Functional and Systemic Linguistics: Approaches and Uses 9783110883527, 9783110127409

Citation preview

Functional and Systemic Linguistics

Trends in Linguistics Studies and Monographs 55

Editor

Werner Winter

Mouton de Gruyter Berlin · New York

Functional and Systemic Linguistics Approaches and Uses

Edited by

Eija Ventola

Mouton de Gruyter Berlin · New York

1991

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter & Co., Berlin.

© Printed on acid-free paper which falls within the guidelines of the ANSI to ensure permanence and durability

Library of Congress Cataloging in Publication

Data

Functional and systemic linguistics : approaches and uses / edited by Eija Ventola. p. cm. — (Trends in linguistics. Studies and monographs ; 55) Includes bibliographical references and index. ISBN 3-11-012740-7 (alk. paper) 1. Systemic grammar. 2. Functionalism (Linguistics). 3. Discourse analysis. 4. Language and education. I. Ventola, Eija. II. Series. P149.F86 1991 91-34972 410 —dc20 CIP

Die Deutsche Bibliothek

— Cataloging in Publication

Data

Functional and systemic linguistics : approaches and uses / ed. by Eija Ventola. — Berlin ; New York : Mouton de Gruyter, 1991 (Trends in linguistics : Studies and monographs ; 55) ISBN 3-11-012740-7 NE: Ventola, Eija [Hrsg.]; Trends in linguistics / Studies and monographs

© Copyright 1991 by Walter de Gruyter & Co., D-1000 Berlin 30 All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Typesetting: Arthur Collignon G m b H , Berlin — Printing: Oktoberdruck G m b H , Berlin Binding: Lüderitz & Bauer, Berlin Printed in Germany

Preface

When the 16th International Systemics Congress was held in Helsinki in 1989, it was the first time that the conference had been organized in a non-English speaking country. At the same time that there was an expansion in a geographical sense, it seemed appropriate to widen the horizons in other respects as well. Thus, the congress called for papers on the latest developments in systemic and other functional theories of language. The purpose was to establish a dialogue with linguists working within, for example, the Prague school linguistic tradition, functional textlinguistics, stylistics, applied, and educational linguistics. The response to the call is reflected in this collection of papers from the congress. Part I begins with general theoretical considerations and then moves on to specific issues and approaches. Enkvist views text production as a dynamic, goal oriented process — as a realization of weighted choices of systems. Such factors as salience, iconicity, patterns of information flow, and syntactic markedness in texts are taken to be controlled by global, hierarchic weightings of decision parameters in the production of various text types. Lemke discusses how the processes of producing a text change the context in which text production continues. Selections at a certain text production stage change the probabilities of system selections which follow. The dynamic reweighting of probabilities of synoptic systems is considered highly significant for the study of text types, i. e., the study of characteristics of various genres and registers. Halliday's chapter discusses how language systems persist and change over time. His view is that to understand these processes, the synoptic and dynamic views on language are necessary. Furthermore, grammars of languages have to be probabilistic in nature and also have to decribe language systems as equiprobable or skewed according to "coding orientations" available for the language user. Danes' chapter sees the main domains of linguistics in language as a system of knowledge, in the domain of language as texts and in that of text processing. He discusses the Prague School functional linguistic view of modelling sentence structure and transitivity roles in grammar and

vi

Preface

semantics. He argues for a semantic classification of predicate structures and proposes a fundamental distinction of predicates into static and dynamic ones, the predicate structures of which will be presented in the form of semantic formulae which are established on a cognitive level. Bateman — Paris' and Teich's chapters introduce the readers to specific computational considerations and developments in systemic linguistics. The Bateman — Paris chapter discusses the application of the notion of "register" to computational text generation. The goal is to generate texts according to the levels and kinds of expertise of the intended readers. The focus is on the implementation of register theory in the experiential and logical functions of language. The chapter introduces new mechanisms which control choices in register and illustrates new progress in research both for computation and linguistics. Teich's chapter discusses the notion of transitivity, and more specifically the treatment of "raising" and "control" verbs and the problems involved with these verbs in the machine translation project of the European Community, EUROTRA-D. Teich illustrates how semantic participant roles, generated from a systemic transitivity network, are used to achieve better translation results, and suggestions are made for a more delicate functional account of raising and control. McGregor addresses a fundamental issue in systemic-functional theory: a rank scale in grammar and the concomitant principle of accountability. The chapter attempts to answer some of the criticisms that have been raised against the notion of rank. McGregor contends that the concept is valid in the domain of experiential metafunction, and furthermore considers it to be universal, although he does not take the ranks themselves or their number to be universal. Part II chapters discuss issues which are partly theoretical and partly applied or educational. They are, however, issues which have to be taken into consideration specifically in the analysis of spoken data and therefore the papers in these chapters have been grouped together. Geluykens', Tsui's, Davies', Fine's and Christies' papers all contribute to the analyses of how meanings are exchanged in conversational contexts. Geluykens' chapter extends the previous presentations of Prague School functional linguistics and systemic linguistics on Given — New distinction. It argues that the previous presentations have not given sufficiently operational and testable characterizations. For analyses of conversational data, he introduces the concept of Recoverability which classifies elements in terms of their derivability from previous discourse.

Preface

vii

Davies' chapter focuses on minimal exchanges in English discourse and the kind of predictions that can be made of exchange structures by using a model of propositional attitudes. The chapter suggests the possibility of defining an "ill-formed exchange" and linking minimal units of discourse to grammatical form. Tsui's chapter examines in detail a speech act class of "assessments", that is, acts where speakers express evaluative assessments. Tsui investigates the socio-semantic choices which speakers have to make when making assessments, and demonstrates their lexicogrammatical realizations with authentic conversational data. Fine is interested in comparing speakers of different abilities in social interaction in terms of how they respond to the dynamic character of creating exchanges in conversation. According to him, in such comparisons three factors have to be taken into account: that creation of social reality through exchanges in conversation involves paradigmatic choices, that conversation has its dynamic dimension, and that the model used for analysis is a probabilistic model. Christie focuses on the generic plane of conversation in educational contexts. She discusses first order registers, which function to operationalize and maintain the teaching and learning activity, and second order registers, which realize the activity and information dealt with by the children as they learn. Christie leads us to adopt a Bernsteinian proposition of considering pedagogic language both as instructional and regulative, and further, to view the operation of first order registers as projecting and determining the second order registers. Part III centres around theoretical and applied issues in written texts, mostly in educational contexts. Couture's chapter takes a critical view on the ways functional theories influence the practices of written communication. Couture argues that often in functional theories, writing practices are advocated which result in "regressive language practices". Yet, she considers some of the new developments in systemic-functional theory particularly capable of revealing pertinent social values and thus benefitting the writing practices. Thibault discusses the relationships between grammar, technocracy, and the noun. By using cognitive linguistics and other scientific language examples as his material, he shows how technocratic values are condensed thematically by grammatical encodings in texts and how in education one needs to have access to these thematic condensations to understand and critically evaluate the values projected.

viii

Preface

Martin takes us further into the pedagogical discourse world with his chapter on discourse of science and history in secondary schools. He looks at how the discourse functions of organizing and explaining are achieved in texts. Differences in the use of lexicogrammatical resources of language, e. g., in nominalization, distinguish the role of technicality in science from the role of abstraction in history. Francis — Kramer-Dahl's chapter is concerned with the generic differences in writing, the differences between a clinical report and a clinical story. The former is a dominant genre in reporting case studies in neuropsychology. The latter is represented by Oliver Sacks' story "The man who mistook his wife for a hat". The authors deal interestingly with the issues of intertextuality emerging from the two exemplary texts. The metafunctional analyses of the linguistic realizations in the texts show that the clinical report is written for an expert audience, sharing highly codified knowledge necessary to decode the meanings in the report. In his text, Sacks wanted to present an alternative way of writing clinical reports, a person-oriented way of writing. Consequently, the clinical report becomes the patient's story. Nwogy — Bloor's chapter investigates variations of thematic organization of information in three related types of medical texts — in abstracts accompanying medical reports, research articles proper, and popularized articles. By using Danes' notion of thematic progression, the authors seek to account for how language is organized in texts to express the writer's attitudes towards contextual realities in texts. The variations in the choice and distribution of thematic progression patterns are found to relate to the influence of contextual factors such as audience, medium, and purpose of discourse in each genre. Hoey's chapter directs our attention to the concept of cohesion. He utilizes previous work done on cohesive harmony in systemic linguistics and argues that cohesion is not only to be understood as global lexical chains of lexical items which occur in texts, but also as cohesion of "whole messages", that is, as coherence. He demonstrates that various coherence processes can be perceived at work between any non-adjacent sentences which have a sufficient number of cohesive ties. The coherence processes indicate how new material in a text is understood on the basis of the constant material. The chapter by Parsons focuses on cohesion and coherence in scientific texts written by foreign students, comparing them to similar texts produced by native speakers. The chapter strives for deeper insight into the difficulties in writing such texts and investigates the validity of chain

Preface

ix

interaction in cohesive harmony as a method of measuring cohesion in a particular genre of texts. Statistical analysis confirms the crucial role of central tokens in chains as a direct measure of textual coherence. The chapter by Drury compares summaries of a journal article written by a native speaker and two non-native speakers of English. The texts were produced as authentic assignments for a postgraduate diploma. The systemic-functional analysis is used, firstly, to discuss the linguistic choices made in the summaries and those made in the original text, and, secondly, to discuss why certain summarizing strategies seem to be identified as more successful in terms of structure and language. The analysis provides insights into developing paraphrasing and summarizing skills of nonnative writers of English at university level. Ventola — Mauranen's chapter discusses the linguistic difficulties nonnative professionals at Finnish universities experience when writing their articles in English and the revision system that operates for correcting writers' texts. Most of the lexicogrammatical problems in writers' articles are corrected by native English speakers who function as revisers. Yet many problems which are textual in nature remain in the articles. The chapter argues that it is necessary to draw both writers' as well as revisers' attention to operating principles of various textlinguistic systems to overcome such problems in academic texts produced by Finns. Systemicfunctional linguistics seems to provide convenient tools for textual improvement of texts by writers and revisers. Helsinki

1990

Eija

Ventola

Contents

Part I

Discourse strategies and discourse types Nils Erik Enkvist Text production and dynamic text semantics Jay L. Lemke Towards probabilistic interpretations M.A.K. Halliday A functional model of the system of sentence structures Frantisek Danes Constraining the deployment of lexicogrammatical resources during text generation: Towards a computational instantiation of register theory John A. Bateman — Cecile L. Paris A treatment of raising and control in systemic grammar Elke Teich The concept of rank in systemic linguistics William McGregor

Part II

Information flow in English conversation: A new approach to the Given — New distinction Ronald Geluykens

xii

Contents

Minimal exchanges in English discourse Eirian

169

Davies

The interpenetration of language as code and language as behavior: A description of evaluative statements

193

Amy Β. M. Tsui

The static and dynamic choices of responding: Toward the process of building social reality by the developmentally disordered Jonathan

Fine

First- and second-order registers in education Frances

213

235

Christie

Part III

Functional theory, scientism, and altruism: A critique of functional linguistics and its applications to writing Barbara

259

Couture

Grammar, technocracy, and the noun: Technocratic values and cognitive linguistics

281

Paul J. Thibault

Nominalization in science and humanities: Distilling knowledge and scaffolding text J. R.

307

Martin

From clinical report to clinical story: Two ways of writing about a medical case Gill Francis — Anneliese

Kramer-Dahl

Thematic progression in professional and popular medical texts Kevin Nwogy — Thomas

Hoey

369

Bloor

Another perspective on coherence and cohesive harmony Michael

339

385

Contents Cohesion coherence: Scientific texts Gerald Parsons The use of systemic linguistics to describe student summaries at university level Helen Drury

xiii

415

431

N o n - n a t i v e writing a n d native revising of scientific articles Eija Vent ο la — Anna Mauranen

457

Index

493

Part I

Discourse strategies and discourse types Nils Erik

Enkvist

1. Communication, information, scenario-building Many linguists have been good at hedging the most fundamental questions that can be posed about communication. One of them is, why do people bother to communicate at all? Presumably because they think it would be good for them if the amount of information they shared with others were greater than it is. Note that "information" is here used in a wide sense, including information about the speaker/writer's wishes, moods, and attitudes. In this sense, phatic utterances too are informative. They may not change the receptor's views of the structure of the physical world but they still increase his information about his communication partner's feelings and attitudes. What, then, do people do when they communicate, that is, when they want to increase the pool of information they share with their communication partners? They can only do this by eliminating paradigmatic alternatives within a system. Information is certainty as opposed to uncertainty, and certainty is achieved through an exclusion of those paradigmatic alternatives that do not hold. If an element has a probability of 1 (or 100 per cent), it has no paradigmatic alternatives and no information content. Conversely, to have information value, an element must have at least one paradigmatic alternative, and the probability of each of the alternatives must be less than 1 (or less than 100 per cent). If we have a code with a finite number of paradigmatic alternative units and if we know the transitional probabilities from one unit to the next, we can even try to quantify numerically the amount of information carried by each choice: so in numbers and, at least to some extent, in alphabetic and phonemic systems. Thus the choice of one letter out of an alphabet of 26 equiprobable letters eliminates the remaining 25, and in so doing gives more information than the choice of one number from a decimal system of 10 equiprobable numbers which only eliminates the remaining 9. But if the system is open in the sense that we cannot count the number of possible alternatives eliminated by a certain choice, we presumably cannot quantify the information with meaningful, non-arbi-

4

Nils Erik

Enkvist

trary numerical precision. Still, like Giora (1988), we can speak of greater or less probabilities and greater or less information content even in open natural-language systems, for instance, by postulating that greater prototypicality equals less information. We can also say that the information increases, presumably monotonously (in the mathematical sense of the word), at least in a text which is well-formed in the sense of lacking correction and repair, as the text goes on and narrows down an increasingly specific state of affairs or scenario or text world. There is a case for arguing that even false starts and corrections add information to their discourse: they may disclose themes, they may show how a speaker's mind operates. Out of such discoursal stimuli, the receptor is expected to build an increasingly restricted state of affairs or scenario or text world around the emerging text. In successful communication the receptor's scenario is supposed to be isomorphic with the speaker/writer's scenario on relevant points. What is relevant depends on the situation and the purpose of the particular act of communication (cf. Sperber — Wilson 1986). If we are to produce meaning, then, we must have at our disposal a system of contrasting paradigmatic elements, as well as a syntax enabling us to build a structure of the elements chosen (a structure consisting of one single element being the minimum message). Finding a meaning will conversely involve relating the elements actually present in the structure to their paradigmatically contrasting, but absent, elements, and interpreting the relations of the elements present within the structure. And how do people eliminate alternatives in natural-language communication? By arranging the units they have chosen out of paradigmatic systems into linear strings. (I shall here ignore the non-linear relations at the interface between sound and phonology: one and the same sound wave may contain cues for more than one phonological unit.) Here we are up against one of the crucial characteristics of natural languages. Human beings cannot escape time, and in their natural languages they cannot escape linearity which is a reflection of time. Just as we are born in time, we live in time, we communicate in time, and we die when our time has come. So, one crucial problem caused by the human condition, by the inescapable linearity of time, is that the hierarchic structures of language — witness the use of tree diagrams in syntax — must be translated into linear strings of symbols. But the linearity of language can also be turned to an advantage when it can be used for iconic purposes. We can add pregnancy to an utterance by making it temporally isomorphic with its

Discourse strategies

and discourse types

5

scenario: we can say Jack and Jill got married and had a baby, as well as Jack and Jill had a baby and got married, and our receptors will at once understand that the events took place in the order in which they were mentioned. In such instances iconicity is an economy device saving space in the discourse, and presumably effort in its processing, both for the sender and for the receiver. If we go in for a non-iconic arrangement we must elaborate the utterance by adding marking signals: we must for instance say, Jack and Jill only had their baby after they got married. In English, certain temporal conjunctions and tenses — remember Jespersen's terminology: before-past, past, after-past, present, before-future, future, after-future (for example, in Jespersen 1933) — serve as such markers of non-iconicity. Such observations seem simple enough. Still they raise a host of interesting problems. One I have already touched on has to do with the mechanisms languages offer to enable us to depart from iconicity when we feel we should. Another has to do with the different degree of iconic freedom allowed by the syntax of different languages. As long as a language relies on word order to express syntactic relations such as subject-of and object-of, we might expect it to yield less scope for the marking of iconicities by word order alone: it must then develop other devices to mark iconicities and information structures which would otherwise violate basic syntactic patterns. Translators and language teachers are among those who must worry about such well-known typological differences. I should perhaps at this point enter a caveat and say that like so many other linguistic terms, "iconicity" too is subject to many interpretations. I have spoken about experiential iconicity (Enkvist 1981) in a narrow sense to indicate an ordering within a text which reflects experiences of the physical world. More recently, "iconicity" has become a fashionable term. It has also been used — and perhaps even overused — for other principles: old information first, crucial information before less important information (Enkvist 1989), socially prestigious referents before less prestigious referents, and other ordering principles. What all these iconic phenomena, in the wide sense of the term, have in common is that they turn the order directly recoverable from the text into an isomorphic representation of order in the world, whether physical, communicative, or social (cf. Haiman 1985).

6

Nils Erik

Enkvist

2. Meaning and interpretability One interesting way of looking into the birth of meaning, or "semogenesis" to use a felicitous term of Halliday's, is to analyze texts that toe the borderline of interpretability. Such texts — and by all means call them "deviant" if you think you know what they deviate from! — are common in, say, certain types of poetry, advertising, and indeed some types of impromptu dialogue. To avoid the small corpus of standard examples already worn out by students of English — Lewis Carroll, Hopkins, e. e. cummings, and so forth — I shall present to you the final movement of the "Ursonate" by Kurt Schwitters, the Dadaist. In Hajny - Wirbelauer (1983) it goes as in (1): (1)

Zätt üpsiilon iks (bewegt) Wee fau Uu Tee äss ärr kuu Pee Oo änn ämm Ell kaa Ii haa Gee äff Ee dee zee beee? Zätt üpsiilon iks (bewegter) Wee fau Uu Tee äss ärr kuu Pee Oo änn ämm Ell kaa Ii haa Gee äff Ee dee zee beee? Zätt üpsiilon iks (einfach) Wee fau Uu Tee äss ärr kuu Pee Oo änn ämm Ell kaa Ii haa Gee äff Ee dee zee bee Aaaaa. Zätt üpsiilon iks (sehr bewegt) Wee fau Uu Tee äss ärr kuu Pee Ο ο änn ämm Ell kaa Ii haa Gee äff Eeee dee zee beeee? (schmerzlich)

If meaning presupposes paradigmatic contrast, contrast is what the reader must look for. Here the conspicuous contrasts lie in the reading instruc-

Discourse strategies

and discourse types

7

tions (bewegt, bewegter, einfach, sehr bewegt), in the appearance of Aaaaa with a falling tone in the third stanza as opposed to the rising stanzafinal beee of the first two stanzas and the long, rising, painful {schmerzlich) beeee of the fourth. Obviously the reader or hearer of stanzas one and two is to wonder, not without emotion (bewegt, bewegter), what happened to the expected A which is so strikingly missing from the reversed alphabet. In stanza three the missing A is found, but in stanza four, painfully, lost again. The theme of the poem is thus an emotionally loaded quest, the joy of finding a looked-for object, and the pain of losing it again, finally and hopelessly. The looked-for, found, and lost A invites interpretation as a universal symbol for all cherished objects, perhaps better than a word with restricted meaning would. Schwitters's use of mere letters of the alphabet, instead of semantically more complex natural-language words and structures, thus brings out this semogenetic contrast in a particularly bare, obvious, and elemental way. There are other meanings that can be teased out of the text — the pun between Ur- in Ursonate and Uhr symbolizing time, the pattern with variations with its parallels with a sonata — but its main thrust obviously rises out of the contrasts between the stanzas. I must resist the temptation to go on quoting more examples showing how semogenesis depends on choices from among paradigmatic alternatives in a system. One observation is in order all the same. All semogenesis goes back to the relentless hunt for meanings that is characteristic of man. Whenever we hear or see a structure that might be a message we try to decode it as best we can, with more or less success depending on the nature of the structure and on the knowledge of languages and of the world that we bring to its decoding. Even texts we recognize as odd, weird, and faulty compel us to such processes of interpretation. Indeed, if we gave up whenever we are faced with a structure which does not completely tally with our preconceptions of a linguistically, semantically, and pragmatically perfect message, society would collapse. We could not understand speakers of other varieties of language. Peculiar voices and speech defects would leave us stymied. We could not manage figurative language and metaphor, we could not empathize with people whose world and behavior differ from our own. It is interesting to speculate a bit further and suggest that different individuals with different temperaments are willing to go through different quantities of effort to decode and decipher messages they find odd or weird. Addicts to modern "deviant" poetry, and perhaps literature-lovers in general, obviously gain pleasure from their mental workouts, their

8

Nils Erik

Enkvist

sparring with the text. Their semantically lazier, more lethargic and complacent fellows are apt to dismiss such texts, even gruffly, as arrant nonsense and an insult to their intelligence. How we can stimulate people to develop their semantic alertness is a moot question for all educators.

3. Linearity and iconicity These observations about discourse as a tool towards increases in shared information and about setting up shared scenarios are by now commonplace, as are the basic principles of iconicity. Let me all the same go on adding some more flesh to the bones of theory and once again cite a basic example showing how a description of spatial phenomena can be mapped into a linear and thus temporal presentation. Let us see in example (2) how the author of the Blue Guide to Greece takes the tourist through Room III of the Byzantine Museum in Athens. (2)

From this aisle opens Room III, devoted entirely to Sculpture of the main Byzantine period (9 — 15C). Against the left (w.) wall are three marble ikons of the Virgin: in the centre, 148, Virgin as fountain of life, a stiff and formal creation of the IOC or 11C; to either side, 147. Virgin 'Hodegetria' (13 — 14C), and 149, Virgin Or ans' (11 — 12C), each surmounted by a marble arch (152, 154) with representations of the Descent into Hell (from 13C FrancoByzantine tombs). Round the lower part of the walls are relief slabs of the 9 — 12C (from iconostaces), showing typical Byzantine motifs influenced strongly by the Orient: 159 Lion devouring a gazelle. Above those on the N. wall are reliefs with mythological subjects, probably used in secular decoration: 176. Hercules and the Erymanthian boar; 177. Gerene, queen of the Pygmies; 178. Centaur playing a lyre. In the centre of the E. wall, 150. Marble plaque with a painting on wax of three apostles, from Moni Vlatadon (Salonika). This is framed in (155) a great arch, surmounted by (251) a slab bearing scenes of the Nativity, both of which formed part of the entrance of a Franco-Byzantine church. Flanking them are two capitals, one (217) bearing monograms referring to Irene, the Athenian empress of Constantinople (A.D. 797 — 802). The S. wall is devoted to reliefs of the Frankish period: 250. St. John the Baptist, from Zante; heads of Venetian doges from Corfu.

Discourse strategies

and discourse types

9

A tourist guide book could in principle be written with many strategies: a chronological strategy arranging sights by age; a biographical strategy arranging the sights by the persons they have associations with; an architectural strategy grouping sights by building styles; and so forth. Here the writer wanted to take his readers on a guided tour, and his strategy might be defined as "stop-look-see": first the tourist should know where to go and stop, and next where to look. Only then can he be told what he is in fact seeing. This stop-look-see strategy is iconically mirrored in the syntax, most conspicuously through fronting of locative adverbials: the pattern is in location χ is y rather than y is in location x. Sometimes the desired information flow is achieved through passivization (this is framed in a great arch...) and sometimes by nominalization of the locative and its use as a subject through the choice of a suitable verb (here, the s. wall is devoted to reliefs... instead of on the s. wall there are reliefs...; other examples of types of nominalizations common in guidebooks are this room has witnessed many important events; the ceiling depicts the battle of Lepanto; etc.). In addition to iconic marking of the stop-look-see strategy, the fronting of adverbials of place serves yet another important and related purpose, namely that of signalling the macrostructural sectioning of the text into text units. (A terminological digression: I have found it expedient to reserve "paragraph" for a typographically marked unit, and to use "text unit" for portions of text, whether discrete or overlapping or embedded, that are definable through text-internal criteria. This brings with it various advantages, including the possibility of discussing paragraphing strategies in terms of how they succeed in reflecting text units.) Each fronted locative marks the start of a new text unit, and if the locatives have hyponymic relations, their hyponymies reflect hierarchic ordering of the corresponding text unit, as in the fabricated example (3): (3)

On top of the hill is the X Palace. To the right of the main entrance is the Museum. The first room on the left contains a collection of armour. Facing the door is the armour worn by A in the battle of Y. On the breastplate you can still see the dent caused by an enemy bullet.

In tourist guide books organized on the tour principle, then, the stoplook-see strategy is iconically manifested as a preferential marking of sections through locative fronting, or, usually for elegant variation, nom-

10

Nils Erik

Enkvist

inalization of the locative and fronting it as a subject of an active construction with a suitable verb, or through passivization. These observations seem simple and perhaps self-evident. Still they pose a serious warning to the more myopic breeds of syntacticians. Those who ignore the specific discoursal functions of a sentence within its text and discourse, and count, for instance, fronted locatives indiscriminately, without stopping to consider their job within their text, will miss a crucial observation. The placing of adverbials of setting, such as locatives and temporals, is neither subject to general rules covering the entire language ("homogenized syntax" is a term I owe to Tom Nunnally), nor does it vary freely. It is in fact governed by text strategy. In the stop-look-see guide book, the principle of iconicity overrules the possible syntactic incertia which postulates that the least marked place of adverbials of setting in English is within the verb phrase. In other words, a certain amount of discoursal force will be needed for their fronting and extraction out of their natural habitat. How we wish technically to model such matters — as frontings; as possible tensions between frontings and a less marked canonical slot; and the like — depends on what religion and sect of syntacticians we wish to join. Soini (1987) has shown how this can be done with an x-bar model in which the placing of the adverbial in the xbar hierarchy reflects its range of movement and thus its chances of satisfying text-strategic requirements. But whatever model we choose it loses its explanatory power unless it is sensitive to text strategy and text type. My mentioning "homogenized syntax" prompts a brief aside on "language as a whole", a concept I used to cite from Bloch's (1953) definition of styles as frequency distributions and transitional probabilities "especially as they differ from those of the same features in the language as a whole." Language-as-a-whole may be convenient fiction, but it will always remain a fiction. It is unfalsifiable because natural languages are open systems. They change and grow from one moment to the next. And then the number of possible utterances and sentences is unlimited. Even if you collect ten million sentences you can never exclude the possibility that the ten-million-and-first might have had an interesting structure missing from your corpus. This, of course, is one of the reasons why so many grammarians have operated with self-generated examples. (To avoid needless discussion, I should perhaps at once give my personal view: there are jobs needing corpora, but there are also situations where corpora need supplementing with generated examples, whose acceptability in the relevant context should ideally be examined and controlled.) This is not to

Discourse strategies

and discourse types

11

deny that if our descriptions are deliberately striving for those levels of low delicacy and high abstraction where the differences between text types get blurred or disappear, heterogeneous corpora and homogenized grammars may have practical merit. Still, language-as-a-whole remains an unfalsifiable fiction, except of course when we are dealing with closed corpora, for instance f r o m extinct languages. I return to my main argument. Some moments ago I instantiated iconicity with one particularly striking text type, namely the stop-looksee tourist guide. It would be easy enough to go on citing examples f r o m other types of texts whose operational character favors iconicity. There are, for instance, cookery-book writers w h o m a k e good use of structures such as into a large kettle put x, y, and z. This is obviously short for first take a large kettle and then put it into x, y, and z. In such instances the fronting of the locative is supported by another general principle, namely end weight: listing all the foodstuffs that go into the kettle results in a long and heavy constituent, whose weight prompts putting it last in its sentence. Here, then, two principles, iconicity and end weight, conspire and work together for the same purpose. T h a t cookery-book writers strive for conciseness can be seen not only in their profiting f r o m iconicity but also in the frequent object deletion, and of course in the traditionally formulaic disposition. This, by the way, raises an interesting point. A reader is expected to know the basic conventions of a text type such as the cookery book. When he reads turn or plunge into cold water in a cookery book, he supplies the ellipsed foodstuff f r o m the context. But if he reads the same instructions in a h a n d b o o k of gymnastics he will interpret the imperatives as directed at himself. Different text types thus have different syntactic conventions, and these intertextual differences must be mastered if one is to become a competent receptor. We need intertextual experience as well as a knowledge of the lexis and syntax, the world, and the ways people behave, if we are to cope (cf. Viehweger's (1989) knowledge systems).

4. Decisions, strategies, conflicts, conspiracies So, in a tourist guide written with a stop-look-see strategy, the iconic ordering specifying place before sight motivates departure f r o m the least marked, canonical word order (x is in place y) through locative fronting (in place y is x). And this came a b o u t after a conflict between the syntactic

12

Nils Erik Erik, ν ist

inertia prompting us to use least-marked structures whenever nothing compels us to do otherwise, and the requirements of a specific text strategy. Such a view of text production is obviously dynamic, not only in its general emphasis on production (and its mirrored reverse, comprehension) but also in its view of a strategy as an outcome of conflicts and conspiracies. This dynamic view of discourse as a result of sequences of choices is nicely apparent in systemic grammar. To what extent present-day systemic grammars and their networks are capable of illustrating conflicts and conspiracies presumably depends on what level of accuracy and delicacy we expect of our definitions of the conflicting and conspiring forces. Still, decisions and choices do not arise unless we give priorities to one alternative above its rivals, and the assigning of priorities is a matter of strategy. Many linguists have written about strategies as if the term were in need of no further specification; yet, let me briefly explain what I understand by a strategy. Whenever we are faced with alternative actions, with paradigmatic alternatives, we must make a decision involving a choice. As we have seen, semogenesis too is a result of choice. In many walks of life and situations, choices may be conscious; when we write an essay for instance, we may even make an inventory of possible expressions, dipping into Roget's Thesaurus perhaps. But in rapid speech, the entire hierarchic complexes of our linguistic choices seem to be automatized rather than results of conscious deliberation. Such automatized competence is akin to our seemingly automatic mastery of syntactic structures. They, too, are part of our linguistic competence but which we, or at least, nonlinguists, are incapable of expressing in the form of explicit verbalized or formalized rules. All choices, then, can be regarded as outcomes of decisions, whether automatized or conscious. Decisions are made in a decision space whose dimensions are determined by the various considerations that affect the decision. Every utterance, sentence, and clause has arisen through a large number of hierarchically ordered decisions, each of which takes place in a specific decision space. We might try to model the dimensions determining the decision space by thinking of them as scalar parameters which have a value and a relative importance or weight. In the stop-look-see strategy of the tourist guide, the iconic representation of the stop-looksee process is so important that it outweighs the inertia that motivates the use of least-marked syntactic structures. Least marked, that is, in the senses implying "most frequent and widely spread in a large sample of

Discourse strategies

and discourse types

13

language involving different text types" as well as "accepted by competent informants as least striking or odd when presented in isolation" (Enkvist 1984: 52). Thus in the stop-look-see guide, the iconicity parameter is set at its highest plus value and weighted strongly enough to prompt the syntax parameter to permit locative fronting, which in a homogenized syntax of that non-falsifiable abstraction, the language-as-a-whole, would appear as a relatively infrequent phenomenon. And once we have made one choice, its results and consequences will be reflected in other, subsequent choices in the discourse. Two corollaries follow. The first has to do with views of strategies within such a decision-based model. A strategy can be seen as a goaldetermined weighting-and-setting pattern of decision parameters. The actual choice of expression, of words and syntagms, and their exponents in speech and writing, can then be seen as a hierarchic sequence of tactical solutions which carry out the strategy. (The military analogy should be obvious.) The second corollary states that in the text and discourse, only the consequences of the winning, that is, preferentially weighted parameters and their settings, are directly exposed to ear or eye. The weaker, losing forces have disappeared. A text, I have sometimes suggested to enliven elementary lectures, is like a battlefield after the battle: the winning army is in possession, the losers have fled and the dead and wounded have been carried away. But the position and appearance of the winners and the traces of operations that can be seen on the battlefield may give the expert a chance of reconstructing something of the battle itself. So, too, the surface appearance of a piece of discourse may give clues to the conflicts and conspiracies that must have gone into its production. All linguists have learned the hard way that human communicative behavior is complex enough to resist simplistic modelling, say, by discourse-generating computer programmes, for instance. This is certainly true of more complex text strategies and their underlying conflicts and conspiracies. It is easy enough to rewrite a piece of description, of a cathedral for instance, with different strategies: stop-look-see, biography, chronology, architecture, and fine arts. My views of text strategies may even be pedagogically expedient as explications of such composition exercises. Also a great deal of effort, some of it suggestive and promising, has been spent by students of artificial intelligence and computer linguistics on devising text-type-and-register-sensitive algorithms for text generation (to cite only a few examples, see, Hovy 1988, 1989; Bateman — Paris this volume). In corpus description and predictions within relatively homogeneous materials we may also profit from the now classic methods

14

Nils Erik

Enkvist

of setting up variable rules (for example, Sankoff 1978). But despite such laudable and necessary efforts we are still far from the day when we can feed a set of pre-lexical, or perhaps basically lexicalized, propositions or predications into a computer together with a comprehensive set of strategies for textualization and a syntax, and then have the machine faultlessly churn out a full range of textual variants, or allotexts, that arise through the different strategies associated with different communicative situations. Before we can thus implement text strategies on the computer we must learn to view texts as arising through conflicts and conspiracies. Such an approach will help us to understand the interactions between discourse structure, text types, styles, and registers. Indeed one of the most urgent tasks in linguistics is to learn more about the ways in which discoursal requirements determine the structures of sentences and clauses. Such studies will be important as preliminaries to all studies of language varieties and alternative formulations of messages. They inevitably enter into linguistics under many labels: stylistics, register studies, pragmalinguistics, politeness, relevance, discourse generation, or whatever. These principles can be applied to concrete examples. We saw that in the tourist guide, the iconicity factor was strong enough to overrule the desire for least-marked syntax. This is obviously a characteristic of the stop-look-see tourist guide as a text type, and perhaps of other text types where iconicity is at a premium. As suggested above in more general terms, such an observation immediately evokes two responses. First, we should try to set up links between text typology and strategic parameter weighting and try to characterize text types in terms of their characteristic weighting patterns. Secondly, we should see what different types of texts share a weight on certain parameters. Another instance: many metrically regular poems achieve their regularity by allowing poetic licence in their word-order patterns. Crabbe's poems yield fine examples of this. In such instances we can say that the metrical parameter, set at the relevant metre, has been given a weight sufficient to override the conflicting claims of canonical, least-marked word order. The stop-look-see tourist guide is one variant of an operating instruction: it tells people how they should behave. The cookery-book is another instance of an operational text with a potential for iconicity. Indeed we might expect operating instructions in general to set a premium on experiential iconicity, on presenting operations in the order in which they are supposed to take place. In fact, we might surmise that one of the characteristics of a successful operational text is maximal iconicity because presenting matters in the text in the same order as they occur in

Discourse strategies

and discourse types

15

time or space makes for greater ease of processing. I presume all of us have gone through frustration, anger, and despair when we have struggled with operational instructions that ignore iconicity. And to make things worse, writers of operating instructions often have false assumptions about what I already know and do not know (which leads to inappropriate information structures in terms of given and new) and an obvious optimism about my, the receptor's, intelligence and manual dexterity. This I have experienced with most diverse instructions from how to assemble a chair to more elaborate computer manuals. Another classic iconicity problem that I already touched upon is the ordering of events in narrative. The basic ordering principle is presumably chronology because events that take place in time are by nature chronologically ordered: in Russian formalist terms, the "fabula" and the "sjuzet" are temporally isomorphic. But narratives often contain elements other than descriptions of actions and events, metatext and descriptions for instance, which may intervene into, and in a sense interfere with, the strict chronological accounts. And then there may be unities in storytelling which the narrator feels are more important than a strictly chronological order (for a description of temporal strategies in narrative, see for example, Genette 1980). Sometimes the ordering has been a problem: to cite just one instance, Scott Fitzgerald's Tender Is The Night sold badly in its first edition, and Fitzgerald then revised it, making the story advance in a more strictly chronological order. As far as the tactical realization of chronological ordering is concerned, languages have many ways of marking shifts where what comes first in the story in fact came later in the world of events. Pragmatic clues also help: we know that people must be born before they go to school, we know we must pour out a glass of wine before we can drink it (unless we drink out of the bottle!), we know we must get up in the morning and go to the office before we seat ourselves behind our desks (see, for example, Enkvist 1985). How much of such detail must be explicitly stated and how much can be left to the receptor's inferencing depends on the storyteller's strategy and on his estimate of the capacities of his hearers or readers. In this perspective the classic model of communication of the behaviorists, Speaker —* Text —• Hearer, was a crude oversimplification. In normal communication, a speaker or writer tries to figure out what his receptor already knows and does not know: figuratively speaking, he explores his receptor with what one might liken to a radar beam, Speaker —> Hearer —> Speaker, and only then does he actually formulate his text. This is so because every message has to signal the distinction

16

Nils Erik Enkvist

between old and new information, partly because of pragmatic reasons, such as those embodied in the Gricean principles, and partly because the structure of the sentence and clause perforce also reflect such distinctions. Faulty receptor adjustment leads to breakdowns in communication and is one reason why some people cannot match their skills in face-to-face communication when they must sit down and write. Such, then, are the data and general principles that we must cope with. In a nutshell summary: the syntactic arrangement of the sentence is not independent but subservient to a text strategy. The discourse is the father of the sentence, not vice versa, and each sentence must serve its discourse and text by confirming to its general strategic principles and thus linking up, deictically and stylistically, with the discoursal situation, and textually with its envelopes within the text. And the text strategy in turn can be seen as the resultant of a set of forces some of which conspire and some of which conflict. In the case of conflicts, the stronger party wins. As to the receptor, to interpret discourse properly, he too must be familiar with the conventional strategies of the relevant text type. He must understand that turn! in a pancake recipe means something different from turn! in a text on physical exercise.

