Construing Experience Through Meaning

Download as pdf or txt
Download as pdf or txt
You are on page 1of 672

Construing Experience Through Meaning

Open Linguistics Series


Series Editor: Robin Fawcett, University of Cardiff

This series is 'open' in two senses. First, it provides a forum for works associated
with any school of linguistics or with none. Most practising linguists have long
since outgrown the unhealthy assumption that theorizing about language should
be left to those working in the generativist-formalist paradigm. Today large and
increasing numbers of scholars are seeking to understand the nature of language
by exploring one or other of the various cognitive models of language, or in terms
of the communicative use of language, or both. This series is playing a valuable
part in re-establishing the traditional 'openness' of the study of language. The
series includes many studies that are in, or on the borders of, various functional
theories of language, and especially (because it has been the most widely used of
these) Systemic Functional Linguistics. The general trend of the series has been
towards a functional view of language, but this simply reflects the works that have
been offered to date. The series continues to be open to all approaches, including
works in the generativist-formalist tradition.
The second way in which the series is 'open' is that it encourages studies that
open out 'core' linguistics in various ways: to encompass discourse and the
description of natural texts; to explore the relationships between linguistics and its
neighbouring disciplines - psychology, sociology, philosophy, cultural and
literary studies - and to apply it in fields such as education, language pathology
and law.
Relations between the fields of linguistics and artificial intelligence are covered
in a sister series, Communication in Artificial Intelligence, Studies that are
primarily descriptive are published in a new series, Functional Descriptions of
Language.
Recent titles in the series:
Educating Eve: The 'Language Instinct' Debate, Geoffrey Sampson
Genre and Institutions: Social Processes in the Workplace and School, Frances
Christie and J.R. Martin (eds)
Language Policy in Britain and France: The Processes of Policy, Dennis Ager
Language Relations across Bering Strait: Reappraising the Archaeological and
Linguistic Evidence, Michael Fortescue
Learning through Language in Early Childhood, Clare Painter
Pedagogy and the Shaping of Consciousness: Linguistic and Social Processes,
Frances Christie (ed.)
Researching Language in Schools and Communities: Functional Linguistic
Perspectives, Len Unsworth (ed.)
Summary Justice: Judges Address Juries, Paul Robertshaw
Ways of Saying: Ways of Meaning: Selected Papers of Ruqaiya Hasan, Carmel
Cloran, David Butt and Geoffrey Williams (eds)

Construing
Experience Through
Meaning
A Language-based Approach to
Cognition
M.A.K. Halliday and
Christian M.I.M. Matthiessen

continuum
LONDON

NEW YORK

Continuum
The Tower Building, 11 York Road, London SE1 7NX
80 Maiden Lane, Suite 704, New York, NY 10038
First published 1999
Reprinted 2000,2001,2002,2006
M.A.K. Halliday and Christian M.I.M Matthiessen 1999
All rights reserved. No part of this publication may be reproduced
or transmitted in any form or by any means, electronic or mechanical,
including photocopying, recording or any information storage or
retrieval system, without permission in writing from the publishers.
British Library Cataloguing in Publication Data
A catalogue record for this book is available from the British Library.
ISBN 0-8264-5152-7
Library of Congress Cataloging-in-Publication Data
Halliday, M.A.K. (Michael Alexander Kirkwood), 1925Construing experience through meaning: a language-based approach
to cognition/M.A.K. Halliday and Christian M.I.M. Matthiessen.
p. cm. (Open linguistics series)
Includes bibliographical references (p.) and index.
ISBN 0-8264-5152-7
1. SemanticsPsychological aspects. 2. Grammar, Comparative and
general. 3. Concepts. 4. Metaphor. 5. Computational linguistics.
6. Cognitive science. I. Matthiessen, Christian M.I.M.
II. Title. III. Series.
P325.5.P75H35 1999
401'.43-dc21
99-19270
CIP

Printed and bound in Great Britain by


Biddies Ltd., King's Lynn, Norfolk

Contents
Preface
Part I: Introduction
1. Theoretical preliminaries
1.1
Meaning base
1.2
Functional grammatics and functional semantics
1.3
Metafunctional diversification
1.4
The scope of the ideation base
1.5
Guiding principles
1.5.1
Approach from grammar
1.5.2
Meaning constructed in grammar
1.5.3
Semogenesis
1.5.4
Semantics in relation to grammar
1.6
Codifying
1.7
Realization
1.8
Grammatical evidence
1.9
Ways of representing semantic organization
1.9.1
Stratification of metalanguage
1.9.2
Logic
1.9.3
Discrimination networks
1.9.4
Componential analysis
1.9.5
Frame-based inheritance networks
1.9.6
Systemic representation
1.10 Organization of the remainder of the book
Part II: The ideation base
2. Overview of the general ideational potential
2.1
Phenomena
2.2
Sequences
2.3
Figures
2.4
Elements
2.5
Participants
2.6
Simple things
2.7
Simple qualities
2.8
Circumstances
2.9
Processes
2.10 Summary
2.11 Construing experience in the ideation base
2.11.1 Typological and topological perspectives
2.11.2 Construing categories: ontogenetic perspective
2.11.3 Construing as locating in a network of relations
2.11.4 Constnial and stratification
2.12 Construal in the grammar: summary of grammatical evidence
2.12.1 Types of figure
2.12.2 Types of element
3.
Sequences
3.1
Natural logic of sequences and prepositional logic
3.2
Expansion and projection
3.3
Projection
3.3.1
Two levels of projected content: ideas and locutions
3.3.2
Propositions and proposals
3.4
Expansion
3.5
The relational character of sequences
3.6
Sequences and text
3.7
Sequences and grammar
3.8
Conclusion
4.
Figures
Two perspectives on figures
4.1
4.2
Composition: domains of experience
4.2.1
Four primary domains
4.2.1.1
Symbolic processing: sensing & saying
4.2.1.2 Doing and being
4.2.1.3 Temporal unfolding
4.2.2
Grammatical realization

ix
1
1
1
3
7
13
15
15
17
17
18
22
25
26
29
30
35
37
39
40
41
47
48
48
48
50
52
58
59
60
62
63
64
65
66
68
72
82
96
98
98
99
104
104
106
108
108
113
117
119
122
124
126
128
128
128
128
128
132
133
134

vi

Contents
4.2.3

5.

6.

7.

Further delicacy
4.2.3.1 Types of sensing
4.2.3.2 Types of being & having
4.2.3.3 Types of doing & happening
4.3
Composition: two models of participation
4.3.1
Particularistic model
4.3.2
Generalized model
4.3.3
Complementarity of doing and being
4.4
Degree of participation
4.4.1
Construing out participants
4.4.2
Degrees of participanthood
4.5
Degree of involvement
Elements
5.1
The primary types of element
5.2
Similarities and differences between participants and processes
5.3
Participants
5.3.1
Things and qualities
5.3.2
Things
5.3.2.1 Configurational roles of things
5.3.2.2 Weight and discursive force of primary categorization
5.3.2.3 Elaboration of things into micro categories
5.3.2.4 Internal structure of things
5.3.2.5 Systematic relationships among things
5.3.2.6 Recurrent semantic principles
5.3.3
Qualities
5.3.3.1 The status of qualities
5.3.3.2 Types of quality
5.4
Processes
5.4.1
Makingsense of time
5.4.2
Patterns of time in English
5.5
Circumstances
5.6
Major motifs in the ideation base
Grammatical metaphor
6.1
Congruent and metaphorical variants
6.2
The nature of grammatical metaphor
6.2.1
Lexical and grammatical metaphor
6.2.2
Semogenetic priority of the congruent mode
6.2.3
Location of grammatical metaphor in the content system
6.2.4
Metafunctional effect of metaphoric shift
6.3
How grammatical metaphor evolves: transcategorization
6.4
Types of grammatical metaphor: elemental
6.5
Syntagmatic and paradigmatic dimensions of grammatical metaphor
6.5.1
Syntagmatic complexity: syndromes of elemental metaphors
6.5.2
Paradigmatic complexity: degree of metaphoricity and steps in
"unpacking"
6.6
Metaphor, transcategorization & rankshift: semogenic resources
6.7
Interpretation of grammatical metaphor
6.7.1
Motifs in grammatical metaphor
6.7.2
Metaphoric instability of relators
6.7.3
Grammatical metaphor and experiential meaning
6.7.4
The significance of grammatical metaphor
6.7.5
Metaphor as ideational reconstruction
6.7.6
Metaphor as ideational reconstruction
6.8
Representing grammatical metaphor
6.8.1
Recapitulation for the purpose of representation
6.8.2
Representing text examples
6.8.3
Grammatical metaphor as elaborating relation between semantic
configurations
6.8.4
The representation of metaphor in the system
Comparison with Chinese
7.1
Historical background
7.2
Some general features compared
7.3
Sequences
7.4
Figures
7.5
Elements
7.5.1
Nominal groups
7.5.2
Verbal groups
7.5.3
Prepositional phrases

137
137
144
147
149
150
153
159
165
165
167
172
177
177
178
182
184
187
188
189
194
197
197
202
205
205
209
213
213
215
217
222
227
227
232
232
235
236
238
242
244
249
250
255
259
263
263
267
268
270
272
275
278
278
281
288
293
297
297
299
301
303
305
305
306
308

Contents
7.6
7.7
7.8

vu
Processes and things
Grammatical metaphor
The meaning base: concluding remarks

Part III: The meaning base as a resource in language processing systems


8.
Building an ideation base
8.1
General
8.1.1
Field and ideation base
8.1.2
Domain models and the general model
8.2
The language of the weather
8.2.1
Introduction
8.2.2
Sequences
8.2.3
Figures
8.2.3.1 Figures of'existence'
8.2.3.2 Figures of'ascription'(attribution)
8.2.3.3 Summary of figures
8.2.4
Elements
8.2.4.1
Processes
8.2.4.2 Participants
8.2.4.3 Circumstances
8.2.5
Commentary on weather forecasting domain
8.3. The language of recipes
>
8.3.1
The specific domain of arecipe
8.3.2
Relation to general taxonomy: overview
8.3.3
Sequences in the culinary domain
8.3.4
Figures in the culinary domain
8.3.5
Simple things in the culinary domain
9.
Using the ideation base in text processing
9.1
Organization of a generation system
9.1.1
Relationship to a client system
9.1.2
Generation system
9.1.2.1
Stratification of resources
9.1.2.2 Processes of instantiation
9.2
An example from recipe generation
9.2.1
Contextual specification
9.2.2
Semantic generation
9.3

9.2.3

Lexicogrammatical generation

9.3.2

Grammatical metaphor as second-order semantic resource

11.

320
320
320
320
323
328
328
333
336
336
338
342
346
347
347
351
351
355
356
359
364
365
370
372
372
372
374
375
382
387
388
389

394

Ideation base supporting text base


9.3.1
Second order nature of textual metafunction

398
398

9.3.3
9.3.4

403
411

Text base stated in terms of ideation base


Summary example of text base

Part IV: Theoretical and descriptive alternatives


10. Alternative approaches to meaning
10.1 Logico-philosophical vs. rhetorical-ethnographic orientations
10.2 Formal semantics
10.3 Cognitive semantics
10.3.1 Two approaches within cognitive semantics
10.3.2 Jackendoff s conceptual semantics
10.4

309
312
314

10.3.3

Fawcett's cognitive model of an interactive mind

11.3.4

The non-transformed relational interpretation

Work on meaning in NLP


10.4.1 Schank's conceptual dependency
10.4.2 Dahlgren's naive semantics
10.5 Stratal perspectives on meaning
Distortion and transformation
11.1 Frames of reference
11.2 The motif of distortion
11.2.1 The notion that language distorts (our thinking about) reality
11.2.2 The notion that lexicogrammar [syntax] distorts semantics
11.3 The status of qualities: an illustration
11.3.1 Some significant transitivity types
11.3.2 Three different transitivity interpretations
11.3.3
The transforming interpretations
11.3.3.1 Subject - Copula - Predicate
11.3.3.2 Subject - Predicate
11.3.3.3 Filling out the paradigm

401

415
415
415
419
425
425
425

428

429
430
435
437
441
441
444
444
446
448
448
452
454
455
457
460

461

viii

12.

Contents
11.3.4.1 The intensive relation: ascription and identification
11.3.4.2 Participants vs. processes
Figures and processes
12.1 Criteria for process typology
12.2 Temporal profile: activities, accomplishments, achievements, and states
12.3 Control (and stativity)
12.3.1 Intentionality
12.3.2 Potency (Chafe)
12.3.3 Power to determine (Dik)
12.4 Foley & Van Valin
12.4.1 Two sets of semantic roles according to specificity/subsumption
12.4.2 Macro-roles
12.4.3 Token-specific roles and verb classification
12.4.4 Dowty's (1979) development of Vendler's verb classes
12.4.5 Verb classes in RRG
12.4.6 Relationship between the two sets of roles
12.5 Inherent indeterminacy, and variation in systemic interpretations
12.6 Conclusion

461
463
466
466
469
476
477
478
485
491
491
492
494
495
497
502
503
505

Part V: Language and the constatai of experience


13. Language as a multifunctional system-&-process
13.1 Types of system-&-process
13.2 Review of modes of meaning - metafunctional and metaphorical
13.2.1 The metafunctions
13.2.2 Ideational metafunction: experiential
13.2.3 Ideational metafunction: logical
13.2.4 Interpersonal metafunction
13.2.5 Textual metafunction
13.2.6 Ideational metaphor
13.2.6.1 Deaf Sign
13.2.6.2 Metaphoric and congruent wordings
13.2.6.3 Origins of grammatical metaphor: the evolution of
scientific discourse
13.3 Indeterminacy and probability in language
13.3.1 Types of indeterminacy
13.3.2 Probability
13.3.3 The significance of indeterminacy
13.4 Polysystemicness
14. Construing ideational models: consciousness in daily Ufe and in cognitive science
14.1 Models embodied in the ideation base
14.1.1 Models at different levels of abstraction
14.1.2 Models at different levels of awareness
14.1.3 Emergence of scientific registers
14.1.4 The shift from folk models to scientific ones
14.2 Sensing: the folk model
14.3 From folks to scientists
14.4 The model in mainstream cognitive science
14.5 Beyond sensing- folk & scientific
15. Language and the making of meaning
15.1 Language and other semiotic systems
15.1.1 The representation of meaning in language, in relation to other
semiotic systems
15.1.2 Socio-semiotic and bio-semiotic systems
15.1.3 The evolution of reality construction
15.1.4 Biological, social and socio-semiotic environments for the
"individuar
15.2 Construing experience: ontogenetic perspective
15.2.1 Protolanguage and language
15.2.2 Generalization, abstraction and metaphor

507
507
507
511
511
512
519
523
528
532
533
535

610
611
611
615

References
Index

619
642

539
547
549
552
557
562
565
565
565
568
569
572
577
586
595
599
602
602
603
605
607

Preface
This book was conceived dialogically : it started as notes on discussions between the two
authors when CM was working at the University of Southern California's Information
Sciences Institute and MAKH was visiting there as a consultant. (The earliest draft is
reproduced as Figure 0-1 on page xiii.) It then developed diachronicaUy from time to
time and diatopically from place to place: the dialogue continued at a summer
school in the midwestem United States; on a tour round England with long walks in
Yorkshire and in Devon; during a visit to Singapore sustained by tropical fruit; in Japan
against the background of the fireworks of autumn colours; and in Australia on beaches
and riversides in northern New South Wales. Critical moments were times spent on the
campus of Bloomington, Indiana; outside the village pub, the Buck, at Buckden,
Yorkshire; and in the countryside around Bellingen, N.S.W. Hie text suggests something
of the interval that elapsed during this 3-B; the chapters were not of course written in the
order in which they appear, and in any case they have all been revised many times before
reaching their present form, but we have not attempted to hide the rather sporadic manner
in which the book came into being.
What has remained constant, throughout this history, is the way we have
conceptualized our topic and its location in a neighbourhood of related theoretical
enquiries. It can be located, perhaps, along the dimensions of a three-dimensional space.
First, we see it as a contribution to the growing body of work in systemic functional
linguistics, work which extends across general theory of language, lexicogrammar and
semantics, text structure and discourse analysis, dld language development language
education, and natural language processing by computer. Secondly, we have tried to orient
it (like systemic functional theory itself) towards the outside: that is, to theorize language
in a way that is relevant to applications in research and practice, focussing on other
people's questions about language at least as much as on questions generally formulated
by linguists. And thirdly, it is outward-racing in another sense: looking from language
out towards its context, to what people do with language (whether this is modelled as
social action, as cognitive process or as some form of abstract value system). Intersecting
these three dimensions we have tried to represent language as the resource whereby the
human species, and each individual member of that species, constructs the functioning
mental map of their phenomenal world: of their experience of process, both what goes
on out there and what goes on in the realms of their own consciousness.
It seems to us that our dialogue is relevant to current debates in cognitive science. In
one sense, we are offering it as an alternative to mainstream currents in this area, since

Preface

we are saying that cognition "is" (that is, can most profitably be modelled as) not
thinking but meaning: the "mental" map is in fact a semiotic map, and "cognition" is
just a way of talking about language. In modelling knowledge as meaning, we ace
treating it as a linguistic construct: hence, as something that is construed in the
lexicogrammar. Instead of explaining language by reference to cognitive processes, we
explain cognition by reference to linguistic processes. But at the same time this is an
"alternative" only if it is assumed that the "cognitive" approach is in some sense natural,
or unmarked. It seems to us that current approaches to neural networks, "connectionist"
models and the like, are in fact more compatible with a semantic approach, where
"understanding" something is transforming it into meaning, and to "know" is to have
performed that transformation. There is a significant strand in the study of language
not only in systemic functional theory but also for example in Lamb's relational
networks whereby "knowledge" is modelled semiotically: that is, as system-&-process
of meaning, in abstract terms which derive from the modelling of grammar.
The semantic perspective enables us to emphasize four aspects of human
consciousness which have been rather less foregrounded in cognitive approaches. One is
that of meaning as a potential, a systemic resource which is deployed in and
ongoingly modified by individual acts of meaning in language. (Whereas most
theoretical work in linguistics since the mid century has focussed strongly on
syntagmatic relations what goes with what, systemic theory has foregrounded the
paradigmatic what is meant in relation to what might be). The second is that of
meaning as growth, a semogenic resource which is constantly expanding in power by
opening up new domains and refining those that are already within its compass. The third
is that of meaning as a joint construction, a shared resource which is the public
enterprise of a collective (whereas "thinking" is essentially a private phenomenon
"located" within the individual). The fourth is that of meaning as a form of activity, a
resource of energy which is powered by the grammar at the heart of every language.
Because we are viewing language from round about, so to speak, rather than from a
disciplinary vantagepoint within linguistics, we have tried to stress what language does
rather than what it is (our notion of the "ideation base" is that of construing experience
through meaning, as embodied in our main title). We have hoped that what we say about
it will resonate with findings from other sources: particularly perhaps those of
neuroscience, with its new understanding of the nature and evolution of human "higher
oider" consciousness. The context within which we have more specifically located
ourselves is that of computer science; not so much that of natural language processing as
a specialized activity within computing, and certainly not with any suggestion that
meaning can be reduced to computation or the brain to a sophisticated computational

Preface

xi

device (we prefer Edelman's analogy of the brain as a complex jungle), but rather taking
the view (following Sugeno) that all computing is essentially computing with meaning
and that "intelligent" computing requires learning to operate with the semantic practices
of natural language.
We naturally hope that our approach will be found relevant by those who are engaging
with language from other points of departure, such as education, literature, philosophy,
psychology and sociology. At the same time we should perhaps caution against carrying
over terminological preoccupations from any of these other disciplines. Terms such as
"system", "functional", "semiotic", "ideational" which have conceptual and often
ideological loadings in one context or another are defined here within the conceptual
frameworic of systemic functional linguistics; we have tried as far as possible to make
them explicit within our own text, and at the least to give references to other sources
which will make their theoretical status clear. (We should perhaps make special mention
of the word "semantic", which in lay parlance means almost the opposite of what it
means in linguistics; when a learned judge says that "the difference is semantic" s/he is
saying that two expressions are equivalent in meaning, whereas to us this indicates
precisely that their meanings differ. Somewhat ironically, the person saying this is likely
to be unintentionally right that is, the two wordings in fact usually do differ in
meaning, often rather more significantly than is being allowed for!)
Condensed into one short paragraph, our own point of departure is the following.
Language evolved, in the human species, in two complementary functions: construing
experience, and enacting social processes. In this book we aie concerned with the first of
these, which we refer to as constructing the "ideation base"; and we stress that the
categories and relations of experience are not "given" to us by nature, to be passively
reflected in our language, but are actively constructed by language, with the
lexicogrammar as the driving force. By virtue of its unique properties as a stratified
semiotic system, language is able to transform experience into meaning. In our attempt
to describe this process, we have deliberately used the grammar as the source of
modelling, because we wanted to show how such a process could take place. We have
confined ourselves, in principle, to how it takes place in English; the theoretical concepts
we have used are general to all languages, but the descriptive categories should be
interpreted in the context of a description of English (there is a brief discussion of
Chinese in Chapter 6). We could, obviously, have located our discourse within any of a
large number of other possible frameworks; and we could have oriented it more towards
issues in language education, in sociology, in the construction of knowledge and so on.
But we found that it was easier to make it explicit by locating it in the general context of

xii

Preface

computational research, since it is here that the notion of modelling natural language is
most clearly expounded.

Even in this context, the idea of modelling meaning is still relatively new. When we
began the work that has evolved into the present study, the prevailing ideology was that
of "knowledge representation"; this was conceived of as a piecemeal accumulation of
individual concepts with certain valences attached and having little or no overall
organization. It was not until the latter part of the 80s that the conception of a semantic
system began to gain ground in natural language processing, and it is this concept that
we have tried to develop here. At the same time we could perhaps observe that it has bee
a concern of systemic theory since the 1960s to model the "content plane" in terms of a
strata! pairing within language, with "semantics" as the "upper model" that is construed
by the lexicogrammar as a whole.
We would like to acknowledge our debt to the many colleagues with whom we have
been able to exchange ideas while working on the present book. In particular, we ate
deeply grateful to Ruqaiya Hasan, to Jim Martin and to John Bateman for detailed
comments on various earlier drafts; and to Bill Mann for envisioning and making
possible the entire enterprise of deploying systemic grammar in a text generation task.
John Bateman has shown the power of a "meaning base" in developing various areas
within natural language processing; and Zeng Licheng has extended our work, both in
theory and in application, implementing many of the central categories that make up the
metalinguistic resource which we draw on and which is being developed and deployed
collectively.

Fig. 0-1: Construing experience: the first design

This page intentionally left blank

Part I:
Introduction
! Theoretical preliminaries
1.1 Meaning base
In this book we are concerned with how human beings construe experience. This means,
first and foremost not experience as an instantial product the particulars of the world
that is around us and inside our heads, the particular individuals, the events of last Friday,
and so on but experience as a resource, as a potential for understanding, representing
and acting on reality. It is in terms of this potential that the particulars of daily life aie
interpreted: they make sense because they are instantiations of this potential.
The construction of experience is usually thought of as knowledge, having the form
of conceptual taxonomies, schemata, scripts and the like. We shall offer an interprtation
that is complementary to this, treating experience not as knowing but as meaning; and
hence as something that is construed in language. In other words, we are concerned with
the construal of human experience as a semantic system; and since language plays the
central role not only in storing and exchanging experience but also in construing it, we
are taking language as our interpretative base.
Our focus will be both theoretical and descriptive. We consider it important that
theory and description should develop in parallel, with constant interchange between
them. The major descriptive component is an account of the most general features of the
ideational semantics of English, which is then exemplified in two familiar text types
(recipes and weather forecasts). We have also made a brief reference to the semantics of
Chinese. Theoretical issues have been raised throughout as they became relevant to the
discussion. Our theoretical base is drawn from systemic-functional linguistics, a
particular functional approach to language developed over the last thirty years or so.1 We
will introduce the theoretical abstractions we need, in the course of our presentation.

For a historical survey, see Matthiessen & Halliday (forthc.); for general discussions of
various aspects of systemic linguistics, see e.g. Benson & Greaves (1985, 1988), Benson,
Cummings & Greaves (1988), Eggins (1994), Fawcett (1980, 1988a), Fawcett & Young
(1988), Halliday (1976), Halliday (1978a), Halliday (1985), Halliday (1993b), Halliday &
Hasan (1985), Halliday <fe Martin (1981), Halliday & Martin (1993), Halliday & Fawcett
(1987), Martin (1992), Martin, Matthiessen & Painter (1997), Matthiessen & Halliday
(forthc.), Steiner (1991). For discussions of systemic linguistics in natural language
processing, see e.g. Bateman (1996), Fawcett, Tucker & Lin (1992), Matthiessen &

1 Theoretical Preliminaries

We feel that our topic is central to general linguistics, and also to cognitive science,
an interdisciplinary formation that includes both natural language processing and artificial
intelligence in its domain; and we have tried to address our discourse to researchers in
both these areas. But at the same time our own approach, both in theory and in method,
is in contradistinction to that of cognitive science: we treat "information" as meaning
rather than as knowledge and interpret language as a semiotic system, and more
specifically as a social semiotic, rather than as a system of the human mind. This
perspective leads us to place less emphasis on the individual than would be typical of a
cognitivist approach; unlike thinking and knowing, at least as these aie traditionally
conceived, meaning is a social, intersubjective process. If experience is interpreted as
meaning, its construal becomes an act of collaboration, sometimes of conflict, and
always of negotiation.
While our general concern is with ideational semantics, we shall also sharpen the
focus by considering in particular one specific research application of this, namely natural
language processing by computer, in particular text generation as an example of this.2
What we are putting forward is a semantic interpretation of the ideational component of
the "environment" of a grammar in text gnration. Typically, the environment of the
grammar consists of two or three parts: (i) a knowledge base, representing the experiential
domains within which the grammar is required to operate, together with (ii) a text
planner, assigning appropriate rhetorical structure to the discourse in terms of some
theory of "register" or functional variation in language, and perhaps also (iii) a third
component specifying features of the writer-audience relationship. These three
components correspond to the three "metafunctions" of systemic theory (see further
Section 1.4 below), (i) ideational, (ii) textual and (iii) interpersonal.
Since we are confining ourselves to the ideational metafunction, this might suggest
that what we are doing is modelling the knowledge base for a text processing system.
However, as already noted, our approach contrasts with representations of knowledge in
that in our own work the experiential environment of the grammar is being interpreted
not as knowledge but as meaning. We have therefore referred to this as a meaning base
instead of a knowledge base.
What is the significance of this switch of metaphor from knowing to meaning? A
meaning base differs from a knowledge base in the direction from which it is
construed In modelling the meaning base we are building it 'upwards' from the grammar,

Bateman (1991), Davey (1978), Patten (19SS), Teich (1995), Winograd (1972, 1983), Zeng
(1996).
2

Since our focus is on the resources needed to support language processing in general, the
ideation base would relate to other aspects of this work such as parsing and machine
translation.

1.1 Meaning base

instead of working 'downwards' from some interpretation of experience couched in


conceptual terms, and seen as independent of language. We contend that the conception of
knowledge' as something that exists independently of language, and may then be coded or
made manifest in language, is illusory. All knowledge is constituted in semiotic systems,
with language as tbe most central; and all such representations of knowledge aie
constructed from language in the first place. (Hence when we consider the knowledge
enshrined in a particular discipline, we understand this by examining the language of the
discipline tbe particular ways of meaning that it has evolved The most obvious
example is perhaps that of scientific taxonomies; but aspects of the grammar arc no less
crucial: this will appear below in our discussion of models of agency and of grammatical
metaphor, among other features.)
This suggests that it should be possible to build outwards from the grammar, making
the explicit assumption that the (abstract structure of) categories and relations needed for
modelling and interpreting any domain of experience will be derivable from those of
language. Our contention is that there is no ordering of experience other than the ordering
given to it by language. We could in fact define experience in Linguistic terms:
experience is the reality that we construe for ourselves by means of
language.
Thus "knowledge" and "meaning" are not two distinct phenomena; they aie different
metaphors for the same phenomenon, approaching it with a different orientation and
different assumptions. But in almost all recent work in this area, the cognitive approach
has predominated, with language treated as a kind of code in which pre-existing
conceptual structures aie more or less distortedly expressed. We hope here to give value to
the alternative viewpoint, in which language is seen as the foundation of human
experience, and meaning as the essential mode of higher-order human consciousness. This
may have more to offer both in theoretical power and in relation to the many practical
tasks for which we need to engage with "knowledge". What we are doing is mapping back
on to language those patterns that were themselves linguistic in their origin.

1.2 Functional grammatics and functional semantics


For the task of constructing such a "meaning base", we shall use a systemic grammar.
A systemic grammar is one of the class of functional grammars, which means (among
other things) that it is semantically motivated, or "natural". In contradistinction to
formal grammars, which are autonomous, and therefore semantically arbitrary, in a
systemic grammar every category (and "category" is used here in the general sense of an
organizing theoretical concept, not in the narrower sense of 'class' as in formal grammars)

1. Theoretical Preliminaries

is based on meaning: it has a semantic as well as a formal, lexicogrammacal reactance.3


Looked at from the formal angle, of course, this means that it is likely to appear
complex; many of the categories are "cryptotypic", manifested only through a long chain
of realizations (a "realizational chain"). Hence it takes a long time and a great deal of
effort to get such a grammar off the ground in any context (such as natural language
processing) requiring total explicitness. Once airborne, however, because it is
semantically natural it has considerable potential as the basis on which to represent
higher level organization provided, that is, such organization is interpreted in
linguistic terms, as meaning rather than as knowledge. For a recent view of formal and
functional approaches to the description of language, see de Beaugrande (1994).
We might refer to the Hjelmslevian notion of the "content plane" as incorporating
both a grammar and a semantics (e.g., Hjelmslev, 1943). Grammar and semantics are the
two strata or levels4 of content in the three-level systemic theory of language, and they are
related in a natural, non-arbitrary way. The third level is the level of expression, either
phonology or graphology. We can draw this model of language as in Figure 1-1. Here the
relationship is not one of 'consists of or 'is a subset of: the concentric circles show the
stratal environment of each level thus lexicogrammar appears in the environment of
semantics and provides the environment for phonology. This ordering of levels is known
as stratification. We have used circles for all three levels to represent the fact that they
are all based on the same fundamental principles of organization: each level is a network
of inter-related options, either in meaning, wording or sounding, which are realized as
structures, based on the principle of rank. These principles will become clear later in our
discussion.
Language, therefore, is a resource organized into three strata differentiated according
to order of abstraction. These strata are related by means of realization. Semantics, or the
system of meaning, is realized by lexicogrammar, or the system of wonting (that is,
grammatical structures and lexical items); and lexicogrammar is realized by phonology, or
the system of sounding.5 For instance, a sequence of figures (a sequence of

The reactance of a category is its distinctive treatment; cf. Whorf (1956), and also Section
1.8 below.
4
Note that the sense of level here is that of 'stratum1 (the term made very explicit in Lamb's,
e.g. 1965, straficaonal linguistics; see also Lockwood, 1972). In linguistics in general,
the term level is used sometimes in this sense, but sometimes in the sense of 'rank* (as in
phrase level) the hierarchy of units according to their constituency potential. (The
ambiguity resides in the overlap between two grammatical relations, those of elaboration
('be') and of extension ('have'), a distinction that will be introduced later in our account of
the relations that construct sequences: see Chapter 3, Section 3.7.2 and Chapter 4, Section
4.2.3.2.)
5

This is the traditional formulation; more properly: semantics is realized by the realization of
lexicogrammar in phonology. For further discussion of realization in reference to the model
presented here, see Halliday (1992) and Matthiessen & Halliday (forthc.).

1.2 Functional grammatics and functional semantics

configurations of processes with participants and attendant circumstances) at the level of


semantics is realized by a complex of clauses at the level of grammar; this, in turn, has
its own realization in the phonology for example, a particular complex of clauses
might be realized by a particular sequence of tones (pitch contours). Between
lexicogrammar and phonology runs the line of symbolic arbitrariness: prototypically the
relation between these two levels is conventional, whereas that between semantics and
lexicogrammar is prototypically natural. What this means is that experience is construed
twice in the content plane, once semantically and once lexicogiammatically. The
ideational meaning base that we are concerned with in this book is a construct that is
'located' within the semantic system that is, at the highest level and realized in the
lexicogrammar.
content

Fig. 1-1: Language as tri-stratal system


A word needs to be said here about the interpretation of the term "grammar" itself.
As used in systemic theory, this term stands for lexicogrammar. The lexical region, or
lexis, is not a separate component, but simply the most "delicate" end of the (unified)
iexicogrammar (Cross, 1992, 1993; Halliday, 1961; Hasan, 1985a, 1987; Matthiessen,
1991b; Nesbitt, 1994).6 There is a complementarity here. Lexis and grammar are not two

* The "lexicon" in the sense of 'dictionary1 is just one view that can be taken on lexis
an
item-based compilation from the lexical stock. The thesaurus is another view, one that is
more indicative of the systemic organization of lexis. See Matthiessen (1991) and Nesbitt
(1994) for further discussion.

