(Amsterdam Studies in the Theory and History of Linguistic Science, Series IV Current Issues in Linguistic Theory)Christian M. I. M. Language Typology a Fun

LANGUAGE TYPOLOGY

AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE

General Editor

E.F. KONRAD KOERNER(Zentrum fr Allgemeine Sprachwissenschaft, Typologie

und Universalienforschung, Berlin)

Series IV CURRENT ISSUES IN LINGUISTIC THEORY

Advisory Editorial Board

Lyle Campbell (Christchurch, N.Z.); Sheila Embleton (Toronto) Brian D. Joseph (Columbus, Ohio); John E. Joseph (Edinburgh)

Manfred Krifka (Berlin); E. Wyn Roberts (Vancouver, B.C.)Joseph C. Salmons (Madison, Wis.); Hans-Jrgen Sasse (Kln)

Volume 253

Alice Caarel, J.R. Martin and Christian M.I.M. Matthiessen (eds)

Language Typology.

A functional perspective.

JOHN BENJAMINS PUBLISHING COMPANYAMSTERDAM/PHILADELPHIA

LANGUAGE TYPOLOGYA FUNCTIONAL PERSPECTIVE

Edited by

ALICE CAFFAREL

J. R. MARTIN

The University of Sydney

CHRISTIAN M. I. M. MATTHIESSEN

Macquarie University

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences Permanence of Paper for Printed Library Materials, ANSI Z39.48-984.

8 TM

Library of Congress Cataloging-in-Publication Data

Language typology : a functional perspective / edited by Alice Caarel, J.R. Martin, Christian M.I.M. Matthiessen. p. cm. -- (Amsterdam studies in the theory and history of linguistic science. Series IV, Current issues in linguistic theory, ISSN 0304-0763 ; v. 253) Includes bibliographical references and index. . Typology (Linguistics) I. Caarel, Alice. II. Martin, J.R. III. Matthiessen, Christian M.I.M. IV. Series.P204.L297 200445'.0--dc22 2004055957ISBN 90 272 4766 8 (Eur.) / 588 559 3 (US) (Hb; Volume : alk. paper)

2004 John Benjamins B.V.No part of this book may be reproduced in any form, by print, photoprint, microlm, or any other means, without written permission from the publisher.

John Benjamins Publishing Co. P.O.Box 36224 020 ME Amsterdam The NetherlandsJohn Benjamins North America P.O.Box 2759 Philadelphia PA 98-059 USA

JB[v.20020404] Prn:1/12/2004; 10:29 F: CI253CO.tex / p.1 (29-76)

Table of contents

List of contributors ix

Foreword xi

Chapter 1Introduction: Systemic functional typology 1

Alice Caffarel, J. R. Martin and Christian M. I. M. Matthiessen1.1 Aims of the book 11.2 Orientation systemic functional language typology 41.3 Mapping language the dimensions of systemic

functional theory 161.4 A systemic functional map of English lexicogrammar 421.5 The global map and approaches to typology 541.6 Sampling to support typological generalizations

across descriptions 581.7 Organization of the book 61References 66

Chapter 2Metafunctional profile of the grammar of French 77

Alice Caffarel2.1 French through time and space 772.2 A short overview of French grammar 782.3 French clause grammar: A metafunctional profile 862.4 Typological outlook 133References 136

Chapter 3Metafunctional profile of the grammar of German 139

Erich Steiner and Elke Teich


Table of contents

3.1 A brief history of German 1393.2 A description of the paradigmatic and syntagmatic organization

of the German clause 1423.3 The mood system and interpersonal structure 1443.4 The transitivity system and experiential structure 1523.5 The theme system and theme structure 1693.6 Conclusion and comments on methodology 180References 183

Chapter 4Metafunctional profile of the grammar of Japanese 185

Kazuhiro Teruya4.1 Introduction 1854.2 Metafunctional preview of the clause in Japanese 1874.3 The mood system and the modal structure 1934.4 The transitivity system and experiential structure 2074.5 The theme system and theme structure 2284.6 Conclusion: Next step into new description

for typological survey 247References 251

Chapter 5Metafunctional profile of the grammar of Tagalog 255

J. R. Martin5.1 Ecosocial context 2555.2 Metafunctional preview 2565.3 transitivity 2585.4 theme 2805.5 mood 2845.6 A note on logical meaning 2965.7 News from somewhere 298References 302

Chapter 6Metafunctional profile of the grammar of Chinese 305

M. A. K. Halliday and Edward McDonald6.1 Introduction 3056.2 Preliminaries: The clause and its constituents 3116.3 Textual metafunction 320


Table of contents

6.4 Interpersonal metafunction 3296.5 Experiential metafunction 353References 393

Chapter 7Metafunctional profile of the grammar of Vietnamese 397

Minh Duc Thai7.1 Introduction 3977.2 The metafunctional description 3997.3 Conclusion 428References 428

Chapter 8Metafunctional profile of the grammar of Telugu 433

V. Prakasam8.1 Introduction 4338.2 Metafunctional preview 4358.3 mood 4378.4 transitivity 4478.5 Theme and Focus 4718.6 Conclusion: Typological outlook 477References 477

Chapter 9Metafunctional profile of the grammar of Pitjantjatjara 479

David Rose9.1 Cultural contexts 4799.2 Metafunctional preview 4819.3 mood 4949.4 transitivity 5069.5 theme 5259.6 Conclusion 532References 534

Chapter 10Descriptive motifs and generalizations 537

Christian M. I. M. Matthiessen10.1 Introduction 53710.2 Descriptive motifs and generalizations 538


Table of contents

10.3 Ideational construing experience 57410.4 Interpersonal enacting social roles & relations 61010.5 Textual creating information as unfolding text 63510.6 Variation across metafunctions: projection 65210.7 Conclusion 655References 663

Index 675

JB[v.20020404] Prn:26/11/2004; 13:30 F: CI253LI.tex / p.1 (43-127)

List of contributors

Alice Caffarel is Senior Lecturer in French and Linguistics in the Department ofFrench Studies at the University of Sydney. She specializes in French grammar andsemantics and has published articles in edited volumes and International jour-nals. She is interested in Register Variation, Corpus Linguistics, ComputationalLinguistics and Stylistics.

M. A. K. Halliday is Professor Emeritus in Linguistics at the University of Syd-ney, Sydney, Australia, Visiting Professor at Macquarie University and HonoraryProfessor at Peking University.

James R. Martin is Professor in Linguistics at the University of Sydney, Sydney, Aus-tralia. His research areas include Tagalog grammar and discourse, discourse seman-tics and theories of context, educational linguistics and systemic functional theory.

Christian M. I. M. Matthiessen is Professor (Chair) in Linguistics at MacquarieUniversity, Sydney, Australia. His research areas include language typology, natu-ral language processing, the modelling of meaning, lexicogrammar, corpus-basedmethodology, and systemic functional theory.

Edward McDonald holds a MA from Peking University and a PhD from MacquarieUniversity. He has taught linguistics and semiotics at Peking University, the Na-tional University of Singapore, and Tsinghua University. His research areas include:the grammar and discourse patterning of modern Chinese; intellectual currents incontemporary China, particularly with regard to ideas about language, culture andnationalism; history of linguistics; syntactic theories; and the semiotics of voice andmusical performance. He has published in both English and Chinese in LanguageSciences, the Journal of Chinese Linguistics, Northern Forum, Histoire, epistemologie,langage, and many edited collections; and has a book manuscript in preparationwith Shen Xiaolong entitled East-West dialogues on language and culture. He iscurrently working as an editor at Chinese Central Television.

V. Prakasam is Professor at the Central Institute of English and Foreign Lan-guages, Hyderabad, India. His research areas include the description of Telugu,systemic functional phonology, stylistics, language and the law and systemicfunctional theory.

JB[v.20020404] Prn:26/11/2004; 13:30 F: CI253LI.tex / p.2 (127-159)

List of contributors

David Rose has worked with Australian Indigenous communities for the pasttwenty years, as an educator, student and community member. His study of thelanguage and culture of Australias Western Desert peoples (The Western DesertCode: An Australian Cryptogrammar, 2001) grew out of this first hand experienceand a concern to accurately portray ways of meaning in the culture. He is cur-rently a reaserch fellow at University of Sydney, investigating language and literacyin Indigenous communities, and trains teachers of Indigenous children in literacyteaching strategies. His research interests have also included technical and scien-tific literacies (Science, technology and technical literacies, 1997; Science, discourseand industrial hierarchy, 1998) and discourse analysis (Martin & Rose 2003).