5. Modeling The next question for the linguist is, how is he to model strategies of discourse and text, and their effects on the text? As terms such as "model" and "theory" have been used differently by different linguists, a brief terminological note may be in order. In this chapter I shall use the term "model" for an explicit presentation of invariances among a selected set of observed data. To be useful beyond these data, a model should permit manipulations with predictive power. A model must be explicit in explaining precisely what goes into it. A model presents a selected set of data: we can build models of ships for decorative purposes and not worry about what exists inside the hull; we can build models for the naval architect's tank and not worry about the rigging and decorative values; or we can build models for navigation schools and omit everything except the bridge and its instruments. Finally, a model must have a purpose, and its purpose governs the selection of features for modeling. Once we agree on such an approach to the definition of models, a theory is definable as a model of the second degree. A theory becomes a model of models. It tells us what principles can go into the selection of

Discourse strategies

and discourse types

17

features for modeling, and what concepts we may use in building our models. We thus need a theory as a kind of recipe or instruction for our model-building. There can be no models without a theory. For those who have followed the development of systemic linguistics, the time may well seem ripe for a more thorough theoretical assessment of its general principles, in comparison with the many other rival theories that have been set up in the past thirty-five years or so. Such an assessment is, of course, beyond the scope of a single chapter and perhaps even beyond the scope of a one-volume book, so what follows should be taken merely as a reminder of some of the points we ought to consider. One such question is the relative merit of a single-system theory versus a polysystemic theory, that is, of a theory which tries to model all of a given language as one single system, and a theory which regards language as a system of systems, each of which forms an entity of its own but is related to, and coupled with, a set of other systems, either in input-output relations or in relations of congruence. Here "input-output relations" should be read as shorthand for a relation where the output of one system is the input of another, and "congruence relations" for relations where one system must know, and adjust to, what goes on in another, without using the entire output of that other system as its own input. An example of a monosystemic theory is that underlying Chomsky's (1957) Syntactic structures. There the rules were organized into one single ordered string with the result that traditional levels of description, such as independent phonemics, disappeared by being scattered in different places in the string of rules. On the contrary, Michael Halliday and his followers are among those who have always insisted on polysystemic descriptions: each system network presumably illustrates the workings of one subsystem. We might note in passing that the Government-and-Binding model, which I understand is still in use, has a polysystemic look. It contains interlinked subsystems labelled as x-bar theory, theta theory, case theory, government theory, binding theory, predication theory, bounding theory, and control theory. Insofar as such "theories" are subsystems, or what I would call models of specific mechanisms of language covered by one single theory, even the Chomskyan world has been going the systemic way. In fact, polysystemic grammars have obvious advantages over monosystemic ones. A polysystemic grammar is open to the addition of new subsystems as needs arise and our knowledge expands. Also, in a polysystemic grammar, deficiencies and breakdowns can often be set right by changes and adjustments within one subsystem, or at worst within those subsystems with which the subsystem concerned has input-output and

18

Nils Erik

Enkvist

congruence relations. On the contrary, falsifications of principles and rules in monosystemic grammars readily compel more extensive revision and even the recantation of principles once regarded as fundamental. Plentiful examples of such revisions and extensions can be found in the history of generative-transformational grammar from 1957 to the present day. Another basic question concerns the modeling of strategies. As soon as we admit that discourses and texts arise, not through an echoing of fixed, predetermined patterns but through complex sets of choices between possible alternatives, we should learn to model sets of purposeful choices governed by strategies. Strategies also become necessary if we wish to model different types of text and discourse, and with them their different styles, because their differences can be regarded as results of different strategies. Further, as I have argued at some length, strategies in turn must be modeled in terms of conspiracies and conflicts between different principles, which can be visualized in terms of parameters that have a value and a weight. To repeat an obvious example: a metrically regular poem which uses unusual word-order patterns can be seen as a result of a strategy where the metrical parameter, set at the value of the relevant meter, was given a weight greater than that of the least-marked word order. At the present stage of our knowledge, discourse-and-textstrategic parameters can, only in a limited and perhaps trivial way, be modeled with, say, numerical values and weights that could be handled by, say, a computer. One obstacle is our lack of automated syntax. But as a programmatic suggestion this parametric view seems worth a try. I am, then, visualizing sets of systemic networks whose choices are weighted by situation-and-discourse-sensitive factors, and in ways far more comprehensive than the attempts at introducing weighted rules into generative grammars. It is the job of theory to make inventories of potentially relevant factors, of the splitting up of language into subsystems, and of the ways in which parameters are given their values and weights. This in turn leads us on to the classic question as to what systems of language are subject to strict rules and what systems are subject to tendencies or even free variation, that is, variation that cannot be explained with available parameters. We all know that certain types of language, impromptu dialogue and speech, poetry, and advertising for instance, may succeed in communication though they sin against patterns, or rules, which syntacticians are fond of calling basic and obligatory. Once we look at what people really do in authentic impromptu speech, as well as in some kinds of poetry and advertising, the line between the

Discourse

strategies

and discourse

types

19

possible and the impossible, between g r a m m a r and u n g r a m m a r gets writ in water. To escape these problems, linguists have long been preoccupied with langue and "competence" rather than parole or "performance". This gave them an excuse to purge their data of the features that embarrassed their preconceptions. As a curious paradox, this was standard practice even a m o n g those structuralists who paid vigorous lip-service to the primacy of speech over writing. Today we know that mechanisms such as pausing, false starts, structure shifts, anacolutha, corrections and repair, repetition, prolongation, verbalized pauses and fillers, hedges, and the like are indeed useful and in some situations essential elements of impromptu speech. They improve the chances of a speaker and perhaps also his hearer to process language in real time. They make it possible to repent and correct what went wrong, and to adjust to rapidly developing shifts and quick changes in the situation and in the communication partner's discourse. M a n y poems and advertisements too are effective precisely because they break against patterns rather than follow them (though this goes to prove that patterns, in the sense of " c o m m o n behavior in a certain type of situation", actually exist; admitting the existence of such patterns is, however, different from insisting that the standards of socially approved, literate writings form a langue fundamental to all communication). Anyway, in the past few decades the ancient debate between analogists and anomalists has surfaced in new and topical ways, and in the linguistic horoscope the anomalists appear in ascendance. For we are increasingly realizing that communicative success is a result, not of the use of faultless structures but rather of successful processing. And successful processing can be triggered off by structures different from those traditionally canonized by linguists as langue. Such observations must, however, not be used to excuse sloppy usage and rude linguistic manners. Every message also carries what Tannen (1986) calls a "metamessage". So, a breach of expectations, for instance in the form of "faulty g r a m m a r " or " p o o r style" in a message aimed at a grammar-and-style-conscious receptor, may stigmatize a writer as uneducated and incapable of handling a j o b demanding literacy. Peas may be just as nourishing if you eat them with your knife. But if you aspire to earn your daily bread as a diplomat, you do wisely in using a fork. Similarly, manners matter, not only at table but also in communication. But we should remember that what is good manners in one kind of communication may spell disaster in another.

20

Nils Erik

Enkvist

6. Decisions, relative success, and relative failure This brings up my final point in way of a summary. If successful communication depends on processing and not only on the structures that trigger off processes, we shall need models and theories permitting descriptions, not only of structures but also of the processes through which structures are chosen, formed, and interpreted. This also applies to structures which prove to carry meaning though they fail to conform to canonized patterns of well-formedness: despite their form they have triggered off successful processes of interpretation. Linguistic processes in turn must be modeled in terms of choices between alternatives, and these choices arise out of decisions. Linguists must thus learn to develop and handle decision theories. And decision theories adapted to linguistics will be related to, but not identical with, those developed by students of business economics or medicine. In business, the success or failure of a decision can be measured in dollars, pounds, or rubles, and in medicine in mortality rates. In authentic communication, success criteria are more complicated and successes harder to measure. Granted, Stolt — Trost (1976) once measured the relative success of marriage advertisements in the Hamburg magazine Die Zeit during the second quarter of 1973 in terms of numbers of replies (which varied from zero to 196). But such neat quantification of perlocutionary effects is rarely possible. In literary communication, the critics and readers are the arbiters; in foreign-language examinations, teachers have certain standards of correctness which they apply to their pupils' texts; and each successful leader, whether in business, the military, or in academe, has presumably developed a communicative style which succeeds in his particular circumstances. In evaluating the relative communicative success of a piece of discourse, which we must do to weight our decisions in discourse production in the best possible way, we are moving from the world of absolute rules into the world of relative success and aesthetic merit. We are moving from linguistics in the narrow sense into rhetoric. This, to me, is just fine. In emphasizing not only syntax but also pragmatics we are moving from codes to usage, from structural abstraction back into real life. And there are some hopeful signs that scholars of the literary persuasion are moving in the direction of linguistics as well. Let us hope our paths will meet.

Discourse strategies and discourse types

21

References Allen, Sture (ed.) 1989 Possible worlds in humanities, arts and sciences. Proceedings of Nobel Symposium 65. Berlin/New York: de Gruyter. Bateman, John A. — Cecile L. Paris this volume "Constraining the development of lexicogrammatical resources during text generation: Towards a computational instantiation of register theory." Bloch, Bernard 1953 "Linguistic structure and linguistic analysis", in: Archibald A. Hill (ed.), 40-44. Chomsky, Noam 1957 Syntactic structures The Hague: Mouton. Enkvist, Nils Erik 1981 "Experiential iconicism in text strategy", Text 1: 97 — 111. 1984 "Contrastive linguistics and text linguistics", in: Jacek Fisiak (cd.), 45 — 68. 1985 "Coherence and inference", in: Ursula Pieper — Gerhard Stickel (eds.), 233-248. 1989 "Connexity, interpretability, universes of discourse, and text worlds", in: Sturc Allen (ed.), 1 6 2 - 1 8 6 . Fisiak, Jacek (ed.) 1984 Contrastive linguistics. Prospects and problems. Berlin/New York/Amsterdam: Mouton. Genette, Gerard 1980 Narrative discourse. Oxford: Basil Blackwell. Giora, Rachel 1988 "On the informativeness requirement". Journal of Pragmatics 12, 5/6: 547-565. Haiman, John (ed.) 1985 Iconicity in syntax. Amsterdam/Philadelphia: Benjamins. Hajny, Peter F. — Horst Wirbelauer (eds.) 1983 Lesekurs Deutsch für Anfänger. Berlin: Langenscheidt. Heydrich, Wolfgang — Fritz Neubauer — Jänos S. Petöfi — Emel Sözer (eds.) 1989 Connexity and coherence. Analysis of text and discourse. Berlin/New York: de Gruyter. Hill, Archibald A. (ed.) 1953 Report on the Fourth Annual Round Table Meeting on Linguistics and Language Teaching. Washington, DC: Georgetown University. Hovy, Eduard Η. 1988 "Planning coherent multisentential text." Information Science Institute Reprint Series 88 — 208, (Reprinted from Proceedings of the 26th Meeting of the ACL.) University of Southern California. 1989 "Pragmatics and natural language generation". Information Science Institute Reprint Series 8 9 - 2 3 3 , (Reprinted from AI Journal, Fall 1989.) University of Southern California. Jespersen, Otto 1933 Essentials of English grammar. London: George Allen and Unwin Ltd.

22

Nils Erik

Enkvist

Lindblad, Ishrat — Magnus Ljung (eds.) 1987 Proceedings from the Third Nordic Conference in English Studies. Stockholm: Stockholm Studies in English. Pieper, Ursula — Gerhard Stickel (eds.) 1985 Studia linguistica diachronica et synchronica ... Werner Winter sexagenario ... Berlin/New York/Amsterdam: de Gruyter. Sankoff, David (ed.) 1978 Linguistic variation: Models and methods. New York: Academic Press. Soini, Annamari 1987 "An X-bar classification of English adverbials", in: Ishrat Lindblat — Magnus Ljung (eds.), 3 1 9 - 3 3 4 . Sperber, Dan — Deirdre Wilson 1986 Relevance. Communication and cognition. Oxford: Basil Blackwell. Stolt, Birgit — Jan Trost 1976 "Hier bin ich — IVY; bist Du?" Heiratsanzeigen und ihr Echo, analysiert aus sprachlicher und stilistischer Sicht. Kronberg/Ts.: Scriptor Verlag. Tannen, Deborah 1986 That's not what I meant. New York: William Morrow and Company. Vichweger, Dieter 1989 "Coherence — interaction of modules", in: Wolfgang Heydrich — Fritz Neubauer — Jänos S. Petöfi — Emel Sözer (eds.), 256 — 274.

Text production and dynamic text semantics Jay L. Lemke

1. Texture and semantic choice When we imagine writing on a blank sheet of paper, it seems to us that all the semantic choices of the language system are initially available to us. With no further information about the context of situation or context of culture, perhaps they are. This is the situation described by a systemic account of the lexicogrammar, beginning from clause rank. We enter a "master system" such as CLAUSE and all the possible clauses of English, all the possible meanings that can be meant with clause-complexes in English, are available for selection. We know that the paradigmatic systems of selection options that provide a systemic account of the lexicogrammar are ideally written without regard to context. If there are inherent weightings in the probabilities of some particular selections relative to others, then those weightings represent in some sense an average over all possible contexts, or a baseline, a default probability, which is then modified and re-weighted when specific contexts are taken into account. These context-sensitive reweightings characterize a register. But suppose we take a slightly different perspective. Imagine that we are about to write not the first clause on our blank sheet, but the second. Or the third. In principle, for each ranking clause, or clause-complex, all possible options are again available to us. But most sequences of clauses which we produce or encounter are not the ones in which each clause is totally independent in meaning of the one(s) which precede(s) it. The sequences of clauses which language has evolved to enable us to produce are by and large texts; that is, sequences of clauses which are to some extent mutually predictive of one another. We know that texts, as opposed to random juxtapositions of clauses, have a quality that may be called "texture" (Halliday — Hasan 1976; Hasan 1985 a), which is fundamentally a semantic property, realized through lexicogrammatical features. A text is, in this sense, a semantic unit (cf. Halliday 1977) in that it is characterizable by a unity of meaning, such that the meanings expressed in each of its clauses have some specific meaning-relations to those expressed in some or all of the others.

24

Jay L. Lemke

Insofar as a second or third clause helps to create a text with the preceding clause(s), not all semantic choices are equally open to us, in the sense in which they may have been in writing a first clause, and correspondingly, not all lexicogrammatical choices are available. More precisely, we can say that there is a re-weighting of selection probabilities that depends, in a first view, on the selections already made in the preceding clause(s). The linkages, the relations of interdependency between clauses or clause-complexes in a text, are not essentially grammatical ones, though they have consequences for lexicogrammatical choices. They are essentially semantic relations. Now we know well that lexicogrammatical selections within clauses and clause-complexes represent potential differences of meaning, and therefore some part of the semantic choices that are made in creating text-as-message (that is, the semantic unit) "pre-select" or condition selections within the lexicogrammatical systems at CLAUSE and lower rank. We can call this lexicogrammatical semantics. But there must be "another part" of semantics: choices which describe the system of options for what kinds of meaning relations will be made between different clauses and clause-complexes of the same text. It is these text-semantic relations which account for the "texture of a text" (and, as I have argued elsewhere, for intertextual semantic ties as well; Lemke 1985). We already know a little about text semantics in this sense. We know that successive clauses and clause-complexes of naturally-occurring texts tend to belong to the same register or "micro-register" (Halliday 1978; Gregory 1985). This is often a necessary, but hardly ever a sufficient condition for them to form a coherent text (Lemke 1985). But the more general semiotic notions that lie behind Halliday's three "metafunctions", and which are at the root of models of register, are still good guides to what sort of relations among the clauses of a text belong in an account of text semantics (Halliday 1978; Lemke 1989 b, in press-a). Their ideational relations can be described by methods such as my own Thematic Analysis (Lemke 1983, 1988 a, 1990 b, in press-b) or Hasan's approaches to lexical cohesion and cohesive harmony (1984 a, 1985 a). Their textual relations, I believe, correspond to those between different functional elements in a Genre structure (cf. Hasan 1984 b, 1985 b) and to relations of the clause-complexing (Halliday 1985: 192 — 251), conjunctive (Martin 1983), and rhetorical structure (Mann — Thompson 1987) types. Finally there are those of the interpersonal type, described in part by an "axiological" analysis of heteroglossia (Lemke 1990 b, in press-a, in press-b), by Martin's work on macropropositions and macro-

Text production

and dynamic text semantics

25

proposals (in press), and by Thibault's (in press) notion of the Global Modal Program of a text. Of course, these three types of text-semantic relations and their corresponding structures and global co-patternings across texts, are intimately interdependent in all real texts. Nearly all the analyses I have cited, as well as much work by many other analysts, combines discussions of elements that are semantically of all three types. This is particularly true in work on genre and ideology. The typology I have just offered is mainly one of convenience, to be used as a guide to remind us of the need to examine relations of all these kinds in our work. With this view of text semantics, we already have a rich literature to draw on in characterizing the kinds of semantic relations between clauses that contribute to the semantic coherence of a text. Different clauses in the same text will likely be cohesive (grammatically and lexically) with one another and exemplify closely related components of one thematic formation (representational consistency). They will make selections such that we will construe them as belonging to functionally related parts of some rhetorical or generic structure (organizational consistency). And they will express a consistency of point-of-view toward addressees and the truth-value and "rightness" of what they affirm and deny (orientational consistency; for all these terms, see Lemke 1989 b, 1990 a: 194 — 206). In all these ways, second clauses and third clauses, and second and third paragraphs, are emphatically not independent, even in their specific lexicogrammatical choices, of the choices made in preceding portions of the same text.

2. Text production and dynamic vs. synoptic perspectives If we do indeed shift our perspective from the independent or "initial clause" view to the interdependent, "medial clause" view, what are the implications for modeling text semantics? First of all, we are moving away from the paradigm of text analysis, which commonly assumes that there is a complete text available for study, and towards the stance taken, for example, in computational linguists' efforts at specifying the processes of text generation. To choose a more theoretically neutral term for the perspective I wish to develop, let me call it simply that of text production. Text production is a process as well as a perspective: it is the process of making text by selecting features (lexicogrammatical and text-semantic) for successive clauses in such a way that the resulting sequences of clauses

26

Jay L. Lemke

form a socially recognizable unit of linguistic action. By this I mean that the resulting text does indeed have a texture: the clauses form a unit with ideational, interpersonal, and textual coherence in the text-semantic sense. They are cohesive, thematically and orientationally consistent, and have a recognizable genre and/or rhetorical structure. It is important to note that the text-production perspective is a dynamic one, while that of text analysis is synoptic (Lemke 1984, 1988 b; Martin 1985; Ventola 1987). A dynamic perspective on meaning is one which enters into the flow of events and asks what is the meaning of an utterance as it occurs and what meaning attaches to how and when it is occurring. Synoptic perspectives, on the other hand, step back outside the flow of events and examine meanings retrospectively, in the full context of a total social event as it has occurred. In the dynamic analytical perspective meanings are taken to be more tentative, more anticipatory — their timeliness and manner more signifying. In the synoptic perspective, the same actions (including utterances) take their text-specific meanings in considerable measure from what follows them in the text, retroactively narrowing and focusing the what-could-be-being-meant of dynamic meaning into the what-evidently-must-have-been-meant of synoptic meaning. The synoptic/dynamic distinction can be applied to the means by which we describe meaning-making as well as to meanings made. A "system" in the lexicogrammar is ordinarily written from the synoptic perspective; that is, it is written from a perspective that stands outside the meaningmaking process, outside text production, and describes options that are "always" there (in principle). We could also introduce into our theory of linguistic meaning-making a notion of dynamic system: one which would describe the options available for a "next" clause, say, at some particular moment of text production. Such a theoretical entity obviously would be simply a tool, not a fundamental construct, since it would vary from clause to clause and from text to text. But it is very important for us to be able to say something about "how" and "why" it varies as it does. If we imagine the description of dynamic systems to be mainly a matter of the dynamic weightings of selection probabilities, then we wish to know how the selections "up to now" condition the probabilities for selections "now". One could imagine studying the process of text production by watching an illuminated wallchart of systems in which the probability of each option at each moment was represented by its brightness. At the beginning of text production, setting aside any knowledge of context of situation

Text production

and dynamic

text semantics

27

or register, illumination would vary across the chart solely to reflect universal inherent weightings within systems. As selections were made in a "pass" through the systems, selected options would brighten and unselected ones darken, but then the entire display would "refresh" itself for the next pass, the next clause. Perhaps the previous selection might be indicated in a special color, but now we would see not the same range of options and weightings as before, but a new set of weightings dependent on the previous selections. As text production proceeded, we would see — at each moment before actual next selections were made — the cumulative influence of past selections on the probability of each next option. To proceed further, we must abandon this idealized fantasy and try to use what we already know of text semantics (cohesion, thematics, genre, rhetorical structure, orientational programs, etc.) to implement the textproduction perspective. Most critically, we need to make use of our theories of the origins of texture, that is, our notions of why people make the semantic units we call texts.

3. Pragmatics, semantics, and social semiotics How can we account for the existence of texts? What is going on that leads to sequences of sayings and wordings that have a unity of meaning? What functions do the text-semantic relations between clauses and clausecomplexes serve? When are they called into use? Our naive notion of this is that language is being used for some social purpose, that there is a socially meaningful action being performed, at least in part through the use of language, and that the way in which it is being enacted leads to the production of a text of a particular type. The text-type is recognizable because of its association with a culturally identifiable type of social action. We insult people, we tell jokes, we perform marriages, we give instructions, we describe events. Many social activity types give rise to one or more characteristic text-types, often called genres. The activities may be complex, consisting of sequences of particular functional acts, and so text-types similarly will usually have a functional structure of meaning elements, with a variety of possible relations among them. This model is useful for many purposes, but it needs some important refinements. First of all, the naive model is too "top-down"; it describes social actions as determining text-types, when we know that language

28

Jay L. Lemke

events and texts co-determine social action reciprocally and dialectically. We know that an utterance (a message-text) is itself an action and can change specific aspects of the context of situation in which it occurs. In this way the situation and the next-action or next-activity also depends dynamically on what has been said and otherwise done before, including what is being said and done now. It is also probably true that linguistic text-types, being evolved, meta-stable social formations in their own right, are not simply responsive to fulfilling the actional function of the moment; they also have a certain "inertia" of their own, such that, once begun, they may carry forward even against the actional needs of the moment, changing the situation and the activity by their performance as a formation and not simply by the effect of particular sayings said in the course of that performance. It is not only context of situation, but also context of culture that is dynamically implicated in a semantic view of text production. It is the context of culture that provides the larger scale social formations we call genres and activity-types. If the genre or the rhetorical formation that has been begun in the initial clauses of a text calls for a reversal or peripety, for a disjunction or contradiction, then from one clause to the next we may see probability weightings for selections in the systems M O O D or A T T I T U D I N A L LEXIS suddenly reverse. The context of culture may provide options for which genres are chosen or how they are implemented, and further condition those options on features of the context of situation. Most theories of pragmatics too naively assume that social acts, including "speech acts", may be defined independently of the semiotic resource systems through which "they" are enacted. It reifies them and then looks for their "expression", rather than defining them in such a way that they are themselves contingent constructions resulting from the meaning-making process. Our folk-theories of the relations between words and deeds, and between intentions and actions, may be interesting as cultural data, but they cannot substitute for more sophisticated accounts of meaning-making practices. Any adequate account of social meaning-making, of human semiosis, must include categories of meaning units (act-types, activity-types, etc.) more abstract than either conventional pragmatics' categories for acts and "intentions" or our linguistic categories of semantic features and relations. We need one unified account of meaning-making practices within which both linguistic text production and other forms of semiosis can be modeled and interrelated. This is true not just because so much

Text production

and dynamic text semantics

29

of human communication and action uses both linguistic and other semiotic resource systems conjointly, but because linguistic semantics itself, in order to avoid the endless regress of mere formalism, must have a functional basis. Semantics must be constructed so as to mediate between the other strata of language and a social semiotic in which socially functional meanings are definable. What does this mean for our account of text production? Essentially, it means that it is social function which must ultimately define the dynamic re-weightings of selection probabilities in lexicogrammar, not simply in the realization of a set of semantic features for each separate clause (though obviously it plays a critical role in that), but most crucially in order to insure that a sequence of clauses "means" as a text.

4. Semiotic formations and dynamic models of text production It is a long way from the social semiotics of meaningful acts and events, through the mediating semantic options for both lexicogrammatical and text-semantic features and relations, to actual selections in lexicogrammatical systems and their realization in clauses and lower rank structures as strings of particular lexical items. That long pathway may be needed for a scientifically satisfactory account of either text production or the meanings of a finished text (and perhaps for the most universal computational text-generation or text-interpretation programs), but otherwise it seems unlikely either that such a comprehensive analysis would be attempted very often, or that people actually do have recourse to such a hypercomplex, multilevel scheme for making or interpreting texts. Most texts are not highly original, whether they are in fact unique or not. They do not need to be original because they need only meet the needs of some local variation on a standardized, well known, and familiar culturally recognizable situation. A complete paradigmatic account of the semiotics of human action is no doubt possible (whether practical or not), but any given human community enacts only a very restricted subset of all the possible action-texts its actional semiotic resources allow. In the same way, of course, most grammatically possible clauses never get said, and most possible sequences of clauses do not form recognizable texts in a given community. There are, in absolute terms, relatively few kinds of social activity and relatively few text-types, genres, thematic formations, social relationships,

30

Jay L. Lemke

issue viewpoints, etc., in a particular community. Mostly we speak in cliches, in textual "boilerplate", in pre-compiled (as the computationalists might say) formations — not word-for-word, by any means (except, for example, in highly ritualized events), but by and large still within narrowly defined limits of selection and co-selection. This habit is not just a function of register or situational specificity; it applies as well to the structural organization of our texts and even to lexical realizations. It is text patterns, text formations, and not just registers or genres that we learn to speak. I have discussed elsewhere the sense in which text formations (or discourse formations, cf. Bakhtin's "speech genres") as instances of the general notion of semiotic formations are complementary to paradigmatic, analytical notions like register or, more precisely, register-specific linguistic systems, which are correspondingly instances of semiotic resource systems (Lemke 1985, 1988 a, 1988 b, 1989 b, in press-b). Characteristically, "formations" tell us what is typically said and done in a community, while "systems" tell us what can be meant with the resources it in fact deploys only in more limited ways. Formations are less general, but more economical modes of description for certain purposes. Formations provide a short-cut on the way from the semiotics of social action to the wording of texts, but they also provide an alternative way of asking our basic question about text production: How is it that the successive semantic selections of the text production processes, each modeled by the changing selection probabilities of dynamic systems, results in a text of a recognizable formation? Whatever information a formation summarizes about the text-semantic relations appropriate to a given text-type, that same information must somehow enter into the shifting selection probabilities of text production. Fortunately, a representation of that information which is wholly compatible with a paradigmatic model is available. Quite a long time ago, before making use of the notion of formations, I proposed that the same kind of culture-specific information could be represented by extending the meta-redundancy notion of Bateson (1972) (see also Lemke 1984: 33 — 41). In essence, meta-redundancy relations describe a hierarchy of conditional probabilities, and this formalism is well suited to describing a social semiotic model of contextual meaning. Social semiotics says that an event (or identifiable feature) A is meaningful insofar as its co-occurrence with other events (or features) B, in a given context C, is statistically predictable. Another way of saying this is that, in context C, A and Β go together more often (or less often) than

Text production

and dynamic

text semantics

31

mere chance would predict. Moreover, since A and Β are paradigmatic selections, the same may be true of other combinations of alternatives to A and to B, and usually is. This is equivalent to saying that in the context C there is a pattern of co-occurrences: not every possible combination of selections is equally likely or equally frequent. The pattern of co-occurrences of the various selections of each sort, then, is part of what "defines" (or constructs) the context C. In another context, there would be a different (or no) co-patterning of the A options and the Β options. Indeed, those other patterns in part define other, alternative contexts to C. The net logic of all this is that contexts are now seen to be simply higher-order features (or events), defined in their turn by patterns of cooccurrence with the patterns of co-occurrence a m o n g the A and Β options. Substituting the information theory term " r e d u n d a n c y " for co-occurrence, a contextualization theory of meaning can be appropriately represented by specifying a hierarchy of redundancies of redundancies (metaredundancies, Lemke 1984: 33 — 41). This model has a pleasing absence of reification of entities at any level — an essential feature of social semiotics. Mathematically, the co-patternings are described as sets of conditional probabilities, and a m o u n t , in fact, precisely to the re-weightings of dynamic systems needed for text production to produce texts of recognizable social formations. The details of this argument are too complex to summarize briefly in this context, but the outcome is that our idealized fantasy of a dynamic semantics of text production can in principle be made as realistic as we wish by including information about context of situation and context of culture. In fact, in order to account for texture, it is precisely this information which must be included. This is simply another way of saying that text semantics is in practice a mediation between social semiotic formations and lexicogrammatical system selections, and is therefore a problem of tractable complexity in the text-production perspective. In principle, of course, it is still equivalent to the fully articulated connections between social semiotic resource systems and the lexicogrammar, but specifying those for every text directly would certainly not be a tractable problem in the text-production (or probably any other) perspective. Thus, context of situation and context of culture, representing the kinds of text-semantic relations that give a text its texture in all the ways we have been learning about in the last several years (that is, cohesion, thematics, genre, rhetorical structure, value-orientations, etc.) can also be analyzed in terms of the constraints they place on any next lexicogrammatical selection at a particular rank. Furthermore, the information

32

Jay L. Lemke

they provide, if complete as an account in terms of formations, should also completely account for the probability distributions of selections across different clauses of the same text.

5. Text production and emergent meaning As I have described the text-production perspective so far, it may seem to be excessively deterministic. That is only because I have been foregrounding the arguments needed to show that this dynamic approach to text semantics is feasible. But what is most surprising about this approach is its radical implications for the relative unpredictability of the creative process of text production. Because social semiotic models require us to take into account the ways in which actions reciprocally influence and alter the contexts of situation which called them into being in the first place, the dynamic model of text production provides a framework for analyzing, not just relatively predictable features, but phenomena of emergent meaning in text. Every act or utterance means in the context which it creates by occurring. That context is always newly emergent from the context that preceded the act or utterance. Very often, the shift in situation is either predictable or of a kind which in effect reinforces the categorization of situation-type, rather than altering it typologically (though it always alters it phenomenologically). But from the text-production perspective, what really matters is whether the total context of situation (including the state of the text-up-till-now) has been altered in a way that will influence the future direction of text production. And that almost always does happen. It is most obvious in longer texts as we write (or speak) them. We set out to write a text of a given type for a given situation. But the text plan at the beginning only specifies certain typological parameters of the textto-be-produced. It is that narrowing which begins to make text production tractable; only some paths can be followed in the immense space of all clauses and all successions-of-clauses (many of which are not even texts). But no text plan (except in rare, ritualized circumstances) pre-specifies the exact final text word-for-word! There are always many functionally equivalent (though systemically distinguishable) ways to meet the specifications of any text-type. Genres, rhetorical formations, field-specific thematic patterns, and orientational "stances" specify "generic" meanings and meaning relations at the level of social semiotic actions (social

Text production and dynamic text semantics

33

signifying practices), which may be instantiated semantically in a variety of linguistically distinct ways. We begin by putting down some words that are "good enough" for present purposes. But we do not write (or speak) semantemes, just as we do not utter phonemes. Every real phonetic utterance has many acoustic features, quite recognizable and even meaningful in some circumstances, which are non-distinctive phonologically. Just so, every realization of the semantic specifications of a text plan contains "incidental" features which were not relevant to those specifications. Each is just one of many ways in which the meaning required could have been made. From each, a more delicately different meaning can be made, but those differences were nondistinctive so far as the previously relevant formation (for example, genre) was concerned. (We see here another consequence of the reduction in meaning potential from systems to formations. If formations specified meaning requirements as precisely as do systems, there would be no specificity to the culture of a given community, and there would be a unique text to each text-type.) But once the meaning requirements of a formation are met by an actual, realized portion of text, the text itself becomes part of the context of situation, and potentially all of its linguistic (and graphic, if written; paralinguistic, if spoken) features, including the previously incidental ones, may now become relevant to how succeeding portions of the text are realized. What we first wrote or said was one way of instantiating a part of the original text plan, and only some of its features were specified by that plan. N o w we must go on to instantiate another part of the text plan, and again it will specify only some features. But how we instantiate those features now depends not just on the text-plan requirements, but also on how we happened to instantiate those of the previous portion(s) of the text. Principles of consistency now come into play that make previously irrelevant, "incidental" features highly relevant for text production. With each succeeding unit of text written, the range of options for fulfilling the remainder of the (potentially changing) text plan diminishes. This is a purely dynamic phenomenon, not a synoptic one. A feature is "incidental" only dynamically, only from the point of view of constraints on selection prevailing as it is selected. Synoptically, such "accidents" will already have been given a more text-significant meaning by having become, retroactively, part of subsequent patterns of selections. In this process, a new type of foregrounding takes place. Features which were essentially unpredictable (and irrelevant) so far as any text

34

Jay L. Lemke

plan (or formation) was concerned are now co-determinants of how further higher-level (that is, formational, text-semantic) semantic features will be realized. This foregrounding now potentially raises the status of these features (and their intertextual associates) to that of meaning elements, which the text plan must be revised to take into account. Such meanings are truly emergent in the text-production process. We all have had this experience in writing. We begin to write about something, and in the course of writing, we happen to choose a particular word when we might just as well have used another; then a particular grammatical construction that also was only one possible way-of-saying, and the result is that two unplanned words are brought into an unexpected juxtaposition and invoke an association from another text, or lead us into a formation not part of our original plan. We decide to stay with this theme or metaphor, or to develop it further, and our text plan changes. In spoken language this sort of process is, for me at least, a major source of previously unmade connections, of serendipities whose probabilities of occurring and being noticed derive from the margins of halfformed formations. It is the creativity of action, in which we are surprised by actions-as-performed, beyond the expectations of actions-as-planned, into new senses of meaning. There is a slippage between the lexicogrammatical underspecification of formational meanings and the unpredictable polysemies of (retrospectively overdetermined) text. That slippage makes a dynamic model of text semantics both tractable and capable of accounting for creatively emergent meaning-making in the use of language and of all the other grammars of action.

6. Relevance to text generation, text analysis, and stylistics Further development of the text-production perspective could provide a useful implementation of dynamic approaches to text semantics in several fields of investigation. In text-generation work, text planning strategies should take into account the post-selection consequences (and opportunities) of "incidental" realizations of text-plan goals. At a level below the globally invariant features of a text plan, there will need to be dynamically responsive strategies that can guide selections in subsequent clauses based on prior selections. This is not merely a matter of later "editing" for stylistic consistency; all textual coherence could be lost from a first version (especially of a long text) without this dynamic "channeling" of selections.

Text production

and dynamic

text semantics

35

Some day, it might even be possible for text-generation programs to "improvise" unpredictable but still meaningfully coherent texts in this way and to, for example, generate new hypotheses, speculate, and construct novel arguments or problem-solutions. This would require dynamic modifiability of even very high-order text-plan parameters ("sub-goals") during text-generation, based on flexible criteria of the potential meaningfulness of "incidental" patterns created and monitored en route to prior goals. In text analysis, the dynamic perspective suggests that part of the total meaning potential of a text (that is, the set of all possible meanings that could predictably be made with it in an interpretive community) derives, paradigmatically, from the other texts a given text-up-to-this-point might have become, if different "incidental" features were selected and/or eventually foregrounded. This implies a principle of prospective intertextuality to complement our more usual synoptic notions of retrospective intertextuality, which applies only to texts regarded as complete(d). Of all the texts that a text-up-to-here might have become, it did become this text. Prospective intertextual ties, that is, those to possible continuations of a text (as well as of these hypothetical texts to still other associations), while conforming to the text formations of a community, have a greater potential for escaping its system of disjunctions, even ac/incidentally (Lemke 1984: 131 —150). Perhaps that is one reason why society encourages people to learn to read far more than it encourages them to learn to write (cf. Kress 1982; Lemke 1989 a). Speaking is, of course, also potentially highly subversive of the standardized use and co-patternings of formations, but it is less readily monitored for its "ac/incidentals" (which are so easily lost forever). Finally, stylistic analysis, whether of texts regarded as literary or otherwise, might benefit from a shift in focus in linguistic work from characteristic (synoptic) patterns, which are likely to be shared with many other texts, to "incidentals" whose later effect on the direction of a text's stylistic choices might reveal something, not only of its uniqueness, but also of the creative process of its composition. Obviously in the case of short texts which have been extensively redrafted (for example, poems), this approach may be limited, but elsewhere it could conceivably be of interest. Where earlier and later drafts of a text are available, however, this method would seem ideally suited. Possibly the greatest use of the approach might be in the development of a "poetics" of prose composition (cf. Halliday 1982).

36

Jay L. Lemke

I hope that in all our work on how meanings are made with text we will remember that the text is a product and a record of meaning-making processes which are essentially dynamic. These processes are social semiotic practices, the signifying practices of a community. It is these practices that make texts and make sense of texts, dynamically, dramatically, moment-to-moment, word-by-word, enacting meaning by words, in moments whose meaning the words make and change.

References Anderson, John (ed.) 1982 Language form and linguistic variation. Amsterdam: Benjamins. Bateson, Gregory 1972 Steps to an ecology of mind. New York: Ballantine Books. Benson, James D. — William S. Greaves (eds.) 1985 Systemic perspectives on discourse. Vol. 1. Norwood, NJ: Ablex. 1988 Systemic functional approaches to discourse. Norwood, NJ: Ablex. Bloome, David (ed.) 1989 Classrooms and literacy. Norwood, NJ: Ablex. Coveri, L. (ed.) 1984 Linguistica testuale. Rome: Bulzoni. van Dijk, Teun — Jan Petöfi (eds.) 1977 Grammars and descriptions. Berlin: Mouton. Flood, James (ed.) 1984 Understanding reading comprehension. Newark: International Reading Association. Gregory, Michael 1985 "Towards communication linguistics", in: James D. Benson — William S. Greaves (eds.), 248-274. Gregory, Michael (ed.) in press Relations and functions within and around language. (Papers from ALRWG International Spring Colloquium, Glendon College, York University, Toronto.) Gregory, Michael — Peter Fries (eds.) in press Discourse in society: Functional perspectives. Norwood, NJ: Ablex. Halliday, Μ. A. K. 1977 "Text as semantic choice in social context", in: Teun A. van Dijk — Jan Petöfi (eds.), 176-225. 1978 Language as social semiotic. London: Arnold. 1982 "The de-automatization of grammar", in: John Anderson (ed.), 129 — 159. 1985 An introduction to functional grammar. London: Arnold. Halliday, Μ. A. K. — Ruqaiya Hasan 1976 Cohesion in English. London: Longman. 1985 Language, context, and text. Geelong: Deakin University Press. [Republished 1989, Oxford: Oxford University Press.]

Text production and dynamic text semantics

37

Halliday, Μ. A. K. — John Gibbons — Howard Nicholas (eds.) 1990 Learning, keeping, and using language: Proceedings of the 8th World Congress of Applied Linguistics, Sydney 1987. Amsterdam: Benjamins. Hasan, Ruqaiya 1984a "Coherence and cohesive harmony", in: James Flood (ed.), 181 —219. 1984 b "The structure of the nursery tale", in: L. Coveri, (ed.). 1985a "The texture of a text", in: Μ. A. K. Halliday - Ruqaiya Hasan, 7 0 - 9 6 . 1985 b "The structure of a text", in: Μ. A. K. Halliday - Ruqaiya Hasan, 5 2 - 6 9 . Kress, Gunther 1982 Learning to write. London: Routledge and Kegan Paul. Lemke, Jay L. 1983 "Thematic analysis: Systems, structures, and strategies", Semiotic Inquiry 3: 159-187. 1984 Semiotics and education. (Toronto Semiotic Circle Monographs Series.) Toronto: Victoria University. 1985 "Ideology, intertextuality, and the notion of register", in: James D. Benson - William S. Greaves (eds.), 2 7 5 - 2 9 4 . 1988 a "Discourses in conflict: Heteroglossia and text semantics", in: James D. Benson — William S. Greaves (eds.), 29 — 50. 1988 b "Text structure and text semantics", in: Robert Veltman — Erich Steiner (eds.), 1 5 8 - 1 7 0 . 1989 a "Social semiotics: A new model for literacy education", in: David Bloome (ed.), 2 8 9 - 3 0 9 . 1989 b "Semantics and social values" WORD 40: 3 7 - 5 0 . 1990 a Talking science: Language, learning, and values. Norwood, NJ: Ablex. 1990b "Technical discourse and technocratic ideology", in: M . A . K . Halliday — John Gibbons — Howard Nicholas (eds.). in press a "Ideology, intertextuality, and the communication of science", in: Michael Gregory (ed.). in press b "Intertextuality and text semantics", in: Michael Gregory — Peter Fries (eds.). Mann, William C. — Sandra A. Thompson 1987 "Rhetorical structure theory: A theory of text organization", in: Livia Polanyi (ed.). Mann, William C. — Sandra A. Thompson (eds.) in press Discourse description: Diverse analyses of a fund-raising text. Norwood, NJ: Ablex. Martin, James R. 1983 "Conjunction: The logic of English text", in: Jan S. Petöfi — Emel Sözer (eds.), 1 - 7 2 . 1985 "Process and text: Two aspects of human semiosis", in: James D. Benson — William S. Greaves (eds.), 2 4 8 - 2 7 4 . in press "Macroproposals: Meaning by degree.", in: William C. Mann — Sandra A. Thompson (eds.). Petöfi, Jan S. — Emel Sözer (eds.) 1983 Micro and macro connexity of text. Hamburg: Buske. Polanyi, Livia (ed.) 1987 The structure of discourse. Norwood, NJ: Ablex.

