Paul describes it? To some, this may seem an adroit way of hoisting the integrationist with his own petard. Dummett evidently fails to see the circularity of appealing to dictionaries and translations in order to establish that words do indeed have meanings and serve to "convey" thoughts. That is, he fails to see that dictionaries and translations are among the most powerful cultural instruments supporting the view of language which The Language Myth rejects. It is as if a sceptic had questioned whether the sun goes round the earth and been recommended to get up earlier in the morning, in order to be able actually to witness the sun beginning its circuit.
30
Dummett, loc.cit.
246
ROY HARRIS
We may perhaps give telementational descriptions (i.e. descriptions couched in telementational terms) if pressed to explain, for instance, how we suppose that the complete stranger accosted in the street managed to give us appropriate directions when asked "Which is the way to the station?" But just as adequate or inadequate - an explanation would be to say that we suppose the stranger we accosted must at some time have learnt how to speak English and knew how to get to the station. The same applies to Dummett's difficulty with the Italian weather report: we could say that he resolved it by learning a word he had not learnt previously. But this no more demonstrates that Hermann Paul must "in some sense" be right than it demonstrates the Tightness "in some sense" of the Dogon theory that speech comprehension is a bodily process akin to the digestion of food.31 Nor does Dummett's discovery of nebbia in his Italian dictionary demonstrate the reality of its telementational function, any more than his finding pizza available in Poggibonsi demonstrates the nutritional value of Italian cuisine. What is fallacious about the telementational fallacy does not turn on the issue of whether or not some psycho-physical chain of events in fact occurs when someone understands what is said. Nor on whether speaker and hearer conceptualize what is happening in telementational terms. The fallacy arises when the telementational description is promoted to the status of a theoretical model. From an integrational point of view, this is to mistake an explicandum for an explication. A more trenchant critic than Michael Dummett is John Joseph (this volume, Ch.l), who claims that my identification of the language myth is itself mythopoeic. In particular, he asks whether we can be sure that Aristotle or Saussure were ever committed to telementation. If these two important historical props are removed, does it not begin to look as if the thesis must be dubious? As far as I can see, Joseph's only argument is an argument ex silentio. If we look to the actual texts of Aristotle, he objects, we discover that Aristotle never says anything about transferring ideas from one mind to another. For Joseph, Aristotle located meaning in the mind, and treated meaning (or, at least, verbal meaning) as conventional; but to suppose that Aristotle thought of linguistic communication as a transfer of ideas would be "an unjustified and pointless logical leap" (p. 24). I would be inclined to agree with Joseph here if the facts were as he presents them. But he has seriously underdescribed the case. Joseph mentions two Aristotelian premisses, but omits a third, which turns out to be crucial. Before turning to that, however, let us simply ask ourselves what might be the possible theories of communication Aristotle had available to select from. In other words, 31
See R. Harris, "On 'Folk' and 'Scientific' Linguistic Beliefs", in S.L. Tsohatzidis, Meanings and Prototypes, London: Routledge, 1990, pp. 449-464.
FROM AN INTEGRATIONAL POINT OF VIEW
247
what theories were up and running in the intellectual climate of Athens in the fourth century BC? Well, there was Plato's account, which posits sempiternal "Ideas" as the ultimate source of meaning. But this account is one which Aristotle clearly rejects, along with the Platonic "Ideas" which are its mainstay. Nor, secondly, does Aristotle anywhere give grounds for assuming that he believes in some natural concordance between vocal sounds and what they signify. So what, apart from telementation, is left? I take Joseph to be saying, in effect, that we just do not know what Aristotle's theory of communication was. Or perhaps hinting that Aristotle may never have considered the question. Or else fought shy of addressing it. At least my interpretation gives Aristotle credit for being more intelligent than that. It seems to me highly unlikely that the founding father of Western logic and author of influential treatises on rhetoric and poetry had not given very careful consideration to the question of how human beings communicate. The evidence that he did so is provided by the three premisses that he spells out at the beginning of De Interpretatione (of which Joseph mentions only two). In other words, given (i) that meanings are "affections of the soul", (ii) that words are merely "symbols" of these (i.e. conventional tokens), and - most importantly (iii) that the external world is the same for all observers, as are the "affections of the soul" derived therefrom, then there is nothing more about communication for Aristotle to explain. Between them, assumptions (i), (ii) and (iii) constitute a telementational theory of communication. It is this that Joseph does not seem to see. The alternative would be to suppose that by mere chance, at the beginning of his introduction to the foundations of logic, Aristotle happened to mention three interlocking premisses from which necessary and sufficient conditions for verbal communication can be derived. I take Aristotle's position to be that there is no mystery about how logic "works", because the grounds for its working are already established in the common practices of verbal communication. It is therefore important to note that if any one of the three initial premisses is denied, then we are left with no basis for telementation and no grounding for logic either. It is perfectly true that Aristotle makes no attempt to elaborate the psychological mechanism involved. But he is teaching logic, not psychology. His concern is simply that no one shall be able to accuse him of propounding a logic which lacks coherent communicational foundations: or, what amounts to the same thing, offering a systematization which allows one person's syllogism to be another's non sequitur. Logic, for Aristotle, has to be interpersonal and public. If it were not, it would be a weak weapon indeed in the human intellectual armoury.
248
ROY HARRIS
Even more surprising, however, is Joseph's questioning of my view of Saussure as a theorist who believes in telementation. That is to say, I might have supposed that what Joseph is complaining about in my reading of Aristotle is that Aristotle never gives an explicit, detailed psycho-physical analysis of the speech act. But when it comes to Saussure, who gives just such an account, Joseph still finds room to doubt whether Saussure fits my description of a telementationist. Here I am simply baffled and can only attempt to reconstruct Joseph's reasoning. Having gone through the French text of the famous "speech circuit" passage, and my translation of it, Joseph seems to wish to fault my interpretation of Saussure on the following specific points: (i) Saussure never claims that in the speech circuit A and Β have to be speaking the same language, (ii) Saussure never claims that A intends to transmit ideas to B, (iii) Saussure never claims that Β interprets A9s utterance as meaning what A originally thought, and (iv) Saussure never claims that this is a model of "communication" anyway. Let us take each of these four points in turn. There is no warrant for assuming that Saussure produced a "speech circuit" theory in which the interlocutors are - or may be - speaking different languages. If A and Β are not ex hypothesi speaking the same language {langue), then the whole purpose of Saussure's account is frustrated. If we are allowed to assume that, for instance, A might have been a monolingual speaker of French and Β a monolingual speaker of Cantonese, then Saussure can no more be taken seriously as a linguistic theorist than Madame Blavatsky. (As in the case of Aristotle, my reading of what the theorist is saying, unlike Joseph's, at least allows the theorist to have a modicum of intelligence.) On point (ii), however, Joseph is right. I am glad to acknowledge this, because I have elsewhere made precisely this point, and used it to argue that Saussurean semantics (unlike the semantics of certain latter-day speech-act theorists) is not intentionalist.32 I think Joseph confuses telementation with intentionalism. There are, of course, theorists who try to give telementational accounts of speech acts in terms of interlocutors' intentions. But I am not among them. Nor was Saussure. When I write in The Language Myth about our total behaviour making "the kind of sense to others that we intend that it should" (p. 165) I am not endorsing any kind of formal attempt to reduce meaning to intentions (e.g. à la Grice), but simply commenting on the lay understanding of what verbal communication generally is about. This affords no more warrant for intentionalist semantics than saying that a batsman's aim is to score runs yields an intentionalist analysis of the cover drive. 32
This is what is wrong with Umberto Eco's reading of Saussure. See R. Harris, Reading Saussure, London: Duckworth, 1987, pp. 26-28.
FROM AN INTEGRATIONAL POINT OF VIEW
249
Point (Hi), again, suggests that Joseph thinks Saussure was intellectually confused about his own analysis. In other words, we are asked to entertain the notion that Saussure regarded his own "speech circuit" as a model which is neutral as between understanding and total misunderstanding. (A says cheval and Β thinks A meant "hippopotamus".) If this were plausible, Saussure would be the only linguistic theorist in the Western tradition who supposed that speaking the same language made no difference to whether people understand one another. Point (iv) confirms all this. If Joseph is right, we have to allow that the Saussurean speech circuit has nothing to do with linguistic communication at all. What, then, has it to do with? Here Joseph hops uncomfortably from one foot to the other. He says he does not wish to "contest" my interpretation of Saussure as a telementationist. But he does not say what alternative interpretation makes better sense. Nor does he bring forward any evidence to suggest that Saussure held a quite different view. As in the case of Aristotle, it seems to me that Joseph's academically fastidious doubts fail to give full weight to the fact that we are dealing with a language myth that is so deeply entrenched in the Western tradition that for more than two thousand years theorists and lay commentators alike have simply taken its validity for granted. The justification they offer is minimal because they assume it needs none: it is a matter of common sense. Robert Borsley and Frederick Newmeyer take issue with me on whether the telementational model of linguistic communication, which they condemn, is actually a model that generative grammar endorses. They say that the only evidence that this is so comes from J.J. Katz and dates from 1966. But they dismiss Katz as not being a typical generativist. That dismissal, needless to say, is very convenient for their argument, because Katz gives a very explicit and unambiguous statement of the telementational model. Borsley and Newmeyer go on to say: "We have found no more recent example of a generative linguist endorsing telementation" (p. 46). In that case, Borsley and Newmeyer have not troubled to look very far. For a start, Katz himself repeats his telementational account in 1972, in a book which explicitly acknowledges his indebtedness to Chomsky {Semantic Theory, New York: Harper and Row, p. 24). But even if we count Katz as an unrepresentative maverick, that hardly explains how one of the most widely used introductory textbooks in linguistics during the 1970s, written by two American generativists, tells students on page 1 that When you know a language, you can speak and be understood by others who know that language. This means you are able to produce sounds which signify
250
ROY HARRIS
certain meanings and to understand or interpret the sounds produced by others.33 How students were expected to understand this statement except in terms of speakers (successfully) communicating messages to hearers it is difficult to imagine. Another generativist no less explicitly committed to telementation is J.A. Fodor, who claims: Verbal communication is possible because, when U is a token of a linguistic type in a language that they both understand, the production/perception of U can effect a certain kind of correspondence between the mental states of the speaker and the hearer.34 Likewise, E. Matthei and T. Roeper in their Understanding and Producing Speech (London: Fontana, 1983) present a "model of the communicational process" which is manifestly a generativist version of Saussure' s, and introduce it with the comment: It takes almost no effort and very little, if any, conscious thought to turn our thoughts into words and sentences in order to communicate them to others; and, likewise, we ordinarily have no trouble in getting at the thoughts that others express in their words and sentences, (p. 13) The list could be extended. Nor can one argue that any allegiance to telementation within the generativist camp is now a thing of the past. Very recently, one of Chomsky's current colleagues at M.I.T., Steven Pinker, states plainly in a much acclaimed book (The Language Instinct, 1994) that the goal of linguistic communication is "to get information into a listener's head in a reasonable amount of time". How is it possible for the speaker to do this? It is possible because, according to Pinker, languages provide both speaker and listener with the necessary verbal equipment. The way language works, then, is that each person's brain contains a lexicon of words and the concepts they stand for (a mental dictionary) and a set of rules
33
V. Fromkin and R. Rodman, An Introduction to Language, 2nd éd., New York: Holt, Rinehart and Winston, 1978, pp. 1-2. 34 J.A. Fodor, The Language of Thought, Hassocks: Harvester, 1975, p. 103.
FROM AN INTEGRATIONAL POINT OF VIEW
251
that combine the words to convey relationships among concepts (a mental grammar).35 Thus equipped, clearly, we can all "get information into" one another's heads in a very "reasonable amount of time" and with remarkably little effort. Pinker even undertakes to demonstrate that, with this equipment, we have an extraordinary form of control over the minds of others. Simply by making noises with our mouths, we can reliably cause precise new combinations of ideas to arise in each other's minds.36 One can only conclude from all this that Borsley and Newmeyer have not read very widely in their own subject area. I note that their own definition of a "generative linguist" is one whose work (whether in grammar or elsewhere) is ultimately indebted to Chomsky's Syntactic Structures (1957) and Aspects of the Theory of Syntax (1965). All the writers I have cited above, including Katz, count as generative linguists under this definition. So Borsley and Newmeyer's ignorance of what generativists have claimed about linguistic communication over the past few decades is truly remarkable. (All the more so since they accuse me of "refusal to attend to what generativists actually say" (p. 64). It seems they have difficulty in following their own advice.) They also cite a passage from Chomsky which allegedly proves that their hero in person explicitly rejects telementation. But what the passage says is vague in the extreme. It simply acknowledges that in practice we use all kinds of clues to interpret what other people say, and that "little knowledge need be shared by the speaker and the interpreter". As an example of the explicit rejection of telementation, this is ludicrous: it does not even address the question. The more careful expositors of telementation (e.g. Locke, Saussure) always hedge the model about with caveats of various kinds. But that it is this model, as I have described it, which anchors all mainstream discussion of linguistic communication throughout the Western tradition I dc not think there can be much doubt. Nothing in any of the contributions to this volume causes me any hesitation in reaffirming this. On the contrary I find corroboration from an unexpected quarter, in the form of Philip Carr's paper (this volume, Ch. 3) . Carr points out (against Borsley and Newmeyer) - and quite rightly in my view - that there is a basic inconsistency between generativist preaching and generativist practice. He argues that the most 35
S. Pinker, The Language Instinct. How the Mind Creates Language, New York: Morrow, 1994, p. 85. 36 Pinker, op. cit., p. 15.
252
ROY HARRIS
recent Chomskyan version of an "I-language" does indeed "commit the generative enterprise to a version of the telementation thesis" (p. 65), and agrees that there is some substance to "Harris's valid accusation that generative linguistics subscribes to the myth of telementation" (p. 66). If Carr is right, Chomsky must be even more confused about language than I had hitherto supposed: that is, he fails to see his own commitment to the myth. The only alternative is to suppose that Chomsky is an extremely competent non-tel em entational theorist, who unfortunately happens to have great (performance) difficulties in saying what he means. (My own opinion is slightly different from Carr's: I think Chomsky has shifted his ground so often that he no longer knows what he means.) It is difficult to make any sense of one of the main planks in the generativist platform - namely, the notorious distinction between competence and performance - unless we take it as tacitly presupposing a telementational scenario of the Saussurean type. That is to say, performance corresponds to parole and competence to langue. A's linguistic competence is manifested in the kind of linguistic performance that functions in the speech circuit both as acoustic product of A's initiative and as auditory "input" to B's comprehension, while B's linguistic competence is manifested in correctly "decoding" that input. To say that A does not need Β at all in order to produce a linguistic performance is trivially true but irrelevant. A might well be standing alone in the middle of the Sahara. But if A had been alone in the middle of the Sahara since birth (and miraculously survived), A would have no linguistic competence to manifest, either qua speaker or qua listener. In short, A would have no language at all. However, my main point in this section is that - whoever we think held it or did not hold it - telementation, as a model of speech communication, will not do even in principle unless coupled with a fixed-code theory of the linguistic sign; and it is precisely this conjunction which generates the internal contradictions of those forms of linguistic analysis which are based upon it. Not only is it impossible to identify the invariant units of form and meaning which the model presupposes, but it is impossible to explain how speaker and hearer could independently come to be supplied with identical sets of such units in the first place. As for testing the model empirically, the tests are blatantly inadequate if any stranger's faultless directions to the station in response to your question are taken to show that the telementational process must have been successfully completed. Yet to insist that something more be shown in order to demonstrate successful communication is manifestly unreasonable. For any further theoretical demand goes beyond the demands we commonly place on language in our daily communicational affairs.
FROM AN INTEGRATIONAL POINT OF VIEW
253
Integrational linguistics must deal with far more and far less than the telementational model requires, because as a model the telementational model demands both too much and not enough.
5. Fixed
codes
As noted above, one of the things a telementational model demands (if it is to explain what Fodor calls "a certain kind of correspondence between the mental states of the speaker and hearer" and Pinker calls "getting information into a listener's head") is a fixed-code view of languages. Such a view is authoritatively embodied in that Western cultural institution, the dictionary, which lays down the lexical basis of the code. Borsley and Newmeyer try hard to wriggle out of the generativist commitment to fixed codes. They admit that "there is a sense in which I-language is in fact a (relatively) fixed code" (p. 48), but claim that it is not a fixed code in Harris's sense. So my objections to fixed codes are "irrelevant to generative grammar" (ibid.). Explaining why this is so, they cite vague waffle from Chomsky about "logical form". This is, allegedly, a linguistic "level" that "interfaces...with conceptual-intentional systems" (ibid.). (Readers, please note the computer jargon.) Thus, it appears, the "(relatively) fixed code" that generativists own up to is a set of correspondences between representations of logical form and representations of phonetic form. In short, they think they have answered the charge that generative linguistics is committed to a fixed-code concept by removing the code to some quite obscure area of abstraction, and redescribing it in thoroughly opaque terms. This is simply playing with words. Obviously what they wish to safeguard is a distinction between "sentence meaning" and "utterance meaning" (which in turn rests on the pre-generative Saussurean distinction between langue and parole). They cheerfully concede to me that "no discovery procedure reliably separates out the two levels of meaning", but then claim that although this "is true" it makes no difference. Right, it doesn't- if you are doing cloud-cuckoo-land linguistics, as Chomsky and his fellow-travellers clearly are. So what this in the end amounts to is, quite literally, that the generativists concede a fixed-code component in their analysis but claim it cannot be faulted because in practice no one can sort out a criterion which could possibly test it anyway. Thus it takes its place alongside the medieval doctrine of humours in the honours list of unfalsifiable theories of human behaviour.
ROY HARRIS
254
We do not have to follow Borsley and Newmeyer through this smokescreen of apologies, however. There is quite plain evidence of the acceptance of fixed codes in the generativist camp. Pinker, again, provides a clear example. His account of how we can "reliably cause precise new combinations of ideas" to arise in one another's minds would not work at all without the assumption that both speaker and hearer attach the same meanings to the same sounds. Pinker is quite explicit about this. The "power" of a word, he says, comes from the fact that every member of a linguistic community uses it interchangeably in speaking and understanding.37 This, for Pinker, is a crucial property of the word. Words are "a universal currency within a community". What does he mean by that, if not that they have fixed values? A word, he tells us, is not merely a person's characteristic behavior in affecting the behavior of others, but a shared bidirectional symbol, available to convert meaning to sound by any person when the person speaks, and sound to meaning by any person when the person listens, according to the same code.38 There could hardly be a clearer demonstration that, in spite of Borsley and Newmeyer's obfuscations, fixed-code theory is still alive, well and thriving at MIT.
6. Making
communicational
sense
For most people, linguistic theory will make little communicational sense until it is based on the assumptions they are forced to make in their own day-to-day communicational practices, until it illuminates those practices, and so acts as a guide to understanding their experience of language. But linguistic theory as so far developed in the twentieth century achieves none of these things. Its business is often conducted in algebraic formulae and technical jargon. Its premisses are philosophical abstractions. Its analyses are largely irrelevant to the linguistic questions which arise in the course of anyone's everyday existence. Its controversies reveal deep divisions among linguists concerning the aims and methods of linguistics itself 37 38
Pinker, op. cit., p. 151. Pinker, op. cit., pp. 151-2.
FROM AN INTEGRATIONAL POINT OF VIEW
255
This last point, at least, is not disputed even by linguists. 'Ίη the field of linguistics", writes F.P. Dinneen, "there is considerable diversity in the definition of the proper object of study, 'language'."39 This is independently confirmed by André Martinet's summing up of the situation: If asked point-blank what the object of their science is, I assume that few professional linguists would hesitate to answer that it is "language". But if asked what they mean by "language" serious divergences would soon appear.40 At first sight it might appear to be little short of scandalous that linguistic theory cannot offer an agreed definition of what language is. To claim language as one's special field of study, but to be unable to see eye to eye with fellow specialists on what that field of study comprises, seems to verge on self-contradiction. Furthermore, since language plays such a vital part in the lives of the great majority of the human race, it is initially baffling that an agreed answer to the question "What is language?" should apparently be so elusive. On closer examination, however, that elusiveness turns out to be not only predictable but enlightening. The basic bone of contention here is often presented as being an issue about whether it is reasonable to expect to find a set of features common to all manifestations of language. An outright rejection of this view was advocated by Wittgenstein: Instead of producing something common to all that we call language, I am saying that these phenomena have no one thing in common which makes us use the same word for all, - but that they are related to one another in many different ways. And it is because of this relationship, or these relationships, that we call them all "language".41 On the other hand, those who do not accept Wittgenstein's claim have alleged that Wittgenstein "gives no coherent evidence"42 to support the negative part of his thesis. Renouncing at the outset any search for "something in common to all that we call language" has accordingly been described as "dogmatic defeatism".43 39
Foreword to C.H. Brown, Wittgensteinian Linguistics, The Hague: Mouton, 1974, p. 7. A. Martinet, "Double Articulation as a Criterion of Linguisticity", Language Sciences, Vol. 6 No. 1, 1984, p. 31. 41 L. Wittgenstein, Philosophical Investigations, trans. G.E.M. Anscombe, 2nd éd., Oxford: Blackwell, 1958, § 65. 42 J. Bennett, Linguistic Behaviour, Cambridge: Cambridge University Press, 1976, p. 4. 43 J. Bennett, loc. cit. 40
256
ROY HARRIS
Those reluctant to be defeatist have tended to justify their optimism by proceeding to identify candidates for the set of common features that Wittgenstein denied. It is ironic that this very conflict over "all that we call language" should exemplify the communicational problems engendered by failure to put the question in an integrational perspective. From that perspective one can already see that there are various questions tangled together here, including "How is one to make sense, in this context, of the words all that we call language?' Assuming that problem solved, however, is there any need for a linguistics which is delimited by reference to "all that we call language" (whatever that might turn out to be)? From an integrational point of view, the confrontation between dogmatic defeatists and dogmatic optimists is maladroitly managed on both sides. Integrational linguistics does not need to accept prior circumscription by some definition of "language", irrespective of whether it be in terms of common features or in terms of some specified set of relationships. To conceive these as the alternatives is already to succumb to a more dangerous linguistic virus than either dogmatic defeatism or dogmatic optimism. It is to assume that a word like language has a hidden rationale, which specialists in linguistics either already understand or else will discover. That assumption itself conceals a question-begging linguistic premiss: that what a thing "really" is may be determined by looking to the use of its designation or designations. As an attempt to provide the study of language with a theoretically neutral delimitation of its field of inquiry, any such assumption is self-vitiating. Linguistics is one subject which cannot be mapped out in advance by direct appeal to the concept "language", since that very concept is part of the subject matter under investigation. The problem was already evident to Saussure, who tried to get round it by claiming that "all definitions based on words are vain".44 The bravado of this appeal to the Platonic notion of "real" or "essential" definition is astonishing, given that Saussure's own attempt to delimit and partition the field of linguistics flatly contradicts it. "It is an error of method," says Saussure, "to proceed from words in order to give definitions of things" 45 and he promptly proceeds to set up a linguistics based on assigning to the French words langue and parole technical definitions of his own invention, offering the justification that these definitions correspond to "real" linguistic distinctions. It is almost as if Saussure qua theorist of language subscribed to Orwell's dictum: "What is above all needed is to let the meaning choose the word, and not the other way about." 46 44
Cours de linguistique générale, 2nd ed., Paris: Payot, 1922, p. 31. Saussure, loc. cit. 46 The Collected Essays, Journalism and Letters of George Orwell, ed. S. Orwell and I. Angus, London: Secker & Warburg, 1968, Vol. IV, p. 168. 45
FROM AN INTEGRATIONAL POINT OF VIEW
257
But this would be a doublethink motto to adopt for any structuralism of the Saussurean variety. Perhaps Saussure thought he had made amends for this definitional chicanery by specifically including as one of the goals of linguistics "to delimit and define linguistics itself'.47 As Saussure states it, this is not so much a goal as a challenge to the linguistic theorist to escape from the Scylla and Charybdis of tautology and definitional regression. The integrational approach offers a way out of this dilemma, for it proposes a quite different way in which linguistics can "make sense". The signpost to this way out is already contained in a much quoted definition which identifies linguistics as "language about language".48 The author of this succinct and crystalline formula, although himself a professor of linguistics, perhaps did not fully realize its implications. If taken seriously, they mean that virtually every man, woman and child alive is, as a matter of daily routine, engaged in linguistics without realizing it. For language about language is inseparable from the ordinary conduct of linguistic intercourse in every society. To ask as simple a question as "What did you say?" and get a simple answer already takes us into the domain of linguistics. If we could neither ask such questions nor answer them, language itself would be a quite different activity from what we commonly understand it to be. To be sure, as an academic subject linguistics may well be interested in asking more sophisticated varieties of question about language. But the more sophisticated questions are based upon and do not belong to a different order from - the everyday questions. A linguistics which cannot make sense of the everyday questions first will have no sound theoretical foundation for dealing with more sophisticated questions in any case. Reflection on the processes of language learning points to the same conclusion. Words are not learnt separately from everything else, but as integrated and integrating elements in a whole continuum of acquiring knowledge. Questions about language are among the first questions prompted in intelligent children by the communication situations in which they find themselves. ("What does that mean?" "What is that called?" "Why is it called that?") Although such questions seem natural enough and unsurprising, the fact of their being asked at all is profoundly significant. They reflect a level of consciousness about language which is intrinsic to its acquisition and mastery. Language is from the very beginning not only a means of communication but a topic of inquiry; and these two aspects of language are intimately interrelated. 47 48
Saussure, op. cit., p. 20. J. Whatmough, Language: A Modern Synthesis, New York: Mentor, 1957, p. 18.
ROY HARRIS
258
Inquiry, through language, is established as a way of improving one's linguistic proficiency. The reflexivity of the process gives language a unique status in our experience of gaining information about ourselves and the world we live in. No form of general linguistic theorizing which does not take this fact as its primum mobile can in the end make sense.