Erich Steiner is Professor at the Universitt des Saarlandes, Saarbrcken, Ger-many. His research areas include the description of German, translation studies,the semiotics of music, language and action and systemic functional theory.

Elke Teich, is Professor at the Technical University, Darmstadt, Germany. She spe-cialises in corpus-based and computational linguistics. She is particulary interestedin the comparative analysis of original and translated English and German texts.She recently published the following book in this area: Cross-linguistic variation insystem and text. De Gruyter: Berlin, New York. (2003).

Kazuhiro Teruya is Lecturer in Japanese, at the University of New South Wales,Sydney, Australia. His research interests include Japanese grammar and discourse,linguistics typology and the modelling of meaning. He is also involved as a re-searcher in the Laboratory for Language based Intelligent Systems at BrainScience Institute, Riken, Japan.

Minh Duc Thai is Lecturer in translation studies at the University of Western Syd-ney, Sydney. His research interests include translation, contrastive analysis andTESOL.

JB[v.20020404] Prn:26/11/2004; 10:19 F: CI253FR.tex / p.1 (43-148)

Foreword

This volume explores the clause grammars of eight languages two from Eu-rope (French, German), five from Asia (Tagalog, Vietnamese, Chinese, Japanese,Telugu) and one from Australia (Pitjantjatjara), covering seven language families.Each language is presented in its own terms, with evidence from a range of spokenand written discourse. The final chapter discusses descriptive motifs and general-izations, drawing on the languages explored in the book but also making referenceto many other languages as well. The descriptions of the individual languages andthe typological generalizations are all based on one particular functional theory systemic functional theory, which is introduced in the first chapter.

The book reflects and consolidates the growth in descriptions of a range of lan-guages based on systemic functional theory, going back to Hallidays fieldwork ondialects of Chinese in the Pearl River Delta in the late 1940s and including a num-ber of new accounts being added in the 1990s. In the last few years, two workshopsfocusing on systemic functional typology have been held at the University of Syd-ney the first in December 1996 and the second in September 1999. During theseworkshops, systemic functional researchers presented and contrasted grammaticalprofiles of a range of languages. This book includes languages discussed at the firstworkshop, but it also adds languages that were not represented at that workshop German and Telugu. The second workshop included continued work on the lan-guages from the first workshop, but it also saw the addition of other languages Danish, Cantonese and Korean in particular. These will be included in a sequel tothe present book, together with accounts of other languages as well.

The book is intended to serve as a contribution to a range of multilingualconcerns. One of the central multilingual concerns is, of course, language typol-ogy and the material has been used in this context. But the book is also intendedto support other multilingual concerns such as language comparison, translationstudies and the development of descriptions of languages that have not yet beendescribed in (systemic) functional terms. The descriptions presented in this bookare all based on rich evidence from discourse; and they have been designed to beuseable in discourse analysis something that has become increasingly importantin translation studies.

In the field of typology, the book is offered as a complement to existing works:it combines accounts of individual languages couched in terms that are sensitive


Foreword

to their own unique discourse concerns with an exploration of descriptive motifsand generalizations. Developing an account of a particular language that strikes abalance between viewing it as unique and viewing it as either just another manifes-tation of language as the general human semiotic or a peculiar, scrambled versionof some dominant language like Latin, English or Chinese is a signficant challengethat can only be met by using the resources of a powerful theory of language(s)and a powerful metatheory of our theory of language(s). The contributors havemet this challenge by drawing on systemic functional (meta)theory. This theoryof language is a general theory in the sense that it provides us with the resourcesneeded for construing human language as one kind of semiotic system withoutbuilding in descriptive assumptions based on one or a small number of languages.The descriptions developed within the general theory are descriptions of particularlanguages descriptions that are grounded in natural discourse. The descriptionsare of course guided by the accumulation of descriptive experience. (i) On the onehand there are models emerging from work on other languages; for example, whenhe began developing his descriptions of English, Michael Halliday was able to drawon his experience with describing Chinese and many have been influenced in theirwork on various languages by his account of English. Such accounts serve as modelor protocol descriptions; but there is an inherent danger in using them the dangerof transferring them to other languages. One way of avoiding this danger is to en-sure that the account being developed is based on natural discourse and then to testthe description by using it in discourse analysis. The accounts presented in this vol-ume have all gone through these two processes. Thus if the descriptive category ofTheme or Subject is posited in the account of a given language, this is only be-cause it represents a plausible interpretation of patterns that emerge in discoursesfrom different registers. (ii) On the other hand, comprehensive descriptions of par-ticular languages provide material for descriptive generalizations concerned withsystems such as those of transitivity, theme, mood, tense/aspect and expansion &projection. Such generalizations should be based not on isolated structural frag-ments but on a sense of how whole systems operate in their environments. Forexample, the category of Subject should not be the focus of investigation; rather anygeneralizations about Subject should emerge from more holistic considerations ofhow languages grammaticalize interpersonal patterns of meaning as a resource forcreating dialogue in the form of systems of mood.

Acknowledgements

Warm thanks to people who have assisted in the editing process Kathryn Tuckwelland Kumi Suzuki, and to Kazuhiro Teruya and Wu Canzhong for help with dia-


Foreword

grams and romanization in the Chinese chapter; Kazuhiro and Canzhong have pro-vided extensive help with the preparation of the camera-ready version of the book.We are also grateful to workshop participants who commented on oral presenta-tions, and students who have commented on various versions of the manuscript ofthis book.

We would also like to thank the School of Languages and Cultures and theSchool of Society, Culture and Performance at the University of Sydney and theDivision of Linguistics and Psychology at Macquarie University for their financialassistance.

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.1 (42-154)

Chapter 1

IntroductionSystemic functional typology

Alice Caffarel, J. R. Martin andChristian M. I. M. MatthiessenUniversity of Sydney / Macquarie University

. Aims of the book

This book is intended as a systemic functional contribution to language typologyboth for those who would like to understand and describe particular languagesagainst the background of generalizations about a wide range of languages andalso for those who would like to develop typological accounts that are based onand embody descriptions of the systems of particular languages (rather than iso-lated constructions). Typology is thus to be understood in a broad sense as thegeneral study of similarities and differences across languages covering not onlytypology in a strict sense as elaborated by, for example, Greenberg (1966, 1978)but also descriptive frameworks embodying generalizations developed to supportthe descriptions of a range of different languages (e.g. Comrie 1981; Shopen 1985;Payne 1997; Whaley 1997). This book is a contribution to this second aspect oftypology, but it differs from previous contributions in important respects.

The book is a unique contribution in at least two respects. On the one hand, itis the first book based on systemic functional theory that is specifically concernedwith language typology. Here it complements previous contributions guided byother theories or approaches such as West-Coast functionalism (as in the workby Sandra Thompson, Paul Hopper and Talmy Givn), Simon Diks functionalism,or any of the approaches presented in Shibatani & Bynon (1995). On the otherhand, the book combines the particular with the general in the description of lan-guages: it presents comparable sketches of particular languages while at the sametime identifying generalizations based on the languages described here as well as onother languages. Here it complements previous contributions that focus on typo-logical generalizations but which do not present accounts of particular languages

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.2 (154-228)

Alice Caffarel, J. R. Martin and Christian M. I. M. Matthiessen

contributions dealing with a range of linguistic domains such as Greenberg (1978),Comrie (1981), Shopen (1985), Payne (1997) & Whaley (1997) as well as contribu-tions that focus on some particular domain such as Comrie (1976, 1985), Hopper(1982), Dahl (1985), Bybee, Perkins & Pagliuca (1994) on tense and aspect, Hopper& Thompson (1982) on transitivity, Blake (1994) on case, Palmer (1986) on moodand modality, Chafe & Nichols (1986) on evidentiality, Givn (1983) on conti-nuity, Lyons (1999) on definiteness, and many others. The focus of the bookon both particular linguistic systems and typological generalizations across suchsystems derives from the systemic functional approach to typology.