38

Jay L.

Lemke

Thibault, Paul J. in press "Interpersonal meaning and the discursive construction of actions, attitudes, and values", in: Michael Gregory (ed.). Veltman, Robert — Erich Steiner (eds.) 1988 b Pragmatics, discourse, and text. London: Pinter. Ventola, Eija 1987 The structure of social interaction: A systemic approach to the semiotics of service encounters. London: Pinter.

Towards probabilistic interpretations Μ. A. K.

Halliday

At the 13th International Systemic Workshop in Sydney in 1987, Jay Lemke spoke about cladistics, the study of how systems change through time; and he outlined models of two of the fundamental processes of change: the evolution of species, and the growth cycle of individual specimens. Here I want to follow up on two aspects of Lemke's observations, one general and the other specific. (His paper was entitled "Semantics, ideology and change".) Let me begin with a very general point. The twentieth century has been the age of the disciplines, when knowledge was organized into subjects each having its own domain, its own concept of theory, and its own body of method. As usually happens in history, what began as a strength, as the opening up of a new potential, gradually — as the potential was realized — became a constraint; and the energy that had gone into creating these structures has had to be diverted into tearing them down again. But all systematic knowledge implies some form of structure; so something has to take their'place; and what are emerging are structures of another kind, this time not disciplinary but thematic. Here the organizing concept is not the object that is being investigated — the "field" of study — but the kinds of question that are being asked. Cladistics is one such theme. In cladistics, the questions are about how systems change; the systems being studied might be cells, or human societies, or galaxies, but the same questions can be raised with reference to them all. The thematic organization of knowledge is not a new idea; the earliest instance in western thought was mathematics, where the theme was that of measuring things. Whatever you were investigating, if you studied it by using measurement you were doing mathematics. In the nineteenth century, evolution became a theme of this kind; it embodied the concept of history, the "arrow of time" which had been missing from the physical sciences and which they were now ready to bring back. But the human sciences were not ready for it. They still lacked the sense of a system, a model of how any form of organization persists; so in the following period they developed their own theme, that of structuralism. Change had to be taken off the agenda so that we could study how human systems

40

Μ. Α. Κ.

Halliday

were organized. Next came the theme of semiotics, the investigation of how things mean; this was clearly a relevant theme for the study of language, one which added a new dimension by contextualizing it within a wider universe of semogenic phenomena. (I am less certain about the status of cognitive science, which seems to me more like a generalized discipline than a theme — macro rather than meta in its relation to the subject-based categories of knowledge.) And then, of course, cladistics, the example from which I started. So in packing up for the move into the twenty-first century we are changing the way knowledge is organized, shifting from a disciplinary perspective towards a thematic one; and this takes me into the more specific line of argument that I want to develop out of Lemke's discussion. This concerns the cladistics of language: the way language persists, and also changes, over the course of time. From the point of view of European structuralism, a human system was a nexus of paradigmatic and syntagmatic relations that could be construed from the way they were manifested in patterns of behavior. Saussure's langue/parole was one of the first formulations of this view. This clearly puts the system itself in a synoptic perspective. It is "held together" by various metaphors: equilibrium, functional load, and the like. By contrast, the behavior — the text, if we are talking about language — is viewed dynamically. This contrast is made explicit in Hjelmslev's "system/process" interpretation. Obviously the system changes; but it moves slowly along the time track, like a glacier, so the synoptic perspective, which enables us to examine it in fine detail, does not noticeably distort it. The text, on the other hand, is like the glacier stream, the water flowing out of the glacier; the natural perspective on text is a dynamic one, especially if our prototype of it is spoken rather than written text. It is clear that there is a complementarity here, and that each terminal — the code, and the behavior — can be viewed from either perspective; and ultimately have to be, as Martin has pointed out (Martin 1985), if the structuralist model is to be retained. Of course, the synoptic and dynamic positions cannot be fixed and determinate; they are acted out in too many ways. As soon as one admits syntagmatic relations into the system (as opposed to modeling it purely paradigmatically), one is taking on a dynamic commitment; while on the other hand syntagmatic relations in text are usually represented as constituent structure, which is a synoptic concept. To take another example, the metafunctional categories can be viewed synoptically as the form of organization of the semantics and lexicogrammar; or else dynamically as currents flowing through a strat-

Towards probabilistic

interpretations

41

ified semiotic system. But questions about change suggest a predominantly dynamic perspective. How do we incorporate change into the structural linguistic concept of a system? Lemke (1984) has characterized human, social systems as "dynamic open systems" which have the property of being metastable: that is, they persist only through constant interaction with their environment, and hence are in a constant state of change. It seems clear that a language is a system of this kind. In what senses, then, does a language interact with its environment? This can only be through its instantiation in text. "The environment" of the system is its context of culture; but the processes whereby it can be said to "interact with" this environment are just those of people talking and listening. There are three different dynamics involved in these processes (Halliday in press). One is that just referred to: the history of the system itself, as it persists by changing through time. This would presumably be change of an evolutionary kind. The second is the history of the individual construing the system: the dynamic of ontogenesis. This seems to be change in the form of growth, maturation, and eventual death. Then, thirdly, there is the history of the instance: the dynamic of the unfolding of a text. The text is a semiotic event, not a system in the way that a language, and an individual "languager", can be said to be kinds of systems. But it is precisely this event, or rather the universe of countless such events, that constitute the system of language and give it its metastable form. But of course these events are not countless. The system may have infinite potential; but it engenders a finite body of text, and text can be counted. Thus, we can make a reasonable assessment of the quantity of text that is typically involved in ontogenesis: how many instances a child of a given age, growing up in a given environment (say, a monolingual family) will have heard of a given phenomenon (say, syllables, or clauses). This provides an insight into how the system is construed by an individual learner: how text becomes system where the dynamic is one of growth. This is one context — that of construing a semiotic system in the sense of learning it, as a child learns its mother tongue — for which we need to know about frequency in text. The other is so that we can explain how the system changes: the mechanism of metastability in a semiotic system, where the dynamic is one of evolution. Consider a physical system such as that of climate. This is instantiated in the form of weather: climate is the langue, weather the parole. Take any one component of climate: say, temperature. The exact temperature at any one place at any one

42

Μ. Α. Κ. Halliday

moment of time seems to be of little significance; but when at many different places the daily minimum and maximum go up by an average of one degree a decade for five decades, we say that the climate is changing. This is not because no such temperatures had ever occurred before; no doubt they had. But now the probability has changed. In fact, every single instance alters the probabilities of the system in some measure; but such perturbations are too small to be taken account of, and mostly cancel each other out. When they build up to form a trend, however, we recognize that the system is undergoing change. Frequency in text is the instantiation of probability in the system. A linguistic system is inherently probabilistic in nature. I tried to express this in my early work on Chinese grammar, using observed frequencies in the corpus and estimating probabilities for terms in grammatical systems (1956, 1959). Obviously, to interpret language in probabilistic terms, the grammar (that is, the theory of grammar, the "grammatics") has to be paradigmatic: it has to be able to represent language as choice, since probability is the probability of "choosing" (not in any conscious sense, of course) one thing rather than another. Firth's concept of "system", in the "system/structure" framework, already modeled language as choice. Once you say "choose for polarity: positive or negative?", or "choose for tense: past or present of future?", then each of these options could have a probability value attached. Shannon — Weaver, in their Mathematical theory of communication (1949), had provided a formula for calculating the information of a system, which I used in an exploratory way for a paper, "Information theory and linguistics", in 1968. This was a valuable concept for linguistics; but structuralist linguists had dismissed it because they had no paradigmatic model and had therefore attempted to relate it to representations of syntagmatic structure, to which it has no relevance at all. Information is a property of a system (in the Firthian sense); not of its individual terms. The system with maximum information is one whose terms are equiprobable; any skewness (departure from equiprobability) involves a reduction in information. Hence a minimally redundant system network would be one in which all systems tended towards equiprobability. Systems of this kind can be designed; but they do not normally evolve. In a human system, minimally redundant does not mean maximally efficient; and human systems, like evolved systems in general, always incorporate a certain redundancy in some form or other, the skewing of probabilities being one. When I was constructing my original system

Towards probabilistic

interpretations

43

networks for English in the early sixties (Kress, ed., 1976), I counted some grammatical frequencies, noting the occurrences of different terms in some of the primary systems, to see if any general principles of frequency distribution suggested themselves. It was necessary at the start to distinguish between systems of two different kinds: recursive systems, and non-recursive systems. (1) Recursive systems: those interpreted in metafunctional terms as "logical" — that is, in which the meaning is construed as a generalized semantic relation between pairs of elements, and thus can be selected iteratively. Languages differ (within limits) as regards which meanings they construe as "logical" in this sense; the typical ones are presumably those in the areas of expansion (/. e., e. g., and, or, then (time), so, then (condition), yet, etc.) and, perhaps, projection (says, thinks). English is unusual in also treating tense in this way, as a serial relationship in time (present relative to past relative to ...).

Such systems could generate indefinitely long strings. In fact they are fairly restrained. But while there may sometimes be a qualitative factor involved in limiting their output, as there seems to be in the case of English tense, in general the restraint is quantitative. If we represent the iterative choice as in Figure 1, then there is a probability attached to these options, and it is skew: one is more likely to stop than to go round again. Counting English tenses in some conversational texts, I found that the ratio was of the order of magnitude of one in ten. Very roughly approximated, for every occurrence of a 4-term tense (e. g., had been going to be working) there were ten 3-term tenses, a hundred 2-term tenses

44

Μ. Λ. Κ.

Halliday

and a thousand 1-term tenses. So I formulated this as a general hypothesis about recursive systems, representing it in a network as Figure 2, where system (I) is the logical-semantic relation (e. g., past/present/future) and system (II) is the option of iterating the choice.

Figure 2. Recursive system with probabilities shown

(2) Non-recursive systems: those of the other metafunctions. In principle these should be able to range over all probability distributions from 0.5/ 0.5 to approximating 0/1. But this kind of spread seemed to me to be highly unlikely. It would be unlikely for all systems to be equiprobable, since this would not leave enough redundancy. (In fact it could be shown to be impossible, since a network is not a simple taxonomy; there are conjunct and disjunct entry conditions in it.) But it would equally be unlikely for systems to take up all possible distributions of probabilities along the whole continuum from equiprobable to maximally skew. On the basis of what little counting I had done, I suggested a bimodal distribution. The hypothesis was that systems tended towards one or other of just two types, (i) equiprobable and (ii) skew, with the skew tending towards a ratio of the order of magnitude ten to one (which I represented for obvious reasons as nine to one, i. e., 0.9/0.1). This cor-

Towards probabilistic

interpretations

45

responds to one interpretation of the concept of marking: type (i), the equiprobable, have no unmarked term, while type (ii), the skew, have one of their terms unmarked. Expected examples of each type, in English, would be: (i)

equiprobable (0.5/0.5) number: singular/plural non-finite aspect: 'to'/'ing' process type: material/mental/relational nominal deixis: specific/non-specific verbal deixis: modality/tense

(ii)

skew (0.9/0.1) polarity: positive/negative mood: indicative/imperative indicative mood: declarative/interrogative voice (verbal group): active/passive declarative theme: Subject-theme/other theme

Two recent studies have been concerned with text frequency and its relation to probability in the grammar: Plum — Cowling (1987) and Nesbitt — Plum (1988). Both are concerned with conditioned probabilities, but where the conditioning is of two different kinds: interstratal in the former, intrastratal in the latter. I shall consider each of these in turn. Plum — Cowling examined 4436 finite clauses from interviews conducted for the Sydney Social Dialect Survey, taking into account (among others) the network of options shown in Figure 3: f— past I— temporal -

TENSE

present L

finite clause

VERBAL DEIXIS

-modal Figure 3. Network for example from Plum — Cowling (1987)

future

46

Μ. Α. Κ.

Halliday

Of the total, 3294 selected temporal deixis and 1142 modal. These were distributed by social class of speaker as shown in Table 1: Table 1. Selection of temporal and modal deixis by social class of speaker [from Plum — Cowling 1987] temporal deixis

modal deixis

Total

Lower working class Upper working class Middle class

891 1323 1080

213 603 326

1104 1926 1406

Total

3294

1142

4436

Thus there is a slight skewing by social class: lower working class favour tense more strongly than the others, and middle class slightly more strongly than upper working class, athough the difference is hardly significant. But when they take the 3294 having temporal deixis and examine the selection of primary tense, excluding a very small number of futures (27 in all, so leaving 3267 past or present), the results are as in Table 2: Table 2. Selection of primary tense (past/present) by social class [after Plum — Cowling 1987] past Lower working class Upper working class Middle class Total

present

Total

306 (35%) 789 (60%) 752 (70%)

579 (65%) 515 (40%) 326 (30%)

885 (100%) 1304 (100%) 1078 (100%)

1847 (57%)

1420 (43%)

3267 (100%)

So while the overall probability of past/present was fairly even (57%/ 43%), the conditioning effect of social class, in the register in question, skewed this in opposite directions: the middle class favored past and the lower working class favored present, with the upper working class located in between. This is a classic manifestation of Berstein's principle of code, or (to give it the more accurate term) "coding orientation". If the linguistic system was not inherently of a probabilistic kind it could not display these sociolinguistic effects. As it is, this kind of quantitative study can reveal important features relevant to the deeper social order.

Towards probabilistic

interpretations

47

An example of such a study, but one with a considerably broader scope, is that of Ruqaiya Hasan, who has been conducting research into how the linguistic interaction between mothers and children shapes the way children learn: their forms of reasoning and of knowing, the ways in which they construe experience in language, and the dimensions of the semiotic space within which their consciousness is constituted and enabled to develop. She has assembled a data base of some 21,000 messages (using that term to label the semantic unit typically realized as a moodselecting clause), consisting of spontaneous conversation, in the home, between mothers and their children of 3—4 years. There are 24 such dyads, 12 with boys and 12 with girls. Hasan has used systemic theory to construct a semantic network representing the total paradigm of significant options, and the entire corpus has been analyzed in respect of the features incorporated in that network. She has then used principal components analysis, as this has been adapted to variation theory, to study the systematic patterns of variation in semantic choice; for example, in the way the mothers answer their children's questions. What the program does is to cluster those sets of features that contribute most to variation in the data (the principal component factors); for example, in parts of the discourse where the mothers are answering their children's questions, a factor consisting of some six semantic features is found to account for a large amount of the variance among the individual dyads. Interestingly, three of these features relate to the children's questions and three relate to the mothers' answers. Once these features are identified it is possible to examine what extralinguistic variables in the population are being reflected in this semantic variation. For these enquiries Hasan has been able to use research methodology developed in variation theory; but her theoretical stance is very different from that of Labov. Labov's variants are by his own definition meaningless; they are phonological or morphological realizations of an assumed semantic constant — and hence arbitrary, in that there is no significance in which group selects which variant (to take the classic example, the prestige variant is in America, — r in England). By contrast, Hasan is concerned with the statistical properties of meaningful choice. The key concept here is Hasan's "semantic variation", the selection of different meanings which are themselves the realization of some constant at a higher level (hence this is code variation, not register variation where there is no higher level constant). But here the higher level constant is located in the social contexts, specifically the socializing contexts of mother-child

48

Μ. Α. Κ. Halliday

interaction; so the options differentially favored by one or other social group are alternative realizations of the same semiotic activity, for example, that of controling the child's behavior, or of helping in the search for explanations. See Cloran (1989); Hasan (1988); Hasan - Cloran (1990). The predominant extralinguistic variables that are found to correlate with this semantic variation are the sex of the child and the social class position of the family. Hasan's results thus strongly support Bernstein's original findings and his general theory, derived as they are from an entirely different data base and underpinned by a much more elaborated linguistic model than was available to Bernstein in the sixties. But the significance of Hasan's work for the present discussion is that the semantic features involved in this variation are realized, non-arbitrarily, through grammatical systems to which probabilities can be assigned. Hasan is showing that the semantic choices made by the mother affect the child's ways of reasoning and of learning; that this is possible is because the lexicogrammatical features that carry these semantic patterns are quantitatively foregrounded, and such foregrounding affects the probabilities of the child's own system. If text frequency was a so-called "performance" phenomenon, it would have no effect on the "competence" of a listener. But the child's potential is affected; not in the sense that it becomes a direct copy of the mother's, but in the sense that the probabilities in the child's system collectively constitute a meaning style that is semiotically and ideologically compatible with that of the mother. Part of the meaning of a grammatical system is the relative probability of its terms. This is so even if these probabilities remain constant across all environments — as in some instances no doubt they do. Where they do not, there may be systematic variation that can be interpreted as the conditioning of these probabilities "from above". This may be either from code or from register; in either case, it is exploiting the potential of language as a stratified probabilistic system. Let us now consider the situation where probabilities are being investigated from the point of view of their conditioning within the grammar itself Nesbitt — Plum (1988) investigated the relative frequency of options in the clause complex system, using data from Plum's sociolinguistic interviews with dog fanciers. The relevant system network is given in Figure 4 (see next page). The total sample was 2733 clause nexuses. The raw scores and percentages for each of the two systems taken separately, (i) interdependency, or "taxis"; (ii) logical-semantic relations, were as set out in Table 3:

Towards probabilistic interpretations

49

Figure 4. Network for example from Nesbitt — Plum (1987)

Table 3. Figures and percentages for taxis and logical-semantic relations [after Nesbitt — Plum 1988] I N T E R D E P E N D E N C Y (TAXIS)

L O G I C A L - S E M A N T I C RELATIONS

parataxis hypotaxis

1918 (70%) 815 (30%)

expansion projection

2284 (84%) 449 (16%)

Total:

2733 (100%)

Total:

2733 (100%)

50

Μ. Α. Κ.

Halliday

When these are intersected, the figures are as given in Table 4: Table 4. Taxis (parataxis/hypotaxis) by logical-semantic relations (expansion/projection) [after Nesbitt - Plum 1988] ' — — L O G I C A L - S E M A N T I C

expansion

projection

Total

parataxis

1656

262

1918

hypo taxis

628

187

815

2284

449

2733

TAXIS

Total

Expressing these as percentages, and summing along both axes, we get Table 5: Table 5. Taxis by logical-semantic relations, with percentages both ways [after Nesbitt — Plum 1988] LOGICALSEMANTIC RELAXATIONS

expansion

projection

Total

TAXIS 86% parataxis

14% 262

1656 58%

73% 77% hypotaxis

187

815

42% 2284

100%

100%

23%

628 27%

Total

100% 1918

449 100%

2733

Thus, if we ask whether the proportions remain the same when the two system are intersected, the answer is that, on the whole, they do. The effect is slightly exaggerated along both dimensions: that is, the combination of "parataxis χ expansion" is 73% of all taxis and 86% of all logical-semantic relations. But the difference between these and the simple proportions (70%, 84%) is not significant.

Towards probabilistic

interpretations

51

When they take the next step in delicacy, however, the picture changes. Table 6 shows the raw scores for taxis intersected with the subcategories of expansion, namely elaborating, extending, and enhancing: Table 6. Taxis by expansion (elaborating/'extending/enhancing) [after Nesbitt — Plum 1988] elaborating

extending

enhancing

Total

parataxis

386

1117

153

1656

hypotaxis

128

42

458

628

Total

514

1159

611

2284

PAN SI O N TAXIS

Nesbitt — Plum sum these as percentages along one dimension only, as in Table 7: Table 7. Taxis by expansion, with percentages showing proportions of taxis within each type of expansion [after Nesbitt — Plum 1988]

TAXIS

—-^JiXPANSION —

elaborating

parataxis

extending

386 75%

hypotaxis

1117 96%

4% 514

100%

153 25%

42

128 25%

Total

enhancing

458 75%

1159 100%

611 100%

They then argue as follows. If you choose "extending", then there is virtually no choice of taxis; we can treat this as "all extending are paratactic". (This corresponds to the traditional category of "co-ordination", which is a complex of parataxis χ extending. The small group of hypotaxis χ extending, such as clauses introduced by as well as, besides, rather than, instead of, would be ignored.) These can therefore be left out of consideration, leaving just the intersection of parataxis/ hypotaxis with elaborating/enhancing. If you choose "elaborating", you are more likely to choose parataxis (the traditional category of "apposi-

52

Μ. Α. Κ. Haüiday

tion"); while if you choose "enhancing", you are more likely to choose hypotaxis (the traditional "adverbial clauses", plus equivalent non-finites). In other words, as Plum — Nesbitt (1988: 21) express it, "elaborating favours parataxis, enhancing favours hypotaxis". This is certainly true — although it would be relevant to point out that elaborating favors parataxis just in the proportion in which parataxis is favored over hypotaxis in expansion as a whole (75% :25% ::: 73% : 27%). But this time if we sum the rows instead of the columns we obtain rather a different picture, as in Table 8: Table 8. Taxis by expansion, with percentages showing proportions of expansion within each type of taxis [based on Nesbitt — Plum 1988]

TAXIS

-~~^EXPANSION elaborating ^ ^ ^

parataxis

extending

23% 386

hypotaxis

67% 1117

20% 128

enhancing

10% 153

7% 42

Total

100% 1656

73% 458

100% 628

From this point of view, parataxis and hypotaxis are identical in their effects on (i. e., as environments for) elaborating, which remains at between a fifth and a quarter of the total (23% in the environment of parataxis, 20% in that of hypotaxis). What changes, in this perspective, is the effect on extending and enhancing; thus: parataxis: favors extending (67%), disfavors enhancing (10%) hypotaxis: disfavors extending (7%), favors enhancing (73%) — with elaborating bubbling along as a minor motif in both. When systems are intersected in this way, one is being treated as the conditioning environment of the other; and since either system can take either role, this gives two complementary perspectives. Nesbitt — Plum's interpretation is: first choose your logical-semantic relation, and that will help to determine the taxis. But it also makes sense to say: first choose your taxis, and that will help to determine the logical-semantic relation. The picture may look rather different from the two perspectives, and for a robust interpretation it may be necessary to take account of both. When taxis is intersected with projection, the picture looks the same whichever system is chosen as environment of the other, as in Table 9:

Towards probabilistic

interpretations

53

Table 9. Taxis by projection, with percentage summed both ways [after Nesbitt — Plum 1988) PROJEC-

locution

idea

Total

TAXIS 87% parataxis

13%

227

35

86%

19% 20%

hypotaxis

80%

37 14%

100%

100%

150

187

185

449

81% 264

Total

100% 262

100%

In other words: locution: favors parataxis (86%), disfavors hypotaxis (14%) idea: disfavors parataxis (19%), favors hypotaxis (81%) parataxis: favors locution (87%), disfavors idea (13%) hypotaxis: disfavors locution (20%), favors idea (80%). These figures, which are very robust, give a clear account of the association between these two systems; they show up the two prototypical kinds of projection, (i) direct speech and (ii) indirect thought. Of course, the figures relate only to this register; but if the pattern turns out to be repeated across other functional varieties then it can be regarded as a feature of the system of the English language at its present stage. Now, this kind of partial association between systems could be perfectly stable; but it could also, as Nesbitt — Plum point out, be a symptom of a change in progress. The hypothesis would be that at some earlier time there was only one system, say, "projected speech / projected thought", the contrast between the terms being realized by two variables: "projected speech" by direct deixis plus a verbal process: "projected thought" by indirect deixis plus a mental process. Subsequently, these realizations were deconstructed so as to be able to vary independently of each other; but the association between them remained in a probabilistic form. This "dissociation of associated variables" is, as I have pointed out elsewhere, an important semogenic resource; and it is virtually inconceivable that it should be other than a gradual process. That is to say, whether or not the situation would eventually evolve to one where the two systems were entirely

54

M.A.K.

Halliday

independent of one another, it would not reach that stage in one catastrophic leap. Let me summarize here the principal types of relationship that we may expect to find between two simultaneous systems. It should be remembered that "simultaneous" means "treated as independent", i. e., networked as in Figure 5: c

Figure 5. Simultaneous systems

From the point of view of a probabilistic grammar, this signifies "anything other than one being fully dependent on the other", which is networked as in Figure 6 (i. e., as a relationship in delicacy):

II ι

Figure 6. System II dependent on system I (related in delicacy)

In other words, simultaneous systems may be statistically associated — that is, only partially independent, their intersection being governed by probabilities which are not simply the product of the separate probabilities of each. In such a case, one system acts as a conditioning environment on the other, so that a term in System I "favors" the choice of a term in System II. Tables 10 (i) — (vii) show some possible distributions of text frequencies that would reflect typical relationships between a pair of simultaneous systems. In (i) —(iii), the two systems are unassociated (fully independent); in (iv) — (vii) there is some association between them. The figures are constructed for binary systems; they can be intrapolated for systems that are ternary or beyond.

Towards probabilistic interpretations

55

Table 10 a (i) — (iii). Intersection of systems I and II where both are (fully) independent systems a

b

Τ

X

100

100

200

y

100

100

200

τ

200

200

400

I II

(i) both have no unmarked term I

a

b

Τ

X

324

36

360

y

36

4

40

τ

360

40

400

II

(ii) both have one term unmarked I

a

b

Τ

X

180

180

360

y

20 *

20

40

τ

200

200

400

II

(iii) I has no unmarked term, II has one term unmarked Table 10 b (iv) — (vii). Intersection of systems I and II where the two are associated (partially dependent) I

m

η

Τ

II Ρ

140

60

200

q

60

140

200

τ

200

200

400

(iv) both have no unmarked term, but m ρ favored, η q favored

56

M.A.K.

\ II

I \

Halliday m

η

Τ

Ρ

140

100

240

q

60

100

160

τ

200

200

400

(ν) both have no unmarked term, but m favors ρ I

m

η

Τ

II Ρ

140

60

200

q

100

100

200

τ

240

160

400

(vi) both have no unmarked term, but ρ favors m I

m

η

Τ

II Ρ

180

20

200

q

20

180

200

τ

200

200

400

(vii) stronger variant of (iv): "reversal of marking"

By investigating this kind of internal conditioning within the grammar, using a large-scale corpus of text, we can greatly improve on the present purely qualitative interpretation, according to which systems are either dependent or simultaneous, and recognize intermediate degrees of "partial association" between systems in a system network. The corpuses are now readily available, for example the COBUILD corpus at Birmingham (see Sinclair, ed., 1987), but the analysis still has to be manual, because no parser can yet assign enough feature descriptions to be usable for the purpose. However, one such pass through a corpus, tagging it for the principal systems in some grammar such as the "Nigel" grammar networks developed at the Information Sciences Institute (ISI), would yield

Towards probabilistic

interpretations

57

material for a great number of valuable studies. It would also be useful, for example, to study transitional probabilities: how the choice of a/b is affected by the choice made in the same system in the preceding clause, or other relevant unit, in the text. The objection has sometimes been raised that, while it is perfectly possible to establish frequencies of occurrence of grammatical features in a corpus of text, these cannot be interpreted as probabilities in the system because every text is in some particular register. This is a spurious objection. Of course, every text is in some register, just as every text is in some dialect — that is, every text is located somewhere in dialectal and diatypic space. This means that the greater the variety of texts examined, the more accurate the picture will be; it requires a lot of observations to approximate to a quantitative profile of the grammar of a language — but that does not render it a meaningless concept. To return to the earlier analogy with climate, every set of temperature readings is made under certain specific conditions — latitude and longitude, height above sea level, time of day, on land or on water, and so on; but it is perfectly meaningful to talk, as everyone is doing, about global temperatures, provided one takes a broad enough sample. These are the probabilities in the climatic system. It is interesting that nobody disputes lexical probabilities; it is accepted that words can be compared for their relative probability of occurrence in the language as a whole, and that this is achieved by examining word frequencies in a wide variety of registers. It is the same principles that are being invoked for probabilities in grammar. In fact lexis and grammar are not different phenomena; they are the same phenomenon looked at from different ends. There is no reason therefore to reject the concept of the overall probability of terms in grammatical systems, on the grounds of register variation. On the contrary; it is the probabilistic model of lexicogrammar that enables us to explain register variation. Register variation can be defined as the skewing of (some of) these overall probabilities, in the environment of some specific configuration of field, tenor, and mode. It is variation in the tendency to select certain meanings rather than others, realizing variation in the situation type. This relates register variation (diatypic) to diachronic variation, since the latter is also variation in the probabilities of the linguistic system — variation along the arrow of time. Sometimes, in the course of this process, a probability will achieve certainty (a value of 0/1), which is a categorical change; and sometimes there will be catastrophic change, as happens when the entire system is creolized. These are the limiting cases (and as

58

Μ. Α. Κ. Halliday

such, both have analogies in diatypic variation). In terms of the cladistic model, linguistic change through time is evolutionary; and this also seems to apply, at least in the case of examples such as the language of science, to the history of particular registers. One of the recurrent motifs of structuralism was the warning against transferring evolutionary concepts onto language; and it was a wise warning, because in many ways, as these had usually been applied, they clearly did not fit. But the concepts are now understood and applied rather differently; and evolution — like other fundamental issues, such as the origins of language, or the relation between language and culture — can be readmitted onto the agenda. It goes without saying that, if we draw on general concepts about systems, and system change, in our interpretation of language, we recognize that semiotic systems differ in important ways from physical systems, and also from biological systems. These differences need to be re-problematized: how, and why, are semiotic systems different? The two critical vectors are surely those of instantiation and realization. The first is the relation between the system and the instance; the second is the relation between strata. If we are talking specifically about language, then the first is the relation between language as system and language as text; the second is that relating semantics, lexicogrammar, and phonology. Saussure conceptualized the first in terms of the opposition between langue and parole, the second in terms of that between signified and signifier. At the same time, Mathesius introduced the concept of "oscillation": variation in language that was both diatypic and diachronic. With a probabilistic interpretation of the system we are, in a sense, combining the Saussurean and the Mathesian perspectives. What is special about semiotic systems is the nature of these two relationships, and in particular, the way they combine qualitative and quantitative effects. This is obviously not something to be discussed in a few closing paragraphs; so I shall just make one point about each. With regard to instantiation, I have used the probabilistic interpretation to try and show how text becomes system, bringing in the analogy from the physical system of climate. So let me emphasize the difference: namely that, in a semiotic system, instances have differential qualitative value, so that one highly valued instance can by itself lead to a change in the system. We might call this the "Hamlet factor", since the main context in which such value accrues is that of literature, and Hamlet is, as the (no doubt fictitious) old lady said, "full of quotations". The Hamlet factor is probably operative only in semiotic systems.

Towards probabilistic

interpretations

59

With regard to realization, all I shall try to do is enumerate some steps in what is inevitably a rather complex argument. The natural sciences, especially theoretical physics, are characterized by very high orders of abstraction. This abstraction, however, is located in the interpretations of physical processes (i. e., the metalanguage), not in the processes themselves; in other words, the systems characterized in this way are semiotic systems, not physical systems. Physical and biological systems are typically compositional, their processes governed by cause-and-effect. (They may include some of the semiotic type also — Jakobson used to cite the genetic code; and some very recent theories treat physical systems as semiotic systems.) Semiotic systems are characterized by stratification, their (internal) "processes" being realizational. Linguistic systems are typically treated as if they were of this compositional, cause-and-effect type: "a consists of b, + b2 + ...; b consists of c, + c 2 + ...", or "a causes b; b causes c; ...". (A common outcry against stratal models is "but which stratum causes which?"). Thus in generalizing the signified \ signifier relationship across the system, we have treated the strata as if they were entities in a universe of this kind: "a is realized by b; b is realized by c; ..." Then, in order to avoid a crude cause-andeffect interpretation, we have tried to remove all directionality from the system, using constructivist metaphors that are made to face both ways: the context constructs the grammar, and the grammar constructs the context. A more promising interpretation of realization is, in my view, Lemke's (1984) concept of "metaredundancy". This is, quite explicitly, a directional model, and there has to be a lower bound: let us call it stratum z. Stratum y is realized by stratum z; y \ z. But stratum χ is realized, not by stratum y but by the realization of stratum y by stratum ζ: χ \ (y \ z). For stratum w, likewise: w \ (χ \ (y \ z)); and so on. Such a relationship is not reducible to a chain of dyadic, cause-and-effect type relations. This makes sense of Hjelmslev's notion of the connotative semiotic, as used by Martin (1985, and elsewhere). It also enables us to escape the constructivist trap, which seems to force an unreal choice between "language expresses reality" and "language creates reality". (I have been using "construe" to try to suggest the upward relationship in the Lemkean model.) I also find it helps me to understand how we can change reality by getting busy on language. There must be some reason why so much of our interaction with our environment begins and ends in grammar. This last section has been, clearly, a trailer; not a systematic presentation of ideas coherently thought through. It is intended mainly to

60

M.A.K.

Halliday

contextualize what went before, since it seems to me that in order to understand the special properties of semiotic systems it is necessary to show that such systems can be represented in probabilistic terms.

References Benson, James D. — William S. Greaves (eds.) 1985 Systemic perspectives on discourse. Vol. 1. Norwood, NJ: Ablex. Bollettieri Bosinelli, Rosa Maria (ed.) in press Language systems and cultural systems: Proceedings of the International Symposium on Bologna, Italian Culture, and Modern Literature. University of Bologna. October 1988. Cloran, Carmel 1989 "Learning through language: The social construction of gender", in: Ruqaiya Hasan — James R. Martin (eds.), 111—151. Fawcett, Robin P. — David J. Young (eds.) 1988 New developments in systemic linguistics. Vol. 2. London: Pinter. Gerot, Linda — Jane Oldenburg — Theo van Leeuwen (eds.) 1988 Language and socialization: Home and school. (Proceedings from the Working Conference on Language in Education, Macquarie University, 17 — 21 November 1986.) Sydney: Macquarie University. Halliday, Μ. A. K. 1956 "Grammatical categories in modern Chinese", Transactions of the Philological Society 1956: 1 7 7 - 2 2 4 . 1959 The language of the Chinese "Secret History of the Mongols". (Publications of the Philological Society.) Oxford: Blackwell. in press "The history of a sentence: An essay in social semiotics", in: Rosa Maria Bollettieri Bosinelli (ed.). Halliday, Μ. A. K. — John Gibbons — Howard Nicholas (eds.) 1990 Learning, keeping and using language: Selected papers from the 8th World Congress of Applied Linguistics. Sydney 16—21 August 1987. Amsterdam/ Philadelphia: Benjamins. Hasan, Ruqaiya 1986 "The ontogenesis of ideology: An interpretation of mother-child talk", in: Terry Threadgold et al. (eds.), 1 2 5 - 1 4 6 . 1988 "Language in the processes of socialization: Home and school", in: Linda Gerot — Jane Oldenburg — Theo van Leeuwen (eds.), 36 — 95. Hasan, Ruqaiya — Carmel Cloran 1990 "Semantic variation: A sociolinguistic interpretation of everyday talk between mothers and children", in: Μ. A. K. Halliday — John Gibbons — Howard Nicholas (eds.). Hasan, Ruqaiya — James R. Martin (eds.) 1989 Language development: Learning language, learning culture. (Advances in Discourse Processes 27.) Norwood, NJ: Ablex.

Towards probabilistic interpretations

61

Kress, Gunther 1976 Halliday: System and function in language: Selected papers. London: Oxford University Press. Lemke, Jay L. 1984 Semiotics and education. (Toronto Semiotic Circle Monographs, Working Papers and Prepublications, no. 2.) Toronto: Victoria Unversity. Martin, James R. 1985 "Process and text: Two aspects of human semiosis", in: James D. Benson — William S. Greaves (eds.), 2 4 8 - 2 7 4 . Nesbitt, Christopher — Guenter Plum 1988 "Probabilities in a systemic grammar: The clause complex in English", in: Robin P. Fawcett - David J. Young (eds.), 6 - 3 8 . Plum, Guenter — Ann Cowling 1987 "Social constraints on grammatical variables: Tense choice in English", in: Ross Steele - Terry Threadgold (eds.), 281 - 3 0 5 . Shannon, Claude E. — Warren Weaver 1949 The mathematical theory of communication. Urbana, Illinois: University of Illinois Press. Sinclair, John M. 1987 Looking up: An account of the COBUILD project. London/Glasgow: Collins ELT. Steele, Ross — Terry Threadgold (eds.) 1987 Language topics. 2. Amsterdam/Philadelphia: Benjamins. Threadgold, Terry — E. A. Grosz — Gunther Kress — Μ. A. K. Halliday (eds.) 1986 Semiotics, ideology, language. (Sydney Studies in Society and Culture 3.) Sydney: Sydney Association for Studies in Society and Culture.

A functional model of the system of sentence structures Frantisek

1.

Danes

An overview of the model

1.1. General assumptions about language and linguistics The large field of interest of present-day empirical linguistics — which, in fact, is a complex of hyphenated sciences — covers three major domains: the domain of language as a system of knowledge (being a component of the overall communicative knowledge); that of texts (optical and acoustic ones, fresh or preserved); and that of text processing, comprising text production and text comprehension/interpretation. These domains are not independent, to be sure, and they rather represent three facets or aspects of one and the same manifold phenomenon, which is set, moreover, in the wide sphere of human communication. One has only to agree with de Beaugrande (1987: 184) that "we might see a natural language as a complex system in which spoken and written texts manifest the coded aspect and the use of those texts the dynamic aspect." Compare also Halliday's concept of language as code and as behavior (1984: 7): "... we are not simply concerned with explaining behavior in terms of the code. More significant ... is the aim of explaining the code in terms of behavior. The system is determined by the process." The factual interdependence of the three domains of language is reflected, naturally, in linguistic research as well. Many years ago Mathesius (1975: 13) duly pointed out the fact that language, being "the sum of the possibilities available to the members of a language community ... for the purpose of communicating through speech, is identifiable from their realizations in particular utterances." In a similar vein a recent statement by de Beaugrande (1987: 163) asserts: "Any linguist can get evidence only from his encounters with texts: the system itself never steps forward in its concrete selfhood. Consequently, the linguist deals with data in whose constitution and interpretation he was inextricably involved." Let us add that this also means that the linguist is involved in text processing as well: when seeking to abstract the systematic knowledge from experience,

64

Frantisek

Danes

he has to comprehend and interpret texts and sometimes to produce them, too. I do not intend to enter heuristic problems here. I only wanted to point to the fact that our engagement in modelling language as system or code implies texts and processing as well. The distinctive specificity of such a modeling consists only in the actual focus and objective of our search. 1.2. The basic features of the model One of the starting assumptions of my syntactic studies is that the domain of sentence/clause appears well-structured enough for modeling, that is, "for being translated into some more systematically organized and generalizable notation" (cf. de Beaugrande 1988: 111). My proposition is to describe the structure of the sentence by means of the correlation of a grammatical sentence pattern and a semantic (propositional or experiential) sentence pattern. Such a correlation of patterns will be called a complex sentence pattern. Thus the English sentence: (1)

The farmer killed a duckling.

appears to be based on: complex sentence pattern (1): NounVAgent Noun 2 /Patient Its Latin structural analogon: (2)

Caius necat leonem.

corresponds to: complex sentence pattern (2): Noun'nominative/Agent υ coi CJ IS 'S so •Ό

ω ε

3 υΟ ΕΛ α

•a

ου c +5-»3 α. ω υο CS ω C 'δ ω U60
ω οί

ε

+-Ι

[BCET : A : 27] B: Teacher training is a good thing to be on. C: I don't want to he a teacher or anything. B: It gives you a year. C: Yeah. B: A bit slow ojf the mark here. C: It's terribly slow, yes. B: Not like you. C: I'm afraid it's like me of late.