7. Language
and
metalanguage
One way in which orthodox linguistics and philosophy of language have marginalized the reflexivity of language is by extrapolating from the lay distinction between speech and reported speech in order to set up "metalanguage" as a dependent domain separate from that of "language". A metalanguage is then construed as a second-order fixed code which allows us to describe first-order fixed codes. The advantage of this move is evidently seen as being that it allows linguistic analysis to escape from complete circularity. Although it is admitted that a language can be - and commonly is - used to describe itself (as in a monolingual dictionary), this is somehow regarded as a special case of the "normal" relationship between a language and a metalanguage. The doctrine of the language/metalanguage division notoriously fails to solve any of the traditional semantic paradoxes involving self-reference. In spite of that its chief tenet, the distinction between "use" and "mention", has acquired quasi-canonical status. Difficulties with the exact formulation of this tenet are brushed aside as minor technical problems which cannot be taken as seriously undermining a distinction that is not only required by formal logic but validated by plain common sense. For who could be so obtuse as to take the sentence Cicero is trisyllabic as saying something about the famous Roman orator rather than something about his name? However, far from making it easier to explain how it is possible that some words can be used to describe other words, the language/metalanguage doctrine complicates that very question absurdly. For if a puzzled British philosopher can be enlightened about an Italian weather report by being told - in English - "The Italian word nebbia means 'fog'", and similarly for any other Italian expression, then it would seem that the English language must include the Italian language as a subpart of itself. Or is the sentence "The Italian word nebbia means 'fog'" not an English sentence after all? If not, which of the other languages of the world does it belong to? It does not take a great deal of reflection to see that once the language/metalanguage doctrine is institutionalized, then either metalanguages proliferate ad infinitum or else any given language must contain both itself and,
FROM AN INTEGRATIONAL POINT OF VIEW
259
potentially, all others. Every fixed code thus conceals a fixed supercode, of which it is but one possible manifestation. Perhaps this might be regarded as a very modest metaphysical price to pay for professional reassurance that our everyday language about language is neither nonsensical nor self-deluding. The trouble is that paying the modest price does not guarantee one's peace of mind anyway. In this respect it is like all dubious insurance policies. From an integrational point of view, the inadequacy of the language/metalanguage doctrine is already signalled in the term metalanguage itself The most recent edition of a well-known dictionary of linguistic terminology defines it as a higher-level language for describing an object of study (or "object language") - in this case the object of study is itself language, viz. the various language samples, intuitions, etc., which constitute our linguistic experience.49 Here we see how the distinction between language and metalanguage is based on the notion that the latter comes as something extra to and later than a prior set of linguistic objects already given. But this is a typically segregationist notion which perpetuates the refusal to acknowledge the integrational character of language. It seems to have escaped most of my critics that the question of metalanguage is closely related to my scepticism on two issues: (i) whether linguistics is a "science", and (ii) whether there exists a "language of science" which has a privileged status vis-à-vis "ordinary language". To both issues I shall return below (Section 10). The more general claim about metalanguage I wish to enter here is the following. Integrational linguistics offers a radically different way of making sense of our ability to talk about what we say. It recognizes that we do not first of all learn a language and only then realize the possibility that there might be something to be said about it. On the contrary, that possibility is already intrinsic to our grasp of our own involvement in language from the very start. David Olson (this volume, Ch. 5) emphasizes the importance of metalanguage, and comments on the role of writing in this connexion. But I think the role of metalanguage is even more fundamental than Olson acknowledges. We gain initial admission to membership of our linguistic community by being given a name. Now the child who knows that he is called Peter knows something more than the dog who knows that he is called Rover. This "something more" is shown by the fact that the child does not merely react to his own name when it is 49
D. Crystal, A Dictionary of Linguistics and Phonetics, 3rd ed., Oxford: Blackwell, 1991, p. 216.
260
ROY HARRIS
uttered in his presence, as the dog does. It is not simply a question of Peter's realizing that he is being alerted, scolded, shouted at, etc. - in short, being brought into the action in some way - by means of the utterance of the word Peter. Or at least, for as long as he realizes no more than this, he knows no more than Rover does. What Peter must also come to realize is that this particular vocable is his name : which means coming to realize that the word Peter stands to him as the word Mary does to his sister, that people refer to him in his absence as Peter, just as he refers to his sister in her absence as Mary, etc. etc. Now this kind of knowledge is in one sense intrinsically metalinguistic; that is to say, in the sense that it involves, very precisely, being able to understand and answer metalinguistic questions like "What is your name?". The moment you realize that Peter is your name - that you are Peter - you are introduced to both the possibility and the necessity of metalinguistic discourse. In short, the metalinguistic dimension of language is not something ontogenetically separate, but is already integrated into the very conditions of linguistic enfranchisement. Without it, there would be no making sense of verbal behaviour. To convince oneself of this it perhaps suffices to imagine what English would be like if it had none of its present metalinguistic designations or predicates whatsoever and there were no corresponding metalinguistic speech acts available to its speakers. At first sight it might seem that this reduced form of the language would still be perfectly adequate for most communicational purposes. People would still be able to say "Hello!", discuss the weather, and so on. But then on further consideration we realize that in this metalinguistically impoverished form of English it would be impossible to ask people discussing the weather - or discussing anything else - to explain what they meant. A lack of metalinguistic terminology is radically unlike a lack of meteorological terminology. Having a language which lacks the latter is like having a map which does not show the annual rainfall, temperature, etc. But having a language which lacks the former is like having a map which has no key at all.
8. Language,
speech and writing
From an integrational point of view, one of the central weaknesses of orthodox linguistics, in both its structuralist and its generativist manifestations, is blind adherence to the doctrine of the "primacy of speech". Or, more precisely, the identification of language with spoken language. This is a theoretical muddle that cannot be laid at Chomsky's door: he simply inherited it - uncritically - from his immediate predecessors. Chomsky, as far as I can see, has no theoretical position at all on the relationship between speech and writing; and this is one of a number
FROM AN INTEGRATIONAL POINT OF VIEW
261
of reasons why his standing as a linguistic theorist is more questionable than his followers suppose. It was Saussure who provided the first theoretical rationale for the restriction of the linguistic sign to the spoken sign: and for Saussurean linguistics this restriction is semiologically essential. (The underlying reasons for Saussure's strategy I discuss in Reading Saussure, pp. 41 ff.) Saussure took his own predecessors to task for failing to realize just how and why this restriction was essential as a component of linguistic theory. In practice, the equation "language = speech" was already accepted by the Neogrammarians, but one has only to read Hermann Paul on writing to realize the distance that separates his thinking on the subject from Saussure's. The result of all this is that modern mainstream linguistics has, overall, taken - and promoted - a badly distorted view of what language is. This view has always been tailored to suit its own professional requirements in the academic market-place. Writing does not feature on the agenda, except to be marginalized. Basso accurately summarizes the position thus: Textbooks continued to include brief chapters on the subject, but this was to emphasize that writing and language were entirely distinct and that the former had no place within the domain of modem linguistics.50 From an integrational perspective, this is entirely unacceptable for various reasons, but here I shall focus on points relevant to the papers by David Olson (Ch. 5) and John Pettersson (Ch. 4). Olson clearly shares my view of the importance of writing, although it is less clear that he shares my reasons, or I his. In a previous article, he advances the claim that writing was instrumental in altering human cognition in such a way as to make possible the birth of modern science. More precisely, he holds that the contrast between texts and their interpretations provided the model, more than that, the precise cognitive categories or concepts needed for the description and the interpretation of nature, that is, for the building of modern science.51 I think this thesis goes too far and is historically unsustainable. Some, however, have gone much further than Olson. For instance, it has been argued that in the 50
K.H. Basso, "The Ethnography of Writing". In R. Bauman and J. Sherzer (eds.), Explorations in the Ethnography of Speaking, Cambridge: Cambridge University Press, 1974, p. 425. 51 D.R. Olson, "Literacy and Objectivity: the Rise of Modern Science", in D.R. Olson and N. Torrance (eds.), Literacy and Orality, Cambridge: Cambridge University Press, 1991, p. 152.
262
ROY HARRIS
difference between Eastern and Western writing lies the explanation of why Europe developed science whereas China failed to. 52 I cannot accept that thesis either. But the first point I want to make about both is that, from an integrational point of view, what they claim is not theoretically impossible: whether in a given instance it actually happened that way is a different question. However, I do think there is an integrationally defensible version of the thesis that writing necessarily has certain cognitive consequences, which I defend in "How does writing restructure thought?".53 But that paper emphasizes that how these consequences work out in particular cultures will depend on the cultural conditions obtaining: in short, the question cannot be decontextualized. Nor am I as sure as Olson appears to be that it is now demonstrably clear that the human mind "has not progressed through a series of stages up to our present one" (p. 100). This thesis seems to me to have more to do with certain currently fashionable forms of political correctness than it has to do with evidence about the remote past of the human race. I do not see how, for example, one could show convincingly that behind the hands that scratched the first proto-calendars analysed by Alexander Marshack54 there were minds essentially no different from mine. In fact, more generally, I cannot see how one draws the distinction between a mind and its mental equipment; and I feel as certain that these first calendar-makers did not have minds equipped like mine as I feel that at the age of two a person's mental equipment is usually considerably poorer than at the age of thirty. The basic reason for this, in integrational terms, is that at the age of two one has to cope with the integration of far fewer activities: consequently the dimensions of meaning accessible to one at that age are far more restricted than they are at the age of thirty. Olson cites Lévy-Bruhl's example of the Huichol equation between corn and deer as typical of a supposedly primitive mentality, one which "ordinary rules of logic" (p. 99) cannot cope with. But if we read Lévy-Bruhl more carefully, we find that the Huichol equation has an integrationally acceptable explanation. It resides, as Lévy-Bruhl points out, in the rituals associated with the harvesting of the hikuli plant, which are held to ensure the success of the corn crop, and the parallel rituals associated with deer-hunting. Now the question can be raised whether these rituals do not already presuppose the "deer = corn" equation. But that is a chicken-and-egg question. We do not know enough about the cultural history of the Huichol to answer it. The fact that 52 R.K. Logan, The Alphabet Effect, New York: Morrow, 1986. 53 R. Harris, "How Does Writing Restructure Thought?", Language & Communication, Vol. 9 No.2/3, 1989, pp. 99-106. 54 A. Marshack, The Roots of Civilization, rev. éd., Mt. Kisco, N.Y.: Moyer Bell, 1991.
FROM AN INTEGRATIONAL POINT OF VIEW
263
seems significant is simply that, however it originated, this is the equation that is maintained by the integration of various practices in Huichol culture. So ultimately we are not dealing with anything profoundly different from my belief that the figures on my bank statement show how much "money" I have got. This is neither more nor less magical than the Huichol equation. Very few people with bank accounts understand the historical process by which this equation came to be established, or the complex integration of financial and commercial activities involved: they are simply brought up to accept it. And they will continue to accept the correlation until it disintegrates; i.e. until economic conditions arise (such as hyperinflation) which entail that by the time I receive my bank statement I have already become far less wealthy than it says (or, pace Holiday, "shows"). And then I shall probably try to do something about it; e.g. stop putting my "money" in the bank and hide it under the bed, or buy diamonds, etc. If the disintegration proceeds far enough, monetary concepts will become meaningless, and society will revert to a basic process of barter (as happens when war or other disasters render the financial equations by which we live untenable). Although Saussure's use of an economic analogy to explain how linguistic signs have a meaning does not prove quite what he wanted it to prove, there is nothing wrong with it in principle.55 The source of meaning in human affairs is always a systematic integration of biomechanically diverse activities. And these integrations commonly lead to questionable equations. I see nothing less viable about the pre-literate Huichol equation between corn and deer than I do about the common literate equation between a spoken and a written form. ("Aren't they 'the same thing' in a different guise?" Well, in one sense they are. But whereas operating with that equation will facilitate programmes and insights for certain purposes, it will just as surely impose cognitive blinkers for others. Here again we encounter perspectives as corridors with invisible walls.) This is why I have no hesitation in supposing that, even for periods of human history about which I can adduce no direct evidence, the introduction of writing must at least have opened up potentially important mental resources that were not available previously. (It should, of course, be noted, as it is by Pettersson, that I do not equate writing with glottic writing, nor suppose that writing "represents" speech. For further disussion of these matters, see La Sémiologie de récriture and Signs of Writing) Pettersson invites me to spell out how a historiographer of writing should apply my theory of the origin of writing. He rightly points out that we cannot assume that our predecessors at any particular point in time "had a conscious notion of the principle of slotting" (p. 95). I entirely agree with him that my 55
See R. Harris, Reading Saussure, London: Duckworth, 1987, pp. 118 ff.
264
ROY HARRIS
theory "will not help the evolutionary-minded philologist to decide whether this or that notational system marks the birth of true writing" (ibid.). But I am not an evolutionary-minded philologist. If I had been, I would never have formulated the theory in the first place. Here I seem to detect another whiff of John Joseph's plea that people who say that smoking causes cancer should stop to consider what will happen to the tobacco industry. My answer, I am afraid, is that I do not think the long-term interests of society are served by the maintenance in perpetuity of the tobacco industry. All I can do is try to persuade them to divert their efforts to the production of something less noxious (see Section 15 below.) Of course, the less noxious product may not give the quick intoxicating effects of inhaling tobacco smoke. But the most important point in all this from an integrational point of view is that in a society which has become fully literate "speech" is no longer the same mode of communication - functionally or cognitively - that it was in the social phase preceding literacy. That is why when modern linguistics turns its back on the existence of writing, it deprives itself of any chance of self-understanding. The paradox is that although linguists nowadays purport to describe spoken language as a primal, autonomous activity, that description is always formulated from a literate perspective. There is no such thing as a pre-literate linguistics. As Sylvain Auroux puts it in his survey of the whole history of the grammatical traditions of the world, far from linguistic analysis being a pre-requisite for the emergence of writing, "les traditions linguistiques apparaissent toujours après la constitution du système d'écriture".56 His conclusion is: "c'est l'écriture qui est l'un des facteurs nécessaires à l'apparition des sciences du langage". 57 This is exactly what, from an integrational point of view, one would expect. Borsley and Newmeyer seem to regard Panini as a counterexample. But he is a very dubious one. Borsley and Newmeyer are evidently under the misapprehension that the earliest glottic writing in Pacini's part of the world dates from the reign of Ashoka. While it is true that no one knows the exact date of composition of the Astâdhyâyï, and informed guesses range over several centuries, I have yet to hear of one which puts it back before the emergence of the Indus valley script. (For an interesting discussion of the internal evidence indicating familiarity with a writing tradition, see Auroux pp. 50-52.)
56
S. Auroux, La Révolution technologique de la grammatisation, Liège: Mardaga, 1994, p. 36. Emphasis in original. 57 Ibid, p. 9.
FROM AN INTEGRATIONAL POINT OF VIEW 9. Verbal
265
behaviour
Through language we make sense of verbal behaviour, whether spoken or written. But through language we also make sense of much else. Part of the trouble with orthodox linguistic theory is that it tries to confine itself to the verbal and ignore the "much else". The integrationist, on the other hand, tries to situate the verbal meaningfully within the totality of relevant behaviour. One of the more curious spectacles of contemporary linguistics is the sight of generativists foaming at the mouth whenever the word behaviour crops up. Part of the explanation for this is that the only concept of linguistic behaviour their theoretical apparatus allows them is the ill-defined Chomskyan notion of "performance", which for them is not the product of the internalized "linguistic rules" which constitute "competence". Borsley and Newmeyer tie themselves in knots trying to maintain simultaneously both (i) that generativists are just not interested in "describing behaviour" (p. 61) (which presumably, as far as they are concerned, would be an enterprise badly tainted with "empiricism"), and also (ii) that generativist claims about the grammaticality of sentences are based on linguistic "data" (which presumably must come from linguistic behaviour of some kind). When one of their number (Lightfoot) rashly makes a quite overt claim about linguistic behaviour, Borsley and Newmeyer try to get out of it by explaining that he really meant to make a claim about linguistic rules instead. In short, they proceed exactly in the manner aptly described by Lakoff some twenty years ago, where the object of the game is for the author to show that he is a Good Guy by demonstrating that what he is doing is studying Competence (rah!) not Performance (hiss!), or in another variation, that he is a Rationalist (yay!) not an Empiricist (boo!).58 One of the difficulties hampering Borsley and Newmeyer in playing this game is that they do not have a very clear grasp of what their own team's position is supposed to be. Thus they object to my characterization of the generativist's hypothetical competence as involving "rules we have no option but to obey" and profess to wonder where I could possibly have got such an idea from. They need have looked no further than p. 8 of Aspects of the Theory of Syntax :
58
G. Lakoff, "Fuzzy Grammar and the Perfomance/Competence Terminology Game", in C. Corum et al. (eds.), Papers from the Ninth Regional Meeting of the Chicago Linguistic Society, Chicago, 1973, pp. 271-291.
266
ROY HARRIS
Any interesting generative grammar will be dealing, for the most part, with mental processes that are far beyond the level of actual or even potential consciousness. It seems fairly obvious that if there are rules operating beyond the level of consciousness then we have no choice about whether to follow them or not. That Borsley and Newmeyer actually fail to understand this point is shown by the way they rush in (pp. 61-2) to produce as counterevidence the mundane fact that it is possible to utter an "ungrammatical" sentence! In short, they do not even know how to apply their own "performance/competence" distinction.59 One critic with even worse hang-ups about "behaviour" than Borsley and Newmeyer sees the integrational approach as downright "behaviourism" (hiss! boo! hiss!). John Hewson {Canadian Journal of Linguistics, Vol. 37 No. 3, 1992, pp. 378-81) describes it specifically as "a return to the Behaviourism of the 30s and 40s". The accusation is odd, all the same, since integrationism is manifestly not based on allegiance to any specific brand of psychological theorizing. What evidently raised Hewson's hackles in particular was a remark in my paper "On Redefining Linguistics" to the effect that B" s opening the door in response to A's verbal request is a response which, although silent, qualifies as "linguistic" no less indubitably than A's vocalization. 60 From this, Hewson immediately leaps to the conclusion that what is being claimed here is "that the overt behaviour following a speech act such as Open the door is the meaning of the act". This, of course, is not my claim and never was. But nor, as far as I know, would it have been the claim of any intelligent behaviourist of the 30s or 40s. Bloomfield, whose name is usually associated with behaviourism in linguistics, goes to some trouble in the famous chapter on meaning in his book Language to establish a point that Hewson has obviously missed; namely, that from a behaviourist perspective it makes no sense to say simply that whatever response a speech stimulus evokes is its meaning. (Bloomfield distinguishes carefully between the distinctive and the non-distinctive features of a speech situation. Only the former, according to Bloomfield, contribute to the linguistic meaning.)6! 59
Various possible reasons for Borsley and Newmeyer's muddle occur to one. They may be confusing rule-following with rule-conformity. Or they may think it follows from producing an "ungrammatical" utterance that you knew what rule(s) you had to break in order to do it. Or they may suppose that refusal to conform to a (specific) rule proves that all rule-conformity is conscious. Which of these or other non-sequiturs they are guilty of is not clear to me. 60 In H.G. Davis and T.J. Taylor (eds.), Redefining Linguistics, London: Routledge, 1990, p. 43. 61 L. Bloomfield, Language, rev. ed., London: Allen & Unwin, 1935, p. 141.
FROM AN INTEGRATIONAL POINT OF VIEW
267
A linguistics which ignores behaviour altogether in favour of the "mental reality" of language is possible, but doomed to solemn self-stultification (as we see in Chomsky's case). On the other hand, a linguistics which tries to confine its attention to the "verbal" part of human behaviour and ignore everything else is attempting to do something that cannot be done. To appreciate the reasons why it cannot be done is already to adopt an integrational perspective. First, there are no autonomous systems of signs; and even if there were they would not include systems of linguistic signs (i.e. systems of the kind that people commonly call "languages"). To convince oneself of this, it suffices to try to imagine how a terrestrial linguist could possibly explain to a Martian how the English demonstratives this and that work without explaining also that human beings integrate the use of these words with non-verbal signs such as pointing, nodding and eye movement. If Martians have no eyes or hands, they will have a problem here. (Dictionaries that try to state the "meaning" of words like this without reference to non-verbal signs simply make a laughing stock of lexicography.62) Those who have made a study of the co-ordination of verbal and non-verbal communication, but from a non-integrational perspective, have found themselves forced to postulate the existence of a communicational subspecies called "paralanguage". No one seems quite sure what this comprises, and it poses an awkward problem precisely because it straddles the boundary between verbal and non-verbal behaviour. Some have taken it to include simply those phonetic features of speech which indicate a speaker's emotional state, attitude, social role, etc.; that is to say, audible indications of information relevant, or potentially relevant, to the communication situation but not part of the linguistic message itself. Others have included also kinesic features accompanying speech, particularly facial expression and gesture. ("Kinemes" and "allokines" have been postulated to accompany "phonemes" and "allophones".) But wherever theorists draw the limits of paralanguage, the basic problem remains. Is there in fact any way of identifying what constitutes "the linguistic message itself, as distinct from its non-linguistic (but informative) concomitants? From an integrational perspective, there is no "linguistic message itself. Or rather, the "linguistic message itself is an artifact of the theoretical perspective and the analytic methods adopted by linguists. There could be no clearer demonstration that this is indeed so than the solution to the paralanguage problem proposed by one authority on the subject: this is to suggest that we should 62
For example: "present or near in time or thought", "nearer at hand or more immediately under observation", "immediate past or future". None of these dictionary explanations would enable anyone to grasp that this object can mean simply the one I point to.
268
ROY HARRIS
recognize a "scale of linguisticness".63 What is revealing about the proposed scale is that at the "most linguistic" end we find phonetic features which can be most easily described in terms of closed systems of contrasts, while at the "least linguistic" end we find features which are difficult to handle in this way. In other words, the scale favours whatever best fits in with the linguist's chosen descriptive methods. There could hardly be a clearer case of the descriptive tail wagging the linguistic dog. An integrational approach at the very least enables us to see why it is so difficult for theorists to propose any non-arbitrary demarcation between verbal and non-verbal communication. Even if one disregards the linguist's prejudice in favour of the "primacy of speech", the same conclusion could be reached by considering writing. For instance, the reader who is at the moment engaged in reading this printed page is taking in a great deal of visual information. But to say how much of it is verbal and how much non-verbal is an impossibility. We might perhaps all agree that it was a verbal convention that a definite article in English should be spelled the. And it certainly makes sense to ask a reader to count the number of words on this page. (Though how the counting might be done is another matter.) But is it a verbal convention that the lines read from left to right, or that the pages of the book should be read from the left-hand cover to the right-hand cover, and not vice versa? Is it even a verbal convention that the word-order on the page should match the sequence of words in corresponding spoken sentences? Is it a verbal convention to indent paragraphs, or number pages? There are no clear answers to these questions, because we do not know where reading begins and reading ends; any more than we know where speaking begins and ends.64 So although the phrase verbal behaviour may be useful, it must not mislead us into thinking that somehow we can begin linguistics with a distinction already given (by Man, God, or Nature) between verbal and non-verbal communication. Having safely got over that initial cultural obstacle, it will doubtless be less difficult to see that there is no reason for treating language as something co-extensive with verbalization. If, when asked to repeat what I said, I repeat it, that is certainly a (partial but relevant) manifestation of my linguistic proficiency. But equally if, when asked to sit down, I sit down, that is no less a manifestation of my linguistic proficiency. (If the former is, then the latter is too: this is the point Hewson fails to see.) 63 D. Crystal, "Paralinguistics", in T.A. Sebeok (éd.), Current Trends in Linguistics, Vol. 12, The Hague: Mouton, 1974. 64 For further discussion of these questions, see R. Harris, La Sémiologie de l'écriture, Paris: CNRS, 1994 and Signs of Writing, London: Routledge, 1995.
FROM AN INTEGRATIONAL POINT OF VIEW
269
Although for various purposes we distinguish with no difficulty at all between linguistic and non-linguistic acts, it by no means follows that there is any one kind of activity that language is; any more than our recognition of the difference between working and not working means that there is any one kind of activity that constitutes work. Many serviceable everyday distinctions may be drawn differently in different circumstances, and this very flexibility is part of their usefulness. It is far from clear that the availability of some universally applicable definition of language is a reasonable expectation for linguistic theorists to have in the first place. It would be even more naive to suppose that this definition, once pinned down, divides what human beings do or know into two non-overlapping categories ("linguistic" versus "non-linguistic") as neatly as the concept "equator" divides the earth's surface into two non-overlapping hemispheres. But surely, it might be objected, even an integrational linguistics must accord some measure of primacy to the verbal component in communication? Certainly, no integrationalist would wish to deny that, at least in Western culture, we commonly think of linguistic acts as typically involving words. Our dictionaries (or at least some of them) go so far as to define "language" in terms of the use of words. Making a speech and writing a letter we regard as linguistic acts, or series of linguistic acts; and such acts are crucially dependent on the words used. A different speech and a different letter would need different words. But none of this provides the basis for a theoretically satisfactory definition of language, for a variety of reasons. One is that the concept "word" itself is very much the product of one particular (European) tradition of language study. The word has no status as a universally recognized linguistic unit. Nor is there any other such unit. (This is why it is rather silly for Borsley and Newmeyer to cite "prodigious feats of grammatical analysis" by illiterates (assiduously coached by linguistic researchers) in support of the orthodox linguistic analysis of those "feats". One might as well go the whole hog and say that as soon as a child can tell the difference between slap and snap it knows what a phoneme is.) Another reason is that to define language merely in terms of the use of words - even if that were acceptable on other grounds - involves a latent circularity. For to understand what "using words" is, for purposes of that definition, it is already necessary to understand words as having a linguistic function. The circularity is not obvious, but it is real nevertheless. It is rather like the circularity of explaining archery as the use of bows and arrows. Anyone familiar with the use of bows and arrows already knows what archery is, even if ignorant of the fact that archery is what it is called in English. The case of language is analogous.
270
ROY HARRIS
We are forced to conclude that "the use of words" is no more perspicuous a notion than "language". There are many cases in which the use of words may have no overt behavioural correlates to enable an observer to recognize it as such. We offer up "silent" prayers and repeat instructions "under our breath". These activities involve words, even though the words do not appear in spoken or in written form. Again, when two people are in conversation it is not only the speaker at any given moment who is involved in the use of words; for the listener too is engaged at that same moment in identifying and making sense of the words uttered. The phrase use of words is in some respects clumsy because it encourages a conflation between making use of one's knowledge of words and putting that knowledge into practice in speech or writing. The politician making a speech is using words in both senses. The audience listening makes use of its own knowledge of words in order to understand what the politician says. Listeners use the words provided by speakers as a basis for constructing their own (listeners') interpretations; and likewise readers use the words provided by writers. In order to clarify the notion "use of words", therefore, it seems that we need to distinguish at the very least between their active and their passive use, as well as between their overt and their covert use. The hang-up about behaviour seems to be central to Philip Carr's position. Unfortunately, I cannot make much sense of it. Carr attributes to me a notion of "linguistic behaviour" which, he says, is the same as Chomsky's. How anyone could come to this conclusion after comparing Chomsky's publications with mine I do not know. "For Harris, verbal behaviour and linguistic behaviour appear to be synonymous" (p. 78). Carr does not seem to have read "Language as Social Interaction: Integrationalism versus Segregationalism", where the point is explicitly made that in the right context a non-verbal act, such as passing the salt, has no less claim to be treated as "linguistic" than a verbal act, such as asking for the salt to be passed. How this could be construed as equating linguistic behaviour with verbal behaviour I cannot see. Carr further supposes that I think "there are such things as words, and they are externally observable in nature" (ibid.). Again, no. Words, as far as I am concerned, are second-order constructs, belonging not to nature but to culture. Those theorists who have seen that understanding what someone says also involves "using words" have sometimes blurred this insight by supposing that understanding must be some kind of automatic computational operation taking place in the brain of the hearer. This supposition, as Peter Hacker observes, should give us pause; for it leads to parallel conundrums about the speaker's understanding.
FROM AN INTEGRATIONAL POINT OF VIEW
271
How does the mind/brain compute the meaning of what its owner says? And when does it do so? Before he says it? Or after? Does he, as quick as a flash, construct a mental representation of what he is about to say before he has said it? If so, how does he understand his mental representation? And how does he know what to construct? Or does he wait until he has spoken before he knows what he means?65 These, as Hacker points out, are absurd questions, and their absurdity suggests that there is something fundamentally wrong with any linguistic theorizing which construes understanding what is said (or written) as internalized symbol-manipulation of some kind. Even when confusions of this order are avoided, the rest is by no means plain sailing. Further problems with the notion "use of words" accumulate around the issue of distinguishing between words and related signs which may in part be substituted for them. Many intellectual advances in the history of ideas have been achieved because human beings have developed specialized forms of symbolism. As a distinguished scientist has observed, "if biologists and physicists could communicate only in dumbshow, there would be no molecular biology today". 66 But are mathematical symbols, tabulations and diagrams, without which it would be equally difficult to conceive of molecular biology, merely surrogates for words? Or do they constitute different and independent modes of presenting information? Again the distinction between verbal and non-verbal communication resists any attempt to draw a simple dividing line. In order to understand the currency exchange rates given in this morning's newspaper, I have to understand a table in which there occur the graphic forms (i) US, (ii) $, (iii) France, (iv) Fr, and (v) 5.659. Furthermore, I have to work out that (v), which occurs in the bottom row of a column headed by (iii), gives me information about how many francs I can expect to get for a dollar. (But I also have to understand that, in Berkeley's sense, there "is no such thing as" six hundred and fifty-nine thousandths of a franc.) How much of all this is verbal information I have no way of telling. Nor does it matter. It does not prevent me from working out what I want to know. To ask what the status of the table is as regards the distinction between verbal and non-verbal communication is just a nonsense question. For a start, I have to understand the significance of the relationship between columns and rows. And in what compartment of human knowledge does that belong? 65
P.M.S. Hacker, "Chomsky's Problems", Language & Communication, Vol. 10 No. 2, 1990, pp. 127-148. 66 P. Medawar, The Limits of Science, Oxford: Oxford University Press, 1984, p. 72.