Systemic functional linguistics is a tradition within functional approaches tolanguage that was developed by M.A.K. Halliday beginning with his work on Chi-nese in the 1950s. He drew originally on the contextualism, prosodic analysis andsystem-structure theory that had characterized work led by J.R. Firth since the1930s in Britain, but later he also incorporated other strands, including ideas fromPrague School functionalism and American anthropological linguistics. (Firthswork already incorporated the ethnographic experience: he had been inspired byBronislaw Malinowskis pioneering work in anthropology from the mid 1910s on-wards.) The systemic functional base means that descriptions of languages are ori-ented towards context, grounded in discourse and focused on meaning: languageitself is interpreted as a meaning potential a meaning potential that embodiesthree different kinds of meaning (ideational meaning, interpersonal meaning andtextual meaning).1

The systemic functional contribution to language typology presented in thisbook must be viewed in the context of the rich and expanding field of typologi-cal linguistics. The interest in similarities and differences across languages goes farback in time, of course; but it was not until the 19th century that systematic workwas undertaken, first by Schlegel and Humboldt inspired both by the intellec-tual theme of evolution and the accumulation of experience from an increasingrange of languages around the world. In the 19th century, language typology wasapproached from below: the focus was on the grammar of words (morphology)and typology was based on word structure the familiar types of isolating, aggluti-native, fusional and polysynthetic languages, later reinterpreted by Sapir (1921) asinvolving two distinct parameters degree of synthesis (from isolating to synthetic)and degree of fusion (from agglutinative to fusional). In this century, the empir-ical approach to typology goes back to the Prague school in the pre-war work byMathesius, Skalicka and Trubetzkoy (see Sgall 1995), with Jakobson (e.g. 1941) de-veloping a kind of universalism that was later taken up in the American context,within generative linguistics in particular.

These early developments are very significant. However, the expansion of thefield in the last 20 years has been particularly impressive. Around 1980, the four-volume publication edited by Greenberg (1978) had just appeared. It was largely

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.3 (228-279)

Introduction

based on Greenbergs (1966) seminal paper presented at the Dobbs Ferry confer-ence on language universals in 1961, and it represented the culmination of the uni-versals and typology project at Stanford University. Greenbergs work introducedthe notion of implicational universals Croft (1995:86) speaks of Greenbergsdiscovery of implicational universals of morphology and word order; but im-plicational universals are part of the descriptive framework designed to interpretlinguistic phenomena and it is these phenomena that are discovered, not the frame-work used to describe them. Other early typological work was available in bookssuch as Li (1976) and Comrie (1976); but one of the standard textbooks in lin-guistic typology Comrie (1981) had not yet appeared and certain breakthroughcontributions such as Hopper & Thompson (1980) were also still unpublished. Bythe late 1970s the early Chomskyan approach to universals and universal gram-mar had of course also been established in the form of conditions and constraintson rules; but it did not have the kind of cross-linguistic empirical foundationprovided by Greenbergs approach.

In the last 20 years in particular, Greenbergs empirical approach has beentaken further within various frameworks, including a number of those surveyedin Shibatani & Bynon (1995). Typology has been linked in a very fruitful way tophylogenetic concerns by means of the concept of grammaticalization, marking areturn at a higher level of understanding to the evolutionary theme of the 19thcentury (e.g. Heine, Claudi & Hnnemeyer 1991; Hopper & Traugott 1993; By-bee, Perkins & Pagliuca 1994; Harris & Campbell 1995). Typology has been linkedto investigations of discourse in different languages, thus expanding the domain oftypology from the system pole of the cline of instantiation (see Section 1.3.2 below)to include the instance pole as well (as in early work on tense and aspect in Hopper1982, on continuity in Givn 1983; and as in more recent work focusing on in-teraction and grammar in Ochs, Schegloff & Thompson 1996 see Cumming &Ono 1997, for a general discussion of the approach to grammar and discourse inwhat they call the discourse-functional approach). The link to discourse has alsomade it possible to advance the interpretation of typological variation in proba-bilistic terms so that we can now see that variation across languages is like variationwithin a language in that it is variation in systemic probabilities. In addition, Hop-per & Thompsons (1980) study of transitivity parameters showed the potentialof system-based typology, supplementing structure-based typology; and it alsoshowed very clearly the importance of focusing on both semantics and lexicogram-mar (i.e. grammar (morpho-syntax) + lexis) in typological statements since allthe transitivity parameters are semantically significant and may be reflected overtlyin the grammar of a language.

In this context of the developing field of typological research, how then is sys-temic functional typology distinctive? It will be easier to answer this question oncewe have sketched an outline of the general systemic functional theory of language

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.4 (279-344)


(Section 1.3) and given a brief example of a systemic functional description of aparticular language (Section 1.4): at that point we will be able to suggest how theresources of systemic functional linguistics can contribute to typological research.Here we will just give a few indications of salient features of a systemic functionalapproach to typology.

i. The sample of descriptions of languages that would form the basis would haveto include rich, comprehensive ones oriented towards meaning and based oninvestigations of discourse, thus ensuring that the features being typologizedcan be motivated independently for each particular language (by reference tonaturally occurring discourse) and that they can be located within the over-all system of each language (cf. Section 1.6 below). The contributions to thepresent volume are brief outlines of such descriptions.

ii. The typological generalizations would be based on a conception of languageas a resource a meaning potential (see Section 1.3.2 below), organized in amultidimensional semiotic space (see Section 1.3.1 below). This would meanoperating with a more highly differentiated conception of language than hasoften been used in typological work (cf. Bateman, Matthiessen & Zeng 1999,for this point in relation to multilingual specifications in general).

iii. One consequence of the multidimensional theory of language is the realizationthat languages are far too complex to be typologized as unified phenomena:typology has to be typology of particular systems (such as tense/ aspectsystems), not typology of whole languages as was done traditionally when lan-guages were typologized as analytic versus agglutinative versus fusional versuspolysynthetic (see Halliday 1966:166168). This is not to say that there cannotbe syndromes of such systemic types either fairly limited or more pervasive,of the kind proposed by G.A. Klimov (for example, his active type is basedon 30 lexicogrammatical features; see Nichols 1992:712); but such typesmust, we believe, be treated as syndromes of individually motivated typologi-cal features rather than as unified types of language, and it must be recognizedthat they do not exhaust the dimensions of typological likeness and differenceacross languages (see for example Martin 1988, on a tenor-oriented syndromein Tagalog).

. Orientation systemic functional language typology

.. Typology as one research application within systemicfunctional linguistics

Many linguistic frameworks see language typology and issues relating to languageuniversals as the central concern or perhaps even the only concern of theo-

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.5 (344-404)

Introduction

retical linguistics. In these approaches their theories are custom-made to addressresearch questions concerning typology and universals. For example, in the Chom-skyan tradition, the theory is typically realized in a formal representation that isdefined in such a way that the representation embodies hypotheses about universalsconstraining what constitutes a possible language (this corresponds to the rep-resentational rather than the theoretical level of the metalanguage of systemicfunctional linguistics; see Section 1.3.7 below).

In contrast, in systemic functional linguistics, language typology (includingissues relating to universals) is only one amongst many research applications (cf.Halliday 1964; 1980; 1994:xxixxxx). Consequently, systemic functional theory hasnot been purpose-built to deal with language typology; but instead it has been de-veloped to be a general linguistic resource that is powerful and flexible enoughto address a wide range of questions about language. Thus the same theory thathas been deployed throughout this book has been applied to areas as differentas: language development how young children learn how to mean; educationallinguistics how people learn language, learn through language and learn aboutlanguage; computational linguistics modelling the generation of written and spo-ken text and modelling translation (machine translation); clinical linguistics working with speakers suffering from aphasia; modelling of language as an inher-ently probabilistic and variable system including not only dialect variation butalso register variation and codal variation; description of particular languages mapping them out metafunctionally and stratally; stylistics and the study of verbalart; translation studies kinds of translation and translation of different metafunc-tional modes of meaning; multimodality the interpretation of semiotic systemsother than language.

This does not mean, however, that typology is a late addition to the systemicresearch agenda. On the contrary, it has been on the agenda right from the start,flowing very naturally out of Firthian linguistics and relating to the Prague Schoolinterest in the characterology of languages (cf. Mathesius 1975, e.g. p. 11). Indeed,in the 1950s M.A.K. Halliday co-authored a paper with Jeffrey Ellis on tense/ as-pect systems in a range of languages, including Chinese, Russian and English. Itwould have been Hallidays first published paper, except the editor of the journal itwas sent to rejected it. And one of Hallidays research investigations at the time wasto survey 37 grammatical features in linguistic area of East and South-East Asia; butthis early work on areal typology was never published (but cf. Halliday 1957:66).There were, however, papers on (machine) translation Halliday (1956, 1962)and on language typology (Halliday 1957, 1966). These early contributions estab-lished a number of important principles of systemic typology (see Matthiessen &Halliday, in prep.: Chapter 5) such as the principle that typology is typologyof particular systems of languages rather than of whole languages, the principlethat typology is generalized comparison, the principle that languages tend to dif-

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.6 (404-485)


fer more at lower ranks (that is, at word rank rather than at clause rank, withgroup/phrase rank as quite variable across languages) and tend to be more con-gruent with one another at higher ones, and the principle that linguistic theory isa general theory of language whereas linguistic descriptions are accounts of partic-ular languages. The rank principle was extended to the dimension of stratificationin a seminal paper by Martin (1983). The early work also included fundamentalconceptions of language critical to modern typology language as an inherentlyvariable system (e.g. Halliday, Macintosh & Strevens 1964; Halliday 1978) and lan-guage as a probabilistic system (e.g. Halliday 1959, in his text-based descriptionof Chinese).

