Integrational Linguistics: A First Reader
Related Pergamon books FIGUEROA HARRE & HARRIS HARRIS KOERNER & ASHER KOMAT...
31 downloads
811 Views
24MB 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
Integrational Linguistics: A First Reader
Related Pergamon books FIGUEROA HARRE & HARRIS HARRIS KOERNER & ASHER KOMATSU & HARRIS KOMATSU & WOLF KOMATSU & WOLF
Sociolinguistic Metatheory Linguistics and Philosophy Introduction to Integrational Linguistics Concise History of the Language Sciences Saussure's Third Course of Lectures on General Linguistics (1910-1911) Saussure's Second Course of Lectures on General Linguistics (1908-1909) Saussure's First Course of Lectures on General Linguistics (1907)
Related journals Journal of Pragmatics Editor. Jacob Mey Language & Communication Editors: Roy Harris and Talbot J. Taylor Language Sciences Editor. Nigel Love Lingua Editors: Teun Hoekstra, John Anderson and Neil Smith Free specimen copies of journals available on request
Language & Communication Library This series features books and edited collections of papers which focus attention on key theoretical issues concerning language and other forms of communication. The history of these issues and their cultural implications also fall within its scope, as do radical proposals for new approaches in linguistics and communication studies. The series has already published important contributions to critical debate both by younger writers and by established scholars, and welcomes more. Inquiries from potential contributors should be addressed to the Series Editor, Roy Harris.
Integrational Linguistics A First Reader Edited by ROY HARRIS Oxford, England and
GEORGE WOLF New Orleans, Louisiana, USA
Pergamon An imprint of Elsevier Science
ELSEVIER SCIENCE Ltd The Boulevard, Langford Lane Kidlington, Oxford OX5 1GB, UK
Library of Congress Cataloglng-ln-Publ1cat Ion Data Integratlonal linguistics : a first reader / edited by Roy Harris and George Wolf. — 1st ed. p. cm. — (Language and communication library series) A companion volume to Introduction to Integratlonal 1ingulsties. Includes Index. ISBN 0-08-043365-0 (hardcover :• alk. paper). — ISBN 0-08-043366-9 (flexlcover : alk. paper) 1. Linguistics. I. Harris, Roy, 1931II. Wolf, George, 1950- . III. Series: Language & communication library. P121.I53 1998 410~dc21
98-34396 CIP
British Library Cataloguing in Publication Data A catalogue record from the British Library has been applied for. First edition 1998 ISBN: 0 08 043365 0 (Hardbound) ISBN: 0 08 043366 9 (Paperback)
© 1998 Elsevier Science Ltd All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means: electronic, electrostatic, magnetic tape, mechanical photocopying, recording or otherwise, without permission in writing from the publishers. ©The paper used in this publication meets the requirements of ANSI/NISOZ39.48-1992 (Permanence of Paper). Printed in The Netherlands.
Contents Chapter sources Introduction
vii 1
PART ONE: LANGUAGE AND COMMUNICATION 1. Language as Social Interaction: Integrationalism versus Segregationalism
5
ROY HARRIS
2. The Integrationist Critique of Orthodox Linguistics
15
ROY HARRIS
3. Making Sense of Communicative Competence
27
ROY HARRIS
PART TWO: LANGUAGE AND THE LANGUAGE MYTH 4. The Fixed-Code Theory
49
NIGEL LOVE
5. A Few Words on Telementation
68
MICHAEL TOOLAN
6. The Dialect Myth
83
ROY HARRIS
7. Integrating Languages
96
NIGEL LOVE
PART THREE: LANGUAGE AND MEANING 8. Three Models of Signification
113
ROY HARRIS
9. Meaning and the Principle of Linearity
126
CHRISTOPHER HUTTON
10. On Inscribed or Literal Meaning
143
MICHAEL TOOLAN
11. Irony and Theories of Meaning STEVE FARROW
159
vi
Contents PART FOUR: LANGUAGE AND DISCOURSE
12. Conversational Utterances and Sentences
171
TALBOT J.TAYLOR
13. Do You Understand? Criteria of Understanding in Verbal Interaction
198
TALBOT J.TAYLOR
14. Analysing Fictional Dialogue
209
MICHAEL TOOLAN
PART FIVE: LANGUAGE AND WRITING 15. The Semiology of Textualization
227
ROY HARRIS
16. Analysis and Notation: The Case for a Non-Realist Linguistics
241
CHRISTOPHER HUTTON
17. How Does Writing Restructure Thought?
252
ROY HARRIS
18. Writing and Proto-Writing: From Sign to Metasign
261
ROY HARRIS
PART SIX: LANGUAGE AND SOCIETY 19. What Has Gender Got To Do With Sex?
273
DEBBIE CAMERON
20. What Makes Bad Language Bad?
283
HAYLEY DAVIS
21. Law Lessons for Linguists? Accountability and Acts of Professional Classification
294
CHRISTOPHER HUTTON
22. Teaching American English as a Foreign Language: An Integrationist Approach
305
DANIEL R. DAVIS
23. What Problems? On Learning to Translate
313
MARSHALL MORRIS
24. Pronouncing French Names in New Orleans
324
GEORGE WOLF, MICHELE BOCQUILLON, DEBBIE DE LA HOUSSAYE, PHYLLIS KRZYZEK, CLIFTON MEYNARD, LISBETH PHILIP
Index
343
Chapter sources 1. Language Sciences, vol. 9, no. 2, 1987, pp. 131-43. 2. The Sixteenth LACUS Forum 1989, ed. by M.P. Jordan, Linguistic Association of Canada and the United States, Lake Bluff, 111., 1990, pp. 63-77. 3. The Foundations of Linguistic Theory: Selected Writings of Roy Harris, ed. by N. Love, Routledge, London & New York, 1990, pp. 112-35. 4. Language & Communication, vol. 5, no. 1, 1985, pp. 1-17. 5. Language Sciences, vol. 19, no. 1, 1997, pp. 79-91. 6. Development and Diversity: Linguistic Variation across Time and Space: A Festschrift for Charles-James N. Bailey, ed. by J.A. Edmondson, C. Feagin & P. Muhlhausler, Summer Institute of Linguistics, Dallas, 1990, pp. 3-19. 7. Adapted from pp. 95-114 of 'The locus of languages in a redefined linguistics', in Redefining Linguistics, ed. by H.G. Davis and T.J. Taylor, Routledge, London, 1990, pp. 53-117. 8. Structures of Signification, ed. by H.S. Gill, Wiley, New Delhi, vol. 3, 1993, pp. 66577. 9. Language & Communication, vol. 10, no. 3, 1990, pp. 169-83 10. Adapted from M. Toolan, Total Speech: An Integrational Linguistic Approach to Language, Duke University Press, Durham, 1996, pp. 24-55. 11. New Departures in Linguistics, ed. by G. Wolf, Garland Publishing, New York & London, 1992, pp. 135-45. Reprinted by permission of Garland Publishing. 12. Language & Communication, vol. 4, no. 2, 1984, pp. 105-127. 13. T.J. Taylor, Theorizing Language, Pergamon, Oxford, 1997, pp. 79-92, and Language & Communication, vol. 6, no. 3, 1986, pp. 171-80. 14. Language & Communication, vol. 5, no. 3, 1985, pp. 193-206. 15. Language Sciences, vol. 6, no. 2, 1984, pp. 271-86. 16. Linguistics and Philosophy, ed. by R. Harre and R. Harris, Pergamon Press, Oxford, 1993, pp. 165-78. 17. Language & Communication, vol. 9, no. 2/3, 1989, pp. 99-106. 18. New Departures in Linguistics, ed. by G. Wolf, Garland Publishing, New York & London, 1992, pp. 180-92. Reprinted by permission of Garland Publishing. 19. Language & Communication, vol. 5, no. 1, 1985, pp. 19-28. 20. Language & Communication, vol. 9, no. 1, 1989, pp. 1-9. 21. Language & Communication, vol. 16, no. 3, 1996, pp. 205-14. 22. Who Owns English?, ed. by M. Hayhoe & S. Parker, Open University Press, Buckingham & Philadelphia, 1994, pp. 68-76. 23. New Departures in Linguistics, ed. by G. Wolf, Garland Publishing, New York & London, 1992, pp. 199-212. Reprinted by permission of Garland Publishing. 24. Language in Society, vol. 25, no. 3 (September), 1996, pp. 407-26. Reprinted by permission of Cambridge University Press.
VII
This page intentionally left blank
Introduction Integrationism has emerged in the past few years as posing a radical challenge to the orthodoxies of structuralism and generativism that have dominated academic linguistics in recent decades. In compiling this selection of texts we have had two main purposes in view. One was to bring together a representative range of papers published over the past fifteen years by scholars exploring an integrational approach to a variety of topics concerning language. Such a collection fills a lacuna in the documented history of twentiethcentury inquiry into language. Most of these papers were originally published independently in academic journals and some as contributions to volumes no longer in print. We invited the authors to revise them where they considered that useful or necessary. The other purpose we had in mind was to provide a companion volume of 'further reading' to Introduction to Integrational Linguistics. These twin publications, the Introduction and the Reader, are Vol. 17 and Vol. 18 in the Pergamon Language & Communication Library series. In our view, which might not necessarily be shared by all integrationists, the originality of the integrational approach can best be summarized under the following six heads: (i) integrational assumptions about the relative priority of the notions 'language' and 'communication', (ii) the integrational critique of the traditional view of what 'a language' is, (iii) the integrational analysis of linguistic 'meaning', (iv) the integrational view of the relation between 'a language' and its 'use', (v) the integrational account of how speech relates to writing, and (vi) the integrational conception of a linguistic community and what membership of such a community may entail. This view of integrationism is reflected in our parallel organization of the two books, Chapters 1 to 6 of the Introduction corresponding to Parts 1 to 6 of the Reader. In our editorial comments on the individual papers, we have not attempted to give a detailed synopsis of each, but to bring out certain thematic links between them. The term integrationism itself does not feature in a number of the contributions we have included, and therefore a brief comment on how the editors understand it seems appropriate here. (For a fuller discussion, the reader is referred to Introduction to Integrational Linguistics.) We take the term integrationism to allude to a recognition that what makes an utterance (or any other form of expression) language is not its conformity to the requirements of a code but its function in integrating other human activities, that integration being what makes communication between one human
2
Introduction
being and another possible. The same applies, furthermore, not only to language but to all modes of human communication. For, contrary to what is commonly assumed in orthodox linguistics, there is no sharp dividing line separating language from other modes of communication, or linguistic behaviour from non-linguistic behaviour. For human beings, a sign is a sign because it has an integrational function in the particular circumstances in which it occurs, and when voluntarily produced by human agency its production is always a creative act on the part of one or more individuals acting in a certain situation. Whatever we recognize as a linguistic sign (by whatever criteria seem appropriate to the occasion) is always a non-linguistic sign as well. The two are never mutually exclusive. Human beings do not inhabit a communicational space which is compartmentalized into language and non-language, but an integrated space where all signs are interconnected. From an integrational perspective, one of the major shortcomings in modern linguistics has been its failure to recognize the integrated character of human communication and its consequent attempt to place language, as an object of academic research, in a self-contained category of its own. This segregation has even been seen as an essential prerequisite for the establishment of linguistics as an autonomous discipline. An unfortunate result is that attempts by linguists to explain how language 'works' in isolation from other signs have produced forms of linguistic communication which, apart from being virtually incomprehensible to everyone except linguists themselves, remove the subject far from the everyday linguistic experience of the lay person. Thus, paradoxically, what is described and discussed by linguists gets further and further removed from language as non-linguists understand it. This situation is itself part of the problem of linguistic inquiry which integrationists look to address in their rejection of the priorities on which orthodox linguistics is based.
R.H. G.W.
PART ONE LANGUAGE AND COMMUNICATION The three papers in this section are about three distinct but interconnected aspects of the relationship between the notions of 'language' and of 'communication'. They may be seen as attempting to clarify possible obscurities or ambiguities in the expression linguistic communication. What requires clarification includes: (i) whether language is plausibly seen as a form of human behaviour that can be analysed at a level of abstraction where no reference is made to its communicational function in particular situations, (ii) whether or not we can give an adequate answer to the question 'What is language?' without also giving an adequate answer to the question 'What is communication?'; and (Hi) to what extent our communicational abilities in practice are seen as depending on our linguistic abilities, or vice versa. Ch. 1 ('Language as Social Interaction: Integrationalism versus Segregationalism') argues that orthodox modern linguistics rests on an unjustified theoretical assumption: namely, that linguistic analysis can be conducted independently of any detailed inquiry about how the language under analysis is actually revealed in the communicational activities of individuals. The orthodox analyst takes it for granted that communication is an unproblematic concept and that whatever needs to be known about language-users and their communicational practices is already sufficiently well-known. Therefore the task facing the linguist is of a different order, consisting essentially in identifying, abstracting and describing the hidden structural patterns that underlie and supposedly 'explain' the observable linguistic behaviour of the users. Such an approach is 'segregational' in the sense that it implies that 'the language can be separated out as an independent object of description, distinguishable from the communicational circumstances - whatever these may be - in which it happens to be manifested or embedded. An 'integrational' approach, on the contrary, assumes that linguistic analysis must focus in the first instance on understanding the communication situation(s) which give(s)
4
PART ONE: LANGUAGE AND COMMUNICATION
rise to any episode of linguistic behaviour. In short, for the integrationist language cannot be decontextualized. Ch. 2 ('The Integrationist Critique of Orthodox Linguistics') sets out the most basic differences between a segregational and an integrational approach. If the study of language cannot be segregated from the study of communicational behaviour, it follows that the orthodox approach to linguistic analysis is flawed at a very fundamental level. For it presupposes, in effect, that linguistic signs are determinate, being components of an abstract system which exists independently of any particular communicational interaction that particular persons might entertain or pursue in particular cases. But this determinacy, according to the integrationist, is the last thing a linguist is entitled to take for granted. Linguistic communication is far more 'open-ended' than the segregational approach assumes, but also far more dependent on particular circumstances. Ch. 3 ('Making Sense of Communicative Competence') focuses on what we are to understand by claims that someone is communicatively 'competent' in a language. If competence is construed along the lines of the generativist distinction between 'competence' and 'performance', we are inevitably led to search for some system of 'rules' governing communicational behaviour. But in practice we judge people's behaviour by its contribution in a particular situation, and not by what systems of rules, if any, people think they are following. (Unless, indeed, we place a higher value on just following rules than on whatever might result therefrom.) The lesson here for linguists is that we should abandon any rule-based notion of competence and substitute for it the notion of communicational proficiency. The difference is important: proficiency, like effectiveness, is not something that can be defined or judged in advance of the situations that call for it, and, a fortiori, not something reducible to rules.
1
Language as Social Interaction: Integrationalism versus Segregationalism ROY HARRIS
'From the fact that language is a social institution, it follows that linguistics is a social science, and the only variable to which we can turn to account for linguistic change is social change, of which linguistic variations are only consequences.' These are the words of a great pioneer of French sociolinguistics, Antoine Meillet (1921: 16-17), cited half a century later by one of the leading methodologists of contemporary American sociolinguistics, William Labov (1972: 263). Unfortunately, it can hardly escape the attention even of a first-year student that this imposing pronouncement contains not just one glaring non sequitur but two; and it is arguable that these twin non sequiturs have continued to flaw sociolinguistic theorizing ever since Meillet first lent his authority to them. From the fact that language is a social institution, it no more follows that linguistics is a social science or that linguistic changes are only consequences of social changes, than it follows from the fact that only in human communities are crops cultivated that agriculture is a social science or that changes in cultivation are consequences of changes in society. Whatever the status of linguistics vis-a-vis the social sciences may be, it is not a status somehow guaranteed in advance by the premiss that language is a social institution (whatever that rather controversial phrase is taken to mean). It is not open to dispute that when a speaker A addresses an utterance to a hearer B, and B, having heard and understood the words A uttered, addresses a reply to A, which A in turn hears and understands, then what we are dealing with is an episode of social interaction. Nor is it open to dispute that we are dealing also with an episode of linguistic interaction between A and B. From the very beginning of modern linguistics, such examples have been taken as illustrating what Saussure was the first to call le circuit de la parole, the 'speech circuit'. On the basis of such examples, at least two fairly uncontroversial but rather modest conclusions may be drawn concerning the role of language in social interaction. One is that no theory of social interaction can afford to ignore language. The other is that no theory of language can afford to ignore the social matrix of which the speech circuit is only one part. Or, to put the point more sharply, whereas we might imagine, a la Rousseau, pre-linguistic communities in which social interaction was somehow carried on by means other than language, the hypothesis of communities either here, or on Mars, or in some other
6
Roy Harris
possible world, in which linguistic interaction was the sole form of social interaction is simply incoherent. Language is an activity which would be meaningless unless the language-users also engaged in other forms of social interaction. This, it might be thought, is a very obvious fact. Precisely for that reason, it is somewhat curious to find that it should be felt obligatory to argue the case, as Silverstein did in an article published as recently as 1977, that certain 'cultural prerequisites', as he called them, need to be taken into account for purposes of grammatical analysis. In other words, if the linguist is to give a full and satisfactory account of native speakers' mastery of their language, that account cannot ignore the speakers' awareness of certain context-dependent social practices that must be presupposed if certain types of linguistic expression are to make any kind of sense at all. It will not be relevant here to discuss in detail the particular case that is Silverstein's main concern, that of indexical expressions. Nor is it necessary to accept Silverstein's analysis of indexicals in order to agree with his more general point that in order to elucidate their linguistic function we cannot appeal to prior-established social roles of speaker, hearer, etc. since the relevant social roles are created by linguistic interaction itself. Thus, for example, it will be irremediably circular to define the meaning of the pronoun / by reference to the role of speaker, if what we understand by the role of speaker involves, inter alia, knowing how and when to use a first person pronoun. In short, there seem to be clear cases where the explication of certain aspects of social interaction is not merely intertwined with but actually co-terminous with and exhausted by the explication of certain linguistic practices. What is interesting is why it should be necessary to have to argue this as a point of linguistic theory. For an answer we must look to that part of social history that comprises the emergence and development of modern linguistics as an academic discipline. General linguistic theory in the course of the twentieth century has seen the emergence of two contrasting approaches. The main-stream approach has been, and remains, a segregational approach. Founded by Saussure, continued in the USA by Bloomfield and today by his generativist successors, 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 integrational1 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 nonlinguistic components. Therefore, according to one view linguistic intercourse is a form of human behaviour which is sui generis; whereas according to the other view that is precisely what it never is and never could be. A more detailed description of the two views would make the point that the segregationalist approach to language typically abstracts from the linguistic community and from the communication situation, and proceeds by setting up 1 The term 'integrational' is here used as proposed in Harris (1981: 165-7). It is a term that has also been used in other ways by a number of linguists, including Guillaume, Katz and Postal, Bailey and, more recently, Lieb; but these other uses have little if anything in common with the present one. In particular, it should perhaps be stressed that for present purposes there is no question of 'integrating' the various components of a language description. Language description of the traditional kind would not feature at all as a concern of the integrational linguistics here envisaged, precisely because 'languages' can only be constituted as describienda by a segregational approach.
Language as Social Interaction
7
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. Consequently for the integrationalist the proposition that language is a mode of social interaction is a simple truism. Whereas for the segregationalist, it is not strictly true at all. For the segregationalist, to speak of language as social interaction is at best a kind of ellipsis, rather like describing alcohol as a social problem. Strictly speaking, the argument goes, alcohol is not a social problem: the social problem is a problem that arises out of the availability of alcohol and what people do with it. Analogously, it may be urged, language itself is not social interaction, even though certain kinds of social interaction arise on the basis of the availability of language and what people do with it. One reason why the opposition between these two approaches has not hitherto played a more prominent part in twentieth-century linguistics is that early on a compromise was struck. The nature of that compromise is enshrined terminologically in the puzzling and ungainly word sociolinguistics. It is a term that gives titular acknowledgement to the social framework of language, but at the expense of suggesting that what sociolinguists study is only one aspect or subdivision rather than the totality of language. This compromise left segregationalists free to insist that the study of language structure rather than the study of linguistic communities constituted the disciplinary heart of linguistics and linguistic theory. But why was such a compromise necessary at all? Two complementary answers may be suggested. One was the anxiety of linguists to establish the credentials of linguistics as a subject that could stand academically on its own two feet. This was a question of institutional politics, of the independence and expansion of university departments, of competition for posts and funding, and similar issues that mark the development of Western higher education in the twentieth century. Linguistics as a subject would on that level have been distinctly less credible if its practitioners had been seen to be in fundamental disagreement with one another about aims, boundaries, and methods. The other answer was the failure of any linguistic theorist to produce a more plausible solution to the problem of linguistic heterogeneity posed by Saussure. That problem is most clearly formulated in a much quoted passage of the Cours de linguistique generate: Language in its entirety has many different and disparate aspects. It lies astride the boundaries separating various domains. It is at the same time physical, physiological and psychological. It belongs both to the individual and to society. No classification of human phenomena provides any single place for it, because language as such has no discernible unity. (Saussure 1922: 25; author's translation)
Saussure's solution to the problem of heterogeneity is well known. It is the famous distinction between langue and. parole, on the basis of which Saussure argued that the study of la langue as a synchronic system must take priority over all other aspects of linguistic study. The effect of the Saussurean solution, as pointed out very clearly by Giglioli (1972: 7), was to divorce linguistics from the study of social interaction. For if every competent speaker possessed la langue as a complete internalized synchronic system, then the testimony of a single individual sufficed in principle to provide the descriptive linguist with all the evidence needed. Furthermore, the testimony of a
8
Roy Harris
single informant seemed the best guarantee of the homogeneity of that evidence. Studying many speakers actually engaged in la parole became at best superfluous and at worst confusing. For all that was necessary for the analysis of the underlying linguistic system was a point by point comparison of linguistic expressions, considered independently of the circumstances of utterance and the disparities between different speakers. This theoretical and practical divorce between the study of la parole in use had profound consequences. One of these was that the integrationalist position never developed into a fully worked out alternative to the segregationalist approach of the post-Saussurean period, but remained in a state of subordinate dependence. A certain amount of lip-service was paid to the existence of an interactional matrix in which language functioned, but then it was to all intents and purposes ignored. Bloomfield's famous behaviourist parable of Jack, Jill and the apple, for instance, recognizes that the meaning of the utterance apple includes the totality of the interaction between Jack and Jill preceding and following that utterance. But then Bloomfield promptly dismisses the study of meanings as impracticable in our present state of knowledge and proposes to concentrate just on the linguistic structure of utterances. This is a typically segregationalist move. The question of whether it even makes sense to propose the linguistic study of utterances in isolation from their interactional context is not allowed to arise (Bloomfield 1935: 22ff, 139ff). If it had been, presumably much of the work that was done on American Indian languages in the first half of the present century would never have been attempted at all. Charles Hockett recalls that it never occurred to him during his work with the Potawatomi that it was relevant in order to describe and analyse their language to have any degree of practical mastery of it himself. He writes of this period in American linguistics, 'Boas and Sapir sent their graduate students for a year (or even a summer or so) of "field work" with a hitherto undescribed language and allowed them to write up their data as a doctoral dissertation' (Hockett 1968: 30), and he recalls the anecdote about Boas who, with a perfectly straight face, asked a student returning from his field work, 'Well did you get the whole language?' The significant point here is that these were anthropologically oriented linguists who should have been the first to grasp the complex integration of Indian languages with Indian culture. They were not, like Saussure's students, working in libraries with texts from the past rather than with human beings in the present. Yet the notion that a few months' field work with informants suffices to sift the 'linguistic facts' from their cultural embedding bears eloquent witness to the extremes to which the segregationalist approach could be taken. An interesting case is that of Sapir, who acknowledges that the normal function of language is to articulate many and varied patterns of social behaviour. Sapir's example is borrowing money. 'If one says to me "Lend me a dollar", I may hand over the money without a word or I may give it with an accompanying "Here it is" or I may say "I haven't got it" or "I'll give it to you tomorrow". Each of these responses', says Sapir, 'is structurally equivalent, if one thinks of the larger behavior pattern.' (Sapir 1949:12). Yet although Sapir realized this, he does not seem to have realized its theoretical significance, or at least not bothered to analyse the theoretical implications. The integrationalist insight is glimpsed, but then not followed through. Sapir apparently fails to see that this kind of structural equivalence is not something
Language as Social Interaction
9
outside language, but something without which what is said would be outside language - that is to say, would be meaningless. It is not surprising that the clearest expression of an integrational perspective on language should have come from one of the leading figures in social anthropology of the interwar period, Malinowski. But 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 readily to truisms with which nobody would disagree. As interpreted by J.R. Firth, Malinowski's claim emerges in the sadly emasculated guise of recognizing an 'outer' layer of contextualization statements that the descriptive linguist is obligated to undertake in order to 'complete' the description of a language. It is like giving a final road-test to the already assembled car. Thus viewed, what Malinowski says is drained of its most radical theoretical content. In America, the attempt to integrate linguistics into the general study of communicative behaviour was pursued most systematically by Kenneth Pike, while in England a similar emphasis emerged in the work of Firth, for whom 'the central concept of the whole of semantics . . . is the context of situation. In that context are the human participant or participants, what they say, and what is going on.' (Firth 1957: 27). In both Pike and Firth, however, one sees a further consequence of the compromise between the segregationalist and integrationalist positions. Although Firth uses the term integration, for him the analysis of that wider integration begins 'when phonetician, grammarian and lexicographer have finished.' In other words, Firth works from utterances 'outwards', and not from the total context 'inwards'. Like Pike, he seems to have conceived of the non-verbal part of communicative behaviour essentially as language carried on by other means. This is evident even terminologically in the case of Pike, who introduced such units as the 'behavioreme' (a term clearly modelled on phoneme and morpheme). Thus in both cases, the approach eventually adopted envisaged an extension of the analysis of language systems to embrace a certain range of related social facts, rather than any rethinking of the basic assumptions underlying the postulation of language systems in the first place. These cases are instructive because they enable us to pinpoint a deeper reason why segregationalism has dominated modern linguistics for so long. It is simply this: that even those linguists most sympathetic to an integrationalist approach to language on the theoretical level were in teaching methodologically committed to segregationalist practices. There was a failure, in short, to come to terms with the fact that a thoroughgoing integrationalism requires us to recognize a principle which may be called the 'non-compartmentalization principle' (Harris 1981: 165). Whatever name we choose to give it, this is the principle that 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. Or, to put it another way, language is not an autonomous mode of communication and languages are not autonomous systems of signs. Integration, in short, is not to be construed on the model of a jigsaw puzzle or construction kit, where we start with separate pieces,
10
Roy Harris
some linguistic, others non-linguistic and then fit them together. On the contrary, the jigsaw puzzle is a typically segregationalist model of how a language works. When the integrationalist thesis is stated in its simplest form, it may appear to leave itself open to an obvious objection. The objection runs as follows. 'But surely plain common sense demands that in the analysis of social interaction a clear distinction be drawn between the use of language and the use of non-linguistic forms of communication. For if I wish to get my neighbour at the dining table to pass the salt, there is all the difference in the world between trying to achieve this by means of pointing or gestural pantomime and trying to achieve it by means of uttering the words "Pass the salt, please". The fact that I may decide both to point and to utter the words as well does not somehow invalidate an absolute difference between verbal and non-verbal interaction.' Does an objection such as this point to the conclusion that in the end the segregationalist approach has common sense on its side after all? On the contrary, when carefully examined, our common-sense grasp of such examples falls a long way short of providing support for the segregationalist position. It is important to see how and why. According to the segregationalist the English sentence Pass the salt, please has an invariant form and meaning which is quite independent of episodes of meal-time behaviour and the interactional exchanges therein. This invariant form and meaning are determined independently by the phonological, grammatical, and semantic rules of the English language. Thus the contribution of the linguist towards explaining such an episode of social interaction will be to explain how these linguistic rules of English make the utterance recognizable as a request to pass the salt. The explanation takes roughly the following form. Because both speaker and hearer have 'internalized' the same set of phonological rules, both of them can identify the utterance as an utterance of a particular string of English phonemes, which means that they are able to distinguish it from utterances of all the infinite number of other strings of English phonemes, whether or not they have ever encountered that particular string before. Second, because both speaker and hearer have 'internalized' the same set of grammatical rules, both of them analyse that identified string of phonemes as representing the same sentence with the same grammatical structure. Specifically, they are able to identify Pass the salt, please as an imperative construction in which the salt is the direct object of the transitive verb pass, the salt being a noun phrase comprising the definite article plus the mass noun salt. Thirdly, because there are rules governing the meaning of each grammatical element and relation, and also of each lexical item slotted into the sentence structure, and because the participants know all of these, they are able to compute a total meaning for the sentence. This total meaning turns out to be something that might be paraphrased roughly as 'Kindly cause the contextually particularized sodium chloride to move in the contextually particularized direction.' The trouble with this, needless to say, is that if that is a correct characterization of Smith's message to Jones at the dining table, it is a complete mystery how the salt ever gets passed at all. In other words, we have now created for ourselves the formidable problem of explaining how Jones manages to understand what it is that this obscurely oblique message is asking him to do. Solving this problem has now become a major industry in the branch of linguistics currently called 'pragmatics': and much effort has been expended on propounding solutions. But why create the problem in the first place?
Language as Social Interaction
11
The point here is not to mount a reductio ad absurdum argument against linguists' analyses of the meaning of sentences. The point is that such an analysis, taken as seriously as you wish, shows rather that the linguist's meaning of the English sentence Pass the salt, please is not the analysandum we were looking for in order to explain the episode of social interaction. A fortiori, no amount of ingenuity in pragmatics will bridge the explanatory gap. Or, to put the matter in Wittgensteinian terms, the language-game we play at the dining table with Pass the salt, please is not a game of skill in combining dictionary meanings into grammatically structured totalities nor, conversely, of resolving such totalities into their constituent parts. The game of formal linguistic analysis is itself a language-game; but not the game we play or even have to know about in order to get the salt. Neither game 'explains' the other; nor does skill in playing one entail skill in playing the other. Once we realize this we realize that when it comes to construing language as social interaction the segregationalist approach puts the cart before the horse. For the integrationalist, on the other hand, priorities are reversed. There is no question of starting from the analysis of a decontextualized verbal abstraction and proceeding thence to explain how a piece of meal-time social interaction works. On the contrary, for the integrationalist it is that piece of social interaction that is one of the established language-games underwriting the abstraction we call the English sentence. From an integrationalist point of view, it would be fruitless to attempt to define the meaning of the words Pass the salt, please without reference to the 'cultural prerequisites', as Silverstein calls them. How complex these are, even in such a trivial case, emerges from the fact that we can imagine various cultures other than our own in which, for example, any one of the following language-games might regularly occur. 1. When Smith said Pass the salt, please, Jones first removed the salt from the salt-cellar, poured it on to a napkin, folded the napkin and then handed the folded napkin to Smith. 2. When Smith said Pass the salt, please, Jones took the salt and sprinkled it on the food on Smith's plate. 3. When Smith said Pass the salt, please, Jones took the salt, and sprinkled it over Smith's head, shouting 'Abracadabra'. There is no need to multiply the hypothetical possibilities further. In these three imaginary cultures, the meaning of Pass the salt, please would be different; and in all three cultures it would be different from the meaning we understand Pass the salt, please to have in ours. From an integrationalist point of view, the meaning is inseparable from the language-game; and it is this view of meaning we have to adopt if we want to understand language as social interaction. Unfortunately, even those who see the radical difference between segregationalist and integrationalist approaches are sometimes seized by a misguided impulse to try to reconcile them, treating them as complementary rather than opposed. What that shows is that in spite of grasping the difference they have not fully grasped its theoretical implications. Trying to combine segregationalist and integrationalist perspectives is, for theoretical incoherence, on a par with trying to combine synchronic and diachronic perspectives. It leads straight to chicken-and-egg conundrums. The language-game of passing the salt is deemed to be explicable on the basis of the meaning of the English sentence Pass the salt, please, which in turn is explained by appeal to the existence of such language-games as passing the salt, which in turn are explained by invoking the meanings of sentences like Pass the salt, please, and so on ad infinitum. The social practices and the language which supports
12
Roy Harris
them are referred back and forth to one another, in a never-ending circularity of explanation. This is what results when an attempt is made to 'complement' the integrationalist view by superimposing a segregationalist perspective upon it. It is easy enough to spot the circularity here, but far less easy to spot the perpetual-motion mechanism that sets the explanatory wheel revolving and keeps it going. That mechanism is nothing other than our talk of 'the English sentence Pass the salt, please'. For once we pose the explanatory problem in terms of second-order constructs of this nature, we have already given the wheel a spin that will keep it turning in perpetuity. But it is not immediately obvious how. For again common sense seems at first sight to support the segregationalist. Surely the integrationalist can hardly expect us to believe that there is no such thing as 'the English sentence Pass the salt, please'. Can we not, from a lay point of view, recognize that sentence, repeat it, quote it, spell it, translate it into French, and so on? How would all that be possible if English sentences were simply fictions of linguistic theory? Once again, however, the deliverances of common sense turn out under scrutiny to support the integrationalist. The integrationalist does not deny that reference to sentences features in everyday talk about language; nor maintain that it is somehow wrong or misleading to describe typical episodes of social interaction in some such terms as the following: 'Smith said to Jones "Pass the salt, please" whereupon Jones passed the salt.' or even: 'Smith said to Jones "Pass the salt, please", whereupon Jones did so.' That is not what is at issue. What is at issue is whether this validates the theoretical assumption that sentences are context-free linguistic invariants that have to be postulated in order to explain why Smith says what he does and how Jones can possibly understand it. What the integrationalist will point out is that by identifying the social interaction between Smith and Jones in these terms we are already choosing to describe what we presume to be one aspect of the participants' contextualized interpretation of that interaction. In so describing it we do not make tacit appeal to some linguistic theory of sentential invariants but overt appeal to how Smith intended and Jones understood the utterance to be communicationally integrated into a particular continuum of activity. We assume no more, in other words, than that if Smith were asked 'What did you say to Jones?' he might truthfully reply 'I said: Pass the salt please', and if Jones were asked why he passed the salt he might truthfully reply 'Because Smith said: Pass the salt, please.' Nothing else is required to validate descriptions of social interaction by reference to sentences. But that validation affords no foothold for a segregational linguistics. On the contrary it renders the linguist's account of the decontextualized sentences not merely superfluous but irrelevant to an analysis of language as social interaction. The segregationalist typically makes several mistakes here. The first mistake is to assume that when we identify metalinguistically what Smith said to Jones as the sentence Pass the salt please we are implicitly referring to some determinate subset of the events that took place in that particular episode of social interaction. And upon this first mistake a second mistake is then built, which is to assume that the correct analysis of the episode is one which compartmentalizes it neatly in two: on the one hand, the subset of events constituting the component Pass the salt, please, and on the other hand the non-linguistic events surrounding it. A third mistake is then added to complete the superstructure of errors, which is to suppose that by applying a set of
Language as Social Interaction
13
analytic procedures valid ex hypothesi for the sentence irrespective of the context we can recover what the participants thought was meant and thus explain the whole episode. In short, there is a total confusion of sentences with utterances, and of the language-games in which we make reference to sentences with the language-games thereby described. A thoroughgoing integrationalism would mean taking seriously the consequences of Sapir's observation that communication is based on structural correspondences between certain forms of behaviour in a situational context. It would mean recognizing that passing the salt when asked to do so is no less a linguistic act than uttering the words Pass the salt, please in order to get it. Both are complementary manifestations of linguistic knowledge, of proficiency in a language-game, and we unhesitatingly treat them so in the communication situations of everyday life. Only a blind acceptance of the theoretical dogma which equates language with spoken language and thus induces the mistake of identifying vocalization as the mark of the linguistic act par excellence will prevent us from recognizing that truth. To make that mistake is on a par with believing that you are not running except during the seconds when your feet touch the ground and only playing tennis at the moments when your racquet hits the ball. If the task of explicating language as social interaction is to be taken as a serious goal, it cannot be tackled either on a basis that equates the use of English with the vocal utterance of English words or on the basis that somehow the English language and other languages, or varieties thereof, are 'given' in advance, and all the investigator has to do is to go out and investigate how they are being used. That positivistic misconception is the inherent flaw in much of what is nowadays called sociolinguistics. It proceeds by trying to correlate a few selected features of speech (very often items of vocabulary or details of pronunciation) directly with social classifications of the speakers (by such parameters as age, sex, income, occupation, education, etc.). Perhaps it could be argued that this is virtually all that is left for sociolinguists to do, once the segregationalist thesis has been accepted that the analysis of linguistic structure is a quite separate enterprise from the description of language use. Whatever the reason, the result is that the tables, graphs, maps, and statistics that are the characteristic tools of the sociolinguist's trade tell us no more about language than they would about fashions in clothes or hairstyles. This is because the patterns they reveal are based on external correlations, the relevance of which to the communicational function of language is itself problematic. To proceed thus is in effect to divorce language from communication in a way that either ignores or trivializes the main theoretical problems concerning both language and communication. Language as social interaction 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 co-monitor with one other person a behavioural continuum along which a succession of integrated events can be expected to occur. To have grasped and be able to exploit these integrational connexions is what makes us communicationally proficient members of a community. That proficiency is not to be explained on the basis of 'knowing the language'. On the contrary, it is only on the basis of that communicational proficiency that one can begin to make sense of the concept of 'knowing the language'.
14
Roy Harris
This takes us back to what Malinowski was already calling in the 1930s 'the dilemma of contemporary linguistics'. Malinowski summed up that dilemma by asking 'Can we treat language as an independent subject of study? Is there a legitimate science of words alone, of phonetics, grammar and lexicography?' Malinowski's answer was 'No'. We may not agree entirely with his reasons; but nothing that has happened in the field of language studies since Malinowski's day suggests that in the end he got the answer wrong.2
References Bloomfield, L. (1935) Language. Allen & Unwin, London. Firth, J.R. (1957) Papers in Linguistics 1934-1951. Oxford University Press, London. Giglioli, P.P. (ed.) (1972) Language and Social Context. Penguin Books, Harmondsworth. Harris, R. (1981) The Language Myth. Duckworth, London. Hockett, C.F. (1968) The State of the Art. Mouton, The Hague. Labov, W. (1972) Sociolinguistic Patterns. University of Pennsylvania Press, Philadelphia. Malinowski, B. (1923) The problem of meaning in primitive languages. In C.K. Ogden & I.A. Richards, The Meaning of Meaning, Routledge & Keegan Paul, London. Meillet, A. (1921) Linguistique historique et linguistique generate. Champion, Paris. Sapir, E. (1949) Language. In D.G. Mandelbaum (ed.) Selected Writings of Edward Sapir. University of California Press, Berkeley. Saussure, F. de (1922) Cours de linguistique generate, 2nd ed. Payot, Paris. Silverstein, M. (1977) Cultural prerequisites to grammatical analysis. In M. Saville-Troike (ed.) Linguistics and Anthropology. Georgetown University Press, Georgetown.
2 An earlier version of this paper was presented at the ICU Conference on Sociolinguistics, Tokyo, 19 July 1986.
The Integrationist Critique of Orthodox Linguistics ROY HARRIS
Introduction The past ten years in linguistics have been marked by increasing scepticism concerning the theoretical foundations on which the majority of linguists have based their descriptions of languages for most of the present century. This critique can be traced back through Firth and Pike to Malinowski and Sapir, but has been significantly influenced more recently by the later work of Wittgenstein and by developments in sociology. The confrontation of theoretical positions involved has been characterized of late by contrasting 'segregational linguistics' with 'integrational linguistics'. What do these terms imply? Is this simply another difference between schools of linguistics (on a par, say, with that between stratificationalists and glossematicians), or is it something which goes much deeper? Does it have implications for the future of linguistics itself? Linguistics, conceived as an independent branch of academic study, embracing the entire range of observations and questions which relate to language, is a relatively recent development in the history of human culture. Its emergence in the particular form with which we are familiar owes much to the patterns of education and research characteristic of Western universities in the nineteenth and twentieth centuries. Linguistic theory has in large part been promoted by the perceived professional need to organize linguistic investigations in a coherent, comprehensive and principled way for academic purposes, rather than by the search for ultimate answers to the most general and far-reaching questions about language that a linguist might ask. The most fundamental of these questions is 'What is language?' As soon as it is asked linguistic theory already faces a problem of focus. For language involves at least three activities. It involves neural activity in the human brain. It involves muscular activity of the body. Third, it involves social activity of the kind which engages individuals interacting with one another. Different definitions of language interrelate these three activities in different ways, and imply different views of exactly which neural, muscular and social activities are involved. Furthermore, if language is viewed not as an activity at all but as a faculty or ability which underlies the activity and makes it possible, the dilemma of conflicting definitions simply reappears at one
15
16
Roy Harris
remove. For there is no way of identifying the language faculty without reference to the activity which it is alleged to sponsor. The solution which has found favour with most theorists since 1900 is to focus neither on the faculty nor on the activity but on something else. The faculty is usually assumed to be universal in the sense of being common to the whole human race, in whatever stage or circumstances of cultural development its separate communities may be. The activity, on the other hand, is usually assumed to be activity carried on voluntarily by particular individuals responsible for their own linguistic behaviour. However, intermediate between the universal faculty and the individual activity there intervenes a third order of phenomena: there are said by the theorist to be systems shared by groups of individuals, but differing from one group to another. These systems, which determine within certain limits the form the activity takes in any particular instance, are languages. By focussing on languages as systems, rather than on the language faculty in general, or on particular language activity, modern linguistic theory made it possible to reserve for linguists a domain of academic inquiry unlikely to suffer unduly from encroachment by the neighbouring disciplines of psychology, physiology and sociology. For language systems, so it was claimed, can be studied as autonomous objects without reference to the circumstances in which they function as means of communication. At the same time, linguistic theory neatly avoided the problem of defining language. Linguistics could proceed without resolving this problem, so long as languages rather than language were taken to be the primary object of investigation. The first theorist to make this strategy the explicit basis of a rationale for linguistics was the founder of the discipline in its orthodox twentieth-century form: Ferdinand de Saussure. To refer to this as the 'Saussurean strategy' carries no implication that it was confined to Saussure and the Geneva school. On the contrary, it came to be adopted by linguists of all schools, irrespective of their attitude towards other aspects of Saussurean theory. Adoption of the Saussurean strategy, however, in turn posed a theoretical problem. Wherein resides the systematicity which languages as such are deemed to possess? The history of twentieth-century linguistic theory has been largely the history of two complementary but conflicting answers to this question. In what sense are languages systems? According to one thesis, the systematicity is internal, in that it lies, for each language, in the relations between the words of that language, their parts and their combinations. According to the other thesis, however, what may appear to be internal systematicity is externally derived from, and maintained by, the wider social practices in which language is involved. A corollary of the first thesis is that linguistic signs are amenable to study in themselves, quite apart from their users and the contexts in which they are used. In this sense languages may be segregated or set apart from language, at least for purposes of academic investigation. A corollary of the alternative thesis, on the contrary, is that linguistic signs are intrinsically integrated into other forms of behaviour, without reference to which they cannot even be identified, let alone studied, either by linguists or by anyone else. The terms 'segregational' versus 'integrational' capture this basic theoretical dichotomy. It is the segregational perspective which has dominated academic language studies for most of the present century, and provided the theoretical guidelines for orthodox
The Integrationist Critique of Orthodox Linguistics
17
linguistics. All the major schools of linguistic theory, from Saussurean structuralism down to contemporary generativism, have been basically segregationalist in outlook. The integrational perspective, on the other hand, has received comparatively little theoretical elaboration, being favoured principally by linguists with strong inclinations towards anthropology and the study of language as part of the broader study of culture. Consequently there has been little in the way of theoretical confrontation between the two perspectives. Moreover, at first sight, there appears to be no need for any such confrontation, since an attractive compromise seems to be readily available. This consists in recognizing a legitimate division of labour. Let segregational linguistics treat languages as autonomous systems, examine their internal relations, and perhaps speculate at a very abstract level about how such systems might be represented in the human brain, while integrational linguistics will concentrate on the external relations between languages and the individuals and communities using them. Unfortunately, the compromise is one-sided. In effect, it assigns to the integrationalist what is left over after the segregationalist has completed the central task of analysing the autonomous language system. Nevertheless, a compromise of this kind was accepted by many linguists whose theoretical leanings were integrational, and it was on this parlous compromise that the modern subdisciplines of sociolinguistics and psycholinguistics were founded. So long as the compromise was accepted, no crisis threatened orthodox linguistics. However, in recent years it has become increasingly evident that the compromise was a mistake in the first place. For the compromise is not theoretically defensible, even though a whole generation of linguists may have thought it was. There are two dovetailing reasons for this. On the one hand, a weak integrational position, for instance of the variety adopted by J.R. Firth and some neo-Firthians, is now seen to be internally inconsistent (Love 1988). It fails to follow through the basic integrational claim that language makes no sense except in relation to the relevant totality of human behaviour. On the other hand, a radical integrationist position (Harris 1981) leads to an outright denial of the assumptions which orthodox linguistics has to take for granted in order to engage in linguistic analysis at all. In short, there can be no theoretical compromise because the segregational and integrational approaches are mutually contradictory. But it has taken linguists a long time to realize the depth of the contradiction. The integrationist case against orthodox linguistics has been most recently and most explicitly argued in the volume edited by H.G. Davis and T.J. Taylor entitled Redefining Linguistics. Although the arguments presented there are not new, the case needed to be laid out in theoretical terms, because the point of the integrationist critique can easily be blunted by sympathetic misunderstanding. Thus, for example, Gardner's remarks on integrationism and generative grammar (Gardner 1985: 221) seem to imply that all would be well if only linguists were prepared to benefit from interdisciplinary co-operation with neighbouring disciplines. Matthews (1985: 7), on the other hand, appears to take the view that integrationism, in common with recent generativist theory, points towards developing a linguistics of the individual rather than a linguistics of languages. Both of these interpretations are rather wide of the mark although each has grasped one facet of the integrationist case.
18
Roy Harris
The radical integrationist position is perhaps best summed up by Davis, when she applies to linguistics a comment borrowed from Wittgenstein's Philosophical Investigations: 'the game we play with it is unregulated' (Wittgenstein 1953: 68). 'Linguistics', says Davis, 'is unregulated' in this sense: that is, open-ended. For the integrationist this is necessarily so, because linguistics is a language game and language itself is open-ended. The integrationist critic therefore sees orthodox linguistic analysis as resulting from a failure by linguists to understand the nature of the game being played. As a result the players are constantly proposing impossible moves. Linguistics in the hands of the orthodox theorist becomes the imposition of determinacy on phenomena which are intrinsically indeterminate and the discovery of rules which do not exist. Integrationism is the first movement in the history of linguistics which does not attempt to extend and refine established methods and concepts of language description, or to provide them with theoretical underpinnings. Its failure to address these descriptive concerns is not, however, an oversight. On the contrary, integrationism as advocated by the contributors to the Davis & Taylor volume calls in question the validity of the whole descriptivist enterprise which has loomed so large in the output of modern linguistics. Far from proposing an alternative descriptive formalization of any kind, the integrationist argues against the descriptivist's efforts as implicitly, albeit unwittingly, misdescribing the way language works. But to abandon language description, it need hardly be stressed, would be for the orthodox camp to abandon linguistics altogether. So the division cuts deep, and could hardly cut deeper. Much deeper, in any case, than controversies such as that between 'hocus pocus' and 'God's truth' linguists. The champions of hocus pocus linguistics - and Firth again comes to mind as an example - never abandoned description as such: they simply admitted that their descriptive systematizations were arbitrary and denied that any one kind of systematization took priority qua description over any other. Hocus pocus linguistics would have better been called 'descriptive pluralism'; for that essentially was what it was. And variety of description could even be held to be of value as highlighting different facets of linguistic structure which no single description could capture. Whereas integrational linguistics, far from being descriptive pluralism, is descriptive nihilism. And while it is true that a number of orthodox theorists (for example, Hockett 1968) have from time to time questioned whether the language system is fully determinate, that has never stopped them from describing selected parts of it as if at least those parts were. But this, from an integrational point of view, is simply an attempt to have your descriptive bun and eat it. For similar reasons, the division between the segregational and the integrational approaches to language goes much deeper than that which separates theorists who differ over whether languages are psychological objects, or biological objects, or semiotic objects, or abstract objects. In spite of their differences, the various theorists agree that these objects are not directly accessible to observation: their existence has to be inferred. What the integrationist questions, precisely, is whether there are any such objects at all. For the integrationist it is possible to do linguistics without assuming that the linguistic universe consists of a large number of discrete objects called 'languages', let alone having to treat each such object as a self-contained system.
The Integrationist Critique of Orthodox Linguistics
19
The philosophical basis of the integrationist position is the thesis that the linguistic universe is populated not by mysteriously unobservable objects called 'languages' but by observable human beings who somehow and sometimes manage to communicate with one another. For the integrationist, however, communication is subject to a universal principle derived from everyday human experience. Human experience is constantly structured and restructured by the need to make sense of present events in the light of past events and vice versa. Language is both a product and a mechanism of this process, by which the ceaseless flow of sensations, perceptions, feelings and judgments which contribute to the mental life of the individual are integrated into a continuum, and a stable framework of beliefs and expectations about the world is constructed and maintained. The human capacity for recognizing, inventing and using signs of various kinds is an essential factor in this process. A sign is any observable feature or complex of features which, by virtue of its integrational function, plays some role in our diverse but continuous practices of making sense. The sole necessary and sufficient condition for the constitution of a sign is our recognition of this role. Signs are fundamental to life in human societies, for every aspect of social organization is in some measure dependent upon them. Where integrational semiology differs from other semiologies is in not restricting the linguistic sign either to speech or to writing. From an integrational point of view, language arises from the creative use of the communicational space in which we live, and speaking and writing are only two of the many human activities which articulate this space. A major mistake made in orthodox linguistics, as far as the integrationist is concerned, is its pervasive conflation of linguistic communication with speech communication, and its consequential attempt to restrict linguistic meaning to meanings of sequences of speech sounds. This restriction is canonized by segregationists as the principle of the 'primacy of speech' (often misinterpreted as being merely a thesis about the relationship between speech and writing; but it is much more than that). The so-called primacy of speech is a vital principle for the segregationist to cling to. Without it, all the other cherished dogmas of orthodox linguistics become manifestly untenable; including such dogmas as the linearity of the linguistic sign and the concept of grammaticality. The central thesis of orthodox linguistics throughout the present century has been that if we restrict our attention to speech signals we shall discover that there is, for each community, a discrete and determinate system of signs in operation, which constitutes that community's language. What is wrong with the 'primacy of speech' principle is just that speech has no primacy in human communication. Speech sounds have no meaning unless and until they can be integrated, just like other signs, into the process of making sense of some particular episode in the continuum of human experience. And this integration demands in general that there should be patterns of activity other than speech which provide the context for speech to be meaningful. Equally, however, speech provides part of the context in terms of which other activities are meaningful. Speech, therefore, is not segregatable as a or the linguistic form of communication: and to suppose otherwise is merely to confuse the study of speech with phonetics.
20
Roy Harris
The segregationist response In response to the integrationist critique, the orthodox segregationist has only a limited number of moves available. Although variations are possible, in the end they reduce to the following strategies. One strategy is to lodge the counterclaim that the integrational approach fails to distinguish between the linguistic knowledge that speakers of a language have and their general knowledge of the world in which they live. Hence the integrationist is unable in principle to establish any scientific basis for linguistics; for without such a demarcation the subject matter for the linguist's investigations remains so ill defined that linguists would be at liberty to vary radically one from another over what to count as linguistic evidence. This type of counterclaim may also be accompanied by the concession that communication does in some sense involve integration; but what the integrationist fails to see is that the integration is between different systems of knowledge - linguistic on the one hand and non-linguistic on the other. A somewhat different strategy is to concede to the integrationist that language is open-ended and linguistic structure in many places weak or unclearly defined, but to maintain against the integrationist that this is perfectly compatible with systematicity. The appeal of this strategy is that it can be made to fit in with the traditional truism that language is in a constant process of change. Thus instances of the kind an integrationist might cite in order to demonstrate the indeterminacy of language can always be explained away as instances in which the system is undergoing evolution. Pursuit of these two strategies in recent years has led to much discussion of such untraditional topics as fuzzy categories, squishes and variable rules. However, as Schnitzer (1990) points out, these manoeuvres in the end reinforce rather than weaken the basic orthodox position that language users have access to a unitary coherent well-formed system. He writes: The existence of approaches involving fuzzy sets (Lakoff 1973) or squishy grammatical categories (Ross 1972, 1973a,b) does not mitigate this claim since these approaches imply not that the system is in some sense nonexistent or not fully known by the speakers, but rather that not all grammatical classification is binarily Boolean. What is intended in such approaches is that inherent in the putative grammatical system which speakers have knowledge of is a degree of indeterminacy and/or gradation.
Schnitzer identifies five theses which remain unaffected by the concessions which orthodox linguistics currently makes in various compromises with its critics. These are (i) that a language may be viewed as a set of sentences generated by a grammar which generates no non-sentences; (ii) that native speakers have access to a system of tacit knowledge which represents the grammar of the language, and on the basis of which they make grammaticality judgments, (iii) that the mental grammar is unitary, (iv) that it is exhaustive in the sense of comprehending all of the speaker's purely linguistic knowledge, and (v) that it is psychologically real, neurophysiologically instantiated, and independent of the various production and comprehension tasks it may be involved in. For orthodox theorists these five theses remain intact, Schnitzer argues, irrespective of
The Integrationist Critique of Orthodox Linguistics
21
whether a theory uses transformations or not, whether it makes use of phrase-structure relationships or dependency relationships, whether it makes use of syntactic categories or syntactic relations or thematic relations or all three, whether it has one or more than one level of description, whether grammars include arrangements or processes or both or neither, whether they admit of apparatus to include 'pragmatic' phenomena, whether they contain derivations, well-formedness conditions, and/or filters, and whether they embody greater or lesser levels of abstraction.
As long as these five theses remain standing, it seems clear that the basic disagreement between integrational and segregational approaches remains unresolved, whatever concessions may appear to have been made on the issues of determinacy and systematicity. All that has changed, from the integrationist standpoint, is the formal guise in which orthodox linguistics presents and promulgates the 'language myth' (Harris 1981).
Three misconceptions of indeterminacy The most important issue which divides integrationist from segregationist theorizing is currently seen as being that of determinacy. But exactly what is at issue is often blurred by prevalent misconceptions of determinacy. To bring these misconceptions out into the open will at least contribute to clarifying the discussion, even if it does not persuade the segregationist to abandon them. Concerning determinacy of grammar, there is much talk at cross purposes which is due to the way in which orthodox linguistics equivocates with the concept of a 'rule'. For example, according to the Grammar of Contemporary English by Quirk, Greenbaum, Leech and Svartvik (1972) there is a rule that in English the only adverbs that can modify other adverbs are 'intensifiers' (of which paradigm examples are very and quite). If it is pointed out that the adverb unbelievably can also modify another adverb, the answer is that in that case unbelievably must be an intensifier too. But the moment that answer is given the status of the alleged rule alters: in effect, the term 'intensifier' is now being implicitly defined as 'adverb which can modify another adverb'. Immediately, the rule becomes unfalsifiable; and the price of unfalsifiability is circularity. The equivocation does not end there, however. According to the Grammar of Contemporary English, the intensifier rule explains why, although His reasoning was theoretically sound is good English, He reasoned theoretically soundly is not. If it is objected that there are speakers of English who find no fault with this juxtaposition of adverbs, the grammarian can always reply either that for those speakers theoretically must be an intensifier, or else that for such speakers theoretically is not an adverbial modifier but a sentential adverb. In either case the rule is saved. Similarly, if there are speakers who accept He behaved extremely offensively but reject He behaved deeply offensively it can always be held that for such speakers deeply is not an intensifier. The point is this. Either you admit that the grammar of English adverbial modification is not well defined; or else you invent grammatical subcategories like 'intensifier' in terms of which it can be presented as being well defined. Now the issue of determinacy does not hinge on whether it is possible in principle to come up with grammatical categories and rules which cover all the cases and leave the rules
22
Roy Harris
inviolate. For the integrationist that is a question of no theoretical interest whatever. If you allow yourself circular rules like the intensifier rule then you will always be able to claim that grammatical structure is determinate, come what may. To suppose that is the issue on which integrationists and segregationists differ betokens one kind of misconception of determinacy. Let us call this the 'definitional' misconception. A somewhat different misconception of determinacy is championed by Lyons (1968), who argues as follows. In describing a given language, the linguist will draw the limits of grammaticality at a particular point. His decision to draw these limits at one place rather than another, if the decision is made consciously after weighing the various alternatives that present themselves, will tend to be determined by two main factors. The first may be referred to as the principle of 'diminishing returns'. It is possible to go a lot further with the distributional subclassification of words than would have been thought feasible, or even desirable, by traditional grammarians. But sooner or later, in his attempt to exclude the definitely unacceptable sentences by means of the distributional subclassification of their component words, the linguist will be faced with a situation in which he is establishing more and more rules, each covering very few sentences; and he will be setting up so many overlapping word-classes that all semblance of generality is lost. This is what is meant by the principle of 'diminishing returns': there comes a point (and where this point is might be legitimate matter for dispute) at which the increase in the complexity of the rules is too 'costly' in proportion to its 'yield', a relatively small increase in the coverage of acceptable and unacceptable sentences. But the second factor is no less important. Since the sentences of the language being described are so numerous . . . one cannot hope to decide for every sentence generated by the grammar that it is definitely acceptable or unacceptable. In fact, one does not have to go very far with the grammatical description of any language before one finds disagreement among native speakers about the acceptability of sentences generated by the rules tentatively established by the grammarian. There is therefore a real, and perhaps ineradicable, problem of indeterminacy with respect to acceptability and unacceptability. It would seem to follow from these considerations that the grammatical structure of any language is in the last resort indeterminate. It is not only that linguists will differ in their interpretation of which constitutes the optimum degree of generality in the scope of the rules and in their evaluation of the acceptability of particular sets of sentences. There is the additional problem that the generation of one set of sentences of a particular type may make the generation of other sentences of a different type extremely difficult to handle within the theoretical framework established for the first s e t . . . We may therefore restate as a general principle which governs all grammatical description . . . the following fact: whether a certain combination of words is or is not grammatical is a question that can only be answered by reference to a particular system of rules which either generates it (and thus defines it to be grammatical) or fails to generate it (and thereby defines it to be ungrammatical). Most writers on grammatical theory would seem to reject this principle. They suggest that the grammatical structure of any language is determinate and is known 'intuitively' (or 'tacitly') by native speakers. This appears to be an unnecessarily strong assumption. It is undoubtedly the case that native speakers will agree that certain sets of utterances 'belong together', or are 'similar' or 'different' in some way. These 'intuitions', in so far as they are ascertainable, are an important part of the linguist's data; and he will try to account for them by distinguishing various kinds of acceptability (or well-formedness) and various kinds of relatedness between sentences. But he need not assume that there will be any very direct correspondence between the 'intuitions' of the speakers and the statements made by the linguist. One should not exaggerate the difference of opinion between linguists on this question. To assert that the grammatical structure of language is in the last resort indeterminate is not the same as to assert that no part of the grammatical structure is determinate. There are many combinations of words (e.g. *They likes she, *The dog bite the man, etc.) which all linguists will characterize immediately, not only as unacceptable, but also as 'ungrammatical' (without necessarily producing a set of grammatical rules). One can say that their immediate reaction is
The Integrationist Critique of Orthodox Linguistics
23
based on an 'intuitive' awareness of the grammatical structure of standard English; one can equally well say that the combinations in question infringe principles of such generality in utterances of standard English that any grammar would necessarily have to take account of them. It is with respect to the less general principles that alternative grammars might differ in their characterization of sentences as grammatical and ungrammatical. And the 'intuitions' of linguists and speakers of the language tend to be unreliable and inconsistent at this point anyway. (Lyons 1968: 152-4)
In short, for Lyons the grammar of English is up to a point determinate, but beyond that point indeterminate; and the point in question can somehow be identified by seeing where different grammatical descriptions begin to diverge. Let us call this the 'common core' misconception of determinacy. For the integrationist, whether there is such a common core is not what is at issue either. There is a third misconception of determinacy which is more simpleminded than either of the foregoing. Even the most diehard defenders of grammatical rules can usually be cornered into admitting that there just is no rule determining exactly which English adjectives form their comparatives by adding -er and which do not; although everyone agrees that some English adjectives aways do and others never do. Similarly, they will often concede that there is no rule in this sense about the use of the English suffix -ness to form abstract nouns. But, they will argue, this does not show that in these areas the grammatical structure of the language is indeterminate. On the contrary, all it shows is that the grammar includes rules which require the language-user to exercise an optional choice. Let us call this the 'compulsory option' misconception of determinacy. This misconception is particularly popular in historical linguistics, because it fits in with the traditional truism that languages are constantly changing. This makes possible a reconciliation of some kind between the theoretical requirements of synchronic and diachronic linguistics, a perennial problem which is part of the legacy of Saussure. Thus points on which the grammar allows the language-user an optional choice can be seen as points at which the language system is potentially open to change. These three misconceptions of determinacy can be summed up in the following three propositions, (i) Linguistic structure is determinate in the sense that, for any given language, grammatical categories and rules can be set up which are comprehensive and admit no exceptions, (ii) Linguistic structure is determinate in the sense that, for any given language, there is at least a common core of inviolable rules of grammar, (iii) Linguistic structure is determinate in the sense that the grammar of any given language determines the limits within which language-users may freely choose one form of expression rather than another. For the integrationist, however, accepting or rejecting any or all three of these propositions does not get to the heart of the matter. This is because, on the integrationist view, grammar is not rule-governed anyway in the sense which has to be accepted in order to make any of these propositions worth debating. There are no rules of grammar: there are only grammarians' rules (which is not at all the same thing). The indeterminacy of grammar, for the integrationist, is a simple corollary of the indeterminacy of the linguistic sign itself.
24
Roy Harris
The indeterminacy of the linguistic sign The indeterminacy of the linguistic sign is the central doctrine of integrationism. The integrationist answer to the question why linguistic signs are not determinate is that the exigencies of human communication demand that they should not be. In other words, indeterminacy serves human purposes better than determinacy. Determinacy has its value in limited areas with specific objectives; but for mediating human communication in the generality of cases determinate systems would be dysfunctional because of their very lack of flexibility. The truth that linguistic signs as used in everyday discourse are indeterminate was recognized centuries ago by the greatest of all empiricist thinkers, John Locke: but it has taken linguists until now to rediscover that truth and think through its implications. A linguistic sign is not a fixed form with a fixed meaning, both of which remain magically invariant across all the communicational episodes in which the sign is used. On the contrary, what constitutes a sign is not independent of the situation in which it occurs or of its material manifestation in that situation. This is one of the basic axioms of an integrational semiology (Harris 1984). It manifestly holds for signs in general, and one of the ironic lacunae in orthodox linguistics is that no theorist has ever explained why the linguistic sign should be treated as an exception. But if the linguistic sign is not an exception, the consequences for linguistics are profound. It becomes vain to look for a fixed code underlying the communicational practices of particular communities; and if this is vain there can be no sense in which linguistic theory needs the concept of 'a language' as its foundation. That would only be necessary if human beings were machines programmed by their maker to accept as communicational input only signals drawn from a fixed inventory of invariant forms, and interpretable only by reference to a fixed inventory of invariant meanings. To believe that this is true, or even approximately true, is to have a remarkably blinkered view of humanity. But this blinkered view is more than congenial to those sciences which can deal with human capacities only on the assumption that the human being must be, in the final analysis, a very complex species of the genus 'computer'. The integrationist contention that there are no such things as languages is open to all kinds of misinterpretations by those who wish to reject it. Here it will have to suffice to make two relevant points. The first is that from the lay observation that horse is an English word, whereas cheval is a French word, it does not follow that there must be two theoretically identifiable objects known as 'the English language' and 'the French language'. How people distinguish words as 'English', 'French' etc. is, from an integrational point of view, an empirically researchable subject; but the fact that they do so has no theoretically privileged status. The second point is closely related to the first. It does not follow, either, that if someone is confident that such-and-such an expression is 'good English', whereas another expression is not 'good English', there must be in operation a set of rules which the good English conforms to and the bad English violates. For the integrationist, value judgments concerning good and bad usage do indeed raise questions to be addressed. But these questions of linguistic correctness are to be addressed by investigating the integrational patterns which give rise to the value judgments in the first place. They are certainly not to be addressed by postulating
The Integrationist Critique of Orthodox Linguistics
25
internalized rules to account for them. For that move is in any case explanatorily vacuous. In linguistic behaviour as in other areas of behaviour, value judgments cannot be explained by appeal to imaginary psychological mechanisms for allocating values. For the integrationist, linguistic inquiry begins not with a search for invariant units underlying speech communication in a given community, but with an investigation of the presuppositions underlying the linguistic inquiry itself. An objection which might be raised to this is that it either treats linguistics as a form of philosophy, or else redefines linguistics out of existence altogether. What will the integrationist reply to this objection? As to the first part, linguistics is a form of philosophy, whether we like it or not: it is that part of philosophy which seeks to elucidate how and where language fits in to the general human scheme of things. As to the second part, one may refer again to Davis's characterization of linguistics as 'unregulated'. Her observation concludes: although linguistics is unregulated . . . it can never be redefined out of existence, only into existence, into an analytic understanding of one's individual linguistic experience.
One might gloss that observation as a call for linguists to begin by examining the foundations of their own metalinguistic practices. The integrationist critique of orthodox linguistics may thus be seen historically as bringing the critical inquiry begun by Saussure full circle. Against the linguistics which Saussure inaugurated and his successors in Europe and America developed, the integrationist levels exactly the same criticism that Saussure himself levelled against the linguistics of his predecessors: that it failed to identify its true object of study. By questioning the availability of any objective viewpoint from which faits de langue, as Saussure called them, may be identified, verified and descriptively systematized, integrationism refocusses attention on the ways in which human beings, as language-makers, construct their own linguistic experience. It subjects both the linguist and the linguistic inquiry to a simultaneous recontextualization within the creative process which is language.
References Davis, H.G. & T.J. Taylor (eds.) (1990) Redefining Linguistics. Routledge, London. Gardner H. (1985) The Mind's New Science. Basic Books, New York. Harris, R. (1981) The Language Myth. Duckworth, London. (1984) The semiology of textualization. Language Sciences 6 (2), 271-86. Hockett, C.F. (1968) The State of the Art. Mouton, The Hague. Lakoff, G. (1973) Fuzzy grammar and the performance/competence terminology game. Papers from the Ninth Regional Meeting. Chicago Linguistics Society. Lyons, J. (1968) Introduction to Theoretical Linguistics. Cambridge University Press, Cambridge. Love, N. (1988) The linguistic thought of J.R. Firth. In R. Harris (ed.) Linguistic Thought in England 1914-1945, pp. 148-164. Duckworth, London. Matthews, P. (1985) Whither linguistic theory? Linguistics Abstracts I (1), 1-7.
26
Roy Harris
Quirk, R., S. Greenbaum, G. Leech & J. Svartvik (1972) A Grammar of Contemporary English. Longman, London. Ross, J.R. (1972) Endstation Hauptwort: the category squish. Papers from the Eighth Regional Meeting. Chicago Linguistics Society. (1973a) A fake NP squish. In C.-J.N. Bailey & R. Shuy (eds.) New Ways of Analysing Variation in English. Georgetown University Press, Georgetown. (1973b) Nouniness. In O. Fujimura (ed.) Three Dimensions of Linguistic Theory. Tokyo. Schnitzer, M.L. (1990) Critique of linguistic knowledge. Language & Communication 10 (3). Wittgenstein, L. (1953) Philosophical Investigations, transl. G.E.M. Anscombe. Blackwell, Oxford.
Making Sense of Communicative Competence ROY HARRIS
I It has become increasingly common to take for granted in discussions of language a theoretical distinction between linguistic competence on the one hand and communicative competence on the other. Various questions might be raised in connexion with this distinction, of which two in particular are of concern here. One question is whether the concept of communicative competence can be made sufficiently precise to be useful. The other is whether the theoretical presuppositions underlying such a distinction are valid. These questions are fundamental because, as will be shown, they bear upon the central issue of the limitations inherent in the adoption of rule-based models in linguistics.
II
There is no doubt that people do often attribute breakdown in communication to what they regard not as linguistic ignorance but rather to ineptness in using one's language for purposes of communication. Recently, for example, the BBC put on a whole series of educational programmes designed to improve the communicational efficiency of ordinary language-users. These programmes concentrated on recurrent types of situation in everyday life where people tend to get at cross-purposes, not because they do not speak or understand English very well, but because they fail to grasp the right way to put across their message. The programme series was accompanied by a publication called Speak for Yourself, and it had the ambiguous subtitle Making Language Work for You. But what the series was trying to do is exemplified fairly clearly by the first scenario in the publication. The title is 'Emergency! Dial 999'. There is a dialogue (presumably scripted) between Teresa and the Telephone Operator which runs as follows. Operator: Teresa: Operator:
'Emergency. Which service do you require?' 'There's a pan of chips on the stove and I'm locked out. I think I can smell smoke.' 'What number are you ringing from?'
27
28
Roy Harris
Teresa: Operator: Teresa: Operator: Teresa: Operator: Teresa: Operator: Teresa: Operator: Teresa: Operator:
'What number . . . number . . . 126 Wembury Place.' 'What telephone number?' 'Oh come quick, quick!' 'What telephone number are you ringing from?' 'Number . . . number . . . 8 . . . 8 . . .' '8 ... yes?' 'No not 88. 839 ... 9 ... 9 ... '839 yes?' '839 1925' '839 1925?' 'Yes, yes.' 'Hold the line. I'm putting you through.'
Following the dialogue, there is an analysis of what has gone wrong. It makes little difference for present purposes that the example used is hypothetical. What has gone wrong is typical enough of such situations. There is panic on one end of the telephone and seemingly impersonal bureaucracy at the other. When the operator asks for the number, the caller evidently thinks that it is the street number she is being asked for. She cannot understand why anyone would want to know the telephone number: it is the address that matters. The fire brigade cannot put a fire out over the telephone. Whereas the operator is following a tested procedure for dealing with all emergency calls, which has established that it is of prime importance to find out first what number the caller is ringing from. But for Teresa this sounds like an irrelevant question or a pure formality: and this apparent irrelevance to her situation increases the panic and decreases her own ability to cope with the question. But nowhere in the analysis of this communication chaos is there any suggestion that Teresa's English is at fault, or the telephone operator's either. They are both fluent native speakers. So there is evidently a difference, the inference will be, between the merely linguistic and the communicative aspects of the interchange. Effective communication requires the right linguistic basis: but mastery of the language is not a sufficient condition for effective communication. To take a somewhat different example, a letter in The Times of 3 October 1980 makes the following humorous complaint. Sir, The latest advertisement on the London Underground tells us that 'The last train leaves later than you think'. Why, in the face of such stiff competition, has the last train been selected for special mention?
Now according to some theorists, a reader of The Times who was equipped simply with a knowledge of the English language would be sorely puzzled to understand the point of this letter. Whereas a communicatively competent reader will immediately understand that what the writer is complaining about is the general unpunctuality of London Underground trains. On the same view, indeed, a person who was merely linguistically competent in English would have been baffled by the London Underground advertisement in the first place. So a fortiori he could not have understood the complaint. This example repays examining in some detail. In the first place it is very typical of the way in which modern advertising tends to take for granted that the consumer will grasp the point without having it spelled out. Secondly, the humour of the
Making Sense of Communicative Competence
29
complaint turns on pretending to take what the advertisement says in a way which is manifestly not what the advertiser intended, even though it appears to be in some sense a 'legitimate' interpretation. Thirdly, the example appears primafacie to support the necessity for drawing a distinction between communicative competence and linguistic competence; although, as will be argued shortly, when one looks at the case in detail one is forced to quite a different conclusion. The humour of the letter to The Times turns on the feigned disingenuousness with which the writer supposes that what the Underground advertisement says is that your idea of when the last train will leave is almost certain to turn out to be wrong, in view of delays that always postpone its departure. But, the logic of the objection goes, it seems that the departure of all Underground trains nowadays is delayed by something or other. So why make a special point of telling us that the last train will be delayed? This is the rhetorical question of the humourist. Whereas in fact we all know that the point of the advertisement was to draw attention to the fact that Underground services are scheduled to run at late hours. They do not close down as early as you might suppose. These conflicting interpretations of 'The last train leaves later than you think' appear to back up the notion that we need to distinguish between linguistic competence and communicative competence. For even if we did come across someone who interpreted the Underground poster in the naive manner mockingly supposed by the writer to The Times, it would somehow go against the grain to say that this naive individual just did not understand English. On the contrary, we would tend to suppose that this person certainly understands English all right. It is just that there has been a failure to grasp the communicational point of this particular advertisement. A third and somewhat different example is provided by a review of the novels of Thomas Hinde by Christopher Ricks. 1 Ricks draws attention to the fact that there is a textual difference between the ending of the novel Mr. Nicholas as printed in the 1962 Penguin edition, and as printed in the 1980 reissue. The book ends in the former case with the following exchange: 'Hallo Peter' 'Hallo Dad'.
In the 1980 version the exchange is the same, except that the word dad is printed in lower case letters, instead of beginning with a capital. At first sight, the difference appears trivial; but not so according to Ricks, who says: You would have to interpret 'Hallo dad' as a telling unique and utterly final act of tacit inner rebellion by this son who seethes with dismay at his cracked ogreish yet commonplace father who has just survived a suicide attempt. You cannot voice the difference between 'Dad' and 'dad', but you can think it, and in the recesses of your head you can withhold the customary respect that would upper-case your father. If it matters whether a poem by T.S. Eliot says 'Jew' or 'jew' although the poem cannot say them differently, it could matter exactly how a son is presented as addressing his father even if, or especially if the distinction would be smoulderingly meant to elude a father's ears.
1
London Review of Books, October 1980, vol. 2, no. 19, p. 6.
30
Roy Harris
This again is an interesting case. No one would claim that it is a generally accepted rule of English that Dad written with an initial capital letter means something different from dad written with a small letter. But if Ricks is right, a sophisticated novelist like Thomas Hinde might expect it to be within the communicative competence of his readers to grasp the significance of that distinction, at least in this context. As a matter of record, Ricks gives it as his opinion that the lower case version is probably a misprint. But for our present purposes whether it is a misprint or not is beside the point. A clear case of a misprint, however, provides a fourth and final example. When a newspaper article in the Business News section mentions the current rise in the price of 'zinz' (as The Times did on 20 July 1968), most readers have little doubt that this word is not the name of some mysterious commodity they have never before heard of, but that the writer is in fact referring to the price of zinc. It could hardly be claimed to be a rule of English, however, which tells us that 'zinz' should be interpreted as a misprint for zinc. Again, insofar as readers understand what the intended message was, it would appear that their communicative competence extends beyond their linguistic knowledge. For if the limits of their understanding were also the limits of their language, the non-existent English word 'zinz' ought in theory to bring about an automatic breakdown in communication. But in practice, as we all know, it need not. This connects with a point raised by Talbot Taylor (1982) concerning the ethnomethodological concept of 'repair' in conversation. 'Repair', as a technical term in ethnomethodological analysis, may be roughly translated as 'correction of mistakes'; and the signs of repair are usually taken to be such phenomena as pauses, hesitations, syntactic discontinuities, and so on. The question is: why are most repairs in conversation initiated by those who made the mistake in the first place? The explanation ethnomethodologists offer appeals to a complex system of conversational rules, governing the circumstances under which it is allowable for the hearer to initiate a repair. However, the low frequency of repair not initiated by the speaker himself might be explained simply by the fact that if the hearer understands what the speaker intended to say, then even if the speaker made a mistake, there is no need for the hearer to draw attention to it by initiating a repair. Unless he is a pedant, or wants to find fault with what the speaker said for some other reason, then there is every reason for the hearer to let the conversation proceed without a hitch. If my drunken host at a party pours me out a glass of sherry from a decanter clearly labelled 'sherry' and says simultaneously 'Can I offer you a Martini?', then if I want the sherry I can afford to ignore his mistake and just say 'Thank you very much'. It is only if I hate sherry and actually want a Martini that I have a communicational problem on my hands. In short, communicative competence, at least in certain types of case, involves not correcting verbal mistakes, but overlooking them. But knowing when it is appropriate to do this goes beyond just knowing the language: or so it would seem. There is no lexical rule or convention of English concerning the words sherry and Martini that tells us under what circumstances one may be substituted for the other without undue hindrance to communication. So there seem to be perfectly understandable reasons why, in the layman's view, it might make sense to draw a distinction between 'knowing a language' and
Making Sense of Communicative Competence
31
'knowing how to communicate'. That lay plausibility is, I think, one of the foundations that has supported the uncritical drawing of a theoretical distinction between linguistic competence and communicative competence. I leave open for the moment the question of whether we need to draw any theoretical distinction of the sort. I proceed first to argue that any such distinction, if required, has to be drawn somewhat less uncritically or less naively than can be justified merely by appealing to the kind of layman's problem that, reasonably enough, provides the material for adult education programmes on BBC television, or to the kind of literal-minded humour that one occasionally encounters in letters to The Times, or to the kind of typographical subtlety that one might sometimes be on the look-out for in literary prose or poetry, or to the fact that sometimes communication can be successful even when the language used is manifestly defective in some way.
Ill Although the term communicative competence as a quasi-technical term is used somewhat differently by different writers, it was originally defined in contradistinction to 'strictly' linguistic or grammatical competence. Psychological, linguistic and ethnographic approaches to speech communication all draw the distinction along much the same lines. For example, Campbell and Wales (1970) recognized three kinds of competence involved in people's everyday linguistic activities, of which communicative competence is only one. It is to be distinguished, they say, from a more general kind of competence that covers the whole range of skills and actions involved when we engage in linguistic performance. On the other hand, it is also to be distinguished from the more restricted kind of competence allegedly described by generative grammars. Campbell and Wales insist upon the point that communicative competence is a competence in its own right: what is involved cannot be relegated to studies of performance. It is 'the ability to produce or understand utterances that are not so much grammatical but, more important, appropriate to the context in which they are made' (Campbell & Wales 1970: 247; italics in the original). This, in their view, is 'by far the most important linguistic ability'; but the competence that generative grammar deals with fails to account for it. A similar distinction was accepted by theorists who advocated a 'dynamic paradigm' in descriptive linguistics. As Ch.-J.N. Bailey put it (Bailey 1973: 30): Those who limit attention to verbal language may wish to think of formal grammatical competence and pragmatic or functional competence as subsets of communicative competence. Formal competence refers to the substance of the grammar; pragmatic competence, to the life and use of the grammar.
In a much quoted passage, Hymes (1967) had drawn the distinction as follows: Linguistic theory treats of competence in terms of the child's acquisition of the ability to produce, understand, and discriminate any and all of the grammatical sentences of a language. A child from whom any and all of the grammatical sentences of a language might come with equal likelihood would be of course a social monster. Within the social matrix in which it acquires a system of grammar a child acquires also a system of its use, regarding persons, places, purposes,
32
Roy Harris other modes of communication, etc. - all the components of communicative events, together with attitudes and beliefs regarding them. There also develop patterns of the sequential use of language in conversation, address, standard routines and the like. In such acquisition resides the child's sociolinguistic competence (or, more broadly, its communicative competence), its ability to participate in its society as not only a speaking, but also a communicating member.
Now in one sense all this was by no means new. Already in the 1930s, J.R. Firth had insisted upon the importance of 'context of situation' for understanding how language works. He wrote (1937: 111) of conversation: It is important to realize that 'meaning' is just as much a property of the people, their 'sets', their specific behaviour, the things and events of the situation as of the noises made.
But the novel way of looking at the matter which was associated specifically with the term communicative competence depended upon presuppositions that Firth would almost certainly have rejected. The new notion of communicative competence was heavily indebted to certain assumptions that are explicable by reference to the stage that transformational theory had reached during the late 1960s. Two assumptions in particular stand out. One concerns the distinction between what belongs to competence and what does not, and the other concerns the role of language acquisition. To take the latter and less important point first, the equation between a child's learning of its native language and the task of the descriptive linguist, together with the assumption that linguistic theory must make it possible to explain how languages are learnt at the same time as analysing how languages are structured, had first been explicitly claimed only in 1965 in Chomsky's Aspects of the Theory of Syntax. Concerning this equation, all that need be said here is that it ignores so many obvious differences between the enterprise undertaken by the linguist and the enterprise undertaken by the child that any assimilation of the two is at best to be treated as metaphorical, if indeed it is coherent at all. The question of what is meant by 'competence' is of more importance. A competence is essentially envisaged as mastery or 'internalization' of a set of rules. Hymes's formulation makes it clear that his distinction between linguistic competence and communicative competence is itself parasitic upon the prior distinction between linguistic competence and linguistic performance. The distinction holds a central place in generativist analyses of language, which treat competence as mastery of the systems of syntactic, phonological and semantic rules that will generate a set of sentences. This immediately raises one problem in connexion with the notion of communicative competence. As communication in practice involves action in indefinitely variable sets of circumstances, with many different kinds of participants and many different kinds of purpose, and because success in communication depends on the contingent satisfaction of a variety of unforeseeable conditions on particular occasions, in what sense can it be claimed that communication is subject to the mastery of rules? The production and comprehension of isolated sentences is, when all is said and done, a relatively limited, specifiable, and homogeneous accomplishment that, at least in the context of Western educational tradition, has long been regarded as reducible to rules. On the other hand success in communication is a
Making Sense of Communicative Competence
33
much vaster, vaguer, more open-ended kind of accomplishment which it is hardly natural to think of in terms of rules at all. There is in any case a great deal that is confused and obscure about the notion of 'rules' in generative grammar; but for present purposes let us accept the generativist's talk of rules as a technically justifiable fagon de parler, descended historically from the traditional pedagogic so-called 'rules of grammar'. Continuing that fagon de parler and extending it to communicative competence, however, raises certain problems. One is the problem pointed out by Matthews (1979) concerning the boundary where, as it were, communicative rules (or 'rules of speaking', as he calls them) take over from rules of grammar. It is, for example, intrinsically unclear to what extent the intonation of utterances should be handled by rules of the kind incorporated in the phonological component of a grammar, and to what extent intonational features fall within the scope of rules of speaking. Another example Matthews cites is anaphora, where similar conditions seem to apply to anaphoric co-reference within sentences as in sequences of sentences. Yet if rules of grammar stop at sentence boundaries, as generativists maintain, it would seem that anaphoric relations beyond sentence boundaries have to fall under rules of speaking. But then it seems that certain rules of speaking will be simply rules of grammar under another name. It is no solution to give up sentence boundaries and extend the province of grammar more generously. For beyond the sentence there is simply nowhere else to stop. Whole conversations, whole novels would have to be generated by rules of grammar. And that, as Matthews remarks, 'merely reduces to absurdity the notion of performing a mental object'. Nor is it any better solution to move in the opposite direction and restrict the province of grammar. If anaphora, for instance, were subsumed entirely under rules of speaking, that would mean that a grammar could not account for certain quite conspicuous features of sentence structure. But if sentence structure is not the concern of grammar, what is? It is important to bear in mind that the kind of rules in terms of which linguistic competence is normally explicated are rules of a particular kind. They are part of an account of bidirectional (F <-> M) pairings that relate linguistic forms and linguistic meanings in such a way that, given the form of a sentence (F), it is possible to specify its meaning or meanings (M), and conversely given the meaning it is possible to specify the form or forms the sentence must take. In principle, there is nothing that confines a system of rules of this type to the description of languages. One could equally well use them for describing the operations of machines. For example, one could draw up a set of rules that would generate the operations of a vending machine, so as to specify all the possible outcomes of inserting various types of coin, and all the possible inputs required to produce a given item of merchandise. The various coin permutations would correspond to combinations of linguistic forms, and the various outputs of the vending machine would answer to the corresponding meanings. Now the point about envisaging communicative competence in similar terms is that simple bidirectional F <-> M pairings would not suffice, because there would now be not two but four elements to take into account: linguistic form, linguistic meaning, context of situation, and contextualized interpretation: (F, M, S, and I). The rules would have to relate these four elements in such a way that when F, M, and S were given I was specified, when I, M, and S were given F was specified, and so on. In other words,
34
Roy Harris
the rules would need to generate the possible interpretations a communicatively competent person would assign to a given sentence in given circumstances, and likewise the possible sentences such a person would produce in order to convey a given interpretation in given circumstances. Such pairings could be made bidirectional by treating F plus S as one unit and relating each such unit to a corresponding I (F & S <-» I) but that would be a different set of rules from that giving the F <-» M pairings for the language. It is worth noting in this connexion that the possibility in principle of codifying rules of the kind just described was emphatically rejected by transformationalists of the early 1960s. For instance, Katz and Fodor (1964: 489) argued that there was no way of constructing a theory of 'setting selection' for speakers of a language, since 'there is no serious possibility of systematizing all the knowledge of the world that speakers share'. To this one might add: 'nor of the knowledge of the world they don't share, either'. In the time since Katz and Fodor adduced this objection, no generative theorist has been able to explain how it might be overcome. With this in mind, let us examine in more detail the hypothesized relationship between linguistic competence and communicative competence. A well known attempt to enlarge the scope of what falls under linguistic competence is Labov's proposal that grammars should incorporate 'variable rules'. This would allow for the fact that speech is not uniform throughout a community, but may vary according to age, sex, status, and other social factors. Such a grammar would generate sentences appropriate to a variety of contexts and, through the mechanism of variable rules, allocate the right kind of sentence to the right kind of context and speaker. To this proposal, however, there are obvious objections. Although the 'enlarged' competence captured by a system incorporating variable rules is presented merely as an extension of grammatical competence, it is plainly nothing of the kind as far as any individual speaker is concerned. That would be rather like supposing that knowing when and where it is socially appropriate to wear sandals is simply an extension of knowing how to tie one's shoelaces. The notion is absurd. In any case, it seems clear that variable rules could not in principle deal with more than a slender portion of all that is involved in turning someone from a linguistically competent into a communicatively competent speaker. To take up the example cited earlier, one might ask what kind of communicative rules there could be that would ensure that the communicatively competent person understood the London Underground advertisement and also understood the point of the complaint in the reader's letter to The Times. As a first step towards answering this question, it might perhaps be urged that at least there is no doubt that the communicatively competent person needs to know more than what the English sentence The last train leaves later than you think means. And if there is something more to be known, then either that knowledge is rule-governed or it is not. If it is not, then it must be just a coincidence that so many people appear to understand what the advertisers are telling them. Indeed, it would seemingly be rash of the advertisers to invest money in such an advertisement if they believed it to be merely a matter of chance whether readers of the advertisement knew how to interpret the message or not. It would be rather like putting up a yellow flag in your front garden, and saying
Making Sense of Communicative Competence
35
that you hoped people would know that it meant you had a very large ferocious dog that delighted in attacking unsuspecting passers-by. But what is open to dispute here is the basic assumption involved in claiming that there is more to be known than the meaning of The last train leaves later than you think. The case for postulating that communicative competence is something extra to mere linguistic competences goes something like this. All competent speakers of English know what The last train leaves later than you think means, regardless of whether they see it on a London Underground poster or not. In other words, there is a semantic invariant - the meaning of the sentence - associated with a formal invariant - the spoken or written words. The relation between the two is such that whenever the formal invariant is present the semantic invariant can automatically be identified by those who know the language. That, precisely, is what knowing a language involves - knowing which semantic invariants to associate with which formal invariants. Both form and meaning are invariant in the quite literal sense that they do not vary between one occasion and another, or from one speaker to another. That has been the position adopted in orthodox linguistic theory at least from Saussure down to the present day. So, the argument runs, the same sentence might be used in quite different circumstances. Suppose Smith believes that the last train from Paddington to Oxford on a Saturday evening leaves at 6.45 p.m., Jones may consult the timetable and correct Smith's misapprehension by saying 'The last train leaves later than you think'. This clearly, has nothing at all to do with London Underground advertisements. But it is the same sentence and, ex hypothesi, has the same meaning, granted that English is indeed both the language of the London Underground advertisement and also the language in which Smith and Jones are conversing. The role that knowing sentence meanings plays in communication is, according to the orthodox view, roughly this. The sentence meaning supplies a context-neutral basis, to which additional information is supplied by contextualized knowledge. Together these two sources of knowledge furnish all that is required for successful communication. It is rather like buying a ready-made suit from a department store. You may have to lengthen the sleeves or shorten the trousers. But these are minor adjustments. The basic tailoring job has already been done for you by the manufacturer, and it was already done before anyone knew that you were going to be the purchaser. Two problems naturally arise at this point. How can the descriptive linguist identify these ready-made meanings that allegedly attach to the sentences of a language? And how do the speakers of the language themselves ever come to learn what these ready-made meanings are? Let us for the moment take for granted that there is a formal invariant that competent speakers of English and descriptive linguists as well can identify as the particular word-sequence The last train leaves later than you think. The questions are: What is the semantic invariant that is associated with this word-sequence in English, and how do speakers ever get to know what it is? One thing that is for sure is that they could not acquire this knowledge by looking it up in some authoritative book, as they might for instance find out what the square root of 49 is by looking it up in the relevant mathematical tables. So they must get to know it in some other way.
36
Roy Harris
What ways are there? They might, perhaps, be told what the meaning is by some authoritative person whom they believe to know. But it seems very doubtful whether any significant proportion of those who read what it says on London Underground posters were ever specifically told, either at school or anywhere else, the meaning of that particular sentence The last train leaves later than you think. The only plausible alternative, if we believe in invariant sentence meanings, is to suppose that somehow most people got to know by working it out for themselves. Furthermore, we would have to suppose not only that all or most English speakers worked it out for themselves, but also that they all arrived at exactly the same conclusion. For otherwise the sentence meaning would not be a semantic invariant. It would vary from one speaker to another. But now how could speakers all arrive at exactly the same conclusion about the meaning of this sentence? They could only do this by abstracting identically from all the particular communication situations in which they had encountered this sentence before, or encountered its constituent words and constructions before. But this would be a kind of collective miracle much more miraculous than if the whole population all voted for the same party at an election. That kind of miracle only happens in countries with political regimes most of us would not wish to be living under. Furthermore, achieving the linguistic miracle would be only a minor difficulty compared with the difficulty of explaining how each individual acquired the communicative competence to interpret correctly the particular instances that supplied the evidence for working out the sentence meaning. In other words, even were the collective miracle plausible, there is a circularity in the explanation. If understanding the London Underground advertisement involved a prior knowledge of the linguistic meaning of the sentence The last train leaves later than you think, how did any Underground user acquire that knowledge except by understanding prior situations in which the same or relevantly similar words were used? But then how did he understand those prior situations without a prior knowledge of the linguistic meanings of those sentences? And so on adinfinitum. Envisaging communicative competence as linguistic competence plus something else makes it impossible to explain how either communicative competence or linguistic competence is ever acquired. On the other hand, if speakers manifestly exhibit the communicative competence necessary to interpret advertisements like The last train leaves later than you think, why is it necessary to postulate that this would have been impossible without knowing both a linguistic and also a communicative set of rules?2 To pursue the question further, one might ask what exactly the context of a typical communicative rule would include. Would the rules have to be so general as to apply equally well to such diverse communicational enterprises as ordering a meal in a restaurant, teaching mathematics to schoolchildren, and commiserating with a bereaved relative? Or, on the contrary, would each separate type of communicative activity call for a separate system of rules? In either case, there are problems. If the rules are general all-purpose rules, it is difficult to see how their content can be other than vague and abstract in the extreme. But if they are specific to different types of communicative activity, then communicative competence itself becomes simply a 2 For some incisive comments on the puzzle of how speakers ever 'come to acquire two different sorts of competence', see Matthews (1979: 86-8).
Making Sense of Communicative Competence
37
cover term for an unrelated series of disparate skills. For example, attempts have been made (notably by ethnomethodologists) to formulate rules for conversation, and this seems to be one kind of communicative rule Hymes had in mind, since the passage previously quoted refers specifically to 'patterns of the sequential use of language in conversation, address, standard routines and the like'. But however one judges the success of such attempts, what is clear is that the systems of rules ethnomethodologists have tried to formulate for conversation certainly would not apply - nor is it claimed that they do apply - to quite different communicational enterprises, such as a minister making a parliamentary speech, or an announcer reading the 9 o'clock news. Indeed, attempts to formulate rules for different types of communicational activity have a long history. In the fourteenth century, Lawrence of Aquilegia produced a complete set of sequential multiple-choice rules for letterwriting, which was intended for use by clerks and secretaries. This in turn was a development of the tradition of medieval rhetorical studies that goes back ultimately to Aristotle and beyond. But a major problem with communicative rules when envisaged in this way is that they fail to capture the central concept of communicative competence. That is to say, it is presumably no use internalizing different sets of rules for different types of communicative activity unless one has also mastered rules that determine which of the different sets is applicable in which circumstances. A politician who tries to apply the rules of conversation in a parliamentary debate, or an orator who produced a forensic speech when what was required was a funeral oration would be just as much a 'social monster' as Hymes's imaginary child who produces any sentence of the language at random. In the case of our London Underground example, it is quite obscure what would be meant by claiming that communicative competence involves mastering a special set of rules for understanding what you read on London Underground posters. Conceivably it might be argued that, for example, there is a rule to the effect that when it says 'The last train leaves later than you think', this is to be interpreted as 'The last Underground train leaves later than you think'. But this seems to be multiplying rules beyond all reason. For it is clear that each and every use of a phrase like the last train implies relevance to some contextually determined circumstance. If we have to postulate separate rules for every conceivable set of circumstances, then it is extremely difficult to see how we ever master the rules of communication at all. In short, once we envisage communicative rules as split up into quite different sets for different purposes, the essential part of the notion of communicative competence is paradoxically lost. For presumably the central thing that an ideally competent communicator must know is how to recognize the appropriate occasion for saying and understanding what. If he cannot recognize this, then he is not communicatively competent, regardless of how many different sets of specific rules he knows. This, then, is a major theoretical objection to such an interpretation of communicative rules: that they involve an explanatory regress. However many sets of rules are postulated, they are useless without some way of determining which set applies when. The problem is precisely analogous to the original problem of producing inappropriate sentences, which led to distinguishing between linguistic competence and communicative competence in the first place. One must conclude that if the point of postulating communicative rules was to avoid the absurdity of depicting
38
Roy Harris
the ideal language-user as a social monster, the move fails. Social monstrosity is not so easily put down. Taming social monsters by theoretically attributing communicative competence to them leaves us with the no less intractable difficulty as theorists - of making sense of communicative competence. To follow this line of thought to its obvious conclusion, one might next ask whether the explanatory regress could not in principle be terminated, by postulating a system of communicative super-rules that had the function, precisely, of accounting for the competent communicator's ability to make the right decisions about which particular set of communicative rules to apply. The snag with this suggestion is that it does not appear to offer a viable possibility. For the super-rules would, in effect, have to be part of some broader explanatory systematization of human social behaviour, and in the absence of any general theory of social rules it is difficult to see that this solution is anything more than an adjournment of the problem sine die. The explanatory regress remains: it is simply passed on to our descendants to resolve.
IV
One may conclude that as long as the notion of competence remains tied to the formulation of generative rules, there is a fundamental difficulty about extending that notion to cover communication in general. For successful communicative behaviour is situationally determined in ways that offer no analogy to the production of grammatical sentences. Unlike the production of grammatical sentences, it depends essentially upon interaction between two or more people, and it is simply unclear how any kind of behavioural adjustment involving two or more parties could be automatically achieved by following a set programme of generative rules. That observation applies not merely to language but to interactional behaviour in general. There are indeed accepted rules for many forms of interactional behaviour. Competitive games provide a much invoked and much abused example. But there are no systems of generative rules for winning such games. If there were, the games would lose their point; or rather they would cease to be competitive games. The task of devising a competitive game that could in principle be won by the mechanical application of a programme of generative rules admits no solution: for the specification of that task implies a contradiction. The same consideration applies, although perhaps less obviously, to forms of cooperative behaviour. There can be rules for this too, but again no fixed programme of generative rules that will automatically ensure success. Keeping in step, for example, as in a parade, might be regarded as subject to rules. But the only rules that govern keeping in step require the parties involved to synchronize the movement of their feet. In other words, these rules define what constitutes keeping in step. They do not generate programmes that, if followed mechanically by the participants, will achieve the required synchronization. For that, each party must adjust from moment to moment to the perceived movements of the others. A's keeping in step with B is dependent on B's keeping in step with A. No system of rules can in principle generate all and only the correct moves for either A or B to carry out separately, because the adjustments that may be required by unpredictable contextual factors cannot be foreseen. In this context, precisely, the potential behaviour of A on the one
Making Sense of Communicative Competence
39
hand and the potential behaviour of B on the other present two very relevant but quite unpredictable sets of contextual factors. This is why theorists who have attempted to specify in any detail rules for communicative procedures - as, for example, in speech act theory - have not tried to produce generative rules for bi-directional pairings analogous to those invoked to specify linguistic competence. Their model has been, rather, the rules governing games, which simply set out necessary and sufficient conditions applying to certain key features of the activity. For example, the rules of Association Football specify, inter alia, what constitutes a goal; but they do not generate a procedure for enabling you to score one except, self-evidently, that you must do whatever is necessary in the match you are playing to bring about the satisfaction of those conditions specified for a goal. Rules of the Association Football variety are sometimes called 'constitutive rules', and the formulation of constitutive rules to define various speech acts has been a serious pursuit of speech-act theorists. Now although sets of constitutive rules are different in kind from systems of generative rules for bidirectional pairings, the question is worth asking whether communicative competence could not be specified in terms of sets of constitutive rules, defining the totality of speech acts available to a linguistic community. (One reason for raising this question is that it has sometimes been suggested that the basic inventory of speech acts is universal, and that moreover the implementation of this basic inventory provides a common foundation for semantics across all known languages. In other words, it is a condition on all languages that they must in some form provide the semantic equipment required to implement certain fundamental types of speech act, because these acts are essential to human communication.) Let us then consider whether the concept of a set of speech act rules offers a viable alternative interpretation of communicative competence. It is certainly a less ambitious interpretation than postulating a system of generative rules that would relate specific sentences to the circumstances of their use. In other words, constitutive speech act rules do not attempt to spell out, for example, what actual words I have to utter in order to make a promise, or to apologize, or to insult you. A discussion of the rules might perhaps include typical examples; but the rules themselves do not attempt to specify all and only the forms of words that could be used to implement a promise, an apology, an insult, etc. They merely state certain conditions that have to be satisfied by the words so used, and these conditions usually relate not to the linguistic forms themselves but typically to the intentions of their users and to the circumstances in which they are used. Searle (1964) has made an important point of the difference between the notion of a rule as it applies to linguistic behaviour and the notion of a rule as it applies to certain other kinds of behaviour. Searle's argument may be summarized as follows. To perform a speech act is to engage in a 'rule governed form of behaviour'. But behaviour in general may be governed by at least two sorts of rule. One sort of rules are the rules that 'regulate antecedently existing forms of behaviour'. An example of rules of this sort would be rules of etiquette regulating interpersonal relationships. The point about these rules is that the interpersonal relationships 'exist independently of the rules of etiquette'. Whether or not it is polite in a certain society or social milieu for a man to take his hat off to a lady does not determine the interpersonal
40
Roy Harris
relationships between members of different sexes. Those relationships exist and will continue to exist whether or not the custom of hat-raising survives. On the other hand, there are also rules of a quite different sort that actually 'create or define new forms of behaviour'. The paradigm case here is that of games. Without the rules of chess, there would be no such activity as playing chess. Searle distinguishes between these two types of rule terminologically by calling the former 'regulative' rules and the latter 'constitutive' rules. The terminology is slightly unfortunate in that Searle claims that constitutive rules also have a regulative function. However, the important difference is that in the case of constitutive rules the existence of the activity is 'logically dependent upon the rules'. Having distinguished these two types of rules, Searle claims that the kind of rulegoverned behaviour we find in speech acts involves constitutive rules, and he proposes that by examining various types of speech act it will be possible to determine what the constitutive rules governing that speech act are. Promising will have different rules from apologizing; apologizing different rules from warning, and so on. Thus the possibility is opened up that a very wide and socially important gamut of linguistic behaviour could come within the scope of sets of rules of this kind. Moreover, this linguistic behaviour is essentially communicative behaviour, as Searle recognizes in observations like the following: 'In speaking a language I attempt to communicate things to my hearer by means of getting him to recognize my intention to communicate just those things.' For example, Searle's rules for promising concern the use of what he calls the promise 'function-indicating' device P. There are five ordered rules, as follows. Rule 1. P is to be uttered only in the context of a sentence (or larger stretch of discourse) the utterance of which predicates some future act A of the speaker S. Rule 2. P is to be uttered only if the hearer H would prefer S's doing A to his not doing A, and S believes H would prefer S's doing A to his not doing A. Rule 3. P is to be uttered only if it is not obvious to both S and H that 5 will do A in the normal course of events. Rule 4. P is to be uttered only if S intends to do A. Rule 5. The utterance of P counts as the undertaking of an obligation to do A.
Now these rules do not at any point actually tell us what form of words P might take. This in itself seems to leave an unsatisfactory gap in the account of communicative competence, because it appears to allow that the competent communicator might not in fact know which particular forms of words could be used to implement a given speech act and which could not. And that is precisely something that is going to affect successful communication. Supposing I have to make an apology in Spanish, it is not much use my knowing merely the general conditions that have to be satisfied for apologies, if I am going to put my foot in it by employing, say, the literal translation of an English apology formula that turns out actually to be an idiomatic insult in Spanish. This gap is plugged in Searle's analysis by appeal to a quite different set of rules, 'the semantical rules of the dialect spoken by S and //'. In other words, the rules for the 'function-indicating' device P depend on the condition that 'the sentence uttered is one which by the semantical rules of the language is used to make a promise'. The
Making Sense of Communicative Competence
41
trouble with this is that now we have come full circle. Making a promise was precisely the notion we originally set out to explicate. But this is not the only problem. What little empirical research has been done on speech act identification seems to cast doubt on the likelihood that there is an underlying inventory of speech acts common to all cultures. Kreckel (1981) points out that the identification of speech acts seems to be bound up with such specific factors as membership of a particular family. This is a claim that indeed appears to be supported by everyday experience. From our personal acquaintance with particular individuals, we are often able to recognize something they say as an apology, or an insult, or an objection, in cases where that interpretation would not necessarily be obvious to an outsider - that is, to someone who did not know the person intimately. If this is so, then clearly it renders impossible in principle the task of specifying all and only the forms of words utilizable as apologies, insults, promises, etc. Communicative competence simply fragments, along this line of analysis, into the detailed items of knowledge associated with membership of particular groups and relationships between particular individuals. Thirdly, the interpretation of communicative competence in terms of speech act rules encounters exactly the same objection as its interpretation in terms of generative rules; namely, it fails to capture the fact that a crucial factor in communicative competence would presumably be knowing how to recognize which set of rules applies to a particular occasion. The arguments previously outlined carry over mutatis mutandis to this alternative interpretation. It is difficult not to conclude, therefore, that the concept of communicative competence so far lacks any viable interpretation in terms of rules. This is not of course to claim that there are no rules operative in communication situations. It is simply to deny on the one hand that any such finite set of rules generates communicative competence in the way that some finite set of rules, allegedly, generates grammatical competence in a particular language; and on the other hand that a merely definitional set of speech act rules could adequately distinguish communicative competence from incompetence. Is there any better sense to be made of the concept of communicative competence if we simply abandon the attempt to define it in terms of rules? I am going to suggest that there is. But first of all I should like to present a quite different but no less fundamental argument that bears upon the distinction between linguistic and communicative competence. The thrust of this argument is that communicative competence as commonly envisaged is not merely an explanatorily redundant concept but involves a confusion between two quite separate ways of looking at linguistic abilities. In her book Children's Minds, Margaret Donaldson quotes the following autobiographical account of a child's first day at school from Cider with Rosie. I spent that first day picking holes in paper, then went home in a smouldering temper. 'What's the matter, Love? Didn't he like it at school, then?' 'They never gave me the present.' 'Present? What present?' 'They said they'd give me a present.' 'Well, now, I'm sure they didn't.'
42
Roy Harris They did! They said: "You're Laurie Lee, aren't you? Well just you sit there for the present". I sat there all day but I never got it. I ain't going back there again!' (Laurie Lee, 1965: 51).