272
ROY HARRIS
Consideration of these cases reinforces the conclusion that the boundaries of human verbal activity are not the limits of language. The former fall well within the latter. Even if we adopt a liberal interpretation of what a word is, and take "verbal behaviour" to include not only the production and comprehension of words, but covert as well as overt uses, there remain large areas of human activity which are linguistic even though they may be non-verbal. Our involvement in language is not limited to the occasions when we ourselves give verbal expression to our own thoughts or feelings, any more than our financial lives are limited to our particular public transactions of buying and selling. Language is an iceberg, of which the words that are audible or visible represent only the tip. The rest of language, lying beneath the verbal surface, is far more extensive. As with the iceberg, however, it is the submerged mass that we must take into account if we wish to explain what projects above the surface. Integrational linguistics is essentially concerned with the whole of language, both above and below the verbal surface. But since metaphors may mislead, it must be pointed out that this "surface" and what lies below it should not be equated with the "surface structure" and "deep structure" of contemporary linguistic terminology, both of which are equally dubious theoretical extrapolations from the sound sequences which speakers utter and listeners hear. (The general public has recently been told by one of Chomsky's acolytes that even the master has now given up the notion of "deep structure", after having harped on it indefatigably for a quarter of a century.67 But who would not wish to applaud the discarding of linguistic illusions?) The integrational approach to language differs from orthodox linguistics of the kind which distinguishes (or distinguished) "surface structure" from "deep structure", because, unlike orthodox linguistics, it acknowledges that the extreme complexity of any distinction between verbal and non-verbal components of human communication poses a fundamental problem for linguistic theory. This complexity can neither be ignored nor be dealt with by arbitrary stipulations. Language does not present itself for study as a neatly disengaged range of homogeneous phenomena, patiently awaiting description by the impartial observer, as is suggested by the misleading expression linguistic data. This expression is much favoured by linguists who pride themselves on their empiricism. (Borsley and Newmeyer clearly regard me as some kind of arch-empiricist; but if they are observant, they will have noted that linguistic data is a term I never use.) On the contrary, language occurs nowhere as "data". Language offers a paradigm case of interference by investigation. The interference arises from the 6
7 Pinker, op. cit., pp. 120-1.
FROM AN INTEGRATIONAL POINT OF VIEW
273
fact that in linguistics language becomes both the object and the instalment of investigation, as well as the medium in which the linguist's conclusions are ultimately formulated.
10. Language,
linguistics
and science
This puts linguistics in a very curious position, the implications of which its professional practitioners have either lost sight of or tried to ignore. Since the first half of the nineteenth century at least, they have repeatedly claimed that their subject was a "science". I realize that my scepticism on this score (see e.g. my 1992 paper "On Scientific Method in Linguistics"68) is something that many linguists find shocking and even, as one of them put it, "disloyal". If linguistics had a trade union I should clearly have been drummed out of it long ago. But more shocking still, at least to some, is my more general scepticism about the "language of science". 69 For this is seen as tantamount to a denial of the possibility of scientific knowledge altogether. Anthony Holiday (this volume, Ch. 6) picks out for commentary my remark: Science, philosophy and language, after all, are presumably words like any others (at least, according to the author of the Philosophische Untersuchungen).70 I find it very interesting that he fastens on this particular observation. What he reads into it clearly scandalizes him. He treats it as "dismissive...of Wittgenstein's methodology", a "rhetorical flourish", an insinuation that "Wittgenstein's prohibition on philosophical theorising was groundless and arbitrary". Where he gets all this from, frankly I cannot see: I do not remember it being in the book that I wrote and he is quoting from. What is clear from his paper, however, is that he and I are not on the same wavelength either about Wittgenstein or about the language of science. Perhaps the problem is that Holiday thinks that in Language, Saussure and Wittgenstein I was trying to write a quite different book. That might perhaps explain why he seizes on just one remark in it, and ignores all the rest. But I do make it very clear in the Preface to LSc&J^what my (very limited) objective was, and I am surprised he makes no mention of that. The fact is that no previous 68
In G. Wolf (éd.), New Departures in Linguistics, New York: Garland, 1992, pp. 1-26. See "The Scientist as homo loquens", in R. Bhaskar (éd.), Harré and his Critics, Oxford: Blackwell, 1990, pp. 64-86. 70 R. Harris, Language, Saussure and Wittgenstein, London: Routledge, 1988, p. 125. 69
274
ROY HARRIS
publication, either on Saussure or on Wittgenstein, had ever drawn attention to the way in which the positions taken by Saussure and Wittgenstein on linguistic issues bring them face to face with problems that show remarkable similarities. It was not my aim to propose an overall interpretation of these parallels: on the contrary, I specifically said that I left open the extent to which they might be significant or deserved further exploration. One of these points on which comparison can be drawn is "Language and Science", which is the title of the chapter from which comes the sentence that so upsets Holiday. Now it would be perfectly reasonable for Holiday, or anyone else, to put forward a case trying to show that the comparisons invoked in this particular chapter are either insignificant or invalid. That would be precisely the kind of investigation the book invites. But, curiously, Holiday makes no attempt to do this: in fact, he does not mention any comparison at all. It is rather as if I had published a study pointing out parallels between Shakespeare and Racine and found myself confronted by a critic who ignored Racine altogether and devoted his entire critique to objecting to a single remark about Romeo and Juliet. Holiday does not actually dispute what I say about Wittgenstein and science: he complains about my "omissions" and "weak perceptions" rather than my "having set down what is straightforwardly false" (p. 112). In short, he does not like my selection of evidence or my reading of that evidence. That Holiday has his own reading of Wittgenstein is clear, and he is perfectly entitled to it. But the Philosophische Untersuchungen is rather like the Bible (which it seems to have supplanted for some philosophers) in the multiplicity of readings that can be got out of it. There is no agreement among experts about what Wittgenstein is actually trying to argue on various crucial points. (Witness the heated debate in recent years about his view of "rules". Or the fact that some commentators detect flagrant inconsistencies between Wittgenstein's account of philosophy and his own practice of it, whereas others deny their existence.) Moreover, as with the Bible, there is no guarantee that the most subtle and sophisticated readings are actually right or the most simple-minded ones necessarily wrong. However, to repeat, the purpose of Language, Saussure and Wittgenstein was never, in any case, to propose some comprehensive reading of either of the two principal figures, but to look for possibly instructive parallels. This inevitably involved focussing selectively on comparable features in the work of both. (It could, doubtless, be argued that the enterprise thus conceived was entirely w/sconceived and should never have been undertaken in the first place; but Holiday does not argue that case either.) Now why do I select "language and science" as one of the topics for comparison? Because, from an integrational point of view, that topic is a litmus-paper test of certain crucial issues in discourse about language. Both
FROM AN INTEGRATIONAL POINT OF VIEW
275
Saussure and Wittgenstein are theorists at odds with their contemporaries on the "scientific" status of their respective disciplines. As I say on p. 122 of LS& W, Wittgenstein's blunt remark "we are not doing natural science" could stand as an epigraph to the chapter in Saussure's Cours entitled "Objet de la linguistique". An integrationist will argue, however, that the question cannot be so easily dismissed. For there are essential aspects not only of language but of communication in general which cannot be investigated without "doing natural science". They are those which fall under the "biomechanical scale" of inquiry (see Section 15 below). What requires careful thought is exactly how doing these necessary bits of natural science is related to the broader semiological enterprise. I confess that it requires more careful thought than I have so far given it. But I am clear that neither Saussure nor Wittgenstein is much help on this. For Wittgenstein the question is of no interest (for reasons which doubtless Holiday can explain better than I can). Saussure, on the other hand, supplies an answer, but it is patently flawed if not question-begging. As I point out in Reading Saussure, he had evidently not thought through the whole complex of problems surrounding the relationship between phonetics and phonology. But these are problems which do not even appear on the horizon in Wittgenstein. If we are not engaged in science, then it follows that we do not need a "language of science" for our investigations. Wittgenstein is perfectly consistent on this. The discussions in the Philosophische Untersuchungen are conducted in "ordinary language". As much cannot be said for Saussure's Cours. Nor can it be said of mainstream linguistics since Saussure's day. On the contrary, linguistics has retreated further and further into the obscure recesses of its own "language of linguistic science", which is now virtually incomprehensible to the general public. It necessitates the production of specialized dictionaries and glossaries, just like the natural sciences. (And this fact is taken by the naive to demonstrate that it is a science.) Whatever else the jargon of linguistics may be, it is certainly not "ordinary language". Sooner or later, therefore, any serious linguistic theorist has to confront the question of the relationship between "ordinary language" and the - or a "language of science". There are basically two options, which I discuss in "The Scientist as homo loquens" and there dub "semantic continuity" and "semantic discontinuity". Wittgenstein's position is absolutely clear: he is a "semantic continuity" theorist. For him, the symbolism of chemistry and the notation of the infinitesimal calculus are merely "suburbs" of ordinary language. Saussure, on the other hand, although he does not state his position on the issue as clearly as Wittgenstein, is basically a "semantic discontinuity" theorist. So also, in practice, are the majority of contemporary linguists.
276
ROY HARRIS
The first point I wish to make about this is simple: that here we have, although most linguists fail to recognize it, a crucial issue in linguistic theory. You cannot seriously set up in the business of "describing language(s) scientifically" unless you are prepared to come clean about the status of your own descriptive language. The integrationist position on this is plain and accords far more closely with Wittgenstein's than with Saussure's. For an integrationist, linguistics and the language of linguistics must either be "lay-oriented" or they are not worth the proverbial row of beans. There is a second point, also discussed in "The Scientist as homo loquens", which is of some importance here. Since the seventeenth century (at least), the development of a "language of science" in the West has been based on what is, from an integrational point of view, a fallacious view of language. To spell it out briefly, scientists have tried to develop their own "fixed code(s)" on surrogational principles. Hence they think of the "language of science" as one which progressively develops in the direction of providing both the basic vocabulary and the ultimately "correct" definitions for terms that "stand for" independently given entities in the external world. (The example I cite in "The Scientist as homo loquens" is the definition of copper) The locus classicus for this conception of a "language of science" is Wilkins's Essay towards a Real Character, and a Philosophical Language (1668). The Royal Society of London held out great hopes for Wilkins's project. These were ultimately disappointed. But, in spite of its practical failure, the principle of Wilkins' project has, paradoxically, remained as a scientific ideal. The chemist who adopts the designation H2O for what the lay community calls "water" is implicitly claiming thereby not only to know what water "really" is but to have displayed this knowledge in the designation (two atoms of hydrogen combined with one of oxygen). Here we come to a distinction between "showing" and "saying" which I could have wished Wittgenstein and Holiday between them had told us more about. Those who think these philosophical issues about language and reality are a far cry from twentieth-century linguistics should read Bloomfield. Here is an important mainstream theorist who does not hide behind the kind of evasion that generativists later adopted. Bloomfield does not hesitate to admit that linguists cannot tell you the meaning of the word love (because science has not (yet) discovered what love is). But linguists can tell you the meaning of the word salt, because scientists know that salt is sodium chloride (NaCl).71 Although I do not agree with Bloomfield on this, I admire his direct confrontation of the issue as much as I deplore his successors' failure to confront it.
71 Bloomfield, op. cit., p. 139.
FROM AN INTEGRATIONAL POINT OF VIEW
277
John Joseph urges me to retract or stop reiterating my scepticism about the scientific status of linguistics. (Is this a condition for re-admitting me to trade-union membership?) He writes: "To drop the denial that it [sc. linguistics] can be a science would...be a move coherent with the general spirit of integrationalism." His proposal is that we should construe the "science" of language as consisting in the rhetorical construction of a variety of accounts (myths, if you prefer) of linguistic behavior - not all equally valid, but subject to evaluation against criteria that are neither independent nor universal, but are themselves part of the same language game within which the construction of theories is taking place (P. 41). Because I value this proposal, I would not wish Joseph or anyone else to construe the following comments as dismissive. First, I see no immediate prospect that other players in the game would accept this interpretation of their role in it. (See e.g. Borsley and Newmeyer (this volume), who clearly see their version of generativism as the only game in town. Apparently their sole concern about integrationism is that innocent onlookers might be misled into believing that there is (at least) one other game being played.) Second, granted that we accept Joseph's proposal, what exactly is the point in any longer claiming that what we are doing is "science"? The rhetoric of the term linguistic science belongs to the nineteenth century. I think the world has moved on since then.
11. Language
and rules
What has certainly survived from the nineteenth century, however, is the rhetoric of rules. "Is Roy hard or soft on rules?" asks Rom Harré (this volume, Ch. 7, p. 145). Well, Rom should know if anyone does, since he has had a longer academic acquaintance with me than any other contributor to this volume. But I am glad he raises the question because it gives me an opportunity to make some further comments on this topic and on my reading of Wittgenstein. If I describe this reply as being "ad Rominem", I do not need to be able to prove to myself that I was not following a linguistic rule; for the simple reason that I know there is no such rule. The expression ad Rominem belongs to no language (which does not stop it from being perfectly viable for purposes of linguistic communication). Someone could, doubtless, claim that I must have been following such a rule or rules without knowing it. They might even come up with the formulation purporting to capture the process by which ad Rominem
278
ROY HARRIS
was "generated". But the notion that we follow rules without knowing we do or even what they are is the beginning of mystification (see Section 9 above). It would be rash of me to claim that my position on language is more Wittgensteinian than Wittgenstein's, but I do not feel that in the end Wittgenstein was entirely successful in carrying out his own admirable programme to liberate our thinking on these matters from the "bewitchment" of language. While I agree with Wittgenstein that "Die Sprache ist ein Labyrinth von Wegen", I think Wittgenstein sometimes got lost in the labyrinth himself. And one of the ways in which it is only too easy to get lost is by overworking this metalinguistic term rule. Rom's interpretation of "what Wittgenstein meant" in § 242 of Philosophische Untersuchungen is highly questionable. First, if what Wittgenstein meant by "agreement" (Ubereinstimmung) was simply "having been subjected to a common regime of training" (p. 141), as Rom claims, why did Wittgenstein not say so? (It is a curious gloss to put on the word: you would not find it in any dictionary.) Second, and more important, if nevertheless that is the right way to interpret it, then as far as I am concerned Wittgenstein's thinking here is very inadequate indeed, for at least two reasons. One is that in language there simply is no common training programme that we are all subjected to. To imagine that such a programme exists and is provided by society itself (even though as individuals we may not realize it) is to invent another extension of the language myth. A no less powerful objection is that to postulate such a programme is in any case merely a postponement of the problem. That is, it now remains to explain how the training programme actually produces the conformities required for communication. And here we glimpse what looks suspiciously like the beginning of a regress. For that explanation itself seems likely to need the postulation of prior uniformities among the trainees. Nor am I very happy with supposing that all Wittgenstein understood effective communication as consisting in, at least in some cases, was "getting something done" (ibid.). (The term used in § 242 is Verstandigung, for which "communication" is not perhaps a fully adequate translation.) But again, if he did, then it seems to me he was not getting very far with the problem. I see no reason at all to retract my remark that "Wittgenstein never tells us where grammar comes from". He doesn't. Of course, we can make hypotheses about where he thought it came from, as Rom does here. But that does not alter the fact that Ludwig himself is silent. And again, I do not care much for Rom's "training" explanation, for reasons already indicated above. Finally, it seems to me that Rom himself provides a very good example of how talk about "rules" can confuse when he bemuses himself with the question of what "the sets of rules" expressing linguistic normativity "represent". He
FROM AN INTEGRATIONAL POINT OF VIEW
279
comes up with the answer: "It can only be the causal processes in a neurophysiological mechanism" (pp. 146-7). A few simple questions are in order here. 1. Is there not an important difference between rules and rule-formulations? 2. Which are we talking about when we say that they "express" something? 3. Why should we suppose that in addition to "expressing" something they also "represent" something? Do they "represent" anything at all, other than the end-product of the metalinguistic process which conjured them into existence? I suspect Rom has here fallen into the usual surrogationalist trap. Having formulated a "rule", we have to find something "real" for this formulation to stand for. In spite of all the above, I do not think I am such a sceptic about rules as I am occasionally accused of being. I am perfectly happy with rules of the kind we find in old-fashioned grammar books. I know how they got there and why. But I am far from happy with any modern metalinguistic mythology that treats rules as more powerful and mysterious things, allegedly "representing" how our brains work, our linguistic "knowledge", our social norms, etc. As soon as I hear talk of this kind, I do not need reminding that the battle against the bewitchment of our intelligence by means of language is far from over. Perhaps it has only just begun.
12. Language,
art and
creativity
The topic of rules leads almost inevitably on to that of creativity. Trevor Pateman (this volume, Ch. 11) makes some interesting suggestions on the subject of creativity and asks us to reflect upon Kant's distinction between determinant and reflective judgement. He describes Kant as wanting to have "beauty in art without a rule book for beauty" (p. 226). The term creativity, to be sure, is much bandied about by linguistic theorists; in particular by generativists, who pervasively confuse creativity with productivity. This muddle, another of our many intellectual debts to Chomsky, is exactly what one would expect from those committed to a mechanistic view of mind. I shudder to think how many times during my academic lifetime I have heard generativists trot out the old "isn't-it-amazing-that-we-can-produce-andunderstand-sentences-we-have-never-heard-before?" routine. If the Nobel committee were to award an international wooden spoon for Asinine Argument of the Century this would certainly be a strong contender. The conclusion to which the argument is intended to lead is that in order to explain such amazing feats as our ability to ask the way to the station it is necessary to postulate a prior mastery
280
ROY HARRIS
of linguistic rules. Following the rules will then account for an otherwise inexplicable but commonplace human achievement. Those who deploy this argument overlook two simple but fundamental objections. First, postulating rules merely postpones the problem and starts us on a cognitive regress. For what then has to be explained is the ability to select and apply the right rules in the right cases. (That, and not rule-following per se, would be the only "creative" part of the business.) Second, producing or understanding an utterance of a form not previously encountered in one's own experience is no more "amazing" than being able to make a variety of sandwich not previously known, walk without guidance down a street one has never before visited, or enjoy a landscape one has never before seen. If human beings were restricted to doing only those things where there were rules to follow or which they had already done before, the race would have become extinct long ago. Nor does homo sapiens need a mysterious, genetically endowed "language organ" in order to engage in the creative activity of talking, any more than a genetically endowed "sandwich organ" is required in order to make sandwiches. All this carries over to art. Being able to paint a picture no one has ever painted before is not "amazing", and having an internal "art organ" is necessary neither for a Rembrandt nor for a child experimenting with its first box of paints. The integrationist view of creativity is, I think, fundamentally different from that which produces the generativist conflation with productivity. In an earlier publication, Newmeyer quotes with approval an observation of Chomsky's which - as it happens - pinpoints the source of that conflation at the same time as it highlights Chomksy's prescriptivist attitude to art. I think that true creativity means free action within the framework of a system of rules. In art, for instance, if a person just throws cans of paint randomly at a wall, with no rules at all, no structure, that is not artistic creativity, whatever else it may be.72
It is interesting to note how Chomsky here slips in the weasel word true to qualify creativity. This allows him discretion to refuse the title of "true" artist to any artist who takes a quite different view of the creative process. (Picasso, not to mention Jack the Dripper, will fare badly here.) The insinuation, clearly, is that there is a "God's truth" about creativity, which Chomsky has found where others have failed. An integrationist would say that throwing cans of paint randomly at a wall may or may not be a manifestation of artistic creativity, depending on the context. 72
Quoted in F.J. Newmeyer, The Politics of Linguistics, Chicago: University of Chicago Press, 1986, p. 79.
FROM AN INTEGRATIONAL POINT OF VIEW
281
Furthermore, for some participants/observers it might be creative, but for others not: for different individuals will contextualize the episode in different ways. But there is in any case no question of the need for a prior rule about paint-throwing to be in place in order for the issue of creativity to arise. A more interesting question - for those who are not committed to mechanism - is whether art and language involve the same kind of creativity. Pateman mentions Croce in this connexion. Appositely, because Croce is, as far I know, the first theorist actually to have identified general linguistics as a form of aesthetics. (The full title of his famous treatise is Estetica come scienza delVespressione e linguistica generate ) Since I have elsewhere quoted with approval Croce's dictum "il linguaggio è perpetua creazione", I suppose it could conceivably be thought that integrationism rests ultimately upon hidden idealist foundations. (NB This would be the exact opposite of Holiday's theory that I suffer from "blindness to a spiritualised, and in the end holistic and unashamedly idealist, view of people and their speech" (p. 133).) So it might be worth pointing out that adopting Croce's dictum does not necessarily mean agreeing with Croce's reasons for advancing it. Nevertheless, I do not mind admitting that I would also endorse the passage that immediately follows the aforementioned dictum in the Estetica: cio che viene espresso una volta con la parola non si ripete e non appunto come riproduzione del già prodotto; le sempre nuove impressioni dànno luogo a mutamenti continui di suoni e di significati, ossia a sempre nuove espressioni. Cercare la lingua modello è, dunque, cercare rimmobilità del moto. Ciascuno parla, e deve parlare, secondo gli echi che le cose destano nella sua psiche, ossia secondo le sue impressioni.73 To the question of repetition, which Pateman has also commented on elsewhere, I will return in Section 13. Here I simply wish to make clear that as a matter of fact I do think that art (however much we argue about the definition of that slippery term) and language involve the same creativity, and that the sense in which language is perpetual creation is no different from the sense in which art is perpetual creation. This is not, however, because I am an idealist in the philosophical sense or hold any brief for the neolinguistica of the Italian school, but because, as an integrationist, I hold that in artistic communication and linguistic communication the source of meaning is the same; namely, the contextualized integration of human activities. The human capacity which makes both art and language possible is the capacity for creating signs. Without signs, 73 B. Croce, Estetica, 10th ed., Bari: Laterza, 1958, p. 164.
282
ROY HARRIS
we have no communication; and without communication we have neither language nor art. Kant's aesthetic, it must be remembered, pre-dates the revolution we now look back upon as the advent of "modern" art. And modern art was mainly founded on deliberately breaking "old rules" (but not "rules" in the Chomskyan sense). I am somewhat sceptical, nevertheless, about Pateman's notion that our aesthetic judgments involve an intellectual effort to find "the rules which will allow us to demonstrate post hoc the beauty we intuited without a rule to guide us". This is not, of course, to deny that artists often indulge in post hoc theorizing to justify their creative acts. A piece of sculpture by Robert Gober, which fetched £35,000 at a recent auction at Sotheby's (I was not the purchaser), consists of a metal plughole mounted on a wall. The artist explains that this work was created from no less than three sinks, but included additionally the incorporation of a crossband over the hole. ("It wasn't from a pre-existing drain because I couldn't f ind one with a cross.") The plughole, as displayed on the wall, is not connected to any drainage system, and this is an essential feature of its aesthetic message. (It is "a window on to another world". "Drains lead us into a netherworld of fluid darkness, leaden with threatening associations and fears of being pulled into the depths of unknown territories.")74 I take this to be a good example of a Kantian reflective judgment (as construed by Pateman) even if the verbal terms in which it is expressed are somewhat heavy-handed. I also agree with Pateman that although we can all be creative, poetry isn't easy. Nor is scuplture, as the case of Gober's plughole clearly shows. If Gober were a Chomskyan, he would now market a kit with which anyone would be able to assemble a wall-mounted plughole out of prefabricated parts by following rules printed on the packet. He would thus demonstrate what is involved in artistic creativity beyond any shadow of doubt.
13. Context No linguistic act is contextless, and every linguistic act is uniquely contextualized (which is not to deny that anything one says or writes is potentially subject to multiple contextualization). Integrational linguistics might perhaps best be defined as any form of linguistic inquiry which gives these two tenets their full theoretical due. Mutatis mutandis, the same goes for integrational inquiry into art, music, cricket and other forms of creative activity. 74
Reported in The Times 30.6.95.
FROM AN INTEGRATIONAL POINT OF VIEW
283
Naomi Baron (this volume, Ch. 8) sees a historical link between the integrationist concept of context and the writings of Malinowski. In this she is right. She is wrong, however, in supposing that the integrationist concept is simply an endorsement of Malinowski's, and that by demonstrating the inadequacies of Malinowski's "context of situation" she has eo ipso exposed a weakness in the integrationist position. The reasons for this are (i) that Malinowski's position was still a long way from the integrationist's, and (ii) that there were crucial ambiguities in Malinowski's own use of the term context. The latter are lucidly exposed in George Wolf's paper "Malinowski's 'Context of Situation'", 75 which Baron does not mention. As Wolf points out, Malinowski changed his view between 1923 and 1935, but Baron does not appear to notice this. Nor does she seem to have read the discussion of contextualization in Ch. 10 of La Sémiologie de l'écriture. Her main objection is to the claim that meaning is "completely context-dependent", which she attributes to Malinowski and which she regards as "too strong". The principle argument she advances is that if meaning is completely context-dependent, then "no communication is possible between people who don't share one another's context of situation" (p. 157). Evidently she regards this latter denial as untenable. In other words, she seems to hold that a theory which makes meaning entirely context-dependent is incapable of explaining how communication works, because communication can take place between people even in the absence of a common situational context. Therefore (the implication is) words must after all have meanings that are context-free. This is a rather odd move to make against Malinowski, but an even odder one to make against the integrationist. For perhaps the first point an integrationist would raise is that "a context" is not an inert, independently given situation that people "share" if they happen to be there together, but do not "share" otherwise. On the contrary, "sharing one another's context" is a notion that, from an integrational perspective, makes little sense at all. We do not "share" anything simply by being physically co-present. Baron's quotation of my remark that "nothing that has happened in the field of language studies since Malinowski's day suggests that in the end he got the answer wrong" (p. 180) is itself a prize example of decontextualization. If the text had been quoted in full, it would have become apparent what "the answer" was. For I was not there approving Malinowski's theory of context at all, but agreeing with his answer ("No") to the question "Can we treat language as an independent subject of study?". In other words, the issue is the validity of a segregational linguistics.
75
Language & Communication, Vol. 9 No. 4, 1989, pp. 259-267.
284
ROY HARRIS
I certainly agree with Malinowski, furthermore, when he states that "the false conception of language as a means of transfusing ideas from the head of the speaker to that of the listener has.. .largely vitiated the philological approach to language". 7 6 This is an unambiguous and forthright denunciation of the telementational model. It seems to me as apposite in 1995 as it was in 1935. (NB I here recontextualize.) The same could be said of his warning concerning the activities of the lexicographer: "The figment of a dictionary is as dangerous theoretically as it is useful practically."77 And the same again of his insistence that, in certain contexts, gesture can be "an integral part of language".78On all these points, I find Malinowski's thinking refreshingly radical and far in advance of his time. But that radicalism was eventually compromised, as it was in the case of Firth. (It may be superfluous to add that the name Malinowski, as far as I can remember, meant nothing to me either as a schoolboy or as an undergraduate.) Naomi Baron's own position, as far as I can judge, falls into the category which I have elsewhere called "weak segregationism". This involves tinkering about with the telementational model in an effort to patch up its most glaring implausibilities. The irony is that Malinowski ended up as a weak segregationist himself. As Wolf points out, he could not in the end bring himself to abandon the notion that words have fixed meanings. Where I think some confusion may arise is over the notion of "sharing a context". There is a lay sense of "sharing" according to which A and Β may have in common an experience denied to C, simply because C was somewhere else at the time. And it is then tempting to suppose that this common experience ("We both saw the gunman shoot the cashier") validates the notion of a context shared by A and B, or even that the context is part of the common experience. How would an integrationist construct a theoretically viable notion of "sharing a context"? The answer to this is that it is not possible if the requirement is that the shared context be exactly the same for those who share it. In other words, we come back here to the mistake involved in supposing that a context is some independently given situation, backdrop or activity, in which, against which, or in the course of which, communication takes place. Being in the same room with X between t1 and t2 is not, from an integrational point of view, an example of sharing a context with X. Nor is walking down the street with X, or playing a game of tennis with X. In all these situations and activities, the context is always different for different participants. This results from a basic biomechanical 76
B. Malinowski, Coral Gardens and Their Magic, New York: American Book Co., 1935, Part IV, p. 9. 77 Ibid, p. 22. 78.Ibid, p. 30.