In this context, it is important to dispel the myth that systemic functionaltheory is, or has been, anglo-centric: this view is simply wrong. If one fails to distin-guish between theory and description (see further Section 1.2.2 below), then onemight be mislead into thinking that the fact there has been a substantial body ofsystemic functional descriptions of English means that the theory is anglo-centric.However, the general theory of language has always been distinct from descriptionsof particular languages. This was established in Firthian linguistics; in his Synopsisof Linguistic Theory, 19301955, Firth (1957:21) writes

Linguists are only just beginning to realize the dangers and pitfalls of person-ification of categories as universal entities. [fn omitted] There is a constantneed to beware of such bogus philosophizing in linguistics.

There is always the danger that the use of traditional grammatical termswith reference to a wide variety of languages may be taken to imply a secretbelief in universal grammar. Every analysis of a particular language must ofnecessity determine the values of the ad hoc categories to which traditionalnames are given. [fn omitted] What is here being sketched is a general linguistictheory applicable to particular linguistic descriptions, not a theory of universalsfor general linguistic description.

In his introduction to the volume in which this essay appeared, Firth characterizedHallidays contribution to the same volume as follows:

Dr. Halliday, whose work is primarily in Chinese linguistics, points out that inBritain we have a long tradition of description of the languages of Asia, Africa,and the Pacific, and gives it as his experience that the general linguistic prin-ciples and theories expounded and put into practice in London and elsewhereby my colleagues and myself are of particular value because most of the workhas been done on non-Indo-European languages.

He goes on to comment:

Such principles and methods must serve all people and all languages, and thegreat majority are non-Indo-European.

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.7 (485-519)

Introduction

The distinction between general theory and particular descriptions has thus beenpart of systemic functional linguistics from the start and has been stated in manyplaces (e.g. Halliday 1961). The general theory was never based on English and wasthus never anglo-centric: most of the work by Firth and his colleagues and studentsin the 1930s1950s was on languages other than English and this was in fact an im-portant force behind the development of prosodic theory as well as other aspectsof Firths system-structure theory. And as previously mentioned Hallidays ownwork started with Chinese rather than with English in the late 1940s. (One of hisfirst research assignments was on a project directed by Professor Wang Li, study-ing dialects of Chinese in the Pearl River delta.) Matthiessen has been amused tonote that while some linguists in the West have characterized systemic functionaltheory as anglo-centric because of their apparent lack of awareness of the linguis-tic range within the Firthian and systemic traditions, scholars in China have toldhim on various occasions that they can see how Hallidays experience with describ-ing Chinese has influenced his description of English! (One very specific examplemight be the recognition in the description of English of the three-term systemof relational clauses intensive, possessive, circumstantial. This contrast is asreal in English, as it is in Chinese; but while it is somewhat cryptic in English,it is more overt in Chinese because of the three general verbs shi be, you haveand zai be at/on/etc.. But there are many more general aspects of his descrip-tion of English that Halliday first worked out in his investigation of Chinese in the1950s, such as the ergative model in the account of transitivity and informationflow in the clause.)

Systemic functional linguistics is also thought to be anglo-centric because Hal-lidays (1985/1994) introduction to functional grammar develops an interpretationof English; but this is again due to lack of familiarity with the central goals of sys-temic work. It has always been held to be very important to develop comprehensiveaccounts that bring out the particular character of a given language (whether itis English, French, German, Dutch, Danish, Chinese, Vietnamese, Akan, Arabic,Telugu, Korean, Japanese, Tagalog, Pitjantjatjara, Gooniyandi, Weri, Zapotec, Aus-lan or any other language), treating it in its own right rather than as a variantof a universal motif that is actually based on Latin, English or some other domi-nant language. Even the early systemic functional framework of the 1960s and thebeginning of the 1970s was applied to various languages, including for exampleBarnwell (1969) on Mbembe, Hasan (1972) on Urdu, Huddleston & Uren (1969)on French, Mock (1969) on Nzema, Prakasam (1972) on Telugu. It is also impor-tant to emphasize that the description of English should be anglo-centric, just asthe description of Chinese should be sino-centric! The point is that the descrip-tions of particular languages should be designed to bring out the special featuresof these languages; they should not make them look like variants of some universalcode derived from English, Latin or some other language with which linguists have

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.8 (519-581)


had abundant experience. For example, the description of the interpersonal clausegrammar of English (see Section 1.4.1) makes it look interestingly different fromthat of French (Section 2.3) and that of Vietnamese (Section 7.2.2); but this is apositive feature, not a negative bug.

The fact that typology has always been only one among many research taskson the agenda for systemic functional theory has important implications for thesystemic functional approach to typology. In purely practical terms, it has meantthat there have been fewer people available to pursue typological questions: theresearch agenda within systemic functional linguistics has never been set only, oreven primarily, by questions internal to linguistics as has happened in theoret-ical linguistics in the US in particular but rather by questions about languages,questions that have often come up in contexts of research application such as ed-ucational, stylistic, clinical and computational contexts. So those of us who haveworked on systemic functional typology have always been engaged with manyother tasks as well.

Typology in systemic functional linguistics has thus always interacted withother research concerns, including other multilingual research concerns such ascomparative studies and translation studies in linguistics and multilinguality incomputational linguistics (cf. Halliday 1957, 1966; Catford 1965; Ellis 1966; Steiner1992; Steiner & Yallop 2001; Matthiessen 2001; Teich 2001), multimodal researchconcerns (Steiner 1988; Kress & van Leeuwen 1996; OToole 1994), and researchconcerned with variation within a given language (cf. Matthiessen 1993); and sys-temic functional work on typology has often been carried out in the context ofsome particular research application such as multilingual text generation (cf. Bate-man, Matthiessen & Zeng 1999; Bateman et al. 1991; Teich 1999).

The special contribution that systemic functional theory can make to typo-logical studies derives in large part from the fact that it is a very general, rich andflexible theory that has been applied in a wide range of research contexts a theorythat has been designed to have more power than is needed for any single researcharea (such as typology) precisely because it has been applied to a range of areas.

.. Theory and description; the boundary between the two

Let us now return to the distinction between the general theory of language andthe descriptions of particular languages. Theory and description are ontologicallyquite distinct in systemic functional linguistics: theory is the theory of humanlanguage (or indeed, by extension, of semiotic systems in general); descriptionsare descriptions of particular languages (or, by extension, of particular semioticsystems). Both theory and description are resources resources for construing lan-guage (theory) and languages (descriptions). The emphasis in the development of

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.9 (581-647)

Introduction

systemic functional theory and of systemic functional descriptions has been on en-riching our potential for interpreting linguistic phenomena, not on constraining it(cf. Halliday 1981).

The general theory is thus a theory for construing language more specifically,for construing language as one kind of semiotic system. In systemic theory, semi-otic systems are interpreted as fourth-order systems in an ordered, linear typologyof four orders of systems (Halliday & Matthiessen 1999:Chapter 13). This typologyis ordered both in terms of complexity and in terms of cosmogenesis (cf. Layzer1990; Delsemme 1998): first-order systems emerged before second-order systems;second-order systems emerged from first-order systems, adding a new level of or-ganizational complexity; and so on. First-order systems are physical, second-ordersystems are biological, third-order systems are social, and fourth-order systemsare semiotic. Each higher order of system represents a new level of complexity:the systems of that order are manifested as systems of a lower order, but in addi-tion they embody a new (emergent) kind of property. Thus biological systems arephysical systems plus life (ability to self-replicate with evolution as the mode ofdevelopment), social systems are biological systems plus value (organization intosocial groups with division of labour), and semiotic systems are social systems plusmeaning. Semiotic systems are simultaneously also social, biological and physicalsystems. At their own systemic level, they are meaning-making systems; their keyprinciple of organization is that of stratification stratification into content andexpression. This principle is, as far as we know, unique to semiotic systems2 incontrast to constituency, which seems to be a feature of systems of all orders.