Donaldson cites this misunderstanding as an example of a teacher's 'egocentric' use of language, showing a failure to realize 'what her words would be likely to mean to a small child' (Donaldson 1978: 18). Donaldson suggested two specific reasons for the communicational breakdown. One reason was that the child did not know - and the teacher forgot to consider that he probably did not know - 'that school is not a place where one normally gets presents'. The second reason was that the child probably did not know the 'idiomatic, adult meaning of the words "for the present'". The first reason sounds as if it relates to communicative competence, whereas the second sounds as if it relates to linguistic competence. In other words, 'that school is not a place where one normally gets presents' is presented as just a fact of life which the child has not yet learnt. The meaning of the expression for the present, on the other hand, is presented as a fact about the English language, in which the child has not yet acquired a full adult's competence. Ignorance of the fact of life and ignorance of the fact of language combine to explain the communicational breakdown in the following way. The child's ignorance of the kind of thing that normally happens at school prevents him from realizing that the words just you sit therefor the present, when uttered by a schoolteacher, are very unlikely to mean 'just you sit there and you will be given a present', even though they could in principle mean that. At the same time, the child's ignorance of another meaning of the words for the present deprives him of the opportunity of seeing an alternative interpretation, which he could otherwise have considered and, had he known what kind of a place school is, judged to be more plausible. According to this interpretation, therefore, the child's expectation is quite reasonable, given his ignorance of the world and his ignorance of the language. The teacher's expectation, on the other hand, is unreasonable, because it failed to take into account, first, that the only source of disambiguation for the remark - namely the significance of its occurrence in a school context - is highly unlikely to be available to a child on his first day at school; and, second, that a child of that age is unlikely to know the expression that is necessary to give the remark its intended interpretation. In short, the teacher is linguistically competent, but communicatively incompetent; whereas the child, through no fault of his own, is both linguistically incompetent and communicatively incompetent as well, at least in respect of the situation with which both find themselves confronted. One might, however, analyse the communicational breakdown in somewhat different terms from those Donaldson chooses, which already to some extent beg certain crucial questions concerning the communicational process. The extent of the question-begging becomes evident as soon as attention is drawn to the consideration that in the event the child would still have been disappointed even if his assumptions about the kind of thing that normally happens at school had been correct. For that would not have altered the fact that on this occasion the teacher did not intend the words just you sit therefor the present to be understood as holding out the implicit promise of a gift. In fact, the general likelihood of receiving presents at school is
Making Sense of Communicative Competence
43
neither here nor there. What matters is that the child understood the teacher to have referred to a forthcoming present, however likely or unlikely, usual or unusual, presents at school might in general be. For that was the only interpretation of her remark available to him, if indeed he had never encountered the phrase for the present meaning 'for the time being'. In Searle's terms, the teacher had unwittingly executed all the linguistically appropriate procedures for promising, in conformity with the semantical rules of the pupil's dialect. According to this alternative analysis, then, the question of contextual disambiguation is an irrelevance: it was not the child's lack of contextual knowledge that occasioned the misunderstanding, nor the teacher's taking it for granted. The source of the trouble was a misleading isomorphism between the teacher's language and the child's. The teacher was using words by which a speaker may express the instruction to sit in a certain place for the time being (sit there for the present); whereas the child assigned a different interpretation to the same sequence of words. In short, an explanation that needs to appeal to two different competences, linguistic and communicative, is redundant. To put the point more sharply, we must be careful not to confuse two different approaches to describing communication, and hence be misled into factorizing out two separate components of misunderstanding. In this example there is only one source of misunderstanding, although it may be seen from two different points of view. Given prevalent patterns of behaviour in British schools, the intended interpretation of the words sit therefor the present as an instruction to the addressee to sit in a certain place for the time being may be seen as considerably more probable than the alternative. But this is a probability that depends on looking at communicational patterns over the linguistic community as a whole; for it is only at that level that it becomes meaningful to assign weightings to different interpretations on the basis of generally relevant contextual factors. In any individual instance, however, the general relevance of the contextual factors in question may easily be overridden by some special circumstance. In language as in all else, probabilities are only probabilities, and no more. What makes no sense whatever is to introduce concepts derived from an analysis of linguistic-behaviour-in-the-mass into the explanation of an individual misunderstanding, where it is the viewpoint of the participants that alone determines the outcome. To say that the child was ignorant of how unusual it is to get presents at school, and hence how unlikely it was that his interpretation of what the teacher said would turn out to be correct, is doubtless true but beside the point. If he had known, it might have made a difference, but only 'might have'. For we often cling to expectations that are very unlikely to be fulfilled, provided we find them sufficiently attractive. To see the child's communicative competence as inadequate to the task in question involves a conceptual muddle between two quite different perspectives on language. This point is important, inasmuch as the perspective from which we endeavour to analyse and explain linguistic phenomena very much determines the limits of the kind of analysis and the kind of explanation for which we can reasonably hope. One of the major criticisms that might be levelled at the current generativist model of linguistic competence is precisely that it makes it extremely difficult to link up in any plausible way our alleged knowledge of a language and our ability to utilize that knowledge for
44
Roy Harris
communicational purposes. This difficulty arises basically because generativist theory, like structuralist theory before it, insists on driving a wedge between languages and language-use. Or, to put it another way, it takes as axiomatic an absolute distinction between linguistic knowledge and knowledge of the world. Its model is a simple bi-planar model. In the Saussurean version, signifiant is matched with signifie. In the generativist version, 'phonetic representations' are matched with 'semantic representations'. In both cases, the link is always between sounds on the one hand, and ideas or concepts on the other. It is a model which has no room for the representation of the outside world, of things, of circumstances, of contexts of use. It treats language as if it existed in some kind of neuro-psychological vacuum. In short, it is a model that fundamentally misrepresents our experience of language; and linguistic theorizing based upon misrepresentation of experience cannot ultimately guide our enquiries about language in the right direction. This misrepresentation cannot be corrected simply by adding on some pragmatic or referential or contextual component, so long as the basic model remains essentially biplanar, and any additional components are treated as mere frills. That is one reason for preferring to adopt a radically different basis for linguistic theory, which I have elsewhere called 'integrational' (Harris 1981). The term 'integrational' is intended to allude to the fact that in real life, as we all know, experience is not neatly compartmentalized into the linguistic and the non-linguistic. The two are integrated. Words are not separate from situations: they are part of the situations, both socially and psychologically. Futhermore, without that essential integration, we could neither learn a language, nor function efficiently as language-users. And it is precisely this integration which, it seems, the notion of 'communicative competence' is an unhappy attempt to come to terms with. But it is a self-defeating attempt because it merely treats communication as something 'extra' to languages and thus presupposes that very separatism that an integrational approach must deny. The plain fact is that no rule-based model of any kind so far proposed can come to terms with the real-life flexibility of communication situations and communicational processes. Nor is there any likelihood that such a model will be forthcoming. If there were, communication would be a radically different kind of enterprise from what it is and what we, as lay communicators, know it to be. Let me conclude by suggesting that there might be something to be gained by dropping the term communicative competence, which has too many misleading theoretical overtones, and substituting for it communicational proficiency. By communicational proficiency I mean simply the ability to cope with the communicational demands and opportunities that situations present us with. What those demands and opportunities turn out to be will depend very much upon the stable cultural components of different civilizations at different times, but also upon the unforeseeable specifics of particular circumstances. There is certainly every reason to believe that communicational proficiency depends on past experience; but there is no reason to believe that what will count as communicational proficiency can be precisely defined in advance of the situations that call for it. And if it cannot be defined in advance, then it cannot be legislated for, either by societies or by linguistic theorists.
Making Sense of Communicative Competence
References Bailey, C.J.-N. (1973) Variation and Linguistic Theory. Center for Applied Linguistics, Arlington, Texas. Campbell, R. & R. Wales (1970) The study of language acquisition. In J. Lyons (ed.) New Horizons in Linguistics. Penguin Books, Harmondsworth. Donaldson, M. (1978) Children's Minds. London. Firth, J.R. (1937) The Tongues of Men and Speech, ed. by P. Strevens. Oxford University Press, London. Harris, R. (1981) The Language Myth. Duckworth, London. Hymes, D. (1967) Why linguistics needs the sociologist. Social Research 34, 632-47. Kreckel, M. (1981) Where do constitutive rules for speech acts come from? Language & Communication 1, 73-88. Katz, J J. & J. Fodor (1964) The structure of a semantic theory. In J. Fodor & JJ. Katz (eds.)TTze Structure of Language: Readings in the Philosophy of Language. Prentice Hall, Englewood Cliffs, N.J. Lee, Laurie (1965) Cider with Rosie. London. Matthews, P.H. (1979) Generative Grammar and Linguistic Competence. London. Searle, J.R. (1964) What is a speech act? In M. Black (ed.) Philosophy in America. Cornell University Press, Ithaca, N.Y., and in J.R. Searle (ed.) The Philosophy of Language. Oxford University Press, Oxford. Taylor, TJ. (1982) Discontinuity in Conversational Speech. D.Phil, thesis, Oxford University.
45
This page intentionally left blank
PART TWO LANGUAGE AND THE LANGUAGE MYTH The papers in this section are concerned with investigating how 'languages' relate to 'language'. The integrationist contention is that orthodox linguistics misrepresents this relationship by relying on a popular but quite mythical account of how languages make linguistic communication possible. This is 'the language myth' of the Western tradition, which linguists in the twentieth century have adopted and adapted as a basis for their own theorizing. Ch. 4 ('The Fixed-Code Theory') analyses one component of the language myth: the assumption that a language, in virtue of its internal structure, establishes correlations between a set of fixed verbal forms and a set of fixed meanings. The combination of the two supposedly provides the community with a fixed code for public use. More specifically, the assumption is that 'it must be possible to say for any expression identified in the dimension of form [. . .] what it corresponds to in the dimension of meaning'. Understanding what another person says involves matching up the linguistic forms that person produces with the meanings the communal code provides. The code thus provides not only a communicational resource for members of the community but also, simultaneously, a standard to which their individual linguistic behaviour is accountable. Some implications of this assumption as it applies in linguistics, philosophy and the law are examined. Ch. 5 ('A Few Words on Telementation') analyses the complementary component of the language myth: the idea that we communicate (linguistically) by transferring the thoughts in our own mind to someone else's, this allegedly being possible because we share with the other person use of the same fixed code. The assumption is that 'when A speaks to B the same idea that A used and encoded into speech is picked out, highlighted, or re-created in B's head'. The notion that this mythical process, telementation, actually occurs every time people communicate is, for the integrationist, 'false, unprovable or irrelevant, and perhaps all three'. It is in any case useless as a basis for serious linguistic work. Ch. 6 ('The Dialect Myth') shows how the language myth in its fully elaborated form (i.e. a telementational view of communication, plus a fixed
47
48
PART TWO: LANGUAGE AND THE LANGUAGE MYTH
code to use for the purpose) has consequences for the way in which dialects and dialectal variation are treated by dialectologists. Three principal concepts of what a dialect is are found in the Western tradition. Orthodox linguistics has opted for a version based upon the postulate that linguistic geography must, at some level of regional or local circumscription, reveal areas where a uniform linguistic system is in operation, to which all speakers within that particular geographical circumscription have equal access. Ch. 7 ('Integrating Languages') sets out an integrationist alternative to the position represented by the language myth. It argues that 'languages' have to be seen not as fixed codes given in advance of communication, but as constructs we derive by generalization from our communicational experience. The crucial idea that gives rise to these second-order constructs is that utterances are repeatable. This idea may be elaborated in various ways and the construct which results may, in the right circumstances, become institutionalized and treated as 'the' language of a community. The construct may also become formally codified in various ways. Attention is drawn to the important role that writing plays in this process in literate societies.
48
The Fixed-Code Theory NIGEL LOVE
Anyone interested in how language is used to communicate must attend to the implications of the fact that certain kinds of language-use may enshrine, and be in part constituted by, a preconceived answer to that question. One way of identifying the preconceived answer to be discussed here is to consider what moves are allowed, and what moves are forbidden, in certain widely played linguistic language-games. Here are three examples. In his book Grammar, F. R. Palmer (1972) observes that the correct grammatical description of a spoken language may differ from that of its written counterpart, and illustrates the point with reference to the morphology of the English noun: If we say the postman came up the street there is nothing to show that it is different from the postmen came up the street. How many postmen? One or more than one? We do not know; the form is the same, phonetically [pousman], in both cases. Of course we CAN provide a pronunciation that makes the difference, a 'spelling' pronunciation; but such a pronunciation is an artificial one used only where there is pressure to resolve ambiguity. To point to this when discussing the spoken form is cheating - it is assuming that the speech must reflect the writing, and with that assumption, of course, the spoken language cannot have a different grammar. (Palmer 1984 [1972]: 31)
What Palmer refers to as a 'spelling pronunciation' is the possibility of resolving the ambiguity of, for example, 'did the ['pousmsn] come this morning?' by replying with something like 'yes, in fact the [pous'men] came', where the unreduced vowel of [pous'men], accompanied by a shift of stress to the second syllable, unequivocally indicates plurality. What requires comment is the contention that such pronunciations are 'artificial', and that to take them into account in a description of the spoken language would be 'cheating' - that is, making an impermissible move in the language-game of stating the grammar of a language. What exactly does Palmer wish to exclude from his description of English noun plurals? Some of the prima facie relevant facts are these. Unstressed vowels in English are commonly subject to reduction to [a]. Compound nouns in -man, such as postman, milkman, boatman, have two relevant features: they are usually stressed on the first syllable, and, provided that the vowel of the second syllable is unreduced, 49
50
Nigel Love
their plurals may be said to be formed by a change in the quality of the vowel in that syllable. These nouns fluctuate between reduced and unreduced vowel in the second syllable in both singular and plural forms in response to a number of factors, such as frequency of use, style of speech, the need to avoid ambiguity. When the vowel of the second syllable is reduced, the singular and plural forms merge. Words like postman and milkman commonly have the reduced vowel, and thus homophonous singular and plural forms: first, because they refer to phenomena often encountered singly, so that occasions on which there might be doubt as to how many were being referred to are likely to be few; secondly, because they are such common words that understanding them does not depend on the perspicuity of their morphological structure. But consider the possibility of doorstep-delivery by tradesmen of other monosyllabically-named articles. If it became fashionable to sell eggs or tea in this way, it might be that new compounds eggman and teaman would come into use. But at first one would expect them to be pronounced exclusively with unreduced vowels, since at this stage the comprehensibility of the word would depend, in part, on the transparency of the compound. The pronunciations [egman] and [timan] would arise only when eggmen and teamen had become established. But this is not to say that there would be a definite time at which one pronunciation dropped out and the other took its place. In some styles of speech the 'full' pronunciation might never be lost; in all styles it would always be latently available for such purposes as ad hoc circumvention of a communicationally awkward case of homophony. All these points appear to be relevant to an account of the singular and plural forms of these words. But all of them, to judge from the remarks quoted, seem to lie outside the scope of 'grammar'. The second example has to do with describing the semantic structure of a language. The ground-rule of this game is that the 'literal meaning' of a sentence is to be distinguished from the role that utterances of it may play in particular communicational exchanges. In one version of the game it is further claimed that this literal meaning of a sentence can be ascertained by determining the set of propositions that are entailed by it. For example, it is argued that, granted the truth of (some utterance of) my son threw a brick at the window, one cannot consistently deny the propositions 'I have a child' and 'my child threw a brick' (Smith and Wilson 1979: 148); or that, given the truth of some utterance of John stole three horses, 'John did something to three horses' must also be true (ibid.: 159). Smith and Wilson are apparently content to determine entailment relations by a process of introspection: other linguists have treated their intuitions about entailments as empirical hypotheses subject to possible disconfirmation by informants' responses in tests of this kind: X: Someone killed the Madrid chief of police last night. Y: The Madrid chief of police died last night. Instructions: Assuming X is true, judge whether Y is true or not. If you think Y must be true, write YES. If you think Y cannot be true, write NO. If you think Y may or may not be true, write YES/NO. If you don't know which answer to give, write?
The Fixed-Code Theory Results: YES NO YES/NO ?
51
96% 0% 3% 1%
(Leech 1974: 92; 1981: 82)
In this language-game, it is open to the players to challenge the viability of the particular examples offered. Thus it might be observed that the idea that 'I have a child' and 'my child threw a brick' are entailed by my son threw a brick at the window will strike as dubious those speakers of English for whom my child primarily means 'my juvenile offspring' rather than just 'my offspring'. Similarly, in the case of John stole three horses, which is said to entail 'John did something to three horses', one might object that there are ways of stealing horses that do not obviously involve doing anything to them at all - illegitimately altering the provisions of a will, for instance. At this stage several moves are available. One is to attempt to patch up one's examples. In the last case mentioned, it might be argued that the entailment holds so long as 'doing something to' horses is not understood to imply some activity directed at them physically, but in a wider sense. Another move is simply to seek less contentious examples. 1 Or thirdly, one might diagnose the problem here as having to do with excessive reliance on introspective judgments, and attempt to put the whole enterprise on a sounder empirical footing. This is the point of Leech's tests with native-speaker informants. A captious critic might still be inclined to point out that, in the case of the example given, if the killer of the Madrid chief of police happened to use a slow-acting poison, then not all English-speakers will feel obliged to concede that the killing and the death must be simultaneous. But Leech is presumably committed by faith in his empirical method to discounting such an objection as based on a misuse of English; for has it not been established, as a matter of empirically tested fact, that to say that X was killed last night implies that X died last night? And that is where this language-game ends. It would simply be 'cheating', to borrow Palmer's word, to point out that it is not just these particular examples that are contentious, that in any given case there will be some particular sense in which one is required to take the form of words held to entail another form of words in order to see that the one does indeed entail the other; in short, that the relation between propositions and the forms of words that express them is not such that the former can be read off directly, in the absence of any context of utterance, from the latter; and that in responding to Leech's test his informants are not telling us anything about the meaning of some English utterance or class of utterances (for there is no actual utterance or class of utterances in question here), but demonstrating their ability 1 Some of the changes in the second edition of Leech's book have presumably been made for this reason. For example, in a discussion of the question of what properties of an entity should be included in an account of the meaning of the corresponding word, Leech observes (1974: 88) that one way of avoiding certain problems is to refuse 'to anatomise the meaning of elephant any further than to define it as "an animal of the species elephant"'. In the second edition (1981) the corresponding passage (p. 84) substitutes dog for elephant as the example. No doubt this substitution was made because Leech remembered, or had it pointed out to him, that living elephants belong to two different species in two different genera. Whether dog circumvents problems of this kind is doubtful. Is it the domestic dog, or canids in general, that are in question? Since, in the abstract, we cannot tell what is in question, the contention that dog means 'animal of the species "dog"' must boil down to saying that dog means 'animal of the species "dog"' in those cases where dog means 'animal of the species "dog"'.
52
Nigel Love
(or, since their answer is arguably wrong, their lack of ability) to play the game of offering an opinion as to the hypothetical relations between forms of words, as considered in the abstract. But to say all or any of this is to go beyond the limits of the language-game in question. The third example is the philosophical problem of 'negative existentials'.2 Negative existential statements are what sentences of such forms as there is no such thing as X or X's do not exist or no such things as X's really exist are used to state. The problem with negative existentials is alleged to be that although some of them seem to be true - for example, carnivorous cows don't exist - it is hard to see how this can be so, given that to deny the existence of something we must indicate what it is the existence of which is being denied. This requires that it be referred to or mentioned: the negative existential statement must be about it. But things which do not exist, so the argument runs, cannot be referred to or mentioned. So if we can deny their existence, carnivorous cows must after all exist. The apparently true negative existential statement is thus either false or not really a statement about what it purports to be about at all; which is paradoxical, since carnivorous cows don't exist not only seems to be used to make a statement about carnivorous cows, but one that is, moreover, true. The next move in this game is to attempt to resolve the paradox. There are two main ways of doing so. The first way depends on drawing a distinction between 'being' and 'existing', such that a denial of existence is not tantamount to a denial of being. Thus, Russell writes: Being is that which belongs to every conceivable term, to every possible object of thought. . . Being belongs to whatever can be counted. If A is any term that can be counted as one, it is plain that A is something, and therefore that A is. 'A is not' must always be either false or meaningless. For if A were nothing, it could not be said not to be; 'A is not' implies that there is a term A whose being is denied, and hence that A is. Thus unless 'A is not' be an empty sound, it must be false - whatever A may be, it necessarily is ... Thus being is a general attribute of everything, and to mention anything is to show that it is. (Russell 1903: 449)
But, taken as denials of existence, not all negative existentials are necessarily false, since, Russell remarks, 'existence is the prerogative of some only amongst beings'.3 So the distinction between 'being' and 'existing' resolves the paradox of seemingly denying the existence of some nameable entity by claiming that the existence which is being denied in such cases is not 'existence' in a sense in which the denial would be paradoxical. The alternative move is to protest that carnivorous cows don't exist is not really about carnivorous cows at all. Negative existentials may seem to be about those things whose existence they deny, but this, so the argument runs, is a mistake arising from the misleading verbal form in which they are cast. Thus, Ryle says: Suppose I assert of (apparently) the general subject 'carnivorous cows' that they 'do not exist', and my assertion is true, I cannot really be talking about carnivorous cows, for there are none.4 2
The exposition of the 'problem' presented here is based on the discussion in Cartwright (1963). It is irrelevant to the present argument that 'two years later . . . [Russell] was very rude indeed about anyone who could hold such a doctrine' (Hacking 1975: 74). 4 Ryle here tries to have his cake and eat it. What is it that he 'cannot really be talking about' ? 3
The Fixed-Code Theory
53
So it follows that the expression 'carnivorous cows' is not really being used, though the grammatical appearances are to the contrary, to denote the thing or things of which the predicate is being asserted. (Ryle 1967 [1932]: 88)
Both of these strategies for resolving the paradox of negative existential statements are unsatisfactory. Exactly how existence is to be distinguished from being is far from clear. In a later paper, Russell suggests that 'inspection would seem to lead to the conclusion that, except space and time themselves, only those objects exist which have to particular parts of space and time the special relation of occupying them' (Russell 1904: 211). From which we are presumably invited to conclude that what distinguishes entities that have being but not existence from those which have both is that the former, unlike the latter, do not occupy 'particular parts of space and time'. But in what sense, in that case, can they nevertheless be said to be? The answer seems to be: in just the s^nse in which it is unparadoxical to say of, for example, carnivorous cows, that they do not exist. But how this is possible is what the distinction between being and existing is supposed to tell us: the paradox is resolved at the cost of substituting for it a terminologically induced mystery. Moreover, it is clear from Russell's remark that 'existence is the prerogative of some only amongst beings' that although to be is not necessarily to exist, it is not envisaged that there may be entities which exist without being. But this seems incompatible with the dictum that 'being belongs to whatever can be counted' - at least, if that dictum is construed as stating a necessary condition of being. For, quite apart from the problem of deciding what language or type of language the counting is supposed to be carried out in (if 'one'-'many' will do as a complete counting system, the class of entities to which being is ascribed will turn out to be different from the result obtained if the counting has to be done in, say, English), it leads to the peculiar conclusion that, for example, gold exists, but is not (except as something other than gold), since, although gold is countable (as one of the metals, for instance), it is not countable as gold. 'How many (of) gold are there?' can only be understood to mean 'how many objects, degrees of purity etc., of gold are there?'.5 The alternative strategy - denying that sentences which seem to be about carnivorous cows are really about carnivorous cows at all - can be dealt with very briefly. For this contention is simply false. If it were not, there would be nothing bizarre about this dialogue: A: 'Carnivorous cows don't exist'. B: 'What was that you said about carnivorous cows?' A: 'I didn't say anything about carnivorous cows; I said that they don't exist'. The next move in the game is either to refute the foregoing arguments against one or other of the two previous moves, thereby re-establishing the status quo ante, or to invent yet another resolution of the paradox. But it is of less interest in the present context to continue with the game than to suggest that there was never really any 'paradox' here in the first place. Let it be granted that expressions like carnivorous cows don't exist have a role in ordinary English usage. Now that role is clearly not to state paradoxes, but, on the contrary, to communicate information which may be true or false. Nor are they commonly felt, in ordinary usage, to have the slightest tinge of the paradoxical about them. In this they differ from expressions which we may take to be genuinely 5
This last point I owe to Professor R. Harris.
54
Nigel Love
paradoxical, such as carnivorous cows eat only grass. Nobody ever feels obliged to preface an utterance like 'carnivorous cows don't exist' with an apology ('It may be paradoxical to say so, but . . .'). That is, the assumption which gives rise to this whole philosophical issue - that true negative existential statements embody a paradox which stands in need of resolution - is an assumption foisted on them by philosophers rather than intrinsic to their use in the language.
II
What do these various language-games have in common? In their superficially different ways they incorporate a clear and consistent view of what, for purposes of theorising about language, a language is and is not taken to be. First, it is as clear as it is on the face of it surprising that a language is not taken to be essentially, or even primarily, a means of communication. For it is hard to see how any of the language-games discussed here could ever have been taken seriously if communicational relevance had been considered a matter of any importance. Let us look at each in turn in this light. The grammarian's insistence that the pronunciation [pousmen] is to be discounted as evidence bearing on a statement of the plural form of postman derives from his view that it is a pronunciation that only occurs in certain special contexts: where the speaker is aware that the form with reduced vowel ([pousman]) would fail to meet his communicational need - that is, when he specifically wants to make it clear that he is referring to more than one postman, and where there are no clues such as the contiguous occurrence in his discourse of words like 'two' or 'lots of. (Consider the absurdity of this dialogue: A: 'Did the postman come this morning?' B: 'Yes, in fact the postmen ([pousmsn]) came', where B intends to observe that more than one postman came, and where he uses what Palmer identifies as the canonical plural form.) Thus forms whose occurrence is to be explained in terms of particular communicational needs seem to be forms which are irrelevant to the grammarian's descriptive enterprise. Similarly, the expression my son threw a brick at the window has a wide range of potential communicational significance, depending on the context of utterance (the speaker might, in uttering it, be lying, or expressing pride, or astonishment, or apology; he might be quoting from a novel, or illustrating the thesis that sentences have meaning in the abstract; he might be referring to a man of forty or a babe in arms, who could be the speaker's son in the sense that (a) he fathered him, or (b) adopted him, or merely (c) married his mother; the brick might be a real brick or a toy; it might or might not be implied that the window was hit by the brick; the throwing of the brick might be metaphorical; and so on). But our inability to say precisely what might be communicated by my son threw a brick at the window is taken as no hindrance to the enterprise of determining, in the abstract, the meaning of that expression. Nor, thirdly, is the fact that there is nothing in the least communicationally vicious about expressions like carnivorous cows don't exist counted as a reason for not embarking on a discussion of the 'paradox' they are held to embody.
The Fixed-Code Theory
55
What a language is taken to be, on the other hand, is a fixed code which, by relating entities in a dimension called 'form' to entities in a dimension called 'meaning', provides language-users with a means of transmitting and receiving thoughts. This is why the structuralist language-describer is reluctant to include within the scope of his description deviations from canonical forms brought about by awareness of communicational difficulties that might arise from use of the canonical form in particular circumstances. For if he were obliged to take account of such variants, the task of exhaustively enumerating or in some sense 'defining' a language's stock of form-meaning pairs would be impossible. Consider the apparently simple and mundane task of listing the plural forms of nouns in a given dialect of English. In the case of postmen we may start with the six variants ['pousman], ['pousmsn], ['pousmen], ['pousmen], [pous'men], [pous'men]. But further possibilities can readily be imagined. For instance, another version of our two-line dialogue might run: A: 'Did the [pousman] come this morning?' B: 'Yes, in fact, the [pousmgn]-plural came', where '[pousm9n]-plural' seems to be yet another plural form of postman. This open-endedness of possible forms is intolerable for the structuralist languagedescriber, for whom languages are essentially fixed codes. If a language is a fixed code relating forms to meanings, then it must be possible to say for any expression identified in the dimension of form (for example, the expression identified by the written sequence my son threw a brick at the window) what it corresponds to in the dimension of meaning. Hence the quest for a procedure whereby some entity might be uniquely identified as its meaning. Types of expression, identified on the plane of form, can be associated with types of function, on the plane of meaning. Expressions formally characterisable as consisting of a noun-phrase followed by a predicate are commonly used to give information about something referred to by the noun-phrase. In a large class of cases the something referred to is assumed or known to exist, as in philosophically uncontroversial expressions like Jersey cows give high-quality milk. Since nounphrases commonly have this function, and since the language is taken to be a fixed code, it is assumed that the utterance in English of a noun-phrase automatically conjures up an entity whose ontological status needs to be accounted for. Hence the urge to resolve the 'paradox' of expressions like carnivorous cows don't exist. Ryle concedes that the sense in which such quasi-ontological statements are misleading is not that they are false, nor even that any word in them is equivocal or vague, but only that they are formally improper to the facts of the logical form which they are employed to record, and proper to facts of quite another logical form. (Ryle 1967 [1932]: 90)
'Proper' and 'improper' here only make sense on the assumption that there is some basic or expected or given relation between the form of an expression and its meaning. So the apparently diverse language-games that have been considered here can be seen as different aspects of just one game, whose fundamental rule is: in your discussion of language, make sure that your manoeuvres are compatible with treating
56
Nigel Love
a language as a fixed code 6 whose properties are to be explicated in the abstract, without reference to the requirements of, or the uses to which it might be put in, actual instances of communication. The account of communication which accompanies the fixed-code theory is basically this. The speaker knows the set of forms comprised by his language, and the set of meanings that correspond to them. If confronted with, for example, the utterance 'I would like a nice juicy apple', he understands it in virtue of knowing the meanings of the forms a, apple, I, juicy, like, nice, would, and the meaning of the composite form consisting of those forms in the sequence I, would, like, a, nice, juicy and apple. Conversely, if a speaker of English wishes to express a desire for a nice juicy apple, he knows that one way of doing so is to utter that sequence of forms.7 Communication between speakers A and B of a language is seen as being possible because A and B both have access to the fixed code of correspondences between forms and meanings which constitutes the language. To communicate with B, A encodes his meanings in the appropriate forms. To understand A, B matches up A's forms with the corresponding meanings.
Ill
Anyone concerned to establish that the fixed-code theory offers the basis for an illuminating account of how language is used to communicate must be able to dispose of at least the following objection. If a language is a fixed code, the code must be fixed before the language can be used. But the fixing of the code (that is, establishing the agreement among users as to the relations between forms and meanings) must itself be done in a language. Any attempt to explain how fixed-code languages can come about thus either runs into an infinite regress or else requires the outlandish claim that although there must once have been at least one language which was not a fixed code, languages of that type have, for unexplained reasons, fallen out of use. As one writer has it: Language not having existed, man had to invent it. Not that the invention was planned: given that such planning would require a level of mental sophistication that only language makes possible, it could not have been. (Platts 1979: 1)
Not only are rebuttals of points of this kind hard to come by in the writings of fixed-code theorists, arguments as to why the fixed-code theory should be thought to be true in the first place are rarely offered. In many modern linguistic discussions, the view that 'language is readily isolatable from its users and characterisable as an independent and self-contained system' has the status of an 'absolute presupposition' (Moore and Carling 1982: 45). Proponents of the theory have consequently not felt obliged to do much more by way of fending off attacks than to point out that, as Dummett puts it, 'our natural reaction is that something along these lines must be right'. Dummett continues: 6 The 'fixed-code fallacy' is one of two components of what Harris (1981) has identified as the 'language myth'. 7 Whether this way of putting it corresponds to how knowledge of this sort might be represented in e.g. a generative grammar of English is not at issue here.
The Fixed-Code Theory
57
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? (Dummett 1982: 9)
The answer to Dummett's question is: yes, he may well be enlightened by adopting such a procedure, but in doing so, far from vindicating the fixed-code theory, he is merely indicating one source of its misleading plausibility. It is true that there are purposes for which it is expedient to treat words as 'having' meanings. One of them is the situation envisaged by Dummett, where what is required is the kind of information about a word provided by a dictionary. But how does the dictionarycompiler come by the information he offers as to the meaning of a word? In the case of a dictionary of last resort (that is, setting aside dictionaries based mainly on other dictionaries) the answer is: by deducing from (preferably, a large number of) contexts in which the word has been used the most plausible interpretation(s) to be assigned to it. The dictionary definition of most words is a summary of what, in the lexicographer's view, is common to the instances of their use in the data on which the dictionary is based. It is a retrospective account of what speakers must be presumed to have meant by a given word in the past. But it does not automatically determine what they will mean by it in the future. Dictionaries display information about past language-use by setting out a list of correspondences between forms and meanings. But that is no good reason for thinking that present language-use is a matter of implementing knowledge of those correspondences. However, it probably is a reason, if only a bad one; in which case, one way of trying to understand the appeal of the fixed-code theory as the basis for an explanation of the mechanics of language-use might be to consider why dictionaries take the form that they do. More generally, it might be fruitful to examine the role of the fixed-code theory in all those institutionalised communicational practices which seem to depend, whether justifiably or not, on taking it to capture a fundamental truth about language.
IV
One such body of practices, and the one which it is proposed to consider here, concerns the regulation of behaviour by enforcing conformity to codes of written laws. Customary procedures connected with the drafting and interpreting of statutes and other kinds of legal instrument enshrine the fixed-code theory in a number of respects. First, the very idea that the legislator's communicational intentions can be precisely captured, without reference to particular situations, by choosing in the abstract an appropriate form of words in which to embody them, implies a faith in the self-sufficiency of linguistic expressions as containers of stable meanings. Secondly, when there arise reasons for doubting whether a law has been infringed, of the kind that might be classified as doubts about matters of law rather than of fact (that is, doubts as to whether some act or omission to act would in principle constitute an infringement), such doubts are not resolved by an appeal for clarification to those who framed the law in the first place. In English legal procedure,
58
Nigel Love
lip-service is paid by appellate judges, when confronted by doubts of this kind, to the need to consider the intentions of Parliament; but this consideration never takes the form of actually asking Parliament what it meant. Instead, the demand is made of the form of words which constitutes the law that it yield a determinate answer to all such questions as may arise. Thirdly, the process of interpreting a written law may sometimes give a result which arguably or even manifestly runs counter to the legislator's intentions. But a conflict between what the legislator meant and what he can be held to have said by using the words that he used is invariably resolved in favour of the latter. The words in which a law is couched are taken in this sense to have a life of their own; they are made to speak for themselves, and what they are heard to say is given priority over the communicational intentions of their original user. It is hard to see how such procedures could become established in a society which did not take it for granted that what a language basically has to offer its users is a context-neutral code for the expression and extraction of determinate meanings. This schematic outline requires amplification. Let us, for illustrative purposes, concentrate on what it is theoretically, in English legal procedure, to interpret a statute.8 Three views of statutory interpretation have jockeyed for attention. They are known as the 'mischief rule', the 'literal rule' and the 'golden rule'.9 The classic description of the mischief rule is to be found in Heydon's Case,W where it was said that for the sure and true interpretation of statutes in general.. . four things are to be ... considered: 1st. what was the common law before the making of the Act, 2nd. what was the mischief and defect for which the common law did not provide, 3rd. what remedy the Parliament hath resolved and appointed to cure the disease of the Commonwealth, and 4th. the true reason of the remedy; and then the office of all the judges is always to make such construction as shall suppress the mischief, and advance the remedy, and to suppress subtle inventions and evasions for continuation of the mischief, and pro private commodo, and to add force and life to the cure and the remedy, according to the true intent of the makers of the Act, pro bono publico.
This approach to statutory interpretation arose before the sovereignty of Parliament was established, when judges still felt able to strike down Acts of Parliament because of their content; and it later gave way to the literal rule. The de facto demise of the mischief rule may be seen from the lack of success of an attempt to revive it in the middle of this century. In the case of Seaford Court Estates vs Asher,'^ Lord Denning had contended that judges should apply the principle enshrined in Hey don's Case: if a judge came across 'a ruck in the texture of the Act', although he must not alter the material of which the Act was woven, he could and should 'iron out the creases'. But in a case which came soon afterwards this was dismissed by the House of Lords as heretical. In this later case (Magor and St. Mellons vs Newport 8
The following sketch draws on Cross (1976). Their designation as 'rules' may be misleading: they are, as will emerge, broad principles rather than rules. 10 (1584) 3 Co. Rep. 7a. 1' [1949] 2 KB 482. 9
The Fixed-Code Theory
59
Corporation),12 Lord Denning repeated and expanded on his remarks in Seaford Court Estates vs Asher: 'We sit here to find out the intention of Parliament and of Ministers and carry it out, and we do this better by filling in the gaps and making sense of the enactment than by opening it up to destructive analysis'.13 But this was roundly condemned by the House of Lords: 'the general proposition that it is the duty of the court to find out the intention of Parliament - and not only the intention of Parliament but of Ministers also - cannot by any means be supported. The duty of the court is to interpret the words that the legislature has used'.14 Moreover, the idea that the court, having discovered the intention of Parliament, should proceed to fill in the gaps in what the legislature had written, was 'a naked usurpation of the legislative function under the thin guise of interpretation'.15 These dicta are prima facie difficult to understand, if 'finding out the intention of Parliament' and 'interpreting the words that the legislature has used' are to be construed as two quite unrelated activities. Since there is obviously a sense in which the legislative function is to enact laws embodying the legislature's legislative intentions, it would be odd if interpreting the laws were held to be an exercise which had nothing at all to do with discovering the intentions. Rather, 'interpreting the words that the legislature has used' is to be seen as the best available politically neutral way of discovering what the legislature intended. That is, on the assumption that Parliament has chosen its words with care, interpreting those words should lead the court to discover what Parliament intended; and in any case, whether that assumption is justified or not, to do other than interpret the words would be to usurp Parliament's role. This is the 'literal rule'. The literal rule is not without exception or qualification. Sometimes, statutory words may be inconsistent or contradictory. In such situations the so-called 'golden rule' has been invoked. This says that: in construing . . . statutes . . . the grammatical and ordinary sense of words is to be adhered to, unless that would lead to some absurdity, or some repugnance or inconsistency . . . in which case the grammatical and ordinary sense of the words may be modified, as to avoid the absurdity and inconsistency but no farther (per Parke B. in Grey vs Pearson).^
But the golden rule allows for only a limited departure from the literal rule: it is taken to sanction the 'modification' of statutory words only where there would otherwise be a logical contradiction.1? So the 'literal rule' is pre-eminent in statutory interpretation in English law. This is confirmed from consideration of the subsidiary rules concerning the use of 'extrinsic aids'. If the task of the courts were to decide what Parliament meant to say, rather than what it actually said, then reference to parliamentary debates, and the use of travawcpreparatories, would not be frowned upon. Yet they are. Parliamentary debates may not be judicially consulted at all;18 and preliminary reports by, for '2 [1952] A.C. 189. 13 [1950] 2 All ER 1226 at 1236. !4 [1952] A.C. 189 at 191. !5 ibid. « (1857) 6 H.L. Cas. 61. 17 See e.g. Christopherson vs Lotinga (1864) 15 C.B. N.S. 809 at 813; Cross (1976: 15). '» R. vs Hertford College, Oxford (1878) 3 QBD 693 at 707; Davis vs Johnson [1979] A.C. 317.
60
Nigel Love
example, royal commissions, may only be consulted to discover what problem the statute was designed to remedy. And this applies even if the report of the advisory body in question contains a draft bill, whose provisions were later enacted without alteration. !9 For 'Parliament, under our constitution, is sovereign only in respect of what it expresses by the words used in the legislation it has passed'.20 The two crucial assumptions underlying the literal rule of statutory interpretation are (1) that there will in general be a close connection between what Parliament means to say and what it actually says, and (2) that what it actually says is determinable, in the abstract, through contemplation of the words used. How do these assumptions fare in practice? What follows is a demonstration of the language-game of disputing the judicial interpretation of a statute. 21 What will be important in the present context is not the particular interpretation arrived at, but its status relative to other possibilities.
The first plaintiff was an orthodox Sikh, who sought to have his son, the second plaintiff, admitted to a private school. The first defendant, the headmaster of the school, refused to admit the boy unless he complied with the school rules, which required pupils to wear the school uniform, including a cap, and to have their hair cut short. But these rules run counter to the habit observed by strict Sikhs of keeping their hair unshorn and wearing a turban. The son brought an action through his father against the headmaster and the company which owned the school, alleging unlawful discrimination contrary to the Race Relations Act 1976, and seeking initially damages, but subsequently only a declaration that unlawful discrimination had occurred. The county court dismissed the claim on the ground that the Sikhs did not constitute a racial group within the meaning of the 1976 Act. The Court of Appeal, upholding the ruling in the county court, dismissed the plaintiffs' appeal - which was, however, allowed by the House of Lords. The main issue was whether the Sikhs constituted a racial group within the definition contained in s. 3 (1) of the 1976 Act, which runs: 'racial group' means a group of persons defined by reference to colour, race, nationality, or ethnic or national origins, and references to a person's racial group refer to any racial group into which he falls.
Since it was not contended that the Sikhs qualified in virtue of possessing any of the other characteristics specified in the statutory definition of a racial group, the crucial question was whether the Sikhs are a group of persons 'defined by reference to ethnic origins'. Although the rulings in the two higher tribunals were different, consideration of the reasoning that led to them shows that all the judges concerned subjected the 1976 Act to interpretations which stand in need of considerable comment. 19 Black-Clawson International Ltd. vs Papierwerke Waldhof-Aschaffenburg A.G. [1975] A.C. 591. 20 ibid, at 638. 21 The case discussed is Mandla vs Dowell Lee ([1983] 2 A.C. 548 (H.L.); [1983] Q.B. 1, (C.A.)).
The Fixed-Code Theory
61
Two speeches in the House of Lords are of interest. The relevant passage from Lord Templeman's speech runs as follows: In the course of the argument attention was directed to the dictionary definitions of the adjective 'ethnic'. But it is common ground that some dictionary definitions constitute the Sikhs a relevant group of ethnic origin whereas other definitions would exclude them. The true construction of the expression 'ethnic origins' must be deduced from the A c t . . . In my opinion, for the purposes of the Race Relations Act a group of persons defined by reference to ethnic origins must possess some of the characteristics of a race, namely group descent, a group of geographical origin and a group history. The evidence shows that the Sikhs satisfy these tests. They are more than a religious sect, they are almost a race and almost a nation. As a race, the Sikhs share a common colour, and a common physique based on common ancestors from that part of the Punjab which is centred on Amritsar. They fail to qualify as a separate race because in racial origin prior to the inception of Sikhism they cannot be distinguished from other inhabitants of the Punjab. As a nation the Sikhs defeated the Moghuls and established a kingdom in the Punjab which they lost as a result of the first and second Sikh wars; they fail to qualify as a separate nation or as a separate nationality because their kingdom never achieved a sufficient degree of recognition or permanence. The Sikhs qualify as a group defined by ethnic origins because they constitute a separate and distinct community derived from the racial characteristics I have mentioned . . . The Sikh community has accepted converts who do not comply with these conditions. Some persons who have the same ethnic origins as the Sikhs have ceased to be members of the Sikh community. But the Sikhs remain a group of persons forming a community recognisable by ethnic origins within the meaning of the Act.22
There are several points of interest here. First, it is odd that anyone should suppose that looking up ethnic in the dictionary could be of any help. The function of the dictionary definition of such a word is to specify its intension (when combined with a noun), not its extension. If you are doubtful, for instance, whether your domestic pet is a cat or a dog, looking up cat and dog in the dictionary is unlikely to suffice to settle the matter. Similarly, whether or not the Sikhs fall within the extension of the expression 'group of ethnic origin' is not to be deduced from what the dictionary says about the meanings of those words. Nor is it easy to see how (pace Lord Templeman) it is to be deduced from the Act. No amount of scrutiny of the wording of the Act will reveal whether the Sikhs are a 'relevant group of ethnic origin'. If the Act itself were capable of yielding this information, the matter would not have required a decision in the House of Lords in the first place - a point which Lord Templeman appears tacitly to concede, in as much as he offers no clue as to the grounds on which we are supposed to accept that the opinion he proceeds to give (that 'for the purposes of the Race Relations Act a group of persons defined by reference to ethnic origins must possess some of the characteristics of a race, namely group descent, a group of geographical origin and a group history') is derived by 'deduction' from the Act. Whatever Lord Templeman means by saying that a group of persons defined by reference to ethnic origins must possess, inter alia a group of geographical origin, it is not something said or implied by the Act. Moreover, if he means that, in order to qualify as a racial group, they must be a group of [one and the same] geographical origin (and share a common 'group descent' and 'group history'), it is unclear why he takes this status to be unaffected by the fact that 'the Sikh community has accepted converts who do not comply with these conditions'. If 22 [1983] 2 W.L.R. 620 at 631-2.
62
Nigel Love
these converts are, after conversion, rightly called Sikhs, then they are, presumably, no less than other Sikhs, persons whose characteristics are to be taken into account for purposes of deciding whether the Sikhs are a racial group. If some Sikhs lack the statutory earmarks of a racial group, it follows that the Sikhs are not a group defined by possession of those earmarks. Just as if a group of prime numbers is defined as consisting of numbers with no integral factors except themselves and unity, it follows that a group containing, for example, '6', is not a group of prime numbers. Lord Fraser, who gave the leading speech in the House of Lords, reasoned as follows: For a group to constitute an ethnic group in the sense of the Act of 1976, it must, in my opinion, regard itself, and be regarded by others, as a distinct community by virtue of certain characteristics. Some of these characteristics are essential . . . The conditions which appear to me to be essential are these: (1) a long shared history, of which the group is conscious as distinguishing it from other groups, and memory of which it keeps alive, and (2) a cultural tradition of its own, including family and social customs and manners, often but not necessarily associated with religious observance . . . A group defined by reference to enough of these characteristics would be capable of including converts, for example, people who marry into the group, and of excluding apostates . . . 23
Opinions will differ on whether this is a reasonable account of what an ethnic group is. One reason for thinking that it is not is that indefinitely many groups might be held to qualify in virtue of fulfilling conditions (1) and (2), some of which (for example, the British royal family) nobody would want to count as a racial group. However that may be, there is a more important objection to Lord Eraser's judgment. For he concludes: if the proper construction of the word 'ethnic' . . . is a wide one, along lines such as I have suggested, the Sikhs would qualify as a group defined by ethnic origins for purposes of the Act.24
Granted, for the sake of argument, that the Sikhs are an ethnic group in terms of Lord Fraser's definition, what is missing here is the reason for supposing that it follows that they are a group 'defined by ethnic origins'. The decision in the Court of Appeal ran counter to that reached in the House of Lords. That is, the Court of Appeal decided that the Sikhs were not a racial group within the meaning of the 1976 Act. But the reasoning in the Court of Appeal had in common with that in the House of Lords a certain lack of cogency. In order to elucidate the word ethnic, Lord Denning MR instanced the Jews as 'the best known example of "ethnic grouping'" in that 'when it is said of the Jews that they are an "ethnic group", it means that the group as a whole share a common characteristic which is a racial characteristic. It means that they are descended, however remotely, from a Jewish ancestor'. 25 This conclusion is puzzling, since Lord Denning has just said that '"a Jew" may mean a dozen different things. It may mean a man of the Jewish faith. Even if he was a convert from Christianity he would be of the Jewish faith'. But if, irrespective of his ancestry, a convert from 23 ibid, at 625. 24 ibid, at 627. 25 [1982] 3 A11ER 1108 at 1112.
The Fixed-Code Theory
63
Christianity can count as a Jew, it is not clear that the Jews are a group 'descended, however remotely, from a Jewish ancestor'. Much depends here on the function of the phrase 'as a whole' in 'the group as a whole share a common characteristic'. For Lord Denning may, by inserting this phrase, be intending to accommodate the possibility that some converts to Judaism may not be Jews by descent, while insisting nonetheless that the Jews 'as a whole' (= 'on the whole') are a group descended from a Jewish ancestor. This seems reasonable: a parallel might be the fact that the presence in universities of undergraduate students who draw their old-age pensions would not necessarily be taken to disconfirm the statement that undergraduates 'as a whole' are a group of people in their late teens or early twenties. Lord Denning appears to agree with Lord Templeman and Lord Eraser that possession of characteristics typical of a group need not be a necessary condition of membership. Reasons for thinking that this view, however acceptable in itself, is irrelevant to the interpretation of the statutory definition in question, will be offered below: what needs to be observed at this point is that if this is indeed what Lord Denning intends, it is rather unfortunate in the context of his judgment in the present case. For the fact that the Sikhs 'freely receive converts from Hinduism' is one that he cites as tending to show that the Sikhs cannot be defined by reference to their ethnic origins. If the Jews are a racial group by virtue of a common ethnic origin, despite the existence of Jews who do not share that origin, it is odd to adduce the fact that there may be Sikhs who do not have the same ethnic origin as other Sikhs as a reason for not counting the Sikhs as a racial group. In the House of Lords, Lord Fraser apparently supposed that, for purposes of establishing whether a group was a group defined by reference to ethnic origins, it sufficed merely to show that it was an ethnic group. In order to arrive at this judgment he was obliged to pass over in silence the opinions given in the Court of Appeal by Kerr and Oliver LJJ. The point overlooked by Lord Fraser was made particularly clearly by Oliver LJ. The essence of his argument was that whether or not one would want to say that the Sikhs constitute an ethnic group, what the statute directs us to look for is not membership of an ethnic group, but membership of a group which (that is, the group not the individual) is defined by reference to ' its ethnic origins' It is true, of course, that the great majority of the individual members of the Sikh faith have a common ethnic origin which is generically Indian (as have a great many Muslims and Hindus), but that, in my judgment, does not mean that Sikhs, Muslims or Hindus are groups 'defined by reference to their ethnic origins'. They are, in my judgment, groups defined by reference to their religious and philosophical tenets to which anyone may belong but which are primarily composed of persons of Indian birth or descent.26
There seems no warrant for reading 'group of persons defined by reference to ... ethnic origins' as 'group of persons defined by reference to its . . . ethnic origins'. But in so far as it insists that constituting an ethnic group is not in itself sufficient to qualify a group as a racial group, this judgment straightforwardly points out what, on one interpretation, the statute says. Similar remarks were made by Kerr LJ: ™ ibid, at 1116-7.
64
Nigel Love The Act does not define a racial group by reference to its members forming part of some 'ethnic' community. It defines it as 'a group of persons defined by reference to ethnic origins'.27
But then the mistake is made of supposing that it follows that the definition of a group by reference to its 'ethnic origin' requires the investigation of the 'ethnic' ancestry and provenance of the group in question.28
This mistake is crucial. Why it is a mistake can be seen from the following analogy. Suppose the question is whether the mammals constitute a class of vertebrates defined by reference to the habit of suckling the young. Investigation of the youngnourishing procedures of mammals will reveal the fact that all extant mammals do indeed suckle their young. 29 But this does not show that the mammals are a group defined by reference to this behaviour. And in fact they are not: the mammals are defined as a class of vertebrates in which the joint between skull and lower jaw is formed by the squamosal and dentary bones. It is true that all the species now living which qualify as mammals by this criterion also suckle their young. But that is a purely contingent fact. The central issue arising from the foregoing discussion is: what construction is to be put on the words 'group of persons defined by reference to ... '1 The point overlooked by all the appellate judges is that the question whether or not a group is defined by reference to a particular set of characteristics may be an empirical one. In this view, determining whether the Sikhs are a group defined by reference to ethnic origins involves discovering what, as a matter of fact, decides who is a Sikh. It is quite irrelevant to argue about whether they are an ethnic group, or have a common ethnic origin, unless it has first been established that this has some bearing on what has to be the case for someone to count as a Sikh. Provided there is agreement as to who does and who does not belong to the group (that is, provided that Sikhhood is a 'well-defined' status), it should be possible to find out by what criteria one is counted as belonging. It might have been that the decisive criterion was ethnic origin. In which case the Sikhs would be a racial group within the meaning of the Act, because they would be a group defined by reference to ethnic origins. But this appears not to be the case, if it is possible (as suggested in several of the speeches reviewed) to become a Sikh by conversion, irrespective of one's ethnic origins. It is much more likely that the necessary condition of Sikhhood is adherence to the tenets of Sikhism.30 in which case the Sikhs will turn out to be, not a group of persons defined by reference to ethnic origins, but a group of persons defined by reference to adherence to the tenets of Sikhism. And if that is so, then even if, per impossibile, it could be shown that every Sikh was of the same colour, race, nationality or ethnic or national origins as every other Sikh, the Sikhs would still not constitute a racial group within the meaning of the Act. For it would not be with reference to these characteristics that they were defined as a group. Analogously, it would be quite vain ™ ibid, at 1121. Mibid. at 1121. 29 Provided that the somewhat perfunctory arrangements made by monotremes are counted as 'suckling'. 30 This formulation is not intended to account for very young Sikhs, whose Sikhhood presumably consists merely in their being the children of persons who adhere to the tenets of Sikhism.
The Fixed-Code Theory
65
to point to the fact that every President of the United States has been male as a reason for concluding that Presidents of the United States are a group defined by reference to maleness. Presidents of the United States are a group defined by reference to the circumstance of having been elected or promoted, to that office. They have, indeed, all been male. But that is not what makes them Presidents of the United States. Given this interpretation of 'defined by reference to' it follows that it is irrelevant that a group may contain members who do not share some typical characteristics of the group, since those characteristics are, evidently, not its defining characteristics. No doubt common usage has it that the truth of the statement that undergraduates are a group of persons in their late teens or early twenties is not impugned by the fact that some undergraduates may be older. But this has no bearing on the question whether undergraduates are a group of persons defined by reference to their being in their late teens or early twenties. Answering that question has nothing to do with ascertaining the age of undergraduates: it has to do with ascertaining what being an undergraduate essentially consists in (presumably, holding a particular academic status in the universities to which they belong). It may be that Parliament intended the Act to be construed in such a way as to allow the judgment handed down by the House of Lords. That judgment can emerge only if the words 'defined by reference to' are ignored (as by Lord Fraser) or (as by Lord Templeman) treated as equivalent to 'recognisable by'. If these manoeuvres are legitimate, Parliament has, arguably, failed to say what it meant. Alternatively, Parliament may have meant what it said. If so, the judgment of the House of Lords in the case ofMandla vs Dowell Lee seems scarcely tenable.
VI
Two interesting questions to ask about the foregoing argument are what it consists in and why it should be taken as decisive. It consists essentially in the adduction of illustrations and analogies of various kinds, designed to show that English usage offers an alternative interpretation of the phrase defined by reference to, different from that appealed to by the House of Lords. But the case for holding that the alternative interpretation is preferable to that appealed to by the House of Lords is no more than insinuated. However adroitly one thinks the weapons of logomachy have been deployed here, the argument rests, in the end, on an opinion as to the meaning to be attached to defined by reference to. And it is hard to see on what grounds it could be claimed that the opinion is authoritative. For the irreducible fact is that the meaning of defined by reference to is not fixed; and if Lord Templeman, for one, thinks it means 'recognisable by', that itself is a prima facie reason for conceding that it does, or may, mean just that.31 This is not to say that the argument, so far as it goes, is unimportant. Indeed there are grounds for thinking that, legally, it goes a long way: that is, for claiming that the House of Lords has in this case violated the rules of the legal language-game. These grounds have been stated at length elsewhere: 32 the main points can be summarised 31 It is not as though Lord Templeman's interpretation is one that could only be put forward by a lunatic, or someone with a grossly defective command of English. 32 In Beynon and Love (1984).
66
Nigel Love
for present purposes by saying that the House of Lords' judgment is incompatible with (1) the judgment in a comparable case arising out of the Race Relations Act 1968,33 and (2) the way in which the contentious words must, on pain of incoherence, be understood in other Acts where they occur. There is little doubt that these arguments are legally compelling. What is interesting about them is their essential similarity to the argument put forward above. Once again, the general form is: X must mean 'Y', because there are other times and places where Xhas meant 'Y'. Attaching weight to such arguments may be sound legal procedure. In fact, it is hard to envisage an alternative procedure which would allow the development of a body of laws which was both coherent and consistent, and adequately stable. Nonetheless, it is reasonable to ask whether it is a form of argument that has much to do with the 'literal rule' of statutory interpretation. For that rule bids us hold that X must mean 'Y', not because X has meant 'Y' on other occasions, but because the literal meaning of Xjust is 'Y'. In the legal language-game the literal rule of statutory interpretation is seen as a tool which, if properly wielded, allows the 'correct' interpretation of contentious statutory words to emerge. If the interpretation of defined by reference to argued for here were ever to prevail, for juridical purposes, over that offered by the House of Lords in Mandla, the official story would be, not that one opinion as to its meaning had been granted priority over another, but that the previous position had been, quite simply, a mistake. Misapplying the 'literal rule', their lordships arrived at the 'wrong' answer. The language is a fixed code, and what we are confronted with here is a local failure to crack it. For anyone less interested in the official story than in what is really happening, the position appears to be this. There is no way of telling, however closely one scrutinises the wording of the Race Relations Act 1976, whether the Sikhs are supposed to count as a racial group or not. It is not just that the statute does not in itself offer a direct answer to that question. It does not unambiguously even suggest a procedure for answering it. Those who want to know whether Parliament would like the Sikhs, as a group, to enjoy the protection against racial discrimination afforded by the 1976 Act will have to look elsewhere (for example, to Parliament) for an answer. What purposes are served by the pretence that the answer is arrived at by consulting the wording of the Act itself? Why is the attempt made, even in a hopeless case like this, to treat the statute as a self-contained source of the information necessary for its application? The importance of these questions lies in the fact that the fixed-code theory seems to underlie the understanding of how language works on which legal procedure is based, and its functions in that procedure therefore demand to be laid bare. If the fixed-code theory has a role to play in the kind of communicational activity exemplified by the writing and interpreting of a law, it cannot simply be dismissed as a sort of naive, empirically refutable mistake made by linguists about the nature of their object of inquiry: for if it is a mistake, it is one that has antecedently been made by at least some of those whose linguistic behaviour is being inquired into. What requires consideration is the extent to which it makes sense to suppose that there can be an objective account of what language is, independent of what language-users 33
The differences between this and the 1976 Act are irrelevant.
The Fixed-Code Theory
67
demonstrate, by their way of using it, that they think it is. No attempt to ground an analytic understanding of how language works can afford to ignore the question of how far, in seeking such an understanding, the linguist can or should emancipate himself from preconceived views on that issue which may themselves actually be constitutive of certain kinds of communicational practice. Acknowledgement - I am grateful to Dr. Helen Beynon for her help with the legal parts of this paper. Dr. Beynon is not responsible for the views expressed.
References Beynon, H. and N. Love (1984) Mandla and the meaning of racial group. The Law Quarterly Review 100, 120-36. Cartwright, R. L. (1963) Negative existentials. In C.E. Caton (ed.) Philosophy and Ordinary Language, pp. 56-66. University of Illinois Press, Urbana. Cross, R. (1976) Statutory Interpretation. Buttenvorth, London. Dummett, M. (1982) Linguistics demythologised. London Review of Books 4 (15), 9-10. Hacking, I. (1975) Why Does Language Matter to Philosophy? Cambridge University Press, Cambridge. Harris, R. (1981) The Language Myth. Duckworth, London. Leech, G. (1974) Semantics. Penguin Books, Harmondsworth. (1981) Semantics: The Study of Meaning (revised ed. of Leech 1974). Penguin Books, Harmondsworth. Moore, T. and C. Carling (1982) Understanding Language: Towards a Post-Chomskyan Linguistics. Macmillan, London. Palmer, F. R. (1984) [1972] Grammar. Penguin Books, Harmondsworth. Platts, M. de B. (1979) Ways of Meaning. Routledge and Kegan Paul, London. Russell, B. (1903) The Principles of Mathematics. Allen and Unwin, London. (1904) Meinong's theory of complexes and assumptions. Mind 13. Ryle, G. (1967) [1932] Systematically misleading expressions. In R. Rorty (ed.) The Linguistic Turn: Recent Essays in Philosophical Method, pp. 85-100. University of Chicago Press Chicago. Smith, N. and D. Wilson (1979) Modern Linguistics: the Results of Chomsky's Revolution. Penguin Books, Harmondsworth.
A Few Words on Telementation MICHAEL TOOLAN
What does saccadic mean; and, if you can't say, how will you go about finding out? How, when you come across the word in the following passage, do you make sense of it? [Our modern, 'information' society] is a society which looks into the screened eyes of its newscasters and need look no further than those saccadic responses to understand that one can no longer believe what the President of the United States says about international terrorism.
Perhaps your first reaction is to the general tenor of these remarks: very literate, alienated, and left, in its focus on the way television has come to facilitate an almost reflex fabrication of facts and values - a tendency we might dub 'telemendacity', or even 'telemendacion', if you get the idea. As you read the sentence, you recognise that familiar feeling of being harrassed by, say, one of those radical intellectuals who write in The Nation; it might be Noam Chomsky writing, you decide (in which case, you have indeed been harrised). But back to saccadic: do you reach for your Webster's, your ageing OED? And find saccade (and no adjectival form) defined as 'a jerk or jerky movement in various specific applications' (OED), with cited instances from horse-dressage, violin double-stopping and involuntary swallowing, none of which is particularly helpful in the present case.1 Now what? Communication breakdown, system failure, a message flashed across your brain: 'THIS SENTENCE CANNOT BE PROCESSED AT THE PRESENT TIME'? Or intelligent situated sense-making, a hazarding that the speaker is referring to the 'packaged speaking', an automated- and mechanical-seeming performance, abrupt, robotic, which we recognize as the hallmark of yet-undropped news anchors and currently popular politicians? The latter, probably. Not 'the latter, of course', because if you don't care enough about the larger purposes of the quoted passage, then you always have the option of turning away uncomprehendingly.
1 If you consult a recent Webster's, or the second edition of the OED, you will find new definitions of saccadic which link it specifically with rapid and jerky eye movement, and indeed it seems that in the twentieth century this usage in relation to eye-movements has come to eclipse the earlier range of uses. In the OED 2nd edition you find that one of the illustrations comes from the writings of Harris's sometime colleague, Jerome Bruner.
68
A Few Words on Telementation
69
Now the entire movement of this scenario is bizarre according to one approach to the nature of language: all this fudging and uncertainty about whether saccadic is a word (perhaps only the noun saccade is, and that an archaic French one - not English at all), and if so about what it means, shouldn't be issues at all. If saccadic is a proper member of the English lexicon, in good standing, it should have a clear meaning and a determinate pronunciation (we should know with certainty whether the second vowel is long or short), and these should be uniformly known (as in simple words like table) or potentially uniformly known. Otherwise however can we explain mutual understanding? If Locke had lived long enough to read the final paragraph of Roy Harris's The Language Machine (1987), from which the quoted sentence above is taken, he would no doubt have flagged that word saccadic as liable to cause misunderstanding and, if used in scientific discourse, in need of what Taylor (1992: 43) has termed - using a word which itself may cause misunderstanding - communicational prophylactics. Among Locke's prescriptions for reliably communicative language may be found the following: don't parrot; follow the best usage; know what the words you use mean; and don't vary word meanings. In sum, Locke implicitly declares, let us have one definitive dictionary for the language, and let us always observe its prescriptions. But why, some may ask; what is at stake here? Locke's answer is that what is at stake is nothing less than communication and the understanding this advances, that is, communicational understanding. But do we really require the kind of uniformity that Locke's prophylactic prescriptions are evidently designed to shore up, in order to secure and advance communicational understanding? Yes we do, if we have the kind of picture of communication that Locke had; and Locke pictures communication as telementation. 'Excuse me?'. Yes, telementation (it's right there, in the Supplement, a little ways down from saccadic): 'that is, the conveyance of thoughts from the mind of the speaker to the mind of the hearer' (Taylor 1992: 30). The word telementation is even more of a novelty than saccadic, having been coined by Roy Harris and tossed into the economy of metalinguistic vocabulary in his book The Language Myth (1981). That book was devoted to analyzing and exorcising the myth which, in Harris's view, has long held Western linguistics captive: the myth that language is telementational in function, and operates via a determinate code of form-meaning pairings. The telementation model regards linguistic communication, in essence, as the faxing of thoughts from A to B with speech (and hearing) fulfilling an equivalent role to that of properly 'handshaking' fax machines. Major poets in the evolution of the telementational myth include Aristotle, Locke and Saussure. Aristotle, for example, assumes that two of the three crucial domains postulated in a triangulated view of linguistic communication - ideas, and objects in the world - are human universals and invariant: they are 'the same for the whole of mankind' (Aristotle, On Interpretation, 1938: 1, quoted in Harris 1981: 9). Only the third partner in the enterprise - words - demonstrably varies, in Aristotle's view, as witnessed by the different languages that different communities use. Fortunately, within any given language, there is the possibility - and indeed the necessity, according to telementational thinking - of all speakers conforming to a single accepted system which pairs particular words with particular ideas. That fixed and uniform pairing is not merely possible but necessary within telementational
70
Michael Toolan
thinking since, where lapses from this standard occur - again, if communication is indeed the faxing of thoughts - they must cause misunderstanding, vagueness, nonsense, and 'chat'. Certainly, fixed pairing must be secured and maintained if linguistic communication is to fulfil its roles of knowledge-dissemination, information-transfer, and the description and analysis of phenomena. It's bad enough that members of one dialect of English speakers understand and react to the directive 'Get off the pavement!' in a sharply contrasting way to that of members from another dialect; but at least we can and must expect and require uniformity and recurrence of form-meaning pairings within any given community. The acme of telementational mythography are the famous pages (27ff.) of Saussure's Cours de linguistique generate outlining the 'speech circuit', complete with schematic depiction of the heads of two women blowing arrows at each other (the figures are undoubtedly women, although why Bally and Sechehaye gave them such avant-gardedly short hair is a minor enigma). Those talking heads give us pause. They are older than the paragraphs among which they sit, or float disembodiedly rather, removed from any discernible context but the context of Saussure's theory, not a single worry-line or gesture or other sign of purpose or goal to be seen, faxing those arrows at each other (more precisely, to their ears, where the arrows take a dog-leg to the brain, to be traded in for ideas), but saying nothing (Harris 1987: 164) and doing nothing. The sketch is emphatically phonocentric: one wonders what a sign-using deaf person makes of it when they come upon it in the Cours, or of the commentary that follows, where, in the course of promoting attention to the mental phases of the speech circuit and disvaluing the physical phases, it is asserted that 'it would be impossible to photograph acts of speech in all their details' (1922: 32). If signing is a deaf community's counterpart of a speaking community's speech, the claim does not obviously apply to all kinds of parole.
B By the standards of modern coursebook reprographics, the sketch is awfully quaint, if not downright hokey: a photograph it is not. And not by accident perhaps: it's hard to imagine those involved in publishing more recent editions and translations of the Cours urging 'We really must supplement that awful sketch with a genuine photograph of two people talking, overlaid by the speech-circuit lines as necessary'. For a genuine photograph, they would quickly remind each other, would be prone to all the interpretive impertinences of readers' imaginations: the readers may start wondering whether the people in the photograph are in fact speaking the same
A Few Words on Telementation
71
language; or some cynic will suggest that what B is actually saying to A is 'I haven't the faintest idea what you're talking about'. And we'd have to be so careful to find suitable interactants: better not have a woman and man speaking, or the 'genderlect' lobby will have us on toast; and of course it can't be a mother-infant pair (what a disaster that would be). No, no, we're really much better off keeping the thing quite schematic, with cameo performances from our ideally situated limbless ladies. Furthermore, it occurs to them, it's as well that this ideal pair don't look much like a Professor and Student, and certainly not the kind of Professor and Student for whom one might invent names like Ferdinand and oh, Albert, say. You know how full of misunderstandings students' notes on lectures are: professor-student communication would be a very shaky example of problem-free telementation. Borsley and Newmeyer (1997) make this same point in a footnote to their recent critique of Harris's work. They are anxious that Harris interprets the Saussure of the Cours holding to a 'stronger form' of the language myth than the F.M.L. might have actually acknowledged; they are only two among thousands of linguists who endure sleepless nights worrying over whether the Saussure of the Cours is the 'real' Saussure, and whether that book represents his 'best work'. Borsley and Newmeyer then warn that the Cours is 'a posthumous compilation from [Saussure's] undergraduates' lecture notes - an important point which the author of Reading Saussure, and of an award-winning English translation of the Cours has evidently entirely overlooked. Posthumous compilations really are of limited reliability, and I understand that in the light of their own reasoning Borsley and Newmeyer are now grappling with a parallel problem they have noticed with regard to the New Testament. Concerning the unreliable transmission of the Cours they add: 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! (1997: 44, n. 2)
A disturbing prospect, indeed. Nevertheless, their disapproval and the possible disapproval of Saussure - 'who perhaps might not have authorized the publication of this text', as Bally and Sechehaye acknowledged in their Preface to the first edition, in 1915 - are irrelevant. Whether we professors like it or not, undergraduates are always likely to draw on their 'scribbled interpretations' when judging their instructors' ideas. And if Borsley and Newmeyer would really hate to have their ideas judged in this way one is inclined to urge them to make a point of never lecturing to undergraduates, since on their own assessment this is always likely to entail the dissemination of misinformation. But of course that doesn't matter, according to Borsley and Newmeyer's implicit reasoning, because undergraduates are not their idea of 'fit judges': it's not that they never want their ideas to be judged, it's just that they want their ideas (and, by extension, Saussure's) judged by the right kind of judges, i.e. one's peers, that is, people who are 'really' speaking the same language as oneself. But Borsley and Newmeyer have not indicated any ways in which Bally and Sechehaye were not Saussure's peers in this sense (if it is allowed that the Father of Modern Linguistics had any peers). What is more telling here is Borsley and Newmeyer's quite explicit acknowledgement of the possibilities of miscommunication and misunderstanding, even in what are arguably the relatively conducive conditions of the lecture hall,
72
Michael Toolan
where notes may be scribbled down. Would Borsley and Newmeyer prefer to have their ideas judged from students' interpretations formed without benefit of writing, one wonders; that is, in a purely oral-aural setting such as that of the communicators in Saussure's speech circuit? One suspects not. In short, whether contemplating writing or speech at the practical or much-maligned empirical level, Borsley and Newmeyer have no blind faith in telementation. Empirically speaking (and even allowing that we may be contemplating an era when certain precocious twenty-oneyear-olds could write dissertations on the original vowel system of Indo-European languages, demonstrating a sophistication of linguistic analysis unparalleled until, in 1957 . . .), let us not be so naive as to imagine that Ferdinand's thoughts are encoded into a linguistic form which is decoded into identical thoughts in Albert's head. Even such synchronic communication is problematic, to say nothing of the difficulties that arise in the diachronic transmission down to today's students. No question, Borsley and Newmeyer acknowledge: empirically speaking, telementation is pure myth.2 But do they acknowledge this in theory, too? Of course, they say. Not that we generativists need to, they add, for telementationism is a defect of a code model of communication, and we generativists reject the code model. But we reject it tacitly, because it relates to something quite outside our area of inquiry: it relates to communication, a ghastly banner to parade under, and one liable to cause you to end up teaching courses on doctor-patient discourse, or children's narratives, or the language of the lawcourts: cultural or political studies carried on by other means, a woefully far cry from hypothetico-deductive science. The latter is what we value, and the latter we pursue, in our studies of acceptable and unacceptable grammatical structure: 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 n o t . . . (Borsley and Newmeyer 1997: 47)
If I may be forgiven for putting words into Borsley's and Newmeyer's mouths, I would suggest that generativists' central question is not only 'How do speakers come to have the grammatical knowledge that they do?', but also ' What is the grammatical knowledge that speakers come to have?' More important though is the need to correct the first question in the quotation above, the one said not to concern generativists but, by implication, one associatable with Harris and what Borsley and Newmeyer might call his language myth myth. For 'How do speakers use language to communicate?' is not in fact the kind of question Harris asks; rather, among his 'central questions' might be said to be What role does language play in communication? 2 At the same time, where does Borsley and Newmeyer's distrust of reliable communication from Saussure to Bally, and from Bally to today's reader, lead? It leads to a generalized scepticism and anxiety about communication which is as unjustified as the absolute and idealized assumption encapsulated in the telementation fallacy from which it depends: it is because, like Locke, Borsley and Newmeyer tactily adhere to the telementational standard that they are so ready to speculate that, on particular occasions, that standard has not been met.
A Few Words on Telementation
73
and
What are speakers doing when they speak to each other? and
What is language? It is in our answers to these questions that the fallacy of telementation needs to be highlighted, when it emerges. Borsley and Newmeyer, I submit, are content to reply to these questions in the following vein: We generativists can't answer and won't answer hopelessly large questions of that kind. But we can tell you about the grammatical structures within the language whose role in communication we couldn't possibly comment on: for example, we have a neat grammatical explanation for why reciprocal pronouns like each other require overt specification of their antecedent when they appear in complement clauses. But don't expect us to explain what the sounds represented in standard orthography by 'Which men does she expect to like each other?' do, or cause, when uttered in some actual situation by A and addressed to and heard by B.
It is hard not to see this as a minimalist or 'residual' (Harris 1987: 89) approach to semantics and meaningful language. By their own insistence, generativists devote scant analytical attention to a phenomenon so external as linguistic communication; their claim then is that their projects are untouched by commentaries directed against the telementational and fixed-code fallacies. But this amounts to an extraordinarily literal-minded defense, to the effect that if certain assumptions are not declared - or if, for that matter, they are explicitly denied - then the paradigm cannot be implicated in such thinking, as if anyone who pleads not guilty must be innocent. In fact this has been the standard response by mainstream linguists - when they have offered a response at all - to Harris's telementation charge. The response is essentially this: fixed-code telementation is so naive and simplistic a picture that no-one would seriously espouse it today. What I shall suggest in reply to this - echoing Harris, I believe - is that the avoidance of open espousal of telementation goes hand in hand, still, with a much deeper-rooted assumption, sufficiently foundational that it is spared scrutiny or even reiteration by researching linguists, that fixed-code telementation encapsulates the essential characteristics of linguistic communication. But before proceeding we should attempt to clarify just who or what it is that the Harrisian critique, in terms of the telementational and fixed-code fallacies, is directed at. Pace Borsley and Newmeyer's suggestion that Harris has gone to inordinate lengths 'attacking (what are taken to be) the historical and theoretical foundations of generative grammar' (1997: 42), it is simply untrue to suggest Harris's critique is specifically directed at generativism, or that he is out to show 'that Chomsky has got it all wrong' (ibid.). In Harris's pantheon of the great pilots of this century who have steered Western linguistics onto the rocks, Chomsky probably ranks third, behind Saussure and Bloomfield, on both of whom Harris has written far more extensively. Of course Chomsky is very much alive and intellectually active, which may make the attention feel different to Borsley and Newmeyer. But Harris's critique of theories -
74
Michael Toolan
is this surprising? - has little to do with the individual authors of those theories (so that whether the author of the Cours is more Charles Bally than Ferdinand de Saussure is of no consequence) but with what he sees as the enduring myth underlying all these theories. For Harris, the breeding ground of this myth is no single scholar but, as he often emphasizes, 'the Western tradition'. Locke and Saussure and Bloomfield and Chomsky are the great midwives of dazzling offspring of the myth. The myth's reach is far more extensive than that of generativism: if generativism as a term of art dies away (as transformationalism has done), and is replaced by, for example, cognitivism, the myth may survive untouched. Borsley and Newmeyer would have us believe that mainstream linguistics has no truck with telementation, and they defy those who suggest it does to demonstrate any cases since the seminal papers on the generative-linguistic treatment of semantics by Katz and Fodor in the 1960s, where idea-transfer has been assumed to be the function of language. 'We have found no more recent example of a generative linguist endorsing telementation', they declare (Borsley and Newmeyer 1997: 46). Perhaps this is simply because the endorsement is ubiquitous. In the interests of brevity, I shall cite evidence of telementational and fixed-code assumptions from just one source - but, arguably, the most crucial contemporary one, since this book has been praised on all sides as a compelling articulation of the real, cognitive interests of contemporary linguistics. I refer to Steven Tinker's acclaimed bestseller, The Language Instinct: The word dog does not look like a dog, walk like a dog, or woof like a dog, but it means 'dog' just the same. It does so because every English speaker has undergone an identical act of rote learning in childhood that links the sound to the meaning. For the price of this standardized memorization, the members of a language community receive an enormous benefit: the ability to convey a concept from mind to mind virtually instantaneously. (Pinker 1994: 83-4) 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 that combine the words to convey relationships among concepts (a mental grammar), (ibid.: 85) Ungrammaticality is simply a consequence of our having a fixed code for interpreting sentences. (ibid.: 88)
A second look at the generativist exemplum supplied by Borsley and Newmeyer is now in order: 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. . .
What's going on here?, the outsider asks. Who is this 'English speaker' that Borsley and Newmeyer refer to? Why the emphasis on her being a speaker: wouldn't she have to be able to read English in order to judge the acceptability of the written sentence 'Does she expect to like each other?' Or is the idea that someone will 'present' her with a spoken utterance which we can take to correspond to the written one specified here, and that it is her reaction when she hears this spoken utterance that Borsley and Newmeyer are talking about? But then wouldn't it be better to call her an
A Few Words on Telementation
75
English hearer than a speaker? This is nitpicking, Borsley and Newmeyer complain: it doesn't matter whether the string is written or spoken, whether she's called the speaker or the hearer. And don't start cooking up some 'real-world context', a rushed telephone conversation, for example, where, sure, you might be able to get away with saying something which might be transcribable as 'Does she expect to like each other?', with the missing parties situationally-retrievable. We're not interested in contextualized examples, or particular individuals processing particular linguistic material: that's our point! We're interested in grammaticality and acceptability as these apply when, frankly, you cut away all the distractions and complications of language integrated into diverse activities. The language we're interested in is autonomous, use-free, the thing in itself. And the only criterion we want to match this worlddetached language against is that of acceptability of the most general kind. At this point, the integrational linguist is surely justified in asking what possible 'acceptability' yardstick the autonomous linguist has it in mind for us to consult, given that they have just emphasized that we need to treat their probe sentences as ideally removed from real or imagined situations, contexts, and human beings. I think it is clear that the acceptability they have in mind relates to pure ideation. That is, we are supposed to inspect the sentence Which men does she expect to like each other? as if it were in vacuo; we are expected to see that it articulates a coherent thought or complex of thoughts, and stamp it acceptable - purely on the grounds that it conveys one or several clear ideas. And on the other hand we are to scan Does she expect to like each other? as if this, similarly, came at us as a pure, context-free stream of ideation, and stamp it unacceptable solely on the grounds that, as a vehicle for ideastransmission it palpably fails. No other communicational grounds, besides ideastransfer, are relevant to this mainstream linguistic notion of acceptability. Thus if one were to suggest that, inter alia, saying Does she expect to like each other might, in relevant circumstances, be a 'very acceptable' means by which haste, anxiety or overeagerness could be communicated, the suggestion would be ruled out as entirely eccentric by the telementationally-minded mainstream. Much the same could be said for sentences like the following classic, which generativist psycholinguists for many years cited as especially revealing, although they now tend to regard it as embarrassingly unnatural: The horse raced past the barn fell.
What is a 'barn fell', the reader of this sentence is thought to wonder; should that be hyphenated; is it at all related to a 'barn yard', or a 'bean field'? The reader has 'no idea' what a 'barn fell' is, and in view of this - so the argument goes - retraces their mental steps along the garden path and belatedly finds the hidden relative clause. As before, all very cerebral and ideational; in fact, nothing but telementational. The sentence belongs to psycholinguists - certainly not to equestrians or farmers - and it is a little ungracious of them to disown it at this late date. Such a sentence is difficult to imagine being uttered in normal purposeful spoken discourse; and if it were uttered, no normal delivery of it would be prosodically silent, in the way the written form is, about the phrasal boundary between barn and fell. But to talk like this is to move down a garden path of no interest to generative psycholinguists - as if the
76
Michael Toolan
sentence referred to a real horse and a real fall, that is, a real situation of language use in which many other things besides ideas-transfer might be of interest. In sum, what unites Does she expect to like each other? and The horse raced past the barn fell, as they join such other prize mutants of the linguistics laboratory as The farmer kills the duckling and Colorless green ideas sleep furiously is their almost pure inutility for any purposes other than the underwriting of telementation. And notwithstanding professed rejections of code-fixity and telementation to the contrary, the generativist methodology depends crucially on 'intuitive acceptability (and unacceptability)' as a heuristic lever, and this intuitive acceptability criterion treats utterances as essentially telementations, either successful ('grammatical') or defective ('ungrammaticar). What does telementation embrace, and what does it discard or ignore? Telementation is, quite bluntly, 'thought transfer': all that it requires and expects is that B 'gets' A's thought. It is therefore not so far removed, in essence, from telepathy, although the latter's roots suggest it involves feeling at a remove, whereas telementation involves thinking another's thoughts, thanks to that other 'sending' them to you. If quite how this is managed is a little hard to visualize, all the better for you: you are not entirely in thrall to the telementational fallacy. If one does try to imagine the 'pure' transfer of thoughts, one usually conjures up images of heads, connected by wires or electrodes, and A furiously intent on 'sending' the thought Mary had a little lamb 'down the wire' to B. The picture is, clearly, not unlike Saussure's one of the limbless ladies. But wait - and this is why language is telementational - A doesn't have to rely on fierce concentration, auras and atmospheres, to send his thoughts to B: he can use the single fixed-code language that he holds in common with B And now, ta da, A simply has to say Mary had a little lamb or Where's the rest of me? or In cognitive psychology the claim that there is a fixed structure of human nature traditionally takes the form of an insistence on the heterogeneity of cognitive mechanisms and the rigidity of the cognitive architecture that effects their encapsulation and B will almost instantaneously 'get' those ideas. (The 'cognitive' sentence just quoted is Fodor's, in a passage quoted in Pinker 1994: 405.) Crucial to telementation, or 'ideas-transfer' is the assumption - which is either false, unprovable or irrelevant, and perhaps all three - that when A speaks to B the same idea that A used and encoded into speech is picked out, highlighted, or recreated in B's head. A has an idea and, via the 'conduit' of language, B receives a copy. In telementation, language enables ideas to be faxed or, in a generous interpretation, re-assembled or re-created; what telementation signally excludes, and what a genuinely anti-telementational approach such as Harris's integrational linguistics includes, is the notion that in situated interactions between As and Bs, all parties draw on language to construct or create ideas. It's all the difference between duplication and invention. Telementation excludes almost everything to do with 'the whole bodily behaviour' of humankind (as Firth termed it: Firth 1964: 19), except our mind/ brains, and is concerned with the latter only insofar as they can be thought of as machines for processing propositions and not, for example, as theatres teeming with emotions. It excludes difference, action, disagreement, emotion, embodiedness and embeddedness. It interconnects with assumptions and confusions in relation to the term 'language', by contrast with the term 'a language'. The former is routinely
A Few Words on Telementation
77
regarded as too vague, general, and protean to be amenable to systematic understanding; only the latter, whose structure can be foregrounded under the rubric of 'langue' or 'competence', is conducive to scientific study; and, in a recent further turn of the screw, only the deep internal workings of a language, studied at a remove from all the 'noise' which contaminates the implementation of linguistic structure in something as 'external' as, say, the English used between people on my busride home tonight, is of linguistic interest. In practice, however, the banished term language reappears at every turn, not least in any and every appeal to empirical evidence.3 Thus there is a systematic ambivalence in the way mainstream linguists use such terms as linguistic knowledge, knowledge of language, and language acquisition. At the heart of all linguistic communication - not just in the use of creative metaphors - lies risk: the risk involved in attempting to get one's interlocutor to see the same picture of some aspect of reality that you, the speaker, believe you see. This element of risk is of course troubling, especially to the telementationally minded, who would wish that language at its most precise and unambiguous might shuttle ideas from mind to mind in a flawless, friction-free manner. On the other hand, and from outside the telementational mindset, the risk involved is part and parcel of genuine linguistic creativity: language can only genuinely go wrong when one is genuinely remaking language as one goes along. A proper respect for the indeterminacies of language should help us to see the importance of interactional risk. In the integrational alternative, no 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. The reverse of iterability 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. In general, from an integrational linguistic perspective, signification is not assumed to be constant across 'repeated' expressions from a single speaker; equally importantly, it is not assumed to hold constant between a speaker and an addressee. There are some standard manoeuvres used for deflecting Harris's 'telementation' criticism. Essentially, the problem is redefined, and then rejection of the redefined phenomenon is documented. Thus, it is claimed that what Harris is railing against is 'the code theory of communication'; and then various emphatic critiques of that code theory are cited, to show that generativists have no sympathy with such a model. The relevance-theoretical work of Sperber and Wilson is often cited at this point, as evidence of a positive rejection of code theory, in favor of a model of communication as ostensive-inferential. But such defenses overlook a countercurrent in relevancetheory argumentation, which emerges from the fact that while relevance theory rejects a code theory of communication, it actually accepts a model of language as a code (albeit one that requires bolstering by various interpretive procedures, including the invoking of a principle of relevance, in order for speakers to communicate). In relevance theory, linguistic communication succeeds because a relevance-shaped process of inferencing can and must operate upon the fixed-code 'thoughts' supplied 3 Empiricism - which in North American theoretical circles seems to have become inseparably bound, collocationally, with the dismissive epithet naive - is, in Borsley and Newmeyer's view, the only principle (and for them a by definition inadequate one) to which Harris's integrational approach is explicitly committed.
78
Michael Toolan
by the language.4 Harris's 'language myth' objects to the fixed-code theory of language; but quite independently he objects to the ideas-transfer theory of communication. In conflating these principles and then concurring only with Harris's critique of the former, commentators reveal just how powerful is the general expectation that nothing but ideas is 'transferred' in language communication, and ideas which are utterly identical in speaker and hearer, at that. One of the most illuminating and crucial distinctions drawn in The Language Myth, that between hearability and mere audibility, is germane to the discussion (a parallel contrast holds between legibility and mere visibility). As Harris notes, orthodox linguistics has attended far too much to audibility, a property conducive to treatment with a mechanistic descriptive apparatus of determinate units, rules, and criteria, and that attention has been to the neglect of hearability. However, When A speaks to B, the prime communicational concern which both share in common is not just the audibility but the hearability of what is said; as when A writes a letter to B the prime communicational concern is not just the visibility but the legibility of what is written. (Harris 1981:178)
Harris pursues the discussion of hearability and legibility via interesting examples of the indeterminacy of form in each domain. He considers the bases of interpretation of a sound transcribable as [zaaaa] produced with a falling intonation, uttered by a bowler in situationally-appropriate circumstances in a game of cricket; and the parallel interpretation of a scrawled figure, in a letter, which might possibly be the word 'and'. In the latter case, he writes: We say, for example, of some dubiously legible word in John's letter, 'It looks like and' - as if we were engaged in an actual process of comparison between the scrawl on the paper and some imagined copperplate prototype, whereas what we are in fact trying to do is to determine how the scrawl is most plausibly interpretable in its context, in such a way as to make sense of John's communication, (ibid.: 178-9)
If we emphasize legibility rather than mere visibility then, as an essential element in interpretation, we must also remember that legibility is to be understood as a matter not of what a written form 'is permitted to be read as', but of what it 'can be, and is, read as'. Harris proceeds to demonstrate the unhelpfulness of any assumed reliance on a prototypical 'and', and shows that prototypicality cannot give an 'and' in the reading task: In some respects the scrawl will be similar to the prototype, and in other respects it will be dissimilar. What we need to know is which similarities matter. But the prototype itself cannot tell us this. It does not incorporate in its own visual contours the criteria relevant for matching it with every other written shape. All we have done is to substitute one problem for another. Furthermore, the substitution leads us nowhere. For we now have to decide in which respect the scrawl would need to be like the prototype in order to determine whether the scrawl is indeed 'like and'. The answer will be that the relevant respects are precisely those in which the scrawl is legible as and. If we can determine that for the scrawl, then we do not need the prototype at all. If not, no amount of visual inspection of the prototype will help, (ibid.: 179) 4 I offer a fuller critique of relevance theory, in relation to integrational linguistics, in Chapter Five of Toolan 1996.
A Few Words on Telementation
79
Harris's chief target here is the fixed-code assumption that there is a prototype form of and in the written language (or of How's that in the spoken language), which would somehow 'tell' the language-user whether a particular encountered form was an instantiation of that prototype, that is, whether the encountered item was an imperfect token of the authoritative and pre-existing type. On the contrary, Harris argues, there is no warrant for assuming such prototypes exist, and even if they did, they would not be able to resolve the question of whether a particular encountered form was a token of the posited type. That crucial question can only be resolved (provisionally) in the perceived circumstances of an experienced communication situation. At the same time it is clear that perception of similarity plays an important role in the local determination of whether some scrawl is an and, that is, in deciding what a particular shape or sound 'counts as'. Readers' and listeners' procedures of comparison and judging similarity are a crucial part of language-comprehension and interpretation. Even to supply the metalinguistic commentary 'It looks like and' is to identify the speaker as an inducted member of a language community, with memories of previous uses of the word and. As Harris remarks: No one supposes that two individuals can somehow achieve instant communication on the basis of words neither has used or heard before. The question is whether it is correct, or useful, or even plausible to represent their past linguistic experience as standing in relation to their present communication situation as the rules of a game to a particular episode of play, (ibid.: 186)
An interactant's encountering of a spoken or written expression does not lead, directly and immediately, to an attempt to match it with a single, one-entry-for-allpurposes lemma in a mental lexicon, which supplies one or several possible meanings for the heard form. That is the essence of mistaken bi-planarity. We do not live in language that way. And when we behave as if we did, we should be discouraged from such fundamentalism, for it is clear that to assume the existence of prototype forms and prototype meanings and to pretend to interpret linguistic communication accordingly amounts to unqualified embrace of fixed-code telementational principles. Such an ill-advised prototype-matching would seem to bedevil generativist linguists' reception of a third word in Harris's lexicon, to which I now turn. The word in question is fascist. Not a nice word, granted, nor a new one. It occurs in a short commentary (on translation) which he published in the Times Literary Supplement of London in 1983. There he called the idea of an ideal speech community which is totally homogeneous (the wording echoes Chomsky's famous formulation) 'a fascist concept of languages if ever there was one' (1983: 1119). Lord, the outcry! That f-word is a fighting word 'if ever there was one'. I don't entirely agree that it was appropriate in the context: surely totalitarian is the more accurate description? But Harris, exercising the freedom of speech that he and Chomsky have championed, chose to use the f-word. And it's hard to fault his strategy. Because - in my reading of this minor current of linguistic historiography by means of that one naughty word in one ephemeral review, Harris got the attention of generativists in a way that none of his earlier booklength critiques had. Inter alia it led to a brief caustic rejoinder in Chomsky (1986: 47), while in the paper discussed here, Borsley and Newmeyer (1997), the word provokes Borsley and Newmeyer to unwitting self-parody when they call it a 'theory' and devote an entire section to what
80
Michael Toolan
they apparently think is knock-down refutation of what they see as Harris's implicit claim. Their section begins: Harris's theory makes a clear prediction about the countries in which structural linguistics (of any variety) should have been the most flourishing and enjoyed the highest degree of official sanction. They are Germany and Italy. (1997: 51)
The 'historical evidence' that Borsley and Newmeyer thereafter parade to show that 'Harris's predictions are completely wrong' is perhaps best not dwelt upon; mindful that Borsley and Newmeyer have fallen into the trap of answering one naughty word with many preposterous ones, I shall risk censure in the other direction by offering no reply to their rebuttal of the fascism 'prediction'. But it would seem that the beginnings of Borsley and Newmeyer's misrepresentation of Harris's remarks lie in their uptake of the word fascist telementationally; that is, as if it, necessarily, meant and conveyed to every reader just one fixed idea. The larger point of Harris's commentary in which the f-word appears is that the post-Saussurean 'idealization' of the homogeneous speech community has knocked cross-cultural language study right off the agenda, at least as far as Linguistics is concerned: bilingualism and translation are taken to be of little and no interest, respectively, to the linguist (the former is of interest only insofar as it furnishes evidence for one account or other of Universal Grammar, the culture-free study of linguistic structures). Let us move beyond these moments of acrimony to a scene of love and communicative harmony, one that occurs towards the close of a play by Brian Friel, The Communication Cord. In this play Tim, a graduate student writing a thesis on language and communication, finds that the empirical specifics of his developing interaction with Claire, in all its warmth and substantiality, contradicts much of what his dissertation has postulated, in theory, about language: CLAIRE: It's been a long time, Tim. TIM: Seven and a half years. The very week I began working on the thesis. CLAIRE: Tim the Thesis. JACK: Pompous bloody idiot. CLAIRE: It's on conversation, isn't it? TIM: Language as a ritualized act between two people. CLAIRE: Yes. TIM: The exchange of units of communication through an agreed code. CLAIRE: Yes. TIM: Fundamental to any meaningful exchange between individuals. CLAIRE: Is it? TIM: That's the theory JACK: (Tentatively) Tim . . . ? CLAIRE: Chat, really. TIM: Yes. CLAIRE: What we're doing now. TIM: But I think that after this I may have to rewrite a lot of it. CLAIRE: Why? TIM: All that stuff about units of communication. Maybe the units don't matter all that much. CLAIRE: I think that's true. JACK: Can you come here for a minute, Tim?
A Few Words on Telementation TIM: We're conversing now but we're not exchanging units, are we? CLAIRE: I don't think so, are we? TIM: I don't think we can be because I'm not too sure what I'm saying. CLAIRE: I don't know what you're saying either but I think I know what's implicit in it. JACK: Tim, I think I'm in a bit of trouble. TIM: Even if what I'm saying is rubbish? CLAIRE: Yes. TIM: Like 'this is our first cathedral' [an apothegm continually used, during the play, to describe the traditional Irish cottage - MJT]? CLAIRE: Like that. TIM: Like 'this is the true centre'? CLAIRE: I think I know what's implicit in that. TIM: Maybe the message doesn't matter at all then. JACK: Tim! CLAIRE: It's the occasion that matters. TIM: And the reverberations that the occasion generates . . . CLAIRE: I feel the reverberations. TIM: I feel the reverberations. CLAIRE: And the desire to sustain the occasion. TIM: And saying anything, anything at all, that keeps the occasion going. JACK: Tim! CLAIRE: Maybe even saying nothing. (Brian Friel 1983)
Just who Brian Friel had been reading when he wrote The Communication Cord is open to speculation (the influence of George Steiner's writings on his play Translations is well-documented: in the present case are there shades of Goffman too, perhaps?). But it is certainly possible to see this closing love scene as another version of the scenario of 'linguistic communication without exchange of units' which Friel previously achieved in one of his most memorable scenes, in Translations, where the Gaelic-using Maire and the English-using Yolland speak love to each other, via and despite their mutual monolingual incomprehension. In Translations the love scene leads ultimately to tragedy; here a similar situation is re-told as farce. And, to be fair to what Friel suggests here, it should be noted that even as Tim and Claire are demonstrating the irrelevance of the exchange of units to one kind of entirely successful communication, Jack is comically failing to communicate to Tim at a more basic, mathetic level concerning the shaky structure of the building they are occupying. The consequence of this communicational breakdown is that, as the final curtain falls, so does the cottage: Tim, leaning too hard on a critical post, literally brings the house down. It makes little difference whether or not we label Tim's successful communication in this scene the interactional, and that which he overlooks transactional, or vice versa for that matter: Friel's key point would seem to be that communication comprises a range of interest and attention, and not mere telementation. Lockean telementation simply has nothing to say to the likes of Tim and Claire, and Maire and Yolland: to put this another way, telementation is a picture of communication that sets no store whatsoever by the fact that individuals at least sometimes love each other, so that saying anything or saying nothing may function equally well communicatively, since the essence of that communication is no particular content, but the occasion and the desire to sustain the occasion. This latter description, jointly formulated by
81
82
Michael Toolan
Tim and Claire, is as direct a gloss of Malinowski's phatic communication as one is likely to find in the theatre, which in turn reminds us of just how fundamentally different are the phatic and telementational conceptions of communication. A thought just 'came' to me which I would like to send to you: do you have the same idea of saccadic now as you did before you began reading this essay? And of telementation? I have suggested in this paper that where mainstream linguistics has even bothered to respond to Harris's fixed-code telementation charge, it has, at best, taken the form of agreeing that linguistic communication certainly cannot be fixed-code telementation and then resuming its theoretical enquiries as if in fact it were (invoking that all-purpose defense, as the need arises, of 'idealization'). What have been striking in their relative absence are responses of the following kind: given that linguistic communication cannot possibly be founded upon fixed-code telementation, what are the ways in which language is not a fixed-code, and what are the ways in which linguistic communication is not telementation? Arguably, these questions are less than paradigm-shifting, and assume that the questioner is setting out from a place still within the labyrinth created by language-myth segregationalism. But if this is the place in which we still find ourselves, then it may be idle to hope to advance our exorcism of the language myth from any other place of leverage.
References Borsley, R. and F. Newmeyer (1997) The language muddle: Roy Harris and generative grammar. In G. Wolf and N. Love (eds.) Linguistics Inside Out: Roy Harris and His Critics, pp. 42-64. John Benjamins, Amsterdam/ Philadelphia. Chomsky, N. (1986) Knowledge of Language: Its Nature, Origin and Use. Praeger, New York. Firth, J. R (1964) The Tongues of Men. Watt, London. Friel, B. (1983) The Communication Cord. Faber, London. Harris. R (1981) The Language Myth. Duckworth, London. (1983) Theoretical ideas. Times Literary Supplement, 14 October, p. 1113. (1987) The Language Machine. Duckworth, London. Pinker. S (1994) The Language Instinct. Harper, New York. Saussure, F. de (1922) Cours de linguistique gentrale, 2nd ed. Payot, Paris. Taylor, T. (1992) Mutual Misunderstanding. Duke University Press, Durham. Toolan, M. (1996) Total Speech: An Integrational Linguistic Approach to Language. Duke University Press, Durham.
The Dialect Myth ROY HARRIS
C.-J. N. Bailey once observed how remarkable it is that in linguistics the term dialect 'can go on being used in a certain sense after investigation has shown that the term refers to nothing that can be found in the real world' (1981: 52). His own work makes an important contribution towards exposing contemporary versions of the dialect myth. The present paper does not discuss Bailey's attack on those versions, or the criticisms to which his own notion of the polylect has given rise (Love 1984), but examines the historical background against which the development of the dialect myth has to be seen. Like the more general 'language myth' (Harris 1981) of which it is one facet, the dialect myth has its ultimate roots in Graeco-Roman antiquity, and its development is part and parcel of European cultural evolution. Before attempting to trace the history of the dialect myth, one preliminary clarification of terminology is necessary. The modern English word dialect, like the word language, has two distinct uses. In one use, both words are pluralizable and can take a definite or indefinite article. In the other use, both words are nonpluralizable and take no articles. As regards the word language, these two uses are a perpetual source of potential and actual confusion in linguistic theory (Harris 1981: Chaps. 1, 2; 1983). The difference between the two uses of the term dialect may also be of some consequence. Wells, for example, in his book Accents of English (1982) proposes to argue the case for English that 'dialect exists, but dialects do not'. Such a thesis would make no sense at all if there were no semantic difference implied between the two uses of the word dialect in that sentence. For Wells is certainly not wishing to argue that the class of dialects happens to have only one member. Initially, controversy can perhaps best be avoided if one speaks vaguely of 'dialect concepts', using that phrase deliberately to avoid commitment to either of the two more specialized English uses of the word dialect. Let us then consider briefly some of the historical background to dialect concepts in the Western tradition. The Greek word didlektos has a complicated and not very well documented history. But one does not have to be an expert in Greek metalinguistics to discover that we must beware of simply equating that term with our modern English term dialect. Liddell and Scott's Greek lexicon (1996) lists various meanings for didlektos, including, in addition to 'dialect', the following: 'discourse, conversation; speech,
83
84
Roy Harris
language, spoken (as opposed to written) language; the language of a country; local word or expression; way of speaking, accent; style (esp. poetical diction)'. Nor is it difficult to discover that very often Greek writers do not seem particularly concerned to draw any sharp distinction between dialects and languages, as for instance in the following passages from Herodotus (Book I, Chap. 139) about the linguistic differences to be observed among the lonians in Asia Minor. They use not all the same speech but four different dialects. Miletus lies farthest south among them, and next to it come Myus and Priene; these are settlements in Caria, and they use a common language. Ephesus, Colophon, Lebedos, Teos, Clazomenae, Phocaea, all of them being in Lydia, have a language in common which is wholly different from the speech of the three cities aforementioned. There are yet three Ionian cities, two of them situate on the islands of Samos and Chios, and one Erythrae, on the mainland. The Chians and Erythraeans speak alike, but the Samians have a language which is their own and none other's. It is thus seen that there are four fashions of speech.
The above is Godley's translation. But the interesting thing about the passage is that Herodotus does not once use the word didlektos or any other term which specifically designates a dialectal variety of speech, even though to our way of thinking the variations Herodotus is discussing cry out for some such specific word. Herodotus simply uses the general word for a language: glossa. Elsewhere (Book I, Chap. 58) Herodotus gives it as his view that the Hellenic peoples have 'always used the same language', as distinct from the Pelasgians, who spoke a language 'which was not Greek'. So the criteria for what counts as 'the same language' seem to be rather variable. From the distinctions that Herodotus draws, it does not appear that comprehensibility plays a very important role as a criterion. But nor, on the other hand, does Herodotus treat each separate settlement as having its own dialect. On the contrary, in the first passage cited he mentions twelve settlements but distinguishes only four dialects. What exactly the linguistic criteria are for distinguishing these four types is not clear. Now to say that the Greeks, at least sometimes or for certain purposes, did not distinguish between dialects and languages is not of course to say that they somehow failed to recognize that Greek and Egyptian were different languages, nor that they somehow failed to recognize the linguistic bond between Ionic and Attic. But it is to say that we find in Greek linguistic thought the first evidence of a very primitive dialect concept, which places the emphasis simply on features which distinguish one linguistic community from another. What at least is clear is that the Greeks of the period when Herodotus was writing conceptualized linguistic differentiation on a geographical basis in such a way as to allow them to say: The inhabitants of place A talk quite differently from the inhabitants of place B, but the same as the inhabitants of place C'; but they did not apparently use this as a basis for distinguishing consistently between terms corresponding to our language and dialect. A move in this direction seems to be detectable in Stoic linguistic theory, where didlektos emerges as a technical term. According to Diogenes Laertius, who puts the matter in not altogether lucid phrases, the Stoics defined didlektos as an expression (lexis) which is stamped on one people 'ethnically and Hellenically' or as 'an expression peculiar to some particular region' . . . e.g., in Attic thdlatta 'sea' (as
The Dialect Myth
85
distinct from thdlassa) and in Ionic hemere 'day' (as distinct from hemerci). If this account is correct, and in particular if it is correct to translate lexis as 'expression' rather than 'manner of speech', it would seem that the Stoics had the concept of what would nowadays be called a 'dialect form'. But that is importantly different from having the concept of a 'dialect'. With the Stoics, too, we find the question raised as to why there should be differences between the Greek spoken in different places; and this is seen as being bound up with the old controversy which goes back at least as far as Plato's Cratylus as to whether names are in origin natural or conventional. One Stoic explanation of linguistic variation which Diogenes cites held that names were indeed natural, at least in the sense that they derived from primitive man's spontaneous vocal reactions to his sense experience, and that consequently it was only to be expected that in different places nature would provide men with different sense experiences, and so their vocal reactions would be different. As Diogenes puts it, 'the air thus emitted was moulded by their individual feelings or sense-presentations, and differently according to the difference of the regions which the tribes inhabited' (Book V, Chaps. 75-6). This is, perhaps, the earliest form of explanation recorded for dialect differences. It is interesting to note that it is a causal explanation of a physical kind: that is to say, dialect differences are held, in origin at least, to reflect actual physical differences between one part of the world and another, which in turn result in different sense experiences for the inhabitants. It is not, obviously, a very convincing explanation unless it can be shown that the geographical distribution of various speech forms did, at some time, correlate with the relevant physical differences between one region and another. No serious attempt was ever made to provide that kind of evidence either by the Stoics or by anyone else in antiquity. It might perhaps be supposed that this primitive dialect concept must be long since dead and buried. But it probably survives quite widely as a lay linguistic assumption in modern Western civilization. That is to say, it is probably far from obvious to many people that the coexistence of the words furze and gorse is in principle any different from the coexistence of the words furze and ajonc. These are just terms used in different parts of the world to designate a kind of bush with prickles and bright yellow flowers. Or, to take a rather trickier example, if in 1902 a Frenchman with no university education had happened to be aware of the kind of information published in that year in Map 27 of the Atlas linguistique de la France, it is very unclear that he would on that basis have had any clear idea of how many languages, as distinct from how many dialects, were spoken in France. He might have discovered, for example, that in Paris the first person plural present indicative of the verb 'to go' was [alo], whereas not many kilometers to the south at Poilly it was [vo], and much further south again at Seilhac it was [ana]. But it is hard to see, by comparing these three forms, how he would have come to the conclusion that [alo] vs. [vo] is merely a dialect difference, whereas [vo] vs. [ana] is a language difference. Very probably many ordinary Frenchmen of his generation would have described the speech of Provencal farmers as a dialect of French, not a separate language; or at least no more a separate language than the speech of the farmers of Picardy or Burgundy. Indeed, academics were for a long time at loggerheads over whether the group of dialects
86
Roy Harris
known as ' Franco-Pro ven9al' should be treated as a third Gallo-Romance language or merely as a regional variety of French. Now what is at issue here is not whether people happen to have been educated to call some differences 'dialect differences' and other differences 'language differences'. What is at issue is whether there are any experiential criteria by which the lay public distinguishes consistently between the word forms of geographically adjacent dialects as opposed to languages. If not, then the lay public in question has basically the same primitive dialect concept that we first find in the writings of the Greeks. If you have no other dialect concept, then the only obvious way of making sense of a distinction between languages and dialects turns on when you reach the point where differences are so numerous that communication simply breaks down But this criterion does not necessarily coincide with the genealogy of languages as traced by the historical philologist. Since the Greeks had no theory of linguistic phylogeny at all, as far as they were concerned that was not a plausible alternative. To sum up very briefly, then, we find at a primitive phase in linguistic thought a recognition of linguistic differences from one region to another; but these are treated as lying, as it were, on the same continuum of linguistic differentiation which separates, say, the speech of a Greek at one extreme from the speech of a Persian at another. Let us call this, for convenience, the 'continuum' dialect concept. This 'continuum' concept stands in contrast with a quite different and more sophisticated concept which is economically captured in the definition of the term dialect given in Crystal's Dictionary of Linguistics and Phonetics (1985: 92): 'A regionally or socially distinctive VARIETY of a language, identified by a particular set of WORDS and GRAMMATICAL STRUCTURES.' Now there are a number of differences between this and the more primitive dialect concept which leap to the eye. Crystal mentions social differentiation in addition to regional differentiation, as if somehow these two things were on the same footing; and he also explicitly mentions grammatical differences in addition to lexical differences. So Crystal's definition in those respects gives us a broader dialect concept altogether. But that is not the most important difference. The contrast relevant for our present purposes is that we are no longer dealing with a continuum concept at all, but with what may be called a relational dialect concept. That is to say, what a dialect is is now defined in relation to what a language is. Specifically, a dialect is conceived of as a particular subvariety of a language. A language is the prior concept, and a dialect is defined in relation to that. Dialect, in other words, has become what logicians would call a two-place predicate. The essential idea is that to be a dialect is to be a dialect of a language, just as to be a father or a mother are essentially relational concepts. You must be a father or a mother of someone else: it is not an empirical but a logical requirement of the definition that there must exist someone else to whom you stand in the relation of being father or mother. So there must exist a language in relation to which a dialect stands as a subvariety. The question 'What is that a dialect of?' must make sense, just as 'Who is he the father of?' must make sense. In neither case can the answer be reflexive. You can no more reply 'It is a dialect of itself than you can say 'He is the father of himself.'
The Dialect Myth
87
This relationship between languages and dialects which the relational concept implies is perhaps best summed up by C. E. Reed in his book Dialects of American English: 'languages normally consist of dialects'. In short, language functions as the collective term embracing a certain group of dialects. Let us now contrast the continuum dialect concept with what appears to be a rather different modern view of the matter, which for convenience may be called the 'aggregate' dialect concept. The 'aggregate' concept approaches the question, as it were, from the opposite direction from the 'relational' concept. It envisages a dialect as constituted out of the sum total of the linguistic practice of a certain group of individuals. In other words, instead of starting with a language and breaking it down into dialects, you start off with individuals, and aggregate their linguistic behavior into dialects. A classic example of this aggregate concept in modern linguistics would be the famous definition proposed by Bernard Bloch (1948: 7-8): 'A class of idiolects with the same phonological system is a DIALECT'. For purposes of this definition, the term idiolect is defined by Bloch as 'the totality of the possible utterances of one speaker at one time in using a language to interact with one other speaker'. Two brief comments about Bloch's definition are relevant to the remainder of this paper. First, it does not really matter here that Bloch defines the term idiolect so restrict!vely as to limit it to one speaker talking to just one other speaker. The term idiolect is often used rather more generously to include the totality of the linguistic equipment employed by any monoglot speaker, regardless of who the speaker is talking to or listening to. Second, it does not really matter either that Bloch picks out the common phonological system as what unites different idiolects into a single dialect. He might just as well, as far as the present argument is concerned, have picked on a range of common vocabulary, or noun declensions, or syntactic constructions. The essential point is that he sees a dialect as an aggregate of idiolects, united by some common feature or set of features. We should note in passing that this Blochian approach to dialects has been carried one stage further by a more recent generation of linguists who are happy to identify dialects by reference to a single feature: for example, whether or not final r is pronounced, or whether or not speakers have an aspirate h in their consonant repertory. It might perhaps be objected that taken to such an extreme, this constitutes an abuse of the traditional term dialect, and other terms have been proposed. C.-J. N. Bailey (1973: 11), for instance, has suggested the single term lect to do duty as to what he calls 'a completely non-committal term for any bundling together of linguistic phenomena'. But it is fair comment to add that even Bloch's less radical definition of a dialect already dispenses with the traditional notion that a dialect has some kind of regional or geographical unity. Of course, it is quite likely in practice that speakers sharing a common phonology live close together. But the point is that they do not have to in order to meet the requirements of Bloch's definition. Their geographical proximity would be a contingent fact, not a criterion. There is one more general point arising out of Bloch's definition which deserves mention here. If a dialect is a class of idiolects sharing the same phonological system, that still allows speakers of the same dialect to vary regionally in their pronunciation. This point has been generalized in the form of the terminological distinction which some theorists draw between dialect and accent. Lyons (1981: 24-5), for example,
88
Roy Harris
claims that the term accent 'refers solely to the way in which the language is pronounced and carries no implications whatsoever with respect to grammar and vocabulary'. But this is manifestly overstating the case. A person who speaks with a broad Yorkshire accent can usually be expected also to use or be acquainted with certain north-country expressions which southerners on the whole would not. Indeed, it would be very odd if someone's English grammar and vocabulary were invariably those of a southerner, but the utterances actually produced were pronounced as if by a native of Yorkshire. If there were communities which showed regional differentiation in respect of pronunciation alone, but complete uniformity in respect of grammar and vocabulary, then Lyons' generalization would certainly hold good for such cases. But it is just as false, in the face of the counterevidence available, to assert, as Lyons did, that the term accent 'carries no implications whatsoever with respect to grammar and vocabulary' as it would be to say that the term bank manager carries no implications whatsoever with respect to income and social status. Any theoretical attempt to distinguish rigorously between accents and dialects is probably a reflection of the covert scriptism which underlies so much of modern linguistics (Harris 1980: 6 ff.). In other words, we are dealing here with one of the most long-lived prejudices in Western education: the belief that linguistic essentials are defined by written forms, to which pronunciation merely gives a superficial and variable vocal clothing. The scriptist assumption is that writing captures the ideal linguistic form: it is what all language would be if only it could. This is not to deny, it need hardly be said, that it is often possible to tell from the way a person reads aloud a passage from the editorial in today's Times where, roughly speaking, he or she comes from. But this simply reinforces the point about scriptism made above. Only in a highly literate community does there arise the possibility of achieving this quite artificial divorce between pronunciation on the one hand and grammar-cum-vocabulary on the other, so that we can present people with the explicit task of providing their own pronunciation for a text of which the grammar and vocabulary are supplied in advance. The highly specialized nature of that task needs no further comment. It remains true that in the everyday case of casual speech, an individual's spontaneous patterns of pronunciation, grammar, and vocabulary are interlinked in a way which affords no pretext for pretending that the accent is one thing and the dialect another, with no inherent connection between the two. To summarize thus far, we have in play three different dialect concepts, which we are distinguishing as the continuum concept, the relational concept, and the aggregate concept (see Figure 1). What will now be argued is that the latter two, in spite of their apparent contrast, are actually variants of a single notion, and it is this notion which lies at the heart of the dialect myth. To explain how the dialect myth arose requires us to concentrate first on the relational version. It begins to take shape in the nineteenth century, with the comparative study of Indo-European; and the crucial factor there is the recognition of historical relatedness between the various branches of Indo-European: Greek, Latin, Germanic, and so on. We find that linguists of the period refer to these various linguistic offshoots as different 'dialects' of Indo-European. Similarly, as Crystal (1985: 92) notes in his entry under the heading dialect, it became customary to refer to Cantonese and
The Dialect Myth
89
Figure 1. Dialect concepts
Pekinese as 'dialects of Chinese', irrespective of the fact that speakers of these dialects may find each other's speech mutually unintelligible. The original motivation for the relational concept of dialect in modern linguistics is, then, the recognition of historical - not communicational - relatedness between speech communities. But purely historical criteria for the application of the term dialect soon became outdated. Already by the end of the nineteenth century it sounded rather odd even in philological circles to say that French and Spanish were both 'Latin dialects', or even 'dialects descended from Latin' because the status of French and Spanish as independent national languages took priority over their historical relationship. The term dialect tended increasingly to be reserved for some contemporary subvariety of a national language. Saussure altered all that by drawing his radical distinction between diachronic linguistics and what he called (perhaps unfortunately) synchronic linguistics. 'Historical linguistics' he readily surrendered to history. But synchronic linguistics he proclaimed as the core of linguistics proper, and furthermore proclaimed it to be a branch not of history but of psychology. Now this meant a rather dramatic alteration in the academic status accorded to the linguist. For psychology was supposed to be concerned with the general principles governing the workings of the human mind, a much more exalted intellectual enterprise than merely collecting scraps of information about the vocal customs of one's ancestors.