FROM AN INTEGRATIONAL POINT OF VIEW
285
constraint not only on interpersonal communication but upon human existence. You cannot be yourself and somebody else as well. Contextualization, for the integrationist, not only varies according to the past experience of the individuals in question (and is thus irreducibly different for different individuals) but is itself a dynamic process. In short, contexts do not precede communication but are constructed in the course of communication. But this in no way precludes agreement by the participants about certain features of the situation. In other words, the integrationist is not trying to deny that, for instance, when A and Β play a game of tennis (which is a form of communication) they are playing with the same ball and on the same court. There are not two balls and two courts, one for each player. On the contrary, it is playing with the same ball on the same court which - among other things makes the communicational roles of the two players strikingly different, but at the same time complementary and interdependent. For in order to play tennis (at least in the conventionally accepted way) it is actually essential that A and Β both recognize that they are playing with the same ball on the same court and act accordingly. However, that recognition, integrationally constructed and articulated, is not something external to the playing, but a contextualized feature of the playing itself. The same applies to human verbal behaviour. Interesting examples of the dynamic construction of context can be seen in the phenomena of rhythmicality (in the sense of Cowley 1994),79 where the continuous voluntary co-operation of two speakers is required in order to achieve quite specific communicational effects. As in the game of tennis, the individuals involved have to anticipate each other's moves in order to continue the "play" successfully. Thus it is by adapting one's speech rhythms or one's movements across the court to those of a partner that, from moment to moment, the communicational process is enacted. The context is dynamically constructed in the sense that, without the continuing contribution of the other "player", the actions of either individual would be meaningless and the process brought immediately to a halt. Thus the integrational construal of the lay notion of "sharing" or joint participation would be the concept of two individuals engaged in communication which involves co-monitoring a given object or course of events, which both agree or accept as given. Furthermore, that agreement or acceptance is not merely a pragmatic but a logical requirement of the enterprise. But even here there is no question of the context being identical for both parties: A and Β may still disagree about what the object consists of, or what is going on in the sequence of events they are watching or participating in. If it were otherwise, there could never be 79
S.J. Cowley, "Conversational Functions of Rhythmical Patterning: a Behavioural Perspective", Language & Communication, Vol. 14 No. 4, pp. 353-376.
286
ROY HARRIS
any situation in which A was convinced that the first service was "in", while Β was no less convinced that it was "out". To insist that because there is only one ball and one court, it must follow that both players share the same context is a simple non-sequitur. It rests on confusing the context with material features of the situation. In other words, as far as the integrationist is concerned, context is not reducible to any "sharing" that can be defined in biomechanical, macrosocial or circumstantial terms. Nor is such a sharing a prerequisite for communication. In order that A and Β be in communication with each other, there does not necessarily have to be any set of conditions that are equally binding on both, or any information to which both have equal access. On the contrary, any such theoretical restriction would limit the concept of "communication" in ways that are quite unacceptable from an integrational point of view. Consider, for instance, the following case. A sends a telephone message to B. It is recorded on B's answerphone. B's secretary types out a transcript and sends it on to B, who is travelling abroad. A and Β have never met. This scenario can be elaborated indefinitely, bringing in other intermediaries. Perhaps A dies before Β receives the message. Perhaps By s secretary translates the message into Urdu. And so on. What this scenario describes is a communicational chain in operation. Faced with such a scenario, a theorist has two basic options: (i) admit that - barring breakdowns - there is indeed communication established between the two individuals at the extreme ends of the chain, however long or devious the chain may be in space or time, or (ii) restrict the concept "communication" to the adjacent sections of the chain which apparently involve direct biomechanical contact of some kind between the intermediaries. The notion of "shared context" comes in here as that "within which" or "on the basis of which" the direct contact takes place. If option (i) is chosen, however, there seems to be no essential explanatory role for a notion of "shared context" to play, because it is superfluous (except under one possibility, to be considered below). If option (ii) is chosen, the theorist has to abandon the claim that A communicates with B. The integrationist here will go for option (i). The only question is whether that option ought to be explicated - and here we come to the one possibility mentioned in the preceding paragraph - as necessarily involving biomechanical contact between every pair of adjacent intermediaries in the chain. But this amounts, obviously, to a backdoor admission that "real" communication requires biomechanical contact, and that the indirect contact between A and Β at extreme ends of the chain is called "communication" only by courtesy. In short, it is a roundabout capitulation to option (ii). A communicational chain is thus analysed
FROM AN INTEGRATIONAL POINT OF VIEW
287
as a sequence in which every link in the chain demands a "shared context" (i.e. that in which the contact occurs), even though there may be no such shared context for the original initiator of the message (A) and the ultimate receiver (B). But why does an integrationist resist this move? For two reasons. One is that the notion of biomechanical contact is itself too weak to bear the theoretical load. In what is generally called "face-to-face" communication this may not pose a problem. But as soon as we move from that even to a telephone call, difficulties begin to arise. The telephone eliminates all communication via sensory modalities other than hearing. But does Β actually hear A's voice? What we know of the technology involved suggests the answer "no". And such an answer immediately puts the notion of biomechanical contact under strain. The second reason is more important, however. To concede that communication must somehow be construed in terms of shared contexts in which biomechanical contact occurs is to reinstate the idea that in the end the explanation of communication rests on identity statements. That is to say, somewhere we have to find transitions between an initial state and a terminal state which do not require further analysis because we can say that something linking the two some bridge that has to be in place and is crossed - simply is one and the same for all parties concerned on both sides of the divide, wherever it occurs. And there explanation comes to a stop. It seems to me that if a theorist wants to make this move at all, it might as well be made sooner rather than later; for instance, at the point where Locke makes it. One then says that the ideas in the minds of A and Β are "the same" (or should be, if communication is to be guaranteed). I take it that it seems attractive to those who are reluctant to embrace such an overt form of telementational theory as Locke's to shift the sameness-requirement from the ideas to some other part of the speech circuit, or even to features of the communication situation. The attractiveness, I imagine, resides in the difficulty of demonstrating the sameness when it resides in internal cerebral goings-on, as contrasted with the plausibility of appealing to sameness when it resides in public, observable actions or states of affairs. ("We all heard what he said, didn't we?") But from an integrational point of view, there is no explanatory mileage to be gained by shifting the sameness-requirement in this way. Integrationism, in other words, resists the tendency that is virtually endemic in the Western tradition to treat identity (sameness) as a bottom-line explanation where communication is concerned (as we see for instance, in the way it underwrites the whole theory of the syllogism). For the integrationist, on the contrary, whenever one encounters the notion that something is "the same as" something else, that is never the end of the road, but a notion that itself requires explication by reference to something else. Consequently, to postulate as part of
288
ROY HARRIS
the explanation of communication a shared contextual factor, or perception, or impression, etc. which A and Β simply have in common can be no more than a deferral of the problem. For this notion that A and Β have something in common is a form of identity statement. Any explanation of communication that resorts to elements allegedly "shared" or "common" as between individuals or contexts tacitly presupposes the validity of some external viewpoint which guarantees the underlying sameness in question. But that is no advance on Aristotle's postulate that the external world is the same for all observers. A theorist who wants to take this way out might just as well give us a reference to what Aristotle said and declare the debate over (or, rather, a non-contest). I move on now to some consequential questions concerning context. The problem that the two integrational tenets cited at the beginning of this section might appear to pose for linguists (those who engage in language about language) is that they preclude systematic descriptions. For, seemingly, there is no way to capture uniqueness in words. Contextualization is Heraclitean. But, an objector will protest, do not words logically commit us to non-uniqueness? (This, incidentally, is the line of attack I think Borsley and Newmeyer should have been pursuing against me, instead of dithering about with unconvincing attempts to cop out of fixed codes.) From an integrational point of view we are all (or most of us) linguists, and at the most basic level the question of contextualization is inextricably bound up with that of iterability. So -I give this example as a no-ball bowled to my critics, who can therefore take a free swing at it - what would we say if anthropologists reported the discovery of a lost tribe in the Amazon jungle speaking a language in which there was no way of asking someone to repeat what had just been said? Do we treat such a report with the same scepticism as we would reports of a community whose members believed that physical effects preceded their physical causes? Does it make sense for us to imagine human beings who cannot ask each other to say "the same" over again? Once we see the pertinence of this question, we see how context and iterability are in turn linked to the arbitrariness of the sign. The connexion between repetition and the arbitrariness of the linguistic sign has been discussed with considerable acumen in a paper by Christopher Hutton. For orthodox linguistic theory, as Hutton points out, arbitrariness is a somewhat paradoxical doctrine, since it affords no explanation of the non-arbitrary connexion which must be deemed to hold between different instances of "the same" linguistic unit. Yet this connexion is equally essential for orthodox linguistic theory.
FROM AN INTEGRATIONAL POINT OF VIEW
289
Is not the instance, x, a sign of the category, X? This question is not posed by Saussure, but it is the key question, for it points us directly to the heart of the ambiguities in our understanding of what signs are.80 From an integrational point of view, on the other hand, it makes perfectly good sense to deny decontextualized iterability; a denial that orthodox linguistics cannot afford to make. Precisely this feature of the integrational position, however, has been criticized by Pateman, who claims that the integrationist argument is "self-defeating". If Pateman is right, any move towards an integrational linguistics is doomed from the start. Commenting on examples offered in The Language Myth, Pateman contends: To make the point that there is no repetition of well or How 's that? we actually have to use, overtly or covertly, and not merely mention, a type, category or form well or How 's that? We may not be able to give a plausible derivation of all the utterances which have counted or would count as utterances of How 's that? and we may be in the dark about how speakers and hearers understand such utterances; but that they make use of some iterable form not defined by temporal co-ordinates seems presupposed by the very critique of that idea.81 Pateman here, I think, makes two mistakes. In the first place, he misconstrues the integrationist position. The point about the example "Well! Well!", 82 cited as counterevidence to the claim that nothing can be repeated, is that it is not very convincing counterevidence, since the two syllables will be intonationally different and the substitution possibilities for each are different. Pateman quotes the integrationist claim as being that repetition "is only partial replication, and even that partial replication is context-bound by succession in time". But he omits the immediately preceding sentence which clarifies that point: The claim is not that speakers cannot produce or recognise instantiations of the same expressions on different occasions, but rather that this ability does not yield a criterion of demarcation between the linguistic and the non-linguistic, nor imply that whatever we say is decontextualisable.83 80 C. Hutton, "The Arbitrary Nature of the Sign", Semiotica, Vol. 75 No. 1/2, 1989, p. 68. 81 T. Pateman, "Philosophy of Linguistics", in J. Lyons, R. Coates, M. Deuchar and G. Gazdar (eds.), New Horizons in Linguistics 2, London: Penguin, 1987, p. 251. 82 The Language Myth, p. 155. 83 The Language Myth, loc. cit.
290
ROY HARRIS
There is a story about the court painter of a Moghul prince who boasted that he could copy any picture by a European artist so faithfully that it would be impossible to tell original and copy apart; but was disconcerted to find that to a European eye the difference was easily discerned.84 What this shows is not the impossibility of exact mimetic replication in painting, but rather that what counts is what to ignore. The Indian court painter got it wrong because his aesthetic involved a (non-verbal) concept of copying which failed to match the European concept. One might say that the repetition in "Well! Well!" is somewhat analogous, in the sense that our ear can easily tell the two apart if we are interested in the differences, but recognize them as the same if we are interested in similarities. The point of the "How's that?" example85 is rather different. Appeals on the cricket field are notoriously variable from a strictly phonetic point of view, but all count as instantiations of How's that?, not because umpires have specially trained ears but because umpires just do not bother to listen for phonetic similarities. Again, in one sense all appeals are the same (examples of How's that?), whereas in another sense they are all different. The counterpart of the Moghul painter's error would be to claim that two appeals on the cricket field may be indistinguishable; whereas the fact is simply that as appeals they do not differ. Nevertheless, they can be made by different players, at different points in the game, etc. The case of How's that? offers a further illustration of the radical difference between an orthodox linguistic approach and an integrational approach to human communicational behaviour.86 The vocalization [zaaaa] cannot, by any stretching of the orthodox "rules" for English pronunciation, be counted as a predictable utterance of the words How's that? Now this cannot simply be "written off' as an extreme or marginal example, even though it may be the case that only a small percentage of people who speak English are acquainted at first hand with the kind of English one might expect to encounter on a cricket field (for the very good reason that many people who speak English have never played cricket or have any interest in acquainting themselves with the verbal behaviour of those who do.) The reason why it cannot be written off is illustrated by the dictionary on my desk, which has no entry between mizzle and mnemonic. That is to say, the lexicographer declined to recognize as an English word one of the most 84
G. Lowry, "Foreign Emissaries to the Mughal Court", The India Magazine, Vol. 2 No. 1, 1981, pp. 34-9. 85 The Language Myth, pp. 176-7. 86 I am indebted to Nigel Love for pointing out to me in some detail that the interpretation of the Laws of Cricket is by no means as straightforward a business as I had at one time supposed. I ought not to have been surprised by this.
FROM AN INTEGRATIONAL POINT OF VIEW
291
commonly occurring vocalizations of assent in the English-speaking community: namely, mm. It is difficult to believe that the lexicographer had never heard anyone say mm, and the form occurs in print. One possible explanation of the omission is that [mm] is treated as a kind of minimal articulation of yes (just as [zaaaa] might be of How's that?). Arguments could be advanced for and against such an explanation. To pursue them here is not the purpose of introducing the example. The purpose is to point out that if linguists are seriously interested, as some of them claim to be, in documenting speech communication, then there is a serious question to be addressed concerning their recognition of certain vocalizations and their resolute indifference to others. (Particularly when some of the vocalizations resolutely ignored are among the commonest in the vocal behaviour supposedly under investigation.) From an integrational point of view, [zaaaa] is no less a vocal sign than [mm], and the reluctance (to say the least) of orthodox linguistics to deal with either is an interesting question for investigation. Pateman's other mistake is that in cutting the ground from under the integrationist he himself falls into the pit of linguistic Platonism. The integrationist will simply deny that there is any such thing as a "form not defined by temporal co-ordinates". So to claim that integrationists cannot discuss examples of repetition without using such forms is merely to beg the question against them. As one of Pateman's critics points out,87 Pateman simply takes for granted that in the speaker's mind language already assumes the form of a set of context-free metalinguistic abstractions. Consequently, for Pateman the denial of iterability becomes automatically incoherent. For the integrationist, however, to say that two examples of well are not the same is no self-contradiction. On the contrary, the integrationist can explain - whereas the Platonist cannot - exactly how both the "sameness" and the "difference" are constituted in terms of the contextualization of linguistic experience. Saussure at least was not a Platonist in this sense. He tried to tackle the problem of variation in repetition by invoking the concept of linguistic structure. It is linguistic structure, according to Saussure, that determines the "space" available for occupation by any given linguistic element. His best example is handwriting.88 He accounts for the latitude allowed to an individual in writing a letter t on strictly structuralist grounds, by claiming that the one essential thing for the writer is "that his t should be distinct from his /, his d, etc." This is no doubt true; but what an integrationist would say is that Saussure fails entirely to acknowledge that context is the most important factor in determining whether a t 87
N. Love, "Language and the Science of the Impossible", Language & Communication, Vol. 9 No. 2, 1989, p. 286. 88 Saussure, op. cit., p. 165.
292
ROY HARRIS
is in fact distinct from an /, a d, etc. The "same stroke of the pen" might do duty as a / in one word, but as an / or a d in another. (It should not be difficult to see how this principle likewise applies to the question discussed above concerning supposedly aberrational vocalizations like [zaaaa] and [mm].) The point being made here goes beyond questions of recognition in the sense relevant to the experimental psychology of reading: some individuals are better at reading "bad handwriting" than others. What is at issue here is the semiological identity of the written character. In other words, the integrationist claim is not merely that "repetition is only partial replication" but also that total replication is not a sufficient condition of repetition. And what counts as "partial" and "total" is in any case itself context-bound. It is here, as Michael Toolan89 has noted, that integrationists part company with post-Saussureans such as Derrida and that the integrationist has something new to say to literary critics about such controversial issues as the meaning of literary texts and the role of the author. It is likewise here, as Toolan argues in the same paper, that integrationists part company with post-Austinian speech-act theory. The crux is the notion of "intention". Both Derrida and the speech-act theorists require the notion of an iterable sign of which the iterability is independent of intentions. Derrida's entire thesis of intention-obscuring iterability rests on the plausibility of his notion of the sign and the claim that the sign is, by definition, repeatable... Derrida begins by saying that for a sign to be a sign it must be readable beyond its original context, but rapidly comes to the conclusion that this entails iterability - the second occurrence of a sign must be identified as the same sign as occurred before... But from an integrational point of view, this is to fall into exactly the same error as orthodox linguists. In saying that iterability itself is "the positive condition of the emergence of the mark", thus a feature of the condition of being a mark from the outset and by the same token an intrinsic undermining of the mark's intentionality, Derrida is making the same mistake of privileging the epiphenomenal (marks taken as "the same" marks) over the phenomenal (new marks in new situations as clues to new intentions) that is incorporated into a theory of language by generativist grammarians. It is not iterability that is the positive condition of the emergence of the mark; the positive condition of the emergence of a (signifying) mark is 89
M. Toolan, "Largely For Against Theory", Journal ofLiterary Semantics, Vol. XIX No. 3, 1990, pp. 150-166.
FROM AN INTEGRATIONAL POINT OF VIEW
293
that an actor takes notice of it, as a purposeful part and exponent of a purposeful activity... The argument between Derrick and Searle about iterability 90 is of interest even if we set aside the fact that, as Searle points out, Derrida has failed to understand Austin's theory of speech acts. Derrida insists that a written communication "must be repeatable - iterable - in the absolute absence of the receiver or of any empirically determinable collectivity of receivers". This conjures up visions of automated computer print-outs spilling out on to the floors of deserted offices, whence all the employees have long since fled to avoid some impending nuclear disaster. Writing, for Derrida, carries on independently of communication, and its capacity to do so is what makes it writing. One could hardly envisage a more emphatic or a more grotesquely mystical commitment to the theoretical position attacked in The Language Machine. By comparison, Searle appears at first to speak with the voice of common sense, observing that Derrida has confused iterability with permanence. For Searle, there is "no getting away from intentionality", whatever form of language we are dealing with, and whatever the level of abstraction. A meaningful sentence "is just a standing possibility of the corresponding (intentional) speech act". So far so good. But Searle, like Pateman, is mesmerised by the type-token distinction. He writes: the logician's type-token distinction must apply generally to all the rule-governed elements of language in order that the rules can be applied to new occurrences of the phenomena specified by the rules.91 Without "this feature of iterability", according to Searle, it would be impossible to generate an infinite number of sentences, and this infinitude is "one of the crucial features of any language". But now it is Searle who is guilty of confusing the issue of iterability. Specifically, he is confusing discourse about parole with discourse about langue. The sense in which the English sentence Where there's a will there's a way contains two indefinite articles is not the sense in which there are two examples of α in every copy of an edition of a text in which this sentence is printed. Can Searle really be convicted of such an elementary error? Unfortunately, he goes out of his way to stress the point, claiming that 90
Glyph, Vol. 1, 1977. J. Derrida, "Signature Event Context", pp. 172-97. J.R. Searle, "Reiterating the Differences: A Reply to Derrida", pp. 198-208. 91 Searle, op. cit., p. 199.
294
ROY HARRIS
without iterability there is no language at all. Every utterance in a natural language... is an instance of iterability, which is simply another way of saying that the type-token distinction applies to the elements of language. To which the obvious rejoinder is: "No. It is not 'simply another way' of saying anything of the kind." Searle is manifestly in a muddle of his own making. Iterability of the kind involved in repeating something previously said or written has nothing at all to do with the quite irrelevant questions of (i) whether what is repeated contains examples of units which might occur in other combinations, and (ii) whether all of these belong to some set of verbal combinations that can be theoretically treated as infinite. In the sentence-type Where there's a will there's a way the second indefinite article is not a repetition of the first: nor in any single spoken or written token of this sentence-type does such a repetition occur. While Derrida confuses iterability with permanence, Searle in turn confuses iterability with the conformity of tokens to a type. The first is like confusing the fact that I have to pay the rent every week with my having a longstanding debt, while the second is like confusing the fact that the same rent is payable every week with the fact that the notes and coins I pay it with this week look more or less identical to the notes and coins I paid it with last week. (On the type-token relation, see Section 14 below.) Derrida's rejection of intentionality is no more than a philosophically elaborated corollary of the error already implicit in Saussure's theory of the autonomous linguistic sign; while on the other hand, as Toolan says, the speech-act theorist, endeavouring to reconcile the contradictory theses that "language is both fundamentally intentional and fundamentally conventional", ends up with the paradoxical notion of "conventionalized intentions". Integrationalism, on the other hand, gets itself into neither of these difficulties because it entails a different notion of signification. In the integrationalist alternative, however, no such assumption of assured, fixed-code reproduction of signification is made. Indeed, the non-iterative nature of language behaviour is made one of its hallmarks and essential characteristics. For in integrationalism, the reverse is assumed: namely that we can never be sure that signification remains constant, from one use of an expression to the next use of the "same" expression.92 This brings us back to our hypothetical Amazonian tribe. If they have no linguistic terms for verbal repetition, and no linguistic practices involving 92 Toolan,loc.cit.
FROM AN INTEGRATIONAL POINT OF VIEW
295
repetition, the question might be asked whether their communication system is a language at all. Perhaps what has been discovered, at long last, is the first example of a community which has not yet developed a language, but is still using some more primitive form of vocal communication. Here we return via another route to the question of defining language. It has sometimes been argued that a minimum condition for counting a communication system as a language is that it should have a syntax. By this criterion, a communication system of the type used by Wittgenstein's builder and his assistant93 is said to fail to be a language. For all they have is an inventory of just four verbal forms corresponding to the four types of building materials which the builder is using. Clearly, the builder and his assistant have no verbal mechanism for dealing with repetition either (and this is not for want of a syntax). There is no word which means "What did you say?" or "Please repeat the last instruction". So if the builder says "Block! Block!" does he repeat himself? Does he want two blocks, or just one? Is he getting impatient? Does he suppose that his assistant may be hard of hearing? These are certainly pertinent questions if we are interested in analysing how the system is being used for purposes of communication. It would obviously be a mistake, however, to suppose that these doubts can somehow be resolved by inspecting the structure of the system itself, which in this case comprises just the four units. Wittgenstein, just as much as Derrida, needs intention-free iterability (which is the same thing as context-free iterability). The problem is where to find it. The integrationist's point is that it cannot be found by listening attentively to the allegedly iterated forms. Allegories are not always apposite in discussing theoretical issues; but this one is. For once we see that examining "the words themselves" supplies us with no clear answer to our query about repetition, we should also see why, from an integrational point of view, both the questions "Is it a language?" and "Is it language?" are misconceived from the start. Language no more repeats itself than history does. But no less either. For language is part of history. (Which is another way of formulating the integrational principle of "cotemporality".)
14.
Cotemporality
It is symptomatic that modern linguistics provides no term for this crucial parity of status between the linguistic and the non-linguistic; an omission easily 93
Wittgenstein, op. cit., § 2.
296
ROY HARRIS
explained by the fact that modern linguistics acknowledges no such concept... The lacuna could be filled if the parity just referred to were designated by the term "cotemporality". Linguistic acts could accordingly be said to be cotemporal in our experience with non-linguistic events and circumstances of all kinds. Cotemporality is the ultimate basis of the distinctions we feel obliged to draw in order to deal metalinguistically with a whole range of so-called "type-token ambiguities", and is also intrinsic to our understanding of such everyday notions as asking questions, stating facts, and giving instructions.94 Fleming (this volume, Ch. 9) points out that "everything truly iconoclastic about both integrational linguistics and ethnomethodological conversation analysis comes down to how they account for the role of time in human action" (p. 200). He contrasts both with "the predominantly detemporalizedstructuralism of conventional sociology and linguistics" (ibid.). I agree with this assessment. Where the two differ - and here again I agree with Fleming - is that the conversation analyst's programme in effect reintroduces the detemporalized type-tcken distinction under another guise. As Hutton observes in Abstraction and Instance, it is interesting that there has been no focussed debate within twentieth-century linguistics as to the nature, applicability and validity of the terms type and tokens For as soon as precise definitions are sought it becomes obvious that they are hard to come by. Peirce's original distinction was - significantly - a typographical distinction; that is, a distinction drawn from the static and quasi-permanent form of language we call "writing". That form of language readily admits of comparisons based on backward-and-forward scanning which is relatively free of constraints imposed by memory and the passage of time. It is an essentially α-temporal distinction, or at least a would-be a-temporal distinction. To generalize such a distinction to the time-bound domain of speech or, even more ambitiously, to the world of signs in toto is an enterprise fraught with conceptual difficulties. By way of illustration, one such difficulty arises with sentences like I'm hungry. According to Bar-Hillel such sentences can never have two tokens with the same reference, since "even if they are produced by the same person, say A, their production takes place at different times".96 One apparent implication of this is that I cannot repeat I'm hungry and mean exactly the same thing by it (for the same kind of reason that I cannot repeat It is now 94
The Language Myth, p. 157. CM. Hutton, Abstraction and Instance. The Type-Token Relation in Linguistic Theory, Oxford: Pergamon, 1990, p. 60. 96 Y. Bar-Hillel, Aspects of Language, Jerusalem: Magnes, 1970, p. 70. Italics added.
95
FROM AN INTEGRATIONAL POINT OF VIEW
297
precisely five seconds past 3 p.m. GMT on 11 November 1990 without one or other announcement being inaccurate). 97 At best, presumably, I could do no more than indicate by the second I'm hungry that I was still hungry. This at first sight seems a coherent interpretation, even if not a particularly convincing one (at least, from the perspective of the lay person). However, as Hutton points out, Bar-Hillel's Heraclitean doctrine (i.e. that the reference shifts continuously with time) runs into trouble over the very definition of what a token is, since there are various candidates for the unique physical entity which the doctrine requires. If five people hear one utterance at the same time, do we have five "perceived" tokens or one concrete token? Bar-Hillel would want to claim that we have just one. Yet if we consider the case of someone who listens to the same recording of the same novel read twice, Bar-Hillel has a dilemma. According to the letter of the definition, we would want to say that we have two sets of tokens, each playing of the recording producing a new set of physical entities. Yet according to its spirit, we should say that the token is the information registered in the tape: this number remains constant. The point is that even the physicalism of Bar-Hillel's definition does not give us firm criteria for deciding when we are dealing with the same token, and when we have a new or additional token.98 Now it might appear that the problem Hutton is posing is trivial in the sense that it merely calls for a terminological decision to be made. Does it matter whether we opt for defining token strictly by reference to the original production of a given mark or sound, or, alternatively, by reference to its reception by readers or hearers? Will not either solution do, as long as we do not confuse the two? The integrational answer is that it does matter, and indeed makes a very great difference. But the integrationist will also point out that neither option can be simply accepted as offered above. For the terms of that offer already beg the crucial question. In presenting it as a choice between focussing on "the original production" of the mark or sound or focussing instead on "its reception by readers or hearers", the use of the pronoun its already slips in the assumption that the it in question is one and the same. Here we see the type-token distinction
97
There are problems about even one such announcement, given that it will take a finite number of centiseconds to utter these words. But let us, for purposes of the example, suppose the announcement counts as accurate if and only if the time designated falls during the utterance announcing the time. Radio and telephone time-checks get round this problem by formulas such as "At the third pip, the time will be..." 98 Hutton, op. cit.,-p. 59.