The linear typology of systems is important in the context of typology. It pro-vides us with a framework for interpreting questions about language and the brain,and about language and culture questions that are specifically concerned withinter-systemic correlations. But it also allows us to focus on individuals and thecollective at different systemic orders of organization (cf. Halliday & Matthiessen1999:610611; Firth 1950; Butt 1991; Lemke, 1995). This relates among otherthings to population thinking in the context of language typology and explo-rations of the evolution of language (cf. Nichols 1992).

It might seem odd that the typology of systems does not include cognitivesystems, but this is no oversight. Phenomena that are interpreted as cognitive inmainstream cognitive science are taken to be biological and/or semiotic in sys-temic functional linguistics (but see Fawcett, for example 1980, for a cognitiveorientation within systemic work). For example, higher-order consciousness isseen as an emergent property in the evolution of the ever more complex brain evolving together with language as a higher-order semiotic (see immediately be-low): Deacon (1992, 1997), Edelman (1992) and Halliday (1995). In contrast tocognitive linguists, we would thus seek to explain cognition by reference to lan-guage and other semiotic systems rather than the other way around; see Halliday

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.10 (647-714)


& Matthiessen (1999) for an attempt to lay part of the foundation. This obviouslyincludes a much more intersubjective orientation (cf. Trevarthen 1987), where theconstruction of knowledge is seen as part of the process of learning how to meanin interaction with others (Painter 1993, 1999).

Language is one distinctive and unique kind of semiotic system what is re-ferred to as a higher-order semiotic (Halliday 1995). It is differentiated from otherkinds of semiotic system by systemic theory in terms of stratification (Section1.3.6) and metafunction (Section 1.3.4): language is interpreted as a tristratal semi-otic (rather than a bistratal one) and as a metafunctional semiotic (rather thanmicro- or macro-functional one). The present kind of language can be assumedto be the third phase in a long evolutionary history of language, very likely start-ing before the last common ancestor we humans share with our closest primatecousins (for the evolutionary perspective, see Matthiessen 1999, forthc.). In termsof stratification, language has evolved beyond the bistratal organization (content/expression) of protolanguage into a tristratal system with a distinct, stratum oflexicogrammar:3 semantics and lexicogrammar are content strata and phonology(sign or graphology) is the expression stratum. In terms of function, language hasevolved beyond the microfunctional organization of protolanguage, where func-tions are complementary but mutually exclusive (making it possible to mean onlyone thing at a time) into a metafunctional system where functions are comple-mentary and simultaneous (making it possible to mean more than one thing at thesame time). Stratification and metafunction are two central semiotic dimensionsin systemic theory and will be discussed in Section 1.3 below.

In systemic functional linguistics, systemic theory thus differentiates languagefrom other kinds of semiotic system, interpreting it as a tristratal and metafunc-tional semiotic a higher-order semiotic. However, systemic theory does not dif-ferentiate among different variants of language such as English and Chinese; thatis the task of systemic descriptions of different particular languages such as theones presented in this book. Thus the fact that both English and Chinese con-strue human experience of time grammatically is a general feature of languageas a higher-order semiotic system: time and other phenomena of experience areconstrued lexicogrammatically within the ideational metafunction. However, thedifferent grammatical models for construing time grammatically that English andChinese have evolved fall within the domain of description: the English temporalmodel is described as a tense system, construed logically within the logical modeof the ideational metafunction (Halliday 1994; Matthiessen 1995a, 1996), whereasthe Chinese one is described as an aspect system, construed experientially withinthe experiential mode of the ideational metafunction (see Halliday & McDonald,this volume). The observation that the tense system of English and the aspect sys-tem of Chinese represent poles on a tense/aspect continuum from the western tothe eastern zones of the Eurasian continent with Russian and other Slavic lan-

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.11 (714-764)

Introduction

guages construing time in mixed tense/aspect systems (cf. Gotteri 1996) constitutesa descriptive generalization, not a theoretical one. The theory is of course used ininterpreting temporal systems of tense and aspect. In particular, the theory spec-ifies two modes of construal the logical mode and the experiential mode (seefurther below), so in describing temporal systems, we can explore them in logi-cal terms or in experiential terms. Thus it appears that the English tense systemis construed on a logical model as temporal serialization, whereas the tense sys-tems of certain languages spoken in Africa are construed on an experiential modelas temporal taxonomy (cf. Section 10.3.1 on the division of labour between thelogical and experiential modes of construing experience in the final chapter). Butthe particular temporal categories that are postulated past, present, future;primary tense, secondary tense; past, non-past; recent past, distant past;and so on are descriptive, not theoretical.

Systemic linguistics thus draws the line between theory and description in sucha way that theoretical assumptions are very general and all the categories of partic-ular languages belong to the domain of description. This is of course only oneway of drawing the line between theory and description. Where the line is drawnvaries considerably across different linguistic schools. For example, outside sys-temic functional linguistics, categories such as NP + VP, Subject, Object and Topichave at one time or another been taken as theoretical primitives and thus been as-sumed to be universal. In systemic functional linguistics these categories wouldall be taken as descriptive rather than theoretical, that is, they would be deployedas categories posited in the description of particular languages.

The systemic view on where the line is to be drawn between theory and de-scriptions has been adopted to ensure that descriptive categories are not merelypostulated and then assumed to have some kind of universal status but that theyhave to be justified in the course of description of every language. Whether or notsuch descriptive categories can be applied in the description of several languages(or even all languages) is an empirical question, not a theoretical one; and it is aquestion to be decided only after the categories have been independently motivatedin comprehensive descriptions of every language. Descriptive categories thus haveto be justified by reference to the patterns of a particular language not by appealto some abstract universal. This principle was already part of Firthian practice andwas articulated by Firth (1957:2122):

Though it is found convenient to employ the words noun, verb, pronoun, parti-cle, for example, it must not be assumed that in all languages, nouns and verbsare to be found as the universalists might express it.

It has been held that in some Melanesian languages the noun-verb dis-tinction is unnecessary. The universalist fallacy is constantly with us. It issometimes said that there are no real adjectives in Swahili, and that adjec-

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.12 (764-827)


tives are really verbs in Japanese. The first step towards adequacy in the higherlevels of linguistic analysis is the same rigorous control of formal categoriesset up and of the terms applied to them, as is now the rule in all forms ofphonological analysis. This does not mean that the analysis of discourse ofthe paragraph and the sentence, for example can be directly developed fromphonemic procedures or even devised by analogy from such procedures. . . .

Reverting to the discussion of grammatical categories in closed systems forany language, the universalist is reminded that the grammatical meaningsare determined by their inter-relations in the system set up for that language.A nominative in a four case system would in this sense necessarily have a dif-ferent meaning from a nominative in a two case or in a fourteen case system,for example.

Firths point was that the grammatical categories of a language are determinedby their inter-relations in the system set up for that language; this is absolutelyfundamental in systemic functional work. The point is illustrated by the systemnetworks that are presented throughout this book in the accounts of the variouslanguages discussed here. Some specific illustrations will be given in the final chap-ter of the book, such as the comparison of the grammar of mental clauses inEnglish, Japanese and Tagalog. One consequence of Firths point is that if categoriessuch as causation or possession are to be compared or typologized across lan-guages, they have to be shown to emerge as language-specific categories that canbe located systemically within the total description of each particular language. (Inother words, the valeur of a category has to be specified within the relevant subsys-tem of language; it is not sufficient just to operate with some signification, possiblyassumed to be universal: cf. Hasan 1985, on the complementarity of valeur andsignification in language.)

We have established that theory and description are kept distinct in systemiclinguistics; but what is the relationship between the two? The relationship can bemodelled in terms of abstraction: the general theory is more abstract than par-ticular descriptions. More specifically, the general theory of language stands ina realizational relationship to descriptions of particular languages: the descrip-tion of a particular language is a realization of the general theory (cf. Matthiessen1995a:5860; Matthiessen & Nesbitt 1996; Teich 1999, 2001). This realizationalrelationship between theory and description is represented diagrammatically inFigure 1.1 on page 17. For example, Subject is a category used in the descriptionof a number of languages (but not all see for example Martins account of Tagalogand Roses account of Pitjantjatjara in this book) and it realizes the theoretical cate-gory of interpersonal structural function at the stratum of lexicogrammar. Giventhis interpretation of Subject, the theory makes the very general prediction that itssemantic correlate will be significant in dialogic negotiation, but the theory doesnot indicate that Subject will combine with Finite in English to form the Mood

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.13 (827-876)

Introduction

element (cf. Section 1.4.1 below). This division of labour between theory and de-scription is appropriate: semantically, Subject will always be a dialogic function;but the combination of Subject and Finite to form the Mood element is specific tocertain languages only (English and German, for example; but not French, Chineseand Vietnamese). Within the realm of description there is of course still room forthe identification of descriptive motifs and generalizations. This is the domain ofdescriptive typology and is the concern of the final chapter of this book. But suchgeneralizations are still descriptive, not theoretical: they are generalizations aboutdescriptive categories; they are not more abstract theoretical ones.