1. Theoretical Preliminaries

different phenomena; they aie different ways looking at the same phenomenon. Some
aspects of this phenomenon of "wording" in language aie foregrounded by viewing it as
grammar, others by viewing it as lexis. Some models of language used in computational
linguistics are lexis based (see tor example, Mel'chuk, 1982; and Sinclair, 1992; cf. also
Becker's, 1975, notion of the phrasal dictionary and the work inspired by Becker's ideas,
e.g. Hovy's, 1988b, text-generator Pauline): all (or most) of what is usually treated
grammatically is subsumed under the concept of lexis. Systemic theory, by contrast,
takes the opposite stance, subsuming all (or most) of what is usually treated lexically
(e.g. in a dictionary) under the concept of grammar. It is outside our present scope to
argue this case in detail; but one of the reasons for doing this is to take maximum
advantage of the naturalness of the grammar in the overall modelling of the system.
In this context, it is useful to remember the origin of the notion of the semantic net
within computational linguistics: when Quillian developed this notion (e.g. 1968), he
used the dictionary as the model. That is, the organization of "knowledge1 embodied in a
semantic net, or 'conceptual* net as it also came to be known, was the organization of
meaning in the dictionary. Our modelling approaches the task from the grammatical end
But the model of lexical organization that emerges when we approach it from systemic
grammar is not that of the dictionary (the lexicon of modem formal grammars), used by
Quillian and most others working on natural language processing, but rather that of the
thesaurus. We shall explain this conception below, in the discussion of delicacy in
Chapter 2, Section 2.11.3.
There is a warning to be issued in connection with the term "grammar". It is not
uncommon in English for the same word to stand both for a phenomenon itself and for
the study of that phenomenon. For example, "psychology" is used to mean both the
study of the "psyche" and the psyche itself (so "feminine psychology" means women's
psychic make-up, not theories of psychology developed by women scholars). In
linguistics, while we do distinguish "language" (the phenomenon) from "linguistics" (the
study of the phenomenon), we fail to make such a distinction with the word "grammar",
which means both the grammar of a language and the study of grammar. To avoid such
pathological ambiguity, we find it helpful to refer to the study of grammar by a special
name, grammatics (see Halliday, 1996). We will use this term from time to time in
order to make it quite clear that we are talking about the model, the theory used to
interpret the phenomenon, and not the phenomenon itself.
Thus we can say that a grammatics is a theory of grammar, while a grammar is
(among other things) a theory of experience. But to show that a grammar is a theory of
experience we use a functional, semantically motivated grammatics, since this allows us
to seek explanations of the form of the grammar in terms of the functions to which
language is adapted. But this closeness of fit between the semantics (i.e. the meaning) and
the grammar does not mean that our grammatics can take over the semantic domain.

1.2 Functional grammatics and functional semantics

Adopting a functional approach enables us to extend the domain of grammar in


significant ways in the direction of semantics not thereby reducing the scope of the
semantics but rather enabling us to investigate how experience is construed in semantic
terms to develop the "meaning base" model that is the topic of the present book.
One essential task for our semantics is that of modelling a particular phenomenon of
the content plane that is known as grammatical metaphor. This is the phenomenon
whereby a set of agnate (related) forms is present in the language having different
mappings between the semantic and the grammatical categories, for example:
alcohol's dulling effect on the brain
alcohol has a dulling effect on the brain
alcohol has the effect of dulling the brain
alcohol affects the brain by dulling it
the effect of alcohol is to dull the brain
the effect of alcohol is to make the brain dull
if one takes/drinks alcohol it makes the brain dull
if one takes/drinks alcohol the/one's brain becomes dull &c.
Since this phenomenon of grammatical metaphor is fundamental to adult uses of
language, we shall take it as a central thrust of our book (see especially Chapter 6). One
way in which we shall seek to demonstrate the validity and power of a semantic approach
is by using it to handle grammatical metaphor, and to show how this pervasive aspect of
the lexicogrammar expands the potential of the meaning base. (Grammatical metaphor is
fundamental in particular to those registers for which natural language processing systems
aie devised. For example, if our task is to build a text generator capable of producing
economic reports, we have to be able to generate text such as the following: While the
March drop in retail sales points to a sharp reduction of growth in consumer credit, the
shortfall of tax. refunds could be partly offsetting.)

1.3 Metafunctional diversification; the ideational metafunction


The content plane of a natural language is functionally diverse: it extends over a
spectrum of three distinct modes of meaning, ideational, interpersonal and textual. These
highly generalized functions of the linguistic system are referred to in our theory as
metafunctions. The ideational metafunction is concerned with construing experience
it is language as a theory of reality, as a resource for reflecting on the world. (For the
distinction of the ideational into logical and experiential see Chapter 13, Section
13.2.) The interpersonal metafunction is concerned with enacting interpersonal
relations through language, with the adoption and assignment of speech roles, with the
negotiation of attitudes, and so on it is language in the praxis of intersubjectivity, as a
resource for interacting with others. The textual metafunction is an enabling one; it is

1. Theoretical Preliminaries

concerned with organizing ideational and interpersonal meaning as discourse as


meaning that is contextualized and shared. But this does not mean processing some preexisting body of information; rather it is the ongoing creation of a semiotic realm of
reality (cf. Matthiessen, 1992, and Section 1.3 below.)
Let us illustrate these categories briefly in reference to the clause. Consider the
following passage concerned with the ontological status of the Velveteen Rabbit, taken
from a children's story; this toy rabbit is confronted by two flesh-&-blood ones:
(I) One sunnier evening, the Rabbit saw two strange beings creep
out of the bracken. (2) they were rabbits like himself, but quite
furry. (3) Their seams didn't show at all, and they changed shape
when they moved.
(4) "Can you hop on your hind legs?" asked the furry rabbit.
(5) "I don't want to," said the little Rabbit.
(6) The furry rabbit stretched out his neck and looked.
(7) "He hasn't got any hind legs!" he called out. (8) "And he
doesn't smell right! (9) He isn't a rabbit at all! (10) He isn't
real!"
(II) "I am Real!" said the little Rabbit.
(12) "The Boy said so!"
( 13 ) Just then . . . (M. Williams, The Velveteen Rabbit)

Each grammatical unit represents a number of selections such as 'relational',


'declarative' and 'unmarked theme' from a large network of systems of choice. These
selections are realized structurally and by means of items of wording. We can illustrate
this with an informal commentary on sentences (1) and (11).
Sentence (1), One summer evening, the Rabbit saw two strange beings creep out of
the bracken, is a simple clause.
Ideationally, a clause construes experience by categorizing and configuring it as
a figure. (1) is a mental clause of perception, with the structure Time: nominal
group (one summer evening) + Senser: nominal group (the Rabbit) + Process:
verbal group (saw) + Phenomenon: clause, nonfinite (two strange beings creep
out of the bracken)'.

1.3 Metafunctonal diversification; the ideational metafunction

Interpersonally, the clause enacts a relationship between speaker and addressee


as a move in a potential exchange; its mood is declarative, with the structure
Mood (Subject (The Rabbit) A Finite (saw))\ and its polarity is positive. It is
the Mood element that embodies the arguability status of the clause: the Rabbit
saw...did he?; the Rabbit saw... didn't he?
Textually, the clause presents the ideational and interpersonal information as a
message a contribution to the text evolving in its context. It is structured as
"Theme (One summer evening)

Rheme (the Rabbit saw two strange beings

creep out of the bracken)'-, the Theme provides a local environment created for
this clause, positioning it in the unfolding text. The Theme is a marked one
(Theme = Subject would be unmarked): more specifically it is a temporal one,
which indicates that the story has shifted time frame and a new episode is about
to start. This time frame is carried forward by default in the ensuing clauses,
where the Velveteen rabbit and the real rabbits take turns at being Theme.
The clause integrates these three metaftinctional perspectives as shown in Figure 1-2.

One summer the Rabbit


evening

two strange beings


creep out of the bracken

saw

textual

Theme

Rheme

interpersonal

Adjunct

Subject

Resi-

Mood

Time

Senser

Process

Phenomenon

nom. gp.

nom. gp.

verbal gp.

clause: nonfmite

ideational

Finite/ Predicator Complement

-due

Fig. 1-2: Metafunctional integration in the structure of the clause (1)


Each metafunctional contribution constitutes one layer in the box diagram. The final
row in the diagram does not represent a layer within the function structure of the clause
but rather the sequence of classes realizing the functionally specified elements within the
configuration of functions. This sequence of classes is called a syntagm to distinguish it
from the (function) structure (see Halliday, 1966). It specifies constraints on the units
serving within the clause the units on the rank below the clause on the grammatical
rank scale. The rank scale determines the overall constituency potential in the
grammar: in English, clauses consist of groups (/ phrases), groups consist of words, and

10

1. Theoretical Preliminaries

words consist of morphemes.7 The units below the clause on the rank scale are all groups
(nominal, verbal adverbial, etc.) or phrases (prepositional phrases), or else clauses that
are shifted downwards on the rank scale to serve as if they were groups or phrases. Such
down-ranking is known as rankshift (Hafflday, 1961; 1985; Matthiessen & Halliday,
forthc.). This has the powerful effect of expanding the resources of grammar by allowing
the meaning potential of a higher-ranking unit to enrich that of a unit of lower rank.
Thus the Phenomenon of the clause above is realized by a rankshifted clause two
strange beings creep out of the bracken; we indicate the rankshifted status of a clause by
special brackets ( E l ) : One summer evening the Rabbit saw ttnvo strange beings creep
out of the bracken^. Such rankshifted clauses construe what we call macro-phenomena.
Sentence (11) is formed by two clauses linked by a relation of quoting, "I am real" and
said the little Rabbit. Let's consider the quoted clause.
Ideationally, the clause is an ascripuve relational clause: 'Carrier nominal
group (/ ) + Process: verbal group (am) + Attribute: nominal group (real )'.
That is, the Attribute 'real1 is ascribed to the Carrier T; T is construed as a
member of the class of real ones a construal that is in conflict with the view
taken by the other rabbits.
Interpersonally, the clause is declarative. The polarity has been switched from
that of the previous clause from negative to positive; and this switch is the
issue of the argument:
He

isn't

real.

am

Real.

The textual metafunction gives the option of presenting the Finite, which carries
the polarity choice, as the news of the message to be attended to; the italicized
am suggests that that is precisely what the Velveteen Rabbit does. That is, he
shifts the New element of this unit of information to a marked position, in this
case to indicate that the news is the contrast between positive am and negative
isn't. In addition, there is also a thematic choice: Subject = / serves as
unmarked Theme. Subject/ Theme = 'Velveteen Rabbit* is in fact the
combination that runs through the quoted clauses in this passage until the
Velveteen Rabbit brings in the boy instead: the boy said so.
The three metafunctional perspectives are diagrammed together in Figure 1-3.

For a discussion of analogous compositional hierarchies in material systems, see Sheldrake


(1988: Ch. 5).

11

1.3 Metaftmctional diversification; the ideational metafuncon

textual

interpersonal

am

real

Given

New

Given

Theme

Rheme

Subject

Finite/ Predicator

Mood

ideational

Complement

Residue

Camer

Process

Attribute

nom. gp.

verbal gp.

nom. gp.: adjectival

Fig. 1-3: Metafunctional integration in the structure of the clause (2)


As already hinted at earlier, the present book is concerned with just one portion of the
higher-level environment of the grammar, that having to do with the ideational
metafuncon. In other words, we are concerned with that portion of the semantics which
"controls" the ideational systems in the grammar, primarily, that of transitivity in the
clause and those of projection and expansion in the clause complex. Transitivity is the
grammar of processes: actions and events, mental processes and relations. It is that part of
grammar which constitutes a theory of "goings-on". Projection and expansion are the
fundamental relations between processes: this is the part of the grammar that
constitutes a theory of how one happening may be related to another. Thus our aim is
towards a general ideational semantics. We may call this the ideational meaning
base, or ideation base for short. It is complemented by meaning bases supporting the
other two metafunctions the interaction base (supporting the interpersonal
metafunction) and the text base (supporting the textual metafuncon). The overall
meaning base thus encompasses three complementary domains:
the ideation base: The ideational semantic resources construe our experience of
the world that is around us and inside us. The phenomena of our experience aie
construed as units of meaning that can be ranked into hierarchies and organized
into networks of semantic types. The units of meaning are structured as
configurations of functions (roles) at different ranks in the hierarchy. For
instance, figures are configurations consisting of elements a process,
participants and circumstances; these figures are differentiated into a small
number of general types figures of doing & happening, of sensing, of
saying, and of being & having.
the interaction base: The interaction base provides the resources for speaker
and listener to enact a social and intersubjective relationship, through the
assignment of discursive roles, the expression of evaluations and attitudes. The

12

1. Theoretical Preliminaries

base includes both the semantic strategies speaker and listener deploy in dialogic
exchanges and the social personae of the interactants. This second component is
a model of the interpersonal and ideational distance between speaker and listener.
the text base: The text base provides the resources that enable the speaker to
produce contextualized discourse and to guide the listener in interpreting it.
These include resources for engendering a wide variety of diverse rhetorical
structures, for differentiating among the different values and statuses of the
components of the unfolding text, and for ongoingly expanding the text so as to
create and maintain the semiotic flow.
The three bases are shown in relation to the metafunctional components of the
grammar (at the rank of clause) in Figure 1-4. Here the three bases aie shown as different
metafunctional domains within the overall meaning base, with the textual one as internal
to the meaning base and oriented towards both the ideation base and the interaction base.
We will return to the relationship between the ideation base and the text base in Chapter
9, Section 9.3 below (for a recent discussion of computational modelling of the
coordination of these three metafunctions, see Zeng, 1993).
There is no separate component of "pragmatics" within our interpretative frame. Since
it emerged as a distinct field of scholarly activity, pragmatics has by and large been
associated with two aspects of language. On the one hand, it has dealt with those aspects
of the meaning of a text which depend on specific instances particulars of the situation
and of the interactants, and inferences drawn from these. But just as, in grammatics, we
do not distinguish between the grammar of the system and the grammar of the
instance a systemic theory is a theory of both, and necessarily (therefore) of the
relationship between them so in semantics we would not want to separate the system
from its instantiation in text. In this aspect, pragmatics appears as another name for the
semantics of instances. And on the other hand, pragmatics has served as an alternative
term for the interpersonal and textual domains of semantics. Here the distinction that is
being labelled is one of metafunction, not of instantiation; but it seems undesirable to
obscure the relationship between ideational meaning on the one hand and interpersonal
and textual meaning on the other hand by locating them within different disciplines.

1.3 Metafunctional diversification; the ideational metafunction

13

Fig. 1-4: The three bases of the environment as seen from grammar
Since our approach is via the grammar, we have taken the boundaries of the grammar
as criterial, using the clause complex the highest rank of ideational organization to
define the scope of the ideational-semantic representation. This is not a necessary
constraint; but it is one that is clearly motivated in tenns of the overall design, and which
may turn out to define the optimal moment of interfacing between the ideational and the
other components. This will depend on subsequent work on the text base and the
interaction base. The constraint does not imply, however, that the scope of ideational
semantics does not extend over sequences longer than a clause complex (see Chapter 3).
1.4 The scope of the ideation base
We have said that the ideation base is a resource for construing our experience of the
world. Such construal is both paradigmatic and syntagmatic. (i) In paradigmatic constnial,
we construe a phenomenon as being of some particular type some selection from a set
of potential types. The ideation base is in fact organized as a network of inter-related
types of phenomena, (ii) In syntagmatic construal, we construe a phenomenon as having
some particular composition as consisting of parts in some structural configuration.
For example, if some phenomenon is construed as belonging to the type 'creative doing1,
it will configure as an Actor, a Process, and a Goal which is brought into existence
through the actualization of the Process, These two modes of construal are related: on the
one hand, syntagmatic organization realizes paradigmatic organization; on the other
hand, types in the network of paradigmatic organization correspond to fragments of
syntagmatic specification this is one way in which such types are differentiated.

14

1. Theoretical Preliminaries

The distinction between paradigmatic organization and syntagmatic organization opens


up one dimension of the ideation base. As we have said, syntagmatic specifications occur
in paradigmatic environments, as realizations of paradigmatic types. But the global
organization of the ideation base is paradigmatic; the paradigmatic network is ordered in
delicacy (subsumption, classification, specialization), from the least delicate (most
general) to the most delicate (most specific types).

Fig. 1-5: The overall dimensions of the ideation base


The ideation base has one other primary dimension, that of instantiation. The
ideation base is not just a repository of particular facts and other instantial meanings but
also a systemic network of meaning potential. For any given domain the ideation base
incorporates not only the known particulars of that domain but also the resources
necessary for assimilating new information. Instantiation refers to the move from the
semantic potential within the general system to instances of this potential within a
particular text (cf. Halliday, 1973, 1977, 1992).' Intermediate between these two on the
instantiation cline are patterns of instantiation that recur in particular situation types
semantic domains located within the overall meaning potential as situated variants of it
(see further Chapter 8, Section 8.1).

The two domains of information have been called TBox and Abox in the KL-ONE tradition.
We avoid these terms since they are misleading the potential is not restricted to
terminological information, nor is the instantial restricted to assertions.

1.4 The scope of the ideation base

In general, the ideation base can thus be thought of as a large semantic 'space'
organized in terms of these two basic dimensions: see Figure 1-5 above. (Note that it is
important to keep delicacy and instantiation distinct. In early work on semantic networks,
they were sometimes neutralized (cf. Woods', 1975, review). The difference is essentially
that between being a type of x (delicacy) and being a token of x (instantiation). Both may
be construed by intensive ascription: cf. Chapter 4, Section 4.2.3.2.)
We will be concerned in the first instance with the general potential; but we shall
move along the dimension of instantiation to illustrate with two domain models, weather
forecasts and recipes. The more genera] semantic potential is realized by selections in the
grammar; but as we move towards the more delicate part of the ideation base, we come to
types of phenomenon that are realized primarily by lexical means.

1.5 Guiding principles


We have mentioned two corollaries of using a functional grammar. One was that the
grammar is 'natural': that is, the forms of the grammar are non-autonomous, non-arbitrary
in their relation to meaning. The other was that the grammar is explained in functional
terms: specifically, in systemic grammar, in terms of the interaction among various
functional constructions of meaning or 'metafunctions': ideational, interpersonal, and
textual. These will be followed up in more detail in Part II.
As already pointed out, these two corollaries together define the concept of the
'ideation base1 which is what we are setting out to explore in the present book.
We can now enumerate certain guiding principles that we are following in this
endeavour. These constitute so to speak our own 'discourse dynamics': not in the sense
that our argument is explicitly presented along these directional lines (that would be
impossible, since all would demand different principles of ordering), but in the sense that
they gave direction to the thinking that has gone into our argument as we developed it.
1.5.1 Approach from grammar
In approaching the semantic environment from within grammar, as opposed to
approaching it from some postulated cognitive or conceptual plane (i.e. in constructing it
as meaning rather than as knowing), we are following the principle from which semantics
first evolved in western thought. By the time of Aristotle there had emerged the
grammatical concept of a word, and of word classes; for example, 'noun'. This concept
was born in the work of the Sophists, in their study of rhetoric, out of the dialectic
between form and function. A noun was that 'about which something is said', thus
embodying the functional concept of a syntactic (Theme-Rheme) structure; and that

16

1. Theoretical Preliminaries

'which inflects for number and case but not for gender', embodying \htformal concept
of a morphological (case and number) paradigm. The category of noun once established,
the question arises of why does a noun appear sometimes in singular sometimes in plural
number, sometimes in nominative sometimes in accusative, genitive or dative case?
These questions are answered with semantic explanations: a noun is the name of a person,
other living creature or inanimate object; a noun is in the plural if it refers to more than
one of these entities; it is in the nominative case if it refers to the 'doer1 in some kind of
action, the accusative if it refers to the 'done-to', and so on. The semantics was construed
by exegesis out of the grammar: both the general conception of meaning as a linguistic
phenomenon and the specific meanings that were constructed by words, their classes and
their variants.
Here we can see both similarities with and differences from our own present approach.
We are also constructing meaning out of the grammar. But there are two significant
differences.
(i) One is that we are not starting from the word, but from larger units of
grammatical organization: clauses and clause complexes (sentences) the
largest units, in fact, that are constructed on grammatical principles.9
(ii) The other is that we are starting not from the oven categories and markers of
the grammar, like case and case inflexions, but from the often covert,
cryptogrammatical relations that are less immediately accessible to conscious
reflection yet constitute the real foundation on which the grammar construes the
world of our experience.10
We could sum this up by saying that grammar provides the mode of entry into
semantics provided the motivating 'grammatics' is broad enough and deep enough. But to
say this already suggests that there are other, more fundamental guiding principles at
work underlying our own current practice. We started by saying that to adopt this
approach means interpreting grammar as 'natural', in the sense that its formal
organization is typically iconic rather than conventional in relation to meaning. We need
to pursue this line of thinking a little further, bringing in the arrows of time.

la systemic terms, the largest that can be exhaustively represented as system networks on the
paradigmatic axis. Cf. Chapter 3 below.

10

This formulation specifically relates to the ideational metafunction. If we were concerned


with the interaction base it would be the cryptogrammar of language as action: enacting
social and personal relationships, power structures of all kinds, institutional attitudes and
value systems and the like.

1.5 Guiding principles

17

1.5.2 Meaning constructed in grammar

In what we have said so far we could seem to be taking an 'essentialist' or


'correspondence' approach to grammar and meaning, according to which 'meaning' preexists the forms in which it is 'encoded' (cf. Lakoffs, 1988, argument against the
'objectifying* view of meaning). In such a view, the grammar is said to be natural because
it evolves to serve an already developed model of experience, a "real world" that has
previously been construed. In fact we are not taking an 'essentialist' or 'correspondence'
approach, and there will be many places throughout our discussion where such an
interpretation is clearly ruled out, as being incompatible with our own conception of
semantics (see for example the discussion of sequences in Chapter 3 below). The view we
are adopting is a constructivist one, familiar from European linguistics in the work of
Hjelmslev and Firth. According to this view, it is the grammar itself that construes
experience, that constructs for us our world of events and objects. As Hjelmslev (1943)
said, reality is unknowable; the only things that are known are our construals of it
that is, meanings. Meanings do not 'exist1 before the wordings that realize them. They are
formed out of the impact between our consciousness and its environment.
1.5.3 Semogenesis
We need, therefore, a further guiding principle in the form of some model of the processes
by which meaning, and particular meanings, are created; let us call these semogenic
processes. Since these processes take place through time, we need to identify the time
frames, of which there are (at least) three.1 '
(i) First, there is the evolution of human language (and of particular languages as
manifestations of this). Known histories represent a small fraction of the total time scale
of this evolution, perhaps 0.1%; they become relevant only where particular aspects of
this evolutionary change have taken place very recently, e.g. the evolution of scientific
discourse. This is the phylogenetic time frame.
(ii) Secondly, there is the development of the individual speaker (speaking subject).
A speaker's history may like that of the biological individual recapitulate some of
the evolutionary progression along epigenetic lines. But the individual experience is one
of growth, not evolution, and follows the typical cycle of growth, maturation and decay.
This is the ontogenetic time frame.

11

There may be more, but these are the ones that matter here.

18

1. Theoretical Preliminaries

(iii) Thirdly, there is the unfolding of the act of meaning itself: the instantial
construction of meaning in the form of a text. This is a stochastic process in which the
potential for creating meaning is continually modified in the light of what has gone
before; certain options aie restricted or disfavoured, while others are emprobabled or
opened up. We refer to this as the logogenetic time frame, using logo(s) in its original
sense of'discourse*.
These are the three major processes of semohistory, by which meanings at
continually created, transmitted, recreated, extended and changed Each one provides the
environment within which the 'next' takes place, in the order in which we have
presented them; and, conversely, each one provides the material out of which the previous
one is constructed: see Figure 1-6.
semogenic processes:

Fig. 1-6: The three semohistories related


As the upward pointing arrow suggests, the individual's (transfinite) meaning
potential is constructed out of (finite) instances of text; the (transfinite) meaning potential
of the species is constructed out of (finite) instances of individual 'meaners'. Following
the downward arrow, the system of the language (the meaning potential of the species)
provides the environment in which the individual's meaning emerges; the meaning
potential of the individual provides the environment within which the meaning of the text
emerges.
The sense in which grammar is said to construe experience will be somewhat different
in each of these three time frames.
1.5.4 Semantics in relation to grammar

In all these histories, the wordings and the meanings emerge together. The relationship is
that of the two sides of the Stoic-Saussurean sign best represented, perhaps, in the

19

1.5 Guiding principles

familiar Chinese figure yin & yang (which is in fact just that, a reprsentation of the
sign):

yang
Thus, to return to our earlier illustration of the noun: what evolved, in the history of the
system, was an entity on the content plane which had a structure as follows:
'a participant in
a process:
conscious or
non-conscious
being1

The relationship between the two sides of the sign is that of realization: thus the
meaning 'participant in a process: conscious or non-conscious being* is realized as the
wording (class of wording) 'noun*.
Suppose we now consider the semogenic processes whereby this potential expands, (i)
We may, for example, construe new participants by creating new thing/name complexes:
thus
'electronic
device for
thinking
with1

Of course, the 'thing' may have been 'there' all along but it is only newly observed and
semanticized:

20

1. Theoretical Preliminaries

Here we are extending the system by constructing new semiotic domains.


(ii) Secondly, we may expand the meaning potential by increasing the semantic
delicacy; for example

Here the semiotic domain has not been extended but rather has been brought into sharper
focus, so that further shades of meaning aie differentiated. A finer grid has been applied to
the given semantic space.
(iii) There is also a third kind of semogenic process which anses from the nature of
the sign itself. Our "sign" is not the Saussurean sign: we are not talking about the
relationship between a word and its phonological representation (between content and
expression, in Hjelmsiev's terms). The relationship is within the content plane, between
a meaning and a wording the non-arbitrary relationship between the system of
semantics and the system of lexicogrammar
This process, then, takes the form of deconstructing the two components of the sign.
How is this possible? This can happen because, once a 'pair' of this kind has come into
being, each component takes on an existence of its own. To pursue the example of the

1.5 Guiding principles

21

complex 'participant ^ noun' above: the category of 'participant becomes detached from
that of noun, so that we can have participants realized by other things than nouns, and
nouns realizing other things than participants.

We now have three meanings instead of one: participants constructed by nouns, as


hitherto, but now contrasting (a) with participants constructed by something other than
nouns, and (b) with nouns constructing something other than participants. (Of course,
neither of these two entails the other; there might be just many to one, not many to
many. To take this actual example, there are many cases of nouns realizing things other
than participants; but relatively few cases of participants being realized otherwise than by
a noun.)
A common variant of this process is that of the dissociation of associated features in
the wording. We could represent this as:

22

1. Theoretical Preliminaries

Here again one meaning has been replaced by three: we now have (say) question-j ^
interrogative x rising tone, question2 ^ interrogative x falling tone, and question3 ^
declarative x rising tone, e.g. 5 she coming ? is she earning? she's coming ?
Typically processes of this type leave their traces in the form of marking, as mated/
unmarked oppositions. The original member of the set remains the unmarked one. (One
could say 'the unmarked term in the system1; but this formulation assumes that the
offspring combine with the parent to form a system. Sometimes they do, but not
always.) In these first set of examples, the unmaiked mapping is that of participant ^
noun; when the noun realizes some other element in the figure, it is a marked variant (see
on grammatical metaphor, Chapter 6 below). Presumably many 'unmarked1 variants
originate in this way, although in most instances we no longer have the evidence which
would enable us to judge.

1.6 Codifying
Semogenic processes of the kinds just described take place in all three dimensions of
semohistory: as the system of language evolves, as children develop their language, aid
as the language of a text unfolds. Hence language embodies the potential for its own
ongoing expansion; and since the system at any moment is the repository of its own
history, we can sometimes recognize disjunctions or interstices that offer a likely context
for new meanings to appear. For example, the 'double -ing ' form of the English verb,

1.6 Codifying

23

which has recently been establishing itself (e.g. being raining, as in it seemed better to
stay at home with it being raining), could have been predicted from a knowledge of the
present state and recent history of the tense system. A change of this kind will propagate
steadily throughout the system: sometimes very rapidly, but more often in an irregular
and rather uneven flow.
Let us refer to this process as that of codifying, noting that as always it is at once
both semantic and lexicogrammatical: there is no implication that meanings are already
there and waiting to be codified. Consider a series of examples such as the following:

What happens here is that a meaning has gradually crystallized, as it were, out of the total
meaning potential of the system so that it can be deployed in codified form instead of
being constructed afresh each time. In an animal that has four legs, each of the
component elements animal four & legs is codified separately, as are the various
grammatical relations involved; but the complex is not codified as a whole. When we
come to quadrupea\ it is. Again, this codifying progression takes place in all the three
dimensions of history: Quadruped evolves later in the system, is learnt later by a child,
and typically at least appears later in the text (cf. an animal that has four legs is called a
quadruped). This process of codifying may take place at any point along the cline from
grammar to lexis, from grammaticalization at one end (cf. Hopper & Traugott, 1993)
to lexicalization at the other. Perhaps the most highly coded meanings are those
which are fully grammaticalized: that is, organized into grammatical systems, such as
tense and polarity in English. This does not mean that they must be overtly signalled in
syntax or morphology; some of them are, but others aie uncovered only through
systematic analysis (cf. Section 1.8 below on covert categories), such as the different
types of process configurations in English.
Lexicalization may take the form of the instantaneous creation of new lexicalized
meanings; like sputnik in 1958 or gazumping sometime in the seventies. But more often
it is the end point of a process of lexical compacting, as in the example of quadruped
above. Since lexicalized meanings do not form clearly defined and bounded systems in the
way that grammaticalized ones do, we might consider meanings of this kind less highly
codified, although the process of codification is the same in both cases.
Somewhere between the two extremes of grammar and lexis we may recognize the
emergence of distinct grammatical structures and lexical classes. In the course of the
history of English the meaning 'it is precipitating1 became highly codified, in that types

24

1. Theoretical Preliminaries

of precipitation came to be lexicalized as verbs (rain, hail snow, sleet, thunder, lighten)
in a unique class having no participants associated with it e.g., it's raining, where the it
functions as Subject but has no role in transitivity. (Note humorous back-formation on
model of Actor-Process: What's it doing ?Raining.)
In general the process of creating meaning involves constructing some kind of
lexicogrammatical generalization - some form of wording that is in some respect
unique. It is not possible to quantify the degree to which any semantic feature or domain
has been codified at any one moment in semohistory; but meanings that are more highly
codified are those that have been to a greater extent condensed and/or compacted, where
'compacting* is generalizing on the syntagmatic axis (e.g. animal that has four legs >
quadruped), while 'condensing* is generalizing on the paradigmatic axis (forming into a
system at some point along the scale of delicacy). The evolution of language (i.e. of
specific languages in their various registers), the learning of language by children, and the
production of language in the form of discourse constitute the historical contexts in
which meanings are continuously being created along these lines.
Of course, not every instance of the use of language involves the creation of new
meanings. The greater part of most discourse consists of wordings which have been
constructed on countless previous occasions in the language, in the individual, and
even in the course of the text.12 When we come across the sentence Rain is expected in
the northern part of the region, falling as snow over high ground we recognize probably
all of it as something that is ready to hand: not only has it occurred in the English
language many times before, but the same writer has probably written it many times
before, and many of these instances could be seen as forming pan of the same discourse
(that is, day-by-day weather reports in a sense constitute one continuous text). The
storing of meanings for repetitive use and reuse is just as important as the potential for
creating new ones. The production of discourse by an individual speaker or writer can be
seen as a dialectic between these two semiotic activities: between (i) recycling
elements, figures and sequences that that individual has used many times before, and so
for him or her are already fully codified, and (ii) constructing new ones that are
being codified for the first time (and some of which may remain codified for
future use especially with a child who is learning the system). Much recent work in

12

Note that text construction is not a matter of 'old words in new sentences'. Obviously, the
higher (more complex) the unit, the more likely it is to be being newly created: sequences
are less often repeated than figures, figures than elements. But (as we have seen) new
elements are being created all the time (again, 'new' in relation to any of the histories
above; for example, the COBUILD project in lexicological research at the University of
Birmingham reports each week on the number of words that have been identified in English
as occurring for the first time); while whole sequences are frequently stored and used over
again. Both creating new meanings and reusing old ones are, of course, equally meaningful
(even 'creative') activities; effective semiosis depends on both.

25

1.6 Codifying

AI and much of linguistics generally has tended to foreground either one or the other of
these two activities, whereas the creation of meaning involves ongoing interaction
between the two.
Thus an interpretation of semantics must account not merely for the system at some
particular point in its evolution but also for the processes by which it got there and the
changes that will shape it in the future. As far as text is concerned, the changes in
semantic styling that take place in the course of a text cannot be dismissed as simply ad
hoc devices for making the text shorter (or longer!), more interesting or whatever, they
should be seen as the operation of general semogenic principles in the specific context
which is engendering and being engendered by that text. If we are attempting to model a
learning system, in which the computer constructs its own general-purpose grammar
from large samples of text data, it will be essential both to model these processes in their
own terms and to set up the semantic and lexicogrammatical representations in terms that
arc compatible with them. We certainly do not claim to have produced a model with
which one could achieve this goal. But we have kept in mind, as a guiding principle, the
need to interpret the system in ways which allow for its possible histories how it got
the way it is, rather than (as seems to us sometimes to be the case) in ways which make
it difficult to see how any system could have arrived at that particular state.

1.7 Realization
We have retained the term 'realization* to refer to the interstratal

relationship

between the semantics and the lexicogrammar: the lexicogrammar 'realizes* the semantics,
the semantics 'is realized by* the grammar. We shall have more to say about this
relationship in general (i.e. as extending to other strata) later on. In any strata! system
(i.e. any system where there are two strata such that one is the realization of the other)
there is no temporal or causal ordering between the strata. It makes no sense to ask which
comes first or which causes which. That would be like taking an expression such as x = 2
and asking which existed first, the x or the 2, or which caused the other to come into
being (it is not like the chicken and the egg, which are temporally ordered even though
in a cycle). There is a sense in which realization is the analogue, in semiotic systems,
of cause-&-effect in physical systems; but it is a relationship among levels of meaning
and not among sequences of events, (In terms of the figures of being we shall discuss in
Chapter 4, the relationship is an intensive one, not a causal circumstantial one.)
Every scientific theory is itself a stratal-semiotic system, in which the relation among
the different levels of abstraction is one of realization. This is to be expected, since all
such theories are modelled on natural language in the first place; and, as we have seen, the
semantics of natural language is itself a theory of daily experience.

26

1. Theoretical Preliminaries

Thus when we move from the lexicogrammar into the semantics, as we are doing
here, we are not simply relabelling everything in a new terminological guise. We shall
stress the fundamental relationship between (say) clause complex in the grammar and
sequence in the semantics, precisely because the two originate as one: a theory of logical
relationships between processes. But, as we have shown, what makes such a theory (i.e.
an ideation base as the construal of experience) possible is that it is a stratal construction
that can also be deconstructed, every such occasion being a gateway to the creation of
further meanings which reconstrue in new and divergent ways. Thus a sequence is not 'the
same thing as* a clause complex; if it was, language would not be a dynamic open system
of the kind that it is. This issue will be foregrounded particularly in our discussion of
grammatical metaphor (see Chapter 6).