The interpenetration

oflanguage

as code and language

as behavior

203

A Compliment is also typically responded to by thanking the speaker for the compliment or by expressing an appreciation, such as That's very kind of you.6 This option is not available in typical responses to other subcategories. For example, a thanking and an appreciation in response to a Criticize would be sarcastic, as can be seen in the following piece of data. (18)

—>

[B : D : Β : 1] Μ has a bad cold and Η couldn't recognize her voice. H: You sound terrible, you sound like a man. M: Thank you.

It should be noted that sometimes, an evaluation directed at a third party could be intended, or taken to be, directed at the hearer. For example: (19)

—>

[B : A : A : 3 : 5] X: I ah walked past your little boy today. He sure is a cute little fella, Kenneth, I must say. I'm sure he's a pain in the ass to live with but ah K: Yeah, he's independent, he's hungry, when he's hungry at times he's a pain in the ass otherwise =

[

X: K: X: K: X:

Yes. =he's alright. Yeah, yeah I like him, he's a cute looking little guy. Yeah, yeah, he he is he's a doll he's he's My god, that's for sure.

In (19), X's positive evaluation of K's son is indirectly a Compliment. This can be seen from the fact that Κ downgrades cute looking little fella to he's alright. Downgrading a Compliment is a typical response because to agree with or to upgrade a Compliment is to make a positive evaluation of oneself and this violates the social norm of modesty. Hence, typically, it is responded to by either a contrary evaluation, or by downgrading (see, Pomerantz 1984). Hence, in the environment of a positive evaluation directed at the hearer, the system of congruent responses operating is as in Figure 9. In (11), we have a negative evaluation of the hearer, that is, a Criticize. To disagree with or to downgrade a Criticize is to make a positive evaluation oneself, or to be defensive. This again violates the social norm of modesty. Therefore, typically, a Criticize is responded to by the hearer

204

Amy Β. Μ. Tsui - disagreement

congruent response/Compliment

- downgrade - thank/appreciate

Figure 9. Congruent responses in the environment of a Compliment

making a similar negative evaluation of oneself, or by upgrading the negative evaluation. In (11), C, by saying that she once dropped the telephone, is in fact making a similar negative evaluation of herself. In the following, D's negative evaluation of C's not returning the books to the library on time is upgraded to disastrous. (20)

—*

[BCET : D : 45] C has been telling D that she has three library books out which are long overdue. D: I'll tell you this, Cathy, if I ever buy a bookshop, or own a library, I'm not letting you take any books out. C: Yeah, I know, ((laughs)) I'm disastrous.

Hence, in the environment of a negative evaluation directed at the hearer, the system of congruent responses operating is as in Figure 10.

- agreement (similar negative evaluation) congruent response/Criticize - upgrade Figure 10. Congruent response in the environment of a Criticize

5.3. In the environment of a speaker-directed

evaluation

In (12), we have a negative evaluation of the speaker herself, that is, a Self-denigration. To agree with or to upgrade a Self-denigration is to make a negative evaluation of the speaker. This is face-threatening and is usually avoided unless the interlocutors know each other very well, as exemplified in (21). (21)

[Fieldnotes] Husband: I've put on a lot since you left. Wife: I know.

The interpenetration

of language as code and language as behavior

205

The hearer's agreement with the speaker's Self-denigration is socially acceptable because they are husband and wife. If their relationship were not so close, a disagreement like You haven't put on an inch, or a downgrading like A little bit, but you look good would be in order. Therefore a Self-denigration is typically responded to by a contrary evaluation, as in (12), or by downgrading the negative evaluation, as in (22) (see, also, Pomerantz 1984). (22)

—•

[Β : Β : A : 3] H: My God, ten years, I can't believe it. It doesn't seem that you were away that long. Maybe it just shows that I'm really getting old. X: Eh, listen, Henry, that's what's happening to all of us. H: Yeah.

In (22), by saying that aging is something that is happening to all of us instead of just to Η, X is alleviating the negative aspect of H's Selfdenigration. Hence, there is a system of congruent responses operating in the environment of a Self-denigration: - disagreement (positive evaluation) congruent response/Self-denigration - downgrade Figure 11. Congruent responses in the environment of a Self-denigration

In (13), we have a positive evaluation of the speaker herself, that is, a Self-commendation. To disagree with or to downgrade a Self-commendation is to make or imply a negative evaluation of the speaker, which is again very face-threatening. Typically, a Self-commendation is responded to by a similar positive evaluation of the speaker or by upgrading the evaluation. In (13), the speaker's Self-commendation is agreed with by the hearer. In (23), M's evaluation of himself as not a bad cook is upgraded to an excellent cook by A. (23)

[Fieldnotes] M: I'm not a bad cook, am P. A: I think you're an excellent cook.

In other words, in the environment of positive evaluation which are directed at the speaker him/herself, there is yet another system of congruent response operating.

206

Amy Β. Μ. Tsui

congruent response/

agreement (similar positive evaluation)

Self-commendation - upgrade Figure 12. Congruent responses in the environment of Self-commendation

As we can see from the above discussion, in the environment of different kinds of Assessment, there are different systems of congruent responses operating. It is interesting to note that the constraints on typical or congruent responses imposed by Assessments directed at the speaker or hearer are greater than those imposed by Assessments directed at a third party. This is because evaluations directed at a third party are less facethreatening than those directed at the speaker or hearer. 7 Therefore, there are more options available in the former than in the latter.

6. System network of facets In the above discussion, we have seen how the options in the semantic network are actualized as behavior in real life situations. We have also argued that they are real options in the social system by showing that the different speech functions, actualizing different combinations of options, prospect different responses. We have left out the choices available in the system operating at "facet" evaluated in the discussion because they do not affect the kind of speech function performed. This is not to say that the facet evaluated is unimportant. When we look at the linguistic exponents of Assessments, we find that they necessarily involve facet evaluated. Let us therefore examine the system network operating at facet. Manes (1983: 98), in a study of how Compliments reflect and express cultural and social values, points out that since Compliments are routinely exchanged even between people who know relatively little about each other, the aspects that are complimented on must be those which are generally accepted by society as positive. She discovers that the most commonly evaluated aspects in the Compliments in her data are: personal appearance, new acquisition, and good work. Similarly, in Assessments, the facets that are evaluated are those which are generally accepted by members of the society as worth evaluating. For example, attitude towards work, as in (8) He's very professional and in (13) I take what I'm doing very seriously.; physical appearance, as in

The interpenetration

of language as code and language as behavior

207

(19) cute looking little guy; performance in work, as in (16): he never disappoints, Graham Greene; personality, as in (15) he's a very nice person; etc. We may draw up a system network of facets evaluated as in Figure 13. voice clothes specific

- hair-style weight ^ age

ι- personal appearance

- general performance facet

- work attitude personality competence

Figure 13. System network of facet (the options given here are by no means exhaustive)

7. Linguistic exponents of Assessments Let us now examine the linguistic exponents of the different subclasses of Assessment in the data presented and see how they reflect options in the semantic network. Table 1 shows examples of linguistic exponents of the various combinations of semantic choices in the data presented. On the basis of the linguistic exponents of Assessments in the conversational data presented so far, we may say that the congruent realizations of Assessments are relational clauses of the attributive type. The feature "target" is realized by the Carrier. In the data, all Carriers are persons; they are all choices from the personal pronoun system. The Carrier in the -S/H option can be an object. For example, This picture is lovely, This concert is marvellous. It can also be a facet of the speaker, the hearer, or a third party. For example, your hair, my work, their work, John's

208

Amy Β. Μ. Tsui

Table 1. Linguistic exponents of semantic choices Semantic choices

Speech functions

Linguistic choices

Target: -S/H Facet: work: attitude Kind: positive

Praise

Target: -S/H Facet: personality Kind: positive

Praise

(15) He's a very nice person.

Target: -S/H Facet: personal appearance: general Kind: positive

Praise

(17) He's a cute looking little guy.

Target: -S/H Facet: work: performance Kind: positive

Praise

(16) He never disappoints, Graham Greene.

Target: -S/H Facet: work: performance Kind: negative

Disapproval

(14) They definitely inflate their findings.

Target: -S/H Facet: competence Kind: negative

Disapproval

Target: Η Facet: work: performance Kind: positive

Compliment

(10) You are in a hell of a good

position.

Target: Η Facet: personality Kind: negative

Criticize

(11) You would drive anybody round the goddamn fucking

half way bend.

Target: Η Facet: personal appearance: specific: voice Kind: negative

Criticize

(19) You sound terrible, you sound like a man.

Target: S Facet: work: attitude Kind: positive

Self-commendation (13) I take what I'm doing very seriously.

Target: S Facet: work: performance Kind: positive

Self-commendation (23) I'm not a bad cook, am P.

(4) He's a serious scholar. (8) He was very professional about it.

(9) You have decisions, such high level decisions like for couple hundred dollars.

The interpenetration

of language as code and language as behavior

209

Semantic choices

Speech functions

Linguistic choices

Target: S Facet: personality Kind: negative

Self-denigration

(12) I don't think I have the guts to make it the subject of my thesis.

Target: S Facet: personal appearance: specific: weight Kind: negative

Self-denigration

(18) I've put on a lot since you left.

Target: S Facet: personal appearance: specific: age Kind: negative

Self-denigration

(22) I'm really getting

old.

painting, etc., are facets of persons whereas this picture and the concert are objects. For the former, the Carrier is a choice from the noun: determiner: demonstrative system whereas for the latter, the Carrier is a choice from the noun: determiner: possessive system. Typically, the Carriers in Compliment and Criticize are realized by you and your + head noun; those in Self-commendation and Self-denigration are realized by I, we and my + head noun, our + head noun; those in Praise and Disapproval are realized by he, she, they, it and his, her, their, its, this + head noun, that + head noun, these + head noun and those + head noun. However, there are instances where the Carriers in the speaker-directed or hearer-directed Assessments are realized by a demonstrative plus a head noun, or a third person (singular or plural). For example, This piece of music is wonderful can be a Compliment if the piece of music was written by the hearer. David is a cute looking little guy can be a Compliment if David is the hearer's son. In other words, the positive evaluation of the piece of music and David is indirectly a positive evaluation of the hearers. There are also instances where the Carriers in Assessments which are directed at a third party are realized by the second person pronoun you, usually in a plural sense. For example, in (9), You refers to the university administration rather than to the hearer. Hence, the utterance realizes a Disapproval. The Attribute is typically realized by some adjective of the attitudinal class (Halliday 1973: 75). For example, serious in (4); professional in (8);

210

Amy Β. Μ. Tsui

nice in (15); terrible in (18); cute in (19); and not bad in (23). In congruent realizations, adjectives describing qualities such as serious, professional, nice, cute, etc., realize positive evaluations. Those describing qualities such as terrible realize negative evaluation. The choice of adjectives is very much determined by the facet evaluated. For example, if the facet work is evaluated, then adjectives like serious, good, outstanding, etc., are likely to be used. If the facet age is evaluated, then the adjectives young, youthful, etc. are likely to be used. When the adjectives are not used in the literal sense, whether they realize positive or negative evaluation is context dependent. As I have pointed out previously in this chapter, the adjective gorgeous, whose superordinates are nice and good (see Collins Co-Build Dictionary 1987: 629), can be used sarcastically to mean awful or ugly. While the congruent realizations of Assessments are relational clauses of the attributive type, there are some which are realized by material processes. The evaluative element, instead of being conveyed explicitly, Table 2. Congruent realization statements Semantic option

Congruent realization

Assessment

Clause: relational: attributive

Target

Carrier

-S/H

Carrier: person = 3rd person pronoun (sing./pl·); Proper noun; 2nd person (pi.) Carrier: object: person-oriented = 3rd person possessive + noun object-oriented = demonstrative + noun

he, she, it, they, Mary, John; you\ his work, Mary's

Carrier: person = 1st person (sing./pl.); object: person oriented = 1st person possessive -I- noun; object-oriented = demonstrative + noun

I, we; my article, our house

Carrier: person = 2nd person (sing./pl.); object: person-oriented = 2nd person possessive + noun

You;

S

Η

object-oriented = demonstrative + noun Evaluation

Linguistic exponent (example)

This picture.

That painting

Your writing This piece of music

Attribute

Positive

Attribute = qualitative Adjective

Negative

Attribute = qualitative Adjective

hair

nice, serious, professional, cute. terrible,

stupid.

The interpenetration

of language as code and language

as behavior

211

is implied by the value judgement attached to the Process in one participant clauses or the Process, Goal, and sometimes Circumstance in two participant clauses. For example, inflating one's findings in a piece of research is generally considered to be dishonest. Hence, to say that someone inflates their findings implies a negative evaluation. Putting on weight is something negative in western societies. To say that somebody has put on weight implies a negative evaluation. However, we must also take into consideration the specific context of situation. If the utterance I've put on a lot since you left were spoken by a husband to his wife who was worried that he might not be able to eat properly while she was away, then it implies a positive evaluation of the speaker himself, hence a Self-commendation. The congruent realization statements associated with the features in the semantic network of Assessments can be presented as in Table 2.

8. Concluding Remarks In this chapter, I have examined a class of conversational acts, Assessments, and I have outlined the semantic options which lie behind them. I have also investigated the actualization of these options in real life situation by examining some conversational data. Although I have only been able to account for congruent realizations of semantic options and although I have only dealt with a small area in conversation, I hope I have succeeded in demonstrating that, as Halliday (1984: 32) points out, behind conversational processes lies a rich network of meaning potential and every instance of a conversational act is the manifestation of a systemic resource, and hence the study of language as code and language as behavior should be unified.

Notes 1. I wish to thank Michael Halliday, Margaret Berry and Christopher Butler for their comments on an earlier version of this chapter. Whatever faults there are in this chapter, however, are my own. I also wish to thank the Chinese University of H o n g Kong and Shaw College o f the Chinese University of H o n g K o n g for funding my trip to present the earlier version of this chapter at the 15th International Systemic Congress held in Helsinki in 1989. 2. The data are conversations between native-speakers of English, both British and American. Face-to-face conversations are labeled [A], [C], [D] and [BCET], and telephone

212

3.

4. 5.

6.

7.

Amy

B.M.Tsui

conversations are labeled [B]. Fieldnote data, labeled [Fieldnotes], are data taken down verbatim immediately after the conversation occurred. [BCET] stands for Birmingham Collection of English Texts. The author wishes to thank the English Department, University of Birmingham, for allowing her to use their data. Transcription notations: Capitalization : prominent syllable. = one utterance latching onto another [ overlapping ( ) indecipherable It should be noted that the labels given to the speech functions are labels for what has been defined rather than definitions in their own right. Here the term congruent is used in the sense of Halliday (1984: 14). According to Halliday, a congruent realization is one which can be regarded as typical. This means one which will be selected in the absence of any good reason for selecting another. Strictly speaking, an expression of appreciation like That's very kind/nice of you is a form of downgrading. In saying that the speaker is being kind, the responder is implying that s/he does not really deserve the compliment. The term "face-threatening" is borrowed from Brown —Levinson (1987).

References Atkinson, J. Maxwell —John Heritage (eds.) 1984 Structures of social action. Cambridge: Cambridge University Press. Brown, Penelope —Stephen Levinson 1987 Some universals in language usage. Cambridge: Cambridge University Press. Collins COBUILD English language dictionary 1987 London/Glasgow: Collins. Fawcett, Robin — M . A . K . Halliday —Sydney Lamb —Adam Makkai (eds.) 1984 The Semiotics of culture and language. Volume 1. London: Pinter. Halliday, Μ. A. K. 1973 Explorations in the functions of language. London: Arnold. Halliday, Μ. A. K. 1984 "Language as code and language as behaviour: A systemic functional interpretation of the nature and ontogenesis of dialogue", Robin Fawcett et al. (eds.), 3 - 3 5 . Halliday, M . A . K . 1985 An introduction to functional grammar. London: Arnold. Manes, Joan 1983 "Compliments: A mirror of cultural values", in Nessa Wolfson — Elliot Judd (eds.), 9 6 - 1 0 2 . Pomerantz, Anita 1984 "Agreeing and disagreeing with assessments: Some features of preferred/ dispreferred turn shapes", in: J. Maxwell Atkinson —John Heritage (eds.), 57-101. Wolfson, Nessa —Elliot Judd (eds.) 1983 Sociolinguistics and language acquisition. Rowley, Mass.: Newbury House.

The static and dynamic choices of responding: Toward the process of building social reality by the developmentally disordered Jonathan Fine

1. Introduction Conversation occurs naturally from an early age and is the basis for the socialization of children into families and other social groups. Conversation is situated in social environments and at the same time constructs those environments. This chapter focuses on how meaning is exchanged in conversation and how the exchange of meaning can be represented. First, the theoretical issues are outlined. These issues include (1) the paradigmatic nature of systemic choices, (2) the dynamic dimension in negotiating conversation, and (3) the use of probabilities to characterize how different speakers choose systemic options. Second, two speakers with different developmental disorders are compared in terms of their responding to an interlocutor. This empirical study draws on the three theoretical constructs of the first section to show how the speakers differ in their typical ways of building on the contributions of their interlocutors in conversation. The conclusion addresses both the social and cognitive implications of the methods for studying conversation. In conversation, meaning is built up by the cooperation and interaction of the speakers. They speak in the context of each other and what is said is interpreted in that context (see Edwards —Middleton 1986 b). Conversational sequences of language are produced one at a time in order. Thus, an account of conversation must show the temporal organization and the meaning that could be and is conveyed at each point in the conversation (Martin 1988 a; compare also, Fawcett —van der Mije —van Wissen 1988). As Ventola (1989) points out, conversations are variable and this variation must be represented in an account of conversation. Studying the variation can, in fact, be a method of systematically relating context to the language used to realize conversation. Since conversation is built up interactively, an account of the sequences in conversation and the dependence of one sequence on another requires that the variation be

214

Jonathan

Fine

studied in terms of an ever-changing and being-constructed context (see, Ventola (1989) for a discussion). Butler (1985 a) presents an account of conversation based on paradigmatic relations representing meaning. Realization rules are used to relate the options chosen from the paradigmatic systems to structures that represent possible syntagmatic relationships. The present study investigates the variation in syntagmatic relationships and non-structural as well as structural dependencies, and concludes that a dynamic dimension is necessary for the description of conversation. The negotiation of meaning in context is ultimately related to larger issues of inter-textuality (Lemke 1989) and the cognitive and social uses of conversation (Edwards —Middleton 1986 a, 1986 b). The variation in the language structures used and how they relate to context can inform the studies of register and genre potential (see, Matthiessen (1988) for an overview and comparison of approaches). An adequate account of conversation should be able to not just show that there are differences in the conversations but explain why the impairments in social functioning are related to the kind of conversation that is produced. For example, this knowledge of specific relations between linguistic patterns and social functioning in psychiatric patients is important for the understanding of the social misfunctioning and how such misfunctioning is ultimately explained etiologically and in terms of classes of patients. There is unfortunately little work in this area (see, the reviews in Fine 1988 b, 1989). One objective of the following study is to present a clear methodology that can describe the variation in diagnostic classes, and relate that variation systematically to the social construction of reality through conversation. Underlying the analysis of language used by different speakers is the need for considering context. This is not just a general principle of looking at language in its social setting, but extends down to the analysis of units of conversation. The Malinowskian and Firthian insights put language into context on the levels of "context of culture" and of "semantics" and "phonology". It is in context that possible meanings can be determined and the meaning of a specific utterance is determined. The analysis of conversation by ethnomethodologists has attempted both to formalize the mechanisms of conversation and to describe the social functions of those mechanisms. Through their social functions, the speaking roles of speakers can be detailed (e.g., Sacks —Schegloff—Jefferson 1974: Schegloff—Jefferson —Sacks 1977). Modern systemic linguistics sees language as a set of resources. This set of resources is, however, sensitive to the

The static and dynamic choices of responding

215

context where language is used, and the realized options help in turn to construct the context for the next use of language. Keeping a sensitivity to context in studying the use of possible meanings from systematically arranged options permits the study of patterns of meanings used by different speakers under different circumstances.

2. Theoretical approach A number of strands of theorizing in systemic linguistics can be profitably applied to conversational analysis. As speakers continue in a conversation, they display their sharing of the encoding and the code but can also influence each other. Moving forward in a conversation requires that the earlier parts are considered and reacted to. Each speaker will presume that the interlocutor will take each new contribution into consideration and will keep track of the earlier contributions. The three components of systemic theory needed to capture these aspects of conversation have been mentioned in the introduction and need to be described: the paradigmatic nature of systemic networks; a dynamic dimension; and the use of probabilities in the analysis of texts. In addition, the link to social context must be outlined to situate the analysis of texts in a wider framework. Fawcett —van der Mije — van Wissen (1988) deal with a model of conversation that is social and interactive and which could be a cognitive model as well. They note that language is a means for problem solving. This perspective can be coupled with that of Edwards —Middleton (1986 b) where the language of conversation is seen as building up joint memories. Moves of challenge and agreement may take place to verify the joint memories. It is through such processes having to do with conversation and memory that social realities are established and continue to develop. Different speakers, and classes of speakers, may construct different realities and enter into different situation types, differences revealed in conversation. The connection should not be of a correlation of an arbitrary linguistic feature with an arbitrary element of a situation type or class of speaker. Rather it should be possible to explain the correlation in terms of the function of the linguistic feature and characteristics of the situation and speaker. From a cognitive perspective, there again should be non-arbitrary relations between the operations needed to create elements of conversa-

216

Jonathan Fine

tion in context and (1) the abilities of classes of speakers and (2) the situations the elements are operating in. An analysis in terms of cognitive processes needs to be done of the mechanisms proposed by systemic theory to underly conversation to demonstrate the specific connections between cognitive processes and features of conversation. 2.1. Paradigmatic

relations

The first of the elements of systemic theory to be discussed is the paradigmatic nature of systemic networks. Systemic networks are to be comprehensive in representing all the possible choices given an entry condition (Halliday 1984). The choices then set up the possible meanings by contrasting one choice with another. Evidence for the reality of the differences among the choices can be provided in part by examining the uses of these systems under conditions that one has reason to believe restrict or constrain the choices (see the work on language development and psychiatric language in Martin 1983; Rochester —Martin 1979; Fine 1988 a; Fine - Bartolucci - Szatmari 1989). The paradigmatic networks do not exist in isolation. Rather they represent what has been socially encoded as relevant meanings. These accepted ranges of possible meanings form a background for the interpretation of speech in particular cases. Combinations of choices may typically be made, and so constitute a patterning accounted for at a higher level of abstraction, register or genre perhaps (see, Halliday 1984; Matthiessen 1988; Ventola 1989). Paradigmatic systems must be preserved as a way of making the description of conversation semantically based. However, to keep the description anchored in the social circumstances and to show the unfolding nature of conversation, meaning must be represented in other ways as well. For example, the identity of the speaker both in terms of social role in the society and in terms of the role in a specific conversation has a bearing on what is meant. There are generalizations to be made both about what certain groups of speakers typically do in constructing actual conversation and how the choices from systems are typically combined. The paradigmatically stated resources (usually interpreted as meanings) will have to be seen in a wider context that includes dynamic modeling of conversation and probabilities that also help to account for the meaning of speakers' contributions to dialogue.

The static and dynamic choices of responding

2.2. Dynamic

217

dimension

Dynamic modeling in linguistics has been proposed by Martin (1985) for a number of reasons. First, the sequences of units in discourse are quite variable and not easily accounted for by structures that are characterized as marked or unmarked. Second, in discourse, and specifically in conversation, there is recursion of non-ranking units; and third, aborted structures are possible in discourse and are well-formed units in themselves. If the representation of dialogue is to capture meaning by choices in a network, then provision must be made for recursion or "end" options (see, Bateman 1989; Fawcett —van der Mije —van Wissen 1988). However, these options have meaning in themselves. Ending a certain conversational pattern (for example, an invitation or the provision of supplementary information in response to an answer) conveys social meaning. What social meaning is conveyed (and is conveyable, thinking of the static potential; see, Martin 1985) depends on the identity of the speakers involved. Aborting an invitation sequence has different social consequences depending on the relative statuses and social relationships of the individuals involved. The meaning of a stretch of language in conversation is partly a matter of the choices from static systems but also partly a matter of the application of dynamic choices made in the context of the developing conversation. The options that are available to specific speakers at given points in conversation depend on the earlier options selected and on the identity of the speakers. It is this complex notion of context (including specifically co-text) that is involved in determining the contrasts available to the speaker and therefore the meaning of the option chosen. In the spirit of seeing text as a semantic unit and not as a large structural (syntactic) unit (Halliday —Hasan 1976), it is considered that the meaning of stretches of language in conversation cannot be represented as independent higher level structures. If this latter approach were taken, then it would not be possible to make generalizations about the meaning of the transition from one move in conversation to another except as they are typically or atypically associated in specific structures. As Martin (1985) and Ventola (1987) point out, it would then be difficult or impossible to account for phenomena such as aborted sequences and recursion that do have functional value. As Fawcett —van der Mije —van Wissen (1988) indicated, though, it is important to preserve the essential paradigmatic characteristic of networks in order to continue to account for meaning. If the operation of the dynamic system is appropriate to the context, then text is formed. Various degrees of non-text may be the

218

Jonathan

Fine

result of choices made at points provided by the dynamic system that do not continue the meaning of earlier stretches of language. Different methods of continuing meaning in conversation which are employed by the two speakers examined will be discussed in Section 3. Having outlined the general characteristics of dynamic modeling, a particular case of conversation will be examined. Speakers start with more or less similar static potentials (as characterized by systemic networks). However, as conversation develops the constraints on successive stretches of language are imposed by the choices made by prior speakers. The constraints need not be (and, perhaps, probably are not) of a structural kind. Butler (1985 b: 93) quotes Berry (1982) as suggesting that syntagmatic relations are needed to explain why "only part of the meaning potential can appropriately be activated at any particular point in the structure of discourse". For the reasons put forth by Martin (1985) and for those of probability outlined in Section 2.3., the syntagmatic relations in conversation are better seen, in many cases, as stemming from dynamic systems rather than representing structures with their implied constituents (for discussions of structure, see, Halliday — Hasan 1976:1 —10; Matthews 1981: 71 - 9 5 ) . In examining conversation as a social, and not just a linguistic activity, we must consider what is accomplished (and again, in more static terms, what is accomplishable) by speakers. Instances of conversation are realizations, and the realizations may exhibit certain relationships or even dependencies, but a full account must also include how those dependencies arose. This leads one to examine what speakers can decide to do as they choose meaning options and follow the realization rules in a context that changes with each additional stretch of language. The control of choices from networks has a number of sources, including the tactics of interaction (as represented for example by flow charts, see Martin 1985; F a w c e t t - v a n der M i j e - v a n Wissen 1988; Ventola 1987, 1989), social aspects of the situation, and cognitive factors (the latter is mentioned, but not developed in Fawcett —van der M i j e - v a n Wissen 1988). The goal should be to model the process of choosing as well as the choices made available by language. In modeling the choosings, a wide array of influences on how individuals use the potential of language is considered. This is not just a competence-performance distinction, since language must in principle be considered to be the product of the interaction of the abstract potentials available and the social and cognitive situation of its use and users. The dynamic dimension reconnects the static potentials of language to the interactive context of its use. The actual use of language

The static and dynamic choices of responding

219

creates social interaction, which in turn is a constraint on how and what speakers can mean and on how any given stretch of language will be interpreted. The actual is thus a factor in influencing the potentials at a later point in the conversation. Bateman (1989) has investigated the use of dynamic systems in grammar from a computational perspective, and a number of the technical issues raised have implications for the study of conversation. He notes that system networks require a traversal algorithm to give meaning to the networks. Such a proposed algorithm requires a number of characteristics including co-selection constraints and markedness conventions to state constraints on the selection of features. These constraints imply a constraint on meanings thus accounting for "holes" in the paradigms generated (see, also, Nesbitt —Plum 1988). From the point of view of the listener, these constraints represent what is not expected and therefore carry unusual meaning (perhaps also grammatical marking, depending on how the realization rules operate on the features selected) when violated. The question of how the meaning of speakers can be represented in dynamic and static networks needs to be examined. Bateman (1989) states that different paths through the dynamic potential network will result in different co-occurrence constraints between synoptic features. If this is the case, then the dynamic network could be representing the expectancies of the listener in conversation and can carry the difference between usual meaning and unusual meaning. In Bateman's approach, choices in the dynamic systems affect later choices (traversal spaces with predetermined constraints). Extending this to conversation, we would want to specify the constraints as probabilities and to allow the termination of a developing sequence (accepting Martin's (1985) argument) as an option that may also be assigned a probability. In investigating the conversational ability of different speakers, it is then possible to study the different paths that select features from the same sets of synoptic systems. Such differences should then represent differences in the meanings of the speakers. "Meanings" here must necessarily apply to how the speaker is interpreted in conversation and depends in part on the expectancies of the hearer. These expectancies are in terms of structures, but also in terms of looser patterning characterized by probabilities in the development of sequences and the selection of combinations of features. The possibility of recursion creates a problem in how meaning is represented in dynamic and static systems. Martin's (1985) claim is that recursions (as well as interruptions) must be provided for. Bateman (1989)

220

Jonathan Fine

allows for recursion in the interpretative process and characterizes reentering a system as re-opening an area of linguistic potential. The point where recursion is possible is a specification of the context for such a potential. A difficulty here is in characterizing the effect in conversation (and probably also in the lexicogrammar) of successive recursions. Repetitions of I don't want to talk about that are not merely responses in particular contexts. The accumulation of five or six such statements starts to indicate to the interlocutor that the nature of the conversation is different from what it was before such a string of repetitions. It would be difficult to argue that the fifth or sixth iteration represents a different structure (compare in lexicogrammatical terms: The very big house, The very, very big house, The very, very, very...very big house). Rather, there must be devices associated with dynamic systems for keeping track of the number of iterations and for accounting for differences in meaning based on the number of interactions expected, given the context in the dynamic network and the wider context of situation. Bateman (1989) makes the point that the dynamic dimension has its own rules and regularities. One of these kinds of rules must be the keeping of a history of choices, not just of features from different systems, but also of recursions through the same system. As Bateman points out, rewiring networks to allow recursion alters the contrastive distribution of options in the network. The examples of the effect of multiple recursions also suggest that the ultimate meaning is not merely a succession of the same features selected repeatedly. This kind of recursion and the meanings it leads to need to be accounted for outside of the synoptic networks. It seems that a historical register in the dynamic dimension is the appropriate place to specify the recursions and their effect on conversational meaning. The dynamic system would then indicate that reopening a synoptic network for the second time is not the same (means something different) as reopening it for the fifth time. In terms of description of language it is important to capture the effect of the second iteration compared to the fifth in a motivated way.

2.3.

Probabilities

In order to account for conversation, the paradigmatic systems and dynamic dimension need to be complemented by a probabilistic element. Introducing probabilities links the conversation to its social sources, permits the description of different intra-organism states, and is useful

The static and dynamic choices oj responding

221

in showing group variation. The description of probabilities also can show the effect of a number of sources of variation on a single linguistic variable (see, for example, Walters —Wolf 1988). These sources of variation may be inter-organism (Fine —Bartolucci —Szatmari 1989) or intraorganism (Bartolucci — Fine 1987). In all cases, though, changes in probabilities on options can be seen as contributing to the meaning of the conversation. That is, for any particular social context there will be expectancies that can be stated as probabilities. Not following such probabilities minimally sends the alarm message of "this is an unusual message". The probabilities help to state the statistical mismatch of the options and the situation. Since the options specify meaning, the statistical mismatch is an unusual mapping of meaning onto the social situation. Grammatical marking also carries such meaning (a structure that is marked is used since there is a specific reason not to use the unmarked structure), but probabilities account for expectations of what should occur even when the specific distinction of marked and unmarked structures cannot be made (for a study of probabilities in interconnected grammatical systems, see Nesbitt —Plum 1988).

3. Analysis of conversation The method of analysis discussed in Section 2 is applied to two conversations with psychiatric patients. The patients differ as to diagnosis. The analysis shows specifically how the speakers select options from systemic networks and the importance of viewing their choices dynamically and probabilistically. The analysis of conversation has been based in the first instance on Halliday's (1984) networks for speech function. These networks represent the highest level of the linguistic system: semantics. In particular, the analysis dealt with the "turn" system at a number of levels of delicacy. Table 1 lists the categories involved. In addition, turns have been classified as immediately following the turn upon which they are dependent (in all cases, responses being dependent on initiating turns) or as a "not first" turn. The turn system, then, has provided a simple but semantic means of studying conversation. It may be possible to modify the system and to clarify some of the options and how they are related to linguistic realizations. However, the system as it stands is usable, and can show the points about dynamic systems and probabilities that are being made here.

222

Jonathan Fine

Table 1. Speech functions — turn system (adapted from Halliday 1984) Turn initiating whyes — no declarative responding direct won't respond will respond doesn't know knows determinate yes no response to wh-question response to alternatives response to list indeterminate either yes or no both yes and no indirect commentary disclaimer supplementary confirming

The data analyzed are conversations of two subjects with a psychiatrist. The conversations were taken from a larger study of Asperger's Syndrome (see Szatmari — Bartolucci — Bremner 1989 for details). One subject was diagnozed as being a high-functioning autistic and the other as a case of Asperger's Syndrome. The interviewer was the same in both cases. The subjects were selected as being clinically typical of the diagnostic categories. Both high-functioning autism and Asperger's Syndrome involve difficulties in social communication, and are apparent early in a child's social and verbal development. Although these clinical categories may not be based on specific etiologies or definitive characteristics, they do represent socially meaningful distinctions among groups of subjects. In future, the methods being developed here could be used to understand the different ways that such groups of psychiatric patients communicate or fail to communicate. The subjects tended to give short answers to the questions they were asked, but, as will be described shortly, there were a variety of speech

The static and dynamic choices of responding

223

functions used by both the interviewer and the patients. At the end of each interview, the psychiatrist asked a series of questions about the subject's symptoms; for example, Have you ever had any problems with your hearing?', What about your eyes? Do they bother you?', Funny taste in your mouth? This section of each conversation is called the "Examination register", and data from it are described separately when relevant. Since there were only two participants in the conversations, with specific and known roles, the coding of the speech functions from the systemic network was straightforward. There were few cases of doubt and these could be resolved by listening to the recordings of the interviews. Some of the results of this analysis will be presented in detail and related to the principles outlined for the analysis of conversation. In a number of cases it is the differences between the two conversations that is interesting. The first speech function to be examined is the response-confirmation. This function is used to acknowledge that the speaker correctly understands the interlocutor, in either local terms of a phrase or a word, or in more general terms of the point being made. An example of responseconfirmation is given in (1). (1)

A: Let's see there's you and do you have brothers and sisters I've forgotten. B: No. A: Just you. (response-confirmation) B: Um hmm. (response-confirmation)

In almost all cases in both interviews, one response-confirmation is followed by a second response-confirmation. This sequence has the effect of ending a topic with the consequence that the speaker of the first response-confirmation can introduce a new topic in the third turn in the sequence. The presence of a second response-confirmation in the turn immediately after a first response-confirmation seems to establish a structure with the specific function of "the current information is confirmed to be shared." With this agreement made clear, a new piece of information can be opened for negotiation. In conversations where the speakers have more equal social roles, it may be likely that the speaker of the second response-confirmation introduces a new topic. The near certain probability of a second response-confirmation following a first response-confirmation, the clear function of the adjacent response-confirmation turns, and the necessity of the second response-confirmation to achieve the function of topic boundary, all argue for treating the sequence as a

224

Jonathan Fine

structure. It is interesting to note that this structurally defined sequence does not discriminate between the two speakers. Both speakers use the sequence in the same way with about the same frequency. This is in contrast to some other sequences to be discussed. In the conversations analyzed here, as in many conversations in psychiatric settings, the interviewer asks questions to start and also frequently to sustain the conversation. Thus a pattern of functions is established where the interviewer initiates and the patient responds. The most frequent initiating function is the wh-initiating function. Canonically the expected speech function in the turn of the patient would be a whresponse which provides the information requested in the initiating turn. In these interviews the "goods-and-services" option in Halliday (1984) is not relevant. As well as the information requested, the respondent may provide other relevant supplementary information. It is this function of the supplementary response that distinguishes the two speakers. The autistic patient very frequently adds supplementary information whereas the Asperger's patient rarely does so. In fact, the autistic patient frequently supplies supplementary responses three turns beyond the wh- initiating function. Example (2) shows how supplementary response can be added in successive turns. (2)

Interviewer: Patient:

Interviewer: Patient: Interviewer: Patient:

How do you think oh what do you think your mother's choice has been was it a good choice or? Well my mother said that she loved my father and still does and I do believe that my father does look after my mother, (supplementary response) Um hmm. (response-confirmation) And as well my father seems to be a good father to my sisters, (supplementary response) Um hmm. (response-confirmation) But I've never been able to have a strong relationship with my father because our beliefs our standpoints are varied and my father himself said... (supplementary response)

A mechanism is thus required for recycling through the response system to produce the sequences of supplementary responses. Of course, the freedom to continue to add supplementary information is partly dependent on the interviewer's turns. In (2), the interviewer is merely acknowl-

The static and dynamic choices of responding

225

edging the supplementary information and not attempting to stop its presentation in following turns. The patient is allowing the interviewer to provide acknowledgements in part by intonation and a slow rhythm. The sequence of turns produced has weak constraints and is dependent on the meshing of the options of the two speakers. The autistic subject frequently supplies supplementary information, and does this in sequences of turns after an initiation by the interviewer. The Asperger's subject rarely (four times in the non-medical part of the interview) adds a supplementary response and never places the supplementary response beyond the turn adjacent to the initiation. In conversation not only do the speech functions of participants have to be accounted for, but the conversational locations must be specified. For each type of initiating function (wh-; yes-no; declarations), the location of the responses was identified as either in the turn adjacent to the initiating turn or in a later, non-adjacent turn. Table 2 gives the percentage of response functions that are found in non-adjacent turns for each kind of initiating turn and for each speaker. Table 2. Percentage of response functions found in non-adjacent turns, by type of initiation function and by speaker

initiating whmitiating yes —no initiating declarative

Autistic

Asperger

59 33 32

53 9 25

Table 2 shows that there is less non-adjacent responding by the Asperger subject compared to the autistic subject but the difference is unlike the almost absolute difference in the supplying of supplementary information. That is, the control of where to put response information seems to be different from the control of what kind of information to supply. Figure 1 presents a system network for the speakers' responses to the different options of initiating. The figure shows the variable rates of responding in a turn that is not adjacent to the initiating turn. As well as a generally greater placement of responding in non-adjacent turns by the autistic subject, the figure shows that the difference between subjects depends on the type of initiation that precedes the response. The autistic subject is most different from the Asperger's subject after yes-no initiations.