The general theory interprets language as a multidimensional semiotic system;we have mentioned two dimensions so far stratification and metafunction andwe will discuss these and other dimensions in Section 1.3 below. The multidimen-sional theoretical organization is projected onto the descriptions of particularlanguages, as illustrated first for English in Section 1.4 of this chapter and then forseveral other languages in the remaining chapters of the book. The description ofa given language will thus fall somewhere within the multidimensional space de-fined by the general theory. As was pointed out above, the general theory has beendesigned to allow for considerable variation. Any particular descriptive categoriessuch as the systems of mood, polarity and tense, the structural functions ofSubject, Predicator, Actor, Goal, Deictic and Auxiliary, and the grammatical classesof nominal group, verb and preposition are realizations of categories defined bythe theory system, structural function, and class in the case of the illustrationjust given; but they are not themselves theoretical categories. Such descriptive cat-egories are posited as part of comprehensive descriptions of particular languagesand as already noted they have to be motivated and justified by reference to the lan-guages in question. Once comprehensive descriptions of particular languages havebeen established, it becomes possible to look for typological generalizations acrossthese descriptions. For example, if the system of polarity has been established inthe descriptions of a large number of languages, we can ask whether it is possibleto make a typological generalization about the relative markedness of the terms ofthe system, positive and negative: is the negative always the marked term? Thisprinciple has informed the present book: the bulk of the book consists of execu-tive summaries of more comprehensive descriptions of a range of languages andthe final chapter attempts to identify descriptive motifs and generalizations.

If we embark on a description of a language that has not previously been de-scribed in terms of systemic functional theory or indeed any other linguistic theory,we can build up the description from scratch using only the general theory as aguide. In many ways, this would be the ideal approach because it would avoid thedanger of imposing (our experience of) a description of another language on thenew language being described a danger manifested in the way that missionarylinguists imposed the categories of traditional descriptions of Latin on languages

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.14 (876-943)


around the world and in the way that modern generative linguists have imposedthe categories of formal descriptions of English on languages around the world.Tozzers warning from the early 20th century in the context of his work on Mayagrammar is still valid:

The Spanish priests did not stop with translations of documents into the na-tive languages but they wrote grammars and collected vocabularies as well.These grammars and dictionaries exist in great numbers. There is hardly a di-alect spoken in Mexico or Central America that has not some sort of grammardealing with the structure of the language. The difficulty met with in usingthese grammars written by the Spanish is the same as that found wherever aprimitive language has been studied and recorded along the lines and with thecorresponding forms found in Spanish, Latin, or some other Indo-Europeangrammar. The Spanish priest thought he had successfully written a grammarof a native language if he had found forms in that language to correspondto every term in his Spanish grammar. The desire to find words which fit-ted the different categories of thought expressed in his own grammar oftenoutweighed his keenness in realizing that many grammatical forms used inSpanish could not be properly expressed in the native language. Parallels weresought for every form in the Spanish or Latin. The investigators usually foundsome native term which seemed to them to conform to the same expression intheir own language. If a native did not seem able at first to give words for thepluperfect tense in his language, the more one insisted that there must be suchforms the sooner the native would give something which superficially seemedto be a pluperfect.

The whole difficulty lies in the fact that it is impossible to build up a gram-mar of a primitive language by following a Latin or Spanish model. This rigidadherence to such a model leads to two defects. Forms are given the investiga-tor, often after repeated questioning, which only vaguely express correspond-ing forms in Spanish or Latin. These are often unnatural and are compoundedso as to express in a most artificial way the idea desired. The second defect isthe greater as scores of native expressions are entirely overlooked and are neverrecorded in the early grammars as there are no forms corresponding to themin Latin. (Tozzer 1921/1977:78)

This is a warning that is supported by the Boasian, anthropological linguistic tradi-tion and it is one which is hopefully heeded by systemic functional linguists (withthe recognition that there are no primitive languages!). Thus when Minh DucThai began to explore the grammar of Vietnamese in systemic functional terms,he made no descriptive assumptions based on English or indeed on any other lan-guage about the textual organization of the clause. Instead he selected Vietnamesetexts from a range of registers, each with a clear, easily identifiable method of devel-opment, and he then analysed the clauses realizing these texts to find out whether

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.15 (943-988)

Introduction

there was any correlation with method of development (thus following Martins1983 model of moving up to the discourse semantics in work across languagesto avoid being trapped in grammatical assumptions inherited from other descrip-tions). It turned out that there was: the beginning of Vietnamese clauses variedaccording to the method of development. This then suggested a thematic interpre-tation of the beginning of the clause in Vietnamese: see further Thai (1998; thisvolume). There is of course ample evidence for making an assumption along theselines including Li & Thompsons (1976) classic contribution to typology in thisarea. But the point was that the description of the Vietnamese system had to bebuilt up from, and justified in terms of, patterns in Vietnamese text. (There are infact two important methodological principles at work here. The first is to base theinterpretation on the language being described; the second is to develop the de-scription of the language by reference to evidence from text instances see Section1.3.2 below.)

However, the type of approach where no assumptions are made based on otherlanguages and where the description of the lexicogrammatical system is built upfrom observations of discursive instances takes a considerable amount of time, soas a practical heuristic, it may be helpful to model the description of one languageon the description of another this is the method of transfer comparison (seeHalliday 1960/6; Teich 1999):

In the comparison of languages we may take advantage of the fact that . . .there are always several different ways of describing the same linguistic phe-nomenon; it is thus possible to adopt the description of one language to thatof another. The aim of this transfer comparison is to draw attention to theresemblances between the two languages. (Halliday 1966:39)

As a complement or as an alternative to transfer comparison, we can draw on apool of typological generalizations. Thus we can attempt to set up a generalizedfunctional map of the grammar of languages organized according to metafunctionand rank (cf. Section 1.4 below). Such a map should identify core systems such asmood, transitivity and theme that are found in one language after another, spec-ifying the envelope of variation within each system (see the final chapter of thisbook). It should also identify systems that expand these core systems but whichare more likely to vary from one language to another systems such as politeness,honorification, localization, determination and systems that expand these coresystems but which are found in complementary variants systems such as tenseversus aspect, modality versus evidentiality.

If the method of transfer comparison is used, it is important to remember thesource of the description and to avoid taking the new description too seriously orusing it as direct evidence for further typological generalizations! A fundamentalway of checking any account of the system of a language is always to apply it in

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.16 (988-1046)


extensive discourse analysis a task that has been central on the systemic functionalresearch agenda since the beginning.4 Thus major systemic functional studies ofvarious languages include lengthy examples of discourse analysis. While there is notenough space in the present volume to demonstrate the use of discourse analysis,many of the contributors have produced longer works that include examples ofgrammar-based discourse analysis.

. Mapping language the dimensions of systemic functional theory

In the previous section, we discussed the distinction between theory and descrip-tion. In this section, we will be concerned with systemic functional theory, and inthe following section, we will turn to systemic functional description.

.. Language as multidimensional semiotic space

As already noted, systemic theory is a resource for construing language as a higher-order semiotic system. Construing language as a higher-order semiotic systemmeans mapping the semiotic resources of language mapping its potential formaking meaning. The cartographic metaphor informs the descriptions of lan-guages in this book (cf. Matthiessen 1995a, on lexicogrammatical cartography).The metaphor works well together with the common way of conceptualizing lan-guage and other systems of meaning in terms of an abstract semiotic space: mapsare comprehensive models of a semiotic space, showing how everything is locatedin relation to everything else. The theory should enable us to see and represent allfeatures of the semiotic landscape of language, including those that are covert andmight not be covered by a traditional map. This is of course particularly importantin the context of typological work, where we have to be able to interpret a rangeof potentially very different languages. The theory provides us with the potentialfor mapping them out in very general terms, scaffolding the particular descriptiondeveloped for each language being investigated.

The systemic functional approach to semiotic cartography is holistic ratherthan componential (see Matthiessen & Halliday, in prep.: Chapter 1, and cf. Capra1996, on these two approaches as alternative strands in scientific thinking): com-prehensive maps are constructed first and then local areas are mapped at a greaterscale of delicacy. Even if the scale of the global map has to be such that the de-tailed features of many domains cannot be discerned at first, the global map makesit absolutely clear where those domains are located in relation to one anotherand in relation to the overall semiotic space. This holistic approach is based on

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.17 (1046-1063)

Introduction

THEORY

descriptive motifs &generalizations acrosslanguages descriptive typology

generalization

description of language A

description of language B

description of language C

description of language D

description of language E

description of language F

...