1.8 Grammatical evidence


Since we are taking the grammar as a point of departure for exploring the organization of
semantics, it may be helpful to say a few general words about the various ways in which
semantic categories may be reflected at the lexicogrammatical level.
In traditional grammar, only certain grammatical categories were taken into
consideration; these categories were (i) overt and (ii) word-based (cf. Section 1.5.1 above).
In particular, inflectional categories of the word such as tense, case, and number were
described and then interpreted semantically. In a functional grammar, while such
categories are not ignored, they tend to play a less significant role, appearing at the end
point of realizational chains. For instance, it is not possible to base a functional
interpretation of number in English simply on the presence or absence of 'plural' as a
nominal suffix (as in grammar+s)\ the category of number is rather more complex,
involving two complementary systems (see Halliday, 1985: 161-2). Similarly, the
general properties of the construal of time embodied in the English tense system are not
revealed by only looking at the overt suffixal past tense marte- (as in laugh+ed); again
the scope of the semantics of tense in English is far greater than this oven word category
would suggest (see Halliday, 1985: 182-4; Matthiessen, 1996). In general our move into
semantics from grammar differs from the traditional one along the following lines.
(i) We consider not only overt categories but also coven ones. The understanding of
coven categories in the grammar is due to Whorf (1956: 88 ff), who made the distinction
between overt categories or phenotypes and covert categories or cryptotypes; he is
worth quoting at some length:
An overt category is a category having a formal mark which is present (with only
infrequent exceptions) in every sentence containing a member of the category. The
mark need not be part of the same word to which the category may be said to be
attached in a paradigmatic sense; i.e. it need not be a suffix, prefix, vowel change, or

1.8 Grammatical evidence

27

other 'inflection', but may be a detached word or a certain patterning of the whole
sentence. ...
A covert category is marked, whether morphemically or by sentence pattern, only in
certain types of sentence and not in every sentence in which a word or element
belonging to the category occurs. The class membership of the word is not apparent
until there is a question of using it or referring to it in one of these special types of
sentence, and then we find that this word belongs to a class requiring some son of
distinctive treatment, which may even be the negative treatment of excluding that
type of sentence. This distinctive treatment we may call the reactance of the
category. ... A covert category may also be termed a cryptotype, a name which
calls attention to the rather hidden, cryptic nature of such word-groups, especially
when they are not strongly contrasted in idea, nor marked by frequently occurring
reactances such as pronouns. They easily escape notice and may be hard to define, and
yet may have profound influence on linguistic behaviour. ... Names of countries and
cities in English form a cryptotype with the reactance that they are not referred to b y
personal pronouns as objects of the prepositions 'in, at, to, from'. We can say I live
in Boston' but not That's Boston I live in it'.

There are many examples of cryptotypes in this sense, both as classes and as systems
(i.e., cryptoclasses and cryptosystems), in our ideational semantics. For example:
process types: doing & happening/ sensing/ saying/ being & having
transitivity model: ergative/ transitive
projections: locutions/ ideas
expansions: elaboration/ extension/ enhancement
number: plural/ non-plural; singular/ non-singular
The concept of reactance is particularly significant for our purposes where it involves a
relationship between an ideational category and features of other metafunctions,
interpersonal or textual. For instance, the interpersonal grammar provides for participants,
within the ideational dimension of the clause, to function as Subjects; but this potential
is not in general open to circumstances, and this is a principal reason for distinguishing
these two classes within the ideational metafonction. Among reactances from the
interpersonal and textual components of the grammar, we could mention the following:
interpersonal:
can/ cannot serve as Subject
can/ cannot serve as 'focus* of alternative question
can/ cannot serve as Wh element

28

1. Theoretical Preliminaries

textual:
can/ cannot serve as Theme
can/ cannot serve as 'focus' of theme predication (t is... that...)
can/ cannot be presumed by substitution/ ellipsis
(ii) The gateway to semantics is the clause rather than the word. Consequently,
grammatical categories will typically be interpreted 'from above', within their context in
the clause or the group, rather than 'from below' within their context in the word. This
has rather far-reaching consequences for the understanding of the semantic systems realized
by the grammar. Systems that are approached 'from above' in this way include:
projection clause complex: traditionally a form of 'subordination' within clause;
reinterpreted as distinction between bypotaxis in clause complex vs. rankshift in
clause, laying the foundation for a semantic distinction between reports and
facts.
transitivity clause: traditionally a word category, transitive = verb taking
object/ intransitive = verb not taking object; reinterpreted as (i) process types
(material/ mental/ verbal/ relational) and (ii) an ergative system (middle/
effective) in the clause.
tense group: traditionally a mixture, because the model was taken over from
Latin with richer word-rank realizations than English, but more recently in this
century often a word category, past/ non-past; reinterpreted (relative to this) as
(i) past/ present/ future and (ii) recursive, with secondary tense.
Taking the clause as starting point facilitates the exploration of cryptotypes: the chain of
realization often starts cryptotypically in the clause, whereas the final stages of realization
at word and morpheme rank are more oven although, as noted in connection with
tense and number, the oven marking is seldom the only factor involved.
A functional grammatics thus allows us to approach semantics from a deeper and
more wide-angled perspective. To this general property, systemic functional grammar
adds another characteristic its paradigmatic orientation (cf. Matthiessen, 1987, for a
discussion of how the development of the semantic environment of a generation grammar
can be based on paradigmatic choice). For instance, while more formally oriented
accounts may approach transitivity patterns essentially in terms of sequences of
grammatical classes such as 'nominal group + verb (+ nominal group)' and speak of
classes of verb followed by one nominal group ('mono-transitive') or two nominal groups
Odi-transitive'), a systemic grammar interprets such sequences in terms of systems of
distinct and contrasting process types (see Halliday, 1985: Ch. 5; Davidse, 1991; Martin,
1996 a, b).

1.8 Grammatical evidence

29

Section 2.11.5 in Chapter 2 below provides a summary of grammatical evidence for


semantic types.

1.9 Ways of representing semantic organization


We have dealt with various theoretical aspects of the nature of meaning in the ideation
base and we now turn to the final issue of our introduction, the question of how to
represent our theoretical account of the organization of meaning in some representational
system. In AI and computational linguistics, this is usually formulated as knowledge
representation; from our complementary point of view, the task is one of meaning
representation. For a systemic functional review of knowledge representation, see also
Steiner (1991: Section 2.6.5).
Representation is, of course, an inherently semiotic phenomenon, and we can gain
some insight into it by considering it first within language itself, before turning to
representation in our own 'metalanguage' (see the beginning of Section 1.9.1 below).
Meaning is realized stratally in two cycles (see Figure 1-1 above). We have noted that the
first cycle, the realization in lexicogrammar, is natural, in the sense of being nonarbitrary: for example, the grammatical constituency structure of a clause provides a
natural representation of the semantic configuration of a process, participants and
circumstances. By attending to grammatical representations, we can thus learn a good deal
about the more abstract organization of meaning at the higher stratum of semantics. We
can learn about the different modes of meaning logical, experiential, interpersonal, and
textual by exploring their different modes of representation in the grammar
chaining, constituency, prosody, and wave (see Halliday, 1979a; Matthiessen & Halliday,
forthcoming). Grammar is thus a hybrid system for representing meaning in the sense of
embodying different modes of representation; but it is this that allows it to maintain a
natural relationship with respect to semantics, with each mode of representation realizing
a different mode of meaning.
Grammatical representations are in turn represented in linguistic expressions
prototypically, in sounding. Here the relationship is more complex than it is between
semantics and grammar, in that it is both natural and conventional. In the interpersonal
and textual domains of content, it is often natural: thus interpersonal content tends to be
reprsentai prosodically by movements or variant levels in pitch, and textual content
tends to be represented by prominence achieved phonologically (e.g. by the major pitch
movement in an intonation contour) or sequentially (e.g. by using distance from initial
position in the clause as a scale of prominence). In the ideational domain, the
representation is usually conventional; but, even here there is a relationship of analogy,
where we find in the sounding modes of organization similar to those of wording (and
therefore of meaning). Systemically, we find that the system construes a phonetic space
notably the vowel space; and that this provides a model for semantic space.

30

1. Theoretical Preliminaries

Structurally, we find that sound is structured both as chains of segments (e.g. rhythmic
units interpreted as syllable complexes) and as configurations of segmentai constituents
(e.g. syllables interpreted as configurations of phonemes).
The phonological representations are still abstract; they have to be manifested in
bodily movements in the ongoing movement of the parameters of the articulatory
system. The sound system thus categorizes bodily processes; and in this respect, it is
similar to the semantic system: both are ways of construing human experience. Meaning
is thus represented by modes of organization that are similar to its own. This parallelism
is even more foregrounded where the modality of expression is spatial, as in the Sign
Languages of deaf communities (see Johnston, 1989; 1992 on AUSLAN). Here the
domain of expression is a 4-dimensional signing space-time in a field of perception shared
by signer and addressee (though clearly perceived from different angles). The spatial
orientation and the shared perception increase the potential for iconicity in the expression;
Johnston (1992) points out: "Despite an oral-aural language being suited to iconically
encode sounds, the fact that our experience as a whole is visual, temporal and spatial
means that a language which has itself visual and temporal resources for representation
has greater means than an auditory one to map onto itself those very visual and spatial
qualities of the world it wishes to represent" Johnston's insight into the power of
representation embodied in Sign is central also to the general challenge of representation
in metalanguage.

1.9.1 Stratification of metalanguage


We noted above that the semantics/ lexicogrammar of natural language is itself a
'realization* (an abstract construction) of daily experience. Likewise, the system we use to
explore the semantics/ lexicogrammar our theory of semantics and our grammatics
is a 'realization* of that pan of daily experience that is constituted by semantics and
lexicogrammar; that is, it is an abstract construction of language. This system is itself a
semiotic one a metalanguage; in Firth's more everyday terms, it is language turned
back on itself (cf. Matthiessen & Nesbitt, 1996). So whereas a language is (from an
ideational point of view) a resource for construing our experience of the world, a
metalanguage is a resource for construing our experience of language.
Metalanguage has the same basic properties as any semiotic system. This means that
it is stratified. It construes language in abstract theoretical terms; but this construal is
in turn realized as some form of representation either language itself, in discursive
constructions of theory, or some form of designed semiotic (system networks,
constituency rules, conceptual networks, logical formulae, and so on). In the environment
of computational work, this level is in turn realized in some form of implementation
(stated in a programming language1 such as LISP, Prolog or C). We could summarize as
follows (see Figure 1-7):

1.9 Ways of representing semantic organization

31

metalanguage construes language at various strata (theoretical construal,


representation, and implementation)
metalanguage construes language in terms of resources (the circles) and
processes using the resources (the arrows), where processes include description,
deduction and compilation.
the relationship between theoretical construal and representation should
preferably be a natural one (note the 'should be'!), while the relationship
between representation and implementation is likely to be more arbitrary (for
instance, it should not matter whether LISP, Prolog or C is used).
the range of phenomena accounted for in metalanguage tends to decrease at lower
levels. (A primary goal of research is always to expand it.)
If we recognize stratification as an aspect of the design of metalanguage, we are in a
position to locate aspects of the overall construal of meaning at the appropriate stratum.
Early work on knowledge representation in the 1960s and 1970s tended not to make this
stratal differentiation; it was under-differeotiated from a meta-theoretical point of view.
Responding to this situation, Brachman (1979) proposed a number of levels to separate
out

different aspects

of the overall statement of knowledge. He identifies

implementational, logical, epistemolgica!, conceptual, and linguistic levels. While these


are not all metalinguistic strata in our sense, it is still clear that he is concerned with the
need to son out different semiotic domains of generalization: the different 'levels' ate
subject to different kinds of tasks and constraints. Brachman's comments on the
implementational level are worth noting:
In implementation level networks, links are merely pointers, and nodes are simply
destinations for links. These primitives make no important substantive claims about
the structure of knowledge, since this level takes a network to be only a data structure
out of which to build logical forms. While a useful data-organizing technique, this
kind of network gives us no more hint about how to represent (factor) knowledge
than do list structures.

32

1. Theoretical Preliminaries

Fig. 1-7: Hie relationship between language and metalanguage

Thus any account of the ideation base has to be metalinguistically stratified It has to
be constructed as a theoretical model out of the resources the theory provides and
according to the constraints imposed by these resources. From a systemic-functional
point of view, this means that the ideation base is construed as a multidimensional,
elastic semantic space. This space is organized as a meaning potential, with an extensive
system of semantic alternatives; these alternatives are ordered in delicacy. Each set of
alternatives is a cline in semantic space rather than a set of discrete categories, and any
alternative may be constituted structurally as a configuration of semantic roles. The

1.9 Ways of representing semantic organization

33

meaning potential is thus differentiated axially into (i) systems of options in


meaning and (ii) structural configurations of roles by which these options ate
constituted.
The meaning potential itself is one pole on the dimension of instantiation: it is
instantiated in the unfolding of text, with patterns of typical instantiation (specific
domains of meaning) lying somewhere in between tbe potential and the instance. At the
same time, this overall ideation base can be expanded by various semogenic strategies,
among which we are foregrounding that of grammatical metaphor.

Fig. 1-8: Stratification in the account of the ideation base


The theoretical model makes considerable demands on the representational system or
systems in which it is realized Such a system must be able to handle the axial
differentiation of the meaning base so that it represents not only structural configurations
but also variation in delicacy, the indeterminacy between options in meaning, the move
in instantiation from potential to instance, and other theoretical specifications. There are a

34

1. Theoretical Preliminaries

considerable number of alt^Tiative representational systems, some of which we will refer


to below. See Figure 1-8 above for a diagrammatic summary of the relationship between
theory and representation in the constnial of the ideation base.
As we have already noted, the representational semiotic may also be language itself
the theory may be represented discursively in the registers) of linguistics. In this case,
the relationship between theory and language is similar to the relationship between
'theme' and 'language' in Hasan's (1985/9) theory of verbal an a relationship where
theory might be construed as a connotative semiotic (in Hjelmslev's, 1943, conception: a
semiotic system whose expression plane is a semiotic system) along lines similar to the
way Martin (1992) construes ideology. We shall return to the representation of theory in
language in another context towards the end of the book (Chapter 14).
Finally, the representational semiotic may also be a diagrammatic one e.g.
network graphs, tree diagrams, our circle diagrams. With such graphic representation, it is
important to ask (i) if the information represented graphically can be restated in some
other form of representation and (ii) if its realizational relationship with respect to theory
is clear. There is a certain danger that the graphic representation is simply assumed to be
natural and transparent even though it depends as much on semiotic conventions as any
other system (cf. Kress & van Leeuwen, 1990, 1996).13 Our own experience with circle
diagrams is that, unless the representational convention has been made explicit, leaders
favour certain types of meaning: the circles are interpreted in terms of 'extension* rather
than in terms of 'elaboration' (see Chapter 2, Section 2.2) which is to say, they arc
lead in constituency terms rather than in realizational terms. In any case, diagrams will
only serve us as Visualizations' as long as they construe a metaphor of abstract space at
the theoretical stratum. (We shall discuss the relationship between metaphorical space in
theory and its representation in diagrammatic form in Chapter 6, Section 6.7 below.)
So far our observations about the meaning base have mostly been at an abstract
theoretical level, though with some graphic representations. Our discussion of semantic
organization in the ideation base will remain at this level in fairly informal terms. The
best metalanguage for talking about language is still language itself; any designed system
such as logic or semantic nets whose mathematical properties aie well-known can only
offer the kind of precision they are intended to have at the price of restricted coverage. In
other words, there is a trade-off between the extent and kind of observations we can make
and the formality of the semiotic system used for the task. Moreover we consider it
important to explore the domain of meaning substantially and in some detail before

13

It has been noted in NLP that in the early days of semantic networks there were too many
ways of using 'links' and 'nodes' (cf. Woods, 1975; also Brachman's, 1978, 1979, emphasis
of the need to define networks algebraically).

1.9 Ways of representing semantic organization

35

judging what forms of representation are most suitable. And for certain aspects of the task
it should be possible to use more than one form of representation. (The Penman Upper
Model [see Bateman et al, 1990] derived from an early version of our ideation base was
represented first in NIKL and then in LOOM, both frame-based inheritance networks; and
ODonnell, 1994, has now re-expressed this model in systemic form.)
At this point we shall briefly outline the major representational options, to wit:

(i) Logic (Section 1.9.2);


(ii) Discrimination networks (Section 1.9.3);
(iii) Componenda! analysis (Section 1.9.4);
(iv) Semantic nets, in particular frame-based inheritance networks (Section 1.9.5);
(v) Systemic representation (Section 1,9.6).
It is possible to use some kind of 'hybrid1 representation; combinations of logic and
semantic nets have been used in computational linguistics and AI, where they are seen to
complement each other (e.g. KRYPTON [Brachman, Pikes & Levesque, 1983], KL-ONE
[Brachman & Schmolze, 1985], NIKL, LOOM [Brill, 1990]). In this connection, natural
language can itself be thought of as a 'hybrid' system of representation; we have already
noted its hybrid nature in the modes of realization of the different metafunctions. The list
above could also be extended: thus it would be relevant to consider feature structures,
typed feature structures, and fuzzy logic. The 'typed feature structure' system, which is
widely used in NLP, is not so much an alternative to the systemic representation, but
rather an additional representational level that could be inserted between the theoretically
informed systemic representation and the lower level of program code (cf. the work on
representing systemic functional information in typed feature structures reported in
Bateman & Momma, 1991). Fuzzy logic foregrounds the ideational indeterminacy that we
shall be emphasizing throughout our account (cf. Zadeh, 1987; Sugeno, 1993).
The choice of representation will always depend on the task at hand. A particular task
might foreground inferencing based on collections of 'instantial facts' in a data base; or it
might foreground taxonomic organization and inferencing based on hyponymy
(classification), and so on. Similarly, the need to analyse ideational meaning in text for
educational purposes and the need to process ideational meaning in an NLP system will
share certain demands on the representational system; but they will also make demands
that point in fairly different directions, e.g. towards perspicuity on the one hand and
towards automatic manipulation on the other.
1.9.2

Logic

Logic is attractive because its properties are well understood in terms of both fonnationrules ('syntax1) and interpretation ('semantics') and it can be used to support inferencing.

36

1. Theoretical Preliminaries

To take a trivial example, if one proposition implies another and the first is asserted, the
second can be deduced
(p->q)&p 1= q
e.g. 'it's raining' -> 'the market is closed' & 'it's raining' 1= 'the market is closed'
However, there are a number of serious problems that have made it very hard to use
logic as the only form of representation of the meaning base.
(1) We need a form of representation that can handle not only 'propositions'
(propositional logic), their internal organization (predicate logic), and their
temporalization and modalization (temporal and modal logic), but also paradigmatic
organization, including taxonomic relations. Standard logic is not designed for the
representation of taxonomies (cf. Samlowski, 1976). One type of logic, sorted logic,
has been developed to represent taxonomic inheritance, with variables assigned to sorts or
types; but it does not constitute a full-fledged representation of paradigmatic order.
(2) Logic operates with a very simple ontology: in standard logic, propositions (p,
q, r etc.), truth functions (&, ->, ~ etc.), individual constants (a, b, c, etc.), variables (x,
y, z, etc.), quantifiers, (V, 3, 3!, etc.) and predicates (F, G, etc.). These ontological
categories are insufficient for the task of representing semantic organization (cf.
Jackendoff, 1983: Ch. 3 and Section 4.1); for example, the truth functions cover a very
small range of the spectrum of logical relations in natural language (cf. Chapter 3,
Section 3.1 below), predicates fail to distinguish among things, qualities and processes,
and it is not clear how circumstances can be accommodated.14 It is possible to increase the
power of logic e.g. by adding temporal, modal, and intensional components, or by
allowing for nonmonotonic reasoning to handle unmarked cases and exceptions; but this
is a slow process since the formal properties that are valued have to be retained. It has
often been noted, of course, that logic falls far short of being able to represent linguistic
theories of meaning; Simon Dik's (e.g. 1986, 1987) response has been to explore a
functional logic that consists of a range of different sublogics reflecting the semantic
diversity of language as a knowledge representation system.

" The problem with the ontology of logic arises primarily when logic is used as the first level
of representation and semantic types are represented directly by logical ones, as was done in
generative semantics, for example. However, a more indirect relationship can be
constructed whereby there is no one-to-one correspondence between semantic types and
logical ones. Thus using Prolog as the coding system at the level of programming does not
impose its ontology on the ideation base. Here a 'line of arbitrariness* relative to the higher
levels of organization is maintained. Compare Brachman's (1979) cautionary note: "A
network implemented as some level should be neutral toward the level above it. For
example, logical nets are "epistemologically neutral", in that they do not force any choice
of epistemological primitives on the language user. Making "concepts" in logical nets,
then, is a mixing of levels."

1.9 Ways of representing semantic organization

37

(3) From the point of view of language, logic misinterprets and displaces a number of
semantic categories. We have referred to the lack of ontological distinctions, which also
has undesirable consequences for semantic structure (see Jackendoff, 1983: Section 4.1),
but in addition to this, logic merges determination and quantification so that their
distinctive contribution to texture is obscured. Quantifier scope has the effect of
scrambling semantic structure and so distorting its relationship to the grammar (cf.
Keenan & Faltz, 1985). Moreover, the interpersonal resource of polarity (positive/
negative) is interpreted in terms of negation as a truth function. While this may be
appropriate for statements (as opposed to other speech functions) in certain registers (viz.
those concerned with deductive reasoning, where logic was first designed and is still used
today), it is certainly not appropriate for language in general (see e.g. Givn, 1979: Ch.
3).

This last point about negation illustrates the general relationship between logic aid
language (a point to be taken up again in Chapter 3, Section 3.1). As a semiotic for
reasoning, logic is registerially quite constrained: it operates under a number of contextual
assumptions about valid reasoning. In contrast, language is a very general system; it can
support many more types of reasoning in a wide range of different contexts. There is
obviously a trade-off here between degree of axiomatization and generality. Logic has its
own contexts of application; problems only set in when we try to use this designed aid
highly constrained semiotic to interpret an evolved and much less constrained semiotic
system such as language. For a highly pertinent discussion of logic and natural language,
see Ellis (1993).
To conclude the discussion of logic, we can try to identify the areas of language that
have been the source of designed logical systems: see Table 1(1). Modem knowledge
representations such as NIKL and LOOM tend to be hybrid systems just like language
itself: they incorporate both logic and inheritance networks. As already noted above, these
different subsystems are all integrated in natural language; as a knowledge representation
system, natural language can be characterized as being 'hybrid' in the sense of being
diversified into different modes of representation for different modes of meaning
(knowledge) and the same holds true of designed systems from the 1980s onwards.
1.9.3 Discrimination

networks

In AI, discrimination networks have often been used to encode taxonomic distinctions,
as in the example of discriminations in the area of sensing shown in Figure 1-9. The
nature of the discriminations may vary from one node to another; but each discrimination
network is a strict taxonomy.

1, Theoretical Preliminaries

38

Table 1(1): Lexicogrammatical systems giving rise to logical systems


grammatical end
expansion
laws, rules of inferencing
the logical
connectives of prepositional logic (&, /,

logical

lexical end

);

experiential

interpersonal

textual

projection
intensional logic
propositional attitudes
transitivity
predication
[i] the predicate-argument structures of predicate
logic ((argl, arg2 PREDICATE));
[ii] the frames of knowledge representation
tense
temporal logic

lexical taxonomy
taxonomic
organization of
predicates
[i]
taxonomies in early
artificial languages; [ii]
inheritance networks in
modern AI/NLP

modality
judgement of truth value
[i] modal operators
in modal logic (Np, Mp ) [cf. also projection
above];
[ii] partitions within networks
reference
constants vs. variables, quantification over
quantifiers in predicate logic ( V B
variables
3! i).

Fig. 1-9: Example of discrimination network


Strict taxonomies of this kind have, of course, a long history: it was this type of
organization that Roget (1852) relied on in compiling his Thesaurus. One of its
limitations is that words in natural language do not have a unique taxonomic location;
Roget overcame this limitation by allowing any word to occur in any number of places.
From the present point of view, strict taxonomies display two major drawbacks.
(1) They are one-dimensional; and while one-dimensional organization may suffice
for certain domains, in the overall semantic system multi-dimensional
organization is the norm.

1.9 Ways of representing semantic organization

39

(2) TTiey only allow us to represent paradigmatic organization; but we need to be


able to represent syntagmatic organization as well. (In a way, this is the reverse
of the problem presented by logic; and as we shall see presently, hybrid forms
of representation have been developed to combine taxonomy and logic.)
1.9.4 Componential analysis
Discrimination networks do not, then, provide the answer to our need. The problem of
multi-dimensional organization in phonological systems was solved by the
introduction of oppositions that can intersect (such as front/ back, open/ closed, and
rounded/ unrounded for vowels); in Tnibetzkoy's (1939) theory, the oppositions were
treated as classificatory dimensions and the values as 'sound properties'. In the late 1940s,
however, Roman Jakobson (1949) reinterpreted these properties as components or
distinctive features of the phoneme (see e.g. Fischer-J0rgensen, 1975: 144-7); this was in
effect a reification of the earlier sound properties. Phonemes were said to consist

of

components Oust as longer phonological sequences consisted of phonemes) instead of


being said to realize terms in phonological systemic oppositions. In other words, a
paradigmatic abstraction (sound property in an opposition) was given a syntagmatic
status (component of a phoneme). Jakobson's notion of component or distinctive feature
was then taken over into generative phonology.
In the 1950s, componential analysis was developed by Goodenough (1956),
Lounsbury (1956) and others as a model of semantic organization, particularly in the
study of folk taxonomies. In the following decade, Katz and Fodor (1963) used semantic
markers as components of meaning in dictionary definitions within a generative grammar.
In componential analysis semantic oppositions are treated in the same way as
phonological ones. So just as rounding in a vowel system is characterized by means of
two components, grounded/ -rounded, and a vowel such as /i/ is said to consist of +high,
+ front & -rounded, so a semantic dimension such as sex is interpreted as +male/ -male
and an item such as 'girl' is said to consist of -male, -adult & +human. This approach
provides for multidimensional matrices for example:

+ adult
-adult

-f male
- human

man
boy

- male

woman
Kirl

+ human

Components may be organized taxonomically; for instance, iadult, tmale and ihurnan
all presuppose + animate. Componential analysis has been applied to various domains
such as kinship and cuisine.

40

1. Theoretical Preliminaries

A comprehensive critique of componential analysis has been provided by Wierzbicka


(1975). Many of the problems with componential analysis arise from misinterpreting the
'components' as if they were constituents of some structure, rather than being
paradigms of abstract features; or, if they are recognized as paradigmatic, from failing to
organize them into networks of systems (cf. Leech, 1974). Without this step, a
componential analysis remains "flat", in that it fails to construe any kind of
multidimensional semantic space (cf. Chapter 2, Section 2.11.1 below).
1.9.5 Frame-based inheritance networks
In AI and computational linguistics, paradigmatic and syntagmatic organization aie
combined in current frame-based inheritance networks. The ancestor of these is the
semantic net developed originally by R. Quillian in the 1960s. Quillian's model was a
linguistic one the dictionary. He constructed a network of nodes and relations; nodes
represented word senses and relations were drawn from dictionary definitions, including
'is-a' (hyponymy or instantiation) and 'has-a' (meronymy). Apart from the greater
definitional rigour achieved in subsequent networks, two later developments ace
particularly significant for our purposes:
(i) nodes were given organization in the form of frames configurations of
roles with specifications of possible fillers of these roles (Value restrictions');
(ii) the 'is-a1 relation was given special status in the network, defined as a
subsumption relation over nodes.
Together these two are essential aspects of a frame-based inheritance network
such as KL-ONE, NIKL and LOOM (see e.g. papers in Brachman & Levesque, 1985, and
Sowa, 1991). Concept frames at a given point in the subsumption hierarchy inherit any
role information associated with concepts higher up the hierarchy. For instance, Figure 110 shows a fragment of the subsumption hierarchy of figures and elements. Certain
concepts have roles associated with them doing/ Actor; directed doing/ Goal; and spear/
Means. Since 'directed doing* is subsumed under (classified under) 'doing', it inherits the
Actor role; and since 'spear* is subsumed under both 'directed doing' and 'doing', it inherits
both the Actor and Goal roles. In addition, the possible class of filler is specified for each
role as a value restriction. These are shown as pointers from the roles to other concepts in
the subsumption hierarchy, in this case pointers to types of element.
Although the example in Figure 1-10 is a strict taxonomy, that is not a restriction
imposed on networks used in e.g. NIKL and LOOM. Any concept may be subsumed by
more than one other concept, co-inheriting their properties. Further, simultaneous
distinctions are allowed; that is, it is easy to represent cross-classification. When two or

1.9 Ways of representing semantic organization

41

more concepts specify another concept, they may constitute a disjoint covering, which
means that their disjunction is exclusive.

Fig. 1-10: Concept frames in subsumption hierarchy, supporting inheritance


1.9.6 Systemic representation
Representations within systemic theory have been designed to realize theoretical
abstractions such as the ones shown in Figure 1-8 above. But the design task has not
been completed; representations are still being explored so as to expand their coverage of
the theoretical resources. In general, explicit systemic representations lag behind systemic
theory (cf. Matthiessen, 1988a), particularly in certain areas such as that of grammatical
metaphor; but this is for a good reason. While the relationship between theory and
representation is a dialogic one, the stratum of theory is the domain where demands on
representation first develop and have to be sorted out relative to the overall theoretical
system.
Systemic representations include (i) system networks and (ii) realization statements,
(i) A system network is an acyclic directed graph, consisting of systems partially
ordered in delicacy (see e.g. Fawcett, 1984; Henrici, 1966; on the use of system networks

42

1. Theoretical Preliminaries

in semantics, see Hafflday, 1973, 1984b; Hasan, 1989, 1996: Ch. 5); on formal and
computational issues, cf. also Bateman & Momma, 1991; Patten & Ritchie, 1987;
Kasper, 1988; Mellish, 1988; Brew (1991); Henschel, 1994; Teich, 1995). Each system
constitutes a choice (alternation, opposition) between two or more terms. These terms are
represented by features, and a system as a whole is a Boolean combination of features:
(1) It has an entry condition, the condition under which the systemic choice is
available. The entry condition may be a single feature or a complex of features,
conjunct and/ or disjunct. These features must serve as terms in other systems.15
(2) It has a set of terms, the options that are available given the entry condition.
The terms are represented by features, which are related by exclusive disjunction.
Collectively, a set of related systems form a system network (since features in the
entry conditions to systems must be terms in other systems). As an example, consider
the systemic representation of the componential analysis given in Section 1.9.4 above:
see Figure 1-11. (This is not a typical example of how system networks are deployed.)
The features of componential analysis can be interpreted as being related through
extension they are the 'atoms' of which meanings consist. In contrast, the systemic
interpretation relates them through elaboration word meanings lie at the intersections
of systemic values in an elastic multidimensional space.
Systems can be implemented or re-represented in various ways; apart from the
implementation in the Penman system, we can mention Patten's (1988) implementation
of systems as production rules (where one system will be implemented as two or more
production rules), Kasper's (e.g. 1987) re-representation in terms of feature structures in
Kay's (e.g. 1979) Functional Unification Grammar, Zeng's (1996) re-representation in
terms of the frame-based inheritance system LOOM, and Bateman & Momma's (1991) rerepresentation in terms of Typed Feature Structures.
If we use systemic representation to encode information in the ideation base, we
factor out the logic of subsumption and alternatives from the syntagmatic realization:
this logic is represented in the system network. If each feature is interpreted as a type,
this gives us a lattice of types. In European structuralist terms, the network is a
representation of the paradigmatic organization of a linguistic system. However, it is also

19

The exception is a special type of system, the logical recursive system. (The logical
system and the kind of structure realizing its selection are the appropriate form of
representation for sequences in the ideation base.) Non-logical systems are purely
declarative terms such as "choice" and "entry condition" do not imply a procedural
interpretation; they merely indicate that the system network can be traversed. For the
interpretation of logical recursive systems, see Bateman (1989). Recently, Zeng (1996) has
developed a new kind of system, a multiple-view system, for representing different but
comparable systemic organization across languages or semiotic systems.

1.9 Ways of representing semantic organization

43

necessary to show how feature choices aie realized structurally; i.e., what arc the
structural properties of the types. This is the province of the realization statement.

Fig. 1-11: Components re-interpreted as systemic terms


(ii) A realization statement is a minimal specification of a piece of structure or
configuration of roles presented in a paradigmatic context; it is always associated with a
particular systemic feature. For instance, the realization statement '+ Senser' occurs in the
context of 'sensing1 in the system of figures; it is the syntagmatic realization of that
feature: a figure of sensing is a configuration of roles one of which is Senser.
The general form of a realization statement is 'realization operator + one or more
realization operands', as in 'insert Senser' and 'confate Medium and Senser1. The operators
used to specify functional structuring are insert, conflate, and preselect. The first operand
is always a semantic function (role); additional operands may be functions or features.
The realization statements we will make use of in our representation of ideation base
information are as follows:16
(1) Presence of functions in the structure: the presence of a function in a
function structure is specified by inserting the function into the structure; the
operation of insertion is symbolized by '+'; e.g. +Actor, +Senser, etc.
(2) Conflation of one function with another: one function from one perspective
is conflated with a function from another perspective they ate identified with
one another. Conflation is symbolized by V; for example, Medium/Senser

The same set is used at the lexicogrammatical stratum, but two additional types are also
needed: order and expand. See Matthiessen & Bateman (1991); Matthtessen & Halliday
(forthc.).