90
Roy Harris
Saussure justified this change of status by proposing an entirely new program for synchronic linguistics. It was based upon the postulate that languages could be conceived of non-historically or a-historically as autonomous, independent systems of signs, which constituted part of the mental equipment of their users at any given time. On this view, it became irrelevant to consider how linguistic forms had evolved over the centuries. All that mattered, synchronically, was their role in a contemporaneous sign system. In other words, the proposal was that synchronic linguistics should concern itself not with languages but with linguistic states (etats de langue). The difficulty with this new program which Saussure proposed for synchronic linguistics was this: what the lay public, not to mention the comparative philologist, normally counted as languages (e.g., English, French, German) could hardly be seen as constituting sign systems in the Saussurean sense, since there was only partial uniformity of linguistic practice among their speakers. Specifically, it had long been recognized by both expert and nonexpert alike, that not every English, French, or German speaker spoke exactly the same English, French, or German as their neighbours in adjacent regions. So it was difficult to suppose that every speaker in fact had 'the same system' in his or her head. Saussure's solution to this problem was simply to deny that the popular language names ('English', 'French', 'German', etc.) were anything more than crudely approximate labels for what, upon investigation, would turn out to be closely related families of sign systems. And to identify these sign systems or 'linguistic states' which, in his view, ultimately underlay family-relations names such as 'English', 'French' and 'German', Saussure appealed to the dialect concept already familiar in the work of nineteenth century comparative philologists. He said of synchronic linguistics (Saussure [1922] 1983: 128): 'The object of synchronic study does not comprise everything which is simultaneous, but only the set of facts corresponding to any particular language. In this, it will take into account wherever necessary, a division into dialects and subdialects.' This remark is immediately followed by the observation that the term synchronic is not sufficiently accurate, and might profitably be replaced by idiosynchronic. In other words, Saussure saw that just as diachronic facts do not constitute a sign system in his sense, neither do synchronic facts constitute an etat de langue simply by virtue of coexisting simultaneously in time. Otherwise the words dog, chien, and Hund would necessarily belong to the same linguistic system, which stands as an obvious reductio ad absurdum of any narrowly chronological interpretation of the distinction between synchronic and diachronic. Now this is a maneuver with quite remarkable theoretical implications, which have often gone unnoticed. The maneuver is induced, in the first place, by the structuralist insistence that a sign system constitutes a theoretical whole, from which alone the identification of constituent parts is possible. Signs, in other words, have no independent existence outside the system. Hence a language could not be merely a historical accumulation of already existing forms, meanings, constructions, etc. which just happened to be in use at a given time in a given community. Saussure's view of 'the system' is a holistic view. It is inconsistent with this holistic view of what constitutes a sign system to suppose that one and the same sign system could simultaneously both include and exclude a given sign. And that is why regional
The Dialect Myth
91
differences between speakers of English or between speakers of French pose a theoretical problem. If we espouse wholeheartedly the holistic doctrine that Saussurean structuralism preaches, then we cannot avoid the following conclusion. If a typical native of Yorkshire uses English expressions unknown to a typical Devonian, or vice versa, then Yorkshire English and Devonshire English must, in Saussurean terms, constitute two separate sign systems. For the one will include signs which the other excludes. This, in theory, makes Yorkshire English and Devonshire English just as distinct from each other as Eskimo from Swahili or Greek from Latin. Saussure clearly saw and accepted this conclusion when he acknowledged that in order to distinguish one linguistic system from another, the linguist might need to identify particular dialects and even subdialects. Or, to put it the other way round, languages would need to be much more uniform and homogeneous than English, French, and German in fact are in order to qualify as coherent systems of signs. So, faced with the heterogeneity of actual languages, Saussure looked for homogeneity to dialects, or even, as he says, to 'subdialects' (whatever that might mean). Bloomfield, too, was aware of the dialect/subdialect problem, and he perhaps faced it more squarely than Saussure ever did. In the chapter entitled 'Dialect Geography' in Bloomfield's Language (1935: 340-1), we find the following revealing pronouncement: The inhabitants of countries like England, Germany, or France, have always applied provincial names to rough dialectal divisions, and spoken of such things as 'the Yorkshire dialect', 'the Swabian dialect', or 'the Norman dialect'. Earlier scholars accepted these classifications without attempting to define them exactly; it was hoped, later, that dialect geography would lead to exact definitions . . . In this respect, however, dialect geography proved to be disappointing. It showed that almost every village had its own dialectal features, so that the whole area was covered by a network of isoglosses. A local dialect from the center of Yorkshire or Swabia or Normandy could be systematically classed in terms of its province, but at the outskirts of such a division there lie whole bands of dialects which share only part of the provincial characteristics. In this situation, moreover, there is no warrant for the initial list of characteristics. If these were differently selected - say, without regard to the popularly current provincial classification - we should obtain entirely different cores and entirely different zones of transition.
The issue of lay vs. linguistic classification of dialects will be taken up again later. For the moment, it suffices to note that Bloomfield (1935: 341) could see no solution to the problem except to observe that 'without prejudice of any kind, we must attribute more significance to some isoglosses than to others'. This, needless to say, begs the question of how one attributes significance to isoglosses. In any case, it hardly answers Saussure's problem of locating geographically an area within which the investigator can find a single etat de langue in full working order. Saussure, on the other hand, had come to a different view. Instead of attributing more significance to some isoglosses than to others, he goes in quite the opposite direction and concludes that there are as many different dialects as there are different places where a language is spoken (1922: 276). But that hardly solves his problem either. In an excellent analysis of the structuralist dilemma in dialectology, Hutton (1982) refers to Weinreich's suggestion that perhaps some linguistic situations may be
92
Roy Harris
inherently more amenable to drawing geographical distinctions than others, and characterizes the issues for theoretical dialectology as being 'those that concern the interrelation of lay and scientific linguistic categories and their relation in turn to social/sociological (likewise lay and scientific) concepts'. Hutton adds, 'Once the linguist steps outside the hermetically sealed zone of homogeneous linguistics, these problems of categorization become immense'. So immense, one might add, that it is no surprise that later theorists tried to avoid them by turning to an alternative which we have already briefly mentioned: namely, the aggregate dialect concept. Unlike some of his American successors, Saussure did not carry the subdivisions of language down as far as the level of idiolects. Why not? The question is a very crucial one, although, again, its implications have often been neglected. (The question is also more complex than this formulation suggests, but there is no space for those complexities here.) For Saussure the concept of a linguistic system peculiar to just one person was as incoherent as the concept of a private language was to Wittgenstein. Any system of signs, in Saussure's view, had to be the property of a collectivity or community, not of an individual. So there had to be some level of social grouping at which the linguistic system existed; and this level must be, if not national, then regional, or even local. That is how and why the dialect concept comes to occupy such a crucial role in Saussurean theory. A dialect, whether it be of a region, or a locality, or a single village, represents, as it were, the basic level at which, in practice, linguistic diversity is reduced to zero. One can see very easily how Saussure's theoretical preconceptions forced him to adopt this position. One can at the same time see how the dialect myth plays an essential supporting role to the language myth. Saussure's retreat from languages to dialects and even to subdialects in order to find a location for his etats de langue is exactly parallel to the generativist's postulation of a 'completely homogeneous speech community'. A 'completely homogeneous speech community' is by definition one in which there are no dialect differences. The motivation is the same in both cases: to eradicate theoretically the possibility of conflicting evidence, i.e., to eliminate linguistic disagreement between speakers. The object of inquiry, for the orthodox theorist, has to be a 'fixed code' (Harris 1981, 1987); what is intolerable is indeterminacy. The dialect myth of modern linguistics amounts, then, to this: to supposing that at some level of regional or local circumscription we shall find linguistic homogeneity, a uniform system to which all speakers within that particular geographical circumscription have equal access. There are a number of interesting comments which could be made about the plausibility of this myth, and about its acceptance or nonacceptance by dialectologists. Here it suffices to make two points of historical commentary. One is to observe how, under the aegis of modern linguistic theory, the old continuum dialect concept acquires new respectability by emerging as an interpretation of the relational concept. That is to say, the orthodox structuralist-generativist theorist finds it expedient to grant that, in principle, there is no qualitative difference separating the distinction between dialects from the distinction between languages. For insofar as languages are reducible to autonomous systems, they are on an equal footing with dialects, insofar
The Dialect Myth
93
as they in turn are reducible to autonomous systems. So the primitive Greek view of dialect differences was perhaps not so naive after all. The second point concerns the further move from dialect to idiolect, which is characteristic of post-Saussurean theorizing. In effect, Saussure's American successors and Bloomfield's, too, abandoned the attempt to locate homogeneous linguistic systems at the dialect level, and sought instead to locate them at the individual level. There was a double historical irony about this move. The first resided in the fact that this amounted to denying that languages, in any ordinary sense, were the subject matter of linguistics at all. This denial is nowadays commonly endorsed. For instance the authors of Modern Linguistics: The Results of Chomsky's Revolution are driven to make the claim: 'we cannot talk of the grammar of English, but only of the grammars of different speakers of English' (Smith and Wilson 1979: 26). So English, in other words, has no grammar; and if it has no grammar, it cannot be a language. Here we see how, if you toss the language-myth coin often enough, it will sometimes come down 'tails'. The other irony attached to the subsequent realization that, in moving from dialect to idiolect, the theorist had not moved far enough. In other words, the speech of one individual is no more homogeneous than the speech of a collection of individuals. For the simple fact is that individuals change their style of speech depending on who they are talking to and under what circumstances. The expressions a schoolboy uses when talking to his headmaster often bear no resemblance to the expressions he uses when talking to members of his peer group in the playground, nor a husband when talking to his wife as compared with his remarks to teammates in a cricket pavilion. So yet a further shift was inevitable: from the idiolect to the style (where style is defined as that subvariety of an idiolect used in a particular set of communication situations). The validity of this paradoxical move is vouched for explicitly by theorists of repute. Lyons (1981: 27), for example, says: There is a sense . . . in which every native speaker of a language is stylistically multilingual. Just as it is, in principle, possible to think of each idiolect as a separate language-system, so it is possible - and no less reasonable [my emphasis] - to think of each distinguishable style as a distinct language-system.
Here we see how the dialect myth in the end gives place to the style myth. All of them - language myth, dialect myth and style myth - belong to the great chain of attempts to identify somewhere or other in the manifold complexity of human speech something that might pass for a determinate system of verbal signs. It will by now be obvious that I believe this search to be vain; and the plurality of unsuccessful attempts at identification I would offer as an indication of that vanity. What, then, is the truth about dialect differences, as opposed to the myth? The truth, as often, is quite simple, and has been long known. Speakers often recognize themselves as typically differing in their speech habits from other speakers of 'the same language' living in a 'different place'. That recognition is the intellectual genesis of dialect concepts. Where such concepts break down is at the point where theorists attempt to generalize in order to construct hypothetically independent systems corresponding to these differences. In short, it is the illegitimate extrapolation of linguistic systems called 'dialects' from the acknowledged recognition of 'dialect
94
Roy Harris
differences' which lands dialectology in the theoretical morass where it currently flounders. My conclusions on this score are, I think, in accord with the view taken by more pragmatically minded dialectologists. But it is clear that the tension between the exorbitant requirements of the theoretical linguist and the actual view of native speakers is a sore point in modern dialectology. A final illustration of this is a paper by Bjorn Jernudd (1968) entitled 'There are no subjective dialects'. This paper actually proposes a methodology for resolving the conflict between how linguists draw dialect distinctions and how native speakers draw dialect distinctions. Jernudd, in fact, concedes that the linguist's techniques for drawing dialect boundaries are arbitrary, no matter whether they are based on mapping isoglosses or, on the other hand, assessing mutual intelligibility between different areas. He cites Wurm and Laycock's finding that in New Guinea, 'in an area encompassing the greater part of the three Highlands Districts, the number of distinct languages encountered there could be said to be 48 or 26 according to what linguistic criteria were applied to distinguish between languages and dialects' (Wurm 1966: 10). One is reminded of similar numerical disparities between different counts of the phonemes of English, depending on the phonemic criteria adopted. To anyone who has read Saussure (but Jernudd never mentions Saussure) all this has an air of deja vu about it. Saussure, before Bloomfield, explicitly conceded that in practice it is impossible to draw geographical lines demarcating one dialect from another. But Saussure was not naive enough to suppose that the problem could be resolved by the kind of experimental technique which Jernudd advocates. This technique involves playing random selections of tapes to randomly selected native speakers, and trying to determine what objectively present linguistic features actually prompt the audience to distinguish between one dialect and another. The last refuge of the dialect myth is to suppose that if only our scientific techniques of analysis were better, then we would indeed discover that - lo and behold - dialects 'really' existed all the time. In a paper published in 1962 Ivic (p. 34) wrote: 'For decades the classic and most controversial question of dialectology has been: 'Do dialects . . . actually exist?'. Many years before, Schuchardt - perhaps the most brilliant linguistic scholar of his day - had given an emphatically negative answer to that question. From my own experience of fieldwork in dialectology, the best evidence I can cite in support of Schuchardt's answer was given to me by an old man whom I asked whether the patois of his Alpine village was the same as that of another locality a few miles distant. I here translate his reply, given in (what I at that time called) 'Valdotain': Is it the same? I would not know how to answer you. Even in this village the younger people speak differently from my generation. And in the next valley perhaps they use words we don't use here. But, for all that, everyone understands everyone else well enough. Is that the same?
By turning my own question back on me, he had made me understand that the mistake lay in the question. What I was asking corresponded to nothing in his own linguistic experience which could provide a determinate answer. When theorists begin to ask unanswerable questions about language (or - which amounts to the same thing - questions which can be answered 'yes' or 'no' as you
The Dialect Myth
95
please) that is the surest indication that in their investigations linguistic myth has taken over from linguistic reality.
References Bailey, C.-J. N. (1973) Variation and Linguistic Theory. Center for Applied Linguistics, Arlington, Virginia. (1981) Theory, description and differences among linguists (or, what keeps linguistics from becoming a science). Language & Communication 1, 39-66. Bloch, B. (1948) A set of postulates for phonemic analysis. Language 24, 3-46. Bloomfield, L. (1935) Language. Allen and Unwin, London. Crystal, D. (1985) A Dictionary of Linguistics and Phonetics, 2nd ed. Basil Blackwell, Oxford. Diogenes Laertius. Lives of Eminent Philosophers, trans, by R. D. Hicks. (Loeb Classical Library, 1925.) Heinemann, London. Gillieron, J., & E. Edmont (1902-10) Atlas linguistique de la France. 13 vols. Champion, Paris. Harris, R. (1980) The Language-Makers. Duckworth, London. (1981) The Language Myth. Duckworth, London. (1983) F. de Saussure, Course in General Linguistics, translated and annotated by R. Harris. Duckworth, London. (1987) Reading Saussure: A Critical Commentary on the Corns de linguistique generate. Duckworth, London. Herodotus, trans, by A D. Godley. (Loeb Classical Library, 1921-24.) Heinemann, London. Hutton, C. M. (1982) Dialectology and linguistic theory. MS. Ivic, P. (1962) On the structure of dialectal differentiation. Word 18, 33-53. Jernudd, B. (1968) There are no subjective dialects. Kivung: Journal of the Linguistic Society of the University of Papua and New Guinea 1, 38-42. Liddell, H. G., & R. Scott (1996) A Greek-English Lexicon, rev. ed. Clarendon Press, Oxford. Love, N. L. (1984) Psychologistic structuralism and the polylect. Language & Communication 4, 225-40. Lyons, J. (1981) Language and Linguistics. Cambridge University Press, Cambridge. Reed, C. E. (1967) Dialects of American English. World Publishing Company, Cleveland. (Reprinted 1973, University of Massachusetts Press, Amherst.) Saussure, F. de (1922) Cours de linguistique ggnfrale, 2nd ed. Payot, Paris. Smith, N., & D. Wilson (1979) Modern Linguistics: The Results of Chomsky's Revolution. Penguin Books, Harmondsworth. Wells, J. C. (1982) Accents of English. Cambridge University Press, Cambridge. Wurm, S. (1966) Papua-New Guinea nationhood: the problem of a national language. Journal of the Papua and New Guinea Society 1, 7-19. & D. C. Laycock (1961) The question of language and dialect in New Guinea. Oceania 32, 2.
Integrating Languages NIGEL LOVE
Giving an integrational account of the concept of a language involves reconciling two apparently contradictory facts. The first fact is that, for many cultures and societies, the world of linguistic phenomena is populated by 'languages', where a language is a fixed system of units (words, sentences, etc.) recurrently instantiated in the utterances (speech, writing, etc.) of members of a given linguistic community which, by serving as a device for encoding and decoding meanings, enables them to communicate their thoughts to one another. This conception is not only ingrained in the Weltanschauung of participants in the cultures in question, but is also underwritten by the academic language studies sponsored by those cultures. And the second fact is that languages, in this sense, seem to have no objectively discernible existence. Indeed, postulating their existence runs counter to certain obvious features of the everyday linguistic experience of members of the societies concerned. What is objectively discernible is that first-order language-use, in whatever medium, involves producing utterances that are unique in respect of their material substance ('form'), their communicational effect ('meaning') and the context in which they occur. The main question to be addressed here is the relationship between unique utterances and the underlying linguistic entities whose existence in some sense or other is acknowledged by both language-users and descriptive linguists. Are these entities given prior to first-order utterances? If so, how are they 'given', and by whom? Or are they indeed abstractions, in the full sense of the word? If so, how are they abstracted, and by whom? If every utterance is unique in respect of both acoustic and communicational effect, there is no hope of identifying or determining the linguistic units that they allegedly instantiate by mere empirical inspection. This fact alone suffices to bring into play the hypothesis that linguistic units are in fact indeterminate. But that assertion is not entirely perspicuous as it stands. What is indeterminacy, in this context? In the most ordinary kind of case, the abstract entity underlying an utterance is specified by producing another utterance which can be taken as identifying the type of which the first utterance was a token. Such a metalinguistic performance may or may not be underpinned by the use of explicitly metalinguistic expressions such as 'word', 'sentence', and so on. So, assuming for the sake of an example that A is
96
Integrating Languages
97
producing a spoken discourse corresponding to this written text, if B asks A to repeat what he has just said, A might say, simply, 'and so on', or he might say 'I used the phrase and so on\ In either case A is specifying, or attempting to specify, what the utterance was an utterance of. An organised practice of systematic talk about the things that utterances are utterances of clearly requires the extensive use of overtly metalinguistic expressions. (A linguistics that stopped short at specifying them merely by repeating utterances would not be very interesting.) Hence the analytic description of linguistic units in terms of phones, phonemes, morphs, words, sentences, etc. So we say that an utterance 'cats mew' is the utterance of a sequence of phones identified by writing certain symbols within square brackets which in turn manifest a sequence of English phonemes, identified by writing a marginally different series of symbols within slant brackets; it is the utterance of a sequence of English morphs, which in turn constitute English words, identified by writing certain symbols in italics, which in turn constitute an English sentence, whose grammatical structure can be shown, for instance by drawing the appropriate tree diagram with labelled nodes. Now it might be said that the difference between identifying a type by repeating an utterance and identifying a type by talking about the words, sentences etc. that the utterance realises corresponds to the fact that if utterances are envisaged as tokens, they must be tokens of at least two distinct kinds of type. In fact, the distinction between a type and a token (that is, between a linguistic abstraction and its concrete realisation) can be drawn in indefinitely many different dimensions. For present purposes this, along with many other issues concerning the type-token distinction, will be ignored. 1 In what follows attention will be focussed on the relation between first-order utterances and any kind of abstraction that first-order utterances might be held to instantiate: it will make no difference (or at any rate for purposes of this discussion it will be held to make no difference) whether cats mew is thought of as an utterance-type, or whether it is thought of as an English sentence (or as a combination of two English words or of three English morphs, or whatever). What is involved in claiming that these abstractions are indeterminate? It should be said at the outset that there is nothing of interest to be gained from arguing about indeterminacy in connection with analyses of particular utterances of theirs that particular utterers might choose to give on particular occasions. That is to say, it is presumably open to a speaker simply to provide (somehow) a determinate identification of the units underlying an utterance of his; and while we may wonder what he thinks he is doing in doing that, there would be little point in denying his right to live in a private world in which his utterances are the surface manifestation of a determinately structured idiolect, if he so chooses. The question of determinacy has to be raised in the light of the idea of a public language whose units, if they exist, are determined or not determined, as the case may be, for all members of the community who share that language. There are at least two rather different claims that might be being made by someone who asserts, in respect of a communal language, the indeterminacy of linguistic units. Let us refer to them as the 'weak' claim and the 'strong' claim.
1 For detailed discussion of types and tokens see Mutton 1990, Harris 1996. Another matter that will not be entered into here is the difference between the type-token distinction and the class-member distinction.
98
Nigel Love
The weak claim is that although a language is indeed to be thought of as a system of abstract units, the system is not determinately identifiable because there are penumbral areas where doubt reigns either as to what the units are, or as to what the limits of the particular language are. A familiar example of the first kind of doubt is engendered by the difficulty of drawing a firm line between homonymy and polysemy. Is there in English a single noun cat, with (at least) both a restricted and a more extended sense? Or are there (at least) two different words which happen to have the same form? On the answer to that question, of course, will depend the size of the inventory of units of English identified at the word-level. Examples of the second kind of doubt are provided by generative grammarians when they stigmatise certain would-be sentences, not with the asterisk that indicates syntactic illformedness, but with the question mark that denotes doubt as to the grammatical status of the item concerned.2 The strong claim is that utterances themselves are not determinately analysable by reference to a set of units, and that consequently, whether that set is determinately identifiable or not is either a question that simply does not arise, or one that only arises on a different plane of linguistic discourse from that which is concerned with the analysis of utterances themselves. The difference between these two claims is that the first finds indeterminacy in the system of units itself, whereas the second finds indeterminacy attaching to the phenomena whose properties, it is alleged, lead to the setting up of a system of units in the first place. Many descriptive linguists are quite happy to admit the weak claim, because they deny, or fail to see, that anything of importance follows from admitting it. For instance, for a variety of reasons it is often difficult to say which phoneme of a language a given stretch of utterance should be referred to. Is the schwa sound in the third syllable of an utterance 'I want to go' an instance of the phoneme that we have in the first and third syllables of banana or of the phoneme that we have at the end of zool As it stands the question is unanswerable, because the answer depends on how abstract a linguistic unit the phoneme is supposed to be, and that is a matter on which phonologists are divided. But the descriptive linguist can and does treat a problem of this kind as a mere local difficulty requiring from him simply a procedural decision as to how to apply his particular concept of the phoneme. As far as he is concerned, it need no more call into question the existence of the phoneme as a unit in terms of which languages are structured than the fact that there are places on the border between Belgium and the Netherlands whose location in one or the other has never been decided calls into question the existence of national boundaries as entities in terms of which the political world is structured. In other words, far from pointing to a need for a radical redefinition of his subject matter, indeterminacy is something that the linguist may well decide that he can live with. He is apparently not ineluctably bound to take it that the indeterminacy of linguistic units demonstrates their non-existence. On the contrary, he may feel free to hold that utterances are for the most part readily analysable as manifestations of abstract units, and that such units can be more or less reliably identified and 2 Also sometimes encountered is the prefix '%', which means 'accepted as well-formed by only a certain percentage of ... speakers' (Radford 1988: 9).
Integrating Languages
99
inventoried at the various levels of linguistic description, and hence described in terms of the two planes of 'form' and 'meaning'. On this weak interpretation of the indeterminacy claim, the fact that linguistic units are in various ways hard to pin down may simply be - and frequently is - admitted without any concomitant admission that indeterminacy destroys the theoretical basis for descriptive linguistics. That this indefensible attitude should be so widely adopted demonstrates the power of the fixed-code theory of languages in alliance with the thought-transference theory of communication. In attempting to tease out the source of that power it is first necessary to admit the cogency of the strong indeterminacy claim (which means that the weak claim falls away or at any rate becomes irrelevant). No viable descriptive science of spoken language can be based on the idea that utterances are to be understood as the outward manifestation of members of a determinate set of underlying abstractions. That is ultimately because utterances are not, as a matter of fact, the outward manifestation of members of a determinate set of underlying abstractions. Given the physical and circumstantial uniqueness of every utterance, the descriptive scientist has no basis for disengaging from the incessant flux of speech the recurrent invariants he is seeking. And the reason he is mistaken in looking for such invariants is that, given the human situation, there is no mechanism whereby language-use could have come to be a matter of producing and understanding tokens of them. The idea of a language as a fixed set of abstractions raises the questions: who fixed them? How? When? It implies prior agreements. But there are no such prior agreements. Language is radically indeterminate, as regards both what is meant and what is said. The use of spoken language involves an incessant process of guesswork as to the significance of the vocal noises we hear one another make, against the background of such general ideas as we may entertain as to the sort of creatures we are and what, in given circumstances, our behaviour is likely to be. We communicate successfully by means of language to the extent that we achieve whatever may have been the purposes for which, on a particular occasion, we used it. But successful communication is not something for which there are criteria external to our own understanding of the interactional situation in which we find ourselves. A language is not a device which, if operated in accordance with the instruction manual, automatically yields something objectively discernible as 'communication'. To the extent that linguists describe structured systems of entities called 'languages' with a view to laying bare the nuts and bolts of the mechanisms with which we communicate, they are in error. There are no such things as languages, in this sense, to be described. This is a reasonable starting-point for an account of the human situation vis-a-vis language. It provides a background against which to raise the question: 'Just what, in that case, are these things called 'languages' that have been thus summarily done away with? Suppose you have a tape-recording of A's utterances so far. Now the strong indeterminacy claim implies the impossibility of analysing those utterances in terms of linguistic units: of saying what the utterances are utterances of, of saying what the phonemes, morphemes etc., are. Yet it is manifestly the case that a competent phonetician could transcribe the recording, and that a competent phonologist could
100
Nigel Love
provide a phonemic analysis of the transcription; that any literate English speaker could, with a certain amount of editing, provide a corresponding stretch of written English prose, of which a syntactician could provide a grammatical analysis. In other words, there is a flourishing practice of analysis of utterances in terms of linguistic units, some phases of which can be undertaken by almost any literate lay speaker of the language in question, let alone by the professional academic linguist. How can this be possible, granted the truth of the strong indeterminacy claim? To answer that question would be to lay bare the dynamic of the fixed-code-cum-thoughttransference theory of language and communication. What must be squarely faced, in attempting to answer it, is that our practice of performing such metalinguistic operations as have just been mentioned is as much a fact about the world of linguistic phenomena inhabited by members of societies such as ours as is the radical indeterminacy, with respect to descriptive linguistic analysis, of first-order utterances; and the two facts must somehow be brought into coherent conjunction if an 'adequate account of linguistic experience in such societies is to be had. What the linguist calls 'a language' cannot merely be dismissed as the result of his having made a lot of mistakes about the nature of his subject matter. Languages must, in other words, be naturalised. One way of approaching this task is to consider certain logically imposed connections between utterances and abstractions. However deep our ignorance about the origin of language, there are certain things one can reasonably say about it which do not depend on access to unavailable historical information. For instance, it can reasonably be said that language must have started when a primordial A first spoke, and a primordial B understood. There is no question of B understanding A's utterance in virtue of being able to relate it to antecedently given abstract units, because there could not have been any such things. The first utterance logically cannot have been the utterance of something antecedently given. So what can 'understanding', at this stage, mean? It means, perhaps, that A's vocal noise elicited from B behaviour that suggested to both A and B that associations (images, memories, etc.) somehow evoked in B by A's noise were similar to the associations of the noise for A. Now all that happened, on the phenomenal level, is that a noise was made and certain behaviour ensued. And, in a sense, that is the beginning and the end of the story of spoken language: ever since, all that has ever happened is that noises have been made and behaviour has ensued. But there is more to be said than that. The origin of language cannot have antedated the capacity in humans for reflecting on their experience. If using language is anything at all beyond making noises, language-use presupposes that capacity. And there is no reason to doubt that A and B will have reflected rather furiously on this utterly novel experience. Thus the birth of language as an object of contemplation follows hard on the heels of the birth of language itself. It is not a necessary feature of this account that A intended his noise to have the effect it had. In fact, it is hard to imagine how he could have, since there was ex hypothesi no established practice of eliciting certain behaviour from others by making vocal noises. But it is no less hard to imagine how even primitive man could have
Integrating Languages
101
failed to entertain the notion - and sooner rather than later - that behaviour-eliciting vocal noises might be made on purpose. The obvious way to test this possibility would be to make a second, similar noise and see if it elicited a similar response. Now there was no fixed criterion for deciding what might count as a 'similar' noise. (Nor has there subsequently ever been.) Nonetheless, if a range of similar noises was tried out and the predicted response, or type of response, was forthcoming, it is hard, again, to see how even primitive man could fail to entertain the idea of a class of noises, or type of noise, which would regularly elicit a class of responses, or type of response. Thus was the first utterance-type invented. The defining characters of this type were not fixed, and never would or could be. Nonetheless, the basic idea of an invariant abstraction - a class or a type - underlying utterances is a necessary feature of any attempt to make sense of even the most primitive linguistic event. If, as will be argued below, the essence of the concept of 'a language', as distinct from 'language', is the idea that there are things which utterances are utterances of, that idea is bound to arise as a result of no more than the human capacity for reflecting on experience, generalising about it, and purposefully attempting to renew it. This is why the alternative to a linguistics based on languages as determinate sets of abstractions cannot be a linguistics of individual utterances themselves. That would leave out of account the fact that for language-users, utterances are utterances of things. It would be absurd to claim that language-use does not involve a distinction between types and tokens at all. To see why that is so, consider a necessary condition of generalising discourse about anything whatever. In order to make any general statement one must indulge in abstraction. Spatiotemporally distinct objects and events must be envisaged as 'the same' in some respect relevant to the purposes of the discourse. Something like a type-token analysis is required. If one is to talk about, say, cats, distinct individual organisms have to be seen as tokens of a type 'cat'. This is true for languages no less than for cats. For instance, for a question such as 'is what time is dinner? a grammatical English sentence?' to be even askable, one must avail oneself of the possibility of citing that form of words to indicate the object about which the question is being asked. This capacity of language to facilitate generalising talk about the world is unremarkable, in the sense that it is so basic to our understanding of how language is used as to make it impossible to imagine what things would be like otherwise. It is a capacity made use of in our everyday talk about language itself. When A mishears B and asks him to repeat exactly what he said, B may understand A to be requesting, not the logical absurdity of an attempt to recapture the original unique utterance, nor even an acoustically identical replica of it, but merely another token of the same type. That is, that utterances can be treated as recurrent instantiations of underlying invariants is an assumption that we take for granted. What is problematic is the further assumption that the analysis into types and tokens - what counts as an underlying invariant, and what counts as an instance of a particular invariant - is uniform for a whole linguistic community, determined for every member of that community by 'the language'. Linguistic theorists sometimes demonstrate how problematic this assumption is, by ignoring it. Consider, for instance, N.V. Smith's study of his son's acquisition of
102
Nigel Love
English phonology (Smith 1973). The general argument is that at all stages what is psychologically real for the child is the adult surface forms. The child's deviant pronunciations of adult words are explained by supposing that the adult forms are the input to a set of 'incompetence rules'; and the narrowing of the gap, as the child gets older, between his pronunciation and the adult's, is a matter of the progressive elimination of those rules. So, for instance, at two years sixty days the child had rules converting [thel9foun] to [dewibu], while at two years 260 days he was saying [thelibu:n], where the number of rules required to take him from [thel9foun] to his own pronunciation is rather smaller. Of interest here is what this account reveals about the linguist's assumptions. For Smith senior offers no explanation of how he is able, quite without hesitation, to take e.g. [dewibu] as a token of the type telephone. The question is not how he knows that [dewibu] means 'telephone', but what authorises him to recognise it as instantiating the English word telephone. It is far from clear that any describer of English in general would admit [dewibu] as an acceptable realisation of telephone, or that English-speakers, in other contexts, would be understood if they used it. Types and tokens are undeniably 'real' for speakers, but what the types and tokens are is something for speakers themselves to decide in particular contexts. When A asks B to repeat exactly what he said, he is speaking metalinguistically. If, on the other hand, he merely asks him to repeat what he said, he is not necessarily speaking metalinguistically; and it is significant that now the response may be a different form of words entirely. All that is demanded in this case is another utterance perceived as similar to the first in respect of playing the same role in the communicational exchange. But that, in the end, is all that A gets even when he makes the more specific demand. He may think he is being offered a token of the same type (not that he would necessarily use that terminology). That is because he takes for granted the type-token distinction as a necessary concomitant of discourse about language. But if one asks what its being a token of the same type actually consists in, there is no answer, except that it conforms to the idea that it is: an idea validated ultimately by nothing more than the circumstances of the communicational interaction within which, and for purposes of which, it is entertained. The idea that language-use is a matter of deploying a predetermined system of types and tokens is so entrenched that the two terms 'language' and 'a language' are often used interchangeably. Weinreich, Labov and Herzog, for example, say (1968: 100) that 'it will be necessary to learn to see language as an object possessing orderly heterogeneity', which only makes sense if for 'language' we substitute 'a language'. The same applies to the title of Katz 1981: Language and Other Abstract Objects. To take another example, the title of Cooper 1975 (Knowledge of Language) suggests that the term 'language' is at least initially being apprehended in its 'mass' rather than its 'count' sense. But then, without further ado, the question of what knowledge of language might be is straightaway construed as the question: what is it to know a language? Similarly, in Linguistic Behaviour, Bennett (1976) starts off his preface by saying: 'this book presents in some detail a view of language - that is, language in general, not any particular tongue'. This at least recognises that there is a distinction to be drawn. But then what immediately follows this preface is a discussion of an imaginary planet whose inhabitants have a language. But if we wish to understand the
Integrating Languages
103
role of the individual in relation to language, it may be worth resisting this facile assimilation, at least until we have attempted to consider from first principles the relation between the two concepts 'language' and 'a language'. The first point is that there is a sense in which 'language' has priority over 'languages'. This is true both as regards the origin of language in general, and as regards the linguistic initiation of the individual; and its truth depends on a simple point of logic. An initial act of understanding language must precede any analysis of the utterance concerned in terms of a language to which it might be held to belong. This is a point of logic in as much as the alternative view - that knowledge of the relevant language (or the relevant part of the relevant language) is a prerequisite for understanding an utterance - leads to an infinite regress. There is no way of getting started at all, unless, as has already been suggested, a primordial A first says something and B understands it. In this view the notion of a language (as distinct from language) arises with the perception - or idea - that utterances are repeatable. To embrace the possibility of 'saying the same thing' is automatically to introduce an abstraction underlying actual utterances: namely, the thing which is held to be repeated. Perceiving utterances as manifestations of underlying 'sames' is not a necessary condition of any use of language whatever. It must be at least possible to understand an utterance without relating it to an antecedently given underlying abstraction. Nonetheless, once one has understood an utterance for the first time, one will entertain the possibility of repeating it. This involves deciding what would constitute repeating it. One fundamental source of variation within a given linguistic community is the fact that what counts as repeating an utterance is not fixed. An utterance does not somehow of itself indicate which of its features require imitation in order for a second utterance to be taken as 'the same'. What constitutes 'saying the same thing' depends on the kind of sameness required. At one end of the scale there may be situations in which saying 'je ne sais pas' counts as saying the same thing as 'I don't know'. Nearer the other end, the speaker of a variety of northern British English who asks a southerner to say [giass] may, or may not, consider that the southerner has failed to respond appropriately if he says [gjcus]. Whether the utterance [giaes] is to be seen as instantiating an abstraction that could alternatively be realised as [gia:s], or whether the abstraction concerned is quite different from the one underlying [gjais], is not determinable in the abstract. A language, as an individual's system of repeatable abstractions underlying language-use, is something that he creates for himself in the light of the constantly shifting situations in which he interprets and produces utterances. At no point, for him, does the system become fixed. (This is tantamount to saying that there is no system.) The abstractions underlying utterances, and hence what utterances count as 'the same', are matters for perpetually revisable decision. The relation between languages, language-users and linguistic variation must therefore in outline be as follows. A language is a second-order construct arising from an idea about first-order utterances: namely, that they are repeatable. Such a construct may be institutionalised and treated as the language of a community. But it remains a construct based on an idea: at no point does it become a first-order reality
104
Nigel Love
for individuals. (Although in a society which teaches its institutionalised construct to its members it may be expected to have a large effect on their first-order behaviour, and may perhaps give rise to linguistic theories which project the construct on to them as the basis for their first-order behaviour.) Individuals, however, entertain the idea that utterances are repeatable. (After all, it is only because they do so that languages could ever be codified.) But the ways in which the idea can be implemented - that is, the abstractions that can be established by implementing it - are not fixed. Hence there arises variation. Variant abstractions may themselves come to be codified. Hence what is perceived as synchronic variation, and eventually, diachronic change. The language-user's capacity to make different decisions as to what an utterance is an utterance of is both a source of variation and a bar to determining the individual's relation (qua first-order language-user) to an abstract system which can only be envisaged at all on the assumption that we already know what utterances are utterances of. Acknowledging this fact must be one necessary step towards a satisfactory conceptualisation of the relations between languages, language-users and linguistic variation. A second step is to recognise that the perspective on the communicational process introduced into modern linguistic theory by Saussure's famous account of the circuit de la parole (Saussure 1922: 27 ff.) is distorted. It encourages the idea that what needs to be accounted for where communication is concerned is the activities of just two individuals, a speaker and a hearer. But languages are characteristically used by communities with many more members than two. Language-users are not just participants in their own communicative acts, but also observers of the communicative acts of others. In respect of his own communicative acts, the individual will entertain shifting and tenuous notions of the abstractions instantiated by his own and his interlocutors' utterances. In the case of communicative acts of which he is a non-participating observer, if he understands them he may well relate the utterances involved to his own conception of comparable utterances in his own personal experience. But the important point is the development of a particular conception of what, in general, linguistic communication is for a whole community. Hence the emergence of a neutral observer's view of the linguistic process as a community-wide process. Such an observer's view is a crucial prerequisite for any idea of a language as a communal possession; and all that it requires is that individuals project on to others the kind of understanding they have developed of the linguistic interactions they themselves engage in. It should be noted that the existence of this 'observer's view', in itself, is quite independent of the possibility of saying, for a large or infinite range of cases, what abstractions are in play in particular interactional episodes, and hence of articulating the supposed structure of a supposed communal language. So long as language is purely spoken language, there is no possibility of doing that, partly because purely spoken language is purely context-bound language, and partly because purely spoken language cannot yield the consistently reliable metalinguistic discourse required for articulating it. The crucial difficulty where generalising orally about linguistic experience is concerned is that we have nothing to do it in except vocal utterances themselves. So
Integrating Languages
105
how do we make clear that some of our utterances are to be taken as utterances, whereas others are to be taken as the names of something more abstract than utterances? One may suspect that primeval attempts at talk about language were considerably hampered by the fact that utterance-type-identifying tokens of a type are simply tokens of that type. Nonetheless, some talk about language must have gone hand in hand with language itself from the very outset (as it does, of course, in the linguistic initiation of individuals today). Utterances are phenomenal events which play a role in interactional episodes. But they are also objects of contemplation. And if they can be contemplated, they can be talked about. But systematic talk about them requires a radically new development. That new development was the use of writing as a linguistic medium. The first point to make about writing in this context is that the very fact it came to be used linguistically at all is evidence of a habit of trying to make sense of our linguistic behaviour by seeing it as involving the recurrent instantiation of abstractions. The notion that a non-phonic medium might be used to set down language already implies the idea of linguistic units as something more abstract than phonic utterances themselves. The inventors of phonetic alphabets have tried to circumvent the inherent abstraction-implying nature of writing by devising more and more detailed modifications of the alphabet in their quest for ever-narrower transcriptions. But the fact is that there is no end to the search for the perfectly narrow transcription. Writing embodies from the outset the idea that spoken languages are systems of abstractions: it cannot handle actual utterances at all.3 Developing a written counterpart to spoken language removes some of the most crucial difficulties attaching to a purely oral practice of metalinguistic discourse. For although type-token ambiguities may arise for writing as for speech, writing provides a firm anchorage for at least one dimension of type-token distinctions, by providing a medium for displaying types which is different from the medium in which the corresponding tokens are produced. It introduces a new level of clarity into attempts to show what the abstractions are, by providing a system of types in terms of which, in literate societies, utterances will henceforth be interpreted. Another important point about writing is that inscriptions are not bound to interactional contexts in the way that speech is. Their physical nature makes them both enduring and portable. But how can these attributes be exploited in a practice of using writing to set down spoken language, given that the interpretation of spoken language is contextually determined? The answer is to combine the idea of utterances as the surface manifestation of underlying abstractions with the idea of a communitywide practice of utterance operating in this way, and simply impose a context-free oral interpretation on written forms. Everyone already knew that his utterances, and presumably everyone else's, were utterances of something. The development of writing is not a matter of revealing what it was that utterances had all along been utterances of, but of providing a basis for saying what, thereafter, utterances would be counted as utterances of. How could it be otherwise? How could a fixed set of units ever be abstracted from the utterances of a purely oral language? Given a certain indefinite range of 3 This is not to deny that an inscription is an utterance, and as such may be understood as a token of a written type.
106
Nigel Love
different noises, and an indefinite range of variation as to their contextual interpretation, how could it be decided that one particular subset of elements from this variety should be gathered together, represented, referred to or named as instances of a single entity embodied in a visually identifiable graphic form - let us say, CAT! It is hard to imagine even a potential answer to this question. It could not have been a matter of the inventors of writing gathering together a variety of different individuals' privately entertained decontextualised linguistic abstractions and then supplying a written entity to label them. It must have been the other way round: providing a written entity which then defines, or - initially - points to, a range of actual and potential utterances with which it thereafter is to be associated. If this is so, then there is a sense in which the linguistic use of writing involves the explicit reinvention or recreation of languages. It might have happened like this. Teaching a writing system involves oral explanation of the written forms. It involves saying what piece of oral language certain graphic marks are, and conversely, indicating what graphic marks the system provides for writing a certain piece of oral language. But as far as the initial phases of the development of a linguistic writing system is concerned, only the first procedure is available. It makes no sense to ask 'how do I write catT until writing has been instituted. The instituting of it must be a matter of, as it were, saying: 'if you do this (sc. write CAT), you write cat'. So what is meant by saying that writing involves the reinvention of language is that what linguistic interpretation is initially attached to a given inscription will depend on what the 'ur-writing demonstator' can get his audience to understand by his oral explanation: that is, what range of images, associations, memories of similar utterances, etc. his meta-utterance conjures up for that particular audience on that particular occasion. This range is unlikely to be the same for each member of his audience, and is even less likely to match the range of images, etc. conjured up by similar utterances on other occasions to other audiences. So what CAT 'is' must at the outset be quite indeterminate. The first embryonic writing systems must have been semi-private modes of communication. But once a system has begun to be worked out, its use can and will come to be taught by means of the idea that particular utterances can be related to particular written forms. At this point writing achieves the object of fixing the identification of utterances (in the limited sense of providing them with something outside speech itself to which they can be referred) by simply laying down what it is your utterance was an utterance of. If you write it CAT, then that is what you said. 4 And given the context-free, community-wide invariance of a writing system, the way is now open to the idea of a language as a context-free communitywide system of signs the indeterminacy of whose manifestation in speech can be explained as a mere imperfection of the oro-aural medium. This idea lies at the heart of modern linguistic theory. Bloomfield, for instance, asserts it in the brusquely straightforward form of what he calls 'the fundamental assumption of linguistics': namely, that 'in certain communities (speech communities) some speech-utterances are alike as to form and meaning' (Bloomfield 1935: 144). For anyone who accepts this waving of the theoretician's magic wand as providing a solution in principle to the problem of how to identify the recurrent 'sames' on which 4 Nothing in this schematic speculation about the development of writing should be taken to imply the historical primacy of the alphabet.
Integrating Languages
107
the kind of linguistic analysis Bloomfield envisages depends, the obvious next question is how it is to be done in practice. Which speech-utterances are alike as to form and meaning? The tenor of Bloomfield's discussion is such as to suggest that, for him, this was either never a serious question or, if it ever was, it has long since been satisfactorily answered - by the inventors of our writing system. Two utterances 'there's a dog howling in the yard' and 'there's a dog howling in the yard' count as 'alike as to form and meaning'. Two utterances 'what time is dinner?' and 'fish is good for you' count as unlike in both form and meaning. Whatever may be the criterion by which these judgements are arrived at, Bloomfield apparently sees no need for the linguistic theorist to examine them. Whoever invented alphabetic writing has already performed the necessary analysis: the linguist need do no more than take over its results as embodied in our standard spelling system. Part of the reason for this attitude is the fact that descriptive linguistics is itself an enterprise conducted in the written medium, which makes it difficult to get beyond the analysis that the medium itself imposes. Bloomfield does not, of course, explicitly admit that his would-be 'scientific' investigation of the structure of languages starts by assuming as unquestionably correct the analysis arrived at by those who first used writing as a linguistic medium. But it is nonetheless clear by implication, as when we are told that although morphemes are minimal meaningful combinations of phonemes, there are 'phonemes which do not appear in any morpheme, but only in grammatical arrangements of morphemes' (Bloomfield 1935: 162). The phonemes of which the morpheme cat consists will, in any actual utterance, be accompanied by one or more 'secondary' phonemes of stress, pitch, intonation, etc. As Bloomfield recognises, there is no 'indifferent or abstract' form which consists of the morpheme cat without any accompanying secondary phonemes. Or, to put the point another way, there is such a form, but it is an abstraction, for which there is only one plausible source, the written word cat. The idea that stress, pitch, etc. are secondary, superficial flesh on a primary skeleton consisting of the morpheme cat in the abstract, seems to arise from the circumstance that orthography does not standardly provide any indication of stress, pitch, etc. Writing drastically reduces one kind of indeterminacy in spoken language. That is to say, within the limits beyond which a writing system fails to offer an unambiguous identification of different spoken units, it eliminates, for literate individuals at least, doubts as to what abstractions they are supposed to refer what they say or hear to. But it does not eliminate indeterminacy of interpretation. That, of course, never could be eliminated merely through the transposition of language into another medium of expression. So what fostered the idea, not only that the units of a language are determinately identifiable, but that associated with them there is a determinate interpretation, or sense, or meaning, such that it can seem reasonable to conceive of communication by means of language as a matter of causing that determinate interpretation, sense or meaning to be evoked in the mind of one's interlocutor? The answer to that question ramifies in many directions. Following one branch would involve pondering the attractiveness of a simple mechanistic explanation of the fact that communication is very often successful. Another aspect of the answer is that determinacy of meaning is an important buttress of the incalculably significant idea
108
Nigel Love
that by using language we can establish genuine knowledge of an outer reality that language represents to us. So determinacy of interpretation is ultimately part of an explanation of how it is that we are so successful at controlling and manipulating our environment. To descend to the linguistic particulars, one prerequisite for such a world-view is a practice of systematically describing languages. Setting out to do that depends, first, on the availability of writing, and secondly, on a transference of the context-independence of the written medium itself to what it is being used to describe. The practice of writing lies behind our capacity to think of a language as an object, detachable from communicational contexts, exportable, explicitly teachable to foreigners, etc. Implied here is the idea of a language as a self-contained system of communication, the meanings of the units of which are, more or less, definable in terms of the system. A language is thus seen as a context-free medium or facility, invented by and in some respects under the control of its users, whereby we can, for instance, unambiguously articulate an understanding of the physical world. Writing made possible the expansion of our communicational universe by encouraging the idea that the vagaries of face-to-face communication in particular contexts were a mere circumstantial superimposition on the use of a system of communication whose determinacy was seen as guaranteed by the fact that linguistic types could at last be unambiguously displayed, discussed and defined. It is only with the advent of writing that the sort of systematic discourse about language that we call 'linguistics' becomes possible. Writing simultaneously invents languages as the objects of systematic linguistic description, and determines the shape of such descriptions. The descriptive medium itself articulates, grosso modo, the analysis which it sets forth. For instance, an introduction to sociolinguistics (Downes 1984: 13-15) sets out, by way of giving a preliminary example of phonetic variation, the following pronunciations of the English word butter: (i) [bAts], (ii) [bAd9r], (iii) [bA^a], (iv) [bA 9 gr], (v) [bAda]. (They are identified as (i) British 'received pronunciation', (ii) Canadian, and e.g. New York upper-middle-class, (iii) London working-class, (iv) west-country British, (v) New York working-class.) Now a crucial fact about the abstraction underlying these five variants of butter is that in spoken English it has no name. If the speaker of British RP is asked to name the yellow substance on the plate before him, he will say '[bAta]'. If asked to say what word he has just used, he will say '[bAts]'. The corresponding forms for the working-class Londoner will be '[bA93]' and '[bA9a]'. The only way to identify a word (or any other linguistic unit) in speech is to produce a particular phonetic realisation of it. There is no superordinate pronunciation which is the pronunciation of 'the word itself, as distinct from one of its phonetic variants. What ties the five forms together - indeed, what makes it possible to see them as different versions of a single entity - is the fact that written English contains the invariant butter. A stable and consistent analysis of utterance-tokens in terms of the types that they instantiate would be impossible without the assistance afforded by writing in the form of a phonetically neutral notation with which to identify types. It is the symbiosis of the medium and the object of study which makes the idea of languages as determinate systems of linguistic units seem so inescapable. Anyone who would make the point that there is no such thing as a determinate unit of spoken English represented by the written form cat runs up against the fact that in order to
Integrating Languages
109
make it he seemingly has to invoke the unit whose existence he calls in question. As Trevor Pateman has put it (1987: 3): 'that we make use of iterable forms not defined by temporal co-ordinates seems presupposed by the very critique of that idea'. Of the many good reasons for rethinking linguistics, one of the most pressing is the urge to extricate oneself from the conceptual straitjacket to which Pateman alludes. Given that getting out of it seems to be by no means straightforward, it might be useful to try to think about how we got into it in the first place. Considerations such as those crudely sketched in the foregoing discussion might indicate whether or how far there was ever any possibility of resisting it. There is a sense in which it must be a mere truism to say that every step along the way was a natural consequence of our attempts to understand and expand our linguistic experience. (Nobody forced us to adopt a certain way of thinking about linguistic phenomena.) But the point about truisms is that they are true; and this particular truth is one whose implications linguistic theory has persistently shied away from. Any attempt to come to terms with language is bound to decontextualise it. In the history of language itself, a systematic decontextualisation was first achieved via the invention of writing: that is, the systematic apprehending of oral utterances as abstractions amenable to representation in another medium. The reason the linguistics that naturally arises from this conceptual advance is inadequately equipped to understand contextualised language itself is that it automatically projects as first-order realia the products of the decontextualisation process that made it possible in the first place. What is needed is a decontextualisation which does not retroject itself on to the original context-bound situation as an explanation of the linguistic aspects of that situation. This seems to require a redefinition of the aspects of an interactional situation which ought to count as 'linguistic', thereby divorcing the descriptive metalanguage from the object of inquiry. The two interlocking ideas that determine the conceptualisation of 'languages' assumed in twentieth-century linguistic theory (i.e. the thought-transference theory of communication and the fixed-code theory of languages) have a natural explanation in terms of the history of language-use itself. It follows that a conceivable reconfiguration of linguistics would be one whereby linguistics was in a sense seen as constituting its own subject matter. This would involve relocating the line between the linguistic and the metalinguistic. 'Languages' as structured systems of abstractions designed to allow throughts to be conveyed from one mind to another should cease to be seen as the uncontentiously 'given' objects of linguistics. On the contrary, the processes whereby 'languages' in that sense come to have whatever existence they do have must themselves be the product of reflection on language, that is, of linguistics in a broader sense. In this conception linguistics started the moment human beings started to get to grips with their first-order linguistic experience, the ratiocinations of latter-day language experts being essentially continuous with that process. To have an understanding of that process would be to have the kind of higher-order perspective on our linguistic affairs whose achievement would be a worthwhile goal for an integrational linguistics.
110
Nigel Love
References Bennett, J. (1976) Linguistic Behaviour. Cambridge University Press, Cambridge. Bloomfield, L. (1935) Language. Allen & Unwin, London. Cooper, D.E. (1975) Knowledge of Language. Prism Press, Dorchester. Downes, W. (1984) Language and Society. Fontana, London. Harris, R. (1996) The Language Connection: Philosophy and Linguistics. Thoemmes Press, Bristol. Hutton, C.M. (1990) Abstraction and Instance: The Type-Token Relation in Linguistic Theory. Pergamon Press, Oxford. Katz, JJ. (1981) Language and Other Abstract Objects. Blackwell, Oxford. Lehmann, W.P. & Y. Malkiel (eds.) (1968) Directions for Historical Linguistics. University of Texas Press, Austin. Pateman, T. (1987) Language in Mind and Language in Society: Studies in Linguistic Reproduction. Clarendon Press, Oxford. Radford, A. (1988) Transformational Grammar: A First Course. Cambridge University Press, Cambridge: . Saussure, F. de (1922) [1916] Cours de linguistique generate. Payot, Paris. Smith, N.V. (1973) The Acquisition of Phonology: A Case Study. Cambridge University Press, Cambridge. Weinreich, U., W. Labov & M.I. Herzog (1968) Empirical foundations for a theory of language change. In Lehmann & Malkiel 1968, pp. 95-195.
PART THREE LANGUAGE AND MEANING The papers in this section deal with aspects of integrational semantics and indicate the ways in which an integrationist approach to questions of meaning opens up different avenues of inquiry from those available in orthodox linguistics under a fixed-code theory of languages. Ch. 8 ('Three Models of Signification') describes the integrational view of meaning by contrasting it with two more widely accepted views. The first (and oldest) of these is the surrogational account of meaning, developed over many centuries in the Western intellectual tradition. According to this, a sign means whatever it 'stands for', either in the 'real world' or in 'the mind'. Words are thus surrogates for something else, either material or conceptual. The second view of meaning is the structural view, according to which what, a sign means is determined solely by its relations with other signs within the same system of signs. The integrational view differs both from the surrogational and from the structural view by treating meaning as depending on the particular circumstances in which a sign is produced, and as having no validity outside that communication situation. Ch. 9 ('Meaning and the Principle of Linearity') considers the way in which the notion of the 'compositionality' of meaning in an utterance depends on the assumption that utterances have a linear structure. For otherwise it would be impossible to assign specific elements of meaning to specific sequences of sound within the utterance. Thus the orthodox assumption is that at some level linear segmentational structure reflects or reveals semantic structure. This in turn is related on the one hand to the Saussurean 'principle of arbitrariness' and on the other hand to the doctrine of languages as fixed codes. But all the 'fixed code' offers is a way of classifying linear segments of utterances as tokens of a certain type, to which the code has already assigned a meaning. In effect, the fixed-code model corresponds to a communicational need for repeatability. It fails, however, to address the no less important communicational need for
111
112
PART THREE: LANGUAGE AND MEANING
recognizing the unique meaning of each and every utterance as and when produced. Ch. 10 ('On Inscribed or Literal Meaning') is an integrationist attack on the traditional notion of 'literal' meanings, the latter being the basis of the orthodox linguist's distinction between the literal and the metaphorical. While it may be possible to cite seemingly uncontroversial examples of the distinction, i.e. of a word being used in one case 'literally' and in another case 'metaphorically ', any attempt to define what constitutes a word's (permanent) literal meaning is beset with problems. The beginning of wisdom here is to see that identifying 'literal meaning' is always a controversial abstraction from a semantic continuum. Ch. 11 ('Irony and Theories of Meaning') comments on the inability of orthodox linguistics to cope with the phenomenon of verbal irony. For here words are apparently 'intended' to mean something other than they 'ought' to mean according to the linguistic code to which they are deemed to belong. This is a problem Saussure failed to address. Wittgenstein may have recognized it, for he seems to admit that a false proposition may be meant to be false and known to be false by both speaker and hearer. Yet this in turn presupposes that a proposition already has a sense that is independent of the facts. And it is difficult to see how there can be any such guarantee. The more we reflect on cases of irony, the more we are led to the conclusion that extracting 'the meaning' of what was said can never be a mechanical process of calculation or something that can be ascertained from a dictionary. This is because individuals create meaning in specific contexts. There is no context-neutral calculus to be applied in determining what is ironical and what is not.
8
Three Models of Signification ROY HARRIS
1. Introduction The three models of signification referred to in the title of this paper all belong to the long tradition of philosophical debate that begins in ancient Greece with Plato and Aristotle, is continued in the great universities of medieval Europe and still remains today as the basic Western framework of academic inquiry into the conceptual foundations of human beliefs about knowledge, reason, meaning and truth. These three models may conveniently be called: (i) the surrogational model (ii) the structural model and (iii) the integrational model. All three may be regarded as components of attempts to answer the general question 'How do signs signify?' Although the term sign may be taken as embracing the broadest possible range of significant objects or events, these three models focus upon signs as used in human communication. So-called natural signs, therefore, will not be considered in what follows. The discussion will be concerned mainly with linguistic signs, but has wider implications, as the example in the final section of the paper is intended to show. It may initially seem puzzling that in the Western tradition the study of signs as such was not recognized as an independent branch of intellectual inquiry until the end of the nineteenth century, or at least did not merit a specific designation of its own. As if by way of overcompensation for this long delay, it then received two designations almost simultaneously. It began to be called semiologie in France and semiotic in the U.S.A. (although this latter term had been anticipated two centuries earlier by the great English philosopher John Locke). 1 The baptismal delay might appear all the more surprising in view of the fact that some of the central topics regarded as falling within the purview of the newly baptised semiologists and semioticians had in fact been discussed many times throughout the centuries by their predecessors. 1 Semiologie, the term coined by Saussure, was first used by him in 1894, thus anticipating by three years C.S. Peirce's use of the term semiotic.
113
114
Roy Harris
Two reasons can, however, be adduced to explain the eventual recognition of the study of signs as a worthy enterprise in its own right. One was the appearance on the Western academic scene during the nineteenth century of a new discipline calling itself psychology, which proposed to study by experimental rather than logical inquiry the workings of the human mind. For the psychologist, signs of all kinds and their interpretation by human beings constituted important empirical evidence. The other reason was that at roughly the same time another new discipline emerged in the universities of Europe, which took as its exclusive subject a comprehensive study of one of the most important types of human sign system. This other new discipline called itself linguistics. The point of intersection of these two new disciplines, psychology and linguistics, was the human capacity for creating and using signs. It is no coincidence that the scholar who invented the term semiologie was a linguist; nor that he identified the future science which was to bear that name as constituting a branch of psychology. 2 The advent of two new disciplines, however, does not explain the previous failure to accord an academically institutionalized recognition to the study of signs. The reasons are complex, but one might attempt to summarize them by saying that from the Greeks onwards the domain of the sign had always been treated merely as that of an adjunct to relationships between more fundamental domains: those of consciousness on the one hand and reality on the other. What happened in the late nineteenth and early twentieth centuries was a Copernican revolution in Western thought. The role of the sign was no longer seen as external and ancillary but, on the contrary, as articulating both the domains of consciousness and reality. The new academic disciplines of psychology and linguistics did not bring about this revolution: rather, they were products of it. We are still living in the aftermath of this revolution, and it is in this historical perspective that any present consideration of these three models of signification must be placed.
2. The surrogational model The surrogational model of signification is the oldest of the three. As the term surrogational implies, the defining feature of this model is that signification is explained in terms of the sign being a surrogate or substitute for something else. According to this theory, the sign 'stands for' what it signifies. This abstract relationship can be interpreted in many different ways. The chips used in a gambling casino are substitutes for money: they are purchased on entering and changed back into cash on leaving. In a strictly physical sense they take the place of money in the transactions that occur at the gambling tables. Quite a different kind of substitution is involved if someone draws a diagram to explain how a machine works. Various lines and shapes in the diagram will stand for different parts of the machine, but they are not substitutes in any physical sense. The motorist who replaces his carburettor by a diagram of a carburettor will find that he has difficulty in starting his car. A different kind of substitution again is involved if, rather than 2
F. de Saussure, Cours de linguistique generate, 2 feme 6d., Paris, 1922, p. 33.
Three Models of Signification
115
delivering an oral message by telephone, I write a letter instead. All these cases seem at least to have in common that they involve a mental operation by which one thing is understood to be replaceable by another, and what the replacement signifies is determined by what it replaces. Signification is thus construed as a surrogational relation between the sign and something else. It is frequently described as a relation of representation. The casino chips are said to represent money. The diagram represents the machine. Writing represents speech. Throughout the Western tradition there is a close association between the surrogational model of signification and the doctrine of mimesis or imitation. The logical connexion is the notion that an ideal substitute should be as similar as possible to what it replaces, at least in the relevant respects. This idea finds its most obvious application in accounting for how we interpret the meaning of certain kinds of art. A portrait, for instance, is said to represent the sitter and is commonly conceived to do so by reason of the visual likeness between the painted image and the person. Similarly, we interpret a map or a scale drawing by reference to likenesses of configuration between what appears on the paper and what it is taken to represent. It is on this basis that we can, for instance, work out from a suitably drawn map whether London is further from Bristol than from Brighton. But likeness is only one way in which the surrogational connexion may be established. There is no necessity for a sign to resemble what it signifies. The Union Jack does not look like the British Isles. Nor does the sound of a fire alarm resemble a fire. In such cases as these the connexion between the sign and what it signifies is often said to be arbitrary; by which one must understand not that such-and-such a sign can be chosen or changed at will, but rather that the form of the sign is not systematically determined by the physical characteristics of what it signifies. Undoubtedly the most interesting but also the most controversial application of the surrogational model is its use in the Western tradition to explain the meanings of words. This can be done in two ways. Words may be regarded as surrogates for physical objects, actions, etc. Alternatively, words may be regarded as surrogates for ideas or mental processes. In the former case the meaning is taken to be the corresponding object, action, etc. In the latter case the meaning is taken to be the corresponding concept. These two approaches may be termed reocentric and psychocentric surrogationalism.3 For the reocentric surrogationalist, verbal discourse is essentially a convenient substitute for laborious and complicated physical action. For example, instead of showing someone how to make a fish pie by going into the kitchen and actually making one, language enables us to give a set of verbal instructions instead. For the psychocentric surrogationalist, on the other hand, verbal discourse is a substitute for mental discourse: that is, speech and writing substitute audible and visible signs for a train of thought taking place in the mind. This is substitution in the rather different sense that we thereby replace something essentially private and invisible by something public and observable. In both cases, however, the terminology of 'representation' is once again applicable. We may describe the 3 For these terms, see R. Harris, The Language-Makers, London, 1980, Ch. 2, where various surrogational positions are discussed.
116
Roy Harris
cooking recipe as representing how to make a fish pie; or as representing the cook's ideas on the subject. A similar distinction can be drawn in the case of nonverbal signs. A portrait may be regarded as representing the sitter, or as representing the artist's view of the sitter. A map may be regarded as representing the terrain, or as representing the cartographer's analysis of the terrain. However, the parallel is more difficult to extend to music and games, where the surrogational concept of signification is in any case less plausible. Beethoven's Fifth Symphony might perhaps be said to represent the musical elaboration of certain concepts in Beethoven's mind, but it is difficult to see in what sense it could be said to represent or stand for any physical object. Similarly, the significance of a certain move in a game of chess might perhaps be explained by reference to the player's thought processes or game strategy, but can hardly be seen as meaningful in virtue of standing for something else that is not involved in the game at all. The various possibilities encompassed in the surrogational model are perhaps best summed up by what is sometimes called the 'triangle of signification'. This was proposed in 1923 by C.K. Ogden and LA. Richards in their book The Meaning of Meaning, a landmark in both semiological and semiotic studies. The three corners of this triangle represent the concept (the apex), the object (the right corner) and the sign (the left corner); or the mental world, the real world and the symbolic world respectively. The sides of the triangle represent relations between these three. In Ogden and Richards' version, the base of the triangle, representing the relation between sign and object, is dotted, to indicate that this is only an indirect and inferential relation, whereas the other two sides of the triangle are unbroken lines, to indicate direct causal relations. However, this feature of the triangle is not essential. The important point is that signification is construed in terms of three interconnected dyadic relations; those between sign and concept, sign and object, object and concept. This 'triangle of signification' provokes various awkward questions. Why is the sign, for example, neither part of the mental world nor part of the external physical world? In what sense can it belong to neither? These and related matters cannot be pursued in detail here. However, it is worth observing at this point that it is also possible to construe the triangle in a broader, non-psychologistic manner. This can be done if at the apex of the triangle we replace concept by system. Now a system may or may not be conceptual. It need not be mental at all. But it may nevertheless link sign and object in the way logically required by a surrogational model of signification. For example, a black-and-white photograph may be regarded as standing in a surrogational relation not only to the object photographed (i.e. that of which it is said to 'be a photograph') but also to the image recorded on the film or photographic plate. The camera, in fact, incorporates a system specifically designed to capture such images, and it is this system which mediates between the photographed object and its photograph, just as, in Ogden and Richards' model, the mind mediates between the symbol and what it refers to in the external world. According to this broader interpretation, then, the mind itself can be viewed as one possible realization of a sign-mediating system, rather than as the system.
Three Models of Signification
117
3. The structural model The structural model of signification is a reaction against the long dominance of the surrogational model in the Western tradition. Specifically, it is a product of the Copernican revolution which ushered in semiologie. Whereas surrogationalism seeks to explain signification in terms of relations between signs and non-signs, the structural model explains it solely in term of relations between signs and other signs. Signs relevantly related to one another are seen as forming structural complexes which confer significance simultaneously on all their constituent members. The particular significance thus conferred on any given sign will be a function of the similarities and differences which both connect and contrast it with other signs in the same structural complex. Thus the structuralist approach yields a quite different account of how words get their meanings. According to the surrogationalist model, the meaning of the word red, for example, is determined by its relationship either with a colour in the visible world which is shown in all red objects, or else by its relationship with a certain colour-concept, which we retain in the mind even in the absence of red objects in our visual field. According to the structuralist model, on the other hand, the meaning of the word red is determined by its relationship to other words in the English vocabulary; first and foremost to other colour-words, such as green, blue, yellow etc. If the structuralist is right, the meaning of the word red would necessarily be different if no such words as green, blue, yellow, etc. contrasted with it, even though the visual world and the human eye were to remain unchanged. If the surrogationalist is right, on the other hand, the meaning of the word red remains constant as long as it is associated with the same colour; and provided this colour does not disappear from our visible world and the human eye continues to see it, then there is no reason for the meaning of the word red to change, irrespective of what may happen to other English words. As originally developed by Saussure, the structural model was specifically applied to languages, but it may be extended and adapted to deal with nonlinguistic signs as well. For example, whether an irregular, unbroken, black line on a map represents a river, a road, a railway or a national boundary is not something which is determined independently of the other cartographic conventions employed for the map in question. On the contrary, a cartographer who wishes to distinguish rivers from roads, roads from railways, and railways from national boundaries must necessarily choose different types of line in each case. If he fails to do so, then whatever his intentions may have been, the lines drawn on his map will be ambiguous. In terms of their signification, the conventions a cartographer uses form a structural complex in which the individual elements do not have meaning independently of the whole. The structural model is also more apposite than the surrogational model in dealing with the dimensions of meaning encountered in non-referential communication, as in activities such as playing games. In fact, analogies between languages and games are often invoked by structuralists in support of their claim that meaning is derived from
118
Roy Harris
within the system of signs and not from external relations. (Saussure, for instance, frequently appeals to comparisons between language and chess.4) Where referential communication is concerned, on the other hand, it might at first sight appear that if a sign is clearly iconic - that is to say, if its physical form bears a sufficiently obvious resemblance to what it is intended to signify - then it becomes independent of any structural complex or of other signs associated with it; and consequently that, although the structural model may be applicable to arbitrary signs, it breaks down or becomes superfluous in the case of non-arbitrary signs. Some theorists, therefore, have been tempted to establish a division of labour between the surrogational and the structural models of signification, reserving the former for iconic signs and the latter for arbitrary signs. However, if there is a division of labour to be established it is by no means clear that it can be established quite so simply. To see why this is so it is instructive to consider the case of the conventional chiming clock, which strikes four times to indicate four o'clock, five times to indicate five o'clock, and so on. Thus the principle on which the clock indicates the hour seems to be clearly iconic. It is, of course, a convention that the hours are divided into cycles of twelve in this method of time-keeping; but, given that convention, the identification of each hour in the sequence by striking the corresponding number of chimes seems to be based on a natural principle rather than on any arbitrary assignment of meanings. However, the question is more complicated than it appears. Suppose a mad or eccentric clockmaker began designing clocks which chimed on a different principle. Instead of chiming once for one o'clock, twice for two o'clock, etc., his clocks chimed twice for one o'clock, three times for two o'clock, and so on, completing the cycle by chiming twelve times for one o'clock. Clearly, although unusual and suggestive of a certainty perversity in the designer, this alternative system is no less efficient and simple than that on which the more familiar chiming clock is based. To realize this is to realize that the familiar system does not after all consist of a set of independently motivated chimes, each having a significance which is permanently guaranteed by its relation with just one of the natural numbers. Thus the structuralist will argue that even signs which have an apparently iconic basis can be shown, upon analysis, to derive their significance not from the iconic relationship but from their membership of an organized structure of contrasts within which they operate. According to the structuralist this holds even for explicitly and consciously mimetic activities, as when, for instance, the artist strives to reproduce as faithfully as possible on the canvas exactly what a particular vase of flowers looks like. The extent to which the artist's attempt is a success - or even could be a success - is not what is at issue. Nor is it denied that in some sense the colours on the canvas have to match the colours of the flowers. The point the structuralist makes is that we cannot interpret the significance in visual terms of this or that patch of red pigment in the artist's painting without, either consciously or unconsciously, looking to and comparing it with other patches of pigment on the same canvas. In short we interpret the various colours of the painting not one by one and independently but in terms of the artist's colour scheme as a whole. It is this overall scheme which confers visual significance 4 A detailed examination of the games analogy is to be found in R. Harris, Language, Saussure and Wittgenstein: How to Play Games with Words, London, 1988.
Three Models of Signification
119
on any individual patch of red, and not a universal colour chart which establishes surrogationally the possibilities of substitution between a painter's pigments and the petals of flowers. From this example a general point can be drawn. The structuralist does not deny that human beings often use signs in order to represent aspects of the world around them. Nevertheless, the structuralist excludes that representational endeavour as irrelevant to any general explanatory model of signification. Signification, for the structuralist, comes from the structure of the signs themselves, and not from anything outside.
4. The integrational model The integrational model is radically different from the other two in that it rejects the traditional theory of communication on which both the surrogational and the structural models are based. This traditional theory is telementational; that is to say, it takes the basic function of signs in human communication as being to convey one person's thoughts to another person. Understanding what someone else means is accordingly construed in terms of being able to access or reconstruct the thoughts expressed by means of the signs employed. Thus a sign, according to the telementational theory, has to be a device which makes it possible in principle to effect this transfer successfully. That is why both the surrogationalist and the structuralist look for relations into which the sign enters which can be known to both parties and then treat these relations as explaining the meaning of the sign. In effect, this is already to assume that the process of communication is in principle unproblematic. Integrational semiology makes no such ambitious assumptions about knowing exactly how we communicate with one another. It starts from the more modest thesis that no act of communication is contextless and every act of communication is uniquely contextualized. However, simple and apparently uncontroversial as this thesis may be, it has far-reaching consequences for a theory of signification. These consequences stem from the fact that the integrationist does not treat the act of communication and its context as two independently given items. In other words, the context is not like a theatre backdrop, against which various events may - or may not - take place. Nor is a particular act of communication something which could have taken place in a variety of different contexts, but just happens to have occurred in this context. Furthermore, the integrationist does not assume that the sign has any existence outside the communication situation which gives rise to it. It follows that the integrational sign is not envisaged as a unit which has determinacy either of form or of meaning. Or, more precisely, from an integrational point of view the only a priori determinacy a sign has is contextual determinacy, and contexts are open-ended. In short, the integrationist has no use for the concepts 'sign system', 'code' or 'language(s)' as these are usually interpreted in contemporary semiology. It replaces these concepts with that of a communicational process which both involves and requires a continuous creation of signs within itself. This dynamic generative process, from an integrational point of view, is the sole source of signification. All other models of signification are inadequate, according to the
120
Roy Harris
integrationist, because even if they were open to no other objection they are at best static models. They fail to account for the dynamic dimension of signification. The integrational model, then, although making more modest theoretical assumptions, claims to have an explanatory potential which both the other models lack: it can accommodate the spontaneous ontogenesis of the sign. How does it manage this? In brief, by opening up the possibility of arguing that the sole necessary and sufficient condition for the creation of a sign is the integration of some new complex of features into an ongoing communicational process as a potential resource for further communication. The story of Ali Baba provides an appropriate parable here. According to the story, when Ali Baba saw the thief make a mark on the door of the house and go away to fetch the rest of the gang, he foiled them by making a similar mark on the door of every house in the street. As a result, when the thieves returned to the scene they were unable to identify the house originally marked. Ali Baba's stratagem shows an intuitive grasp of the integrational model of signification. He realized that signs are created by their contextual integration. Consequently a sign is automatically destroyed if its contextual integration is altered. Or, more exactly in this particular instance, the first sign is effaced by reintegrating the mark in question into a new contextual pattern. Semiologically, this is equivalent to removing all traces of the robber's original mark. The robber, on the other hand, makes the typically surrogationalist mistake of supposing that he has established a fixed correlation between the mark and the house he has previously identified. A structuralist might argue that this mistake and the way Ali Baba exploits it provide no less support for the structural model of signification than for the integrational model. But to this the integrationist might reply that this is an illusion. For there is no semiological structure given in advance of the actions performed by the actors in the story. Rather, what the actors do creates and destroys both the sign and its significance. Thus it is quite unlike the structuralist account of how words have meanings. For on that account it would be impossible to explain how an already established sign could be altered or disestablished simply by multiplying its tokens.
5. Signification, epistemology and science Having now compared and contrasted these three models of signification, we may proceed to inquire what their respective epistemological implications are. Perhaps the first point to be made is that none of the three is epistemologically neutral. Each tends to support or be conducive to the adoption of a different epistemological stance, and hence a different view of human rationality. In this connexion it will be instructive to examine briefly the positions advanced by three European philosophers: Hobbes, Derrida and Austin. If Hobbes is singled out for attention here among the many Western philosophers who have taken a surrogationalist approach to language, it is because in Hobbes we find 'the beginning of the conventionalist theory of necessary truth'.5 According to Hobbes there are only two kinds of knowledge: there is knowledge of facts and there 5 W. Kneale and M. Kneale, The Development of Logic, rev. ed., Oxford, 1984, p. 312.
Three Models of Signification
121
is science. The difference between the two is that factual knowledge is given directly by observation and retained by memory, whereas scientific knowledge is conditional, 'as when we know, that, if the figure shown be a circle, then any straight line through the centre shall divide it into two equal parts.'6 Scientific knowledge, for Hobbes, 'is the knowledge required in a philosopher; that is to say, of him that pretends to reasoning'. Hobbes's answer to the question of where scientific knowledge comes from involves an appeal to language and to the signification of words. The first truths were arbitrarily made by those that first of all imposed Names upon Things, or received them from the imposition of others. For it is true (for example) that Man is a Living Creature; but it is for this reason, that it pleased men to impose both names on the same thing . . . Now Primary Propositions are nothing but Definitions, or parts of Definitions, and these onely are principles of Demonstration, being Truths constituted arbitrarily by the Inventors of Speech, and therefore not to be demonstrated.7
According to Hobbes, words are either names or parts of names. Names are names of things, and to name something is to impose a name upon it. The paradigm example of naming for Hobbes is contained in the Biblical account of how Adam gave names to the birds and animals in the Garden of Eden. Affirmations are simply names appropriately joined together, as in the case of Man is a Living Creature, which joins together the two names Man and Living Creature. In every branch of science, therefore, it is important to start with clear definitions. Seeing then that truth consisteth in the right ordering of names in our affirmations, a man that seeketh precise truth had need to remember what every name he uses stands for, and to place it accordingly, or else he will find himself entangled in words as a bird in lime twigs, the more he struggles the more belimed . . . By this it appears how necessary it is for any man that aspires to true knowledge, to examine the definitions of former authors; and either to correct them, where they are negligently set down, or to make them himself.8
From this it is quite clear that for Hobbes signification cannot be divorced ultimately from human intentions. It is in the end up to human beings to make sure that words mean what they intend them to mean, and this can be done by establishing clearly what thing or things in the external world any given name is to be understood as standing for. The quintessential surrogationalism of Hobbes's view of language is summed up with his customary lucidity in the well known epigram: words are wise men's counters, they do but reckon by them; but they are the money of fools .. .9
Had Hobbes been a structuralist he would have been unable to produce an epistemological argument along these lines for at least two reasons. In the first place, as a structuralist he would have held the relationship between names and things to be 6
Thomas Hobbes, Leviathan, London, 1651, Ch. 9. Italics in the original. Hobbes, De Corpore, 1655. Cited in Kneale & Kneale 1984, pp. 311-2. 8 Leviathan, Ch. 4. 9 Leviathan, loc. cit. 7
122
Roy Harris
irrelevant to establishing verbal meanings, and therefore could not have sponsored a scientific programme in which definitions are based on correlations with selected features of the real world. In the second place, for a structuralist there could be no question of defining or redefining the meanings of single words to suit particular purposes, since the meanings of all the words in a language are structurally interdependent. An alteration to any one necessarily has repercussions throughout the vocabulary. For the typical structuralist, linguistic signs are first and foremost units of an extensive and complex system of interrelations, and as such are protected from interference by individual users of the language. This autonomy of the linguistic sign was already emphatically affirmed at the very birth of linguistic structuralism by Saussure: No individual is able, even if he wished, to modify in any way a choice already established in the language. Nor can the linguistic community exercise its authority to change even a single word. The community, as much as the individual, is bound to its language.10
The philosophical implications of this affirmation are carried to their logical conclusion by Derrida, who argues that the meaning of a text must be independent of any intentions on the part of its author or interpretation by its readers. And insofar as speech, like writing, involves the use of words, this will likewise be true of what is said. In short, signification has nothing to do with communication. It is a property of the sign itself, not a function of the way the sign happens to be used by particular individuals. Here Derrida finds himself in alliance with modern generative linguistics, which also stresses the difference between the knowledge of language systems as such (idealized as 'linguistic competence') and their actual use by members of the linguistic community ('linguistic performance'). Consequently, the description of a language as envisaged by generativists requires a context-free identification of both the forms and the meanings of its sentences. Derrida's attack on the notion of signification as communication of intentions deploys the following argumentative strategy. In order to qualify as writing a sign must in principle be able to function in a way that is not spatio-temporally tied to the presence of its producer. That this is not an unreasonable requirement is shown by the fact that in practice writing can be read in the absence of the writer or even after the death of the writer. For this to be possible, Derrida argues, the signs in question must be systematically iterable. (A letter can be copied by someone other than the original writer. A text can be published and republished many times after the author's death.) Similarly, writing is not bound to the presence of or reception by the particular addressee or addressees for whom it may have been originally intended. From this Derrida concludes: My communication must be repeatable - iterable - in the absolute absence of the receiver or of any empirically determinable collectivity of receivers.11
10 11
F. de Saussure, Cours de linguistique generate, 24me eU, Paris, 1922, p. 104. J. Derrida, 'Signature Event Context', Glyph 1, 1977, pp. 179-80.
Three Models of Signification
123
But if writing can thus be divorced absolutely both from sender and receiver, what it means cannot depend in any way on their contribution, or indeed on any particular contextualization. Derrida then generalizes these conclusions from writing as valid for 'all orders of "signs" and for all languages in general'. This generalization, clearly, calls in question any simplistic notion of the transmission of knowledge by means of signs, as it also calls in question the whole Hobbesian concept of 'science'. Derrida's arguments are open to various objections which do not call for discussion here. The point to be stressed, as in the case of Hobbes, is that these arguments would not be available in the first place were it not for the adoption of a certain model of signification, which in Derrida's case is the structural model. The case of Austin is somewhat different from the previous two. Instead of setting out from an integrational model of signification, Austin's arguments move towards an integrational position without ever quite getting there. What is clear enough, however, is that Austin is trying to integrate context systematically into a theory of speech acts. For this he is criticized by Derrida, who appears to hold that a systematic account of contexts is an impossibility. Austin's basic tenet is that the 'force' of an utterance depends on the context, and that contextual conditions determine, in general, both how we can 'do things with words' and what these things are. An immediate casualty of Austin's theory of speech acts is any straightforward dichotomization of declarative pronouncements into 'true' and 'false'. According to Austin, a claim like 'Lord Raglan won the battle of Alma' is neither true nor false; or at least it is pointless to insist that it be so judged. For it is in fact an exaggeration, and as such will be acceptable in some contexts but perhaps not in others. Austin writes: It is essential to realise that 'true' and 'false', like 'free' and 'unfree', do not stand for anything simple at all: but only for a general dimension of being a right and proper thing to say as opposed to a wrong thing, in these circumstances, to this audience, for these purposes and with these intentions.12
Never before in Western philosophy had the concept of truth been so radically relativised or so sweepingly. But exaggerations were for Austin a relatively unimportant class of statements which cannot be accommodated by the true-false dichotomy. More significant were cases like 'I apologize'. This is not true or false either: according to Austin it is a 'performative'. The core of Austin's linguistic theorizing comprises an analysis of different kinds of performative use and an attempt to specify the contextual conditions required for successful performance.13 Where this endeavour falls short of adopting an integrational model is that, although recognizing the importance of contextual factors, Austin nevertheless appears to take for granted that sentences do have invariant meanings which are context-free. Without that assumption, indeed, it would be difficult to make sense of the characterization of 'Lord Raglan won the battle of Alma' as an 'exaggeration'. 12
J.L. Austin, How to Do Things with Words, Oxford, 1962, p. 144. Performance in Austin's sense, it should be noted, has nothing to do with performance in the generativist sense (where it stands opposed to competence). 13
124
Roy Harris
Likewise, all Austin's speech acts rely on the availability of sentences which need no further explanation or justification, irrespective of whether they are being used in particular contexts as warnings, threats, suggestions, commands, etc. Austin, in short, treats context as a set of variable backdrops against which invariant signs are produced with different effects, depending on the backdrop selected. He has no concept of an integration in which sign and context reciprocally and uniquely define each other; no concept, in other words, of an integration that creates at one stroke a sign-in-context. Nevertheless, Austin's speech act theory, compromise though it is, takes us considerably beyond either the surrogational or the structural models of signification. It at least brings us as far as a point where the integrational model can be seen as the next destination for those interested in an epistemology which takes the internal logic of Western theories of the sign one stage further.
6. Conclusion As an examination of these three models demonstrates, theories of signification are highly relevant to - and often intrinsically implicated in - discourse concerning human knowledge, particularly scientific knowledge, and discourse concerning human rationality in general. Implicitly or explicitly, they set limits to human rationality and categorize human activities in various ways. They thus provide a basis for value judgments of many kinds, including moral judgments and aesthetic judgments. Nothing reveals this more clearly than a consideration of aesthetic judgments, as a final example will illustrate. Consider the case of the Parthenon. On one (perhaps fairly uncontroversial) level we can make sense of the Parthenon as a building erected for a particular purpose at a particular period in a certain cultural context, and we can see it as reflecting in various ways the concerns of its builders and users. As a product of its time, therefore, the Parthenon has a historical signification, and whether or not this is clear to us will depend more or less directly on how well informed we are concerning the time and place in question. The building we see today is a 'sign' of past activities and beliefs of a society now extinct. On this level, it is doubtless easier to make sense of the Parthenon than of, say, Stonehenge or Lascaux. But if we wish to proceed to another level of judgment, of the kind commonly called 'aesthetic', it soon begins to matter whether we accept the Parthenon as a 'sign' of another kind; and this involves our acceptance or rejection of certain models of signification. For example, from the surrogational point of view adopted by Hobbes, to ask what the Parthenon means is a question that does not make much sense at all. We are at a loss to correlate this building with any other feature of the physical world in such a way as to be able to say 'The Parthenon stands for this other thing'. From a structural point of view, on the other hand, the question not only makes perfectly good sense but admits of a detailed answer. This answer will be based on the 'language' of Classical architecture and will explain how the various features of the building cohere to make a particular architectural statement. This explanation will
Three Models of Signification
125
involve comparing and contrasting the building both with other buildings that employ the same architectural language and with alternative possibilities that might have been realized in the case of the Parthenon itself. From an integrational point of view the question also makes sense but demands a different kind of answer and a more complex one. Fortunately, the rough outline of that answer was provided half a century ago by the American philosopher John Dewey, and it is worth citing in full. A work of art no matter how old and classic is actually, not just potentially, a work of art only when it lives in some individualized experience. As a piece of parchment, of marble, of canvas, it remains (subject to the ravages of time) self-identical throughout the ages. But as a work of art it is recreated every time it is esthetically experienced. No one doubts this fact in the rendering of a musical score; no one supposes that the lines and dots on paper are more than the recorded means of evoking the work of art. But what is true of it is equally true of the Parthenon as a building. It is absurd to ask what an artist "really" meant by his product; he himself would find different meanings in it at different days and hours and in different stages of his own development. If he could be articulate, he would say "I meant just that, and that means whatever you or any one can honestly, that is in virtue of your own vital experience, get out of it." Any other idea makes the boasted "universality" of the work of art a synonym for monotonous identity. The Parthenon, or whatever, is universal because it can continuously inspire new personal realizations in experience.14
Perhaps some integrationists would think that Dewey places too much emphasis on the viewer as the arbiter of meaning and not enough on the complexity of the factors involved even in the viewer's recognition of the Parthenon as a building. Be that as it may, Dewey's account of the matter squares with the integrationist's insistence on 'individualized experience' as the ultimate locus of significance. By comparing these three interpretations, we see how the model of signification provides not only the basis for a particular value judgment but a basis for categorizing value judgments in general. Nor should this be surprising. For value judgments are not determined independently of the patterns of signification which permeate and organize our understanding of the world in which we live.
14
J. Dewey, Art as Experience, New York, 1934, pp. 108-9.
Meaning and the Principle of Linearity CHRISTOPHER HUTTON
Linguistic theory is dominated by models that treat linguistic structure as a hierarchy. Each level of description is expressed as a linear sequence of elements, the nature of which is determined by the particular grammatical model. In conceptualizing the linearity of the linguistic sign, the linguist draws on an unspoken analogy between the line of units written out on paper and a linear conception of time. The linguist may insist that the unit of grammatical description (the 'system-sentence') is abstract, yet the abstraction could not be grasped without some tacit appeal to a common-sense parallelism between a left-to-right sequence on paper and time conceived of as a linear arrow. What complicates the picture of linear structure is the notion of hierarchy. Some sentences can be formalized as the linear structure Del + N + V + Det + Adj + N, but also as a linear structure NP + VP. They will also be expressed as S or by whatever unit is designed as the highest. Thus at one end of the hierarchy - the dominant node - there is no linear structure, whereas each respective level of structure will be represented as a linearity of a different order. The linearity on a higher level is translatable without residue into that on the lower level and vice versa by means of rewriting rules. At one end of the hierarchy - the 'top' - we have a single unit which defines the upper limit of the system, at the other the 'ultimate constituents', be they conceived of as words, morphemes, phonemes or features. This property of language is generally referred to as 'compositionality' (Crystal 1985: 62). The principle of the linearity of the linguistic sign is, with the principle of arbitrariness, presented as fundamental to the study of language in Saussure's Cours de linguistique generate. As with the principle of arbitrariness (e.g. Coseriu 1967), scholars have pointed to the lineage of the notion of linearity. Aarsleff (1982: 157) cites Condillac as follows: 'If all the ideas that compose a thought are simultaneous in the mind they are successive in discourse: it is languages that provide us with the means of analyzing our thoughts.' It is not clear, however, that Saussure's conception of linearity is precisely the same as that ascribed to Condillac, since linearity can be seen either as concealing a unitary concept (Condillac) or as expressing or reflecting the concatenation of autonomous signs (Saussure). For Saussure, elements in the linear chain are independent from each other to a degree since they enter into their own set of paradigmatic relations.
126
Meaning and the Principle of Linearity
127
Linearity has not been granted the attention afforded to arbitrariness or to the Saussurian dichotomies (Harris 1987: 69), either in histories of linguistics or in discussions of Saussure. Lepschy (1982: 48) writes that the 'cryptic and inconsistent remarks on arbitrariness which appear in the Cours have stimulated far too much comment' and that the second principle, the principle of linearity, is 'more interesting' (Lepschy 1982: 49). The Cours' basic statement of the principle of linearity runs as follows (Saussure 1983: 69-70 [103]): 1 The linguistic signal, being auditory in nature, has a temporal aspect, and hence certain temporal characteristics: (a) it occupies a certain temporal space, and (b) this space is measured in just one dimension: it is a line.
It is not clear how Saussure arrives at the conclusion that the linguistic sign is analogous to a line. He takes it as 'obvious', yet fundamental since 'the whole mechanism of linguistic structure depends on it' (loc. cit.). We should not assimilate linguistic signs to visual signs which 'are observed to coexist in space without confusion'. The speech chain is a 'continuous ribbon of sound, along which the ear picks out no adequate or clearly marked divisions' (Saussure 1983: 102 [145]). It is only through meanings that the formal patternings in the speech chain become available to us. The principle of linearity therefore embodies a principle of classification. Henry (1970: 88) argues that there can be no question of the linearity of the phoneme and that the linear conception of time invoked by the Cours is inadequate. He rejects the simplifying notion of time as an arrow: 'le temps est plutot con$u maintenant comme un contenant universel, un milieu indefini, qu'une ligne est absolument incapable de figurer'. For Henry, there is a Gestalt perception involved ('aperception en bloc') which cannot be reduced to a linear sequence of individual elements. He concludes (Henry 1970: 92) that the notion of linearity is not a fundamental one, arguing that the linear structure of language is a consequence of the contingent physical structure of the organs of speech and as such is not of profound interest in the study of 'le probleme fondamental des rapports entre signifiant et signifie'. Lehmann (1987: 341) argues on similar lines: 'For an accurate understanding of the aim of our inquiry - language - we must incorporate the improved conception of time achieved by Albert Einstein, Werner Heisenberg and their successors.' For Saussure, a linear sequence is a sequence of signs (signifier-signified pairings). For Condillac, the linearity conceals the underlying similarity between the conceptual structure of visual and linguistic signs, since the thought 'has no succession in time'. Saussure would thus regard a sentence-length unit as a sequence of thoughts rather than a single, holistic thought. In this limited and specific sense, the relation between signifier and signified is not arbitrary, since it is created by and creates a conceptual reality that is bound to it. Following this line of argument, Benveniste (1939: 25) argues that the relation between signifier and signified is not arbitrary but necessary: 'il y a entre eux [signifiant et signifie] symbiose si etroite que le concept "boeuf' est comme 1'ame de 1'image acoustique bqf. Whatever the merits of this metaphor, the consequence is clear: on the logic of the Cours, the divisions in 1
Numbers given in square brackets are references to the pagination of the 1922 edition of the Cours.
128
Christopher Mutton
conceptual space are parallel with the divisions in formal space, viz. the famous analogy between the linguistic sign and a sheet of paper (Saussure 1983: 111 [157]): 'Just as it is impossible to take a pair of scissors and cut one side of paper without at the same time cutting the other, so it is impossible in a language to isolate sound from thought, or thought from sound.' Benveniste argues for the virtual fusion of the signifier and signified (Benveniste 1939: 25): 'le concept (signifie) "bceuf' est forcement identique dans ma conscience a 1'ensemble phonique (signifiant) bof. Comment en serait-il autrement? Ensemble les deux ont etc imprime dans mon esprit; ensemble ils s'evoquent en toute circonstance.'2 Bolinger takes a similar position (Bolinger 1949: 55): It has never been contended that complex utterances are arbitrary in the same sense in which arbitrary has been applied to morphemes. When I say The fire consumed the house I 'might as well' say mabu, an arbitrary sign to symbolize the entire occurrence. The fact is, however, that in my language experience parts of the utterance correspond to parts of the event - and the whole utterance is to that extent not arbitrary, for it is articulated in some such way (remote as you please to call it) as the event itself. Here meaning and form affect each other. Now there is no reason why this habit of non-arbitrariness, of point-to-point correspondence, should stop at the level of complex utterances. It continues to the level of morphemes and beyond. And herein, at the floor of language where phonologists and morphemicists have made their stand and where signs have been pictured as unimpeachably arbitrary, we are challenged to find proof that language is still systemic.
Bolinger argues that linguistic - semantic - change can be influenced by meanings attached to sub-morphemic segments by a process of association. He gives the example of English words beginning with sh- that are associated with the concept of 'mess' (Bolinger 1949: 58).3 Bolinger illustrates that units on a lower level than the conventional morpheme can plausibly have meanings attached or associated with them. This has nothing to do with the principle of arbitrariness, unless it is being postulated that the sequence sh- has the intrinsic meaning of 'mess' independently of any particular language. More importantly for the present discussion, he assumes that divisions in linguistic structure correspond to divisions in the structure of the events described. The important intellectual content represented by the notion of linearity is the hypothesis that it offers the structure of thought as expressed or revealed in language. For Bolinger and Benveniste, as for Saussure, linearity reflects semantic structure. A similar notion of linearity is crucial to the so-called Sapir-Whorf hypothesis of linguistic relativity. Put loosely, this asserts that linguistic divisions reflect conceptual divisions. Sapir, in his discussion of the sentence The farmer kills the duckling', begins by asserting (Sapir 1921: 82-3): 2 This is criticized by Nehring (1950: 2) on the ground that such a definition precludes our remembering a form and not a concept, and because it implies that we could not explain how our words come to change their meanings. See also discussion in Lerch (1939), Pichon (1940-41), Sechehaye et al (1940-41) and Gardiner (1941). There is in any case no logical contradiction between the notion that the link between signifier and signified is arbitrary, and its being 'necessary'. 3 Firth (1930, 1957: 44) discussing similar phenomena employs the term 'phonaesthetic function'. Whorf s early work includes the development of a theory of 'oligosynthesis' which argued for the semantic value of sub-morphemic elements in Aztec and Maya (See Carroll's introduction to Whorf 1956, reprinted 1987). For other bibliography, see Bolinger (1949: 59 fn).
Meaning and the Principle of Linearity
129
A rough and ready analysis discloses here the presence of three distinct and fundamental concepts that are brought into connection with each other in a number of ways. [. . .] We can visualize the farmer and the duckling and we have also no difficulty in constructing an image of the killing. In other words, the elements farmer, kill and duckling define concepts of a concrete order.
Goffman (1981: 205) cites research that seems to support this kind of formalconceptual segmentation: 'Apparently almost all pauses occur at word boundaries, suggesting that words are encoded from thought into speech in whole word clumps (Maclay and Osgood 1959)'. But this is in the final analysis circular, since words are generally defined as units that are bounded by potential pauses. It is in effect a play on definitions: the lengthening of the vowel 'within a word' could with equal justification be regarded as a pause. A concept that might usefully be compared to the Saussurian concept of linearity is Firth's notion of collocation. A collocation can be defined as a habitual syntagma, such as 'silly ass' or - in Lear's limericks - 'old man' (Firth 1957: 195). Firth distinguishes between 'general or usual collocations', which are in some sense a property of 'the language', and 'technical or personal collocations': Firth (1957, loc. cit.}: Just as phonetic, phonological, and grammatical forms well established and habitual in any close social group provide a basis for the mutual expectancies of words and sentences at those levels, and also the sharing of these common features, so also the study of the usual collocations of a particular literary form or genre or of a particular author makes possible a clearly defined and precisely stated contribution to what I have termed the spectrum of descriptive linguistics, which handles and states meaning by dispersing it in a range of techniques working at a series of levels.
It is evident that the principle of linearity has little or nothing to do with the fact that word-order can be 'significant', since this is a contingent rather than a necessary feature of language. As Harris remarks (Harris 1987: 74): 'From a Saussurian point of view, languages would still be languages even if free word-order were universal and words themselves were invariably monosyllables with a fixed structure which never contrasted in respect of the internal order of consonants or vowels.' The principle of linearity implies that, to put the matter simply, words in the speech chain cannot be spoken simultaneously. They have to be said one after the other, even if the order in which they appear has no particular bearing on the meaning. You can't say two words at once (Saussure 1983: 121 [170]): Words as used in discourse, strung together one after another, enter into relations based on the linear character of languages. Linearity precludes the possibility of uttering two words simultaneously. They must be arranged consecutively in spoken sequence. Combinations based on sequentiality may be called syntagmas.
Hockett (1987: 4) expresses a similar idea. 'Given two words, U and F, the only possibilities are that U precede F and that F precede U. They cannot be simultaneous; they cannot be lined up in space instead of time.' In raising linearity to a fundamental principle, Saussure appears to have conflated langue with parole (Harris 1987: 73), for temporality and sequentiality belong to the world of actions and intentions, rather than to the impersonal impassivity of langue. One might argue that for Saussure the principle of linearity is a feature of langue
130
Christopher Mutton
reflected in parole. The divisions in the speech chain are divisions represented in langue. It is in this sense that the principle of linearity is fundamental: it is the key to the analysis of speech syntagmatically (and therefore paradigmatically), for it asserts that speech reflects sequentially the formal and conceptual divisions of langue. The exact nature and mechanism of the relation between langue and parole is a separate and ancillary issue. Clearly, speech is a temporal activity. The question of what makes up the sequence of speech is, however, conceived of in terms of meaning. And the question of meaning is determined with reference to the discrete categories of langue which, as Harris points out, find their realization in writing, thereby reflecting the domination of writing ('scriptism') in the conceptual schema of the Cours (Harris 1987: 78). The Cours recognizes that the details of the principle of linearity are problematic (Saussure 1983: 123 [173]): Where syntagmas are concerned, however, one must recognize the fact that there is no clear boundary separating the language, as confirmed by communal usage, from speech, marked by freedom of the individual. In many cases, it is difficult to assign a combination of units to one or the other. Many combinations are the product of both, in proportions which cannot be accurately measured.
The notion of syntagma applies 'not only to words, but to groups of words, and to complex units of every size and kind (compound words, derivative forms, phrases, sentences)' (Saussure 1983: 122 [172]). We should ask though whether there is a radical break at some point between units of a phonetico-phonological kind and units of a morphological-lexical-phrasal kind. The Cours seems to imply that units on the level of morphology are the basic elements of syntagmas (Saussure 1983: 121 [170-1]). But are we to think of the internal structure of morphemes as radically different from that of units higher up the hierarchy? The Cours states (Saussure 1983: 39 [63-4]): The ear is what tells us that a particular sound is a b or a t, etc. If it were possible to film every movement of mouth and larynx involved in articulating a sequence of sounds, it would be impossible to detect subdivisions in that series of articulatory movements. [.. .] The sequence of sounds we hear is not divided into segments of equal duration, but into segments identifiable as auditory units.
It is not ultimately clear, therefore, how we are to make the distinction between a 'lexical' item (i.e. a unit of any length that is stored 'whole' in the lexicon), an item produced compositionally by application of the rules (i.e. an item that is not stored 'whole' in the memory) and a combination or sequence of items that is 'purely' individual. The category 'idiom' is created in relation to the notion that the approximate or average 'length' of stored unit is 'word', since an idiom chunk is wordlike semantically, yet seems combinatorial structurally. The meaning of the whole - the idiomatic meaning - cannot be derived from the meanings of the parts plus the rules of combination. But even if we accept the rules-units distinction, there is no reason why a unit which could be constructed by the rules actually is so constructed. There is no necessity that structures be compositional, or that two tokens of the same type be compositional in the same way. Just because a sentence seems to
Meaning and the Principle of Linearity
131
follow the rules, so that the meaning of the sentence is retrievable from the meaning of its parts plus the rules of combination (though what this involves has never been made precise), this does not imply that it could not be a holistic, lexical item. The notion of linearity is also closely related to the concept of language as a 'fixed code' which is used to convey determinate meanings (Harris 1981: 9-10). The speech chain is held to be in principle uniquely divisible into constituents on various levels of a structural hierarchy (ibid.: 57). The search for the fixed set of sounds issued in the development of phoneme theory. It became, in effect, the search for phonic units which would play for speech the same role as the letters of the alphabet played for writing. Not until the mid 1930s was it pointed out that, given any set of utterances in a language, it will always be possible to derive conflicting analyses of the underlying set of phonemes [Chao 1934]. This conclusion, one might have supposed, would have devastating consequences for the 'fixed code' hypothesis. [. . .] Far from it. Competing phonemic analyses continued to proliferate, on the confident assumption that what now needed to be determined was which one of the analytically possible phoneme systems the speakers of the language were 'really' using.
The notion that a linguistic structure such as a sentence has a single structure independent of speaker and context has been termed 'the uniqueness fallacy' (Hockett 1984).4 Hockett, however, is not arguing for some form of structural subjectivism, since (Hockett 1984: 42) 'certainly there must be large areas of practical agreement among the speakers of a language'. He stresses the role of the hearer in retrieving grammatical structure, though the utterance may include 'clues' as to its structure. 'Hearers have to interpret these clues, and if speakers want to be understood, as they normally do, they must make the clues interpretable' (ibid.: 42-3). Hockett's argument, however, is not as radical as it might seem. What he argues is that utterances may have an indeterminate structure, in that we may not be sure which of a few structural possibilities the utterance instances. This is far from implying the indeterminacy of linguistic structure and is more akin to the notion of 'fuzzy' grammar developed in linguistics by Ross among others.5 As a thought-experiment, we can hypothesize a different form of English ('Condensed English') that would be parallel to, but distinct from, regular English ('Explicit English'). We can for example stipulate that the sentence 'The man saw the dog' is to be represented by 'A', just as Bolinger suggests that the sentence 'The fire consumed the house' could be represented by an arbitrary sign 'mabu' (Bolinger 1949: 55). Thus in all contexts where one might say 'The man saw the dog' one could also say 'A'. 'A' is a translation, a substitute, a sign of the sentence. Now it might be argued that 'A' is simply a long way round to reaching the structure of the sentence. On hearing 'A' we would effect a translation from 'A' to The man saw the dog'. All we have done is slow down the 'normal' process whereby the sequence is uttered by the speaker and revealed to the listener. 'A' is on this analysis simply an unnecessary mediation. But this argument is insufficient. There is a well-known story about prisoners who have been together in captivity so long that they simply call out the numbers of jokes rather than go to the trouble of telling them over and over again. There is no reason in principle why 'A' could not become a 'direct' linguistic sign, 4 5
See also Hockett (1987: 65, 89). See Ross (1973), Lakoff (1973) and discussion in Newmeyer (1986: 123-5).
132
Christopher Mutton
i.e. one where no further translation would be necessary. In effect, the sentential unit is no longer to be perceived as 'relatively arbitrary' (or 'partially motivated'), but as fully arbitrary, the property generally ascribed to individual words or morphemes. We can see that the principle of arbitrariness is not simply a statement about the relation between form and meaning, but about the level on the linguistic hierarchy at which these basic formal and conceptual divisions are made: the monomorphemic word is held to be arbitrary (since it has no internal structure), whereas the compound-word and the sentence are both relatively motivated, i.e. only partially arbitrary.6 It might be objected that a language that developed a set of meta-signs for its sentences (and further meta-meta-signs for sequences of meta-signs) would be impossible or extremely difficult to learn, and that the simplification in length would be compensated for by a loss of economy in representation. We would need a different symbol for The woman saw the man', The dog saw the woman', etc. What is more, we would have only a limited number of 'short' symbols at our disposal. If, for example, numbers were assigned to sentences, then we would ultimately be speaking in very high numbers. Conversations would run something like this: '659999?' '455565!' The argument that such a language would be a strain on the memory does not necessarily invalidate it as a theoretical possibility. We can conceive, for instance, of extensions to the human memory made possible by advances in bio-technology. The Chomskian would argue that in any case the number of potential sentences is infinite, and that therefore no mapping between the sentences of a language and a symbolic system would be possible. However, the alleged unbounded character of natural language, its ability to make infinite use of the finite means (Chomsky 1965), relies on a prior distinction between competence and performance. The system of applying numbers or letters or whatever to sentences or conversations would not be required to put a number to an infinitely long string, since no such utterance can be produced. There is an important difference between 'very long' and infinite (or unbounded). The transition from Explicit English to Condensed English would simply create a language with a much larger lexicon and a different form of sequential structuring. For the argument that Explicit English makes infinite use of finite means would apply just as well (or poorly) to Condensed English. The structure of the sentence or utterance articulated in 'spaces' or pauses between the words or in morphemic divisions does not necessarily give us any special insight into the conceptual structure. Even if one accepts the 'bi-planar' or 'form-content' duality of linguistic structure, one need not be committed to the notion that word boundary equals concept boundary, for a sentence can be reduced to a smaller 'substitute' (non-linear) segment, 'A'. Put another way, in what sense can we demonstrate that the conceptual units identified in conventional semantics are 'atomistic'? How can we make a distinction in principle between different levels of linguistic structure? Have I not heard the sentence The man saw the dog' at least a few dozen times in my life, certainly more than I have heard instances of the word 'properispomenon', of which I also happen to know 'the meaning'? It is therefore plausible to argue that we could have a language which conveyed the same information as our own but which did not have the same 'internal structure'. Now it is evident that there will still be a question of linearity to be considered in this meta-language (which I am assuming would, however, become in use a new 6
See Harris (1987: 132) for a criticism of Saussure's notion of relative arbitrariness.