298
ROY HARRIS
pulling itself up by its own linguistic bootstraps. The crucial theoretical question is precisely the validity of that assumption of identity. Bar-Hillel's dilemma - and the dilemma for any "physicalist" attempt to define token - arises from the fact that all physical events are time-bound: yet what the type-token distinction requires - or appears to require - is a set of criteria justifying identifications of sameness independently of time. This results in desperate attempts to square the circle, by keeping the temporal dimension out of one's definition of what a token is. At least Bar-Hillel attempts to tackle the problem. More often, linguistic theorists choose to ignore it. As Hutton puts it: The type-token relation is granted a sacrosanct status that reflects a fundamental division between language and non-language." But at this point a critic of the integrational position might perhaps renew the objection that even integrational linguists will need to acknowledge such a division. Is it not already acknowledged, not merely implicitly but explicitly, in the statement of the principle of cotemporality itself? Are we not told, as in the passage from The Language Myth quoted above, that it is a question of "parity of status between the linguistic and the non-linguistic"? And does not this in turn cast serious doubt upon the integrational claim that the distinction between language and non-language is itself problematic? No. These would be difficulties only if integrational linguistics were conceived of as a simple replacement for orthodox linguistics; that is to say, as a new "science of language" which could do everything orthodox linguistics does or did, only on a sounder theoretical basis. But one of the basic messages of integrational linguistics is that the distinction between the linguistic and the non-linguistic can be drawn in a variety of different ways, from different perspectives, in different contexts and for different purposes.^ 0 There is no unique way of drawing it, nor any one way of drawing it that automatically takes preference and hence can be treated as foundational for a "science of language" or for any privileged form of discourse about language. On the contrary, the integrationist will insist that an important priority for linguists is to re-examine and re-evaluate the discourse of linguistics itself Here the principle of cotemporality has far-reaching implications, in particular for understanding what is nowadays described as "the history of linguistics". It is no 99
Hutton, op. cit., p. 60. Edwin Ardener makes essentially this point in his highly instructive comparison of the Ibo term aka and the English hand. "Social Anthropology, Language and Reality", Approaches to Language, ed. R. Harris, Oxford: Pergamon, 1983, pp. 143-156. 100
FROM AN INTEGRATIONAL POINT OF VIEW
299
coincidence that the Magna Carta of modern linguistics, Saussure's Cours, begins with what purports to be a history of the subject. 101 Writing history - or rewriting it - is a favoured form of self-justification for academics and politicians. As Sylvain Auroux shrewdly observes, "pour définir son propre statut historique, on n'est jamais si bien servi que par soi-même!".102 And where linguistics is concerned it becomes particularly difficult to distinguish histories of the subject from histories of the politics into which they are inextricably meshed. Julie Andresen, commenting on what has hitherto been included in and excluded from the received history of American linguistics, points out that "all linguistic theories are political". 103 A fortiori, all histories of linguistics are political, inasmuch as they necessarily require decisions about what to count as linguistic theories in the past and how they are presented as relating to linguistic theories in the present. To put it another way, the application of the principle of cotemporality to the discourse of linguistics itself requires us to recognize that there is no timeless way of talking about language or languages. Toolan's integrationist dictum that "meaning is always 'now"'104 applies first and foremost to and within linguistics. "Now", it might be objected, is itself a temporal concept. But that objection immediately highlights the theoretical objective which is close to the heart of the integrationist endeavour. "What is time?" said jesting Harris, and did not stay for an answer. But since he is more amenable than Pilate, let us oblige him to tarry just a little for the benefit of both his critics and their readers. Let us put the question: where, and when, is this encounter between Harris and his critics taking place? It would be question-begging to reply "here and now". That would be merely a deferral of the explanation. Is the encounter taking place in the pages of this book, where the words of the participants are on display? Is it taking place in the year of publication? Will it always be taking place from now on, for as long as copies of the book survive and there are readers to read what we wrote? My guess is that some critics will not like any of these options. They might even argue that the encounter is not an event at all, although they would be hard put to it to dispute that others may write narratives which treat it as such. Intellectual history is replete with similar stories. ("In the December 1997 issue of the Journal for Fatuous Expostulation, Professor R. de Cule definitively 101
R. Harris, Reading Saussure, London: Duckworth, 1987, pp. 4-13. S. Auroux, La Révolution technologique de la grammatisation, Liège: Mardaga, 1994, p. 7. 103 J.T. Andresen, Linguistics in America 1769-1924. A Critical History, London: Routledge, 1990, p. 6. 104 Toolan, loc. cit. 102
300
ROY HARRIS
refuted... etc.") And these narratives specify times and places at which intellectual events, allegedly, occurred. I think an integrationist can give a reasonably sensible answer to the question. It runs roughly as follows. We make sense of temporal and spatial assignments in accordance with the demands of the particular communication situation to which they relate. Philosophers are wont to blow this up into an enormous, fundamental metaproblem about "the nature" of space and time. When, for example, Kant asserted that "time is not an empirical conception" {Critique of Pure Reason, I, § 5) he was talking the kind of impressive gobbledegook that has bemused philosophers for centuries. There is no reason whatever to take it seriously. The only concept of time (or space) integrational theory needs is the concept that any reader of this page already has. That is to say, a concept dictated by the activities which "reading" integrates. This integration imposes, for instance, serial optical scanning of some kind: "what it says" on the page cannot all be taken in at a glance. Time is what this serial processing "takes", and space is where the material to be processed is presented. You know that you are familiar - or unfamiliar - with what is involved by your success - or failure - in reading the text. But this processing is itself temporally integrated into other activities. You can put the book down for the moment and carry on at a later time. Or you may never get round to that resumption at all. It may fall into the category that Maurice Bowra once described as "books you cannot bring yourself to pick up again". With his usual acumen, he had put his finger on a crucial (integrational) feature of the relationship between texts and readers.
15. Towards an integrational
linguistics
The Language-Makers, The Language Myth and The Language Machine are, all three, books mainly about the linguistics of the present and the past. The phrase integrational linguistics looks towards the future. (Borsley and Newmeyer speak scornfully of a "promissory note" (p. 64).) It would be a pity if that forward look should be misdirected or the promissory note misinterpreted for want of a more perspicuous phrase. So it may be as well to comment here briefly on some possible misperceptions of the integrational approach. One is represented by Gardner's comments on integrationism and generative grammar, which seem to imply that what the integrationist is calling for is integration between linguistics and neighbouring disciplines.105While it is true 105 H. Gardner, The Mind's New Science, New York: Basic Books, 1985, p. 221.
FROM AN INTEGRATIONAL POINT OF VIEW
301
that more co-operation between linguists and colleagues in neighbouring fields might be welcome, it would only be welcome from an integrational point of view if it led to linguists questioning the orthodox theoretical assumptions of their own discipline. If, on the contrary, it merely led to a reinforcement of those assumptions and hence to a narrowing of horizons in future linguistic inquiry, it would be far from welcome. Integrational linguistics, I repeat, does not purport to be some kind of ecumenical movement in language studies. Nor does it even envisage establishing what Auroux and Kouloughli have proposed as "a reasonable research programme in philosophy of linguistics". 106 For this programme seems to involve a division of labour between linguists and philosophers, whereby philosophers have the task of teaching linguists which linguistic issues are empirical and which theoretical, what constitutes "proof, what constitutes linguistic "explanation", and so on. The trouble with such a programme from an integrational point of view is the dubious notion that such advice could be impartially offered by philosophers with no linguistic axe to grind at all. In fact it would seem that many contemporary philosophers have exactly the same professional blind spots about language as many linguists have; and the blind leading the blind is not exactly a hopeful scenario for the future of any discipline. Another possible misunderstanding is represented by Matthews's view that there is a convergence between integrationism and recent generativist work, in that both point towards developing a linguistics of the individual, rather than a linguistics of languages. 107 It is true that the adoption of an integrational perspective might be expected to refocus attention on the individual as the ultimate locus both of linguistic judgments and of linguistic responsibility. But that is a far cry from the current direction of generativist linguistics, which simply idealizes the individual as the locus of postulated systems of linguistic rules. Third, nor should integrational linguistics be envisaged as a potential alternative "school of linguistics" (if that involves proposing alternative systematizations of phonology, grammar, lexicology and semantics). On the contrary, an integrational approach would question whether there is any serious theoretical justification for attempting such systematizations at all. (This is not to deny that there are perfectly reputable purposes for which linguistic systematizations of some kind are required, including educational purposes. But these needs can be, and have always been in the past, met by systematizations purpose-built to deal with the particular needs in question. The theoretical input to 106 S. Auroux and D. Kouloughli, "Why Is There No 'True' Philosophy of Linguistics?", Language & Communication, Vol. 11 No. 3, 1991, pp. 151-163. 107 P. Matthews, "Whither Linguistic Theory?", Linguistics Abstracts, Vol. 1 No. 1, 1985, p. 7.
302
ROY HARRIS
such systematizations has usually been minimal and highly questionable: but that did not prevent them from serving their purposes. Moreover, that such systematizations exist and may be influential in forming people's linguistic perceptions makes them a relevant topic of inquiry for the linguist.) Fourth, there is the widespread misconception that all an integrational linguistics is really calling for is a kind of improved sociolinguistics. Thus Borsley and Newmeyer find it hard to see anything original in the integrational programme at all. They imply that, in their opinion, it is just another version of the "ethnographic approach" to communication advocated by Hymes. This assimilation is revealing. For a generativist, indeed, the two must be well nigh indistinguishable. But what that shows is just how blinkered the generativist outlook is. Fifth, nor does the integrational approach imply some kind of glorification of what is sometimes called "folk linguistics". Integrational linguistics does indeed assume that linguistics is a lay-oriented inquiry; but that does not mean that every lay linguistic concept automatically acquires a privileged status. Again, on the contrary, there is no assumption that certain linguistic concepts have to be taken for granted in order to have a basis for the identification, classification and description of all linguistic phenomena. From an integrational point of view, there are no "privileged" linguistic concepts. The Dogon view that words are body fluids is as much open to question as the European view that words are meaningful noises. Linguistic concepts, from whatever source, are themselves seen as standing in need of analysis, as part of the human communicational process to be studied. Sixth, some people I have met seem to make the equation "integrationism = theoretical subversion". They regard me as a kind of "anti-everything" man, an intellectual anarchist. So the only thing that interests them is not the integrationist position as such, but whether integrationist criticisms οf their favourite enterprise can be rebutted. Again, Borsley and Newmeyer are fairly typical of this reaction. They are generous enough to concede that my "advocacy of an investigation of the semiology of written texts" is "reasonable" (p. 63). But their generosity is a function of their narrow-mindedness. The fact is that generativists are not in the least interested either in semiology or in written texts. Consequently, they are at a loss to understand anyone who sees generativism itself as a misguided attempt to mathematicize linguistics, an attempt which in the end tells us nothing new about language at all. On the other hand, it is evident why Borsley and Newmeyer would resist such a view, for it implies that generativists are engaged in a quite pointless exercise (albeit one displaying considerable ingenuity in the realm of formalization). The irony is that generativists have not stopped to consider exactly what formalization {any formalization) entails in linguistic terms. Thus
FROM AN INTEGRATIONAL POINT OF VIEW
303
they do not realize that formalization already involves question-begging assumptions about the human faculty generativists claim to be investigating. In short, they cannot see their own professional spectacles. Maybe they do not actually realize they are wearing any. This last point (the construal of integrationism as theoretical anarchism) relates to the claim that integrationism falls a long way short of providing for contemporary linguistics what Love calls "a Kuhnian 'normal science' to occupy the academic troops". 108 John Joseph makes a similar point when he says that I have failed to "articulate the kind of practical program on which the linguistics industry could continue to survive, let alone expand" (p. 9). His comparison with the tobacco industry is apt. I am not at all sure I am interested in the survival of "the linguistics industry" as it operates at present: it generates too much intellectual pollution and too many cultural health hazards. Theorists, in any case, should be judged by their theorizing, not by the applications which have been - or might be - made of their theories (whether by themselves or by others). It would seem to me absurd, for example, to criticize Saussure for never having published a comprehensive synchronic analysis of French (or any other language). The fact that Plato founded no state on Platonic principles does not in the least detract from the value of Plato's political theorizing. Furthermore, the success or failure of others in applying Saussurean or Platonic ideas is equally irrelevant to their theoretical assessment. Even a theory with no conceivable applications may nevertheless be a sound theory. On the other hand, applications merely demonstrate applicability; and it is notorious that the applicability of a theory is no guarantee of its soundness. The widespread human preference for easily applicable theories, regardless of whether they are sound or unsound, has always been one of the most formidable obstacles to intellectual progress. Linguistics enjoys no special dispensation which makes it an exception to considerations of this order. Nor do I see why anyone should accept that unless it can provide a rescue package for the industry, or at least work for the troops, linguistic theorizing has to be stigmatized as negative. For in general there is no ground for claiming that, in the history of ideas, the exposure of error and the questioning of received wisdom may not qualify as positive contributions. (If they do so fail, then Socrates for one lived in vain.) However, I also see no reason why language studies should not be conducted in a manner that is free from the errors and academic vices that modern orthodox linguistics has perpetuated. (For an example of how one might, in a particular situation, set up an integrationally based language-teaching programme, see D.R. 108 N.L. Love, "Transcending Saussure", Poetics Today, Vol. 10 No. 4, 1989, p. 817.
304
ROY HARRIS
Davis, "Teaching American English as a Foreign Language: An Integrationist Approach".109Such a programme might well be regarded as controversial or even subversive; but that is a different question from whether integrationism can lead to practical implementation.) My critics therefore seem to me to confuse (i) opening up, on a viable theoretical basis, new initiatives in language studies (which I think I have shown to be possible) with (ii) laying down a methodology - i.e. mechanical application procedures - (which I have no wish to do). The dangers of (ii) are well illustrated in David Fleming's paper. The case history he takes is that of conversation analysis. Fleming comments: "Harris's work, while devastating in its critique of conventional linguistics, has failed to produce the kind of 'positive' methodology employed by conversation analysts (p. 205)". This observation would be even more apposite if the term positivist had been substituted for positive. What happens when theorists take it upon themselves to supply a methodology is that the resultant analyses proceed, solemnly and inevitably, to "reveal" a structure in the "data" that reflects, point by point, the "system" that is already tacitly incorporated in the methodological procedure. As Fleming points out, what Sacks and his associates in effect do in their pursuit of the "rules" of conversational turn-taking is not to examine each individual linguistic act as a unique, contextualized, communicational episode, but, on the contrary, "impose on that act hypostasized invariants that exist independently of it". In this way, as he says, "the decontextualized abstractions of structuralism are thus resurrected in full force" (p. 199). The end product is the very opposite of what an integrational approach seeks to achieve. That is to say, conversation is treated "as a virtually autonomous domain". It is precisely because I do not wish to make the same kind of mistake that, as an integrational theorist, I stop short of supplying anything that could be construed as an integrational methodology. Nevertheless, if integrational "work for the troops" is what is needed, how in very general terms - might it be provided? First of all, it would have to be recognized that if language studies are to have any serious theoretical basis, that basis must be semiological. In other words, an integrational linguistics presupposes an integrational semiology, in which "language" is no longer conceived of as a self-contained field of inquiry because to conceive of it thus is already to divorce it unrealistically from the whole range of other human communicative activities. A ground-clearing start might be made by abandoning the orthodox distinction between "core" linguistics and its ancillary subdisciplines in favour of 109
In M. Hayhoe and S. Parker (eds.), Who Owns English?, Buckingham: Open University Press, 1994, pp. 68-76.
FROM AN INTEGRATIONAL POINT OF VIEW
305
distinguishing between various scales or levels of linguistic inquiry, depending on the mode of our involvement in the communicational processes of daily life. This would be necessary in order to replace a "normal science" which focuses on describing the structure of postulated language systems by a "normal science" which focuses on our construction of linguistic experience (admittedly a far more complex enterprise, since acts of communication are infinitely diverse; but one not to be confused - for reasons touched on above - with the current concerns of speech-act theorists or discourse analysts). For this purpose, one would need to recognize at least three scales of linguistic inquiry, which might be called (i) "macrosocial", (ii) "biomechanical", and (iii) "circumstantial".110 The justification for the first of these would be that acts of communication typically depend on and contribute to large-scale communicational patterns which the communicating individuals themselves cannot possibly survey in totality, and over which they have no control. (So for any given act of communication there must always be questions about what macrosocial patterns it presupposes or exemplifies.) The justification for the second scale would be that acts of communication typically involve physiological mechanisms operating in part at least below the level of conscious awareness (however this is defined) of the individuals involved, and exploit physical processes beyond their understanding. (So for any given act of communication there will always be questions for "normal scientists" to ask about its physical and physiological basis.) The justification for the third scale - self-evidently would be that acts of communication typically involve the contextualized integration of macrosocial and biomechanical factors for interactional purposes. In a "normal science" organized along these lines, the relevance to linguistics of macrosocial and biomechanical inquiries would depend on their contribution to questions arising initially for investigation on the third scale, where the inquiry is concerned with elucidating the act of communication as a function of the individual's contextualized experience. These three scales between them furnish a conceptual framework on the basis of which the academic troops could at least be allocated a parade ground, if not a whole terrain for manoeuvres, in linguistic analysis. Part of this terrain would indeed already be familiar territory to them, inasmuch as the macrosocial scale would deal with factors relegated by orthodox linguistics to such sub disciplines as dialectology and sociolinguistics, while the biomechanical scale would concern 110
The designations macrosocial, biomechanical and integrational were originally proposed as general semiological terms for scales of inquiry in "The Semiology of Textualization", Language Sciences, Vol. 6, 1984, pp. 271-86 (reprinted in N. Love (ed.), The Foundations of Linguistic Theory, London: Routledge, 1990, pp. 210-26). I now think it is less misleading to call the third scale circumstantial, reserving integrational to cover the approach as a whole.
306
ROY HARRIS
physiological, physical and perceptual factors of the kind long studied by the orthodox phonetician, the experimental psycholinguist, and others. What would be unfamiliar, and perhaps initially difficult for the troops to grasp, would be the difference between the questions asked by the new "normal science" and the old. For new macrosocial and biomechanical questions would arise from the new theoretical rationale underlying the relationships between the macrosocial, the biomechanical and the circumstantial scales; and these relationships would dictate radical changes of approach. Nevertheless, the overt form of the questions posed might remain the same or deceptively similar. To take a simple biomechanical example, it would no longer make much sense to suppose that the right way to investigate the operation of certain consonant distinctions was to calculate the mean value and standard deviation in voice-onset time in the pronunciation of isolated words, as obtained experimentally in the laboratory from a selected group of informants assembled for that purpose.111 Why not? Because from an integrational point of view, that information tells us little if anything about how such a distinction operates and is maintained in practice, and even less about whether or not it is an important distinction to maintain. Speakers do not normally go around reciting aloud lists of words such as pill, bill, pin, bin, pit, bit, etc. into microphones. Having machines in the phonetics laboratory that can measure voice-onset time in centiseconds may well be useful for engineers designing a telephone system. But that is because a telephone system would be useless if its users could not distinguish what their interlocutors were saying at the other end of the line. On the other hand, telling speakers that in order to articulate a given distinction clearly one ideally aims at a difference in voice-onset time of so many centiseconds contributes no more to the speakers' ability to cope with the communicational world in which they live than telling them how many cubic centimetres of daily liquid intake is on average necessary for purposes of survival helps them realize what to do when they are thirsty. The contribution, such as it is, that experimental phonetics and similar areas of science can make to an integrational study of language lies rather in establishing, insofar as it is possible to do so by the techniques currently available at any given time, within what parameters of a physical or physiological nature certain forms of communication are possible, and at precisely what points, for purely physical or physiological reasons, communication is bound to break down. At present little progress has been made towards establishing these operative parameters other than in artificially contrived laboratory situations. But this is like training the troops to shoot at an enemy who always obligingly stands still. 111 The example is from A. Woods, P. Fletcher and A. Hughes, Statistics in Language Studies, Cambridge: Cambridge University Press, 1986, Ch. 1.
FROM AN INTEGRATIONAL POINT OF VIEW
307
John Taylor (this volume, Ch. 10) raises an apparently biomechanical question when he writes of the connexion between language and memory. Infants are not born with a knowledge of this or that language; languages have to be learned, slowly and gradually; the product of the learning activity has to be stored in the brain, i.e. in memory; and using language, whether in production or reception, involves retrieving some aspects of stored knowledge from memory. No matter how we may ultimately characterize a speaker's knowledge of his language, this knowledge, uncontroversially, resides in memory, and must, therefore, be consistent with some plausible model of human memory. (p. 208) It may seem cavilling to say that this is an "apparently" biomechanical question. But it is far from clear that what Taylor claims as uncontroversial is uncontroversial in the least. The cavil is prompted by Taylor's hasty assumptions about the biomechanics involved. From an integrational point of view, there is no warrant for the presuppositions already slipped in with the terms Taylor uses, such as product, stored, retrieving, etc. Here linguistic "knowledge" is evidently treated as a set of static items, stacked up somewhere in the brain. This is already to regard memory as a mental warehouse; i.e. the problem is whether the warehouse is a "plausible model of human memory" at all. That belief may, on the contrary, be nothing but an inherited corollary of telementational theories of communication. In which case, the "consistency" which Taylor demands would be nothing more than a way of insisting that the linguist subscribe to the telementational fallacy at one remove. Thus at least two notes of caution are in order here. First, Taylor makes no mention of the fact that the (alleged) mnemonic function of language is a well-worn topos in the Western linguistic tradition. Memory was not regarded as a prerequisite for language: on the contrary, language was regarded as a mechanism facilitating memory. Before we take on trust any assumptions about the notion that linguistic knowledge "resides in memory", it would be as well to examine in some historical detail how that idea ever emerged. One suspects that we are dealing not with an attempt to supply an explication for empirically observed "facts" about human memory, but to provide part of an α priori epistemology. Second, it would be naive to assume in any case that memory processes are, from a biomechanical point of view, simple processes of retrieval. How sure can we be that the effort to recall a forgotten name is like looking for a lost object (i.e. in biomechanical as distinct from metaphorical terms)? If an integrational approach can contribute anything to these questions, it would be to suggest that what we call "memory" is probably not the mechanical operation of
308
ROY HARRIS
some independent retrieval system, but the creative construction of a schema for integrating the past with the present. Whether that suggestion can be turned into any kind of programme for empirical research on language and memory I rather doubt. On the macrosocial scale, the position at first sight might appear to be quite different. Cannot an integrational état-major at least equip the troops here with weapons already stockpiled under the orthodox regime? Is there not a solid, well-documented repository of sociolinguistic facts and methods of investigation? Unfortunately, no. The arsenal of orthodox sociolinguisties was stocked on the basis of segregational assumptions. m Tout est à refaire. The difference between what has been done hitherto and what needs to be done is rather like the difference in cartography between a geophysical map of the United Kingdom and a psychophysical map showing the inhabitants' assumptions about the geography of the land they inhabit. On the former, the distance between Manchester and Glasgow will appear as greater than the distance between Manchester and London. On the latter, it may well appear as less. To take another example, it would no longer be regarded as adequate from an integrational point of view to investigate dialect differences by plotting isoglosses to show the ways in which neighbouring communities characteristically differ from one another in details of pronunciation, vocabulary, etc., or by trying to quantify the distribution of such features among groups defined on the basis of age, sex, social class and other variables. For this is already to make the basic segregational assumption that "dialects exist"113, and that the only descriptive task for the linguist is to go out with notebook, tape-recorder, etc. and identify them. What would be required, rather, is an investigation of what speakers recognize in the flux of linguistic variation in which they live as being "dialectal". In other words, the macrosocial diversity in a linguistic community would be investigated as a function of the communicational significance attached to the perception - or non-perception - of variations by the members of the community themselves. And this in turn would be relativized to the communicational context of that perception. There is undoubtedly enough "normal science" even in this single subprogramme to occupy the academic troops in drills and manoeuvres for a long time to come. Similarly, I venture to think that my book Signs of Writing (London: Routledge, 1995) lays the foundations for a new grammatology. In this book the study of writing is based - for the first time - on an explicit, coherent 112
See, again, "Language as Social Interaction: Integrationalism Versus Segregationalism". R. Harris, "The Dialect Myth", in Development and Diversity: Language Variation across Time and Space: A Festschrift for Charles-James N. Bailey, ed. J. Edmondson, C. Feagin and P. Muhlhausler, Dallas: Summer Institute of Linguistics, 1990, pp. 3-19. 113
FROM AN INTEGRATIONAL POINT OF VIEW
309
and internally consistent set of principles. What is proposed is a theoretical basis on which it will be possible for future generations of students to analyse all forms of graphic communication in ways that are far more interesting than is possible through the traditional approach to writing. I believe this has a wide range of potential applications in fields as diverse as computer programming and teaching children to read. At the very least, it offers at last a way out of the dead end into which the study of writing has been shunted by orthodox linguistics throughout most of the present century. Moreover, there are already signs that such inquiries are beginning to be regarded as of interest to linguists, even if those interested do not always appreciate the conceptual break with orthodox segregationalism which this interest implies, or see that the long-term consequence for their discipline is that it must surrender the hallowed equation between language and speaking, and surrender at the same time to the humanities what belongs to humanity. What the troops might find yet more difficult to cope with would be the persistent intrusion into their drills and manoeuvres of awkward questions about the authority, the politics, and the morality of the activities in which they were so assiduously engaged. As some linguists have learned to their cost, "collecting linguistic data" and "writing grammars of the local language" are not always regarded by suspicious foreign governments or populations as the innocent occupations those bland descriptions suggest. But the fact that missionary linguists sometimes end up in tribal cooking pots is no more surprising than the fact that even well trained troops sometimes end up as casualties. This is simply the sharp end of a reminder to the effect that there is no such thing as a socially neutral professional linguistics, any more than there is a socially neutral professional militarism. What the above proposals do not supply, however, is a working schedule. The principal reason why structuralism and generativism have proved so popular in the linguistics industry is that - like historical linguistics before them - both are easy to teach and provide formulas which make it possible for young graduate students to do fairly self-contained pieces of research as doctoral dissertations. ("The vowel system in Wallahwallah", "The subordinate clause in Glouglou", "The acquisition of consonant clusters by two-year-old Plonque twins", etc.) This kind of work can be generated ad infinitum, and jobs and careers secured on the basis of it, without the workers ever having to bother their heads about the nature of the enterprise. A competent university supervisor can show them "how to do it" in a matter of weeks: all the student has to do then is "get on with it". The reason why an integrational approach holds out no hope of such an easy life is that to take an integrational approach in the first place requires some fundamental rethinking about what you are doing; and this in turn requires the
310
ROY HARRIS
exercise of more intellectual muscle in the workplace than many graduate students are capable of. The first thing they would have to grasp, as Nigel Love has put it, is that if you take this approach, linguistics becomes "its own subject matter". 114 You are required "to relocate the line between the linguistic and the metalinguistic", or at least to think very hard indeed about where that line runs. And there are no simple formulas for doing this, as, fortunately, there are for counting phonemes or writing rules for passivization. All this is bad news for the academic empire-builders who so often run university departments: they are the last people likely to be converted to integrational linguistics. Keeping the academic troops busy and the doctoral dissertations churning out, however, would in itself be an unworthy objective in the long run. We already have more than enough "schools" of linguistics to keep it all going until governments, funding agencies and the public begin to wonder if it isn't rather a waste of resources. (There are, some would say, clear signs that they are beginning to wonder that already.) If an integrational linguistics is worth pursuing, it is as a pragmatics of self-understanding and a basis for lay linguistic therapy rather than as part of a university curriculum. For simply to elaborate a new metalanguage and give it academic employment in lectures and learned periodicals would, in the end, be to occupy no more than an "insular possession": What matters it if a man has acquired the tongue to perfection when he puts it to such a purpose?... The scholar is lost to all but himself, estranged from those who should be his flock, and his glistering accomplishment...but the mirror to his own desires and vanity.115
114 N. Love, "The Locus of Languages in a Redefined Linguistics", in H.G. Davis and T.J. Taylor (eds.), Redefining Linguistics, London: Routledge, 1990, pp. 53-117. 115 Timothy Mo, An Insular Possession, London: Picador, 1986, p. 65.
Epilogue The core of Roy Harris's contribution to language study consists of a rejection of the assumptions underlying mainstream theorising, together with an outline of the principles of a conceptually sounder approach. Joseph, for one, not only believes that these two aspects are separate and separable, but that the criticism of the "orthodoxy" might be dispensed with altogether: u not only is Harris's historical exposition filled with red herrings - it is itself the biggest, reddest herring of all, deflecting attention from his original thinking on language and communication and his program for redefining linguistics" (p. 38). Is this a tenable view? The programme for redefining linguistics may for convenience be repeated here in its most succinct published form: Integrational linguistics may be defined in terms of the acceptance of three principles which differ from those widely accepted by linguists belonging to one or other of the orthodox modern schools. These three principles are: (i) the integrational character of the linguistic sign, (ii) the indeterminacy of linguistic form, and (iii) the indeterminacy of linguistic meaning. Principle (i) implies that linguistic signs are not autonomous objects of any kind, either social or psychological, but are contextualised products of the integration of various activities by individuals in particular communication situations. The continuous creation of linguistic signs to meet the exigencies of communication constitutes the first-order linguistic process which is the primary object of study in integrational linguistics. Principles (ii) and (iii) together imply that languages are not fixed codes but second-order social constructs of an intrinsically open-ended, incomplete and variable nature. They are consequently not amenable to the standard bi-planar analyses imposed by orthodox linguistics. The definition proposed is derived from two basic axioms of integrational semiology. The integrational approach is not restricted to those communicational activities traditionally designated by the term language, and a fortiori not to those traditionally designated by the term speech. The two semiological axioms are: Axiom 1. What constitutes a sign is not given independently of the situation in which it occurs or of its material manifestation in that situation.