44

1. Theoretical Preliminaries

means that Medium (ergative perspective) and Senser (transitive perspective)


apply to the same element of a figure.
(3) Restriction on the type of phenomenon that can serve a particular function:
this is stated by preselecting one or more features from the unit serving that
function; preselection is symbolized by , e.g. Senser: conscious being, which
means that the participant serving as Senser is restricted to the type 'conscious
being*. (Since 'conscious being' is more delicate than 'participant', the fact that a
participant serves as Senser can be inferred from the statement 'Senser
conscious being*. In general, it is only necessary to state the most delicate or
specific restriction along any subsumption path in the system network.)
Figure 1-12 shows an example of a systemic representation, with the graphic
conventions for the system network. Preselection is shown by means of a pointer leading
from the function (role) being restricted to the feature that it is restricted to.

Fig. 1*12: Example of systemic representation, graphic mode


The example is basically the same as we used to illustrate the frame-based inheritance
network in Figure 1-10 above. As the two examples illustrate, there is a basic similarity
between frame-based inheritance netwoiks and system networks, in their common concern
with types and subtypes. These types are related by Boolean operators: a given type may

45

1.9 Ways of representing semantic organization

be a subtype of a single type, a conjunction of other types or a disjunction of other types.


Types are distinguished in terms of structural properties: each type will typically have
some structural consequence, in specifying some aspect of a configuration of roles
(functions, slots). Each role may in turn be restricted as to what type can serve that role
(value restriction, preselection). Table 1(2) below summarizes how this general
organization is manifested (1) in frame-based inheritance networks used for representing
knowledge' and (2) in system networks used for representing lexical and grammatical
information.
Table 1(2): Frame-based inheritance networks and systemic representation

basic notation
nodes
network logic
relation
between node
and structure
specification
unit:
structure
role restriction

frame-based
inheritance network
frame-based inheritance
network
concept frames (with
roles)
Boolean classes
identical (with a concept
frame as node)
concept frame:
configuration of roles
value restriction of role

fillers

systemic representation
system network
features (with associated
realization statements)
Boolean systems (with
input & output features)
feature realized by
specifications of
contributions to structure
semantic unit
configuration of functions
preselection of functions

The example in Figure 1-12 is a representation of a fragment of the ideational


potential. However, given the summary of the theoretical account of the ideation base at
the beginning of this section (cf. Figure 1-8), there are additional representational
challenges. We shall mention four such challenges briefly here.
(i) The first representational challenge is the need to handle the dimension of
instantiation. As a process, instantiation can be represented as involving traversal of
the system network and activation of realization statements (cf. Matthiessen & Bateman,
1991: Section 6.5, on the grammatical generation algorithm). The instance is thus a set
of features (semantic types) selected, with associated realizational specifications an
instantial pattern over the semantic potential. However, instantiation also defines a scale
between the potential and the instance, with intermediate patterns of instantiation. We
will introduce these later as register-specific domain models (see Chapter 8). Such
patterns of instantiation will need to be represented: perhaps as systemic probabilities, or
as domain-partitions within the overall ideation base (cf. Matthiessen, 1993b, for some
discussion).

46

1. Theoretical Preliminaries

(ii) The second representational challenge is the need to model how the overall
ideation base is expanded by grammatical metaphor. It must be shown how
metaphor adds junctional types to the ordinary types illustrated in Figure 1-12. We shall
not solve this representational problem in the present book; but we shall suggest ways of
thinking about it (see Section 6.8).
(m) The third representational challenge comes from outside the account of the
ideation base itself: the ideation base has to be related to the other metafunctional
modes of meaning, the interaction base and the text base. Hie interaction base will
include alternative 'projections' of the ideation base to account for the relationship
between speaker and addressee. We shall not attempt to take up this particular challenge;
the text base is discussed briefly in Chapter 9, Section 9.3 below.
(iv) The fourth representational challenge concerns the non-discreteness of the
various systems that construe semantic space (see Chapter 13, Section 13.3). The
semantic types represented by the features of systems in the system network do not
constitute discrete Aristotelian categories; they are values on semantic clines (cf.
Halliday, 1961) core regions, to use the metaphor of semantic space. We can bring
this out by adopting a topological view on meaning (cf. Martin & Matthiessen, 1991);
we can also explore the possibility of interpreting features as names of fuzzy sets (cf.
Matthiessen, 1995a).
In what follows we shall use mainly the systemic form of representation, although it
will be helpful at points to refer to frame-based inheritance networks. Our representational
practice is informal; the representations would be defined formally if required to support
various operations such as inference of inherited properties. We should also emphasize
again that representation is at one strata! move below theoretical specification; while it
takes us closer to a form of specification that can be implemented in an NLP system, the
total coverage decreases. There are various types of theoretical information that can be
stated in terms of language or a diagrammatic representation but which we do not yet
know how to represent in terms explicit enough to be implemented in a computational
system. One instance that arises at various points in our discussion is that of topological
specifications and the related notion of clines (scales) between types. It is thus essential
to construe our interpretation of the ideation base multi-stratally: representation at any
single level is bound to be partial, and will be so whichever type of representation we
choose to deploy. But this is not an undesirable state: the theory should always push
towards the expansion of the representational resources.

1.10 Organization of the remainder of the book

47

1.10 Organization of the remainder of the book


Here in Part I, we have now introduced the major themes of our book: the notion of a
meaning base and the approach to it from grammar. We shall elaborate on these as
follows in Parts II through V.
EL The meaning base: We begin by presenting an overview of our description of
the meaning base. We then discuss the major semantic categories sequences,
figures and elements in more detail and explore the semantics of
grammatical metaphor. After our description of the ideational semantics of
English, we turn to Chinese to identify the major similarities and differences.
ID. The meaning base as a resource in text generation: Next, we show how the
meaning base can be used as a resource in a language processing system,
particularly in a text generation system. First we discuss how two particular
'domains' can be modelled in the terms introduced in Pan H. Then we locate the
meaning base in the overall organization of a text generation system, discussing
for example how it interacts with the grammar.
IV. Theoretical and descriptive alternatives: After having presented the details of
the meaning base and shown how it can be used, we compare and contrast
certain key points with other approaches within linguistics and computational
linguistics, in theoretical as well as descriptive terms.
V. Language and the constnial of experience: In the final part, we take, our account
of the meaning base as the basis for a discussion of the role of language in the
construal of experience. We look at how children learn about the world how
their 'world view1 expands and changes as their linguistic resources develop.

Part II:
The ideation base
2. Overview of the general ideationai potential
In this second part of the book, we shall describe the ideation base how the
phenomena of our experience aie construed as categories and relationships of meaning.
Our primary concern here is descriptive rather than theoretical or representational; but, at
the same time, theory and representation have to be grounded in a comprehensive
description if they are to account for the full range of meanings in the meaning base. The
descriptive scope of our account is essentially the most general part of the ideation base
potential (cf. Chapter 1, Section 1.4); Figure 1-5 showed how this relates to the overall
ideation base. In Chapter 8, we shall then illustrate how domain models can be developed
and related to the general potential.
We shall present the meaning base first in summary fashion, in the form of a system
network accompanied by a brief explanation and exemplification. This will include an
initial account of grammatical metaphor as this is interpreted in the systemic model,
showing how it may be formally defined by reference to the systemic organization. For
greater clarity, the system network is introduced and built up in a number of steps,
allowing for a brief commentary to be inserted at each step.
2.1 Phenomena
A phenomenon is the most general experiential category anything that can be
construed as part of human experience. The phenomena of experience are of three orders of
complexity: elementary (a single element), configurational (configuration of elements,
i.e. a figure) and complex (a complex of figures, i.e. a sequence) see Plate 1.
While figures are said to consist of elements and sequences are said to consist of
figures, the 'consist-of relation is not the same: elements are constituent parts of
figures, functioning in different roles; but figures form sequences through
interdependency relations. We will return to these different types of organization
below. The typical representation of sequences, figures and elements in the grammar
is as in Figure 2-1.

2.1 Phenomena

Fig. 2-1: Typical realization of sequences, gures and elements


Examples:
sequences
Rain ending from the west, becoming partly sunny.

49

2. Overview of the most general ideational potential

50

Take 8 hard-boiled eggs, chop finely, msh with 3 tablespoons of


soft butter, and add salt and pepper.

figures
rain ending from the west
becoming partly sunny
take 8 hard-boiled eggs
chop finely

elements
rain, ending, from the west, take, 8 hard-boiled eggs, chop,
finely

2,2 Sequences
A sequence is a series of related figures. Consequently, sequences are differentiated
according to the kinds of relations figures can enter into temporal (x happened, then y
happened, etc.), causal (x happened, so y happened, etc.), and so on; in the most general
terms, the relations yield the following type of sequence: see Plate 2.

Plate 2: Types of sequence


In any pair of figures related in sequence, one figure may (i) expand the other, by
reiterating it, adding to it or qualifying it; or (ii) project (report, quote) the other by

2.2 Sequences

51

saying it or thinking it In either case, the two may be either equal or unequal in status,
or semantic weight. Some examples:
(1) expansion: add & unequal
The heat wave in the although northern Florida
Southwest will weaken will remain hot.
slightly,

(2) expansion: add & equal


Highs will be mid-80s to
mid-90s,

but parts of Texas could


reach the 100s.

(3) expansion: reiterate & unequal


The rest of the South will
be mostly dry and sunny,

with only isolated showers


in Florida.

Sequences are organized by interdependence relations and they aie indefinitely


expandable. Consider the following two examples, which are substeps in procedures:
If the rotor is not pointing to the mark, pull the distributor
part way out, turn the rotor some, and reinsert the distributor
until the rotor points to within a half inch of the mark.
Add the remaining ingredients, stir to coat the chicken well and
continue until a thick sauce has formed and the chicken is tender.
These procedures illustrate the logic1 of the organization of sequences. The
organization of the second sequence can be diagrammed as in Figure 2-2. The
inteidependency relations linking the units in the chain are marked by arrows between
pairs of links.
The phenomenon diagrammed in the figure is organized as a temporal sequence (and
[then]') of three process configurations, 'add', 'stir1, and 'continue to stir1, two of which
are further expanded, one purposively ('[in order] to') me other temporally ('until'). The
temporally related unit is itself an additive sequence ('and [also]'). The sequence might
have been expanded further, since it is organized as a chain of interdependent units; and in
this respect it contrasts with the configuration in Fig. 2-4a below.

52

2. Overview of the most general ideational potential

Fig. 2-2; Representation of a sequence

2.3 Figures
A figure is a representation of experience in the form of a configuration, consisting of a
process, participants taking pan in this process and associated circumstances. There are,
of course, indefinitely many kinds of process in the non-semiotic world; but these aie
construed semiotically, according to the way in which they configure participants, into a
small number of process types being, doing, sensing, and saying. The first three of
these have clearly defined subcategories: see Plate 3.
Then, each figure may be either projected (by another figure, or in some other way) or
else not; and if projected, it may be an idea or locution or something else but that
'something else* embodies grammatical metaphor and will be discussed later on.
Examples:
(1) (being: ascribing) & non-projected
Rain and thunderstorms will extend today from New England to the
upper lake region and North Dakota.
Morning skies will be partly cloudy today.
The taste is very pleasant and salty and it
content.

has a high iron

53

2.3 Figures

(2) (sensing: thinking) & non-projected


Cloudy skies are forecast today for the New York metropolitan
area.
Variable cloudiness is expected tomorrow.

(3) (doing: doing (to)) & non-projected


A warm front may bring scattered showers or thunderstorms to the
northern Tennessee Valley.
Rain will fall in the North West.
Melt the butter in a saucepan and add the onion.

Plate 3: Types of figure


The principle of organization of a figure is different from that of a sequence. As we
have seen, a sequence is constructed by interdependency relations of expansion and
projection. In contrast, a figure is constructed as an organic configuration of parts. Each
part stands in a specific relation to the figure as a whole. The parts of a given
configuration are (i) a nuclear process, (ii) one to three participants of different kinds

54

2. Overview of the most general ideations! potential

taking part in the process, and (iii) up to around seven circumstances of different kinds
associated with it.1
Participants arc inherent in the process; they bring about its occurrence or
mediate it. There are a number of specific ways in which a participant may take
part in a process; it may act out the process, it may sense it, it may receive it, it
may be affected by it, it may say it, and so on. The different configurations of
participants are the bases for a typology of process types. The distinction
between participants and circumstances is a cline rather than a sharp division,
but it is semantically quite significant
Circumstances are typically less closely associated with the process and arc
usually not inherent in it. They specify the spatial or temporal location of the
process, its extent in space or time (distance or duration), its cause, the manner
of its occurrence, and so on.
Grammatically, the nuclear process, its participants, and its circumstances arc
typically represented as constituents in the transitivity structure of a clause: see Figure 23.
In knowledge representation, a configuration is typically represented as a concept
frame with a number of roles (slots; as in Brachman's KL-ONE; e.g. Brachman, 1978;
Brachman & Schmolze, 1985; cf. also Steiner (1991: 121-31) for a review of framerepresentation see Chapter 1, Section 1.9.5 above). Configurations can be represented
in two ways by means of such a frame.
(i) The configuration itself may be taken to constitute the frame (cf. for example
Anderson, 1983: Chapter 3), much as in the grammatical constituency diagram in Figure
1-2 above. The process is represented as a role (slot) in the frame in the same way as
participants and circumstances: see Figure 2-4a.

Both participants and circumstances have been discussed extensively in terms of (deep) cases
in linguistics and computational linguistics, but usually without a distinction between the
two. The process is typically not given a deep case.

55

2.3 Figures

Fig. 2-3: Typical realization of processes, participants and circumstances

Fig. 2-4a: Configuration as frame


(ii) Alternatively, the process part of the configuration can be treated as the frame
itself, rather than as one of the roles of the frame: see Figure 2-4b

56

2. Overview of the most general ideational potential

Fig. 2-5: Figure with role types and role fillers

2.3 Figures

57

In either case, the roles have restrictions on what semantic types they can be realized
by so-called "value restrictions". These restrictions are themselves types from other
parts of the semantic taxonomy not included in the figure. The two representations make
different claims about a configuration. In particular, the first approach allows the process
to be a role that is filled by another concept frame which may have its own internal
organization. As we will see, the process part of a configuration may indeed have internal
organization. At the same time, the second approach is simpler and it may be sufficient
for many purposes.
If we use the first approach, we can see that a configurational frame is made up of
three kinds of roles, (i) the process role, (ii) participant roles, and (iii) circumstance roles.
Each type has a default filler from the hierarchy of simple phenomena, respectively
processes, participants, and circumstances: see Figure 2-5 above.
As already mentioned, there is a typology of figures based largely on the nature of the
particular types of the process, participant, and circumstance roles; the most general part
of this is shown in Figure 2-6. Each type of figure has its own set of more delicately
specified roles with particular value restrictions. For example, figures of doing have an
Actor, which in turn is a participant (rather than some other kind of phenomenon).
(Notice that the specification of the filler of the Phenomenon role in a sensing figure is
more general than 'participant' any type of phenomenon can be sensed: see Chapter 4,
Section 4.2.1.1.)
Table 2(1): Participant roles and their fillers for figures of three types

2. Overview of the most general ideational potential

58

Fig. 2-6: Subtypes of figures and the participant role value restrictions
The role restrictions for sensing, saying, and doing ate summarized in Table 2(1)
below. These role restrictions represent a kind of metaphysics of English transitivity.
For example, according to English, ideas and locutions cannot act on things, but there is
no general restriction on what kinds of things may act on other things. Not only
persons, but also inanimate things and abstractions may kill people (a figure of doing):
the rifleman/the rock/his stupidity killed cousin Henry. In contrast, if we had built our
model according to the demands of the transitivity grammar of Navajo, we would have
had to rank things in terms of their capacity to act upon other things; e.g., an inanimate
cannot act upon an animate thing (Witherspoon, 1977).

2.4 Elements
As we have seen, elements fill the roles of figures. Participant roles aie filled by
participants (things or qualities), circumstance roles by circumstances (times, places,
causes, etc.), and the process role by a process. There are correlations here between the
taxonomy of configurational phenomena and that of simple phenomena: Table 2(1)

59

2.4 Elements

above summarizes the value restrictions on the fulos of the different participant roles of
the four process types. The elements of a figure are of three kinds: (i) the process itself
(action/event, process of consciousness, or relation), (ii) a participant in that process,
or (iii) a circumstantial element or circumstance. Example:
participant

process

participant

a separate
weather
system

will
brins

thundershowers

process
malt

circumstance
to the eastern
Dakotas and
Minnesota

participant

circumstance

the butter

in a saucepan

circumstance
in the morning

As already noted above, processes are realized by verbal groups, participants by


nominal groups, and circumstances by adverbial groups or prepositional phrases. In
addition to the three types of element that serve in figures, there is (Hie further type of
element the relator: see Plate 4. Relators serve to construe logico-semantic relations
of expansion between figures in a sequence (cf. Chapter 3, Section 3.4); they are realized
by conjunction groups. We shall discuss them briefly in Chapter 5, Section 5.1, and then
again in the context of grammatical metaphor in Chapter 6.

Plate 4; Types of element

2.5 Participants
Participant roles in figures are filled by elements of the type 'participant'; they aie
phenomena capable of taking on a participant role in a process configuration, e.g.
bringing it about or being affected by it. They are further differentiated accoiding to two
parameters: see Plate 5.

60

2. Overview of the most general idealional potential

Plate 5: Participants

Macro-participants arc all metaphorical and will be left out of consideration for the
time being. Simple participants may be things or qualities; for example,

2.6 Simple things


Some 'simple things' aie metaphorical; the remainder are referred to as 'ordinary', and
these are either conscious or non-conscious (this is the distinction that is actually made in
the semantic system, not animate/inanimate or human/non-human): see Plate 6.

61

2.6 Simple things

Plate 6: Simple things


Table 2(2): Examples of participants
conscious
material: animal:
higher
material: animal:
lower
material: object

general
person, man, woman,
boy, girl, baby
horse, stallion, mare,
foal; dog, bitch, puppy
ant, butterfly, slug

meteorological

bouse, rock, car,


hammer

scattered clouds

material: substance water, air, tea, sand


material:
abstraction
semiotic:
institution
semiotic: object
semiotic:
abstraction

history, mathematics
government, school

culinary
cook ['you']
[only as ingredients]

air, cloud, sunshine


a slow-moving
weather system
weather bureau

book, document, report, forecast


film, picture, painting,
symbol
notion, idea, fact,
chance
principle

[ingredients:]
potato, onion, stem,
root, [implements:]
knife, pan
[ingredients:] fat,
sugar, pure
heat, taste, colour

recipe

Non-conscious ordinary things are distinguished along more than one dimension, but the
categorization given here can be taken as primary, in the sense that it is the one that
seems to have the clearest reactances in the grammar.
Most of the participants in the meteorological texts are in fact metaphorical; but there
are a few which illustrate these categories: a slow-moving weather system (abstraction),
ice (substance), scattered clouds (object), weather bureau (institution). In contrast, most
of the participants in the culinary texts are non-metaphorical (we refer to this as
congruent). They are concrete objects and substances that can be chopped, added,
sprinkled, and poured; vegetable fat, sugar, pure, spinach, stems; or used as

62

2. Overview of the most general ideational potential

implements: knife, saucepan. These and other examples are tabulated in Table 2(2)
above.

2.7 Simple qualities


Qualities characterize things along various parameters, as in green cabbage : red cabbage:
see Plate 7.

Plate 7: Simple qualities


Again, a number of simple qualities are metaphorical; of the remainder, the "ordinary"
qualities, one subtype is qualities of projection and the other is qualities of expansion.
(As we will show in subsequent chapters, the categories of projection and expansion ye
very prevalent in the organization of the ideation base. We discuss their application to
qualities in Chapter 5, Section 5.3.3.2 below.) Examples from the weather repon texts:
likely in rain is likely (qualities of projection); hot, humid, sunny, dry (qualities of
expansion: sense-measure); one to three inches of [rain] (qualities of expansion:
quantity); metropolitan (qualities of expansion: class); high [pressure] (qualities of
expansion: sense-measure [abstract]). These examples and others are tabulated in Table
2(3).

63

2.7 Simple qualities

Table 2(3): Examples of qualities


meteor- culinological ary
happy, angry; likely, likely
certain
elaborat- identity
similar, different
metropoling
attribut- class
wooden, stone,
medieval, urban, rural itan
ion
dead, alive; male,
status
female
brown,
senseheavy, light; green, hot,
measure red, blue; soft, hard; humid,
soft
rough, smooth; loud, sunny, dry
quiet
propens- difficult, naughty,
helpful
ity
quantity few, many; one, two
one, two
extendadditional,
alternative,
ing
contrasting
enhancprevious, subsequent;
interior, external
ing
general

qualities of
projection
qualities of
expansion

2.8 Circumstances
Circumstances fill circumstantial roles in figures. We can recognize two simultaneous
distinctions; see Plate 8. One concerns the type of circumstantial relation construed; the
primary contrast is between circumstances of projection and circumstances of expansion
and within the latter we distinguish those of elaboration, extension and enhancement Hie
other concerns the experiential complexity of the circumstance; circumstances are either
'simple' or 'macro', the fonner being more truly elemental while the latter are more like
figures. We shall discuss these distinctions in Chapter 5, Section 5.5 below and just
mention a few common types here.
Among simple circumstances, the most usual are those of time, place, manner-quality
and intensity, all of which are circumstances of enhancement Examples: I skies will be
partly cloudy] today (locative: time), increasingly (manner: intensity), widespread
(locative: place), easily, carefully (manner: quality).
Macro circumstances are those which are made up of a special type of figure having
another participant inside it, for example (circumstances of enhancement): (locative: place
[abstract]) in the low to mid 60s, (locative: place) from the northeast, (manner: quality)
at 15 to 25 m.p.h., (locative: place) throughout the northern Rockies, (locative: place)
in a casserole, (locative: place) in a hot oven, (extent: duration) for 10-15 minutes,
(manner: means)wrA a clean absorbent cloth.

64

2. Overview of the most general ideational potential

Plate 8: Circumstances

2.9 Processes
Processes serve in the most central or nuclear role in a figure; they embody the
temporal properties of a figure unfolding in time: see Plate 9. Other than metaphorical
processes, the process element is either polar (positive/negative) or modal (some
intennediate degree between positive and negative); it may embody phase, or aspect; and
it will refer to past, present or future time. Polarity and modality derive from the
interpersonal perspective on the process.

Examples from the meteorological texts: [yesterday] was [sunny] (past polar); [skies]
will be [clear tonight] (future polar); [scattered showers] may develop [south west]
(future modal); [temperatures] are expected to be [in the high 80s] (present polar phasal);
[coastal sections] could get [an inch or more of rain] (future modal).
Examples from the culinary texts: [this soup] comes [from Northern Thailand]
(present polar); simmer [for 15 minutes] (polar), continue to boil (polar phasal);
continue cooking and stirring [for 15 minutes] (polar phasal); [they] will not require
[any further cooking] (future polar).

65

2.9 Processes

- future
Plate 9: Processes

2.10 Summary
Plate 10 below shows the lattice as described up to this point, incorporating all the
categories introduced above (but still omitting those embodying any grammatical
metaphor). Here is an example of a meteorological text interpreted in terms of the
semantic features presented so far (see Figure 2-7).
New York Area
Morning
partly

kia

will

partly

cloudy

today,

becoming

sunny by afternoon. High temperaturas will ba in the

low 70s. Skies will be clear tonight. Low temperatures will be in


the middle 50s.
We interpret the first sentence (in bold) as a sequence of two figures; they arc related
by an extending relation (shown by the arrow in the diagram in Figure 2-7). Both are
figures of 'being'; sensory qualities ('cloudy', 'sunny1) ate ascribed. The first figure
consists of four elements (whose features are shown in boxes): one process, two
participants and one circumstance. The second figure consists of three elements: one
process, one participant and one circumstance.

66

2. Overview of the most general ideational potential

Fig. 2-7: Interpretation of part of meteorological text


2.11 Construing experience in the ideation base
We have presented an initial sketch of the ideation base as a resource for construing our
experience of the world around us and inside us. Hie focus of this sketch was the most
general system of semantic types such as 'figure', 'being-&-having', 'participant',
'conscious being'. These semantic types aie categories to which phenomenological
instances are ascribed; they thus embody the fundamental principle of generalizing across
individual phenomenological variation. And they are located somewhere in delicacy
between the most general type, the all-inclusive class of phenomenon1, and the most
delicate types we can recognize as being codified lexically in English.

2.11 Construing experience in the ideation base

Plate 10: General options in lattice

67

68

2. Overview of the general ideationai potential

Categorization is often thought of as a process of classifying together phenomena that


are inherently alike, the classes being as it were given to us by the nature of the
experience itself. But this is not what really happens. Categorizing is a creative act: it
transforms our experience into meaning, and this means imposing a categorical order
rather than putting labels on an order that is already there. As Ellis has expressed it,
categorization consists in grouping together sets of phenomena that are essentially not
alike (1993: Ch. 3). Tliere would be indefinitely many ways of construing analogies
among different elements in the total flux of experience; what our semantic resources
enable us to do is to construe those analogies which yield categories resonating with what
as a species, and as members of a particular culture, we have found to carry material and
symbolic value.
Naturally our categorization tends to be oriented towards phenomena on a particular
scale, those that lie within the bandwidth of those phenomena which are most readily
accessible to our senses and which we engage with in day to day existence. This is the
realm which impinges most closely on our physical, biological and social being; the
semantic system has evolved with this as its primary semogenic environment. Lying
beyond this are the micro- and the macro-worlds which are accessible to us only by
instrument and by inference. It is a powerful demonstration of the potential of the
semantic system that it readily fashions new meanings that model these experientially
remote domains. What is less obvious, perhaps, is that it has had to be equally
resourceful in modelling the ongoing changes in our social environment, where both the
overall social order and our local interpersonal networks are constantly being modified and
realigned.
2.11.1 Typological and topological perspectives

So far we have sketched the most general organization of the ideation base as a
typology of semantic types or classes. This is one theoretical perspective on the
organization a perspective realized in certain conventional systems of representation,
such as the system network (see Chapter 1, Section 1.9 above). It allows us to bring out
quite clearly the global organization of the ideation base as a resource for construing
experience; and it also allows us to show how the semantic types in this global
organization are interrelated, as in the case of participants filling Actor and Goal roles in
figures of doing. At the same time, there is another theoretical perspective on the
organization of meaning the topological perspective. Here meaning is construed in
terms of a spatial metaphor we can view the ideation base as an elastic,
multidimensional semantic space. This metaphor is already familiar in discussions of
meaning; we find it in Trier's notion of semantic fields, in the distinction between core
meanings and more peripheral meanings, in specifications of semantic distance, and so
on.

2.11 Construing experience in the ideation base

69

The notion of a vowel space (with its 'cardinal vowels') provides a familiar analogy.
As a material construction, it is limited to the three dimensions of physical space; but as
a physiological space, it accommodates variation along a number of dimensions, and
brings an elasticity to the expression plane that is in some extent analogous to the
metaphorical elasticity that we are ascribing to the plane of the content.
We shall relate typology and topology to one another as complementary
perspectives on meaning, and then we shall say a few words about the value of keeping
the topological perspective in view. Let us start with our analogy from the expression
plane: a vowel system can be construed both typologically as a set of systems, e.g. two
systems 'front/ back' & 'open/ closed' defining four vowel values, and topologically as a
two-dimensional space with four focal (core, cardinal vowel) locations. These two
perspectives are related to one another in Figure 2-8. Each system in the typology
corresponds to a dimension in the topology. We can thus say that the two simultaneous
systems correspond to a two-dimensional space. The systemic terms, or values,
correspond to regions within the vowel space along one of its dimensions; an intersection
of two systemic terms such as 'front* & 'open' is a region located along two dimensions.
If we add further systems in the typology, e.g. rounding ('rounded/ umounded'), nasality
('nasal/ non-nasal1) and tongue root position ('neutral position/ advanced position'), these
will correspond to further dimensions in the vowel space. With tongue root position we
are still maintaining a reasonably congruent relationship between our representation of
the vowel space and the oral cavity in which vowels are articulated, since advanced tongue
root position is simply a global shift of the whole space; but with nasality we are
beginning to use our representation more metaphorically, since the control of airflow
through the nose is not a feature of the oral cavity.
We can now consider a comparable example from the content plane of language
from the ideation base. In our description of sequences, we recognized two simultaneous
systems: the relative status of the figures ('equal/ unequal') and the kind of relationship
between them ('projecting/ expanding1). This account constitutes the typological
perspective on this region of the ideation base; it is mapped onto the topological
perspective in Figure 2-9. Sequences arc thus construed as a two-dimensional region
within the overall semantic space. The correspondences are the same as those already
noted for the vowel space.

70

2. Overview of the general ideational potential

Fig. 2-8: Correspondence between typological and topological perspectives on


vowels
What is the relationship between these two perspectives on meaning? So far in work
on semantics only the typological perspective has been able to be realized in a formal
system (cf. Chapter 1, Section 1.9 above); the power of the topological perspective still
derives largely from the metaphor of space within theories of meaning. For present
purposes, we can say that the topological organization construes a semantic space,
creating the correspondences shown in Table 2(4).
Both perspectives are valuable. The typological perspective allows us to gain insight
into the organization of meaning through the network, both as theoretical metaphor and
as a system of formal representation (again, cf. Chapter 1, Section 1.9). The topological
perspective gives us complementary benefits in foe first instance, the general notion
of a multidimensional elastic space. We have indicated in the rightmost column some
implications that derive from adopting the topological perspective. The general motif here
is that of indeterminacy (see further Chapter 13, Section 13.3). For example, we can
show how regions of meaning overlap (e.g. doing & happening overlapping with sensing
in an area of 'sensing as activity'). We will use informal topological diagrams at various
points in our discussion to bring out this central feature of the ideational system (see e.g.
Figure 4-5 in Chapter 4 below).

71

2.11 Construing experience in the ideation base

Fig. 2-9: Coirespondence between typological and topological perspectives on


sequences
The last row in the table above also represents indetenninacy, but this is
indeterminacy of a particular kind, relating to the probabilistic nature of the semantic
system. The types in the semantic system are instantiated according to probabuity values;
these aie manifested as relative frequencies in text The equivalent in the spatial
interpretation of meaning would be curvature or chreodization

Chreodization

embodies time and represents the change of systemic probabilities over time (see e.g.
Waddington, 1977, Sheldrake, 1988: Ch. 6, for discussion).
Table 2(4): Typological and topological correspondences
typological

topological
topological equivalent implication of
topological perspective

typology

topology

system (systemic variable,


parameter)
simultaneous systems

dimension in space

systemic term (semantic type)

intersecting dimensions
nw)ti4tmfitionality
region (along dimension in space)

probability of systemic term


partial association

chreodization of dimension
curves in space

(multidimensional) space with entent


(which can be extended) and regions
located within it (which are at
different distances from one another)
continuous rattier than discrete (cline,
scale),
more or less central locations in the
region along the relevant dimension,
related to other regions in relative
proximity with indeterminate
boundaries

72

2. Overview of the general ideational potential

2.11.2 Construing categories: ontogenetic perspective

Categorization has received much attention in recent work in linguistics, especially


within the framework of cognitive science. The received tradition was the classical or
Aristotelian conception of a category as a type definable in terms of necessary and
sufficient conditions its essential characteristics rather than its accidental ones. One
central aim of cognitiveiy oriented linguists has been to show that the Aristotelian
conception does not apply to semantic categories, which have to be retheorized in terms
of prototype theory. The initial impetus came from a series of well-known experiments
in categorization by E. Rosch and by Labov (1973) and from Berlin & Kay's (1969) study
of colour terms across languages; Wittgenstein's notion of family resemblances had
provided an earlier insight within a philosophical frame of reference (cf. Ellis, 1993).
Since there are now a number of summaries of these positions, and of the critique of
the Aristotelian conception (e.g. Lakoff, 1987; Taylor, 1989), we will not review the
discussion here. From a systemic-functional point of view, the intellectual context was
rather different. Systemic-functional theory was not oriented towards philosophy and logic
(see further Part IV on different theoretical orientations in work on meaning), and the
Aristotelian conception of a category did not figure as a traditional frame of reference that
therefore had to be rejected. Already at the outset of the theoretical work that was to
become systemic-functional theory, a major descriptive focus was on intonation (see e.g.
Halliday, 1963a, b, 1967), a domain that clearly cannot be construed in Aristotelian
terms; and the notion of cline was part of the theory already in the first major statement
(see Halliday, 1961). Further, the system was conceived of as a probabilistic one even in
proto-systemic work (see e.g. Halliday, 1956).
Systemic-functional work on 'categorization* has thus not engaged with the
philosophical tradition; nor has it tended to proceed by experimental methods. Rather, it
has been concerned with how meaning is construed in naturally occurring text, e.g. in
child language studies (Halliday, 1975 onwards; Painter, 1996) and in factual writing (e.g.
Halliday & Martin, 1993; Harvey, in prep.). Here one of the questions has been what
resources are available in the semantic system for construing new meanings for
'category development', both in the ontogenetic time frame and in the logogenetic time
frame. One central semantic motif that has emerged from these studies is that of
elaboration: both in relational clauses (realizing figures of being & having) and in
complexes, especially nominal ones (the traditional notion of apposition, realizing
sequences of participants). Here elaboration is at work e.g. in 'distilling* new meanings
over extended passages. That is, the semantic system includes a theory of how meanings
are construed; and this theory is itself a resource for construing new meanings. Thus
when a young child says cois also aren't people (see below and Halliday, 1991, for the
construal of cats), he is using a type of figure of being & having to construe a taxonomic
relationship in the ideation base, so that the membership of 'cats' (construed as Carrier) in

2.11 Construing experience in the ideation base

73

the class of 'people' (construed as Attribute) can be probed, in this case to 'outclassify'
cats from the class of people. Figures of being & having of the intensive and ascriptive
kind thus construe, among other things, the taxonomic relation among classes or types in
the semantic system.
The ideation base is thus a resource both for construing experience and for
construing its own construal of experience. It has the potential for expanding
itself precisely because it includes a theory of how meanings are construed When children
begin to make the transition from proto-language into language (that is, when they
begin to develop the system of the mother tongue, typically early in the second year of
Ufe), these resources for self-construal are not yet in place. The first things that aie
construed by naming aie individuals; there is as yet no potential for taxonomies of
general classes. But children soon take the critical step of generalizing across individuals
(see e.g. Halliday, 1993a, on generalization). From a lexicogrammatical point of view,
this means that naming has been generalized from individual names ("proper nouns") to
class names ("common nouns"). Painter (1996) provides a key to the understanding of
how linguistic resources are deployed in 'categorization* in language development,
drawing on her longitudinal case study of one child, Stephen, between 2 1/2 and 5 years.
The first stage in categorization is naming individuals as members of classes;
instances of the visual experience shared by the young child and his father or mother aie
ascribed to some general class of experience by means of a figure of being. At about 2
1/2, Stephen produced examples such as: Stephen (examining pattern on a rug): That's a
square. What's that? Mother: That's a circle. Here some perceived phenomenon of
experience is brought "into intersubjective focus" by being referred to exophorically pointing verbally, so to speak, to some feature of the material setting, sometimes
accompanied by or replaced by a pointing as a gesture. This phenomenon is construed by
Stephen as the Carrier of the figure of being, and is ascribed as a member of some general
class of experience, construed as the Attribute of the figure. Stephen is 'importing'
experience of instances into the semantic system by ascribing them to general classes in
that system. This is an act of naming, and later this act itself gets named by call (see
also Halliday, 1977, on calling as an early example of language being turned back on
itself).
Children thus build up experience as meaning, in contexts such as the one exemplified
above. Construing experience as meaning means locating classes such as squares and
circles somewhere in the semantic system, both locally as terms in systems and also
more globally in the ordering of these systems in delicacy. Painter comments: "...
through the naming utterances where Stephen was practising signification, he was also
necessarily construing the things of his experience into taxonomies". So Stephen also
construes the attributes of semantic classes, attributes that will help him sort out the
organization of the semantic system. Examples from about 2 1/2: (1) Mother: What cars

74

2. Overview of the general ideational potential

have you got there? Stephen: There's afire engine one with a ladder on; (2) Mother:
What did you see at the zoo ? Stephen: Elephants; they got big trunks.
When Stephen's meaning potential has gained critical semantic mass, he begins to
construe its own internal organization explicitly in an effort to sort out taxonoma
relations within the system. Painter gives the following example from about a year later
in his life, at approximately 3 1/2 years (Stephen is examining animal jigsaw puzzle
pieces):
Stephen: There isn't a fox; and there isn't - is a platypus an
animal?
Mother Yes
Stephen: And is a seal is an animal (sic)?
Mother. Yes (shepherding S to bathroom)
Stephen: And is er- er- er- erMother: You do your teeth vftiile you're thinking

Again, the resource for construing 'categories' is the intensive ascriptive figure of
being; but now both the Carrier and the Attribute are meanings internal to the semantic
system. That is, Stephen construes a taxonomic relationship between e.g. 'seal' and
'animal' by construing them as Carrier + Attribute:

Painter comments: "The importance of this development is that it constitutes a move on


Stephen's pan from using language to make sense of non-linguistic phenomena to using
language to make sense of the valeur relations of the meaning system itself." We can
diagram the contrast between these two steps in construing experience as categories of
meaning as in Figure 2-10.