Meaning and the Principle of Linearity
133
language, just as in mastering a foreign language one operates more effectively without the mediation of one's native language). Thus 'ABX' might mean The man saw the dog. It barked at him loudly. He took no notice of it and continued on his way.' The notion of sequentiality is clearly present. But if we free ourselves from the constraints of memory, there is no reason in principle why there should not be a meta-meta-sign 'P' that would stand for 'ABX'. (Clearly since there is a limited number of letters, we would have to use numbers or invent a new alphabetic system.) None the less, in this hypothetical language, we would still presumably have sequences of characters, the 'size' of the units being determined by the 'forward reach' of our mental capacities. We might conceive of a situation where a language evolves by becoming increasingly more 'compressed'; interaction would take place at a much greater speed, relative to previous states of the language, though this would not necessarily be perceptible to the speaker. To continue the logic of this argument, a whole conversation that in the former 'conventional' speech would take hours could be finished in a second, if both participants agreed on the 'number' of the discourse. To say that it is implausible for such a situation to arise, we would have to show what difference there is in principle between, say, the ability of a speaker to 'put a certain form to a certain thought', so that the utterance 'Please open the window' expresses the wish that the window be opened, and a single unit utterance equivalent to War and Peace in all its details and nuances, which will similarly reflect the aspirations and conceptual schemas of the speaker. The larger the conversational unit or discourse segment to be 'translated', the greater the number of items that would have to be known. To include utterances the size of War and Peace it would have to be staggeringly large. The purpose of this exposition is to point to the fact that we have no means of being precise about the internal complexity of our discourse, just as we have no means in principle for stating the point at which such a 'new' language becomes implausible, provided we allow ourselves unlimited memory. Nor does the dialogic structure of conversation necessarily affect the model by limiting the upper plausible size of the unit. Since conversation would proceed at a (in relation to the past) much more rapid pace, it would develop its own 'natural' rhythms of turntaking and we could imagine conversation proceeding at (phenomenologically) the same pace as in the former 'Explicit English'. To the outside, neutral observer who remained the same over time, there would be a clear difference in the nature of the two languages (Explicit and Condensed English), but to the participants there would not be. Similarly, the fact that we can now cross the world in a matter of hours does not necessarily imply that we feel or experience our lives to be moving hundreds of times faster than the lives of those who travelled by horse. Only by constructing ourselves a neutral or quasi-neutral vantage point, or by experiencing the fact of change (living through the invention of the telephone and speaking directly to contacts with whom one had previously communicated with only by letter) can we experience an acceleration, a feeling which will diminish with the passage of time and will be absent in the next generation. But this objection expresses precisely the point of this discussion: we have no conception whatsoever of the level of complexity of our discourse. Or rather, the conception that we have is dictated by the division of the speech chain into the units that correspond to orthographic words, with some allowance for morphemic structure and for 'idiom chunks'. Let us say that I wish to
134
Christopher Mutton
translate a single lexical item from one language to another by means of a simple, 'lexical look-up' program that I have devised. The program will give 'chat' if asked to translate 'cat' into French, by a process of lexical mapping. But suppose I wish to translate War and Peace from Russian into English. One possibility would be that in the memory of the computer (in the 'lexicon') would be the entire text of an English translation of War and Peace, this being accessible under a single 'lexical' entry. Thus the program could give an instant 'translation' of the text, by the same process of lexical retrieval that enabled the machine to give 'chat' for 'cat'. It will be objected that the machine has not really translated War and Peace, since obviously the translation done previously has simply been entered into the memory of the computer. But of course the same could be said of 'cat' and 'chat'. The basic structure of the operation whereby the computer completed the translation task would be identical. Either both operations are translations or neither is. The point of the distinction between Condensed English and Explicit English is to dramatize the fact that we cannot assume we have any abstract grasp of the complexity of the utterances that we employ. We are still in the grip of the powerful lexicographic metaphor for linguistic memory; we think of speaking as a form of reaching into a static reservoir of perpetually self-renewing items which remain unchanged by our experience and personal history, and only undergo change in harmony with wider macro-social processes. Let us imagine that each time we hear an utterance or utter an utterance it is accorded a separate entry in the mental lexicon: thus if I hear The man saw the dog', I record an entry which specifies the context, the spatio-temporally unique character of the utterance. That is to say, I remember the occasion, the speaker, the addressee, etc. After a number of years, I would have perhaps a number of entries of the 'form' 'The man saw the dog'. Let us suppose that I hear a 'new' utterance of this 'form'. How will I understand it? Presumably, on the conventional model of linguistic memory, by some process of look-up and rule analysis, based on my linguistic competence. But let us imagine that I in fact possessed no such linguistic competence, that it had been removed surgically or been lost through some trauma. Could I not still understand the utterance on the basis of my recollection of other, 'similar' utterances? It will be objected that I will have no means of linking the new utterance to my memory, since this utterance will be different in key respects to the other utterances. The speaker, location, etc. will be different; the 'form' of the utterance will be different (intonation, stress, accent, pausing, facial gestures, hand gestures, affect, reference, place in a conversational history etc.). How will I be able to relate this new utterance to the old ones? Two points should be made here: firstly, the theorist who argues for the existence of an autonomous linguistic competence independent of linguistic memory is in no better position to explain how I understand a given utterance, since in order to map the utterance onto an underlying system-sentence, the utterance must first be perceived as a token of a particular type. But how can this be done before the meaning of the utterance has been provided by linguistic competence? The meaning is the key to the structure, as Saussure made clear. There is no way to pass from the 'etic' to the 'emic'. Secondly, we do not have to assume a look-up model of comprehension at all, for what underlies it is a model of language based on written forms, or more precisely, printed forms from the same font or typeface, where the system provides that the two printings of the 'same' letter in the font will be the 'same' shape from the
Meaning and the Principle of Linearity
135
point of view of the system (i.e not if examined under a magnifying glass). Printing provides what Ong (1982: 127) characterizes as an 'exactly repeatable visual statement'. The 'look-up' model of comprehension is ultimately paradoxical: I have to know the form, before I can find the meaning, but the meaning is the key to the form. Yet on the second model, i.e. where no autonomous linguistic competence is assumed, how would I know which of the recorded utterances to relate the heard utterance to, even assuming I could 'crack the code' and map it onto the (perhaps very large numbers of) tokens of the same type stored in memory? How would I decide whether to look-up sentence-tokens or word-tokens, and would the wordtokens be cross-referenced to the sentence-tokens? Would the particular utterance of the word 'dog' that I heard on occasion x be cross-referenced to the recorded token of the sentence 'The man saw the dog' and would that sentence in turn be crossreferenced with the larger conversation that would also be stored under a separate entry? In fact we would need an unbounded number of entries, given that there is no logical upper limit to the number of entries that could be made corresponding to the divisions into levels of a conversation. There is no limit to the number of ways we can divide up a conversation and make 'lexical' entries corresponding to these divisions. Both the autonomous model of competence and the 'utterance-storage' model of linguistic knowledge fail to give an account of the way in which we understand utterances. The 'utterance-storage' model, however, is not an implausible model of how linguistic memory works, i.e. it implies that linguistic memory works the same way as any other form of memory, in remembering actions and events as individual actions and events, not as members of an abstract set. This is not to say that memory does not involve a process of generalization: but generalization involves precisely the opposite qualities to that of an abstract set where all the members are identical. The notion of a Condensed English is intended to provide a framework for thinking about the nature of linguistic structure. It poses the question of the transparency of discourse. In general, it is assumed that linguistic utterances are divided by speakers and hearers into their constituent parts for the purposes of communication. But this is to bring etymology into synchrony. Who is to say where the synchronic divisions are, and on what ground are we to segment the speech chain? For what purposes and in what contexts is this done? Linguistic theory at present offers, to paraphrase Wittgenstein, 'analysis idling'. It has no pragmatic teeth. The notion of Condensed English was introduced to illustrate how we might conceive of a language that was structurally distinct from Explicit English yet semantically and pragmatically equivalent. However, further questions arise in relation to this hypothetical language. The question of semantic and pragmatic equivalence is merely begged: how can we be sure that such a hypothetical language would in reality be 'semantically and pragmatically equivalent'? Have we not in fact created a different language, as different from Explicit English as French or Hopi? For it is evident that we define differences between linguistic systems on the basis of an analysis we make of the sequential structure of the signifier. But this analysis is made on the basis of the assessment of meaningful divisions in the speech chain (or rather, in the hypothetical speech chain). There is a chicken-and-egg problem here, for writing gives the analysis to us. If the hypothetical Condensed English became a 'reality', would we not similarly construct a semantics on the basis of the divisions given by the notation-
136
Christopher Mutton
system? Thus just as the meaning of The man saw the dog' is 'The man saw the dog', so the meaning of 'A' would be 'A'. The Sapir-Whorf hypothesis amounts to the constatation that linguists will analyse the semantic structure of different languages as being different, on the basis of their identification of different structural features. This itself achieved on the basis of the reduction of the language under study to writing (phonemic analysis, morphophonemic analysis or whatever). Condensed English would be as foreign to Explicit English as Hopi. Universalist linguists then take the further step of attempting to reduce the different analyses to a further, metasystem (of universal features) which will be expressed in the writing-system of whatever language the linguist happens to be working in. Condensed English is as like or unlike Explicit English as Hopi or French depending on your point of view: if you are a conceptual relativist, you will point to the non-isomorphism of the structures as implying semantic non-isomorphism. If you are a universalist, you will reduce the semantic structure of Condensed English to Explicit English by effecting a translation: 'A' = The man saw the dog'. The relativist will point to the structural difference 'we dissect nature along lines laid down by our native languages' (Whorf 1987: 213), the universalist to the fact of equivalence. The controversies in generative grammar over the role of semantics were in effect disputes about the validity of certain paraphrases, e.g. the paraphrase of 'kill' as 'cause to die'.7 What Tarski showed was that the neatest way out of the knot of semantics was to construct an isomorphism between object-language and metalanguage: truth thus becomes a universal property (a mapping) of particular (language-specific) relations defined internally to the system: 'snow is white' is true if and only if snow is white (Tarski 1944). The semantic theorist has three options available: to define words in terms of other words, in terms of things or in terms of ideas (the last two are labelled respectively 'reocentric and psychocentric surrogationalism' in Harris 1980: 33-78). But ultimately all of these approaches rely on the lexical segment as identified by the orthography. We have offered to us the pre-packaged divisions given by the orthography and enshrined in dictionaries, metalinguistic practices of various kinds, linguistic theory, 'common-sense' notions that linguistic units are recurrent and so on. The problem for the linguist in this is not that linguistic structure is inscrutable, but that the work of linguistic analysis seems to be already complete. The orthographic divisions give us the conceptual divisions, and vice versa. It would be foolish to deny the 'validity' of the segmentation of speech offered by orthography and dictionaries. For writing and dictionaries are fully integrated into the linguistic practices of literate societies and the metalinguistics of even their non-literate members. And all languages can potentially be written down. The notion of a writing system is based on the same reification of language as the notion that one can repeat what someone else said. Now clearly there will be differences between literate and non-literate societies with regard to the keeping of communal records, the construction of legends and history etc. and the preservation of cultural materials. But the wish to record the past, to create overarching macrocontexts where meanings are stable, is common to both. Just as we can show that lineages recited by specially designated persons in oral societies change and are modified in order to fit the changing 'political' circumstances (Ong 1982: 487 On the 'kill/cause to die' controversy, see for example Wierzbicka (1975). For a partisan but clear history of these and other disputes see Newmeyer (1986).
Meaning and the Principle of Linearity
137
49), so we can observe how history is continually rewritten in Western societies, even given the existence of vast amounts of written materials documenting the past. Ong (1982: 22) discusses Parry's (Parry 1971) work on the Homeric texts. Parry argued that the Iliad and Odyssey were in essence a concatenation of formulae, i.e. that the basic constituent units were not of word-length but phrase or sentence-length. This theory met strong opposition, at least in part because it seemed to be an attack on Homer's 'creativity'. What Ong's discussion illustrates is that the question of the sequential division of texts is not independent of cultural-specific questions about methods of composition, notions of creativity, modes of transmission and recording and so on. One approach to sequential structure and language that does not rely exclusively on the reification of lexical segments is Conversational Analysis (CA). This approach was developed out of interactionalist sociology, in particular ethnomethodology (Garfmkel 1967; Heritage 1984) and undertakes the detailed analysis of 'mundane' or everyday conversation. With its emphasis on 'sequencing' (Schegloff 1968) and 'turn-taking' (Sacks et al 1974) this school clearly offers a different perspective on the conventional linguist's notion of linearity, though whether the two notions are in conflict is difficult to assess. Meaning is not equated with the application of lexical definitional meaning to an objectified context, i.e. with instances of abstraction. Conversational analysts are not particularly interested in conventional linguistic parts of speech analysis, though they use transcriptions based on a modification of conventional English orthography. 8 They have available to them the divisions offered by the orthography but, unlike conventional semantics, they do not seek to refine this segmentation in search for the essential lexical-orthographic invariants underlying speech. Garfinkel's original critique of Parsonian sociology involved what Heritage (1984: 33) terms 'the dilemma of reflexivity'. The analyst had a choice between adopting 'the actor's view and thereby allowing the individual as a source of change in the system with the risk of indeterminism', or viewing interaction as a system, 'a table of organization that operates as a set of impersonal forces that shove the individual around here and there, while taking it as a matter of factual interest that he is correctly aware or not of what is happening to him' (Garfmkel 1952: 145, cited in Heritage 1984: 33). Shared knowledge is accessible, transparent; actions are willed and publicly meaningful. Their meaningfulness is 'locally' or indexically 'displayed'; one utterance is an interpretation of a previous one, the subsequent utterance is a confirmation or a refutation of the understanding shown by the reply. Heritage (1984: 257) writes: The second speaker can determine the adequacy of the analysis displayed in his or her turn by reference to the next action of the first speaker. This observation can be developed by noticing, firstly, that one 'third turn' option which is open to the first speaker is the explicit correction or repair of any misunderstanding which was displayed in the second speaker's turn.
Meanings are available to interactants who are not viewed as behavioristic automata or 'judgemental dopes' (Heritage 1984: 110-5). Their judgements 8 A summary of these conventions can be found in Atkinson and Heritage (1984: ix-xvi). The conventions are discussed in Jefferson (1984).
138
Christopher Mutton
reflexively constitute social action. While not hypothesizing the existence of a fixed linguistic code, conversational analysts have turned increasingly to look for rules and procedures underlying conversational interaction and the maintenance of social order.9 They have developed their own metalinguistic terminology which has become a tool for the division and classification of conversation: 'adjacency pair', 'insertion sequence type' (Schegloff 1972), 'closing', 'delay', 'turn'. One could see these terms, therefore, as the logical complement to the lexical-alphabetic model of linear classification: the development of interactional analogues to the traditional part of speech, i.e. 'parts of conversation'. It is evident that the ethnomethodological CA theorist would not subscribe to the notion of meaning that related linear divisions in the signifier with conceptual or meaning divisions. For CA, language is essentially a shorthand. There is a disparity between saying and meaning: 10 'An utterance is thus the starting point for a complicated process of interpretative inference rather than something which can be treated as self-subsistently intelligible' (Heritage 1984: 140). Garfinkel and Sacks (1970: 342) discuss the gap between saying and meaning in the following terms: The interests of ethnomethodological research are directed to provide, through detailed analyses, that accountable phenomena are through and through practical accomplishments. We shall speak of 'the work' of that accomplishment in order to gain the emphasis for it of an ongoing course of action. The work is done as assemblages of practices whereby speakers in the situated particulars of speech mean something different from what they can say in just so many words, that is, as 'glossing practices'.
Garfinkel and Sacks give further explanation as follows (Garfinkel and Sacks 1970: 344): The idea of 'meaning differently than he can say in so many words' requires comment. It is not so much 'differently from what he says' as that whatever he says provides the very materials to be used in making out what he says. However extensive or explicit what a speaker says may be, it does not by its extensiveness or explicitness pose a task of deciding the correspondence between what he says and what he means that is resolved by citing his talk verbatim.
Goffman (1981: 8-9) presents this characteristic of 'naturally occurring talk' as a form of ellipsis. Thus the exchange AI: 'are you coming?', 62: 'I gotta work' can be viewed as a contraction of AI: 'Are you coming?', BI: 'No.', A2: 'Why aren't you?', 62: 'I gotta work.' The ethnomethodologist would stress that ellipsis is intrinsic to all forms of talk, that explanations or instructions or interpretations can always have an 'and so on'. For example, Garfinkel's 'conversation clarification experiment' required the student investigator to insist that their interlocutor 'clarify the sense of his commonplace remarks' (Heritage 1984: 80; Garfinkel 1963, 1967). The experiment showed that further clarification can always be asked for, but that such clarification is deemed 'sanctionable' if the explanation requested exceeds what is necessary 'for all practical purposes'. 9
This notion of conversation is discussed critically in Taylor and Cameron (1987: 99-124). See Taylor and Cameron (1987: 83) for a discussion of Grice and the notion that speakers mean 'more than they actually say'. Morris (1981) is an account of linguistic practice in Puerto Rico based on the notion that there is a systematic disjunction between what people say and what they actually 'mean'. 10
Meaning and the Principle of Linearity
139
Goffman therefore seems to suggest that conversation can be elliptic, but that there is a level on which a satisfactory form of explicitness can be formulated. For the ethnomethodologist, this explicitness is locally produced, and requests for an interlocutor to clarify a statement beyond what seems reasonable result in moral sanctions and are seen as aggressive or deviant behavior. In two-party conversation 'much that is being talked about is not mentioned, although each expects that the adequate sense of the matter being talked about is settled' (Garfinkel 1963: 221, cited in Heritage 1984: 81). When we consider the issue of segmentation, we therefore have a dilemma. It would seem that the ethnomethodological view of conversation speaks precisely to the notion that the words themselves do not exhaust the meaning of the utterance, that there is no direct correlation between saying and meaning, between form and content. But in a sense this breach between saying and meaning is a product of the investigator's action or observation. In the case of the 'conversation clarification experiment', it is clear that speakers did feel that they had said what they meant in so many words. They find requests for elucidation puzzling and ultimately morally sanctionable. Interactants may accuse each other of reading into or taking too 'literally' remarks or observations. So we require after all a sense in which words mean just what they say, because interactants themselves require it. Put another way, how can the analyst claim that there is a breach of discourse between saying and meaning? Only by asserting that words have literal meanings and that in speech the 'real' meaning is not captured by the contextual or situational meaning of the actual words used. The notion that speakers cannot say what they mean in so many words implies that literal or segmental meaning is an inadequate indicator of contextual or interactional meaning, but it affirms that there is a sense to the notion of literal meaning, if only by contrast. In order to repeat something or to carry it over from one context to another, we need to analyse it. Firstly, we remove it as an object from the surrounding (preceding, subsequent and simultaneous) action; secondly, we give it our voice, our accent, we may omit certain pausings, paraphrase to some extent, etc. If we say that someone used a certain word, we have segmented their discourse for a certain analytic purpose. There is therefore a (constructed or pseudo-) iconic relation holding between, for example, an utterance and the citation of an utterance. We are invoking the equation 'cat' = 'cat' by defining an utterance as an instance of cat. We create an ideal iconic triangle from type to token to repeated token in order to give coherence to the world. The perspective in time from one instance to another, plus the conflating (ideal orthographic) identity of 'cat' = 'cat' combine to give the (impression of) meaning. The meaning of 'cat' is 'cat'. But this ideal meaning is both inside and outside the interactional discourse: inside since it is 'constructed' and applied within, outside since it is constructed to lie outside so as to be applied within. The importance of the topic of linearity lies ultimately in the challenge it offers to our understanding of the notion of lexical item, and of the 'pre-packaged' units of speech. It emerges that from a certain perspective there are no means available to the linguist to draw a distinction between phoneme and word, or even between phoneme and the entire text of War and Peace. Both can be seen as lexical items. What we can learn from this is that the concept that we have of lexical item, for all its intuitive power, is largely empty of analytic content. Ethnomethodology, in positioning itself in contrastive opposition to this notion (of the lexical segment with its meaning) is
140
Christopher Mutton
defining itself against that emptiness. The complexity of language is not one of formulations and rules; it derives from the twin demands on language that it remain the same over time (that it present an identity across contexts, i.e. that it abolish time and context in some sense) and that it express the individuality, the uniqueness of each utterance (i.e. that it express time, change and context). We need no further refinements on that alphabetic principle, 11 but alternatives to it. We need to think our way past the analysis offered by the alphabet and the limited metaphors of linear space and time that it offers us and in this way we may come to understand better the complex of notions and ideals that it represents. As Harris (1986: 157) states: 'From [our descendants'] point of view [.. .] the "origin of writing", may turn out to be a point in the history of human sign-systems which we ourselves, in the late twentieth century, have not yet reached.' The challenge for linguistics is one of conceptualizing correctly the twin faces of language: its function in the maintenance of social and semiotic coherence (the maintenance of identity over time), and its function in expressing the individual and unique meaning of each utterance. This duality is not adequately expressed in the abstraction-instance dichotomy postulated by the linguist in the segmentation of speech, nor in the distinction between micro- and macro-social that underlies the ethnomethodologist's notion of 'local' meanings in the 'here and now'. We need to examine critically the physical and temporal metaphors that are derived from and feed into the notion of linearity and the alphabetical-lexical model of language.
References Aarsleff, H. (1982) From Locke to Saussure: Essays on the Study of Language and Intellectual History. University of Minnesota Press, Minneapolis. Atkinson, J.M. and J. Heritage (1984) Structures of Social Action: Studies in Conversational Analysis. Cambridge University Press, Cambridge. Benveniste, E. (1939) Nature du signe linguistique. Acta Linguistica 1, 23-29. Bolinger, D. (1949) The sign is not arbitrary. Boletin del Instituto Caro y Cuervo 5, 52-62. Chao, Y.-R. (1934) The non-uniqueness of phonemic solutions of phonetic systems. Bulletin of the Institute of History and Philology, Academia Sinica 4, 363-397. Chomsky, A.N. (1965) Aspects of the Theory of Syntax. MIT Press, Cambridge, Massachusetts. Coseriu, E. (1967) L'arbitraire du signe: zur Spatgeschichte eines aristotelischen Begriffes. Archivfurdas Studium derneueren Sprachen und Literaturen 204, 81-112. Crystal, D. (1985) A Dictionary of Linguistics and Phonetics. Blackwell, Oxford. Derrida, J. (1978) Writing and Difference. Translated and annotated by Alan Bass. University of Chicago Press, Chicago. Firth, J. (1930) Speech. Benn, London, (1957) Papers in Linguistics 1934-1951. Oxford University Press, London. Gardiner, A. (1944) De Saussure's analysis of the linguistic sign. Acta Linguistica 4, 107-10. Garfinkel, H. (1952) The perception of the other: a study in social order. Unpublished Ph.D. dissertation, Harvard University. (1963) A conception of, and experiments with, 'trust' as a condition of stable concerted actions. In O.J. Harvey (ed.) Motivation and Social Interaction, pp. 187-238. Ronald Press, New York. 11 On the alphabet, see Derrida (1978), Ong (1982), Harris (1980 1986), Havelock (1976) and Logan (1986).
Meaning and the Principle of Linearity
141
. (1967) Studies in Ethnomethodology. Prentice-Hall, Englewood Cliffs, New Jersey. . and H. Sacks (1970) On formal structures of practical actions. In J.C. McKinney and E.A. Tiryakian (eds.) Theoretical Sociology, pp.338-66. Appleton-Century-Crofts, New York. Goffman, E. (1981) Forms of Talk. University of Pennsylvania Press, Philadelphia. Harris, R. (1980) The Language-Makers. Duckworth, London. (1981) The Language Myth. Duckworth, London. (1986) The Origin of Writing. Duckworth, London. (1987) Reading Saussure: A Critical Commentary on the Cours de linguistique generate. Duckworth, London. Havelock, E. (1976) Origins of Western Literacy. Ontario Institute for Studies in Education, Ontario. Henry, A. (1970) La linearite du signifiant. In Linguistique contemporaine: hommage a Eric Buyssens, pp. 87-92. Editions de L'Institut de Sociologie, Brussels. Heritage, J. (1984) Garfinkel and Ethnomethodology. Polity Press, Cambridge. Hockett. C. (1984) The uniqueness fallacy. In J. Copeland (ed.) New Directions in Linguistics and Semiotics, pp. 35-46. Current Issues in Linguistic Theory 32. Amsterdam Studies in the Theory and History of Linguistic Science. John Benjamins, Amsterdam. (1987) Refurbishing our Foundations: Elementary Linguistics from an Advanced Point of View. Current Issues in Linguistic Theory 56. Amsterdam Studies in the Theory and History of Linguistic Science. John Benjamins. Amsterdam. Jefferson, G. (1984) Caricature versus detail: on capturing the particulars of pronunciation in transcripts of conversational data. Tilburg Papers on Language and Literature 31. University of Tilburg, The Netherlands. Lakoff, G. (1973) Fuzzy grammar and the performance-competence terminology game. Papers from the Ninth Regional Meeting of the Chicago Linguistic Society, pp. 271-91. Department of Linguistics, University of Chicago, Chicago. Lehmann, W. (1987) Time. In A.G. Ramat, O. Carruba and G. Bernini (eds.) Papers from the 7th International Conference on Historical Linguistics, pp. 339-48. Current Issues in Linguistic Theory 48. Amsterdam Studies in the Theory and History of Linguistic Science. John Benjamins, Amsterdam. Lepschy. G. (1982) A Survey of Structural Linguistics, 2nd ed. Andre Deutsch, London. Lerch, E. (1939) Von Wesen des sprachlichen Zeichens: Zeichen oder Symbol? Acta Linguistica 1, 145-61. Logan, R. (1986) The Alphabet Effect. St Martin's Press, New York. Maclay, H. and C. Osgood (1959) Hesitation phenomena in spontaneous English speech. Word 15, 19-44. Morris, M. (1981) Saying and Meaning in Puerto Rico: Some Problems in the Ethnography of Discourse. Pergamon Press, Oxford. Nehring, A. (1950) The problem of the linguistic sign. Acta Linguistica 6, 1-16. Newmeyer. F. (1986) Linguistic Theory in America, 2nd ed. Academic Press, Orlando, Florida. Ong, W. (1982) Orality and Literacy: the Technologizing of the Word. Methuen, London. Parry, M. (1971) The Making of Homeric Verse: The Collected Papers ofMilman Parry, ed. by Adam Parry. Clarendon Press, Oxford. Pichon. E. (1940-1) Sur le signe linguistique. Acta Linguistica 2, 51-2. Ross, J. (1973) Nouniness. In O. Fujimura (ed.) Three Dimensions of Language Theory, pp. 137-238. TEC Corporation, Tokyo. Sacks, H., E. Schegloff and G. Jefferson (1974) A simplest systematics for the organization of turn-taking for conversation, Language 50, 696-735. Sapir, E. (1921) Language: An Introduction to the Study of Speech. Harcourt, Brace and World, New York. Saussure, F. de (1983) Course in General Linguistics, translated and annotated by Roy Harris. Duckworth, London. Schegloff, E. (1968) Sequencing in conversational openings. American Anthropologist 70, 107595.
142
Christopher Mutton
(1972) Notes on a conversational practice: formulating place. In D. Sudnow (ed.) Studies in Social Interaction, pp. 75-119. Free Press, New York. Sechehaye, A., C. Bally and H. Frei (1940-1) Pour 1'arbitraire du signe. Acta Linguistica 2, 1659. Tarski, A. (1944) The semantic conception of truth. Philosophy and Phenomenological Research 4, 341-75. Taylor, T. and D. Cameron (1987) Analyzing Conversation: Rules and Units in the Structure of Talk. Pergamon Press, Oxford. Whorf, B. (1987) Language, Thought and Reality, ed. by J. Carroll. MIT Press, Cambridge, Massachusetts. Wierzbicka, A. (1975) Why "kill" does not mean "cause to die": the semantics of action sentences. Foundations of Language 13, 491-528.
10 On Inscribed or Literal Meaning MICHAEL TOOLAN
Literal, conventional, and utterance meaning An informal characterization of literal meaning might be: The conventional meanings of words of a language, and the meanings of sentences in that language, where any sentence-meaning is derived from a complex synthesis of the meanings of its composite words.
Such a characterization seeks to distinguish literal meaning from conventional sentence-meaning. The distinction is clear in the contrast between two interpretations of the sentence: John got the sack.
The conventional intepretation, sensitive to the usual idiomatic use of get the sack, might offer the paraphrase 'John was fired'. But the literal meaning purports to express more 'basic' or foundational meanings, those residing first in a sentence's component words, and so would produce the paraphrase 'John fetched/received the (nonrigid) container'. One assumption is that idiomatic meanings are some kind of overlay of the literal meaning, such that a conventional sense, attached to a fixed sequence of words taken as an unanalyzed unit, supplants the literal interpretation of the component words. Idiomatic conventional meanings thus occupy an in-between area, being neither literal nor fully metaphorical. The definition of literal word- and sentence-meanings above amounts to an appeal, in determining meanings, to the information enshrined in a reliable grammar and dictionary of a language. But such an appeal may often yield several possible literal meanings for a sentence. Consider: Can you pass the salt?
The literal meaning of the sentence may vary considerably, given the multiple conventional meanings of can and pass, as should be evident if we imagine the sentence used in a chemistry laboratory, or in a medical interview with a patient, besides the dinner-table setting. Thus, polysemy too (in addition to idiomaticity)
143
144
Michael Toolan
overlays literal meaning. It is important to note however that even in the lastmentioned context of use, at the dinner-table, the literal paraphrase will be 'Are you able (is it possible for you) to convey the salt?', i.e., something rather different from its probable conveyed- or speaker-meaning, which would not be an enquiry about ability or possibility at all, but rather a polite request. Thus literal meanings and conveyed- (or speaker- or utterance-) meanings are in principle quite different from each other. Indeed much contemporary linguistic pragmatics purports to explain how language users seemingly proceed inferentially from decontextualized literal meanings of sentences to their conveyed meanings in particular contexts of utterance. Even thus situated, however, the characterization of literal meaning given above retains many problems. The definition refers to 'meaning-synthesis', a variant term for the projection, combination, or predication-analysis that semanticists have postulated as the necessary means for capturing the semantic representation of sentences. But just how such a synthesis might take place in a language-user's mind - and even the assumption that it does at all - remains highly controversial. Before pursuing the issues involved, a more formal definition of literal meaning may now be in order, to set beside the earlier informal one: Literal meaning is always language-specific. The meaning of an utterance of a sentence S of language L is said to be literal if it is only composed of the meanings of the words and phrases in S in accordance with the syntactic conventions in L. It is, however, not always clear what the meaning of the words and phrases in S actually is, because the words may have different meanings, and because their meaning often depends on the context C of the utterance. For example, the meaning of indexical expressions such as /, here, now in an utterance, or the meaning of anaphoric expressions such as he, then, that in an utterance depends on C. Even words such as enough, but, otherwise, big, can, and many others, have a context-dependent meaning, i.e., their meaning includes a context variable x. (Wunderlich 1980: 298)
This formulation introduces the crucial issue of context-dependency. The determination of utterer's meaning in using a particular sentence can be seen to rest, typically and heavily, on principles or phenomena that are in a sense both semantic and pragmatic. The expressions are grammaticized, distinctions built into the language (hence semantic); but their specific sense and reference can only be determined in relation to some concrete context. Within the latter area fall (at least) deixis and anaphora and, probably, presupposition. But there are in addition other relatively systematic principles at work, which it is arguable we use in the determination of utterance-meaning: within this area lie proposals concerning conversational implicature and cooperativeness and speech act theory (or, as a recent counter-proposal, relevance theory), and the designs of talk identified by conversational analysts.
Searle on literal meaning relative to a background In fact Wunderlich's description of literal meaning would seem to leave very few sentences sufficiently context-free to have a purely literal meaning. While that might suit anti-literalists it is clearly not what Searle and other speech act theorists have in mind when they speak of literal meaning. They might complain that in writing of 'the
On Inscribed or Literal Meaning
145
meaning of indexicals', and similarly, Wunderlich has not distinguished sentence from utterance: there is a specifiable literal meaning of words such as /, here, and enough, and without this the derivations and determinations of particular contextbound referents and values for these words, in actual utterances, would be impossible. Or, as Searle has put it, literal meaning as such is not affected by the particular reference of an indexical, or the truth or otherwise of a particular statement in context: the truth conditions may vary but the literal meaning cannot. As indicated above, for Searle literal meaning is not entirely context-free. He has emphasized a certain 'contextualizedness' of literal meaning: The notion of the literal meaning of a sentence only has application relative to a set of background assumptions, and furthermore these background assumptions are not all and could not all be realized in the semantic structure of the sentence in the way that presuppositions and indexically dependent elements of the sentence's truth conditions are realized in the semantic structure of the sentence. (Searle 1979: 120)
But this necessary background is hardly of the kind stipulated as required by thoroughgoing contextualists. For example, he has subjected the sentence 'The cat is on the mat' to detailed scrutiny. He imagines the cat and mat floating freely in outer space; or, on earth, suspended separately but contiguously from wires; or with the mat poking stiffly out of the floor and a drugged cat perched on the elevated end of it. It transpires that whether it is literally true to say that the cat is on the mat in any such cases depends on context-specific assumptions as to what might constitute being on the mat. Searle offers such outlandish contexts in order to advance his thesis that the literal meaning of a sentence and its associated determinate truth conditions only have application relative to potentially variable background assumptions that are quite different from the familiar semantic/pragmatic parameters such as indexicality, change of meaning, and presupposition. Furthermore these background assumptions are too numerous to have a place in the semantic structure of the sentence, and besides, the very statement of those assumptions, in literal terms, relies on yet other background assumptions in order to be intelligible. This account may cause confusion. On the one hand, by conceding that literal meaning varies and is context-bound only in relation to the sort of differences of background that most of us would regard as remote from typical utterancecomprehension (cats perched on vertical mats, the supply of a mile-wide hamburger upon ordering 'a hamburger' in a restaurant), Searle seems to be tenaciously protective of literal meaning as a firm foundation for utterance-interpretation, in all 'normal circumstances'. On the other hand, in proceeding to add that such crucial background assumptions are too numerous and endless to permit semantic characterization, he seems more radical than ordinary contextualist pragmaticists who attend to the indexicals, presuppositions, etc. that are believed to admit principled specification. Searle's position amounts to saying that the truthfulness of asserting 'The cat is on the mat' will always involve a kind of provisionality. He represents this provisionality as a provisionality 'relative to background assumptions' without much clarification. But some clarification can be attempted. A first crucial background assumption is an awareness, on the part of both speaker and addressee, that even the
146
Michael Toolan
conventional meanings of 'cat', 'on', 'mat', 'hamburger', etc., are multiple. While relatively institutionalized meanings of 'on', 'hamburger', etc. are all potentially available for adducement, there is typically an implicit ranking of those possible meanings in any particular context of situation, and interactants usually operate on the 'in good faith' assumption that there is considerable orientation to, and awareness of, this context or (as Searle calls it) background. In the restaurant, when the customer asked for a hamburger, did she mean a mile-wide hamburger? Certainly not. As socialized members of the hamburger-eating world, we know what to expect when we order a hamburger in a restaurant, just as surely as the waitperson and cook know what is expected. Furthermore, if we contrast the idea of a hamburger in a fastfood establishment and in an independent restaurant, all the relevant parties know that, in the latter context, hamburger has a somewhat different sense. On the other hand, is it not the case that a mile-wide hamburger is still a hamburger? In a sense, yes it is, we have to concede. But it is not a hamburger in the sense (or in any of the senses) that a customer could in good faith assume as one to which the customer and waitperson would be mutually oriented when the order was made. There is then, for each speaker of the language familiar with hamburgers, a preferential or probabilistic ranking of the meanings of 'hamburger', those rankings (each speaker has their own, nor can we assume identity of rankings by any two speakers) having grown up entirely on the basis of previous experience (the linguocultural experiences of eating out, of categorizing foods, of glossing the language of menus, etc.). As we proceed through the interactions that punctuate our lives (and even the business of identifying some point as the close of one interaction and some other point as the commencement of another interaction is provisional), we are constantly making provisional assessments of the current gestalt - of where we are at now, and of what will likely be understood (what probable sense will be made or taken) if our talking or actions now orient to hamburgers, or cats on mats. An integrationally-minded starting-point will be that each speaker carries in memory a fluid field of probable and possible meanings of every word of which they have had some experience (confining the discussion to previously-encountered words, at this stage, and also postponing the question of how one processes the stream of ongoing experience so as to identify particular words). It would be a mistake to stress that this is cultural experience or social experience, as if to imply that these were contrastive with something to be termed linguistic experience. Experience is experience, whatever its provenance: even having encountered a word only in a dictionary may leave experiential traces in the memory, which may be drawn upon in subsequent encounters. In the course of any situation in which a particular word (e.g., 'hamburger') is used, a complex network of interconnecting assumptions and preferences provides the provisional and revisable justification for the senses of 'hamburger', and of, for example, Give me a hamburger, medium rare, with ketchup and mustard, but easy on the relish that the speaker, in interactional good faith in mutual orientedness, intends.
On Inscribed or Literal Meaning
147
We go through any new interaction armed with the provisional knowledge and construction of the world derived from previous encounters. That knowledge amounts to a running commentary to ourselves which, in the case of ordering a hamburger, might give rise to interpretive processing dramatizable along the following lines: This is a fast-food establishment (again), isn't it? It is open (again), isn't it? I have come here to eat (again), haven't I? This individual in front of me across the counter (it is a counter, isn't it?) (again) is (and knows that he is, doesn't he?) taking orders, isn't he? Etc., etc.
With the copious use of the word again here, the issue of repetition or iterability looms large. But no great theoretical weight should be placed on the everyday term again. When we assume ourselves to be going through routines or normative activities 'again', we are not in fact misled into imagining that the present activity and communication could possibly be truly identical with previous experiences, or that here and now we are once more going through that experience first undergone there and then, or that a present encounter could be merely a mechanical replication of an earlier one - even in the emphatically routine-laden patterns of communication in fastfood restaurant chains. At the very least, we know ourselves to be different and removed from the person we assume ourselves to have been in the past; our very memories of that previous encounter (or those previous encounters), on the basis of which remembrances our sense of repetition is postulated, are equally guarantor that the present encounter, however many ways analogous with previous ones, is not pure replication. In short, we remember the gap: the fact of the orientational shift, and the fact that we remember it, are crucial. If to remember the gap is as important as is here claimed, what might be the consequences of failing to do so? In such a dysfunctional condition, where an individual fails to register experiential difference and the corollaries of temporality and linearity, a psychosis of lived repetition may be apparent, of the kind reported in the case-histories of Oliver Sacks. The again's that we attribute to the elements in our ongoing interactive lives by no means imply identity of current features with prior ones, but rather an assumed and trusted similarity with previous, remembered features. The rational assumptions interactants make as to the similarities between a current experience and earlier ones are fundamental to this account of meaning in interaction. This making sense and pattern of past experience, in memory, may be the basis for the important role in language-use and language-learning that Tannen and others have attributed to repetition: see Tannen (1987), and a special issue of the journal Text (ed. Johnstone 1987), on the same topic. From an integrational perspective, the term 'repetition' remains in need of qualification, being a potentially misleading abbreviated way of referring to perceived relatedness. A similar interactional good faith in mutual orientedness, within a network of assumptions and preferences, is the provisional grounding for the waitperson's interpretation of the 'Give me a hamburger . . .' request. But, to repeat, speaker's intended meaning and addressee's interpreted meaning are, by definition, distinct (i.e., separate); they are also potentially rather different from each other. In short, interactants are guided in utterance-construction and utterance-interpretation by their current sense of the likely uses of such words as
148
Michael Toolan
- in the example under discussion - 'hamburger' and 'give' in such situations as fastfood restaurants, and so on.
Componentialism Componential analysis is only one of many accounts of how the semantic interpretation of sentences proceeds. Nevertheless it probably remains the most influential, the one most widely accepted within mainstream linguistics as at least promising explanatory adequacy. (For trenchant criticism of one of the other most influential theories, truth-conditional semantics, see Harris 1981.) The following characterization of semanticist analysis (from Rumelhart 1979: 81) is a fair representation of the theory: The traditional program of semantic analysis (cf. Katz & Fodor 1963) provides a set of meanings for the individual lexemes of the language and then provides a set of rules of composition whereby the individual meanings of the lexemes are combined to form the meaning of the sentence. Likewise, for any discourse, the meanings of the individual sentences of the discourse can be combined to form the meaning of the discourse. Arguably, this program of semantic analysis can provide a reasonable account of the conveyed meanings (that is, what the listener understands upon hearing the sentence uttered in context) of many sentences in English.
Where conveyed meanings evidently depart rather far from literal meaning, as in metaphors, semanticists want to resist both a solution involving postulating multiple senses, and any adducement of extra contextual criteria, and insist upon the continued validity of their methods of analyzing words into core components of meaning, and the systematic combinations of such components. In the case of metaphor, semanticists (Cohen 1979; Levin 1977) hold to the need for an ordered process of cancellation of incompatible features from the metaphorical vehicle as it is applied to the literal tenor. Other important domains where conveyed utterance meaning is said to depart from the literal meaning of sentences are indirect speech acts and irony. Although the focus of much attention in the 1960's (see especially Katz and Fodor 1963; Bierwisch 1967 and 1970) componential semantic theories have failed to live up to expectations. Aitchison's verdict on the attempt in Schank 1972 to specify the semantic primitives underlying common verbs is representative of the general conclusion: It seems unlikely that Schank has specified 'primitives' in any psychological sense. He has simply found a convenient way of describing 'family resemblances' between the groups of words he was dealing with . . . It is important to distinguish between useful ways of describing things and structures which are likely to exist in speakers' minds. Judging from the work of Schank, there is no reason to believe that verbs are assembled out of pieces in the human mind - even though we might well adopt his system if we were trying to write a computer program which explained relationships between words in a helpful, though non-realistic way. And the problems found with Schank's primitives - difficulty of deciding what they are, their decomposability, and their incompleteness for specifying meaning - occur in most other attempts to identify primitives (e.g. Jackendoff 1983). (Aitchison 1987: 79)
On Inscribed or Literal Meaning
149
At its core, a componential semanticist perspective claims that sentence-meaning is objective, invariant, and relatively context-free, that to share a language is to share or be potential sharers of - an inventory of determinate words and their meanings, the contents of that inventory remaining constant regardless of changes of interactional setting, mood, or purpose. Word meanings, capturable by a relatively delimited set of componential features variously applying to each item in a language's lexicon, and the predictable meanings of grammatical combinations of those word meanings, are uniformly shared by all the members of a speech-community, who employ this fixed bi-planar code (tying determinate forms to determinate meanings) to communicate. Such an assumption of form-meaning determinacy is at the core of most versions of semantics, even though semanticists are ready to acknowledge the innumerable departures from that determinacy in contextualized speech practice. For in all such practice (or performance), it is argued, the foundational system is built upon, so that particular, context-bound additional meanings are conveyed by utterances. Semanticists thus postulate a crucial distinction between sentences and utterances, sentence-meaning and utterance-meaning, and so on. But they argue that the second of each of these pairs depends entirely on the first, that the first is the necessary precondition of the second, and should be the focus of linguistic study, being part of the language proper, or competence, the second merely part of language-use, or performance. Relatedly, semantics or semantic representation (or logical form) is a distinct enterprise, fully a part of (rule-governed) grammar. The 'rough and ready distinction' re-stated in Leech (1983: 6) remains widely adhered to - 'Meaning in pragmatics is defined [in relation to speech situations], whereas meaning in semantics is defined purely as a property of expressions in a given language, in abstraction from particular situations, speakers, or hearers' - particularly in its indication that semantics unequivocally attempts less than pragmatics does (or, in semanticists' own terms, seeks a properly constrained theory of sentence-meaning, about which falsifiable claims may be made). The chief reason to mention componential semantics is to indicate that here, as in Searlean pragmatics, there is a clear swerving away from 'the total speech act in the total situation' (Austin 1962), so as to make meaning, literal or otherwise, a manageable concept within an autonomous and decontextualized linguistics.
The Pragmaticist Tradition At least partly in competition with the semanticist tradition, pragmaticists hold that meaning can only be fully determined by adducing certain delimited aspects of context: word- and sentence-meaning, for adequate and coherent determination, must itself be determined in relation to aspects of interaction that will only emerge from within the relevant speech encounter, rather than being given essentially by the lexicon or grammar. Such contextual aspects include information specifying who speaks (the deictic orientation point), and who or what the referents intended by particular expressions are. Other relevant factors include judgements as to whether interlocutors are being cooperative (seemingly truthful, adequately informative, relevant, suitably brief and orderly): apparently intentional departures from those norms are then assumed to be creative exploitations conveying extra inferences or
150
Michael Toolan
calculable implicatures. For those who espouse a speech act theory version of pragmatics, other factors such as the orderly application of principles and procedures for associating utterance-meaning with sentence-meaning will be germane: for example, application of the theory of indirect speech acts, and the lexical force hypothesis so that sentence-meaning, locution, or sense (what is said) may be mapped onto utterance-meaning, illocution, or force (what is intended and conveyed by what is said). But again, as in semantics, some notion of literal meaning seems crucial for these versions of pragmatics. In so far as their project is more ambitious than that of semanticists, being the goal of identifying the principles and conventional forms by means of which intentional meanings are conveyed in performance, their work diverts attention from the problematic notion of literal meaning in the process of attempting to specify, in predictable ways, actual meanings. That is to say, declared interest in interlocutors' actual meanings brings in train such an enlarged attention to the multiple linguistic and extralinguistic dimensions of utterances (as these arise in response to the particular perceived needs, interests and goals of the equally heterogenous set of individuals within a speech community) that doubts grow concerning the role of literal meaning in such complex negotiations. Nevertheless, as in the work of Searle and as is evident in various remarks in Austin (1962: 94ff.) and Grice (1975: 44, 50), some notion of literal or conventional meaning remains a model-shaping element in pragmatic accounts. Whatever idealization called literal or conventional meaning is retained by the theorist, the crucial need is that such constructs be depicted much more explicitly as a matter of cultural habituation and contingency than has hitherto been the case. The degree of resistance to the relinquishing of a more-than-contingent characterization of sentence meaning (vs. utterance meaning, etc.) is vividly illustrated in some remarks of Rorty, critiquing the 'against theory' position of fellow pragmatists Knapp and Michaels (1982). For Knapp and Michaels authorial meaning and textual meaning are always one and the same thing, and there is never - genuinely as it were, as distinct from within the contrivances of theory - a separateness of the one from the other, for analysts to explicate. This Rorty sees as an ill-judged rejection of the foundational distinction, accepted by Grice along with many others, between sentence-meaning and utterance-meaning. Supporting Grice, Rorty claims the distinction opens up a logical space - 'the space in which one asks the traditional interpretive question "Granted that the sentence means such and such, did its author use it to mean that on this particular occasion?"' (Rorty 1985: 460). And that, for Rorty, is a 'useful' question, and indeed is of a piece, he implies, with 'Grice's handy distinction between more and less familiar contexts in which to place words' (1985: 461), and a distinction between two sets of intentions, 'the ones normally had by users of a sentence and some special ones had, or possibly had, by an individual user' (1985: 464, n.4). An immediate doubt must concern whether the three distinctions Rorty upholds here (sentence meaning vs. utterance meaning; more familiar vs. less familiar; normal intentions associated with use of a sentence vs. special intentions) are all of quite the same kind. But in any event in their rejoinder to Rorty, Knapp and Michaels declare
On Inscribed or Literal Meaning
151
themselves if not more pragmatist then at least less rationalist than he is: his 'logical' space between sentence and utterance is for them purely empirical. In our view . . . normal intentions are just frequent particular ones. The 'space' between these two sets of intentions is not logical but empirical. A text means what its author intended it to mean whether or not other authors on other occasions use the same marks (or noises) to say the same thing. What people normally mean when they shout 'Fire!' might well be that something is burning. On some special occasion, however, someone might shout 'Fire!' and mean by it 'Discharge your weapon'. In the first case, the speaker has an intention frequently had by other speakers on similar occasions; in the second case, not. Indeed, there could be a third case in which a speaker might shout 'Fire!' and mean by it something that no one had ever meant before or would ever mean again. But in all three cases the relation between meaning and intention would be the same: the sentence would mean only what its speaker intended. (Knapp and Michaels 1985: 469)
Minor adjustments to the foregoing formulations could be proposed (to emphasize that no-one can actually shout the written form 'Fire!', but only sounds that might be transcribed as 'Fire!'; and to suggest that 'sentence' is Knapp and Michaels' shorthand way of referring to a situationally-embedded use of language), but in large part this account accords with an integrational linguistic view. At the same time the integrational approach concurs with Rorty's idea that distinctions such as the sentence meaning/utterance meaning one are 'useful' and 'handy'; but it denies that they are appropriate as axiomatic foundations of language theory. The conclusion that examination of particular cases of utterance-interpretation invites (and in the last analysis, it is supposedly as idealizations from a range of such particular cases that our abstract categories derive their status) is that both of the terms 'literal meaning' and 'metaphorical utterance' are precipitations from authorized uses, and have no other more permanent grounding, no foundational status in some autonomous and transcendent system we call 'the language'. Literal meanings, of utterances which we metalinguistically term words or sentences, are the most established and authorized uses of expressions. It just so happens that, within these historical conditions and given such and such distributions of power to signify authoritatively through the language, such and such uses have become accepted, recognized, established, habitual, and codified in dictionaries as 'literal meanings'. Equally and contrariwise, uses of expressions in ways that are in no way established or habitual are potentially (creatively) metaphorical; and creative metaphorical meanings comprise a subset of such non-habitual non-accepted uses, where the expression is used in a way which is seemingly or actually unique. At the immediate level of language use, then, uniqueness is argued to be pervasive. Accordingly there might appear to be a clash between this view and the claim here that metaphorical utterance or meaning stands out vis-a-vis literal and habitual meanings, in its seeming uniqueness. But the point to emphasize here is that the entire system of categorization explored in this and the following chapter, in terms of the literal, the metaphorical, words, sentences, etc., is thoroughly metalinguistic. It is at this culture-reflective level of metalinguistic characterization that we can contrast literal meaning as authorized and established, with metaphorical meaning as uncodified use. And words that the dictionary records as n-ways polysemic are words with n literal meanings.
152
Michael Toolan
Integrational linguistics As indicated above, both standard semantic and pragmatic approaches hold to the theoretical importance of the notion of literal or conventional meaning in the comprehension of conveyed meanings. This position is an orthodoxy which pervades the literature, and has been widely taken up in neighbouring disciplines. Clark and Lucy (1975), for example, writing from a background in psychology, propose the following three-stage model (reported in Rumelhart 1979: 82) for comprehension of indirect requests (and here we should remember that most requests - in fact most speech acts - are indirect by this definition): A person must: (a) determine the literal meaning of the utterance (presumably by a simple composition of lexeme meanings); (b) compare the literal meaning so determined with various contextual and conversational rules to decide if the literal meaning could also be the intended meaning; (c) if the literal meaning is determined inappropriate, apply additional rules to determine the indirect, or conveyed, meaning.
The integrational linguistic response to this would take the form of close consideration of the stages of interpretation postulated above. Though sceptical of any secure boundary between the literal and the non-literal of the kind assumed in stage (a), that stage might be passed over and attention focussed on the phrasing and assumptions of stage (b). For here the difficulties inherent in so many such 'stepwise' accounts of comprehension is displayed. A first but minor complaint is that stage two asserts we decide if the literal meaning could be the intended meaning of the utterance: this is misleadingly tentative, for by definition literal meaning is always a possible reading of an utterance. The major complaint follows from making what seems to be an inescapable assumption, namely that what the authors mean to say here is that we in fact decide if the literal meaning is (not 'could be') the intended meaning. They do not specify how we decide this question, but were they to attempt to do so it would soon emerge that any such decision would rest on a comparison: a comparison of the computed literal meaning with a prior understanding of what the utterance, in context, meant. Without some such prior understanding, the implicit task specified by these stages, whether it is phrased in terms of comparison or appropriacy or reconciliation-of-anomaly, simply could not be undertaken; but if we do have that prior understanding, these stages are superfluous. The integrationist does not deny a role for principles of comparison and appropriacy in the ordered and incremental rational determination of utterancemeaning in context. What is contested is whether these roles amount to a simple and routine manipulation of fixed form-meaning pairs in a way that is delimitable without consideration, as more than a secondary modifying component, of context. Indeed, rather than referring directly to context, the interpreter in Clark and Lucy's account is said at stage (b) to invoke 'various contextual and conversational rules'. These rules are not specified, nor is it clear how an interpreter would know which ones to invoke. In short, the account as given is unworkable. With Rumelhart, it is assumed that an adequate theory of language comprehension will be able to deal equally directly with conventional meanings, metaphor, irony, and other so-called indirect utterances,
On Inscribed or Literal Meaning
153
rather than promoting a dichotomized account in which one set of meanings (metaphor, irony) must be calculated via a core (literal meaning): The processes involved in the comprehension of nonfigurative language are no less dependent on knowledge of the world than those involved in figurative language. Any theory rich enough to generate the meanings people actually assign to nonfigurative language is rich enough to deal with figurative language as well. (Rumelhart 1979: 84)
Integrationists tend not to express the comprehension-process in quite the psychological terminology of Rumelhart, who talks of the hearer selecting and verifying some configuration of schemata (from, presumably, an open set of such schemata) that offers a sufficient account of the situation. But they share his contention that a standard bottom-up view of meaning-construction, with nonlinguistic knowledge only ever coming into play after language-based determination of possible meanings, seems to receive no confirmation in interactional practice. For one thing, that view operates upon the illicit assumption that the world is already and definitively categorized into domains recognizable as 'linguistic' and 'non- or extra-linguistic'. In an integrational linguistic spirit, then, and quite counter to the compositionalist position noted earlier, I suggest that we learn and store lexical items, and even whole utterances, with contexts attached. The norm, in such expression-cum-context learning, will be for a range of contexts to be associated with any particular expression, and a range of expressions to be associated with any particular context. The likelihood of different speakers of a language having identical combinations of expressions and contexts, to be invoked in their determinations of what is meant by any particular situated utterance is consequently very slim: this picture relates difference in language-understanding directly to variation in life-experience and memory. Clearly this suggestion only addresses the first assimilation, as it were, of new experience (linguistic or otherwise). We have every reason to believe that these relatively 'uninterpreted' interactional memories are subsequently sorted, matched, backgrounded and virtually forgotten, or foregrounded into mental models, schemata, scripts, notions of genre and grammaticality and contextual appropriateness, etc. etc., with the passing of time and the development of interactional (including linguistic) proficiency. Speculations concerning these kinds of ranking and sorting, concerning the architecture of memory cannot be developed here, nor need they be: the crucial and initial premise, in rejection of reliance on a notion of literal meaning, is that learning the meanings of expressions should scarcely be called a learning at all, but should be understood as an experiencing. And again, that experiencing is not one of meanings of expressions, but an experiencing of expressions integrated within particular occasions of use. Whatever sorting and extrapolation occurs, occurs after the fact, and is performed by the reflecting thinking language-using individual; it is in no way given in or by the language. What is undoubtedly given, by other languageusers rather than the language, and what will thus be experienced and added to the individual's mental modellings, is abundant feedback and example as to how the language is ordinarily used. One influential cognitivist proposal which may be viewed as just such a normative pressure is prototype theory, a theory which in addition
154
Michael Toolan
might seem to sidestep some of the problems that attach to a standard notion of literal meaning.
Prototype theory In the last twenty years an ingenious and seemingly empirically-based account of normativity in concept-interpretation has emerged from cognitive psychology: Rosch's prototype theory. The way this theory has been elaborated however, and the tendency - among some psycholinguists - to assume a collective agreement on the degrees of typicality attributed to particular exponents of any given word or concept, are an indication of the power of the notion that in language there are given basic meanings. For Rosch's part, it should be noted that her later papers have withdrawn some of the more speculative but problematic claims of earlier papers, in favour of a more guarded position: she argues, for example, that rather than there being prototypes as such, what her research has actually displayed are 'judgments of degree of prototypicality' and various prototype effects; prototypes 'do not constitute any particular processing model for categories . . . [nor] a theory of representation for categories' (1978: 40-41). These qualifications notwithstanding, Aitchison (1987) is representative of others in endorsing Roschian prototype theory in the course of her enquiry into the nature of the mental lexicon. She agrees that there is no great oddity in our use of the term tiger to refer even to tigers with only three legs, lacking stripes, and so on. Similarly, there is nothing that defies explanation in our use of the verb 'see' in cases where, apparently, the mind fails to register what the eyes focus upon (as in 'I've seen that picture a hundred times but never noticed it'). These are simply nonprototypical instances of tigers and seeing. We use fuzzy-edged words all the time, the argument runs, in situations where one or another such 'typicality conditions' (such as, of tigers, that they have stripes) is broken. But the danger of this way of representing matters lies in the apparent assumption that the prototypicality of tigers, seeing, etc. stays intact and is inviolable, and that somewhat secondary typicality conditions are what are met or flouted in extended uses of words, metaphors, and so on. The prototype is often characterized, in the literature, as the best exemplar or central member of a category, with the implication that the 'centrality' of this exemplary type inheres in its having one or more criterial attributes (not attributes that all members of the category must necessarily bear, but nevertheless ones which are focal or central to the category). Similarly, Lakoff invokes Wittgenstein's notion of 'family resemblance', but casts this as 'the idea that members of a category may be related . . . without all members having any properties in common that define the category' (1987: 12 and passim). In both cases it looks as if an attempt is being made to have one's cake (category-membership is not a matter of category-wide sharing of specific properties) and eating it too (categories are welldefined, and structured, relative to a core or central exemplar which bears exemplary attributes). Notice the awkwardness of the scope expressions in Lakoff s gloss of 'family resemblance': 'Without all members having any properties in common'. This reflects Lakoff's wish (which is likely to be shared by any cognitivist theorist of prototypes) to cling to the idea that most category-members will share some properties. It may also invite the following kind of inference:
On Inscribed or Literal Meaning
155
Although it is not the case that all members have any properties in common that define the category, still it is the case that some members have properties in common that define the category.
But a more radical gloss of 'family resemblance' would be simply to the effect that, in such categories, no properties are shared by all members, and none of the properties shared by only some of the members defines the category. The implication is that there are certain core typicality conditions which necessarily apply, whenever the word tiger or see is invoked. But where does prototypicality come from, except from normatively-constrained usage? A similar reification to that which turns literal meaning into an unchanging given is in danger of extending to prototype theory also. Thus Aitchison asks, of the sentence The price of mangos went up: 'Is went up a metaphor, since the price did not literally travel up a hill? Or is this simply the use of go up with a typicality condition broken, since going typically involves travelling between two physical points, and covering the distance in between?' (Aitchison 1987: 144-5). Whether Aitchison seriously intends these explanations to be disjunctive is unclear, since earlier she states that the basic mechanism behind metaphor 'is simply the use of a word with one or more of the 'typicality conditions' attached to it broken' (ibid.: 144). (This would seem to make the explanation offered in the second sentence quoted a precondition for the applicability of the explanation given in the first, and not an alternative to it.) But it is with the premise here that objections begin: the premise that went up is not typically used to refer to increase. The one word signally absent from discussions of prototypicality, which is instead characterized in terms of cognitive grasp or modelling, is use. The role of use, and frequency of use, continues to be undervalued even in conspectual studies (e.g. Lakoff 1987) which invoke aspects of the later Wittgenstein's thinking. Thus Lakoff notes that Rosch's prototypicality findings do not force the conclusion that membership of categories is ordinarily gradable, so that, e.g. a penguin counts as a not-very-birdy bird; rather, the findings are still consistent with the idea that categories are rigid, and that a penguin is fully a member of the bird category. And yet, Lakoff adds, 'that category must have additional internal structure of some sort that produces these goodness-of-example ratings' (1987: 45). One is bound to ask why additional internal structure must be involved; a perfectly adequate alternative account would point to the role of use and habituation (the frequency, especially in conventional situations, of reference to robins and robin-like birds in all matters to do with birds, and the relative rarity of reference to penguins, ostriches, etc.) together with, in the case of color-naming and -recognition, neurophysiology. The ease with which we make sense of talk of prices rising or going up, of people 'coming out' of comas, or of teenagers being 'under' age, may chiefly be a reflection of the frequency with which these phrases are encountered used in just these ways. In that case, the reification called 'semantic prototype' has no more work to do than that called 'literal meaning'. When Aitchison claims that ' coming out typically involves physical movement, and under typically involves a physical position underneath' (1987: 145), it looks increasingly possible that prototypicality is simply newfangled literalism - the literalism of historically-oriented dictionaries which report that the earliest written uses of expressions like come out, under, go up were to denote various physical movements and locations. The breadth of historical development of
156
Michael Toolan
expressions from originally concrete and physical uses to (even predominantly) abstract and derived ones is a rich field (see Traugott 1985 and Hopper and Traugott 1993), but the historical linguist's perspective is not that of the ordinary languageuser, for whom the use of coming, going, up, down etc. as physical denominators may not be so typical as to be more prominent or core than various other common uses. Concluding remarks The purpose of this chapter has been not only to resist the notion of literal meaning as an essentialist delusion or an instrument of enforced communication, but also to consider in detail what the notion entails, and what consequences follow, for one's view of language and agency. I have argued that literal meaning is always a controversial abstraction from linguistic (and, particularly, written language) behaviour; while heuristically and pedagogically useful, it remains ontologically and epistemologically suspect if not misleading. It is a myth, as Rommetveit notes, and so is not to be dismissed lightly: 'The myth of literal meaning addresses the basic semantic riddle of how the Many may become One. It tries to explain how language can bring about a spiritual union of the Many, how one state of intersubjectivity or 'shared meaning' is attained by linguistic means in encounters between different subjective worlds (1988: 13). One function of the myth is to act as a bulwark against the inroads upon the sociocultural order that continued reflection on Lockean uncertainties about intersubjective communication might cause. And as Rommetveit also emphasizes, literal meaning is an important myth since in our highly literate societies it is 'a reality lived by enlightened laymen under subtle influence from stories told by prominent scholars of semantics' (ibid.: 15). We can and should extend the burden of responsibility, however, beyond the ranks of prominent scholars of semantics - particularly if we keep in mind that literal meaning is simply, as it were, the quintessence of more widely-invoked authorized abstractions, viz., conventional meanings. The representing of the world in the terms authorized by conventional meanings and their most ossified forms, literal meanings, is something we are all implicated in. But, like Orwell's egalitarian animals, some are undoubtedly more authorial and implicated than others - lexicographers, editors, publishers, advertising copywriters, media stars, journalists, guardians of 'good' English, writers, teachers - just as some activities are more authorial and implicating than others - composition classes, word games, news bulletins, editing, lectures, the discourses of the more influential professions, and so on. Literal meaning is an important myth particularly within literate societies with developed traditions of legal, literary and linguistic studies recorded in written language. While it is a myth, it is also real, with large consequences for the conduct of various societal functions. However, the ontological status of literal meaning's reality is not of that foundational order, essential to the very emergence of language, which standard linguistics (and legal studies, etc.) requires, and under the compulsion of that requirement, presupposes. At the heart of all linguistic communication - not just in the use of creative metaphors - lies risk: the risk involved in attempting to get one's interlocutor to see the same picture of some aspect of reality that you, the speaker, believe you see. This
On Inscribed or Literal Meaning
157
element of risk is of course troubling, especially to the telementationally minded, who would wish that language 'at its most precise and unambiguous' might shuttle ideas from mind to mind in a flawless, friction-free manner. On the other hand, and from outside the telementational mindset, the risk involved is part and parcel of genuine linguistic creativity: language can only genuinely go wrong when one is genuinely remaking language as one goes along. A proper respect for the indeterminacies of language should help us see the importance of interactional risk, as Derrida has noted: 'Does the quality of risk [or infelicity, voided performative, through shift of context/circumstances] admitted by Austin surround language like a kind of ditch or external place of perdition which speech could never hope to leave, but which it can escape by remaining 'at home', by and in itself, in the shelter of its essence or telosl Or, on the contrary, is this risk rather its internal and positive condition of possibility? Is that outside its inside, the very force and law of its emergence?' (Derrida 1977: 190).
References Aitchison, J. (1987) Words in the Mind: the Mental Lexicon. Blackwell, Oxford. Austin, J. (1962) How to Do Things with Words. Oxford University Press, Oxford. Bierwisch, M. (1967) Some semantic universals of German adjectivals. Foundations of Language 3, 1-36. (1970) Semantics. In J. Lyons (ed.) New Horizons in Linguistics, pp. 166-84. Penguin, Harmondsworth. Clark, H., and P. Lucy (1975) Inferring what is meant from what is said: a study in conversationally conveyed requests. Journal of Verbal Learning and Verbal Behavior 14, 56-72. Cohen, L. J. (1979) The semantics of metaphor. In A. Ortony (ed.) Metaphor and Thought, pp. 6477. Cambridge University Press, Cambridge. Derrida, J. (1977) Limited Inc abc . . . Glyph 2, 162-254. Grice, H. P. (1975) Logic and conversation. Syntax and Semantics 3: Speech Acts, ed. by P. Cole and J. Morgan, pp. 41-58. Academic Press, New York. Harris, R. (1981) Truth-conditional semantics and natural languages. In T.E. Hope, T.B.W. Reid, R. Harris and G. Price (eds.) Language, Meaning, and Style: Essays in Memory of Stephen Ullmann, pp. 21-37. Leeds University Press, Leeds. Reprinted in N. Love (ed.) The Foundations of Linguistic Theory: Selected Writings of Roy Harris, Routledge, London, 1990, pp. 96-111. Hopper, P. and E. Traugott (1993) Grammaticalization. Cambridge University Press, Cambridge. Johnstone, B. (ed.) 1987. Special issue on repetition. Text! (3). Katz, J., and J. Fodor (1963) The structure of a semantic theory. Language 39, 170-210. Knapp, S. and W. Michaels (1982) Against theory. Critical Inquiry 8 (4), 723-42. (1985) Pragmatism and literary theory III: a reply to Richard Rorty, What is Pragmatism?. Critical Inquiry 11 (3), 466-7. Lakoff, G. (1987) Women, Fire and Dangerous Things. University of Chicago Press, Chicago. Leech, G. 1983. Principles of Pragmatics. Longman, London. Levin, S. (1977) The Semantics of Metaphor. Johns Hopkins University Press, Baltimore. Rommetviet, R. (1988) On literacy and the myth of literal meaning. In R. Saljo (ed.) The Written World, pp. 13-40. Springer Verlag, Heidelberg. Rorty, R. (1985) Pragmatism and literary theory II: philosophy without principles. Critical Inquiry 11 (3),459-65. Rosch, E. (1978) Principles of categorization. In E. Rosch and B. Lloyd (eds.) Cognition and Categorization, pp. 111-44. Academic Press, New York.
158
Michael Toolan
Rumelhart, D. (1979) Some problems with the notion of literal meanings. In A. Ortony (ed.) Metaphor and Thought, pp. 78-90. Cambridge University Press, Cambridge. Schank, R.C. (1972) Conceptual dependency: a theory of natural language understanding. Cognitive Psychology 3, 552-631. Searle, J. (1979) Meaning and Expression. Cambridge University Press, Cambridge. Tannen, D. (1987) Repetition in conversation: Toward a poetics of talk. Language 63, 574-605. Toolan, M. (1996) Total Speech: An Integrational Linguistic Approach to Language. Duke University Press, Durham and London. Traugott, E. (1985) On regularity in semantic change. Journal of Literary Semantics 14, 155-73. Wunderlich, D. (1980) Methodological remarks on speech act theory. In J. Searle, F. Kiefer, and M. Bierwisch (eds.), Speech act Theory and Pragmatics, pp. 291-312. Riedel, Dordrecht.
11 Irony and Theories of Meaning STEVE FARROW
From their inception in the early years of this century, theoretical linguistics and philosophy of language have made a fundamental assumption about meaning and communication: understanding what someone says (or writes) is a process of calculating the meaning from the signs produced. The appropriateness of what is said, and the intentions and beliefs of the speaker (or hearer) were, from the very beginning, either ignored, or it was denied outright that they could affect the meaning of a sentence or proposition. This assumption will be examined with particular reference to the concept of verbal irony. Ferdinand de Saussure revolutionized linguistic theory when his Cours de linguistique generate was published posthumously in 1916. In that work, communication was described as a mechanical process of 'transmission' and 'association': Suppose, then, we have two people, A and B, talking to each other:
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, 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 B 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. (Saussure 1983: 27-28)
159
160
Steve Farrow
'Concepts' are 'expressed' by 'sound patterns', and to understand what A has said, B performs (or undergoes?) a physiological process (when the sound pattern is transmitted from ear to brain), and a psychological process in the brain, when the brain matches the sound pattern 'with the corresponding concept'. Note that B does not have to judge whether or not A believed what he said, or decide whether or not A intended to deceive B. The process of communication is complete when the somewhat mechanistic processes in B's head have run their course and the 'corresponding concept' has been produced: B has then understood what A said to him. At about the same time that the Cours was published, Ludwig Wittgenstein, the most influential philosopher of the twentieth century, was working on the Tractatus Logico-Philosophicus, which was eventually published in German in 1921, the English translation appearing a year later. In this work Wittgenstein declared that 'a proposition has a sense that is independent of the facts' (Wittgenstein 1961: 4.061); if those 'facts' include the fact that, for example, B did not believe what A said, or the fact that A did not intend to deceive B, then Wittgenstein's account of communication begins to display a certain 'family resemblance' to Saussure's. The similarity becomes even more marked when Wittgenstein declares that a proposition 'is understood by anyone who understands its constituents' (Wittgenstein 1961: 4.024). Once again, no reference to intentions, beliefs, or the context of the utterance is required: meaning is a property of the symbol, and understanding a sequence of symbols is to know which symbols have which properties. Saussure in the Cours never discusses intentionality or linguistic appropriateness: such considerations are clearly felt to be irrelevant to the interpretation of linguistic signs. Indeed, Saussure could hardly have argued otherwise without admitting that psychology and sociology (or anthropology) were branches of linguistics, and that Saussure was not willing to do: By proceeding thus one opens the door to various sciences - psychology, anthropology . . . which are to be distinguished from linguistics. These sciences could lay claim to language as falling in their domain; but their methods are not the ones that are needed. (Saussure 1983: 24-5)
For linguistics to be a science, according to Saussure, it must have its own object of study, and concepts like intention would have necessitated the unwelcome intrusion of the psychologist into linguistics. The young Wittgenstein, on the other hand, did feel called upon to address such issues. In the Tractatus he poses an interesting question: 'Can we not make ourselves understood with false propositions just as we have done up till now with true ones? So long as it is known that they are meant to be false'. (Wittgenstein 1961: 4.062). Philosophical commentators have made heavy weather of this passage (Black 1964: 179; Stenius 1960: 173-6), but none appears to have considered the possibility that Wittgenstein is in fact confronting the problem of irony. The communicative event postulated by Wittgenstein fulfils the three criteria for a speech act to be ironical: the proposition is inappropriate ('false'), it was 'meant' to be inappropriate, thus there is the requisite intention, and it is 'known' to be inappropriate - there is no mention of deceit, as there would be if the speaker was simply lying. Wittgenstein's answer to his own question is startingly emphatic:
Irony and Theories of Meaning
161
No! For a proposition is true if we use it to say that things stand in a certain way, and they do; and if by 'p' we mean ~p and things stand as we mean that they do, then, construed in the new way, 'p' is true and not false. (Wittgenstein 1961: 4.062)
Let us take an example to make sure that we have understood Wittgenstein's position: Imagine a pair of weary travellers stepping from the air-conditioned cool of a plane at Kai Tak airport into the heat of the Hong Kong summer; the first traveller remarks to his companion that 'It's cold here', and his friend laughs and nods in agreement. Here we have a false proposition that was meant to be false and known to be false by both speaker and hearer; according to Wittgenstein, this ironical utterance could not be understood. Wittgenstein's account of the situation outlined above presumably runs as follows: We have here a case of linguistic change; p ('It's cold here') no longer means what it used to mean (it now means 'It's warm here'), and hence it is an illusion to think that something was communicated by means of a false proposition: p ('It's cold here') has changed its meaning (to 'It's warm here'), and is, in fact, true. But such an argument encounters the obvious objection that irony and linguistic change are categorically distinct phenomena: no lexicographer of English is going to revise his definition of the English word cold simply because someone makes an ironical comment about the weather. Wittgenstein has little option but to insist that the sarcastic weather-commentator is either a liar or a fool; needless to say, such a judgement would appear to most competent speakers of English as being not just misguided but positively perverse. It is not difficult to understand why Wittgenstein puts himself through such contortions in order to get around the problem of irony: his account of the proposition is at stake. Wittgenstein argues that the proposition must have 'a sense that is independent of the facts' (Wittgenstein 1961: 4.061). If under 'facts' we include the state of mind or intention of the speaker (whether he is being serious or ironical), as well as the state of affairs in the world being referred to, then the possibility of irony shows us that propositions do not have meanings that are 'independent of the facts'. To cite our earlier example: if the weather at the airport had indeed been cold then it is difficult (but not impossible) to imagine that the utterance 'It's cold here' would have meant 'It's warm here'. And if the weather had been warm, but the speaker had not been joking (perhaps he was feverish, or he was used to the intense heat of the Sahara Desert) then the proposition 'It's cold here' could presumably have been interpreted as meaning 'It's cold here'. Irony reveals to us that linguistic communication does not allow us to rigidly distinguish between signs, states of affairs, and communicative intentions. One possible response to these Wittgensteinian dilemmas would be to distinguish between propositions and utterances, and to supplement the account of the proposition with an account of how that proposition is to be understood in a context. So, for example, Michael Dummett's description of how a sentence is understood might be supplemented by an appeal to H. P. Grice's work on conversational implicature. Dummett argues that A theory of meaning will. . . represent the practical ability possessed by a speaker as consisting in his grasp of a set of propositions; since the speaker derives his understanding of a sentence
162
Steve Farrow
from the meanings of its component words, these propositions will most naturally form a deductively connected system. (Dummett 1976: 70)
Grice adds a number of 'maxims' to this 'set of propositions', one of which is the 'maxim of Quality': 'Do not say what you believe to be false' (Grice 1975: 46). An ironical utterance flouts this maxim, but Grice argues that hearers will assume that, at a deeper level, the speaker is co-operating with them, and so the hearer will interpret the utterance whilst making due allowance for the ironical intentions of the speaker: Irony. X, with whom A has been on close terms until now, has betrayed a secret of A's to a business rival. A and his audience both know this. A says 'X is a fine friend'. (Gloss: It is perfectly obvious to A and his audience that what A has said or has made as if to say is something he does not believe, and the audience knows that A knows that this is obvious to the audience. So, unless A's utterance is entirely pointless, A must be trying to get across some other proposition than the one he purports to be putting forward. This must be some obviously related proposition; the most obviously related proposition is the contradictory of the one he purports to be putting forward.) (Grice 1975: 53)
In other words, B hears A say that 'It's cold here', and further decides that A does not intend to deceive B, and hence is not a liar (as he would be if he did not believe what he asserted, but he did intend to deceive B). Criteria would have to be specified by reference to which B makes judgements about A's beliefs and intentions. In short, Dummett's account of how the meaning of a sentence is derived from the meanings of its component words would be supplemented by sets of criteria relating to intention and belief; the final (possibly ironical) meaning of the utterance would be derived from the speaker's knowledge of these criteria, and his judgement as to whether the criteria had been satisfied. The difficulty with this manoeuvre is that one can no longer claim to be giving a derivational account of how the meaning of an utterance is calculated. That is, criteria relating to intention and belief would necessarily be defeasible, for there may be no external manifestation or symptom that A holds a certain belief or has a certain intention, but it does not follow that A does not have such an intention or belief. A person may believe that God exists, but never reveal that belief in either word or deed; similarly, a person may intend to write a best-selling romantic novel, but never reveal that intention either. Thus, A may say that 'It's cold here' but give no indication of his intentions or beliefs by keeping, as we say, a perfectly straight face. The hearer may take a guess, judge that A is being ironical, and so interpret the utterance as meaning 'It's warm here'. But guesswork is not a derivational process, and the propositions that speakers are said to implicitly know could hardly be said to 'form a deductively connected system'. The meaning of the original sentence could, we may grant for the sake of argument, be 'calculated' from the symbol, but the intentions and beliefs of the speaker could never be calculated in the same way, for we may be mistaken in our belief that A is being ironical, even though he has a smirk on his face as he utters the words 'It's cold here' (he might be in pain, or be afflicted with a nervous tic), but we could never be mistaken about a calculation or a genuine process of derivation: we could never subsequently discover that in this context 2 + 2 = 3, and not 4.
Irony and Theories of Meaning
163
Dummett demands of any theory of meaning that it yield a theory of understanding: Any theory of meaning which was not, or did not immediately yield a theory of understanding, would not satisfy the purpose for which . . . we require a theory of meaning. For I have argued that a theory of meaning is required to make the workings of language open to view. (Dummett 1975: 100-1)
But no theory of meaning, in the sense of a deductively connected series of propositions, could show that each and every ironical utterance was understood as a result of a derivational process of understanding. Such an understanding would be achieved after a series of more or less reliable guesses; in which case, the theory does not yield a theory of understanding, and so the rationale for the philosopher's theorizing collapses. Possible responses to the problem of irony have been either to deny that an ironical utterance could be understood, or to try to develop an expanded theory of meaning which allows one to calculate the significance of the context, add that to the meaning of the sentence (which has been derived from the meaning of its constituent words), and to compute from this the total meaning of the utterance. The first of these alternatives failed to square with our everyday communicational experience which tells us that irony is an ever-present possibility and indeed actuality. The second alternative failed to square with the demand that a theory of meaning should be a deductively connected system that explains how utterances are understood. The Gricean addendum was not a theory in this sense, but amounted to little more than a reminder to the effect that intention and belief are relevant to the concept of irony; but we have been aware of that for the last two millennia. A third position as regards irony is that adopted by J. J. Katz in his paper 'Literal Meaning and Logical Theory' (1981). Katz criticizes Dummett for demanding too much of a theory of meaning by requiring that it yield a theory of understanding; such a requirement, argues Katz, 'has the effect of pushing the boundaries of semantics beyond language into matters of everyday manners and morals, psychology and sociology, and so on', (1981: 207) - a position, we might note, not wholly dissimilar to Saussure's. Instead, Katz proposes that a theory of meaning need not yield a theory of understanding, but that it be a theory of the meanings understood. All aspects of manners, morals, psychology, etc. required to explain understanding and use now fall outside the domain of semantics. (1981: 207)
Meaning, for Katz, 'is to be accounted for compositionally as a function of the meanings' of the words comprising a sentence; any 'talk about intentions and background assumptions expressing information about context is idle' (1981: 232). Katz is not denying the possibility of irony, for he concedes that 'almost every word has ironic as well as literal uses' (1981: 210); nor is Katz denying that a theory of meaning is a deductively connected system that allows a hearer to derive the meaning of a sentence from the meaning of its component words. What is being denied is that we can have a theory of how that sentence is understood in a context, or rather if there were such a theory then it would be psychological or sociological rather than semantic. Katz does not wish to challenge the orthodox assumptions; he merely
164
Steve Farrow
wishes to point out that some theorists have been a little too ambitious: we may have a theory of the sentence, but not a theory of the utterance. And indeed, such a position does appear to have a certain attraction. The young Wittgenstein, Dummett, and Grice would all have agreed that a theory of the sentence or proposition is not just a possibility, but a prerequisite to any further account of how that proposition is understood in a context. The proposition p means 'p', but if we subsequently discover that it was uttered ironically then it actually meant '~p'; but in order to determine that we had to first know that p meant 'p', and only then could we judge that it was false, or inappropriate in some other way. Katz tries to provide what all the theorists require: the meaning of the sentence or proposition in what Katz describes as the 'zero' or 'null' context (1981: 217). How it is that we actually understand an ironical utterance is an issue that the Katzian linguist does not feel called upon to address. But such talk of a 'null' or 'zero' context should give us pause, for a 'zero' or 'null' context is the linguist's equivalent of Orwellian Doublethink. One may talk of the 'usual' or 'normal' context, or one may say that we simply do not know what the context was for a certain sequence of inscriptions, in which case we do not even know if the inscriptions are meaningful at all. But a 'normal' context is still a context, as is an unknown context; there simply is no such thing as a 'null' or 'zero' context. Furthermore, the distinction between the sentence (or proposition) and the utterance is equally invalid: a proposition is always either inscribed or uttered, and it is always either inscribed or uttered in a context. Katz's 'null context' is a stale rehash of the young Wittgenstein's 'unasserted proposition' (Wittgenstein 1979: 96). Even if sense could be made of the idea of a 'null context', that still leaves the problem of how to determine the meaning of a 'context-free form'. Any such determination or assignment of meaning would presumably have to begin from observation of how words are used in contexts; but anyone observing such usage would see that, for example, granted the possibility of irony, cold could mean either 'hot' or 'cold'. Katz sees this as an unanswered objection to 'use' theories of meaning (Katz 1981: 210), but it is in fact a reductio of his own position: how could we assign a meaning to a 'context-free form' if, in actual context, the meaning is indeterminate?
In Wittgenstein's later work the problem of irony appears to arise once again. It occurs in a section where Wittgenstein is arguing that there is nothing hidden 'behind' our signs, or at least whatever is hidden (in the mind) is irrelevant to our understanding of signs: If I give someone an order I feel it to be quite enough to give him signs. And I should never say: this is only words, and I have got to get behind the words. (Wittgenstein 1958: 503; emphasis in original)
But the point to make here is that the person giving the order may have been speaking ironically: in some sense we do have to 'get behind words'. Wittgenstein anticipates the objection:
Irony and Theories of Meaning
165
But if you say 'How am I to know what he means, when I see nothing but the signs he gives?' then I say: 'How is he to know what he means, when he has nothing but the signs either?' (Wittgenstein 1958: 504; emphasis in original)
Such a reply, however, is curiously weak: Wittgenstein should have argued that one does not 'see nothing but the signs', but also 'tone, gesture, or known circumstance' (Sedgewick 1948: 5). Both speaker and hearer generally have equal access to these phenomena, and the speaker is surely in a privileged position to know whether he is being ironical or not. In no sense does either party see 'nothing but the signs'. At this point we might remind ourselves that in the Philosophical Investigations Wittgenstein had likened linguistic communication to the playing of a game: I shall also call the whole, consisting of language and the actions into which it is woven, the 'language-game'. (Wittgenstein 1958: 7)
We should also remind ourselves that in most games, for example chess (Wittgenstein's favourite analogy), the intentions of the participants are irrelevant to the playing of the game: all that matters, in the case of chess, is how you move your pieces. To bemoan at the end of a disastrous game that you never intended to get into such a mess would be, at best, a futile (and perhaps unsporting) gesture. It is significant that if you wish to adjust the position of a piece but do not intend to make a move in the game, then you must utter the words J'adoube: failure to do so means that, according to the rules, the player did intend to move the piece. The difference between chess and language is that, in chess, one cannot utter those words without meaning them; the sincerity of your utterance is irrelevant to the game; the same cannot be said of language. In other words, the comparison of language with games like chess makes it difficult for Wittgenstein to acknowledge that intentionality has a crucial role to play in linguistic communication; difficult, but not impossible: Make the following experiment: say 'It's cold here' and mean 'It's warm here'. Can you do it? And what are you doing as you do it? And is there only one way of doing it? (Wittgenstein 1958: 510; emphasis in original)
The implication, clearly enough, is that one can say 'It's cold here' and mean 'It's warm here' (Because Wittgenstein enquires as to what happens as you 'do it', and as to how many ways there are of 'doing it'), but we are given no clues as to how this semantic transformation is to be achieved. The earlier discussion of intentionality and the reiterated comparison with games would seem to have required Wittgenstein to argue that such a transformation could not be achieved, except by means of specific stipulation or genuine linguistic change. We are unsure as to whether Wittgenstein is finally admitting that irony is a possibility, or whether he is maintaining his earlier Tractatus position that irony is a form of language change. What is certain is that the reader is left with the unmistakable impression that 'the first philosopher of the age' was hedging his bets on the question of irony.
166
Steve Farrow
The account of meaning and understanding to be found within Wittgenstein's Tractatus and Saussure's Cours has shaped the form of most subsequent theorizing in philosophy and linguistics: language is a hidden calculus and the task of the philosopher or linguist is to discover the enormously complicated 'tacit conventions on which the understanding of everyday language depends' (Wittgenstein 1961: 4.002). As we have seen, however, this model of language and communication is unable to give a satisfactory account of the phenomenon of irony. In recent years some philosophers have claimed that Wittgenstein's later work, in particular the Philosophical Investigations, heralds a radical departure from previous conceptions of language (Baker and Hacker 1980, 1984, 1985; Hacker 1988): speaking a language is akin to playing a game of chess, in which all players consciously know the rules, and know that their partners know the rules. In the sphere of language, nothing is hidden. But once again irony presents a difficulty: the intentions of the speaker are relevant to the concept of irony, but those intentions may on occasion be hidden. When playing chess and most other games, the concept of intention has little or no role to play: the unfortunate footballer who kicked the ball into his own net did not intend to score such a goal, but that lack of intention will make no difference to the final score. In short, games are a form of human activity to which only certain aspects of behaviour are relevant; communication, on the other hand, is an activity in which all and any aspects of behaviour and circumstance may be relevant. Irony is a phenomenon that threatens to demolish the shaky conceptual edifices that constitute contemporary theories of meaning. Particular individuals create meaning in specific contexts, and the ironical intention of the speaker or writer is not something to be judged after a meaning has already been calculated: the intention of the speaker or writer is inseparable from what she or he meant, and to see that is to understand why linguistic communication could never be a mechanical process of calculation, regardless of whether we are discussing a private calculus of hidden rules, or a public set of rules known to all speakers of a language.
References Baker, G.P. and P.M.S. Hacker (1980) Wittgenstein: Meaning and Understanding. Blackwell, Oxford. (1984) Language, Sense and Nonsense. Blackwell, Oxford. (1985) Wittgenstein, Rules, Grammar and Necessity. Blackwell, Oxford. Black, M. (1964) A Companion to Wittgenstein's Tractatus. Cambridge University Press, Cambridge. Dummett, M. (1975) What is a theory of meaning? In S. Guttenplan (ed.) Mind and Language, Oxford University Press, Oxford. (1976) What is a theory of meaning? (II). In G. Evans and J. McDowell (eds.) Truth and Meaning, Oxford University Press, Oxford. Grice, H.P. (1975) Logic and conversation. In P. Cole and J.L. Morgan (eds.) Syntax and Semantics 3: Speech Acts, Academic Press, New York. Hacker, P.M.S. (1988) Language, rules and pseudo-rules. Language & Communication 8 (2), 15972. Katz, J.J. (1981) Literal meaning and logical theory. Journal of Philosophy 78, 203-33.
Irony and Theories of Meaning
167
Saussure, F. de (1983) Course in General Linguistics, transl. R. Harris. Duckworth, London. Sedgewick, G.G. (1948) Of Irony, Especially in Drama, 2nd ed. University of Toronto Press, Toronto. Stenius, E. (1960) Wittgenstein's Tractatus. Blackwell, Oxford. Wittgenstein, L. (1958) Philosophical Investigations, transl. by G.E.M. Anscombe, 2nd ed. Blackwell, Oxford. (1961) Tractatus Logico-Philosophicus, transl. by B. McGuiness and D. Pears. Routledge and Kegan Paul, London. (1979) Notebooks 1914-1916, ed. by G.H. Von Wright and G.E.M. Anscombe, transl. by G.E.M. Anscombe. Blackwell, Oxford.
This page intentionally left blank
PART FOUR LANGUAGE AND DISCOURSE These papers set out the case for questioning the orthodox assumption that linguistic discourse is simply the practical exercise of linguistic knowledge (where linguistic knowledge is equated in the first instance with mastery of the grammar and vocabulary of one's native language). The integration1st rejects the notion that discourse is merely 'language use', i.e. involves no more than an individual's selection and implementation of units and relations already provided by 'the language' (just as, for example, playing chess involves the individual players in choosing and carrying out moves already defined by the rules of 'the game'). Discourse, for the integrationist, cannot be analysed independently of the communication situation in which it takes place, and this is where the analogy between language and games like chess breaks down: there are no discourse 'rules' which cover all situations. Stereotyped discourse does indeed occur, but in stereotypical situations. Most discourse is not of this kind and, more importantly, the potential of discourse is not limited in this way. Ch. 12 ('Conversational Utterances and Sentences') challenges the notion prevalent in discourse studies that there is a perspicuous relationship of some kind between the organization of conversation and linguistic structure, i.e that what is said in conversation is related by rulegoverned correspondence to the principal structural units recognized in descriptions of 'the language' used. This assumption might be plausible if, for instance, every single utterance in conversation corresponded to what grammarians call a 'sentence'. But this is far from being the case. Even worse, not only do we fail to converse in complete sentences or parts of sentences, but the utterances we produce fail to obey the grammatical 'rules' that they apparently should obey. So there is a mismatch of a very blatant kind between the structure of conversation and (what orthodox linguists claim to be) the structure of the language.
169
170
PART FOUR: LANGUAGE AND DISCOURSE
Ch. 13 ('Do You Understand? Criteria of Understanding in Verbal Interaction') takes this further and points out that the mismatch extends to matters of understanding as well. It is in practice impossible to 'read off' the meaning of an utterance from its apparent linguistic structure. Nevertheless, in discourse it is commonly taken for granted that participants can often tell whether their utterances have been correctly interpreted or not. It would be difficult to explain how they do this unless there were more to discourse than is revealed by an orthodox linguistic analysis of the verbal message they receive. This is confirmed by the reasons speakers give for believing that what they said has been understood. The ways in which they justify such beliefs is heavily dependent both on the circumstances obtaining when the original utterances were produced and on the circumstances in which the justifications are called for. Ch. 14 ('Analysing Fictional Dialogue') turns attention to the ways in which discourse is perceived and presented by those whose profession it is to 'reproduce' it artificially, e.g. novelists, playwrights, and similar artists in creative fiction. What can be inferred from studying fictional dialogue is that the same principles and patterns are at work as in 'natural' conversation and, specifically, that it is misleading to think of the creativity involved as resulting from the application of obligatory and specifiable rules. Furthermore, it is a mistake to think of conversation as being necessarily goal-directed at all, i.e. as even analogous to the proceedings of a committee, where it is often possible to sum up at the end what has been agreed, or where there has been failure to agree, and on what. Thus it would be fruitless to attempt to typologize 'conversational acts' on the assumption that it is possible in principle to correlate utterances with speakers' purposes and achievements in any systematic and contextneutral way.
12
Conversational Utterances and Sentences TALBOT J.TAYLOR
Introduction Any grammarian who has analysed a lengthy corpus of transcribed conversational speech will know that before such an analysis may be done, the data must be 'idealized'. That is, before the grammatical analysis may begin, the transcribed speech must be modified so that the utterances take on the appearance of sentences similar to those typically encountered in written discourse. For instance, should the analyst come upon a typical stretch of speech like the following, he/she would not normally begin the analysis before making certain alterations to the text. (1) It's just a subject which is - . sort of. basically well-known to all mathematicians and is a tool which . one uses whenever required and is - at the [ha:] . at the is the logical foundation of quite a lot of . quite a lot of things (S.I.6, 849)1
To idealize this transcribed speech, the analyst will delete those parts of the transcription which keep the utterance from looking like a standard written sentence. For instance, the passage and is - at the [ha:]. at the is the logical foundation would never appear in a written text. So, the analyst would probably idealize this passage to the form 'and is the logical foundation'. Given the traditional support within linguistics for Saussure's principle of 'the primacy of speech', it is revealing to ask why the analyst would decide to delete part of his spoken data so that it might conform to written language style. A contributory reason is that, according to conventional standards, the unidealized passage cited above is simply not grammatical. The sequence and is - at the [ha:] . at the is the logical foundation could not form part of what any grammarian would call a standard grammatical sentence of English. This is true regardless of the theoretical 'school' to which the grammarian belongs. Indeed, one part of the sequence, [ha:], could not be taken as a token of any word-form of contemporary English. Thus, to leave such a 1 All references in this form refer to specific tone units in Svartvik and Quirk (1979). The examples quoted have, to a certain extent, been simplified. In some respects this was unavoidable, given the specially designed typography used by Svartvik and Quirk. My intention was also to make the examples as simple and readable as possible. All of the examples used are selected from speakers who did not know their conversations were being recorded. Further description of the conversations, their participants, and the recording and transcriptional techniques may be found in Svartvik and Quirk (1979).
171
172
Talbot J. Taylor
sequence unaltered in an utterance would render that utterance grammatically unanalysable. There is no need to go into the details of the various methods of grammatical analysis in order to defend this claim. Since the utterance both contains a form which does not approximate to any English word-form and incorporates sequences which could not form a part of any English sentence, it is clear that any grammatical analysis which takes the sentence and the word as its basic elements would not find the above sequence acceptable. One need only point to the odd combinations such as article plus verb (the is) and article plus non-word (the [ha:]) to confirm such a conclusion. Grammatical analyses are not designed to parse verbal sequences which are not sentences of English. In his introduction to linguistics, Charles Hockett argues for such editing as a prerequisite to grammatical analysis (Hockett 1958: 141). He produces a transcribed sample of real conversational speech. This sample, similar to the passage from the Lund Corpus given above, includes a typical amount of both filled and silent pauses, repeats, false starts and other features characterized by Hockett's expression 'hemming and hawing'. Following this sample, he offers an edited version of the same transcription. The edited version, from which the features of 'hemming and hawing' are deleted, is said to convey 'much the same meaning' as the unedited version. Indeed, Hockett argues, 'this edited version is implicit within the original'. It is such edited versions which Hockett claims to be the phenomena deserving of the linguist's attentions. Recent research suggests that much can be learned about a person through a close examination of his unedited speech. The particular ways in which he hems and haws, varies the register of his voice, changes his tone quality, and so on, are revealing both of his basic personality and of his momentary emotional orientation. But since (if our assumption is correct) phenomena of these sorts are not manifestations of the speaker's linguistic habits, it is proper to ignore them in the study of language, basing that study exclusively on edited speech. (Hockett 1958: 143-4)
Nevertheless, it should not be surprising that certain constraints are imposed on the grammarian's process of idealization. For instance, it is not within the grammarian's brief to render all ungrammatical utterances into grammatical sentence form. Furthermore, he must also edit sequences which, if taken without alteration, could possibly pass as grammatical strings. To take the former point first, it may be seen that ungrammatical utterances such as (2) would not merit idealization in the same way as (1). (2) There be two girls at the door, (invented)
Most grammarians will leave utterances like (2) intact in their analyses and will simply register the fact that, in the dialect being studied, (2) is standardly judged to be ungrammatical. Presumably, if similar usage of the copula was in fact found to be common, the grammairian might well question the standard claim of its ungrammaticalness. But s/he would never, for instance, edit it to the form (2a). (2a) There are two girls at the door.
Conversational Utterances and Sentences
173
To do so would be deemed a falsification of the data, a claim not ordinarily felt to be applicable to the type of editing Hockett recommends. So we may conclude that it is not merely the fact that (1) is ungrammatical that leads the analyst to edit it to 'ideal' form. Ungrammaticalness is not a sufficient condition to cause the grammarian to edit the form of conversational speech. If we consider examples (3) and (4), we may see that ungrammaticalness does not provide a necessary condition either. (3) I don't think. I mean he's very organized. (S.I.8, 914) (4) I'd paint for I'd paint forever if I had the time. (S.I .8, 853)
In all probability, examples (3) and (4) would be idealized to (3a) and (4a) by the grammarian. (3a) I mean he's very organized. (4a) I'd paint forever if I had the time.
However, as they stand, (3) and (4) are not ungrammatical. If (3) was not altered, it could be analysed as the two sentences (3b) and (3c). (3b) I don't think. (3c) I mean he's very organized.
Similarly, the utterance transcribed as (4) could be analysed as the two grammatical sentences (4b) and (4c). (4b) I'd paint for. (4c) I'd paint forever if I had the time.
Thus it appears that there are criteria other than ungrammaticalness which may lead the analyst to alter a transcribed utterance into an idealized sentence-form amenable to grammatical analysis. The precise criteria which do, in fact, guide the grammarian's editing of 'raw' conversational data is a topic of obvious interest for the chronicler of practical scientific method, but it has little relevance to our present concerns. What does concern us is the use to which this picture of the practical grammarian's task has been put in the conceptualization of the interface between linguistic competence and linguistic performance.
The target sentence hypothesis In his article 'The Production of Speech' John Laver (1970) argues that 'hemming and hawing', such as that in example (1), is not noticed by the hearer in perceiving speech. Much like the analyst who deletes parts of a transcription in order to idealize it and thus make it amenable to grammatical analysis, the hearer is thought by Laver to 'edit' what he hears.
174
Talbot J. Taylor
The conscious perception of speech in some sense regularizes and idealizes the actual data of speech. (Laver 1970: 73)
A similar argument is proposed by Bernd Voss in a paper on hesitation phenomena and perceptual errors. He suggests that the hearer's idealization process is similar to the one operative in reducing speech to writing. (Voss 1979: 130)
The hearer is assumed to have 'in store' the units of his language and the possible sentential sequences of such units. So, when he hears an utterance, he attempts to take it as a token instance of one of the stored sentences. During the hearer's decoding of an utterance, then, he is thought to be forming hypotheses as to what sentence-type the 'incoming' utterance is a token of. If, having formed one such hypothesis, he can then take the remainder of the utterance as conforming to the predicted sentence-type, he will do so. In this case, 'performance difficulties' of the kind - the logical foundation of quite a lot of. quite a lot of things - would be unconsciously edited by the hearer in order that he might take the whole utterance (1) as an instance of the grammatical sentence (Ib). (Ib) It's just a subject which is basically well-known to all mathematicians and is a tool which one uses whenever required and is the logical foundation of quite a lot of things.
That is, the hearer is thought to edit incoming utterances in much the same way as the grammarian who edits speech in reducing it to analysable, written form. By doing so, both hearer and analyst are able to assign a structural description to the competence sentence assumed to underlie the performance utterance. A similar argument might be made regarding ellipsis. That is, in the same way that utterances including 'hemming and hawing', repetitions, non-fluence, etc. (all of which we will be referring to here as 'discontinuity'), might be thought to require the hearer to delete superfluous elements of the utterance in order to reduce it to grammatical sentence form, utterances involving the ellipsis of verbs, subjects, or other essential elements might be thought to oblige the hearer to add the 'missing' elements to the utterance in order to make it a fully grammatical, and thus comprehensible, sentence. So, (5a) and (5c) would require their hearers to add various words in order to derive the underlying sentences (5b) and (5d): (5a) In the car. (5b) I left the keys to the house in the car. (5c) Before. (5d) Tom met Mary before he decided to devote his life to music.
Thus, while discontinuity can be taken to require editing in order to expunge superfluous elements from the utterance, ellipsis can be taken to require a complementary editing procedure involving the addition of missing elements. Both types of editing, under this view, would be essential to the hearer's need to derive the sentences underlying the speaker's utterances.
Conversational Utterances and Sentences
175
If we adopt this assumption (i.e. what will here be called the 'target-sentence hypothesis': that, like grammatical analysts, hearers have to edit incoming speech in order to analyse it in terms of its sentential grammatical structure), then it would appear that knowledge of how to edit utterances into sentences is an essential part of a competent speaker/hearer's conversational competence. That is, most hearers can be assumed to know that, in the appropriate contexts, a speaker's utterance of 'tomorrow' is to be edited to Tomorrow is my birthday' and that 'I saw Susan at Mary at the debate' is to be edited to 'I saw Mary at the debate'. If they lacked this knowledge, it might be presumed, they would find much of ordinary conversational speech unintelligible. Furthermore, if we assume that hearers implicitly edit utterance tokens into sentence form, then we must also assume that, if all hearers of the same utterance edit it to the same sentence form, there are shared rules or procedures which guide their editing. Consequently, it becomes of interest to the study of conversation to determine what these rules or procedures are. In this context the research on the perceptual errors made by non-native speakers of English takes on a relative importance. In the aforementioned article, 'Hesitation phenomena and perceptual errors', Voss claims that his experiments show that nonnative speaker/hearers of English have difficulties with the task of editing performance utterances into 'target' competence sentences. Sometimes they do not realize that editing should be done; sometimes they edit improperly; and sometimes they edit where no editing is required. Referring to the task of 'reconstructing' incoming utterances into appropriate sentence form, Voss claims: This, however, is a task which although usually no problem for the native speaker is typically more difficult for the non-native speaker. (. . .) The non-native speaker will find it difficult to know in each case whether a given stretch of acoustic information is part of the speaker's performance that can be disregarded, or whether, if the reconstruction is to be correct, it needs to be accounted for. (Voss 1979: 130)
In an experiment performed in 1975, Voss asked 22 German students to transcribe a passage of spontaneous English speech. All the students were training to be teachers of English and had undergone five to seven years of instruction in English. The experiment was based on the aforementioned assumption that speech perception consists of a matching between a predicted competence sentence and the incoming acoustic information. The students were asked to transcribe the passage as accurately as possible - in the sense of providing an orthographic rendering. (Voss 1979: 131)
It was assumed that the errors in the transcript - identified by a comparison of the students' transcriptions with a transcription made and checked by a native English speaker - should reveal the general causes of perceptual difficulties for the non-native speakers. In addition, No instructions were given how to handle hesitations. The assumption was that they would be idealized out. This turned out on the whole to be true . . . (Voss 1979: 132)
Voss found that 32.6% of the errors made in the non-native transcriptions were
176
Talbot J. Taylor
likely to have originated in a misinterpretation of hesitation features. (Voss 1979: 136)
He groups the hesitation-linked errors into two types. The first type consists of cases where the subject has taken a repeat or a filled pause to be an instance of a word or a part of a word. For instance, the utterance and you can - ring off a a great string was interpreted by one subject as 'and you can ring offer a great string'. What happened here appears to be that the hearer took the utterance of off followed by the repeated indefinite article a a as an instance not of a repeat but of a non-hesitation-interrupted sequence consisting of the two words 'offer' and 'a'. That is, the first a was taken as a second syllable attached to off to make the word 'offer'. Similarly, a great string of erm of activities was mistaken by one subject for 'a great string of an activities'. Here a filled pause is taken for the article 'an'. The second type of error attributable to unfamiliarity with hesitation phenomena in English consists of those errors when the subjects mistook a word or part of a word for an instance of a repeat or a filled pause. Two cases are cited. In the first, the subject took the utterance one of the problems - erm - of people who work . . . for 'one of the problems people who work ...'. Here, the preposition of appears to have been mistaken for a filled pause and so was idealized out of the subject's transcript. In the second instance, the utterance . . . such as - boat building was transcribed as 'such as building'. Voss suggests that the subject here took building to be a correction of boat since both begin with the same consonant. Thus the subject is supposed to have interpreted the utterance in the same way that one might (correctly) interpret the utterance / work at the top of a large boat building on the 37th floor as 'I work at the top of a large building on the 37th floor'. In his own terms, what Voss's data seems to indicate is that the non-native hearers often did not know the signals given by the English speakers regarding when a section of the speaker's utterance needed to be edited or how it needed to be edited in order that it might match the appropriate idealized competence sentence. As Voss sees it, sometimes they edited where no editing was required, and sometimes they did not recognize cases where editing was called for. And yet native speakers do not appear to encounter these difficulties. Consequently, if provisionally we accept Voss's reasoning, it seems possible to conclude that there are procedures which guide the hearer in establishing which competence sentence (or 'target sentence') a discontinuous utterance is to be matched with. The non-natives in Voss's test appeared to lack the linguistic knowledge required for the operation of these procedures. Furthermore, because lack of this knowledge seems to have kept the non-native hearers from performing as competently as a native hearer of English, this knowledge can, in some sense at least, be rightfully termed part of the 'knowledge of English'. In this case it is not surprising that in recent years more attention has been drawn to the task of determining an explicit formulation of editing rules. Much of the recent interest in editing stems from a paper read by William Labov to the 1966 meeting of the Linguistic Society of America (Labov 1966). In this as yet unpublished paper Labov argues that verbal performance is not nearly so ungrammatical as might be concluded from a superficial appraisal of conversational speech. By the application of a simple set of editing rules, it can be seen that less than 2% of conversational
Conversational Utterances and Sentences
177
utterances are truly ungrammatical. Much of Labov's paper consists in the formulation of a possible set of such editing rules.2 A second source of recent interest in editing rules comes from the growing amount of money and research-energy invested in the projects to develop computer programs for the processing of natural speech. If it will ever be possible for a computer to parse and interpret spontaneous speech, this achievement will require a program which incorporates editing rules. A recent paper by Donald Kindle (1983), an ex-student of Labov, proposes a set of editing rules to be incorporated into the computer parsing program 'Fidditch' developed by Marcus (1980). Of greatest importance to our purposes is the fact that Kindle (1983) adopts the now familiar assumption that any set of rules which can be shown to help a computer process speech must be similar to those used by the competent human language-user. Thus Kindle's work on editing rules is intended as a contribution to the study of the editing rules human hearers use in transforming spontaneous yet flawed verbal performance into grammatical target sentences. Why is the notion of editing rules, and the related target sentence hypothesis, so crucial to linguistics and to conversation analysis? The answer is editing rules provide the necessary apparatus for linking a model of linguistic competence with the evidence of conversational performance. Without such a link it might be more easily argued that a model of competence, conceived as knowledge of the rules of a sentence-grammar, is irrelevant to an account of the everyday task of understanding conversational behaviour. For linguistics standardly assumes that communication consists in the transfer of meaning from speaker to hearer and that this transfer is possible because the speaker and hearer share knowledge of the same abstract system fixing the relations between meanings and sentences, i.e. the grammar of the language. Thus by producing an utterance which instantiates a particular sentence of the language, I can be sure that, providing my hearer shares my knowledge of the grammar, he will understand my intended meaning. In such a picture of communication, much depends on the hearer's ability to establish the connection between the utterance I produce in conversation and the sentence of the language which encodes the meaning I intend to communicate. If my hearer cannot determine which sentence my utterance is supposed to instantiate, then my effort to communicate will be in vain. For communication to succeed, the hearer must be able to identify the sentence-type of which my utterance is a token. Now, since communication only rarely breaks down, it seems reasonable to suppose that hearers generally have no trouble identifying the sentence-types of the utterances they hear. And, since hearers do not appear to have any more trouble understanding elliptical and non-fluent utterances, then it seems plausible to suppose that they have methods for deriving the appropriate sentence-types from the elliptical
2 Labov's paper may have been provoked by Chomsky's assumption (e.g. Chomsky 1965) that one reason children could not learn a language by imitating the speech of their parents is that so much of adult verbal performance is characterized by ungrammaticality, self-correction, attention shifts, slips of the tongue and other such 'performance errors'. But if this is the spur to Labov's editing hypothesis, then it is perhaps understandable that the paper was never published; for it would have implied that, although children may not require the sort of innate grammatical knowledge that Chomsky assumes, nevertheless they must be assumed to have innate knowledge of something like editing rules. For how else are children to learn the editing rules required to correct the (superficially) ungrammatical utterances of their parents?
178
Talbot J. Taylor
and non-fluent utterances they hear. In this case, the recent interest in editing and ellipsis rules is unsurprising. However, if it should turn out that no coherent sense can be made of any hypothetical formulation of editing rules, then the manifest regularity of communicational success becomes evidence against the assumption that, in order to understand an utterance, a hearer must identify the sentence-type of which it is a token. And yet without this assumption it becomes unclear how a sentence-grammar model of competence can shed any light on the ability of humans to participate in conversation interaction.