226

Jonathan

Fine whAutistic Aspergers

initiating

yes-no Autistic Aspergers

declarative Autistic : : : : : : Aspergers === = = Figure 1. Proportion of responses that are non-adjacent to the given types of initiating functions for each speaker

After an initiating function, the response may be a direct response (the supplying of the precise information required) or an indirect response (supplying other information). This choice of direct or indirect response (see, Halliday 1984) may also be a method of distinguishing speakers and may be sensitive to aspects of context and speakers that are different from the other mechanisms discussed. The autistic speaker uses the indirect responses of response-commentary, response-disclaimer and response-supplementary (implicating an answer) occasionally after each type of initiating speech function without first using a direct response. Examples (3) —(5) illustrate the pattern. (3)

Interviewer:

Patient: (4)

Interviewer: Patient:

(5)

Interviewer: Patient:

Yes - well you seem to be oh fine person in what way do you feel you're not on the usual track? (initiating wh-) Well. ..it's a bit that's somewhat of a difficult question to answer, (response-commentary) In what way are you and your father different? (initiating wh-) Well...I don't know the best way to start that. (response-commentary) I see - what do you read beside the scriptures? (initiating wh-) I'm not into non-fiction. Well I'll start that again. I'm not into fiction because it just doesn't interest me at school. I read novels because it's part of the course but I don't do so voluntarily (responsesupplementary)

The static and dynamic choices of responding

227

The Asperger's subject never uses indirect responses in the regular conversation without first producing a direct response. The indirect responses display the speaker's management of conversation and convey the expectation that the interlocutor will understand the implications of the indirect responses. Figure 2 presents a system network for responding. The options are presented at a low level of delicacy with just [DIRECT] and [INDIRECT]. For each of these options, the proportion for each of the two speakers is given. The autistic subject chooses [DIRECT] less frequently and [INDIRECT] more frequently than the Asperger subject. Conversation is dynamic, however, and the routes back to the entry condition for the responding system indicate that, after an [INDIRECT] option is chosen, the speaker can re-enter the responding system to add another response. (In fact a similar route back to the beginning of the responding system can be drawn from the [DIRECT] response option. Recycling after direct responses takes place for both speakers but will not be elaborated on further.) The percentages before the route recycling into the responding system indicate the rate of choosing to add another response. After 50 percent of his indirect responses, the autistic subject re-enters the system to add another response. In contrast, the Asperger subject never re-enters the responding system after an indirect response. This subject chooses the [END] option 100 percent of the time. The difference between the two subjects is not just in the rate of indirect responses (the autistic subject more than the Asperger subject) but also in the probabilities of re-entering the response system after indirect responses. [DIRECT] Autistic Aspergers

ι—> Responding —>

U

[INDIRECT] Autistic : Aspergers =

50% 100% — end

Examination register" Autistic 0 Aspergers

33%

Figure 2. Proportion of [INDIRECT] versus [DIRECT] choices

228

Jonathan Fine

The only occasions when the Asperger subject uses such indirect responses are at the end of the conversation when the interviewer has been asking many medically related questions (for example, What about your eyes?; Do they bother you?·, Do you see anything (unusual) at all?', Funny taste in your mouth?·, Funny feelings in your body in any way?) Within the "Examination register" the Asperger subject produces indirect responses, such as in (6) and (7). (6)

Interviewer: Patient:

(7)

Interviewer: Patient: Interviewer:

Just these voices once in a while? What do you think they are? (initiating wh-) I think they 're. I hear them they remind me of the people on soap operas the a they talk gossip, (response-supplementary) So soap operas are something you do watch uh? Um Hmm. Which one is your favorite? (initiating wh-)

The options chosen in the "Examination register" are shown separately in Figure 2 and can be compared to the earlier part of the interview. In this register, the autistic subject never responds indirectly (compare this to his reasonably frequent indirect responses earlier in the conversation). In contrast, the Asperger subject greatly increases his rate of indirect responses. Furthermore, the Asperger subject changes the rate of reentering the responding system after an indirect response. In the earlier part of the conversation, the Asperger subject never chose a second response option after an indirect response, but in the "Examination register" he re-enters the responding system after one third of his indirect responses. The change in register has altered the probabilities for choosing the [INDIRECT] option from the responding system and also the probabilities of recycling through the responding system to add another response. The change of probabilities was not uniform for the two speakers. Rather, the speakers changed the pattern of options in different ways when the register changed. Furthermore, the differential changes took place in both the synoptic choice of [DIRECT] or [INDIRECT] and in the dynamic dimension represented by the recycling through the responding system. The indirect responses seem to be controled by the kind of speaker (in this study the difference between the autistic and Asperger subjects is clear), and the register. A full description would have to specify the probabilities for direct and indirect responses at particular points in conversation. These probabilities would have to taken into account the

The static and dynamic choices of responding

229

identity of the speakers, the social relationship between them, and a history of functions in the developing conversation. Information about adjacent turns is not sufficient, and it seems that the dependence of nonadjacent turns is quite loose. Nevertheless, there is social meaning in the difference in choice between a direct and an indirect response. A summary of the data and their implications is presented below in (a)-(d). a. There are sequences in conversation that are tightly constrained and that carry specific meaning, the confirmation — confirmation sequence. All components of the sequence must be present to achieve a given meaning. That is, there is a structural relationship. b. Responses to initiating functions may be expressed in terms of a single instance or a sequence of supplementary responses. There must be a device for allowing successive selections of supplementary responses after initiating functions. The use of such a device depends on the speaker and leads to a non-structural relationship among the functions produced. c. Responses may be direct responses to initiating functions or they may be indirect. Indirect responses give supplementary information, comment on the initiation, or evade an answer (response — disclaimer). These indirect responses provide metacomments on the organization and status of the conversation. d. Responses may be realized immediately adjacent to initiations or realized somewhat later in the conversation. For this reason, a device that keeps track of the functions produced and their speakers is needed. The probabilities of the different response options are dependent on the development of the conversation at the entry into the responding system and upon speaker and register characteristics.

4. Conclusion This study has proposed some methods for understanding how conversations work. The three chief characteristics are paradigmatic networks that represent contrast in meaning, dynamic systems that account for the building of sequences, and probabilities that account for the expectations of participants in conversation. Underlying this study of conversation is the objective of understanding the conversational ability of speakers in social contexts. The study of conversation must then be sensitive to the context of the conversation and the development of that context by the conversation itself. By comparing different types of speakers we can see

230

Jonathan

Fine

how the systems and probabilities for conversation are exploited differentially. That is, the method of analysis is shown to be relevant to social differences in carrying on conversation. The data show that choices speakers make from the systems may pattern similarly and so reinforce each other in characterizing types of speakers. For example, the autistic speaker, in contrast to the Asperger speaker, is characterized by both using supplementary indirect response and on other specific occasions answering yes —no initiatives with sequences of declaratives. From the wider perspective of conversation as building social reality, the analysis of conversation must show that construction of conversation represents the mutual affirmation of the rules of conversation and the substance of conversation. Conversation is an indication of agreement that the text under construction is a shared social entity, that is, that can serve as the basis for further sharing of views of reality. Edwards — Middleton (1986 b) stress the fact that conversational discourse realizes joint remembering and develops social cognition-understanding of the world that is shared. There is also a cognitive side to this account of conversation. The characteristics of paradigmatic systems, dynamic systems, and probabilities are characteristics that the human mind must process in conversation. Speakers must be able to keep track of the cycles through the systems. Sometimes the choices from the systems may build conversational structures with specific constraints, constituents, and meaning in the conversation. Other selections of choices may create more loosely related sequences for which the social effect at any one point may be difficult to specify. The mind should have the ability to process both types of sequence both in production and comprehension. It must have the ability to store information and to continually update the context and the nature of the sequences involved as each new step in the construction of conversation takes place. The processing must also take into account knowledge of the probabilities of choices at any point in the creation of conversation. These probabilities represent expectations of the speakers involved, given the local context, the wider situation and the identities of the speakers. The analysis of conversation can be used as a tool for understanding the cognitive differences between speakers. For example, similar variation in responses in two contexts (after wh-initiations and declarative-initiations) may indicate a common cause of the variation. If substantiated, this common cause can then be related directly to the social effect of the conversational pattern.

The static and dynamic choices of responding

231

The probabilities or weightings on choices from systems may be set by the specification of genre (cf. Matthiessen 1988). The assigning of probabilities from genre implies that different genres are associated with expectancies that the reader/hearer has. The writer/speaker then constructs texts in relation to the expectancies of the reader/hearer. It would be interesting to investigate whether the probabilities that characterize register and code (see Martin 1983) are different from those that characterize genre. Could a code or a register difference, stated as changes in probabilities, be mistaken for a genre difference (or a specification of a particular unintended genre) and thus lead to misunderstanding? This question suggests the investigation and comparison of many more texts of different kinds. Conversation, and more generally text, is seen as, not a structural unit, but a combination of choices made from networks that are related by probabilities to elements of the social situation. Speakers negotiate and mutually construct the reality in specific contexts partly by relating to the frequencies commonly found and expected. The negotiation itself is a dynamic process that establishes frequencies (see Lemke 1989 and elsewhere). Probabilities on dynamic systems capture the process of constituting text as different speakers add their contributions. Following probable choices should result in easily shared meaning and texture that is accessible. The relation of one stretch of language to another should be apparent. As argued by Martin (1985) and outlined in Fawcett —van der Mije — van Wissen (1988), the systems for conversation must allow for recursion and aborted sequences. These two options should also then be assigned probabilities. The probability of aborting a sequence is certainly used by the hearer in interpreting a conversation. Furthermore, by having a record of the number of recursions, a speaker and hearer should adjust the probabilities for the next possible recursion and also the social meaning that would be assigned to its occurrence or non-occurrence. Halliday (1985) notes that options are not equiprobable, and these differences in probability are a guide to the meanings involved. As well as probabilities on choices that reflect extra-linguistic facts, the probabalistic relationship of one stretch of language to another must also be considered. In grammatical terms, Bateman (1989), developing points from Martin (1988 b), notes that there are co-selection constraints that can be characterized by markedness conventions. These markedness conventions imply that some combinations of meanings are not possible. By adding probabilities to options in systems, not just markedness relations

232

Jonathan Fine

but other looser constraints can be stated. Earlier choices from systems (and perhaps also the realization rules applied) will have an influence on the probabilities of choices in later sequences. This, then, is another argument for keeping a history of the choices made. Although it has not been explored yet, there may be a limit to the extent one sequence may influence another in a conversation. These influences may in some cases be structural (for example, invitation — acceptance pairs), but in other cases may have weaker and perhaps noncontiguous constraints. Nevertheless, violation or manipulation of these probabilities will have an effect on hearers. A limiting case may be the one where one move always follows another, and the interpretation of the latter is always in terms of the former. In these cases an argument can be made that the two moves are in fact structurally related (for example, invitation — acceptance/rejection). Incorporating probabilities into system networks thus offers a number of advantages. Individuals and groups can be compared; the fit of speakers with the expectancies typical of social contexts can be described; the typicality and meaning of dynamic patterns can be described as they develop; and non-structural constraints such as marking can be considered. Using probabilities helps to make the modeling of conversation cognitive as well as social and linguistic. An account of conversation that includes paradigmatic systems of choices, dynamic systems, and probabilities must also relate to the lexicogrammatical system. The lexicogrammatical system will realize the systems used to describe conversation. This relationship between the two kinds of descriptions is not arbitrary and itself requires a study that starts with detailed examinations of conversations (see Halliday (1984) for suggestions of the fit of lexicogrammar with a model of conversation).

References Bartolucci, Giampiero—Jonathan Fine 1987 "The frequency of cohesion weakness in psychiatric syndromes", Applied Psycholinguistics 8: 67 — 74. Bateman, John 1989 "Dynamic systemic-functional grammar: A new frontier", Word 40: 263 — 286. Benson, James D. —William S. Greaves (eds.) 1985 Systemic perspectives on discourse 1. Norwood, N.J.: Ablex. 1988 Systemic functional approaches to discourse: Selected papers from the 12th International Systemic Workshop. Norwood, N.J.: Ablex.

The static and dynamic choices of responding

233

Berry, Margaret 1982 "Review of M . A . K . Halliday, 1978", Nottingham Linguistic Circular 11: 64-94. Butler, Christopher 1985 a "Discourse systems and structures and their place within an overall systemic model", in: James D. Benson —Williams S. Greaves (eds.), 213 — 228. 1985 b Systemic linguistics: Theory and applications. London: Batsford. Edwards, Derek —David Middleton 1986 a "Conversation with Bartlett", Quarterly Newsletter of the Laboratory of Comparative Human Cognition 8: 79 — 89. 1986 b "Joint remembering: Constructing an account of shared experience through conversational discourse", Discourse Processes 9: 423 — 459. Fawcctt, Robin P.— Anita van der Mije —Carla van Wissen 1988 "Towards a systemic flowchart model for discourse structure", in: Robin Fawcett —David Young (eds.), 1 1 6 - 1 4 3 . Fawcett, Robin R — M . A . K . Halliday —Sidney M. Lamb —Adam Makkai (eds.) 1984 The semiotics of culture and language I: Language as social semiotic. London: Pinter. Fawcett, Robin R — David Young (eds.) 1988 New developments in systemic linguistics 2: Theory and applications. London: Pinter. Fine, Jonathan 1988 a "Cognitive processes in context: A systemic approach to problems in oral language use", in: Erich Steiner —Robert Veltman (eds.), 171 — 182. 1988 b "Review of Bent Rosenbaum —Harley Sonne", Applied Psycholinguistics 9: 283-287. 1989 Review of Ruth Wodak — Pete van de Craen (eds.). Studies in Second Language Acquisition 11: 96 — 97. Fine, Jonathan —Giampiero Bartolucci — Peter Szatmari 1989 "Textual systems: Their use in creation and miscalculation of social reality". Word 40: 6 5 - 7 9 . Fine, Jonathan (ed.) 1988 Second language discourse: A textbook of current research. Norwood, N.J.: Ablex. Fine, Jonathan —Roy O. Freedle (eds.) 1983 Developmental issues in discourse. Norwood, N.J.: Ablex. Halliday, M . A . K . 1978 Language as social semiotic: The social interpretation of language and meaning. London: Arnold. 1984 "Language as code and language as behavior: A systemic-functional interpretation of the nature and ontogenesis of dialogue", in: Robin P. Fawcett et al., 3 - 3 5 . 1985 An introduction to functional grammar. London: Arnold. Halliday, Μ. A. K. —Ruqaiya Hasan 1976 Cohesion in English. London: Longman. Halliday, Μ. A. K . - R o b i n P. Fawcett (eds.) 1988 New developments in systemic linguistics 1: Theory and description. London: Pinter.

234

Jonathan Fine

Lemke, Jay 1989 "Semantics and social values", Word 40: 3 7 - 5 0 . Martin, James R. 1983 "The development of register", in: Jonathan Fine —Roy O. Freedle (eds.), 1-39. 1985 "Process and text: Two aspects of human semiosis", in: James D. Benson — Williams S. Greaves (eds.), 2 4 8 - 2 7 4 . 1988 a "Hypotactic recursive systems in English: Toward a functional interpretation", in: James D. Benson —William S. Greaves (eds.), 240 — 270. 1988b "The meaning of features in systemic linguistics", in: M . A . K . Halliday — Robin P. Fawcett (eds.), 1 4 - 4 0 . Matthews, Peter H. 1981 Syntax. Cambridge: Cambridge University Press. Matthiessen, Christian 1988 Organizing text: Rhetorical schemas and generic structure potential. [Unpublished MS. A paper presented at the 15th International Systemics Congress, East Lansing, Michigan, August 1988.] Nesbitt, Christopher —Gunther Plum 1988 "Probabilities in a systemic functional grammar: The clause complex in English", in: Robin P. Fawcett — David Young (eds.), 116 — 143. Rochester, Sherry R. —James R. Martin 1979 Crazy talk: A study of the discourse of schizophrenic speakers. New York: Plenum. Rosenbaum, Bent —Harley Sonne The language of psychosis. New York: New York University Press. Sacks, Harvey —Emanuel Schegloff—Gail Jefferson 1974 "A simplest systematics for the organization of turn-taking for conversation", Language 50: 696 — 735. Schegloff, Emanuel —Gail Jefferson —Harvey Sacks 1977 "The preference for self-correction in the organization of repair in conversation", Language 53: 361 —382. Steiner, Erich —Robert Veltman (eds.) 1988 Pragmatics, discourse and text: Some systemically inspired approaches. London: Pinter. Szatmari, Peter —Giampiero Bartolucci — Rebecca Bremner 1989 "Asperger's syndrome and autism comparisons of early history and outcome", Developmental Medicine and Child Neurology 31: 709 — 720. Ventola, Eija 1987 The structure of social interaction. London: Pinter. 1989 "Problems of modeling and applied issues within the framework of genre", Word 40: 1 2 9 - 1 6 1 . Walters, J o e l - Y u v a l Wolf 1988 "Integration of first language material in second language comprehension", in: Jonathan Fine (ed.), 1 9 3 - 2 0 5 . Wodak, Ruth —Pete van de Craen (eds.) 1987 Neurotic and psychotic language behaviour. Clevedon: Multilingual Matters.

First- and second-order registers in education Frances

Christie

1. Introduction: pointing directions In a paper I gave at the Fifteenth Systemics Congress in Sydney (Christie 1987), I argued that a teaching —learning activity of the kind typically found in schools might be thought of as an instance of a curriculum genre, and I examined field and schematic structure in one instance of a curriculum genre — namely, the morning-news genre. In this chapter I want to take further my consideration of the concept of the curriculum genre. I intend on this occasion to turn to an instance of what I term a writing-planning genre and, rather more fully than in 1987, I shall seek to demonstrate the operation of register in such a genre. Specifically, I shall suggest that two registers operate to realize the various elements of schematic structure of a curriculum genre. Further, I shall suggest that the two registers operate in such a way that the first-order register "projects" the second-order register, and that the nature of the second one is significantly determined by the first one. In addition, in demonstrating the operation of the two registers, I shall seek to show how such a demonstration supports a number of observations made by Bernstein (1986) with respect to the operation of what he called a "pedagogic discourse". In explaining the ways in which a culture transmits its knowledge and methods of working to the young, Bernstein argued that a pedagogic discourse should be thought of in terms of an "instructional discourse" and a "regulative discourse". The former referred to the "specialized competencies (of a culture) and their relation to one another", while the latter referred to the "specialized order, relation, and identity" — the manner of ordering things, in fact — in which pedagogical practices occurred.

2. Genre, register, and language Some definitions of terms as they will be used here will be in order. Firstly, then, we should note that the notions of "genre" and "register" as they will be used here, are understood as separate semiotic systems,

236

Frances

Christie

following the work of Martin (1985). Genres, Martin writes, represent "linguistically realized activity types" (1985: 250) in a culture, and they can thus be thought of as artefacts of the context of culture (Malinowski 1935 [1967]) in which persons operate. Registers, on the other hand, are of the contexts of situation (Malinowski 1923) within a culture, and the kinds of choices made with respect to register at any given time are actually constrained by the genre selected. That is to say, register choices are made in using language with respect to field, tenor and mode, but these sets of choices are themselves in part a condition of the type of genre selected with which to get things done. For example, in Englishspeaking cultures people not uncommonly give lectures, but a lecture genre is not normally selected for dealing with such subjects as housecleaning. When people are being socialized into house-cleaning, genres other than lectures are actually selected. The field itself would constitute a reasonable one for a lecture, but the tenor and mode values in an English-speaking culture do not combine to produce the powerful monologic text type which is a lecture. A recent statement, summarizing what is generally argued with respect to genres reads thus: "Genres are referred to as social processes, because members of a culture interact with each other to achieve them; as goal oriented, because they have evolved to get things done; and as staged, because it usually takes more than one step for participants to achieve their goals" (Martin - Christie - Rothery 1987: 59). A great deal of work will be required before anything like a complete account of the genres and associated potential register choices in an English-speaking culture is developed, although since Martin (1985) first appeared, studies by Ventola (1987); Plum (1988); Christie (1989); Rothery (1990); and Poynton (1990) have added substantially to the total body of research evidence about genres of both spoken and written modes. In particular, more is known about the ways in which genre choices appear to constrain the kinds of register choices made for given contexts of situation. Thus, the study of morning-news genres (Christie 1989) demonstrates that the activity of morning-news giving in the junior primary school is generated in such a way that only certain types of register choices may be taken up. Children are required to draw upon fields of personal experience, but these choices must be of experiences that are celebratory and/or happy. Experiences that are "unpleasant", or at least not celebratory or happy in character, are quite actively discouraged by the teacher. Furthermore, in terms of tenor relationships, the personal expriences

First- and second-order

registers in education

237

taken up must be constructed in a manner which affirms the status of the morning-news giver as well-mannered, child-like, and "innocent". In other words, a romantically conceived notion of the young child actually drives the operation of the curriculum genre that is morning news, and it has important consequences for the register choices that may legitimely be taken up with respect to that genre. Turning to the relationship of language to register and genre, this will be understood here in the manner suggested in Figure 1. genre register language Figure 1. Language, register, and genre (Martin 1985: 250)

As the figure is intended to suggest, genre, register, and language each operate on a different communication plane. In the bottom right hand corner lies language itself, functioning as a meaning system. Above it, and standing in a parasitic relationship to language, lies register. It is parasitic because it must select from language in order to realize its meanings, since it has no form of expression of its own. Similarly, genre is said to stand in a parasitic relationship to register, in that it finds expression through language. Thus genre and register are said to "stack up" in relation to language, accounting for the manner in which they are represented in Figure 1. Overall then, given the model of genre, register, and language proposed here, it is suggested that a curriculum genre is realized in choices made in register and that these are themselves realized in choices in language. In order to track the operation of the various elements of schematic structure found in a curriculum genre, one actually tracks the operation of register in the genre. More accurately, as already indicated, I shall argue that one actually tracks the operation of two registers in the creation of the curriculum genre. Curriculum genres constitute instances of socially significant activities in which persons engage in western cultures in order that people be educated. By its very nature, educational activity is a process of initiation into ways of dealing with, and of ordering, experience which is valued in a culture. The particular significance of curriculum genres is that they represent settings in which students — often, but not necessarily always,

238

Frances

Christie

quite young children — are initiated into aspects of their culture. It is because of this characteristic of initiation, as we shall see, that two registers will be found to operate in the curriculum genre.

3. First- and second-order registers Register will be defined as in Figure 2, representing a model based upon the work of Martin (1986); Poynton (1984, 1985, 1990); and Plum (1988). FIELD

set of activity sequences oriented to a global institutional purpose: e. g., dog breeding, sailing, medicine, shopping.

TENOR 1. power 2. contact 3. affect

formality of interlocutors' relations as modulated by: power and solidarity relations degree of involvement in relationship love-through-hate predisposition of interlocutors.

MODE 1.

spatial and temporal "distance" scalcs: distance between interlocutors as affecting aural and visual feedback

2.

distance between language and the social activity in which it plays a part (language-in-action to languagc-as-reflection)

Figure 2. Martin and Poynton's revised register categories (Plum 1988: 43)

Briefly, field is defined as social activity built around what Martin (1986) has defined as "activity sequences oriented to some global institutional purpose." Tenor is defined, following Poynton (1984, 1985, 1990) in terms of: power, relating to power and solidarity relations; contact, relating to the degree of involvement in a relationship; and affect, relating to the predisposition of the persons in the relationship ranging from love through to hate. Mode, following Martin (1984), is thought of in terms of spatial and temporal distance. The first sense involves the relationship of participants in the construction of the discourse, ranging from face to face conversation, through the various permutations caused by uses of aural and/or visual aids such as telephone, television, radio, and out to the various relationships achieved through the written mode, from intimate letters to materials written for remote and often unknown audiences. These may be thought of as arranged on a scale. The sense of temporal distance may be also thought of in terms of a scale, ranging from language which is part of action as in a football match, on to the language of commentary upon the match, and ultimately through a number of other

First- and second-order

registers in education

239

ways of using language less and less closely related to action — towards language for reflection upon experience, in fact. A first-order or pedagogical register has the function of operationalizing and maintaining the teaching —learning activity that takes place within a curriculum genre, while the second-order or "content register" realizes the acitivity and/or information which is to be dealt with by the children as they learn. It should be noted that the term "content" is not ideal, since it carries the unintended sense of some "packaged" information to be passed on to the learners. What is really at issue in teaching a second-order register is teaching the students involved to recognize some of the activities associated with a particular field. Such a field might include, for example, how to conduct scientific experiments as in a natural science and write up the findings, how to research and amass data as in the social sciences, or perhaps how to craft a literary piece as in English literary studies. I use the terms "content register" and "content field" because no other acceptable alternatives seem available, and provided the reservation is borne in mind, they will suffice. Activities such as educational ones appear to be characterized by their tendency to select from a range of different fields of human experience and enquiry, and to relocate aspects of those fields in other contexts whose purposes are that some will teach and others will learn about those fields. It is the demands of the teaching and learning activities which primarily seem to determine the manner of treatment of the other fields. It is for this reason that I propose the working of first- and second-order registers. It is worth noting that in arguing for the operation of pedagogic discourse Bernstein (1986: 210) represented the relationship of his instructional and regulative discourses as "instructional discourse/ regulative discourse", by which he intended to suggest that the instructional discourse was "embedded" within the regulative discourse. Two matters were at issue in representing the relationship of the two in this way. Firstly, Bernstein intended to suggest the operation of a fundamental "recontextualizing principle" such that the discourses of other contexts are appropriated and relocated in the educational discourse. Secondly, he intended to suggest that it was the regulative discourse which dominated, significantly determining much of the character of the instructional discourse. The first-order or pedagogical register I propose in fact relates closely to Bernstein's regulative discourse, while the second-order or "content" register relates to his instructional discourse. I shall not represent the

240

Frances

Christie

relationship of the two in terms of the notion of "embedding", since that is a technical term used in the systemic functional grammar (Halliday 1985) with a very different meaning. However, I shall suggest that we may characterize the relationship of the two registers in terms of projection, a term borrowed and used metaphorically from the systemic functional grammar. Technically, the notion of projection relates to one of the two possible sets of relationships between clause complexes (Halliday 1985: 192 — 251). The other relationship Halliday proposes is that of expansion. Clauses are said to project when: "the secondary clause is projected through the primary clause, which instates it as (a) a locution or (b) an idea" (Halliday 1985: 196). An instance of a projecting clause of locution is provided in she said she was going home, where it is clear that the second clause (technically in a relationship of hypotaxis) is projected through the former clause. An instance of a projecting clause of idea is provided in she thought she would go home. According to Halliday (1985: 251), clauses which are projected are "categories of the language, not of the real world". That is to say, in the case of a locution, what is represented has previously been represented in language. In the case of an idea, that which is represented has previously been a meaning of the "inner world" of thought, and hence not directly of the "outer world" of action. What is projected through the primary clause in both cases is thus a re-expression of some other representation of reality. In this sense, the projected clause involves a kind of relocation. In view of what I have already said of the relationship of the first- and second-order registers, my reasons for selecting metaphorically the notion of projection from Halliday's work should be clear. Educational enquiry, I have already suggested, following many of Bernstein's observations on the matter, takes significant discourses — a systemicist would say significant fields of human experience — from "outside the school" as it were, and relocates these "within schools" in order that the young shall learn. The process of relocation may be represented as a kind of projection.

4. Identifying the systems in which the two registers are realized Some general principles with respect to genre and register having been established, it becomes necessary to indicate the linguistic features in which I propose to argue that the two registers are realized. Martin's

First- and second-order

registers in education

241

(1985) model of a tri-stratal systemic functional grammar, as set out in Figure 3, lies behind this discussion. CONVERSATION STRUCTURE CONJUNCTION REFERENCE LEXICAL C O H E S I O N

TRANSITIVITY THEME MOOD group LEXIS and word systems

TONALITY TONICITY TONE foot and syllable prosodies phoneme systems

discourse

lexicogrammar

phonology

Figure 3. Outline of a tri-stratal systemic functional grammar with central systems on each stratum noted (Martin 1985: 249)

The third stratum, labeled phonology, need not detain us. Looking to the middle stratum, that of the lexicogrammar, I shall select for examination the "theme" and "transitivity" systems, both of them important systems involved in realizing the meanings of my chosen text. On the third stratum, that of the discourse semantic, I shall select the "conversation structure" system as one of the systems instrumental in holding the text together. A brief explanation of each will be necessary, though space will not permit a detailed account. Examining the significance of theme in the clause, Halliday defines it as: "the starting-point for the message; it is what the clause is going to be about" (Halliday 1985: 39). Three types of theme choices are recognized, responsive to the three metafunctions recognized in the systemic functional grammar: textual, interpersonal, and experiential. As we shall see, theme choices do differ in realizing each of the two registers and hence also in realizing the various elements of schematic structure of the writing planning genre. Transitivity refers to the manner in which a clause represents a process, whether that be a process of doing, happening, feeling, or being. According to Halliday (1985: 101), a transitivity process consists potentially of three components: (i) the process itself; (ii) participants in the process; (iii) circumstances associated with the process. As we shall also see, choices in transitivity differ in realizing the two registers and hence the elements of schematic structure. Conversation structure as the term is used here draws most heavily upon the work of Berry (19881 a, 1981b, 1981c); Ventola (1988); and Martin (1986). When we consider conversation structure, we consider discourse as dialogue: how it is that spoken texts become conversations, and the kinds of speech acts in which persons engage to create such

242

Frances

Christie

conversations. I shall draw upon Berry and Martin in particular in identifying the principal "moves" in a conversational exchange, and upon Ventola's recent work on "move complexes". I shall not attempt to sketch in all the types of moves identified, but confine myself to those pertinent to my discussion: K l , K2, DK1, A2, [Al: NV], and [Al: N V . . . + ] . Exchanges may be knowledge- or action-oriented. Thus, where a person offers information, the exchange is knowledge-oriented, and the move taken up is that of K l (Primary Knower) as in my mum buys Home Pride. Where, as is often the case in the classroom, a question is asked by a person who actually knows the answer, the move is known as a DK1 (Delayed Primary Knower) what's this on the shoulder? Where a person responds to such a question, the move is known as that of a K2 (Secondary Knower): Australia. Where an exchange is action-oriented the move is known as an A2 move, because the person speaking is actually guiding the actions of another: move back to your seats please. In this situation, where no rejoinder is required, the response is represented thus: [Al: NV], indicating that the person(s) involved do not speak. Sometimes, in classrooms in particular, the A2 move is delayed, as when a teacher advises the class that in a few minutes they will start writing. This move is represented thus: [Al: N V . . . + ] , With respect to Ventola's (1988) notion of the move complex, I shall observe that this is a tool she has developed in order to show how a series of moves could be represented as a complex unity, rather than as a series of discrete or independent moves. In building the relations within move complexes, she proposes that we borrow Halliday's notions of "elaboration" (represented with " = "); "extension" (represented with " + " ) ; and "enhancement (represented with " χ "), with respect to the relations between clauses. An instance of extension is represented as in (1): (1)

Mrs S., Jodie W.'s dad told Jodie not to lose them else we'll be in real strife.

Kl Kl

1 +2

In (1), the numbers 1 and 2 denote the two moves, while the symbol " + " indicates the presence of a move complex.

First- and second-order

registers in education

243

5. The writing planning genre: some general observations Text A in (2) has been selected from a very large body of data collected in the course of a three year study (Christie 1989) in which I investigated the teaching of writing to young children in the first three years of their schooling. In considering the "schematic structure potential" of the writing planning genre, I have argued that the most minimal instance of the genre may be represented thus: TCTTS'T, where TO stands for "task orientation", TS stands for "task specification", and Τ stands for "taks", while " λ " indicates sequence. Like most other genres, including the morning-news genre (Christie 1989), the most significant area of variation in schematic structure lies in the "middle" part of the genre. Sometimes there will be a "task reorientation", sometimes a "task respecification", and sometimes even two "task specification" elements. In the interests of conserving space and time, however, I have deliberately chosen, for the purposes of this chapter, to look at a text which offers the most minimal expression of schematic structure. In fact, Text A is set out with the three elements of schematic structure indicated. (2)

Task Orientation "opening" Teacher:

Well now today we're going to talk some more about the things we have been talking about for the last couple of days. We were looking at different people's uniforms weren't we? We looked at the crossing lady's uniform, and we talked to the lady about why she wears it. And for each of the uniforms we've looked each of the uniforms we've looked at, we've asked some questions. What were the questions, Christian?

"body" of schematic structure element Christian: Who they were, what they wear, why they wear it. (He is actually reading this from the board.) Teacher: Yes, why do people wear uniforms? Danielle: Tells who they are. Teacher: Yes, it tells us what their job is. Now Mrs. P. (another teacher) has lent us this uniform. Look at this peaked hat. Danielle: That's just like my uncle's.

244

Frances

Christie

Mirko: Teacher:

Christian: Teacher: Christian: Teacher: Mirko:

Teacher: Christian: Teacher: Gabriel: Aaron: Teacher: Aaron: Christian: Teacher:

Cindy: Diana: Teacher: Aaron: Teacher: Aaron: Carrie: Teacher: Christian: Aaron:

Is that Mrs. P.'s husband's? Yes. See the peaked cap, the stiff piece here in the front. It's a dark colour isn't it? A dark blue, you'd call it. Now look at the jacket He wears a white shirt. See the white stripes on the sleeves. I know why he wears those things. (He means the stripes.) Yes, they show who he is don't they? Do they have a badge here? (He points to the patch where a badge has obviously been removed.) Yes, Mr. P. has taken it o f f . And you sew them on. What's this on the shoulder? Australia. Australia. And he has special buttons. Mrs. S., there's two sets of buttons, and two sets of button holes. Mrs. S., how do you take the stripes o f f ! With a pair of scissors. This is a double breasted jacket. That's why there are all those buttons. Seel (She holds it up.) There's buttons on the sleeve. They look like silver coins — the little ones. Now let's look at the train driver's uniform. That's Jodie W.'s dad's. Look, there's some badges here, some of them in the pocket. Mrs. S., Jodie W.'s dad told Jodie not to lose them. Else we'll be in big strife. (Said with a slight shudder.) Well, we'll pin them on here, on the jacket, then they can't get lost. All of the badges are different. 'Cos they'd be for something different. (He means each has a different purpose.)

First- and second-order

Teacher:

registers in education

245

(Holds up the green jacket.) See, this is a green jacket. I wonder if he'd wear a green shirt. He would wear a blue shirt. Christian: What else would he wear? Teacher: Slacks. Diana: Yes, he would. Teacher: Grey slacks. Stephen: Teacher: Yes, I think they are grey. Well, so that's a train driver's uniform. Now let's look at this one. Who wears this kind of uniform ? A nurse. Stephen: Teacher: Where do you find nurses? Who do they work for? Stephen: A dentist. Another child: A doctor. Teacher: Yes, that's right, and what does she do? She's a helper. Diana: That's Jodie's mum's. Olivera: She goes to school .... She's learning to be a Cindy: nurse. Teacher: Yes, she's really a nurse's aide. Stephen: An aide. What's that? Christian: She's got to clean the people. Teacher: See what it says on the badge here. It's got R.N. A. Aaron: That's "Real Nurse's Aide". Teacher: No, it's "Registered Nurse's Aide". She's a registered nurse's aide. It means she helps the nurses to look after the people in hospital. See, it's a blue uniform, and I think she wears white stockings, and a white cap. Sometimes they have a cape. It's red. Diana: Yes, good. I think she'd have a red cape too. Now Teacher: let's look at the two uniforms Belinda's brought in. They belong to Mr. and Mrs. S. They both work for the Home Pride Bakery. See their uniforms hanging up here. They make the bread. Tony: My mum buys Home Pride. Carrie: Look at Mr. S.'s shirt. It's brown. Teacher:

246

Frances

Christie

Mirko: Teacher: Aaron: Teacher: Christian:

Teacher:

Olivera: Teacher: Christian: Teacher: Chorus: Teacher: Aaron: Teacher:

Christian: Stephen: Teacher:

Teacher:

It's something like an army shirt. See the writing. What does it sayl Home Pride. Why do they put it on the hats'} They can't put the writing up here. (He gestures to his chest.) They have to put it here. (He points to his own head.) Why do they have to wear a hat in the bakery when they make the bread? Vanessa? (Vanessa does not reply.) They might have to stop the sun. It might get in their eyes. No, you wouldn't get sun in a bakery. Christian? So they look nice. No. Oh come on everyone. What's on your head? Hair. And why do you need to cover that up if you're making bread in a bakery? To keep it tidy. No, oh, can't anyone think better than that? What would happen if their hair fell out when they were making the bread"! It'ud get in the bread. There was paper in my bread once. Yes, it'ud get in your bread, and that wouldn't be very nice, would i f . (A brief burst of animated talk here about eating bread with hair in it; too loud and too many speakers to catch it. The teacher ignores it.) See, Mrs. S. wears a dress, and Mr. S. has a shirt and trousers. And they're all the same colour, aren't they? It's a dark brown. They wear those to make the bread in the bakery.

Task Specification "opening" of element Teacher: All right, now you've had time to look at all these uniforms. We should ask Belinda to thank her mum and dad for letting us see their uniforms. Now remember what we said before when we

First- and second-order

registers in education

247

looked at other uniforms and wrote reports about them. Well, we're going to write about these uniforms today. You can choose any one you like: the policeman's, the train driver's, the nurse's, or the one from the Home Pride Bakery. " b o d y " of element Now what do we have to remember to write about? Aaron: Who they are and what they wear. Yes. Teacher: Task Teacher:

See it's on the board there: Who they are. What they wear. What it looks like. Why they wear it. (Above these on the board is written the heading " U n i f o r m s " though the teacher does not read this out. The children get up and move to their desks, where the teacher has previously placed their writing books.)

In a writing-planning genre the teacher initiates the learning activity. In the task orientation she prepares the children for writing by having them consider some "content field" for writing. The task-orientation element is, in fact, always the longest element in the genre and the one in which the children participate most frequently. In mode terms, it is here that the language is used most for reflection upon experience. The task specification is the element in which the teacher identifies the task for writing. This element is shorter than the task orientation, and this is normally because the teacher devotes nothing like the same amount of time to defining the task for writing as she does to talking about the "content field" for writing. In other words, as the children are moved towards the actual writing task in which they are to engage, the language increasingly assumes the character of language for preparation for action. This is a nice instance of the "invisibility" of language in educational contexts. At the very point at which the nature of the writing task is supposed to be developed so that the children may learn it, the actual issues involved in undertaking that task become most implicit. That is to say, for the purposes of their writing, the children are left to deduce the linguistic choices they must make in order to write the genre of concern. An enduring problem of English-speaking cultures generally — not uniquely of the sub-cultures of schooling — is that language is so often

248

Frances

Christie

an invisible and hence unacknowledged part of social processes. While this can be shown to create difficulties in many contexts of situation, it does create particular difficulties in educational ones. The final element, the task, operationalizes action and it always accompanies significant movement on the part of the teacher and the children. It is thus very much language for action, and hence it is always the shortest element of schematic structure.