2.11 Construing experience in the ideation base

75

visually shared experience

Fig. 2-10: Construing categories externally and internally


In bis fourth year, Stephen begins to provide causal evidence for categorization, by
adding an enhancing figure. The figure that is linked causally specifies attributes that ate
critical to the construal. For example: Mother refers to Bond airship as 'spaceship
balloon' Stephen: Not a spaceship - an airship cause a spaceship has bits like this to
stand it up; Stephen (pointing at page numbers): Thai's fifteen because it's got a five;
that's fourteen because it's got a four. These causal relations are internal (Halliday &
Hasan, 1976), i.e. oriented towards the interpersonal act of communication itself (1 know
it is an x, because it has feature y'X rather than external ('it is an x, because it has feature
y'). In other words, Stephen is attending to the act of construal itself 1 construe it as x
because of y1. At this stage in the development of Stephen's construal of experience, the
construal has become something that is not only shared, but can also be explicitly
negotiated and argued about
Also in his fourth year, Stephen begins to extend his deployment of figures of being
to include not only the ascriptive mode ('a is a member of x') but also the
identifying mode Cy equals x') so that he can explicitly relate meanings in his
ideation base in the form of definitions. For example, in Balance means ftyou hold it
on your fingers and it doesn't #01, he construes a single class of abstract participant,
"balance1, as something defined in terms of a sequence of figures, 'you hold it on your
fingers and it doesn't fall off: see Figure 2-11. He is now in a position to construe new
meanings in terms of the semantic system itself; that is, they do not necessarily have to
be 'imported' from direct extra-linguistic experience. As Painter points out, this is yet
another expansion of his semantic resources preparing him for educational learning in
later life where experience is very often vicarious experience, constructed entirely in
educational discourse.

76

2. Overview of the general ideational potential

Fig. 2-11: Construing the abstraction 'balance' within the ideation base

Stephen's definition of 'balance' is on a developmental path towards the kinds of


definition used in educational and scientific discourse definitions such as the dictionary
definition of 'cat4: a cat is a carnivorous mammal long domesticated and kept by man as a
pet or for catching rats and mice (adapted from Webster's Seventh New Collegiate
Dictionary; cf. Nigel's construal of 'cat' discussed below). Hie definition of 'cat' involves
a move to a less delicate category ('carnivore1 or 'carnivorous mammal', which can be
reconstrued as 'mammal that eats meat' cf. 'cats like things that go' in the Nigel
dialogues below) plus a qualification by a downianked sequence of figures of doing: see
Figure 2-12. In both examples, the definition construes a token-value relation (Chapter 4,
see Section 4.23.2) between a fairly delicate semantic type that is lexicalized within the
lexicogrammar and a restatement of this type by means of other resources in the ideation
base. The restatement draws more on the resources towards the grammatical end of the
scale, so that in the definition a lexicalized token is construed as a
grammaticalized value.2 This entails a shift from construing experience through
depth in the experiential taxonomy towards construing experience through expansion in a
logical sequence. In the first example, this only involves a downianked sequence of
figures; in the second example, this involves a move towards a less delicate category plus
a downianked sequence of figures. The first example can actually be interpreted as
meaning 'the condition when you hold it on your fingers and it doesn't fall off, showing
that the downianked sequence has the class meaning of 'condition'. See Harvey (in

The ideational potential for construing such relations of restatement within the ideation base
does not at all imply that meaning is analysed by means of decomposition into semantic
primitives, (i) Semantic types (such as 'balance', 'cat') are not decomposed into their
defining glosses within the ideation base; rather, they are construed as standing in an
intensive token-value relationship to these glosses, (ii) The semantic types are construed
internally to the ideation base according to their location in the elaborating taxonomy and
can be restated in various ways, (in) The semantic types may also construe extra-linguistic
categories of experience; that is, they may have signification outside the ideation base.

2.11 Construing experience in the ideation base

77

prep.) for a review of studies of the development of dfinitions, a survey of different types
of definition, and a general account of their role in technical discourse.

Fig. 2-12: Construing 'cat1 within the ideation base according to a dictionary
definition
We wl leave her account here (see also Painter, 1993, for Stephen's further
development), and refer to one additional example from a child's language development to
see how one particular category may be ongoingly construed over time. The example is
that of Nigel construing the category of cats, taken from Halliday (1984c) and discussed
in Halliday (1991). The following dialogues occur over a period of 8 months, from 2;10
to 3;6:
Text 1: Nigel at 2;10;22
Nigel
Mother
Nigel
Mother
Nigel

And you (=1) saw a cat in Chania Falls.


Yes, you saw a cat in Chania Palls.
And you picked the cat up. Mummy, do cats like meat?
Yes, they do.
Do cats like bones? Do cats like marrow?

2. Overview of the general ideational potential

78

Text 2: Nigel at 2;10;26


Nigel

Mother
Nigel

Can I stroke the cat? You (= 'I') want to stroke the cat
...you want to scratch it ... it's drinking itsmilk . . .
it's moving its tail out ... it'smoving its tail outside
... it's scratching ... it's putting its tail up ...
what's this?
I don't know; I suppose it's its elbow.
It's waggling its tail ... it's lapping it with its tongue
. . . you can go near its elbow . . . you can go near its
elbow . ..butyou can't go toonear itsface .. .because
it thinks you might take away its milk ... it was just a
bit frightened ... it thinked that you might take away its
milk ... has it finished its milk?

Text3:Nigelat2;ll;5
Nigel
Father
Nigel
Father
Nigel
Father
Nigel
Father
Nigel
Father
Nigel
Father

(thinking about "The house that Jack built') What is a


rat?
It's a sort of big mouse.
Does the rat go when the cat has killed it?
No, it doesn't go any more then.
Why did the cat kill the rat?
Cats do kill rats.
Why do they?
(formula) You'll have to wait to understand that till
you're a bit bigger.
No I can understand it now.
Well, cats just like to eat rats and mice.
Why do they like to eat them?
They just do.

Text 4: Nigel at 2;11; 15


I Nigel
| Mother

Why did the cat go out? Mummy, why did the cat go out?
It gets fed up, having its tail squashed.

Text 5: Nigel at 3;0;26


Nigel
Father
Nigel
Father
Nigel
Father

How do the cat's claws come out?


They come out from inside its paws Look, I'll show you.
Does it go with its claws?
Not if it's going along the ground.
And not if it's climbing up a tree.
Yes; if it's climbing up a tree it does go with its claws .

Text 6: Nigel at 3;2;7


Nigel
Father

Will the cat eat the grape?


I don't think so. Cats like things that go, not things
that grow.

2.11 Construing experience in the ideation base

79

Text7:Nigelat3;5;12
Nigel

Cats have no else to stop you from tros s ing them . . . cats
have no other way to stop children from hitting them ...
so they bite. Cat, don't go away! when I come back I'll
tell you a story. (He does so.)

Text8:Nigelat3;6;12
Nigel
Mother
Nigel

Can I give the cat seme artichoke?

Well she won't like it.

Cats like things that go; they don't like things that
jprow.

Text9:Nigelat3;6;14
Nigel
Father
Nigel
Father
Nigel
Father
Nigel
Father
Nigel
Father
Nigel
Father
Nigel
Father

I wish I was a puppet so that I could go out into the snow


in the night. Do puppets like going out in the snow?
I don't know. I don't think they mind.
Do cats like going out in the snow?
Cats don't like snow.
Do they die? (He knows that some plants do.)
No they don't die; they just don't like it.
Why don't puppets mind snow?
Well (hesitating) ... puppets aren't people.
Yes but . . . cats also aren't people.
No, but cats are alive; they go. Puppets don't go.

Puppets do go.
Yes, but you have to make them go; like trains.
Trains have wheels. Puppets have legs.
Yes, they have legs; but the legs don't go all by
themselves. You have to make them go.

Here the location of 'cat1 in Nigel's overall ideation base is construed directly by
means of figures of being & having (as in the earlier examples taken from Painter)
cats are alive; cois also aren't people', but the location of the category is also construed
relative to other categories through the participant roles it takes on in various figures. For
example, it is related in the figures of doing in certain ways to the young investigator
himself, in other ways to other animals and food, and in yet other ways to the parts of
its own body; and in one clause Nigel takes on the role of Actor with the cat as
Recipient: see Table 2(5).

2. Overview of the general ideational potential

80

Table 2(5): Cats in figures of doing


Goal:

Nigel
cat

creature:
'mouse*
food:
vegetable &
fruit; milk

Actor :
Nigel

pick up; stroke,


scratch;
(you/ children) +
Cross, bit
take away +
possession: milk

cat

creature:
'mouse*

food:
vegetable &
fruit; milk

(bite + [children])
scratch
move, put up,
waggle; lap +
body part: tail
kill; eat
eat + grape
drink, finish +
milk

This small sample suggests that cats aie construed in terms of a fairly clear
semantic world order: Nigel can act on cats, whereas they can act on 'mice' (i.e. rats
[= 'a sort of big mouse'] and mice) and on food, and they can (intentionally) act on
themselves. On the other hand, 'mice' and food are construed only as Goal relative to the
cat as Actor. (This kind of order is also important in Text 9 in the construal of puppets:
puppets move only through human agency.)
The constnial of cats is filled out in other types of figure: they can be construed as
Sensers (so endowed with consciousness) in processes of thinking and liking; they can be
construed as Phenomenon (with the child as Senser); and they can be construed as
Receiver (with the child as Sayer). The construal of cats as participants in figures is also
filled out by expansion in sequences, whereby the cat's actions can be qualified by
conditions, reasons, and so on: 'it does not go with its claws > if it's going along on
the ground'; 'does the rat go > when the cat has killed it'; [implicit:] 'the cat killed the
rat > because cats just like to eat rats and mice'.
The network of relations that emerges from these everyday dialogic construals of cats
is diagrammed in part in Figure 2-13 above (which uses some of the general semantic
types identified in our survey of the ideation base, together with a number introduced ad
hoc for the present example). This is of course only a small fragment of an ongoing
construal of experience where 'cat1 is just one category in the overall system; but it does
indicate how experience is construed as a network of related categories. The network is
not fixed or rigid, of course; it is constantly being reviewed, renovated, and expanded.

2.11 Construing experience in the ideation base

81

Fig. 2-13: The networking of cats


Figure 2-13 shows the various participant roles as elements in various types of figure.
As well as being figures in their own right, such configurations can also be used as
qualifications of a class of participant, such as cat as figures serving to qualify that
category with respect to its own role in the figure. Here the figure is being viewed in a
participant perspective; this is achieved by downranking the figure so that it requires the
status of a property (cf. Section 2.11.3 [3] below). Grammatically, this perspective would
be realized in the nominal group as a Qualifier of some Thing e.g. a cat is a creature
that drinks milk, kills & eats mice, climbs up trees ... . In other words, these
relationships can also be construed explicitly as parts of definitions, as we showed above.
Finally, we may note that the figures in which 'cat1 serves a participant role may form
sequences, as in it gets fed up, having its tail squashed Such sequences are also part of
the construal of cats. Indeed, Nigel tries (unsuccessfully, as it happens) to find out why
cats kill rats and mice, so that he might establish a causal sequence.
We can now sum up what can be learnt about construal from the ontogenetic
perspective:

82

2. Overview of the general ideational potential

(i) Initially the child consumes experience of phenomena that are in, or are brought
into, a shared visual field; once constituted into meaning, the experience can be
shared, validated and scaffolded dialogicaUy in collaboration with other members
of his/ her meaning group.
(ii) The earlier experiences ate clearly situated by virtue of the child's perception.
But <mce the process of constnial has been established, experience can be
generalized in the form of semantic classes, and classes of classes; it can be
further explored in terms of the semantic system itself; it can be developed
vicariously through discourse, and extended to include purely abstract categories.
(iii) Categories are construed in the network of the ideation base in terms of
different kinds of relations they enter into: [1] globally, they are construed in
terms of taxonomic elaboration: they form pan of a system that is located
somewhere in delicacy within the meaning base; [2] locally, they may be
construed in terms of meronymic extension: they may form part of a local
meronymic taxonomy, such as the cat and its body parts; [3] transcategorially,
they may be construed in terms of the roles they play in some other type of
category, such as the cat serving in certain participant roles of figures.
(iv) Categories are thus located within the ideational network. The network defines
a multidimensional, elastic space; and locations within this space are not fixed,
clearly bounded regions but rather regions with core or focal areas and more
peripheral areas that shade into one another.
(v) The ideation base is a resource for construing extra-linguistic experience (such
as visual experience). But it is also a semogenic resource for construing itself,
since it is built up out of the kinds of relations it itself construes relations
such as intensive ascription. Once critical semantic mass has been built up, new
categories can be construed internally within the system of the ideation base.
2.11*3 Construing as locating in a network of relations

We have seen how Nigel construed the category of 'cat' by probing its location in a
network of relations in his ideation base. Rom the typological point of view, construing
experience in terms of categories means locating them somewhere in this network of
relations. When we examine this network more closely, we find that there are actually
three types of network involved: (1) taxonomic in the strict sense (i.e. based on
hyponymy, 'a is a kind of x/ x subsumes a, b, c1), (2) taxonomic in the extended sense
(Le. based on meronymy, 'd is a part of y/ y has parts d, e, f ), and (3) eco-functional (i.e.
based on selection, *g has function m in environment z/ environment z comprises
functions m, n, p, and function m may be filled by g, h, j'). Of these three, the first

2.11 Construing experience in the ideation base

83

provides the global organizing principle of elaboration in delicacy; while the third relates
paradigmatic organization to organization on the syntagmatic axis.
[1] Taxonomic elaboration
Any option (category, semantic type) is thus located in the network relative to other
options, first within its own system, and secondly in terms of the location of that system
relative to others. For example, 'participant1 is one option in the system 'participant/
process/ circumstance1; and that system is a more delicate elaboration of 'element*. We
noted above in reference to Painter's work that the resources of figures of being are
deployed in construing taxonomic relations. There is also a generalized set of nominal
categories for construing steps in the delicacy hierarchy e.g. type, kind, class; and
some taxonomic regions have specific categories of their own, for example: brand, model,
make, issue; genus, species, family.
Delicacy is a uniform ordering from most general to most delicate; but along this
scale, semantic systems differ both in the number of distinctions at any one degree of
delicacy and in the overall delicacy that is achieved. There are specific differences
associated with particular taxonomic regions. For instance, humans and higher animals
are much more highly elaborated than lower animals. But there are also general differences
in the nature and degree of taxonomic elaboration associated with different 'bands' in
delicacy. Such general taxonomic principles are probably best known for folk-taxonomies
in the domains of plants and animals, diseases, and the like (e.g., Berlin, Breedlove &
Raven, 1973; Conklin, 1962; Frake, 1962; Slaughter, 1986); these fall within 'element:
participant: thing' in the ideation base. The maximum steps in delicacy in a folk
taxonomy are kingdom (unique beginner), life form, basic (generic) level, specific level,
and varietal level. These steps are by no means always present in every particular
taxonomy; and they have different characteristics, summarized in Table 2(6) (for
taxonomic examples, see Leech, 1974).
Berlin (1972: 53) characterizes the elaboration of labelled taxonomies as follows:
Generic names are fundamental and will occur first. These will be followed by the major
life-form names and specific names. At yet a later period, intermediate taxa and varietal
taxa will be labelled. Finally the last category to be lexically designated ... of any
ethnobotanical lexicon will be the unique beginner. The suggested sequence can be seen
diagrammatically as follows:

... no temporal ordering is implied for some categories. Thus no claim is made as to the
priority in time of specific names over major life form names. On the other band a

84

2. Overview of the general ideational potential

claim is made that a language must have encoded ai least one major life-form name and
one specific name before the appearance of intermediate and varietal named taxa.

The most favoured 'band' in delicacy is that of the basic level: it is taxonomically
most highly elaborated, it tends to be learnt first by children, and it construes categories
with syndromes of usually salient functional and perceptual properties. This is where
Nigel's categorization of 'cats' fits in and also Stephen's probing of 'seal', 'platypus1, etc..
It has been pointed out that basic level categories aie most distinct in terms of how
human beings interact with them (for a review, see e.g. Taylor, 1989). For example,
there are various ways in which we interact with cats in particular; but no distinctive
modes of interaction with animals in general. Indeed, Nigel's exploration of 'cats'
suggests that an important aspect of construing this basic level category is what roles he
can take on in relation to the cat of his immediate experience (including telling it a
story!). In addition, the "basic' degree in delicacy may have a special status in the
instantiation of categories in text. In her study of lexicalization in the pear stories,
Downing (1980) found that the basic level was by far the most preferred one; these are the
figures for concrete, nonhuman things (p. 106):
subordinate
5%

basic
93%

superordinate
2%

Such findings aie interesting because they give some indication of what factors to
consider in lexicalization in text generation: a given phenomenon can, in principle, be
construed at any point along the scale of delicacy. They are also interesting because they
point to the relationship between the system and the instance: from the perspective of the
system, we can observe that the basic degree of delicacy is most highly elaborated; from
the perspective of the instance, we can observe that it is most frequent. However, we have
to allow for considerable variation within a language in particular, functional or
registerial variation within the system according to the context of use, as in the variation
between the spoken system of everyday life and the written system of science to which
we turn next. (We also have to allow for variation across languages. We shall discuss
some differences between Chinese and English in Chapter 7 below.)

2.11 Construing experience in the ideation base

85

Table 2(6): Model of folk taxonomy


kingdom

life form

(unique
beginner)

(kind)

few in number;
polytypic

generic (basic)

specific

varietal

large in number;
core in folk
taxonomy

most in sets of two


or three

rare in folk
taxonomies

primary lexemes

seondary lexemes
highly salient;
among the first
learned by
children

The ideation base embodies not only folk taxonomies but also a range of taxonomic
models such as those used by experts and by scientists (see Chapter 14, Section 14.1).
These are all variants within the overall ideational system. We will come back to the
principle of variation in our discussion of recipes and weather forecasts (see Chapter 8)
and in our discussion of the polysystemic nature of the ideation base (see Chapter 13,
Section 13.4). Here we will just say a few words about the significance of such variation
for the taxonomic aspect of categorization.
The move from folk taxonomies towards scientific ones involves both an increase in
steps in delicacy and a change in the criteria used for classification. Wignell, Martin &
Eggins (1990) give examples from the classification of roses and of birds of prey (see
their discussion for original references). For roses, folk taxonomy and scientific
taxonomy can be contrasted as in Table 2(7).

2. Overview of the general ideational potential

86

Table 2(7): Roses


scientific
taxonomic
order:
kingdom
division
class
sub-class
order
family
genus
species
variety

Latin

English: scientific

Spermatophyta
Angiospermae
Dicotyledoneae
Rosales
Rosaceae
Rosa
Rosa setigera
Rosa setigera
tomentosa

seed plants
flowering plants
dicots
rose order
rose family
rose
wild climbing rose
special wild climbing
rose

English:
folk

folk taxonomic
order:

plant

unique beginner

(flowers )*

life form

rose

basic level
specific
varietal

* Note: this is only a partial equivalence


For birds of prey, they contrast a folk taxonomy with a "birdwatchers' vernacular
taxonomy". We superimpose these to give a sense of the difference; see Figure 2-14.
These examples show the difference in degree of delicacy quite clearly; an increase in
delicacy reflects the move in the direction of scientific knowledge. At the same time, the
criteria for categorization also change, from overt entena that are accessible to the naked
eye to covert criteria available only through the application of scientific techniques. The
change in taxonomic criteria is very clear from a review of the early taxonomic stage of
science in Europe in the 16th and 17th centuries: see Slaughter (1986).
So far, our discussion of taxonomic elaboration has focussed on those steps in
delicacy that tend to be construed lexically in the lexicogrammar. But in the overall
meaning base, lexically construed folk and scientific taxonomies do not start at the
highest degree of generality in delicacy; they are ordered in delicacy arto those systems
that are construed grammatically. For example while 'plant' is the "unique beginner" of a
particular folk taxonomy, there are several steps in the construal of things that are more
general than this steps that are constructed in the grammar, in categories such as
'thing', 'conscious/ non-conscious thing1, 'countable/ non-countable thing1 and the like.

2.11 Construing experience in the ideation base

87

Fig. 2*14: Folk and expert taxonomies

In general, then, we can say that the move in delicacy in the ideation base from 'most
general* to 'most delicate' is construed lexicogiammatically as the move from 'grammar'
to 'lexis': see Figure 2-15. This is of fundamental significance in the construal of
semantic categories. The early part of the scale of delicacy is construed in the grammatical
'zone'. This zone provides the resources of grammatical schematization for construing
more delicate categories: those categories are realized lexically but construed according to
the systemic parameters of the grammar. For example, the grammar of the nominal group
provides a schema for construing various delicate categories of things, by classifying,
describing, ordering and other such strategies.
Semantic types in different taxonomic regions are distinguished according to different
criteria; and they have different sets of roles. But the different criteria and sets of roles arc
construed within the grammar. We can exemplify these, taking domains a little more
delicate than those presented in Plate 9 above.

88

2. Overview of the general ideational potential

Fig. 2-15: The scale of delicacy in the ideation base construed lexicogrammatically
(i) Phenomena within different taxonomic regions are classified according to different
kinds of grammatically construed criteria. For example:
(a) perception (figures: subtype of sensing): according to means of perceiving: see
(means: eyes)/ hear (means: ears)/ smell (means: nose)/...
(b) transformation (figures: subtype of doing): according to result, e.g. break
(result: into pieces)/ melt (result: into liquid state)/ shrink (result: smaller)/
pulverize (result: powder, dust)/...
(c) motion (figures: subtype of doing): according to manner, location, purpose,
e.g. fall (location: downwards)/ stroll (purpose: for pleasure)/ flee (purpose: to
escape)/ walk (manner: with legs, fairly slowly)/...
(d) higher animals (elements: subtype of conscious thing, at the taxonomic depth
of species): according to epithets of age and sex, e.g. cow (sex: female, age:
adult), bull (sex: male, age: adult), calf (age: nonadult).
(e) artefacts (elements: a subtype of object): according to material, purpose, e.g.
containers (purpose: to contain substances) barrel (material: wood), basket
(material: cane or other woven material), basin (material: metal), bowl (material:
earthenware or glass).
(ii) Phenomena within different domains have different grammatically construed
structural roles associated with them. For example:
(a) perception: perceiver & phenomenon being perceived;
(b) concrete thing (elements: subtype of object): various epithets, specifically of
physical dimensions such as size, shape, weight, colour and age.
(c) weight (elements: subtype of quality): tensor, showing degree of intensity.

2.11 Construing experience in the ideation base

89

Within the lexical zone, we have already referred to the differentiation according to the
degrees in delicacy; in particular, the 'basic1 degree of delicacy stood out as most highly
elaborated. There are also grammatical reactances indicating the taxonomic differentiation
between the "basic1 degree of delicacy and lower degrees of delicacy, at least in cases where
the taxonomic relationship is of a particular kind For example, the less delicate category
may be construed by a mass noun whereas the more delicate, basic degree categories may
be construed as count nouns: see Table 2(8).
Wierzbicka (1985: 321-2) identifies this phenomenon and shows that it is not
arbitrary. Her explanation is in fact that the relationship between e.g. 'furniture* and
'chair, table' as super-category and subcategory is not the normal 'kind of relationship
between e.g. 'bird' and 'swallow, magpie', but rather a grouping of different kinds
according to similarity in use:
Thus, supercategones such as bird or tree are 'taxonomic', i.e. they belong to
hierarchies of kinds (where each 'kind* is identified on the basis of similarity between
its members); supercategories such as crockery, cutlery or kitchenware are not
taxonomic they include things of different kinds, grouped on the basis of contiguity
and/ or similarity of function, not on the basis of similarity of form.
We take delicacy to include both types of the relationship of supercategory to
subcategory both the "taxonomic" and the "non-taxonomic" ones. However, it would
seem that with the relationship between 'furniture' and 'chair', 'table' &c. we are on the.
borderline between elaboration and extension. We now turn to meronymic extension in
the ideation base.
[2] Meronymic

extension

The ideation base construes phenomena as organic wholes that may take on roles in
other kinds of phenomena; but it also deconstrues many such organic wholes into their
component parts. When these component parts are phenomena of the same type
participants (e.g., chair: legs, seat, back), figures (e.g., baking a cake: stages in the
procedure), this is known as meronymy (or meronymic taxonomy; cf. meros 'part'; see
e.g. Cruse, 1986: Ch. 7).3 We find local taxonomies of this kind, often interlocked with
hyponymies; for example, see Figure 2-16. Taxonomies thus embody the two types of
expanding relationships we mentioned at the beginning of this section extension and
elaboration. The meronymic type of taxonomy is extension, whereas the hyponymic type
is elaboration.

This has sometimes been called partonymy, in apparent ignorance of the established term
meronymy.

90

2. Overview of the general ideational potential

Table 2(8): Count and mass nouns in taxonomic relation


kind (life form')
^ mass noun
furniture
silverware; cutlery
bedlinen
clothing

basic degree of delicacy


^ count noun
chair(s), couch(es), table(s), bed(s), ...
spoon(s), fork(s), knife (knives), ...
sheet(s), blanket(s), pillow case(s), ...
shirt(s), sock(s), vest(s), ... ; pants, trousers, slacks, ...

Fig. 2-16: Meronymic taxonomy

And this is in fact how we shall model them as local meronymic taxonomies,
applicable to some particular region of the hyponymic taxonomy, and
not as one global metonymy superimposed on our taxonomy. There would seem to be Ear
too many discontinuities to create a global meronymy; meronymies tend to occur only
where there are contiguous parts of an independent whole.4 Thus while concrete objects
are regularly construed meronymically (with the human body as both a representative
example and a model for other mtonymies), substances are not; substances are extended
through measure ('unit of) rather than through pan ('part of). Similarly, taxonomies tend
to be more extended meronymically for concrete regions than for abstract ones (although

On a fairly large scale these discontinuities are embodied in the differentiation among the
natural sciences specializing in domains such as the body, the atom and the universe.

2.11 Construing experience in the ideation base

91

even things in an abstract region can have parts, e.g. aspect of an idea). Thus there is a
generalized set of categories such as part, element, component, aspect; and also a
generalized set 'facets' of spatial or temporal orientation, top, bottom, side, front, back,
middle, centre; beginning, middle, end. There are more specific variants for parts of
particular concrete things, such as facade, roof, wall [of a building]; ceiling, floor, wall
[of a room]; slope, peak [of a mountain]; limb, trunk, root, bark [of a tree]; skin, core,
pips [of an apple]; crust, crumb [of bread]; preface, epilogue [of a book].
The examples of meronymies given above come from the region of 'participant1
within the ideation base. It is not clear to what extent, or in what sense, meronymic
taxonomies extend beyond that region. We can certainly recognize that processes have
phases (cf. the discussion of phase in Chapter 4, Section 4.4) *begin to do, keep
doing, stop doing'; but it is not immediately clear that these form a process meronymy
analogous to the parts of a participant. Although we might reconstnie he began to dance
metaphorically as the beginning of his dance on the model of 'the beginning of the
book', this is a metaphorical reification of the process 'dance* and we have to be cautious
in interpreting the implications for the congruent process 'dance'. If we probe a little
further, we can see that process phase is concerned with the occurrence of a process in
time its temporal unfolding: "begin to do* means 'begin to be actualized (to occur) as
doing in time1. In contrast, participant meronymy is not tied to the existence of a
participant in referential space.
Turning to figures, we can note that a given figure may be restated in a definition as a
sequence of less delicate figures: see Table 2(9).5

Adapted from J. Anderson & E. Hanna. 1985. The new Doubleday Cookbook. Gaiden City,
NY: Doubleday & Company, Inc.

92

2. Overview of the general ideational potential

Table 2(9): Figure defined as sequence of less delicate figures


Token = figure
sb caramelize sugar
sb baste food
sb cream butter or
other fat
sb marinate meat, ...
sb steep tea leaves, ...

Value = sequence [of figures]

sb heat sugar
until it melts
and turns golden
sb ladle or brush drippings, liquid, butter, or sauce over food
in order to add flavour
as it cooks
and to
prevent dryness
sb beat butter or other fat either solo or with sugar or other
until it is smooth and creamy
ingredients
sb steep meat, fish, fowl, vegetables or other savoury food
in a spicy liquid several hours
until food absorbs the
flavouring
sb let tea leaves, coffee grounds, herbs, or spices stand in
until their flavour is extracted
hot liquid

We can interpret the figures to the right as consisting of a sequence of (less delicate)
figures. There is certainly a change in focus some happening either may be construed
as a single figure (at a fairly high degree of delicacy) or may be *blown up' as a
sequence of figures (at a lower degree of delicacy); and this change of focus is a
significant option offered by the ideation base. It has been recognized in work on
knowledge representation; and it is a significant variable in the move between
commonsense models and uncommonsense ones. However, while sharing this change in
focus with prototypical participant-based meronymies, the relationship between figure and
sequence is unlike these meronymies in other respects. Prototypical meronymies do not
involve a change in delicacy; and the categories they relate together are phenomena of the
same kind (as noted above). Here the relationship is again one of restatement; for
example: '[[for somebody to caramelize sugarDmeans ([that s/be heats it until it melts and
tujms golden brownfl.
While the choice of level* in hyponymic elaboration is the choice in delicacy of
categorization, the choice of level in a meronymic taxonomy is the choice in
delicacy of focus. The focus is typically on the whole (i.e., the most inclusive region
within the meronymy) even if a specific part is particularly important (cf. Langacker,
1984). We can see this in the way the involvement of elements in figures is represented:
He switched on the light (with his right hand)
[rather than: His right hand switched on the light]
She held the ticket with her teeth
[rather than: Her teeth held the ticket]
He patted the dog (on its head)
[ratherthan: His hand patted the dog's head]
He saw the alien (with his own eyes)

2.11 Construing experience in the ideation base

93

[rather than: His own eyes saw the alien]


Let's paper [the walls of] the living room
He put the vase on [the top of] the table

That is: the whole is involved in the figure as a participant, and if the part comes into the
picture as well it tends to be construed as a circumstance (Means, Location) or as a
separate participant (Goal, as in Nigel's observations about the cat quoted above: it's
putting up its tail, rather than us tail is putting up it is the whole organism that is
endowed with agency).6 There is clearly room for alternative choices; for example,
figurative representations of conscious processing may construe the whole person as a
privileged body part (and in some languages, e.g. Akan, this is a regular, non-figurative
feature of certain figures of sensing):
His death broke her heart ('devastated her1)
His behaviour turns my stomach ('upsets me')
The news blew my mind ('surprised me')

As an example, in the major portion of his novel The Inheritors, dealing with life
before homo sapiens, Golding always represents the act of perception as involving a
sense organ rather than the whole person. This is a departure from the norm of the
language; it owes its effect, that of deconstructing the whole person as a potentially
independent agent, precisely to this departure from the norm, as in his nose examined this
stuff and did not like it (see Halliday, 1973: 117, 124).
Such figurative construals all depend on extension a whole is being construed in
terms of a part (synecdoche; pars pro tow). They occur in various environments;
additional examples are:
I ' m happy to see so many new faces in the audience.
I've got too many mouths to feed,
he man at table 5 is the chicken curry.