DESCRIPTIONS

theory of (human)language as a

higher-order semioticrealization (abstraction)

Figure 1.1 The relationship between theory and description

systems thinking rather than on the Cartesian analysis that informs the com-ponential approach.

The componential approach has been the dominant one in western scholar-ship, going back at least to Descartes and the early phase of modern science (whereit served as a way of coping with the complexity of the phenomena being observed).It has been characteristic of a great deal of work in linguistics including formallinguistics in the 20th century; one central manifestation of this approach has beenthe focus on constituency analysis. In contrast, the holistic approach has developedon the periphery of the componential mainstream. It has informed the develop-ment of ecological thinking in biology and of contextual thinking in anthropologyand linguistics; and in the second half of the 20th century it has been embodied insystems thinking. Holistic thinking is characteristic of systemic functional linguis-

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.18 (1063-1122)


tics; two central manifestations of this approach are the focus on context and theforegrounding of systemic organization.

The cartographic concern in systemic functional linguistics is thus one centralaspect of the holistic approach. A systemic functional map of language construes itas being organized along a number of semiotic dimensions. Two of these dimen-sions have already been mentioned stratification and metafunction; these are thedimensions along which language can be distinguished from other semiotic sys-tems. But the total semiotic space of the resources of language is multidimensional.In the remainder of this section, we will introduce the dimensions of systemicfunctional theory that will be needed in this book.

The dimensions organize either the whole system of language in context oreach stratal subsystem. The former may be called global and the latter local: seeTable 1.1. The dimensions intersect with one another to organize the multidimen-sional semiotic space of language in context. This space can be mapped by means ofa series of two-dimensional matrices (see Matthiessen & Halliday, in prep.: Chap-ter 2); since the bulk of this book is concerned with lexicogrammar, we will onlypresent one such matrix the lexicogrammatical metafunction-rank matrix to beintroduced in Section 1.4 below.

.. Instantiation (system to text)

Imagine that were embarking on the description of a language that is being de-scribed scientifically for the first time. Since there are no existing descriptions ofthe system of this language, our main source of information will be what we canobserve. What we can observe is the language as text as language functioning incontext. (Note that the term text covers all modalities; it may be spoken, writtenor signed.) The texts we observe will be very varied in terms of what contexts theyoccur in, how they are organized, how long they are, and so on; and our task willbe to collect a significant range of texts, taking note of their contexts and makingsure we include the contexts that are critical to daily life. These texts must in thefirst instance be naturally occurring ones so that we can be sure that we have notdisturbed the phenomena we are trying to study as we would be sure to do if wewere to elicit examples based on, for example, categories in another language.

We can of course make our goal the description of each text as a phenomenonin its own right. The result would be a collection of commentaries on the model ofexplication de texte; but we would not be able to generalize beyond the individualtexts. If our field of study was meteorology, this would be analogous to observingthe weather on a number of occasions and then going on to produce commen-taries on each instance that had been observed without generalizing across all theseinstances (much as we would in a narrative but unlike what we would do in a sci-

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.19 (1122-1191)

Introduction

Table 1.1 Semiotic dimensions of systemic functional theory

Dimension Regions within dimensionglobal stratification context / language [(discourse) semantics / lexicogrammar /

phonology (graphology, sign)]instantiation potential (system) subpotential | instance type instance

(text) [i.e. a cline between potential and instance]metafunction [language:] ideational (logical & experiential) / interpersonal /

textual[context:] field / tenor / mode

local rank [semantics:] variable according to register and metafunction[lexicogrammar:] clause / group-phrase / word / morpheme (asin e.g. English)[phonology:] tone group / foot / syllable / phoneme (as in e.g.English)

axis paradigmatic (systemic) / syntagmatic (structural)

entific report). But meteorologists generalize beyond the instances of weather thatthey have observed in order to describe weather patterns and even the climate.Weather patterns and the climate are not different phenomena from the weather:they are all part of the same realm of meteorological phenomena; they differ onlyin generality. A weather pattern is nothing more than an accumulation of a numberof instances of weather; and the climate is nothing more than an accumulation ofa number of weather patterns. By the same token, todays weather here in Sydneyis nothing more than an instance of Sydneys climate.

Todays weather and Sydneys climate are related to one another as instanceto potential along a (meteorological) cline of instantiation: todays weather is aninstance of the climate; the climate is the potentialization of the weather distilledout of innumerable instances of weather. The relationship takes the form of a cline,not of a dichotomy: weather and climate are the end poles of this cline; weatherpatterns and subclimates fall on a range somewhere in between these poles. Asobservers, we can measure properties of the weather temperature, pressure, hu-midity, and so on. We cannot observe the climate directly; we have to infer allpatterns beyond the instance pole of the cline of instantiation.

The cline of instantiation is a dimension that organizes systems of all kinds physical systems like that of meteorology, biological systems, social systems andsemiotic systems. In the realm of semiotic systems, text lies at the instance endof the cline. Text is semiotic weather; but what about the semiotic climate,weather patterns and subclimates? There are in fact clear semiotic analogies. Thesemiotic climate is the overall linguistic system; it is the meaning potential of alanguage.5 Thus a text instantiates the linguistic system; and the linguistic systempotentializes innumerable texts.

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.20 (1191-1256)


Instantiation can be stated probabilistically: terms in the systemic potentialwill be instantiated in text with a certain degree of probability; for example, in thesystem of polarity (positive / negative), the probability that positive will beinstantiated is 0.9 while the probability that negative will be instantiated is 0.1(see Halliday & James 1993). But probability is nothing but the potentializationof relative frequency in text, so any new instantiation of one term or another in theunfolding of text will perturb the system. Systemic probabilities provide us with away of identifying subtle similarities and differences across languages similaritiesand differences which would not be seen if we were to operate with only qualitative,categorical statements (which turn out to be just the limiting case of a probabilityof 1 or 0; cf. Givn 1979:2628, 5153, on Subject and definiteness).

Text and system are the poles of the semiotic cline of instantiation. Intermedi-ate between these two poles are patterns that are more general than specific textsbut more specific than the overall system. These patterns have been described astext types, registers or genres. (These terms refer to categories that are different;but for present purposes we can say that they refer to the region intermediate be-tween text and system.) Text types (registers or genres) range over regions alongthe cline of instantiation; they should be conceived of as fuzzy families rather thanas strict taxonomies.

As observers of language, we investigate properties of texts and move beyondtexts along the cline of instantiation to characterize text types or the overall lin-guistic system by inferring generalizations based on particular texts. This is themethodology of corpus linguistics: we collect a sample of texts or text extracts thatis extensive enough to be representative of some higher point along the cline ofinstantiation of a text type or a family of text types, or of the overall system. Howfar we move along the cline of instantiation towards the system pole is of coursea matter of choice. J. R. Firth was wary of moving too far away from the instancepole of what we now theorize as the cline of instantiation; he preferred to operatewith what we might call the generalized instance and developed the notion of re-stricted languages. However, once the cline of instantiation has been developed asa semiotic dimension within our overall theory, there is less reason to worry aboutmoving all the way to the system pole: the extent to which it is possible to move to-wards that pole while still developing reliable descriptions depends on how muchtext can be observed.

The cline of instantiation is an important theoretical dimension in linguisticresearch in general and in typological research in particular because it makes itpossible to locate text as a source of data in relation to generalizations abouttext types and the overall system. It also makes it possible to compare registerranges across languages (cf. Rose 1998, 2001) and across diachronic variants ofone language (cf. Steiner & Teich, this volume) while at the same time ensuringthat languages are typologized in terms of comparable registers. However, this pic-

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.21 (1256-1334)

Introduction

ture has been clouded by two conceptual dichotomies postulated by linguists in thelast century.

At the beginning of the century, Saussure distinguished between langue andparole; and around the middle of the century, Chomsky distinguished betweencompetence and performance. Parole and performance are, in different ways,conceptualizations of text and they are thus located at the instance pole of the clineof instantiation, and langue and competence are, again in different ways, concep-tualizations of the system and are thus located at the potential pole of the cline; butboth Saussures and Chomskys proposals had the effect of directing subsequentresearch away from the instance pole (with certain important exceptions such asthe work undertaken and inspired by J. R. Firth; one early example of typologicalinterest is Mitchell 1957). Saussurean structuralist research focused on langue andChomskyan generative research on competence; both traditions ignored text. Thereaction has tended to be a move to the other extreme focus on the instancepole combined with a rejection of the system, as has happened in both post-structuralism and conversation analysis, with the latter influencing West-Coastfunctionalism. This seems to us to be an example of the typical swing between the-sis and antithesis in the history of ideas; the synthesis is the reconceptualizationof the dichotomy as a cline the cline of instantiation.