Aims Given the importance to linguistics of the target-sentence hypothesis and the reliance of that hypothesis on the notion of editing rules, it would seem essential to conversational study to discover a coherent formulation of rules for editing. The main aim of this paper is to forestall any further research on such hypothetical formulations. It would be a waste of the researcher's valuable time. I will argue that no coherent formulation of editing rules could account for a wide range of utterances occurring in conversational speech, utterances which do not, however, appear to pose any problems of understanding for their hearers. To reach these conclusions, I will attempt to show that, in order to fulfil their function of deriving appropriate target sentences, editing rules must obey a variety of general constraints on their form. Secondly, I will argue that any proposed formulation of editing rules which does obey these necessary constraints will not be able to account for the evident interpretability of various types of typical conversational utterances. This double-bind will then lead to the conclusion that, since the target-sentence hypothesis relies on the notion of editing rules and yet that notion seems to be practically incoherent, the target-sentence hypothesis should be rejected. In the following discussion I will focus my analysis on the formulation of editing rules for utterances with 'discontinuity' ('hemming and hawing', false starts, selfcorrections, etc.). I will later argue that the obstacles to that task are basically the same as those facing attempts to formulate editing rules for ellipsis, although certain features of ellipsis exaggerate the difficulties.
Editing rules and their constraints We may begin by identifying two phases in editing. We may imagine that the hearer is somehow able to determine both when an utterance should be edited as well as how it should be edited. Earlier studies of the editing of discontinuous utterances have assumed that speakers produce phonetic editing signals when their utterances require editing and that competent hearers have no trouble recognizing these signals. For instance, Labov (1966) assumes the speaker's editing signal takes the form of 'a markedly abrupt cut-off of the speech signal (which) always marks the site where self-correction takes place'. In order that we may turn our attention to the more complicated problem of formulating rules to guide the hearer in editing, I will ignore
Conversational Utterances and Sentences
179
for the moment the issue of the editing signal. Let us assume for present purposes that speakers do produce editing signals when editing is required and that these signals consist in a discontinuity in the otherwise continuous production in the speech flow (i.e. 'a markedly abrupt cut-off of the speech signal'). This will then permit us to distinguish three components in the source utterance: 3 the discontinuity serving as editing signal, the speech prior to the discontinuity (hereafter the 'pre-discontinuity'), and the speech subsequent to the discontinuity (hereafter the 'post-discontinuity'). It will be convenient in beginning this discussion to focus on one typical example of a discontinuous utterance in need of editing. In this way hypotheses may be easily formed and, if necessary, easily rejected before testing them against a wider range of data. Imagine a speaker producing the following discontinuous utterance. (6) I've watched him for a long - longer than you have.
We might say that it is intuitively apparent that the target sentence for (6) is (6a). (6a) I've watched him for longer than you have. What sort of editing rule might we envisage for transforming (6) into (6a)? A first suggestion might be that the hearer simply compares the source utterance with a mental list of all the possible competence sentences of English. Indeed, sentence (6a) would be among the latter. However, how is the hearer to know which of many possible competence sentences is, in this instance, the appropriate target? There are other possible selections, such as those listed under (6b). (6b) I've watched longer than you have. I've watched him longer than you. I've longer than you have.
If we assume that hearers do edit discontinuous utterances, then we must also assume that they have a method by which they can tell which grammatical sentence is the appropriate target. Merely comparing the discontinuous utterance with a list of possible target sentences does not solve that problem. Furthermore, such an ad hoc method would permit different hearers of the same utterance to arrive at formally, and therefore semantically, contrastive target sentences. Given that discontinuous utterances do not often appear to result in communicational breakdown, then a method which leads different hearers to contrastive interpretations of the same utterance seems worthy of outright rejection. Editing rules must unambiguously pick out the appropriate target sentence. It might here be argued that a hearer could easily select the appropriate target sentence from those possibilities listed under (6a) and (6b). This could be accomplished easily because the hearer would already have formed an idea of what the speaker intended to say. There is a simple reply to this proposal: if the hearer can, in this way, determine directly from the speaker's discontinuous utterance and its context what was meant, then there is no reason for the hearer to perform the now superfluous task of editing that (already interpreted) utterance to target sentence form. 3 I have chosen the terminological pair 'source utterance/target sentence' to suggest the parallel in translation studies to 'source language/target language'.
180
Talbot J. Taylor
If he has already understood, then any subsequent transformation of the form of the utterance would be pointless. The target sentence hypothesis is based on the assumption that a hearer needs to discover, by means of editing, the competence sentence underlying a discontinuous utterance and that the reason he needs to do this is so that he may then apply his knowledge of the grammar of the language to derive the meaning from the competence sentence. There would be no need of editing nor of sentence grammars if the hearer could determine the speaker's meaning directly from the discontinuous utterance itself. So far, then, we have come upon two principles governing any formulation of editing rules. They must guide the hearer unambiguously to the target sentence, and they must not presuppose the hearer's prior understanding of the utterance's meaning. I will refer to these principles respectively as the principle of systematicity and the form-before-meaning principle. As will later be shown, certain formulational constraints follow from these two essential principles of editing. It is evident that in order to derive the target sentences (6a), an editing process is required which deletes certain units from the pre-discontinuity (here marked with an X): (6c) I've watched him forXtfeftg • • • longer than you have.
We may imagine the hearer as deleting units from the pre-discontinuity in order to be able to take the post-discontinuity as a continuation from the pre-discontinuity: i.e. to take longer than you have as a continuation from I've watched him for. The relevant question is: which constraints guide the hearer in his choice of elements to delete? Why should he not delete as in (6d)? (6d) I've watched ^i fo^l^Jg - longer than you have
The result is a competence sentence ('I've watched for longer than you have') but is, nonetheless, not appropriate. In (6d) the hearer deletes too much of the prediscontinuity. The deletion of him is in excess. He also deletes unsystematically. In deriving (6b) from (6), it would seem that a rule which might guide the hearer is as follows: delete 'from right to left' in the pre-discontinuity until the postdiscontinuity may be taken as a grammatical continuation from the pre-discontinuity. The example (6c) conforms to this rule. The command 'from right to left' ensures a systematic order in the editing while the proviso, 'until the post-discontinuity may be taken as a continuation from the pre-discontinuity', is needed to ensure that the deletion be the minimum necessary to arrive at a continuous grammatical sentence as target. That is, to give one hypothetical description of the editing process, we may imagine that elements of the pre-discontinuity are deleted - starting with the last word and working back one by one to the beginning of the source utterance - until the remainder of the pre-discontinuity followed by the post-discontinuity approximates a satisfactory target sentence. Analogously, the source utterance can be thought of as a section of magnetic tape,
Conversational Utterances and Sentences pre-discontinuity
181
post-discontinuity
We may then imagine that the post-discontinuity is pushed back over the end of discontinuity until a grammatical sentence, formed by the splicing of the two parts of the tape, is arrived at. b
I've
watched
c him
d for
e a
f
^
l
2
3 4 5
5 .£ >
long^ c > longer than you have
So, first point (1) in the post-discontinuity is joined with (f) of the pre-discontinuity. But the resultant string - 'I've watched him for a long longer than you have' - is obviously not a grammatical sequence. Consequently the post-discontinuity must be pushed back further into the pre-discontinuity. If point (1) is made then to coincide with (e) in the pre-discontinuity, the resulting string - 'I've watched him for a longer than you have' - is still ungrammatical. Not until point (1) (note: always the beginning of the post-discontinuity) is joined to (d) does the grammatical sentence (6a) finally appear. The process may similarly be thought of in terms of deletion. That is, first the last word in the pre-discontinuity string is deleted, but the resultant combination of the remainder of the pre- and post-discontinuity is ungrammatical. Only when the last two words of the pre-discontinuity are deleted - i.e. (d)-(f) - do we arrive at the grammatical sentence formed by ((|>)-(d) followed by (l)-(5). It is important to note that the process deletes words of the pre-discontinuity in reverse order of their production and then stops as soon as a grammatical sequence appears. For instance, if this were not the case and (b) to (f) were deleted first, a grammatical sentence (6e) would appear. (6e) I've watched longer than you have.
Similarly, deletion of (c) through (f) would produce a grammatical sentence: (6f) I've watched him longer than you have. Although there is hardly if any difference in meaning between (6f) and (6a) the principle that deletion proceeds one-by-one in reverse order of production (what will be called the 'mimmalness constraint') must be retained so that grammatical but unsatisfactory sequences like (6e) are not derived by hearers. At this point it seems appropriate to formulate the following hypothetical editing rule. Minimal deletion rule If a break in the continuity of speech occurs, (1) take the post-discontinuity as a continuation of the pre-discontinuity utterance, but
182
Talbot J. Taylor
(2) If (1) is inapplicable due to its creation of an ungrammatical sequence, then delete only as much of the end of the pre-discontinuity as is needed to make (1) applicable, insuring that the order of deletion is the reverse order of the production of the pre-discontinuity. This minimal deletion rule is meant to apply equally to editing involving pauses, repetitions, insertions, corrections and other types of 'non-fluency'. An examination of the minimal deletion rule reveals that it incorporates a certain number of constraints. (a) The minimalness constraint. This constraint ensures that only as much is deleted of the pre-discontinuity as is required to allow the post-discontinuity to be taken as a continuation of the (remainder of the) pre-discontinuity. (There may, of course, be no remainder.) (b) The continuity constraint. When the edit is complete, the post-discontinuity must be taken as a continuation from the (remainder of the) pre-discontinuity. (c) The deletion constraint. The minimal deletion rule implies that any element of the source utterance which is deleted in the application of the rule does not form a part of the target sentence. (d) The grammaticality constraint. This constraint ensures that editing is not complete until the (remainder of the) pre-discontinuity followed by the postdiscontinuity forms a grammatical sequence, as defined by the rules of the language. (e) The identification constraint. This constraint ensures that the minimal deletion rule is applied if and only if an editing signal, in the form of a discontinuity in the speech flow, occurs in the speaker's utterance. The minimal deletion rule further specifies that all deletion proceed 'from right to left' in the pre-discontinuity. I will not insist on this constraint, as other editing hypotheses (e.g. Kindle 1983) may have a different yet equally efficient way of constraining the order of deletion in the edit. These five constraints, required by the more general principles of systematicity and form-before-meaning, form the obligatory basis of any coherent editing hypothesis. In order to support this claim, the following discussion will consider the need for each individual constraint. Without the grammaticality constraint, there would be no need to edit any utterance. Speakers frequently produce discontinuities in their utterances where no editing of the pre-discontinuity is in fact required. We need only to think of the regular occurrence of hesitation pauses (cf. Maclay and Osgood 1959), as in the examples below, to see that this is so. (7) You're on this . [a] senate committee of course aren't you. (S.I.26, 1215) (8) I don't know whether you . noticed. (S. 1.26 1229)
Because these utterances are grammatical, though discontinuous, no editing of the pre-discontinuity is required. On the other hand, were there no grammaticality constraint, hearers would never be led to delete any elements of pre-discontinuity sequences. That is, hearers, must see that there is something 'wrong' with the syntactic form of an utterance, in addition to the occurrence of a discontinuity, for them to perceive the need for editing.
Conversational Utterances and Sentences
183
The continuity constraint is required to ensure that the hearer takes what comes after the editing signal (until a sentence boundary) as a bonafide part of the target sentence. Otherwise the hearer might also delete the post-discontinuity, as in the following example. ., (discontinuity) (9) I've only got fiveXKl 5>£ht y>&rs>Cgp ar^vay. (S. 1.1, 315)
In addition, the continuity constraint ensures that the hearer knows how to put the remaining elements of the utterance 'back together again' after the process of deletion. Finally, the continuity constraint is itself presupposed by the minimalness and grammaticality constraints. The identification constraint, as we have already seen, is required so that the hearer knows when (and when not) to edit. The hearer is not supposed to edit simply whenever s/he feels like it, regardless of the intentions of the speaker. Instead s/he must be constrained only to edit when the speaker produces an appropriate editing signal. Without such a constraint, the notion of a systematic editing procedure would be impracticable. The minimalness and deletion constraints have been left until last because not all who have worked on editing agree on the need to constrain the hearer in these ways. For instance the editing rules proposed by Donald Hindle incorporate neither of these constraints. The editing rules Hindle (1983) formulates amount to the following instructions to the hearer. 1. If an editing signal occurs and the surface strings on either side of the signal are identical, delete the first (as in example 10). 2. If (1) is not the case, but the two words on either side of the signal are instances of the same type of sentential constituent, delete the first (as in example 11). 3. If neither (1) nor (2) apply, go back into the pre-discontinuity until you find a constituent of the same type as the first constituent in the post-discontinuity. Delete the former (as in example 12). (10) Well if they'd - if they'd had a knife I wou - wouldn't be here today. (Hindle 1983, ex. 7a) (11)1 was just that - the kind of guy that didn't have - like to have people worrying. (Hindle 1983, ex. 8) (12) I think that you get - it's more strict in Catholic schools. (Hindle 1983, ex. 9)
Perhaps the main problem with Hindle's proposed rules is that they do not combine the constraints of minimalness and grammaticality. Consequently, nothing tells the hearer (or the computer) which has identified an editing signal not to edit only as much as is required to obtain a grammatical target sentence. The point of these combined constraints is to keep the hearer from editing where no editing is required, in spite of the presence of an editing signal. For instance, Hindle's rules (but not the minimal deletion rule) would inappropriately oblige the hearer to edit the following (a) examples to the forms listed under (b), in each case because of the misleading presence of an editing signal.
184
Talbot J. Taylor (a) I suspected always that Delaney would be late . that Chomley would be on time and that this would produce a nice staggering of their arrival on your desk. (S.I.I, 124) (b) I suspected always that Chomley would be on time and that this would . . . (a) This is to run their coordinating machinery - you know to establish the standards and all that kind of thing. (S.I.I, 553) (b) This is to establish the standards and all that kind of thing. (a) And he was boasting about all this stuff they'd been using of Lawrence and . George Eliot. (S.I.I, 837) (b) And he was boasting about all this stuff they'd been using of George Eliot.
In addition, Kindle's rules would probably mis-edit the following example (13a) because it could not identify the constituenthood of the only partly produced form [bro:] and thus would produce the target sentence (13b). (13a) I've done mine and I. [bro:] thought I brought it down (S.I .4, 243). (13b) I thought I brought it down.
Now, in one sense Hindle can accept (13b) as the target sentence for (13a), because he explicitly rejects the operation of a deletion constraint. The status of the material that is corrected by self-correction and is expunged by the editing rules is somewhat odd. I use the term expunction to mean that it is removed from any further syntactic analysis. This does not mean however that (it) is unavailable for semantic processing. (Hindle 1983:3)
In other words, Hindle can accept (13b) as the target for (13a) because he can always insist that, although deleted by the editing rules, the string 've done mine and I is still available for semantic processing. Hindle cites the work of the ethnomethodologist Jefferson (1974) in support of the 'odd status' his rules assign to deleted material (Hindle 1983: 3). In this case it is interesting that another member of the ethnomethodologist school of conversation analysis, Charles Goodwin, adopts the same fence-sitting position on the deletion principle. For Goodwin (1979), words in an utterance which are deleted through editing play no role 'in a unit on another level of organization necessary for properly understanding that stream of speech' (Goodwin 1979: 98, fn. 3) (i.e. in the target sentence derivable from the source utterance). Nevertheless, much of Goodwin (1979) as well as Goodwin (1981) is concerned with revealing the interactional function of such deleted elements! The fence-sitting position adopted by Hindle and the ethnomethodologists is incoherent for the simple reason that a verbal string cannot be deleted from the syntax of an utterance while remaining in its semantics. The argument in support of editing rules is based on the assumption that, in order to understand an utterance, hearers need to determine the grammatical sentence-type underlying it. If hearers can understand the meaning of parts of the utterance which do not function in the underlying target sentence, then there is no reason for them to edit in the first place. For they seem able to understand without the intervening aid of target sentence form. The result is that Hindle is left in the quite mysterious position of holding both that a hearer can understand, e.g. a sequence of five or six words to which his internalized processing programme assigns no syntactic representation, and that hearers need to edit utterances to grammatical sentence form in order to understand them.
Conversational Utterances and Sentences
185
It thus seems more reasonable to conclude that, if an editing hypothesis is to explain how a hearer equipped with knowledge of the grammatical sentences of a language is able to understand discontinuous utterances which are, at least superficially, ungrammatical, then the deletion constraint must be incorporated into any precise formulation of that hypothesis. The same is true of the minimalness constraint, which ensures that the editor only deletes the minimum that is required to produce a grammatical target. Nevertheless, it is not at all inconceivable that Hindle's editing program could itself be edited so as to incorporate some version of the minimalness constraint. And his opposition to the deletion constraint is not essential to the formal coherence of his proposed program. There seems to be a strong case for incorporating each of the five listed constraints (as well as the two more general principles) into any proposed formulation of editing rules. Indeed, a proposal which failed to incorporate any one of the contraints would be incoherent or unworkable. For a wide range of the discontinuous utterances to which it was applied, such a proposal would either mis-edit the source utterance to an inappropriate target string, or, if it ignored the deletion or identification constraints, it would fail to serve the purpose for which editing rules are required, viz. to form an explanatory bridge between actual verbal performance and grammatical models of competence. In spite of this conclusion, we will now argue that, essential though they are to any editing hypothesis, the incorporation of the aforementioned constraints would prevent any editing hypothesis from being able to account for the hearer's manifest ability to understand a wide range of typical conversational utterances. In this case, the very constraints which any set of editing rules must obey, if it is to account for one domain of the explanandum, serve to prevent those rules from being able to account for another domain of the explanandum. We may begin by looking at the drawbacks with the continuity and deletion constraints. I have already argued for the need of these constraints if we are to conceive of editing rules as determining a systematic procedure for hearers to derive the underlying target sentences, and thus the meanings, of discontinuous utterances. However, the price of these requirements is that any set of editing rules which incorporate these constraints is not able to derive an appropriate target sentence (and hence a meaning) for the following sorts of examples (14) That's . Velasquez' Pope Innocent the fourth . a copy of. (S. 1.4, 434) (15) We could all slant atheistic slogans slant. chant, (overheard in conversation) (16) The interview was - it was all right. (5.1.3, 305)
The continuity constraint holds that, after all deletion has been carried out, the postdiscontinuity is to be taken as a continuation from whatever is left of the prediscontinuity. However, with examples such as (14) and (15) the appropriate target sentences, (14a) and (15a), could not be derived. (14a) That's a copy of Velasquez' Pope Innocent the fourth. (15a) We could all chant atheistic slogans.
186
Talbot J. Taylor
Instead, because of the continuity constraints the sequences Velasquez' Pope Innocent the fourth and atheistic slogans would be deleted during editing. And, because of the deletion constraint, the semantic content of those sequences would not be derivable from the resultant target sentences. In fact, the continuity principle would prevent any edit arriving at a grammatical target sentence for (14). Thus, contrary to intuition, (14) would be uninterpretable to the hearer. Example (15) would simply be edited to We could all chant, leaving out the information regarding what would be chanted; and yet, again, it seems intuitively obvious that no hearer would experience any difficulty in interpreting (15) appropriately. Example (16) poses a problem because, in order to arrive at a continuous, grammatical target sentence, the first three words of the utterance would be deleted in editing, resulting in (16a) as the target. (16a) It was all right.
The objection to (16a) is that it does not provide the subject pronoun with an antecedent (and one does not occur in the preceding context). Consequently, given the deletion constraint, the intuitively implausible conclusion is forced on us that the hearer could not determine to what the pronoun refers. In example (17) an edit under the constraint of the continuity principle would delete not only, producing the target sentence (17a). However, this target leaves out the first half of the not only . . . but also structure which does not seem lost to an intuitive interpretation of the utterance. Thus the deletion and continuity constraints seem here to produce a target sentence which is not an appropriate paraphrase of the source utterance. (17) I'm very suspicious of the press generally and I can tell you because - not only I mean that's one case that you've given but also in their reporting of erm affairs foreign affairs . . . (example taken from Crystal 1980: 158) (17a) I'm very suspicious of the press and I can tell you because that's one case that you've given but also in their reporting of foreign affairs.
(In fact, the source utterance (17) appears to resist any systematic edit into an appropriate target sentence. This does not, however, make it any less easy to understand.) Finally, in the following example the continuity constraint would ensure that the sequence because ofsessionals would not occur in the target sentence even though it is intuitively relevant to an interpretation of the source-utterance. (18) I had a seminar today in which people hadn't read the stuff because of sessionals hadn't read the play. (S.I.4, 1082)
The minimalness constraint holds that only as much is deleted of the prediscontinuity as is required to allow the post-discontinuity to be taken as a continuation of the pre-discontinuity. However, the minimalness constraint can often result in a target sentence which incorporates too much of the source utterance. Thus, this constraint would lead to (19a) and (20a) as targets for (19) and (20).
Conversational Utterances and Sentences
187
(19) He's in the Yiddish [dipa:] department of Yiddish literature. (S.I.2, 741) (19a) He's in the Yiddish department of Yiddish literature. (19b) He's in the department of Yiddish literature. (20) It was sort of the ... uh certain speech patterns, (example 9.3 from Labov 1966) (20a) It was sort of the certain speech patterns. (20b) It was sort of certain speech patterns.
Although (19a) and (20a) are the grammatical target sentences that would be produced under the operation of the minimalness constraint, it is obvious that (19b) and (20b) would be more appropriate. A similar situation may be seen with the following example. (21) It was during the fixty-six sixty-four. (S.I.2, 1089)
From the context in which this utterance is embedded, it is apparent that the speaker means to say here 'It was during sixty-four': i.e. it was during the year 1964. Similarly, it appears from the context that the hearer understands that the speaker means 'during 1964'. But, if we suppose the hearer to be following a minimalness principle, s/he should not have derived a target sentence which could be so interpreted. A strict application of the minimalness principle to (21) would merely result in the deletion of the pause between fifty and six. There is no reason why the speaker cannot be thought to be saying that the event referred to was 'during the fifty-six sixty-four'. To make sense of such an utterance we need only imagine that 'the 5664' refers to some event - e.g. the annual meeting of a futurist society. Similarly, if the hearer edits only the from the utterance, a next-to-minimal deletion, the target sentence - 'it was during 5664' - makes perfect sense if taken within the context of a discussion of a science fiction novel about the 57th century. Or if the hearer deletes both the and fifty, a less minimal edit, the sentence would make perfect sense within a discussion of the middle ages. In other words, the minimalness constraint will sometimes impose too much of a restriction on the editing procedure. For the appropriate target sentence may not always be the first that the editor arrives at during editing. However, no satisfactory alternative method can be developed for constraining the amount of the prediscontinuity to be edited. Kindle's constituency constraint (see discussion above) sometimes deletes too much and sometimes too little. No other way to regulate the amount to be deleted in the pre-discontinuity seems possible unless we presuppose the hearer's advance knowledge of what the speaker intends to say. For instance, how can the hearer determine that (22a) is the appropriate target sentence for (22), even though a more minimal edit would produce the two complete targets in (22b), unless we assume that somehow he knows in advance that (22a) is what the speaker means? And yet if the hearer already knows, before editing, what the speaker means, then he need not bother editing. (22) If you take a statistical analysis of the people who pass you'll find that it is this question . which - they're passing on that question. (22a) If you take a statistical analysis of the people who pass you'll find that they're passing on that question.
188
Talbot J. Taylor
(22b) If you take a statistical analysis of the people who pass you'll find that it is this question. They're passing on that question.
Even greater than the problems arising from imposition of the continuity, deletion and minimalness constraints are those which arise from the identification constraint. The identification constraint presupposes, in general terms, that the hearer will apply editing procedures if and only if an editing signal occurs in the speaker's utterance. Furthermore, the identification constraint must assume that the speaker's editing signal takes some sort of recognizable phonetic form, specifically a discontinuity in the speech flow (a silent or filled pause, an abruptly cut-off speech signal, or some such 'hesitation filler' as well, that is, etc.). Kindle, for instance, recognizes the need for this constraint without stating how it can be satisfied. The self-correction rules specify how much, if anything, to expunge when an editing signal is detected. The rules depend crucially on being able to recognize an editing signal, for that marks the right edge of an expunction sight. For the present discussion, I will assume little about the phonetic nature of the signal except that it is recognizable, and that, whatever their phonetic nature, all editing signals are, for the self-correction system, equivalent. Specifying the nature of the editing signal is, obviously, an area where further research is needed. (Hindle 1983:4)
Two sorts of difficulties might arise as the result of the identification constraint. First, potential editing signals (however defined in phonetic terms) often occur where no editing is required, as in the following example. (23) And there was [em] - lecturer in German - a lecturer in philosophy - [em] . the president. of the college - the treasurer of the college - and another - bitchy . iceberg of a woman. (S.I.3, 456)
However, provided we assume that the minimalness constraint operates, these examples should not raise problems for the identification constraint for, although the hearer will assume that editing is required whenever the appropriate phonetic signal occurs, the minimalness constraint will, in such cases as (23), prevent the deletion of anything more than the editing signal itself. However, a more serious problem arises from the fact that editing is often required when no signal has occurred, as in the following examples. (24) . . . and is - at the [ha:] . at the * is the logical foundation of ... (S.I.6, 849) (25) she said it's * the basic truth about men . is that men like to be with other men. (S.I.3, 755) (26) And then they rang up about six months later * three months later. (S.I.3, 134) (27) I think * I think this is really [em] - this - feels fairly sound. (S.I.3, 277) (28) I could * I would * I had -joined the staff temporarily. (S.I.6, 299) (29) They've been offered accommodation in the college . which they've [teik] which they've actually going to * I don't understand American systems of. who pays for whose education * (S.I.13, 284) (see also last example in Section 4.5.1)
At each asterisk in the examples above, editing is required, and yet no phonetic editing signal recognizable to the transcriptionists of the Lund Corpus occurs. Such examples are not hard to find; and yet their regular occurrence confounds the whole
Conversational Utterances and Sentences
189
editing hypothesis. For how is the hearer, let alone a computer, supposed to know that such examples need editing and to determine where the editing should begin if there is no signal in the source utterance itself indicating its need of editing? Furthermore, if hearers can understand these utterances without being told that they require editing, then why should we suppose that, in order to understand utterances where discontinuities do occur, hearers must edit to grammatical target sentence form? It would not be hard to argue that we have already seen enough problems with the minimalness, continuity, deletion and identification constraints to shatter any hope that an editing hypothesis might be constructed incorporating these constraints. The grammaticality constraint is inseparable from the other constraints for, much as the identification constraint concerns when to begin an edit, the grammaticality constraint formulates a finishing point for an edit. When enough deletion has been performed so that the remaining sections of the utterance may be combined to form a grammatical sequence of the language, editing is then complete. Thus, it is in order to achieve grammaticality that editing takes place, and when it is achieved, there is no further need for editing. The need for grammaticality, as we have seen, is based on the predominant assumption that speaker/hearers know the rules for constructing and interpreting the grammatical sequences of the language. Thus, when a hearer is confronted by an ungrammatical string of words in an utterance, the only way he can succeed in deriving the meaning of that utterance is by first determining the appropriate grammatical sequence into which it may be edited. Thus we could not have a practicable editing hypothesis without the grammaticality constraint. Nevertheless, we cannot have a practicable editing hypothesis with the grammaticality constraint either. For there are far too many utterances spoken in informal conversational speech which cannot be systematically edited to grammatical target sentence form and yet which intuitively seem easily interpretable and which do not appear to present their hearers with any interpretational difficulties. The following examples, which are by no means rare (cf. Brown 1980, Crystal 1980), provide instances of the sort of problems posed by the grammaticality constraint. (30) It's just as good as setting Virginia Woolf and much more to where - these people want to read. (S.I. 1,904) (31) She said that affects the whole of living and the whole of the sort of [am] you know going . I mean also of course this oil the oil heating. (S.I. 13, 357) (32) But in actual fact of course their central heating oil . is probably roughly about the same price as ours because we're not [am] for very cheap oil. (S.I.13, 372) (33) I mean I'm sack them. (S.I.5, 646) (34) How long do you suppose a life of a fur has. (Brown 1980) (35) That is a suggestion for which I am all. (Brown 1980) (36) I'm very suspicious of the press generally and I can tell you because - not only I mean that's one case that you've given but also in their reporting of erm foreign affairs. (Crystal 1980) (37) Really unless there's something wrong with the candidate from their college . why she shouldn't get it can you - this make sense to me sort of loyalty to their own. (S.I.3, 266) (38) We have some lovely moments when . a kid comes into court and he's got an air gun - and - [am] - it's happened twice that [am] . they've been forgiven - and one time he was given an absolute discharge. (S.2.13, 321)
190
Talbot J. Taylor
(39) Now Sete is one of these beautiful towns that has . a little bit you known I mean your typical . sort of [am] . Mediterranean . sea produce cafes and restaurants on the quay - and then . you go back sort of up you know the it rises steeply through you know sort of cobbled ways up . into the hills. (S.2.13, 757) (40) It's sort of very narrow streets and a forum and the rest of it and it's incredibly beautiful sort of spot that we've found. (S.2.13, 777) (41) The thing about Sete being not only is it quite an attractive . town - it's one of the . few places that had . at least four restaurants. (S.2.13, 785) (42) It's built as heavily and . a . long time ago as it can be really. (S.2.13, 869) (43) But it was quite a blizzard cause Sabre was out in the back garden and this huge Alsatian -1 mean [ris] sort of covered in [ s] . in in flakes of snow . it was really like a sort of beautiful wolf in the Arctic. (S.2.13, 1049)
The point about these utterances, and the many others like them in conversational speech, is that they cannot systematically be edited to grammatical target sentences; nevertheless, they provide no real obstacles to understanding (cf. Brown 1980: 27). It is thus instructive to see the logical corner into which such evidence forces the proponent of any editing hypothesis. He can say that the interpretability of such uneditable utterances as these does not disprove the claim that hearers use editing rules for simpler utterances. This is true. However, the question then arises: why should they? If hearers have no trouble understanding such 'confused' (from the perspective of grammaticality) utterances as (30)-(43) without needing to apply editing rules, then why should we assume that to understand less 'confused' utterances hearers require the use of editing rules? Hearers would then be viewed as some sort of very cantankerous cripples who, although they have no trouble sprinting 100 yards in under ten seconds, require the assistance of crutches when walking to the corner shop. At this point there is an obvious strategy of rebuttal which the proponent of editing rules will be tempted to employ. A familiar strategy in social science methodology, it might be called the 'divide and conquer' method. For it is open to the proponent of editing rules to argue that I have made the mistake of grouping together two types of utterances, i.e. two domains of the explanandum, which would be best treated separately. In other words, it could be argued that it is unfair to take the resistance to systematic editing by some utterances as evidence that a simple editing procedure, obeying the constraints discussed above, is not in fact used by hearers to edit more straightforward utterances. Such an attempted rebuttal would maintain that it is not inconceivable that hearers use one type of systematically constrained editing procedure (e.g. the minimal deletion rule) for some utterances, and another as-yetunformulated editing procedure for the more difficult utterances. My mistake, then, was in trying to formulate a general editing rule accounting for all utterances. The problem with such an attempted rebuttal is that it must rely on the assumption that hearers could determine which set of editing rules to apply to which utterances. They would have to know upon hearing an utterance whether or not it should be edited by a rule incorporating (e.g.) the minimalness constraint. But how could they determine which rule to use unless they somehow knew in advance the intended target sentence? Since it would be admitting defeat for proponents of editing rules to hold that the hearer knows what the intended target is in advance of his choice of editing rule, his only recourse is to claim that the editing signal itself somehow
Conversational Utterances and Sentences
191
indicates which editing rule is to be applied. However, the evidence of discontinuous utterances argues against this possibility, as do the example utterances listed above as problems for the identification constraint. For these utterances do not incorporate any editing signal at all. The divide-and-conquer strategy not only presupposes that the hearer can recognize the criteria for that division. It also assumes that some sort of editing rules obeying the systematicity and form-before-meaning principles can be formulated for the more complex discontinuous utterances. But since these principles presuppose the five constraints discussed above, the evidence provided here argues against that possibility. It is perhaps worthwhile at this time to glance at the complementary phenomenon of ellipsis. The obstacles which face any attempt to formulate rules for editing ellipsis are much the same as those for rules for editing discontinuity. Consider the following conversational extract. Tl T2 T3 T4
G. M. G. M.
I can't figure out how to start on a hill By using the handbrake. And your left foot on the foot brake? No, the clutch.
In each of the last three turns there is a great deal that the hearer would have to add to the utterance if, in order to understand it, s/he had first to edit it to grammatically complete sentence form. The second turn, for instance, would have to be edited to the target form (A), 'You start on a hill by using the hand brake' or something similar. But to arrive at this target sentence the hearer, in this case G, could not simply follow a formal procedure. For that might well result in the inappropriate target (B), 'By using the handbrake you figure out how to start on a hill'. In order to determine whether (A) or (B) is the appropriate target form, the hearer would need to determine in advance what M meant by that reply. But if s/he could do that then there would be no need to edit to grammatical sentence form, since the utterance would already have been understood and the edit would be pointless. Similarly, in order to determine whether the appropriate target for T4 is (C) or (D), the hearer would have to know already what M meant. A purely formal editing procedure, which did not rely on the hearer's advance understanding of the source utterance, could never determine which of the two was most appropriate. (C) You start on a hill by using the handbrake and the clutch. (D) You start on a hill by using the handbrake and your left foot on the clutch.
Indeed, it seems intuitively quite absurd to suppose that, in order to understand T4, a complete sentence such as (D) has first to be computed. It is revealing that this absurdity is the same many schoolchildren feel when they are told that they must not answer the question 'Who was the King of England during the American revolution?' simply by saying 'George the Third'. Instead, they are told to reply 'The King of England during the American revolution was George the Third'. A similar dilemma faces attempts to formulate rules for editing ellipsis and rules for editing discontinuity. No comprehensive formal rules may be devised without relying on the hearer's prior understanding of the elliptical or discontinuous utterance.
192
Talbot J. Taylor
But if the hearer already understands the utterance, then there is no need to edit it to grammatical sentence form.
Grammaticality and communication If the conclusion to the argument presented above is that hearers have no need of editing rules, then we can expect to be asked how hearers - without the help of editing rules - could make sense of ungrammatical, non-fluent and elliptical utterances. That is, the question will inevitably arise: how can hearers understand such disorganized and confusing utterances? The first thing to remark about such a question is that it attributes 'disorganization' and 'incompletion' only to utterances which do not approximate to written language style. The question further privileges written language by assuming that writing is more transparently interpretable, i.e. less 'confusing'. The written language (or 'scriptist') bias has a long history in the study of language, and it reflects the continuing influence of the prescriptivism of traditional rhetoric on descriptive linguistics. In the present instance, it leads to the assumption that the differences spoken language style exhibits, in contrast to written language style, are obstacles to spoken communication. So, since spoken communication is so regularly successful, hearers must have methods for overcoming those obstacles: e.g. by editing spoken language to written language form. In other words, the scriptist bias causes us to perceive a puzzle in spoken language and therefore also to seek explanations for how that puzzle never really poses a problem to conversationalists. The trouble is that the puzzle the scriptist sets himself appears to be unsolvable with the techniques of a linguistics based on scriptist assumptions. In this case a better solution would seem to be to work towards a non-scriptist linguistics which does not take written language style to be the required form underlying all communicational expression. With regard to the questions of syntagmatic structure on which this paper has focused, the suggestions offered in Brown (1980) indicate the direction of at least one possible way to proceed. In his article 'Grammatical incoherence', Keith Brown argues that hearers have little trouble understanding 'grammatically incoherent' utterances such as (30)-(43) above. The syntactic anomalies included in such utterances not only do not interfere with the success of the speech acts in which they occur, but also, he claims, they are probably undetected by either their speakers or hearers (Brown 1980: 27). Referring to example (34), Brown presents a plausible argument in support of his claim. (34) How long do you suppose a life of a fur has?
Having provisionally characterized the production of a fur as a cognitive blend, based on the model of a hair (the speaker is holding up one hair from the fur of an animal), Brown first hypothesizes that the surface structure of the whole utterance might be the result of the blending of the grammatical processes underlying the related structures (34a), (34b) and (34c). (34a) How long a life does a hair have? (34b) How long a life has a hair?
Conversational Utterances and Sentences
193
(34c) How long is the life of a hair?
Brown eventually argues against accounting for (34) in terms of a blending of the surface structures of (34a-c). This would be inappropriate, he says, because the mechanism which would be required to regulate such a blending would have to be very complex. It would be quite absurd to suppose that a speaker could not manage his speech production mechanisms well enough to produce a grammatically wellformed sentence yet could manage a complex mechanism required to blend three intricate surface structures into an ungrammatical output. Instead Brown argues It is perhaps less complex to hypothesise that what is involved is a set of abstract grammatical processes. The processes involved I have represented as 'interrogative inversion' (shared by (34a) and (34b)); DO support (shared by (34a) and the interpolated do you suppose), the formation of the correct form of have (shared in different ways by (34a), (34b)); the formation of constituents (shown by (34a), (34b), and (34c)). Suppose that some monitoring function is satisfied, if the appropriate processes have been carried out - and in a sense, as you can see, they have: we have DO support (but only once in do you suppose rather than twice), we have interrogative inversion, we have a form of HAVE, and we have the interpolation of do you suppose at a place that is consistent with some appropriate constituent break. (Brown 1980: 30)
A more general version of the point Brown makes would be to say that utterance (34) fulfils a variety of syntactic requirements, viz. those needed for the purpose of communicating the intended message (achieving a successful speech act). To fill these requirements a number of operations must be performed: DO support, inversion, constituent structuring, etc. These could be carried out in different ways; but as long as they are individually performed in some way, it does not matter that the ways chosen for each particular operation do not form a grammatically coherent fit with the ways chosen for the other operations. To illustrate, let us imagine that four types of operation (A, B, C, D) must be performed in order to achieve the desired effect X. For each of these types of operation, there are three different ways of accomplishing them. These are represented by subscripts. So we have AI, A2, AS; BI, 62, 63 ... DS. Four operations must be chosen, one from each letter set. If, for instance, A2, 62, C i and Dj are chosen, X is achieved. The same result applies to AS, 62, C2 and DI, and so on for all the possible permutations. However, imagine now that there is more to performing these operations than simply achieving X. Throughout history, certain conventional methods have been established for achieving X. In an imaginary community, at a particular moment, it may be 'the done thing' to achieve X by choosing subscripts in their proper numerical order: e.g. AI, 62, Cj and 04. According to this convention, although another combination - say, A^,B\, Cj,D\ may achieve X, it does not do so by the socially accepted method. This illustration appears to capture what is at issue in Brown's grammatically incoherent, but communicationally successful, utterances. It may be that communication is achieved by (34) because each of the types of grammatical processes required for that utterance are performed. However, the choice of variants within those types does not result in a combination which could conventionally be called 'grammatical'. Consequently, (34) appears, on paper at least, to be grammatically incoherent. No editor would allow it into the pages of his publications.
194
Talbot J. Taylor
Many teachers and linguists would notice it in the speech of others and - if pedantic would correct it. But what is of interest to the target sentence hypothesis is that the question of the grammaticality of (34) appears to be separate from that of its communicational efficacy. Thus, it does not seem unreasonable to conclude, as Brown does, that hearers (attempting to interpret a speaker's utterance) and speakers (monitoring their speech for its communicational efficacy) will be undisturbed by such 'grammatical incoherence'. They may indeed not even notice it. If successful communication is their primary concern, there is no reason why they should. Should they add to their monitoring a requirement that the speech be grammatical - and this is quite common, say, in a 'public speaking' context - then they might well notice such incoherence. But in ordinary conversational situations, i.e. those situations in which language is mainly used, they will have no need to. Another example: if I am playing in a trial for the tennis team, I will do more than simply try my best to beat my trial opponent. I will also try to impress the selectors with the variety of my stroke play, the strength and consistency of my service, my continuous awareness of the game's possibilities, my ability to adjust to different situations and to react appropriately to pressure, and so on. I must do this if I am to convince the selectors not only that I can win on my home court against a familiar opponent but that my game has both the breadth and constancy so that it should adapt well to playing against opponents of differing styles on a variety of court surfaces. The selectors will be looking for more qualities in my style of play than the simple capacity to beat the player opposing me today. However, if the same opponent and I are playing with no selectors present and with a large bet riding on the result, all that matters then is that I win. It is irrelevant whether, in order to do so, I rely on my knowledge, e.g. of this particular opponent's weakness against unorthodox shots with peculiar cuts and spins. In order to win the bet it is sufficient for me simply to defeat the opponent any way I can. The extra requirement of impressing the selectors with my style of play is absent in this situation. Only the minimum requirement of winning matters. It is possible to conceive of ordinary conversational interaction analogously. In everyday conversation what matters most, it could be argued, is that communication is achieved, to the mutual satisfaction of the parties concerned. As long as this criterion is fulfilled (or, as long as the interactants have no reason to believe that it has not been fulfilled), the question of the grammaticality of the utterances produced is irrelevant. The standard of grammaticality which is applied to written language (and which sentence-based grammarians take to underlie language competence in general: for speaking and for writing) does not apply to everyday conversational speech. If this conclusion is accepted, then it may be seen that, in an ordinary conversational setting, the presence of a certain amount of 'grammatical incoherence', of discontinuity, of ellipsis, of self-correction, of 'hemming and hawing', etc., is both natural and untroublesome. In the same way, it is to be expected that, when no selectors are present, my style of play will be sacrificed to the achievement of the more important goal of winning. It would be both artificial as well as irrelevant to measure my style of play, in the betting context, against the selectors' conventional notion of 'ideal style', since that notion is irrelevant to the accomplishment of what, in that context, is the task at hand. The same conclusion might be drawn regarding the
Conversational Utterances and Sentences
195
sentence-based grammarian's application of written language standards of grammaticality to conversational speech. To summarize the conclusions drawn from this reflection on Brown's data: it appears that two questions may be separated in the analysis of conversational speech: the question of communicational efficacy and the question of grammaticality. It is one thing to speak effectively, another to speak in conformity to the conventions for written language style. This does not mean that 'speaking grammatically' never overlaps with 'speaking effectively', but only that the satisfaction of the requirements for one does not imply satisfaction of the requirements for the other. With example (34) - as well as most of the examples included in this article - we have seen that one does not have to speak grammatically in order to speak successfully within a particular communicational context. Similarly, we may imagine someone who speaks perfectly grammatical utterances, 'worthy of prose', yet who fails to communicate in a situation because his utterances do not fit the communicational needs of the moment. We have seen that to understand a non-fluent and ungrammatical utterance the hearer does not in fact need to transform that utterance into grammatical sentence form. The target sentence hypothesis and its subordinate apparatus of editing rules thus appear to be superfluous to any account of how communicational understanding is possible. Furthermore, the argument presented in Brown (1980) suggests that an account of the contribution made by syntagmatic relations to the communicational efficacy of an utterance does not need to rely on an intervening notion of grammaticality. This is certainly not the place to begin the exposition of a new theory of syntax, independent of the notions of grammaticality and sentencehood. In any case, there are a variety of ways that the requirements on such a syntax might be answered. Brown's suggestion of syntactic processes underlying the construction of an utterance is only one possibility. Others are conceivable; what is important for our present purposes is the understanding that the following two questions should not be equated. A. How must a language-user structure his output in order to make it grammatical? B. How must a language-user structure his output in order to make it communicative? Both questions, it may be seen, demand contextually determined answers. How one answers the first depends on whether the output is spoken or written, telegraphed or shouted, used in an advertisement or in an instruction manual. The answer to the first also depends - as we all well know from letters to the newspaper and arguments with our friends - on who has asked and who is answering the question. But, finally, it is important to realize that in many situations - when the overriding concern is not grammar but communication - the first question is simply irrelevant, or, irrelevant until someone decides to make it explicitly relevant. The same context-bound character must be attributed to answers to the second question. Whether an utterance is or is not a communicational success depends on who uttered it and on who they were speaking to. It also depends on what each interactant's criteria are for success at that particular moment and place. If a speaker inaudibly mumbles something and obtains the glass of wine he desired, this may be counted, in some situations, by some people, as a communicational success. Others
196
Talbot J. Taylor
may call it luck. But the point is that the question of the communicational efficacy of an utterance is not decidable once-and-for-all by fixed inter-subjective criteria. The criteria change from interaction to inter-action and from individual participant to individual participant; and the same criteria can be applied different ways in different contexts. In this case, the answer to the question 'How must language-users structure their output in order to be communicative?' is not formulable in terms of abstract rules which generate all and only the set of communicative utterances. For what is communicative at one moment, for one person, may not be at the same or at a later moment, for a different person. At the very least, the attempt to answer question B may be informed by a consideration of prevalence of discontinuity and ellipsis in conversational speech. For this will show that the question of how a communicative utterance must be structured is not answerable in terms of a specification of what the final verbal output must look like. An examination of the utterances that conversationalists actually produce leads instead to the conclusion that the final output can look like almost anything, i.e. that no fixed limits may be drawn determining what utterances must be like in order to be communicative. As the above discussion of discontinuity suggests, in order to make their utterances cohere, speakers can draw on a potentially limitless range of resources: from gesture to 'paralinguistic' and poetic features and from situational context to assumptions of prior experience. For making one's utterance cohere is a fundamentally creative act, and that creativity is not explicable in terms of a simple choice between the instantiation of one fixed string of abstract elements or another. In the end nothing prevents the speaker from incorporating anything at all into his attempt to make his utterance cohere syntagmatically. The distinction, crucial to modern linguistics, between linguistic knowledge and non-linguistic knowledge and between linguistic and discourse or conversational structure knowledge can only serve to prevent the linguist from recognising this fact. At the same time, we must not assume (we have no reason to assume) that the criteria according to which the speaker takes his or her utterance to cohere are the same criteria employed by the hearer. For the act of interpretation is no less creative than the act of speaking, and the creative act is fundamentally the act of an individual. Thus, the communicational efficacy, like the syntagmatic coherence, of an utterance is not only not determinable in advance, by the grammar of the language, it is also not determinable once-and-for-all for all participants and observers alike. The more linguistics enforces abstraction from time, from persons, and from situations on its understanding of verbal behaviour, the clearer but, at the same time, the less relevant its analysis of verbal form becomes. The linguistic perspective on non-fluent and elliptical speech provides only one example of this general law. Another is provided by the analysis of conversational structure in terms of abstract units and their combinatorial rules. The problem with analysing behaviour in terms of the instantiation of fixed strings of abstract elements, whether they be words or conversational 'acts', arises when one considers the huge number of communicationally successful strings produced which, under such an analysis, are incorrectly formed. This is the problem which a sentence-based linguistics encounters with discontinuity, with 'grammatical incoherence' and with ellipsis. The postulation of editing rules may thus be seen as an attempt to patch up the sentence-based analysis just at the point where it connects with reality: that is, at
Conversational Utterances and Sentences
197
the point where the well-formed strings generated by the grammar are compared with the utterances produced and understood by real people in communicative interaction. Within conversation analysis, Griceans and ethnomethodologists have sought to tackle the very same problem by developing new ways of conceiving of rules for the combination of conversational acts. The focus of the Griceans on normative principles and maxims and of the ethnomethodologists on accountability may be seen as products of the concern to bring together the abstract models of conversational organization with the creativity and spontaneity of actual instances of conversational behaviour. The linguistic analysis of syntax would do well to follow their examples. At the same time, conversation analysis should learn not to take for granted sentencebased syntax as the starting point from which the study of 'structure-above-thesentence' is to begin. What research into conversation shows is that many of the basic assumptions underlying the study of verbal interaction, including those supporting such traditional linguistic domains as syntax, have to be reconsidered; and this reassessment must be performed in the light of conversation analytic discoveries about what speakers and hearers really do with words, and not just what grammarians, following an in-built scriptist bias, have for centuries been telling us that they do.
References Brown, K. (1980) Grammatical incoherence. In H. Dechert and M. Raupach (eds.) Temporal Variables in Speech. Mouton, The Hague. Chomsky, N. (1965) Aspects of the Theory of Syntax. MIT Press, Cambridge, Mass. Crystal, D. (1980) Neglected factors in conversational English. In S. Greenbaum, G. Leach, and J. Svartik (eds.) Studies in English Linguistics for Randolph Quirk. Longman, London & New York. Goodwin, C. (1979) The interactive construction of a sentence in natural conversation. In G. Psathas (ed.) Everyday Language. John Wiley, New York. (1981) Conversational Organization. Academic Press, New York. Hindle, D. (1983) Deterministic parsing of syntactic non-fluencies. In Proceedings of the 21st Meeting of the Association for Computational Linguistics. Mimeo. Hockett, C. (1958) A Course in Modern Linguistics. Macmillan, New York. Jefferson, G. (1974) Error correction as an interactional resource. Language in Society 2,181-99. Labov, W. (1966) On the grammaticality of everyday speech. Paper presented at the Linguistic Society of America annual meeting. Laver, J. 1970 The production of speech. In J. Lyons (ed.) New Horizons in Linguistics. Penguin, Harmondsworth. (1969) An Introduction to Theoretical Linguistics. Cambridge University Press, London. Maclay, H. and C. Osgood (1959) Hesitation phenomena in spontaneous English speech. Word 15. Marcus, M. (1980) A Theory of Syntactic Recognition for Natural Language. MIT Press, Cambridge, Mass. Svartvik, J. and R. Quirk (eds.) (1979) A Corpus of English Conversation. Liber, Lund, Sweden. Voss, B. 1979 Hesitation phenomena and sources of perceptual errors for non-native speakers. Language and Speech 22.
13
Do You Understand? Criteria of Understanding in Verbal Interaction TALBOT J.TAYLOR
Perhaps the most important, and surely the most perplexing, concept associated with the role of the hearer in verbal interaction is that of understanding. Indeed, it might be argued that how theorists conceptualize what it is for a hearer to understand an utterance constitutes one of the fundamental pillars supporting the edifice of modern linguistic thought. For on the concept of understanding depend the related notions of communication and language; and it is these notions which serve a foundational role in determining the goals and methods of modern language study. Consequently, if we are to search for the foundations of modern thinking about language, about communication, or about the role of the hearer in verbal interaction, and if we are to pursue this search with the purpose of examining the strength of those foundations, we could hardly do better than to begin by examining modern assumptions about the nature of understanding. In what has arguably been the most influential work in the history of linguistic thought, John Locke's Essay Concerning Human Understanding, verbal communication is represented as a form of telementation: i.e. as the transmission of thoughts from the mind of the speaker to that of the hearer (Locke 1690, Book III, Ch. 1, Section 2). Understanding a speaker's utterance, then, is viewed as a mental event occuring when the hearer derives from the utterance the thought which the speaker intended to convey by it. To make Words serviceable to the end of Communication, it is necessary that they excite, in the Hearer, exactly the same Idea, they stand for in the mind of the Speaker. Without this, Men fill one another's Heads with noise and sounds; but convey not thereby their Thoughts, and lay not before one another their Ideas, which is the end of Discourse and Language. (Locke 1690, Book HI, Ch. 9, Section 6)
This telementational conception of the nature of communication and understanding has cast an enduring spell on the development of Western linguistic thought (cf. Harris 1981; Taylor 1981, 1992, 1997) however, it was certainly not original to Locke. Indeed, it can be found lurking in the linguistic reflections of Aristotle, St. Augustine, the Modistae, the Port Royal grammarians, Hobbes and many others. But 198
Do You Understand?
199
what does appear to be original to Locke's reflections on telementation, and has since become a central factor in his influence on linguistic thought, is his concern with what he called the 'imperfection of words'. We can never know, Locke argues, if the ideas we signify by certain words are the same as our hearers signify by the very same words. Consequently, we can never be certain that our hearers receive the thoughts we intend by our utterances to convey. That is, the 'imperfection of words' consists in the fact that, because the understanding of words is a private, mental event, they do not provide speakers with a means of knowing whether their words are being correctly understood. Locke's concern with the imperfection of words focuses on what we may call the question of the 'intersubjectivity of understanding'. Since, under Locke's telementational view of communication, understanding is a private, mental event, my understanding of a word or utterance is not observable by you or anyone else. Consequently, the speaker cannot be sure if the hearer's understanding of utterances is what the speaker intends them to be. It was never Locke's concern in the Essay to do more than draw attention to the threat that the imperfection of words poses to the intersubjectivity of understanding; he did not attempt to trace the implications of that threat for the study of language and communication (although he did suggest that such a study, called 'semiotic' or 'the doctrine of signs', should form a fundamental part of the new sciences: Locke 1690, Book IV, Ch. 21). Instead, his overriding concern was to suggest how, for the purposes of understanding in the sciences, the imperfections of language could be remedied. This intention conforms with what appears to have been the primary purpose of the Essay, viz. to perform the conceptual ground-clearing required for the laying of solid, empirical foundations to the investigations of the Royal Society (cf. Aarsleff 1982: 24ff.). Locke argued that communicational intersubjectivity could be assured in scientific discourse if speakers clearly defined all complex ideas in terms of the simple ideas of which they are composed. The intersubjectivity of our understanding of words signifying simple ideas was assured; so, if all words signifying complex ideas were defined in terms of their component simple ideas, the threat posed by the ordinary imperfection of words could be averted. On these grounds, and in response to this view of language and understanding, Locke built his scientific method. The implications of Locke's remarks on the imperfection of words were left to those that followed him to consider. In the eighteenth century, theorists of the origin of language, such as Condillac, Monboddo and Tooke, endeavoured to explain how intersubjectivity of understanding was achievable by the ordinary use of language, in spite of the imperfection of words. They argued that in the natural origins of sign-using lay the required guarantee of the mutual understanding of signs. A rediscovery of the natural sources of our common heritage as language-users, they argued, would provide us with a notion of language free from Locke's doubts about intersubjectivity. In the twentieth century, Saussure countered this naturalist response to Locke with a conventionalist account of how ordinary language guarantees the intersubjectivity of understanding. From Saussure's conventionalist point of view, we may be certain that all speakers of the same language link the same signifies with the same signifiants because that connection is arbitrarily imposed on them by the conventions of their language. Saussure's reply to Locke's worries about the
200
Talbot J. Taylor
intersubjectivity of the connection between words and ideas was to argue, in effect, that speakers and hearers do not possess any ideas other than those given to them by the signs of their language. Two speakers of the same language must then, by definition, mean and understand the same things by the same words, for meaning and understanding are language-given properties of the mind. This conventionalist solution to Locke's worries about understanding has endured until the present day, although now and again (and with increasing force in recent years), it has been complemented by naturalist arguments reminiscent of the reasoning of the eighteenth century Lockeans. It is important to see that both eighteenth-century naturalism and Saussurean conventionalism have adopted the same strategy in attempting to remedy the problem raised by Locke's doubts about the intersubjectivity of understanding. This strategy aims to formulate a picture of the essence of language that makes it clear how, by the use of language, ordinary speakers can safely convey their thoughts to hearers. For naturalism, the essence of language lay in its genetic roots in natural signs; for conventionalism, that essence was to be found in its arbitrary and supra-individual form as a socially imposed norm. Consequently, since the eighteenth century it is the concept of language which, one way or another, has been assumed to hold the key to Locke's puzzle about intersubjectivity. The Lockean tradition, has, to the present day, taken it for granted that a proper conception of the nature of language, once formulated, would simply make the puzzle of the intersubjectivity of understanding disappear. Therefore, it is on the enduring drive to preserve Locke's telementational concept of understanding and to protect it from his doubts about its intersubjectivity that we should today focus our attention if we hope to explain the source and inspiration of modern theories of language and if we seek to understand modern linguistic theory's neglect of the role of the hearer in verbal communication. However, while post-Lockean linguistics has endeavoured to explain what language must be for it to be a vehicle capable of conveying an individual's thoughts, it has neglected a related issue. Although Locke was the first to face theoretical dilemmas regarding the intersubjectivity of understanding, ordinary speakers are faced every day with the need to be understood and to know that they have been understood. And yet ordinary speakers do not appear to be stymied by the theoretical complexities surrounding the notion of understanding. Indeed, they will sometimes complain that they have been misunderstood, while at other times they will claim that they have been correctly understood. Yet linguistics, in spite of its concern with the intersubjectivity of understanding, has neglected to investigate the basis for speakers' knowledge of their hearers' understanding of their utterances. By what criterion do speakers judge their hearers' understanding? This should be the first question asked, for the ordinary layman apparently has always had the practical solution to the Lockean puzzle that has troubled linguistics and determined the course of its theoretical development for nearly 300 years. Although naturalism and conventionalism offer explanations of how telementational success can be achieved, their explanations have no bearing on the ordinary communicator's practical need for and reliance on means for determining communicational success. For ordinary communicators are not aware of the theories of naturalism and conventionalism. Furthermore, if, in order to find out if he had been correctly understood, a speaker had to determine if his hearer had received the
Do You Understand?
201
thought he had intended to convey, then no speaker could ever be certain that he had been correctly understood. But it would be absurd to claim that no speaker ever knew if his utterances were understood. It is not the case that, as speakers, we are constantly puzzled about whether our hearers understand us. And, not only would it be absurd to insist that, because of a lack of determining evidence, speakers never know if they are being correctly understood, it also would render obscure ordinary communicational practice. Speakers do occasionally claim that they have or have not been understood, in spite of Locke's worries about the theoretical incoherence of such claims. In this case it should be of great interest to the student of language to investigate the grounds on which ordinary communicators make such claims. Given that, as speakers, we often find ourselves of the opinion that we have, or have not, been understood by our hearer(s), as professional students of language we ought to make it a primary concern to discover the criteria applied in forming such opinions. We need to investigate the reasons speakers will provide in support of such assertions as 'he (you) did not understand me when I said It is worth emphasizing the importance of this question. A modest claim for its importance would point out, at the very least, that it is essential to the study of conversational structure to investigate the criteria conversationalists invoke in justifying claims that they have been understood. A fundamental premise of conversation studies is that speakers ordinarily try to be understood. Indeed, it is not hard to find instances of speakers complaining if they feel they have been misunderstood. So, if speakers do care whether they are understood and if their own behavior in conversational interaction is in some part responsive to their perception of their communicational success, then it should be important to analysts of conversational structure to examine the reasons conversationalists provide for taking a hearer to have understood or to have misunderstood. At least one of the burgeoning schools of conversation analysis has recognized the importance of investigating when and why conversationalists believe themselves to be understood (or misunderstood), the reasons given for those beliefs, and their interactional consequences. Ethnomethodological analysis of conversation recognizes the role that lay accounts of understanding play in conversational structure, and this may be taken as at least one source of their interest in repair, in formulations, in understanding checks, in securing agreement, and in reference specificity (cf. for instance Atkinson and Heritage 1985). A much less modest claim for the importance of investigating criteria of understanding might take its cue from Wittgenstein's dictum, 'An "inner process" stands in need of outward criteria' (Wittgenstein 1953, section 580). That is, it might be argued that the ordinary concept of understanding, and the criteria which determine the application of that concept, should provide the foundation for its theoretical formulation. Furthermore, the ordinary concept of understanding is manifested in what we do, specifically in how we use 'understanding' and other related terms. If we are to construct a solid theoretical account of understanding, and on those foundations build a theory of language, we should, by this reasoning, first turn our attention to the commonplace use of 'understanding', e.g. in the justifications speakers provide of claims that their hearers did or did not understand their utterance.
202
Talbot J. Taylor
Such a methodological reversal would turn on its head the priority which the Lockean tradition confers on the theoretical explanation of understanding. But this reversal would be warranted by the fact that the dilemmas encountered by the theoretical explanation of the concept of understanding are manifestly not encountered by the ordinary speaker's practical use of that concept. And, in turn, this might result in demonstrating that Locke's worries about the intersubjectivity of understanding are groundless, arising from the theoretical preconception of communication as a form of thought-transfer. In the end, a revised notion of understanding, based on the investigation of ordinary communicational practice, might allow us to form a radically original, yet nonetheless empirical, account of the function of language in communicational interaction. (An example of the philosophical interest in criteria of understanding may be found in recent work in the Wittgensteinian tradition, cf. especially Baker (1974) and Richardson (1976).) In order to begin this investigation, we might first ask ourselves: what sorts of reasons do speakers provide when they support a claim that a hearer (H) understood or failed to understand them? On reflection it would seem that they come up with a wide variety of reasons, invoking many different kinds of evidence in criterial support of their claims. (In the following the original utterance will be abbreviated by 'U'; 'S' will be the speaker of that utterance and will take a masculine pronoun; 'H', referred to by a feminine pronoun, will be the hearer of U whose understanding has been queried; 'E' will be the reason(s) S provides in support of his claim that H understood U; 'Q' will be the person asking S why he feels that H understood U.) We might imagine S providing the following type of justification: (1)1 told her to close the door, and she got right up and did it; so of course she understood what I said.
Here, as E, S refers to H's behaviour following U. This might be expected to be a very typical sort of justification of an understanding-claim. At the same time, however, we should note that, in some contexts, H's lack of behaviour could be taken as evidence of her understanding. (2) I told her to leave the door open; and, since she didn't go ahead and close it, she must have understood me.
So, it would seem possible both for particular behavioural acts, as well as the absence of such acts, to be used as evidence in support of an understanding-claim. Related to the latter type, we might imagine a typical sort of justification in which S points out that, since H made no sign of misunderstanding, she must have understood. For many (perhaps all) U, there do not appear to be any fixed limits as to the sorts of behaviour by H a speaker can invoke as evidence of H's understanding. For instance, we might imagine a situation in which S said: (3) Since she left the room when I told her to close the door, she must have understood me. She always objects when I treat her as a subordinate.
Do You Understand?
203
In fact, it would seem that there will always be some possible context for any sort of E to be invoked in support of a claim that H understood U. The type of E chosen will thus be dependent on the situational context. Consider: (4) I know she understood me because, when I told her to close the door, she took out her handkerchief and blew her nose. She always feigns indifference to even the smallest of my requests.
At the same time, we should note that, depending on the context-of-utterance, it is not hard to imagine types of E that could be invoked either in support of a claim that H understood U, or as evidence against it. Consider again the evidence provided in (1) and (4). (5) She must not have understood me when I told her to close the door because she got right up and did it; yet she always makes a point of ignoring me when I tell her to do something. (6) Since she blew her nose when I told her to close the door, she obviously hadn't understood me.
These examples indicate what might be called the 'practical ambiguity' of an E invoked in support of an understanding-claim. No E provides independently conclusive proof that H understood U since, given the appropriate situational context, that same E could be used to support a claim that H had not understood U. And, at the same time, it is always possible, given the appropriate situation, to respond to S's use of E in support of an understanding-claim: 'But what makes you think that shows she understood?' Of course, a very common reason which speakers provide in support of understanding-claims involves reference to H's verbal behaviour following U. This might be taken as a subcategory of the behaviour-invoking justifications already discussed. (7) Since she replied to me that she wasn't the one who left the door open, she must have understood my complaint that Fido had escaped from the kennel again.
Again, it is possible to imagine a situation in which S invoked the same E defence of a claim that H had not understood U. (8) She must have misunderstood me as accusing her of being the cause of Fido's escape since she immediately replied that she wasn't the one who left the door open.
Two particular types of verbal response by H should be mentioned as commonly cited evidence of understanding. We might imagine S citing H's production of what he takes to be an acceptable paraphrase of U as indicative of her understanding. (9) She must have understood me when I said Reagan was unpredictable because she replied correctly that I really meant he was a hypocrite.
Again we could easily imagine the same evidence being used to argue that H had not in fact understood U. The second type of verbal response often citable as evidence of
204
Talbot J. Taylor
H's understanding is the reply 'Yes' to S's query 'Do you understand?'. Speakers will often support their claims that H had understood by pointing out that, when asked if she did, she had replied affirmatively. But, at the same time, given the appropriate context, it is possible to imagine S taking H's simple reply of 'Yes' as evidence that she had not really understood. Many other types of evidence which a speaker might offer in support of an understanding-claim could be imagined. For present purposes, it is only necessary to mention one more. (10) Of course she understood what I said. I've known H for fifteen years and her English is as good or better than mine.
In this invented example, S does not refer to H's behaviour in supporting his claim that she understood U; rather he refers to his familiarity with H and with her command of the language in which U was spoken. We might also imagine S indicating that H had been listening attentively, or that the room was perfectly quiet, or that he knew her hearing was fine, or that she was quite familiar with the topic, as well as many other examples of E where neither H's verbal nor non-verbal behaviour is invoked to support the claim that she understood U. Indeed, the evidence invocable as criterial support of an understanding-claim is far from being restricted to H's behaviour in response to U. It might well seem an impossible task to limit it at all. There is no point in pretending that the imagined examples just discussed constitute satisfactory evidence for the investigation of criteria of understanding. Not only ought we to widen the survey of examples to include explanations of the meaning of utterances, objections to understandings, and more, but above all we ought to direct our attention not to invented but to naturally occurring examples. For instance, the use of an invented example makes us rely on decontextualized acceptability judgements, i.e. when we decide intuitively whether a proposed act would or would not be a way a speaker might justify the understanding of U. Imagined examples of justifications of understanding claims are far from being acceptable substitutes for the real thing. Indeed, there is every reason to suspect that empirical samples of conversational data will reveal even greater variability, ambiguity, and indeterminacy than imaginary data; although, on the other hand, in the empirical study of justifications of understanding-claims, problems will arise as to how to determine what should count as an instance of the category sought. It is all too easy, as an armchair linguist, to imagine that what people actually do with language conforms to our theoretical preconceptions. If the study of criteria of understanding is as important to discourse analysis and linguistics as has been argued above, then those criteria should be the subject of empirical, not just speculative, investigation. On the other hand, I would maintain that some important benefits may be gained through reflection on artificial data. At the very least, we may begin to deal with the sorts of questions that will be raised concerning how the data should be interpreted. That is, we have the opportunity now of forestalling future attempts to force the data into inappropriate theoretical moulds. In the remainder of this essay I will discuss and argue that interpretations of criteria of understanding that will inevitably be raised. In the empirical investigation of criteria of understanding, we might conceive of ourselves as looking for the conditions which competent speakers know must be
Do You Understand?
205
satisfied for it to be appropriately asserted of a hearer that she understands U. This quest may be dissolved into two main assumptions. 1. For any S, H, E and U, the connection between (a) the assertion of E, and (b) the assertion that H understood U is a rule-governed connection, given in the language.
In support of this assumption, one might invoke the general Wittgensteinian maxim that the criteria justifying an assertion of an utterance, i.e. its 'assertion-conditions', are given in 'grammar'. The communicational competence of the speaker of (e.g.) English would then include knowledge of the assertion-conditions, for any English U, of 'she understood U'. 2. Knowledge of the rule-governed relationship between (a) and (b) is tacitly applied by any S in judging his hearers' comprehension of utterances.
That is, when producing U, S unconsciously applies his knowledge of the assertion-conditions of 'H understands U' to his perception of H's response to his utterance. This latter assumption might be taken to explain why it is that speakers are not perpetually in doubt whether their hearers understand them. Tacit application of language-given assertion-conditions would thus supply speakers with an inter subjective criterion of communicational success, a criterion which Locke's discussion of the imperfection of words had seemed to banish forever. Given these two assumptions, the empirical investigation of criteria of understanding might be taken to hold the promise of providing a 'window' on the mind of ordinary speakers. An empirical instance of a speaker justifying an understanding-claim would be interpreted as an explicit version of a rule-governed process performed implicitly by all speakers, during the course of every verbal interaction: i.e. the use of language-given criteria to judge whether their hearers understand what they are saying. At the same time, these assumptions would bring the study of criteria of understanding into line with what some take as given in the study of ascriptions of pain and other 'private' phenomena. The assertion-conditions of an utterance like ' Y is in pain' are taken to be a rule-governed feature of the language, knowledge of which (a) is tacitly shared by all its speakers and (b) informs their implicit judgements regarding Y's ordinary behaviour. So by studying the evidence people invoke in support of pain ascriptions, we might think ourselves to be discovering the tacit criteria all speakers of our language unconsciously apply in judging whether someone is or is not in pain. It is tempting to think that, given the two foundational assumptions of the grammar-given determinacy of and the tacit use of assertion-conditions for understanding-claims, a new theory of conversational structure, and indeed of communicational interaction, could be constructed. However, although it is doubtless true that, for some, what we might call this 'Utopian' picture of criteria of understanding has a definite attraction, a consideration merely of our invented data quickly reveals that the real world of verbal interaction is far from being organized on Utopian principles.
206
Talbot J. Taylor
For what is immediately apparent is that the ways in which speakers justify claims that their utterances have been understood are heavily dependent both on the contexts in which those utterances were produced as well as on the contexts in which their justifications are produced. Although the Utopian view may leave some room for the assertion-conditions of English sentences to vary according to the contexts in which those sentences are uttered, it could not permit the assertion-conditions to vary according to the contexts in which they themselves are invoked. And yet even our brief glance at the types of justifications speakers produce indicates that the nature of those justifications depends on (1) S's interpretation of the context of U (call this UC). (2) S's interpretation of the context in which he is asked for E (call this context QC). (3) What S knows about H. (4) What S knows about the person who asked for a justification (Q). (5) Why S thinks, in QC, he is being asked for a justification, by Q, of his claim that H understood U. (6) What interactional objectives S thinks he can accomplish in his response to Q. (7) What S thinks Q knows about H, U and UC. (8) What S thinks Q knows about S. (9) What S thinks Q's interpretation of QC is. (10) How much S wants to convince Q of H's understanding of U.
It would not be difficult to expand on this list. What becomes increasingly apparent is that what S will invoke as E is not independently decidable or limitable simply by reference to U and to H's behaviour in response to U. Instead, the justification of an understanding-claim is an organic feature of the context in which it occurs (and of S's interpretation of that context). Consequently it is illegitimate to make assumption 1 above: the assertion-conditions for 'H understood U' are not 'given in grammar'; they are not determinable independently of the context in which the justification of 'H understood U' is supplied. It might help at this point to return to the analogy with justifications of pain-ascriptions, where X says Y is in pain and Q asks him how he knows. It seems unreasonable to suppose that, in response to Q, X pays no attention to what he knows about Q, to why he thinks Q is asking him this question, to the circumstances in which the question is asked, to what he, X, knows about Y ('I've known her for thirty years, and she couldn't fake pain if her life depended on it'), to what X thinks Q knows about Y ('Oh come on! Y pretend she's in pain? You know she could never do that'), to the circumstances in which the question is asked (Y is writhing on the floor during the production of a play. X says she's really in pain. Y asks how he knows that's the case and not that she's just acting), and so on. Furthermore, we must not forget that many justifications of understanding-claims (like those of pain-ascriptions) are potentially ambiguous. They might just as easily be used to justify a claim that H had not understood U (cf. invented examples 1, 4, 5 and 6 above), and in no case do they provide incontrovertible proof that H understood U. Instead they are best viewed simply as instances of what S feels will accomplish the task requested of him in the situation given. How S views that task will depend on the individual S, on the context in which it is requested, on who
Do You Understand?
207
requests it of him, and on the particular utterance, hearer and utterance context to which it refers. It would seem then that the assertion-conditions for 'H understands U' are neither fixed nor determinate. Instead, it seems more plausible to assume that, in justifying an understanding-claim, S does not follow any rule but acts creatively according to his interpretation of the interactional context. Given the argument against taking criteria of understanding as governed by language-given rules, it is clear that Assumption 2 should also be rejected. There is no possibility of speakers tacitly applying language-given criteria if those criteria cannot exist. Yet, for many there will still be an attraction to the notion of speakers constantly assessing their hearers' behaviour to determine if it indicates understanding or misunderstanding. 'After all', they will say, 'why would speakers ever complain that their hearers misunderstood what they had said if they had not, at least unconsciously, been assessing their hearers responses to their utterances?' While the force of this argument may be admitted, it should however be pointed out that a speaker's assessment of his hearer's understanding will be dependent on the context (UC) in which that utterance and its response occurs. At the same time, we must recognize that that context is not identical with the context (QC) in which the speaker is asked to justify his claim that H understood U. Thus we may not infer the characteristics of the context-dependent, unconscious assessment of H's behaviour, an assessment performed ('on-line') in the context UC, from the characteristics of the context-dependent justification performed in the context QC. Similarly, we have no grounds for assuming that, for any utterance U, the criteria according to which one speaker will unconsciously assess a hearer's understanding of U are the same as those another speaker would employ in assessing the same hearer's understanding of U. Indeed, speculation about the criteria unconsciously employed in private assessment of understanding can only lead us straight back to Lockean worries about the intersubjectivity of mental phenomena to which we, as observers, have no access. What we, both as observers and as interactants, do have access to are the justifications of understanding-claims that speakers and others produce as a functioning part of particular interactions. It is important that, as students of language, we turn our attention to their study, as well as to the study of explanations of, objections to and corrections of particular, situated uses of the concept of understanding. It is equally important that we, like the interactants themselves, take what we observe at face value, and not simply as surface images of 'hidden' internal processes. To conclude: the telementational picture of communication, passed down to us from Locke, takes the understanding of an utterance to be an unobservable, private, mental event. According to this picture, then, we can never know if our hearers understand what we say to them. Yet this is manifestly not the case: speakers do often object that they have been misunderstood, and they will occasionally produce justifications for their objections. I have argued that it is important for students of language and understanding to focus greater attention on the reasons speakers (and others) will provide for their assertions that a hearer understands an utterance. For these reasons constitute the lay, common-sense discourse of understanding which holds the answer to Locke's telementational picture of communication and the sceptical doubts to which it leads.
208
Talbot J. Taylor
We should, however, be wary of interpreting the empirical instances of these justifications in a way that only leads back to Lockean scepticism. We would risk that danger were we to interpret actual examples of justifications of understanding-claims as explicit illustrations of implicit reasoning mentally performed by speakers during every interaction. In doing so, we would be replacing the Lockean notion of understanding as a private, mental event with a new picture of criteria of understanding as private and mental. And a similar sceptical problem would arise: do all speakers, even of the same language, use the same criteria in judging the understanding of the same utterance? Such a turn of events would only show that we had not learned the lesson of the failure of the Lockean notion of understanding. If understanding and/or criteria of understanding are private, mental phenomena then we have no way of dispelling doubt whether different individuals possess identical versions of those phenomena. But ordinary speakers and hearers do not share these doubts. What this should teach us is that the communality of understanding, and indeed of criteria of understanding, is established in public, observable practice. The intersubjectivity which is the essence of communication is a part of communicational interaction itself and not a phenomenon to be explained by the identity of the private, mental events of different individuals. It is because the intersubjectivity of understanding and of criteria of understanding is a public, interactional phenomenon that speakers do not regularly question its existence. Consequently, we must learn not to study the public face of verbal interaction simply for the value it supposedly has as a window on the private events of the mind, where the 'real' and essential activity is occurring. Communication is not a private, internal process; it is a public, co-operative activity.
References Aarsleff, H. (1982) From Locke to Saussure. Athlone Press, London. Atkinson, M. and J. Heritage (1985) Structures of Social Action. Cambridge University Press, Cambridge. Baker, G. P. (1974) Criteria: a new foundation for semantics. Ratio 16 (2). Harris, R. (1981) The Language Myth. Duckworth, London. Locke, J. (1690) An Essay Concerning Human Understanding, ed. by P. H. Nidditch. Oxford University Press, Oxford, 1975. Richardson, J. (1976) The Grammar of Justification. Sussex University Press, London. Taylor, T. J. (1981) Linguistic Theory and Structural Stylistics. Pergamon Press, Oxford. (1987) The place of Charles Bally in the Lockean tradition. In H. Aarsleff etal (eds.) Papers in the History of Linguistics, pp. 607-14. John Benjamins, Amsterdam and Philadelphia. (1992) Mutual Misunderstanding: Scepticism and the Theorizing of Language and Interpretation, Duke University Press, Durham, N.C. (1997) Theorizing Language: Analysis, Normativity, Rhetoric, History. Pergamon, Oxford. Wittgenstein, L. (1953) Philosophical Investigations, transl. by G.E.M. Anscombe. Blackwell, Oxford.
14
Analysing Fictional Dialogue MICHAEL TOOLAN
We may not in fact understand each other but we have to keep up the pretence, since it is a faith in communication, rather perhaps than communication itself, that marks our claim to social interaction. M.A.K. Halliday, Grammar, Society and the Noun Conversation is like playing tennis with a ball made of Krazy Putty that keeps coming back over the net in a different shape. Morris Zapp, in D. Lodge, Small World
The argument of this paper is that several procedures for the analysis of naturallyoccurring conversation are also valuable to the study of dialogue in fictional texts. Fictional dialogue is an artificial version of talk, partly shaped by a variety of aesthetic and thematic intentions and conventions. But I believe that my working hypothesis is sustained by the brief commentaries on texts I present here: that crucial structural and functional principles and patterns are at work in fictional dialogue as they are in natural conversation. My interest then is in developing a stylistics of fictional conversation, drawing eclectically on the work of a variety of theorists. While these theorists represent different perspectives, their views can be made to converge on the data, so as to provide mutual support for a particular assessment of a dialogue. My main sources are various articles in the Conversation Analysis tradition, e.g. Sacks et al (1974), selective and modified use of the Birmingham discourse analysis propounded by Burton (1980); and the work of Grice (1975). In attempting to describe the structure and coherence of talk, I believe it is misleading to think in terms of a discourse creativity that is governed by obligatory and specifiable rules (a parallel to what some see at work in sentence production). Rather than talk being characterized by rule-governed creativity, what seem more evident are preferred patterns of talk-sequences. Furthermore the specific goals of talk, and the interpretations of patterns, will themselves be locally-determined by the particular interests of co-conversationalists. That is why the theoretical emphasis of Conversational Analysis (henceforth, c.a.) is attractive, where interaction-structures are posited as methodical solutions to ongoing conversational 'problems', and the
209
210
Michael Toolan
orderliness of conversation is understood as an achieved orderliness (West and Zimmerman 1982). The fictional dialogues I will discuss are Melville's 'Bartleby' and Hemingway's 'Cat in the Rain', but I will begin with a brief critical review of the models and orientations I use in this stylistics of fictional dialogue. The only developed model known to me that claims some degree of comprehensiveness, and which has been specifically applied to fictional dialogue, is that contained in chapter 7 of Burton (1980), which is a development of the Sinclair and Coulthard (1975) approach to discourse. I believe that the models have many inadequacies, but are also pioneering work, where whatever shortcomings that emerge are themselves valuably clarifying. In the earlier chapters of her book, Burton very effectively adopts an eclectic 'limited' approach in analysis, but then rejects this in search of something more rigorous. I hope to provide some reasons for spurning that rigour as deathly, even though I endorse her introduction of Supporting and Challenging Moves, as a valuable recognition that conversations are often powerstruggles between participants who adopt various strategies for dissent, in pursuit of the objective that the conversational outcome is consonant with their own conversational goal (the terminology here is that of Hymes 1972). Using this model in an earlier paper, I attempted a Move-analysis of the conversation in the famous Christmas dinner scene in Joyce's Portrait of the Artist as a Young Man (Toolan 1987). I found a quite remarkable preponderance of Challenging Moves within the talk. Integrating the patterning and distribution of these Moves with a variety of other types of evidence, I suggested that the conversation appeared to be manipulated by antagonistic co-conversationalists, who apply their sensitivity to various features of Recipient Design to anti-collaborative purposes. I particularly noted both a series of approaches to conversational breakdown (developed mutually by the antagonistic participants); and, relatedly, various strategies of 'topic suppression', used by other participants so as to avoid such breakdown and to re-orient the talk in consensual, non-controversial directions. But in attempting to use the Burton model of description I encountered many difficulties, some of which I review in what follows. As in the Sinclair and Coulthard (1975) model, Burton posits a rank-scale of discoursal units identifiable in interaction: Transaction, Exchange, Move and Act. Five of the seven Move-types she identifies are properly 'Conversational' (the other two occur at Transaction or Exchange boundaries). These five Conversational Moves, occurring in partially predictable sequences in particular Exchanges, are labelled Opening, Bound-Opening, Re-Opening, Supporting, and Challenging. Working with these it is possible to achieve a preliminary characterization of the development of a conversation in Move terms. The level of Move seems the key interactive one, usually coterminous with speaker change, and 'the minimal free interactive unit' (Burton 1980: 124). On the other hand Moves are actually realized by Acts, allegedly 'the simplest units of all to identify' (Burton 1982: 102), which 'carry all basic functions' (ibid.). The claim is that the array of 19 acts can replicably and unambigously identify all the basic functions in any conversation. Or, put another way, the innumerable ways in which conversations proceed can be reliably, testably, and informatively reduced to just 19 distinct functions.
Analysing Fictional Dialogue
211
I now feel that only along notional lines, as a kind of provisional rhetoric of the norms of (British?) conversational interaction (as some analysts perceive them, on the basis of severely limited data - formal classroom interaction, Pinter plays, etc.) can such models stand. Any stronger claims, to the effect that well-formed conversation (whatever that might be) obligatorily conforms to partially predictable sequences of just these acts and moves, will not be supported by impartial examination of any mixed collection of conversations. That stronger claim reflects what Levinson (1983: 286) has called the 'premature theory construction' to which Discourse Analysis is prone. Nor is such a claim supportable in principle, since it is an implicit denial of the extemporising creativity of actual conversationalists. Insisting on a stronger claim for the validity and power of the model involves the analyst in running the risk of producing analyses which are merely artefacts of the system. I thus view models such as Burton's as partly ethnographic, with considerable pedagogical potential. Predictably, conversation analysts have more forcefully acknowledged that their work characterizes cultural norms of interactive behaviour: their theoretical grounding is ethnomethodological. On the central topic of Supporting and Challenging Moves, Burton contrasts these on the grounds that the former are topic-developing, or compliant with the previous turn, while Challenges are topic-diverting, or non-compliant. Intuitively this seems a plausible way of talking informally of trends in self:other interaction, parallel to the broader patterns of accommodation vs resistance discussed in theories of social psychology (see e.g. Giles and St. Clair 1979). But it assumes many things - not least that we all know, in a uniform way and without specific contexts of utterances, what will count as a topic, a diversion and an act of compliance. In defining Challenges and Supports in terms of attributed functions such as diversion and compliance, the burden of identification is not resolved but simply shifted onto the opaque functional terms. These proposals also tend to discount the fact that the observing analyst cannot inspect the intentions of speakers as they take their turns at talk. And they make no provision for the 'double' analysis that would be needed even to represent speakers' intentions in a dyadic conversation: speaker A's repeated intended Supports may be perceived by speaker B as Challenges, so that a separate description would seem required for each speaker. (Note that this is a quite different type of 'double coding' from that mentioned by Burton (1980: 142).) Finally, how can this model cope with the possibility that a speaker at a particular time may have multiple, even conflicting intentions? It is misleading to assume that not only the analyst but all speakers can inspect and will agree on the intentions (and functions) of each speaker's contributions: an idealizing myth of transparent human behaviour. The entire procedure can only be rescued if it is emphasized that only the analyst's judgements as to speakers' intentions are being reported, and that the analyst qua analyst occupies an awkward marginal ground half in and half out of the conversation (cf. Harris 1981: 46-7; Levinson 1983: 287). Burton further distinguishes Challenges from Supports on the basis of three criteria: the breach or maintenance of (a) discourse framework, (b) topic-establishing requirements, and (c) Labovian validity preconditions. Under discourse framework, she refers to the expectation of certain acts being followed by certain 'partners', as
212
Michael Toolan
though the first were an initiator and the second a response (the analogy with c.a.'s notion of 'adjacency pairs' is clear, and instructive). Paired acts, where nonoccurrence of the second after appearance of the first is said to constitute a Challenge, include the following: initiating informative elicitation directive accuse
responding acknowledgement reply react excuse
An initial dissatisfaction here may lie with the indiscriminate lumping-together of these four pairs, where it seems that varying degrees of prospective force inheres: for me at least, elicitations and directives are much more noticeably requiring of particular appropriate responses than informatives and accusations are. But more damaging to the proposal here is the simple fact that initiate-response maintenance may easily involve a Challenge. Any response act after an elicit one, for example, is said to constitute a Support, despite the misleading descriptions that this would give rise to, as in: What shall I do this afternoon? Go away and drown yourself.
elicit reply + directive?
OPENING SUPPORT?
We must also qualify the general discourse framework claim that, for example, directives and reacts form a well-tied pair, tightly interconnected by the 'gravitational' pull of their prospective and retrospective forces. Rather, a react is a preferred second-pair-part following a directive, not an obligatory one. The shift in emphasis here simply echoes the abandoning, in c.a., of unjustifiably stipulative specifications of adjacency-pair rules, in favour of greater emphasis on conditional relevance and preference organization. If it is accepted that Challenging and Supporting are chiefly a notional contrasting of fundamental dynamics of talk, the way is open for more discursive speculation as to the bases of these dynamics. My suggestion would be that these moves be viewed as orientations to the two fundamental roles mutually recognized in all speech interaction, Self and Other, respectively. Supports are really dependent moves, defmitionally supportive of or deferential to Other's move, whether that was an Opening, Bound-Opening, Challenge, or whatever. Challenges, by contrast, I see as intrinsically supportive of Self, perhaps oriented away from Other(s). The near-identity of act-structure description between Openings and Re- and Bound-Openings reflects the fact that these latter two are thirdturn alternative Moves designed to develop the orientation of the Opening Move while adequately attending to the Support or Challenge in the immediately-preceding second turn. All thus hinges on the Opening Move. As to whether that is Self or Otheroriented, this cannot be specified in advance. But looking at Burton's possible exponents of the obligatory head slot in Opening Moves - informs, elicits, directives or accusations - informs look to me Other-oriented (co-operatively telling what isn't already known by Other), while elicits and directives look Self-oriented (requesting
Analysing Fictional Dialogue
213
or ordering some compliance with Self s interests or wishes). Whether accusations are a very common or significant Opening Move seems doubtful. The argument for the centrality of elicits, directives, and informatives - at least at Exchange openings - gets support from two quarters: 1. Burton has Labovian validity preconditions for just these three acts (and no others). 2. These three acts are fairly clearly discourse counterparts of the three major grammatical sentence-types, interrogative, imperative and declarative; and these are remarkably widespread in the languages of the world, if not universal. But I'm not arguing for a strong formrforce correlation here, rather sharing Levinson's tentative recognition of the association of these three types (Levinson 1983: 242, 274-5) But a deeper problem persists: how can we be sure that we can identify, say, directives? Once a wedge is driven between forms and meanings, as was effected by Austin (1962), such that we acknowledge that interrogative forms do not necessarily perform the act of asking questions, that declarative forms do not necessarily perform the act of making statements, there is no basis in principle for denying that there is no fixed method for identifying, for example, all and only the well-formed directives in conversation. Whether or not we wish to adopt Austin's categorization of the three aspects of a speech act, it seems that Discourse Analysis models typically derive their descriptions solely on the basis of the conventional illocutionary force associated with an utterance. But as soon as one looks at utterances in sequence in a conversation, and attempts to characterize 'what is going on' in the interaction, the analyst - like the participants - is obliged to consider the possible perlocutionary effects (intended or not) that may be felt by succeeding addressees, and which may help shape their own subsequent conversational contributions. Speakers (like analysts) build hypotheses as to just what are the illocutionary forces and perlocutionary effects of their own and others' utterances, as the talk proceeds. The problem remains, however, that perlocutionary effects are often indeterminate, unpredictable, unspecifiable or unmodellable outside of the particular interaction in which they arise. On reflection it is clear why there is a degree of success in the Sinclair and Coulthard model of formal classroom interaction, even with their focus on illocutionary acts. The explanation is that such classrooms give rise to very ordered interaction, which is not mutually ordered. All the ordering or controlling is done by the powerful architect of the talk, the teacher. And the description is built around the teacher's assumed illocutionary acts (the 'guesswork' here in identifying those acts is that of informed insiders - teachers themselves). Where pupils' perlocutionary reactions do not fit the teacher's intentions, they are easily supplanted by a teacher's Re-Opening Move. As Burton puts it (1980: 141) 'The teacher is in control of structural choices . .. [and] . . . of content right through the hierarchy'. In practice, given their failure to emphasize that it is teacher-control that designs formal classroom talk, Sinclair and Coulthard's overly prescriptive tendency to assert that certain utterances are obligatory tokens of just particular speech act types is frequently on show. Thus in their analysis of interrogatives they give the example of: Can you play the piano, John?
214
Michael Toolan
asserting that this will be interpreted as a command (a directive?) 'if there is a piano in the room'. But one can think of many cases where such an interpretation would be mistaken, and might create havoc in the classroom. There are just too many other context-bound requirements that will also have to apply before John's prompt engagement with the piano would be felt to be appropriate behaviour. Suppose John is a new pupil, for instance, or a new colleague (in such cases that old favourite riposte might be proffered: 'I don't know, I've never tried'). Similarly, while Sinclair and Coulthard claim that 'I can hear someone laughing' is a declarative functioning as a 'command to stop', it might simply be a teacher's annoyed aside, remarking on distant laughter, somewhere outside the room. The formalism here then is overly restrictive. But elsewhere this study seems to be moving in a more illuminating direction. On p. 113 the authors make the crucial point that the structuring of contributions to a lesson is a learned social behaviour, that 'speaking in class' is a learned behaviour. The implications of this for discourse modelling are profound, and relate directly to an earlier point - that in formal classroom interaction the teacher is entirely in control. Similarly, on p. 114 the authors suggest a cross-cultural comparison of a British school with one in a Commonwealth country, so as to identify any differences in 'the expectations and norms that govern teacher-pupil interactions in different cultural groups'. And on the chimera of well-formedness, we should note Sinclair and Coulthard's scepticism (p. 120) over 'the likelihood of a useful concept of well-formedness in discourse, above the smallest units'. On the same page they express their preference for a procedure of 'continuous classification' rather than any macro-structuring: there is an ethnomethodological orientation here that seems not to have been sustained in later work. The temptation to make misleading and destructively ambitious claims for discourse analysis is always at hand. There is no warrant, in my view, for declaring that linguistic analysis of conversation 'accounts for' effects or phenomena in the data (Burton 1980: 68). Our descriptions merely attempt an account of, they do not account for. The 'the model accounts for the data' approach very easily gets re-cast as extraordinary claims for the 'explanatory power' of the model, and for its 'generative' power (all well-formed conversations are derived from this model, and so on). Rather we must return to acknowledging that our modellings of 'what is going on' in conversation are context-based (i.e. data-based) - if not context-bound - descriptions or accounts. For such reasons, I am more attracted to the first half of Burton's book, where what she dismisses as 'studies' are presented, than the second half, where 'analyses' are offered, constituting a shift towards 'a full and adequate linguistic analysis' (Burton 1980: 93). Part of the rationale for this shift is a search for a replicable method (particularly for use with students). And yet some of the evidence here points in the opposite direction. In her opening chapter she comments very effectively on the interaction in Pinter's Last to Go using Labov's hypothesis of AB knowledge: my judgement is that if students were 'primed' with this hypothesis, they would typically come to conclusions quite consonant with her own. But the doubtful replicability of the Act-Move taxonomy is vividly highlighted if we compare Burton's own labelling of Pinter's The Dumb Waiter, in the penultimate chapter of her book, with her later
Analysing Fictional Dialogue
215
analysis of the same interaction in Carter and Burton (1982). In one stretch of talk alone, (Moves 21-39 in the later paper), what were formerly seven distinct Exchanges are now treated as one, where, at the Act level, there are changed descriptions of 12 of the 21 Acts. What was previously an 'acknowledge' is now an 'evaluate', what was previously an 'inform' is now a 'reply', and so on. All these changes are made without comment. The hollow foundation of Burton's approach is a defence of 'total' stylistics (p. 94), which is a familiar variant of the biplanar fixed-code myth. Discursive observation-cum-description is down-graded as 'theoretically unsatisfactory from the linguistic-stylistician's point of view' (p. 95). But it is that point of view itself, the dominant paradigm of orthodox linguistics (see Harris 1981) and orthodox stylistics (see Taylor 1980), which is the suppressed source of dissatisfactions - the myth of fixed-code telementations, and fixed inter-subjective meanings. We need to abandon any narrow assumption of being able to specify acts in terms of (closed or open class) forms at all. Using the language of traditional grammar, acts can only be specified positionally, not formally, in terms of sequential constraints and patterns, or what c.a. calls conditional relevance. And the positional orientation must be resolutely functionalist, specifying what the analyst takes the particular act to be counting as, for the participants, in the ongoing talk. To be distracted by linguistic form in this area is to fall victim to the principle one began by denying - the assumption that 'what is said' is 'what is done'. To be sure, the Birmingham models are functionalist to a degree, and the notion of prospective/retrospective force is a valuable contribution to developing a positionalist description (see also Burton 1980: 20). But their 'microcosmic functionality' (Burton 1980: 120) needs to be wedded to a broader attention to function within context-of-utterance. Relatedly, the tendency to conceptualize conversation as 'a joint production' needs to be amended. It's more a joint producing, with no product or text at the close of the conversation. Goals may have been reached, but that's a different issue (and B may not even be aware that A's goal has been reached). There is not, in typical conversation, anything like the minutes of a committee, binding interactants to what the Chair sees as the 'key achievements' of the talk. There needs to be a renewed recognition that the (often unpredictable) structurings of conversation must be related to its instrumentality, its use in getting things done. Coulthard and Ashby (1976) are quite correct in saying 'Equal participants negotiate in subtle and complex ways for the right to speak and the right to control the direction of the discourse or to introduce new topics'. But it should be added that, typically, all such negotiations, and the talk itself, are not the ultimate concern. The strategies and struggles are designed to help achieve various goals or needs (physical, social, psychological). Given all the problems arising from expecting any simple Act-and-Move model to be applicable to a large range of conversations in an informative, replicable and principled way, I believe stylisticians - and analysts in general - should content themselves with doing the possible. For me this involves relating, where seems appropriate, various patterns, phenomena, conversation-shaping principles, etc., to the selected data, fictional dialogue. One does so in the expectation that such principles may contribute to an account of how and why, as readers, we have the intuitions we do, make the judgements we do, concerning fictional dialogues.
216
Michael Toolan
In the course of my discussions of my chosen texts, I shall frequently refer to the Gricean hypothesis, that we normally take a Cooperative principle to be underlying our interlocutors' utterances, in which requirements of truthfulness, informativeness, relevance and clarity are attended to (or 'flouted' only with good reason). Grice's picture, as he himself has emphasized, is partial and incomplete. It may be that (in Western culture at least) cooperativeness is a more widely applied norm of conversational behaviour than any other, but this does not mean that other principles - e.g. of submission, coercion, resistance, tact - do not exist, nor that they are not quite central to understanding particular types of interaction. The fault lies less with Grice than with those who have taken what were intended as descriptions, and treated them as ground-rules, with an increasingly prescriptive colouring - the fate of all truly influential descriptions. If one keeps in mind that the maxims are requirements for maximally efficient, rational, and cooperative talk, i.e. that the maxims are themselves an attendance to certain interests rather than others, they remain very valuable. I should now like to discuss, briefly, possible uses of the orientations referred to above, in my reading (in the broadest sense) of the selected fictional dialogues. In addition to those orientations, however, we need to pay greater attention, in our analyses, to speech-event contextual constraints (Hymes 1972), and to contextual cues (Gumperz 1977), as well as to the staging and graphological representation of talk, especially since certain features of natural talk - e.g. prosody, overlap, repair, routine opening and closing sequences - are usually only very sketchily presented in written texts. In any event, as I hope to have suggested, while Discourse Analysis models remain rather problematic in their application to 'casual' conversation and its fictional representation, both Gricean principles and c.a. orientations are useful in their specification of normal behaviours, against which we can set these fictional instances of motivated departure.
'Bartleby' Melville's famous story records the curious life and death of a mysterious and unconventional scrivener who appears to become wholly - even fatally - dependent on the increasingly reluctant patronage and protection of a wholly unmysterious and conventional lawyer. It is the lawyer himself, a prosperous Master in Chancery, who narrates the tale. I will focus solely on the two parts of the famous problematic exchange which recurs with slight variations through much of the story, and which concludes the following extract: It became necessary to examine [quadruplicates of a lengthy legal document] . . . I called [three of my four clerks] from the next room, meaning to place the four copies in the hands of my four clerks, while I should read the original... I called to Bartleby to join this interesting group. 'Bartleby! quick, I am waiting'. I heard a slow scrape of his chair legs on the uncarpeted floor, and soon he appeared standing at the entrance of his hermitage. 'What is wanted?' said he, mildly. 'The Copies, the copies', said I hurriedly. 'We are going to examine them. There' - and I held toward him the fourth quadruplicate.
Analysing Fictional Dialogue
217
'I would prefer not to', he said, and gently disappeared behind the screen, (pp. 477-8)
The enigmatic reply, 'I would prefer not to', with which the office clerk, Bartleby, answers a range of requests from his employer, the Master in Chancery, both compels interpretive analysis and defies it. Bartleby's conversational behaviour is intriguing since his replies to the lawyer appear to attend to all four Gricean maxims and yet the lawyer understandably feels him to be maddeningly wn-cooperative. His responses are consistently honest, informative, relevant, and concise, while also conveying that Bartleby's goals and interests remain different from the lawyer's. And, in addition, Bartleby's replies seem not to trigger informative conversational implicatures or inferences. They 'feel' cryptic, but are hard to fault as insufficiently informative. Merely taking note of the 'basic' prepositional content of Bartleby's responses seems to be 'all we know or need (in Bartleby's view) to know'. Even at the most basic level, of deciding whether or not Bartleby's reply is a rejection of the Master's directive or not, there can be uncertainty. One can easily imagine a hard-working but outspoken subordinate informing his superior, even as he complied with instructions, that he didn't like doing so. Bartleby's utterance has the potential to convey two entirely contrasting implicatures. Compare A Bl B2
Please do this at once. I would prefer not to, and will not. I would prefer not to, but will comply.
Significantly, the ambivalence between a Bl and B2 import to Bartleby's utterance is eventually probed by the Master (p. 482), but inconclusively: 'Bartleby', said I, 'Ginger Nut is away; just step around to the Post Office, won't you? (it was but a three minutes' walk away), and see if there is anything for me'. 'I would prefer not to'. 'You will not?' 'I prefer not'.
And yet every reader is struck by the oddity of Bartleby's replies: they are noticeables, dispreferreds, marked or abnormal responses. This is partly because it is simply a cultural dispreferred to deny or reject a reasonable request (see Pomerantz 1978). But the oddness we feel in Bartleby's behaviour is good evidence of the way we must - like it or not - attend to the socio-political setting of his utterances. His replies are then recognized as noticeables because we know from experience (pragmatics in a non-technical sense) that employees who evade the work assignments of their bosses risk dismissal. We are constantly aware of the (buckled) employer-employee frame of the Bartleby:lawyer interactions: it is one of the crucial parameters for assessing the 'communicational relevance' (Harris 1981) of what is said. Within that employer-employee frame, assumptions of true cooperation, of the mutually-supportive interaction of equals, must sound rather hollow. As we proceed through the Master in Chancery's many efforts to engage Bartleby in particular tasks, we may wonder what distinguishes a request from a directive. In Speech Act theory, it is the felicity condition that the speaker believes the addressee might not mind doing the act requested. This formulation covers a multitude of veiled
218
Michael Toolan
coercive sins, in our exploitation of others. For very often, we may suspect, the powerful unreflectingly proceed on the assumption that their subordinates 'do not mind' doing the work they are set, and need only to be requested - a kind of Whorfian backgrounding of the facts of control. But Bartleby quite conspicuously (however unwittingly) forces the velvet-gloved hand. His initial replies to the lawyer might be seen as those of someone hearing directives as mere (declinable) requests or offers. Having so declined these 'requests', Bartleby's behaviour forces the lawyer to expose to us (if not also to himself) his true directive intents. Typically, Bartleby's (apparently unconscious) carnivalizing of the established power relations also dismantles the myth underlying capitalist relations - the enabling fiction maintained by the propertied classes (the J. J. Astors admired by the Master in Chancery) that there is nothing coercive or directive in the system, that they believe workers 'not to mind' being wage-slaves. Bartleby's persistent preferences for denial compel a developing recognition (again, in readers more than in the lawyer) that what was taken to be consciousness was false consciousness. Applying Burton's category of Labov-style preconditions for a valid request, we can see that the defectiveness of the lawyer's moves as requests stems from his complacent and false assumption that the addressee will not mind complying with the request that is made. Yet another way of examining Bartleby is via Austin's illocutionary/ perlocutionary distinction. The Master in Chancery's utterances clearly seem to have (to us observers) the illocutionary force of ordering or directing (even if Bartleby initially sees them as offers). But whatever their illocutionary force - and Bartleby does seem to recognize that force - there seems to be a minimal perlocutionary effect, on Bartleby as addressee, by making these utterances: he is not persuaded or cowed or caused to become engaged in the specified task. As for Bartleby's response, 'I would prefer not to', this is very hard for the Master - or Burton, or Grice - to handle. In Burton's terms, a fitting response to the lawyer's 'directive' would be a non-verbal compliant 'react' Act. But Bartleby wreaks havoc with the Support/Challenge binarism, and the Gricean distinction between maxim-observation and maxim-flouting. He fastidiously observes Burton's criteria for producing a Supporting move, and seems to acknowledge that the directive is appropriate and that the lawyer is justified in making this claim on him. So far, so supportive and cooperative. But as for complying in quite the way the lawyer would wish, this he would prefer not to do. The lawyer's 'consternation' (p. 476) should be echoed by the discourse analyst's, in face of what is both a Support and a Challenge. Some kind of baffling, innocent exploitation of conversational resources is being effected. In c.a. terms, the response is a dispreferred second, suitably modulated by observance of negative politeness markers but, notably, lacking any mitigating account as to why the preferred second cannot be performed. The best that the Burton system could do would be to suggest that Bartleby's response is an 'informative' Act, hence a Challenge since it is 'unexpected and inappropriate' (p. 151). But such a labelling doesn't fit well: Burton stipulates that informatives have the sole function of providing information. Clearly, Bartleby's reply does more than that - it is in addition a well-designed deflection of a request/directive. Again, the model is too inflexible - the binarism of expected vs unexpected lacks delicacy and plausibility of description. In sum, Bartleby's responses are excellent examples of the
Analysing Fictional Dialogue
219
unexpected departure in actual conversational practice. Part of the mesmerizing enigmatic power of Bartleby's reply - a conversation-stopper of the best design - is its impenetrability, reflected in the difficulty in describing it within extant, normative discourse models. One way of responding to the Bartleby data, which might be adopted by some Discourse Analysis systems-builders, is to argue that Bartleby's responses are the oddities that prove the rule or system (see Whorf on proving the rule). But my point is rather that in Bartleby the potential for limited interactional 'success', for mismatch, non-compliance, the interrelation of all sorts of principles beside cooperativeness, is simply peculiarly highlighted: it is in fact there, behind the authorially groomed and polished sentences of typical fictional dialogue, all the time. And we ignore this at our peril, for so ignoring these phenomena fosters two quite crucial misunderstandings: 1. We then tend to see both real people and fictional characters, whether 'flat' or 'rounded', as unproblematically readable, typically (and ideally) informative (i.e. confessional), succinct, unambiguous, single-minded, and with easily summarized goals and aspirations. Individuals are reduced to the condition of analyst-readable programs. 2. We also tend to assume a simplistic picture of the nature of language, as a fixed shared code, devised (somehow) for the purposes of mutually attending to a shared inventory of concepts (again, on bi-planar telementation, see Harris 1981). We need to re-cast our descriptive models to accommodate a cline or continuum of possible acts in particular contexts, ranging from most expected to least expected. (In fact, such is the diversity of communicational needs in different situations, a multiplicity of distinct but interrelated clines or axes will probably need to be integrated, if we wish to build a reasonable model of meaning potential in discourse structuring.) If there are degrees of grammatical acceptability, then so too are there degrees of conversational cooperativeness, and of conversational coherence. But we won't be able to posit these divorced from (implicit or explicit) criteria and contexts of use. It is perhaps worth noting that in that branch of linguistics closest to discourse analysis, sociolinguistics, recent theory has concentrated intensively on just this issue of modelling the clines and variability of actual usage (by way of variable rules or implicational scales or whatever).
'Cat in the Rain' Hemingway's short story renders a few brief scenes from a marriage. A young American couple are staying in an Italian hotel where, on a rainy afternoon, the husband (George) seems happy to lie on the bed reading, while his more restless wife talks on of the things she wants but lacks. Her apparently unengaged sympathies become focussed on 'a poor kitty out in the rain'. Having failed to retrieve it herself, it is unclear at the close whether the cat brought to the couple's room by the maid (at the instigation of the padrone) is the same one for which the wife had earlier pined. My particular interest in the story concerns the seeming emotional and temperamental gulf between the couple, and the way this is conveyed by and reflected in their stilted uncooperative talk:
220
Michael Toolan
She opened the door of the room. George was on the bed, reading. 'Did you get the cat?' he asked, putting the book down. 'It was gone'. 'Wonder where it went to?' he said, resting his eyes from reading. She sat down on the bed. 'I wanted it so much,' she said. 'I don't know why I wanted it so much. I wanted that poor kitty. It isn't any fun to be a poor kitty out in the rain'. George was reading again. . . . She laid the mirror down on the dresser and went over to the window and looked out. It was getting dark. 'I want to pull my hair back tight and smooth and make a big knot at the back that I can feel', she said. 'I want to have a kitty to sit on my lap and purr when I stroke her'. 'Yeah?' George said from the bed. 'And I want to eat at a table with my own silver and I want candles. And I want it to be spring and I want to brush my hair out in front of a mirror and I want a kitty and I want some new clothes'. 'Oh, shut up and get something to read', George said. He was reading again. His wife was looking out of the window. It was quite dark now and still raining in the palm trees. 'Anyway, I want a cat', she said. 'I want a cat. I want a cat now. If I can't have long hair or any fun, I can have a cat.' George was not listening. He was reading his book. (pp. 325-6)
In 'Cat in the Rain' there are several exchanges where we may suspect that George's contributions are at least recalcitrant, if not challenging. The conversational pattern of that recalcitrance is worth delineating. In George's responses, as in other areas of the story (are there two cats? are the nationality and language contrasts significant? and so on), the text seems to be deliberately inexplicit. But the cumulative evidence points to the following: 1. The wife is not merely 'making conversation'. Whether or not the cat and the hairstyle are themselves symbolic of deeper wants (i.e. regardless of just what it is that is wanted), her utterances seem not to be intended to be seen as mere expressives: they are indirect requests performed by asserting one of the felicity conditions on the act, but with no explicit mention of an addressee). But it may be that George pretends to treat these utterances as defective informatives, therefore ones he need not attend to. He might view them as defective on the Labovian grounds that he believes his wife knows that he already knows and is not interested in the information that she wants a kitty, that she would like to change her hairstyle, and that she wants silver, candles and spring. But if some of his responses do suggest George trying such a strategy in the hope of closing the talk (e.g. his silence, and his unenthusiastic 'Yeah?') the sheer persistence and repetitiveness of the woman's contributions support the need for an indirect-request interpretation of them. 2. These indirect requests are quite clearly addressed to the inattentive husband, notwithstanding the incidental involvement of the hotel-keeper and maid. Of the hotelkeeper it is reported that the wife 'liked the way he wanted to serve her'. And it is precisely some such 'wanting to attend to or serve her' that seems lacking in George's behaviour (conversational and otherwise). Reflecting this, George never initiates exchanges, except on one occasion, a ritual politeness enquiry, on the wife's return to their room, as to whether she has found the cat. 3. George's responses to his wife's indirect requests are almost all noticeably dispreferreds - silence, or some minimal contribution with slight prospective or talk
Analysing Fictional Dialogue
221
maintaining force. Not only are they not compliances with the requests, they are not even apologetic accounts for non-compliance. As the woman no doubt realizes, some of these requests make unrealistic demands: the preparatory condition that George is in a position to supply silver (not to mention spring!) probably does not hold. In such cases, accommodation in a broad sense, marked by a sympathetic response, is the conventional preferred. But George, in his efforts to get back to his reading, seems to be often doing the bare minimum of attentive, cooperative talk-monitoring; and relatedly, he is frequently reported to have withdrawn speaker-oriented gaze. 4. Faced with such uncooperativeness, the wife's lengthy repetitive turns may be interpreted as designed to keep off silence. Silence is well-recognized as a conversational dispreferred, one which the wife knows George would be only too happy to introduce and observe. 5. The above evidence of the lack of fluency in the talk only points up the broader conflict here, as to what the two people are doing: the wife wants to be conversing, but the husband wants to be reading. The Gricean base of cooperative behaviour is being undercut even within the broad frame of a collaborative partnership (the couple are married, share a room, talk to each other): the doing of 'such as is required' specified by Grice, which must mean crucially 'such as is mutually felt to be required', is repeatedly disregarded. The disregard persists not least because, it appears, there is no mutual acknowledgement here of 'what is required'. And in nonroutine non-ritual talk there is the ever-present threat that disagreement over 'what is required', previously unnoticed, may emerge and trouble the interaction. But then in non-routine talk a major attraction is the recognized lack of a required structure and content. It is a kind of living dangerously done in the hope of communicating, of 'getting through'. Linking Michael Halliday to Morris Zapp, we eschew silent scepticism and suspicions that our interlocutors have some unfair advantage, in favour of a faith that spontaneous conversation is a game of skill and chance, worth playing. It is common to emphasize that talk typically involves a mutual negotiation, in and through talk, so that each party's interests and face are attended to. If that is a preferred norm, it has to be stressed that interaction frequently and noticeably departs from it, as the two texts examined suggest. In the Bartleby/Master dialogues, there is a striking paucity of negotiation on the scrivener's part. Behind his responses stands a particular curtailed attentiveness to his superior's needs and goals: that much having been allowed, Bartleby makes almost no concessionary modifications or adjustments in later turns, notwithstanding the Master's evident discomfort expressed in intermediate turns. It is very largely only the Master - and he far more than the other clerks - who attempts various failing strategies of negotiation, prior to a series of baffled capitulations. But the price to be paid by uncooperative dependents in institutionalized settings (school-children, employees) may be very high. When Bartleby is finally cast out by the Master, he wastes away in a prison for the destitute. It emerges that the clerk's 'pallid hopelessness' may have been first nurtured by his working - cruel irony - in a Dead Letter Office, the graveyard of unattended written first turns.