312
LINGUISTICS INSIDE OUT
Axiom 2. The value of a sign is a function of the integrational proficiency which its identification and interpretation presuppose. These two axioms apply to all signs in interpersonal communication, and more generally in the human environment; for they also apply to natural phenomena to which human beings assign a semiological value (as, for example, in medicine and meteorology). The second of the two axioms quoted refers to "integrational proficiency". This is best defined in the following way. Human beings bring to their firstorder communicational task, which is the creation of signs, capacities of various kinds. These capacities depend on three different types of factor, which may be called "biomechanical", "macrosocial" and "integrational" [= "circumstantial"]. Biomechanical factors pertain to the organic and neuro-physiological mechanisms which underlie communicative behaviour and their exercise in particular physical circumstances. Macrosocial factors pertain to culture-specific patterns of organisation within which communication situations occur. Integrational [= circumstantial] factors pertain to the fitting together of all these within a particular set of circumstances in ways which make sense to the participants involved. The integrational assumption is that any episode of linguistic communication can be analysed in terms of these three sets of factors, and must be if the analysis is to be adequate, in the sense of answering to the actual experience of the participants. An analysis which fails to do this fails to capture the phenomenon, because the integration of these various sets of factors is precisely what constitutes the phenomenon. (Harris 1993c: 321-2) What do critics make of this? Joseph comments approvingly that "the strength of the integrationalist programme lies in its commitment to a diversity of perspectives and sources of insight into language and communication". Borsley and Newmeyer see little that is new in it, but admit that they "do find aspects of Harris's program to be of interest". Fleming's paper is conceived as a contribution to the development of that programme. With the doubtful exception of Baron, not one of the contributors who mention it at all takes issue with the theses on which integrationalism rests. Nor is it clear that those theses have been challenged in other quarters. It may be assumed, therefore, that members of the "orthodox modern schools", although they may not be rushing to espouse integrationalism, have no particular quarrel with it. It seems to threaten no vested intellectual interests. And that, presumably, is because it is not seen as being in intellectual competition with those schools. This is arguably the most striking feature of the debate between Harris and his critics. How does this state of affairs come about?
EPILOGUE
313
In effect, Harris's proposal is that we take seriously Saussure's idea that linguistics should be part of semiology - "une science qui étudie la vie des signes au sein de la vie sociale" (Saussure 1916 [1922: 33]). This means that the study of language is, or should be, intrinsically bound up with the study of communication. Now the "orthodox modern schools" either have no interest in communication or else treat communication as the subject matter of some discipline other than linguistics proper: in this respect Philip Carr merely puts forward an extreme version of a doctrine adopted in milder form by many others. That language is used to communicate is hardly open to denial (pace Carr on Chomsky, whose apparent assertion to the contrary is perhaps best read as flowing from a proposed technical redefinition of the word "language"), but this fact is seen as tangential at best to the concerns of those who wish to describe and analyse linguistic systems and entirely irrelevant to those preoccupied with Chomsky's "I-language". Hence their bewilderment and irritation when Harris attributes a theory of communication to them, and their benign indifference to his proposal of an approach squarely based on attending to what is involved in using language to communicate. From an orthodox standpoint, languages (i.e. determinable idiosynchronic sign-systems) are simply given. Their existence is not liable to empirical disconfirmation. To think otherwise, and to object to the orthodoxy on the ground that their existence is in fact empirically disconfirmed, is to be guilty of some mixture of "behaviourism", "empiricism" (i.e. "the rejection of abstractions on the basis of the impossibility of deriving them by a mechanical procedure from a series of observations" - see Borsley and Newmeyer, p. 59), and refusal to accept the necessity for science to be based on "idealisations". Put in somewhat different terms, what is "given" for the orthodox linguist is the "ideal speakerlistener, in a completely homogeneous speech-community, who knows its language perfectly" - which, as Chomsky immediately goes on to observe, is no more than a reformulation of "the position of the founders of modern general linguistics". Moreover, "no cogent reason for modifying it has been offered" (Chomsky 1965: 3-4). On the basis of one or another version of this idealisation, orthodox linguistics proceeds to theorise about the structure of languages, considered in abstraction from the communicational purposes for which they are used and the communicational processes in which they are implicated. It is hard to see what might in principle be allowed to count as a "cogent reason" for "modifying" this position, but it is easy enough to understand how Harrisian integrationalism, in itself, might be seen as failing to offer one. In the first place, its fundmental tenets are no more subject to empirical test than those of the orthodoxy. For instance, although Harris's assertion that linguistic signs are not autonomous objects of any kind is incompatible with the orthodox
314
LINGUISTICS INSIDE OUT
counter-assertion that linguistic signs are autonomous objects, we are not confronted by a conflict between two statements about the nature of certain antecedently identified portions of the world's furniture called "linguistic signs" that might be resolved by careful inspection of the objects in question. On the contrary, whether or not one takes linguistic signs to be autonomous is one factor involved in deciding what counts as a linguistic sign in the first place. Is there a standpoint from which it could be demonstrated that one decision is "right" and the other "wrong"? Secondly, treating signs as autonomous offers the important pragmatic advantage of providing orthodox linguistics with a definite object of study. This, ultimately, is the basis for the attractiveness of the orthodoxy: that it points the way to a specific programme of activity for linguists. Harris makes no bones about conceding that integrationalism does not perform this pragmatic, discipline-internal function. For one thing, every episode of language use is to be treated as a unique event, analysable in terms of the biomechanical, macrosocial and "circumstantial" abilities brought to that event by the participants. In any particular case one may expect some of these abilities to be either idiosyncratic in themselves, or deployed in an idiosyncratic way not explicable without reference to the personal life-history of the individual concerned, and perhaps not even then. Contemplation of such points lends an ironic air to the complaint Harris once voiced (1980: 162) about the "dramatic expansion of the boundaries of linguistics" required by Bloomfield's conception of semantics. Moreover, expanding the boundaries in this way may threaten to reduce linguistics to the collecting of indefinitely many anecdotes. So the choice between an autonomous linguistics and an integrational linguistics may appear to be just that - a choice. From the orthodox linguist's point of view, there is nothing particularly objectionable about the key propositions of integrationalism, but nothing particularly compelling either. The thesis that linguistic signs are not autonomous seems no more decisively demonstrable than its antithesis. If integrationalists think they can erect a viable, institutionalisable practice of language study on the basis of that thesis, then let them get on with it. In fact, it is hard to see that it does offer the basis for any institutionalisable practice, and Harris for one apparently has no interest in seeking to establish any such thing. So there appears to be nothing here to distract autonomists from their own concerns. Hence their unconcerned, laissezfaire attitude to Harris's proposals. In this context Harris's hostile appraisal of orthodox linguistics, far from being dispensable, is crucial. For his argument is that the choice is not between a semiologically-based linguistics and one that happens to prefer to treat the linguistic sign as autonomous, but between an overtly semiological linguistics
EPILOGUE
315
and a pseudo-autonomous linguistics covertly and unwittingly in thrall to an untenable theory of communication. Harris's critique of the orthodoxy may be seen as based on identifying a key fallacy that, once accepted, ineluctably drags in its train the whole suite of errors (as Harris sees them) on which what passes for a modern science of language is founded. That key fallacy is the telementational theory of communication. The other component of the "language myth", the fixed-code theory, is a corollary of the telementational theory, in that fixed codes are required to make telementation possible. But the fixed-code theory flies in the face of linguistic variability. So if it is to be even remotely plausible, some means of identifying the alleged fixed code is required. Hence the doctrine that, broadly speaking, the code has already been identified for us by our writing practices, and the concomitant perversion of our understanding of the relationship between writing and speech. There is also a requirement for some story as to how language-users might come into possession of the fixed code. Hence the de-humanising doctrine of the "language machine". There are of course other ways of understanding the relationship between the various parts of Harris's critique, but there is no doubt that if linguistics is not in fact committed to a telementational theory of communication, his case against the orthodoxy is in serious trouble. So those of the foregoing chapters that discuss telementation, and Harris's response to these discussions, are of considerable interest in the present context. Telementationism in modern linguistics, according to Harris, starts with Saussure. Joseph argues that "the evidence for Saussure's telementationism is ambiguous", and that in any case "he did not have a 'theory of communication ' at αll' (p. 28). What Joseph says about Saussure and telementation is echoed by what Borsley and Newmeyer say about generativism and telementation: that (i) generativism (or at any rate Chomsky) explicitly rejects it, and (ii) that "the goal of generative grammar is not now and never has been to explain communication" and consequently "is under no obligation to propose an alternative to the telementational conception". These waters are somewhat muddied by the third essay that discusses telementation. Carr concedes that generative linguistics, as currently practised, is wedded "to a version of the telementation thesis". But it is clear enough that Carr's conception of the telementation thesis itself is idiosyncratic: ... talk of manifestation, realisation or externalisation of linguistic expressions is telementational talk: if... expressions are manifested, they "become" observable, are transmogrified into, or encoded in, observable physical events. If this were not being claimed by generativists, there would be no sense in the notion, much appealed to by generativists, of "Primary Linguistic Data" (PLD): this data,
316
LINGUISTICS INSIDE OUT
crucial to the current generative conception of language "acquisition", is claimed to be sensory and linguistic. In short, generativists currently claim that acoustic events may have linguistic properties, an idea which is positively medieval in its occult nature: in the late twentieth century, we ought to be asserting that acoustic events have only acoustic properties. Alas, the myth of telementation is alive and well and embedded in the PLD idea. (p. 71) What Carr seems to be espousing is a conception of the "linguistic" that would completely disconnect anything properly called "linguistic" from sensible physical events of any kind. Whether this is (in another sense) sensible is not a matter for discussion here; the relevant point is that "talk of manifestation, realisation or externalisation of linguistic expressions" is surely not eo ipso "telementational talk". Now it is perfectly possible, non tali auxilio, to quote passages from generativists that seem to imply a telementational understanding of communication. Harris quotes several such passages, and having done so is content to treat Borsley and Newmeyer's claim that generativism has no truck with telementation as refuted. But the interesting question is how the claim ever came to be made, given the textual evidence to the contrary. Harris scathingly suggests ignorance on Borsley and Newmeyer's part, but there is an alternative possibility - that the passages Harris quotes are taken, not as embodying or implying a particular theory of communication, but as unremarkable and entirely anodyne statements as to what communication is. Something to this effect is overtly suggested by Joseph, in respect of Saussure's alleged status as a telementationalist: [Saussure's] speech circuit is based on no more elaborate assumptions than that "people communicate" and "words have meanings". If these things are so, then whenever someone tells someone else something...the actual utterance (as phonation) is sandwiched between meaning as speaker's intention and meaning as hearer's interpretation. If this is telementation, then anyone who believes that words have meanings, or that communication can occur, is a telementationalist. That seems ultimately to become Harris's position, and it threatens to turn the whole concept of telementation into a red herring. (pp. 33-4) This, of course, is grist to Harris's mill: the telementational theory, he would say, is so firmly rooted in Western thinking about language that it has come practically to have the status of a mere explanation or definition of the word "communication". But would that satisfy the critics?
EPILOGUE
317
Saussure, Joseph says, never discusses communication - hardly even uses the word. Therefore Saussure has no theory of communication and a fortiori is not a telementationalist. In respect of generativism, Borsley and Newmeyer put a similar point the other way round. Chomsky, they say, explicitly rejects telementation as a theory of communication; furthermore, generativists are not in the business of elaborating such theories. That surely settles the matter. Or does it? The crux here is the legitimacy of critical discourse conducted at a level of generality at which it may be appropriate to attribute to thinkers positions that they would consciously deny adopting. In other words, what is at issue is the validity of the second of Ernest Gellner's "two quite distinct criteria for what a man believes, or for his doctrinal classification": (a) self-characterisation, and (b) interpretation of his centre of gravity. As far as criterion (a) is concerned, what a man says about himself is final. He ought to know... Criterion (b) is more complex and less conclusive... What makes a thinker significant...is that his work...contains some central ideas which pervade and inspire the corpus as a whole. Just what those ideas are...must in the nature of things be in greater or lesser measure a matter of interpretation... (Gellner 1985: 14) Harris's critics are applying Gellner's criterion (a). Harris, however, is operating with criterion (b), in conjunction with the idea that "every linguistic theory presupposes a theory of communication" (Harris 1987a: 205). That is to say, the pretheoretical phenomenon confronting anyone who would engage in analytical inquiry into language, according to Harris, is the fact of linguistic communication among human beings. Accordingly, whatever the theorist might overtly say, his approach to language necessarily implies some view of how this happens. Denying that, by criterion (a), a given theorist has any theory of communication at all simply misses the point. Every linguistic theory presupposes a theory of communication. That is the key both to Harris's critique of the orthodoxy and his integrationalist proposals. It emerges from a discourse in which he is concerned to make general statements of a kind that, unless handled with more care than he can always be bothered to apply, are likely to be vulnerable to charges of oversimplification, polemical hyperbole and, in general, a cavalier attitude to the "academic fastidiounsess" Harris attributes to Joseph. Accusations involving disparaging use of the term "rhetoric" can be found in both Joseph's and Borsley and Newmeyer's papers. A trivial illustration of the style is to be seen in the quotation from Harris's reply that stands as an epigraph to this volume. The academically fastidious may be
318
LINGUISTICS INSIDE OUT
inclined to ask how many linguists, in fact, have ended up in tribal cooking-pots? Who are they, and when and where did they meet their fate? A less trivial example is quoted by Borsley and Newmeyer (p. 56): The normative grammarian is no sooner evicted from paradise than his intellectual stock-in-trade is confiscated by his evictors. The very rules he was condemned for peddling turn up on the linguist's stall, sometimes - but not always - with new labels attached. (Harris 1987b: 128-9) Assertions of this kind are easy meat for any critic not a priori sympathetic and only too pleased to have an excuse for not considering whether behind the fanciful mode of expression there lies a valid point. Less trivial still is the charge that Harris is sometimes careless about defining terms whose clarity is seemingly crucial to the argument: Joseph's discussion of the confusion of surrogationalism with nomenclaturism is a case in point (for another instance see Love 1989a: 800, n. 2). Harris's rhetoric, let it be conceded, is not altogether apt for the task of persuading the ill-disposed to attend to his message. But that is ultimately a matter of small importance. The important question is whether there is any way of putting the message such that orthodox linguistic "scientists", given that selfdescription, would consider themselves obliged to take it seriously. If the answer is "no", as Harris himself seems to imply, the debate with his mainstream critics is in the end as much about the validity and legitimacy of institutionalising inquiry along certain lines as it is about the nature of language. It is, to that extent, a philosophical debate, in which Harris's views on the philosophy of linguistics are intertwined with his views on language in a way that establishes for him a position in (relation to) linguistics reminiscent of that of the later Wittgenstein visà-vis philosophy. Now the philosophy of philosophy is itself philosophy; philosophers cannot point to a disciplinary boundary that might provide an excuse for discounting metaphilosophy of unwelcome import. But is there any intellectually respectable reason to deny that linguists are in the same boat? Contemplating Harris's answer to this question, and how he would justify that answer, is one way of seeing the essential unity of the positive and negative sides of his work.
References Aarsleff, Hans. 1982. From Locke to Saussure. Minneapolis: University of Minnesota Press. Allen, W. Sidney. 1953. Phonetics in Ancient India. London: Oxford University Press. Andresen, Julie T. 1990. Linguistics in America 1769-1924: A Critical History. London: Routledge. anon. 1980. Review notice of R. Harris (1980). Choice, November, 390. Ardener, Edwin. 1983. "Social Anthropology, Language and Reality". R. Harris (ed.) 1983.143-56. Atran, S. 1990. Cognitive Foundations of Natural History. Cambridge: Cambridge University Press. Auer, Peter & Aldo di Luzio (eds.). 1992. The Contextualization of Language. Amsterdam & Philadelphia: John Benjamins. Auroux, Sylvain. 1994. La Révolution technologique de la grammatisation. Liège: Mardaga. & D. Kouloughli. 1991. "Why Is There No 'True' Philosophy of Linguistics?". Language & Communication 11/3.151-163. Austin, J.L. 1975. How to Do Things with Words. 2nd edn., ed. by J.O. Urmson & M. Sbisà. Cambridge: Harvard University Press. (1st ed. 1962) Baker, Gordon P. 1988. Wittgenstein, Frege and the Vienna Circle, Oxford: Basil Blackwell. & P.M.S. Hacker. 1980. Wittgenstein: Understanding and Meaning. Oxford: Basil Blackwell. & . 1985 .Wittgenstein: Rules, Grammar and Necessity. Oxford: Basil Blackwell. Bar-Hillel, Yehoshua. 1971. Aspects of Language. Jerusalem: Magnes. Baron, Naomi S. 1984. "Computer Mediated Communication as a Form of Language Change". Visible Language 18/2.118-141. . 1986. "Language, Sublanguage, and the Promise of Machine Translation". Computers and Language 1/1.3-19. . 1992. Growing Up with Language. Reading, Mass.: Addison-Wesley. . 1994. "Do Words Have Meanings? Dictionaries, Definitions, and Context". Semiotica 99.147-162. Barsalou, L.W. 1983. "Ad Hoc Categories". Memory and Cognition 11.211-227.
320
LINGUISTICS INSIDE OUT
Barwise, Jon & John Perry. 1983. Situations and Attitudes. Cambridge, Mass.: MIT Press. Basso, K.H. 1974. "The Ethnography of Writing". Explorations in the Ethnography of Speaking, ed. by R. Bauman & J. Sherzer, 425-32. Cambridge: Cambridge University Press. Beach, R., J.L. Green, M.L. Kamil & T. Shanahan (eds.). 1992. Multidisciplinary Perspectives on Literacy Research. Urbana, Illinois: National Council of Teachers of English. Becker, A.L. 1979. "Text-building, Epistemology, and Aesthetics in Javanese Shadow Theatre". The Imagination of Reality: Essays in Southeast Asian Coherence Systems, ed. by A. L. Becker & Aram A. Yengoyan, 211-243. Norwood, N.J.: Ablex. Bender, J. & D.E. Wellbery. 1990. The Ends of Rhetoric: History, Theory, Practice. Stanford: Stanford University Press. Bennett, Jonathan. 1976. Linguistic Behaviour. Cambridge: Cambridge University Press. Berlin, B. & P. Kay. 1969. Basic Color Terms. Berkeley: University of California Press. Bernstein, Basil. 1971. Class, Codes, and Control, vol.1. London: Routledge & Kegan Paul. Bertinetto, P. M. 1994. "Phonological Representation of Morphological Complexity: Alternative Models (Neuro- and Psycholinguistic Evidence)". Cognitive Linguistics 5.77-109. Bhaskar, Roy. 1975. A Realist Theory of Science. Leeds: Leeds Books. Bierwisch, M. 1967. "Some Semantic Universals of German Adjectivals". Foundations of Language 3.1-36. & R. Schreuder. 1992. "From Concepts to Lexical Items". Cognition 42.23-60. Billig, M. 1987. Arguing and Thinking: A Rhetorical Approach to Social Psychology. Cambridge: Cambridge University Press. . 1991. Ideology and Opinions: Studies in Rhetorical Psychology. London: Sage. Bley-Vroman, Robert D. 1989. "What is the Logical Problem of Foreign Language Learning?" Linguistic Perspectives on Second Language Acquisition, ed. by Jacqueline Schachter & Susan D. Gass, 41-68. Cambridge: Cambridge University Press. Bloomfield, Leonard. 1924. Review of Saussure 1922. Modern Language Journal 8/5.317-9. . 1927. "On Recent Work in General Linguistics". Modern Philology 25.211-230. . 1933. Language. New York: Holt, Rinehart & Winston. (1935, London: Allen & Unwin.)
REFERENCES
321
Boas, Franz. 1911. "Introduction". Handbook of American Indian Languages. Washington: Government Printing Office. Borsley, Robert D. 1992. Review of Davis & Taylor (eds.) 1990. Lingua 88.67-71. & F. J. Newmeyer (this volume, 42-64). "The Language Muddle: Roy Harris and Generative Grammar." Boyarin, J. (éd.). 1993. The Ethnography of Reading. Berkeley, Cal.: University of California Press. Brandt, D. 1992. "The Cognitive as the Social: An Ethnomethodological Approach to Writing Process Research". Written Communication 9.315-355. Bright, William. 1988. "Written and Spoken Language in South Asia". On Language: Rhetorica, Phonologica, Syntactica: A Festschrift for Robert P. Stockwell fom his Friends and Colleagues, ed. by Caroline Duncan-Rose & Theo Vennemann, 22-38. London: Routledge. Bromberger, Sylvain. 1992. "Types and Tokens in Linguistics". On What We Know We Don't Know, by S. Bromberger. Chicago: University of Chicago Press. & M. Halle. 1989. "Why Phonology is Different". Linguistic Inquiry 20/1.51-70. & M. Halle. 1992. "The Ontology of Phonology". On What We Know We Don't Know, by S. Bromberger, 209-228. Chicago: University of Chicago Press. Brown, Penelope & Stephen Levinson. 1978. "Universals in Language Use: Politeness Phenomena". Questions and Politeness, ed. by Esther N. Goody, 56324. Cambridge: Cambridge University Press. Burke, Kenneth. 1969. A Rhetoric of Motives. Berkeley: University of California Press. Burton-Roberts, Noel C. 1994. "Ambiguity, Sentence and Utterance: A Representational Approach". Transactions of the Philological Society 92/2.179-212. & Philip Carr. In preparation. "I-language and the Nature of Phonology". Centre for Research in Linguistics, University of Newcastle. Butler, Samuel. 1912. The Notebooks of Samuel Butler, ed. by H.F. Jones. London: Fifield. Repr. London: Hogarth Press, 1985. Button, G. 1991. "Introduction: Ethnomethodology and the Foundational Respecification of the Human Sciences". Ethnomethodology and the Human Sciences, ed. by G. Button, 1-9. Cambridge: Cambridge University Press. Cairns, Helen & Charles Cairns. 1976. Psycholinguistics. New York: Holt, Rinehart & Winston. Chafe, Wallace. 1985. "Linguistic Differences Produced by Differences Between Speaking and Writing". Literacy, Language, and Learning: The Nature and Consequences of Reading and Writing, ed. by D. R. Olson, N. Torrance & A. Hildyard, 105-123. Cambridge: Cambridge University Press. Chomsky, Noam. 1957. Syntactic Structures. Janua linguarum series minor 4. The Hague: Mouton.
322
LINGUISTICS INSIDE OUT
. 1961. "Some Methodological Remarks on Generative Grammar". Word 17.219-239. . 1965. Aspects of the Theory of Syntax. Cambridge, Mass.: MIT Press. . 1966. Cartesian Linguistics. New York & London: Harper & Row. . 1972. Language and Mind. Enlarged Edition. New York: Harcourt Brace Jovanovich. . 1975a. "Questions of Form and Interpretation". Linguistic Analysis 1.75109. . 1975b. Reflections on Language. New York: Pantheon. . 1979. Language and Responsibility, Based on Conversations with Mitsou Ronat. New York: Pantheon. . 1980. Rules and Representations. New York: Columbia University Press. . 1981. Lectures on Government and Binding. Studies in Generative Grammar 9. Dordrecht: Foris. . 1986a. Barriers. Cambridge, Mass.: MIT Press. . 1986b. Knowledge of Language: Its Nature, Origin, and Use. New York: Praeger. . 1988. Language and Problems of Knowledge: The Managua Lectures. Current Studies in Linguistics 16. Cambridge, Mass.: MIT Press. . 1990. "On Formalization and Formal Linguistics". Natural Language and Linguistic Theory 8.143-147. . 1991a. "Linguistics and Adjacent Fields: A Personal View". The Chomskyan Turn: Generative Linguistics, Philosophy, Mathematics, and Psychology, ed. by Asa Kasher, 3-25. Oxford: Blackwell. . 1991b. "Linguistics and Cognitive Science: Problems and Mysteries". The Chomskyan Turn: Generative Linguistics, Philosophy, Mathematics, and Psychology, ed. by Asa Kasher, 26-55. Oxford: Blackwell. . 1992. "A Minimalist Program for Linguistic Theory". MIT Occasional Papers in Linguistics No. 1. . 1993a. "A Minimalist Program for Linguistic Theory". The View from Building 20: Essays in Honor of Sylvain Bromberger, ed. by Kenneth Hale & Samuel Jay Keyser, 1-52. Cambridge, Mass.: MIT Press. . 1993b. "Naturalism and Dualism in the Study of Language and Mind". International Journal of Philosophical Studies 2/2.181-209. . 1995. "Language and Nature". Mind 104/413.1-61. Cicourel, A.V. 1974. Cognitive Sociology: Language and Meaning in Social Interaction. New York: The Free Press. Clahsen, Harald. 1988. "Parameterized Grammatical Theory and Language Acquisition". Linguistic Theory in Second Language Acquisition, ed. by Suzanne Flynn & Wayne O'Neil, 47-75. Reidel: Kluwer. & Pieter Muysken. 1986. "The Availability of Universal Grammar to Adult and Child Learners". Second Language Research 2.93-119. Coe, M.D. 1992. Breaking the Maya Code. London: Thames & Hudson.
REFERENCES
323
Comrie, Bernard. 1988. "Coreference and Conjunction Reduction in Grammar and Discourse". Explaining Language Universals, ed. by John A. Hawkins, 186208. Oxford: Basil Blackwell. . 1989. Language Universals and Linguistic Typology, 2nd ed. Chicago: University of Chicago Press. Conklin, Harold C. 1959. "Linguistic Play in its Cultural Context". Language 35.631-636. Conway, G. 1989. Wittgenstein on Foundations. Atlantic Highlands, N.J.: Humanities Press International. Cook-Gumperz, J. (ed.). 1986. The Social Construction of Literacy. Cambridge: Cambridge University Press. Coulter, J. 1991. "Logic: Ethnomethodology and the Logic of Language". Ethnomethodology and the Human Sciences, ed. by G. Button, 20-50. Cambridge: Cambridge University Press. Cowley, S.J. 1994. "Conversational Functions of Rhythmical Patterning: A Behavioural Perspective". Language & Communication 14/4.353-76. Croce, Benedetto. 1958. Estetica, 10th ed. Bari: Laterza. Crystal, David. 1981. "Improvising Language", review of R. Harris 1981a. The Listener, 17 September, 311. . 1991. A Dictionary of Linguistics and Phonetics, 3rd ed. Oxford: Blackwell. . 1974. "Paralinguistics". Current Trends in Linguistics, ed. by T.A. Sebeok, 12/1.265-95. The Hague: Mouton. Davidson, Donald. 1980. "Actions, Reasons and Causes". Essays on Actions and Events, by D. Davidson, 3-19. Oxford: Oxford University Press. Davis, Daniel R. 1994. "Teaching American English as a Foreign Language: An Integrationist Approach". Who Owns English?, ed. by M. Hayhoe & S. Parker, 68-76. Buckingham: Open University Press. Davis, Hayley G. & Talbot J. Taylor (eds.). 1990. Redefining Linguistics. London & New York: Routledge. Deigh, John. 1983. Review of R. Harris 1980. Ethics 93/2.420. DeCastell, S., A. Luke & K. Egan (eds.). 1986. Literacy, Society, and Schooling: A Reader. Cambridge: Cambridge University Press. Deprez, Viviane. 1992. "Raising Constructions in Haitian Creole". Natural Language and Linguistic Theory 10.233-278. Derrida, Jacques. 1977. "Signature Event Context". Glyph 1.172-97. Derwing, Bruce L. & Terrance M. Nearey. 1986. "Experimental Phonology at the University of Alberta". Experimental Phonology, ed. by John J. Ohala & Jeri J. Jaeger, 187-210. Orlando: Academic Press. , Terrance M. Nearey & Maureen L. Dow. 1986. "On the Phoneme as the Unit of the 'Second Articulation'". Phonology Yearbook 3.45-70. Diamond, Cora. 1991. The Realistic Spirit: Wittgenstein, Philosophy and the Mind. Cambridge, Mass.: MIT Press.