The dichotomous conceptualizations of the cline of instantiation also hadrepercussions within typological research. Most classical typological work hasbeen concerned almost exclusively with the system pole of the cline of instantiationand text has played little or no role. This changed within typological work basedon West-Coast functionalism, with studies such as Hopper & Thompson (1980),Hopper (1982), Givn (1979, 1983), DuBois (1987), and Thompson (1988). Herethe concern with discourse analysis emerging from the 1970s was combined withtypological work. Ultimately this seemed to lead to a crisis of confidence for a num-ber of scholars because they began to question how far it is possible to generalizebeyond (small samples of) text to make generalizations about (what we have herecalled) the linguistic system. (These are questions we can recognize from both post-structuralism and conversation analysis; and they echo Firths concerns from themid-20th century mentioned above.) The solution is, we believe, to model the re-lationship between text and system as a cline, while rejecting dichotomies that treatthe two poles as different kinds of phenomena, and to relate regions along the clinequite explicitly to contextual patterns.

The cline of instantiation is also important in the context of typological workbecause it defines different domains of semogenesis the creation of meaning(see Halliday & Matthiessen 1999). At the instance pole of the cline, semogene-sis is manifested as logogenesis; this is the creation of meaning in the course ofthe instantiation (unfolding) of text. At the potential pole of the cline, semogene-sis is manifested as phylogenesis; this is the creation of meaning in the course of

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.22 (1334-1381)


the evolution of the linguistic system. Within typology, there has been a growinginterest in phylogenesis (see for example Harris & Campbell 1995); and grammat-icalizaton, the evolution of grammatical patterns, has received particular attention(e.g. Hopper & Traugott 1993; Bybee, Perkins & Pagliuca 1994; Traugott & Heine1991). The cline of instantiation helps us see how grammaticalization at the sys-tem pole of the cline emerges from logogenetic patterns: distilled out of relativefrequency in text, systemic probabilities gradually change (the classic study show-ing this was Ellegrds (1953) investigation of what we might now call the gradualgrammaticalization of do in English as a general auxiliary in different grammaticalenvironments).

In terms of linguistic cartography, we can obviously develop maps of any re-gion along the cline of instantiation; but the most comprehensive map will be asurvey of the overall meaning potential of a language: this map will allow us to lo-cate text types and texts within the total semiotic space and it will make it possibleto reason about what would constitute a corpus of instances that is large enoughand registerially balanced enough to shed light on the system. The accounts pre-sented in this book all map the overall linguistic system located at the potential poleof the cline of instantiation; but they are based on investigation of a large number oftext instances and also include descriptions of sample texts (which can be assignedto text types), showing the system being instantiated in particular contexts.

.. Axis (kind of order: system and structure)

How do we map the meaning potential of a language, the subpotential of a texttype or the flow of instantiation of a text? To answer this question, we have tointroduce another semiotic dimension that of axis. Axis is the familiar distinctionbetween two modes of semiotic organization paradigmatic and syntagmatic, alsocalled system and structure or more informally choice and chain. We could mapthe meaning potential in terms of either of these two modes of axial organization.Both are in fact involved in specifications of the potential. For example, if we saythat a clause has the structure Theme Rheme (the functional element Themefollowed by the functional element Rheme), we are stating the structural potentialof the clause. But it is the paradigmatic axis that allows us to map the meaningpotential of a language because this mode of organization takes the form of globalnetworks of contrasting options represented by means of the system network ofsystemic functional theory.

Because the systemic mode of organization allows us to map the meaning po-tential of a language, systemic functional theory foregrounds system (paradigmaticrelations) over structure (syntagmatic relations) as the point of departure for lan-guage description, whereas most other theories have foregrounded structure over

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.23 (1381-1448)

Introduction

system although the situation has begun to change with the emphasis on featurestructures in frameworks such as HPSG. This special orientation gives rise to adistinctive organization in the design of the model, for which we will provide abrief introduction6 here.

Were in a bakeshop and overhear the server say Its right there in front ofyou. How might we take an interest in this? One way would be to consider whatwas said in relation to alternatives. The server has chosen to state something,instead of asking something or telling the client to do something:

Its right there in front of you. statingIs it right there in front of you? asking if Whats right there in front of you? asking whatBe right there (by six). telling to

And we could consider how these choices are related to one another. How is askingdistinguished from stating? Within asking, how is asking if distinguished fromasking what? How is telling to different from any of these? On this tack, we are ininterested paradigmatic relations in precisely how what is heard (the instance) isrelated to, and contrasts with, what could have been (the potential).

Another way to take an interest would be to look into what was said andconsider its own internal structure the parts of whats there. Based on the punc-tuation, for example, there appears to be a strong bond between the parts of Its,which we could confirm by adding a tag to the clause; the tag would pick up justthese elements (shown in bold):

Its right there in front of you, is it?

And right appears to go with there, since we can leave out right or substitute herefor the whole phrase and still have a plausible sentence:

Its there in front of you.Its here in front of you.

Pursuing such procedures (cf. Harris 1946; Wells 1947) we can arrive at an analysisof structure. On this tack, we are interested syntagmatic relations in how theparts of what is heard are related to one another.

In linguistics, the strategies for modelling system in relation to structure arevarious (see Hockett 1954; Robins 1959, for discussion). In SFL, the modellingstrategy is to treat structural relations as derived from paradigmatic ones thatis, to treat structure as a means for realizing choice. To illustrate this, lets takethe English system of mood. Suppose this time, we overhear a little more of theconversation in the bakeshop.

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.24 (1448-1522)


[Server] Why do you wonder? Its right there in front of you. Take one. Canyou reach them?

Now we have a range of alternatives instantiated, to which we might assign tradi-tional Latinate labels as follows:

Its right there in front of you. stating declarativeCan you reach them? asking if interrogative: polarWhy do you wonder? asking why interrogative: whTake one. telling to imperative

More to the point is the basis for this labelling, once we import it into an anal-ysis of English grammar. In this regard we might note the absence of a Subjectand a verb realizing tense or modality in the imperative, and the presence of bothof these features in the declarative and interrogative. We might also note that theSubject and realization of tense or modality is sequenced differently in the declar-ative (Subject preceding) and interrogative (Subject following). In addition thereis the presence of an initial interrogative word in the wh interrogative. On theseand other grounds we might suggest that we need to do more than simply list thechoices were reviewing here we need to be more explicit about their interde-pendency. A notation for expressing these relationships as an image is presented asFigure 1.2. Read from left to right this diagram opposes imperatives to indicatives,divides indicatives into declaratives and interrogatives, and further divides inter-rogatives into polar and wh varieties (the square brackets in the network indicatealternatives; e.g. if indicative, then declarative or interrogative but not both andnot neither).

Alongside information about how mood choices are related to one another,this network of systems specifies their structural consequences. These realizationstatements are included in boxes just under the feature they realize.7 Thus, if theclause is indicative, it contains a Subject function, and a Finite function realizingtense or modality (+ means include that is, present in the structure of theunit in question); if it is declarative, the Subject precedes the Finite, whereas if itis interrogative, the Finite precedes the Subject (^ means is followed by); andwh interrogatives have an additional Wh function, which conflates with somethingelse (Subject, Complement or Adjunct; / means conflate (unify) with anotherfunction) and comes first (#^ means initial position). More delicate features(to the right) inherit realizations from less delicate ones (to the left); this helps toclarify the nexus between system and its structural output in the model.

On the basis of this network and closely related reasoning and systems, we canprovide structural analysis of particular examples. For Can you reach them? for ex-ample, we can recognize a Finite function realizing modality (can), followed by theSubject (you); these are the two functions that are replayed in English tags (e.g.

JB[v.20020404] Prn:26/11/2004; 10:29 F: CI25301.tex / p.25 (1522-1589)

Introduction

Figure 1.2 English mood some key systems and realizations

for Australian English, Can you reach them, can you?). These are followed by thePredicator

Documents

(Amsterdam Studies in the Theory and History of Linguistic Science, Series IV Current Issues in Linguistic Theory)Christian M. I. M. Language Typology a Fun