222
Michael Toolan
Concluding remarks If one opts for an eclectic 'intermittent' conversational analysis, interweaving informal commentary with formalist characterization, one is constrained by the fact that several of the features focussed on by c.a. - such as closings and openings - are not often prominent, or rendered fully, in fictional dialogue. In addition, while c.a. rightly focusses on everyday conversational routines, fictional dialogue is very often non-routine. Embedded in and constitutive of a story, it is part of what author and reader take to be extremely tellable material. And it may well be designed (in ways that make it less like a transcript of ordinary talk) so as to enhance its tellability. Despite such complications, I believe a stylistics of literary dialogue remains a valid undertaking, but that stylistics must reject any attempted inventory of conversational acts which stipulates entirely context-free correlations of particular utterances with particular conversational purposes. Such a model issues in arbitrariness, and is quite inappropriate in the domain of literature and literary response, where originality should be especially valued and expected. I also share Pratt's (1986) scepticism concerning the special status claimed for Grice's Cooperative Principle, and suggest that a broad expectation of, and contrivance of, Relevance, is the key dynamic of conversation, regardless of whether particular conversationalists happen to adopt cooperative, antagonistic, submissive, directive, or other principles. Generalizable though the Cooperative Principle may be, as a logical helpfulness from which departures are noteworthy, very many conversational interactions will also and (undisregardably) carry the marks of participants' differential status, power, temperament, and so on. At the very least, we should recognize that the cooperation that Grice has proposed as the macro-principle is itself observed due to the constraining of conversationalists by other and varying strategies or maxims in varying circumstances: strategies of coercion and submission, of tact, respect, and so on (cf. Leech 1983; Brown and Levinson 1978; Brown 1980). In their monitoring and working at talk interactants are always making efforts to be relevant (though clearly, and for various reasons, those efforts vary), and to see relevance in others' contributions. From this arises the nearest counterpart, in interaction, to the notion of grammatical ill-formedness: a judgement by one or more participants that the turns (if, indeed, the speakers attend to any turn-taking system) lack interconnectedness, relation and direction - are merely a psychotic babbling. Again, given that different speakers may choose to take every twitch and scratch, every palatalized inhalation or vowel nasalization as relevant, there can be no fixed context-free criteria for identifying relevance. The term 'cooperative' is itself too Other-oriented, supportive and helpful; 'relevance' is suitably Janus-faced, like all talk: the degree to which a speaker speaks to his own interests, while maintaining a veneer of grudging attention to the other speaker's assumed interests, is left to be determined. We are returned to Halliday's provocative observation, which heads this essay, so alert to the possibility that varying perceptions of relevance reflect varying faiths in communication itself.
Analysing Fictional Dialogue
223
References Austin, J.L. (1962) How To Do Things With Words. Clarendon Press, Oxford. Brown, P. (1980) How and why are women more polite: some evidence from a Mayan community. In S. McConnell-Ginet (ed.) Women and Language in Literature and Society. Praeger, New York. Brown, P. and S. Levinson (1978) Universals in language usage: politeness phenomena. In E. N. Goody (ed.) Questions and Politeness. Cambridge University Press, Cambridge. Burton, D. (1980) Dialogue and Discourse: A Sociolinguistic Approach to Modern Drama Dialogue and Naturally Occurring Conversation. Routledge and Kegan Paul, London. Carter, R. and D. Burton (eds.) (1982) Literary Text and Language Study. Edward Arnold, London. Coulthard, M. and M. Ashby (1976) A linguistic description of doctor-patient interviews. In M. Wadsworth and D. Robinson (eds.) Studies in Everyday Medical Life. Martin Robertson, London. Giles, H. and R. St. Clair (1979) Language and Social Psychology. Basil Blackwell, Oxford. Goffman, E. (1981) Forms of Talk. University of Pennsylvania Press, Philadelphia. Grice, R.P. (1975) Logic and conversation. In P. Cole and J. Morgan (eds.) Syntax and Semantics 3: Speech Acts, pp. 41-58. Academic Press, New York. Gumperz, J. (1977) Sociocultural knowledge in conversational inference. In M. Saville-Troike (ed.) Linguistics and Anthropology, pp. 191-211. Georgetown University Press, Washington. Halliday, M.A.K. (1967) Grammar, Society and the Noun. University College, London. Harris, R. (1981) The Language Myth. Duckworth, London. Hemingway, E. Cat in the Rain. In The Essential Hemingway. Penguin, Harmondsworth. Hymes, D. (1972) Models of the interaction of language and social life. In J. Gumperz and D. Hymes (eds.) Directions in Sociolinguistics. Holt, Rinehart and Winston, New York. Labov, W. (1970) The study of language in its social context. Studium Generale 23, 66-84. Leech, G. (1983) Principles of Pragmatics. Longman, London. Levinson S. (1983) Pragmatics. Cambridge University Press, Cambridge. Lodge, D. (1984) Small World. Seeker and Warburg, London. Melville, H. Bartleby. In J. Leyda (ed.) The Portable Melville. Penguin, Harmondsworth. Pomerantz, A. (1978) Compliment responses: notes on the cooperation of multiple constraints. In J. Schenkein (ed.) Studies in the Organisation of Conversational Interaction. Academic Press, New York. Pratt, M.L. (1986). The ideology of speech act theory. Poetics Today 7, 59-82. Sacks, H., E. Schegloff. and G. Jefferson (1974) A simplest systematics for the organisation of turntaking in conversation. Language 50, 696-735. Sinclair, J.McH. and M. Coulthard (1975) Towards An Analysis of Discourse. Oxford University Press, London. Taylor, TJ. (1980) Linguistic Theory and Structural Stylistics. Pergamon Press, Oxford. Toolan, M. (1987). Analyzing conversation in fiction: The Christmas dinner scene in Joyce's Portrait of the Artist as a Young Man. Poetics Today 8 (2), 393-416. (Reprinted in R. Carter and P. Simpson (eds.) Language, Discourse and Literature, Unwin Hyman, London, 1989, pp. 195212.) West, C. and Zimmerman, D. (1982) Conversation analysis. In Scherer, K. and P. Ekman (eds.) Handbook of Methods in Nonverbal Behaviour Research. Cambridge University Press, New York.
This page intentionally left blank
PART FIVE LANGUAGE AND WRITING
Although orthodox modern linguistics lays great emphasis on the primacy of speech, it has always been, as integrationists point out, the linguistics of a literate society; and specifically a linguistics of alphabetic literacy. The papers in this section explore a wide range of questions concerning the role of writing in human communication. Ch. 15 ('The Semiology of Textualization') examines the relationships between texts and the physical objects that bear them, arguing that the text cannot be analysed or understood independently of its organization as part of the object. The integrationist treats this phenomenon as an example of mutual contextualization. It is a pervasive feature of the way language is used in contemporary society, where textualization accompanies the making and marketing of virtually all manufactured products. Orthodox linguistics, which regards writing merely as a reflection of speech, has no way of dealing with mutual contextualization of this kind. Ch. 16 ('Analysis and Notation: The Case for a Non-Realist Linguistics') raises the question of whether systems of written notation can be 'true' representations of linguistic structure. This is a crucial issue for any linguistics which draws a basic distinction between analysing linguistic structures and describing them, since the latter is itself a communicational enterprise and usually takes the form of a written text of some kind. Inasmuch as descriptions purport to describe things other than themselves and to be accurate rather than inaccurate, the descriptive linguist seems committed to some form of surrogational semantics (see Ch. 8). The notations used must in some sense 'stand for' linguistic items given independently and antecedently. But exactly what the relation needs to be between the symbols of a written notation and the linguistic structure described thereby is highly problematic. For competing systems of notation abound in modern linguistics, while some are claimed to be simply 'variants' of others. What needs to be recognized is that the creation of
225
226
PART FIVE: LANGUAGE AND WRITING
linguistic notations through writing is itself a process that alters perceptions of language. Ch. 17 ('How Does Writing Restructure Thought?') addresses a question much discussed in cultural psychology. It is commonly claimed that societies where writing is well established have a different mental outlook on the world from pre-literate communities. But exactly what this difference consists in remains elusive. This paper proposes that the key to it resides in the fact that writing introduces a level of verbal conceptualization which detaches words from their human sponsors. A written text exists as a continuing cultural artefact in its own right, irrespective of who originally wrote it (which may be unknown). Thus for a literate society the pre-literate equation of language with speech is destroyed. For speech now becomes integrated with a biomechanically different mode of communication, which depends on visual rather than auditory signs. Ch. 18 ('Writing and Proto-Writing: From Sign to Metasign') considers the question of the origin of writing in human culture and argues that one of the important cognitive prerequisites for the emergence of wnting is a grasp of the possibility of 'synoptic equivalences' between visual signs. A single sign can thus function as a metasign for a series of other signs. Unless this resource had been available, the development of writing as we know it could never have taken place. There are two basic types of visual sign that can be used for recording information, 'emblems' and 'tokens', neither of which presupposes linearity. The key question is: how does synoptic reduction arise naturally in the course of visual communication? It is suggested that this may well have taken place very early on in human culture (Upper Paleolithic period) at the stage of 'proto-writing', where there is evidence of the repeated re-use of the same surface for a succession of visual messages superimposed on one another.
15
The Semiology of Textualization ROY HARRIS
By a text is meant, for present purposes, an inscription, carving or similar configuration of visible marks in some notation conventionally taken as representing words, letters, or other linguistic elements. It does not include oral linguistic material of any kind. A text thus defined usually requires a surface of some kind on which to appear, in contradistinction to spoken language, which merely requires a physical medium, such as air, for transmission. This requirement perhaps suggests a possible division of the procedures involved in the production of texts into two broad classes, depending on whether the surface and the materials used to mark it are man-made or not. It is perfectly possible to scratch a text on a natural surface, such as a rock face, by using a suitably shaped natural object, such as a stone or flint. On the other hand, in many civilizations we know of, a characteristic form of progress has involved the development of specially prepared surfaces (such as parchment or paper or wax tablets) designed specifically to receive texts, and likewise the development of special instruments (such as the pen and the stylus) to mark those surfaces with a text. Over the course of time, more primitive techniques of textualization tend to survive alongside more sophisticated ones. Modern civilization shows no sign as yet of the abandonment of graffiti as a popular form of protest in urban environments. It is, indeed, not only popular but communicationally effective. The point was put rather neatly a few years ago by a message chalked, amid many others, on the outside wall of Balliol College in Oxford at a time of widespread student discontent. It said simply: These walls have a greater circulation than Cherwell' (Cherwell was the name of the Oxford student newspaper). The inscription of graffiti on public buildings always seems to have been a recognized form of self-expression, at least in Europe; and in certain cases modern scholarship owes a great deal to the evidence it has provided. (The graffiti of Pompeii would be a case in point.)
227
228
Roy Harris
II Texts are sometimes relegated to a relatively minor role in the artefact of which they form part. On the other hand, there are cases in which the presence of the text may be regarded as a primary reason for the construction of the artefact in the first place. A wide variety of possible relationships between the two must be allowed for. When Shah Jahan built the Taj Mahal, the best craftsmen were recruited for the great work from all over India and Central Asia. They included, we are told, not only master masons and journey-men, stone-hewers, inlay workers and goldsmiths, but also calligraphers. Why did Shah Jahan need to employ calligraphers to build a mausoleum? Their expertise was required for the drafting and execution of the Koranic texts which were to appear on its walls. The work of Shah Jahan's calligraphers can still be admired today. The verses are picked out in coloured stones set in a white marble ground - not perhaps the commonest form of calligraphy. But these texts have a carefully planned and integral place not only in the conception of the Taj Mahal as a monument but in its physical structure as a building. As texts deliberately inserted into a predominantly non-textual environment of great aesthetic merit, they presumably qualify semiologically as what have been called 'semantic enclaves' (Wallis 1973). The term semantic enclave is defined by Wallis as 'that part of a work of art which consists of signs of a different kind or from a different system than the signs of which the main body of that work of art consists'. Thus, on Wallis's definition, the French passages in the Russian text of Tolstoy's War and Peace would be semantic enclaves. The words 'Johannes de eyck fuit hie 71434' in the portrait of the Arnolfini would be a semantic enclave. The balloon saying 'I wish Batman were here' issuing from the heroine's mouth in the comic strip would be a semantic enclave. Wallis's particular concern is with inscriptions in paintings. He makes it clear that by speaking of such inscriptions as 'semantic enclaves' he wishes to emphasize that 'these enclaves are autonomous entities within those paintings in which they occur, that they have a different semantic structure.' The term semantic enclave thus defined raises a number of awkward questions. One is 'What counts as a work of art?' A second closely related to it is 'Why is it necessary to restrict the concept of a "semantic enclave" to works of art in any case?' It is difficult to see how, for example, Tolstoy's use of French in War and Peace is intrinsically different from, say, a bilingual speaker's use of both Russian and French in the course of conversation. Communication of all kinds presents us with examples of semantic enclaves, irrespective of whether or not they occur in what is judged to be a 'work of art'. Problems of a different order are raised by the term enclave itself, which seems to take it for granted that it will in any given instance be clear what is included in what. The part-whole relation is notoriously difficult where different semiologies are concerned. For example, the front page of The Times on 15 July 1982 featured a picture of a middle-aged man extending a clenched fist, below which appeared the caption: 'Sir Peter Parker: "ready for long siege'". Now it is clear enough that in some sense the picture of Sir Peter Parker was part of that issue of The Times. But was it an 'enclave' ? And, if so, by what criteria? Does one assume that because in
The Semiology of Textualization
229
any issue more square inches of The Times are occupied by newsprint than by reproduction of photographs, this automatically relegates any photograph in The Times to 'enclave' status? But then how does that in turn affect the status of the caption? Is the caption, as a mere appendage to the photograph, on that account an enclave in an item predominantly pictorial? How, in that case, would one count the quoted words 'ready for long siege', which do not make much sense except by reference to the accompanying front-page article which reported Sir Peter Parker's decision to close the British Rail network in response to a strike by train drivers. To put the problem in more general terms, how does one decide whether a picture supplements a text, or a text supplements a picture, or whether both are equal partners in a complementary relationship, or whether each independently complements the message contained in some third item? To appreciate the complexity of what Wallis describes as 'forms of co-operation of image and writing' it suffices to draw attention to the fact that very often textualized artefacts presuppose - and, indeed, epitomize - whole episodes of social history. For example, the text on a recently issued British postage stamp reads simply '14p' and '29 July 1981'. The first of these legends appears in the top right-hand corner immediately beneath a profile silhouette of a woman's head. The second appears in the bottom left-hand corner, reading vertically instead of horizontally, and is superimposed upon a head-and-shoulders picture of a smiling young man accompanied by a smiling young woman. The semiology of all this is extremely recherche but there can hardly be any doubt that we are expected to recognize, for instance, the fact that the young man is the eldest son of the lady whose profile appears in the top right-hand corner, that the young woman is this man's bride, and that 29 July 1981 was the day on which they were married. How this in turn relates to the function of postage stamps is another story again. Or, to take a quite different example, the design of a well-known cigarette packet features a distinctly maritime iconography, with a stylized sailing ship on one side of the packet and an anchor on the reverse side. A band of gold on navy blue running round the top of the packet echoes a motif found in the design of jackets of officers of the Royal Navy. The most prominent part of the text on the packet features the words 'SENIOR SERVICE' printed in navy blue. What all this has to do with the function of the packet as a container would be quite obscure to anyone who knew nothing about the social stereotype presented by the naval officer in the context of early twentieth-century British culture, the paradigms of marketing psychology in the tobacco industry, the existence of other naval brand images, and so on. It is doubtless a far cry from designing an Islamic mausoleum in the seventeenth century to designing such things as postage stamps and cigarette packets in the twentieth. But postage stamps and cigarette packets have in common with the Taj Mahal at least two things: first, they are products of human design, having been constructed for a specific purpose, and secondly, they are textualized. That is to say, part of their design function is reflected in the linguistic material which is incorporated into their structure. The text on the postage stamp will usually indicate the postal tariff for which it is valid, sometimes the country of origin, and sometimes other information as well. The text on the packet of cigarettes will usually indicate the brand and the maker. The mausoleum, the postage stamp and the cigarette packet serve very
230
Roy Harris
different purposes in human affairs, but in all three cases textualization has been utilized as one of the design resources available to the designer. Nowadays this resource is exploited on an ever-increasing scale. It is coming to be increasingly difficult in Europe, the United States, and many other countries to go into a shop and buy manufactured articles which do not, somewhere or other, have a text of some kind printed on them or attached to them. It may be only the manufacturer's name, or the country of origin. Or it may be a text of a much more extended and complex kind. For a variety of reasons, it has come to be a normal practice that manufactured objects bear a verbal message or messages which, it is assumed, the purchaser or user should be able to understand. This reliance on a textually competent public in the marketing of consumer goods goes hand-in-hand with a parallel reliance on the management of social order through the use of texts of all kinds. They range from signs which say 'No Smoking' to census and income-tax forms. We live to an ever-increasing extent in a world which takes for granted a social competence organized by means of textualized artefacts of many different varieties, and the ubiquity of these textualized artefacts is one of the most conspicuous features of modern life. The twentieth-century expansion of the role which texts of all kinds play in people's lives is commonly associated, and rightly so, with the dramatic increase in literacy brought about by modern systems of education. But it would be a mistake to equate the communicational proficiency which textualization demands simply with the ability to read and write. The semiology of textualization goes - and always has gone - far beyond those basic skills. To take a simple illustration of this point, it is not sufficient to have enough Arabic to be able to read the words inlaid on the walls of the Taj Mahal. It is at least as important to grasp that these are quotations from a sacred work and to know which the work in question is. Anyone who failed to understand that would fail fully to understand the communicative purpose of the texts and their role in the design of Shah Jahan's building. Furthermore, our modern concept of literacy is essentially the product of a certain educational tradition, which already incorporates certain attitudes towards the written word, its function, and its social value. Seen in a broader historical and anthropological perspective, literacy involves no more than a subset of the kinds of practical knowledge, skills and training which textualization in a given culture may take for granted. Textualization as such is not parasitic upon literacy. On the contrary, literacy in our modern sense is an outgrowth from the development of one particular set of techniques employed in textualization. In textualization we invariably find combined what are usually regarded as two separate but distinctively human achievements: language and tool-making. Textualization integrates graphic surrogates for elements of speech with the design of man-made objects serving a variety of social purposes. But in different forms of textualization the relationships between text and artefact are different. Writing originated as the systematization of a highly flexible technique of textualization adapted to the individual requirements of particular languages and language communities. It was for centuries, until the advent of modern reprographic techniques, the most artefact-neutral variety of textualization, as well as being the most highly specialized.
The Semiology of Textualization
231
The different types of relationship between text and artefact are worth careful examination, because they can tell us a great deal about what is presupposed concerning the communicational proficiency communities take for granted. Such a study must recognize the fact that textualization in any form could not exist unless communication extended beyond mere mastery of languages. It would be putting the cart before the horse to think that the kind of proficiency textualization involves is simply an offshoot or corollary of anyone's knowledge of English, or of Japanese, or of any other language. Nor, on the other hand, does it seem plausible to trace this proficiency to any specific genetically endowed capacity. For the particular cultural practices associated with different forms of textualization cover an enormously disparate range. We find that typically different behavioural routines tend to be associated with different types of textualized artefact. For instance, to take a simple case, no one on buying a book peels off the dustjacket, screws the dust-jacket into a ball, and throws it into the nearest wastepaper basket. Yet that is quite a common form of behaviour with certain other kinds of protective wrapping: for example, with paper bags and packaging of various kinds. But one does not do that with book jackets, at least not in contemporary Western culture. Anyone who did that would immediately mark himself out as coming from an alien culture: quite literally, we would interpret this as an indication that he was not used to books, that he did not really understand what kind of object he was dealing with. The implications of communicational proficiency for normal patterns of behaviour extend much further than one might without reflexion suppose. It would be possible to give many, many other examples, ranging from the routines viewers use in operating a television set to the routines motorists use in following road signs. If you come to a road junction, and you find a fingerpost, it is no use being able to read the word Oxford on the arm, unless you also know what a fingerpost is. A Martian invader who could read English tolerably well but knew nothing at all about the organization of traffic in England might conceivably suppose that the post was some kind of shrine dedicated to a deity called Oxford, or that it marked the grave of a local worthy of that name, or any other of a hundred and one logically entertainable hypotheses, which just happen to be totally wrong. In short, it is quite literally part of one's communicational proficiency to be able to recognize a signpost as a signpost, which involves knowing the appropriate behavioural routines associated with that object in its normal cultural setting. Modern linguistic theory, however, does not deal with things like signposts. It deals only with the linguistic entities which might or might not be represented on a signpost. In short, it is committed to divorcing the text from the object. By implication, it dismisses the object as irrelevant to the identification of the linguistic sign. Whereas the plain fact is that anyone who does know what a fingerpost is is also in a position to identify the word Oxford on the arm of the post as one particular kind of linguistic sign, a place name. This immediately supplies a more plausible explanation as to how we are able - and indeed expected - to decode the signpost message than supposing, for instance, that it is a question of treating the single word Oxford as an elliptical surface form actually representing the deep structure of some such sentence as This is the road to Oxford, or even perhaps If you want to get to Oxford, follow the direction in which this arm of the post is pointing.
232
Roy Harris
Nor can current speech-act theory give a very convincing account of the function of signposts. It is certainly true that any hypothetical dunderhead who did not know what a signpost was, and hence failed to grasp why the word Oxford appeared on the arm, would have failed to grasp the illocutionary force of the message. Specifically, he would have failed to identify it as a direction, even though he might know that Oxford is the name of a certain university city. But, in terms of speech-act theory, this would be simply on a par with failing to take the words 'This bull is dangerous' as a warning (Austin 1962: 62). There is an important difference, in other words, between the kind of relationship speech-act theory attempts to deal with and the kind of relationship which is central to textualization as such. This we shall come to shortly. First, however, let us note the general reason why analyses of the kind offered by linguists and speech-act theorists are likely to throw little if any light upon the semiology of textualization. As far as linguistics is concerned, the existence of textualized artefacts is merely incidental. There is nothing special about them or about the way they function as signs. Current speech-act theory, like modern linguistic theory as a whole, simply disregards as irrelevant to its concerns the specific physical forms in which a text might be instantiated. The possible varieties of those forms are classed alike as tokens of the same type, and it is the postulated type in question which is taken to be the true object for analysis. This amounts to dismissing from consideration a whole range of factors which can be of semiotic relevance and which are, moreover, regularly exploited in the design of textualized artefacts. For this reason alone, a different approach becomes necessary if we wish to come to terms with what is unique about textualization as communication. A theory of textualization must accommodate from the outset the fact that in textualized artefacts the text always functions as a sign in ways which are not exhaustively described by giving a merely 'linguistic' account of what the text says, of the linguistic forms used, and of their meanings as contrasted with other forms available in the language. Linguistic considerations in this narrow sense are important; but it is no less important to consider what kind of object the artefact is. For that is a consideration which cannot fail to have been foremost in the mind of whoever designed the artefact and decided on its textualization in the first place. In short, a theory of textualization must assume that language and object are, as it were, potentially equipollent in the communicational enterprise. Otherwise, the text would be superfluous or the object superfluous. A text, as a graphic configuration of some kind, has to appear on or in something. And recognizing what that something is must be assumed to be as potentially relevant to the message as recognizing the linguistic form of particular marks which appear there. One reason why it is easy to lose sight of this fact must be mentioned here. Western education has conferred supreme importance upon one particular variety of textualized artefact: the book. As a result, educated people commonly fail to see that the book represents a by no means archetypal form of textualization. In the case of books, the primary purpose of the artefact is merely to act as a vehicle for the text. Since the presence of the vehicle is intellectually discounted by the Western reader, to the point of virtually being ignored, there is no incentive to analyse in any detail the
The Semiology of Textualization
233
rather complex relations which obtain between text and artefact. It remains nonetheless true that it is impossible to comprehend the text unless one knows how to use the artefact, and part of our communicational proficiency vis-a-vis books consists in understanding the mechanisms of this particular form of textualization. For example, we have to understand the convention whereby text follows on consecutively in the standard Western book from bottom recto to top verso of the same sheet of paper. That is not a convention which normally applies to a newspaper. So anyone who tried to read a newspaper according to the convention for books would find it extremely difficult to make sense of what he read. It is part of our communicational proficiency as educated readers to grasp this and many other differences between books and newspapers as textualized artefacts. Furthermore, we have to understand that books as such do not employ any uniform set of principles governing the organization of a text even within the same language community. That is another reason why textual competence is not to be equated with mastery of any language in the usual pedagogical sense. For instance, anyone who supposed that the same principles applied to an English telephone directory as to a novel by Charles Dickens might find it almost impossible to cope with either one text or the other. It is part of the communicational proficiency of someone who can use a telephone directory to grasp, for instance, the fact that each entry presents a self-contained item of information which does not have to be related to its immediate successor. Whereas anyone who supposed that this applied to the individual sentences of a novel would automatically fail to understand even the first page of Oliver Twist. Books themselves, moreover, do not invariably exemplify the general relationship whereby priority is assigned to the text as 'information', and non-priority to the object as 'vehicle'. Paradoxically, books are also among the most striking examples of a relationship of quite the reverse kind: that is to say, where the text is, as it were, simply a vehicle for the object. We see the recognition of this relationship carried to what many would regard as an absurd degree in those examples of modern fine printing which remain forever unread, because to open the pages would be automatically to lower the market value of a collector's item. It would be naive to think that such books happen fortuitously to become collectors' items. On the contrary, they are for the most part designed and produced to be that. This is simply one illustration of a more general point. The study of textualization requires methods which will enable us to analyse these rather complex and varied relations between text and object: but not as relations constituting a separate semiotic organization for each case or type of case. To treat the integration of text and object as a closed system would be just as misguided as to treat the text alone as a structure isolable from that of the object. In both instances, this would be to deny the most important organizing principle operative in textualization - namely, that the text serves communicational needs which relate to the cultural function of the object and are in that sense external to it. But such needs, it seems, vary unpredictably across cultures. On what general basis, then, might an analysis of textualization proceed? The next section of this chapter will attempt to sketch the outlines of an answer to this question.
234
Roy Harris
III Broadly speaking, it seems that we need an approach to semiological facts which distinguishes three scales. These three scales correspond to three possible perspectives on communication. I shall call them 'biomechanical', 'macrosocial' and 'integrational'.i The biomechanical scale is concerned with investigation of phenomena which pertain primarily to the organic and neuro-physiological mechanisms which underlie communicative behaviour. Traditional articulatory phonetics, for example, is a typical biomechanical investigation in this sense. Biomechanical phenomena are often phenomena which escape the conscious analytic attention of those engaged in communication, even though the participants may have finely tuned control of and reactions to the various processes involved. Few people apart from phoneticians, for example, are consciously aware of what precisely they have to do with their vocal apparatus in order to make the requisite distinctions between the sounds their speech utilizes: but they are quite expert at pronouncing the sounds nonetheless. They have acquired the requisite biomechanical proficiency. The macrosocial scale is concerned with phenomena which, again, participants may be not consciously or only dimly aware of. But in this case their lack of awareness will relate to the fact that as participants they are not in any position to grasp a much broader pattern, of which their own communicational behaviour forms only part. Linguistic change, as traditionally studied, is a typically macrosocial phenomenon. Only certain aspects of it ever enter into the ordinary individual's experience of language, even though their own behaviour is instrumental in bringing such change about. Macrosocial features of communication show up clearly only when communicative behaviour is studied en masse. Nonetheless, macrosocial proficiency of certain kinds is important if individuals are to achieve their communicational purposes, and insensitivity to macrosocial features can give rise to serious communicational breakdowns. Individuals may simply fail to realize, for example, that some feature of their communicational behaviour is unacceptable or liable to misinterpretation once they move outside their familiar communicational environment into situations where they are in contact with individuals of a different racial or social background. The integrational scale is concerned with the fitting together of various patterns and features of communicational behaviour in ways which make sense to the participants. The articulation of events in a particular communication situation or type of situation cannot be structured randomly, or without due regard for the physical and social conditions obtaining. Here, one is dealing with the results of biomechanical and macrosocial proficiency used to further communicational objectives. Studies on the integrational scale are concerned essentially with how people have to integrate various forms of proficiency in order to achieve interactional aims in given situations. It is, in short, the study of communicational proficiency, envisaged as a semiologically focused hermeneutics. As such it shares in common with all hermeneutic approaches a concern to explain human actions and reactions primarily by reference to the relevance they have for the participants themselves. 1 I now think it preferable to call the third scale 'circumstantial' and reserve 'integrational' as a general term to designate this approach as a whole.
The Semiology of Textualization
235
These three scales correspond to three different perspectives not only on communication but on all forms of human behaviour. The assumption is that communication involves activity of some kind by the participants, and this activity like any other - is constrained in various ways by factors of three different kinds. They are: what the human being is physiologically equipped to do, what the human being is collectively conditioned to do, and what the human being is individually aiming to do in given circumstances. Now, any textualized artefact may be analysed in terms of the type of communicational proficiency it presupposes, and this can conveniently be described by reference to the three scales just mentioned. Such an analysis will provide, in effect, a systematic account of the communicational relevance of the various features of the text, and their relations to the object in question. For example, if I go into any supermarket or grocery store in Oxford and buy a carton of milk, I am likely to find that the textualization of this object conforms to certain fairly stringent constraints along the three scales we are considering. First, the macrosocial scale. I am likely to find that the text on the carton is in what I was brought up to call 'English'. This relates very directly to the marketing intentions of the manufacturers. It may well be that their product is manufactured abroad, in a non-English-speaking country. But the language selected for textualization will be a consumer-oriented choice. If they choose Hungarian or Eskimo for the text on the carton, it is very likely to remain a long time unsold in an Oxford shop: for a very good reason. The chances of an Oxford purchaser being able to understand what the carton contains if the text is in Hungarian or Eskimo are very much reduced. If he is not sure what the carton contains, he is less likely to buy it. If he doesn't buy it, the shop-keeper will not want to stock it. The manufacturer will lose money. In short, here we have a case where textualization is directly related on the macrosocial scale to the economic and marketing structures of the society in question. Next, the biomechanical scale. I am likely to find that the text on my carton comes in print of various sizes, and these sizes are fairly directly related to biomechanical factors. The fact is that the human eye cannot easily scan printed characters of less than a certain size, which depends on variables in the relevant viewing conditions. This has nothing to do with language conventions. It is a biomechanical matter, of the same order as the factors which the optician takes into account when testing people's eyesight. Just as the manufacturer cannot afford to textualize his carton with words the customers cannot comprehend, equally he cannot afford to use words the customers cannot read. However, what they can read depends to some extent on the distance they are from the text, and whether or not they are standing still. Hence, it is quite typical on a textualized carton to find text in different sizes of print, some of which can be read at a distance or while walking by the object, whereas other sizes presuppose that one is holding the object steady in one's hand; that is to say at a distance from eye to object of twelve or eighteen inches. Now in order to understand the reason for that, one has to move to a third scale of analysis: the integrational. The reason for having different sizes of print on the object in question relates to integrational factors arising from the type of communication situation in which the text is expected to function. At the crudest level of analysis, one can distinguish between two types of function which the words on the carton might be expected to
236
Roy Harris
have. I will call these two types of function 'identificational' and 'amplificational'. The identificational function will be that of telling you what the carton contains. The amplificational function will be that of telling you certain facts about the contents thus identified. The point to note there is that there is no macrosocial or biomechanical distinction involved between the two functions: we are dealing here with purely integrational matters. And this involves asking questions like 'How does a member of late twentieth-century European civilization set about the task of finding out which of the various cartons, bottles, tins and packets he might be confronted with contain the contents he is interested in?' The answers to such questions turn out to be extremely complex. But they are essential to giving any realistic account of communicational proficiency in a culture such as ours. As regards the size of print on modern milk cartons, two of the relevant integrational considerations are these. The purchaser will often be confronted with the task of picking out visually which of a large number of roughly similar cartons or bottles on the shelf or tray in a shop display contain the type of liquid he wants. It is only when that initial problem of identification has been solved that there is any point in moving on to subsidiary questions, such as how much it costs, what quantity the carton contains, how long it will keep, and so on. So there is an integrational priority in the order of questions to be asked from the customer's point of view. But that would not per se explain the different type sizes in the carton text. The crucial factor is that in the relevant communication situation the customer often seeks to answer the first question while moving around the shop on a general tour of inspection. He does not stop immediately upon entering and start picking up the objects nearest to him in order to read the labels. If he did, shopping would take up so much time that there would be nothing left for any other form of activity during the average 24 hours. It follows that, given the biomechanical constraints on legibility, one might expect the largest size of print to be allocated, all other things being equal, to that portion of the text which answers the question which has integrational priority for the customer. Casual observation suggests that this expectation is, if not invariably, at least very commonly fulfilled. Again, the reason is quite directly related to the economics of communication between producer and consumer. If the customer looking for milk in a shop cannot quickly identify the producer's carton as one containing milk, then he is less likely to buy it. So there is a convergence between the communicational interests of the producer and the communicational interests of the consumer. This maximizes the effectiveness of having part of the text picked out typographically in a manner determined by the biomechanics of reading under certain situationally determined conditions. On the other hand, print size is not the only relevant factor. Lettering shapes and colour contrasts may play a part in giving prominence to a certain portion of the text. Nor is it inevitable that the identificational function should be allocated primarily to the text at all. Pictorial or iconic representations of various kinds may to a large extent reduce the identificational function of the text to an ancillary role. The total design of the carton as communication will depend on the interplay of a variety of considerations, which will in turn depend on how the particular product is marketed, and what consequential importance its textualization has. There is no way in which the textualization of the artefact can be explained in isolation, and hence no way in
The Semiology of Textualization
237
which the communicational proficiency implied in that textualization can be explained in terms of 'purely linguistic' factors, even if we could separate out the 'purely linguistic' factors from others.
IV
I should like to draw attention to a number of consequences and implications which this approach to textualization has for semiological analysis in general, including linguistic analysis. Before doing so, it is perhaps worth re-emphasizing that the scalar distinctions we are drawing are quite general ones They apply, for instance, no less to the textualization of a milk carton than to the textualization of the Taj Mahal. In fact, if we compare the milk carton to the Taj Mahal, the parallels are striking. On the macrosocial scale, Shah Jahan had no option but to have his texts in Arabic, given the political and religious structures prevailing in seventeenth-century Agra. The choice of any other language would have been tantamount to a declaration of unorthodoxy of a very serious kind. On the biomechanical scale, Shah Jahan's designers showed that they were perfectly well aware of the fact that legibility is in part dependent on viewing distance. They turned this to their advantage by making the text serve a dual function in the design of the fa?ade. As the visitor approaches the front of the mausoleum, the eye cannot at first distinguish bands of script from bands of abstract floral patterns. This biomechanical fact allows both to be used decoratively. They function visually to contrast at a distance with the brilliant white of the marble walls, and pick out certain architectural features and proportions. As the visitor draws closer, however, he begins to see that some of those bands are not merely decorative. From the garden below the main platform on which the mausoleum stands he already begins to identify the principal text which frames the central porch. The word 'identify' is important, for here we move from biomechanical to integrational considerations. The visitor is manifestly intended to identify the text, rather than to read it. Its sheer length, calligraphic complexity and disposition pose severe problems for the eye of a reader, whereas casual scanning by the eye of a visitor thoroughly familiar with the Koran will suffice to identify the well known passage in question. Furthermore, anyone not familiar with the Koran would in any case be perplexed to understand why the words in question appear there at all.2 They make no obvious reference to death, mortality or grief, as one might perhaps expect of an inscription on a mausoleum. On the contrary, they speak of 'warning a people whose fathers were not warned and therefore lived in heedlessness' and of sentencing them to live in unbelief. The relevance of the text cannot fail to baffle anyone who does not know that these are the opening verses of the famous sura 36, Ya Sin, reputedly described by Muhammad himself as 'the heart of the Koran', and traditionally the text recited to the dying throughout the Muslim world. Thus the transition by which, to the visitor's eye on approaching the mausoleum, what began as a delicately decorative 2 For comments on the Koranic text I am indebted to my colleague D.S. Richards of the Oriental Institute, Oxford.
238
Roy Harris
ornamentation gradually resolves itself into the words of Ya Sin, is itself an architect's metaphorical expression of the transition from the awareness of visual beauty to the awareness of imminent death. That metaphor is articulated on the integrational level with a subtlety that marks the work of a great artist. To treat such a text as forming a 'semantic enclave' within the architectural semiology of the building as a whole is theoretically unsatisfactory for a number of reasons which should already, in the light of the foregoing discussion, be evident. One might perhaps summarize some of these reasons by making the general point that it is absurd to suppose that elements from different semiotic systems can simply be juxtaposed randomly and interpreted separately as independent sets of signs. That would be as fatuous as trying to analyse code-switching in a bilingual conversation on the assumption that what was being said in one language must make sense independently of what was being said in the other. In short the notion of an 'enclave' in the manner defined by Wallis is inadequate precisely because it implies that what is contained within the enclave is, as Wallis puts it, an 'autonomous' entity. A second general point is that by treating a passage of text as autonomous, we often find ourselves at a loss to explain quite conspicuous features of its organization. For example, there would be no explanation available for the different sizes of print on a milk carton. Nor can we hope to explain why a Koranic text happens to be written so as to form three adjacent sides of a rectangular panel unless our analysis can somehow take into account how and why the text is designed to outline the doorway of a building. Divorcing the text from the object invariably leaves us in a theoretical impasse. For then we must either ignore certain quite deliberate features of textual organization as being semiologically irrelevant; or else we must seek their semiological relevance in a system which does not belong to the text per se, but adventitiously uses textual features for quite different purposes. Neither of these unwelcome alternatives comes to grips convincingly with the fact that textualization itself is a process which physically integrates word and object. This integration is not an occasional consequence but an intrinsic feature of the phenomenon. If, on the other hand, we adopt an analysis which distinguishes from the outset between biomechanical, macrosocial and integrational factors as having potential relevance at all levels of semiological organization, there is no such theoretical impasse. What this means, however, should not be underestimated. It means that we are committed theoretically to two important propositions. These two propositions stand in approximately the same relation to semiology as Saussure's famous principles of linearity and arbitrariness stand to Saussurean linguistics. They are the basic axioms of an integrational semiology. The first is that what constitutes a sign is not given independently of the situation in which it occurs or of its material manifestation in that situation. The second is that the value of a sign is a function of the integrational proficiency which its identification and interpretation presuppose. Acceptance of these two axioms has very far-reaching consequences for semiology which can only be briefly indicated here. Perhaps the most obvious is that it follows that there is an important sense in which the sign is not arbitrary, and the investigation of this non-arbitrary domain automatically becomes one of the major tasks of an integrational semiology.
The Semiology of Textualization
239
Secondly, it follows that for an integrational semiology signs, as part of human behaviour, cannot be considered in abstraction from the specific ways in which they are involved in communicational interaction. In other words, there is no priority of signs over the use of signs. Sign behaviour as such cannot be treated simply as the exercise of individual choice from among a pre-determined inventory or system of signs. On the contrary, the status of being a sign is itself relative to a communication situation and determined by relevant features of that situation. Signs, in brief, are defined for an integrational semiology by communicational relevance in a situation and not by criteria for membership of some previously established typology. What makes textualization a practice of particular interest for an integrational semiology is the fact that it documents in a remarkably detailed and concrete way the situational factors which communities are prepared to treat as relevant in conferring sign status upon an arrangement or configuration of visual elements. In spite of the great diversity of textualization, what unites its many forms and justifies treating textualization semiologically as a single modality is the following fact. In all cases, it is the physical relationship guaranteed by the material construction of a particular artefact which is used as the basis for establishing semiotic significance. The design of textualized artefacts makes intrinsic use of that physical relationship, and that is why no theoretically satisfactory analysis of textualization can afford to neglect the integration of text and object. There is a formal and functional interdependence between the two which determines the whole gamut of semiotic possibilities which can be realized in any particular instance. Semiologically, that physical relationship is translated into one of complementary contextualization: the text is automatically contextualized by the artefact at the same time as the artefact is by the text. Neither contextualization is - or could be - prior to the other, since both are manifested in one and the same physical realization. Seeing and exploiting the communicational possibilities of that relationship was one of the greatest intellectual steps Homo sapiens ever took. Textualized artefacts are, in this sense, an entirely original creation in the history of man's development as a maker of his own environment. Considered from the point of view of an integrational semiology, textualization offers a rich field of research which has largely been neglected by linguists and others, because their own theoretical stance obscures the existence of this field and access to it. As a result, we still have no good descriptions of what people are expected to know and be able to do in order to achieve average communicational proficiency in any one of the hundreds of different areas of standard communicative behaviour involving texts. Yet such descriptions would seem indispensable if contemporary programmes of education, for instance, are to be saved from falling increasingly out of touch with the linguistic world in which people actually live today. A first step in this direction would be to attempt a classification of textualized objects on the basis of the situations in which they commonly function, the relations exhibited between text and object in each type of case, and the communicational implications involved. Only in this way can semiology keep abreast of the creativity by which modern society is constantly evolving new forms of communication and forging new semiological links between texts and other symbolic systems.
240
Roy Harris
References Austin, J.L. (1962) How To Do Things With Words, ed. by J.O. Urmson and G.J. Warnock. Oxford University Press, Oxford. Wallis, M. (1973) Inscriptions in painting. Semiotica 9, 1-28.
16
Analysis and Notation: The Case for a NonRealist Linguistics CHRISTOPHER HUTTON
The theological controversies that led to the schism between Protestantism and Catholicism in the Christian church that we know today concerned in part differing symbolic interpretations of the bread and wine offered in communion and wider issues about the possibilities and dangers of representing that which is deemed eternal and holy within the confines of time and space. Reading accounts of these controversies (Pelikan 1971-89), one is struck by the subtlety of many of the theoretical positions taken, a subtlety that is lost in the crude summary often given of the debate as being between the belief that the flesh and blood of Christ is actually 'in' the bread and the wine and the belief that the act of communion is a symbolic eating and drinking. This paper is concerned with a more mundane, but closely related, 'theological' question, that of whether systems of notation are 'true' representations of linguistic structure. The chief concern of this discussion will be with conceptualizations of the relation between notational representations and their 'real world' analogues. Linguists have on the whole maintained implicitly and explicitly a firm distinction between analysis and notation. Such a distinction is clearly implied if we are able to deny the importance of an analysis on the grounds that it is a notational variant of one previously proposed. One could, however, see F. W. Householder's 'hocus-pocus' linguist (as distinct from the 'God's truth linguist') (Householder 1952: 260) as arguing that linguistic analysis involves the creation of a notation-system. Householder's version of the distinction runs as follows: The theory of the 'God's truth' linguists . . . is that language has a structure, and the job of the linguist is (a) to find out what that structure is, and (b) to describe it as clearly, economically, and elegantly as he can, without at any point obscuring the God's truth structure of the language. The hocus-pocus linguist believes . . . that a language (better, a corpus, since we describe only the corpus we know) is a mass of incoherent, formless data, and the job of the linguist is somehow to arrange and organize this mass, imposing on it some sort of structure (which must not, of course, be in any striking or obvious conflict with anything in the data).
Just as one would not normally talk of a notation-system being incorrect, but rather impractical or ill-conceived, so the hocus-pocus linguist denies that a 241
242
Christopher Mutton
description or analysis tells us anything verifiable or falsifiable about the true nature of language. For Householder, the hocus-pocus linguist holds a particular belief about the nature of language, that it is disordered in some sense. According to Sampson, the hocus-pocus linguist asserts the apparently more radical thesis that 'the real nature of language is somehow ineffable', a belief he characterizes as 'strange' (1980: 72). Strange or not, the hocus-pocus linguist holds that we cannot invoke descriptive fidelity as a criterion for choosing between different representations. In the light of this, Sampson questions whether the pure 'hocus-pocusser' would have any incentive to continue in linguistics. It is not totally clear from Householder's and Sampson's respective expositions what the hocus-pocus linguist actually believes, or indeed if there is a single position corresponding to this label. We should note that Householder uses the phrase 'in conflict with the data' in defining the hocus-pocus linguist, implying a more 'realistic' hocus-pocus than that defined by Sampson. Firth, as Sampson reports, was happy to accept the label (1980: 231) and indeed criticizes Malinowski as follows (1968: 154): The word 'utterance' seems to have had an almost hypnotic suggestion of 'reality' which often misleads him [Malinowski] into the dangerous confusion of a theoretical construct with items of experience. The factors or elements of a situation, including the text, are abstractions from experience and are not in any sense embedded in it, except perhaps in an applied scientific sense, in renewal of connection with it.
What we mean in Firth's case by hocus-pocus depends therefore on how we understand 'renewal of connection'. Evidently some referential connection is envisaged between notation-system/analysis and the data. Firth offers a warning against the reification of analytic categories and the assumption that they are somehow to be found 'in' human behaviour. However, the notion of renewal of connection is clearly suggestive of some direct mapping between the linguist's categories and something else given independently of them. Robins in his General Linguistics (1971: 41) discusses three approaches to abstraction which he equates with realism, conceptualism and nominalism respectively. Realism involves the belief that 'the constants abstracted in an analysis, if the analysis is correct, are in some way inherent in the actual material of the language under analysis'. The second view is 'that the linguist's abstractions are also abstracted, albeit unconsciously, by native speakers of the language as the result of their having learned the language in childhood, and are part of the content of the speakers' minds or brains'. This is the conceptualistic or mentalistic version of God's truth. Robins' formulation of the third view is the most interesting. It involves the belief that the linguist's abstractions have no other status than as part of his scientific terminology, and are justified by their utility in stating regularities and making predictions about the forms of utterances, but have no express relevance to the speakers' mental or cerebral states or equipment, though certain correspondences between the two are not ruled out and are prima facie quite probable.
Robins writes (1971: 42) that
Analysis and Notation
243
all three of the standpoints mentioned above possess a degree of validity . . . but the third view is in many ways the easiest to adopt.
Linguistic descriptions should have a predictive and explanatory value . . . different analyses of the same data and different procedures of analysis are to be evaluated and judged in merit against one another by their fulfilling the standards of scientific work, exhaustiveness, consistency and economy, and are in all cases justified by their successful application to further data. (1971: 43)
However, it is evident that Robins, like Firth, cannot be classified as a hocuspocusser. Robins introduces the concept of 'exponent' and the Firthian notion of renewal of connection, then concludes with the following statement (1971: 44): Strictly speaking, in such a scheme of analysis one should not talk of such and such elements or items (literally) 'occurring' in a text or in a language, or of phonological elements as 'found in' or as 'constituting' grammatical or lexical elements. Such forms of statement are, rigidly within the terms of the theory, more accurately made by saying that the abstractions concerned are statable at the different levels and that the abstractions at two or more levels may show specific congruences of exponency. However, provided the simpler and more 'realistic' language of 'occurrence' and the like is recognized as a convenient metaphor and no more, its use is well justified by the resultant convenience and brevity. It will, in fact, be frequently adopted, with the proviso just mentioned, in this book.
Although Robins, following Firth, might seem to embrace some form of hocus-pocus belief about linguistic descriptions, this is severely modified by terminological practice and by the introduction of terms such as 'exponent' and 'renewal of connection'. The Firthian hocus-pocus approach amounts to the observation that to talk of people 'saying phonemes' is a kind of category mistake. This should be distinguished from a hocus-pocus position which would deny that there can be a direct mapping or renewal of connection between the analysis and the data. Firth intended to clarify the referential relation between analysis and data, not to deny that any such relation exists. A more purely hocus-pocus form of argument that the would-be hocus-pocusser might adopt runs as follows. The representation of a sentence in a particular notation offers us a picture. Just as a picture of a horse does not need to be a depiction of any particular horse in order to function as a picture, so the representation of a structure does not need to depict any particular sentence or any particular utterance to function as a representation. A picture of a horse is in some sense no less a horse than a horse running around a field (in the same way that a unicorn is no less an animal than a cat, merely a mythical animal - after all we can just as easily draw a unicorn as a cat) and that similarly, the representation of a sentence is as much an actual sentence as any utterance on a particular occasion. A picture of a horse does not have to be the portrait of a particular horse and we cannot tell by inspection of a picture whether it is or not (unless it says so underneath), and the picture of a sentence does not have to be the portrait of a particular utterance. Linguists produce a specialized form of utterance just as painters produce pictures. To ask what the relation is between the picture and reality is to misunderstand the nature of the picture. The picture of a horse does not stand in any particular relation to a particular horse (unless it is explicitly a portrait),
244
Christopher Mutton
nor to horses or other pictures of horses. The representation of the sentence does not stand in any particular relation to a particular utterance. This is not to say that there is nothing to say about these relationships, rather that there is no way to classify the relationship in the simple 'bi-planar' or dualistic terms of representation and reality. The picture is itself an object and therefore sufficient unto itself in an important sense. This view will be discussed further below. Unlike the hocus-pocus linguist, the 'God's truth' linguist is committed to the belief that there is a correct answer to questions about linguistic structure. However, the role of the notation-system for a realist theory of this kind is far from clear. Are we to conceive of the relation between notation and reality in terms of some semiotic category? Is it, to borrow Peirce's terms, symbolic, indexical or iconic or some combination? Presumably the realist linguist must have a 'surrogationalist' (Harris 1980) notion of what a notation is: the notation must 'stand for' or 'refer to' or 'denote' a structure which exists independently of it and antecedent to it. What kind of surrogationalist is the realist linguist? Rom Harre defines realism as follows (1972: 91): 'if, in addition to verbal reference, an act of demonstration can be performed by which a hypothetical entity is pointed out or indicated, then that entity must be said to exist'. Naming and pointing emerge as the twin foundations for a realist philosophy of science. If we take this at face value, we must assume that terms within the scientific theory ultimately derive their meaning from ostensive definition (Harre's paradigm case is the discovery of viruses). In Peirce's terms, this is a blend of the symbolic with the indexical: symbolic because of the dimension of naming and because of the generalization of the pointing; indexical because pointing on a particular occasion (and its more passive analogue, looking and seeing) involves physical contiguity. Our attention is drawn to some individual thing. Thus the representation of a particular sentence in a notation is the name of a class of objects (utterances) which can be individuated by some kind of pointing. What then is the status of the notational representation of a sentence within a realist philosophy of linguistics? If the notation-system is made up of names which derive their reference from pointing, it follows, given the arbitrary nature of the sign, that we can have two different notations that represent correctly the same sentence. The object associated with the sentence to which ultimate reference is made would be any one of the utterances associated with, to use Lyons's term, the 'system-sentence' (1977: 29). It does not matter whether we represent the sentence by tree-diagrams or by bracketing conventions. As long as one can be translated into the other without residue then they are synonymous names for the same sentence. But where, it might be asked, is the analysis in all this? Can naming and pointing separately or in conjunction be said to constitute analysis? The linguist might answer that a grammatical notation-system can for example express the fact that whole classes of superficially different utterances have the same structure, so that 'The cat saw the dog' and 'The dog saw the cat' would have the same representation within the notation-system. And classes of superficially similar utterances, such as 'Swindling loan-sharks can be dangerous' are in fact to be represented by different structures. Thus the notation-system expresses an analysis and represents, ultimately, objects in the real world which can be pointed to or at least be observed, heard or recorded. Notation-system and analysis are independent, for the same notation-system can be
Analysis and Notation
245
used to express different analyses, and different notation-systems can be used to express the same analysis. A different option open to the realist linguist is to profess the doctrine of the 'true' iconicity of the notation, i.e. to argue that the notation-system offers a picture of the data. Whereas the 'pure' hocus-pocusser as defined above believes in the selfsufficiency of the notation as picture, the iconic realist ascribes the qualities of a portrait to the notation. The representation of a sentence is a portrait of a class of utterances or of one particular utterance. The most prominent articulation of a picture theory of this kind is to be found in Wittgenstein's Tractatus (1977). In order to represent correctly, according to Wittgenstein, the representation must offer a picture of what the world (in this case, linguistic structure) is like. Each atomic element of the notation corresponds to an element of the structure depicted. Hacker gives the following account of this (1986: 60): The elementary proposition is a logical picture of an elementary state of affairs. It is composed of simple unanalysable names having various logico-syntactic forms. The names have a meaning, namely the simple objects in reality to which they are connected . . . Their logicosyntactical forms mirror the metaphysical forms of the objects they stand for.
The iconic realist might argue against the non-iconic ('symbolic' or 'arbitrary') realist that the relation between notation and notated must have a strong pictorial or iconic element. For if this relation is held to be arbitrary, we need some further explanation of how to map the notation onto the object, opening up a regress. To avoid the regress, the linguist must regard the notation as a picture. To achieve the surrogational aim of direct mapping, the analysis as expressed in the notation must show the structure as it really is. The non-iconic realist might object that the representation of a sentence in the notation system of a particular linguistic theory is not a picture of the data, but a representation of the important relations to be found there. A phonemic analysis of a stream of spoken speech is not a picture of the speech, but a statement of certain structural relations. These are necessarily abstract, since the analysis is not isomorphic with the data (if it were, it would not be an analysis). Thus the analysis is abstract, true to the data, and expressed in a non-iconic notation system. The structural relations expressed in the notation-system could be equally well-expressed in any number of trivially distinct notations. It might seem that for the realist the distinction between pictorial and non-pictorial (or iconic and non-iconic) representation breaks down at some point and that this is it. The relation between an S and an NP node that it dominates, indicated by a line joining them in a tree-diagram or by a bracketing relation of some kind, could be seen either as some kind of pictorial representation, or as a conventional symbolization of an abstract relation. The depiction of the relation of S to NP may or may not be seen as pictorial; however it clearly makes no sense to say that the notation-system is a picture of the data if by that is meant that the replacement of the notation NP by some other letter combination involves a change of analysis, a new picture of the data. As long as new notation is translatable directly into the old, then there is no difference between them. In this sense - for the realist - the notation-system is clearly arbitrary,
246
Christopher Mutton
and while it expresses an analysis, it is nonetheless in principle distinct from that analysis. However, the criterion of translatability, simple as it might seem, is somewhat difficult to apply in practice. Is bracketing notation directly equivalent to treediagrams, such that every relation expressed in the one is also expressed in the other? (It could be argued that the fact that generativist linguists use tree-diagrams rather than bracketing notations has directly affected the evolution of generativism. For the treediagram allows domains or relations to be defined visually which in a different notation-system would be less than salient.) What is the relation of writing as a system of notation to phonemic analysis, phonetic transcription (broad and narrow) or to shorthand? Does the fact that we could translate a piece of written English into some form of phonemic analysis (basing our translation on some form of pandialectal phonology) imply that the two writing systems are notational variants? Or is one a notation without an analysis and the other a notation with an analysis? Can we say that two notational systems are notational variants if we can translate from one to the other without performing any kind of analysis? The question of the distinction between a notation-system and analysis needs to be addressed at this point. Is a secretary performing an analysis of a letter when taking it down in shorthand from dictation? The shorthand is a notation, moreover a notation parasitic on the standard alphabetico-lexical writing systems employed in the West. The case could be made nonetheless that the transcription of speech into written form of whatever kind is an analysis, since the secretary must make a decision about what is dictated within the conventions of the notation-system being operated. It is possible for the secretary to get it wrong. Thus it might be argued that in the case of shorthand or of the transcription of speech into written form (e.g. for the purposes of recording a debate in written form), we are dealing with a form of linguistic analysis, albeit one at a fairly low level. Ambiguities, mishearings, problems of completeness (whether to transcribe so-called 'false starts' to utterances) may occur which pose analytical problems. The problem of definition that we are faced with here is a familiar one to philosophers and others concerned with conceptual clarification, i.e. with creating and defending certain distinctions and casting doubt on or denying the importance of others. It is clearly open to a philosophically-minded linguist to argue that the secretary who takes shorthand or the journalist who transcribes a speech is making an analysis. The fact that this analysis does not, once the technique is mastered, require profound reflection on the nature of the utterances to be transcribed can be attributed not to the straightforwardness of the task, but the efficacy of the training that has produced a level of proficiency, and the cultural achievement of the creation and diffusion of writing systems. This might suggest one objection to calling shorthand, or indeed writing, an analysis of English. In one sense it is evident that the analysis is not being done in situ, but, as it were, on the shoulders of the cultural and intellectual circumstances that brought about the creation of writing systems. The writing system offers in the abstract an analysis of English, and the secretary is simply the humble executor of a tradition. The analytic contribution of the secretary is so minimal as to render the term meaningless. The real analysis was done by the Greeks or whoever, and we can simply lie back and enjoy it, just as we can turn on a light or watch television without
Analysis and Notation
247
any grasp of electricity or the principles of the transmission and reception of images. On this analogy, the secretary is simply switching on the writing-system (be it shorthand or ordinary English spelling), and using general cognitive abilities and motor skills in the service of a particular goal. No one would suggest that to switch on the television is to engage in profound analysis. Of course, in some sense everything involves some kind of analysis. I have to analyse my visual field in order to locate the television, and then analyse the television in order to locate the on-off switch. But there is nothing specifically 'televisual' about this analysis, if analysis it be. By analogy we might say, there is nothing linguistic in the secretary's shorthand analysis of the dictated speech. The secretary is using technology, and in this is as far from being a linguistic analyst as someone changing a light-bulb is from being a scientist. The question can then be posed as follows. Are linguists simply turning on the technology, taking dictation in a different kind of shorthand, or are they offering new insights into linguistic structure? Are they notating or analysing? Grammatical theories of most kinds attempt to represent the structure of linguistic types by offering analyses in terms of constituents. These constituent analyses are expressed in hierarchical fashion, with the top node giving the highest (largest, most inclusive) category. This category is sub-divided or rewritten into successively smaller categories, with no necessary (a priori) constraint on the number of levels and the 'size' or 'depth' of the lowest level. All formal theories of grammar share this hierarchical property, irrespective of whether they use tree-diagrams or brackets or some other notation to commit structures to paper. In generativist linguistics and in schools that have developed out of or parallel to generativism, the aim was to make this hierarchical framework as 'powerful' as possible, and to develop constraints of various kinds to ensure that a high level of generality would be achieved (one basic structure would account for a large number of sentences, as in early work on the active-passive relation). The power of the model (its capacity to generate a large number of different structures) would be modified through constraints specified within the notation. The formulations of the distinction between deep and surface structure (and the successors to these notions) opened up the possibility for endless permutation in notation and different forms of rule-constraint: high rule generality (e.g. one rule, 'move alpha', constrained in various ways) with the need for powerful constraints, or many specific rules requiring only very limited 'fine-tuning'. As has been often pointed out, the history of these formal models suggests that simplification achieved in one area of the grammar leads to complications in other areas. Generative linguistics offers a test case for a realist philosophy of linguistics, for it was avowedly hypothetico-deductive and hence driven by a heuristic approach to notation and analysis. The question of falsification has therefore been central to its concerns. Chomsky's commitment to a realist philosophy of science is expressed in the hope that physical structures and processes in the 'mind/brain' will be discovered that reflect the analyses of linguists. (See Chomsky 1980, 1988 for summaries of Chomsky's views, and Hacker 1990 for a critique of the term 'mind/brain'.) But the validation of current linguistic theories, he implies, cannot yet be achieved on this basis, given the lowly state of our understanding of the workings of the brain. Instead, the abstract notion of linguistic competence has served as a heuristic, with
248
Christopher Mutton
grammars aiming to achieve descriptive adequacy in addition to observational adequacy. Verification is deferred, and simultaneously transferred to the linguist's own theoretical postulate, 'competence'. The notation offers a surrogational account of an abstract entity. The spectre of notational variants clearly haunts any model of linguistic analysis that aims for descriptive adequacy. For how can the true description be found within the welter of competing formalisms, all of which could claim to capture the linguistic competence of the speaker-hearer? Chomsky's answer was to invoke a further level of adequacy, so-called 'explanatory adequacy'. The notion of explanatory adequacy was designed to deal with the fact that 'the range of alternative descriptively adequate grammars are notational variants of each other' (Smith and Wilson 1979: 246). The need to propose a further level of adequacy arose for Chomsky because of his belief that the ultimate ontological validation or verification of his theories lay in the longterm achievements of neurophysiology or some similar scientific discipline. This left any theory vulnerable in the short-term to being swamped by notational-variants. Lacking external criteria for verification, Chomsky opted for simplicity as the criterion of explanatory adequacy. Thus, paradoxically, Chomsky's physicalist surrogationalism led to the choice of system-internal (or notation-internal) measures as criteria for distinguishing good analyses from bad. Realism led in effect to a heuristic hocus-pocus. The distinction between notation and analysis was maintained only by drawing on an as yet unrealized validation from neurophysiology. Whereas Chomskian realism has led to a hocus-pocus free-for-all (with the criterion of simplicity offering the only very general constraints on linguistic models), other forms of realism are inevitably compromised by the difficulties of proving or demonstrating the referential success of the theory (i.e. the difficulty of finding the correct notation-system to interpret the first, 'true' notation-system). Pateman (1987a, b), in his advocacy of realism for linguistics, argues that there are real processes and cause and effect relations of a general nature specific to language, but that they are often obscured or cancelled out in the real world by extraneous factors. We cannot reduplicate laboratory conditions, but if we could, we would be able to identify independent cause-effect patterns. There are isolatable linguistic factors causing language change, but for practical and ethical reasons we cannot create the ideal conditions under which they can be observed. This kind of realism is best characterized as Platonic realism, i.e. not realism at all. What Pateman offers is a first cousin of the competence-performance distinction, a distinction that was itself formulated as a retreat from cognitive realism into some form of 'abstractionalism', i.e. Platonism. Realism leads ineluctably to a hocus-pocus heuristic (deferred verification) or to some form of Platonism. Katz's summary of different approaches to linguistic abstraction shows one form of the merger between realism and Platonism (1985: 203): The conceptualist criticized the nominalist for confusing competence and performance: the speaker/listener's knowledge of language with the speech resulting from the exercise of this knowledge. The Platonist criticizes the conceptualist for confusing the speaker/listener's knowledge of the language with the language that the speaker-listener has knowledge of. The nominalist's constraints require faithfulness to the facts of speech; the conceptualist's require
Analysis and Notation
249
faithfulness to the facts of knowledge. Only Platonist constraints require faithfulness to just the facts of language.
The merger is expressed with the postulation of a transcendent set of entities: 'the facts of language'. To paraphrase D. H. Lawrence on Galsworthy, realist theories enter the linguistic temple as iconoclasts, but stay to worship at the altar of idealization. The philosophy of linguistics has a clear choice between some branch of realism and a non-surrogational view of linguistic notations/analyses. If we conceive of linguistic notations as non-surrogational pictures various conclusions follow. Firstly, in some sense, the notation must then be seen as the analysis. It then becomes evident that there are limits to the amount of analytical mileage we can squeeze out of a notation system. Wittgenstein writes (1978: para. 297): Of course, if water boils in a pot, steam comes out of the pot and also pictured steam comes out of the pictured pot. But what if one insisted on saying that there must also be something boiling in the picture of the pot?
There are limits to the analysis of pictures, and no refinement of technique will give us an answer to the (nonsensical) question of whether there is boiling water (or an implied picture of boiling water) in the picture of the boiling pot. There is a strong analogy here with realist linguists who, bewitched by their own notation-systems, have drawn a pot and proceeded to look for the boiling water. The non-surrogational view of the notation does not imply (necessarily) that there exists no linguistic reality outside the notation-system, that the notation-system is fused phenomenologically with its signified, i.e. the language. But it does involve the recognition that any attempt to grasp that extra-notational or alternative reality will involve some form of new notation. English spelling and English phonetic transcriptions of various kinds do not all share some mystical common referent, 'English' or 'the utterances of English'. They offer in part competing and in part complementary pictures of what English is. But any attempt to transcend the notation-systems must involve the creation of some further reification in a notation. The point is that to deny that a picture of a horse is related in some surrogational way to horses is not to deny that there are such things as horses 'in the real world'. It does, however, involve a denial that horses in the real world are transcendent signifieds. For what is being argued here is that the phrase 'horses in the real world' is itself a non-surrogational picture, and is, in this sense, no less a notation than the 'visual' picture. For a realist philosophy of linguistics, analysis involves the pointing to and ('correct') naming of a transcendent set of objects or structures: a bizarre embrace between physicalist realism and Platonism. The objection will surely be made that there are no constraints to be put on linguistic analysis if we conceive of notation systems as non-surrogational pictures, that any old picture will do. This is the argument that is laid against the hocus-pocus view of linguistic description. From the verificationalist's point of view, the argument is of course sound. We cannot say on verifiable grounds which of several linguistic analyses is better than which. But if we deny that the notation offers a direct, surrogational representation of speech, the question of validation of the notation
250
Christopher Mutton
system does not arise in these terrns at all. At present, linguists hold a variety of surrogational theories about analyses while being unable to prove the superiority of one school or methodology over another, which is a self-defeating state of affairs to say the least: verificationalism without verification. It is not that objections on the grounds of lack of verification are wrong, they are besides the point. Recognizing this involves understanding not only the limitations of notation systems, but also of their power in conceptualizing language. Little or no work has been done in exploring new and different kinds of notation-systems qua notation-systems, and linguistics has not strayed very far from the alphabetic notation-system of European languages. The notion that linguistics should be involved in the creation of new writing and notational systems might seem paradoxical, given the avowed aim of modern linguistics to study speech, but it offers one way out of the temple of idealization for the realist and involves the recognition that notation systems do not just create analyses, they also create new forms of language, the most notable example of which are writingsystems themselves (Harris 1986). It also involves the recognition that linguists, no less than economists and sociologists, change the nature of their objects of study. Harris (1989: 104) argues that writing is a technology that 'introduces a level of verbal conceptualization which detaches words from their human sponsors'. Only through writing does it become possible for Socrates to ask questions like 'What is justice?' The existence of writing also makes the sort of questions linguists ask about the referential status of their notation systems possible. But just because you can ask a question, that does not mean you can answer it. One could imagine that the terminology of discourse analysis might become part of everyday metalinguistics, and if 'opening up a closing' or some other technical term of conversational analysis (see the essays in Sudnow 1972) became part of general parlance, it would be true to say, not that discourse analysts had scored a surrogationalist 'bull's eye' with their terminology, but they had changed, in a small way no doubt, the nature of conversation itself. A linguistics that ultimately left language as it found it would be no kind of linguistics at all.
References Chomsky, N. (1980) Rules and Representations. Blackwell, Oxford. (1988) Language and Problems of Knowledge. MIT Press, Cambridge, Mass. Firth, J. R. (1968) Selected Papers of J.R. Firth 1952-1959, ed. F. Palmer. Longman, London. Hacker, P.M.S. (1986) Insight and Illusion. Clarendon Press, Oxford. (1990) Chomsky's problems. Language & Communication 10, 127-48. Harre, R. (1972) The Philosophies of Science: An Introductory Survey. Oxford University Press, Oxford. Harris, R. (1980) The Language-Makers. Duckworth, London. (1986) The Origin of Writing. Duckworth, London. (1989) How does writing restructure thought? Language & Communication 9, 99-106. Householder, F. (1952) Review of Z. Harris, Methods in Structural Linguistics (1951), International Journal of American Linguistics 18, 260-8. Katz, J. (1985) An outline of Platonist grammar. In The Philosophy of Linguistics, ed. J. Katz, pp. 172-203. Oxford University Press, Oxford. Lyons, J. (1977) Semantics, vol. 1. Cambridge University Press, Cambridge.
Analysis and Notation
251
Pateman, T. (1987) Philosophy of linguistics. In New Horizons in Linguistics 2: An Introduction to Contemporary Linguistic Research, ed. J. Lyons, pp. 249-67. Penguin Books, London. Pateman, T. (1987b) Language in Mind and Language in Society: Studies in Linguistic Reproduction. Clarendon Press, Oxford. Pelikan, J. (1971-89) The Christian Tradition: A History of the Development of Doctrine. 5 vols. University of Chicago Press, Chicago. Robins, R.H. (1971) General Linguistics: An Introductory Survey. Longman, London. Sampson, G. (1980) Schools of Linguistics: Competition and Evolution. Hutchinson, London. Smith, N. and D. Wilson (1979) Modern Linguistics: The Results of Chomsky's Revolution. Penguin Books, Harmondsworth. Sudnow, D. (ed.) (1972) Studies in Social Interaction. Free Press of Glencoe, New York. Wittgenstein, L. (1977) Tractatus logico-philosophicus. Logisch-philosophische Abhandlung. Suhrkamp, Frankfurt. (1st edition 1921.) Wittgenstein, L. (1978) Philosophical Investigations. Blackwell, Oxford.
17
How Does Writing Restructure Thought? ROY HARRIS
'Writing' is a concept which has undergone a remarkable transformation in the Western intellectual world over the past 30 years. In the wake of influential work by Derrida (1967a, b), Goody (1968), Havelock (1963, 1976, 1982), McLuhan (1962, 1964), Parry (1971) and others, writing is no longer regarded as a mere substitute for speech, or as a useful way of preserving and transmitting knowledge, but as an active and powerful cultural agency in its own right. What distinguishes the literate from the preliterate society is no longer seen as being the possession of a superior communications technology which has overcome the intrinsic limitations of the spoken word and makes it possible to accumulate records and accounts ad infinitum. These advantages, long recognized within the traditional view of writing, now tend to be regarded as merely incidental and external. According to the, modern view, the essential innovation which writing brings is not a new mode of exchanging and storing information but a new mentality. Propositions such as 'Writing restructures thought' and 'Writing restructures consciousness' (Ong 1982) have become almost cliches in current discussions of literacy. So much so that it may now be timely to pause and consider exactly what truth, if truth it is, these cliches purport to capture. Whatever truth it may be, there can be no doubt that it is often accompanied, and perhaps obscured, by an accretion of misconceptions and fantasies which might be called, to avoid any harsher term, 'romantic'. These include a romantic view of the origin of writing, a romantic view of the role played by the alphabet in the history of writing, and a romantic view of the contribution made by the Greeks to alphabetic writing. To disentangle the thesis that writing restructures thought from these and similar encumbrances is an essential preliminary step. It should be - but probably is not - superfluous to insist at the outset that there are forms of writing which have nothing to do with language at all, let alone with the 'representation of speech'. Those who claim that writing restructures thought, however, usually have in mind verbal as opposed to non-verbal forms of writing, and this restriction will be taken for granted in what follows. The romantic view of the origin of writing is a very ethnocentric view, and specifically a Eurocentric view. It is a view summed up with admirable concision in
252
How Does Writing Restructure Thought?
253
the eighteenth century by Charles Davy, who described it as the communis opinio of the day: Writing, in the earliest ages of the world, was a delineation of the outlines of those things men wanted to remember, rudely graven either upon shells or stones, or marked upon the leaves or bark of trees. (Davy 1772)
In short, writing originated as an evolutionary adaptation of pictorial representation. Thus writing systems which retain vestiges of pictorial figures are more 'primitive', and those which have none more 'advanced'. This thesis, in one version or another, survived as received wisdom among twentieth-century authorities on writing (Cohen 1958; Diringer 1962; Fevrier 1948; Gelb 1963) and was taught as such to students of linguistics (Bloomfield 1935), for it fitted in admirably with the orthodox linguistic doctrine of the 'primacy of speech'. The history of writing was seen accordingly in terms of a universal progression from pictograms to logograms to phonograms, culminating eventually in the alphabet. Given this scenario, writing systems such as Chinese, which had failed to progress to the alphabetic stage, were seen as inferior or 'retarded', an opinion still widely held to this day. As Havelock reminds us, when Boswell objected to describing the Chinese as 'barbarians', Johnson's reply was 'Sir, they have not an alphabet'. (One might perhaps have expected him to add: 'And they have no grammar, Sir, either.') The romantic view of the origin of writing has not gone unchallenged (Harris 1986). But inasmuch as challenging it leads automatically to calling in question the position of the alphabet as the crowning achievement in human attempts to devise writing systems, the challenge is likely to be met with astonished incomprehension. For those who regard the alphabet as 'the constitutional framework of our culture' (Levin 1987: 114) anyone who appears to undervalue alphabetic writing is likely to be seen as someone who 'neither appreciates nor even understands the intellectual achievement that went into it' (Levin 1987: 114). According to Levin, an acquaintance with Latin, Greek or Hebrew alphabetic writing opens up for the sophisticated reader endless vistas of unique audio-visual aesthetic pleasure, of being able 'to drink in and relish, hour after hour, page after page, how neatly, how beautifully each of these did justice to the sounds . . .' (Levin 1987: 116). Small wonder that those accustomed to indulge in such delights have little patience with the unromantic question of whether the letters of the alphabet represent sounds at all. The same scenario which bills Chinese writing as a system devised by 'barbarians' predictably reserves the noblest role for the Greeks. The romantic view of the Greek contribution to writing may be summarized as follows. The Greeks did not just add letters for vowels to a borrowed Phoenician inventory of letters for consonants. They did something far more profound. By a typical stroke of Greek analytic genius, they rethought the whole question of how to represent spoken language by means of visible marks and came up with what modern phonologists later rediscovered as the phoneme principle. In other words, the Greeks converted the alphabet into a phonemic notation. Among champions of this alleged Greek achievement the most eloquent is undoubtedly Havelock, for whom the Greeks were the first people to devise a 'true' alphabet.
254
Roy Harris
Atomism and the alphabet alike were theoretical constructs, manifestations of a capacity for abstract analysis, an ability to translate objects of perception into mental entities, which seems to have been one of the hallmarks of the way the Greek mind worked. (Havelock 1982: 82)
Furthermore, the Greeks were not the sole beneficiaries of this revolutionary insight, for 'the new system could identify the phonemes of any language with accuracy'. The cultural consequences of this were far-reaching in Havelock's view. Thus the possibility arose of placing two or several languages within the same type of script and so greatly accelerating the process of cross-translation between them. This is the technological secret which made possible the construction of a Roman literature upon Greek models - the first such enterprise in the history of mankind. (Havelock 1982: 85)
The views briefly delineated above are open to objection on various counts. Not only do they romanticize the history of writing in conformity with the cultural biases of Western education and in so doing carry conjecture far beyond the bounds of available evidence, but they make it all but impossible to come to grips with the question of how writing restructures thought. In order to correct this distortion of perspective, it is necessary to bear three points in mind. First, it no more follows from the fact that various writing systems include what appear to be recognizably iconic characters that writing is derived from pictorial representation than it follows from the fact that various writing systems include characters recognizable as simplified geometric shapes that writing is derived from geometry. To conclude otherwise would be to confuse form with function. The question of the origin of writing must not be conflated with the question of the ancestry of particular written characters (Harris 1986). Second, since we cannot know exactly how the inventors of the alphabet (whoever they were) pronounced their native language, there is no basis for discussing any original correspondence between the characters and the sound system. Furthermore, any argument which depends on taking primitive forms of the alphabet as evidence for a putative sound system would be patently circular. Third, attributing the discovery of the phonemic principle to the Greeks on the evidence afforded by Greek alphabetic writing is like crediting the inventor of the kettle with discovering the principle of the internal combustion engine. If the Greeks were unaware of the phonemic principle they could hardly have made it the basis of the Greek alphabet. Mysticism at this point comes to the rescue of the romantic hypothesis, and the Greeks are said at least to have achieved 'an unconscious phonemic analysis' (Robins 1979: 13). But what kind of cognitive feat is an 'unconscious phonemic analysis'? It hardly helps the romantic case to say that the Greeks hit upon the phonemic principle without realizing it. That will not do for at least two reasons. First, it contradicts the romantic story that the alphabet is a work of analytic genius. Second, the phonemic principle is not the kind of thing it is possible to hit upon without realizing it, any more than it is possible to grasp the Newtonian law of gravitation without knowing you have done so. That is precisely why we attribute a grasp of the law of gravitation to Newton, and not to his many predecessors in human history who had been hit on the head by falling apples. If we look soberly at the historical evidence we are forced to conclude that if at any time in Graeco-Roman antiquity some unsung genius did work out a version of
How Does Writing Restructure Thought?
255
the phonemic principle, then either that insight died with its author, or else those to whom it was imparted as a useful principle for devising writing systems must have entirely failed to understand the message. The facts which force us to this conclusion are the following. First, phonetics was manifestly the weakest branch of Greek linguistics (Robins 1979: 24). People who were no better informed about the mechanisms of articulation than the Greeks were not likely to have understood enough about speech production to arrive at a clear understanding of the phonemic principle. Furthermore, if the Greeks had understood this principle, or even groped towards it, it is quite mysterious why they never discuss it. For example, it is certainly relevant to the topics debated by Socrates in Plato's Cratylus: but it is quite clear from that discussion that in Plato's day the sharpest minds in Greece, far from grasping the phonemic principle, were still floundering with the crudest notions of sound symbolism as an explanation for the structure of human speech. This is a devastating piece of testimony against the romantic hypothesis. It admits only two interpretations. Either the institution of the Greek alphabet had nothing to do with the phonemic principle; or else it had, but by the time of Socrates the connexion had already been forgotten. Either way, it is clear that from Plato onwards the phonemic principle simply is not grasped in Greek discussions of language. This must cast serious doubts on any thesis to the effect that the Greek system of letters embodies the world's first example of phonemic analysis. No less devastating is the internal evidence afforded by comparison of the two major applications of alphabetic writing in Graeco-Roman antiquity: namely, to record the classical Greek and Latin texts that survived as the basis for Western education. Inspection of the Greek case in isolation might possibly lead one to view it as an example of phonemic notation, if one were prepared to overlook certain inconsistencies, including the muddle it makes of aspirate consonants. What gives the game away is the Latin case, which fails to cope with the phonemic structure of the Latin vowel system altogether. Latin spelling treats the long and short vowels as allophonic variants, which they are not. Now if the Greeks had understood the phonemic principle, it is inconceivable that the Romans did not, especially those Romans most thoroughly versed in Greek culture, and who wrote specifically on linguistic topics (for example, Varro and Quintilian). Quintilian, who knew everything the Greek grammarians had written, actually raises the question of the adequacy of the alphabet; but it is quite clear that he had no conception of interpreting the alphabet in terms of the phonemic principle. The only way to salvage the romantic hypothesis would be to retroject the Greek discovery of the phonemic principle back into a remote past, and postulate an ancestral dialect with exactly the inventory of phonemes to match the postulated inventory of original letters. But this postulation is entirely gratuitous, for a more plausible account of events is readily available. This 'unromantic' account has been given many times; and no doubt it is rather humdrum compared with the attractive story about Greek analytic genius. This alternative is simply that the Greeks found that unless they added some letters for vowels, the Phoenician alphabet of consonants could not be used for Greek in any satisfactory way. The reason for this does not take much genius to discover. It is a reflection of the fact that Indo-European languages have a typically different morphological structure from Semitic languages. One must
256
Roy Harris
stress here the term morphological: the crucial factor has nothing to do with phonology at all, much less with the alleged Greek capacity for 'unconscious phonemic analysis'. Debunking romantic fantasies about writing would hardly be worth anyone's time did they not threaten to conceal from us a more realistic picture of how writing restructures thought. The threat arises in a number of ways, of which the following are two. First, there is the inculcation of anachronistic views about the relationship between language and art, and about the social functions of language and art, which are comprehensible only after the advent of writing. Second, there is the perpetuation, in an extremely subtle form, of the 'cultural chauvinism' (Ong 1982: 18) of the nineteenth century, which assumes that the zenith of human intelligence and cultural achievement is represented by and reflected in the kind of education available to a member of the upper classes of white, Western industrialized countries. In other words, we are dealing with an attitude which assimilates the plight of the poor illiterate Westerner to the plight of the poor preliterate savage, even while romanticizing it as 'primitive', 'innocent', 'ancestral', 'close to Nature', etc. This comes out in the form of insisting that the mental difference between literacy and non-literacy has to do with memory. Allegedly, oral cultures are obsessed with the problem of handing down in oral formulaic capsules pearls of communal wisdom which would otherwise be forgotten. By contrast, literate cultures do not have to impose this mnemonic burden on themselves, because they can just write it all down. So they are free to use their collective minds in other ways. The restructuring of thought which writing brings is based on this supposed liberation of psychological space from the duties of storage. The clearest statement of this theory comes from Havelock, who applies it specifically to the 'literate revolution' in Greece; but his argument, if valid, would clearly apply far more generally. He writes: The alphabet, making available a visualized record which was complete, in place of an acoustic one, abolished the need for memorization and hence for rhythm. Rhythm had hitherto placed severe limitations upon the verbal arrangement of what might be said, or thought. More than that, the need to remember had used up a degree of brain power - of psychic energy - which now was no longer needed . . . The mental energies thus released, by this economy of memory, have probably been extensive, contributing to an immense expansion of knowledge available to the human mind. (Havelock 1982: 87)
This thesis about the limitations of memory has significant affinities with common nineteenth-century views of 'linguistic impoverishment': for instance, the view that primitive languages have impoverished vocabularies, the view that the uneducated farm labourer has an impoverished mastery of the vocabulary of English by comparison with the squire's son, and the view that dialects are impoverished versions of a hypothesized standard language. The common factor uniting all these views is that literacy broadens the mind and widens the intellectual horizons; because the human memory is finite, and those of us who cannot relieve the memory by recourse to writing or by having access to written records are condemned to trying to remember what would otherwise be forgotten. Writing, in brief, was the technology by which the human mind was at last freed from the cultural burden of constant oral repetition.
How Does Writing Restructure Thought?
257
Whatever plausibility this theory has begins to sag as soon as the contention that literacy broadens the mind is called into question. Like all new technologies, writing was a mixed blessing in human history. In many respects, literacy can narrow the mind just as easily as broaden it. Furthermore, as Socrates was aware, it can be argued that the effect of reliance on writing is not to liberate 'psychological space' but merely to weaken the memory. In addition, writing itself imposes 'storage requirements' on the mind. Learning one's letters is not a simple matter of familiarizing oneself with a couple of dozen arbitrary shapes. (This is the simplistic notion which seems to underlie the common contention that an alphabet is superior to a syllabary because it reduces the number of symbols. On the contrary, substitution of an alphabet for a syllabary may well increase memory load. If reduction of the basic number of symbols were a criterion of efficiency, then clearly the binary system would be the optimum system of numerals. Significantly, only computers find this to be the case. Human beings don't.) Nor does the theory that literacy makes possible a release of 'psychic energy' provide any specific answer to the question of exactly how thought is restructured by writing. If it is true that writing restructures thought, that is merely a particular case of the more general truth that all new intellectual tools restructure thought. The abacus restructured human thought much more profoundly than the alphabet. The camera restructured human thought much more profoundly than the calculus. The question with every new intellectual tool is always: how does this innovation make possible or foster forms of thought which were previously difficult or impossible? The key to understanding how writing restructures thought lies in seeing how writing facilitates a variety of forms of autoglottic inquiry. The actual development of these facilities then depends on other factors in the cultural equation as it obtains in a particular historical context. Nevertheless, this does not make it impossible to formulate certain generalizations about the process itself. Inevitably, the introduction of writing reshapes the whole framework of communicational concepts available to a community. It does so simply by destroying the equation of language with speech. But to describe the revolution in these terms is already to describe it retrospectively from the vantage point of literacy. Less misleadingly, we might say that 'it was the invention of writing that made speech speech and language language' (Harris 1983: 15), although any endorsement of that formula would be vacuous unless it took account of the various ways in which the distinction between language and speech may be drawn in literate societies. (For a discussion of this distinction see Harris 1983.) The same point may be put in slightly different although no less retrospective terms as follows. The restructuring of thought which writing introduces depends on prising open a conceptual gap between sentence and utterance. This is the locus for the creation of 'autoglottic' space; and it is into this autoglottic space that the syllogism is inserted in the Western tradition. Without that space, no formalization of the kind we now call 'logic' would be possible. (Exactly how this and cognate formalizations develop depends, as noted above, on other factors in the cultural equation. One can see this by comparing different cultures. In Europe the advent of writing led to the formalization of logic before the formalization of grammar. In India it was the other way around: first the formalization of grammar and only then the
258
Roy Harris
formalization of logic. This is an example of the kind of intercultural difference which the romantic view of writing obscures.) Writing is crucial here because autoglottic 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. The Artistotelian syllogism, like the Buddhist Panchakdrani, presupposes writing. It is sometimes argued (Ong 1982: 78ff.) that preliterate cultures have available an alternative model of unsponsored language in the form of proverbs, songs, etc. (where these are traditional, i.e. essentially 'authorless'). But this is to overlook a number of important differences. For instance, proverbs acquire vicarious sponsorship by being cited appropriately by parents, elders and others noted for wise speech. Songs likewise are the responsibility of those who sing them. In short, there is no pragmatic divorce in these cases between a linguistic object and an episode of discourse. More importantly, however, proverbs, songs, etc. offer no general model of unsponsored language; precisely because it is not the case that any utterance fulfils the requirements of a proverb, song, etc. Whereas writing affords exactly that dimension of generalization: in principle, any utterance will have a written counterpart in a culture which has an adequately developed script. The creation of autoglottic space depends not merely on the fact that written words have a physical existence which is independent of their author's existence, but also, and more crucially, on the fact that writing offers a form of unsponsored language which is not limited to particular categories of speech act or verbal practice. Thus in a literate culture it is relatively less difficult than in a primary oral culture to distinguish consistently what is said and what is meant from the person who said it and the occasion on which it was said. In a primary oral culture there are no genuinely autological forms of verbal knowledge because there is no technology by means of which words and their relationships can be decontextualized at will. Writing constitutes such a technology: it thereby introduces a level of verbal conceptualization which detaches words from their human sponsors. It is the availability of this level of conceptualization which makes it possible for Socrates to ask questions like 'What is justice?'. If this is correct, then the role of literacy in the evolution of Greek thought is quite different from the role assigned to it within the 'romantic' perspective. The pivotal movement in the intellectual development of ancient Greece is not accounted for simply by pointing to the introduction of writing, or to its 'interiorization' (whatever that may mean), and even less by anachronistic construals of the Greek alphabet as a phonemic notation. The contrast between Homeric patterns of thought and Socratic strategies of inquiry is striking but inexplicable, until and unless we introduce a different factor into the picture. This factor is political. It has to do on the one hand with the advent of democracy, and on the other hand with the emergence of the Sophists. The pivotal balance is swayed by the assimilation of all forms of intellectual inquiry to political debate. The paradox of Classical Greek culture lies precisely in that assimilation. At the very moment when, according to the romantic view, a new and powerful restructuring of thought became available, Greece renounced it and apparently reverted to an essentially oral strategy for acquiring knowledge. Socrates, according to the adulatory account which Plato has left us of his activities, is the
How Does Writing Restructure Thought?
259
supreme apostle of what in the Indian tradition would be called the sabda pramana. Plato, although in many respects Socrates' faithful disciple, ultimately reached a different view. As Julius Tomin demonstrates in a recent paper (Tomin 1988), it can be argued that the conflicting attitudes towards writing to be found in Plato's works are explicable chronologically by reference to the failure of the philosopher's political experiment in Sicily. Having begun his career as a Socratic sceptic about writing (as witnessed in the Phaedrus), Plato eventually ends up in the Laws as a convert to the new technology. Laws must be set down in writing: the lawgiver must not only write down the laws, but in addition to the laws, and combined with them, he must write down his decisions as to what things are good and what bad; and the perfect citizen must abide by these decisions no less than by the rules enforced by legal penalties. (Laws, 823A)
One might take the argument a stage further by pointing out that although Plato treats laws as requiring sponsors in practice (i.e. legislators, kings, etc.), the criteria by which laws are to be justified in effect treat laws as examples of unsponsored language: that is to say, laws are precepts having virtues or demerits which are to be evaluated independently of their sponsors or their sponsors' intentions. Without the autoglottic space afforded by writing in a society which can survey, as Plato says, 'not only the writings and written speeches of many other people, but also the writings and speeches of the lawgiver' (Laws, 858C), it would be difficult indeed to evaluate what 'the law' says as distinct from what the law-giver says. That surveyability, which places tyrants and poets on a par by substituting texts for both, is the contribution which writing makes to the structuring of conceptual space in a literate culture.
References Bloomfield, L. (1935) Language. Allen & Unwin, London. Cohen, M. (1958) La grande invention de I'ecriture et son evolution. Klincksieck, Paris. Davy, C. (1772) Conjectural Observations on the Origin and Progress of Alphabetic Writing. London. Derrida, J. (1967a) L'Ecriture et la difference. Editions du Seuil, Paris. (1967b) De la grammatologie. Editions de Minuit, Paris. Diringer, D. (1962) Writing. Hutchinson, London. Fevrier, J.G. (1948) Histoire de I'ecriture. Payot, Paris. Gelb, IJ. (1963) A Study of Writing, 2nd ed. University of Chicago Press, Chicago. Goody, J. (ed.) (1968) Literacy in Traditional Societies. Cambridge University Press, Cambridge. Harris, R. (1983) Language and speech. In R. Harris (ed.) Approaches to Language. Pergamon Press, Oxford. (1986) The Origin of Writing. Duckworth, London. Havelock. E.A. (1963) Preface to Plato, Harvard University Press, Cambridge, Mass. (1976) Origins of Western Literacy. Ontario Institute for Studies in Education, Toronto. (1982) The Literate Revolution in Greece and its Cultural Consequences. Princeton University Press, Princeton. Levin, S. (1987) Review of Harris 1986. General Linguistics 27, 113-7. McLuhan, M. (1962) The Gutenberg Galaxy. University of Toronto Press, Toronto.
260
Roy Harris
(1964) Understanding Media. McGraw-Hill, New York. Ong, W.J. (1982) Orality and Literacy. Routledge, London/New York. Parry, M. (1971) The Making of Homeric Verse. Clarendon Press, Oxford. Robins, R.H. (1979) A Short History of Linguistics, 2nd ed. Longman, London. Tomin, A. (1988) With the Phaedrus on his mind. Unpublished paper.
18
Writing and Proto-Writing: From Sign to Metasign ROY HARRIS
The cognitive complexity of writing has long been recognized by both linguists and psychologists. Much attention has been paid in recent years to the cognitive consequences of writing not only for the individual but also for society, and the view is now widely accepted that in some fundamental sense 'writing restructures thought'. Exactly what this restructuring is taken to imply is another matter. 1 But neither this nor any other alleged consequence can be fully understood unless situated in a developmental sequence which clearly identifies the relationship between cognitive consequences and cognitive prerequisites. To focus on the former at the expense of the latter is simply to invite confusion. Writing is, uncontroversially, one of the uses of visual signs. Exactly how it is defined from a semiological point of view will depend on one's adoption of a certain theoretical perspective. But regardless of the particular perspective adopted, writing cannot plausibly be supposed to emerge ex nihilo. There will be a developmental sequence from 'proto-writing' to writing which must be accounted for, both ontogenetically and phylogenetically. The Aristotelian theory of writing, which treats the written form simply as a visual metasign representing a previously given vocal sign, has gone virtually unchallenged throughout the Western tradition. But although this theory makes the metasemiotic function criterial for writing, it offers no account of the phylogenesis of the visual metasign itself. The hypothesis of an evolution from primitive pictography does not help here, because the problem is precisely to account for the crucial transition of the primitive pictograph from sign to metasign. The present chapter argues that one of the important cognitive prerequisites for the emergence of writing is the recognition of what will here be termed 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. An important question, therefore, is how the concept of a synoptic metasign arises, and under what circumstances prior to the development of writing such signs could occur 'naturally' in the processes of graphic communication. 1
See Ch. 17 above.
261
262
Roy Harris
The relevance of memory to the cognitive functions assigned to writing has been recognised from very early times. In Plato's Phaedrus the practice of writing is criticised as likely to weaken the memory. 2 But it is highly unlikely that this criticism was grounded in any experimental evidence. If it was, neither Plato nor any of his spokesmen bothered to produce it. It seems more likely that we are dealing here with an inference grounded in ethnoscience, which simply assumes that any technology must necessarily undermine whatever human skills it displaces. Parallel assumptions in our own day would be that television discourages reading, or that the use of calculators will result in a generation of children deficient in the skills of elementary arithmetic. The validity of this kind of 'deficit theory' is not what is at issue here. Nevertheless, the existence of such a theory in a society of the kind we know Plato's Athens to have been shows clearly enough that the main utility of writing was assumed to be that it provided a visible check on what would otherwise need to be consigned to memory. It must not be forgotten, however, that for the Greeks of Plato's day writing meant the use of a script. That equation cannot be expected to throw any light on the origin of writing, or on the stages of protowriting that preceded it. If a script is defined as any system of regular correlations between speech and visible marks which provides for the identification of the former on the basis of the latter, it is highly unlikely that at any time or place a script has ever been developed except as the result of a conscious, deliberate attempt to construct one. The task is too specific and too complex for a script to come into existence 'by accident'. But in order to construct a script, it is first necessary to have the concept of 'writing', unless we are willing to regard that concept itself as either a descendant or a sibling of the concept 'script'. This suggestion seems implausible for a variety of reasons. The anthropological evidence shows clearly enough that other forms of graphic representation are more widely distributed than the use of scripts, and in societies which practice both drawing and writing the terms designating these two activities are not infrequently etymologically identical. In short, historically speaking, scripts seem to be a specialized form of writing, rather than writing a generalized development of script. The problem of the origin of writing therefore has two aspects, which should not be confused. One of these concerns the differentiation of writing from other forms of graphic communication, including drawing. This question is addressed by theories which deal with the birth of the 'scriptorial sign'.3 The other concerns the evolution of cognitive structures underlying this differentiation. It is the latter question which is the chief concern of the present chapter. The cognitive structure which has been the chief focus of theorists who deal with the relationship between speech and writing is 'linearity'. Whether the term linearity itself is well chosen might be disputed, but it has generally been accepted since at least the time of Saussure, who explicitly identified it with temporal succession. If this is correct, to say that an utterance or an inscription has a linear structure is to say, among other things, that the mind grasps it as a connected sequence of units, each of which has a unique position in terms of accompanying units which either precede or 2
R. Harris, The Origin of Writing, Duckworth, London, 1986, p. 19. 3 Ibid. pp. 55-6, 77, 123ff.
Writing and Proto-Writing
263
follow. The exact visual analogue of temporal succession is assumed to be that of points along a single continuous line, which may proceed from right to left, left to right, top to bottom, etc. as the case may be. Typical scripts are linear in the sense that they are interpretable as representing certain units occurring in the temporal flow of speech, and representing those units as a succession of written signs occurring in exactly the one-dimensional spatial ordering which corresponds to the auditory succession of the units represented. This correspondence seems so obvious and so natural to anyone acquainted with writing that its cognitive complexity may easily escape attention. It might seem that no explanation is required for the emergence of linearity in association with writing. But this is too simplistic. For neither of the two basic varieties of autonomous visual sign which appear to underlie all forms of graphic communication is intrinsically linear. Emblem and token are terms which have been proposed for these two categories of sign.4 The former are signs of which each recurrence designates the same symbolic content, while the latter are signs of which each recurrence designates a separate instance or example. (Thus a national flag is an emblem: two Union Jacks do not symbolize two nations, but one and the same nation. A notch on a tally stick is a token: two notches represent two separate items which the tally records.) What is the relationship between tokens, emblems and linearity? Token-iterative recording employs a one-to-one mapping between tokens and items recorded, whereas emblematic recording can replace this by an appropriate emblem or combination of emblems. Neither, however, presupposes linearity as the basis of graphic representation. Tokens can be grouped in a variety of different ways, and emblems can combine in a variety of forms. (For instance, in representing the information 'ten sheep', it is possible to juxtapose two separate emblematic signs for 'ten' and 'sheep', or to position the sign for 'ten' inside the sign for 'sheep', or to position the sign for 'sheep' inside the sign for 'ten', or simply to superimpose one sign on the other.) Linearity is only one of a number of semiologically possible modes of arrangement for both emblems and tokens. A script, however, as defined above, is a visual sign system designed to permit, precisely, a replication of certain oral patterns in writing. It does so, typically, by treating given graphic configurations simultaneously both as emblems and as tokens, in relation to the oral signs they signify. (As the written record of an utterance, 'Boys will be boys' tells us that the word boys was uttered twice, and the words will and be once only.) In this sense, the linearity of the written sentence presupposes the availability of both emblems and tokens, whereas neither emblems nor tokens as such presuppose the availability of linearity as a semiological mode of arrangement. Can we identify, then, a more primitive mode of representing speech, which simply dispenses with linearity? In a famous experiment conducted sixty years ago, Luria demonstrated that with the help of pencil and paper children were able to recall and distinguish dictated sentences more accurately than when relying on aural memory alone. 5 What was remarkable about the children in Luria's experiment was that none of them could write. To an adult eye, the marks on the paper were mere scribbles. *Ibid. pp. 131ff. A.R. Luria, The development of writing in the child', in M. Cole (ed.), The Selected Writings of A.R. Luria, Sharpe, White Plains, N.Y., 1978, pp. 145-94. 5
264
Roy Harris
What Luria's experiment suggests is that, were it not for limitations of memory, the visual recording of speech would certainly not require the use of script. Luria distinguishes various stages in the development of 'scribbling' as a child's aid to oral recall. In the first stage the scribbling is undifferentiated and does not appear to have any mnemonic function. In the second stage, although still apparently undifferentiated, the scribbles somehow enable the child to identify and distinguish between dictated items. We checked this and indeed discovered that these scribblings actually were more than just simple scrawls, that they were in some sense real writing. The child would read a sentence, pointing to quite specific scrawls, and was able to show without error and many times in succession which scribble signified which of the dictated sentences. Writing was still undifferentiated in its outward appearance, but the child's relationship to it had completely changed: from a self-contained motor activity, it had been transformed into a memory-helping sign.6
At a third stage, the scribbles begin to become differentiated in a manner which corresponds to differences in the sound of the dictated item: e.g. longer sentences are transcribed by longer squiggles, and the shape of the squiggle suggests the rhythmic sequence of the spoken words. At a fourth stage, the differentiation begins to correspond to specific differences of content between the dictated items. And so on. At what stage can we say that the child is actually making a record of what it hears, or that the squiggle is an attempted reproduction - albeit rudimentary and holistic - of the utterance? A written record of speech, it could be argued, simply is a set of marks which can be 'read'; that is, translated back as required into the same oral sequences that originally prompted them, irrespective of exactly how this is done. Different writing systems work on different principles to make this translation possible; but it would be question-begging to insist a priori that a sentence cannot be written down unless the written marks reflect in detail the linear structure of the oral sequence. One reason for this is that the linear structure of an utterance cannot be taken as something perspicuously 'given', and another is that the notion of representing a linear structure is not perspicuous either. This is in part because of the problematic nature of equivalences in language between linear and non-linear structures. It seems impossible to dispense with such equivalences in any systematic enterprise of linguistic description, but their exact status is far from clear. In an important recent contribution to the long-running debate among linguistic theorists concerning the linearity of the linguistic sign, Hutton? has pointed out that there is no upper limit in linguistic communication to the reduction of linear structures to equivalent non-linear structures. This means that, were it not for limitations of memory and practical inconvenience, there is no reason why English should not be replaced by a system of communication in which any given English sentence or other complex sign was represented by a single, non-complex sign or character. Such a system would be synoptically equivalent to English. Synoptic reduction has long been utilized in the construction of codes for telegraphic and other purposes. It is also presupposed in our everyday use of abbreviations and titles. For instance, 'He sang Rule Britannia' is understood as 6 Luria, op. cit., p. 159. 7 See above, Ch. 9.
Writing and Proto-Writing
265
meaning not that he sang just those two words, but as meaning that he sang the song which begins 'When Britain first, at heaven's command, Arose from out the azure main . . .' Now it is merely a contingent fact that the title Rule Britannia itself happens to be syntagmatically complex. Disregarding that fact as in principle irrelevant, the title may be regarded as a synoptic reduction or substitute for the song, in the sense that it 'stands for' a linguistically complex item which could be formulated in extenso. Some writers in the Western tradition (Home TookeS is a conspicuous example) have treated synoptic reduction as one of the fundamental linguistic processes. At its crudest, this amounts to treating, for example, the noun square as an abbreviation of some complex phrase such as rectangular figure with four equal sides, or the word brother as a substitute for male sibling. In this sense, the notion of synoptic reduction is intimately associated with theories of definition and paraphrase. Complementary to the notion of synoptic reduction is that of analytic expansion. Just as the word square may be treated as a synoptic reduction of rectangular figure with four equal sides, so may rectangular figure with four equal sides be treated as an analytic expansion of square. Viewed from this perspective, the linearity of the linguistic sign comes to be seen as indissolubly associated with its arbitrariness. That is to say, linearity presents itself as the natural mode of articulation of signs which resist further sequential analysis into significant units, while at the same time these ultimate significant units are seen as being 'arbitrary' in nature precisely because no account of what they mean can be given in terms of the decomposition of their signifiants into smaller units sequentially combined. Saussure's erection of the principle of arbitrariness and the principle of linearity into the twin foundational principles of modern linguistics betokens his intuitive recognition of the profound connexion between the two, but also shows his failure to recognize that both are simply complementary aspects of a single principle. In fact, he gave priority to the principle of arbitrariness over that of linearity, whereas it is arguable that he should have done the reverse. In other words, he did not get as far as seeing that arbitrariness, for his purposes, could have been defined in terms of linearity, but not linearity in terms of arbitrariness. In visual, as distinct from oral, communication linearity plays a different role; or rather, a variety of different roles. For example, token-iterative recording, universally practised in some form or other in all human communities which have progressed as far as the keeping of accounts or reckonings (for whatever purpose) may employ linearity as a semiotic device, but not in the way speech does. The linearity of the abacus is not the linearity of the sentence. It does not matter whether you count the beads on a string from left to right or from right to left. Nor is the linear relationship of any given bead to any other bead on the same string significant. Nevertheless, in just the same way as synoptic reduction in language can replace a complex of signs by a single sign, so a number of beads on one string can be replaced by a single bead on another string for purposes of reckoning. In speech, as Hutton points out, the linear analysis of an utterance into a succession of discrete sequential signs can be construed in at least two different ways. It can be construed, following Condillac, as an imposed analytic elaboration of 8 For a recent discussion, see R. Harris and T.J. Taylor, Landmarks in Linguistic Thought: The Western Tradition from Socrates to Saussure, 2nd ed., Routledge, London, 1997, Ch. 12.
266
Roy Harris
a unitary concept or percept. Or it can be construed, following Saussure, as reflecting a concatenation or configuration of separate concepts. But whichever construal is adopted, there is an assumption that 'linearity reflects semantic structure'. It may be observed, however, that exactly the same holds for the linearity of the abacus. To count a row of beads is to conduct a linear analysis based on the assumption that each unit has a given semantic value. In the simplest case, each bead has the arithmetic value 'one'. But other conventions are possible. One bead may have the value 'ten'. Or each group of two beads may have the value 'one'. So there is nothing specific to linguistic representation which accounts for this semiotic property of linearity. The development of mathematical notation of the kind in which the linear sequence 333 is understood as meaning 'three hundred and thirty three' would have been impossible without the intellectual breakthrough which consists in realizing that the 'place' of a sign in a linear sequence can be allotted a significance systematically related to its significance when considered 'in isolation'. (Isolation here is the limiting case of a linear structure in which the position of the unit is defined by having no unit which precedes it and no unit which follows it.) One important difference, however, between the linearity of the sentence and the linearity of the abacus is that the latter can easily be replaced by other forms of visual representation fulfilling exactly the same semiotic function. For example, beads arranged linearly on a string can be replaced by dots randomly grouped within an enclosing circle. But the reason why this is so is that there is no difference between counting the number of beads on a given string and counting the number of dots lying within a given circle. In the case of the sentence, on the other hand, there is nothing corresponding to a general computational procedure which could be applied equally well to a linear as to a non-linear arrangement of units. (Once we scramble the wordorder in the sentence John and Mary both disliked Bill there is no way the message can be recovered other than by guesswork.) It is intrinsic to the equivalence between beads on a string and dots in a circle that the string and the circle function emblematically, and that the token-iterative representation take this emblem as its semantic basis. In the case of the abacus, one has to 'remember' whether the beads on this particular line are 'ones' or 'tens', etc. In the case of the dots inside circles, one likewise has to 'remember' which is the 'unit' circle, which is the 'tens' circle, etc. Otherwise the necessary synoptic reductions break down and one's reckoning system collapses. Similarly, the child in Luria's experiment has to 'remember' which oral sequence its scribble 'stands for'. In effect, the scribble functions as an ad hoc synoptic reduction of the complex linear utterance. What is interesting from the point of view of cognitive psychology is the apparent fact that scribbling actually helps recall. What is interesting from the point of view of a theory of writing is the fact that a visible mark may be reliably correlated with a given utterance without the assistance of a prior convention. A script, by contrast, is a specific convention or set of conventions for achieving such correlations, and in many cases for extending them to other possible utterances in a given language. When operating with a script, one has to 'remember' what each sign 'stands for'. Opting for a linear script is convenient because it obviates having to remember in addition what each place in the sequence 'stands for' (as is essential in mathematical notations of the type mentioned above).