There are two tendencies at work here. One is a general tendency within the ideational
metafunction to involve wholes rather than parts, especially where the whole obviously
functions organically (like a person). The other is a pressure from the textual metafunction:
since in English, first position in the clause has a distinct in the configuration as text (as
'point of departure1: grammatically, the Theme of the clause cf. Figure 1-2 in Chapter 1,
Section 1.3 above), these wordings allow the "whole person0 to function as the point of
departure rather than just a part of the body. Compare here pairs such as he has brown hair/
his hair is brown, I have a headache/ my head aches where the first is typically preferred.

94

2. Overview of the general ideational potential

In discourse, the whole may serve as the anaphoric domain of a pan ('bridging1, cf. Clark,
1975):

The view of the mountain was breath-taking. The slopes were


covered with snow and reflected the last purple rays of the sun.
[3] Eco-functional selection
Taxonomic elaboration and meronymic extension form complementary networks of
relations that together make up the paradigmatic organization of the ideation base. They
define, as we have noted, options in delicacy of categorization and in delicacy of focus.
The third type of network involved in the organization of the ideation base which we
identified above serves to relate paradigmatic organization and syntagmatic organization.
Specifically, it relates syntagmatic functions or roles associated with paradigmatic types
to the paradigmatic types that can serve in these functions. For example, the syntagmatic
function 'Senser' is associated with the paradigmatic type 'sensing', and it is related to the
paradigmatic type 'conscious being', since only participants of this subtype can serve as
Sensers. Figure 2-6 above shows this relation, together with a set of other such relations.
We shall refer to such network relations as eco-functional selection in order to
indicate that they specify the syntagmatic environment of semantic types by showing
them as selections for syntagmatic functions. Such selections have been referred to as
"pie-selections", but in order to avoid any connotations of temporal sequence, we prefer
the term "selection" for such relations in the ideation base.7
Selections are bidirectional. We have already seen that functions associated with
paradigmatic types constitute 'structural properties' by which these types are distinguished
and that selections provide further information about these 'properties'. Thus Figure 1-12
above illustrates how a figure of speanng is characterized by reference to the function of
Means, which selects for the thing 'spear'. Conversely, semantic types which are selected
are characterized by the functional environment within which they are selected Thus the
weapon 'spear1 is characterized by its potential for serving as Means in a figure of
spearing. Similarly, part of the meaning of a 'conscious being' is that it has the potential
for serving in the Senser role of a figure of sensing (cf. Figure 2-6).
The network of selection is inter-axial in that it relates syntagmatic specifications
to paradigmatic ones; it 'cuts across' the paradigmatic organization of the ideation base,
establishing correspondences between paradigmatic types through the syntagmatic
fonctions associated with them, as we have just illustrated by reference to Figures 1-12

As noted in Chapter 1, Section 1.9.6, preselection corresponds to the notion of Value


restriction' in frame-based inheritance networks.

2.11 Construing experience in the ideation base

95

and 2-6. We can trace one further example from sequence via figure to element: see Figure

2-17.

Construing a category thus includes locating it not only taxonomically and


meronymically but also eco-functionally, as in the case of 'conscious being1 just referred
to. We noted above in connection with the cat example how the fact that a participant
serves a participant role in a figure can be viewed either from the perspective of the figure
or from the perspective of the participant (i.e. as a property of the participant). If cats kill
mice, the ideation base accommodates the view from the angle of the figure: 'cats kill
mice'; but it also accommodates the view from the angle of the participant: 'animals that
kill mice1. Here the figure in which cats participate as actors has been construed as if it
was a property, so that the category of cats might be construed within the system of the
ideation base by means of the definition 'cats are animals that kill mice'. (In formal
systems of representation, this can be expressed by means of lambda abstraction.) There
is a general semogenic strategy at work here. We referred to 'sequences', 'figures' and
'elements' as three orders of complexity in Section 2.1. The shift in perspective
means that configurations of meanings that are of a particular order of complexity can be
accessed through selection not only in their normal environment (within phenomena of
the next higher order of complexity) but also within phenomena of lower orders of
complexity. Through this semogenic strategy of opening up the possible domains of
selection, a great deal of experiential complexity can be imported into the construal of a
participant.

2. Overview of the general ideauonal potential

96

Fig. 2-17: Successive links in selection network


2.11.4 Construal and stratification

A phenomenon of experience is construed as a category in the ideation base by being


given a location in the three semantic networks descnbed in the preceding section. This
taxonomic, meronymic and eco-functional location is the category's value (valeur) in
relation to the other categories within the ideation base. By being assigned a value
internal to the ideauonal system, a semantic category is also being related to categories
that lie beyond semantics itself. On the one hand it is being related to categories within
systems that lie outside language but which the semantic system interfaces with; on the
other hand it is being related to the grammatical categories in terms of which it is
realized. Figure 2-18 shows these three strata! angles on the process of constnial. Let us
explore these relationships in turn.
We saw in the examples of how categories are first developed that young children will
typically construe concrete phenomena that are part of the field of visual perception they
share with their interactants. In other words, they are construing into linguistic meanings
their experience of the material world as it is construed in the categories from another
semiotic system, viz. (visual) perception. These extra-linguistic categories are construed
as the signification of the semantic categories of the ideation base always in some

2.11 Construing experience in the ideation base

97

particular situation when the child first engages with them. To construe experience of
concrete phenomena as meaning is thus to construe some signification which lies outside
the ideation base as a value which is internal to the ideation base system. Part of the
power of categorization is that extra-linguistic phenomena that are quite varied in
signification can be construed as alike in value.
'signification' in external system

Fig. 2-18: Stratal angles on constnial


Children are able to construe the semantic system because they start in situations with
a material setting that is shared cf. above, and see Hasan (1985b: 25 ff.) on
Malinowski's contribution to this insight As they build up their ideation base, they can
begin to construe categories internal to the system out of existing ideational values; and
they can begin to move into the abstract domains of a purely symbolic world, where the
significations of semantic categories are abstract categories in social and socio-semiotic
systems. But precisely because these abstract categories are construed as meanings, they
can still be built up, negotiated and validated in collaboration with other members of the
meaning group. The move from the realm of concrete phenomena to the various realms
of abstract phenomena is made possible through the homogenizing power of meaning.
The semantic categories arc themselves construed by means of realization; they are
constructed within the grammar and the lexis of a language. If we model the ideation base
as a semantic space, we are foregrounding one aspect of the construction of meaning in
language, namely the way in which lexicogrammar construes our experience of the world
in the guise of multidimensional matrices or grids. This is an important feature of
language as a semiotic system, an inevitable consequence of the principle of
arbitrariness: since the forms of expression are arbitrary, they impose discontinuity on

98

2. Overview of the general ideational potential

the content. We have to decide about any given instance whether it is singular or plural,
temporal or causal, possible, probable or certain; whether it is bus or a van, smiling
or grinning, cloudy, misty or foggy. But the semantic categories themselves (seen from
above, as it were) are much more fluid and indeterminate than their realizations in
wording imply. Hie notion of semantic space allows us to adopt a complementary
standpoint from which we can view these phenomena topologically, bringing out the
inherently elastic quality of the dimensions involved, and gaining a deeper insight into
the semogenic processes by which the meaning potential is ongoingly remoulded in the
history of the system, of the user, and of the text.
2.12 Construal in the grammar: summary of
evidence

grammatical

In Chapter 1, Section 1.8, we discussed the nature of grammatical evidence for the
organization in the ideation base, pointing out that we have to consider both oven and
coven (cryptotypic) features and that we have to take account of reactances from all three
metafunctions. In Chapters 3 to 5, we shall discuss sequences, figures and elements in
further detail; meanwhile to conclude the present chapter we summarize the salient
features in the grammatical organization of figures and elements.
2.12.1 Types of figure

There are four types of figure doing & happening, sensing, saying, being & having.
Broadly speaking, these are constructed in the grammar as follows:
(i) Within the ideational metafunction, each is realized congruently by one
particular transitivity type: doing & happening ^ material, sensing ^ mental,
saying ^ verbal, and being & having ^ relational. These have various
reactances, such as the number and nature of participants and the unmarked
present tense selection (see Table 2(10)).
(ii) Relating to the interpersonal metafunction: in any given register there may
be typical correspondences between the type of figure and the speech function;
e.g. in procedural registers, material clauses are typically imperative, relational
ones indicative.
(iii) Relating to the textual metafunction: different types of figure are presumed
in different ways and have different potential for textual prominence; e.g. only
material clauses are substituted by the pro-verb do (toAvith).

2.12 Construal in the grammar: summary of grammatical evidence

99

2.12.2 Types of element

Tbere are three types of semantic elements serving in figures processes, participants
and circumstances. Broadly speaking, these are constructed in the grammar as follows.
(i) Within the ideational component: they are realized by different classes of
units:
process verbal group
nominal group
group
participant nominal
circumstance adverbial group; prepositional phrase
Participants tend to be inherent elements of a figure; circumstances are typically
optional.
(ii) Relating to the interpersonal metafunction: participants can serve as
Subject; circumstances and processes cannot Furthermore, participants and
circumstances can serve as WH-etements, but processes cannot (if the process is
being questioned, a participant element has to be construed as a Range: what...
do?\
(iii) Relating to the textual metafunction: participants and circumstances can
both readily serve as Theme (though their potentials differ); processes only
rarely, other than in imperative clauses. Participants and (more restrictedly)
circumstances can serve as referables identified by referring expressions, but
processes cannot as with WH-interrogation, they have to be construed as a
Range together with the pro-verb do as Process: do it/that (see Halliday &
Hasan, 1976: 125).
These properties are summarized in Table 2(11).
The view from grammar brings into relief a number of relevant factors relating to
participants and circumstances. There are certain special subcategories of these, in the
grammar, which are distinguished by the fact that they embody features of interpersonal
or textual meaning:
(1) interpersonal: questioning
interrogative: who, what, when, where, how far, how long, how, why
(2) textual: cohesive
(i) referring, personal: he, she, it, they; demonstrative: this, that, now,t then,
then,
here, there, thus
(ii) generalizing: lexical items such as person, creature, thing, stuff, affair

Table 2(10): Grammatical (ideational) properties of different types of figure

rote in sequence
projecting clause:
nature of Medium

nature of second
participant
(Agent or Range)

material

mental

verbal

relational

Actor, Goal:
participant (simple or macro)

Senser
participant: conscious

Sayer
participant: symbol source

Carrier, Value:
participant or projected figure;
circumstance

nominal group;
clause: non-finite

nominal group

nominal group

Acton Range:
participant

nominal group
Attribute
Beneficiary

nominal group;
clause: projection

Verbiage:
participant
linguistic
nominal group

(condition or result)
Recipient or
Client
Place-directed

present-inpresent

various
(if ascriptive)

[except for o speciai subtype]


Receiver
Matter

somewhat restricted
circumstances
directionality:
two way or
oneway(-)
TENSE:
unmarked present

Phenomenon:
participant or projected figure

nominal group;
clause (projection or expansion;
finite or non-finite)
adv. group;
prep, phrase
[Depends on
subtype]

simple present

Matter

Table 2(11) Grammatical evidence for differentiation of elements

(i) ideational

process

participant

circumstance

TRANSITIVITY: role in
figures

Process

participant roles: Agent,


Medium, etc.

circumstantial roles: Cause,


Location, etc.

class of unit

verbal

nominal

adverbial;
prep, phrase

DEIXIS

TENSE
(temporal)

MODIFICATION

grammatical items

INDIVIDUATK^

DETERMINATION (reference
spatial)
lexical & grammatical items

lexical & grammatical items

(except for circumstantial


gangs)

Proper names
as well as class names
(ii) interpersonaJ

MOOD:
WH-SELECTION

(excqrtifcoostnjedasRange:

fawkat&c)

who, what, which

when, where, how long, how


far, how, why ...

Wh/
MOOD:
Subject/
(iii) textual

THEME:
Theme/
REFERENCE

{veryaarrfv,
ttWt &r tefcragve clauses)
(except ff <x*$tnK<f a$ Hangs;
<fc*f&c>
he/she, it, this, that, ...
[Note range of subtypes]

(but more limited:)


now, then, here, there, thus,

2. Overview of the most general ideational potential

102

These are significant in their own right because they are critical to the construction of
discourse: the textual ones provide internal cohesion, while the interpersonal ones
construe dialogic speech roles. They have a further significance in that they reveal by
reactance the major subclasses within the general classes of participant and circumstance,
as shown in Table 2(12).
Table 2(12): Grammatical evidence for differentiation of participants and circumstances

participant

circumstance

conscious

non-

animal

conscious

institution
obiect
substance
abstraction

interpers.
WH-TYPE

textual
REFERENCE

who?

s/be/ thet

what?

it/ they

LEXICAL
COHESION:
general noun
person, fellow,
chap
animal, creature,
beast
thing, object
stuff
matter, affair,
business

time

when?

now/ then

place
distance
duration
manner
cause

where?
how far?
how long?
how?
why?

here/ there
thus, this way

These special subcategories have the effect of construing other elements as "referable"
that is, of enabling them to retain their semiotic identity for subsequent access and
hence as it were authenticating them (e.g. Don't give me any more of that peanut butter! I
can't stand the stuff.). This applies primarily, though not exclusively, to participants (cf.
Webber, 1987, on reference to phenomena other than things). At the same time it allows
us to recognize not only participants of the 'simple' type (cf. Plate 5) but also the 'larger'
elements known as macro- and meta-phenomena. Macro-phenomena are figures
dowraanked to function as ordinary elements; meta-phenomena are figures projected as
elements of a second order (see Chapter 3, Section 3.3 below). Halliday & Hasan (1976),
where reference to macro-phenomena is called extended rfrence, and reference to metaphenomena is called reference to facts, cite ambiguities which bring out the difference
among elements of these various kinds:
(i) extended reference to macro-phenomenon:
They broke a Chinese vase.
( i ) That was valuable. (phenomenon: thing the vase)
(ii) That was careless, (macro-phenomenon the act of their breaking of the
vase)

2.12 ConstruaJ in the grammar: summary of grammatical evidence

103

(ii) reference to tact1 to meta-phenomenon


It rained day and night for two weeks. The basement flooded and
everything was under water.
(i) It spoilt our calculations, (meta-phenomenon (fact)the fact that it
rained so much upset our predictions)
(ii) It spoilt our calculations, (macro-phenomenon the act of raining
destroyed physical records)
Note how this relates to the shift of perspective from figure to participant (cf. Section
2.11.3 [3]): a gure such as 'catch + mouse' becomes referable as a macro-phenomenon
(Your cat's caught a mouse - It's never done that before), and by the same token can
enter into participant-based construal through the system of modification (Table 2(11)):
(cats are) creatures that catch mice.
This brings to an end our overview of the general ideational potential of the semantic
system. In the next four chapters we shall fill out further details of the resources of the
ideation base. We begin by mapping out the resources for construing experience in the
congruent mode:
sequences are discussed in Chapter 3;
figures are dealt with in Chapter 4; and
elements are described in Chapter 5.
In Chapter 6, we explore the metaphorical mode of construing experience interpreting this
mode as a reconstrual of the model construed in the congruent mode. Rounding off this
part of the book, we then sketch the outlines of the ideation base of Chinese and raise the
issue of variability across languages in how they construe experience.

3, Sequences
We introduced sequences in Chapter 2, Section 2.2 above. Here we shall characterize them
in more detail, increasing the delicacy of the account to cover both the types of sequence
and their mode of organization.
A sequence constitutes a model of how figures can be related. One prominent form
of this relationship, which has been foregrounded in various guises in science and logic,
is that of cause & effect, whereby experience is given a causal interpretation. But that is
only one among many such possible relationships, which taken together can be said to
constitute the logic of natural language.

3.1 Natural logic of sequences and propositional logic


Sequences might be said to constitute the 'natural logic' equivalent of propositional logic
that is, the evolved system for reasoning about relations of cause, conditionality,
etc. from which propositional logic has been derived by design. Thus we have parallel
series such as:
propositional logic

natural logic

p&q
pvq
p > q

pandq
porq
p so q; if p then q

But, as has often been pointed out, the two are not translation equivalents; for example,
material implication (p - q) applies even when the rendering in ordinary language seems
odd, and disjunction in logic is either inclusive or exclusive whereas natural disjunction is
non-committal. Since propositional logic is a designed system, its relations are codified
and defined (typically in truth-functional terms1). In contrast, sequential relations have
evolved. A certain type of relation will have a core the prototypical representatives of
that type; but there will also be more peripheral representatives and 'grey areas' where one
type shades into another.
There is another important difference between propositional logic and the natural logic
of sequences. While there is only a very small handful of truth-functional connectives in
propositional logic (conjunction, disjunction [exclusive or non-exclusive], implication),
there is a very wide range of sequential relations in language all the more specific
varieties of projecting and expanding. A summary of these is given in Table 3(1).

Propositional logic is interpersonally invariable. Unlike natural logic, it is only concerned


with statements, or rather since language is concerned with validity rather than truth
with the philosophical version of what are statements in natural language.

105

3.1 Natural logic of sequences and prepositional logic

Table 3(1): The types of natural and prepositional logic


propositional
logic

natural logic of sequences


projection
expansion

say
think 2
reiterate
(elaboration)
add
(extension)

qualify
(enhancement)

addition

conjunction

alternation
replacement
time

disjunction
_.

space
cause-condition
reason
purpose
concession
condition

..
..
..
..

manner

--

..

implication

...

Both the difference in scope and the difference in 'definability' can be explained in
functional terms. Sequences have evolved in the interpretation of human experience in
general; consequently, they have to be flexible and powerful enough to cope with a large
amount of variation, and the implicit 'definition1 of each relation (i.e., its location in the
semantic system along various dimensions) is the evolving distillation of innumerable
instances where it is invoked (for a revealing account of how sequences may construe
rationality in everyday talk, see Hasan, 1992). In contrast, the truth-functional
connectives of propositional logic have been designed for a very restricted purpose the
kind of deductive reasoning western philosophy came to focus on and their definitions
are fixed by reference to values of "true" and "false" (by means of truth tables).
In a restricted register, such as those of weather forecasts and recipes (see Chapter 8
below), it may be possible to define sequential relations more precisely; but that is only
because these registers are special cases, rather in the same way that the applications of
propositional logic are special cases.
One might try to characterize sequences by reference to propositional logic; this would
give some indication of what type of knowledge representation system is embodied in

Projecting figures of the 'think1 type 'know1, 'believe', 'want have been represented
outside standard logic within intensional logic.

106

3 Sequences

this domain of ideational semantics. However, such a characterization would be likely to


distort our understanding of sequences, because sequences cover and organize a
considerably larger domain of the ideational semantic space than prepositional logic
attempts to do.

3.2 Expansion and projection


Instead of referring to formal logic, let us ask what kind of construal of experience is
embodied in semantic sequences. Throughout the semantic construal of human
experience, there is a differentiation between two orders of reality: between the everyday
reality of our material existence on the one hand and on the other hand the second-order
reality that is brought into existence only by the system of language. This is a contrast
between semiotic phenomena, those of meanings and wordings, and the first-order
phenomena that constitute our material environment. (Note that the linguistic processes
themselves, as apprehended by our senses, are part of the first-order reality; second-order
reality is formed of the meanings and the wordings that these processes bring into being.)
This differentiation is embodied in relations of sequence in the following way. Either
a sequential relation expands one figure by adding another one to it, the two still
remaining at the same phenomenal level; or the sequential relation projects one of the
two figures onto the plane of second-older, semiotic phenomena, so that it enters the
realm of metaphenomena (meanings or wordings). This is the distinction between
expansion and projection that we introduced briefly in Chapter 2, Section 2.2. For

example, expansion: highs will be mid-SOs to mid~90s


but parts of Texas could
"parts of Texas could reach the
reach the 100s; projection: the forecast predicts
100s". We can explore this distinction through the conventions of comic strips.
Expansions are typically represented pictorially by means of consecutive frames 'x
then y', 'x so y', 'x meanwhile y', and so on (the relationship is sometimes spelt out in
writing, e.g. Meanwhile... ). We can even identify the three different kinds of expansion,
those of elaborating, extending, and enhancing. When further detail is given in a picture
of a magnifying glass, this is elaboration; where two frames are joined to form a
continuous picture, this suggests extension; otherwise, consecutive frames display a
relation of enhancement But in all cases the representation is pictorial throughout. In
contrast, projections involve two modes: the projecting figure is represented pictorially as
the symbol source, but the projected figure is represented linguistically and this secondorder reality is typically named within a 'balloon*. (There are two types of balloon; see
further below.) The two types of relation can be shown as in Figure 3-1.
We could say that the world view that is constructed in this way is (quite reasonably,
from our point of view) focussed on humans, with human consciousness occupying a
privileged place. Thus typically only humans can project into second-Older reality.
However, since human consciousness is the locus of semiotic activity, it has the power

3.2 Expansion and projection

107

of interpreting as metaphenomenon that which is manifested by some other, non*


conscious symbolic source. Thus while "sensing" (that is, semiotic activity that is
unmanifested, like thinking) does require a human senser, saying can be ascribed to a nonhuman as well as to a human sayer (cf. he thinks
the moon is a balloon vs. he
says/ the book says

the moon is a balloon). Figure 3-2 shows the relationship

between the two orders of metaphenomena and the presence of consciousness in the act of
projection.

Fig. 3-1: The two basic types of sequence

108

3. Sequences

Fig. 3-2: Consciousness in projection and the two levels of metaphenomena


In the next two sections we shall consider projection and expansion in turn.
3.3 Projection
3.3.1 Two levels of projected content: ideas and locutions

We have suggested that the relation of projection sets up one figure on a different plane of
reality we refer to this as the second-order or semiotc level. This second-older level of
reality is the content plane of a semiotc system (cf. our characterization of the
linguistic system in Chapter 1, Section 1.2 above). That is to say, the projected figure is
projected in the form of 'content1: see Figure 3-3.

109

3.3 Projection

Fig. 3-3: A figure construed as content by projection


We have seen that the content plane is stratified into two levels semantics (the
level of meanings) and lexjcogrammar (the level of wordings). Consequently, we would
expect projections to be located at either or both of these levels, and this is indeed what
happens: a projected figure is either a meaning or a wording: see Figure 3-4.
In (i) the 'content1 of his thinking is the meaning. It remains internal to his
consciousness and unrealized, i.e. unworded. In (ii) the 'content* of his saying is the
wording. It has been externalized, realized as a wording. We will refer to these in the
context of projection as (i) ideas and (ii) locutions. The contrast between these two
statuses of projected figures is typically reflected in the shape of balloons in comic strip
projections: see Figure 3-5.

110

3. Sequences

Fig. 3-4: Projection at the two levels of content

Fig. 3-5: Indication of projection in comic strips


Let us look more closely at these two modes of projecting, in both of which the
projecting figure represents symbolic processing, the processing that brings the other
figure into symbolic existence. Either the projection takes the prototypical form of

3.3 Projection

111

semiosis: it is presented as verbal, shared, an exchange or joint construction of meanings


(e.g. Harriet said 'me feed cat?'); or it is fashioned into a derived semiotic form, unshared,
interiorized, and without any meaning being exchanged (e.g. Harriet thought 'me feed
cat?'). In the first, the projecting figure is one of saying; the projected is referred to as
locution. In the second, the projecting figure is one of sensing; the projected is referred
to as idea. Ideas are projections which are sensed, locutions are projections which aie
said.
But there is another variable intersecting with the above. We pointed out in Chapter
2, Section 2.2 that whenever two figures are related in a sequence, they may be either
equal or unequal in semantic weight. We illustrated this interdependency system with
reference to expansion; but it applies also to projection. The projecting and the projected
figures may have equal status in the sequence: this relation is that of quoting, as in
Harriet said/thought "Shall I feed the cat?". Or they may have unequal status: this
relation is that of reporting, as in (Harriet asked/wondered) whether she should feed the
cat. We can relate this to the fact that projection is the creating of a second-order reality.
In reporting, the status of the two parts is unequal the projected figure is dependent
on the projecting: hence the projected figure is clearly construed as belonging to a
different, second-order plane of reality a reality that is made of meaning, as it were. In
quoting, on the other hand, the two have equal status as independent figures; the
projected figure is thus projected as if it was still pan of the same first-order reality.
There are two consequences of this. The first is that, in quoting, the form of wording
is still that of the first-order realm of experience: "Shall I feed the cat ?" taken by itself,
there is nothing about this to show it is a projection. In reporting, on the other hand, the
form of wording is clearly marked as being of the second order whether she should feed
the cat has lost its first-order semantic features and hence has to be projected by words
that specify its particular speech function: asked/wondered, not said/thought.
The second consequence is that although the two variables, locution/ idea and quoting/
reporting, can combine in either of the two possible alignments, there is a natural default
condition, which is that locution goes with quoting and idea goes with reporting, as
shown in Figure 3-6. The reason for this is clear where the first-order phenomenon is
one of saying (prototypically shared semiosis), the projected figure can be presented as if
it was also of the same order: Harriet said + "Shall I feed the cat?" (reversible as "Shall I
feed the cat?" + said Harriet). Where the first-order phenomenon is one of sensing
(unshared semiosis), the projected figure has no counterpart on the first-order plane of
experience and cannot be naturally presented as if it had; so, Harriet wondered + whether
she should feed the cat. But there is always the possibility of semogenic extension by
cross-coupling; so we also find the corresponding marked alignments: locution/ report
Harriet asked + whether she should feed the cat, and idea/ quote Harriet wondered + "Shall

112

3. Sequences

feed the cat?" - the latter being again reversible. See Halliday (1985: Chapter 7) and
Nesbitt & Plum (1988).
projecting: type

projected: status
quoted

reported

locution
idea

Fig. 3-6: Default alignments (projection)


Projections of ideas have played a special role in the extension of standard logic
within intensional logic (or, more resthctedly for knowing and believing, within
epistemic logic) to allow for reasoning in this domain. One of the features of the
projection of ideas that has received special attention in logically-oriented approaches to
meaning is its "referential opacity". If a speaker knows that Henry is the king, and
construes somebody else's belief as Thomas thinks Henry is a nice man, the referring
expression Henry cannot be replaced by the king although the speaker knows they have
the same referent' 'Thomas thinks the king is a nice man' cannot be inferred from
'Thomas thinks Henry is a nice man' since the identity of Henry and the king has not
been established in the projected world of Thomas's consciousness.
The projection of ideas has also played an important role in hearer modelling in AI
and computational linguistics. From a systemic-functional point of view, hearer
modelling corresponds to one aspect of the interaction base: the interactants in a dialogue
have to be able to assess and model the way their experiences diverge. This is one
measure of the overall interpersonal distance between them. In hearer modelling, this
tends to be construed ideationally by means of cognitive projection, where the projected
idea is represented as a separate partition or space within the overall knowledge (see e.g.
Allen, 1987: Ch. 15; Kobsa & Wahlster, 1989; Bafflm & Wilks, 1991): 'the speaker
believes
the hearer knows
...'. Such models may involve long projecting
sequences (often called "nested beliefs"), as in the example in Figure 3-7 'a I believe
'd it's boiling'.
' that you believe
'y that I believe
In general, there is no temporal constraint between the projected and the projecting
figures: the projected figure may be past, present or future relative to the projecting one.
The dimension affected by projection order of reality, not time. However, there is a major
dichotomy between two types of projection which does have temporal implications. To
explore it, we have to look briefly at a fundamental interpersonal category, that of mood.

113

3.3 Projection

Fig. 3-7: Sees of cognitive projections


3.3.2 Propositions and proposals

Interpersonal semantics is centrally concerned with varieties of symbolic exchange


(Halliday, 1984b, 1985: Ch. 4). Here, as in other places in the meaning base, the system
is organized in such a way that it creates a difference between non-symbolic reality and
symbolic reality, between phenomena and metaphenomena. Hie "commodity" that is
being exchanged in interpersonal dialogue is either semiotic or material: it is either one
that is construed by language itself information or it is one that exists
independently of language goods & services. In the first case, language
constitutes the exchange; in the second, it facilitates the exchange of a nonlinguistic commodity. For example:
giving
demanding

information proposition
statement
I've

done the laundry.

question
Have you done the laundry?

goods & services proposal


offer
I can do the laundry.

command
Do the laundry!

Information is either given or demanded; in either case it is encoded as a


proposition.3 Similarly, goods & services are either given or demanded, but the
linguistic act is a mediating one, specifying the (typically non-linguistic) action that

The term proposition is not used here in its sense in logic, but in contrast with proposal (see
immediately below).

3. Sequences

114

embodies the exchange, as an offer to do or a demand that something should be done.


This is a proposal for a deed, one that commits either the speaker (giving) or the
addressee (demanding); together with a third type, which is a combination of these two,
viz. suggestion: Let's do the laundry.
The interpersonal system for dialogic interaction thus creates a fundamental distinction
between 'content* as proposition and 'content* as proposal. This distinction is then
reflected in the ideational system of projection: a figure is projected in one or other of
these two modes, as a proposition or as a proposal. The two categories combine freely
with ideas and locutions:
content: meaning
(idea)
content: wording
(locution)

proposition
that he had
he said
done the laundry
"I have done
he said
the laundry"

proposal
she told him
to do
the laundry
"do
she said to him
the laundry! "

Projected proposals are non-actual, or uninstantiated: that is, the occurrence of a


projected proposal is always future in relation to the figure that projects it. In contrast,
propositions are actual, or instantiated: that is, the occurrence of a proposition is located
in actual time (which may be past, present or future).
Grammatically the distinction between propositions and proposals is constructed as
follows. When the projection is reported, propositions are realized by finite bound
clauses, i.e. clauses that select for primary tense or modality; and proposals are realized
by irrealis (infinitival) non-finite bound clauses taking the contrasting examples from
above: (he said) that he had done the laundry : (she told him) to do the laundry. When the
projection is quoted, propositions are realized by indicative clauses, i.e. clauses that select
for primary tense or modality; and proposals are realized by imperative clauses taking
the contrasting examples from above: (he said) 7 have done the laundry" : (she said to
him) "do the laundry". These realization^ patterns are summarized below:
proposition

proposal

reporting

bound: finite

bound: non-finite: irrealis


(she told him) to do the laundry/
(she wanted) him to do the laundry

quoting

(he said/ thought) that he had done


the laundry
(she asked/ wondered) whether he had
done the laundry
free: indicative
(he said) "I have done the laundry"
(she said/ asked) "have you done the
laundry?"

(she said to him) "do the laundry! "

free: imperative

As was said above, in quoting, where the projected clause retains its mood, the general
verb say can be used whatever speech function is being projected In reporting, on the
other hand, the projected clause is no longo* specified for mood; its speech function is

3.3 Projection

115

signalled by the verb in the projecting clause (asked, ordered, etc.). This gives the
projected element more of an ideational status (cf. its treatment in traditional grammar as
"object" of the projecting verb), and opens up the way to a series of agnate expressions
such as (the king ordered) "Execute him!" / thai he should be executed / him to be
executed/him executed/his execution.
There is thus a parallel between reported propositions/ proposals and quoted ones. But
with offers (proposals: giving goods-&-services) there is a difference in respect of
realization. Offers, in English, aie not giammaticalized in the mood system; that is,
while the other categories, statements, questions and commands, have corresponding
mood categories in tbe grammar (declarative, inteirogative and imperative), offers do not.
They may be realized by any of tbe mood categories; for example:
declarative: I can do the laundry.
interrogative: Shall I do the laundry?
imperative: Let me do the laundry!

Significantly, the indicative clauses realizing offers are modulated; they select for an
imperative modality of readiness or obligation (see e.g. Halliday, 1985: Section 10.4).
Quoted offers naturally retain the property of being variously realized in the mood
system:
declarative: She said: "I can do the laundry."
interrogative: She said: "Shall I do the laundry?"
imperative: She said: "Let me do the laundry!"

However, reported offers can always be realized in the same way as reported commands
as in the case of reported propositions, the distinction in orientation between giving
and demanding is realized by the projecting clause; for example:
command: she told him

to do the laundry

offer she offered (promised; threatened)

to do the laundry

The category of repotted proposal is thus realized generally as a perfective non-finite


clause when it is reported.

We saw that quoted offers may be realized by modulated indicative clauses. This would
seem to be an exception to the generalization that indicative clauses realize propositions.
However, it is a principled one: the type of indicative clause involved is precisely the
type that lies closer in the interpersonal clause grammar to imperative clauses
modulated indicative clauses, i.e. those with an imperative modality. As can be expected,
this is then also a possible realization of commands; it is a metaphorical strategy for

3. Sequences

116

expanding the meaning potential, typically to vary the tenor between speaker and listener.
For example:
she told him: "Do the laundry!"; "You should do the laundry."
she asked him: "Could you do the laundry?"
Commands can be reported in the same way:
she told him
she told him
she asked him

to do the laundry.
he should do the laundry.
whether he could do the laundry.

To sum up the distinctions discussed so far. In a projecting sequence, a figure of


thinking or saying projects another figure as an idea or a locution, either of which may be
a proposition or a proposal: see Figure 3-8.

Fig. 3-8: The possibilities of projecting sequences


We can now turn to the other primary type of sequence expansion.