6. The linguistic realization of the two registers in the writing planning genre I shall summarize the principal features of the operation of the two registers in realizing the elements of schematic structure. These are set out in some detail in Table 1. Table 1. The operation of the "pedagogical" and "content registers" in Example (2), an instance of the writing —planning genre Task Orientation "opening": "pedagogical register" is foregrounded (i) (ii)

teacher monologue Theme choices: textual themes: continuatives point directions and/or link to previous activities well now. topical themes: identify class members as part of operationalizing activity we. (iii) Transitivity processes to operationalize activity are behavioral: we 're going to talk some more about the things... we were looking at different people's uniforms we looked at the crossing lady's uniform we've asked some questions (iv) Principal participants identify class members: we (v) Secondary participants and/or circumstances identify aspects of the "content" for writing: about the things [[we have been talking about for the last couple of days J ] different people's uniforms about [[ why she wears it]] for each of the uniforms [[we've looked at]] (vi) Conversation structure: both a) teacher K1 moves which provide information: well now today we're going to talk some more about the things we have been talking about for the last couple of days Kl 1 we were looking at different people's uniforms weren't we? K1 = 2

First- and second-order registers in education we looked at the crossing lady's uniform and we talked to the lady about why she wears it and for each of the uniforms we 've looked at we've asked some questions and b) a teacher DK1 move designed to initiate dialogue "content register": what were the questions?

249 Kl 1 Κ1 + 2 K1

+3

and the entry to the DK1

"body": "content register" is foregrounded (i) (ii)

dialogue Theme choices:

textual Themes: a) some teacher uses of continuatives to carry the activity forward e.g., well we'll pin them on here, but also b) teacher uses of s t r u c t u r a l realized in conjunctions, mainly temporal or additive, to build links in the development of the "content field" e. g., then they can't get lost', see it's a blue uniform and I think she wears white stockings and a white cap. and c) some limited use of structurals by the children: else we'll be in big strife. interpersonal themes: a) frequent teacher use of W H interrogatives functioning as both interpersonal and experiential theme choices: e.g., "why do people wear uniforms?, c) teacher A2 moves to direct activity: now let's look at the train driver's uniform (said as she displays it) and d) frequent student K2 moves to respond to teacher DK1 moves: Australia e) student K1 moves to provide information. There was paper in my bread once Task Specification "opening": "pedagogical register" is foregrounded (i) (ii)

Teacher monologue Theme choices: textual themes: continuatives point directions and/or link to previous activities: all right; now; well. topical themes: identify class members, signaling that direction is being given to their activity: you; we. (iii) Transitivity processes operationalize activity, and they include: possessive: now you've had time [[to look at all these uniforms]] verbal: we should ask Belinda to thank her mum and dad for [ [letting us see their uniforms]] cognition: now remember what we said before... behavioral: we're going to write about these uniforms today affect: you can choose any one [[you like]] ... (iv) Principal participants identify class members: you; we.

250 (ν)

Frances

Christie

Secondary participants and/or circumstances mainly identify aspects of the "content field" for writing: all these uniforms', other uniforms; about these uniforms, a) while one only identifies the genre to be written: reports. and b) occasional uses of vocatives Mrs. S., Jodie W.'s dad told Jodie not to lose them.

experiential themes: a) identify aspects of the "content field" and are used frequently by the teacher, e. g., he has special buttons; that's a train driver's uniform, but b) they also feature as the most frequent theme choices in the children's talk: e. g., my mum buys Home Pride; she goes to school. (iii) Transitivity processes build aspects of the "content field", and they include: material: he wears a white shirt verbal: tells who they are perception: see the peaked cap, the stiff piece here in the front attributive: this is a double breasted jacket possessive: that's Jodie's mum's (iv) Principal participants identify aspects of the "content field": he wears a white shirt they look like silver coins — the little ones all of the badges are different (v) Secondary participants and/or circumstances also identify aspects of the "content field": who wears this kind of uniform? where do you find nurses? why do they have to wear a hat in the bakery? (vi) Conversation structure: a) frequent use of teacher DK1 moves to provoke talk about the "content field": what's this on the shoulder? b) teacher K1 moves to provide information: this is a green jacket. (vi) Conversation structure: a) use of teacher K1 moves to inform the children of their task: all right now you've had time to look at all these uniforms we should ask Belinda to thank her mum and dad for letting us see their uniforms K1 b) use of teacher A2 moves to direct activity: now remember what we said before when we looked at other uniforms and wrote reports about them (points to board) A2 c) use of a teacher DK1 move to draw the children into dialogue: now what do we have to remember to write about? DK1 and d) an [Al: NV] move from the children when directed to look to the board. e) an [Al: N V - f . . . ] move from the children when told of the writing they are to do a little later. "body": "pedagogical" and "content registers" converge (i)

dialogue but little student participation

First- and second-order registers in education

251

(ii)

Theme choices: textual themes: continuative: yes interpersonal/experiential themes: who; what (iii) Transitivity processes are identifying and material, helping to identify aspects of the "content" for writing about: identifying: who they are material: what they wear (iv) Principal participants identify aspects of the "content" to be written about: they (v) A secondary participant also identifies an aspect of the "content field": what (vi) Conversation structure: a) a K2 move from a student: who they are and what they wear, and b) a K1 moves from the teacher: yes Task "content register" is foregrounded but pedagogical activity is generated through changed physical disposition (i) (ii)

teacher monologue Theme choices: no textual themes no interpersonal themes two topical themes, one guiding the actions of the children and realized in an imperative: see, and a second one, identifying the series of questions to be answered in writing: it (iii) Transitivity processes There is one primary one, an intensive process: it's on the board there: 'who they are, what they wear, what it looks like, why they wear it.' (iv) The principal participant identifies the series of questions for writing: it (v) The circumstance "spells out" out the series of questions: on the hoard there: 'who they are, what they wear, what it looks like, why they wear it.' (vi) Conversation structure: a) a teacher A2 move: see b) a teacher K.1 move and c) an [Al: NV] move from the children.

The "opening" of the Task Orientation is signaled in teacher monologue: here the teacher points directions, selecting some continuatives to do so {well now). Her experiential theme-choices identify class participants {we, you), as part of alerting the children that it is their behavior that is at issue at this point. Her choices of transitivity processes are behavioral, and used to operationalize activity {today we 're going to talk some more

252

Frances

Christie

about the things we have been talking about ...). In terms of conversation structure the teacher takes up a series of K1 moves as she informs the children of what is to happen. The entry to the "body" of the element is signaled with an entry to dialogue, also signaling entry to engagement with the "content field". The teacher uses a DK1 move to effect the entry into dialogue (what were the questions!:). Once into the "body" of the element, the topical theme choices realize not class participants, but persons and/or items of the "content field" (that's just like my uncle's). Continuatives are still sometimes used by the teacher, but she also makes more frequent use of structural, realized in conjunctions, mostly temporal or additive (then they can't get lost). The children, it will be noted make no use of continuatives, and while they do sometimes use structurals, they use fewer of them than the teacher. That seems to be because it is her role to point directions and make links between observations made about the "content", while it is the children's role to talk about the "content". The theme choices taken up most often by the children are in fact experiential, as they engage with the "content". Transitivity processes are various, as one might expect. Some are material and have to do with the activities of those who wear the uniforms under discussion (he wears a white shirt), while others are instances of relational processes, building features of the uniforms and those who wear them (and he has special buttons). In terms of conversation structure, the teacher makes frequent use of DK1 moves to guide talk, as well as a number of A2 moves to guide actions (look at this peaked hat). The children make no uses of DK1 moves, and they use a number of K2 moves as they respond to questions (tells who they are), though occasionally they use K2 moves as well to elicit information (do they have a badge here?). They do use K1 moves also on occasion, though these are quite often ignored by the teacher, whose prerogative it is to take note of, or to ignore, offerings from the children (for example, there was paper in my bread once). They also take up a number of [Al: NV] moves as the teacher guides their observations of the uniforms. Turning to the task specification, the "opening" of this element is signaled through a foregrounding of the pedagogical register. The teacher takes up a monologue, and the process of pointing directions is again signaled through uses of continuatives in textual theme position (all right), while topical theme choices once again identify class members (you and we). Transitivity choices, which are varied in the monologue, are all intended to operationalize activity (for example, now remember what we

First- and second-order

registers in education

253

said before...). The principal participants in the processes identify class members, while the secondary participants identify aspects of the "content" for writing (all these uniforms). In terms of conversation structure, the teacher uses a K1 move to signal the beginning of the "opening" and an A2 to direct activity. A DK1 move signals the entry to the "body" of the element: now what do we have to remember to write about?. The only response required from the children in the "opening" of the Task Specification is an [Al: NV] move. Once into the "body" of the element (which is minimal indeed) the pedagogical register and "content register" converge. There are no textual themes, apart from a use of a continuative (yes). Interpersonal/ experiential theme choices introduce two questions intended to guide writing, both of which employ material processes (who they are and what they wear). The child who offers these takes up a K2 move to do so. The task element is constructed entirely in teacher monologue. There are no textual or interpersonal themes, and the two topical themes are realized firstly in a transitivity process used in the imperative mood (see); and secondly in a choice (it) which establishes the questions to guide writing. The other principal transitivity process apart from the one already identified is an intensive one, also helping to establish the questions to guide writing. With respect to conversation structure the teacher opens the element with an A2 move, and follows this with a K1 move. The children are required to make an [A2: NV] move in response. It is notable that the teacher does not actually tell the children to stand up and move back to their seats. That is because the action required of the participants is so fully understood by them that it requires no linguistic realization at all. Overall, the children's participation is most marked in the task orientation, while their contribution in the task specification is considerably reduced, and they are not required to make any contribution in the final element. Indeed, were they to attempt much conversation in this last element, the teacher would tend to cut them off. In general, it is the foregrounding of the pedagogical register which signals the start of an element, while some expression of the "content register" occurs in the "body" of the element. The "content register" finds most considerable expression in the task orientation — the element for reflection upon experience. It finds somewhat reduced expression in the task specification, and here some convergence of the pedagogical register and "content register" occurs. However, the task for writing is developed poorly, since it remains largely implicit to the context of situation. In the last element,

254

Frances

Christie

where the participants are galvanized into action, it is the "content register" that is expressed in the clearest manner linguistically, though its very realization accompanies the taking of steps which actually satisfy the requirements of the pedagogical register: the participants get up to move to their seats for writing.

7. Conclusion I began this chapter by arguing that a first-order register projects a second-order register in a curriculum genre. I have sought to demonstrate this general proposition by reference to one instance of a writing planning genre taken from early childhood education. I suggested further that in demonstrating the manner in which the second-order register is projected by the first-order register, I would provide linguistic evidence for Bernstein's general proposition that in educational contexts a pedagogic discourse applies, such that a regulative discourse "embeds" an instructional discourse. In providing the linguistic evidence I have also, hopefully, drawn attention to the important function of education as a process of initiation. All students of any age are engaged in a process of being inducted into some field, or fields, of human endeavour. It is the teacher's particular responsibility to accept this and to guide students into a principled understanding of a "content field", taken, as we have argued, from some other context of situation, "outside the school", and relocated for the purposes of school learning. The point may seem an obvious one to make. Yet at this juncture in the history of English-speaking educational practices at least, I suggest it needs to be asserted very strongly. That is because a great deal of advice to teachers on the design and implementation of the school curriculum — the literacy curriculum in particular — has for some years now extolled a view of an education as some kind of journey of "personal discovery" or "independent enquiry". Educational theories, often supported by theories of childhood psychology, would see young learners as "growing" through "development of personal and/or independent capacities" where the latter are seen as somehow innate and even rather private. Where such a model of the learner applies, the teacher's role is that of "facilitator" only, pursuing a rather neutral and essentially non-interventionist role as the learner develops. Such models have been unhelpful, producing generations of teachers reluctant to take an overtly interventionist role in their teaching. Indeed,

First- and second-order registers in education

255

they are often simply unable to do so, since — like the teacher in the analyzed text — they have become increasingly deskilled and unaware of the critical ways in which language is used to shape human experience, and hence to structure significant fields. In saying this, by the way, I stress that I intend no criticism of the particular teacher, who was, after all, extremely generous in allowing me into her classroom, and I knew her to be a concerned person. Above all, teaching should be seen as a deliberate act — one in which the teacher defines goals and works confidently towards these with his/ her students, as they slowly master an expanding range of fields and an associated range of genres in order to deal with these fields.

References Benson, James D. —William S. Greaves (eds.) 1985 Systemic perspectives on discourse. Vol. 1. (Advances in Discourse Processes. Vol. XV.) New Jersey: Ablex. 1988 Systemic functional approaches to discourse: Selected papers from the 12th International Systemic Workshop. (Advances in Discourse Processes. Vol. XXVI.) New Jersey: Ablex. Bernstein, Basil 1986 "On pedagogic discourse", in: John G. Richardson (ed.), 205 — 240. Berry, Margaret 1981a "Systemic linguistics and discourse analysis: A multi-layered approach to exchange structure", in: Malcolm Coulthard —Martin Montgomery (eds.), 120-145. 1981 b "Towards layers of exchange structure for directive exchanges", Network 2: 23-32. 1981 c "Polarity, ellipticity, elicitation and propositional development, their relevance to the well-formedness of an exchange", Nottingham Linguistic Circular 10, 1: 3 6 - 6 3 . Christie, Frances 1987 Field and schematic structure in the morning news genre. [Unpublished paper given at the International Systemic Workshop, University of Sydney, 24—28 August, 1987.] 1989 Curriculum genres in early childhood education: A case study in writing development. [Unpublished Ph. D. dissertation, University of Sydney.] Coulthard, Malcolm —Martin Montgomery (eds.) 1981 Studies in discourse analysis. London: Routledge and Kegan Paul. Halliday, M . A . K . 1985 An introduction to functional grammar. London: Arnold. Malinowski, Bronislaw 1923 "The problem of meaning in primitive languages", in: C . K . Ogden —K.A. Richards (eds.), 2 9 6 - 3 3 6 .

256

Frances Christie

1935 [1967]

Coral gardens and their magic. Vol. 2. London: Allen and Unwin. [Reprinted as The language of magic and gardening. Bloomington: Indiana University Press.] Martin, James R. 1984 "Language, register and genre", in: Language studies: Children writing: B. Ed. course study guide. Geelong, Australia: Deakin University Press, 21 —30. 1985 "Process and text: Two aspects of human semiosis", in: James D. Benson and William S. Greaves (eds.), 248-274. 1986 English Text: System and Structure. [Unpublished MS. A draft for teaching purposes, Department of Linguistics, University of Sydney.] [To be published by Benjamins.] Martin, James R. —Frances Christie —Joan Rothery 1987 "Social processes in education: A reply to Sawyer and Watson and others", University of Sydney Working Papers in Linguistics 5: 116 — 152. Ogden, C. K . - I . A. Richards (eds.) 1923 The meaning of meaning. (International Library of Philosophy, Psychology and Scientific Method.) London: Kegan Paul. Plum, Guenter 1988 Text and contextual conditioning in spoken English: a genre-based approach. [Unpublished Ph. D. dissertation, University of Sydney.] Poynton, Cate 1984 "Forms and functions: Names as vocatives", Nottingham Linguistic Circular 13: 1 - 3 4 . 1985 Language and gender: Making the difference. Geelong, Australia: Deakin University Press. 1990 Address and the semiotics of social relations: A systemic functional account of address forms and practices in Australian English. [Unpublished Ph. D. dissertation, University of Sydney.] Richardson, John G. (ed.) 1986 Handbook of theory and research in the sociology of education. New Jersey: Greenwood Press. Rothery, Joan 1990 'Story' writing in primary school: Assessing narrative type genres. [Unpublished Ph. D. dissertation, University of Sydney.] Ventola, Eija 1987 The structure of social interaction: A systemic approach to the semiotics of service encounters. London: Pinter. 1988 "The logical relations in exchanges", in: James D. Benson —William S. Greaves (eds.), 51 —72.

Part III

Functional theory, scientism, and altruism: A critique of functional linguistics and its applications to writing Barbara

Couture

1. Introduction Functional linguistic theory is a scientific metadiscourse. It treats language as an observable object to be explained in general terms which identify its social functions. In this chapter, I argue that functional language theory goes beyond the boundaries of scientific explanation to assert a moral philosophy. My claim is that functional theories of language are essentially altruistic, projecting a society where conventional language behavior operates for the social good. Through this projection, functional theory both implicitly approves social linguistic norms and regards divergent linguistic behavior as aberrant. In the following pages, I examine some categorical assumptions of functional theories of language which support both a "scientific" approach to textual analysis and an altrustic view of social convention. Through citing applications to writing, I argue that functional theory advocates conventional expression and discounts written communication as an agent of change — a consequence which encourages regressive language practices. At the same time, however, I demonstrate the explanatory power of functional theory to both validate and expose the moral meaning of language acts. I open my argument with a discussion of the problematic epistemology of functional language theory which conflates the scientific claims of linguistic theory with the moral claims of philosophy (Section 2). I continue with an elaboration of three paradigms in functional theory that both establish its scientific base and project a benevolent view of accepted rhetorical practices, a view which encourages their use to invalidate minority or individual expression (Section 3). In closing, I comment on new trends in the development of functional linguistic theory and its application to writing. These new directions enhance and strengthen the capacity of functional theory to reveal social values — the goal which appropriately unites all humanistic studies (Section 4).

260

Barbara

Couture

2. Functional theory as science and philosophy The term "functional linguistic theory" encompasses a variety of theories and models all of which explain how language operates as social behavior. In their attempt to explain social behavior, functional theories can conflate the aim to describe with the aim to evaluate, and hence generalize about what language ought to be rather than what it is. This kind of conflation has historical roots dating to the conflict between formalist and sociological theorists of language function. Formalists restricted explanations of linguistic function to grammatical categories while sociolinguists explained language function in terms of social values (see, Titunik 1973: 1 8 3 - 1 8 5 ) . The categorical difference between the premises of formalist and Marxist theory disguised their common objective to forward a philosophical argument. The formalists used logical and statistical analysis of linguistic form to demonstrate "objective" proof that literary aesthetic exists in a text apart from human evaluation. Having found a formal pattern to correlate with human judgment, they concluded that the former dictated the latter. The Marxists also sought an objective explanation for human choice. Relying on the premise of "dialectical materialism", they aimed to demonstrate scientifically that human struggle results from a clash of ideologies socially constructed by language. Personal deprivation and individual powerlessness were thus explained as environmental inevitabilities rather than as consequences of individuals' actions (see, Lock 1976; 1; H o r k h e i m e r - A d o r n o [1972]: 1 - 1 6 ) . Formalist and Marxist theory both sought to evaluate human behavior, yet buried this aim under the guise of scientific description. The subtle disguise of evaluation in description may be an inevitable outcome of scientific analysis of social behavior. Science holds that the world can be evaluated through empirical study of its regularities; "predictable patterns of behavior" explain what the natural world is and means. Analyses of social behavior inevitably fail to locate predictable patterns because "there is a complexity in and between humans that allows no penetrating summaries" of their interactions (Borgmann 1984: 70). Identified "patterns" of social behavior may stand for a moment as factual descriptions, but as such they say little or nothing about the significance of this activity. The meaning of human behavior is not revealed through study of its regularities, but it is rather exposed through study of its significant instances — those events which point to something of value, something that directs and changes us.

A critique

of functional

linguistics

and its applications

to writing

261

The search for regularity and pattern in social behavior cannot be viewed as totally misguided. Our ability to conduct business in the world, to love others, to laugh, and to dream with them depends on our hope that human behavior is predictable and constant in some operational sense. Our discovery of linguistic conventions for transacting business, for loving and being loved, and for communicating with friends enables us to succeed and thrive in this world. But in recognizing the utility of these conventional behaviors we should not mistake them for the values of security, love and friendship we seek to obtain through them. It is this erroneous perspective on things of ultimate value that scientific study of language function maintains. As I shall demonstrate in the remainder of this chapter, functional language theories mistake pattern to be the focal point of linguistic behavior, its ultimate significance. This assumption leads functional theories to privilege as valued expression language which is socially harmonious, ordered, and coherent. This privileging underlies three premises which govern the linguistic laws or patterns established by functional language theory: 1. Language constructs social harmony. 2. Language organizes space and time. 3. Language polarizes reality. Through these three orientations, functional language theories validate conventional meanings, structures, and ideologies. They celebrate language's capacity to create and maintain a stable society, and, by default, they discourage social change. And in their particular application to writing, functional theories undermine the potential of text to advance moral behavior.

3.

Critique of the premises of functional language theory

3.1. Language constructs social harmony Functional theories interpret language as the medium which orchestrates the interplay of diverse and multiple meanings in human societies. This positive, helpful objective motivates functional explanations of verbal texts. Functional theorists view the text as a token of positive social exchange. Halliday (1980: 12), for instance, compares verbal text to the giving of a gift "in which there is a shift of focus from donor to recipient in the course of the exchange, or rather from the giving to the receiving."

262

Barbara

Couture

Though he notes that the exchange can be interrupted at any point and the gift rejected by the beneficiary, his description of textual form features its internal stability and positive intent. He describes the text's unique reference to multiple semantic systems as harmoniously ordered: "The clause is orchestrated as melody (the experiential component, constellations of different notes), as harmony (the interpersonal component, an ongoing modal progression), and as rhythm (the textual component, the beat which organizes the sound into a coherent whole)" (Halliday 1979: 78). The ideational, interpersonal, and textual functions are cooperative in social discourse, much as the psychic functions of the ego, superego, and id are cooperative in the operation of the personality, or as the state, the schools, and the church are cooperative in creating a meaningful society. Other functional theorists also highlight the potential of text to systematically order multiple meanings. Jakobson, for instance, claims that functional dominance orders meanings in text, placing less prominent meanings in a supporting role. Jakobson (1960: 356) explains the poetic function in literary text: the "[p]oetic function is not the sole function of verbal art but only its dominant, determining function, whereas in all other verbal activities it acts as a subsidiary, accessory constituent." Barthes refers more directly to "orchestration" when explaining the articulation of meanings in a "readerly" text, that is, one which expresses "polyphonic" meaning in such a way as to limit the reader's interpretation: "The area of the (readerly) text is comparable at every point to a (classical) musical score.... [T]he readerly [text's] tonal unity is basically dependent on two sequential codes: the revelation of truth and the coordination of the actions represented..." (Barthes [1974]: 28, 30). The "readerly" text, to which the "writerly" text is Barthes' anarchic alternative, coordinates multiple meanings in a predictable and "irreversible" way. Functional theory interprets the socially valued text as organized, coherent, and harmonious — in short, well-mannered — and it sets out to demonstrate this fact scientifically. It is a short move from demonstrating that the socially valued text is well-mannered to believing that reading and writing such texts ensures good behavior. And to encourage this position is to perpetuate what Graff (1979: xix) has called the "literacy myth". In a longitudinal study of social conditions in mid-nineteenth century Ontario, Graff shows that literacy training primarily ensured the continuance of behaviors considered beneficial to both individuals and society. Through literacy, one class maintained hegemonic control over

A critique

of functional

linguistics

and its applications

to writing

263

another with the latter's willing assent: "The poor were taught that to succeed, they must be self-restrained, obedient, and cooperative; the rich would then treat them with respect" (Graff 1979: 34, 45). Because literacy was believed to encourage good behavior, both educators and the courts disproportionately accused illiterates of criminality, even when the crime, more often than not, was mere "idleness' 1 (Graff 1979: 267). This attitude, Graff asserts, still persists today. We yet believe that reading and writing ensure a well-mannered populace — a good society: language in and of itself constructs the social conscience. It is extremely difficult, I believe, for functional linguistic theory to divorce itself from this position; in fact functional theory tends to assert that language structures both social and psychic consciousness. Functional theories claim that the word is the medium of exchange among ideologies, that ideology is fundamentally social, and that the individual conscience is fashioned by the social milieu. One of the more startling early articulations of this position was taken by Volosinov in Marxism and the Philosophy of Language (1973: 13): The only possible objective definition of consciousness is a sociological one.... The individual consciousness is nurtured on signs; it derives its growth from them; it reflects their logic and laws. The logic of consciousness is the logic of ideological communication, of the semiotic interaction of a social group. If we deprive consciousness of its semiotic, ideological content, it would have absolutely nothing left.

In adopting this view of individual consciousness, Volosinov justified socio-functional approaches to discourse analysis as part of the Marxist project to expose the ideological causes of class struggle. A recent educational permutation of Marxist functional theory — social constructionism — overtly directs writers to adjust their individual expression to social expectations (Bruffee 1986: 111). Social constructionism, Bruffee (1986: 784) argues, encourages community or collaborative learning in the classroom, and it urges us to abandon the cognitivist notion that "writing is primarily an individual act". As empirical proof of the inadequacy of the cognitivist position, Bruffee (1986: 785) cites sociological studies of science which "document instances" where "what scientists actually knew gradually changed as the community of knowledgeable peers they belonged to demanded change in the language of the articles they were writing." The hegemonic control of community thinking over individual thought implied by such a finding is not explored by Bruffee; nor does he consider mindless acquiescence to social pressure as a possible consequence of collaborative learning and writing in the class-

264

Barbara

Couture

room. Rather he emphasizes the humanizing potential of situating meaning in positive social interaction and problem-solving (Bruffee 1986: 788). Linguistic theories extend the social constructionist claim to assert that writing alters individual thinking. In a genuinely helpful article written over two decades ago, Pike summarized a number of principles from his theory of tagmemic linguistics, demonstrating how they can be used profitably by students to organize written composition. He concluded (1964: 83) that written language and its study structures cognition: "If one assumes that thought itself is not fully structured until it is articulated through language — a view which I would personally hold — then an analysis of language forms would feed back on an analysis of thought structure." An even bolder assumption about the influence of linguistic structure on thinking underlies the teaching of sentence-combining, a method of improving student writing through the practice of transforming and combining kernel sentences to produce a variety of embeddings. Though the long-term influence of sentence-combining on the quality of student writing has been debated, the claims about its immediate effectiveness have been impressive: "[Sjtudents who worked with exercises in sentencecombining produced an increase of 21 percent in 'critical transform types' — the kind of structures that seem to mark skilled writing" (Strong 1973: xii). Some educators have believed that "sentence combining practice will very likely improve reading ability and thinking" and strengthen "the concept of writing as a profound mental activity, not simply as a spontaneous transcription of oral speech" (Cooper 1975: 72). All such claims are based on an analysis of structures in writing which are favored by school teachers — a fact which confirms that what sentence combining actually ensures is conformity to a social norm. The instrumental value of conformity has recently been invoked by functional linguists who promote writing instruction in the conventional genres. Christie (1985: 31) asserts that language learning amounts to training in "appropriate behavioural patterns". She views an emphasis on individual expression in the writing classroom as a "confusing" objective which "rests upon certain serious misconceptions about the individual and about the nature of language in general, and of writing in particular" (Christie 1985: 36). It is through "equipping children with a knowledge of the language patterns they need to grasp, and through which they will express themselves," she argues, that teachers will help students become "original and creative" (Christie 1985: 31 —36). Through this claim Christie in effect asserts that human creativity is subject to the

A critique

of functional

linguistics

and its applications

to writing

265

mastery of appropriate conventions. Applications of functional theory to writing have supported implicit advocacy of socially accepted language: functional theory concludes that language as it is constructs social harmony. This position assigns creative men and women an awesome task: they must see in an idealized view of society a new and better social order and define it within the constraints of conventional expression. Ironically, social change will be granted by individuals who recognize and create a unique destiny in societies which reject them — such individuals disrupt convention and demand a new order. Such dynamism is devalued by theory which regards language as the orchestrator of social harmony. And such theory takes disempowerment and alienation for granted, confirming that individuals are inescapably defined by the schools which teach them, the industries which employ them, and the societies which govern them.

3.2. Language organizes space and time Another way in which functional theories promote an altruistic vision of existing society is through imposing a regular structure on time and space: language is shown to order space and time through creating spatial hierarchies of constituency and temporal protocols of choice. Spatial order is achieved through part-to-whole ranking in functional theory. Systemic grammar, for instance, relates parts to wholes within the ranks of morpheme, word, group, clause, and sentence (Berry 1976: 7); clause-relational analysis similarly ranks rhetorical constituents at the levels of argument and support, or summary and detail (Hoey 1983: 134—167); and various theories of rhetoric define textual genres as consisting of individual components. The narrative, for instance, includes an orientation, a complication, a resolution, and a coda (Labov —Waletzky 1967: 12 — 44); and the scientific report includes a problem statement, method, results, and conclusions (Glassman — Pinelli 1985: 10 — 11). In addition to viewing meaning as spatially structured, functional theories structure meaning in time. In describing text at the sentence or discourse level, they label constituents as they perform a function in a time-bound social process. In systemic theory, clauses name agents who perform actions with or without certain goals, instruments, and beneficiaries. The participants' sequential roles and activity are categorically structured. Transitivity identifies the nature of the action as mental or material; mood the interpersonal tenor of the act as declaratory or

266

Barbara

Couture

hortatory. Theme accounts for development in time, identifying a point where activity begins and suggesting the sequence which will bring it to an end. Nearly all representations of genre in functional theories identify chronological stages and sequences, such as the Opening Transaction, Eliciting Transaction, and Closing Transaction of a sales agreement; or the Situation, Unexplained, Procedure, Findings, Interpretation, and Explanation of a research report (for a discussion, see Ventola 1989: 132 — 133). At each stage in the generic structure proposed, various options are open to the speaker (or writer), but the overall scheme compels an initial choice which will signal that the text has begun and compels a final choice which will signal that a patterned sequence has come to a close. To represent reality in terms of part-to-whole relationships in space and beginning —middle —end transactions in time is to put specific limits on linguistic creativity and discovery. In an essay on the future of science and the limits of scientific growth, Medawar (1984: 80) notes that partto-whole or deductive thinking "merely makes explicit information that is already there." Deductive thought "is not a procedure by which new information can be brought into being." Similarly, inductive reasoning, which reports the limited observation of events over a terminal period, cannot help us discover new truths about "first and last things" — things that concern us for all time (Medawar 1984: 82). If such thinking is limited, is language itself to blame? Is language structured so that it truly inhibits us from expressing ideas in any scheme other than part-to-whole or beginning —middle —end? Or rather have we limited our explanations of language behavior in these ways? In The Domestication of the Savage Mind, a critique of social anthropology and sociolinguistics, Goody (1977) suggests that language does not proscribe part-to-whole or sequential thought, but rather some of its written forms do. Particularly, tables and matrices, the very stuff which gird sociolinguistic theories of cultural melaning. These devices seek pattern and ignore variation. Of the use of tables in anthropological interpretations of storytelling in ancient societies, Goody (1977: 119) claims: "Any model we may erect is simply an averaging of the variations in the number of versions we happen to have collected; it is a statistical artifact, not the authorized version." The worst consequence of reducing such complex linguistic behavior to a table is the suggestion that repeated features are in fact valued; a storyteller may be admired for giving "a new twist" to an old tale, rather than for presenting dutifully a sequence of familiar features (Goody 1977: 117).

A critique

of functional

linguistics

and its applications

to writing

267

We can easily make the same criticism of functionalist descriptions of written genres. The analyses of expository discourse offered by Hoey in On the Surface of Discourse reveal the repeated formulaic features of published texts. Writers and teachers are advised to study the models because they show how written discourse "works". Hoey (1983: 1) offers his research "to those who want to mend either their own or others' damaged discourses", suggesting that the purpose of model building in discourse analysis is to discover ways to correct structures that are incorrect. One wonders what might be called an incorrect structure. Is it one that does not follow the received genre for a particular kind of communication? And if so, how does a variation come to be regarded as a mere indiscretion rather than a clever foray into new forms of expression? We favor received genres because they confirm established cultural values, not because they make room for new information. Studies of the sociology of science verify this claim, demonstrating that report genres disguise the actual sequence of events in laboratory work. In a meticulous longitudinal study of the revisions of a scientific paper, Knorr-Cetina (1981) demonstrates that the final paper reduces the linguistic activity surrounding laboratory work (including a great deal of oral discussion, written records of laboratory data, written observational logs, and written outlines and drafts of reports) to schemes that allow it to be related to other formal reports. For instance, the results and discussion section of a report "effectively denies the interdependence of methods and results which ruled laboratory reasoning [at the time the experiment was conducted] by relating the results not to the process of production, but to other results" (Knorr-Cetina 1981: 123). Objects under study in a laboratory, in fact, become recognized as discrete structures only after language has been manipulated to signify them in this way. The culturally dominant rhetorical patterns in science are designed to produce "facts", which, though valued, may be quite distinct from "truth". In fact, conventional scientific reporting imposes an idealized temporal order on the conduct of laboratory activity and creates a distinct spatial structure for objects under study to establish their reality as fact. Scientific progress can be made when scientists remove qualifications from such statements and cast them as declarative fact. Latour and Woolgar (1979) confirmed this phenomenon in their study of the laboratory discovery of the structure of TRF, a hormone. At one point in their experimental work, one of two competing laboratory groups investigating the structure of T R F simply changed a qualified statement about

268

Barbara

Couture

the hormone's structure to an unqualified assertion. The laboratory had previously claimed that T R F is "like", or similar to, "the synthetic compound Pyro-Glu-His-ProNH2". They now claimed: "Guillemin and Schally have established that T R F is Pyro-Glu-His-Pro-NH2" ( L a t o u r Woolgar 1979: 147). This simple change in syntax allowed both labs to end a debate about T R F ' s structure and seek more funds for a new arena of research — the manufacture of T R F as a substance. Functional theories celebrate the capacity of language to order time and space in the ways shown here. Yet in citing how language so manipulates the environment, functional theories fail to address the consequence of imposing order where none may be. This problem is rarely acknowledged by teachers who apply functional theory in writing instruction. How does a writer represent experience as a continuous state of becoming, unfetered by the spatial constraint of community boundaries and the temporal constraint of human mortality? Writing that is so startlingly different would be dismissed by current applications of functional theory as "non-writing" or as a deviant aberration of an existing genre, simply because the "facts" constructed by the theory to describe existing genres do not validate such deviant expression. And how does a writer question the merit of reducing a complex technical investigation to the instrumental "fact-producing" discourse commonly practised in the sciences? To do so is definitely risky business. Yet risky writing may be required to keep a writer's eye on the truth. Functional theory — whose object is to reveal language's meaning potential — sadly fails to address such dilemmas.

3.3. Language polarizes

reality

Functional theories assert that reality is either rationally ordered or unpredictably chaotic. This polarized perception of the world persists in grammatical, rhetorical, and literary functional theory. In Halliday's functional theory (1985), a rational reality is shown to emerge through grammatical structure. When the subject of a clause, logical agent, and informational theme share the same syntagmatic slot, as in the example The duke gave my aunt this teapot, sensibility reigns (Halliday 1985: 35). Disintegration or a kind of chaos results when paradigmatic and syntagmatic choices are not parallel, as in This teapot my aunt was given by the duke. Further, in Halliday's theory, grammatical structure is shown to relate congruously or incongruously to the outside world. Congruent expression corresponds to "literal" meaning (Halliday

A critique

of functional

linguistics

and its applications

to writing

269

1985: 321); as an example, Halliday (1985: 322) offers the "unmarked" literal expression: Mary saw something wonderful which contrasts with the "marked" metaphorical expression: Mary came upon a wonderful sight. Such distinction presumes that a rational reality exists which is consistent with unmarked, literal language; and that grammatical metaphors exist outside that reality as incongruous perceptions. A polarized representation of reality as either rational or incongruous is reflected in both practical rhetoric and literary theory. Writing teachers advise students to compose texts which create a rational order, backing up their claims with empirical research on readability. They can cite tests of reading comprehension which demonstrate that "the active form of a statement leads to easier recall and verification than the passive" (Klare 1977: 3) and that sentences are easier to comprehend when topics and agents are named at the beginning (Huckin 1983: 97). Or they can cite functional theory and thus advise writers to present familiar information before new information; to put topics before comments (Selzer 1983: 82); and to highlight themes through repeating them in headings and topic sentences at the beginning of paragraphs (Huckin 1983: 101). All these linguistic devices project a rationally ordered world. Many applications of functional theory to practical writing suggest that writing is dysfunctional when it does not project a rational order. Reports work best in organizations, Mathes —Stevenson (1976: 31) insist, when they reflect the rational sequence of problem-solving: a report should first state the problem, then elaborate the assignment, and then identify its purpose to propose a solution or a recommendation. Yet in a study of professional writing in fifteen corporations Brown —Herndl found technical professionals to repeatedly prefer writing which reflected the actual chronology of their personal work rather than the rationalized problem-solving scheme preferred by their managers. Interviews revealed that managers found this behavior inappropriate and unprofessional, complaining, as one did, that writers continued to develop "some kind of chronological recitation of what they did, as opposed to selecting from it those key things that will influence a decision-maker to accept their recommendation" (Brown —Herndl 1986: 20). In contrast, the engineers they supervised felt that their narratives better described the "process" of their technical work, both creating a good defense for their professional decisions and providing additional detail that may be inspiration for new solutions (Brown — Herndl 1986: 20). In short, their narratives of personal activity conflicted with the problem-solving focus which is the preferred "reality" of organizational settings.

270

Barbara

Couture

Functional rhetorics often view conflict between the aims of the writer and the audience as a dissonance to be resolved and erased in writing, rather than encouraged as a potential source of new information. For example, Mathes —Stevenson (1976: 50) write in their textbook: "[T]he unreflective report writer, especially one who has failed to put on his rhetorical hat, mistakenly designs his report in terms of what he has done, that is, from particular to general [and] produces a report which is not effectively designed to be read by audiences who have solely an instrumental interest in what he writes." A social ideology which presumes that business professionals only have instrumental purposes in reading is hardly disguised here. Apparently, the aim of the functional rhetoric is to teach writers to remove dissonance between the reality perceived by themselves and their readers by discounting the writer's perspective. Literary theory, in contrast, views the creation of dissonance as an inventive strategy. As Kent (1986: 50) points out, the inventive literary text, as opposed to the automatized or ordinary text, is more likely to display paradigmatic or syntagmatic foregrounding. This amounts to countering the readers' expectations for a conventional representation of reality by presenting a different paradigm. Such happens, Kent (1986: 54) claims, in Henry James' "The Beast in the Jungle," where opening expectations for the typical sequence of events in a Gothic romance are not fulfilled in the end; the Gothic features serve instead as a metaphor for the protagonist's "psychological condition". New realities are invented by the reader of modern literature, Barthes [1977] claims, because these "writerly" texts force the reader to alter their conventional perception of reality. He contrasts classical or "readerly" texts which affirm socially accepted views of reality with modern "writerly" texts which disrupt conventional perspectives: Classical language is a bringer of euphoria because it is immediately social. There is no genre, no written work of classicism which does not suppose a collective consumption .... In [modern literature], Nature becomes a fragmented space, made of objects solitary and terrible, because the links between them are only potential. Nobody chooses for them a privileged meaning, or a particular use, or some service; nobody imposes a hierarchy on them, nobody reduces them to the manifestation of a mental behaviour, or of an intention, of some evidence of tenderness, in short. (Barthes [1977]: 49, 50.)

Barthes' literary theory, like functional grammar, presumes that language represents reality and chaos as polar opposites; they are alternate extremes. The former substantiates the social milieu while the latter aggravates individual discontent.