Writing and Proto-Writing
267
The differences between the earliest stages of the development traced by Luria can hardly be described as the emergence of a 'writing convention' of any kind. Yet it is clearly crucial to the developmental sequence which results in the child's control of the correlations between scribbles and oral sentences. To account for what happens, Luria speculates that the child somehow attributes significance to the external relations into which the scribble enters. In itself no scribble meant anything; but its position, situation, and relation to the other scribbles, i.e. all those factors together, imparted to it its function as a technical memory aid.9
There is, however, another possible explanation; namely, that the 'memory' of what the scribbles 'stand for' is somehow laid down in the physiological process of inscription, in such a way as to be recovered or triggered by visual inspection of the inscribed trace. This we might call motoric recording, as distinct from the schematic recording which appears at the 'differentiated' stage. It would be characteristic of the former that particular traces are 'one off efforts, and characteristic of the latter that comparison between two or more different traces is involved. Even if the scribbles are lines of some kind, it cannot be assumed that they have a linear structure in any semiological sense. They are neither 'proto-tokens' nor 'proto-emblems'. They are not proto-tokens because clearly the child's scribble does not represent indiscriminately any sentence at random. On the other hand, they are not protoemblems because the child has not developed any consistent set of correlations between traces and utterances. Both emblem and token presuppose a level of abstraction and systematization which are absent from whatever cognitive processes are involved in the child's use of these scribbles as mnemonic aids. Nevertheless, insofar as the scribble effectively recalls the utterance, it functions as a holistic protosign. Phylogenetically, the corresponding problem of the proto-sign has to be envisaged differently. The child in Luria's experiment is not only given paper and pencil, but is already forced by the circumstances of the experiment to grapple with the problem of producing a pre-conventional graphic sign which will function as a synoptic reduction of a given oral message. In Aristotelian terms, the child has successfully devised a visual metasign, but only because the metasemiotic function was externally imposed. The question which arises as regards the prehistory of writing is how an illiterate community could, in the natural course of events, arrive at a comparable stage: i.e. grasp the metasemiotic potential of inscribed marks. It has recently been suggested by Russell that apparently randomly scratched stone and slate plaquettes recovered from Upper Paleolithic sites in Europe are evidence of the beginnings of writing. 10 (Plaquettes are defined as slabs 'with parallel faces under 20 cm across and 4 cm thick'. 1!) Some of these plaquettes seem to have
9
Luria, op. cit., p. 159 P.M. Russell, 'Plaques as paleolithic slates: an experiment to reproduce them', Rock Art Research, Vol. 6, No. 1, 1989, pp. 40-1, 68-9. 11 P.O. Bahn and J. Vertut, Images of the Ice Age, Windward, London, 1988, p. 76. 10
268
Roy Harris
been used over a period of time and present 'a maze of superimposed lines'. i2 For Russell the significance of these plaquettes is that they are found in situations which indicate everyday domestic purposes, rather than association with art, magic or ritual reserved for special occasions. Russell has also demonstrated by experiment that such plaques could be 'wiped clean' for repeated use by the application of water or a mudand-water solution. Consequently, their use for daily record-keeping of a more or less ephemeral kind is highly plausible. Bahn and Vertut make the crucial point that at sites such as Enlene there were thousands of blank plaquettes available, so it cannot be supposed that the same plaquette was re-used simply for reasons of economy; 13 while Hadingham notes that at at least five unconnected sites (Labastide, Gb'nnersdorf, Parpallo, Limeuil and La Marche) there is evidence of 'engraving fragments of animals as well as whole figures, and of superimposing outlines, sometimes to an almost undecipherable extent'. 14 The reference to 'fragments of animals' cannot fail to suggest that in the Upper Paleolithic period visual synecdoche was already a recognized semiotic device. These plaquettes throw a rather different light on the development of protowriting. Any object with a surface which can be marked has the potential for use in communication in two possible ways. (i) The object itself retains 'semiotic neutrality', merely providing a convenient surface for the inscription of a sign or signs. (ii) The object itself functions as a sign, as well as acting as the bearer of a further (inscribed) sign or signs. The latter possibility subdivides at its simplest into: (iia) The object itself functions as emblem and the inscribed mark as token, or (iib) The object itself functions as token and the inscribed mark as emblem. In the case of Russell's plaquettes, which seemingly derive their significance from use in some domestic context, what is interesting is the evidence of repeated re-use, which eventually leaves behind a cumulative superimposition of traces. At each reuse, this residue is temporarily obliterated to make way for the currently relevant inscription, which will in turn make way for another in due course. The point about re-use made by Bahn and Vertut does not conflict with the theory that both (i) and (ii) were semiological possibilities utilised by the 'artist-scribes' responsible. In other words, in the Upper Paleolithic period we are already dealing with a stage in the evolution of graphic representation at which the sign is conceptualized as having an existence which is separate from the surface on which it is inscribed. This divorce between sign and object is an advance of the greatest significance. No less significant, however, is that this advance goes hand in hand with another. A purely adventitious consequence of continual re-use is that each plaquette acquires a cumulative residue of traces. This residue represents, in effect, a synoptic reduction of the whole sequence of 'messages' which had been previously inscribed and which, with the aid of memory, could be reconstructed. 12 Russell, op. cit., p. 68. Hadingham (1979: 240) describes some of the plaquettes at La Marche as showing 'outlines superimposed in an apparently haphazard fashion' or 'a mass of meaningless scribbles and "parasitic" lines'. For an illustration of a particularly dense superimposition of plaquette engraving see the example from Gonnersdorf reproduced in Bahn and Vertut (1988: 49). '3 Bahn and Vertut 1988, p. 78. 14 E. Hadingham, Secrets of the Ice Age, Heinemann, London, 1979, p. 266.
Writing and Proto-Writing
269
Thus a single non-linear graphic configuration becomes directly iconic of a temporal sequence of signs: exactly the pre-condition required for the development of writing. The semiological significance of this development is easily overlooked in retrospect, precisely because modern civilizations are accustomed to more sophisticated types of visual sign, and in particular to the sophistications of linearity. It is therefore worth while drawing attention to the following points. 1. The superimposition of two (or more) independent signs A, B ( . . . ) which produces a compound configuration X automatically results in a situation in which X can subsequently be 'read' as a metasign for the sequence of original signs A, B (...). 2. By recontextualizing the original signs in this way, the configuration X confers a significance on what may have been originally a fortuitous relationship (i.e. the sequence A, B . . .). It thus provides a cognitive model for the development of metasigns representing any arbitrary sequence of signs already given. 3. 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 B to the form X, but at the same time provides the paradigm for such a convention. (Henceforth, A + B can intentionally be represented as X.) 4. Speech provides no analogue to the formal relationship between X, A and B, but numerous analogues to the semantic relationship (since X can be read as a synoptic reduction of A + B). Thus the graphic form X is already suggestive of a way of conceptualizing a semantic relationship between signs which would also be applicable to speech. (It seems significant, for example, that in Sumerian writing the sign for 'eat' is a composite sign consisting of the signs for 'mouth' and 'food', while the sign for 'drink' is likewise a composite sign consisting of the signs for 'mouth' and 'water'. In both cases the component signs exist independently of the composite sign, but in neither case is there any formal (phonetic) correspondence in speech between the words in question. Hence, in effect, the writing system makes possible a semantic analysis of a vocal sign which has no counterpart in the formal constituency of the vocal sign itself. This possibility is already implicit in the metasemiotic equivalence 'X = A + B'.) With these far-reaching implications, the Paleolithic technology of the slate provides us with the most plausible explanation to date of the 'accidental' emergence in human history of a new type of compound visual sign having a cognitive potential directly relevant not only to the emergence of writing but also to the concomitant conceptual analysis of the spoken word.
This page intentionally left blank
PART SIX LANGUAGE AND SOCIETY
Unlike the previous sections, this one does not bring together papers devoted to a common topic, but attempts to illustrate the diversity of topics that arise for consideration within an integrational framework for the study of language. The link in these cases is that all the papers throw light in one way or another on relations between language and society. They examine linguistic issues that can be seen as controversial, whether for the individual in relation to society, or for one social group in relation to another social group, or for one society in relation to another society. Ch. 19 ('What Has Gender Got to Do With Sex?') challenges the notion that the grammatical categories linguists employ are neutral, objective and devoid of ideological significance. The case in question is that of grammatical gender. Attempts to provide purely formal justifications for this category, to explain it in terms of 'markedness', etc. are examined and found wanting. All these perpetuate the mythology that language is an autonomous fixed code functioning independently of the society in which it operates. Ch. 20 ('What Makes Bad Language Bad?') confronts the social issue of swearing as a language taboo. The paper considers various linguistic accounts that have been given of swear-words, drawing attention to the different status they appear to have in speech and in print, and pointing out that 'swear-word' does not seem to be recognized as a linguistic category until the Victorian period. Ch. 21 ('Law Lessons for Linguists? Accountability and Acts of Professional Classification') considers language and the law as areas of professional expertise. As professionals, academic linguists are less accountable to society than lawyers, doctors and many other 'experts'. The linguist's analytic categories do not have the kind of social relevance that is helpful in tackling the linguistic difficulties that may present themselves in the case of legal texts. But both lawyers and linguists deal with the
271
272
PART SIX: LANGUAGE AND SOCIETY
interpretation of allegedly fixed codes that turn out not to be as fixed as they appear. Ch. 22 ('Teaching American English as a Foreign Language: an Integrationist Approach') describes the application of integrational principles to a specific language-teaching problem. The key to its solution is the contextualization of two linguistic roles - those of teacher and learner - in relation to a certain linguistic subject-matter, identified as 'American English'. This in turn has to be contextualized by reference to the learners' prior expectations of what 'American English' is, which involves demythologizing notions of 'standard' language, linguistic variation and linguistic stereotypes Ch. 23 ('What Problems? On Learning to Translate') outlines an integrational approach to the traditional practice of interlingual translation. For the integrationist there can be no question of simply mapping from the structure of one language on to the structure of another, or the 'relexification' of one and the same experience in two different linguistic forms, or the 'transference' of a message across linguistic and cultural boundaries. These are all concepts of translation generated by adopting a fixed-code view of communication. An integrational approach to translation recognizes as a basic premiss that translation itself, as a communication process, is not the substitution of one set of linguistic signs for an alternative set but the contextualized creation of new signs. Ch. 24 ('Pronouncing French Names in New Orleans') examines a social and linguistic difficulty which is rarely recognized. It concerns the identification of an individual's family name, and hence the individual's self-identification as a member of society. The problem arises when, for cultural reasons, uncertainty arises as to how the name in question is pronounced. This paper analyses the ad hoc metalinguistic and other strategies that may be brought into play in dealing with such problems. It concludes that the 'facts' about someone's name are not fixed by any linguistic code but are created in and by the communication processes in which production of and reaction to the name are involved.
19
What Has Gender Got to Do With Sex? DEBBIE CAMERON
Linguists tend to assume that the grammatical categories used in linguistic descriptions are neutral, objective and devoid of ideological significance. Regularities of language use, such as social and stylistic variation, may call for an explanation in sociopolitical terms: but few scholars would dream of looking for similar explanations of descriptive concepts like 'noun' or 'sentence'. Occasionally this common-sense attitude is criticised, often by researchers in socalled 'exotic' (i.e. non-European) languages, on the grounds that traditional categories are ethnocentric and beg the question whether all systems are like 'standard average European'. But even here, what is being appealed to is some sort of inherent linguistic typological difference (even though it may have cultural correlates); the implication is that we should let our categories somehow arise out of our data, rather than adopting certain traditional categories a priori. This just reinforces the idea that description is, or should be, innocent. Yet linguistic analyses, even when they relate to familiar phenomena, do not arise in a vacuum. Not only do they build, in many cases, on a long tradition of folklinguistic speculation, prescriptivism and pedagogical grammar, they are also produced within a specific social and historical context. They cannot escape from ideology, and they cannot but be implicated in it. In this paper, the interaction of ideology and linguistic description will be explored in relation to ideas about the grammatical category of gender. This is a category with a very long history (it was first used by the linguistic scholars of Greece) and it will be argued that discussions of it have always engaged with, and been influenced by, sociopolitical beliefs about the nature and relative importance of the two sexes. It will also be argued that the evolution of gender systems and certain conventions regulating their use should in some cases be seen as articulating these same beliefs. There is a complex and interesting interaction between the linguistic (what people say) and the metalinguistic (how they theorise what they say) which linguistic theory participates in but fails to acknowledge. The argument will focus on two parallel debates in the history of language study. One is the long controversy in Indo-European scholarship over the origin and significance of gender; the other is the more recent question of whether gender distinctions in English are sexist and in need of reform. In both cases linguistic 273
274
Debbie Cameron
scholarship defended the social and linguistic status quo; but the arguments used to do this were dramatically different in each, reflecting different historical conceptions of sex difference and its connexion with language. The two debates must be examined in some detail; but first it is necessary to define gender and to distinguish between its 'grammatical' and 'natural' variants. Modern analyses of gender define it in terms of concord: it is a property of words that requires other words to agree with it. Thus substantives in many languages can be divided on the basis of article and adjective concord and/or pronominalisation into classes (the original Greek word for gender means 'kind'). Traditionally these were labelled masculine, feminine and neuter, although many 'exotic' languages possess more than three classes. The definition in terms of concord is applicable to any language in principle, but in practice analysts recognise different sorts of gender. In particular they retain a traditional distinction between those languages where gender is 'grammatical' and those where it is 'natural'. The difference can be summed up in the following formula: grammatical gender \sformal whereas natural gender is semantic. In languages with grammatical gender, such as most of the well known IndoEuropean languages, the standard account tells us that gender assignment to the masculine, feminine or neuter class cannot be determined with reference to semantic features [± animate] and [± male]. To English speakers learning grammatical gender languages this can be a source of irritation and of humour (cf. Twain 1880): English is a natural gender language in which, again according to the standard account, these semantic features are the only relevant criteria for gender assignment.1 Pronominalisation in English depends on sex-reference, which means that some nouns are always masculine or feminine in virtue of their meaning (e.g. bachelor and wife) while others, the 'common gender' nouns (e.g. friend) may vary, taking their gender from the sex of the referent on a given occasion. It is important to note that this standard account already provides an answer to our initial question, 'what has gender got to do with sex?' It tells us that natural gender has everything to do with sex, and grammatical gender nothing to do with it. How clear cut and accurate this dichotomy is will be considered below. The point that grammatical gender is purely formal is given particular emphasis in the modern literature (Fodor 1959; Arndt 1970) and in textbooks (Lyons 1968). This reflects the fact that it has only just become conventional wisdom. Theorists of the past may have disagreed about its precise significance, but they took for granted some connexion between gender and sex. This is clear from Istvan Fodor's classic article on the origin of grammatical gender (Fodor 1959). Fodor himself is a vehement opponent of the idea that gender has anything to do with sex. He points out that gender assignment is semantically illogical, and that the same concept may be assigned to different classes in different languages. Gender languages often use lexical differentiation for pairs like bull and cow, rather than exploiting masculine and feminine parallels. Thus sex and meaning 1 "The standard account' here means the account given in standard grammars of English such as Quirk (1972) or Strang (1968). Both these texts assert a 'close' connexion between sex and gender in English; the hedge is apparently because of exceptions like ship (inanimate but feminine) and child (naturally male or female, but grammatically able to be neuter). It should also be pointed out that not every description employs the same terminology (for instance, Quirk et al use the term dual personal gender for my common gender, and this is probably more accurate). I have kept to one terminology for the sake of consistency and to make cross-language comparisons easier.
What Has Gender Got To Do with Sex?
275
must be separate from gender and form. But in his discussion of the views put forward by earlier scholars, Fodor reveals that his own views are unusual (though as we have seen, they have since become standard). Grammarians and philologists for centuries assumed that gender was a special case of sex reference. As one commentator puts it: For two thousand years there has been an unresolved question: whether word endings of nouns and pronouns, as well as articles, are an extension of the qualities regarded as male and female. (Janssen-Jurreit 1982: 280)
The Greek Sophists, for example, who are usually credited with originating the notion of gender, believed that the gender of a word must reflect its essential qualities. It worried them that words with similar meanings should be of different genders, and Protagoras went so far as to recommend reform in cases where synonyms fell into different classes. Later theorists may not have been in favour of intervention, but they did feel such cases were anomalous and required an explanation. On the face of it this assumption was rather problematic, since gender assignment did not seem to follow any obvious natural principle: yet rather than reject the assumption itself, theorists constructed ingenious explanations that linked sex and gender in 'deep' or indirect ways. The simplest proposed link was a historical one. Languages had once followed a consistent pattern based on natural gender (i.e. semantics/sex reference) but diachronic processes of sound change and analogy had disordered and obscured the pattern, producing synchronic anomaly. Another explanation was rather more sophisticated. Its proponents, including Grimm, claimed that the anomalies were only apparent, for underneath the lexicon did follow some sex differentiating principle. Grimm believed that grammatical gender was a later stage of natural gender, which developed when speakers of a language passed from mere recognition of male and female beings to postulate a male/female principle in whose terms anything might be abstractly classified. The nature of this principle could be inferred from the resulting classification itself: thus Grimm tells us: The masculine means the earlier, larger, firmer, more inflexible, swift, mobile, productive; the feminine the later, smoother, smaller, the more still, suffering, receptive. (Grimm, Deutsche Grammatik, 1831; quoted in Janssen-Jurreit 1982: 292)
This kind of circular explanation was often used for explicitly political purposes. The semantic underpinnings of grammatical gender provided 'independent' evidence for contemporary theories of sex difference, the qualities of a grammatically feminine noun fitting neatly with prevailing ideas of 'natural' femininity. And indeed the order of priority was sometimes reversed so that male and female characteristics were assigned on the basis of grammatical gender. Theodor Hippel, an eighteenth-century German thinker, asserted that women could not be rational because the German word for 'reason' is masculine (Janssen-Jurreit 1982: 295). History does not record whether he felt rationality to be the exclusive property of women in France, since the French word raison is grammatically feminine! A later theory discussed by Fodor is that of Wundt (1900), which suggests gender classifications are based on a concept of relative value. Thus those languages
276
Debbie Cameron
which classify men with big animals and women with small ones, or men with gods and women with valued animals, followed by a class of disvalued creatures, make use not directly of sex reference or the semantic features maleness and animacy, but of a common-sense ranking principle. There is still an indirect appeal to sex-reference in this theory, of course: it is taken for granted that male and female beings are of different values, and that males outrank females. It is of interest to note that Wundt's intuitions on this point find a clear parallel in several cultures which exploit the ranking 'masculine, feminine, neuter' in their use of language in order to mark the inferiority of women to men, or of some women relative to others. In Konkani, for instance, women of low status (unmarried girls and widows) are referred to by neuter pronouns, while their married sisters are accorded the feminine. In France and Germany, successful women refuse to be denoted by grammatically feminine occupational titles on the grounds that such feminisation downgrades the office to which it is applied: . . . the successful madame prefers to be le Docteur, le Professeur, 1'Ambassadeur and le Philosophe, even with the succeeding il which is required in formal texts. (Corbett 1983)
It has also been pointed out that lay perceptions of gender value can influence the assignment of neologisms. French feminists cite pairs like chaise and fauteuil or maison and chateau, in which the grander item is masculine and the humbler feminine; they also maintain that among less educated French speakers this kind of semantic ranking overrides formal considerations, so that a word like autoroute with two feminine roots may nevertheless be accorded the masculine because 'motorway' is a masculine sort of thing (Corbett 1983). Fodor also mentions with more approval a theory that comes closer to saying grammatical gender has no basis in sex-reference. The theory has a number of variants, irrelevant for present purposes, but essentially it claims that gender differentation in Indo-European came about in two stages and was syntactically motivated. The first distinction, according to this account, was between animate and inanimate nouns. It was necessitated by the free word order then normal in IndoEuropean, which was obscuring the difference between subject and object nouns in a sentence. This problem could be alleviated by introducing a word ending for inanimate accusatives: no one would suppose that inanimate nouns were acting as agents in a sentence, and therefore the ending would be unproblematically identified with the patient. Possession of the accusative ending would eventually become the criterion for membership of a class of inanimate nouns, the neuter gender. Yet even this account cannot wholly escape sex-reference. For when asked to explain why the animate class later subdivided into masculine and feminine genders, the theorists fall back on a 'natural gender' explanation. Certain word endings became feminine, they suggest, because they were associated with common Indo-European words for females (e.g. the lexical items meaning 'wife' and 'mother'). Alternatively, the endings associated with sex-referential pronouns he and she spread to nouns, again for purposes of syntactic marking in a language without fixed word order. All these various theories rely on notions of sex and sexual difference. The world is assumed to be divisible into male and female, the male is presumed to be more
What Has Gender Got To Do with Sex?
277
important than the female, and above all gender is thought to need an explanation in terms of sex either in the abstract or in the distant past. Yet if we reject the essentialism of a Grimm or a Wundt, the persistent connexion that is evident here between sex and gender has to be seen in ideological terms. This involves rejecting the standard account of grammatical gender as purely formal. It cannot be true that grammatical gender has nothing to do with sex, if a long lay and scholarly tradition that it is a matter of sex allows speakers to see it as a marker of particular values and meanings. This is precisely what is happening, for example, in Konkani and in French. The scholarly tradition we have been discussing, a tradition that spans the centuries from Aristotle to nineteenth-century comparative philology, is founded on a belief that grammatical gender has a natural basis; it is determined not by form but at some deeper level by meaning, and may reflect the reality of masculine and feminine attributes. As we have seen, linguists no longer hold this belief, though lay languageusers appear to subscribe to it still. A combination of factors, some linguistic and some social, have caused it to be abandoned. At the same time, however, there has been a symmetrical shift in linguists' beliefs about natural gender. Instead of claiming that grammatical gender is 'really' natural, modern linguistic theory maintains that at least some aspects of natural gender are 'really' grammatical after all. The point of this manoeuvre is to deny that ideological factors enter into gender assignment. Gender is 'natural' in a new scientific sense, not because it mirrors nature but because it inheres in the structure of the language. Like other linguistic analyses, this one did not originate in a social vacuum. It is in fact a response to feminist criticisms of 'he/man' or 'sexist' language. In the last few years a considerable body of work on this topic has been produced (cf. Miller and Swift 1979, 1980; Spender 1980; Martyna 1980), most of it dealing with English, and one major source of feminist discontent is precisely the use of the masculine gender to denote generic or unspecified referents. Commentators observe that these alleged generics are frequently either misleading (as in the case of abortion statutes which refer to the pregnant 'person' as he} or else they are not true generics at all: Man is unique among the apes because he grows a long beard, and it is to this that he owes his superior intelligence.2
Feminists have used such examples to illustrate the absurdity of English gender and have called for reform on the grounds that it will make the language more accurate (Miller and Swift 1980: 8). Moreover, they can muster considerable theoretical support for this aim from the standard linguistic account. If gender in English is natural, determined by sex reference alone, how can anyone object to the feminist project of not using the masculine to refer to females? The reformers are only trying to secure what descriptions of English maintain is actually the case: 100% congruence between gender and sex. In fact, however, linguists do not generally support the feminist cause, and some have opposed reform in the strongest terms, developing two separate arguments, both of which condemn it for 'purely linguistic' reasons. This is important: opposition 2
Women's Journal, December 1981.
278
Debbie Cameron
must be seen to be on the side of reason, above ideology. As indeed must language itself. The first argument against reform might seem to be more characteristic of lay ideas about language than of expert scientific opinion. It is summed up by a recent newspaper article thus: Only Canute's courtiers would deny that language is a living thing which has been adapting organically to changing needs and mores since personkind's first Neanderthal grunt. (Baistow 1982)
Apparently linguistic reform is unnecessary, since faced with a disputed territory like the English pronominal system, 'language' will eventually resolve matters in accordance with 'changing needs and mores' (whose needs is another question). This view is also apparent in the pronouncements of certain linguists, albeit with less emphasis on the 'living organism' notion of language: As more and more women have taken up activities formerly reserved for men, languages that make gender distinctions in their nouns have had to accommodate their occupational titles. As long as it was unthinkable for anybody but a man to be a doctor, the word for 'doctor' could remain comfortably masculine. But the new state of affairs forced a decision: either feminines had to be coined to match every masculine, or the gender of the noun had become a formality. The languages facing this problem - Polish for example - are still trying to decide in each case which of two variant forms to adopt. (Bolinger and Sears 1981: 256-7)
It seems the linguist is unwilling to consider that language both affects and is affected by norms and political structures. For changes of the kind described here do not just magically happen: they emerge, rather messily, from a process of argument and struggle. Polish speakers, like speakers of current English, are likely to be aware of the conflicts brought about by social change, and will be influenced in their discussions of linguistic usage by the factors we have already noted in relation to French occupational titles, for instance a feeling that the feminine title downgrades the office and draws undue attention to the sex of the incumbent. (The same thing is often said of alternants like chairman/woman/person; the feminine is felt to sexualise the office in a way the masculine does not. The reasons for this will be discussed below.) Clearly, such considerations are ideological and not primarily linguistic, but this is concealed by the phraseology which has 'languages' facing 'problems'. The complex forces that shape and control language cannot be acknowledged, for it would then be difficult to deny that sexism was one of them. The second anti-feminist argument is altogether different, and it explicitly contests the usual equation of English gender and sex reference on which, it will be recalled, the feminist argument itself implicitly relies. According to this argument 100% fit between sex and gender is a goal as undesirable as it is unattainable, because gender in English is merely grammatical. A summary of this position comes from the linguistics faculty at Harvard University. Theology students had protested about the use of masculine pronouns to refer to God, and about 'sexist' language in general; they were answered in the university newspaper, the Harvard Crimson, as follows:
What Has Gender Got To Do with Sex?
279
Many of the grammatical and lexical oppositions in a language are not between equal members of a pair, but between two entities one of which is more marked than the other . . . For people and pronouns in English, the masculine is unmarked and hence is used as a neutral or unspecified term. The fact that the masculine is unmarked in English (or that the feminine is unmarked in the language of the Tunica Indians) is simply a feature of grammar. (Miller and Swift 1976: 92)
This new account claims that there is no ideological significance in the fact that the masculine is used as a generic: it is a linguistic fact and completely accidental (since after all, other languages use the feminine in the same way). The argument has been repeated several times when the question of sexist language is treated in the sociolinguistic literature (cf. Dubois and Crouch 1978; Crystal 1984). Resting as it does on the theoretical notion of marking many linguists find it a convincing argument; so it is worth examining the meaning of the terms marked and unmarked in the context, and looking more closely at the justification for using them at all. Given that the 'marking' argument undermines the standard account of English gender as semantic and not formal, it is surely of great interest that it has been accepted so speedily by many linguists. The notion of marked and unmarked entities can be used to make either a strong claim or a weaker one. The strong claim concerns universal tendencies in natural language: it is alleged that some elements turn up more frequently or earlier than others in the languages of the world. For instance, it is more common for languages to have the vowel /u/ than the vowel /y/. Languages with /y/ will normally have /u/ as well, whereas the reverse is not true. Thus vowel systems with /y/ are 'more marked' than those without. The Harvard faculty's claim about gender cannot be of this strong variety however, because some languages are said to have the masculine unmarked and others the feminine. Theirs is a claim about the grammars of single languages. What then is meant by asserting that 'for people and pronouns in English, the masculine is unmarked'? This question can be answered by looking at the use other linguists have made of the notion unmarked in relation to single systems. One criterion for labelling some variant unmarked is precisely that the variant in question can be used generically, i.e. in a way that subsumes marked variants. In the case of English gender this is circular: the masculine is generic because it is unmarked, but it is unmarked because it is generic! We are no further on with the argument. A more promising and indeed commoner criterion is that of relative neutrality of meaning. This is used for instance by Halliday in his treatment of English intonation (Kress 1976: 229) where he states that a falling nucleus on the final lexical item of a tone group co-extensive with a clause is the maximally neutral option, and thus unmarked: all other possibilities are defined partly by contrast with this neutral one and thus are marked. But if we apply this idea to the case of English gender, we encounter a puzzle: why is the meaning of a masculine form more neutral than that of the corresponding feminine, and for whom? Certainly many speakers do feel the masculine to be less specific, as has already been observed in the case of chairman vs chairwoman. In common usage, a woman can be a chairman but a man cannot be called a
280
Debbie Cameron
chairwoman. But how would the Harvard linguists prove that this is due to some inherent property of English grammar? The truth of the matter must be that the unmarked status of the masculine gender in English is not 'simply a feature of grammar'. In fact there is a great deal of evidence bearing on the origin of the generic masculine, at least where pronouns are concerned: Bodine (1975) documents the various arguments and prescriptive writings by which the 'neutrality' of the masculine gender was established. The precedence of masculine over feminine forms was asserted by Thomas Wilson as early as 1553, and in 1850 the view that 'man embraces woman' (i.e. that the masculine gender is the more comprehensive) found its way into the statute book. But it was to nature and not to grammar that prescriptivists appealed: grammatical custom and practice actually had to be changed in order to accord with the dictates of 'natural' male superiority. There was, and indeed still is, a concerted effort on the part of prescriptivists and pedagogues to eliminate the generic they of much ordinary speech. To claim that the unmarked status of masculine pronouns is a feature of grammar confuses description and prescription by naturalising a historical process whose determinants were social and political. It might be objected here that while Bodine's research casts doubt on the markedness argument as it relates to pronouns, it does not touch the other category mentioned by the Harvard linguists, namely nouns denoting persons. It has already been pointed out that, anecdotally, masculine job titles are felt to be more acceptable than feminine equivalents even in cases where the feminine form does not include a trivialising or diminutive affix like lady doctor or usherette. An examination of quality newspapers and BBC radio broadcasts confirms that the neutral pronoun formula, he or she, is more common than neutral titles such as spokesperson, sportsperson etc. Women are still quite frequently referred to as chairman, spokesman and so on; while any attempt to substitute chairwoman and spokeswoman fequently attracts comment to the effect that these feminine forms call undue attention to the sex of the individual concerned (this complaint is never made about the traditional masculine forms). The Harvard response to this would presumably be to claim that formally masculine nouns need not be semantically masculine. In the cases outlined they are capable of being truly generic and neutral as to sex reference. This argument is elaborated from an explicitly feminist point of view by Black and Coward (1981). Black and Coward argue that it is part of patriarchal ideology to equate women with sexuality, and that the consequent asymmetry between women and men (women are necessarily sexual creatures, men are not) is both reflected and recreated in linguistic representations. Thus masculine forms are neutral, because men are the sexless sex, the norm of humanity when sexuality is not at issue. But the cause of masculine neutrality lies essentially outside language, in more general matters of ideology and belief. There are several linguistic phenomena to which Black and Coward's analysis may usefully be applied. For instance, it may help to explain the phenomenon of 'semantic derogation' (Schulz 1975), a process by which originally unisex words, when semantically associated with women, become sexualised and pejorated (e.g. tramp, mistress, tart, harlot}. The analysis also throws light on data about common gender nouns in English. According to the standard account, these can be either masculine or feminine
What Has Gender Got To Do with Sex?
281
depending on sex reference in any particular context. But consider the following examples from newspapers: (1) The lack of vitality is aggravated by the fact that there are so few able-bodied young adults about. They have all gone off to work or look for work, leaving behind the old, the disabled, the women and the children.3 (2) A coloured man subjected to racial abuse went berserk and murdered his next door neighbour's wife with a machete, Birmingham Crown Court heard today.4 (3) Fourteen people, three of them women . . . 5
Here the 'common gender' words adult, neighbour and people are used as if they were inherently masculine, although they have no particular male connotations, either formal or semantic. If the words women and wife were to be replaced by men and husband, moreover, we would have three sentences that would never occur in a newspaper, and which most speakers of English would find rather odd. What are we to make of this? Clearly, it is not that the words adult, neighbour and people cannot be used of females: yet women may be treated as special cases, whereas men exemplify the norm. And this can hardly be 'simply a feature of grammar', for as Black and Coward observe, it is not just the gender which is marked here, but rather the category of persons to which it refers. The conventions of linguistic representation have evolved so that gender both marks and reinforces the anomalous position of women in the real world. It should now be clear that the relationship between sex and gender is a good deal more complex than the 'all or nothing' of conventional analyses, and is always mediated by ideas of sex difference and male superiority. So called 'natural' gender in English is not fixed entirely by sex reference but also reflects a variety of ideologically motivated prescriptive practices and folklinguistic beliefs. Conversely, the assignment of 'grammatical' gender in many languages does not depend only on formal criteria. There is a felt connexion with semantic properties of words, and these too are categorised according to ideas about sexual characteristics. As well as illuminating the relation between sex and gender, the above discussion sheds light on the relation between linguists and their data. Evidently there have been significant shifts in the way gender has been analysed and explained; the explanations popular at a given moment can be seen, furthermore, to be linked with particular attitudes to women. This suggests that linguists are perhaps less disinterested than they might wish. Their scholarship has social origins, and even political consequences. Yet so long as it is realised, this is no bad thing. For language itself has social origins and political consequences: ways of speaking and writing, and ways of thinking about those activities, are inextricably bound up with other forms of social organisation. Linguists have long attempted to minimise the implications of this by inventing abstractions to describe (langue, competence) and relegating 'social context' to the subdisciplinary fringes of 'pragmatics', 'sociolinguistics' or 'ethnography of speaking'. As the case of gender shows, this can lead to a dead end; it is because they saw language as an autonomous system unconnected with other kinds of behaviour 3
Sunday Times, 24 January 1982. Guardian, 9 March 1982. 5 Oxford Mail, 12 January 1982.
4
282
Debbie Cameron
that the Harvard faculty put forward a technical explanation of the generic masculine which ignored its historical roots in prescriptive practice and ultimately in ideology. In fact, linguists participate in ideology even as they obscure its workings. In denying the social basis for linguistic usage, they perpetuate an important part of our ideology: a facile and impoverished idea of language as a fixed, transparent code, perfectly adapted to human needs but at the same time completely detached from the conflicts of human existence.
References Arndt, W. (1970) Non-random assignment of loan words: German noun gender. Word 26. Baistow, T. (1982) Force-fed on pidgin pie. Guardian, 13 December. Black, M. and R. Coward (1981) Linguistic, social and sexual relations: a review of Dale Spender's Man Made Language. Screen Education 19. Bodine, A. (1975) Androcentrism in prescriptive grammar. Language in Society 4. Bolinger, D. and D. Sears (1981) Aspects of Language. Harcourt, Brace and World, New York. Corbett, A. (1983) Cherchez la metaphor. Guardian, 18 February. Crystal, D. (1984) Language in Church. The Tablet, 16 June. Dubois, B.L. and I. Crouch (1978) American minority women in sociolinguistic perspective. IJSL 17. Fodor, I. (1959) The origin of grammatical gender. Lingua 8. Janssen-Jurreit, M. (1982) Sexism: The Male Monopoly of History and Thought. Pluto Press, London. Kress, G. (1976) Halliday: System and Function in Language. Oxford University Press, London. Lyons, J. (1968) Introduction to Theoretical Linguistics. Cambridge University Press, London. Martyna, W. (1980) Beyond the he/man approach: the case for non sexist language. Signs 5. Miller, C. and K. Swift (1979) Words and Women: New Language in New Times. Pelican Books, Harmondsworth. (1980) The Handbook of Non-sexist Writing. Women's Press, London. Quirk, R., S. Greenbaum, G. Leech and J. Svartvik (1972) A Grammar of Contemporary English. Longman, London. Schulz, M. (1975) Semantic derogation of women. In Thorne and Henley (eds.) Language and Sex: Difference and Dominance. Newbury House, Rowley, Mass. Spender, D. (1980) Man Made Language. Routledge and Kegan Paul, London. Strang, B. (1968) Modern English Structure. Edward Arnold, London. Twain, M. (1880) The awful German language. In Teacher (ed.), The Unabridged Mark Twain, vol. 2. Running Press, Philadelphia.
20
What Makes Bad Language Bad? HAYLEY DAVIS
Introduction 'Bad language' is a subject which frequently attracts a certain amount of critical attention from linguists and from the media. The present paper analyses and compares three attempts by linguists to explain various features of the phenomenon. Two of these accounts take an integrationalist approach to the subject. All three are found wanting. The main points at issue may be identified by reference to the following well publicized case in the U.K. press. On 17 December 1987 The Independent printed an article which referred to a complaint made by Kelvin MacKenzie, the editor of The Sun. Mr MacKenzie was apparently disconcerted by an article that The Independent had previously published on 10 December 1987. The report in question referred to a confrontation on the cricket field between Mike Gatting, the captain of England's cricket team and Shakoor Rana, the Pakistani umpire. This incident culminated in Rana allegedly calling Gatting a 'fucking cheating cunt'. I wish to complain in the strongest possible manner at the disgusting language I found on the back page of The Independent today,
Mr MacKenzie's letter to the Press Council said. I know that all newspapers had this quote, but declined to use it as they were family newspapers. (The Independent, 17 December 1987)
This statement by MacKenzie raises a number of important questions. It appears that in this instance it is the publication of certain specific words that caused consternation. MacKenzie is 'disgusted' by the fact that a newspaper had printed this quotation; we are given no indication as to whether the editors or journalists who had transmitted it were similarly offended. Moreover it is interesting that Mackenzie was shocked by the printing of Rana's words (terms which many say denigrate women), but not by a report the previous day of Gatting's description of Rana as 'a shit umpire' (The Independent, 9 December 1987). This is doubly significant, in that it was Gatting's words and not Rana's which were the focus of attention on the cricket
283
284
Hayley Davis
pitch, and which led to the story being covered by the press in the first instance. The incident, The Independent reported, began on the field with Rana accusing Gatting of cheating and foul and abusive language.l He also alleged that Gatting had called him 'a shit umpire'. If so it most certainly went beyond the bounds of acceptable behaviour from an England captain. Rana claimed afterwards that it was the 'worst language' he had ever experienced from a player and said that the whole thing would be subject of an official complaint to the Pakistani Cricket Board, the Test and County Cricket Board and the England tour management. (The Independent, 9 December 1987)
Rana, in fact, 'refused to stand unless Gatting apologised for "abusive language". Gatting has said he will apologise if Rana apologises for calling him a "cheaf" (emphasis mine, The Independent, 11 December 1987). Whether this counted for Gatting as 'abusive language' is unclear. Although the actual words used were obviously at the heart of the incident, according to the editor of The Sun it was objectionable in this instance for a newspaper with a 'family' readership to tell its readers what words they were. The rationale behind this is not immediately clear. Most members of the community who have the skills of literacy required to read The Independent would almost certainly have had prior acquaintance with the words in question. It would be curious to object that perhaps some members of the community had never seen these word in print. For their printed form is presumably no more 'disgusting' than their oral form. Nor does their appearance in print in a newspaper in any obvious way sanction or validate their use in general, or imply approval of such use. (At the subsequent Press Council inquiry, MacKenzie maintained that such language acquired 'legitimacy' as a result of appearing in respectable newspapers; but the Press Council, which rejected the complaint, evidently did not agree.) Appeal to legal considerations throws no light on the matter either. It is difficult to maintain that the editor of The Independent might have contravened the Obscene Publications Acts of 1959 and 1964, since the test for obscenity is similar to that laid down in a mid-Victorian case, where it was held to be whether the matter in question had a tendency to deprave and corrupt those who were open to such influences and into whose hands it might fall. Under the recent Acts, the prosecution must prove that the matter, taken as a whole, would tend to deprave and corrupt persons likely to read or see or hear it; hence a book is not now obscene merely because it contains a few obscene words or sentences, (de Smith 1977: 479)
Nor could the editor be charged with the common-law offence of obscene libel. On 2 July 1954 in The Old Bailey, Justice Stable summed up a case in which Stanley Kauffmann's novel, The Philanderer, was a subject of prosecution on this charge. Remember the charge is a charge that the tendency of the book is to corrupt and deprave. The charge is not that the tendency is either to shock or disgust. That is not a criminal offence. The charge is that the tendency of the book is to corrupt and deprave. Then you say: 'Well, corrupt or deprave whom?' To which the answer is: 'Those whose minds are open to such immoral influences and into whose hands a publication of this sort may fall'. (Kauffmann 1952: 278) 1
Some reports allege that Gatting also called Rana 'a bastard'.
What Makes Bad Language Bad?
285
You have heard a good deal about the putting of ideas into young heads. Really, members of the jury, is it books that put ideas into young heads, or is it Nature? (ibid.: 280)
Thus, under the categories of this act, the worst that The Independent might have done was to 'shock or disgust'. Why this should be so simply brings us back to the question of why certain words are considered 'bad language'. At the Press Council hearing, Mackenzie submitted that The Independent should have printed c... instead of cunt. Why the former should, if correctly understood, be found less offensive than the normal spelling is puzzling. Mackenzie's contention in any case makes an interesting contrast with that of Sir William Emrys Williams in the Lady Chatterley trial. As a witness in his capacity both as a literary expert and a director of the defendant company, he claimed that to substitute asterisks for the fourletter words in D. H. Lawrence's novel 'would make the thing just a dirty book' (Encounter, February 1962, p. 101). Very little research has been done on 'bad language' as a linguistic phenomenon, and what has has been largely unsatisfactory. None of this research has dealt with the main issues which the reporting of the Gatting incident raises: namely, (a) the different attitudes to 'bad language' in speech and print, (b) why certain words, which are known to the majority of the population, arouse the strong reactions which they do, and (c) what it is about certain words which makes just those words objectionable; i.e. what makes 'bad language' bad. The following section of this paper examines three attempts by linguists to throw light on the problem of 'bad language'.
The sociolinguistic approach The orthodox sociolinguistic approach takes as its basic premise the assumption that the use of certain linguistic devices, in this case lexical items, is correlated in a fairly systematic way with social indices, (sex, class, age, etc.) and situational variables, (formal vs informal situations, etc.). The correlations are then 'explained' in terms of such causal factors as prestige (covert/overt), masculinity, permissiveness. Surveys conducted by sociolinguists on this issue had previously concluded that women swear less than men. The explanation offered was that since such language is non-standard, women, being more conscious of propriety and upward mobility, try harder to avoid using such terms. Trudgill (1979) however, has shown that it is not simply a matter of sex, and that class is an important variable: 'working class' people, including women, tend to use more stigmatized forms. He claims that 'the type of word that is tabooed in a particular language will be a good reflection of at least part of the system of values and beliefs of the society in question' (Trudgill 1979: 30). When, however, he has to explain away the fact that taboo words often have synonyms which are not similarly tabooed, he claims this is due to an irrational response to the form of the word (ibid.: 31). Risch in a recent article (Risch 1987) studies the derogatory terms ('dirty' words) that women use to refer to men. Her method was to have three female interviewers who noted informants' responses to the following question:
286
Hayley Davis
There are many terms that men use to refer to women which women consider derogatory or sexist (broad, chick, cunt, piece of ass, etc. . . .) Can you think of any similar terms or phrases that you or your female friends use to refer to males? (Risch 1987: 355)
The interviewers set the ball rolling by mentioning a few terms that they themselves used. 279 responses were elicited from 44 students. 'A classification system for 50 variant terms emerged quite naturally from the data obtained' (ibid.: 356). Risch was thus able to use headings in classifying the data which included reference to 'birth', 'ass', 'dick', 'boys', 'animal', 'other' (sic). Her survey, she concludes, questions many assumptions that relate to females and the use of such derogatory terms. The results show that the standard/nonstandard distinction does not (pace Trudgill 1979: 30), have any explanatory force here, since the majority of the informants were middle class. Rather, she considers the important factor to be the total subculture of females. 'It is doubtful whether any response would have been elicited in the presence of male interviewers' (Risch 1987: 356). And so, the frequency and variancy of response that results from the study calls into question the assumption that women are more prone to use standard forms of speech, and suggests that the standard/non-standard distinction is more appropriately applied to the contrast between public vs private discourse than to that of the speech patterns of women vs the language use of men. (ibid.: 353)
Like Trudgill, however, Risch explains this phenomenon in terms of covert prestige. 'It could be that nonstandard linguistic varieties carry some value of covert prestige for communities of young females' (ibid.: 357). Risch apparently sees nothing odd in the thesis that 'bad language' is non-standard, and the associated implication that it is impossible to swear in standard English. Risch's explanatory problems are compounded by her covert semantic assumptions. She appears to believe that every English word has a determinate and inherent meaning: hence some uses of 'dirty' words in her data have the appearance of semantic anomaly. She is puzzled, for instance, by the application to males of terms she regards as having meanings semantically appropriate to females, and vice versa. Having overheard a young male student on the campus referring to a female as a 'dick' she suggests that perhaps these terms are losing their allotted features of [+ male] or [+ female]. Finally, Risch does not explain why some of the responses, such as 'hunk' or 'mama's boy', should be counted 'dirty words' at all. Part of the trouble with her analysis is that it does not even entertain the possibility that not all dirty words are dirty all the time. She rests content with the vague claim that they are 'forbidden in most communicative contexts' (ibid.: 353). But this will not do. Once the importance of context is realized, one is led to see that any approach of the orthodox linguistic variety has no means of coming to grips with the underlying question 'What makes bad language bad?' Rather it assumes the existence of 'bad language' as a sociological given, and endeavours to account for its use.
What Makes Bad Language Bad?
287
The feminist approach Cameron in Feminism and Linguistic Theory clearly recognizes the importance of context and is critical of the orthodox approach, which she claims oversimplifies linguistic explanations. 2 Instead of viewing language as an abstract object, which is an ineffectual position for an active feminist to adopt, she suggests looking at how language is used to oppress women. Cameron advocates integrationalism as the theoretical foundation most congenial to her own perspective. The orthodox linguistic view of meaning sees language users looking up meanings in some internalised dictionary. The integrational view, on the contrary, sees them as creating meaning in specific contexts, negotiating where necessary in order to achieve as fully as possible their communicational aims. (Cameron 1985: 140) There is pretty well no limit to the novel situations humans may encounter, and therefore the communicational demands which may be made on them are almost limitless. To meet these demands, speakers and hearers engage in a constant renewal of language. They employ words in a flexible, innovative and often playful way, and are creative in their interpretations of other speakers, (ibid.) It is clear that without the indeterminacy that stops us communicating telepathically we would not be able to adapt our language to the novel situations we need it for; imperfect communication is the price we pay for a creative and flexible symbolic system, (ibid.: 142)
Therefore, she says, 'male control over meaning is an impossibility'. And since conversation is a 'highly contextualised phenomenon', it is unwise for us to 'generalize about it on the basis of so gross a variable as sex' (ibid.: 42). However, these statements are difficult to reconcile with her opinions on verbal taboos: generally speaking, taboo words tend to refer to women's bodies rather than men's. Thus for example cunt is a more strongly tabooed word than prick, and has more tabooed synonyms. Even words like bugger and arsehole whose reference is male are insulting because they connote homosexuality, which is not only taboo in itself but associated with femininity as well, (ibid.: 76)
Such words, she continues, are verbal violence against women, expressing both our essential qualities in patriarchy (repositories of sexuality, prostitutes) and male women-hatred, which makes women afraid, (ibid.: 78) The existence of so many insulting words for women, many of them meaning the same thing, has a significance over and above what it tells us about cultural beliefs. It is, in fact, itself a form of social control, (ibid.: 77)
This position however, is open to just as many objections as the orthodox sociolinguist's approach. It is prima facie unclear how, for example, umpire Rana 2 Since the first publication of this article, Cameron has revised her book Feminism and Linguistic Theory. The second edition (1992) seems to have taken into account many of the criticisms made here and elsewhere of the 1985 edition.
288
Hayley Davis
could possibly be trivializing and having control over women by calling Gatting a 'cunt'. Or how The Independent could be similarly guilty by reporting this. Furthermore, to state that some taboo words have a 'male reference' apparently endorses a semantics which postulates invariance and determinacy of lexical meaning. Cameron seems to ignore usages of the kind exemplified in 'that was a fucking good game' or 'it's been a bugger of a day', where the claim that 'male reference' is involved is quite unconvincing. It may be that although Cameron advocates an integrationalist approach to language she has not fully grasped the theoretical implications of integrationalism. It involves rather more than the thesis that speakers 'create meanings in specific contexts', employing words 'in a flexible and innovative and playful way and are creative in their interpretations of other speakers'. For all this amounts to is a rejection of determinacy of meaning. Cameron still accepts the determinacy of form assumed by the orthodox view of languages as autonomous objects. For her taboo words seem to be determinate formal entities (bugger, fuck, etc.). Integrationalism, however, rejects both determinacy of form and determinacy of meaning. An integrationalist, rather, considers the contextualized totality of the communicative experience, which necessarily entails considering the linguistic and the non-linguistic as interrelated aspects of the situation, along with the assumptions the participants make about their own and others' intentions. From the integrationalist point of view, 'sexist language' is not a linguistic category that can be formally identified (e.g. by reference to a dictionary, or by word lists), but an analytic category of one's individual linguistic experience. Cameron's claim, therefore, that 'imperfect communication is the price we pay for a creative and symbolic system' is, from an integrationalist point of view, rather mystifying: we are given no indication as to what 'perfect' communication would be like or what it would entail. Cameron, despite her claims to the contrary, is here espousing linguistic orthodoxy. This is borne out by her view as to how a feminist linguistic theory should be construed. A feminist linguistic theory, in my opinion, is a theory that links language with sex in two ways: it spells out the connection on the one hand between language and gender identity, and on the other hand between language and women's oppression. If we are to have a useful dialogue, therefore, all feminist linguistic theories must make it clear where they stand on four basic questions. First, what are we talking about when we talk about language? Secondly, what do we mean by women's language (or indeed men's language)... (ibid.: 163) We cannot leave the definition of language to the vagaries of context as we would in ordinary conversation, because we are trying for something less hit-and-miss than the cross-purpose talk of an ordinary exchange, (ibid.: 164)
This 'solution', however, is at odds with integrationalism, which holds no brief for such categories as men's language or women's language. Cameron's avoidance of 'offensive and sexist language' (ibid: viii) is likewise questionable. Although, Cameron claims, men cannot control meaning in the sense that there is nothing to control, they have controlled metalanguage. The fact that our education system has on the whole been regulated by men, especially in the writing of grammar books, dictionaries and statutes, has left us with the view of languages as fixed codes.
What Makes Bad Language Bad?
289
The institutions that regulate language use in our society, and indeed those of most societies, are deliberately oppressive to women. Men control them, not in the rather mystical sense that they are said to control meaning, by making esoteric semantic rules or possessing the vital signifier, but simply because it is the prerogative of those with economic and political power to set up and regulate important social institutions, (ibid.: 145)
An examination of some of the Oxford English Dictionary's entries for stigmatized words excluded from earlier editions in question supports Cameron's claim. fuck: . . . for centuries and still by the great majority, regarded as a taboo-word; until recent times not often recorded in print but frequent in coarse speech. V - trans use (rarely used with a female subject) To copulate with; to have sexual intercourse with. Used profanely . . . coarsest equivalent of DAMN . . . N - A person (usu. a woman) considered in sexual terms. cunt: 1) The female external genital organs. Its currency is restricted in the manner of other taboo-words. 2) Applied to a person, esp. a woman, as a term of vulgar abuse. Vb . . . act of copulation . . . a person, usually a female considered in sexual terms.
These are manifestly men's definitions. Most dictionaries which have recorded these words define them in a similar fashion (Cf. Partrige's Dictionary of Historical Slang). Grose's revised Dictionary of the Vulgar Tongue defined 'F k', merely as 'to copulate' and 'C**t' lightheartedly (?) as 'A nasty name for a nasty thing'. Although these definitions might be held to support the claim that at least some common tabooed terms are recognized - even by men - as being offensive to women, they throw little direct light on the nature of the offence. It is not as if it were impossible to imagine a society in which such terms were complimentary rather than derogatory.
The speech-act approach Harris (1987) attempts to deal with the ontology of swearwords. In so doing he offers an explanation for the prohibition on their mention as well as their use. His differs from previous accounts in considering it necessary, in order to understand this double prohibition, that one must first explicate the speech act of swearing. He criticizes the two main explanations of why bad language is bad. The first, which he terms the 'Graeco-Roman' account is given its classic formulation by Plato: If one is disputing with another in argument, he shall either speak or listen, and he shall wholly refrain from abusing either the disputant or the bystanders. For from these light things, words, there spring in deed things most heavy to bear, even hatreds and feuds, when men begin by cursing one another and foully abusing one another in the manner of fish wives. (Laws, XI, p. 461)
This, Harris claims, leaves a number of facts unaccounted for, merely saying that bad language is bad because it is anti-social, associated with an inferior class of person and likely to cause 'a breach of the peace'. It does not explain why an
290
Hayley Davis
embargo on the use of such words extends also to mentioning them. Nor does it explain why the actual class of swearwords includes the expressions that it does. 'Swearing (in language L) involves using the expressions a, b, c, etc. Why those? Because, damn it, those are the recognized swearwords in language L' (Harris 1987: 181). The other view, the 'Judaeo-Christian' view, condemns language which 'takes the Lord's name in vain'. This, apart from not being able to solve the puzzles unanswered by the 'Graeco-Roman' account, does not cover the notorious 'fourletter' words. Harris's own hypothesis gets off the ground by comparing the incident reported by Thomson of a two-year-old aboriginal child exclaiming 'Devil! Excrement foul! Excrement foul!' when interrupted in the act of suckling, with a monkey signing 'dirty Roger' when being refused to be let out of her cage, and a Bloomfieldian child saying 'I'm hungry' to delay bedtime. From these seemingly disparate situations Harris concludes that 'semiotic displacement is the operative mechanism in the ontogenesis of swearing' (ibid.: 185). In each case, he continues, 'it is a naive confidence that one is in a possession of a verbal formula which is bound to produce the desired result'. 'It therefore represents a primitive attempt to extend the use of learned signs as an indirect means of control over one's environment' (ibid.). As it is a strategy based on deception, 'Both fear and guilt carry over from the primitive situation to the socially established practice, and account for various irrational' aspects of aversion to swearing, concomitantly with its widespread survival' (ibid.: 186). This, Harris claims, explains the fact that when such words become socially institutionalized, 'swearwords become unmentionable precisely because institutionalized swearing is the unique and marginal case where locution and illocution are one: the utterance is the deed and the deed is the utterance' (ibid.: 187). However, Harris's account closes one explanatory gap only to open up others. It seems, in fact, that Harris has as odd a conception of swearing as Wittgenstein had of games (Wittgenstein 1953: 83). For it is unclear in what sense a report in The Independent, or the mention of swearwords in this article, is swearing. Yet it seems that for Harris it is 'a necessary and sufficient condition' of swearing that it involves the use of swearwords. (Pythagoras could not have sworn by the number four unless the Greek of his day already had 'swearwords' as a lexically institutionalized category (Harris 1987: 186).) Thus even the mention of institutionalized swearwords is swearing, come what may. 'To say "Damn!" is to swear whether you intended to or not' (Harris 1987: 187). This makes it extremely difficult for Harris to explain the fact that there are some words which are only considered to be objectionable in certain contexts (cock, prick, God, Christ, etc.). Harris also needs to explain how the ontogenetically primitive mechanism becomes the developed social practice. For the task of deciding when a sign has been 'displaced' or 'misused' is fraught with theoretical problems. Moreover, granted that there are some cases in which we could call X 'a deliberate misuse of a sign', we are given no codes of practice to decide whether we are witnessing an instance of swearing. For example, calling someone a 'pig' involves semiotic displacement and this particular sign has become lexically institutionalized in our language. But is it a swearword?
What Makes Bad Language Bad?
291
Discussion There are at least two different types of behaviour which are subsumed under the term swearing. It is important not to conflate them. In the passage from Plato which Harris cites we are not being informed about swearwords, but about abusing someone in a quarrel. It is not the use of any specific term which causes offence. The Greeks and Romans did not seem to have this kind of taboo on any words qua words. Rather it is the illocutionary intention or perlocutionary effect that is being discussed in the Laws. The same is true of the Judaeo-Christian view. The uttering of God's name in itself was perfectly acceptable, but when his name was taken in vain this incurred 'divine displeasure'. Referring to profane swearing, Sagarin writes: the individual words are not tabooed, what is prohibited is their utterance in combination with other words, in a given context, to convey a thought or to incite an action which is condemned in society. (Sagarin 1962: 23)
Originally, as Sharman and Silving tell us, swearing was a declaration of sincerity. With the ancients oaths were employed in guarding as efficiently as they could the public conscience and public security . . . To render the practice less capricious and incontinent, a notion of an individual property or trade mark in oaths came to be perceptibly encouraged. The specific appropriation of some distinctive oath raised the presumption that it implied an unequivocal pledge of sincerity. (Sharman 1884: 68)
The swearing of oaths in court has its ultimate rationale in the presumed sincerity of the commitment undertaken 'under oath'. However, even this solemn form of swearing was in certain periods held to be objectionable, and citations were given from the Bible as evidence in support of this view. Swear not at all; neither by heaven; for it is God's throne. Nor by the earth; for it is his footstool; neither by Jerusalem for it is the city of the Great King. Neither shalt thou swear by thy head, because thou canst not make one hair white or black. But let your communication be, Yea, Yea; Nay, Nay: for whatsoever is more than these cometh of evil (Matthew, 5, 33-37)
It is not until 1601 in England that we find the first Parliamentary legislation against 'usual and common swearing'. This was introduced in the Commons but dropped in the Lords, and it may be significant that this occurred at the time when there was a great deal of religious unrest. It also coincided with the impact of printing on the dissemination of texts, which had the effect of making people more aware of linguistic standards. Literature of the 17th century leaves no doubt that swearing as a social habit was as prevalent as it was widely deplored. Bunyan, for instance, in The Life and Death of Mr Badman included a diatribe against swearing. 'Vain and sinful swearing is light and wicked calling of God, etc., to witness to our sin a foolish attesting of things' (Bunyan 1680: 164). However, one sees that still the actual words used are besides the point, it is the affront to God that is the sin.
292
Hayley Davis
It was not until 1745 that the first law in England was passed against swearing. Fines were imposed both for 'taking God's name in vain in common discourse' and for 'jestingly or profanely' using the name of the holy trinity, 'or any of the persons therein' in any stage play or show (Blackstone 1769: 59-60). Blackstone explains this as follows: the temporal courts resent the public affront to religion and morality on which all governments must depend for support. (Blackstone 1769: 58-9)
Its introduction coincided with the first major attempt to standardize, refine and 'fix' the language.3 With the changes in English social values in the nineteenth century, attitudes towards swearing changed. Social propriety began to take precedence over religious susceptibilities in the condemnation of 'bad language'. The class of objectionable expressions expanded accordingly to include designations of certain parts of the body and bodily functions. These changes also coincided with the establishment of Victorian lexicographical theory and practice. Fuck and cunt had been included in dictionaries of an earlier age (e.g. in John Florio's A Worlde of Wordes (1598), Stephen Skinner's Etymologicon Linguae Anglicanae of 1671 and Nathanial Bailey's Universal Etymological English Dictionary of 1721. But the inclusion of such words was felt to be incompatible with the new status of the dictionary as a historical repository of the 'best' usage and an authoritative instrument for universal education. A fact overlooked in the linguists' analyses examined above is that the first attested use of the term swear-word was not until 1883. This calls in question whether the swearwords as such enter the sociolinguistic picture at all before the late nineteenth century. In brief, the recognition of a class of 'swearwords' arises in a social situation in which it is taken for granted that such words are not to be included in dictionaries and are therefore banished from use in reputable publications. The exclusion becomes part of an implicit definition of what a swearword is, which is in turn part of a certain view of the function of language in society. The emergence of the Victorian concept 'swearword' (as distinct from the concept 'oath') reflects a social conscience which would like to believe that language is a direct reflection of morality, but is uncomfortably aware in its own case of an awkward discrepancy between proclaimed standards of behaviour and common linguistic practice. In order to accommodate this discrepancy, swearing is assimilated to the commission of other linguistic faults (such as misspellings, mispronunciations, mistakes of syntax). The swear-word is not a new part of speech but a new linguistic lapse. It is this assimilation which makes 'bad language' bad language (as distinct from offensive behaviour) and at the time (as in the case of 'bad grammar') holds out the prospect of its eventual elimination by education. Whether they realize it or not, sociolinguists who currently interpret 'bad language' as a variety of non-standard usage are perpetuating a typically Victorian form of linguistic prescriptivism; while theorists who conflate the history of swearing with the history of swearwords make the complementary mistake of projecting the same Victorian prescriptivism anachronisticalry back into the past. 3 See Baugh and Cable (1951).
What Makes Bad Language Bad?
293
We are now living in the immediate aftermath of this particular episode in the social history of English and the linguistic perplexities consequent upon further changes in social values. The 'swearword' is still recognized as a lay linguistic category, but it has a number of interesting features. Although it is a category supported by a special metalinguistic terminology (including such expressions as four-letter word, dirty word, effing, blinding, etc.) it is essentially an open-ended category. Preliminary research (cf. Davis 1992) suggests that currently there are many areas of disagreement among English speakers over the classification of particular vocabulary items as swearwords. What is not in doubt, however, is the importance of the role which this category plays in the conceptual framework within which 'swearing' is situated as a possible speech act in contemporary English, and the no less important role which is assigned by interlocutors to contextual factors when identifying such speech acts. What makes bad language bad is a question on which more research is needed than linguists have hitherto suspected. This is yet another case where the 'linguistic facts' of the matter are far more complex than the linguistic accounts currently on offer would lead one to suppose.
References Bailey, N. (1721) Universal Etymological English Dictionary. London. Baugh, A. C. and T. Cable (1951) A History of The English Language, 3rd edn. Routledge & Kegan Paul, London. Blackstone, W. (1769) Commentaries on the Laws of England 1765-1769. University of Chicago Press, Chicago. Bunyan, J. (1680) The Life and Death ofMrBadman. J. M. Dent (Everyman Edition), London. Burchfield, R. (1972) A Supplement to The Oxford English Dictionary. Clarendon Press, Oxford. Cameron, D. (1985) Feminism and Linguistic Theory. Macmillan, Basingstoke. Davis, H. G. (1992) Words: an Integrationalist Approach to Word-Identification and WordCharacterization in a Literate Society, D.Phil, thesis, University of Oxford. Florio, G. (1598) A Worlde ofWordes. London. Grose, F. (1811)A Revised Dictionary of The Vulgar Tongue. Papermac, London. Harris, R. (1987) Mentioning the unmentionable. International Journal of Moral and Social Studies 9 17^ 88 .£, 1 / j-oo.
Partridge, E. (1984) Routledge Dictionary of Historical Slang and Unconventional English, ed. by P. Beale. Routledge & Kegan Paul, London. Plato (192 ) The Laws of Plato, ed. E.B. England. Manchester University Press, Manchester. Kauffman, S. (1952) The Philanderer. Penguin, London. Risch, B. (1987) Women's derogatory terms for men: that's right 'dirty words'. Language in Society 16, 353-8. Sagarin, E. (1962) The Anatomy of Dirty Words. Columbia University Press, New York. Sharman, J. (1884) A Cursory History of Swearing. Reprint 1968. Burt Franklin, U.S.A. Silving, H. (1958) The oath. Yale Law Journal 63, 329-90. Skinner, S. (1671) Etymologicon Linguae Anglicanae. London. Smith, S. A. de (1977) Constitutional and Administrative Law. Penguin, London. Sparrow, J. (1962) Regina v Penguin Books Ltd. Encounter 18, 35-43. Trudgill, P. (1979) Sociolinguistics. Penguin, London. Wittgenstein, L. (1953) Philosphical Investigations, translated by G.E.M. Anscombe. Basil Blackwell, Oxford.
21
Law Lessons for Linguists? Accountability and Acts of Professional Classification CHRISTOPHER HUTTON
The rise of new disciplines and new areas of academic specialization is often interpreted as a sign of vitality. The rapidly expanding area of language and law can be viewed as a welcome indication that linguists are increasingly interested in the complex ways in which societies conduct their linguistic business. But this search for involvement carries with it a danger, for in seeking to explain society's needs to society, or in assuming the mantle of professionalism in relation to a particular set of social practices, linguists potentially expose their own in-house practices and beliefs to the scrutiny of a wide range of professionals in fields outside academia. They also make themselves accountable in some general sense to the general public. The question is: can linguistics as a whole meet this challenge? By 'linguistics' I mean in this context the discipline as a whole, i.e. the academic mainstream with its core of theoretical assumptions. In particular, I am thinking of the contributions by linguists to the recent Washington University Law Quarterly volume 'What is meaning in a legal text?' (Proceedings 1995). Members of the professions such as lawyers, doctors, psychiatrists and social workers work within many different kinds of constraints in their professional lives. They must meet standardized accreditation criteria, and are judged by their interaction with processes and procedures that have real world outcomes (e.g. trials, operations). They need to master a range of different professional and lay discourses in order to talk to fellow professionals and to be able to communicate about their 'in-house' knowledge to the general public, their clients. Doctors, for example, talk both to other doctors and to patients about medical matters. Lawyers talk both to other lawyers and to their clients. While it is true that linguists talk to other linguists, their 'clients', as such, are typically students at various levels, and usually there is no real analogue in their professional experience to doctor-patient interaction. Students are socialized into the language of the linguist; they are novice professionals, at least potentially. 'Pure' specialists such as academic linguists are individually and collectively judge and jury over teaching and research in their own discipline. Thus, while academic linguists might consider themselves professionals, and might even be regarded as such by non-academics (for example in marketing surveys, sociological categorizations, etc.), they rarely have the kinds of professional skills that are exercised at the intersection of 294
Law Lessons for Linguists?
295
academia and society. In this sense, they are not professional experts: their professional standards and professional knowledge are largely discipline-internal. Linguistic theories are created, debated, rejected or affirmed primarily by other linguists, and career advancement in linguistics, as in other academic careers, depends on peer review, on what 'the field' has to say. This is not necessarily to imply the theoretical beliefs of doctors or lawyers are better grounded in empirical reality than those of linguists; rather that there is no automatic challenge to linguists' professional language that arises out of their daily professional experience. They do not have to explain themselves, or what they are doing, to others qua specialists. Academic linguists do occasionally volunteer to cross over the boundary of their own intra-disciplinary language and write accounts of language and linguistics for a general audience. Two recent examples of the genre are Deborah Tannen's works on discourse style, You Just Don't Understand! (Tannen 1990) and Talking from 9 to 5 (Tannen 1994), and Steven Pinker's The Language Instinct (Pinker 1994). Tannen's orientation is avowedly therapeutic; the point of her works for the general audience is to enable people to communicate better, to become aware of different discourse styles and to improve the quality of people's lives. Pinker's book, of course, makes no such claim for itself; its problem is how to translate the language of formal academic linguistics into a language which a lay reader can understand. In a review of Pinker, Roy Harris (1995) argued that this dilemma creates, in effect, two rhetorical voices, which he labelled 'Pinker the Popularizer' and 'Pinker the Psycholinguist'. Given the task Pinker has set himself, these two voices can never be saying the same thing: 'The problem for Pinker the Popularizer is that he has set himself an impossibly Quixotic task. He is trying all the time to translate Mitspeak the arcane algebraic jargon of MIT linguistics - into everyday English.' But, concludes Harris, 'there is no translation' (Harris 1995: 274). For Tannen, the risk of simplification is balanced by the benefits of reaching a larger audience and the possibility of effecting useful change. Her writings for the general public have however been criticized for taking a relativistic view of male and female discourse styles, and for thereby downplaying sexism. In a reply to a critic of You Just Don't Understand! (Troemel-Ploetz 1991), Tannen recalls the 'truism that a scholar who breaks rank by writing for the wider community will incur the wrath of academic colleagues' and offers this factor as an explanation for Troemel-Ploetz's 'hate-filled diatribe' (Tannen 1992: 252-3). Tannen also argues that there is no discontinuity between her views as expressed in her academic writings, and those as expressed in her writings for the general public, making the point that 'the framework comes directly from the work of John Gumperz' (Tannen 1992: 252). Whatever the rights and wrongs of this particular case, it is far from clear that Tannen, any more than Pinker, can remain both entirely faithful to her academic discourse and categories, while at the same time addressing directly a lay audience. There must be two Tannens, just as there are two Pinkers. This is not to take anything away from Tannen's success in bridging the worlds of academic linguistics and self-help literature. But, while the distance in terms of academic discourse is greater in Pinker's case than in Tannen's, there is nonetheless a process of translation. Tannen argues that her expertise in therapeutic linguistics derives from, or is grounded directly in, her expertise in academic linguistics. But the demands of the therapeutic genre, not least as a rhetoric that asks the reader to change his or her
296
Christopher Mutton
behaviour, make that impossible. To make the book 'popular', it must appeal to the general reader, or to those readers who understand themselves as sharing the basic daily problems of society. The therapeutic context demands this, not least because diagnosis is part of cure, and diagnosis is performed by the reader as an act of selflabelling. The therapeutic text must relate to its audience in an entirely different way than the academic one, and the desire of the audience to recognize themselves in the text shapes the kinds of generalizations that the book must make. Therapeutic texts build on stereotypes and popular categorizations, reinterpreting them and explaining them in terms of the therapeutic theory. Tannen skilfully negotiates a path between the academic and the popular, but in order to communicate about language to a wider audience, she cannot entirely dismiss popular stereotypes and popular categorizations. To take a simple example, the communicationally challenged husband reading his newspaper at the breakfast table and ignoring his wife, is a popular stereotype of male behaviour, familiar from the cartoons of the New Yorker and the world of James Thurber (see Thurber 1945). Such cliches do not exist independently of actual behaviour. At the very least they are a lens through which behaviour is judged and classified. The therapeutic popularizer must use these images, perhaps reinterpreting them or rationalizing them in terms of 'culture' or hidden presuppositions, perhaps challenging them. But they cannot be ignored. There is nothing in academic sociolinguistics to tell anyone what a good relationship is, nor is there any framework within which one could evaluate the success or failure of a therapeutic intervention, or show why a particular marriage, or marriages in general, are worth saving. Tannen is not in the same position as a Freudian psychoanalyst who is 'professionally' trained to make evaluative judgments of this kind. This is not to say that Freudian therapy is better or worse than any other approach, including Tannen's, simply that there is no professional structure for what Tannen does within academic linguistics. Tannen has made her own innovative synthesis of academic linguistics and the self-help literature, and in that lies her considerable achievement, rather than in the direct application of the insights of linguistics to a set of social problems. One way to make this clear is to point to the parallels between Tannen's therapeutic writings and those of General Semanticists and their successors in areas such as Neuro-Linguistic Programming (NLP). The movement of General Semantics, founded by Alfred Korzybski (see Korzybski 1933), views language as an autonomous social force in its own right, one which needs to be scrutinized and guarded against, lest it mislead us into confusing words and things, into mistaking the map for the territory. All kinds of social ailments, including wars, are held to arise from 'semantogenic' causes. Recognizing this involves knowing that 'what causes the disorder in my life is language. Not things that "happen to me" . . . not things other people "do to" me . . . not faults I myself have or bad things I myself do ... but language' (Elgin 1995: 9; ellipses in original). The same presupposition underlies Tannen's account of communicational difficulties, in that she implies strongly that the existence of ethnic and gender styles is in and of itself a problem. In You Just Don't Understand! Tannen remarks that 'she understands and is in sympathy with those who wish that there were no differences between men and women - only reparable social injustice' (Tannen 1990: 17), implying that language creates a set of problems of its own. The recognition that, for example, there exist gender styles in
Law Lessons for Linguists?
297
conversation can prevent arguments over nothing, i.e. problems created purely by language. A couple who sincerely love each other may nonetheless have conflicts, since language does not operate as a clear window onto their thoughts and beliefs; rather it is a distorting lens. Understanding this will mean that 'at least you'll be arguing about real conflicts of interest rather than fighting styles' (Tannen 1990: 187). When we come to look at linguistics and the law, a set of similar issues and problems can be raised. While linguists and lawyers can and should debate the interrelation of their two specialties, it can nonetheless be argued that the central categories and beliefs enshrined in 20th-century linguistics cannot be directly applied to the understanding and interpretation of linguistic behaviour, actual 'texts'. While disciplines like stylistics have sought to establish a metalanguage deriving from linguistics, one to be applied to the interpretation of individual texts, the enterprise has at its heart a fundamental contradiction. Texts, in the widest sense of the word, are communicative acts which do not recur, and which require interpretation. Their interpretation can never be definitive, since we cannot lay down in advance what it is we wish to understand about a given text and how we should go about the task of interpreting it in light of that goal. Within linguistics, units such as words, sentences, speech acts, and conversational moves are all viewed as recurrent. Linguistic analysis identifies the recurrent elements in linear structures, and seeks to reduce all categorical ambiguity. Thus, a linguistic analysis which argued that a particular recurrent segment was, let us say, a preposition would, all else being equal, be better than one which argued that it sometimes was and sometimes was not. In linguistics, the way in which a text is approached is laid down by the fundamental aims of the discipline. Those aims include the investigation of linguistic structure, including conversational structure, and the identification of invariant elements. They do not include any way of grasping the unique aspects of a speech event. A number of objections might be raised to this questioning of the relevance of linguistic concepts for interpretative disciplines like literary criticism and the law. Firstly it could be argued that linguists manifestly do contribute to disciplines such as stylistics and forensic linguistics, and that their expertise in, for example, analyzing transcripts, is a worthwhile counterweight to the often highly questionable assumptions under which courts operate in interpreting quoted speech, or in reading transcripts. However, the point being argued here is that the theoretical concepts of linguistics, its core beliefs as a discipline, can have nothing to say to the question of what a particular utterance meant on a particular occasion. They simply do not provide expertise in this. This does not preclude academic linguists, like Tannen, constructing particular metalanguages or procedures in response to particular situations: giving evidence in court, looking at transcripts, identifying individual voices, etc. But these activities cannot consist simply of applying academic expertise in grammar or discourse to 'real life' situations. That expert discourse is expertise in the idealizations of linguistics, in the abstractions that linguists have made away from actual language behaviour. This done, linguists cannot then apply back their categories onto the interpretation of behaviour, for interpretative problems are precisely the element that is idealized out.
298
Christopher Mutton
Following on from this, the linguist might object that linguistic analysis is, in this, no different from chemical analysis. The concepts of chemists, philosophers of science tell us, are to no small extent built on presuppositions about the material world. These presuppositions, combined with the methodology employed in a given research instance, predetermine, to a considerable degree, the kinds of results that can be achieved. The scientific method is also set up so as to look for invariance, to look for replicability, and to idealize away from the contingencies of the individual case to reach generalizations about chemical processes. Technology is the application of these generalizations to a practical end. While granting this in principle, one would have to point out that there is no equivalent in linguistics to the scientist's replicable experiment. There is no parallel to publishing a set of results and having the method repeated by other colleagues in order to check whether the result obtained was due to a set of general processes, or a freak of the particular conditions on a particular occasion. There could be no controversy in linguistics of the order of the 'cold fusion' controversy. This is not to say that the scientific method crudely summarized here is not open to objections from philosophers of science, or that it guarantees objective truth; rather it is to point to a key difference between linguistics and the natural sciences. Linguistics methodology in the post-discovery procedure era consisted of the use of intuitions about whether linguistic structures were grammatical structures in a given language, and the application of theory-internal categories and theory-internal heuristics (notably the criterion of simplicity) to those data. Linguistic intuitions were a means into the system, not a means of analysing the system. In this sense linguistic analyses are subject neither to empirical confirmation or disconfirmation, nor are they accountable to social judgments about their correctness, relevance, plausibility or importance. Linguistics has, in effect, developed a new realm of nature for itself, treating language as sui generis and developing for it a correspondingly unique methodology and terminology, shared neither by the social sciences nor by the natural sciences. Linguistics is perhaps the most '19th century' of the academic disciplines taught in universities today. When the linguist asserts to the lawyer that, like chemistry and physics, '[w]e're a science too' (Michael Geis, quoted in Proceedings 1995: 821), the lawyer (Clark Cunningham) replies that '[l]aw hasn't claimed to be a science, at least since the late 1880s'. A further objection from the linguist might run as follows: it is true that linguistics involves the idealization of linguistic behaviour and looks to find certain invariant categories in that behaviour. But linguists no more falsify the complexity of reality than do chemists or lawyers. Courts have to decide for example if a particular set of actions on a particular occasion fall under a particular pre-determined category (let us say, 'burglary'), and they must use a complex set of operations and professional expertise to arrive at a judgment. That judgment may be disputed and is liable to correction to a degree (depending on the contingencies of the legal system, avenues of appeal, etc.), but there is no certainty that these and other similar acts of classification are correct. The application of legal categories and labels such as 'guilty' is, like the linguist's determination of the nature of a segment ('preposition'), a matter of professional expertise and experience and is constrained in both cases by indefinite but real checks and balances in the overall discourse. Decisions, professional acts of classification, are continually subject to the scrutiny of the relevant interpretative
Law Lessons for Linguists?
299
community, that of lawyers in the one case and linguists in the other. Neither lawyers nor linguists have a monopoly of the truth, and both could learn from each other, and benefit from the chance to examine each other's presuppositions about language. Lawyers, it could be argued, reduce the world to the categories that they need to operate the law, and try to identify those factors that are irrelevant. In the case of a trial, the procedure is set up so as to eliminate facts that are not germane to the matter at hand and to interpret those that are. It is a discovery procedure for the application of legal categories to the mess of everyday life. It substitutes 'guilty' and 'not-guilty' for the lay categories wrong and right, and does not pretend to capture or appeal directly to the lay sense of justice. Even those critics who question the gap between the law and justice would not argue that the procedures of the law should reflect fully a lay sense of justice or a simple moral duality between right and wrong, good and bad. Similarly, the linguist applies a certain set of categories to the 'mess' of the world, and does not, or at least should not, pretend that they exhaust the complexity of the phenomena under study. Nor would the linguist deny that these procedures and categories are often remote from lay conceptions of language. The issue at stake can be explored through Cunningham et a/'s discussion of Cardozo's 'famous paradox of law' as summarized by Solan (1993: 208, n. 1). Cardozo pointed out that the law must be both flexible enough to allow for new cases, but rigid enough to provide for stability of interpretation and reliable definitions of rights and responsibilities. However, Cardozo went a step further to depict an ideal state of affairs where this dilemma would not arise (Cardozo 1921: 143): No doubt the ideal system, if it were attainable, would be a code at once so flexible and so minute, as to supply in advance for every conceivable situation the just and fitting rule. But life is too complex to bring the attainment of this ideal within the compass of human powers.
For Solan (1993), as reported by Cunningham et al (1994: 1567) this wistful fantasy is realized in actuality, 'this "ideal code" describes the reality of language'. Solan sees language as an ideally flexible system that can cope with any new situation. Solan finds in ordinary language an ideal system that contains within it in advance all the possible worlds that it will be required to confront. Understandably, the authors are reluctant to endorse this view wholeheartedly, given, as they point out, that miscommunications do occur. For Solan's view, evidently, is a fantasy about language. If it were true, there would be no problem about interpretation in the law; nor in any of the many other areas of ordinary life, i.e. all, where problems of communication arise. Nonetheless, Cunningham et al do accept that the fields of syntax, semantics and pragmatics have made considerable progress, that progress consisting of 'finding and analyzing predictable order in the seemingly infinite variety of speech' (Cunningham etal 1994: 1568). The methods of linguistics, therefore, 'can also assist judges in finding and analyzing predictable order in the complex textual issues which so frequently make cases hard' (ibid.}. The logic of this is as follows. To the extent that linguistics has made progress in finding the 'predictable order' beneath the infinite variety of speech, they will be able to assist judges in finding that same order in the texts which they confront. As linguistics progresses it will discover more and more of that order, until one day it will crack the code of
300
Christopher Mutton
language in its entirety. When that day dawns the textual problems of judges will be over. This question of the scientific status of linguistics is clearly a crucial one in the eyes of many linguists who claim that they are able to advise lawyers about language. Cunningham et al (1994: 1568) argued that linguists follow an accepted scientific method: 'they deduce contrasting predictions of opposed hypotheses and measure those predictions against observable aspects of reality'. However, the measuring of those predictions against observable reality is carried out, I would argue, according to theory-internal criteria, especially the criterion of simplicity. Alternatively, the prediction is simply a plausible observation: 'if you have interactions in which people have conceded their willingness to do something, you will not find utterances of the form, Will you do X? or Would you do X?' (Michael Geis, quoted in Proceedings 1995: 902). Cunningham et al seem to concede this point by continuing as follows (1994: 1568): Because the hypotheses and their predictions do not typically involve observable reality, and because the reality that linguists try to describe is the set of abstract or complex rules underlying language use, linguists have to make inferences from people's linguistic behavior under both natural and controlled conditions.
In other words, the procedures of linguists are quite unlike those of natural scientists, if scientific method involves testing hypotheses against observable aspects of reality. The methods of linguistics are not like those of the natural sciences simply because linguistics is not a natural science. Making inferences or generalizations about behaviour is a quite different activity from testing hypotheses. Cunningham et al (1994: 1568) however, argue as follows: They [linguists] analyze the contexts in which people naturally use words and constructions, in order to infer the senses of those words or the discourse function of those constructions. Linguists also investigate the judgments made by native speakers about whether a given sentence could mean something particular or whether it would be natural to use it to mean anything at all. While intuition guides the critical first steps of linguists, their hypotheses are always subject to empirical testing.
The authors give as examples of empirical studies the use of corpora, such as NEXIS, and the use of questionnaires to survey informants' views on the meaning of particular terms. In particular, they discuss the vagueness of the term 'enterprise' as used in the RICO (Racketeer Influenced and Corrupt Organization) statute in the United States, and criticize the 'new textualist' tendency to rely on dictionaries as a guide to ordinary usage (ibid.: 1565, 1591). Compared with 'analysis of a particular textual problem by a trained linguist', dictionaries are 'a crude and frequently unreliable aid to word meaning and usage' (ibid.: 1563). What should we make of these claims by linguists? For there is a world of difference between the suggestion that judges use corpora to think about meaning, and a fantasy that linguistics will one day be able to crack the code of language and provide lawyers with a scientific metalanguage for talking about texts. Leaving aside the systematic ambiguity (i.e. confusion) between language and metalanguage in Cunningham et al (for even if we conceded that language could cope with all
Law Lessons for Linguists?
301
situations, why would this mean that linguistics could capture that ability?), and their neglect of the distinction between competence and performance, interpretative problems cannot be solved by using the metalanguage of linguistics, not least because that is the last thing that metalanguage was designed to do. Cunningham et al seem themselves to recognize this when they conclude that 'the linguist's skills do not make her more authoritative than any other native speaker, such as a judge, as to whether any particular use of a word is appropriate' (ibid.: 1596). What then can the linguist offer? The linguist can, however, use the NEXIS search to augment the judge's thinking about possible uses of the word, to suggest distinctions among uses that might escape the judge's notice due to the unconscious nature of language comprehension, and to inform the judge of uses entirely current and acceptable in the relevant speech community, of which the judge might not be aware.
The key word here is 'unconscious'. The linguist's role is presented here as that of bringing into the judge's consciousness usages and shades of meaning that might otherwise be passed over. But it is simply misleading to assert that language comprehension is unconscious. Legal interpretation is a conscious process, to the extent that any process can be. Of course there is always room for more awareness, and the complexity of ordinary usage is one area in which no doubt awareness could be raised. But what special insight do linguists bring to these matters? What is 'ordinary language', and what is its role in the law? A dogma of linguistics with regard to the unconscious nature of language comprehension is inserted here into what is otherwise an unexceptional plea for judges to get more advice about ordinary language and its interpretation. I would suggest that social workers, literary critics, criminals, and lexicographers are in as good as or better position to offer this kind of advice. What could a linguist tell someone whose partner had said to them: Tm not in love with you, but I do love you', beyond jargonizing an already patently transparent interpretative problem? Interpreting such statements is a highly conscious matter, and that is the kind of interpretation that lawyers do. There are two crucial differences between academic linguists and lawyers. Firstly, linguistics just simply is not about the interpretation of individual linguistic acts. It does not ask: 'What did X mean when she said to Y that PT, it asks 'What kind of shared knowledge did X and Y have such that P could be understood when said by X to T. If Y did not understand X, then it is assumed that either they do not share a common system of communication (the code) or that some other extraneous, nonlinguistic, factor must have intervened (noise, inattention, etc.). Even if the speechact theorist seems to ask: 'What did X mean when she said to Y that P?', that question concerns the status of the utterance in relation to a taxonomy of discoursal moves. The speech-act theorist's interest is predetermined by the linguist's rules of the game. But the legal interpreter cannot prejudge what it is about the meaning of a particular utterance that may be relevant to acts of legal classification; nor can they determine in advance what evidence may be relevant to achieving an interpretation. This point is made in the following way by Paul Campos in a commentary on the recent Washington University conference 'What is meaning in a legal text? A dialogue
302
Christopher Mutton
among scholars of law and linguistics' (Campos 1995: 971). Campos illustrates his point as follows. Imagine if one were to ask how much a particular tiger in a zoo weighs, and were to receive the reply that 'in general' tigers weigh 375 pounds. Campos comments that a generalization such as this, while it may be informative, does not answer the question as asked: Now suppose my informant were to insist that in a sense he has already told me what I want to know, because all tigers have two weights: 375 pounds (their theoretical tiger weight) and N pounds (the particular and hence variable weight of any individual tiger). The strangeness of this statement is obvious enough when what one seeks to know is what the tiger weighs. What theories of textual interpretation overlook is that it is just as peculiar to answer someone's questions about the meaning of a particular text with the assertion that, although we don't know what this text means as an utterance, we can tell you what it means as a sentence. Texts cannot have meaning 'in the abstract' any more than tigers can have abstract weights. The meaning of a text, like the weight of a tiger, is an empirical question, and no theoretical generalizations about what tigers weigh or what texts mean will give you the answer to either question.
The interpretation problems of the law are in this sense much closer to those of the ordinary language user than those of linguists. A second difference concerns the question of accountability. Lawyers are professionally trained to apply abstract rules and procedures to individual cases qua individual cases and are ultimately accountable for the answers they give to society, to the court of public opinion, to the scrutiny of legislators, to the lay categories of right and wrong. This is true to some extent even in non-democratic countries. However nebulous, uncertain and often contradictory this process, it is evident that the decisions of lawyers are not taken in a social vacuum. Lawyers are responsible, sometimes directly, sometimes remotely, for what they do, in that legal decisions are scrutinized in newspapers and are judged against the beliefs of ordinary people, social commentators, academic critics, politicians, etc. about what is right and wrong. Under certain social conditions these beliefs may be translated eventually into legal change in one direction or another. This is not to say that the world's legal systems are models of democratic accountability, nor that lawyers are more sensitive and caring than linguists. It is simply to make the point that there are non-professional means of judging the decisions of the law; 'society' is not merely a passive observer of the legal process. Legal decisions are translated into everyday moral terms of right and wrong and judged against them. No doubt this process is often unfair to the judiciary, who must on occasion violate their own private standards of right and wrong in the name of legal procedure, or who resent their decisions being translated into the terms of a simple morality play. But it is intrinsic to the sociology of the law. Linguists simply do not operate in this way. Their professionalization is rhetorically grounded in a rejection of everyday views about language. For the linguist, the court of public opinion is no more than a kangaroo court. Lawyers, by contrast, have some interest in being seen to promote not only justice as defined within the law (the intra-disciplinary notion of justice) but also justice as society understands it (the lay notion). Linguists are under no such pressure. Under professionalization, linguists have attempted to award themselves the label 'scientist', and such a label has precluded any such accountability to public opinion, except in the
Law Lessons for Linguists?
303
indirect way that academic institutions, and therefore academic disciplines, have in particular political contexts to justify themselves to the communities that fund them. In brief, linguists are not accountable or responsible to non-linguists, broadly speaking, for their acts of classification. Lawyers are. Even if linguists could give a better answer than lawyers to the question of whether 'vehicle', in the prohibition 'No vehicles allowed in the park', is a hyperonym of 'ambulance', the fact remains that they are accountable to no one for such acts of classification. No amount of empirical research into the everyday meaning of the term 'vehicle' can settle this question, not least because it is a moral question bound up with a definitional one. The two can no more be separated in law than they can in everyday life. Linguistics as a discipline has simply nothing to say on moral questions such as these. Even the much-maligned general lexicographer is held accountable by the community at some level, since dictionaries are seen as laying down facts of language that, unlike linguists' analyses, are to be applied in everyday situations. This point is suggested by William Eskridge's remarks comparing the way linguists and lawyers look at problems of interpretation (Proceedings 1995: 851). Eskridge sees linguists 'treating the statutes as communicated acts, words on a page, simple commands, whereas the law professors were treating statutes as more than that, either as social policy or as some kind of public or natural law'. In the recent Washington University Law Quarterly volume, the underlying assumption on the part of the linguists was that linguistics will ultimately inform substantially how the law operates, and how lawyers work with language. There is scarcely a hint that lawyers may be able to help linguists to do linguistics better. This is evident from the basic menu for the conference, which set as topics for discussion either the analysis of the differences between the way the two disciplines treat key words such as 'meaning', 'interpretation', etc. or discussion of how linguists can contribute to the various textual problems of the law (Levi 1995: 779-81). Indeed, how could lawyers be expected to teach linguists anything? How could a lawyer help a linguist decide if a given segment was or was not a preposition or an adjacency pair? This question only arises within linguistics and is only answerable to the satisfaction of other linguists. The discipline of linguistics as a whole can only make a contribution to interpretative issues by accepting, at least in principle, some form of social accountability for its acts of classification. As long as linguists seek to accrue all the trappings of professionalism, such as academic status, prestige, influence on neighbouring disciplines, a masterful and opaque terminology, with none of the costs (the scrutiny of those acts of classification, public debate, criticism, challenges to those ideas from ethical and other standpoints, a critique of opaque terminology), the profession will continue to experience a sense of crisis in its relationship with society, a crisis that threatens the existence of linguistics departments in the United States and worldwide. This might seem an unduly pessimistic welcome to a new area of academic inquiry. But it is precisely because such new areas of inquiry offer the possibility of a more meaningful set of disciplines devoted to the study of language that we need to beware of filling other people's bottles with our stale old wine. Many recent writings on language and law achieve this, and in so doing offer a challenge to the dominant discourse of 'core linguistics'. Any new metalanguage must arise out of a genuine
304
Christopher Mutton
attempt by linguists to understand the beliefs, practices and customs of lawyers and non-lawyers in relation to language, not out of an attempt at 'intellectual colonization', mutual or otherwise (see Mootz 1995: 1010-11). To the extent that linguists and lawyers can face up to this challenge, I believe we can afford to be cautiously optimistic.
References Campos, P. (1995) This is not a sentence. Washington University Law Quarterly 73, 971-82. Cardozo, B. (1921) The Nature of the Judicial Process. Yale University Press, New Haven. Cunningham, C., J. Levi, G. Green, and J. Kaplan (1994) Plain meaning and hard cases. Review of Solan 1993. The Yale Law Journal 103, 1561-625. Elgin, S.H. (1995) You Can't Say That to Me! Stopping the Pain of Verbal Abuse - an 8-step Program. John Wiley, New York. Harris, R. (1995) Translating Mitspeak. Review of Pinker 1994. Essays in Criticism 45, 272-9. Korzybski, A. (1933) Science and Sanity, 3rd edition. The International Non-Aristotelian Library, Institute of General Semantics, Lakeville, Conn. Levi, J. (1995) Introduction. What is meaning in a legal text? A first dialogue for law and linguistics. Proceedings of the Northwestern University/Washington University Law and Linguistics Conference. Washington University Law Quarterly 73, 771-83. Mootz, F. (1995) Desperately seeking science. Washington University Law Quarterly 73, 1009-23. Pinker, S. (1994) The Language Instinct. W. Morrow, New York. Proceedings 1995. Proceedings of the Northwestern University/Washington University Law and Linguistics Conference. Washington University Law Quarterly 73, 769-970. Solan, L. (1993) The Language of Judges. Chicago University Press, Chicago. Tannen, D. (1990) You Just Don't Understand! Women and Men in Conversation. W. Morrow, New York. (1992) Response to Senta Troemel-Ploetz 1991. Discourse and Society 3, 249-54. (ed.) (1993) Gender and Conversational Interaction. Oxford University Press, London. (1994) Talking from 9 to 5: Women and Men in the Workplace: Language, Sex and Power. Avon Books, New York. Thurber, J. (1945) The Thurber Carnival. Harper, New York. Troemel-Ploetz, S. (1991) Selling the apolitical. Review of D. Tannen 1990. Discourse and Society 2, 489-502.
22
Teaching American English as a Foreign Language: An Integrationist Approach DANIEL R. DAVIS
This chapter presents the conception and formulation of a course, 'The Structure of American English', taught at the English Department at the University of Hong Kong and discusses why and how an integrationist approach to the study of American English can illuminate students' experience of language in the political and sociocultural context of Hong Kong. The relevance of this chapter extends to the teaching of linguistics and varieties of English in both first and second language situations, and the thesis can be summarized as follows: Linguistics and 'standard languages' should not be taught as a set of techniques of analysis and objects for analysis. Instead, they should be taught as a set of cultural, intellectual practices arising in a particular historical, cultural, intellectual, and above all, political context. By making this (sometimes difficult and frustrating) shift, language and linguistics teachers can empower their students, first, by de-reifying and de-mystifying language and linguistics, second, by situating linguistics within various political and intellectual discourses, recognizing linguists as responsive actors within these discourses, and finally, by making students aware of the political and social consequences of their own acts as linguists. 'English, Whose English?' (the title of the conference for which this chapter was written as a paper) is a controversial question in at least two ways. First, because it is a loaded question, demanding an answer which stakes a claim on the language. Second, because it makes the theoretical assumption that 'English' (whatever it may be) is the sort of entity which can be possessed. The question thus begs another: 'English, What's English?' The integrationist critique of linguistics begins with a similar question, 'What is a language?', as posed by linguistic theorists (see Harris 1980). Linguistic theorists advance this question not in a genuine spirit of inquiry, but as a rhetorical ploy to introduce their own definitions of a language. Integrationism sees language (and linguistics, being a kind of language) strictly as an integral aspect of human behaviour, and therefore gives priority to identifying the intellectual and cultural contexts of various answers (explicit or assumed, sincere or otherwise) to the question, 'What is a language?' For an integrational linguistics, the definition of language and a language is fluid and dependent on context; the integrationist 305
306
Daniel R. Davis
enterprise is essentially to identify and re-identify the contexts significant to linguistic and metalinguistic behaviour. My purpose here is not to define the terms 'English' or 'a language', but to identify this as the centre of the discussion. In the case of the specific example presented by this chapter, I wish to start the discussion by opening up the definition of 'American English' and leaving it open, turning instead to the contexts which bear on this definition. The first context to be addressed is that of the students of the course, with particular reference to the language in the University of Hong Kong and in Hong Kong in general. Students arrive at the Arts Faculty at the University from a society which is multilingual, with languages such as Cantonese, English, Putonghua, Chiu Chow, Hakka, Shanghainese, Tagalog, Portuguese, and Hindi spoken by different groups within the community. There is a high degree of variation within the English spoken in Hong Kong, owing to its second-language status for most of the community, and to the widespread geographical origin of first-language speakers. At the same time, there is widespread awareness of the political and economic significance of language in general and English in particular. This has to do with the role of English within the colonial history of Hong Kong, but also with the rapid modernization and development of the Hong Kong economy since the Second World War, and concerns for the future of Hong Kong in post-1997 Asia. Perceptions of English are arguably quite complex: it is seen as essential to the economic success of Hong Kong within the Pacific; it is part of the modern cosmopolitan identity of Hong Kong; but it can also be perceived as the heritage of colonial and imperialist rule. There is a considerable amount of insecurity about language in Hong Kong, especially among that part of the population attending university. This is witnessed in the perennial debate about language standards in education, seen in the South China Morning Post and the Hong Kong Standard newspapers, and in the question asked of academics, 'What do you think of the students' standard of English?' It is within this linguistic and social situation that the course, The Structure of American English', has been offered in the Arts Faculty at the University of Hong Kong, where it has been very popular. The students' reasons for taking the course fall into two groups: the first centres on the perceived utility of American English for doing business with Americans in Hong Kong; the second may be that students have various contacts with North America such as the emigration of relatives and the import of American popular culture to Hong Kong. When asked, students tend to refer to the former reason (the economic importance), or to express an interest in learning the features of American English. There are a number of ways to design a course on this subject for an audience with these expectations, in this linguistic situation, each with its own drawbacks. One could give a structural account, identifying the features of American English. This has the advantage of meeting the students' expectations, and there are very useful texts such as Trudgill and Hannah (1982) which are readily available. However, this strategy becomes in practice a comparison with other varieties rather than a true structuralist description of 'the American Language'. More importantly, this strategy plays into some of the language myths which already beset the students, the myth of a standard language especially. On the one hand, the students have a large investment in the concept of a standard language, since their success in examinations has placed
Teaching American English
307
them at the top of the educational ladder in Hong Kong, and this will determine their place in society. On the other hand, acceptance of the mantle of the standard language places them in a permanent crisis of identity. Is their English of a sufficient standard? Is it necessary to defer to speakers of English as a first language? Does conformity to a standard result in a transferral of loyalty from Hong Kong or China to some other place, especially if it results in a usage noticeably different from that of their peers? A structuralist approach gives them a method for learning a variety of English, feature by feature, but in doing so it denies them the possibility of addressing the political situation it has landed them in. A different way to structure the course would be to give a socio-historical account of the development of American English, explaining the diachronic development of various features of American English in society (usually relating specific features to historical events or social trends or processes). Again, there are easily obtainable books such as Marckwardt (1980), McCrum el al (1986) and Dillard (1992). Unfortunately (and perhaps surprisingly), such an approach meets neither the students' expectations nor their interests. They do not see the relevance of such an approach, and I would advance the explanation that the social history of American English assumes a set of values embodied in the American experience to which they have little if any access and to which they find it difficult to relate. Furthermore, because they rarely have a background in American history, this approach tends to become a history lesson rather than a study of language. The important point about both of these strategies is that the role they expect of the teacher is as an insider, an American, initiating the students as outsiders into the rules and values of the entity 'American English'. This is perhaps a role which the TEFL teacher is required to play at times, but in this context the role is not appropriate, for the following reasons. First, the students have relatively fluent English. Second, the initiation is not part of a training in linguistics at the University level. Finally, and more importantly, it is not acceptable ideologically, in that it propagates myths of language which deny the students possession of the language. It assumes that language is inherently and entirely national in its constitution, and in doing this ignores the issues of identity (both political and cultural) bound up in language. The initiation relies on linguistic and cultural stereotyping without recognizing that it does so. It is for these reasons that I decided not to adopt either of these strategies as a framework for the course, but to formulate an integrationist strategy. As I have said above, such an approach must begin with contexts, and the discussion up to now can be said to embody three positions or contexts: the students', my own as teacher, and that of the subject matter. The students, for the reasons mentioned earlier, are interested in features of American English. My own position (as a teacher of university undergraduates and as a linguist) is to encourage reasoned and critical thought, writing and discussion in the students, centring on the subject and its definition. The subject, for its part, presents a number of possible approaches to students, which are to be found in its literature and which can (and should) be incorporated into an overall integrational approach. This should identify the concerns and intellectual and political context of each text. Criticism of the texts in which these various approaches are found is the proper beginning of an integrational discussion. But for the purposes of integrational teaching, the students' expectation must first be
308
Daniel R. Davis
addressed. This amounts to a simultaneous introduction to and attack on the notion of linguistic features. The text used as the object of criticism for this is Trudgill and Hannah (1982), International English. This book presents so-called standard varieties of English by means of lists of phonological, grammatical, lexical and orthographical features. The text is valuable in that it is simple enough for the students to digest quickly, but it is sufficiently detailed to capture some of the complexities and contradictions involved in setting up standard varieties. Its use of minimal pairs and the notion of phonemes to set up a sound system for each variety points up its debt to structuralist discovery methods, and ultimately to structuralism itself. By setting the linguistic feature within this intellectual tradition, it can be shown that, at least within structuralist theory and methodology, the linguistic feature has no existence apart from the network of relations in which it is embedded, and in fact, can be said to be constituted by the system as a whole. Since pure Saussurean structuralism does not allow direct mapping between features in different systems, the theoretical problem for varieties of English is that either each variety is a system in its own right, in which case its features cannot be compared to those of another variety, or, each variety is seen to exist within one system, in which case it is to be banished to the realm of parole as the idiosyncrasies of an individual or group of individuals within the community. Nor can the use of the term 'national standard language' remove this theoretical problem, since what makes one dialect the standard is not part of synchronic linguistics as defined by Saussure. There is no linguistic reason to equate the standard language with the linguistic system. Linguistic features which derive from structuralist method are thereby shown to pose problems for the linguist trying to use them to define different varieties of English. The next part of the attack on linguistic features in the course involves the examination of texts in order to identify varieties of English within them. As an exercise, this requires students to apply their reading of International English and its features to selected texts from various origins, trying to state the origin of text on the basis of linguistic features. This exercise achieves a number of tasks. First, it satisfies the students' curiosity about linguistic features, and points up how difficult such identification of variety on the basis of features is. Second, it shows that so-called 'American English' features exist outside of 'American English', and that 'non-American English' features exist within 'American English'. That is, not only are there theoretical difficulties with the notion 'linguistic feature', but there are also serious difficulties in applying the notion to linguistic behaviour, even if a 'practical' or non-theoretical stance is taken regarding the existence of linguistic features. Distinctions such as standard/non-standard are shown to be just as useful (that is, of very limited usefulness) as the American/British or American/non-American distinction in accounting for the language of a text. The exercise usually reveals certain stereotypes which the students have formed about American English, the crudest being that all non-standard English is American English (for example, that the form 'ain't' and the double negative are exclusive to American English). This hypothesis can be easily shown to be a stereotype by including an English dialect text. But the point is not to ensure that the students' stereotypes of or hypotheses about American English converge with those in Trudgill and Hannah; rather it is to show that all hypotheses about linguistic features and varieties involve stereotyping. Variation within and across national borders can be
Teaching American English
309
used to show that all national varieties are political constructs based on the stereotyping of features. It is undeniable that in a dialect continuum some isoglosses for certain features may be shown to run along a national boundary; but the perception that these are the significant aspects of variation upon which to concentrate is ultimately determined by political orientation. This raises difficult problems about the political implications of standard language varieties (and in fact, the entire method of using varieties as a means of rationalizing linguistic variation) for individual identity. A final point brought up by the exercise is that authors (and speakers) can of course manipulate features (as they perceive them) to achieve various ends. The result of this two-pronged introduction to and attack on features is, it is to be hoped, that the students learn both features (as they desired originally) and, more importantly, their theoretical and practical limitations. Also, it should be noted that an attempt is made to address their possible experience of American English in Hong Kong, first through literature, and second through their everyday life. They are asked, as another assignment, to make a phonetic transcription of a news broadcast and to say whether the news reader is British, American, Canadian, or a speaker of some other variety. This is not to be interpreted as an attempt to make latter-day Professor Higginses of them, but to illustrate by direct appeal to their (almost) everyday experience, the fact that judgements are made on the basis of perceived accent, and that these judgements often have political and social significance, particularly if the newsreader is from Hong Kong but has lived overseas in Britain, Canada, the United States, or Australia. (As an aside, it should be mentioned that many of the students arrive at university unable to distinguish between American, British, Canadian and Australian accents, perceiving them all as 'foreign'. I am indebted to K. Bolton for suggesting and corroborating this observation.) The course then turns to other approaches to American English as manifest in a variety of texts. The texts are used to situate the study of American English within various political and intellectual discourses. The texts are Marckwardt (1980) (an example of what I call the 'social historical' approach), Hans Kurath (1972) (the 'dialectological' approach), William Labov (1972) (the 'sociolinguistic' approach), J. Dillard (1972) (the issue of Black English), and Noah Webster (in Crowley 1991) (the 'political' approach). Each passage has been selected because it exemplifies or characterizes a set of assumptions about language and because it addresses certain arguments, often unstated, within the study of American English. The students are asked to work out what the linguistic assumptions are within each text, and to consider these assumptions in the light of the argument of each text. The aim of the exercise is first, to give them some familiarity with linguistic issues in the study of American English and second, to illustrate the Saussurean structuralist principle that, where language is concerned, the point of view creates the object of study. The passage from Marckwardt is taken from the chapter entitled 'Colonial Lag and Levelling'. In it the assumption is made that there is a direct correlation between linguistic change or stability (and the resulting features of a variety) and specific aspects of social change and stability. This is not too controversial (although it is not strictly structuralist), nor does it produce too striking a view of language, but it clearly plays into an argument about the authenticity and 'originality' of American English vis-a-vis British English. For Marckwardt, American English is linguistically conservative just as American society is socially conservative; when there are
310
Daniel R. Davis
instances of clear innovation on the part of American English, Marckwardt takes great pains to situate this within sixteenth- to eighteenth-century development in British English. Since he has earlier argued that the language of the first colonists is the language of Shakespeare and Marlowe, one can easily make out that his argument is concerned with validating American English as closer to the language of the cultural icons of the English speaking world. This is in answer to a criticism (unstated in this text) of the innovative, 'nouveau', uncultured aspect of American English (and the influence it has had on other varieties of English in the twentieth century). This sort of re-situation and recovery of American English within the English cultural world has its linguistic counterpart in Kurath's article. His ostensible problem is to explain the relation between American English dialects and British English dialects, but the article can also be read as an answer to the question, 'How many structural linguistic systems exist within the Anglo-American English-speaking community/ies, one or two?' The difficulty with the overt problem is that some of the same variation occurring within British English dialects occurs within American English dialects (though the dialects themselves cannot be said to have been transplanted), yet, structurally speaking, American English observes for the most part the same mergers and splits that 'Standard British English' (to use Kurath's term) does. Kurath's solution is to argue that British English and American share the same set of phonemes, although these are in different lexical distribution in the two varieties. Regional variation is said to be sub-phonemic. What is interesting about this claim is that it argues that there is one sound system of English, equivalent to the standard, and that there is in fact one English language. This is the only position which a structuralist could afford to take in dealing with variation, as any admission of variation within the system proper would allow the possibility not only of two systems, American and British, but of any number of systems based on different dialects, with the result that the English language would disintegrate into a very large number of separate systems, or that it would consist of one system with so much internal variation as to bring into question its status as a system in any sense. Kurath's point of view has created the English language as he sees it. Issues of language definition and status also arise in 'The Social Motivation of a Sound Change'. In this article Labov (or at least, the early Labov) attempts to explain linguistic change in its social setting. In the particular case of centralized diphthongs in Martha's Vineyard, Labov relates a high frequency of centralized diphthongs to a political orientation towards the island and its values, in opposition to those of the mainland. However, despite recognizing the social and political setting of variation and change, Labov denies the claim by one of the islanders that the language of Martha's Vineyard is a different variety, a 'separate language within the English language' (Labov 1972: 29). For him, the definition of language and variety is still an issue of structural facts, not a question of political interpretation, and it is the province of the trained linguist. I find it instructive to point out to students that, on the basis of linguistic features, the language Labov describes in Martha's Vineyard is as different from the 'US English' described by Trudgill and Hannah as their 'Canadian English' is. The definition of a language or variety is dealt with in 'Black English and the Academic Establishment', in Dillard (1972). Dillard makes the case that in earlier
Teaching American English
311
discussions of the English of Blacks in the United States, linguists were misled by their political prejudices in saying that a distinct variety, 'Black English', did not exist. He goes on to discuss the political and educational implications of the status of Black English as dialect or variety. Clearly, it is of extreme importance for him to argue that Black English is a separate variety of English, both in terms of its structure and its historical origin, and that this distinct status merits special educational treatment for the speakers of this variety. From this chapter it can be seen that the political point of view as much as the linguistic point of view creates the object of study (and its history), and further, that a linguist is to be held responsible for both his politics and the political implications of the analysis of language which he puts forward. The final text used in the course is Noah Webster's essay on reforming spelling, 'An Essay on the Necessity, Advantages and Practicability of Reforming the Mode of Spelling, and of Rendering the Orthography of Words Consistent to the Pronunciation' (Appendix to Dissertations on the English Language 1789, reprinted in Crowley 1991). Webster's approach to language can be said to be political: it is clear from his discussion and examples that he assumes language to be a political and social institution. He makes the specific point that the distinction between American and British English can be reinforced by changing the spelling system (which for Webster is part of the language), and that this change would result in pride in American literature, language, and culture in opposition to (and freed from dependence on) British literature, language, and culture. For Webster, political reality is at least in part based on language, and a linguist is one who can change language to reinforce a political reality. Alone among all the other linguists dealt with in this chapter and course, he recognizes both the responsibility and the power of the linguist. He alone recognizes his role as actor within a political and linguistic discourse; he alone illustrates (and indeed requires) that the relation between point of view and language is reciprocal: the point of view creates the object of study creates the point of view. . . One might ask, how is this course 'integrational'? First, it de-segregates language, in the first place by calling into question segregationist linguistic categories and a segregationist system of language (i.e. the structuralist position), and in the second by re-siting language ('American English') within the students' experience of English in Hong Kong. Second, the course studies the metalinguistic behaviour which constitutes and defines the terms 'American', 'English', and 'American English'. Third, it considers the responsibility and power of linguists as political agents within an intellectual, social, cultural and political situation, and finally, it extends this role to students. One can consider this form of empowerment in terms of cost and benefit. The first cost is that the notion of 'standard' English is lost; the benefit is that students may realize the contingency of this notion according to various purposes within a political situation. This could be of considerable value in Hong Kong society. Second, the status of the linguist as trained expert putting forward description and analysis of language, is put in question. Against this cost, one may place the benefit that by abandoning aspirations to this status, students may realize their own political power and responsibility as linguists. Also, as a return for weakening the notion of systematicity, one may find that the Saussurean principle of point of view is perhaps
312
Daniel R. Davis
a better position from which to appreciate the nature of human language. Third, the credibility of stereotypical national identities formed on the basis of stereotyped linguistic features has been brought down. They are of course still in existence and available as screens to hide behind for the individual who so chooses. But they are no longer shelter, only cover. To leave these stereotypes is to heighten one's sensibility to issues of language and identity: different political manifestations of the individual are made possible. Finally, the course clearly limits the power of analytical methodology in linguistics. On the credit side, it develops one's critical faculty towards linguistic theory and language studies. During the last tutorial of the course, I usually ask students to consider what significance these issues in the study of American English may have for the linguistics of English in Hong Kong. So far, reactions vary, with some quite enthusiastic, others perceptive and cautious, and others uninspired and unresponsive. It is perhaps too easy to make facile comparisons which may obscure rather than illuminate their linguistic and political experience. At any rate, my aim is not to expose them to a political ideology, but to an ideology about the political role of the linguist. It is their decision whether to adopt this role, or to turn to other forms of linguistics, and the roles they provide. Acknowledgements I would like to thank Professor R. Harris, Mr K. Bolton, and Dr C. Hutton for reading this paper and commenting upon it.