324
LINGUISTICS INSIDE OUT
Dinneen, Francis P. 1974. Foreword to C.H. Brown, Wittgensteinian Linguistics. The Hague: Mouton. Dingwall, William 0 . 1965. Transformational Generative Grammar: A Bibliography. Washington: Center for Applied Linguistics. Dirven, R. & John R. Taylor. 1988. "The Conceptualization of Vertical Space in English: The Case of Tall". Topics in Cognitive Linguistics, ed. by B. RudzkaOstyn, 379-402. Amsterdam: Benjamins. Dresher, B. Elan & Aditi Lahiri. 1991. "The Germanic Foot: Metrical Coherence in Old English". Linguistic Inquiry 22.251-286. Drew, P. & John Heritage. 1992. "Analyzing Talk at Work: An Introduction". Talk at Work: Interaction in Institutional Settings, ed. by P. Drew & J. Heritage, 3-65. Cambridge: Cambridge University Press. Dummett, Michael. 1982. Review of R. Harris 1981a. The London Review of Books 4/15, 19 August, 9-10. Duranti, Alessandro. 1988. "Ethnography of Speaking: Toward a Linguistics of the Praxis". Linguistics: The Cambridge Survey, vol. 4: Language: The SocioCultural Context, ed. by Frederick J. Newmeyer, 210-228. New York: Academic Press. Duranti, Alessandro & Charles Goodwin (eds.). 1992. Rethinking Context: Language as an Interactive Phenomenon. Cambridge: Cambridge University Press. Edwards, J. A. 1993. "Principles and Contrasting Systems of Discourse Transcription". Talking Data: Transcription and Coding in Discourse Research, ed. by J.A. Edwards & M.D. Lampert, 3-31. Hillsdale, N.J.: Lawrence Erlbaum. Emonds, Joseph E. 1985. A Unified Theory of Syntactic Categories. Studies in Generative Grammar 19. Dordrecht: Foris. . 1986. "Grammatically Deviant Prestige Constructions". A Festschrift for Sol Sapor ta, ed. by Michael Brame, Heles Contreras & Frederick J. Newmeyer, 93-131. Seattle: Noit Amrofer. Epstein, Samuel D., Suzanne Flynn & Gita Martohardjono. 1994. "Explanation in Theories of Second Language Acquisition". Unpublished ms. Farrell, T. 1993. The Norms of Rhetorical Culture. New Haven: Yale University Press. Figueroa, Esther. 1994. Sociolinguistic Metatheory. Oxford: Pergamon. Firth, J.R. 1957. "Ethnographic Analysis and Language with Reference to Malinowski's Views". Man and Culture, ed. by Raymond Firth, 93-118. London: Routledge & Kegan Paul. Fodor, Jerry A. 1979. The Language of Thought. Cambridge, Mass.: Harvard University Press. . 1981. "The Present Status of the Innateness Controversy". Representations. by J. Fodor, 257-316. Cambridge, Mass., MIT Press: & Z. W. Pylyshin. 1988. "Connectionism and Cognitive Architecture: A Critical Analysis". Cognition 26.5-71.
REFERENCES
325
Frege, Gottlob. 1966. Translations from the Philosophical Writings of Gottlob Frege, ed. by Peter T. Geach & Max Black. Oxford: Blackwell. . 1977. Logical Investigations, ed. Peter T. Geach. Oxford: Blackwell. Frentz, T. S. 1985. "Rhetorical Conversation, Time, and Moral Action". Quarterly Journal of Speech 71.1-18. Fromkin, Victoria & R. Rodman. 1978. An Introduction to Language, 2nd ed. New York: Holt, Rinehart & Winston. Gaita, R. 1991. "Language and Conversation: Wittgenstein's Builders".Wittgenstein: Centenary Essays, ed. by A. Phillips Griffiths, 101-115. Cambridge: Cambridge University Press. Garfinkel, H. 1967. Studies in Ethnomethodology. Englewood Cliffs, N.J.: Prentice-Hall. . 1991. "Respecification: Evidence for Locally Produced, Naturally Accountable Phenomena of Order, Logic, Reason, Meaning, Method, etc. in and as of the Essential Haecceity of Immortal Society (I) - An Announcement of Studies". Ethnomethodology and the Human Sciences, ed. by G. Button, 10-19. Cambridge: Cambridge University Press. Gardner, H. 1985. The Mind's New Science: A History of the Cognitive Revolution. New York: Basic Books. Garman, M. 1990. Psycholinguistics. Cambridge: Cambridge University Press. Gee, J. P. 1989. "Literacy, Discourse, and Linguistics: Introduction". Journal of Education 17.1-176. Geisler, C. 1994. Academic Literacy and the Nature of Expertise: Reading, Writing, and Knowing in Academic Philosophy. Hillsdale, N.J.: Lawrence Erlbaum. Gelb, I.J. 1963. A Study of Writing, 2nd ed. Chicago: University of Chicago Press. ( l s t e d . 1952) Gellner, Ernest. 1985. Relativism and the Social Sciences. Cambridge: Cambridge University Press. Germain, Claude. 1979. The Concept of Situation in Linguistics. Transi. B.J. Wallace. Ottawa: University of Ottawa Press. Givon, Talmy. 1979. On Understanding Grammar. New York: Academic Press. . 1984. Syntax: A Functional-Typological Introduction, vol. 1. Amsterdam: John Benjamins. . 1990. Syntax: A Functional-Typological Introduction, vol. 2. Amsterdam: John Benjamins. Gladwin, T. 1970. East is a Big Bird: Navigation and Logic on a Pulawat Atoll. Cambridge, Mass.: Harvard University Press. Goldsmith, John. 1976. "Autosegmental Phonology". Ph. D. dissertation, MIT. . 1992. "A Note on the Genealogy of Research Traditions in Modern Phonology". Journal of Linguistics 28.149-164. Goodenough, W. H. 1956. "Componential Analysis and the Study of Meaning". Language 32.195-216.
326
LINGUISTICS INSIDE OUT
Goody, Jack & Ian Watt. 1968. "The Consequences of Literacy". Literacy in Traditional Societies, ed. by J. Goody, 27-68. Cambridge: Cambridge University Press. Greenfield, Patricia. 1972. "Oral or Written Language: The Consequences for Cognitive Development in Africa, the United States, and England". Language and Speech 15.169-178. Gumperz, John J. & Dell Hymes (eds.). 1972. Directions in Sociolinguistics: The Ethnography of Communication. New York: Holt, Rinehart, Winston. Habermas, Jurgen. 1976. "Was heiBt Universalpragmatik?" Sprachpragmatik und Philosophie, ed. by Karl-Otto Apel, 174-272. Frankfurt am Main: Suhrkamp. Hacker, P.M.S. 1986. Insight and Illusion: themes in the philosophy of Wittgenstein, revised edition. Oxford: Clarendon Press. (1st ed. 1972.) . 1990a. Wittgenstein: Meaning and Mind. Oxford: Basil Blackwell. 1990b. "Chomsky's Problems". Language & Communication 10/2.127-148 Haegeman, Liliane. 1990. "Understood Subjects in English Diaries: On the Relevance of Theoretical Syntax for the Study of Register Variation". MuItilingua 9.157-199. Halverson, J. 1992. "Goody and the Implosion of the Literacy Thesis". Man, 301317. Hanzeli, Victor Egon. 1969. Missionary Linguistics in New France. The Hague: Mouton. Harré, Rom. 1993. Social Being, 2nd ed. Oxford: Blackwell. Harris, J. & G. Lindsay. 1993. "There is No Level of Phonetic Representation". UCL Working Papers in Linguistics 5.355-74. Harris, Roy. 1970. "Semantics and Translation". Actes du Xe congrès international des linguistes, vol. II, ed. by A. Graur, 461-467. Bucharest: Editions de la Académie Socialiste de Roumanie. Reprinted in Love (ed.) 1990.60-65. . 1972. "Words and Word Criteria in French". The History and Structure of French: Essays in Honour of Professor T.B.W. Reid, ed. by F.J. Barnett et al., 117-133. Oxford: Backwell. Reprinted in Love (ed.) 1990.44-59. . 1973a. "Performative Paradigms". Transactions of the Philological Society 1972, 44-58. Oxford: Blackwell. Reprinted in Love (ed.) 1990.66-78. . 1973b. Synonymy and Linguistic Analysis. Oxford: Blackwell. . 1977. "Semantics, Performatives, and Truth". Journal of Literary Semantics 6.63-75. Reprinted in Love (ed.) 1990.79-95. . 1978. Communication and Language. Oxford: Clarendon Press. . 1980. The Language-Makers. London: Duckworth. . 1981a. The Language Myth. London: Duckworth. . 1981b. "Truth-Conditional Semantics and Natural Languages". Language, Meaning and Style: Essays in Memory of Stephen Ullmann, ed. by
REFERENCES
327
T.E. Hope, T.B.W. Reid, R. Harris and G. Price. (eds.), 21-37. Leeds: Leeds University Press. Reprinted in Love (ed.) 1990.96-111. . 1983a. "Language and Speech". Approaches to Language, ed. by Roy Harris, 1-16. Oxford: Pergamon Press. . 1983b. "The Speech-Communication Model in Twentieth-Century Linguistics and its Sources". Proceedings of the Thirteenth International Congress of Linguists, ed. by S. Hattori & K. Inoue, 864-869. Tokyo. Reprinted in Love (ed.) 1990.151-157. . 1983c. "Theoretical Ideas". Times Literary Supplement, 14 October, 119. . 1984a. "Must Monkeys Mean?". The Meaning of Primate Signals, ed. by Rom Harré & V. Reynolds, 116-137, Cambridge: Cambridge University Press. Reprinted in Love (ed.) 1990.158-179. . 1984b. "The Semiology of Textualization". Language Sciences 6.271286. Reprinted in Love (ed.) 1990.210-226. . 1985. "Saussure and the Dynamic Paradigm". Developmental Mechanisms of Language, ed. by Charles-James N. Bailey & Roy Harris, 167184. Oxford: Pergamon Press. . 1986. The Origin of Writing. London: Duckworth. . 1987a. "Language as Social Interaction: Integrationalism versus Segregationalism". Language Sciences 9/2.131-143. Reprinted in Love (ed.) 1990.197-209. . 1987b. The Language Machine. London: Duckworth. . 1987c. Reading Saussure. London: Duckworth. . 1988a. Language, Saussure, and Wittgenstein: How to Play Games with Words. London: Routledge. . 1988b. "Murray, Moore, and the Myth". Linguistic Thought in England 1914-1945, ed. by Roy Harris, 1-26. London: Duckworth. . 1988c. "Preface: Across the No Man's Land". Linguistic Thought in England 1914-1945, ed. by Roy Harris, ix-xi. London: Duckworth. . 1989a. "Degree-0 Explanation". Behavioral and Brain Sciences 12.344345. . 1989b. "How Does Writing Restructure Thought?", Language & Communication 9/2-3.99-106. . 1990a. "Making Sense of Communicative Competence". Love (ed.) 1990.112-135. . 1990b. "On Freedom of Speech". Joseph & Taylor (eds.) 1990.153-161. . 1990c. "On Redefining Linguistics". Davis & Taylor (eds.) 1990.18-52. . 1990d. "On 'Folk' and 'Scientific' Linguistic Beliefs". Meanings and Prototypes, ed. by S.L. Tsohatzidis, 449-64. London: Routledge. . 1990e. "The Scientist as homo loquens". Harré and his Critics, ed. by R. Bhaskar, 64-86. Oxford: Blackwell. . 1990f. "The Dialect Myth". Development and Diversity: Language Variation across Time and Space: A Festschrift for Charles-James N. Bailey, ed.
328
LINGUISTICS INSIDE OUT
by J. Edmondson, C. Feagin & P. Muhlhausler, 3-19. Dallas: Summer Institute of Linguistics. . 1992. "Writing and Pro to-Writing: From Sign to Metasign". New Departures in Linguistics, ed. by G. Wolf, 180-192. New York: Garland Publishing. . 1993a. "Saussure and Linguistic Geography". Language Sciences 15.114. . 1993b. "What is Philosophy of Linguistics?" Linguistics and Philosophy: The Controversial Interface, ed. by Rom Harré & Roy Harris, 3-20. Oxford: Pergamon Press. . 1993c. "Integrational Linguistics". Actes du XVe congrès international des linguistes, ed. by A. Crochetière, J.-C. Boulanger, C. Ouellon, 321-23. SteFoy: Presses de l'Université Laval. . 1993d. "The Integrational Approach to Meaning". Paper delivered at Carnegie Mellon University, Pittsburgh, Pennsylvania, April, 1993. . 1994a. La Sémiologie de l'écriture. Paris: CNRS Editions. . 1994b. "Semiotic Aspects of Writing". Schrift und Schriftlichkeit I Writing and Its Use, Handbjucher zur Sprach- und Kommunikationswissen- schaft Band 10/1, ed. by Gunther H. Halbband & 0 . Ludwig, 41-48. Berlin: Gruyter. . 1995. Signs of Writing. London & New York: Routledge. (ed.). 1983. Approaches to Language. Oxford: Pergamon. & Talbot J. Taylor. 1989. Landmarks in Linguistic Thought: The Western Tradition from Socrates to Saussure. London: Routledge. Harris, W.V. 1989. Ancient Literacy. Cambridge: Cambridge University Press. Harris, Zellig S. 1944. "Simultaneous Components in Phonology". Language 20.181-205. Hedelin, L. & E. Hjelmquist. 1988. "Preschool Children's Mastery of the Form/Content Distinction in Spoken Language". Growing into the Modern World, ed. by K. Ekberg & P. E. Mjaavatn. Trondheim: University of Trondheim, Norwegian Centre for Child Research. Henry, Alison. 1992. "Infinitives in a For-To Dialect". Natural Language and Linguistic Theory 10.279-302. Henson, Hilary. 1974. British Social Anthropologists and Language. Oxford: Clarendon Press. Heritage, John. 1984. Garfinkel and Ethnomethodology. Cambridge: Polity Press. Herskovits, Melville. 1953. Franz Boas. New York: Charles Scribner's Sons. Hewson, John. 1992. Review of Davis & Taylor (eds.) 1990. Canadian Journal of Linguistics 37/3.378-81. Hintzman, Douglas J. 1986. "'Schema Abstraction' in a Multiple-Trace Memory Model". Psychological Review 93.411-428. . 1988. "Judgments of Frequency and Recognition Memory in a MultipleTrace Memory Model". Psychological Review 95.528-551.
REFERENCES
329
Hjelmslev, Louis. 1942. "Langue et parole". Cahiers Ferdinand de Saussure 2.2944 Holiday, Anthony. 1993a. "Teaching to Rule: Wittgenstein and Educational Practice". Perspectives in Education 14/2.157-179. . 1993b. "Conversations in a Colony: Natural Language and Primitive Interchange". Pretexts 4/2.3-19. Hopper, Paul J. 1987. "Emergent Grammar". Berkeley Linguistics Society 13.139157. . 1988. "Emergent Grammar and the Apriori Grammar Postulate". Linguistics in Context: Connecting Observation and Understanding, ed. by Deborah Tannen, 117-134. Norwood, N.J.: Ablex. . 1992. "Times of the Sign: Discourse, Temporality and Recent Linguistics". Time & Society 1.223-238. Hudson, Richard. 1984. Word Grammar. Oxford: Blackwell. . 1990. English Word Grammar. Oxford: Blackwell. Hutton, Christopher. 1989. "The Arbitrary Nature of the Sign". Semiotica 75/12.63-78. . 1990. Abstraction and Instance: The Type-Token Relation in Linguistic Theory. Oxford: Pergamon. Hymes, Dell. 1974. Foundations in Sociolinguistics: An Ethnographic Approach. Philadelphia: University of Pennsylvania Press. Jackendoff, Ray. 1990. Semantic Structures. Cambridge, Mass.: MIT Press. Johns on-Laird, Philip N. 1987. "The Mental Representation of the Meaning of Words". Cognition 25.189-211. Johnston, Paul. 1989. Wittgenstein and Moral Philosophy. London & New York: Routledge. Joseph, John E. 1990. "Ideologizing Saussure: Bloomfield's and Chomsky's Readings of the Cours de linguistique générale". Joseph & Taylor (eds.) 1990.51-78 . 1994. "Saussurean Tradition in Twentieth-Century Linguistics". The Encyclopedia of Language and Linguistics, ed. by R.E. Asher & J.M.Y. Simpson, 7.3664-3669. Oxford: Pergamon. & Talbot J. Taylor (eds.). 1990. Ideologies of Language. London & New York: Routledge. Karanth, P. & Suchitra, M.G. 1993. "Literacy Acquisition and Grammaticality Judgments in Children". Literacy and Language Analysis, ed. by R. J. Scholes, 143-156. Hillsdale, N.J.: Erlbaum. Katz, Jerrold J. 1981. "Literal Meaning and Logical Theory". Journal of Philosophy 76.203-233. . 1966. Philosophy of Language. New York: Harper & Row. & Jerry A. Fodor. 1963. "The Structure of a Semantic Theory". Language 39.170-210. Reprinted in The Structure of Language: Readings in
330
LINGUISTICS INSIDE OUT
the Philosophy of Language, ed. by J. A. Fodor & J. Katz, Englewood Cliffs, N.J.: Prentice-Hall, 1964, 479-518. Kay, Martin, Jean Mark Gawron & Peter Norvig. 1994. Verbmobil: A Translation System for Face-to-Face Dialog. Stanford, California: Center for the Study of Language and Information. Kayne, Richard S. 1984. Connectedness and Binary Branching. Studies in Generative Grammar 16. Dordrecht: Foris. Keil, F.C. 1989. Concepts, Kinds and Cognitive Development. Cambridge, Mass.: Bradford Books/MIT Press. Kempson, Ruth M. 1988. "Logical Form: The Grammar Cognition Interface". Journal of Linguistics 24.393-432. Kindaichi, Haruhiko. 1978. The Japanese Language, transi, and annotated by Umeyo Hirano. Rutland, Vermont: Ch. E. Turtle Co. Kintgen, E. R., B.R. Kroll & M. Rose (eds.). 1988. Perspectives on Literacy. Carbondale: Southern Illinois University Press. Lahiri, Aditi & William Marslen-Wilson. 1991. "The Mental Representation of Lexical Form: A Phonological Approach to the Recognition Lexicon". Cognition 38.245-294. Lakoff, George. 1973. "Fuzzy Grammar and the Performance/Competence Terminology Game". Papers from the Ninth Regional Meeting of the Chicago Linguistic Society, ed. by C. Corum, T.C. Smith-Stark & A. Weiser, 271-91. Chicago. Langacker, Ronald W. 1967. Language and its Structure: Some Fundamental Linguistic Concepts. New York: Harcourt, Brace & World. . 1987. Foundations of Cognitive Grammar. Vol. 1. Stanford: Stanford University Press. . 1990. "A Usage-Based Model". Concept, Image, and Symbol: The Cognitive Basis of Grammar, by R.W. Langacker, 261-288. Berlin: Mouton de Gruyter. . 1991. Foundations of Cognitive Grammar. Vol. 2. Stanford: Stanford University Press. Lee, E., N. Torrance & D.R. Olson. 1994. "The Emerging Concept Of Text: Can Young Children Distinguish the Exact Words from a Paraphrase?". Paper presented at the meeting of the American Educational Research Association, New Orleans, Louisiana. Lee, J. R. E. 1991. "Language and Culture: The Linguistic Analysis of Culture". Ethnomethodology and the Human Sciences, ed. by G. Button, 196-226. Cambridge: Cambridge University Press. Leech, Geoffrey N. 1983. Principles of Pragmatics. London: Longmans. Levelt, Willem J. M. 1989. Speaking: From Intention to Articulation. Cambridge, Mass.: MIT Press. Levinson, S.C. 1983. Pragmatics. Cambridge: Cambridge University Press.
REFERENCES
331
Lévy-Bruhl, L. 1926. How Natives Think. London: George Allen & Unwin. 1st ed. 1910. Liberman, Mark & Alan Prince. 1977. "On Stress and Linguistic Rhythm". Linguistic Inquiry 8.249-336. Lightfoot, David. 1989. "The Child's Trigger Experience: Degree-0 Learnability". Behavioral and Brain Sciences 12.321-375. Lightfoot, David. 1991. How to Set Parameters: Arguments from Language Change. Cambridge, Mass.: MIT Press. Lloyd, G.E.R. 1990. Demystifying Mentalities. Cambridge: Cambridge University Press. Logan, R.K. 1986. The Alphabet Effect. New York: Morrow. Love, Nigel. 1989a. "Transcending Saussure". Poetics Today 10/4.793-818. . 1989b. "Language and the Science of the Impossible". Language & Communication 9/4.269-87. . 1990a. "The Locus of Languages in a Redefined Linguistics". Davis & Taylor (eds.) 1990,.53-117. . 1990b. Editor's Introduction. Love (ed.) 1990.1-21. (ed.). 1990. The Foundations of Linguistic Theory: Selected Writings of Roy Harris. London & New York: Routledge. Lowry, G. 1981. "Foreign Emissaries to the Mughal Court". The India Magazine 2/1.34-9. Lucy, John. 1992. Language Diversity and Thought. Cambridge: Cambridge University Press. Luria, A.R. 1976. Cognitive Development: Its Cultural and Social Foundations. Cambridge, Mass.: Harvard University Press. Lynch, M. 1985. Art and Artifact in Laboratory Science: A Study of Shop Work and Shop Talk in a Research Laboratory. London: Routlege & Kegan Paul. Mackinlay, W.E.W. 1901. "Memorandum on the Languages of the Philippines". Journal of the Royal Anthropological Institute 31.214-218. Mailloux, S. 1989. Rhetorical Power. Ithaca: Cornell University Press. Malcolm, Norman. 1989. "Language Game (2)". Wittgenstein: Attention to Particulars, Essays in Honour of Rush Rhees, ed. by D. Phillips & P. Winch, 3544. London: Macmillan. Malinowski, Bronislaw. 1915. "The Natives of Mailu". Transactions and Proceedings of the Royal Society of South Australia 39.494-706. Adelaide. . 1920. "Classificatory Particles in the Language of Kiriwina". Bulletin of the School of Oriental Studies 1/4.33-78. . 1922. Argonauts of the Western Pacific. London: Routledge & Kegan Paul. . 1923. "Supplement I: The Problem of Meaning in Primitive Languages". The Meaning of Meaning, by C.K. Ogden & LA. Richards, 296-336. New York: Harcourt, Brace & World. . 1935. Coral Gardens and Their Magic, 2 vols. London: Allen & Unwin.
332
LINGUISTICS INSIDE OUT
Marshack, Alexander. 1991. The Roots of Civilization, rev. ed. Mt. Kisco, N.Y.: Moyer Bell. Martinet, André. 1984. "Double Articulation as a Criterion of Linguisticity". Language Sciences 6/1.31-8 Matthei, E. & T. Roeper. 1983. Understanding and Producing Speech. London: Fontana. Matthews, Peter H. 1985. "Whither Linguistic Theory?". Linguistic Abstracts 1/1.1-7 McClelland, J. L. & D. E. Rumelhart (eds.). 1986. Parallel Distributed Processing: Explorations in the Microstructure of Cognition, vol. 2. Cambridge, Mass.: MIT Press. Medawar, P. 1984. The Limits of Science, Oxford: Oxford University Press. Mester, E. Armin. 1994. "The Quantitative Trochee in Latin". Natural Language and Linguistic Theory 12.1-62. Miller, Roy Andrew. 1982. Japan's Modern Myth. New York: Weatherhill. Mo, Timothy. 1986. An Insular Possession. London: Picador. Moerman, Michael. 1988 Talking Culture: Ethnography and Conversation Analysis. Philadelphia: University of Pennsylvania Press. Morais, José. 1985. "Literacy and Awareness of the Units of Speech: Implications for Research on the Units of Perception". Linguistics 23.707-721. , Paul Bertelson, Luz Cary & Jesus Alegria. 1986. "Literacy Training and Speech Segmentation". Cognition 24: 45-64. , Luz Cary, Jesus Alegria & Paul Bertelson. 1979. "Does Awareness of Speech as a Sequence of Phones Arise Spontaneously?". Cognition 7.323-331. Morrison, K. F. 1990. History as a Visual Art in the Twelfth-Century Renaissance. Princeton, N.J.: Princeton University Press. Mothersill, Mary. 1984. Beauty Restored. Oxford: Oxford University Press. Muhlhâusler, Peter & Rom Harré. 1992. Pronouns and People. Oxford: Blackwell. Muller, Friederich Max. 1861. Lectures on the Science of Language. London: Longmans. Nagao, Makoto. 1989. Machine Translation: How Far Can It Go?, transi. Norman D. Cook. Oxford: Oxford University Press. National Research Council. 1965. Language and Machines: Computers in Translation and Linguistics, Report by the Automatic Language Processing Advisory Committee. Washington, D.C.: National Academy of Sciences. Nelson, J.S., A. Megill & D.N. McCloskey. 1987. The Rhetoric of the Human Sciences: Language and Argument in Scholarship and Public Affairs. Madison: University of Wisconsin Press. Newmeyer, Frederick J. 1978. "Prescriptive Grammar: A Reappraisal". Approaches to Language: Anthropological Issues, ed. by William C. McCormack & Stephen A. Wurm, 581-594. The Hague: Mouton (paper presented at the IXth International Congress of Anthropological and Ethnological Sciences, Chicago, October 1973).
REFERENCES
333
. 1983. Grammatical Theory: Its Limits and its Possibilities. Chicago: University of Chicago Press. . 1986. The Politics of Linguistics. Chicago: University of Chicago Press. Nida, Eugene. 1959. "Principles of Translation as Exemplified by Bible Translating". On Translation, ed. by Reuben A. Brower, 11-31. Cambridge, Mass.: Harvard University Press. Niyekawa, Agnes M. 1991. Minimum Essential Politeness: A Guide to the Japanese Honorific Language. Tokyo: Kodansha International. Ochs, E. 1979. "Transcription as Theory". Developmental Pragmatics, ed. by E. Ochs & B.B. Schieffelin, 43-72. New York: Academic Press. Olson, David R. 1977. "From Utterance to Text: The Bias of Language in Speech and Writing". Harvard Educational Review 47/3.257-281. . "Literacy and Objectivity: The Rise of Modern Science". Olson & Torrance (eds.) 1991, 149-64. . 1993. "How Writing Represents Speech", Language and Communication 13.1-18. . 1994. The World on Paper: The Conceptual and Cognitive Implications Of Writing and Reading. Cambridge: Cambridge University Press. & N. Torrance. 1987. "Language, Literacy, and Mental States". Discourse Processes 10/2.157-167. Olson, David R. & Nancy Torrance (eds.). 1991. Literacy and Orality. Cambridge: Cambridge University Press. Orwell, George. 1968. The Collected Essays, Journalism and Letters of George Orwell, ed. by S. Orwell & I. Angus. London: Seeker & Warburg. Parkinson, G.H.R. 1977. "The Translation Theory of Understanding". Communication and Understanding, ed. by Godfrey N.A. Vesey. Hassocks, Sussex: Harvester Press. Pateman, Trevor. 1987a. Language in Mind and Language in Society. Oxford: Oxford University Press. . 1987b. "Philosophy of Linguistics". New Horizons in Linguistics 2, ed. by J. Lyons, R. Coates, M. Deuchar & G. Gazdar, 249-67. London: Penguin. . 1991. "Beauty". Key Concepts: A Guide to Aesthetics, Criticism and the Arts in Education, by Trevor Pateman. London: Falmer Press. Paul, Hermann. 1891. Principles of the History of Language, transi. H.A. Strong. London: Longmans & Green. Pawley, A. & F.H. Syder. 1983. "Two Puzzles for Linguistic Theory: Nativelike Selection and Nativelike Fluency". Language and Communication, ed. by J. C. Richards & R. W. Schmidt, 191-226. London: Longman. Pears, D.F. 1987-88. The False Prison: A Study of the Development of Wittgenstein's Philosophy, 2 vols. Oxford: Oxford University Press. Percival, Keith. 1984. Review of R. Harris 1981a. Unpublished ms., University of Kansas.