3.4 Expansion

117

3.4 Expansion
Expansion is a highly generalized type of relation, whereby one figure is joined lgicosemantically to another figure by a relator to form a sequence of the same order of reality.
It will be easier to characterize expansion at the next step in delicacy, in terms of its
immediate subtypes: elaborate (reiterate), extend (add), and enhance (qualify):
(i) elaboration is a (partial) identity relation between figures: one is identified
with another with a difference in perspective (it matters a lot; it plays an
important role) or one is included under another as an example (it plays an
important role; e.g., it provides the infrastructure). These are clearly related to
one another: identity is the limiting case of inclusion and inclusion is partial
identity.
(ii) extension is an additive relation between figures: a sequence is made bigger
by the addition of another figure. This may involve pure addition ('and': he is too
young and he doesn't speak the language) or addition with an adversative feature
('and yet': he speaks the language but he is too young ). As a variant of addition,
we also have alternation (he is too young or else he is just immature).
(iii) enhancement is a circumstantial or qualifying relation between figures: it
is, in a sense, extension plus a circumstantial feature 'and' + time ('and then',
'and at the same time', etc.), 'and* + manner ('and in the same way', 'and
likewise'), 'and' + cause ('and therefore', etc.), etc.: it is autumn, so the leaves are
turning brown.
Expansion can be thought of as construing another dimension of experience, so that
superimposed, as it were, on the construal of a figure a basic fragment of experience in
the form of a quantum of change (event, action, behaviour &c.) is the construal of
a nexus between two figures, such that one such fragment is non-randomly (i.e.
meaningfully) cumulated with another,
From one standpoint the limiting case of expansion would be accumulation in a
temporal sequence (hence our general term "sequence" for the product of this construal): 'a
happens, then x happens'. This gives value to a as the temporal circumstance of x.
From this we could derive a wide range of more complex enhancing relations: variations
on the simple temporal sequence ('after, before, at the same time, immediately after, &c.)
and further circumstances such as cause, condition, concession, and their subcategories.
We shall not try to enumerate these here; they arc familiar as categories at the level of
lexicogrammar (see Halliday, 1985: Ch. 7; Matthiessen, I995b: Ch. 3). These ate
'enhancements', multiplying one figure by another, as it were. But figures may also be
added to one another, making them part of the same story without assigning any kind of

3. Sequences

118

logical priority to either: 'x as well as/ instead of/ in contradistinction to a '. We have
refened to these as extensions. And there is the third type of accumulation where the
logical relation is that of 'equals': 'x is the same figure as a'. Here at this end the limiting
case is a simple repetition; this may be further elaborated in such a way that one figure is
reworded as another, or else further clarified or brought out by an example.
There is no sharp line between a figure and a sequence of figures: a quantum of
experience is not defined before it is construed, and the grammar rather sets up a dine
from sequence to simple figure. (This is, in fact, a variable across languages; we
shall return to this point from a cross-linguistic point of view in Chapter 6.) For
example:
1 sequence of 2 figures:

Tarzan pulled the rope; so it broke.


Tarzan made the rope break (by pulling it)
Tarzan broke the rope.

1 figure:

The rope broke.

The following passage of spontaneous conversation exemplifies the principal


categories of expansion (note than it also contains one example of projection):
levels of nestin e In sequence

xp

+2
+2
a

+2

a
xp
1
=2

fleures In sequence
[1] if you go into a bank

[2] if you walked through to Barclays for instance

2
1
"2

[3] and said


[4] will you look after my investments
[5] or even if you went off to Hambros or one of the
famous banks of that nature
[6] they would every year value your securities
[7] as we do of course
[8] but they would charge you
[9] they'd send you a bill for a percentage of what they
were worth

KEY: parataxis [equal]: 1, 2; hypotaxis [unequal]: a, f); elaboration: =; extension: +;


enhancement: x. (Taken from Svartvik & Quirk, 1980:430.)
Relations of expansion are typically realized in the grammar by conjunctions or
conjunctive expressions linking a pair of clauses, either paratactically or hypotactically
(e.g. that is, in other words; and, but, or, also, besides; so, yet, then, when, if, because,
unless). Some of these may realize more than one category; for example, but may be
adversative 'and yet' (extending), or concessive 'and in spite of this1 (enhancing); while
may be additive 'and in addition' (extending), or temporal 'and at the same time'
(enhancing); or may be alternative 'or else' (extending: alternatives in the external world,
like take it or leave it), or restating 'in other words' (elaborating: alternatives in the world

3.4 Expansion

119

'internal' to the discourse, like they one reduced to the smallest size, or microminiaturized. ). Ova-laps in realization of this kind show that the primary categories we
have set up do in fact shade into one another; in particular, extending in some sense
occupies a space intermediate between elaborating and enhancing, and shares a fuzzy
borderline with each (cf. Chapter 2, Section 2.11.1 and Chapter 13, Section 13.3).
Finally, the distinction between expansion and projection is less determinate than we
have suggested. The iogico-semantic relation of condition, which is prototypically
construed as a form of enhancement, could also be construed as a kind of projection; and
this is also brought out in the grammar. Conditions specify a potential and actualizable
but non-actual situation. This potential situation can also be set up through projection:
If the power supply fails, what's the best thing to do?
Supposing the power supply fails, what's the best thing to do?
Say the power supply fails, what's the best thing to do?

Words such as supposing

and assuming are verbs of projection which have come to

function as conjunctions in conditional figures; while other words such as imagine and
say retain more of their projecting force. Sometimes even variants of the same word have
come to differ a little in their location on this cline: for example, suppose and assume
seem closer to projection than their corresponding participial variants. This is an
uncertain region in which a figure hangs in the air, so to speak, suspended between the
hypothetical material plane and the semiotic one.

3.5 The relational character of sequences


In the overview, we introduced two distinct types of structure the relational
organization of sequences, and the configurational organization of figures (see Chapter 2,
Sections 2.1, 2.2 & 2.3). In contrast to figures, sequences are not constructional units.
We con specify the range of projecting and expanding relations available for further
developing a sequence, but we cannot specify where a sequence has to come to an end
that is, we cannot specify a sequence as a unit whole with a conventional configuration of
parts. Thus if we have expanded one figure we can always repeat the operation:
A then B
A then B then C
A then B then C then D . . .
In contrast, a figure is a unit with a finite number of elements:
B follows A

120

3.Sequences

Hence a sequence can be indefinitely complex, whereas a figure cannot. For instance,
consider the sequence from the culinary domain (Highton & Highton,4 p. 156) shown in
Figure 3-9.

Fig. 3-9: Sequence of culinary operations

N.B. Highton & R.B. Highton. 1964. The Home Book of Vegetarian Cookery. London
Faber and Faber.

3.5 The relational character of sequences

121

This sequence is expanded step by step: one 'operational' figure is qualified by another,
which is in turn qualified by another, and so on. As the example illustrates, there may be
internal nesting. That is, a sequential relation may relate not only to a figure but also
to a sequence of figures. The sequence of culinary operations in Figure 3-9 is worded in
the text as follows:
Prepare a whole head

and

carefully. Make a stuffing

cook

for

only 10 minutes.

Drain

(see section on stuffings) which you

like, and which will go well with cauliflower. Place a round of


it about 1-inch thick on the bottom of a fire-proof serving-dish.
Fill all the spaces between the stalks with the rest of the
stuffing and place the cauliflower stalk downwards on the
stuffing, if there is any stuffing left use it to fill in around
the base.
Cover the cauliflower with a greased paper ad bake in a moderate
oven 325 F. - 350 F., M3-4 for 30-40 minutes. That is, until
the cauliflower is cooked and the stuffing is hot. Remove the
greaseproof paper and cover with a coating sauce to complement
the stuffing, such as: . . .
The example illustrates a common type of sequence used to encode experience with
operations a procedure. The relations in a procedure are typically expansions: (i)
additive: addition and alternation; (ii) qualifying: temporal and causal conditions. A
procedure, like any sequence, is potentially infinite. The only reason the procedure above
stops is that the cauliflower surprise is ready to serve. The procedure does not end because
a unit boundary has been reached; on the contrary, because the structure is relational, the
procedure could be developed indefinitely.
One important type of procedure is the algorithm of computational specifications.
Here is an example of a pre-computational algorithm for searching a tree, breadth-first,
construed linguistically as a procedural text (Ban- & Feigenbaum, 1981: 56-7).
1. Put the start node on a list, OPEN, of unexpanded nodes.
2. Remove the first node, n,

from OPEN.

3. Expand node n - generating all its immediate successors and,


for each successor m,

it m

represents a set of more than one

sutproblem, generating successors of m

corresponding to the

individual subproblems. Attach, to each newly generated node, a


pointer back to its immediate successor. Place all the new nodes
that do not yet have descendants at the end of OPEN.

122

3.Sequences

4. If no successors were generated in (3), tta*n


a. Label node n unsolvable.
b. If the unsolvability of n

makes any of its ancestors

unsolvable, label these ancestors unsolvable.


c. If

the start node is labeled unsolvable, exit

with

failure.
d. Remove from OPEN any nodes with an unsolvable ancestor.
5. Oth*rwi*, if any terminal nodes were generated in (3), then
a. Label these terminal nodes solved.
b. If the solution of these terminal nodes makes any of
their ancestors solved, label these ancestors solved.
c. If the start node is labeled solved, exit with success.
d. Remove from OPEN any nodes that are labeled solved or that
have a solved ancestor.
6. Go to (2).

The default relation 'and then* is left implicit in this text; conditioning ('if; if not') is
made explicit. Algorithms are often displayed by means of both verbal instructions and a
simple diagrammatic semiotic, the flow chart.
Thus sequences are formed by binary logico-semantic relations which may relate either
single figures or sequences of figures. Certain relations may impose constraints on the
phenomena being related; for example, a projecting relation can only obtain when the
first figure is one of sensing or saying. But the range of diffrait logico-semantic
relations is highly varied, so that constraints tend to be specific to particular subtypes.

3.6 Sequences and text


Sequences impose a certain order on our experience in terms of the relations that connect
happening with another. Hence sequences can be used to store information about the
world in the form of organized text 'this is how to change tyres on your car', 'this is
how to make cauliflower surprise', etc.. Such texts often fall into a clearly recognizable
text type, such as procedures, proofs, explanations, and episodic narratives. Not all texts
are as highly regulated as these; but it is usually possible to make some prediction about
the kinds of sequence, and the complexity to which sequences extend, in most of our
culturally recognized modes of discourse.
Texts and sequences are of the same order of abstraction; both are semantic
phenomena. A text is a piece of language that is functional in context. It draws on the
ideational meaning base but it involves the full metafunctional spectrum; i.e. there are
interpersonal and textual contributions as well. Since text draws on the ideational

3.6 Sequences and text

123

meaning base, sequences are one principle for organizing text. For example, the culinary
procedure for making cauliflower surprise constitutes one text Many text types are
heavily influenced by patterns in the meaning base they can be seen as 'macro-figures',
i.e. as expansions of figures by means of logico-semantic relations. This is not to say
that the relationship between organization in the meaning base and discoursal
organization is always one-to-one even when a text is organized according to an ideaiional
sequence. In particular, a text may leave to be inferred certain steps that would be
specified hi the sequence in the meaning base (e.g. to make explicit the inferential
processes involved).
As already noted, text organization also draws on the interaction base; and there are
certain parallels with the ideation base with respect to sequences. The interaction base
certainly has text fonning resources that are uniquely interpersonal. In particular, it has
the resources for the collaborative exchanges that are embodied in the notion of
interaction for producing dialogue jointly by means of coofdinated moves alternating
between the interlocutors (see e.g., Bateman, 1985; Berry, 1981; Halliday, 1984b;
Martin, 1992). But these interpersonal moves may also form sequences of moves in a
way that is similar to the formation of sequences of figures in the ideation base. Typical
examples involve motivating condition ('I invite you to accept x, if you want x'; for
example: If you're thirsty, there's beer in the fridge) and evidence (1 think, infer/ you
should believe x because y1; for example: John's in Germany because I just talked to
him).
Certain text types are heavily influenced by such interpersonally oriented sequences.5
This is the case with persuasive text as in the following advertisement:
[1] To get to the top you have to go to the right school.
[2] The first thing we do with all new pilots, no matter how
qualified, is send them bacX to school. [3] They must pass
advanced and rigorous regular checks before they can fly with us.
[4] Exacting standards. [5] Another reason why we're Australia's
leading airline.
[6] Australian Airlines
[7] "You should see us now."

In terms of Rhetorical Structure Theory (Mann & Thompson, 1987; Mann, Matoneasen &
Thompson, 1992), the relations here include enablement, motivation, and evidence
interpersonally oriented relations (see Mann & Matthiessen, 1991).

124

3. Sequences

Here the nuclear passage in the advertisement is [7] getting the reader to feel inclined
to see Australian Airlines; but to achieve this, [1] through [5] are also included. They
stand in an interpersonally oriented reason relationship to [7] a motivation
relationship. (Hiere are also ideationally oriented sequence relations, of course; for
example, [3] is internally organized as a temporal sequence they must pass advanced
and rigorous regular checks
before they can fly with us.)
We can generalize as follows: the logical resources for forming sequences have
evolved in the environment of ideational meaning as sequences of figures. But these
highly generalized resources can then also be applied in an interpersonal environment to
form interactional sequences. (We should note here, however, that evidence from language
development studies suggests that the logical-semantic relations are first construed in
interpersonal contexts: see Phillips, 1985, 1986; Halliday, 1993a; and Pan V below.) In
producing a text, we may use either or both, depending on the nature of the text: see
Figure 3-10 and see also further Chapter 9 below on the role of the text base in
organizing text.

Fig. 3-10: Support in the organization of text from ideationally and interpersonally
oriented sequences

3.7 Sequences and grammar


As we noted in Chapter 1, Section 1.5, semantics and grammar evolve together (in all
three senses of semohistory we discussed). In the present context that means that

3.7 Sequences and grammar

125

sequences and clause complexes evolve together. The basic principle is that a sequence is
realized by a clause complex. But the two may become dissociated from one another.
(i) On the one hand, a sequence may extend beyond a single clause complex, as the
culinary sequence in Section 3.5 above clearly shows. That is, the general potential is
simultaneously semantic and grammatical; but in the creation of this particular text, this
potential may be taken up semantically to create a sequence that is more extensive than
the clause complexes realizing it. Here is a very simple example of a sequence for making
"new potatoes with lemon1 (Highton & Highton, p. 185), which is realized by two clause
complexes:

The text reads as follows:


Prepare and cook the potatoes in sea-salted water to which has
been added lexton juice. Drain and serve with melted butter, a few
shreds of lemon rind, a squeeze of lemon juice and chopped
parsley.
The sequence has been grouped into two clause complexes in the grammar. However,
this grouping is by no means arbitrary; it serves to indicate the two major phases of the
preparation of new potatoes with lemon. In other words, once sequence and clause
complex have become partly disassociated so that one sequence does not automatically
imply one clause complex, the decision how to associate them in realization becomes a
meaningful, significant choice. (The significance may vary from one register to another,
but the principle that the choice is meaningful is quite general.)
(ii) On the other hand, a clause complex may in principle correspond to a figure rather
than a sequence. This happens through grammatical metaphor of an interpersonal kind

126

3. Sequences

(see Halliday, 1985: Section 10.4): an interpersonal modality that would be realized
congruently as a modal auxiliary (can, may, will &c) or a modal adjunct (perhaps,
probably &c) is 'upgraded' to the status of a projecting clause in a clause complex; for
example:
I don't suppose li there's very much

'there is probably not*

I think II I might have walked out too from all

the accounts

1 probably might have

(in) In addition, one or more of the figures in the clause may be realized by something
'less than* a clause. One major source of this is ideational grammatical metaphor, which
will be discussed in detail in Chapter 6 below. But there are also other cases such as
circumstances of Role with temporal implications for example: as a child, he was very
shy 'when he was a child, he was very shy' (cf. Chapter 8, Section 8.2.2 below on
prepositional phrases with with in weather reports). The grammar forms complexes at
ranks below the clause, of course: melted butter, a few shreds of lemon rind, a squeeze of
lemon juice and chopped parsley. It is always possible to interpret certain instances of
these as sequences that have been 'shrunk' by the grammar because they share one or more
elements. Thus Henry and Anne went to the store might be interpreted as a simple figure
if they went to the store together but as a sequence if they went there separately.

3.8 Conclusion
We have identified two primary types of sequence (ways of relating figures), projection
and expansion; and with both of these the figures that are related may be either equal or
unequal in status. To say that these relations are between figures means in principle that
they hold between figures as a whole; that is, given a pair of related figures the domain of
the expansion or projection relation between them is each of the figures in its entirety.
But in some instances, some subdomain may be particularly implicated. This is perhaps
especially true with elaborating sequences the grammar of hypotactic elaborating
clause complexes tells us as much: the elaborating dependent clause includes a relative
reference expression (if the clause is finite) and the clause is placed immediately after the
domain that is being elaborated, whether that is the whole clause (Mary could never feel
comfortable with him, which was perfectly understandable) or some element within the
clause (Mary, who was very sensitive, could never feel comfortable with him).
The limiting case is that where what is being expanded is not a figure but an element
of a figure, in which case instead of a sequence of figures we get a sequence of elements,
realized by group or phrase complexes. Thus instead of Schank conceived of scripts and
Abelson conceived of scripts as a solution to this problem, a sequence of figures where
there are two processes of conception, we get Schank and Abelson conceived of scripts as

3.8 Conclusion

127

a solution to this problem, a single figure with a sequence of elements (serving as


Semer of a single process). Putting this in terms of the grammar, the sequential
relationships remain constant, but the rank at which the complex occurs depends on the
domain being related: see Figure 3-11.

Fig. 3-11: Different domains of logico-semantic relations


One question that arises here is the extent to which the semantics and the grammar are
in phase with one another. It is conceivable that a semantic sequence of figures could be
realized in the grammar by a single clause with complexes at group/word rank i.e., the
sequencing is downranked in course of realization. One argument in favour of exploring
this possibility is that it would be possible to son out the potential ambiguity of
examples such as Henry and Anne went to the movies grammatically we have a
group complex (Henry and Anne), but semantically, it could be either a sequence of
figures (on the reading 'separately': 'Henry went to the movies and Anne went to the
movies') or a sequence of elements (on the reading 'together': 'Henry and Anne went to the
movies'). This was, of course, an area investigated in terms of transformational grammar
with different deep structures posited for the readings 'separately* and 'together'. It is clear
that there ate grammatical group complexes which cannot be expanded semantically as
sequences of figures; for example, Henry and Anne chatted is not likely to be a
grammatical compression of 'Henry chatted and Anne chatted'; rather, it is agnate with
Henry chatted with Anne. In such cases the problem of ambiguity does not arise.
Sequences at lower ranks than that of figures retain the logical mode of realization in
the grammar. Expansion and projection thus flow throughout the system, forming
sequences. This is in fact an instance of a general principle: expansion and projection are
trans-phenomenal categories in the sense that they are manifested over the system
as a whole not merely in different logical environments across ranks but also
expericntially. For example, projection manifested within a sequence: Brutus said Caesar
was ambitious; projection manifested within a simple figure: According to Brutus,
Caesar was ambitious. This feature is particularly exploited when the system is expanded
through grammatical metaphor: see Chapter 6 below.

4. Figures
4.1 Two perspectives on figures
As we expressed it in the last section, a figure is a basic fragment of experience that
embodies one quantum of change. As such, it is like a little drama it is a constellation
of actors and props; and it unfolds through time. We can recognize here two
complementary perspectives on a figure: composition, and time.
Compositionally, figures are phenomenal units that are formed by configurations of
other phenomena (elements). Being "units" means that they are constituted as organic
wholes with functionally distinct parts. (For an interpretation of participant and other
roles in figures in the context of a "theory of activity", see Steiner, 1988a, 1991.) In this
respect they differ from sequences, which are not compositional units but loci of serial
expansion and projection.
Concomitantly, figures take place in time; but the temporal aspect of a figure is
typically construed in association with one particular element, the process. We shall
therefore deal with this aspect of figures under the heading of process, in the context of
our discussion of elements (see Chapter 5, Section 5.4 below). For the remainder of this
chapter we concentrate on the compositional aspect of figures.
There are two interlocking aspects of the configuration of a figure:
(i) the domain of experience to which the figure belongs; and
(ii) the nature of the interaction among its participants.
To put this another way: as a theory of experience, the semantic system of figures
embodies two subtheories: one concerning different domains of experience and one
concerning the ways in which participating phenomena can interact We deal with the
first of these in Section 4.2 and with the second in Section 4.3.

4.2 Composition: domains of experience


4.2.1 Four primary domains
The system of figures construes experience as falling into four broadly conceived domains
of goings-on: doing (including happening), sensing, saying and being (including having)
. Each type of figure has its own set of participant roles: see Table 4(1).
4.2.1.1 Symbolic processing: sensing and saying
Projecting sequences differentiate rather sharply between figures of sensing and saying on
the one hand and figures of doing and being on the other by selecting figures of sensing
and saying as the ones that have the special power of setting up other figures as second-

129

4.2 Composition: domains of experience

onto, seiniotic reality. That is, projecting sequences construe figures of sensing and
saying on two levels, the level of sensing/ saying itself and the level of the content of
sensing/ saying. As we put it in Chapter 3, the projecting figure represents symbolic
processing, processing that brings another figure into symbolic existence. Figures of
symbolic processing involve the symbolic process itself (thinking, saying, etc.) and a
participant engaged in the symbolic processing, as in 'Symbolizer;' she + Process:' said/
thought
that he had left. The projected symbolic content is either a proposition
to leave; she
(she said/'thought
he had left) or a proposal (she asked him
wanted

him to leave).

Table 4(1): Types of figure and participant roles


projection

Process
doing

Actor

she

is playing

Actor

she

is polishing

Actor
she
Actor

is giving

she
sensing

is building

Senser

she

knows

Range
the piano
Goal
the piano
Goal
a book
Goal
a house
Phenomenon
his father

Recipient
to her brother
Client
for her brother

Senser

she
saying

knows

Saver

she

says

that his father has


arrived
Verbiage
a few words

Sayer

being &
having

she

says

Carrier
she
Token

is

she

is

Receiver
to her brother
Receiver
to her brother

that his father has


arrived

Attribute
a lawyer
Value
his lawyer

"Symbolic processing" is a generalization across sensing and saying that foregrounds


the fact that they can both project. But sensing and saying differ in the level of
projection: sensing projects interior content, ideas; saying projects exterior content,
locutions. The level of the projected content determines the typical status of the projected
content: locutions may be either quoted or reported, with quoting being favoured in many
types of discourse; in contrast ideas are typically reported and only rarely quoted. That is,
ideas are construed as being further removed that locuations from experience that is
shared.
Projection thus construes a distinction between interior symbolic processing (sensing)
and exterior symbolic processing (saying). The distinction between 'interior' and 'exterior*
is reinforced by the internal organization of figures of sensing and saying.

130

4. Figures

(i) Sensing and saying construe the "Symboiizer" along different lines. The interior
Symbolizer of sensing is construed as a participant engaged in conscious processing;
hence it is endowed with consciousness by virtue of serving in a figure of sensing. Thus
in an example such as the thermometer thinks it is 35 degrees, the Symbolizer has to be
interpreted as if it was a conscious being.
The "Symbolizer" of a figure of saying often is a conscious speaker. However since
saying is exterior rather than interior symbolic processing, the Symbolizer of saying,
unlike that of sensing, is not restricted to human consciousness; it may also be any kind
of symbol source, a 'semiotic thing* such as institutions, documents and instruments of
measurement (see Halliday, 1985:129-30). Thus alongside examples such as
In the hospital's newsletter/ he tells of one patient who stopped
a two-week-long bout.

we also find
The British medical Journal The Lancet recently reported a
study at Oxford university's John Radcliffe Hospital.
And while the thermometer thinks it is 35 degrees requires a metaphorical reading, the
thermometer says it is 35 degrees does not
We recognize the difference between a Symbolizer of sensing and a Symbolizer of
saying by calling them Senser and Sayer, respectively.
(ii) Figures of saying construe the addressee of exterior symbolic processing in the
form of a participant, the Receiver, as in She told/asked/commanded him...; She said to
him/ asked of him .... In contrast, interior symbolic processing cannot be addressed;
figures of sensing cannot be configured with a Receiver.
Here there is a subtle difference between sensing and saying in their grammatical
realization by verbal and mental clauses that project reports. In the verbal case, the
Receiver is a grammatical constituent of the verbal clause, as in she told him
to
leave. Being a constituent participant, it can serve as Subject in the verbal clause: he was
to leave. However, although the mental case, e.g. she wanted him
told (by her)
to leave, looks similar, it is not, since mental clauses do not have a Receiver. Hie
him to leave.
element him is a constituent of the reported clause: she wanted
Consequently, it cannot serve as Subject in the reporting mental clause; we cannot get He
was wanted (by her) to leave. The difference also shows up clearly when the projected
clause is in the passive voice; contrast the incongnience of she told the car to be washed
with the acceptability of she wanted the car to be washed. The analysis shows the

4.2 Composition: domains of experience

difference: she told the car


Receiver) : she wanted

131

to be washed (where the car gets interpreted as


the car to be washed1

iiii) Directionality. Saying is construed as proceeding from Sayer to Receiver (she


asked/told/commanded him 'she addressed him'). In contrast, sensing embodies two
complementary perspectives: either the Senser's involvement in the sensing ranges over
the Phenomenon or the Phenomenon is construed as impacting on the Senser's
consciousness (she likes the design : the design pleases her).
The demain outside this conscious-semiotic centre of the ideational universe is then
quintessentially either active (doing) or inert (being): see Figure 4-1.

Fig. 4-1: The basic types of figure seen as different domains of experience

As always, there are departures from the general principle. For instance, while she asked/
told/ persuaded/ implored/ encouraged/ promised/ threatened the car to be washed is
incongruent, she ordered the car to be washed is perfectly fine, which suggests the mental
model of the constituency boundary between the projecting and toe projected clause he
ordered
> the car to be washed. This gets support from the fact that the car was ordered
to be washed is odd.

132

4. Figures

There is of course a great deal of indeterminacy here, including such borderline cases
as those where sensing and saying aie construed as forms of action (and therefore cannot
project), e.g. watching, listening, chatting, speaking. These properties reflect the
borderline location of such processes in the overall semantic space. We shall return to
these borderline cases below.
4.2.1.2 Doing and being

Doing and being do not preclude the involvement of a conscious participant; but they
do not require it and hence do not have the effect of endowing a participant with
human-like consciousness. They can be differentiated in terms of two parameters, time
and energy, both of which are involved in their actualization.
(i) Time. Doing involves a change over time of occurrence (including maintaining a
state in spite of force for change). The change may take place along any one of a number
of dimensions: (a) circumstantial: spatial (motion or disposition, concrete or abstract); (b)
intensive: qualitative (colour, size, shape, solidity, etc.), quantitative (increase, decrease);
(c) possessive (transfer of ownership, loss or accretion of parts); (d) existential (creation
or destruction). In contrast, being does not depend on any change over time. As a figure
of being unfolds over time, the only change is that embodied in the temporal unfolding of
the process itself (see Section 4.2.1.3 below). The nature of the actualization will be the
same at any point in time.
(ii) Energy. Doing requires some input to occur. This will typically come from one of
the participants involved, the doer (as with voluntary motion); but the source of energy
may also be outside the figure (as with falling). No input of energy is required with a
figure of being.
Figures of doing and being can be interpreted as complementary perspectives on a
'quantum of change*. Construed as doing, the change appears as a change in the thusness
of a participant. Construed as being, the change appears as an achieved or attainable
so
result. Consider a causal consequence such as [doing:] lie washed it'a
[being:] It was clean*. This quantum of change may be construed as two figures, as it is
here (He washed il, so it was clean.). Alternatively, it may be construed as one figure, in
which case it may adopt either point of view. If construed as doing, he washed it clean,
the figure is elaborated with a result. If construed as being, he made it (be) clean, the
figure is enhanced with an agentive Atthbutor. The wording he cleaned it embodies both
perspectives in a single process. See Figure 4-2.
This complementarity between doing and being is most clearly brought out by the
ergative model of figures, which will be discussed below in Section 4.3 (see 4.3.2

4.2 Composition: domains of experience

133

"Generalized model", esp. Figure 4-11). It is not confined to those figures where the
being' is of the intensive (qualitative/ quantitative) type; we also find pairs such as
(circumstantial) I put it on the shelf/ it's on the shelf
they covered the floor with a carpet/ the
carpet covers the floor.
(possessive)

she's given him a new car/ he has a new car

TTiese more delicate types are presented hi Section 4.2.3 below. Here we should note
finally that the category of doing includes events, so the figure is one of 'doing &
happening'; while the category of being includes (i) being in some circumstantial
relation, and also (ii) having, itself a special case of (i).

Fig. 4-2: Agency and resultative attribution


4.2.1.3 Temporal

unfolding

Whatever the mode of occurrence of any figure, it will always unfold in time. This
temporal unfolding is construed as an inherent property of the process itself, realized
grammatically as tense and aspect; it thus serves to validate the distinction between
process and participant. Whereas on the one hand in its manifestation as process, the
figure unfolds in time, in its manifestation as participant, on the other hand, it persists
through time whether or not the participant undergoes a change of state. The limiting
case is a creative or destructive process, such as writing or erasing a symbol, through
which a participant comes into being or ceases to exist. Figure 4-3 illustrates this
process/ participant complementarity.
The complementarity can also be seen in the different kinds of deixis (relation to the
here & now) associated with processes on the one hand and participants on the other. A
process is made finite it is pinned down in time, with point of reference in the act of
speaking. A participant is made determinate, being held in a location within a referential
space. This same distinction also appears in the temporal unfolding of a text, where
participants have the potential to persist as discourse referents, but processes are excluded,

4. Figures

134

unless they aie turned into honorary participants through the use of grammatical
metaphor (see Chapter 6 below).

Fig. 4-3: The complementarity of participants and processes in the temporal unfolding
of figures
4.2.2 Grammatical realization

We have characterized the distinctions among the different types of figure in semantic
terms. Let us now relate them to the grammar of TRANSITIVITY the ideational
grammar of the clause where the semantics of figures is construed (cf. the summary in
Chpater 2, Section 2.12.1 above). The different semantic types, sensing, saying, doing &
happening and being & having are realized in the grammar of transitivity as shown in
Figure 4-4 (cf. Halliday, 1985: Chapter 5). Thus doing & happening are realized as
material clauses, sensing as mental ones, saying as verbal ones, and being (at, etc.) &
having as relational clauses. The different process types are not signalled overtly in the
grammar; they are covert or cryptotypic categories and emerge only when we consider
their reactances (cf. Chapter 1, Section 1.8), shown in italics in Figure 4-4. The
grammatical reactances for the figure types include:
Directionality of process: many mental processes are typically bidirectional,
appearing in two opposite configurations (/ like it/ it pleases me; cf. detest/
revolt; fear/frighten; remember/ remind, notice/ strike). It is thus possible to
construe conscious processing either as the Phenomenon impinging on the

4.2 Composition: domains of experience

135

Senser's consciousness (he music pleases him) or as the Senser's consciousness


having the Phenomenon as its domain (he likes the music). Neither material nor
relational clauses display this dual directionality.
Nature of participants; In a mental clause, the Senser is endowed with
consciousness s/he thought the moon was a balloon but not normally it
thought the moon was a balloon. This constraint does not apply to any of the
participants in material or relational clauses. While the Senser is heavily
restricted in this way, the other mental participant, the Phenomenon, is entirely
unrestricted: it can be not only phenomenal ( she remembered the old house) but
also macro-phenomenal (act: she remembered him coming down the stairs) or
meta-phenomenal (fact she remembered that thev had pen happy in the old
house). Participants in a material process cannot be meta-phenomenal. For
instance, while it is possible to demolish not only concrete things such as
buildings but also abstract things such as ideas and arguments (she demolished
the house/ their ideas/ his argument), it is not possible to demolish "metathings" (we do not find she demolished that the earth was flat).
Unfolding in time: In material clauses, the unmarked present tense is presentin-present (he is mowing the lawn; Ftajjuiag the job), whereas with the other
process types it is the simple present (mental: she believes he's mowing the
lawn; relational: heuajsa lawn mower; I'm. busy)',
Participation: Material clauses have a special pro-verb, do (to/ with), as in
what he did to the lawn was mow it. This does not occur in mental clauses:
what he did to the story was believe it; nor in relational ones: what he did to the
lawn-mower was have it.
Projection: Mental and verbal clauses are distinct from material and relational
clauses in that the former can project ideas and locutions (quote or repon; see
Chapter 3 above). These represent the 'content' of sensing and saying, as in

David thought

the moon was a balloon, where the relation of projection

is represented by an arrow. Verbal clauses are distinct from mental clauses in


that the Sayer is not necessarily an entity endowed with consciousness; and in
verbal clauses there may be a further participant, the Receiver, which is not
found in a mental clause.
As with all systems in language,

any given instance will be more or less

prototypical; and there may be subtypes lying intermediately at the borderline of the
primary types. The grammar construes the non-discreteness of our experience by
creating borderline cases and blends. One such area is that of behavioural processes
(Halliday, 1985: 128-9): "processes of physiological and psychological behaviour, like

4. Figures

136

breathing, dreaming, smiling, coughing". These can be interpreted as a subtype of


material processes or as a borderline category between material and mental. They include
conscious processing construed as active behaviour (watching, listening, pondering,
meditating) rather than as passive sensing (seeing, hearing, believing). Like the Senser in
a mental clause, the 'Behaver' in a behavioural one is endowed with consciousness;
whereas in other respects behavioural clauses are more like material ones. Like material
clauses (but unlike mental ones), behavioural clauses can be probed with do: What are
you doing ? I'm meditating but not I'm believing. Furthermore, behavioural clauses
normally do not project, or project only in highly restricted ways (contrast mental:
the moon was a balloon with behavioural: David was
cognitive David believed
meditating
the moon was a balloon);2 nor can they accept a 'fact' serving as
Phenomenon (mental: David saw that the others had already left but not behavioural:
David watched that the others had already left ). In these respects, behavioural processes
are essentially part of the material world rather than the mental one. Many of them are in
fact further removed from mental processes, being physiological rather than psychological
in orientation.

Fig. 4-4: The process types in the clause grammar

For the special case of quoting by a behavioural process, as in "You're late again" , she
frowned , cf. the discussion of saying in Section 4.3.

4.2 Composition: domains of experience

137

Such borderline cases, in which the pattern of reactances does not conform exactly to
that of a major type, are typical of grammatical systems in general. Figure 4-5 represents
the overall semantic space construed by figures, taking account of some of the areas of
indeterminacy. Some of these will be followed up in the next section; for further
discussion, see Martin & Matthiessen (1991).