A critique

of functional

linguistics

and its applications

to writing

271

Applications of functional literary theory to classroom writing accept and encourage this polarized view. In an unusual textbook called An Alternate Style, Weathers (1980) argues that there are two writing styles: grammar A "has the characteristics of continuity, order, reasonable progression and sequence, consistency, unity, etc." (1980: 6); grammar B, in contrast, is an "alternate grammar" and has the "characteristics of variegation, synchronicity, discontinuity, ambiguity, and the like" (1980: 8). Weathers views the use of grammar Β as liberating, because it allows writers to escape conventional forms which may present a false picture of an ordered reality. Yet, this alternative forces writers to adopt a "variegated, discontinuous, fragmented grammar of style [in the belief that it accurately] corresponds to an amorphous and inexplicable universe ..." (Weathers 1980: 12). In short, writers have two choices: to regard reality as ordered, harmonious, and stable, or to view it as chaotic and alienating. Functional theory as it has been applied to writing leads us to deny a world where ambiguity coexists with certainty, movement with inaction, purposefulness with randomness, and hateful rage with perfect love. The inventive metaphor is incongruent with reality; the narrative technical report a thoughtless protest against a more instrumental order; and writerly literature a salute to the alienating milieu of modern times. Applications of functional theory to the teaching of writing encourage monolithic expression which reduces the complex texture of human existence to the sad dichotomy of order versus chaos — each admired separately in their socially appropriate contexts. Such a retreat, I believe, eventually will triumph in ensuring that a recidivist discourse, committed to documenting and preserving current social practice, regardless of its merit, will dominate among students, lawmakers, teachers, and, sadly, linguists as well.

4. Functional theory redeemed Throughout this discussion, I have sought to demonstrate that the scientific premises of functional theory emphasize the social value of conventional meaning, the structured nature of that meaning, and the correspondence between language and a polarized reality. Further, I have claimed that these emphases altruistically project a benign society, maintained through accepted language practices. If functional theories of language remain dedicated to validating tradition, we will continue to dismiss linguistic expression which denies it. I believe something inherent

272

Barbara

Couture

in the functional approach makes looking back to what was more significant than projecting what can be. And this limits scholarship in language function in important ways. If linguists describe meaning as it is socially structured, how can they account for individuality without discounting it as idiosyncrasy or downright ornery behavior? Pratt (1987) argues that this is a quandary perpetuated in sociolinguistics. She cites as an egregious example Labov's criticism of the speech of a "black middle-class speaker, asked by a white interviewer to give his views on the supernatural". As Pratt (1987: 56) reports, Labov bemoans the fact that the "speaker 'fails' to speak in BEV [Black English Vernacular], and instead produces the 'turgid, redundant, bombastic and empty' English of the American middle class." In short, he asserts arbitrarily that community affiliation must dictate individual behavior, a position consistent with racist ideology, if we are to give it its least charitable interpretation. Pratt concludes that a "critical project" which grants authority to community approval over individual choice not only overlooks forms which are not authorized by community assent, but also actively marginalizes them as less valid. She notes that the lack of attention paid to "speech forms associated with women and women's spheres", for instance, "is symptomatic ... of an extraordinary, really pathologically narrow conception of what 'the normal system' or 'straightforward communication' is" (Pratt 1987: 55). Why do theories which should explain all kinds of meaning invite such recidivist applications? And if linguists claim all meaning is generically structured, how can they account for language which maintains structural fluidity or appears as a structure in progress? How can they explain, for instance, the new literary forms emerging among women writers? In a quick review of the writing style of Gertrude Stein, Ntozake Shange, Toni Morrison, Jill Johnson, and others, Stanley —Wolfe (1983) conclude that women are creating their own writing style. Women's language emphasizes a becoming, a process, rather than a fixed point, a stasis. They claim: "As women strain to break through the limits of English, certain patterns begin to emerge, recurrences of similar syntactic ways of ordering perception that is always moving and often contradictory" (Stanley — Wolf 1983: 136). How can this be understood by theory which persistently divides experience into parts and wholes, and beginnings, middles, and ends? And finally, if linguists assume that reality is rationally ordered and that deviation from that order creates incongruity, how can they fairly evaluate language which changes that order? Or worse, how can they

A critique

of functional

linguistics

and its applications

to writing

273

admit the value of change at all? In her criticism of sociolinguistics, Pratt claims that the desire to see social reality as coherent "displaces other quite compelling social logics" (1987: 52). She argues that analyses of classroom discourse which interpret classroom speech as repeating cycles of regular interchanges must necessarily be overlooking something: "For indeed, classrooms are supposed to be places where things change all the time, where pupils do and say different things from one day to the next because education and socialization are going on" (1987: 52). I believe that functional theories can counter the urge to relapse into the commonplace, the already tried, and the accepted through focusing not only on language as it already has been said, but also on language as it is yet to be. This, I believe, will be their greatest challenge: to explain the linguistic moment where the social confronts the individual, the structured confronts the fluid, and the established confronts the novel. New concepts which accept the merging of static and dynamic orders are already changing the development of systemic functional theory. Halliday's (1985) theory of grammatical metaphor, which I cited earlier; Nesbitt —Plum's (1988) elaboration of probabilistic grammar; and Martin's (1985) and Ventola's (1987, 1989) explanations of genre as social process are three promising developments which in their infancy already show extraordinary potential to reveal the power of language in movement. Analysis of grammatical metaphor — albeit the current unfortunate distinction between congruent and incongruent expression — can uncover the linguistic sources of dynamic meaning in both literary and rhetorical text. For example, grammatical metaphor accounts for the impact of Toni Morrison's repeated expression "the crawling-already? baby" in her startling novel Beloved (1987). This phrase describes a child whose death at the hand of her mother is ambiguously suggested at the beginning of the novel, and whose life is uncertainly returned at the end. The expression points to this questionable status through conflating an elliptical clause with a noun phrase. Grammatical metaphor also can recover the linguistic history of metaphors created by organizations to mediate change or spread "new ideologies". In business settings, metaphor is used to promote managerial philosophies, negotiate conflicts in value systems, and important values from domains outside the organization. In short, metaphors are "linguistic artifacts" used "as tools for control" (Czarniawska-Joerges— Joerges 1988: 179 — 187). Through unpacking grammatical metaphors we can better assess the alternate realities which they foreground for the

274

Barbara

Couture

writer and reader alike. Consider, for instance, the value of analyzing the grammatical structure of these metaphorical statements taken from business settings: One has to increase the opportunities for people to come nearer to the decisions (and the reverse); There are to be short lines between execution and decision; Decision makers must anchor themselves better in reality (Czarniawska-Joerges — Joerges 1988: 187). Through identifying "literal" grammatical equivalents for such utterances, we can study textual foregrounding, that is, the device by which a secondary meaning is brought forward through unusual syntax or lexical choice. Another way functional theory can address language change is through the study of probabilistic grammar, that is, the assessment of the likelihood of one grammatical choice being made over another in the environment of other choices. In a study of oral texts collected in fifty sociolinguistic interviews, probabilistic analysis allowed Nesbitt —Plum (1988: 11, 26) to conclude that the grammatical "combination of parataxis and locution and hypotaxis and idea" opposes the less probable combination of "parataxis and idea and hypotaxis and locution". This gave empirical support to Halliday's theoretical observation that the former combination of choices from syntactic and logical-semantic systems is more common than the latter combination. Probabilistic grammar can be employed to test the validity of such observations over time, thus providing a quantitative and qualitative description of language change. One of the strengths of probabilistic grammar is that it implicitly concedes the limitations of paradigmatic modeling through highlighting the dynamic implications of choice: a single expression may in fact represent a choice from system "A" followed by a choice from "B", or a choice from "B" then "A", and possibly the simultaneous choice from both. This latter option might not be impossible for the "chooser", a human being whose psychic capabilities for producing language are largely unknown. But it can be prohibited in the "either/or" apparatus of current systemic theory. This is not a cause for despair; it is only a humble reminder that it is not the theory that governs language potential, but rather quite the opposite. Probabilistic grammar, it seems to me, comes very close to examining the process of sequential and simultaneous choice as it is controled by speakers and writers. It is an exciting and highly worthwhile development. Finally, I am encouraged by the genre studies of Martin, Ventola, and others who have expanded the potential of systemic theory to evaluate text in the process of becoming. In a speculative essay, Martin (1985: 251) suggests that genre provides an "underlying semiotic" for text con-

A critique

of functional

linguistics

and its applications

to writing

275

struction; it directs the stages of a text's formulation. Functional linguists traditionally have represented genres as schematic structures. As Martin (1985: 251) defines them, schematic structures show "the positive contribution genre makes to a text: a way of getting from A to Β in the way a given culture accomplishes whatever the genre in question is functioning to do in that culture." To regard genre as an underlying semiotic, Martin (1985: 2 5 8 - 2 5 9 ) concludes, is to suggest that it has both static and dynamic dimensions. Genre can be realized statically in actual text and manipulated dynamically in the composing process. A model of genre as an underlying semiotic could take two forms: one depicting the resulting text and another the process of producing a text. Research in systemic linguistics has just begun to attempt dynamic representations of genre. Ventola's experiments (1987) with a flow chart schematization of conversational service encounters demonstrates the promise of this method. At any given stage in the production of this genre, the flow chart can handle the possible and probable moves of the primary and secondary "knowers" (that is, holders of information on the topic at hand) in the exchange (Ventola 1987: 73 — 74). To date, flow chart analysis has been applied only to conversational structures. It seems to me that such representation of "dynamic systems" has promise for depicting the process of composing and reading written language as well. Dynamic representation may be used, for instance, to compare writers' selections from synoptic systems to produce, say, a laboratory report or a feasibility study. Probability analysis might be superimposed here to suggest the combination of patterns and their sequence which is more probable for the experienced writer than for the novice, or more probable for the writer in a stable job situation than for the writer facing organizational change. Representation of genre as a dynamic system also has exciting implications for studies of the reading process. In my own research, I have proposed a schematic representation of the interpretive "moves" a reader makes when proceeding through a text. The scheme treats the text as a conversation; it analyzes textual forms as they point to underlying systems of meaning, presumably shared by intra- and extra-textual participants. The scheme also suggests the intertextual exchange among a given text and others in its field which contribute to its meaning. Such a representation attempts a linguistic articulation of theories of reader response and textual dialogism (Couture 1987: MS). What all of the new developments in systemic functional theory share is the capacity to represent text as a process whose outcome is directed

276

Barbara

Couture

both by choices of linguistic form and choices from semiotic systems which form represents. And this by implication points to the ultimate potential of the "chooser", the human being who at some point decides to employ a particular meaning or another in the process of composing or interpreting a text. It is this element — the human "chooser" — that makes language possible and asserts its social function. It is human choice that privileges one meaning over another or accepts shades of meaning. It is human choice which obliterates the neat discreteness of the systems we construct to represent how language works. It is human choice which creates from a static text a dynamic interpretation, allows product to become process, conflict to become compromise, presumed chaos to become meaningful, and seeming hopelessness to become hope. It is human choice which rejects language that favors social systems as they are and which voices the beginning of change. In the end, my criticism of the premises of functional theory and its application to writing is not a call to revolution. The analytic potential of functional theory is — as I am sure my examples have shown — highly useful. Rather than a call to revolution, this chapter is, I hope, an invitation to experimentation. Functional theory can do more than it has done, and it can do so by daring to explore the uncertain, by daring to speculate about the human values beyond its scientific premises and, hence, about the ways language may change. Such work will come about, I think, when we approach language as a process under human control, one which at every step in its composition and interpretation has the potential to achieve the social altruism which we believe to be its promise.

References Allerton, D. J. —Edward Carney —David Holdcroft (eds.) 1979 Function and context in linguistic analysis. Cambridge: Cambridge University Press. Anderson, Paul V.—John R. Brockmann — Carolyn R. Miller (ed.) 1983 New essays in technical and scientific communication: Research, theory, practice. Farmingdale, N Y : Baywood. Barthes, Roland 1953 Writing degree zero. (Trans. Annette Lavers —Colin Smith). N e w York: Hill and Wang. [1977] [Reprinted N e w York: Hill and Wang.] 1970 S/Z. Editions du Seuil: Paris. [1974] [Reprinted: S/Z. (Translated by Richard Miller.) N e w York: Farrar, Straus, and Giroux.]

A critique of functional

linguistics and its applications to writing

277

Benson, James D. — William S. Greaves (eds.) 1985 Systemic perspectives on discourse 1. Norwood, NJ: Ablex. Berry, Margaret 1976 Introduction to systemic linguistics 2: Levels and links. New York: St. Martin's Press. Borgmann, Albert 1984 Technology and the character of contemporary life: A philosophical inquiry. Chicago — London: The University of Chicago Press. Brown, Robert L. —Carl G. Herndl 1986 "An ethnographic study of corporate writing: Job status as reflected in written text," in: Barbara Couture (ed.), 11—28. Bruffee, Kenneth A. 1986 "Social construction, language, and the authority of knowledge: A bibliographical essay", College English 48: 773 — 790. Christie, Frances 1985 Language education. Victoria: Deakin University. Cooper, Charles 1975 "Oral and written composition", English Journal 64: 72 — 74. Couture, Barbara 1987 A description of the exchange functions in written discourse. (14th International Systemic Workshop. University of Sydney, Sydney, Australia. August 2 4 - 2 8 . ) [Unpublished MS.] Couture, Barbara (ed.) 1986 Functional approaches to writing: Research perspectives. London —Norwood, NJ: P i n t e r - A b l e x . Czarniawska-Joerges, Barbara — Berward Joerges 1988 "How to control things with words: Organizational talk and control", Management Communication Quarterly 2: 170—193. Fabb, Nigel —Derek Attridge —Alan Durant —Colin MacCabe (eds.) 1987 The linguistics of writing: Arguments between language and literature. New York: Methuen. Fawcett, Robin P. — David Young (eds.) 1988 New developments in systemic linguistics. London: Pinter. Glassman, Myron —Thomas E. Pinelli 1985 "Scientific inquiry and technical communication: An introduction to the research process", Technical Communication 32, 4: 8 — 13. Goody, Jack 1977 The domestication of the savage mind. Cambridge: Cambridge University Press. Graff, Harvey J. 1979 The literacy myth: Literacy and social structure in the nineteenth-century city. New York: Academic Press. Halliday, M . A . K . 1985 An introduction to functional grammar. London: Arnold. 1980 How is a text like a clause? Nobel Symposium on Text Processing. Stockholm. August 1 - 1 5 . [Unpublished MS.] 1979 "Modes of meaning and modes of expression", in: D . J . Allerton — Edward C a r n e y - D a v i d Holdcroft (eds.), 5 7 - 7 9 .

278

Barbara

Couture

Helm, June (ed.) 1967 Essays on the verbal and visual arts: Proceedings of the 1966 Annual Spring Meeting of the American Ethnological Society. Seattle: University of Washington Press. Hoey, Michael 1983 On the surface of discourse. London: Allen and Unwin. Horkheimer, Max —Theodor W. Adorno 1944 Philosophische Fragmente. [1972] [Reprinted: Dialect of enlightenment. (Translated by John Cummings.) New York: Herder and Herder.] Huckin, Thomas N. 1983 "A cognitive approach to readability", in: Paul V. Anderson —John R. Brockmann —Carolyn R. Miller (eds.), 9 0 - 1 0 8 . Jakobson, Roman 1960 "Closing statement: Linguistics and poetics", in: Thomas A. Sebeok (ed.), 350-377. Kent, Thomas 1986 Interpretation and genre: The role of generic perception in the study of narrative texts. Lewisburg: Bucknell University Press; London —Toronto: Associated University Presses. Klare, George R. 1977 "Readable technical writing: Some observations", Technical Communication 24, 2: 1 - 5 . Knorr-Cetina, Karin D. 1981 The manufacture of knowledge: An essay on the constructivist and contextual nature of science. Oxford: Pergamon Press. Labov, William —Josua Waletzky 1967 "Narrative analysis: Oral versions of personal experience", in: June Helm (ed.), 1 2 - 4 4 . Latour, Bruno —Steve Woolgar 1979 Laboratory life: The social construction of scientific facts. Beverly Hills: Sage. Lock, Grahame 1976 "Introduction", Essays in self criticism. (Translated by Louis Althusser.) London: Lowe and Brydone. Martin, J. R. 1985 "Process and text: Two aspects of human semiosis", in: James D. Benson — William S. Greaves (eds.), 2 4 8 - 2 7 4 . Mathes, J . C . —Dwight W. Stevenson 1976 Designing technical reports: Writing for audiences in organizations. Indianapolis: Bobbs Merrill. Medawar, Peter 1984 The limits of science. London: Oxford University Press. Morrison, Toni 1987 Beloved. New York: New American Library. Nesbitt, Christopher —Guenter Plum 1988 "Probabilities in a systemic-functional grammar: The clause complex in English", in: Robin P. Fawcett — David Young (eds.), 6—38.

A critique of functional

linguistics and its applications to writing

279

Pike, Kenneth L. 1964 "A linguistic contribution to composition". College Composition and Communication 15, 82 — 88. Pratt, Mary Louise 1987 "Linguistic Utopias", in: Nigel Fabb —Derek Attridge —Alan Durant —Colin MacCabe (eds.), 4 8 - 6 6 . Sebeok, Thomas A. (ed.) 1960 Style in language. Cambridge, Mass.: The M . I . T . Press. Selzer, Jack 1983 "What constitutes a 'readable' technical style", in: Paul V. Anderson —R. John Brockmann —Carolyn R. Miller (eds.), 71 —89. Stanley, Julia Penelope —Susan J. Wolfe 1983 "Consciousness as style; style as aesthetic", in: Barrie Thorne —Cheris Kramarae — Nancy Henley (eds.), 125 — 139. Strong, William 1973 Sentence combining: A composing book. New York: Random. Thorne, Barrie —Cheris Kramarae — Nancy Henley (eds.) 1983 Language, gender, and society. Rowley, Mass.: Newbury House. Titunik, I. R. 1973 "The formal method and the sociological method ( M . M . Baxtin, P.N. Medvedev, V. N. Volosinov) in Russian theory and study of literature", Marxism and the philosophy of language, in: V.N. Volosinov (ed.), 175 — 200. Ventola, Eija 1987 The structure of social interaction: A systemic approach to the semiotics of service encounters. London: Pinter. 1989 "Problems of modelling and the applied issues within the framework of genre", Word 40: 1 2 9 - 1 6 1 . Volosinov, Valentin N. 1973 Marxism and the philosophy of language. (Translated by Ladislav Matejka — I. R. Titunik.) New York —London: Seminar Press. Weathers, Winston 1980 An alternate style: Options in composition. Rochelle Park, NJ: Hayden.

Grammar, technocracy, and the noun: Technocratic values and cognitive linguistics Paul J. Thibault

1. The "cognitive-science" paradigm and technocratic discourse The emergence of a "cognitive-science" paradigm in linguistics and discourse studies over the past twenty years or so has now reached hegemonic proportions. In its technical discourse, this paradigm typically combines the fields of cognitive and instructional psychology, artificial intelligence, and an increasingly individual-centred linguistics. With its positivistic and experimental traditions, it rejects any overt appeal to value judgements in the pursuit of theoretical propositions which can be shown to be "true" or "false" according to empirical criteria. Some recent work in systemic linguistics has begun to discuss the axiological dimension of discourse meaning, in ways which build on, and extend, Halliday's interpersonal metafunction in conjunction with Mikhail Bakhtin's (1981) writings on the evaluative orientation of all forms of discourse (Lemke 1987 [1990]; Thibault 1988, 1989). Lemke's (1987 [1990]) paper on technocratic discourse in education policy-making documents argues that the technocratic discourse of policy making is articulated through combining and condensing two distinct thematic formations: one, the technical discourse of "scientific fact", free of value judgements and human interpretation; and the other, the discourse of expert knowledge, "value-free" and voiced by a technocratic managerial elite for policy making and decision making in the solution of social and educational problems. We can say, following Lemke, that the axiological orientation of technocratic discourse remains implicit, whereby it succeeds in constructing a seemingly neutral metadiscourse for formulating social policies and decisions. This orientation conceals the way technocratic discourse constitutes a project for engineering social change, without making explicit the valuecriteria behind the practices it recommends in order to bring this about. In this chapter, I will analyze some texts from the cognitive-science paradigm and its extension into educational practice. My purpose is two-

282

Paul J. Thibault

fold. First, I want to show how the lexicogrammatical and discourselevel features of these texts index the social practices and positionings which they participate in and help to enact. Second, I want to discuss a broader social-ideological problem raised by the cognitivists' ways of talking about language and the individual. The heteroglossic alliance of the technical registers of cognitive and instructional psychology, statistical analysis, and educational practice on the one hand; and the discourse of educational policy making on the other, is successfully bringing about a shift towards a "value-free" technocratic discourse of "neutral" facts and their implications for educational practice. In other words, scientific or technical registers are hybridized with the discourse of a managerial social elite in the interests of a technocracy which seeks to impose its own models of social planning and control by bypassing any discussion of its own, or alternative values and politics. The discourse of "cognitive science" is just such a hybrid.

2. Technical discourse and the nominal group The principal lexicogrammatical features of technical registers have been noted by a number of systemic linguists (e. g., Halliday 1967 [1977], 1988; Lemke 1987 [1990]). These include (i) a preference for non-actional (relational) processes rather than processes of material action; (ii) extensive use of nominalization as a means of "packaging" the processparticipant relations into a single grammatical entity; (iii) a relatively low incidence of animate human Agents, especially those which index the communicative and other roles of the researchers themselves in the text — their place being taken by abstract and reified nominalizations. The extensive use of nominalization also means that a high proportion of the lexical content of the text is encoded in these grammatical forms. Halliday (1967 [1977]: 13, 1988: 169) has shown that technical registers use nominalization as a means of constructing causal and other types of logical relations between processes within the grammatical structure of a single clause. Here is an example from a cognitive linguistics text: (1)

Communication is a process involving two devices. (Sperber —Wilson 1986: 1)

information-processing

This clause contains two nominalized processes, which are related to each other by a relational : identifying process in the verb is. The first of these nominal groups, i.e., communication, is a verbal noun, which packages

Technocratic

values and cognitive

linguistics

283

into the one grammatical entity an entire set of (potential) processparticipant relations. These can be expanded into a clause-level activitystructure of the following type: "x communicates to y." The second nominal group enables the various processes and participants encoded to be packaged as a single grammatical entity as in Figure 1. a

process

Deictic : T h i n g

[[involving

two information

processing

devices]]

Qualifier Process:

Attribute

Relational

Figure I. N o m i n a l group, s h o w i n g experiential structure

These process-participant relations are related to each other through the Deictic-Thing-Qualifier structure in the experiential semantics of their nominal group. Potentially, this can be expanded into a clause-level set of process-participant relations, though it does not follow that there is always a one-to-one mapping of the one onto the other. In Example (1), the relational process in the verb is expresses a logical relation between the two sets of process-participant relations in the two nominal groups which it links. Here, this is a relation of expansion, whereby the second defines, adds to, or qualifies, the meaning of the first. Example (2) shows the same tendency: (2)

The performance of humans in comparing the strength of unlike assumptions is thus a powerful indication that strength, as a basic psychological concept applied to assumptions, is comparative rather than quantitative. (Sperber —Wilson 1986: 81)

The semantic Agent humans which does the "performing" is here postmodifier of the nominalized Head performance. This is then related (by is an indication) to a nominalized clause, itself also relational, in which the human presence is totally obliterated. Example (3), taken from the field of biochemistry, illustrates the ubiquitous "research shows" formula and its variations in technical registers (Lemke 1987 [1990]: 1 2 - 2 1 ) . (3)

The appearance of the same type of ionizing group observed in the VjK profile suggests that the ionizing group observed in the V/K

284

Paul J. Thibault

profile is involved with catalysis rather than substrate ( M o r r i s o n - S t o n e 1986: 832).

binding.

In examples like (3), it is the process of "researching", "studying", etc., which is reified, while the human Actor who did the researching or studying is downgraded in grammatical status, or simply not specified. Consequently, it is the reified nominalized process which takes on the semantic qualities of a participant. An entire activity, process, or process-participant relation, may be "referred to", or indexically presupposed, simply by being named with a single grammatical entity, the nominal group. Thus, as Halliday (1988) shows, a process-participant complex can be thematized as a nominal group in the clause in which it occurs, and then causally or otherwise logically related to some other such complex in a highly economical fashion. Lemke (1987 [1990]: 10) refers to this grammatical process as "condensation", which means that the complete activity structures can only be recovered by readers familiar with the relevant intertextual thematic formations. This is just one of the discourse strategies which technical registers use to divide the world into those who are initiated into its specialized ways of making meaning and those who are not (Lemke 1987 [1990]: 1 9 - 2 1 ; Wells 1960: 218). We shall now consider how the nominal group participates in the standard thematic-semantic patterns which are made in the technical discourse of cognitive science.

3. Thematic condensation in the discourse of cognitive science Let us consider Example (4), from Schank — Abelson's Scripts, plans, and knowledge: (4)

Plans are responsible for the deliberate behaviour that people exhibit. Plans describe the set of choices that a person has when he sets out to accomplish a goal. In listening to a discourse, people use plans to make sense of seemingly disconnected sentences. By finding a plan, an understander can make guesses about the intentions of an action in an unfolding story and use these guesses to make sense of the story. (Schank — Abelson 1977: 428)

We might say — as indeed the cognitivists do — that this passage "presumes a [technical] background knowledge" for its understanding.

Technocratic

values and cognitive

linguistics

285

But this "knowledge" consists of linguistically realized patterns of meanings, which this text may share with other field-specific texts — perhaps a very restricted number (Thibault 1986 a: 93 — 94). How are we to "understand" such an apparently simple item as plans'? This nominal group has generic and presuming reference, whereby it is assumed that the reader already knows the general class of "plans" to which it refers. We must be familiar with the thematic formations of the community where "plans" does not refer to architectural plans, cartography, or business projects, but to the "hierarchical structure underlying behaviour" in the discourse of cognitive psychology. Lexico-semantic studies (e.g., Lyons 1988; Coseriu — Geckeler 1981) have shown how individual lexical items may have a range of meanings according to the particular semantic fields in which they occur. But this depends on their taxonomic status. Lemke (1987 [1990]) has shown how lexico-semantic interpretations depend on the semantic "valeur" of an item in a particular field-specific intertextual thematic formation. This also explains the difference between the so-called "literal" and "figurative" uses of the same item. There are field-specific semantic valences which activate the item's "literal" meaning potential, and others which activate its "figurative" meaning potential (Thibault 1986 b). In Example (4), plans has such a field-specific semantic valence, which depends on the reader's ability to activate the field-specific meanings of the relevant technical register. Thus, the reader would need to know that the notion of "reinforcing feedback" in cognitive psychology is relevant to the field-specific meaning of the process noun plans. Additionally, the presuming reference of the Head noun in the nominal group the set of choices that a person has when he sets out to accomplish a goal can only be "understood" when placed in relation to notions like the "stimulus", "information", or "control" which are compared and tested in the feedback process (see Miller —Galanter —Pribram 1960 [1969]: 5 4 8 - 5 4 9 ) . Notice how the terms I have used to "explain" or contextualize those in (4) are themselves derived from the technical discourse of cognitive psychology. The items in (4), and the thematic-semantic relations among them, can be assigned to a common intertextual pattern for connecting PLANS, CHOICES, GOALS, INTENTIONS, and ACTIONS. A given text may reconstruct these networks of thematic-semantic relations by appropriate lexico-semantic and ideational-grammatical choices. Note that the capitals are used to indicate that PLAN, etc., function as superordinate thematic (not lexical) items in the field-specific formation of "Cognitive psychology". This means that other lexicogrammatical

286

Paul J. Thibault

locutions may substitute for them in the same formation (Lemke 1988: 85). What is required of the reader is a familiarity with the ways in which these items enter into specific ideational semantic relations with other items in the intertextual thematic formations of the technical discourse. All texts (spoken and written) only mean what they do through the relations which text-users construe between a given text and other texts in some intertextual formation. Lemke (1987 [1990]: 14) points out that it is a question of the "degree of condensation": in technical discourse the number of unrealized thematic items and relations — which are relevant for the contextualization of those items actually present — is considerably greater than for other discourse types. Let us consider Example (5), a paragraph from Robert de Beaugrande's book Text, discourse, and process: Towards a multidisciplinary science of texts. In the final chapter, the author puts forward some general proposals, based on his cognitive model of "text-processing" and "text-understanding", for teaching writing. The value-chains which are visually highlighted in (5) will be discussed in section 4: (5)

[1 ] The focus of a theoretically well-founded methodology of writing should be upon MOTIVATION A N D DECISION. [2] Learners who acquire workable standards for evaluating their own prose as a protocol of decisionmaking need not rely constantly on the teacher's feedback. [3 a] Instead, they can compare their text to their current motivations and goals [3 b] and revise inadequate decisions accordingly. [4 a] In this fashion, untrained writers can distribute their attention selectively during several phases [4 b] rather than trying to manage all writing operations successfully in the first run. [5] Presumably, this latter proceeding simply overloads processing resources. [6 a] We must therefore break the writing task down into sufficiently small subtasks [6 b] which any learners can manage. [6 c] irrespective of their prior experience and social background. [7 a] As training progresses, [7 b] the ability to co-ordinate more and more subtasks at once should rise in the same fashion [7 c] that active storage can hold more material [7 d] when larger, better integrated "chunks" are formed [...]. (de Beaugrande 1980: 286)

We can see how, in the very first nominal group The focus of a theoretically well-founded methodology of writing, it is as if the "methodology" does the "focusing" rather than the researcher (see above). Again, the reference

Technocratic

values and cognitive

linguistics

287

here is presuming. The key items motivation and decision are explicitly defined intratextually earlier in the book. The graphic prominence of these two words — both process nouns — foregrounds the fact that they are key thematic items, condensing activity structures which can be expanded into a range of social activities, such as "personal and scholastic interests"; "discussion with teacher"; "deciding on a writing topic"; etc. Some expansions will be more specific to the technical field than others; clearly, motivation and decision could be given many alternative meanings which do not fit into the field-specific requirements of this technical register, and these then raise a whole series of questions concerning the taken-for-granted practices and assumptions which the items entail. These issues are not raised in the text; instead the two process nouns are relexicalized in terms of a thematic formation already operating there, viz., the nominal group workable standards for evaluating their own prose as a protocol for decision-making. This monologic discourse strategy makes it impossible to relate these to any thematic-semantic alternatives. But this is not the main condensation in this nominal group. Whose standards? Worked by whom? For whom? How? To answer these questions, we must look further down in (5), where it is proposed that the writing task be "broken down" into sufficiently small subtasks which any learner can manage. What these subtasks are must be adduced on the basis of the wider intertextual thematic formations, for they are not spelt out in (5). But we can construe them from the thematic relations pivoted on the item plan. Thus: [Learner chooses series of hierarchically organized subactions to accomplish his/her goal]. The verbal noun training in the next sentence does not make explicit who the semantic Agent is. Is it [Teacher trains student] or is it [Student trains him or her self]? I would argue that this ambivalence is functional in the text in ways I shall discuss in Section 6. These examples illustrate the way in which the meaning(s) of any thematic item must be construed through its relations with other items in some wider formation, whether or not such items are realized in the given text. Note how the thematic formation "Cognitive psychology" interacts with another, which I shall gloss as "Writing pedagogy". Thus, my expansion of training in relation to the latter might also be rewritten as either [Teacher plans student's writing] or [Student plans his or her own writing]. Thus, I would not wish to suggest that the discourse of cognitive psychology is a thematically "pure" discourse, for it is itself a "hybrid" of various registers.

288

Paul J. Thibault

But what is the central ideology which these two texts share? These texts exemplify what Habermas (1968 [1970]: 81) identifies as "purposiverational action, which refers to either the organization of means or choice between alternatives." Habermas argues that this is linked to the "rationalization" of society and administration through scientific and technological development, adding that "it removes the total social framework of interests in which strategies are chosen, technologies applied, and systems established, from the scope of reflection and rational reconstruction" (Habermas 1968 [1970]: 82). Therefore, rather than "reflecting upon" and "rationally reconstructing" factors like "prior experience" and "social background" (de Beaugrande 1980: 286), we adapt these to our own needs, of which we are the self-managers. The de Beaugrande passage moves between the abstract exposition of the cognitive psychological principles involved and advocacy of their deployment in writing pedagogy. Thus the wider framework of social interests and values is bracketed out of the discussion, while responsibility for the "plans", "decisions", and "motivations" is shifted onto the individual, thereby absolving the teacher and the wider social processes involved from any central role in these. The discourse of cognitive psychology entails a model of "action regulated by its own results as the conjunction of rational decision and instrumental action" (Habermas 1968 [1970]: 87). This has led, through the development of technology, to the "step-by-step objectivation of the elements of that very system" (Habermas 1968 [1970]: 87).

4. Axiological meaning in the discourse of cognitive science We shall now discuss the three value-chains which I have identified in (5), in order to suggest how this text constructs a more global axiological orientation. The "Writing pedagogy" chain is given in bold; the "Cognitive psychology" chain is given in double-underline; and the "Social factors" chain is given in triple-underline. (Note: In (5) major and minor clause level constituents are numbered thus [1]; embedded (rankshifted) clauses are not numbered independently of the constituent in which they occur. Some locutions have been assigned to two value-chains and this is indicated in the coding of these according to the above key.) The "Writing pedagogy" chain shows a clear progression from "what should be" to "what can be", i. e., from the nominal group a theoretically well-founded methodology of writing to the ability to co-ordinate more and

Technocratic

values and cognitive

linguistics

289

more subtasks at once. Clearly, this is given high positive value in relation to the goals of "learning to write". The "Cognitive psychology" chain corresponds to the technical value-orientation of Example (5). Clearly, this is also given high positive value on account of its being the privileged technical discourse, in terms of which the "Writing pedagogy" chain is defined. The "Social factors" chain is much less represented and lexicosemantically would be hard to define on independent grounds. What matters is the way it is put into a particular kind of relationship with the two more dominant chains in the text. Taken separately, these three value-chains might all appear to be positively valued. Thus, a well-founded methodology of writing is positive, as are motivation and decision, and social background, at least from the middle-class orientation adopted here. But Example (5) constructs a specific metadiscursive stance among the three chains. The technical value-orientation of the "Cognitive psychology" chain privileges the "Writing pedagogy" chain and downgrades the "Social factors" chain. Nowhere in (5) does this occur explicitly. For instance, there are no overtly negative uses of interpersonal lexis which indicate a negative evaluation of the third chain. Instead, these valueorientations are construed in, and through, the metadiscursive relations among the three chains. The technical discourse of "Cognitive psychology" is strongly articulated both thematically and lexicogrammatically to the "Writing pedagogy" chain, which thereby takes on the high value of the former. In this way, "Writing pedagogy" is equated with thematic items such as motivation and decision from the "Cognitive psychology" chain. This occurs quite directly through the "equating" of nominal groups by means of the relational process verb be (see above). The high technical value of "Cognitive psychology" is then positively linked with something which would, generally speaking, be considered a worthy cause, i.e., enabling people to learn to write. At the same time, the "Social factors" chain is low in status, or even marginalized, relative to individualistic factors such as motivation, decision, and goals. The technical value-orientation of the "cognitive psychology" discourse assumes that it is the "motivations", etc., of the individual learner which are the variables which can be altered. Lemke (1987 [1990]) shows that all technical discourses privilege that which can be manipulated by the technical procedures of that discourse. In this way, the "Writing pedagogy" discourse is articulated to the strongly individualcentred claims of its technical partner. It might be said that anything which can improve the "motivations", "decision making", etc., of the

290

Paul J.

Thibault

individual in the pedagogic process is a worthy cause from which anyone can only benefit. Taken on its own terms, this is a value-orientation which is likely to be undisputed by most people. But notice that this text downgrades the importance of the third chain, which is backgrounded and only very weakly articulated in relation to the other two. Thus, factors such as teacher's feedback, prior experience, and social background (the only members of this chain in Example (5)) are not shown as important to the "Writing pedagogy" discourse, a fact which is likely to conflict with the perceptions of many teachers and pupils, and of people from social backgrounds different from those assumed by the dominant technical values in the text. The strongly co-articulated relation of the first two chains and the weakness of their relations to the third chain may be shown in Table 1 on the basis of lexicogrammatical and cohesive criteria which I have adapted and modified from Hasan's (1984) notion of cohesive harmony. The four principal lexico-semantic sets are represented as vertical chains of tokens. The tokens are numbered according to the clause they belong to. These semantic groupings are based on particular axiological (interpersonal semantic) domains. The horizontal lines show the interaction among the chains on the basis of interpersonal semantic criteria, which link items from two lexico-semantic chains on the basis of their cooccurrence in the same clause or group structure. This is done on the basis of the interaction between the value-chains and the modal orientation (obligation, ability, unmodalized) in the verbal group. The "Writing pedagogy" and "Cognitive psychology" chains show many such links and are therefore said to be strongly co-articulated on the basis of both axiological domains and modality. The "Social factors" chain is articulated only once in relation to modality (obligation), and not at all in relation to the "Writing pedagogy" and "Cognitive psychology" chains. The interaction of interpersonal and textual criteria in the establishment of these axiological orientations shows the field-like prosodies of their intersectings, mergings, overlappings, reemergings, and disjunctions. Any two chains linked by an arrow are in an interpersonal semantic relation glossed as follows: I. II. HI. IV. V.

= = = = =

"Writing" + "Writing" + "Writing" + "Cognitive" "Cognitive"

obligation ability unmodalized + obligation + ability

Technocratic values and cognitive linguistics

291

Table 1. Coarticulation of value-chains in the de Beaugrande text "Writing"

Modulation

3 a they 3 b (they) 4a untrained

"Cognitive"

3 a can compare biggesf, in (3) in hobby activities —• on hobbies·, and in (5) They had, apparently, to a greater extent than the other girls, embraced that pattern of life goals which they believed to be characteristic of adults. —* They had, apparently, embraced the pattern of life goals which they believed to be characteristic of adults to a greater extent than the other girls. The changes are minimal in terms of the kinds of problems discussed in Section 5.4.1. Only one referential change was made: that pattern —• the pattern. No thematic changes were suggested. Surprisingly, even other girls in (3) remained uncorrected. This was, in fact, also true for many other occurrences of other girls in the same text. It appears that, at least on this occasion, the reviser was processing the text sentence by sentence rather than textually. A textual revision would have involved paying attention to cohesive factors between text participants and thematic patterns. In general, many of the texts in the study would have been improved by simply checking that the major text participant discussed did actually appear in the head position of the nominal groups which realized the topical themes in constant theme patterning, provided that the writer seemed to have aimed at this patterning. Again, it appeared that the surface repairs made by revisers — this time concerning reference — did not actually improve the text very much. To sum up, this section has discussed some of the problems Finnish writers seem to have when building up reference chains throughout the text, and when referring to certain nominal groups, especially when they appear in thematic positions. The problems seemed to remain even after the native reviser's checking of the text. It appears that writers as well as revisers look at texts mainly sentence by sentence. Finnish writers' focus on sentences is understandable, since most likely when they were first learning English the teaching traditionally emphasized sentences as gram-

482

Eija Ventola—Anna Mauranen

matical units. Moreover, earlier the Finnish school system did not offer any instruction in extended writing; today composition writing is, however, included in the syllabus. Reference in English appears in general to be problematic to Finns (for further discussion, see Ventola forthcoming; Mauranen forthcoming). But many of the problems could be solved if instruction on how reference interacts with thematic patterns were used to help writers improve the cohesiveness of their texts. The initial research reported here on the reference and thematic patterning in Finnish researchers' texts also suggests a need for training language revisers in observing writers' patterns of reference and theme and suggesting reformulations to the writer. 5.5. An example of a textually motivated rewriting of a text This final section discusses the possible effects the textual changes suggested in the previous sections might have when implemented in a text. An example text was selected randomly from the data. It discusses protective measures taken at workplaces from the perspective of physics, and its title was "Selection of the filter shade number in welding". The global generic structure elements realized in the whole text were typical of an empirical report: introduction, materials and methods, results, discussion, references, and appendices. The main argumentation revolves around the factors influencing the choice of the best protective filter shade for welders. The part that we shall look at in detail here is the results section, where the writer explicitly summarizes the results of his experiments. Three different versions of the text will be presented. Example (12) is the original text written by a Finn; Example (13) is a revised version by a native English speaker; and Example (14) is a version which the authors of this chapter have rewritten, consciously playing with the textual systems introduced and discussed earlier in this chapter. After the description of the different versions and their characteristics, some reader reactions to the changes will be briefly commented on. (12)

Results (1) Welder's age and welding experience distributed normally, but the luminances viewed through filter plates distributed log-normally (fig. 1). (2) The latter was expected, because the glare sensation is known to be a logarithmic function of the source luminance.