334
LINGUISTICS INSIDE OUT
Perelman, C. 1982. The Realm of Rhetoric, transi. W. Kluback. Notre Dame: University of Notre Dame Press. Perlmutter, David M. 1992. "Sonority and Syllable Structure in American Sign Language". Linguistic Inquiry 23.407-442. Pettersson, John Soren. 1991. Critique of Evolutionary Accounts of Writing. Reports from Uppsala University Linguistics, RUUL 21. Uppsala. . 1992. "Rebus Interpretations of Ancient Writing Systems". Paper presented at The 13th Scandinavian Conference of Linguistics held in Roskile, Denmark, January 1992. To appear in the proceedings. . 1994. "Evolutionary Accounts of Writing and the Disobedient History of Scripts". Language & Communication 14.129-153. Pinker, Steven. 1994. The Language Instinct: How the Mind Creates Language. New York: Morrow. Postal, Paul M. & Geoffrey K. Pullum. 1986. "Misgovernment". Linguistic Inquiry 17.104-110. Potter, M.C. & B.A. Faulconer. 1979. "Understanding Noun Phrases". Journal of Verbal Learning and Verbal Behavior 18.509-521. Potter, J. & M. Wetherell. 1987. Discourse and Social Psychology: Beyond Attitudes and Behaviour. London: Sage. Pylyshyn, Zenon W. 1984. Computation and Cognition: Towards a Foundation for Cognitive Science. Cambridge, Mass.: MIT Press. Quirk, Randolph. 1980. Review of R. Harris 1980. The Times, 27 May, 9. Read, Charles, Y. Zhang, H. Nie & Β. Ding. 1986. "The Ability To Manipulate Speech Sounds Depends on Knowing Alphabetic Spelling". Cognition 24.3144. Reed, Lisa. 1992. "Remarks on Word Order in Causative Constructions". Linguistic Inquiry 23.164-172. Reis, Marga & Inger Rosengren. 1992. "What Do WH-imperatives Tell Us About WH-movement?". Natural Language and Linguistic Theory 10.79-118. Roberts, Ian. 1985. "Agreement Parameters and the Development of English Modal Auxiliaries". Natural Language and Linguistic Theory 3.21-58. Roberts, R. H. & J.M.M. Good (eds.) 1993. The Recovery of Rhetoric: Persuasive Discourse and Disciplinarity in the Human Sciences. Charlottesville: University Press of Virginia. Robinson, E., H. Goelman & D.R. Olson. 1983. "Children's Relationship Between Expressions (What Was Said) and Intentions (What Was Meant)". British Journal of Developmental Psychology 1.75-86. Rohr, H.M. 1994. Writing: Its Evolution and Relation to Speech. Bochum: Universitatsverlag Dr. Norbert Brockmeyer. Rumelhart D. E. & J. L. McClelland (eds.). 1986. Parallel Distributed Processing: Explorations in the Micro structure of Cognition, vol. 1. Cambridge, Mass.: MIT Press.
REFERENCES
335
Rundle, Bede. 1990. Wittgenstein and Contemporary Philosophy of Language. Oxford: Blackwell. Russell, Bertrand. 1903. The Principles of Mathematics. London: Allen & Unwin. . 1905. "On Denoting". Logic and Knowledge, ed. by R.C. Marsh, 41-56. London: Allen & Unwin. . 1959. "Mr Strawson on Referring". My Philosophical Development, by B. Russell, 238-245. London: Allen & Unwin. Sacks, H., E.A. Schegloff & G. Jefferson. 1974. "A Simplest Systematics for the Organization of Turn-taking for Conversation". Language 50.696-735. Sampson, G. 1981. Review of R. Harris 1981a. The Times Higher Education Supplement, 16 October, 21. . 1987. Review of Rumelhart & McClelland 1986. Language 63.871-886. Sandmel, Samuel, M. Jack Suggs & Arnold J. Tkacik, O.S.B. (eds.). 1970. The New English Bible, with the Apocrypha: Oxford Study Edition. New York: Oxford University Press. Sapir, Edward. 1963 [1933]. "The Psychological Reality of Phonemes". Selected Writings of Edward Sapir, ed. by David G. Mandelbaum, 46-60. Berkeley: University of California Press. Saussure, Ferdinand de. 1922. Cours de linguistique générale, publié par Charles Bally & Albert Sechehaye, avec la collaboration de Albert Riedlinger, 2 e édition. Engl. transi, by Roy Harris, London: Duckworth, 1983. . 1972. Tullio De Mauro (éd.). Edition critique du Cours de linguistique générale de F. de Saussure. Paris: Payot. Schegloff, E.A. 1979. "Identification and Recognition in Telephone Conversation Openings". Everyday Language: Studies in Ethnomethodology, ed. by G. Psathas, 23-78. New York: Irvington Publishers. . 1991. "Reflections on Talk and Social Structure". Talk and Social Structure: Studies in Ethnomethodology and Conversation Analysis, ed. by D. Boden & D.H. Zimmerman, 44-70. Berkeley: University of California Press. . 1992. "To Searle on Conversation: A Note in Return". (On) Sear le on Conversation, compiled by H. Parret & J. Verschueren, 113-128. Amsterdam: Benjamins. Scholes, R.J. & B.J. Willis. 1991. "Linguists, Literacy, and the Intensionality of Marshall McLuhan's Western Man". Olson & Torrance (eds.) 199L215-235. Scribner, Sylvia & Michael Cole. 1978. "Literacy without Schooling: Testing for Intellectual Effects". Harvard Education Review 48/4.448-461. . 1981. The Psychology of Literacy. Cambridge, Mass.: Harvard University Press. Searle, John R. 1969. Speech Acts: An Essay in the Philosophy of Language. Cambridge: Cambridge University Press. . 1977. "Reiterating the Differences: A Reply to Derrida", Glyph 1.198-208.
336
LINGUISTICS INSIDE OUT
. 1980. "The Background Of Meaning". Speech Act Theory and Pragmatics, ed. by J. R. Searle, F. Kiefer & M. Bierwisch, 221-232. Dordrecht: Reidel. Shapin, S. 1984. "Pump and Circumstance: Robert Boyle's Literary Technology". Social Studies of Science 14.481-520. Shlonsky, Ur. 1992. "Resumptive Pronouns as a Last Resort". Linguistic Inquiry 23.443-468. Shotter, J. 1993. Conversational Realities: Constructing Life Through Language. London: Sage. Silverman, D. 1993. Interpreting Qualitative Data: Methods for Analysing Talk, Text, and Interaction. London: Sage. Smith, E. E. & D. L. Medin. 1981. Categories and Concepts. Cambridge, Mass.: Harvard University Press. Smith, F. 1989. "Overselling Literacy". Phi Delta Kappan 70.352-359. Spengler, Oswald. 1961. The Decline of the West, abridged ed. by Herbert Werner & Arthur Helps, transi. C.F. Atkinson. London: George Allen & Unwin. Sperber, Dan & Deirdre Wilson. 1986. Relevance: Communication and Cognition. Oxford: Blackwell. Starosta, S. 1988. The Case for Lexicase: An Outline of Lexicase Grammatical Theory. London: Pinter. Street, B. 1984. Literacy in Theory and Practice. Cambridge: Cambridge University Press. Suchman, L. 1987. Plans and Situated Actions: The Problem of Human-Machine Communication. Cambridge: Cambridge University Press. Suner, Margarita. 1992. "Subject Clitics in the Northern Italian Vernaculars and the Matching Hypothesis". Natural Language and Linguistic Theory 10.641672. Taylor, Talbot J. 1990. "Normativity and Linguistic Form". Davis & Taylor (eds.) 1990.118-48. . 1992. Mutual Misunderstanding: Scepticism and the Theorizing of Language and Interpretation. Durham: Duke University Press. & D. Cameron. 1987. Analyzing Conversation: Rules and Units in the Structure of Talk. Oxford: Pergamon. Taylor, John R. 1992. "How Many Meanings Does a Word Have?". SPIL (= Stellenbosch Papers in Linguistics) 25.133-168. . 1994. "The Two-Level Model of Meaning". Linguistische Behchte 149.3-26. Thomason, R. H. 1974. "Introduction". Formal Philosophy: Selected Papers of Richard Montague, 1-69. New Haven: Yale University Press. Toolan, Michael. 1990. "Largely For Against Theory". Journal of Literary Semantics 19/3.150-166. Tsohatzidis, Savas L. 1995. "What Lack Needs to Have: A Study in the Cognitive Semantics of Privation". Language and the Cognitive Construal of the World,
REFERENCES
337
ed. by John R. Taylor & Robert E. MacLaury, 81-93. Berlin: Mouton de Gruyter. Vygotsky, L. 1963. Thought and Language. Cambridge, Mass.: MIT Press. Waldron, T.P. 1980. Review of R. Harris 1980. Times Literary Supplement, 11 July, 785. Watson, John B. 1924. Behaviorism. New York: W.W. Norton & Co. Whatmough, Joshua. 1957. Language: A Modern Synthesis. New York: Mentor. White, Lydia. 1985. "Universal Grammar as a Source of Explanation in Second Language Acquisition". Current Approaches to Second Language Acquisition: Proceedings of the 1984 University of Wisconsin-Milwaukee Linguistics Symposium, ed. by Barbara Wheatley, Ashley Hastings, Fred R. Eckman, et al, 43-69. Bloomington: Indiana University Linguistics Club. White, J. B. 1985. Heracles' Bow: Essays on the Rhetoric and Poetics of Law. Madison: University of Wisconsin Press. Whorf, Benjamin Lee. 1956. Language, Thought and Reality. Cambridge, Mass.: MIT Press. Wierzbicka, Anna. 1992. Semantics, Culture, and Cognition. Oxford: Oxford University Press. Wilks, Yorick. 1987. "Bad Metaphors: Chomsky and Artificial Intelligence". Noam Chomsky: Consensus and Controversy, ed. by S. Modgil & C. Modgil, 197-206. Lewes: Falmer Press. Wittgenstein, Ludwig. 1922. Tractatus Logico-Philosophicus, transi. C.K. Ogden. London: Routledge & Kegan Paul. 2nd ed. transi. D.F. Pears & B.F. McGuinness, 1961 .1929. "Some Remarks on Logical Form". Proceedings of the Aristotelian Society, supp. vol. 9.162-171. . 1953. Philosophical Investigations. Oxford: Blackwell. . 1958. The Blue and Brown Books: Preliminary Studies for the Philosophical Investigations. Oxford: Basil Blackwell. 2nd ed., 1969. . 1953. Philosophical Investigations, transi. G.E.M. Anscombe. Oxford: Basil Blackwell. 2nd ed., 1958. . 1967. Zettel, ed. G.E.M. Anscombe & G.H. von Wright, transi. G.E.M. Anscombe. Oxford: Basil Blackwell. 2nd ed., 1981. . 1980. Culture and Value, ed. G.H. von Wright & H. Nyman, transi. P. Winch. Oxford: Basil Blackwell. Wolf, George. 1989. "Malinowski's 'Context of Situation'". Language & Communication 9/4.259-67'. Woods, Α., Ρ. Fletcher & A. Hughes. 1986. Statistics in Language Studies. Cambridge: Cambridge University Press. Wundt, Wilhelm. 1916. Elements of Folk Psychology , transi. Ε. L. Schaub. New York: Macmillan. Zimmerman, D. H. & D. Boden, D. 1991. "Structure-in-Action: An Introduction". Talk and Social Structure: Studies in Ethnomethodology and Conversation
338
LINGUISTICS INSIDE OUT
Analysis, ed. by D. Boden & D.H. Zimmerman, 3-21. Berkeley: University of California Press. & R. Schreuder. 1992. "From Concepts to Lexical Items". Cognition 42.23-60.
Index alphabet, alphabetic writing 6, 45, 52-6, 84, 87, 91, 97-8 analogy 216-7, 222 Anscombe, G.E.M. 108n., 255n. anthropology 99, 101, 132, 152, 155-7, 160, 167-8 arbitrariness of the linguistic sign (Saussure) 34, 288 Aristotle 21, 22, 39, 43, 125, 132, 288; and: metaphor 105;psychocentric surrogationalism 15, 39; telementation 4, 23-4, 35, 243, 246-9 art: see language Augustine 16-18, 112, 125, 126-7, 148 Auroux, S. 264-5, 299, 301 Austin, J.L. 100, 148, 158, 185, 292 Bacon, F. 2 , 4 , 10,21 Baker, G.P. 110, 112 Bar-Hillel, Y. 296-8 Baron, N.S. 7, 157, 158-9, 160, 161, 172, 183a, 283-4, 313 beauty 226-7, 279, 282 behaviourism 19,216,266-7,313 Bennett, J. 255-6 Bernstein, B. 163-4 Bhaskar, R. 60, 273n. Bierwisch, M. 214,220 Billig, M. 190, 192, 205 Bloomfield, L. 2, 21-2, 34-5, 62, 89, 139, 156, 209, 227; and: analogy 216; antimentalism 34-5; behaviourism 4, 17-19, 266-7; empiricism 62; linguistic
meaning 276-7, 314; reocentric surrogationalism 18; Saussure 241 Boas, F. 155-6 Boden, D. 186, 187, 193, 195, 200, 202 Borsley, R.D. 5, 6 4 a , 67, 77-8, 82, 229a, 230, 249, 251, 253-4, 264, 265-6, 269, 272-3, 277, 288, 300, 302-3, 312, 313, 315,316,317,318 Bromberger, S. 65, 71-6, 78-81 Burton-Roberts, N.C. 65-6, 68, 83a Button, G. 190, 193 Cameron, D. 183a, 185a, 194, 199 Carr, P. 5-6, 65-6, 68, 8 3 a , 251-2, 270, 313,315-6 Champollion, F. 88-9 Chomsky, A.N. 51-2, 60, 63, 71, 80, 82, 83, 137, 139, 184, 197, 200, 217, 226, 266; and: analogy 216; art 280; communication 46-7; creativity 279-80; deep structure 272; fixed codes 48-9; Ilanguage 5, 47, 48, 50, 65-6, 68-9, 70, 78, 81, 252, 313; the ideal speakerlistener 61, 189, 313; linguistic behaviour 77, 269; linguistic knowledge 69; linguistic performance 265; linguistic rules 62, 146, 282; Logical Form 253; the sociopolitical dimension of language 50; telementation 317; the U.S. government 229a; writing 260 cognition 48, 67, 69, 70, 79, 95, 99-100, 146, 164, 182, 187, 215, 261-2 cognitive grammar (Langacker) 215, 222
340
LINGUISTICS INSIDE OUT
Cole, M. 104-5, 161 communication 11 et passim; as telepathy 66-8; theories of 26, 28, 30, 31, 33-5, 45, 247, 314-7; see also telementation computational linguistics 51 computers 44, 149, 159, 172-4, 212 Constantin, E. 28, 29, 33 context 7, 151-69, 171, 175-6, 179-80, 181, 184, 196, 197, 201-3, 205, 228, 261, 281-8, 294 contractualism 2, 12 conversation analysis 7, 62, 152, 182-207 passim, 296, 304; see also ethnomethodology cotemporality, principle of (Harris) 201, 203, 205, 295-6, 298-9 creativity 227, 279-82 cricket 136-7, 146, 226, 282, 27, 290 Croce, B. 226,281 Crystal D. 259n., 267n. Davis, H.G. 6 4 a , 266n. Derrida, J. 292-5 Descartes, R. 22, 125, 131 dialect 238, 308 Diamond, C. 110, 112 dictionary 19-20, 43, 100, 159-60, 164, 184, 197, 238, 245, 253, 258, 267, 275, 290 doublethink (Orwell) 256 Drew, P. 186, 191, 197, 202 Dummett, M. 244-6, 258 Duranti, A. 62, 152 Durkheim, E. 187, 189, 198 empiricism 5, 49, 59-62, 66, 82, 117-8, 265,272-3, 313 ethnomethodology 7, 152, 182, 185-96, 198-90, 201-4, 207, 296; see also conversation analysis
Firth, J.R. 166-7 fixed codes, languages as 2-4, 5, 11, 22, 43, 45, 47-51, 65, 184, 223-5, 253-4, 258-9, 275, 287, 311, 315; see also language myth Fleming, D. 7, 296, 304, 312 Fodor, J.A. 43, 148-50, 215, 250, 253 folk linguistics 302 Frege, G. 110a, 112-6, 123, 244-5 Gardner, H. 44, 235, 301 Garfinkel, H. 186, 189, 191, 196, 199, 202 Gawron, J.M. 152, 172-4 Gazdar, G 232n., 289n. Gelb, I.J. 84-91, 96, 97 generative grammar, generativism 5-6, 4264, 65-83, 216, 234, 249-54, 260, 265-6, 277, 301-3, 309, 315-6 Genesis 13-15 grammar, grammatical theory 3, 6-7, 43, 47, 48, 117, 136-7, 146, 156, 159, 167, 168, 169-72, 174, 176, 178, 179, 234, 236, 238, 251, 279, 302, 309; see also generative grammar grammar (Wittgenstein) 124, 140-2,278 Hacker, P.M.S. 106a, 110, 111n., 112, 123a, 270 Halle, M. 65,70,72-6, 78-81 Harré, R. 7, 136, 277-9 Harris, R. 1 et passim; and: art 227-9; conversation analysis 182-207; generative grammar 42-64, 65-83; the language myth 9-41; Malinowski 15181; Wittgenstein 106-35,136-50; writing 84-98, 99-105; on: context 28295; cotemporality 295-300; fixed codes 253-4; language, art and creativity 27982; linguistic rules 277-9; metalanguage 258-60; science, language and linguistics
INDEX 273-7; telementation 243-52; verbal behaviour 264-73; writing 260-4 Hegel, F. von 99, 226 Henson,H. 156-7 Heraclitus 288, 297 Heritage, J. 186-7, 189, 191, 192, 193, 196, 197, 202 Hewson, J. 266 Hilbert,D. 138-9 Hintzman, D. 208-12, 214-5 Holiday, A. 6, 106a, 111n., 127n., 263, 273-6, 281 Hopper, P. 63, 182n., 200 Hudson, R.A. 217,218,219 Hutton, C. 288-9,296-8 Hymes, D. 62, 233n., 302 I-language (Chomsky) 47-8, 50-1, 63, 6770, 74, 78, 81-2, 251-2 idola fori (Bacon) 2, 4, 10, 12, 21 instrumentalism 2, 12 integrational linguistics, integration(al)ism (Harris) 7, 10-11, 22, 45, 52, 62, 64, 98, 108, 154, 180, 182-3, 184-5, 188, 195, 198, 199, 203, 204-7, 229-310 passim, 312-5 intentionalist semantics 248-9 Italian 57-8 iterability 289-90, 292-5 Japanese 151,168-72,177-9 Jefferson, G. 189, 191, 192, 194, 197-8, 203 jokes 106-7,133-4 Joseph, J.E. 4, 32, 35, 64, 246-9, 263, 277, 303, 311, 312, 315, 317, 318 Kant, I. 99, 109, 123, 125, 137, 226, 279, 282, 300 Katz, J.J. 43, 46, 220, 249
341 Kay, M. 152,172-4 Lahiri, A. 56, 59 Langacker, R.W. 215-6,219,221-2 language: acquisition (and/or learning) of 3, 16-17, 71, 112, 159, 209, 217-8, 257-9; as action or activity 18, 147, 152, 154, 166, 182, 184, 186, 201; and art 8, 2268, 279-82; and memory 307-8; and time 200-3, 295-300 language change 59, 142-4, 236, 238 language game (Wittgenstein) 109, 124, 126-8, 129, 132, 134, 142-4, 185 language machine (Harris) 3,12, 44, 293, 315 language myth (Harris) 2, 4, 9-41, 42-5, 51, 65, 233-6, 248, 278, 315; see also fixed codes; telementation language orienteering (Baron) 158-60, 163, 180 langue (Saussure) 27, 28, 32-6, 140, 235, 237, 241, 252, 253, 256, 293 Lee, J.R.E. 187, 188, 193 Leibniz, G.W. 13, 114 Levinson, S.C. 177, 186n., 189, 192, 193, 194, 196, 202 Lévy-Bruhl, L. 99,262-3 lexicalism 219 Lightfoot, D. 54n., 59, 265 linguistic behaviour 22, 60, 66-8, 70, 72-3, 77-8, 80, 82-3, 182-203, 264-73 linguistic competence 60, 70, 185, 197, 218, 235, 252, 265-6 linguistic performance 60, 65, 66, 68, 76, 235, 252, 265-6; see also linguistic behaviour linguistics: and freedom of speech 63, 228n.; history of 9-41, 51-2, 299; Indian 53; as an industry 9-10, 303; orthodox 182, 183, 190, 207, 230, 233-5, 238-40,
342
LINGUISTICS INSIDE OUT
241, 258, 261, 264, 272, 288-92, 298, 301, 303-4, 309, 312-4, 317; as a science 1, 40-1, 43, 60, 214, 231, 232, 242, 259, 273, 277, 298; Western tradition of 21, 249, 251, 287, 307 literacy 100-2, 104-5, 155, 161-2, 182, 264 Locke, J. 13-14, 15, 22-24, 27-8, 30-7, 43, 148, 243, 244, 251, 287 logic 105, 112-20, 122-4, 136, 139, 187,
names 13-16, 116-24, 142; see also nomenclaturism Neogrammarians 243, 261 Newmeyer, F.J. 4-5, 4 6 a , 51, 5 9 a , 6 1 a , 67, 77-8, 82, 229a, 230, 249, 251, 2534, 264, 265-6, 269, 272, 277, 280, 288, 300, 302-3, 312, 313, 315, 316, 317, 318 nomenclaturism 4, 13-22, 112, 127, 318;
247, 258, 262 logical form (Wittgenstein) 111, 123-4 Logical Form (Chomsky) 48-9, 69, 253 Love, N.L. 11, 25, 41, 6 4 a , 154, 157, 163, 179, 183a, 185, 188, 229a, 230a, 231a, 290a, 291a, 303, 305a, 310, 318 Luria, A.R. 104, 146 Lynch, M. 186a, 189, 193, 203
natural vs non-natural H;see also surrogationalism Norvig, P. 152, 172-4
machine translation 152, 159, 172-6, 17980 Malinowski, B. 152-7, 161-7, 171, 175-6, 179-80, 283-4 Matthews, P.H. 228, 233a, 236, 301 meaning 19-21, 49, 103, 121-2, 140, 148, 151-69, 171-2, 179, 180, 184, 188, 196, 220, 224, 241, 245, 247, 249, 254, 263, 267 Medin, D.L. 209-10 memory: multiple-trace model of 8, 208-25; primary 210, 211, 213-4; secondary 210, 211, 213-4, 217, 224; see also language mentalism 34-5, 215; see also surrogationalism, psychocentric metalanguage, metalinguistics 6, 129, 160, 258-10, 279,291, 295, 310 minimalism (Chomsky) 65-8 Moerman, M. 186a, 191 Muhlhâusler, P. 183a, 308a
Olson, D.R. 6, 55, 100, 103, 161, 259, 261-2 ordinary-language philosophy 117-8, 245 orthography 55-6 Oxford University, linguistic studies at 52, 230-33 Pânini 53, 264 paralanguage 267 parole (Saussure) 27, 28, 32-4, 35, 140, 235, 237, 241, 252, 253, 256, 293 Parsons, T. 186, 187, 198 Pateman, T. 8, 4 8 a , 6 0 a , 279, 281-2, 289, 291, 293 Paul, H. 243a, 245-6, 261 Pawley, A. 219 Pears, D. 106a, 108a, 111 Percival, K. 44a, 60, 64 Pettersson, J.S. 6, 89, 95, 96, 261, 263 phoneme, phonemic theory 52, 55-6, 269, 310 Phonetic Form (Chomsky) 48, 65-6, 68-70, 72, 75-6, 78, 82-3, 253 phonology, phonological theory 53-4, 56, 70, 72-5, 79-81, 213, 222, 223, 275, 301 phonography 85, 87, 89, 90, 97; see also writing Pinker, S. 250-1, 253-4, 272a
INDEX Plato, Platonism 15-17, 20, 23, 33, 39, 72, 114, 123, 125, 132, 247, 256, 291, 303 positivism 118, 304 postmodernism 21 prescriptivism 43, 45, 56-9, 61 psycholinguistics 56, 218, 223 psychological reality 45,215 Pylyshyn, Z.W. 61,215 rhetoric, integrational (Fleming) 205-7 rhetorical constructionism 40 Roberts, I. 58a, 59 Roche, J.J. 137-8 Rohr, H.M. 84, 94 rules 7-8, 43, 58, 61-2, 77, 136-7, 140-7, 149, 183, 186, 196, 197, 215-9, 223-4, 226-8, 265-6, 274, 277-80 Russell, B. 112-24, 133 Sacks, H. 189, 191, 192, 194, 197, 198-9, 202, 303 Sampson, G. 215, 234 sandwich organ 280 Sapir, E. 55, 156 Saussure, F. de 2, 15, 21-2, 43, 106, 137, 189, 198, 201, 209, 216, 227, 230, 2401, 243, 246, 248-9, 251, 256-7, 261, 263, 275-5, 289, 291, 299, 301; and: historical linguistics 190, 202; the history of linguistics 299; intentionalist semantics 248-9; the language myth 434; the langue/parole distinction 140, 253; the linguistic sign 259-60, 263, 294; nomenclaturism 13; orthodox linguistics 234; the scientific status of linguistics 274-6; segregationism 184; semiology 314; structuralism 32-4, 36, 51-2, 109, 291; telementation 4, 24-38, 243, 246, 249,251-2,315,316-7 saying vs showing (Wittgenstein) 109-10,
343 112-4, 126,263,276 scales of integrational linguistic inquiry (Harris): biomechanical 275, 305, 306, 307-8, 312; macrosocial 305, 306, 308, 312; circumstantial 305,306,312 Schegloff, Ε. 186η., 189, 191, 192, 194, 197-8, 202, 203 science: language of 259, 273, 275-6; and philosophy 6-7, 106-35, 273-7; see also linguistics Scribner, S. 104, 161 scriptism 45, 232 Searle, J.R. 158, 220, 293-4 segregational linguistics, segregation(al)ism 10-12, 22, 41, 184-5, 188, 195, 201, 242-3, 259, 283, 284, 308, 309 semiology 260, 275, 292, 304, 311, 313-5; of written texts 63, 302 sentence (vs utterance) meaning 49, 253 sign 6, 23-4, 27, 36, 85, 89-94, 98, 102, 183-4, 187, 190, 224, 260, 267, 271, 294,296,312-5 Smith, E.E. 209-10 Socrates 16, 107, 303 sociolinguistics 185, 223, 302, 308 sociology 32, 185-7, 189, 191, 193, 195, 201, 235, 296 speech, primacy of 260, 264, 268 speech circuit (Saussure) 24-30, 32-4, 36, 248-9, 252 Spengler, Ο. 124 Sperber, D. 45,47 structuralism 34-37, 44, 51-2, 54, 184, 187, 199, 200, 201, 207, 236, 260, 304, 309 surrogationalism 2, 4, 12-22, 23, 148, 276, 279, 318; psychocentric 4, 13, 14, 15; reocentric 14,15,17-18; see also nomenclaturism Syder, F.H. 219
344
LINGUISTICS INSIDE OUT
synchronic linguistics, synchrony 1, 29, 35, 50, 183, 201, 235, 241, 301, 313 syntactic theory, syntax 54, 56, 70, 75-6, 117, 122, 146, 174, 200, 217, 223, 295 Taylor, J.R. 7, 213, 307-8 Taylor, T.J. 13, 15, 21, 23, 25, 35, 4 2 a , 64n., 182n., 183a, 186a, 189, 194, 199, 266a telementation 2, 4, 5, 11-12, 22-38, 43, 457, 48-50, 65-8, 71-2, 83, 153, 184, 224, 229, 243-52, 284, 287; see also communication, theories of tennis 285-6 token (vs type) 70, 71, 139, 293-4, 296-8 Toolan, M. 292,294,299 Torrance, N. 103,262a truth-conditional semantics 49a Tsohatzidis, S.L. 212, 246a type (vs token) 71, 139-40, 293-4, 296, 298 universal grammar 47, 68, 6 9 a Vygotsky, L. 104, 149
Watson, J.B. 18-19 Wilson, D. 45,47 Winch, P. 108a, 134a Wittgenstein, L. 3, 6-7, 20, 21-2, 106-50 passim, 136-7, 139, 255, 295, 318; and: denotation 121; grammar 140-2, 187; language change 142-4; language games 109, 124, 126, 128-9, 150, 185; the language of science 273-6; logic 114, 122, 187; logical form 111-2,124; nomenclaturism 127; philosophy vs science 106-35; surrogationalism 13, 16, 18 Wolf, G 229a, 273a, 283-4 writing 2, 84-98, 99-105, 161-3, 165, 175, 180, 193, 230, 260, 261-4, 267-9, 291-3, 296, 309, 315; history (evolution) of 84, 88-90, 95-6; origin of 92-5, 264; as representation of speech 2, 91, 97, 264; see also phonography Zimmerman, D.H. 186, 187, 193, 195, 200, 202