References Crowley, T. (ed.) (1991) Proper English? Readings in Language, History, and Cultural Identity. Routledge, London. Dillard, J.L. (1972) Black English: Its History and Usage in the United States. Random House, New York. (1992) A History of American English. Longman, London & New York. Harris, R. (1980) The Language-Makers. Duckworth, London. Kurath, H. (1972) Studies in Area Linguistics. Indiana University Press, Bloomington & London. Labov, W. (1972) Sociolinguistic Patterns. University of Pennsylvania Press, Philadelphia. Marckwardt, A.H. (1980) American English, rev. by J.L. Dillard. Oxford University Press, Oxford. McCrum, R., W. Cran & R. MacNeil (1986) The Story of English. Viking, New York. Trudgill, P. & J. Hannah (1982) International English: A Guide to Varieties of Standard English. Edward Arnold, London.
23
What Problems? On Learning to Translatei MARSHALL MORRIS
The essential problem for aspiring translators is not, as they are inclined to believe, the foreign language as such. If they only had more of the words that make it up, they imagine, the meanings of texts would come clear. This impression is very like that of the newcomer to an unfamiliar society. What one needs to get along, one feels, is the names of the tangible things which make up the world. If one just knew what things are called then one could make one's way. And so one can, up to a point. But one reaches that point rather quickly and can easily get stuck there, imagining that whatever one encounters in the foreign culture and language, one's home culture and language will provide at least a rough and ready equivalent. This way of thinking takes for granted that experience of the world is essentially one, that languages and the ways they are used are essentially the same. There are realms in which this is more true than others - some aspects of international business come to mind - but the translator is exceptionally well placed to appreciate how many areas of human life are taken in, experienced and expressed in irresolvably distinct ways. The issue is not whether something will be lost, a nuance or a shade of meaning, for these certainly will, but whether something more important, the overall sense of a text, its meaning in its context, will be permanently obscured even by the careful and respectful moving of each linguistic brick. Randolph Haynes used to say that Spanish is one of the easiest languages to begin and one of the most difficult to master. He was, of course, talking about language as far more than the words that inventory the tangible world, which only point to meanings, as the Spanish sociolinguist Manuel Alvar has told us. In language nosotros no transmitimos realidades, sino sonidos que nos permiten descubrirlas, y en ocasiones, identificarlas. (Alvar 1979: 83) 1 This paper originated in the experience of teaching Spanish-English translation at the University of Puerto Rico, for some twenty years, an experience which has proven deeply stimulating to me. As a young teacher, at Oxford for graduate study, I remember asking a don known not only for his original work but for his translations from French if we might meet to talk about the problems of translation. He looked at me curiously and asked 'What problems?' This paper is dedicated with affection and gratitude to Dr. Rodney Needham, Professor of Social Anthropology, Oxford University, and to the students I count myself fortunate to have had in my care, who do in fact encounter problems in learning to translate. An earlier version of this paper was presented as 'What Problems? Between the Lines in Translation', at the conference 'Linguistics Redefined. An Interdisciplinary Colloquium', convened by Hayley Davis at Northeast Missouri State University in the Spring of 1989.
313
314
Marshall Morris
Beginning translators make their mistakes largely because they think they are supposed to be good bricklayers when in fact the job they have taken on requires an eye for the architecture as well, unlike Lewis Thomas's termites, who instinctively build the right structure! The essential problem that aspiring innocent translators face is not the oneness of the world, or the sameness of experience that they imagine can be relexified, language by language. It is the differences we experience in contact with another society which happily afford us meaningful surprises. But beginning translators, intent on finding the word or phrase that corresponds to the 'same' experience in their own language, overlook and miss the significance of what might otherwise surprise them. They are in this sense resistant to surprise. A first lesson they must learn is to recognize and learn not to stifle their sense of surprise, which offers them their essential and most trustworthy and certainly most personal line of inquiry for discovering what there is beyond the words that preoccupied them in the beginning, beyond the words that easily become too closely linked to the familiar concepts of the familiar world of the translator's home language and culture. The value of surprise is that it points to something so important as to be invisible, like the water the fish swims in or the air we breathe: we have all 'learned the world', if not as a smooth and coherent thing, then at least as a reasonably comprehensible, reasonably consistent set of things, persons and experiences that 'go together' for us in particular ways. These allow us to reason out what we do not see first hand, to fill in the gaps, and so make sense of experience. We cannot see everything, hear everything, experience everything ourselves: we depend on slight clues to suggest whole worlds of familiar or probable experience, so that a person in a particular place can be expected to be engaged in a particular thing, with the intentions and motives associated with these circumstances, places and events. In effect, we depend on a system of hypotheses about language and society and the way they work. This system is provided to us without our willing it, by the historical experience of our society and its language, by its way of life, into which we are thrust at birth and in which, over time, we become quite expert. This system is not ordinarily visible to us, nor are we critically aware of it, though experience can teach us to be so. As ordinary users of a language we are not obliged to bring our system of connections and hypotheses into critical view, except as we grow and the system itself evolves. As translators, however, we are confronted over and over again with our lack of expertise in the second language and culture, with the fact that our home-grown intuitions and our personal calculus of what is appropriate do not quite work in that other world: Von Frische's honey bees dancing in an alien hive! The dependability of our home culture and language leads us to make use of them in contexts where it is not justified. For the traveller, it is to interpret the facial expressions of the persons one meets in the strange bazaar as one would familiar faces at home. 2 For the translator, it is making all the right connections, at lightening speed, but in the wrong place, for each text the translator deals with speaks out of a different tradition, with different names for different things that make up the world, things which connect differently in thought, which point to different constellations of 2 Judith Nine broke important ground in contrasting forms of Puerto Rican and mainstream American non-verbal communication. The Puerto Rican schoolchild lowering her eyes out of respect for her teacher may be interpreted by the Anglo teacher as wishing to hide something (Nine Curt 1976).
On Learning to Translate
315
character, motivation, intention, to beings for whom the meanings are necessarily different, too. One issue then, is the limits of awareness (Bolinger 1980, Silverstein 1981), which affects even linguistically skillful people. Not knowing what we do not know is a general human problem, but translators are under strict and constant professional obligation to listen for clues, for any hint in the texts they translate, that there is something there they have not understood. They must take careful note when their expectations are defeated, when the original text becomes blurred, when a swarm of linguistic forms or a single odd usage points to something they have not anticipated, do not understand. A particular instance of this concerns us here. We are accustomed to thinking of language and society as in some sense distinguishable. After all, we talk a lot about language, study it in schools, have familiar genres of written language, and so forth. Naive translators often restrict their attention to what looks like language in their own tradition, in particular to written texts. Though they routinely supply from their experience of society what would be required for understanding in their own language, they make no adjustment in their thinking for the fact that the text before them, however much it may look like their own familiar forms, derives from a different society and experience of life. They may need to supply quite different sorts of contextual matter, and this consciously, to understand what the author of the other text intended to be understood, and also what the common reader would easily understand. Here is a kind of limit to the idea that we should trust not the teller but the tale. No tale is told with absolute completeness, the listener must always supply what language points to. The translator-listener must not only supply what the listener in the original language would normally supply from experience of that other society, but also specifically refrain from supplying what would not be supplied. And in a sense, business letters, government documents, personal accounts, words spoken in court, all are tales. Translators must understand what is being attempted, what is being done with language in the text: who is doing what with whom, and why, and how the writer would like his readers to respond to these things. Translators are helped in this not so much by the particulars of their own language and society as by the fact that their lives are complexly patterned. By analogy, they can expect similar complexity of pattern, yet too often they do not. Why? The specific ways we analyze and teach language, and the tools we use dictionaries, grammars and so forth - not only illuminate but can interfere with our understanding, impose a kind of pattern that diverts attention as one tries to listen to the language of the text, which is someone else's language, after all. The continuity of the writer with his world may be lost to the translator in the choppy reduction that the formal study of language implies. These wonderful tools of analysis, instruction and reference, monuments to the cumulative richness of a tradition of thought, all essential in translation, may also have the effect of authorizing translators to think only in their terms, and so constitute a limitation on the imagination. They may look for the parts, and the configurations of parts, that characterize the home language, and they may easily miss seeing different kinds of parts and different configurations. They look for language to do for the author what it does for them, and so bend whatever they encounter to fit what is
316
Marshall Morris
familiar. They may not notice where the fit is odd or where many small oddnesses make up a systematic difference between two systems of thought, between two experiences of life. In this way, unnoticed, expectations can lead to distortions in translation, distortions being more dangerous than errors which are often incongruous and visible. If it is difficult to correct for expectations regarding cultural form or lexicon, it is even harder to correct for the subtle formative power of one's grammar, system of discourse and experience of language use.
An integrational view of translation In thinking about how people learn to translate and the difficulties they encounter, I find Roy Harris's views on the matter helpful (Harris 1983). 'Non-integrational theories', he says, are of two kinds: 'transference theories' and 'replicational theories'. Transference theories treat translation as a set of procedures for extracting a 'message' from a text in language A and re-encoding it in language B. The 'object' of translation is the message, not the text itself. Replicational theories treat translation as a set of procedures for replicating (as far as possible) in language B (features of) a text formulated in language A. The 'object' of translation is the text itself.
Both assume that the signs of the languages are invariant, and that the problems of translation come from a lack of match between the invariant signs. 'Integrational theories', in contrast, assume that 'in human communication (linguistic) signs are never in variants': If translation is a communicational process, it follows that translation itself creates new signs. Consequently any account of translation in terms of (matched or mismatched) invariant signs is to be rejected. For the integrational theorist, the identity of a sign is constituted solely by its communicational integration with other relevant variables (including other signs) in the context of a specific communication situation. Translation must therefore be a process which supplies a simultaneous interpretation of the identities of two sets of signs (the 'original' set and the 'translation' set), neither 'given' in advance of the other.
Difficulties in learning to translate do derive from assuming too great a stability in language, in one's own and in the relations of one's own with another language. This common sense view of language also assumes the necessity or at least the primacy of the forms with which one is familiar - an assumption that linguistic and conceptual change over time in one language and contact with other languages and societies both undermine. Not everyone manages to free his mind of the grip of these assumptions even in his or her own language. In translating this makes for unwarranted literalness, so Roy Harris's criteria for the 'integrational translator' are doubly ambitious and
On Learning to Translate
317
difficult to achieve, particularly as they concern not only one's expectations about language but also one's approach to comprehending experienced I would like to provide a couple of examples of the problem of understanding when one does not know some particular contextual social facts, which are visible in an oddness of language but may pass unscrutinized by the translator whose mind is set, as it were 'on automatic pilot'. They have to do with underlying assumptions of particular peoples at particular times and places.
A modern Japanese example The Japanese psychiatrist Takeo Doi speaks of how he became aware of a systematic difference between Americans and Japanese when he was studying in the United States (Doi 1981). His supervisor would do him some little favor or other, and he would hear himself saying not Thank you', but 'I'm sorry'. Was this just bad translation? Doi turned this over and over in his mind and discovered that his expression arose spontaneously from a distinct Japanese experience of human relations, systematically different from the American experience. The Japanese form was meaningful in a system of human relations in which, Doi explains, one is constantly attempting to establish personal relationships of intimacy analogous with those of family. The expression of regret at having even inadvertently imposed on another, like other formal expressions, serves to demonstrate the interest, deference, concern, and ultimately dependence appropriate to such a relationship. The American expression marked, in contrast, a more distant relationship of 'equalizing reciprocity' in which individual independence is highly valued. The fact that Doi heard and could reflect on the inappropriateness of his literal translation made it possible for him to think about the ways human relations are structured in the two cultures, as Roy Harris's ideal 'integrational' translator would do.
An English example, remote in time A second example is Roland Voise-Valayre's experience of reading the brief accounts of cases which constitute the historical record of Medieval English courts (VoiseValayre 1991). He found it puzzling, for example, that a woman who accused a man of rape and proved her case might herself be hauled off to prison. How could that be? Voise-Valayre found himself re-reading the accounts thinking he had missed something or perhaps not understood the language, but he had in fact understood. He was to discover, in the end, that the reason for what seemed to him, as it must to us, an absurd outcome, is to be found in the underlying assumptions of the time and place, which were not stated in the accounts themselves. If a woman became pregnant 3 The anthropologist Edmund Leach (1978: 662-3) characterized the two poles of the problem in this way: '. . . In my world words are, alternatively, evanescent patterns in the flow of breath and conjured images in the mind which, in Joycean fashion, are forever melting into other words and images to produce what I understand by the word 'thought'. But in [an opponent's] world it would seem that words are hard surfaced objects in the world out there, with impermeable skins and carefully labelled contents (meanings) like the bottles in a chemist's dispensary; and woe betide anyone who suggests mixing up the contents according to any formula which is not written out on a doctor's prescription! And so too with language'.
318
Marshall Morris
after being raped, she was judged to be culpable of false accusation since it was believed that a woman could not conceive unless she was a willing partner. The language of the accounts as such, even if translated with the greatest exactitude, fails to convey what any person of that time and place would immediately understand. And this is surely true of language generally, since its efficiency and economy would be lost if all the pertinent presuppositions had to be made explicit whenever someone wished to say something. To my mind, these examples pretty well do away with the arbitrary division between language and society and help us on to the problem that most concerns us here. If we find it difficult to understand something that occurred, say, centuries ago in the history of our own legal and moral system, how much more difficult it is to grasp what is going on in the exotic societies, so-called. This was the subject of LevyBruhl, who wrote on what was then called 'primitive mentality' (Levy-Bruhl 1975). Modern anthropology continues to be concerned to understand the reasoning behind what are to us exotic forms of life and thought, and to show, in language we can understand, what it is like to live in such societies. That is, it is concerned with a kind of 'translation of culture', to use Evans-Pritchard's felicitous phrase (Beidelman 1971). It is only partly a matter of translation of language, as we usually think of it, since it attempts to make the lived experience of the other kind of life accessible and even reasonable to us in our own terms - short, that is, of doing to Homer what Chapman did! To do this, certainly early on in the work, the social anthropologist has no more essential tool than his own sense of difference from the culture he is studying, his capacity to be surprised. He goes beyond this, as his work progresses, to compare his findings with work among peoples who are in some important anthropological sense comparable. And if he is fortunate, his initial perceptions of difference will become insights into or even a systematic understanding of the form of life of another people. When he writes up his ethnography, he may make explicit use of his sense of difference, his surprise, though this is not the usual convention of anthropological writing. For most people, the exotic character of anthropological study obscures the fact that when another form of life becomes comprehensible, and so comparable to one's own, it inevitably raises important questions about one's own society and about human society generally. (Not all of us are able and willing to see these comparisons, or to see them through to their logical conclusion, but they are there if we wish to be instructed.)
When language misleads Certain conventions as to the use of language must puzzle the integrational translator. I am thinking here of the assumption, common in our long, text-based tradition, that people speak propositionally. The defeat of this assumption affords no end of surprise, of course, when real utterances are being translated. I am thinking, too, of conventions of directness and indirectness (Morris 1981). And of the uses of silence (Jamin 1977, Tannen and Saville-Troike 1985). And I am thinking of truth-telling. The latter is of special interest in trying to see how the experiential and ideological
On Learning to Translate
319
ordering of society directly and profoundly shape the use of language, and may serve as an antidote to the translator's belief that his problems are located, in any simple way, in the peculiarity of vocabulary or idiom. With our tradition as the frame of reference, one may well begin reading a text with the idea of a language-user who 'has spoken or written with perfect command of the language and perfectly truthful intention' (Collingwood 1939: 31), but end up with a very different idea indeed, perhaps even believing that such a language-user is only a fictional character, if one for whom we have understandably strong feelings in our tradition. I would like to refer, here, to the work of Juliet Du Boulay, Michael Gilsenan, and Tom Beidelman, who have all written on Mediterranean societies, because their data runs counter to assumptions the translator may make: that language 'says what it means and means what it says' (Harris 1981). In these societies, to be a good and a successful person, one must lie. And one must learn to do it well. Honor and shame are crucial ordering concepts in society and the members of families, in their different roles, must place the integrity of the family above all other values, particularly above individual values. In being loyal to the family in word and deed, one is obliged to represent the circumstances of the family to outsiders in ways that protect it from the damage and the evil that outsiders can do. In these societies, one creates and maintains one's self, and sets and protects the status of one's family, perhaps for generations into the future, by cunning, deceit and the control of information. Take an example from Du Boulay. Families in the Greek village she studied lived in sharp and hostile competition. The standards by which personal and family honor were judged were impossibly high, and mockery involving questions of honor were rife - the mockery of others being sufficient to diminish one's value and the status of one's family in the village. Children were trained in curiosity about others, and in secrecy in regard to their own family's doings. It was important to know how to deceive others by lying and to trick others into revealing their secrets. Lying is thus a phenomenon which is built into the structure of social relations, and is a vital safety-valve for the tensions that would otherwise accumulate under the pressure of an ideal code and the eagle eyes of an inquisitive community. And the way in which the villager personally relates to his lying, being culturally conditioned by the same context that produces the lie, produces no disharmony between himself and his understanding of his own honour. In spite of this, however, it is not that the villager is unaware of the difference between telling the truth and telling a lie ... It is that truth has many manifestations, and a lie on one level may legitimately be accepted as a way of revealing a truth on a higher level. (Du Boulay 1974:193)
Beidelman and Gilsenan find, among Homeric Greeks and modern Lebanese Arabs respectively, that men must be eternally vigilant in defining their social character in talk and social intercourse, chiefly with other men. One's status is never secure but must always be defended against trickery involving matters of honor. Beidelman examined the Odyssey and the Iliad for insight into gift-giving practices and related discourse, and found that . . . one's standing could be jeopardized, lost or enhanced by victory, defeat, improvidence, luck or unacceptable conduct. (Beidelman 1989: 230)
320
Marshall Morris
The competition here is both physical and verbal, and losses in one sphere authorize one to use cunning in the other. One loses simply by failing to put matters at risk. One cannot drop out claiming to be above such struggles. One must remain agonistically involved. The public nature of exchanges, the need for validation by others, is intense for Homeric Greeks. Respect, dignity, honor, shame, are attributes conferred or denied by others. One needs an audience, (ibid.: 249)
Relentless public scrutiny and competition, and jostling for highly relative prestige obliges one to be cunning - to make use of theft, and trickery and, in language, lying and deceit - to assure a place in the society of men. Though this is held to be in contradiction to the strictures of religion, as Gilsenan observes, people have no choice but to participate or risk mockery, which again means loss of honor, relegation to inferior status and even what he calls 'social death' (Gilsenan 1976: 198, 203). The systematic behavior reported by these anthropologists - translated, with reservations, as lying and deceit - turns out not to be in any simple way destructive of, but is even constitutive of persons, family and society. I think we can take two lessons from this. One has to do with the reflection to which such knowledge leads us. The other has to do with the dilemma of such conventions for translators. We are led to think about the nature and uses of lying, about how groups and societies are composed through the use of language, about the definition and constitution of persons, about the shaping of character and about what is being called the 'social construction of emotions' (Harre 1986). We are led to think about lying in our own society and to question the assumptions of our tradition about the nature and uses of language, and in particular of truth-telling. The translator - or in cases such as these, the anthropologist - knows that the translation even of the word lying is not and cannot be exact. What is being called lying in these ethnographies is a practice from some other, equally coherent experience of human life. The succinct translation is inexact, and all too powerfully meaningful in our own tradition. To get the translation right, it might take, as Peter Riviere has often noted, as much as 600 pages to 'translate' some central concept of an alien people. That is the length of Evans-Pritchard's ethnography of the Nuer, the whole of which is necessary to make the word kwoth - approximately 'spirit' understandable to us (Evans-Pritchard 1940). The translation 'lie' presents to the English reader not the Mediterranean concept, experienced as part of a family's strategy for survival, but our own tradition's concept. This distorts, however unwillingly, the experience of the people in the tradition in which the original figures. It is questionable whether it is even possible in such cases to grasp that other experience of life except vaguely and imprecisely, as Naja Shamaa indicates in her study of the difficulties of translating from what she calls rhetorically extravagant, 'overstated' Arabic into 'understated' English (Shamaa 1978: 48). It is of course ironical that we perceive the problem as a vagueness and imprecision in the other language rather than in our own or in ourselves (Morris 1984). Malinowski himself was inclined to this deficiency:
On Learning to Translate
321
In a primitive tongue, the whole grammatical structure lacks the precision and definiteness of our own. (Malinowski 1930: 300)
What are the implications for the integrational translator of a systematic convention of lying and deception? For the non-integrational translator, there is, of course, much less of a problem. To have translated the word-items of which the original is made up is thought sufficient, and I cannot say how many times, when questioning a nonsensically literal translation, I have heard the phrase, 'But that is what it says!'. Full stop. The integrational translator, in contrast, is faced with a problem of understanding and representation of considerable complexity, one that is meaningful to us today. On the one hand, it is the kind of problem that Smullyan sets his readers in his manylayered logical games, with the great difference that the moments when we must understand clearly come fast and heavy with human implications - as they do for the characters in Le Carre's chilling novel, A Perfect Spy. The world of language for the integrational translator resembles the one Stephen Tyler describes as . . . the dark underside of communication - of the interpretive potential of cooperation withheld, hedged, or made deliberately ambiguous through irony, vagueness, white lies, and outright prevarication, reminding us, in other words, of that messier and more interesting world of talk where speakers and hearers do not resolutely seek clarity, consensus, and agreement in harmonious communion, but exploit ambiguities, hint at meanings, create misunderstandings, and conspire against one another, seek some advantage in a slip of the tongue, or hear hidden insults in a pause or tone, or hide their intentions and meanings as they reveal them, or reveal them in the act of hiding them. This is the tale of the everyday world where meanings are in the spaces between words and in the contexts words evoke rather than in the facts they may purport to describe. (Tyler 1978: 1)
This, of course, to the despair of the translator, for even when he understands perfectly, as is unlikely, for there is always more going on than the language actually records, and even when he translates 'with perfect command of the language and perfectly truthful intention', as our tradition would oblige him to do, he constantly discovers anew that language only points toward meanings, as Alvar has said. This world is not unfamiliar to us, either. It is certainly not the exclusive realm of the 'primitive'. Read, for example, Dwight Bolinger's thorough expose of the dangers of lying, deceit and self-deception in our own society in his Language - The Loaded Weapon (1980b) or the introductory chapters of F.G. Bailey's The Prevalence of Deceit (1991).4 And to judge by Neil Postman's claims, we have gone the Greek villagers one better. The effect of television and show business on modern American discourse has brought about the destruction of coherent contexts in which such ideas as contradiction are useful in ascertaining the truth of a matter. Lacking this, and the close human context of the village, The credibility of the teller is the ultimate test of the truth of a proposition' (Postman 1985: 102). 4 Skepticism about the possibility of what is said being meant or what is meant being said is certainly broad enough: Larissa Lomnitz says of modern Mexican political ritual, 'En Mexico no hay hechos, solo interpretaciones' (Lomnitz, Lomnitz and Adler 1990), and a Russian adage runs as follows: 'A thought pronounced is a lie . . . " I am grateful to Claudio Lomnitz, of New York University, for the first reference and to Leonora Chernyakhovskaya, of the Moscow International School of Translation and Interpreting, for the second.
322
Marshall Morris
What are translators of texts, then, to do when the translation of culture seems to overwhelm frail language? What they have always done, I suppose. Look about for whatever imaginative resources are available, say what can be said as exactly as possible, take care to suggest the rest... Translators depend to an inordinate degree on the sophistication and particularly the good will of their readers. So do the members of any society in their ordinary doings. It is to this fragile and vacillating structure that we entrust the transmission of our most precious heritage - in everyday life and in committing ideas to paper - for we have no other, no stable, fixed and invariant form to comfort the aspiring translator. The existential dilemma of the speaker, the translator, the anthropologist and the linguist is one.
References Alvar, Manuel (1979) Del glosano al diccionario automatizado. Boletin de la Academia Puertorriquena de la lengua espanola (San Juan) 7(1). Bailey, F.G. (1991) The Prevalence of Deceit. Cornell University Press, Ithaca. Beidelman, Tom (1989) Agonistic exchange: Homeric reciprocity and the heritage of Simmel and Mauss. Cultural Anthropology 4 (3), 227-59. (ed.) (1971) The Translation of Culture. Essays to E.E. Evans-Pritchard. Tavistock, London. Bolinger, Dwight (1980a) Fire in a wooden stove: on being aware in language. In Leonard Michaels and Christopher Ricks (eds.) The State of the Language. University of California Press, Berkeley. (1980b) Language - The Loaded Weapon: The Use and Abuse of Language Today. Longmans, London. Collingwood, R.G. (1939) An Autobiography. Oxford University Press, Oxford. Doi, Takeo (1981) The Anatomy of Dependence. Kodansha International, Tokyo. Du Boulay, Juliet (1974) Portrait of a Greek Mountain Village. Clarendon Press, Oxford. Evans-Pritchard, E.E. (1940) The Nuer: A Description of the Modes of Livelihood and Political Institutions of a Nilotic People. Clarendon Press, Oxford. Gilsenan, Michael (1976) Lying, honor and contradiction. In Bruce Kapferer (ed.) Transaction and Meaning: Directions in the Anthropology of Exchange and Symbolic Behavior. Institute for the Study of Human Issues, Philadelphia. Harre', Rom (1986) The Social Construction of Emotions. Basil Blackwell, Oxford. Harris, Roy (1981) Editorial Foreword to Morris (1981). (1983) Notes on an integrational theory of translation. Personal communication. Jamin, Jean (1977) Les Lois du silence: Essai sur lafonction sociale du secret. Francis Maspero, Paris. Leach, Edmund (1978) Reply to correspondence from David Bradley. Man, n.s. 13 (4), 662-3. L6vy-Bruhl, Lucien (1975) The Notebooks on Primitive Mentality. Basil Blackwell, Oxford. Lomnitz, Larissa, Claudio Lomnitz and Ilya Adler (1990) El fondo de la forma: actos publicos de la campafia presidencial del PRI, 1988. Kellog Working Paper in Latin American Studies. Kellog Institute, University of Notre Dame, South Bend, Indiana. Malinowski, Bronislaw (1930) The problem of meaning in primitive languages. In C.K. Ogden and LA. Richards, The Meaning of Meaning. Harcourt, Brace, New York. Morris, Marshall (1984) Everyday language, common sense and conventions: the impression of elusiveness and inexactness in another language, (m.s.) (1981) Saying and Meaning in Puerto Rico: Some Problems in the Ethnography of Discourse. Pergamon Press, Oxford. Nine Curt, Carmen Judith (1976) Non-Verbal Communication in Puerto Rico. Northeast Center for Curriculum Development and the National Assessment and Dissemination Center for Bilingual Education, Fall River, Mass.
On Learning to Translate
323
Postman, Neil (1985) Amusing Ourselves to Death: Public Discourse in the Age of Show Business. Viking, New York. Silverstein, Michael (1981) The limits of awareness. Sociolinguistic Working Paper No. 84. Southwest Educational Development Laboratory, Austin, Texas. Shamaa, Naja (1978) A Linguistic Analysis of Some Problems Inherent in Translating Arabic to English. D.Phil, thesis, University of Oxford. Sperber, Dan (1975) Rethinking Symbolism. Cambridge University Press, Cambridge. Tannen, Deborah and Muriel Saville-Troike (eds.) (1985) Perspectives on Silence. ABLEX, Norwood, N.J. Tyler, Stephen (1988) Foreword to Ivo Strecker, The Social Practice of Symbolization: An Anthropological Analysis. Athlone Press, London. Voise-Valayre, Roland (1991) Vers une economic de 1'ordre. Cahiers Charles V. Publications de 1'Universite de Paris VII, Paris.
24
Pronouncing French Names in New Orleans GEORGE WOLF, MICHELE BOCQUILLON, DEBBIE DE LA HOUSSAYE, PHYLLIS KRZYZEK, CLIFTON MEYNARD, LISBETH PHILIP
Introduction The present paper aims to extract some general conclusions from a body of data concerning pronunciations and mispronunciations of French names in the city of New Orleans, remarks by their bearers on their attitudes to these latter, and on bearers' various ways of correcting interlocutors. 1 The data were gathered primarily by telephone interviews. The task of the interviewer was: (i) to inform name-bearers of the nature of the project, described as a study of the pronunciation of French names in New Orleans, (ii) to ask bearers if they would be willing to have their names included in the study, (iii) in case of an affirmative response, to ask bearers (a) how they pronounce their names, (b) if they ever hear mispronunciations of their names, (c) if yes, if they correct people who mispronounce, (d) if they do correct, how they might go about doing so, and (iv) again to ask bearers if they mind their names being used in the study. An initial question arose as to what the criteria are for a French name. The initial answer to this was an informal one: if the name looked obviously French, it was placed in the list. Usually, this could be justified by appeal to the name-ending; certain endings are French from a historical-linguistic point of view, thus -elles, -ard, -eau, -ieu, -ain, -ault, etc. Sometimes mistakes were made, and these were corrected by the bearers: e.g. Auld, Bute, were claimed to be Scottish, Crossin Irish, Buitron Spanish, Guitart Catalan, Burlet and Dussel German, and Zatarain Basque. In these cases, the word of bearers was taken, and no attempt was made to 'argue them out of the belief in the national affiliation of the name. On the other hand, this rendered the affiliation of a name with a nation sometimes a difficult matter to determine, most clearly in cases where the bearer's family came from a border region such as Alsace where a bearer might well claim a 'French' name was 'German', or vice versa, or where the name had been taken into another country (Orillac2), sometimes early on 1 We are grateful to Sheila Embleton, Dennis Preston, and William Bright for many helpful suggestions which have led to our improving the content and organization of this paper. 2 Bold face indicates that the name can be found in the Appendix. Both in the text and the Appendix, where a stress mark is missing from the phonetic transcription of the pronunciation, this indicates that the place of the stress in the pronunciation was not determinable.
324
Pronouncing French Names in New Orleans
325
(Bayard 3). It also happened that the bearer was not aware that the name was French, and this raised ethical issues, for example in cases in which the bearer was African-American and retained an ambivalent, or possibly hostile, attitude about any claim to the name which could recall the past. Thus, the Frenchness of a name was potentially a problematic notion. Yet it was not necessary to arrive at a hard and fast definition, since subjective attitudes toward names came into account at all levels as determining them. It might have been possible to rely on the definitions 'statistically common in France (e.g. found in a certain percentage in French telephone books)', or 'was considered French by the bearer when introduced into New Orleans'. But in fact, the Frenchness of the name was treated for the purpose of the interviews as a working hypothesis which might later be shown to be unwarranted. Very often the name was 'clearly French', i.e. it looked French, could be found in numerous French telephone books, was taken as French by the bearer, and posed pronunciation difficulties consonant with a clash of different phonetic traditions. In other cases - and perhaps ultimately - whether the name was in fact French or not was secondary to whether it posed pronunciation problems for the bearers and/or the interlocutor; and 'French' was merely a convenient label to apply to it. In short, if the interviewer and/or the bearer thought it was French, or thought it might be French, and no reasons could be found for excluding it from being French, then it was treated as French. One of the primary concerns of this paper is to take an initial look at the terms in which name-bearers manage the pronunciations of their names. That is to say, given that the pronunciation of French names poses frequent problems in New Orleans, bearers are often faced with the task of correcting people who mispronounce, and of conveying what in their view is the correct pronunciation of the name. Such a situation provides an example of a case in which the language of every day requires to be referred to so that it can proceed in its communicational role. One result is what may be called a 'lay metalanguage', that is, terms and concepts persons who are not professional linguists use to refer to and to manage the language which is their daily communicational vehicle. As a preliminary to a survey of concepts of lay metalanguage, it will be useful to turn our attention to two examples of concepts of metalanguage as used by professional linguists, so as to provide a point of comparison for the subsequent discussion.
Two theoretical discussions of pronunciation Some useful reflections on the notion of pronunciation of items initially foreign to the phonetic habits of a given group of speakers are offered by Martinet (1977). Martinet considers the conditions under which French speakers pronounce foreign words which have been taken into French. One of Martinet's first points is that monolingual French speakers take their pronunciations of these words not directly from the foreign model, but rather from those French speakers who have some knowledge of the foreign language which provides the model, and who have already adapted that model to French phonetic habits. One example of this is the English word laser, taken into French not as /laze/, which would conform to such French words as jaser, baser, and
326
George Wolf et al
raser, but as /lazeR/, which follows rather the pattern of ver, mer, fier, and amer. Martinet makes a related point concerning the (Parisian) French pronunciation of foreign words containing the cluster nd. He says: 'In Parisian French, the cluster [nd] is perfectly pronounceable between vowels . . . But [nd] does not exist word-finally. A realization such as [wikend] is thus a distortion of French phonology by those who have at least some knowledge of English' (1977: 81). Three phrases here are particularly revealing with regard to the boundary between phonetics and phonology: 'perfectly pronounceable', 'non-existent word-finally', and 'a distortion of French phonology'. 'Perfectly pronounceable' is opposed phonetically to 'perfectly unpronounceable', e.g. the 'law of three consonants' in French which prevents clusters of more than three consonants from occurring. Thus, in rien de special [rjedspesjal]ffap is 'perfectly pronounceable', whereas in quatre de speciales *[katdspesjal] the tdsp would be 'perfectly [i.e. physically, articulatorily] unpronounceable'. 'Non-existent word finally' is not relevant to the same order of constraint, for there is nothing articulatorily difficult about pronouncing a word-final nd in French as evidenced by the widespread pronunciation of weekend. Now, this is either because the French don't like to have nds at the end of their words, or it just so happens that nd does not occur word-finally, in which case it is a historical accident. But in either case there is nothing preventing the French - indeed they have now done so - from introducing [nd]s word-finally, should they decide it suits them. Finally, 'a distortion of French phonology'. This may, similarly to the word-final case, mean one of two things. A final nd might be a statistical aberration in French; or, it might offend 'the sense of the language'. In the statistical interpretation, it may be that statistics should be taken to exercise a certain power over the language. In the second interpretation, speakers themselves have power over their language, for an offensive arrangement will be consciously avoided. An interesting point of comparison here would be that of the pronunciation of French [y] and [R] by English speakers. Would their introduction into English be articulatorily unpronounceable; is their absence merely a historical accident with no implication about their pronounceability by speakers; or would their introduction offend against our sense of English? Martinet's reflections are useful in that they help us more clearly to separate the phonetic from the phonological, and to clarify what we mean by both these terms. We seem to be presented with a spectrum, at one end of which is physical impossibility (or, if one prefers, improbability) - such as an apico-laryngeal consonant - and at the other end of which is a set of subjective impressions, with a large number of historical accidents ranging over the middle. Often it is uncertain - as is the case with a spectrum - whether there are boundaries between these divisions. A second view, this one incorporating reflections on the influence of spelling on pronunciation, is provided by Kokeritz (1964: 137-45). Here we face the specific issue of the relationship, if any, between a given pronunciation and a written form. An example of a 'spelling pronunciation' is provided by the word falcon. The traditional pronunciation of this word, as Kokeritz notes, is [foksn] o ['folksn]. In the 1950s the Ford Motor Company introduced a new car and called it the 'Falcon'. Familiarly in cases such as this, the introduction of a new product on the market necessitates, if the word for it is unfamiliar, a marketed pronunciation decided
Pronouncing French Names in New Orleans
327
upon by the company (cf. Hyundai, Croissanwich). In the present case the decision was to call the car the ['faelksn], which Kokeritz calls 'an obvious spelling pronunciation'. What does this mean? To call it 'an obvious spelling pronunciation' would appear to imply that, let us say, the spelling ale somehow naturally entails the pronunciation [aelk]. Here we think, as Kokeritz notes, of the words balcony and talcum. (One also hears the pronunciation ['tolkgrn], but perhaps this is a non-spelling pronunciation.) It is an interesting fact also that the words which contain ale come for the most part from Latin words which were all pronounced [aelk] (or [alk]; cf. It. balcone.falcone). So one thing we could say is that one 'standard English pronunciation' of ale is the result of an historical accident. Now if this is true, then - and here the case recalls Martinet above - it cannot be said that ale entails the pronunciation [aelk], but that it merely so happens that those words have retained their historical pronunciation. Now it also happens that, as Kokeritz points out, the word falcon was pronounced [fokan] or [folksn] before the advent of the car. If so, then this in itself would prove that ale does not entail the pronunciation [aelk]. But there is a further fact of interest. When an advertising representative of the company was asked why the new pronunciation was being introduced, the answer was that [Taelkan] 'appears in larger sections of the country, and by virtue of its common usage is more identifiable' (Kokeritz 1964: 137). We might ask: if [fokan] or [folkan] is the pronunciation of people who also read, what makes [faelkan] a spelling pronunciation? According to The Oxford Dictionary of English Etymology (Onions 1966), falcon entered the language in the thirteenth century in the form faucon, a form clearly Anglo-Norman in spelling and doubtless in pronunciation ([fok-]). The word was respelled in the fifteenth century on the Latin model. Later, talc(um) was taken over from either French talc ([talk]) or Medieval Latin talcum (doubtless the English pronunciation was, as it still is, [taelkam]) in the sixteenth century, and balcony from Italian balcone ([balkone]) in the seventeenth. One question then is, if the AngloNorman pronunciation [fo-] was retained when the spelling falcon was adopted, why did [folkanRffoksn] not become the spelling pronunciation of falcon, and why were the pronunciations of the later words (-[al]-) not considered aberrations? On the other hand, if the motor company representative is right, and [faelkan] was a widespread pronunciation at the time of the introduction of the Ford Falcon, then it is tempting to assume that the pronunciation [faelksn] came into England along with the re-spelling, and that the older pronunciation was only retained in a deliberately conservative manner by people who would be likely to engage in falconry. But if this turned out to be the case, then ['faelksn] would not have been a spelling pronunciation originally since the pronunciation came with the spelling (and had existed as such for centuries); and ['faelksn] would not have been a spelling pronunciation later since the assumption is that the pronunciation was not remodelled on the spelling, but was already in place. In another example, Kokeritz states that 'Not a few words and groups of words can be shown to have acquired their present pronunciation through the influence of the spelling at some time in the past both in British and American English. We may point to the reappearance of [t] in pestle, often or of [n] in kiln, the use of [1] in British solder . . .' (ibid. : 138-9). Once again the consonants in question here were originally pronounced: Lat. pistillum, solidare, Old Engl. oft, cylene < Lat. culina. Secondly, it is of interest that in all these words the consonants were later suppressed
328
George Wolf et al
for reasons having nothing to do with spelling, showing that it was possible to change the pronunciation of a word quite apart from considerations of spelling. That is to say, one could see the consonant in the word and still 'leave it in silence'. Therefore, at no time, apparently, was one obliged to follow the spelling. In this connection we may ask: why is the t pronounceable in pestle but not in castle, in often but not in soften, the / in solder but not in folk, balm, talk, walk! It is difficult to avoid the conclusion that the criteria for spelling pronunciation are not at all obvious; which is to say that there is no evident possibility of telling when a pronunciation is 'following' spelling, and when it just coincidentally appears to parallel it. Part of the reason for this is that we do not know what 'following a spelling' or 'pronouncing a spelling' really means. Rather, these phrases tend to beg the question of what the 'pronunciation' of a spelling is. Why should pestle and not castle be the model for a certain pronunciation? It could just as easily be the case that the 'correct' pronunciation of t is in fact silence. It is well known, for example, that h cannot be pronounced in French. Would, then, the French spelling pronunciation of hibou ('owl') be [hibu]? An irony here is that it has been clear for centuries that there is no precise correspondence between English spelling and pronunciation. The Great Vowel Shift wrought havoc, and after that spelling reformers bravely, but futilely, sought to rectify the situation. Yet if this was known, and if one encountered a word whose pronunciation one did not know, it would be risky, not to say contradictory, to believe it reliable to base a correct pronunciation on the written word. Rather, the required strategy would be to seek the help of someone who seemed to know how to pronounce the word. The conclusion is that it may seem that one can 'pronounce letters', but one knows that a given letter is pronounceable only because the pronunciation exists already quite apart from any written form. Therefore, a pronunciation would appear to be based not on letters, but rather on a tradition of pronunciation. It is for this reason that a so-called spelling pronunciation of words like pestle cannot be a coherent notion unless an accepted pronunciation is already in place. But this pronunciation itself would ex hypothesi not be based on spelling.
Field observations of speakers' remarks on pronunciation To these deployments of theoretical metalanguage in the clarification of what is involved in pronunciation we may add another set of notions, which have been seen to emerge from problems of pronunciation in actual communicational exchanges. These are notions from which lay speakers have been observed to draw in the management of their daily communicational activity. A fertile ground for doing this is provided by the city of New Orleans, where, owing to its French heritage, a large number of people have French names. Because, however, a knowledge of French phonetic habits has virtually disappeared there (as it has not among the Cajuns who live in the country), there is often scope for confusion about how to pronounce a given name. Consequently, there is a frequent need for strategies for solving the communicational problem of how to pronounce a name which is either aurally or visually unfamiliar.
Pronouncing French Names in New Orleans
329
(i) The general problem: pronunciation of unfamiliar written forms A number of pronunciations of French names in New Orleans follow what appear to be 'spelling pronunciations', i.e. there appears to be a conscious attempt to represent certain letters by a corresponding sound. Hence one mispronunciation of De Gray is [da'grui], an attempt to cover every letter of Dubuisson is evident in ['dubwisan], and one pronunciation of Jambon is ['dsaembon]. But this principle can be seen to be inconsistently followed in Andrieu (['aendru]), Catoire ([kaetwae]), Credeur ([kra'dur]), Dupleix (['dupleks]), Ecuyer ([ekwije]), Laguaite ([la'get], mispronounced [la'gwat]), and Quezergue ([ks'zer], [ki'zer], mispronounced ['knzare], [krs'vergi], ['kezor], [ka'zerki]). For in these cases, even if there is a desire to follow the spelling, it is not necessarily clear how to do so. At the other extreme, in the cases in which the pronunciation of the name shows a consciousness of French tradition, it may happen that the spelling is not appealed to as a model for the oral form, but may be ignored. For example, although one mispronunciation of Duquesnay ([du'kezne]) does attempt to follow the s, there is no apparent anxiety about the qu, universally [kw] in English, and the pronunciation fduksne], as well as one other mispronunciation ([du'ken]) betray no constraint imposed by the spelling. Moreover, any systematic theory of spelling pronunciation will encounter formidable obstacles when attempting to account for the articulatory solutions to the unfamiliar pronunciations of a majority of French names which are considered by their bearers to be unsuccessful, such as Andrepont (['aendrspojnt]), Ardenaux (['orsano]), Areaux (['aeru]), Bayle (['beli], [bojl], [bqjl], ['bali], [bel]), Bourquard (['bugard]), Courvoisier ([kgvszir]), Dusaules ([da'sael], Faucheaux ([foffeks]), Geathreaux (['goGraks]), Labruyere ([la'brunjel), Orgeron (['orsgan]). Given these examples, the relationship between spelling, oral performance of what is taken to be spelled, aural (or cognitive) experience of that performance, visual experience of spelling, and the experience of being faced with having to perform something unfamiliar would all appear to need to be taken into account in any attempt to explain the phenomenon of pronunciation.
(ii) Categories of non-bearers' experience of names To begin from the point of view of the person who does not bear the name but must pronounce it, the perceptual field is divided into (i) sound, (ii) writing, (iii) a relationship between (i) and (ii). The pronouncer is also faced with the evident requirement to pronounce, but this is not so much perceived as entered into as an act which, once begun, must take care of itself for better or worse (see below). As for (i), the (English-speaking) hearer may recognize a sound as a familiar name, e.g. [smi6]. This will generally be regarded as posing no barrier to producing the name when required. On the other hand, the hearer may not recognize the sound, as hypothetically in the case of the French pronunciation of the name Argeanton ([aRsato]). The [an], [30], [to] may either not be recognized at all by a monolingual English-speaker, or if minimally recognized may be felt to be too foreign to put the
330
George Wolf et al
effort to repeat it within sufficient reach of success. If, moreover, the speaker has little concept of spelling, the forced result will be able to rely on few cues as crutches. As for (ii), quite apart from a heard sound, the reader may recognize a written form, say Smith (or Jones), and if so will be able to produce a pronunciation without problem. Or the reader may not recognize the form - an extreme example would be a name in a foreign script, but even a name such as Cuiellette or Calongne would serve. Potentially the scope for visual confusion might appear to be greater than that for aural confusion; however, in the case of French names usually not more than one unfamiliar vowel- or consonant-cluster is present, and it can frequently be passed over. Indeed, a frequent strategy in this case is to ignore what is unfamiliar, to substitute something familiar, and to press ahead with the rest, hence [kju'let] or [ks'lon] for the above names. An unfamiliar foreign script would provide no basis for performance. The case of (iii) is prima facie more complex. The hearer/reader may recognize a written form and a sound as going together (the written form as the spelling of the sound, and the sound as the pronunciation of the written form), and can pronounce accordingly; thus Hebert: ['eber] (in New Orleans this can be taken as known; see below). Secondly, the hearer/reader may recognize the sound but not the written form. (Frequently this would mean that the hearer/reader does not recognize the sound and the written form as going together; but ex hypothesi the hearer/reader takes the forms as related in this case). The hearer/reader has two choices: (i) to produce the familiar sound and either ignore the written form, or revise the interpretation of it; (ii) to ignore the familiar sound and attempt to match the written form. The reader of Jacques, for example, may be familiar with the pronunciation fooJc], but may none the less say feoks] or [dsaks], perhaps interpreting the familiar pronunciation as defective. Thirdly, the hearer/reader may recognize the written form but not recognize the sound taken to go with it. Take the example of Aucoin, pronounced (here hypothetically for the sake of argument) [okwe] by its bearer. In this case the hearer/reader may try to match the bearer's pronunciation (and pronounce e.g. ['okwaen]), ignoring his or her own previous pronunciation (['okwin]) of the written form, possibly re-interpreting it as mistaken. Or the written form may be pronounced as recognized (['okwin]), ignoring what was heard. Finally, the hearer/reader may be presented with a written form and a spoken sound, neither of which provide a recognizable basis for assured pronunciation. Examples here might be the names of French cities Rouen ([RWQ]) and Niort ([njoR]; imagine e.g. a situation in which the written name occurs in a list, the bearer points out the name in the list and utters it; the reader/hearer would be facing a written form and the pronunciation simultaneously, but recognize neither). In this case the hearer/reader may try to match the sound and ignore the written form; or ignore the sound as hopeless and attempt to pronounce the written form. There are two further strategies: the hearer/reader can negotiate the form (ask for a repetition, make attempts at it and ask for help, etc.); or the hearer/reader can seek a way to avoid the issue and be silent. The above offers an informal framework of ways in which non-bearers experience names they may have to pronounce. Pronunciation, however, must also be approached from the point of view of the bearer of the name who, if the name is unfamiliar, as is often the case with French names in New Orleans, is in a position of
Pronouncing French Names in New Orleans
331
either ignoring the situation entirely (which for practical reasons may not always be possible), or of having to manage the public use of the name.
(iii) Bearers' attitudes toward the pronunciations of their names There are a number of aspects to the public management of the pronunciation of names. One of the first issues to face is the bearer's relationship to the name and/or its pronunciation. For example, it can happen that a wife will say, if asked how she pronounces her name, that the name is 'really her husband's'; for it cannot be assumed that they will pronounce it the same way. (It was a matter of public knowledge that the former mayor of New Orleans Sidney Barthelemy said [bor'Oalami] while his wife said [bor'Oetemi]. His was virtually the universal public pronunciation of the name, yet one news anchor-person always used the wife's pronunciation.) One wife (Alexcee) said that her husband pronounced the name better than she did; another called into the background: 'Honey, how do you pronounce our name?' (Lepree). Another thing which cannot be assumed is that every bearer is committed to one, or any, pronunciation. Some bearers actually avoid committing themselves to a pronunciation, waiting to see how the interlocutor will pronounce it, or will volunteer that it should be pronounced. In the case of Credeur, a bearer said 'I'm not really sure what the correct pronunciation is to tell you the truth'. And a bearer of Durocher said 'I use it the way people say it'. Occasionally, as in the mayor's case, different pronunciations will exist within a single family. According to one Fouquet bearer, 'Half the family says ['foke], half ['fuke]'. Distortions of any adopted pronunciation are a chronic phenomenon in New Orleans, and it can happen that the bearer will adopt the distortion (this happens also in cases of distorted spellings). Bressard was for one bearer Bessard before he went into the Armed Services. Also there, one bearer of Catron adopted the pronunciation ['ketrsn] because ['kaeptarikaetrsn] caused too much confusion. But it can also happen that a bearer, unable to prevent or avoid a particularly unlucky distortion, but unable to adopt the distortion, is forced to abandon the name. Virtually all two-syllable names in New Orleans beginning with Fag- (e.g. Faget ([fa'se]) or Faug- (e.g. Faugeaux ([fago]) will be pronounced [Taegat] at one time or another, sometimes consistently, producing a difficult situation for the bearer. The situation became such for one bearer of what was spelled Fagout in the telephone book but turned out to be correctly spelled Fagot, that the family decided to have the name legally changed to the husband's grandmother's name (Santos). Other bearers have varying attitudes to what they regard as mispronunciations of their names. Quite frequently the bearer does not mind mispronunciations and ignores them. Although Bazile is mispronounced 'plenty a'times' (often [bra'zil]), one bearer 'mostly just lets 'em go on with it'. The bearer of Petitpain, who claimed to hear 'everything but 'petty larceny", seemed resigned to mispronunciations. The bearer of Lalande said she was too old to care ('I'm 97 years old and I don't let little things upset me'). And the bearer of Laguaite was extremely accomodating: 'I tell them I'll answer to 'L' if they get that much!' Other bearers do mind
332
George Wolf et al
mispronunciations, but ignore them anyway. A bearer of Buffet accepts what he gets, but prefers ['bAfgt] because people tend to make fun of [bA'fe]. It may happen also that a bearer appears to be unaware of mispronunciations. The bearer of Houin, claiming the pronunciation ['hjuin], but initially called [hue] by the interlocutor, when asked if her name was ever mispronounced, said no, and that the interlocutor had pronounced it correctly. The wife of a bearer of Maureau said Til let you talk to my husband; he's the ['moro]'. The husband claimed the pronunciation [moro]; but when asked if he ever heard the pronunciation ['moro], he said no. Conversely, some bearers may either refuse or fail to answer to mispronunciations of their names. A bearer of Dejoie, while in the Navy, would amuse himself by refusing to answer to frequent mispronunciations of his name during roll call, then when roll was finished object to not having been called. A more serious case was that of a bearer of Voisin. Having spoken only French until age ten, and having only heard the French pronunciation of his name ([vwaze]), he entered the Army. After some weeks, he asked a friend when his name would be called in roll. The friend replied it had been called all along. The man subsequently learned he had been declared AWOL. Some bearers, on the other hand, will explicitly claim that one pronunciation is right, and will cling to it. In most cases this leads them to correct people who mispronounce. Many, however, will not correct at all times. A bearer of Dureau 'wouldn't correct anyone socially, but [would] if it were business or legal'. A bearer of Orgeron only corrects 'if they say 'Oregon". And a bearer of Buisson candidly admits: 'I do it for different purposes, and what effect I can get; for clout'. The implication is that many bearers are conscious of interlocutors' feelings when faced with correcting them. This is evident in cases such as Brulet, whose bearer admits: 'I say 'it's pronounced [brule]' and they're usually insulted'. A Brugier bearer corrects only if it's important: 'Mostly people you'll never see again, so why make them miserable?' However, correction is not a guarantee of altering the interlocutor's pronunciation. Some may adhere to a mistaken pronunciation through strength of preconception or through inattention to the correction. As a bearer of Fryou said, 'I correct it but it don't do any good. I just tell 'em it's ['friju] and they go right back to ['frqju]'. Two final categories of speaker are sheltered from problems concerning mispronunciation of their names. One includes speakers who are publicly known. In two cases this concerned sports celebrities: e.g. the New Orleans Saints' former quarterback Bobby Hebert (['eber]; see above), whose name was frequently pronounced ['hibart] by sportscasters before he became well known. The other case was the news anchor-person Margaret Dubuisson (['dubisan]), who had much trouble before her name was pronounced correctly twice a day on prime time television in the introduction to the evening news, as well as in advertisements for the evening news at other times during the day and night. The other category includes the rare bearer who finds the entire notion of mispronunciation offensive, and anyone who would bring it up offensively intrusive. One bearer of Dutreix objected to being asked if anyone mispronounced his name, and said 'If they make a mistake they make a mistake. Don't worry about it!'
Pronouncing French Names in New Orleans
333
(iv) Bearers' strategies for managing public pronunciation of their names Depending on the bearer's attitude to the (mis)pronunciation of his or her name, he or she will inevitably be faced with having to put that attitude to practical test. When this occurs, and assuming the bearer does not simply ignore the situation, he or she will accordingly have recourse to various ways of conveying the pronunciation of the name. A vast majority of speakers proceed on the assumption that the, or a, pronunciation of their name can be conveyed to an interlocutor should they choose to convey it, whether to correct or simply to inform. And a number of devices are revealed as being available for this purpose. Perhaps the most evident of these is the device of utterance or repetition. That is, the bearer takes it that the pronunciation can be learned or conveyed by virtue of the interlocutor or bearer saying it or repeating it. This can be done spontaneously. Or it may be done in response to the interlocutor asking for help (Jacquillon: 'Most people don't even try to pronounce it but will ask what it should be'). Or, finally, it may be done in response to an unsuccessful attempt at the name by the interlocutor (Duquesnay: "They are usually struggling to pronounce it and are happy for my assistance'). For some names, the device is available of adducing a homophone or a rhyme for all or part of the name. One way of proceeding is to mention the homophone(s) or rhyme(s) which cover(s) the entire name. For example, Argeanton can be covered by '"ar" like "car", "gin", "ton" as in "ton of coal'". (Others: Barre: 'It's like "Green Beret'". Cresson: 'You know Lee Press-On Nails? It sounds like that'. Faugeaux: 'I tell them to say [sings] "do-re-mi-fa-go" or sometimes tell them it's a lot like ['forgo] as in Wells Fargo'.) (A well known visual case of this is the refrigeration company Robert, whose logo, painted on company trucks, is a picture of a bear rowing a boat.) The homophone or rhyme can also be alluded to without mentioning it, as in Bourdeu ('like the city'), Brou ('like beer'), De Gruy ('it's just like the temperature'). A homophone or rhyme may cover part of the name, and the rest be adjusted by mentioning letters: Avril ('Take "April", take away the p and put the v and you have it'.), Huguet ('Drop the h and "you get" out of town'.). Or the homophone or rhyme may cover part of the word and the rest be assumed to be thereby clarified without further device: Barilleau ('You know the singer Barry Manilow?'), Dejoie ('It's a little like "days" and then the end'.), Jambon ('It's just like jam that you eat, and then bon could be either [ban] or [bon]'). Or it may happen that the bearer has only a homophonous scrap to cling to, as in one example of Deroche, whose bearer will say simply 'shhh'. A third device used by bearers is spelling. Some consider that the pronunciation can be conveyed merely by virtue of the interlocutor seeing the spelled name. A name such as Duthu or Faucheaux is taken by its bearer to be 'pronounced just like it's spelled'. Or the bearer will actually spell out the name, thereby 'conveying' the pronunciation, as in Viger: ' v as in 'victory', i-g-e-r'. Or the bearer, aware that the spelling of the name is not a sure guide to its pronunciation, will spell not the name but the name's envisioned pronunciation: Dupuy: 'It's spelled d-u-p-u-y but pronounced d-u-p-w-e'. Letters, however, can also be more consciously used as metalinguistic devices instead of as mere items in spelling recitation. For example, the bearer of Casteix tells people 'the / is silent'. If the interlocutor makes Castillon
334
George Wolf et al
into a four-syllable name, the bearer will point out that 'there is no / [after //] in the name'. Finally, the bearer may refer to the homophonous part of another word whose spelling is analogous to the corresponding part of the name, as in the case of Casbergue whose bearer tells people it is 'like catalogue: the g-u-e is [gs]'. Often the entire problem of finding devices can be side-stepped by using a language-name as a metalinguistic term. Thus, the pronunciation of Hezeau can be conveyed merely by saying 'It's French with a French ending'. This can sometimes be enough to ensure understanding. Another way of alluding to, and thereby conveying, the correct pronunciation is to eliminate a better-known mistaken rival which sounds like it, as when the bearers of both Ardenaux and Ardeneaux say 'it's not 'Arceneaux".
Toward lay metalanguage The lay metalinguistic devices sketched briefly above form an initial framework for comparison with linguists' discussions of pronunciation. Two points may be made: 1. Monolingual speakers do not seem to conceive of pronunciation problems in such terms as 'physical/articulatory barriers'. Such barriers would only appear to arise in 'pedagogical' contexts in which the interlocutor was being 'taught' the correct French pronunciation of a name. Only in such a context would the interlocutor 'discover' that he or she 'could not say' a French r or a French u. But outside of these specialized contexts speakers appear directly to target sounds which 'will do', as in the case of Calongne ('standard' French pronunciation [kaloji]) pronounced [ks'larn]. With regard to phonotactics, this shades into the issue of subjective attitudes and 'the sense of the language'. For example, it is a fact that English words do not end in [n]; but speakers are not physically prevented from pronouncing [p.] at the ends of words (it would be possible to end an utterance of onion at ra); rather, they may sense that it is 'not right' to do so, hence they will tend to substitute a sound that is familiar ([on], [arn]). It is in these terms that speakers approach 'the phonology of the language'. 2. It is clear that bearers' pronunciations of their names, the spellings of those names, and those names' mispronunciations by others are interrelated in very varied ways. There can be no direct correlation between pronunciation and spelling because the importance which bearers accord to the oral versus written tradition of their names is never constant; it not only changes from speaker to speaker, but as we have seen can vary for a single speaker depending on circumstances. One bearer will count as a correct pronunciation what another bearer of the same name will count as a mispronunciation (Amedee: ['aemgdeM'semsdi], Dauzat: ['doza]/['dozaet]; but they may change their views of this in some contexts. Given these considerations, we may tentatively try to define pronunciation as a notion which is used by speakers in daily communicational interaction: Pronunciation is a mode of utterance mediated for the speaker or the hearer by (i) the notion 'how an item sounds' in conjunction with the notions 'correctness' and 'incorrectness', 'good' and 'bad'; and by (ii) the ways in which such a mode of utterance can be publicly managed.
Pronouncing French Names in New Orleans
335
Such a definition emphasizes that pronunciation is a normative concept in which a certain category of physical activity is mediated by notions which speakers carry around with them and in terms of which they interpret their experience. This is related to a Saussurean point: that there can be no bare speech sounds apart from concepts, since such sounds are defined in terms of them; and on the other hand there can be no bare concepts apart from sounds since it is in terms of sounds that the concepts themselves are defined (Saussure 1922: 155). Thus, pronunciation becomes a focus of interrelated notions which speakers use to solve certain types of communicational problem. That network of notions can be informally reduced to some basic types, of which the following are central: First, the notion 'pronunciation' is itself a notion which speakers use to manage communication. Whether with respect to names or to words in general, it is a device which concerns the public nature of communicational units, and the importance to speakers of those units. The vast majority of people interviewed understood something by this term. What varied was what bearers understood by 'mispronunciation': often they would include in mispronunciation intentionally frivolous (or malicious) distortions of their names, interpreting the request for information as a prompting for a personal revelation - as opposed to a report of failed attempts based on ignorance or lack of skill. But in one way or another pronunciation is a concept they all use. A second device used by speakers is the alphabet. This is a prime example of a lay metalinguistic device, because it provides a manageable set of objects with which most people are familiar by virtue of an education which has taught them this set precisely in terms of its connection to oral speech. Thus, the alphabet emerges as an entirely separate domain of metalinguistic devices which appear inherently to include the notion of an essential connection between speech and writing. The objects have a convenient 'order', can be 'recited' in a completely conventional and universally recognizable manner, and each have names allowing them to be referred to, and thus giving them an extra dimension of manipulability. The alphabet gives speakers an easy set of tools for directly managing speech. A third concept is a kind of loosened and expanded version of the alphabet, the strictly visual form of which is the rebus. Here, instead of the restricted set of alphabet letters, we have at our disposal the entirety of things and concepts for which there are words. Here, unlike the alphabet, there is no conventional order, nor homogeneity of visual form, but on the other hand the notion of similarity of sound between metalinguistic object-name and pronounced item in the language is fundamental. And while the stringing together of homophonous items is a kind of 'spelling', the items which spell are designed to be in a relationship of identity with what is spelled. The success of this device depends on familiarity with the meaning attaching to the sound of the metalinguistic object-name (e.g. it is useless adducing the sound of beret if the interlocutor has never heard of one: they will need directions how to pronounce that word in turn.) A fourth concept available as a metalinguistic device is that of norm. Without a norm ('If they make a mistake, they make a mistake. Don't worry about it!'), anything people say (which will then be regulated for purely practical reasons) will be accepted without comment, and there will remain nothing 'linguistic' to talk about. Without the 'tyranny of the alphabet' (Harris 1986: ch. 2) it may be that the mere oral
336
George Wolf et al
imitation of others would provide a sufficient basis for successful communication. That is, the ability and propensity to imitate or repeat the sound of what someone has uttered would be taken as a linguistic universal (cf. Love 1990: 99 ff. [above, pp. lOlff], who makes a slightly different point). It may be that only when speakers are distracted by the question 'How would that be spelled?' is their confidence in their purely oral ability shaken. But be this as it may, it remains the case that we operate verbally with constant appeal to a norm, and hence we conceive of what people say, and how they say it, as 'right' or 'wrong', 'good' or 'bad'. A fifth useful concept is that of silence. In many cases silence can be used as a way of managing communicational exchange. In the case of an ignored mispronunciation, silence can serve to further the communication by preventing it from being disrupted. Or silence may be used in another way, as goading an interlocutor into self-correction. Silence can communicate assumptions, as when an unforthcoming but expected response prompts a reevaluation. Silence can protect. In the case of names, silence can also be used as an index of failure: a lack of response may mean a failed pronunciation. The above discussion, which may be taken as preliminary to a larger study, is designed to provide evidence that pronunciation embodies the action of uttering sound as mediated by an interrelated network of normative concepts in the possession of speakers. The interpenetration of the subjective with the objective informs every aspect of the study. There was no possibility of treating bearers as 'neutral agents of information', or interviews with them as a kind of ore from which to separate out the pure metal of scientific fact from the slag of informants' views about their own linguistic behavior. Rather, the facts were in part constituted by bearers' views of their and others' linguistic behavior. Here, the situation is the reverse of Labov's 'observer's paradox', according to which one expects knowledge to be gained by observing certain facts, while in reality any knowledge of those facts is taken out of the realm of accessibility precisely by virtue of their being observed (Labov 1972: 612, 209). For us, by contrast, the facts are constituted in and by the process of observation, and by virtue of the interaction between interviewer and bearer, as between bearer and interlocutor generally. In a sense, they are negotiated, but more, they are created in and by communication itself, involving as they do the network of concepts by virtue of which people communicate.
References Harris, R. (1986) The Origin of Writing. Duckworth, London. Kokeritz, H. (1964) Spelling pronunciation in American English. In D. Abercrombie etal (eds.), In Honour of Daniel Jones, pp. 137-45. Longman, London. Labov, W. (1972) Sociolinguistic Patterns. University of Pennsylvania Press, Philadelphia. Love, N. (1990) The locus of languages in a redefined linguistics. In H.C. Davis & TJ. Taylor (eds.), Redefining linguistics, pp. 53-117. Routledge, London. (Adapted as Ch. 7 above.) Martinet, A. (1977) La prononciation frangaise des mots d'origine etrangere. In H. Walter (dir.), Phonologic et socieU (Studia phonetica 13), pp. 79-88. Didier, Montreal. Onions, C.T. (1966) The Oxford Dictionary of English Etymology. Clarendon Press, Oxford. Saussure, F. de (1922) Cours de linguistique generale, 2*™* ed. Payot, Paris.
Pronouncing French Names in New Orleans
337
Appendix: select list of names with notes on comments by their bearers The following is a list of names referred to in the main body of the text. It consists of a representative sample from a larger data base of 984 interviews conducted almost all by telephone in the city of New Orleans (Orleans and Jefferson Parishes; see Michele Bocquillon, Ava Daniel, Debbie de la Houssaye, Phyllis Krzyzek, Darnell Lopinto, Clifton Meynard, Lisbeth Philip, George Wolf (1994) An alphabetical list of French names in New Orleans, with pronunciations and mispronunciations (in IPA script), and comments on these as reported by the bearers, Unpublished ms., Department of Foreign Languages, University of New Orleans). In the list, transcriptions in parentheses represent pronunciations which bearers consider to be mistaken. Alexcee [a'leksi] ([a'leksis]) Said her husband pronounces it better than she does. Amedee 1 ['aemade] (['aemadi]). Amedee 2 ['aemsdi] (['asmade]). Andrade 1 [aen'dradi] (['ondred], [an'dredi]) 'I tell them "That's not how you say it" and then have them say it after me.' Andrade 2 [aen'dradi] (['aendred], ['aendredi], [aen'dredi]) She doesn't bother to correct. Expressed strong irritation that a good friend of hers who is well educated can't seem to get the pronunciation of her name right. Andrepont 1 ['aendrspont] (['aendrspojnt]) He breaks it into syllables and gets people to pronounce it syll. by syll. Andrepont 2 ['aendrapant] (['aendrepant], ['aendrspojnt]) Spells it for them and pronounces it again. Andrieu 1 ['aendru] ([an'dreju], ['andre]) Just tells them it's ['aendru]. Andrieu 2 ['aendru] (['andre]) 'I tell them the mispronunciation is a whole different name. My name is just like Landrieu except you drop the L. Most people have heard of [former New Orleans mayor] Moon Landrieu.' Andrieu 3 ['aendru] Many mispronounce. Never just says the name, always says it then spells it. 'You would think in New Orleans people would know how to say it, but they dont. It's not all that common a name, with street signs or anything. We've given up on the real pronunciation a long time ago. Still, people have to hear it or they never say it right. That surprises me since this is New Orleans.' Antrainer ['ontrenar] Spells it. Tells people to think: first name is Ann, last name Trainer. Ardenaux ['ordano] (['arssno]) Says the name is not Arceneaux. Ardeneaux ['ardsno] (['ardsneks], f'arssno]) Says the name is not Arceneaux. Areaux 1 ['aerio] (['aero]) 'I laugh at them.' Areaux 2 ['aerio] (['aero]) 'It's the same as Breaux, only Breaux has a B.' Argeanton ['ordsmtan] 'It's easy to pronounce: "ar" like "car", "gin", "ton" like a ton of coal.' Armingeon ['armirjo^sn] (['armsnd^an], ['armiqdsan]) From Alabama; he said that one of his relatives changed the name to Armington; often he would begin to spell the name, get to ge and people would automatically finish it with ton. Aucoin 1 ['okwaen] (['okojn], ['okojn]). Aucoin 2 ['okwaen] (['okojn], ['oke], ['okojn], ['okwe]). Avril 1 ['evrsl], ['evnl] (['aevnl]) She says ['evrsl] then spells it. She says it is mispronounced as if spelled Avrille. Avril 2 ['evrsl] ([a'vril]) 'Do you repeat it?' 'I do sometimes but they can't get it right; sometimes I spells it. I say 'Take April, take the p away and put the v and you have it." Barilleau ['baerslo] (['basrslu], ['baerali]) She corrects by saying 'You know the singer Barry Manilow? . . . " Barre [bae're] ([baeri], [bar], [baer]) To correct she says 'It's like 'Green Beret". Barrosse ['baros] ([bg'rosi]) He says it is said [bar] like 'bar' and [ros] like 'Ross.' Bayard 1 ['bojordj, ['bejord] (['bojard]) When they lived in the country they went by ['bajard]. In the city they normally go by ['bejord]. The husband works for an oil company, and he couldn't
338
George Wolf et al
get them to use ['bejard], so they let it go. They can still tell they are getting a call from New Iberia if someone asks for ['bejard]. Bayard 2 ['bejard] (['bqjord]) 'I just tell them to say "bay" and then "yard" like in your yard. It's easy and people get it.' Bayard 3 ['bojard] (['bejard]) He doesn't mind if people don't get it right; it is often mispronounced. He knows his name is not the normal French pronunciation. His family was 'part of the 1066 invasion of England.' They settled in Ireland for several centuries before coming to America. He considers himself more Irish than French. Bayle [bel] (Cbeli], [bojl], [bajl]) He usually just has them repeat the name. 'I have used "bale of hay" or "bale out".' Bazile [ba'zil] ([brs'zil]) The name is mispronounced 'plenty a'times. I mostly just lets 'em go on with it.' Benoit [benwo] ([ba'nojt]) He breaks it into syllables, then uses exaggerated mouth movements. Berteau ['berto] (['bertaew]) He breaks it into 'bear' and 'toe.' Bourdais ['bordez] 'Bore days.' Bourdeu ['bordo] (['bordu]) 'It's like the city.' Bourquard ['bokord] (['bugard]) He tells people to use q instead of g. He says the French say [burkword]. Boutillier 1 [butalie] ([bantilis], [butisr]) He tells people to say [but] as in boot and to add 'illier.' Boutillier 2 [butilje] ([butuliar]) Just makes them repeat, or doesn't say anything. Bouzon [buzon] ([buzaen]) 'It depends on whether or not they need to get it straight.' Claims that when people spell her name they tend to change the o to an a. Bressard [brazord] ([bs'sord]) When her husband went into the Service they changed his name to Bressard from Bessard. Her husband never changed it back to the original form. Brossard ['brossrd] (['brusord], [bra'sord]) 'I'll say 'Broussard, and leave out the M'.' Brou [bru] ([braew], [bro]) 'It's [bru] like "you" or like beer.' Brouillette [bru'jet], [bru'let] Doesn't like to correct. She asked how the interlocutor would pronounce it; the interlocutor said that's what he was going to ask her. She said 'they'll say [bru'jet], [bru'let]; I like [bru'jet].' Her notion of the pronunciation of her name seemed entirely dependent on how others pronounce it. Brugier ['bnrye] (['brugsr], ['bargsr], ['brudjir]) Corrects 'only if it's really important; mostly people you'll never see again, so why make them miserable?' Brulet [bru'le] ([bru'let]) 'Like a silent t. I say "It's pronounced [bru'le]" and they're usually insulted.' Buffet ['bAfat] ([bA'fe]) Doesn't correct. Prefers ['bAfst] because people think [bA'fe] is funny and make jokes. Buisson ['bjuson], ['bwisa] (['busan], f'busian]) 'To someone who does not know N.O. culture I say it's either one or the other. I do it for different purposes, and what effect I can get; for clout.' Calongne [ka'larn] ([ka'lon]). Casbergue ['kassberg] (['kaesbsrgu]) Corrects by saying it's like 'catalogue': the gue is [g]. Casteix ['kaesteks] (['ka;snet], ['kaestin], ['koteks], ['kaesta]) 'I tell them the i is silent, then it is easy to pronounce.' Castillon [kaes'tilan], ['kaestalan] ([kaes'telo], [kass'tiljan]) 'If people are having a hard time with [kaes'tilan], I use the simpler f'kaestslon]. But if they make it a 4-syllable name, I say there is no i in the name.' Catoire 1 [kaetwo] ([katori], ['kaetsren]) Usually just lets it go. 'After all these years, it doesn't bother me anymore.' Catoire 2 [kaetwa;] ([ks'tori], ['kaetari], ['ketwae]) 'Imitate the pronunciation.' Catron ['kaetren], ['ketran] (['kastra], ['ketron]) When her father was in the Service he began to use the alternate form because people would get tongue-tied trying to say Captain ['kaetrsn]. He became Capt. ['ketren]. Champagne [Jaem'pen] No mispronunciations. 'It's just Champagne, c-h-a-m-p-a-g-n-e.'
Pronouncing French Names in New Orleans
339
Courvoisier [kurVwasje] ([kavazir]) 'Most of the people get it right down here. They butcher the spelling more than the pronunciation. I usually repeat by pronouncing the name correctly. I say the name and they say "Whoa, you better spell that".' Couvillon ['kuvian] ([ka'viljsn]) 'Pronounce it as if it was spelled Couvion. That's how we learned it in school like when you had to break it down into phonics. You can spell it quicker if you write it down instead of thinking. It's the Is that throws it off.' Both parents spoke Acadian French. Crais [krez] 'If they spell it they'll put a g on the end. It's pronounced as it looks.' Credeur [loredsr], [kra'dur] Also legitimate are [kredoer], [kredeir] Didn't correct people in the Service. 'I'm not really sure what the correct pronunciation is to tell you the truth. I tell everybody it's fkredsr]; then I'll tell 'em how to spell it. In Lafayette the old people who still speak French call it [kredoer]. Younger people who don't speak French say f'kredar]. It's [kredeir] in real French.' Cresson ['kresan] 'Just. . . you know Lee Press-On Nails? It sounds like that.' Cuiellette [kju'let] Think of [dsu'let] [Gillette] razor blades; instead of [cFsu'let] it's [kju'let]. Danjean 1 ['dasndsin] (['dasa]) 'The pronunciation is ['dasa] but here in the city they go by ['daendjin] . . . A lot of people mispronounce the name as ['dejon], which is a common name around Lafayette.' Danjean 2 ['dasa] 'Most people in the city when they see it pronounce the New Orleans way. When people hear ['daendsin] they think it is two words, with a hyphen.' He retains ['dasa] even though it is universally pronounced ['daendjin] in New Orleans. Dauzat 1 ['doza] (['dozaet]) Repeats. Dauzat 2 ['dozaet] (['doza]). De Gruy 1 [di'gru] ([di'gri], [di'gre]) 'Just say it; if they hear it, they got it.' De Gray 2 [ds'gri] ([ds'gru], [ds'gqj], [da'grui]) 'It's just like the temperature.' De Gruy 3 [de'gri] ([da'gru], [ds'grui], [dg'gre], [ds'gri]) To correct people, she pronouces it clearly, but sometimes they still don't get it. She says she thinks it's a mind set that tries to deal with the spelling but doesn't know how. De La Houssaye [dslo'huse] ([delshusi], [delahaewsi]) 'I just repeat it.' Dejoie 1 [deswa] ([dedipj], [dedsan], [ded^ojs]) He tells people to pronounce three letters at a time. He had a hard time, especially with the phone company, to correct their mistake of wanting to make two words of the name, capitalizing they. Dejoie 2 [dejwa] ([ded^ui], [didjoj], [did^o]) When she corrects people they will often remark: 'But it doesn't look like that.' She tells them to look again. 'It's a little like "days" and then the end.' Dejoie 3 [deswa] ([didjoj], [dedsoj], [ded^an]) He said he had a lot of fun with his name while in the navy. During roll call, he would refuse to answer to frequent mispronciations, and then when the roll was finished he would object to not having been called. Deroche [da'roj] Says 'shhh.' Dubuisson ['dubasan] (['dubwisan], [du'bojsan], ['dubs/an], [du'plantie]) Many people ask her to pronounce it for them. She simply repeats. The spelling problem is worse. Duplantier [du'plaje], [du'platje] ([ds'plaentis], [du'plesis], [du'plaentar]) 'I just tell them it's [du'plqfe] or [du'platje].' Dupleix ['dupleks] ([dupli], [du'pleksis]) 'I say my last name is not [dupli] or [du'pleksis], it's Juliette ['dupleks].' Dupuis [du'pwi] (['dupis], [du'pri]) Does not usually correct unless they ask. Repeats, usually. 'It's likep-w-e . ..' Dupuy [du'pwi] 'It's spelled d-u-p-u-y but pronounced d-u-p-w-e.' Duquesnay ['duksne] ([du'kezne], [duken]) "They are usually struggling to pronounce it and are happy for my assistance.' Dureau [da'ro], ['djuro] 'Some family members say ['djuro] like "bureau", others have a more French accent like [da'ro]. [ds'ro] is how I say it. Almost no one gets it correct. Over the telephone I'll say "This is Mrs [ds'ro]". Socially I wouldn't correct anyone. But if it were in business or in legal . . .'
340
George Wolf et al
Durocher [dero'Je], [da'rojsr] 'The pronunciation is [dsro'Je]; other people say [da'rojar]. I use it the way people say it. The latter is a lot more common here. I just tell them how it's said. He was from Thibodeaux, and had a Cajun accent. Avoided answering the question how he pronounces the name. Dusaules ['dusolz], [da'solz] ([ds'sael], [da'saeli]) 'The Americanized way? It's ['dusolz] or [da'solz]. That's our pronunciation. It think it used to be [da'sael], [da'saeli], something like that. They tend to spell it with De.' Corrects. 'In a polite way. If they tore it up, it's ['dusolz]. Without being ugly about it.' Duthu ['du6u] (['dudu], ['dudi]) 'I tell'em it's pronounced just like it's spelled.' Dutreix ['dutre] 'If they make a mistake they make a mistake. Don't worry about it!' Ecuyer [ekwije] ([ekwiji]) Usually people who telephone for the first time say 'E...E...E...' and he pronounces it. This doesn't bother him. Faget [fase] ([fajgst]) Repeats it. If someone still can't understand [fase] she will use [faset] to see if that helps. Faggard [faegard] ([faegat], [fs'gard]) Just lets it go. Sounded as though it was slightly frustrating to have it pronounced [faegst] but would not say how much. Fagout This was misspelled; it is really Fagot. They just changed it since it has been mispronounced so often as [faegst]. Did not want her children to have to live with all the bad jokes. Now their name is Santos, her husband's grandmother's name. Faucheaux 1 ['foje] (['fojo]) Says it's pronounced just like it's spelled, so not too many people mispronounce it. If they do say it wrong, she spells it and says it for them. Faucheaux 2 [foje] ([fotfeks]) Says she has more of a problem with people misspelling her name. The two most common examples are Fouchay and Fouche. Faucher [foje] ([faewtfsr], [fokar]) 'Just pronounce it right. It's easy once you hear it. Some people want to make the soft ch into a hard k sound.' Faucheux [foje] ([fotfeks], [fojo]) Grew up in St John Parish where she spoke French before she spoke English. Name used to be pronounced [fo|0]. Some of her husband's family have settled on [fojo]. It doesn't bother her if people mispronounce the name, but it does upset her husband who corrects people with a clear pronunciation. Faugeaux [fago] ([fs'go], [forgo], [fagies], [faego], [faegst]) 'It's like in the jingle that most people know from "The Sound of Music". I tell them to say [sings] 'do-re-mi-fa-go' or sometimes tell them it's a lot like [forgo] as in Wells Fargo.' Foucheaux [foje] ([fojo]) Doesn't correct. The son lives out of town and says [fojo]. Fryou [friu] ([frqju]) Gets mispronounced 'on account of the spelling of it. I correct it but it don't do any good; I just tell 'em it's ['friu] and they go right back to [frqju].' Geathreaux ['go6ro] (['goOraks]) 'I tell people it's like "Go throw the ball".' Hebert ['eber] (['ibart], ['hsrbsrt]) Depends if he sees the person more than once. If it's a waiter or in a department store, doesn't bother correcting. Hezeau 1 ['hezo] (['hozo], ['hizo]) Tells people who mispronounce that it's a French name with a French ending. Hezeau 2 [hezo] (['hekson]) Most of the time people have a hard time pronouncing it when they see the spelling, when they see h, z, and u. Houin ['hjuin] Interlocutor pronounced it [hug] and when the bearer was asked if her name was mispronounced she said no and that the interlocutor had pronounced it correctly. Huguet 1 [Ju9et] ([ha'gwet]) 'Drop the h and "you get" out of town.' Huguet 2 [juget] 'Pronounce like "you" and "get".' Jacque [d^ok] (['dsaekju]) 'It's pronounced like "jock" or as in Jacques Cousteau.' Jacques baks] (foaka], bake], bakes]) 'I tell them to think of someone they've probably heard of, baks] Cousteau.' Jacquillon ['joklio] ([dja'kwiljsn]) 'Most people don't even try to pronounce it but will ask what it should be. It helps to tell thern^the Is have the same sound as in court-bouillon.' Jambon ['dsasmbon], orig. babo] (['d^aenbasn], ['d^emsan]) 'It's just like jam that you eat, and then bon could be either [ban] or [bon]. We like [ban].'
Pronouncing French Names in New Orleans
341
Jaquillard 1 ['cfeaekwilard] 'They have all kinds of ways to say it - almost anything with an "ard" or "nard". Most people want to put a c in it.' She said she couldn't give specific errors. She spells it for them and points out the q and the fact that there is no c. Jaquillard 2 ['dsaekalard] ([dsg'kilard]) Sometimes they drop the J and say ['aegalord]. Her husband volunteered that often people think he is saying his name is Jack Lard. Laguaite [la'get] ([la'kwet], [te'gwat]) 'I tell them I'll answer to "L" if they get that much!' Even after she pronounces it, people still stumble around trying to make sense of the spelling. Lalande [Island] ([b'len], [Is'lendi]) 'I say it for them, but they sometimes still don't get it. It doesn't really bother me though. I'm 97 years old and I don't let little things upset me.' Lepree [la'pri] 'Honey, how do you pronounce our name?' People get it confused and say 'Dupree'. This alternate form is not treated as a mispronunciation but as a separate form. Maureau ['moro], [moro] (['miro]) 'Just pronounce it.' The wife answered the phone and in response to the query said Til let you talk to my husband—he's the ['moro].' After talking with the husband who said his name was [moro], the interlocutor asked him if he ever hears the pronounce ['moro]. He said no. Mayeaux 1 ['meju], ['mejo] Most people pronounce it one of the two ways the family uses. There were mistakes made on birth certificates, and the family has just lived with it. Mayeux 2 ['mejar] (['mejo]) 'I tell them it's like'Mayor Barthelemy'. Mayronne [meron], [ma'ron] ([me'roni]) Sometimes when she is shopping by phone and they can't find her name, they find it listed under the Rs (May Ronne). Orgeron ['osara] (['orsgan], ['orjaran]) Corrects if they say ['orsgan], not if they say ['orsaran]. Orillac ['orslask] Is from Panama, where the pronunciation is [oriclsak]. Here, she let's people say what they want. Pertuit [pertwit] ([pertuit], [pertu]) 'I tell them it's "purr" like a cat and "tweet" like a bird.' Petitbon ['petiban] Says it should be ['pstibon]. Not often mispronounced. Her husband was an NFL football player, so his name is well known. Petitpain [petipaen] ([pstitpae], [pitsrpasn]) Gets 'everything but "petty larceny".' Lets it pass. Quezergue 1 [kfzer] (['knzare], [kra'vergi]) Corrects by just saying it right. Many people try to pronounce the gue. Quezergue 2 [ks'zer] (['kezar], [kju], [ks'zerki]) 'I write it out and then I write out how it should be pronounced: Ca-zair.' Viger t'visqj] (['viger], ['vqjger]) Corrects by spelling: 'V as in "victory", i-g-e-r.' Voisin [vwaze] (['vojzm], [waze], ['vwazin]) Corrects by saying 'I will spell the first name for them and then pronounce it.' Her husband was born in Grand Caillou, didn't speak anything but French until age ten, and hadn't heard anything but the French pronunciation of the name. He entered the army and after a long time said to a friend, 'When are they going to call my name?' Friend said 'They've been calling it all along!' He turned out to be AWOL.
This page intentionally left blank
Index of topics abstraction 96, lOOff, 156, 242, 297 abstraction, autoglottic 258 abstractions, determinate sets of 101 alphabet 252ff, 335 alphabetic principle 140 amplificational function 235 anthropology 17, 160, 262, 318 arbitrariness 115, 118, 126ff, 132, 244-5 arbitrariness and linearity 126ff, 265 autoglottic autoglottic abstraction 258 autoglottic inquiry 257-8 autoglottic space 259 bad language 283ff behaviour 6, 8, 16, 17, 100, 166, 196, 202, 211,214, 281, 291, 296ff behaviour in-the-mass 43 behaviour in communicational interaction 201 behaviour, bodily 76 behaviour, communicative 9, 40, 234 behaviour, conversational 177, 197, 216-7 behaviour, co-operative 38, 221 behaviour, first order 104 behaviour, human 6, 17, 211, 235, 239, 242, 305 behaviour, interactional 38 behaviour, linguistic 40, 66, 77, 156, 2978, 306, 336 behaviour, metalinguistic 306 behaviour, patterns of 231 behaviour, regulation of 57 behaviour, rule-governed form of 39 behaviour, social 8, 38 behaviour, standards of 292 behaviour, systematic 320 behavioural routines 231 bi-planar 79, 149, 215, 244 biomechanical scale 234ff communication 3ff, 27ff, 49ff, 54, 69, 86, 99, 119, 159-60, 177,208 communication and language 198 communication as transmission and association 159 communication, code model of 72, 77 communication, economics of 236 communication, grammaticality and 192ff
communication, graphic 261 communication, human 19, 316 communication, intersubjective 156 communication, language and 3ff, 54 communication, linguistic 3, 19, 69, 73ff, 104 communication, meaning and 159 communication, non-referential 117 communication, open-ended 33 communication, perfect vs imperfect 288 communication, problems of 299 communication, referential 118 communication, spoken 192 communication, successful 336 communication, synchronic 72 communication, textualization as 232 communication, thought-transferrence theory of 99, 109 communicational communicational activity 66, 328 communicational behaviour 234 communicational competence 27ff, 205 communicational difficulties 55, 296 communicational effect 96 communicational efficacy 194ff communicational exchange 50, 102, 328, 336 communicational intentions 57-8 communicational interaction 201, 205, 239, 335 communicational intersubjectivity 199 communicational needs 219 communicational practices 24, 57 communicational problem 329, 335 communicational process 120 communicational proficiency 13,44, 23Iff communicational prophylactics 69 communicational purposes 234 communicational relevance 54, 217,239 communicational role 325 communicational space 9, 19 communicational success 195, 200, 205 communicational understanding 69 communicational units 335 communications technology 252, 259 communicative 195-6 communicative acts 297 communicative behaviour 234
343
344
Index of topics
communicative competence 27ff communicative contexts 286 communicative intentions 161 communicative interaction 197 communicative rules 36, 37 competence 77, 281 competence/performance 301 competence sentence 174, 176 competence, linguistic 177 competence, linguistic vs communicational 27ff conceptual clarification 246 context 12, 13, 305 context-based 214 context-dependent 6, 144, 207 context of situation 32, 33 context of utterance 51 context, situational 203 context, social 281 context, total 9 contextual contextual constraints 216 contextual criteria 148 contextual cues 216 contextual disambiguation 42 contextual factors 38, 123 contextualization, complimentary 239 contextualized contextualized act of communication 119 contextualized interpretation 33 contextualized speech practice 149 contextualizedness 145 conversation 30 conversation analysis (CA) 137, 201, 209ff, 250 conversation, fictional 209ff conversation, naturally occurring 209 conversational conversational behaviour 177, 216-7 conversational breakdown 210 conversational competence 175 conversational implicature 144, 161, 217 conversational interaction 194, 201, 209ff conversational moves 297 conversational pattern 220 conversational performance 177 conversational purposes 222 conversational resources 218 conversational rules 30 conversational speech 171ff conversational speech, discontinuity in 174ff conversational structure 297 cunning 319-20 deceit 319ff derogatory terms 285-6
descriptive descriptive adequacy 248 descriptive nihilism 18 descriptive pluralism 18 determinacy 18,21ff, 119 determinacy of form 119, 149 288 determinacy of interpretation 108 determinacy of meaning 107,119,149, 287-8 diachronic linguistics 89 dialect 83ff, 310 dialect/subdialect problem 91-2 dialect vs accent 87-8 dialect, Greek linguistics on 83ff dialect, lay vs linguistic classification of 91, 93-4 dialect concept, aggregate 87ff, 92 dialect concept, continuum 86ff, 92 dialect concept, relational 86ff, 92 dialectology, structuralist dilemma in 91 dictionary 143, 288-9, 292, 315 dilemma of reflexivity 137 dirty words 285-6 discourse discourse analysis 211, 213, 250 discourse creativity 209 discourse modelling 214 discourse style, male vs female 295 discourse, system of 316 discourses, intellectual 305, 309 discourses, political 305, 309 dynamic paradigm 31 editing rules 177ff education 32 288, 292, 335 ellipsis 138, 174, 191, 196 emblem vs token 263ff emblematic recording 263 English 305 English, American 305ff English, Black 311 English, Condensed 131ff English, dialects of British 310 English, knowledge of 176 English, standard varieties of 286, 308 epistemology and science 120ff gtatdelangue 90-2 ethnocentric categories 273 ethnography 320 ethnography of speaking 281 ethnomethodology 30, 37, 137ff, 197, 201 experience, linguistic 25, 96, 146, 212, 288 explanatory adequacy 248 fails de langue 25 family resemblance 154-5 feminist linguistic theory 288
Index of topics fictional dialogue 209ff first-order first-order behaviour 104 first-order language use 96, 104 first-order linguistic experience 109 first-order reality 103 fixed code 24, 49ff, 73, 77, 79, 82, 92, 100, 109, 131,215,219,282,288 folk linguistics 273 four-letter words 285, 290 games 38ff, 117, 165-6 gender 273ff gender and form 275 gender and sex 273ff gender, grammatical 274 gender, natural 274, 276 gesture 196 graffiti 227 grammar 50, 177, 180 grammar, a 20, 23, 50, 143, 316 grammar, a, not rule-governed 23 grammar, a, indeterminacy of 24 grammar, formalization of 257 grammar, knowledge of 180 grammar, pedagogical 273 grammar, rules of 33 grammars 288, 315 grammatical grammatical analysis 171 ff grammatical categories 273 grammatical custom 280 grammatical features 308 grammatical gender 274ff grammatical incoherence 196 grammatical knowledge 72 grammatical practice 280 grammatical rules 10 grammatical theory 247 hearability vs audibility 78 hesitation phenomena 174, 176 heuristic approach 247 holistic view 90-1 hypothetico-deductive linguistics 247 hypothetico-deductive science 72 iconic characters 254 idea transfer 74, 76 ideal speech community 79 idealization 171ff identificational function 235 ideology 273,278,280-2 ideology and linguistic description 273 ideology, patriarchal 280 idiolect 87,93,97 idiom 318
345
idiom chunks 133 idiosynchronic 90 illocutionary illocutionary acts 213 illocutionary force 213, 218, 232 illocutionary/perlocutionary distinction 218 imitation 115,336 indeterminacy 2Iff, 77, 97ff inscription 227 integrational 5ff, 15ff, 146, 311 integrational approach 283, 287-8, 305ff integrational function 19 integrational linguistics 15, 75ff, 109, 152ff, 305 integrational model of signification 113, 119ff integrational patterns 25 integrational scale 234ff integrational semantics 111 integrational semiology 19, 24, 238-9 integrational semiology, axioms of 238 integrational view of translation 316ff intentionality 160, 165 intentions 121-2, 159ff, 209, 211 intentions and beliefs 159, 162 intentions, communicational 57-8 interaction 147, 209ff, 300, 336 interaction-structures 209 interaction, classroom 213-14 interaction, communicational 205, 335 interaction, communicative 157 interaction, conversational 138, 194 interaction, doctor-patient 294 interaction, linguistic 5-6 interaction, ordered 213 interaction, social 5ff interaction, verbal, role of the hearer in 198ff interactional interactional aims 234 interactional context 8, 105 interactional episodes 105 interactional good faith in mutual orientedness 146 interactional matrix 8 interactional memories 153 interactional practice 153 interactional proficiency 153 interactional risk 77 interactional situation 99 introspection 50 intuitions 50 irony, verbal 148, 159ff iterability 77, 147 iterable forms 109 iterable signs 122
346
Index of topics
judicial interpretation 60 knowing the language 13, 31, 103 knowledge 120 knowledge of English 176,231 knowledge of facts 120 knowledge of grammar 180 knowledge of language 77 knowledge, linguistic 20, 77, 102, 176 knowledge, linguistic vs non-linguistic 196 knowledge, scientific 121, 124 knowledge, transmission of 123 language 3ff, 15, 76-7, 83, 96ff, 252, 325 language acquisition 77 language and art, social functions of 256 language and communication 3ff language and law 294ff language and meaning 11 Iff language and sex difference 274ff language and society 5, 314-5 language and speech 257 language and tool-making 230 language as ability 15 language as a distorting lens 297 language as an autonomous system 281 language as autonomous social force 296 language as creative process 25 language as faculty 15 language as fixed code 219,282 language as hidden calculus 166 language as living organism 278 language as means of communication 54 language as mode of action 9, 15 language as social interaction 5ff language change 20, 23, 104 language differences 86 language faculty 16 language game 11, 18, 49, 54-5 language is open-ended 18,20 language is radically indeterminate 99 language myth 21, 47ff, 72, 77, 82 language study, cross-cultural 80 language study, history of 273 language use 49, 100 language use, first order 96, 104 language vs a language 76, 102-3 language, abusive 284ff language, bad 283ff language, code of 299-300 language, communal 97 language, definitions of 15 language, formal study of 315 language, idea transfer as function of 74 language, men's 288 language, ordinary 299, 301 language, origin of 100
language, real nature of 242 language, sexist 279 language, spoken 105, 227 language, standard 306ff language, unsponsored 258 language, women's 288 languages 96ff languages and games 117 languages as abstract objects 18 languages as biological objects 18 languages as psychological objects 18 languages as second-order constructs 103 languages as semiotic objects 18 languages as systems 16, 17, 96, 98, 151 languages, fixed-code theory of 109, 288 languages, national 89 languages, no such thing as 24, 93, 99 langue/parole 7, 129-30, 281 law 57ff, 294ff law, linguistics and 294ff lay lay audience 295 lay categories 299, 301 lay communicators 44 lay discourses 294 lay ideas about language 278 lay linguistic category 293 lay metalanguage 325 lay metalinguistic device 334ff legal procedure 57ff legibility 236-7 legibility vs visibility 78 linearity 126ff, 262-3 linearity and arbitrariness 126ff, 265 linearity and semantic structure 266 linguist, accountability of the 294ff linguist, political role of the 311-2 linguistic linguistic analysis 237, 24Iff linguistic behaviour 40, 66, 156, 297, 306 linguistic behaviour, idealization of 298 linguistic change 161, 234, 310 linguistic communication 3, 19, 69, 73ff, 104 linguistic communication and games 165-6 linguistic communities 5 linguistic competence 247 linguistic creativity, risk and 157 linguistic description 99, 243 linguistic differentiation 86 linguistic evidence 20 linguistic experience 25, 104, 109, 288 linguistic facts 8 linguistic features 306ff linguistic form 33 linguistic homogeneity 92-3 linguistic impoverishment 256
Index of topics linguistic inquiry 25 linguistic intercourse 6 linguistic intuitions 298 linguistic knowledge 20, 77, 102, 176 linguistic knowledge vs general knowledge 20 linguistic meaning 33 linguistic performance 32 linguistic practice 90 linguistic prescriptivism 292 linguistic reform 279 linguistic representation 280-1 linguistic sign 16, 24, 113, 122, 126ff linguistic structure 23, 77 linguistic systems 93 linguistic theorists 101 linguistic theory 6, 16, 24, 232, 288, 295 linguistic vs communicative competence 29, 31,35,37,42 linguistic units 96 linguistics 108, 300ff linguistics and standard languages 305 linguistics as a form of philosophy 25 linguistics as a language game 18 linguistics as science 160 linguistics methodology 298 linguistics, accountability in 294ff linguistics, decontextualized 149 linguistics, descriptive 31, 35 linguistics, diachronic 89 linguistics, dilemma of 14 linguistics, folk 273 linguistics, forensic 297 linguistics, fundamental assumption of 106 linguistics, generative 3 Iff, 44, 92ff, 122, 136, 246-7 linguistics, God's truth 18, 241ff linguistics, Greek 255 linguistics, historical 23 linguistics, hocus pocus 18, 24Iff linguistics, integrational 15, 75ff, 109, 305 linguistics, integrationist critique of 305 linguistics, methods of 299 linguistics, non-realist 24Iff linguistics, orthodox (mainstream) 6, 15ff, 73ff, 148, 253, 286 linguistics, rule-based models in 27 linguistics, scientific status of 300 linguistics, segregational 5ff, 15 linguistics, sentence-based 196 linguistics, synchronic 89-90 linguistics, teaching of 305ff linguistics, theoretical 159 linguistics, therapeutic 295-6 literacy 230, 252, 256-7, 284
347
literate revolution 256 Locke's puzzle 198ff logogram 253 lying 319ff macrosocial scale 234ff macrosocial proficiency 234 marking 279 meaning 8, 11 meaning and communication 159 meaning and linearity 126ff meaning, conventional 143 meaning, conveyed 144, 148 meaning, determinacy of 107, 288 meaning, dictionary 11,61 meaning, idiomatic 143 meaning, institutionalized 146 meaning, language and 11 Iff meaning, lexical, determinacy of 287 meaning, literal 50, 139, 143ff, 316 meaning, relative neutrality of 279 meaning, sentence 35-6, 143, 150 meaning, speaker 144 meaning, theory of 161ff meaning, transfer of 177 meaning, utterance 144, 150 metalanguage 132, 136, 288, 297, 300, 303 metalanguage/object language 136 metalanguage, descriptive 109 metalanguage, lay 325 metalanguage, scientific 300 metalanguage, theoretical 328 metalinguistic 273 metalinguistic commentary 79 metalinguistic discourse 104-5 metalinguistic expressions 96-7 metalinguistic operations 100 metalinguistic performance 96 metalinguistic practices 25, 136 metalinguistic system of categorization 151 metalinguistic terminology 293 metalinguistic terms 334 metalinguistic vocabulary 69 metalinguistics 83, 250 metasemiotic function 261 metasign 132, 26Iff metasign, phylogenesis of 261 metasign, synoptic 261 metasign, visual 261, 269 mimesis 115 motoric recording 267 mutual orientedness 146 names 121, 324ff names, bearers' attitudes toward pronunciation of 33Iff
348
Index of topics
names, bearers' strategies for managing public pronunciations of 333ff names, written forms of 330ff names, natural vs conventional 85 national identities 312 national languages 89 naturalism vs conventionalism 199-200 negative existentials 52-4 non-compartmentalization principle 9 normative concepts 336 notation, analysis and 24Iff oath 292 object language/metalanguage 136 obscenity 284 observer's view 104 observer's paradox 336 oral oral imitation 336 oral patterns and writing 263 oral performance of written words 239 oral repetition 256 oral speech 335 ordinary language 299, 301 orthographic words 133 paradigmatic relations 126 paralinguistic features 196 parole 7, 8, 70, 308 performance difficulties 174 performance, conversational 177 performative 123 perlocutionary effects 213 phatic communication 82 philosophy philosophy of language 159 philosophy of linguistics 244, 249 philosophy of science, realist 244, 247 phoneme 10, 97-8, 107, 127, 254, 308, 310 phonemic phonemic analysis 245-6 phonemic notation 255, 258 phonemic principle 254-5 phonetic representations 44 phonetic traditions 325 phonetics 19 phonetics and phonology, boundary between 326 phonetics in Greek linguistics 255 phonocentric 70 phonogram 253 phonological phonological features 308 phonological rules 10 phonological system 87 phonology phonology, French 326
phonology, pan-dialectal 246 pictogram 253 pictograph 261 plaquettes 267ff Platonism 248 poetic features 196 pointing 244 polylect 83 polysemy 98, 143 pragmatics 10, 149, 281, 299 preliterate society 252 preliterate cultures 258 prescriptivism 273, 280 primacy of speech 19, 253 primitive mentality 318 primitive languages 256 printing 291 pronunciation 88, 324ff pronunciation, definition of 335 pronunciation, distortion of 33Iff pronunciation, spelling 326ff proposition 160-1, 164 proposition vs utterance 161 prosody 216 proto-emblem 267 proto-sign 267 proto-token 267 proto-writing 26Iff prototype theory 79, 154ff psychology 16, 25, 89, 114, 154, 160, 211, 229, 256-7, 261 realism 242ff recipient design 210 regional differentiation 86 relativist vs universalist 136 relevance, principle of 77, 222 repair 30, 216 repetition lOlff, 333, 336 replicability 298 representation 115, 243-4 representation of a sentence 243ff representation of a structure 243 representation of spoken language 253 representation, linguistic 280-1 representation, pictorial 253-4 representation, pictorial vs non-pictorial 245 representation, semantic 44, 149 risk 156-7 rule 33ff, 209 rule-based models 27 rule, golden 58ff rule, literal 58ff rule, mischief 58ff rules-units distinction 130 rules, communicative 36, 37
Index of topics rules, constitutive 39 rules, conversational 30, 37 rules, generative 88ff rules, internalized 25 rules, internalization of 32 rules, speech act 39ff rules, variable 20, 34, 219 schematic recording 267 scientific method 173, 199, 298, 300 scientific discourse 199 scribbling 264 script 262, 266 script, foreign 330 scriptism 88, 192 second-order constructs 12 segregational linguistics 5ff, 15 semantic semantic derogation 280 semantic enclave 228-9, 238 semantic features 274 semantic interpretation 148 semantic representations 44, 149 semantic rules 10, 40 semantic structure 128 semantics 73, 136, 299 semantics, integrational 111 semantogenic causes 296 semiological analysis 237 semiological organization 238 semiology (semiologie) 113,117, 228ff, 261 semiology of textualization 228ff semiology, integrational 19, 119,238-9 semiotic 113, 244 semiotic coherence 140 semiotic displacement 280 semiotic neutrality 268 sentence 97, 273, 297 sentence-based linguistics 196 sentence meaning 35, 143, 150 sentence vs utterance 145, 150, 257 sentence, competence 174, 176 sentence, theory of the 164 sequentiality 133ff sex 285,287 sex and gender 273ff sex and meaning 274 sex difference and language 274 sex reference 275ff sexist language 279, 288 sign 19, 113ff, 165, 199, 200, 232, 261ff sign and metasign 261 ff sign-in-context 124 sign systems 90-1 sign-using 199 sign, arbitrary nature of 115, 118, 244 sign, complex vs non-complex 264
349
sign, definition of 19 sign, displaced 290 sign, iconic 118 sign, integrational model of the 119ff sign, invariant 316 sign, linguistic 16, 24, 113, 122, 126ff sign, linguistic, indeterminacy of 23-4 sign, misused 290 sign, natural 113,200 sign, non-arbitrary nature of 238 sign, non-linguistic 117 sign, scriptorial 262 sign, spontaneous ontogenesis of 120 sign, structural model of the 117ff sign, visual 261ff signification signification and human intention 121-2 signification, structural model of 113, 117ff signification, surrogational model of 114ff signification, dynamic dimension of 116 signifier/signified (signifle/signifiant) 127-8, 199, 265 silence, uses of 318, 336 simplicity, criterion of 248, 298, 300 social social action 138 social activity 15 social behaviour 38 social coherence 140 social context 281 social differentiation 86 social interaction 5ff social organization 19, 281 social practices 6, 16 social sciences 298 society 294 society, academia and 295 society, experiential and ideological order of 318 sociolinguistics 5,7, 13,281,285,296 sociology 15, 16, 137, 160 sociology of law 302 speech 19, 115,252 speech act 123, 150, 160, 193, 213, 289ff speech-act rules 39ff speech-act theory 39, 123-4, 144,217-8, 232, 289, 301 speech act, indirect 148 speech and writing 19, 262, 269 speech circuit (circuit de la parole) 5, 70, 72, 104 speech communication 19, 25 speech community, homogeneous 80, 92 speech event 297 speech signal 178 speech sounds 19, 335
350
Index of topics
speech, conversational 17Iff speech, infinite variety of 299 speech, representation of 252 speech, spoken 245 spelling 33 Iff statutory interpretation 60ff stereotyping 308, 312 stigmatized words 289 structuralism 308 style 93 styles, ethnic 296 styles, gender 296 stylistics 209, 215, 222, 297 surrogationalism 121 surrogationalism, psychocentric 115, 136 surrogationalism, reocentric 115,136 swearing 285, 289ff swearwords 289ff symbol 160 synchronic synchronic communication 72 synchronic linguistics 89-90 synchronic system 7 synchronic variation 104 synoptic reduction 264ff synoptic equivalence 261 syntactic discontinuities 30 syntagma 129-30 system 116, 151, 177 system of connections and hypotheses 314 system of discourse 316 system-sentence 244 system, la langue as a synchronic 7 system, language as an autonomous 281 system, legal and moral 318 system, visual sign 263 system, writing 246, 253 systems of abstractions, languages as 105 systems, decontextualized 7 systems, linguistic 93 teaching 305ff telementation 68ff, 119, 157, 198, 215 text 227ff, 297 texts, meaning of 313 textualization 227ff thought thought transfer 55, 69, 76, 99, 100, 109, 119, 198, 202 thought, restructuring of 252ff time, linear conception of 126-7 token 135, 263ff token-iterative recording 263 token vs emblem 263ff translation 245, 313ff translation and surprise 314 translation of culture 318
translation, replicational theories of 316 translation, transferrence theories of 316 truth-telling 318 turn-taking 137 type/token 96ff, 105, 177, 232 understanding 100, 198ff understanding-claim 203ff understanding, intersubjectivity of 199 understanding, mutual 69 understanding, theory of 163 universalist vs relativist 136 universals 69 usage, good and bad 25 utterance 96ff, 123, 134, 171ff, 243, 245, 318, 333, 335 utterance construction 147 utterance interpretation 147,151 utterance meaning 144ff utterance-storage model 135 utterance vs sentence 145, 257 utterance, conversational 17Iff utterance, metaphorical 151 utterance, proposition vs 161 utterance, theory of the 164 utterances, communicational efficacy of 196 utterance, communicative 196 utterances, communicationally successful 193 utterances, confusion of sentences with 13 utterances, discontinuous 171ff utterances, repeatability of 104 utterances, semantics of 184 utterances, syntax of 184 variation 103, 308ff words 97,297,313 words, four-letter 285, 290 writing 19, 105ff, 115, 122-3, 252ff, 261ff, 329 writing and memory 262 writing and proto-writing 26Iff writing system 106-7, 136, 246 writing, Aristotelian theory of 261 writing, cognitive consequences of 261 writing, cognitive prerequisites of 261 writing, Greek contribution to 253 writing, history of 253-4 writing, memory and 262 writing, non-verbal forms of 252 writing, origin of 26Iff writing, romantic view of 252ff