Non-native

writing

and native revising

of scientific

articles

483

(3) Table 1 presents the age, welding experience, luminance and filter shade characteristics of the subjects. (4) When examining the shade numbers of the filter plates as a function of the welder's age a slight change towards lighter filters was observed. (5) The number of young welders was 36, whose average age was 20.7 years and welding experience an average 1.1 years. (6) Young welders and aluminum welders viewed at much lower luminances than the whole group of subjects. (7) Luminances increased both with the age of subjects and the welding experience (fig. 2 and 2). (8) The luminance difference was very significant between both young welders and whole group, and aluminum welders and whole group at 0.1 risk level. (9) The mean luminances were 540 cdjm2 for young welders, 401 cdjm2 for aluminum welders, 1586 cdjm2 for women, and 1448 cdjm2 for men. (10) The mean luminance for whole group of subjects was 1462 cdjm2. (11) There was no significant differences between the mean luminances of whole group, men, and women. (12) The questionnaire study showed that 69% of welders used fiber-board type welding helmet and the rest of the subjects used fiberglass type. (13) 21% of welders were fully contented with their welding helmets. (14) 8 welders had mounted a piece of leather on the lower edge of the welding helmet to protect their neck from radiation. (15) Welders complained of inadequate protection of the neck and heaviness of the fiberglass helmet. (16) The drawbacks of the fiber-board helmet were lack of space on the inside and a feature of bending the edges. (17) The edges bended outward when the helmet got wet. (18) The advantages of the fiber-board helmet was without exception its lightness. (19) The helmet with liquid crystal filter, which transmits visible light and darkens within 3 — 10 ms after the arc ignition, were highly approved by the subjects. In its original form the text in (12) is difficult to read and includes many realizations which are not native-like. The Finnish writer has difficulties in article use, lexical choices, and the construction of phrases and sentences. But the major difficulty lies in relating the expressed meanings logically to one another. For example, at the beginning of the text, the writer contrasts the welders' age and experience distributions with the luminances, which are, as we are told in the second clause, consistent with his expectations. The but in sentence (1) implies a contrast, although clearly the writer has simply meant to recount the distributions, and then offer a commentary on one of the differences only.

484

Eija Ventola — Anna Mauranen

In the second paragraph, it is similarly hard for the reader to conceptualize which variables and groups interact. Note, for example, how sentence (3) introduces the variables but not the experimental groups. The groups are, however, often (although not systematically) taken as themes. The point of view changes constantly, as themes and rhemes seem to have been constructed on a stream-of-consciousness principle. Similar features can also be pointed out in the third paragraph, which deals with various kinds of protective welding helmets and their benefits and disadvantages. Neither the benefits nor the disadvantages are grouped systematically. Rather, the helmets are discussed in such a fashion that the reader very soon loses track of what particular helmet was focused upon at a given point, and how this helmet differs from the others, and further how welders evaluated each helmet type. Example (13) is the language reviser's version of (12). (13)

Results (1) Welder's age and welding experience distributed normally, but the luminances viewed through filter plates distributed log-normally (Fig. 1). (2) The latter was expected, because glare sensation is known to be a logarithmic function of the source luminance. (3) Table 1 presents the age, welding experience, luminance and filter shade characteristics of the subjects. (4) On examining the shade numbers of the filter plates as a function of the welder's age a slight change towards lighter filters can be observed. (5) The number of young welders was 36, their average age was 20.7 and average welding experience 1.1 years. (6) Young and aluminum welders worked at much lower luminances than the group of subjects taken as a whole. (7) Luminances increased both with the age of subjects and the welding experience (Fig. 1 and 2). (8) The luminance difference was very significant between both young welders and the whole group, and aluminum welders and whole group at the 0.1 risk level. (9) The mean luminances were 540 cd/m2 for young welders, 401 cd/m2 for aluminum welders, 1586 cd/m2 for women, and 1448 cd/m2 for men. (19) The mean luminance for entire group of subjects was 1462 cd/m2. (11) There was no significant differences between the mean luminances of whole group, men, and women. (12) The questionnaire study showed that 69% of welders used fiber-board type welding helmet and the rest of the subjects used the fiberglass type. (13) 27% of welders were fully contented with

Non-native

writing and native revising of scientific articles

485

their welding helmets. (14) Eight welders had mounted a piece of leather on the lower edge of the welding helmet to protect their necks from radiation. (15) The welders complained of inadequate neck protection and weight of the fiberglass helmet. (16) The drawbacks of the fiber-board helmet were a lack of space on the inside and the tendency for its edges to bend. (17) The edges bent outward when the helmet got wet. (18) The advantage of the fiberboard helmet for all the workers who used it was its lightness. (19) The helmet with a liquid crystal filter, which transmits visible light and darkens within 3—10 ms after the arc ignition, were highly regarded by the subjects. The language reviser has made altogether 24 changes to the original text. Most of the corrections concern the use of the article (eight corrections, out of which six were additions of articles). Six of the changes have to do with nominal group structure; for example, in sentence (6) the whole groups of subjects > the group of subjects taken as a whole; and in sentence (15) inadequate protection of the neck > inadequate neck protection. Some lexical items have been changed: for example, in clause (10) whole > entire; in (19) approved > regarded. Spelling, the number of some nouns, tense in some verbs, and some prepositions have also been changed, although the incongruence between was referring to differences in (11) was missed by the reviser. In all, the corrections seem similar to those that revisers typically make, as described in Section 4 of this chapter. The changes naturally did improve the text somewhat. But whether the changes actually made the text more native-like textually and improved the comprehensibility of the text for the reader is questionable. Example (14) is the present authors' rewritten version of (12). (14)

Results (1) In the present study the welder's age and welding experience were found to be normally distributed. (2) Furthermore, as Figure 1 shows, the luminances which were viewed through filter plates were distributed log-normally. (3) This could be expected, because it is known that glare sensation is a logarithmic function of the source luminance. (4) Table 1 presents four variables: age, welding experience, luminance and the filter shade used. (5) The values of these variables are first presented for all subjects and then separately for men, women, young welders and aluminium welders.

486

Eija Ventola —Anna

Mauranen

(6) The values of men and women on all four variables did not differ significantly from those of the group as a whole, or from each other. (7) In contrast, the values of young welders (N = 36), whose average age was 20.7 and average welding experience 1.1 years, differed from the whole group in two respects. (8) Firstly, the younger welders tended to use darker filter plates than the entire group. (9) Secondly, as shown by Figures 1 and 2, they worked at lower luminances. (10) Their mean luminance value was 540 cdjm2 as opposed to 1462 cdjm2 of all the subjects. (11) The difference was highly significant at the 0.1 risk level. (12) The values of the last group, the aluminium welders, were similar to the mean values of the whole group, except with respect to luminance. (13) Their mean luminance value of 401 cdjm2 was similar to that of young workers, being thus lower than that of the group as a whole. (14) The difference from the total group mean was again highly significant at the 0.1 risk level. (15) One of the aims of the questionnaire was to find out what type of a helmet the welders used and whether they found the helmet satisfactory. (16) Two major types were found: 69% of the subjects used a fiber-board type welding helmet and 31% a fiberglass type. (17) As to workers' assessment of their helmets, it was found that only 21% were fully content with the type they were using. (18) The disadvantages associated with the most common type, the fiber-board helmet, were that it lacked space inside and that its edges tended to bend outward when the helmet got wet. (19) The advantage of this type was considered to be its lightness. (20) In contrast, the fiberglass type was found to be heavy. (21) Furthermore, the workers complained that it did not protect their necks adequately. (22) Therefore eight of the welders had mounted a piece of leather on the lower edge of the welding helmet to give their necks better protection from radiation. (23) In addition, the welders were generally highly satisfied with those helmets which had a liquid crystal filter. (24) The filter transmits visible light and darkens within 3—10 ms after the arc ignition. Example (14) has been rewritten with the aim of making the text easier to read. Owing to some additions of elements and changes of sentence structure, the number of sentences has increased by five. The paragraph organization has been changed once, when a text-predicting and textexplanatory paragraph has been added to start the results section. For

Non-native

writing and native revising of scientific articles

487

thematic reasons, some information which originally appeared in numerals has in the new version been verbalized and vice versa. The first category of changes here is the addition of reader-orienting material and material which links back to what has preceded in the article. This effect has been achieved, for example, by adding a marked theme, In this study, to the beginning of the section. The results are thereby contextualized. Similarly, comparisons, careful references, and connectors have been used to build up cohesion between the contrasted groups; see, for example, was similar to that of young workers, being thus lower than that of the group as a whole in sentence (13); again in sentence (14); and the whole of sentence (15). The second type of additions consists of those which help the reader to anticipate what will follow in the text. For example, the additions in sentences (4) and (5) orient the reader to the oncoming list of information: Table 1 presents four variables'. ...; In two respects ... in (7); and Two major types were found ... in sentence (16); and As to workers' assessment of their helmets ... in sentence (17). As can be seen, thematic choices and reference choices can be used effectively to improve the systematic unfolding of the text. As the third kind of text-clarifying devices, more connectors were introduced into the text to make the interrelationships between clauses explicit and show the global unfolding of the text as discourse. Thus in contrast is an addition to sentence (7), marking the semantic relationship of contrast explicitly. The use of connectors offers further support for such text-predicting references as In two respects in (7). The connectors firstly and secondly, in sentences (8) and (9) respectively, indicate the locations of the predicted text parts where the differences are expressed. In the last paragraph, the negative aspects of using the fiberglass helmet are discussed from sentence (20) onwards. In (20), the beginning of the discussion of this helmet type is marked by the connector in contrast; the contrast is made between the positive aspects discussed earlier and the negative aspects to be discussed next. The explicit internal conjunction furthermore in (21) indicates that there is more to be said about the negative aspects, and the added connector therefore in (22) marks the follow-up to the workers' complaints. A conscious effort was also made to change the thematic patterning, or thematic development, of the original text. For example in the third paragraph, the thematic pattern has been made more systematic by taking the groups and the variables discussed as themes, and the values obtained by the groups as themes, for example, in sentences from (6) —(14).

488

Eija Ventola — Anna

Mauranen

The three versions of the text, from (12) —(14), were shown to a group of Finnish teachers of English and to a group of native English teachers. The Finnish group saw all three text versions, whereas the English group saw only the two revised ones. We did not want to show the nativespeaker teachers a text which was not lexicogrammatically correct; therefore they only read (13) and (14). Both groups considered the textually reworked version, Example (14), to be the best. The Finnish teachers commented that it was only after having read the textually reworked version of (14) that they understood what the author actually meant, and they found the connector additions especially useful. The English group also considered (14) to be better than (13). The last paragraph of Example (13) was felt to be poorly organized, and the informational hierarchy overall to be more clearly presented in Example (14). The reorganization, in their opinion, made the text more readable, and they also thought that the language in the reworked version in (14) was more natural. But they did note that the text had become lengthier and that possibly the conciseness of the original (12) might correspond more closely to the expectations of the author's colleagues in the field than Example (14) would do. Experts in the field were not asked to read the three versions. But the reworked version, Example (14), was presented to the original author of the text, to ensure that no ideational content had been changed when textual changes were made. The author was also asked whether the reworked text fulfilled the aims he had set for his text and whether the content was, from his point of view, adequately expressed in the new version. The author considered the reworked version to be better than his own, or that of the language reviser, and decided to use Example (14) in his further elaboration of the article. Thus the textual changes which were made in Example (14) seemed to satisfy the original author's feeling of improvement.

6. Conclusion This chapter has reported some preliminary findings from a study on the linguistic difficulties that Finnish scientists and scholars face when writing academic texts in English. When interviewed, the writers seemed to expect that the native speakers of English to whom they sent their papers for revision would help them to improve the quality of their writing substantially. When revised articles were linguistically analyzed, however, it

Non-native

writing and native revising of scientific articles

489

was found that the texts were not often adequately improved by revision. The most obvious errors in the texts were corrected and the texts improved lexicogrammatically, but many cohesion problems remained. Thus, there seemed to be a mismatch between the writers' expectations and the actual revision practice. To see what cohesive problems remained after revision and how these might be solved by further textual improvements, a number of texts by Finnish writres were analyzed in detail. The first major area where Finnish writers' problems were apparent was in the use of connectors, which Finnish writers used less frequently than native writers in comparable texts. They also employed a relatively limited set of connectors, favoring some connectors excessively at the expense of variety. The revisers were mostly content to offer alternatives to the most frequently used connectors and to correct erroneously placed connectors within sentences. Usually the revisers did not add connectors to the texts, and thus the writers who under-used connectors were not helped. The Finnish writers also seemed to differ from their English colleagues in the handling of thematic development patterns. Their texts showed less thematic variation than the native writers' texts did. Typically, when writing in English, although not in their native language, Finnish writers would keep to one method of development within a passage, either a constant theme or a linear pattern, whereas native speakers employed different methods. Finnish writers also used fewer textual themes and had less lexical cohesion between themes. Thus, it appears that certain simplification processes, similar to the simplified connector uses, also dominated thematic choices. As far as the revisions were concerned, no conscious attention was paid by revisers to writers' thematic patterns. Sometimes revisers' grammatical corrections even ruined the writers' attempts at thematic and other discourse effects. Finally, Finnish writers had problems with the use of reference. Choices of articles were made unsystematically; for instance, the writers often left out a definite article where it should have realized a specific reference. The result was that the reader was not sure whether a presumed, specific participant was being tracked through the text or whether the reference was general. Often the writers also selected minor text participants instead of the major ones as heads of the nominal groups realizing themes. Thus the major participants occurred in less significant positions in themes. In revising, article use was frequently corrected in the papers. But a textual aspect was lacking in most revisers' work on references and so the consistency of writers' references to text participants was not checked

490

Eija Ventola — Anna Mauranen

systematically. Writers' unexpected shifts from specific to generic reference were left uncorrected. Revision of texts seemed to proceed from clause to clause, frequently without awareness of textual cohesiveness. To illustrate the effects of the textlinguistic approach introduced, an example text was written by the authors of this chapter. The rewritten text was considered an improvement on the original and also on the ordinary revision by a number of experienced teachers of English and, most importantly, by the original writer. The example shows how the textlinguistic tools introduced in this article can be applied to improve texts, and what this kind of textual approach might offer for writers and revisers. The general conclusion from this study is that a textlinguistic perspective would be beneficial for developing researchers' writing skills and the system of language revision. Finnish academics have had fairly traditional training in English during their school years, with teaching primarily oriented to sentences, not to texts. Good skills in written English are needed among Finnish academics and professionals, and training for writers needs to be organized at tertiary level. Educational authorities in Finland should be encouraged to organize such training. Writers would benefit from instruction which would provide textlinguistic tools for coding meanings coherently in texts. Building up writers' text-construction competence also benefits the revisers' work. The approach presented here can also be used for developing revision practices and providing training for revisers, bringing the reviser's work and his role closer to a text editor's role. Writing and revising are closely related processes. Consequently, both writers and revisers of scientific papers would benefit from having basically the same tools for their work. This would enable successful cooperation.

References Benson, James D.—William S. Greaves (eds.) 1985 Systemic perspectives on discourse I. Norwood, New Jersey: Ablex. Christie, Frances (ed.) 1984 Language studies: Children writing. Reader. Geelong, Victoria: Deakin University Press. Clyne, Michael 1987 "Cultural differences in the organization of academic texts", Journal of Pragmatics 11: 2 1 1 - 2 4 7 . Connor, Ulla —Robert B. Kaplan (eds.) 1987 Writing across languages: Analysis of L2 text. Reading: Addison-Wesley.

Non-native writing and native revising of scientific articles

491

Danes, Frantisek 1974 "Functional sentence perspective and the organization of the text", in: Frantisek Danes (ed.), 106-128. Danes, Frantisek (ed.) 1974 Papers on functional sentence perspective. (Janua Linguarum, Series Minor, 147.) The Hague: Mouton. Dudley-Evans, Tony (ed.) 1987 "Genre analysis and E. S. P.", in: ELR Journal 1. Birmingham: University of Birmingham, English Language Research. Eggington, William G. 1987 "Written academic discourse in Korean: Implications for effective communication", in: Ulla Connor — Robert B. Kaplan (eds.), 153 — 168. Enkvist, Nils E. — Kaj Wikberg—Ann-Charlotte Lindeberg (eds.) forthcoming Nordtext symposium 1990. Äbo: Äbo Akademi Press. Fries, Peter 1981 "On the status of theme in English: Arguments from discourse", Forum Linguisticum 6: 1 —38. Galtung, Johan 1979 "Deductive thinking and political practice. An essay on Teutonic intellectual style", in: Johan Galtung (ed.). Galtung, Johan (ed.) 1979 Papers on methodology II. Copenhagen: Ejlers. Gunnarsson, Britt-Louise (ed.) 1987 Facktext. Malmö: Liber. Hakulinen, Auli —Fred Karlsson 1979 Nykysuomen lauseoppia. [Modern Finnish Syntax.] Helsinki: Suomalaisen kirjallisuuden seura. Halliday, M . A . K . 1985 Introduction to functional grammar. London: Arnold. Halliday, Μ. A. K. — Ruqaiya Hasan 1976 Cohesion in English. London: Longman. Kaplan, Robert B. 1966 "Cultural thought patterns in intercultural education", Language Learning 16: 1 - 2 0 . Källgren, Gunnel 1979 Innehäll i text. [Content in Text.] (Ord och stil, Spraksamfundets skrifter 11). Stockholm: Studentliteratur. Lauren, Christer—Marianne Nordman (eds.) 1987 Special language: From humans thinking to thinking machines. Clevedon, Philadelphia: Multilingual Matters. Lautamatti, Liisa —Sauli Takala (eds.) 1986 AFinLan vuosikirja 1986 (Publications of the Finnish Applied Linguistics Association 43). Jyväskylä: AFinLa. Martin, James R. 1983 "Participant identification in English, Tagalog and Kate", Australian Journal of Linguistics 3, 45 — 74. 1984 "Language, register and genre", in: Frances Christie (ed.), 21—29.

492 1985

Eija Ventola—Anna

Mauranen

"Process and text: Two aspects of human semiosis", in: James D. Benson — William S. Greaves (eds.), 2 4 5 - 2 7 4 . Mauranen, Anna forthcoming "Reference in academic rhetoric. A contrastive study of Finnish and English writing", in: Nils E. Enkvist —Kaj Wikberg—Ann-Charlotte Lindeberg (eds.). Mauranen, Anna —Eija Ventola 1989 Englannin tieteellisen tekstin kirjoittamisen kurssi ja kielentarkastusseminaari. [A Seminar on Writing Scientific English and Language Revision — A Report.] Helsinki: Helsinki University Language Centre. [Unpublished MS.] Neuendorff, Dagmar 1988 Textsorte und übersetzen. [Unpublished MS. A paper presented in a seminar on empirical research in intercultural studies and translation in Savonlinna, Finland, 1988.] Regent, O. 1985 "A comparative approach to the learning of specialized written discourse", in: Philip Riley (ed.), 1 0 5 - 1 2 0 . Riley, Philip (ed.) 1985 Discourse and learning. Papers in applied linguistics and language learning from the Centre de Recherches et d'Applications Pedagogiques en Langues (C.R. A.P.E.L.). London: Longman. Schröder, Hartmut 1986 "Teksti — argumentaatio — kulttuuri. Ajatuksia tieteellisen tekstin kontrastiivisesta diskurssianalyysistä" [Text — argumentation — culture. On contrastive discourse analysis of scientific texts], in: Liisa Lautamatti —Sauli Takala (eds.), 9 1 - 1 0 9 . Swales, John 1981 Aspects of article introductions. (ESP Research Reports 1.) Birmingham: Aston University. 1987 "Utilizing the literatures in teaching the research paper", TESOL Quarterly 21, 1: 4 1 - 6 8 . Ventola, Eija 1987 The structure of social interaction. London: Frances Pinter. forthcoming "Text and reference", in: Nils E. Enkvist —Kaj Wikberg—Ann-Charlotte Lindeberg (eds.). Ventola, Eija —Anna Mauranen 1990 Tutkijat ja englanniksi kirjoittaminen. [Researchers and Writing in English.] Helsinki: Helsinki University Press. Villemoes, Anette 1987 "En tekstlingvistik analyse af en teknisk tekst" [Textlinguistic analysis of a technological text], in: Britt-Louise Gunnarsson (ed.), 16—42. Winter, Eugene 1977 A clause-relational approach to English texts: A study of some predictive lexical items in written discourse. (Special issue of Instructional Science 6,1.) Ylönen, Sabine —Dagmar Neuendorff—Gottfried Effe 1987 "Zur Kontrastiven Analyse von medizinischen Fachtexten. Eine diachrone Studie", in: Christer Lauren —Marianne Nordman (eds.), 203 — 224.

Index

abstraction 309, 3 1 3 - 3 1 5 , 332, 351, 436, 448 acceptance 174, 179 accountability total 121 — 122, 135 acronym 317 activity types 28 adjunct 1 2 5 - 1 2 6 , 128 advcrbials 10,127 agreement 172, 1 8 0 - 1 8 1 , 186, 189, 201 corrective 180, 181 internal 1 8 3 - 1 8 4 modified 180 preferred 201 revised measure of 172, 186 standard measure of 172, 180, 186 total 193 assessments 193. 196, 207 association partial 53 attribute 2 0 9 - 2 1 0 axiological dimension 281, 288, 290, 296

behavior 193, 2 6 0 - 2 6 1 , 266

carrier 2 0 7 - 2 1 0 chain identity 417 interaction 3 8 6 - 3 8 9 , 4 1 7 , 4 2 5 - 4 2 7 significant 4 2 5 - 4 2 6 , 4 2 6 similarity 417 choice 12, 20, 20, 2 3 - 2 5 , 208 (see also·, decision) cladistics 39 clause 1 1 1 - 1 1 2 , 124, 240, 314 classification 316 — 322 code, 46, 63, 193 code variation 47 coding orientation 46 cognition 69 cognitive content 66 — 67, 69 cognitive science 281, 284, 288, 293 coherence 25, 385, 396, 412, 415

coherence processes 396, 412 — 413 lexical 3 9 6 - 3 9 7 syntactic 397 discoursal 397 cohesion 25, 385,412, 415 cohesive chain 3 8 5 - 3 8 9 , 4 1 2 cohesive harmony 385,412,417 — 418 cohesive ties 385, 390, 394, 396 lexical 444 collaborative learning 263 communication 3 — 5 communicative dynamims 374 computational linguistics 81, 83, 85, 117, 219 Explainable Expert System 89 C O M M U N A L 85 N I G E L 85, 92, 94, 101 P E N M A N 85, 9 1 - 9 3 , 100, 117 SLANG 83 condensation degree of 286, 294, 301 conditioning 48 environment 52 conflicts 11 — 16 conjunction 4 4 3 - 4 4 5 , 4 6 3 - 4 6 8 , 489 connectors 4 6 3 - 4 6 8 , 489 conspiracies 11 — 16 constituency 107, 123, 125, 129, 135 content field 247 context 71, 1 4 3 - 1 4 4 , 214, 218, 229 of culture 23, 28, 31, 214, 236, 431 of situation 34, 28, 3 1 - 3 2 . 214, 236, 431 control 107, 110, 112, 1 1 3 - 1 1 4 conversation 213 — 214, 229 analysis 215 structure 241—242 correlation coefficient 421 creativity 34

decision space 11 — 16, 20 definition 3 1 1 - 3 1 2

494

Index

deixis 46 modal 46 temporal 46 delicacy 51 dependency 107, 135 development disorders 213 downgrading 202 dialogue system 194 diathesis 67 disagreement 181, 201 (see also: agreement, non-agreement) discourse academic 436, 437 classroom 273 of history 3 1 3 - 3 1 5 journalistic 370, 373, 379 medical 370, 379 metadiscourse 289 moral-practical 293 pedagogic 239, 254, 292, 308 of political philosophy 391 scientific 307, 3 1 1 - 3 1 3 , 333, 339, 3 7 0 379, 415, 457 significance 171, 173, 179, 186 social 262 stratum 241 technical 282, 286, 289, 292-293, 301, 309, 317, 332 technocratic 281, 292-293, 301 theoretical 293 disjunctions 35 dispreferred 201 distance 146, 148 dynamic (see: perspective)

elaboration 311—312 empathy 67 end weight 11 English as a second language 431, 434, 457 native speakers of 416, 431, 457, 459 — 461, 470 non-native speakers of 416, 431, 451, 457, 470 equivalence lexical 397

syntactic 397 ethnomethodology 214 evaluative judgement 193 — 194 negative 1 9 6 - 1 9 8 positive 196-198 exchange 171-171, 241 - 2 4 2 , 355 aborted 217, 231 minimal 170, 177 ill-formed 170 well-formed 171 expansion 43, 49 — 52, 396 elaborating 51 extending 51 enhancing 51 discourse 397

facet 198, 2 0 6 - 2 0 7 features, situational 97 flow charts 218 formalist theory 260 formations 29 — 31 (see also: text) frequency 41 fronting 9, 11 function markers 126—127 functional linguistic theory 259 — 261, 265, 268, 271, 274 functional sentence perspective 369, 373 function positions 65 functional stages 437 — 442

genre 24, 27, 30, 231, 235, 264, 266, 273, 274-275, 308, 339, 360, 370, 432, 461 463 curriculum 235 explanation 308, 334, 436 exposition 308, 433 narrative 365, 340, 348 morning-news 236 recount 334 report 308, 333-334, 339 scientific reporting 367, 346, 370 — 371, 373-379, 415 written 267, 307, 352, 415, 431 givenness 142—144, 349 (see also: information)

Index Government and Binding 17, 108 grammar 92 intricateness of 347 — 348, 449 lexical functional grammar 108 monosystematic 17 polysystematic 17 probabilistic 274 subgrammar 93, 98 systematic 11, 86, 107, 121, 265, 273 grammatical feature preselection 91 grammatics 42 Gricean principles 16 group 134—135 nominal 283, 449

hesitations 156 heteroglossia 24 hierarchization 67, 69 hybrid formation 295 hybridization 361 hyponymy 9

iconicity 5, 8 — 11, 14 implication sequences 333 implicit correction 181 — 183 impromtu speech 18 inferability link 160 informant test 416, 4 2 0 - 4 2 1 information 3 - 5 , 42, 98, 141, 143, 145, 170 given 141, 145, 349, 373 irrecoverable 145, 151, 153, 156, 161, 349 new 141, 145, 349, 373 recoverable 145, 147-153, 161, 349 instantiation 58 interdependency 48, 69 {see also\ taxis) interference 146, 465 interjections 124-125, 128, 130 interpenetration 193 interpretability 6 — 8 intertextuality 35, 214, 284, 286, 295

jargon 317

495

kind 198 knowledge 307, 333 background shared 141

langue 19, 58 language 63, 69, 235 as behavior 63 as code 63 as system of knowledge 63 language revision 457 — 461, 466—468, 4 7 6 - 4 7 8 , 4 8 1 - 4 8 2 , 485 left-dislocation 146 — 153 lexis attitudinal 28 lexical density 347, 352, 442, 449 lexical rendering 417 lexico-semantic studies 285 linearity 4, 8—11 linguistic sign 129 literacy curriculum 254 myth 262 literary theory 270 — 271

machine translation 107, 116 Eurotra 107 KOMET 107, 116 'makes sense'-principle 71 markedness 231 —231, 354 Marxist theory 260, 263 meaning 6 - 8 , 7 1 - 7 2 , 219, 261 discourse 183-184, 281 emergent 32 — 34 figurative 285 language-specific 66, 69 lexical 72 linguistic 67 literal 71, 285 potential 300-201 metaphor experiental 331 grammatical 269, 273-274, 298, 3 0 0 301, 315, 326-328, 331 - 3 3 2 , 397, 400, 407, 432, 4 4 9 - 4 5 2 ideational 331

496

Index

interpersonal 331 lexical 327 logical 331 metafunctions 24, 40, 117, 121, 135, 241, 262, 331, 341 ideational 24, 169, 241, 341 experiential 91, 117, 121, 3 4 1 - 3 4 7 , 442 logical 43, 48, 91, 128, 1 3 1 - 1 3 2 , 347-349 interpersonal 24, 125, 128, 130, 169, 241, 281, 3 5 5 - 3 5 9 , 442 textual 24, 117, 126, 128, 131, 169, 241, 3 4 9 - 3 5 5 , 442 metaredundancy 30 — 31 method of development 350, 353 models 1 6 - 1 9 , 29, 254 cognitive-instrumental models of rationality 293 dynamic 2 1 7 - 2 2 0 user models 82 mood 28, 3 5 6 - 3 5 7 modality 117, 125, 3 5 7 - 3 5 9 modulation 331 move complexes 242 mutual correspondencies 65

network 12, 44, 48, 96, 196, 215, 2 2 6 - 2 2 7 dynamic 219 register 83, 9 6 - 1 0 0 semantic 47, 196 static 219 paradigmatic 216 — 217 nominalization 9, 282, 294, 298, 301, 307, 315, 326, 331 - 3 3 2 , 345, 351 - 3 5 2 , 451 non-agreement 181 (see also: agreement)

object delition 11 ontogenesis 41 oscillation 58

paradigmatic (see: relations) paraphrasing 431 parole 19, 58

participants 65, 6 8 - 6 9 , 7 6 - 7 7 , 108, 282, 341-342 -er-role 342 -ed-role 342 passivization 9 perspective active 66 dynamic 2 5 - 2 7 , 33, 35, 4 0 - 4 1 , 214, 2 1 7 - 2 2 0 , 229, 273 passive 66 synoptic 25 — 27, 33, 40 phrase 1 3 4 - 1 3 5 plagiarism 431, 452 poetic function 262 politeness constraints 383 polysemy 66 postposition 126 — 127 pragmatics 27, 82, 169 predicates, classes of 74—76 static (states) 74 dynamic (ongoings) 74 mutations (events) 74 predictability 141, 144, 174 preposition 127 principle components analysis 47 probability 42, 48, 215, 219, 2 2 0 - 2 2 1 , 2 2 9 - 2 3 2 , 2 7 3 - 2 7 5 , 357 conditioned probability 45 equiprobability 42, 44—45, 231 (see also: system) lexical 57 process 74, 282 types 1 0 8 - 1 1 0 , 3 1 9 - 3 2 0 , 3 4 1 - 3 4 7 , 445 - 448 cognitive 216 social 236 projection 43, 4 9 - 5 3 , 119, 240, 3 4 8 - 3 4 9 , 445 direct speech 53, 119 indirect thought 53, 119 propositional argument 110 prosodic field-like prosodies 290 overlays 135 psychiatric patients 214, 221 purposive-rational action 228

question-test 373

Index raising 107, 110, 112 rank 1 2 1 - 1 2 2 , 265 hypothesis 121 scale 1 2 1 - 1 2 2 shift 122, 135 re-accentuation 293 readability 269 realization 58, 92, 144, 195, 2 0 7 - 2 1 1 , 221, 248, 328, 379, 451 congruent 328, 330, 345, 444, 448 incongruent 328, 330, 345, 351, 444, 448 realizational operators 98 statements 97 reasoning 323 recursion 217, 2 1 9 - 2 2 0 , 231 reduction 396 lexical 397 redundancy 42, 44 metaredundancy 59 reference 350, 4 7 8 - 4 8 2 , 489 chain 478 referent 147, 351 text participant reference 478 register 24, 47, 57, 81, 8 3 - 8 5 , 92, 214, 235, 301, 381 field 57, 238, 333, 434 mode 57, 238, 333, 434 tenor 57, 238, 434 pedagogical 252 technical 282, 284 relations linking 135 also: conjunction, connectors) logical-semantic 48 (see also: expansion, projection) paradigmatic 40, 42, 85, 96, 214, 216, 229 scopal 135 syntagmatic 40, 214 repetition 394, 465 replacement 394, 465 response 183, 200, 2 2 3 - 2 2 4 , 2 2 6 - 2 2 7 , 229 congruent 200 — 202 indirect 2 2 7 - 2 2 8 rheme 67, 141, 349, 369, 3 7 1 - 3 7 2 , 448 (see also: theme) rhetorics 20, 270, 461

497

saliency 141, 143 scenario-building 3 — 5, 160 semantic choices 208 features 47 roles 108, 1 1 3 - 1 1 4 semantics 24, 2 7 - 2 9 , 169 text semantics 24, 31 semiosis 28 semiotics 27, 2 9 - 3 0 , 40, 462 connotative 59 social 27, 2 9 - 3 0 semogenesis 6 — 7, 12, 53 sentence combining 264 order 389 sentence patterns grammatical 64, 68, 73 semantic 64, 73 complex 64 system of 65 sequence objective 374 subjective 373 skewness 42, 45 social constructionism 263 — 264 social heteroglossia 293 sociolinguistics 272 speech functions 199, 208, 2 2 1 - 2 2 3 (see also: acceptance, agreement, assesment) stratification 59 structure 42, 108 conjunctive 323 — 325 generic 266, 3 5 4 - 3 5 5 , 357, 360, 461 463 information structure 115 multivariate 133 schematic 243, 2 4 8 - 2 5 1 , 375, 432, 4 3 7 442 thematic 354, 370 univariate 133 stylistics 35 substitution, pro-form 397 summarizing 431, 436 synonymy 66 synoptic (see: perspective)

498

Index

syntagmatic {see: relations) system 30, 39, 42, 63, 72 dynamic 26, 41 grammatical 42, 48 non-recursive 43—44 recursive 43—44 selection probabilities 24, 26 semiotic 58, 235 simultaneous 54 static 217

tagmemic linguistics 264 tailoring 8 1 - 8 2 take-up assent 172, 186, 189 corrective 1 8 1 - 1 8 2 failure 172 modified 172 take-up consent 171, 186, 189 corrective 1 8 2 - 1 8 3 failure 171 modified 171 taxis 4 8 - 5 3 , 274, 348 hypotaxis 4 9 - 5 3 , 3 4 8 - 3 4 9 parataxis 49 — 53, 348 taxonomy 308, 3 1 6 - 3 1 7 , 333, 435 teacher monologue 253 technicality 309, 311 - 3 1 3 , 315, 317 tendencies 65 restrictional 65, 70 selectional semantic 65, 70 — 72 tense 43 text 40, 63, 2 6 1 - 2 6 2 , 283 allotext 14 cotext 71 formations 30, 281, 284, 286 generation 25, 29, 3 4 - 3 5 , 81, 93, 107 metatext 15 processing 63 production 23, 25 — 27, 32 — 34 strategy 9, 11 - 1 6 typology 14, 27, 308 unit 9 texture 23, 26, 332 thematic progression 369, 371 — 373, 379 — 380, 468, 470

theme 67, 115, 141, 170, 241, 349, 3 5 0 355, 369, 3 7 1 - 3 7 2 , 3 7 9 - 3 8 1 , 432, 4 4 2 - 4 4 5 , 4 6 8 - 4 7 8 , 489 thematization 67, 376 thematic formations 281, 284, 286 (see also: method of development; thematic progression; structure) thesis statement 438, 441 tokens central 417, 4 2 1 - 4 2 2 peripheral 417 relevant 417 significant 426 significant central 426 traversing 100 transference 396 lexical 397 transformation 397 transitivity 1 0 8 - 1 1 0 , 241, 432, 4 4 5 - 4 4 8 network 110, 116 turn-taking 149, 2 2 1 - 2 2 3

uncommon sense 311 unit complexes 133 — 134 universal 67, 136 upgrading 201

valency 65, 72 — 73 grammatical 65, 72 potential 65 semantic 65, 72 zero 65 value-chain 293, 295 variation experiental 91, 95 diacronic 58 diatype 57 free 18 logical 91, 94 semantic 47 theory 47 (see also: register) verb finite 65 semi-auxiliaries 112 — 113 stage-of-process 113

Index word-order, canonical 11 writing academic 415, 431, 457 assignments 431 instruction 243, 247, 253, 264, 2 6 8 269, 271, 2 8 7 - 2 9 0 , 307, 415, 431, 4 8 1 - 4 8 2 , 490

popular scientific 339, 466 textual rewriting 4 8 2 - 4 8 8 (see also: discourse, genre)

x-bar model 10, 108

499

m m

m m m m m m m m

m m m m m

Vladimir Ivir Damir Kalogjera (Editors]

Languages in Contact and Contrast

Essays in Contact Linguistics 1991. XII, 502 pages. Cloth. ISBN 311 012574 9

(Trends in Linguistics. Studies and Monographs 54]

The invited papers in this collection on contact linguistics deal not only with the effects of linguistic borrowing and mutual influence of linguistic systems, but also, in the broader sense of the term, with the bilingual speaker who is the locus of language contact. Both approaches to contact linguistics are represented in this volume, the first exemplified by analytical papers presenting material from several Indo-European languages that cover different aspects of linguistic description, and the latter by examinations of bilingualism and foreign-language acquisition. The group of theoretically oriented papers examine the rationale of contact and contrastive linguistics, their relationship, their place among the linguistic disciplines, possible models of description, and methodologies. Also included are papers which take a broader perspective and consider the social and cultural context in which language contact occurs.

m m m

mouton de gruyter Berlin · New York

m Mary E. McGroarty Christian J. Faltis (Editors) m Languages in School and m Society m Policy and Pedagogy m m m m m

m m m m m

1991. X, 570 pages. Cloth. ISBN 311 012576 5 (Contributions to the Sociology of Language 58)

The six sections in this collection of 30 articles demonstrate the scope of current work in applied linguistics and suggest some of its applications to broader social and educational issues. Part I deals with theoretical and methodological advances in sociolinguistics, while Part II focusses on second language pedagogy. Recent trends in classroom language research are discussed in Part III, with Part IV concentrating on bilingual education for minority students. Part V addresses the assessment of second language proficiency. In Part VI, the connections between home and school contexts for language learning are examined. The volume illustrates the comprehensive nature of applied linguistics in its approach to understanding basic questions about how and why social context affects language learning and use.

m m m

mouton de gruyter

m

Berlin · New York