Fig. 4-5: The overall semantic space of figures with principal areas of indeterminacy
and complementarity
4.2.3 Further delicacy
4.2.3.1 Types of sensing
We have presented the distinction among figures as a primary cut in delicacy. To
illustrate how the delicacy of the analysis can be increased, we will discuss figures of
sensing. Here projection turns out to be a major characteristic distinguishing between
different types. Sensing projects ideas into existence; the projection may take place either
through cognition or through desideration, for example (from Pinter, The Birthday Party):
I just thought

I'd tell you that I appreciate it.

4. Figures

138

I think

I'll give it up.

They want

me to crawl down on my bended knees.

Thus the idea 111 give it up' is created by the process of thinking; it does not exist prior
to the beginning of that process. Similarly, the idea 'me to crawl on my bended knees' is
brought into hypothetical existence by the process of wanting. In contrast, perceptive and
emotive types of sensing cannot project ideas into existence. That is, ideas do not anse as
a result of somebody seeing, hearing, rejoicing, worrying, grieving or the like. However,
these two types of sensing may accommodate pre-existing projections, i.e. facts (see
Halliday, 1985: Section 7.5); for instance:
It assures me U that I am as I think myself to be, that I am fixed,
concrte]). (Pinter, No Man's Land)
I was impressed, more or less at that point, by an intuition |[that
he possessed a measure of sincerity the like of which I had never
encountered]!. (Pinter, No Man's Land)
We heard [that you kindly let

rooms for gentlemen]!. (Pinter, The

Birthday Party)

Thus 'that I am fixed, concrete' is construed as something already projected (hence we


could add assures me of the fact that) and this fact brings about the emotion of assurance.
These features are summarized in Table 4(2).
Table 4(2): Ideas and facts
idea projected Into existence pre-projected fact
cognitive &
desiderative

they thought > the earth was flat


they wanted > the earth to be flat
Cognitive processing and
desiderative processing create ideas

emotive &
perceptive

fettt pen*ptfv procwsiag *&4


eqtouve pfoceoint don't; ****

they are activated by facts.


they rejoiced (at the fact) that the
earth was flat :
(the fact) that the earth was flat
pleased them
they heard (the news) that the earth
was flat

The grammar thus draws a fairly clear line between cognition & desideration on the
one hand and perception & emotion on the other. The former can create worlds of ideas,
the latter cannot at least, according to the theory of consciousness embodied in the

4.2 Composition: domains of experience

139

grammar of English. Now, this difference with respect to the ability to project is one of a
set of properties that collectively serve to differentiate perception, cognition, desideration
and emotion as the major subtypes of sensing. These properties are summarized in Table
4(3).

Table 4(3): Distinctions among subtypes of sensing

140

4. Figures

We will comment on the properties very briefly one by one.


(1) Projection of ideas. While both cognition and desideration project ideas, they
project ideas of different kinds (see Section 4.2 above). Cognition projects propositions
ideas about information that may or may not be valid: he believed/ imagined/ dreamt
> that the earth was flat. In contrast, desideration projects proposals ideas about
action that has not been actualized but whose actualization is subject to desire: he wanted/
intended/hoped for
> her to leave. Projection is the critical link between sensing
and saying; cf. property (4).
(2) Phenomenality. A process of sensing may range over or be caused by a
metapbenomenon, i.e. by apre-projected fact serving as Phenomenon, as in (the fact) that
she is late worries me. The two types of sensing that can involve a Phenomenon of this

4,2 Composition: domains of experience

141

metaphenomenal type aie the ones that cannot project, namely perception and emotion.
That is, while perception and emotion cannot create ideas, they can 'react to* facts. In this
respect, they are like certain relational clauses such as (the fact) that she is late is a worry/
worrying cf. (8) below.
(3) Metaphor for modality. Both cognition and desideration can come to serve as
metaphors for the interpersonal system of modality for modalization and modulation
respectively alongside congruent realizations such as modal auxiliaries and adveibs
(Halliday, 1985: Section 10.4). That is, a number of processes of cognition can stand for
probabilities / think : probably, I suppose : perhaps; and a number of processes of
desideration can stand for inclinations and obligations / want : should, I insist : must.
For instance:
I think that in a sense you've had to compromise, haven't you?
(CEC 387) 'in a sense you've probably had to compromise, haven't you'
Neither perceptive nor emotive sensing can serve as metaphors for modalities.3
(4) Verbal causation. Both cognition and desideration may be brought about
through verbal action: have told you that : you know that :: I have persuaded you to :
you intend to. There are no related verbal types causing perception and emotion,
(5) Directionality. Processes of emotion are typically bidirectional. They can be
construed either as the emotion ranging over the Phenomenon or as the Phenomenon
causing the emotion as in / like Mozart's music (the 'like' type) : Mozart's music
pleases me (the 'please' type); cf. Figure 44 above. Here the grammar of English
construes a complementarity between two conflicting interpretations of emotional
processes, with opposing angles on whether we are in control of our emotions, as if
neither one by itself constitutes a rounded construction of experience. Processes of
desideration are not bidirectional; here there is no 'please1 type, only the like' type. Here
the grammar upholds the view that we arc in control of our desires. Cognitive and
perceptive processes may be bidirectional but favour the 'like' type perception almost
exclusively so; 'please1 type perception such as the noise assailed my ears seems quite
marginal.
(6) Construal as behaviour. Sensing is not construed in the grammar as
activity. But, as already noted above, certain types of conscious process may be construed

Emotion is related to interpersonal attitude I rejoice that she's returned : she has, happily,
returned. Unlike modality, attitude is not an assessment of the validity of a clause
(grammatically it is not a Mood Adjunct). Rather, it is a comment on the information
presented in a clause.

142

4. Figures

not only as sensing but also alternatively as a kind of doing as behaviour (as if active
sensing). For instance:
Stanley (urgently): Look
McCann: Don't touch me.
Stanley: Look. Li*tn a minute. (Pinter, The Birthday Party)
Anna: Lirtea. What silence. Is it always as silent?
Deeley: It's quite silent here, yes. Normally. You can hear the
sea sometimes if you listen very carefully. (Pinter, Old Times)
Here look, touch, listen are verbs in behavioural clauses rather than mental ones; they
are construed as activities controlled by an active Behaver. Hie difference is suggested
quite clearly in the last example You can hear the sea sometimes if you listen very
carefully. All the modes of perception may be construed either as behaviour or as
sensing. One significant grammatical difference is that present behaviour would normally
be reported as present-in-present (the present progressive) What are you doing? I'm
watching the last whales of August. but present sensing would not / (can) see the
whales in the distance.4 Another one is that only sensing can involve a Phenomenon of
the metaphenomenal kind. As long as the 'phenomenon' is of the same order of existence
as ordinary things, there is no problem with either process type; we can both see and
watch macro-phenomena: / saw/ watched the last whales leave the bay. But while we can
say / saw that he had already eaten we cannot say / watched that he had already eaten,
which includes a metaphenomenon. This is the borderline between the mental and
material domains of experience. There aie some behavioural processes that are agnate to
cognitive ones (pondering, puzzling, meditating) but none that are agnate to desiderative
or emotive ones. (Behavioural processes of giggling, laughing, crying, smiling and the
like are outward manifestations of emotions; but they aie not active variants of inert
emotive processing such as rejoicing, grieving, and fearing.)
(7) Phase. The different types of sensing have somewhat different potentials for
unfolding in time. With perception and cognition we have various categories of duration,
inception, and the like: e.g. (perception) glimpse, sight, spot as well as see; (cognition)
discover, realize, remember as well as know. But similar distinctions do not seem to
obtain with desiderative and emotive processes.
(8) Agnate ascriptive process. Processes of perception are unique among the
different types of sensing in that they are agnate to a set of relational processes of

Notice also the difference with respect to ability: there is little to choose between / can see
birds in the sky and / see birds in the sky, but / can be watching birds in the sky and / am
watching birds in the sky are quite distinct in fact the former would most probably be
interpreted as usuality I sometimes watch ...'.

4.2 Composition: domains of experience

143

ascription, those which ascribe an Attribute in terms of the way in which it presents
itself to our sense, as in Madam, you'll look like a tulip.
(9) Construal as Attribute of ascription. With many processes of emotion,
there is an alternative construal of the emotion as a quality that can be ascribed as an
Attribute to a Carrier in a relational clause; and this alternative exists for both the like*
type and the 'please' type. Thus I'm afraid of snakes is an ascriptive alternative to the
mental I fear snakes; similarly, in the other direction, snakes are scary and snakes scare
me. This relational type of alternative exists for some cognitive and deskterative
processes, but it is much more productive with emotive ones. Analogous attributes in the
domain of perception seem always to involve potentiality (visible, audible)-, cf. footnote
4 above.
(10) Scalability. Related to the possibility of construing emotion as an Attribute
is the possibility of scaling or intensifying emotive processes: many qualities can be
intensified. We find sets of processes differentiated essentially according to degree of
intensity scare : terrify, horrify ; and emotive processes can be intensified by means of
adverbs of degree such as much, greatly, deeply. These options are also open to some
cognitive and desiderative processes, although not to perceptive ones; but intensification
is an essentially emotive characteristic.
(11)

Reification. Finally, when the different types of sensing are construed

metaphorically as things, they are reified in different ways. Perception, cognition aid
desideration are reified as bounded, i.e. countable things, such as sight(s), thought(s),
plan(s), whereas emotions aie reified as unbounded things, i.e. masses, such as anger,
fear, frustration.s That is, emotion is construed as boundless like physical resources
such as water, air, iron and oil (cf. Halliday, 1990). Indeed, one can see from Lakoff &
Kovecses's (1987) discussion of the cognitive model of anger in American English that a
number of the metaphors for anger construe it as concrete mass (e.g. as a fluid contained
in the body: He was filled with anger, She couldn't contain her joy, She was brimming
with rage). In being construed as unbounded mass, emotions are again more like qualities
(cf. the unbounded strength, height, heaviness, redness).
As always in language, the picture that emerges from a consideration of a multiplicity
of properties is far from simple; it is multifaceted. But it is possible to bring out certain
salient features of the system of sensing as suggested in Figure 4-6. Emotion seems to be
closer to quality-ascription than to a prototypical process; it arises from, but does not
create, projections. In contrast, perception is essentially closer to behavioural processes.
Cognition and desideration are different from both in that they can project (i.e., bring the

A few processes of cognition are also unbounded, e.g. knowledge, realization, understanding.

4. Figures

144

content of consciousness into existence), can stand for modalities, and arc not in general
like either behaviour or ascription; they may be interpreted as the most central classes of
sensing. Cognition is arguably closer to perception than desideration is there are
certain cross-overs like see in the sense of 'understand' alongside its basic sense of visual
perception, and both can be construed in an active mode as processes of behaviour.

Fig. 4-6: The orientations of the different types of sensing


4.2.3.2 Types of being & having
Figures of being & having construe relations between participants. They construe the
same overall range of relations as expanding sequences, and the basic subtypes also
correspond to the subcategories of expansion, viz. elaboration, extension, and
enhancement. Once we recognize that the semantic system construes phenomena
according to trans-phenomenal (fractal) principles (see Chapter 3, Section 3.8 and above;
see also Chapter 5, Section 5.6 below), it will seem natural that figures of being &
having construe relations in such a way that they resonate with the semantic types
manifested in expanding sequences. They do not construe an arbitrarily different theory of
relations.
In a figure of being & having, one participant may thus elaborate, extend, or enhance
another one. (i) One participant elaborates another one along the dimensions of
delicacy, realization, or instantiation. In other words, the elaboration sets up a
relationship either of generality (delicacy), of abstraction (realization), or of token to type

4.2 Composition: domains of experience

145

(instantiation): see Table 4(4). There is another variable whereby elaboration involves
either identity or membership along the dimension in question. (We saw in Chapter 2,
Section 2.11,2 how in building up the grammar of ascription, Stephen started with
instantiation and then moved on to include delicacy.)
Table 4(4): The elaboration of amphibians
identity identifying:
Token + Process + Value
delicacy
(hyponymy)
realization
instantiation

frogs, toads and salamanders are


('constitute') the amphibians :
the amphibians are ('are constituted
by') frogs, toads, and salamanders
groda is ('means') 'frog' :
'frog' is ('is meant by1) groda

membership ascriptive:
Carrier + Process + Attribute
frogs are amphibians
(amphibious)

This (specimen) is a frog

Elaborating figures can thus be used to construe hyponymic taxonomies. In other


words, they arc, among other things, a theory of the systemic organization of the
meaning potential itself; and, by virtue of this fact, they can be used to elaborate it
further. For example:
The fuels of the body mr carbohydrates, fats and proteins. These
are taken In the diet. They are found mainly In cereal grains,
vegetable oils, meat, fish and dairy products. Carbohydrates ar
the principal source of energy in most diets. [ . . . ] Fats make up
the second largest source of energy in most diets. [ . . . ]
(ii) One participant extends another in a relation of composition, possession or
association. As with elaboration, there is also the intersecting variable of identity or
membership: see Table 4(5).
Extending figures can thus be used to construe meronymic taxonomies. In other
words, they are (among other things) a theory of constituency, semantic composition, and
other meronymic relations in language; so they can be used to create further relationships
of the same kind. For example, the following paragraph establishes a meronymic
taxonomy for sentences in Hawaiian:
Sentences are sequences bordered by periods, question marks, or
exclamation points. In Hawaiian they can be thought of as simple,
vorbloss, or coitplex. The most common sinple sentence conmimtm
of

verb phrase noun phrase (s) . Verb phrases contain verbs as

their heads;

verbs are

defined

on the

occurrence with the particles marking

basis

aspect,

of

potential

especially

ua

(perfective aspect) . Noun phrases contain nouns or substitutes


for nouns; these are names of persons or places, or are defined on

4. Figures

146

the basis of potential

occurrence after

the article ka/

ke

(definite), or the preposition ma 'at . (Samuel H. Elbert & Mary


Kawena Pukui* 1979, Hawaiian Grammar, Honolulu, The University Press of Hawaii, p.
39.)

Table 4(5): Types of extension


identity identifying:
Token + Process + Value
composition the sports centre
(meronymy) comprises (consists of)
four buildings :
Four buildings make up the
sports centre.
possession
The students own all the
equipment :
All the equipment is owned
by the students (is the
students ' )
association The training features
videos of leading athletes

membership ascriptive:
Carrier + Process + Attribute
The sports centre has
(includes) a health clinic

The students have lockers

The training involves a lot


of hard work

Videos of leading athletes


are featured by (feature
in/ acconpany) the training
Elaboration and extension are agnate, one with the other; they offer alternative modes
of construal, often with very little apparent difference. Thus we may find both
(1)Elaborating 'Man' is +male, + adult & + human ;
'Man* is a male, adult & human being
(2)Extending 'Man1 consists of +male, + adult & + human :
'Man' has the features +male, +adult & +human
But if we technicalize these alternatives, they do constitute significantly different
approaches to the interpretation of meaning cf. Chapter 1, Sections 1.9.6 and 1.9.4
above. In this book and elsewhere in

systemic-functional

work, elaborating

interpretations tend to be taken further than in many other approaches: this means
emphasizing realization, delicacy, and identities across metafunctions to supplement the
traditional emphasis on constituency and composition.
(iii) One participant enhances another along a circumstantial dimension of time,
space, cause, condition and the like. Table 4(6) illustrates the categories of time and
cause.

4.2 Composition: domains of experience

147

Table 4(6): Types of enhancement

time
cause

identity identifying:
membership script tve:
Carrier + Process + Attribute
Token + Process + Value
Severe floods followed the
Severe floods ensued
rain : The rain was followed
by severe floods
Heavy rain caused floods :
Severe floods resulted
Floods were caused by heavy
rain

Thus enhancing figures may be used to construe arrangements or orderings in space or


time, such as chronologies, maps or structures. For example, the following extract from
an account of the structure of skeletal muscles construes locations through enhancing
figures:
he fibrous connective tissue proteins within the tendons
continue in an irregular arrangement around the muscle to form
a sheath known as the epimysium (epi = above; my * muscle) .
Connective tissue from this outer sheath extend into the bod/
of the muscle, subdividing it into columns, or fascicles (e.g.,
the "strings" in stringy meat) . Each of these fascicles i thus
urroundftd by its own connective tissue sheath, known as the
porimysium (peri = around) . (Stuart Ira Fox, 1984, Human Physiology,
Dubuque, Iowa, Wm. C. Brown Publishers.)
Enhancing figures construing temporal and causal ordering play an important role in
constructing knowledge in a metaphorical mode, as the following example illustrates (see
further Chapter 6 below):
The divergence of inpulses from the spinal cord to the ganglia,
and the convergence of impulses within the ganglia, usually
rults) in the mass activation of almost all of the
postganglionic fibers, (opcit.)
4.2.33 Types of doing & happening
Throughout the history of the study of language, in all the major traditions, grammarians
and philosophers have focussed primarily on figures of doing. Certain subtypes have been
fairly well explored: usually those having some special structural feature, such as figures
involving transfer of possession ('giving') where there is an additional participant role
(Beneficiary, recognized in traditional grammar as indirect or dative object). And figures of
doing in their very simplest form (John ran, Mary threw the ball} have remained for more
than two millennia as the foundation of the theory of transitivity. But there has been

148

4. Figures

little attempt at a systematic treatment of the total range of material clauses with their
intersecting features and subtypes.
Here we shall refer to three major distinctions that have traditionally been recognized;
and then take one further step based one our earlier observation (Section 4.2.1.3) bringing
together figures of doing and being. The subtypes that have been generally recognized in
grammar are (1) intransitive/ transitive; (2) within intransitive, action/ event; and
(3) within transitive, effectum/ affect urn. The first is the distinction between doings
that involve only a doer (intransitive: John ran) and those that also involve something
'done to1 (transitive: Mary threw the baity, realized respectively as Actor + Process, Actor
+ Process + Goal. The second is that between an intentional act by an animate (typically
human) being (John ran) and an unintentional action or inanimate event (John fell; rain
fell). We shall discuss both of these further in the section that immediately follows
(Section 4.3). The third is the distinction between a Goal that 'exists* prior to the doing
of the deed (affectum: Mary threw the ball) and one that is brought into existence by the
doing (effectum: Jack built a house). We shall use this distinction, referred to as
dispositive/ creative, to explain figures of doing in terms of their outcome in other
figures.
As pointed out earlier, doing is a process of change involving time and energy. Such
change implies an outcome; the outcome may be of various kinds, but it is always such
that it can be construed as another figure. We can therefore examine what kind of figure
emerges as the outcome of the one under investigation. (1) If the process is creative, the
outcome is that some entity comes into existence: such a figure may be construed as a
doing with effectum, as in he baked a cake; but it may be simply a creative happening
such as icicles formed. In either case the outcoming figure is one of being (more
specifically, existing):
he baked a cake

outcome: 'there exists a cake'

icicles formed

outcome: 'there exist icicles'

(2) If the process is dispositive, the outcome is more variable; it may be either (i) a figure
of doing (more specifically, doing [to]/ happening), or (ii) a figure of being (more
specifically, being [at]/ having):
(i) the cat chased the mouse

outcome: 'the mouse ran'

(ii) the boys mended the roof

outcome: 'the roof was whole'

John gave his sister a violin

outcome: 'John's sister had a violin'

We have seen that figures of being, other than the existential, may be elaborative
(intensive), extending (possessive) or enhancing (circumstantial). This enables us to

4.2 Composition: domains of experience

149

recognized further subcaegories of doing according to the nature of the figure being
brought about: see Table 4(7).
Table 4(7): Subtypes of figures of doing according to outcome
subtypes of figures of doing (classified
according to outcome)

existential

being
&having

expanding

outcome

examples
happening

doing (to/with)

tomatoes are
growing

John's growing
tomatoes

'there + be +
tomatoes'

elaborating

the rat died

extending

Jenny's
received an
award

they've given
Jenny an award

'Jenny + have
+ award'

enhancing

he moved to
Canberra

the government
moved him to
Canberra

Canberra'

the kite flew

Bobby flew the


kite

'kite + fly'

doing &
happening

the cat killed the 'rat + be +


dead'

rat

'he + be in +

Notice that in some cases the outcome is embodied in the clause by which the figure
is realized; for example in middle variants of the doing & happening type (the outcome of
John ran is 'John + run'), and in clause with resultative elements (Attribute, Role) such
as I'tt boil the eggs head (outcome: 'eggs + be + hard'), Let's appoint Fred timekeeper
(outcome: Tred + be + timekeeper').

4.3 Composition: two models of participation


So far we have explored figures in terms of how they categorize experience into particular
types or domains, showing how this typology extends in delicacy. The next step is to
specify what modes of participant interaction the semantic system of figures engenders.
Tbere ate two models of paiticipation-in-process embodied in the semantic system of
English
(i) One is particularistic: it diversifies our experience of participant interaction into
four domains doing, sensing, saying and being.
(ii) The other is generalized it unifies our experience of participant interaction
across the different domains.
The system thus strikes a balance in the construal of figures between unity and
diversity between differentiating one aspect of experience from another and

4. Figures

150

generalizing over the whole. These constitute distinct but complementary perspectives:
see Figure 4-7.

Fig. 4-7: Particularistic and generalized interpretation of figures


4.3.1 Particularistic model
In what we are calling its "particularist" modelling the grammar is categorizing
experience for us (or we ate categorizing experience through our grammar) by construing
a small number of different types of figure, differentiated according to what kind of
process is taking place and what kinds of participant are involved in what relationships
to each other and to the process.
What is the principle on which the grammar categorizes experience? In the most
general terms, as we have seen, the principle is that all phenomena can be interpreted as
falling within a small number of broad experiential domains:
those happening "inside", within the realm of our own consciousness;
those happening "outside", in the perceptual world that lies around us;
those that ate not kinds of happening at all, but rather kinds of being and of
relating to something else.
We have referred to these as, respectively:

4.3 Composition: two models of participation

151

(1) figures of sensing or, more inclusively (since 'languaging' is treated as a


distinct phenomenal realm), (1) figures of sensing and
(2) figures of saying;
(3) figures of doing or, more explicitly (since the word 'doing' might suggest
intentionality), figures of doing & happening;
(4) figures of being or, more accurately (since 'having* is construed as a kind
of relative 'being'), figures of being & having.
Each of these types of figure has its own special character, as revealed by the way it is
organized in the lexicogrammar. We are not attempting to spell out here the grammatical
features by which they are differentiated (for a general account of process types in the
grammar, see Halliday, 1985: Chapter 5; Matthiessen, 1995b: Chapter 4; Davidse, 1991;
1992a,b,c; 1996a,b). But we shall characterize them briefly in semantic terms with
reference to the forms of participation involved.
(1) Figures of sensing. Here there is one participant, the Senser, who is construed as a
conscious being engaged in "inert" conscious processing ("sensing", as distinct from
conscious processing as a form of active behaviour). This may involve another
participant, the Phenomenon, which enters into the consciousness of the Senser (or is
brought into (mental) existence by the Senser1 s conscious processing). Alternatively, the
Senser's conscious processing may project another figure within the same sequence.
(2) Figures of saying. Here one participant, the Sayer, is involved as the originator of
a process of symbolic (semiotic) activity, or "saying". There may be another participant
the Receiver, whose role is that of 'decoding1 what is said. What is said may itself be
construed as a further participant, the Verbiage; or else it may be projected as another
figure within the same sequence. Finally, there may be a participant functioning as Target
of the saying process.
(3) Figures of doing. Here there is one participant, the Actor, that performs the
process in question; and this process may then impact upon another participant, the Goal
(or may result in bringing the Goal into (material) existence). Other participants that may
be present are the Beneficiary, the one that derives "benefit" from the process; and the
Scope, the one that defines the domain over which the process extends.
(4) Figures of being. In the limiting case, there is only one participant, the Existent;
but generally there are two participants, the one being related by the process to the other.
They may be being related by ascription, as Attribute to Carrier; or by identification, a
rather complex relationship involving two pairs of participant roles: Identifier and
Identified, and Token and Value. These latter intersect with each other, so there are two

152

4. Figures

possible role combinations: (i) Identified/ Token and Identifier/ Value; (ii) Identifier/
Token and Identified/ Value.
The grammar places different constraints on these participant roles, in terms of what
categories of element are typically associated with them, how they relate to the process
itself, and so on. What concerns us here, however, is the particular categorization of
experience that underlies each of these types of figure.
Sensing is clearly modelled as a process of human consciousness, with the Senser as a
human being so much so that merely coming to occupy that role is sufficient to
endow the participant in question with human-like consciousness. The Phenomenon, on
the other hand, is given a somewhat ambivalent status: in one of its guises (as in Do you
like those colours?) it seems to be just a part of the environment; but in its other guise
(as in Do those colours please you ?) it seems to be playing a more active role.
Why does it give this impression? Partly no doubt because of the agnate form Are
you pleased by those colours? where the Phenomenon those colours is brought in
indirectly, like an instrument or means. But this is part of a larger syndrome whereby, on
the one hand, there are other related 'sensing* figures like How do those colours strike
you ?, where the verb strike suggests a fairly violent kind of action; and on the other hand
the prototypical form of a 'doing' figure seems quite analogous to these, as in Were those
boys hitting you? (with those boys as Actor, you as Goal).
The 'doing1 figure is based prototypically on a schema we might refer to as "action and
impact". There is always an Actor, the participant that performs the Process; and in an
example such as the boys -were jumping, the Process stops there that is all there is to
it. But in examples such as the boys were throwing stones, or the stones hit the wall
the Actor's performance of the Process extends beyond, so as to 'impact' on another
participant this is the one known as the Goal (see Figure 4-8). In the typical case (the
"active voice", in grammatical terminology), the clause unfolds iconically, reflecting the
movement of the impact from Actor to Goal.6 And, as we saw above, the latter may then
be followed by representation of the outcome of the impact a resultative Attribute (he
knocked it flat), a circumstance of Role (he cut it into cubes), or a circumstance of
Location (he threw it into the corner).

This iconicity is, however, easily overridden by the textual metafunction, which has its own
mode of iconic realization (see e.g. Halliday, 1979a; Matthiessen, 1988a; 1990b; 1992).

4.3 Composition: two models of participation

153

Fig. 4-8: Actor impacting Goal through Process of doing


In figures of saying, the Sayer is the symbolic source: prototypically human, but not
necessarily so (e.g. the instructions tell you to switch it off first). The Process is
symbolic; but here too there is a subtype of figures of saying that imparts a similar sense
of action and impact, those where the Sayer 'does something to' another participant by
means of a verbal process, as in Don't blame the messenger, Everybody praised her
courage. We refer to this participant as the Target; and again we may note a partial
analogy with figures of doing (though only partial for example, such figures cannot
take a resultative Attribute or other representation of the outcome).7
But these "impacting" figures of sensing and saying are only submotifs within these
two overall types. The general motif of figures of sensing is 'conscious processing'; that
of figures of saying is 'symbolic processing1. And in figures of being, where we might
characterize the general motif as 'relational ordering', there is no trace of a submotif of
impacting at all. What we find, if we try to take in the picture as a whole, is a kind of
focussed model in which the essentially human processes of consciousness, and the
prototypically human processes of symbolic action, constitute the experiential centre;
while the two other types of figure, that of doing on the one hand and of being on the
other, lie on opposite sides of this centre: the one (doing) lying towards the pole of the
concrete, with the experience construed as 'this impacts on that1, the other (being) lying
towards the pole of the abstract, with the experience construed as 'this is related to that'.
This is one model construed by the ideation base. At the same time, it also construes
a contrasting, complementary model in which all figures are treated as alike. This is the
topic of the following section.
4.3.2 Generalized model
The particularistic model, then, comprises a set of submodels: (i) impacting, (ii)
conscious processing, (iii) symbolic processing, (iv) relational ordering. The model that

In the idiomatic expression praised her sky high, sky high is a circumstance of Manner; cf.
praised her highly (not praised her high).

4. Figures

154

generalizes across these various domains of experience is different from any one of these
particular submodels. It sets up one central participant that is common to all processes.
This is the participant through which the process comes to be actualized. We refer to it as
the Medium: see Figure 4-9.

Fig. 4-9: The medium of the unfolding process


Medium and Process form the nucleus of the whole figure (see further Section 4.4
below) that part of the figure which is essential to the complementarity of unfolding
and persisting (cf. Figure 4-3 above). The participant functioning as Medium may be
affected in various different ways, depending on the particular domain the 'trace' may
by physical, mental, and so on; but the status of Medium generalizes across these
domains.
The model thus construes a nuclear figure consisting of a process unfolding through
the medium of a participant. This then makes it possible to construe a further variable:
namely, the causal origin of this unfolding. The Medium's actualization of the Process
may be construed as being brought about by a further participant the Agent: see
Figure 4-10. If the figure is construed with an Agent, it is other-agentive; if it is
construed without an Agent, it is self-agentive.

Fig* 4-10: Actualization of Medium + Process caused externally by Agent


For example:
Several alternatives have developed :
We have developed several alternatives (Agent we)

4.3 Composition: two models of participation

155

The door opened :


Catherine opened the door (Agent Catherine)
The scones should cool on a wire rack :
You should cool the scones on a wire rack (Agent you)

She rejoiced (at the news) :


The news delighted her (Agent the news)
The batter will be very heavy :
ltd s amount will make the batter very heavy (Agent this amount)

Semantically, the Medium is the participant through which the process is actualized.
It is in the combination of Medium + Process that we find the complementarity we spoke
of earlier between the temporal unfolding (the Process) and the atemporal persistence (the
Medium). This close bonding of Medium and Process is manifested in a number of ways.
(i) Of all the participants, the Medium is the most restricted in terms of the range of
phenomena that may function in that role. We can see this in relation to the general types
of figure:
type of figure
doing
sensing
saying
being

range of phenomena functioninc at Medium


phenomenon (of any kind but not metaphenomenon, i.e. not fact)
conscious being
symbol source
phenomenon (of any kind, including metaphenomenon)

We can also see this in relation to more delicately specified subtypes such as:
type of figure
shine
ache
decant
neigh
dress

ranee of phenomenai functioning ai Medium


heavenly body
body part
wine
horse
salad

In other words, whatever the type of figure, the participant that is most closely bonded
with the Process is the one that takes on the generalized role of Medium; it is this that is
in a relation of mutual expectancy with the Process. This is not to say that only horses
can neigh, but rather than anything that neighs is thereby endowed with horse-hood.
(ii) In the taxonomy of figures, the nature of the Medium is more criterial than that
of any other participant or circumstance. For instance, if we consider processes such as
'strew, spill, pour, sprinkle', it is the Medium, not the Agent which enables us to
differentiate among them (cf. Hasan, 1987) (cf. sprinkle + salt, spill/ pour + water,
coffee; strew + flowers); similarly with 'bend, straighten, flatten; melt, fieeze, evaporate,
condense; crack, break, shatter' and so on.

4. Figures

156

(iii) The manner of performance of a process may vary, in which case it is the
Medium by which it is typically determined. This may be a major variation in the mode
of actualization, for example 'open + door, open + account, open + eye* where the process
is respectively mechanical, verbal or physiological; or simply a minor difference in the
means that is employed, e.g. "brush + teeth, brush + clothes'. Some examples:
'control access*
'(de)stabilize structure*
'suspend above ground*
'disrupt integrity*
'remove extraneous matter*
' travel by mounted conveyance*

Process + Medium
open/ close + door
open/ close + eyes
open/ close + account
stand up/ collapse + building
stand up/ collapse + argument
hang + prisoner
hang + painting
break + glass
break + equipment
brush + teeth
brush + clothes
ride + horse
ride + bicycle

In many cases, the difference in the manner of performance is the basis of a lexically
codified (cf. Chapter 1, Section 1.6 above) taxonornic distinction; for example:
'remove protective cover*

'grow towards fulfilment*

Process + Medium
skin + rabbit
peel + fruit
unwrap + parcel
grow up + child
ripen + fruit
age + wine
mature + policy

These are some of the ways in which the close bonding of Medium with Process is
made manifest in the grammar of the clause, such that the two together constitute what
we have referred to as the clause nucleus. Semantically, the nucleus construes the centre
of gravity of a figure, the focal point around which the system of figures is organized.
When we describe the Medium as "actualizing" the Process, we are really saying that the
unfolding is constituted by the fusion of the two together there can be no Process
without an element through which this process is translated from the virtual to the actual.
In the grammar therefore the Medium appears as an obligatory element die only
element that has this status in the clause. This does not mean that we will find a nominal
reflex of the Medium made explicit in the syntagm of every clause; there are various ways
hi which the Medium may be present as a cryptotypic feature rather than as an overt
form. Nevertheless its presence is required in some guise or other; and this distinguishes
the Medium from all other participants in the figure. We shall go on to consider certain
of these other participants in Section 4.4 below.

157

4.3 Composition: two models of participation

As Figure 4-7 above shows, the generalized participant roles of Medium and Agent
correspond to different sets of roles in the particularistic model one or more for each
type of figure. For example, in a figure of saying the Medium corresponds to the Sayer,
whereas in a figure of sensing it corresponds to the Senser. We summarize the
correspondences in Table 4(8), where the generalized participant roles are represented in
columns. These include one role, that of Range, which we will refer to below.
Table 4(8): Correspondences between participant roles in the (i) generalized and (ii)
particularistic models
(i)

(u)

Agent
doing

doing
(to/with)

sensing

'liking'
'pleasing*

saying
being

existential

ascriptive
identifying

Medium Range
Actor

happening

Scope

The river is
overflowing its
banks

Lakes aire forming


The river is forming
lakes
The river is making
Initiator
Actor
lakes form
Scope
The rain's making
Initiator
Actor
the river overflow
its banks
Phenomenon The cat likes mice
Senser
Mice please the cat
Phenomenon Senser
Verbiage
He's telling the cat
Sayer
a story
There were heavy
Existent
showers
Existent
The cold front
Creator
brought heavy
showers
Attribute
Carrier
The cat's hungry
Attribute
Carrier
Attributor
The thought of mice
makes the cat hungry
Value
Token
The cat is our
hungriest family
member
Assigner
Value
Token
The thought of mice
makes the cat our
hungriest family
member
Actor

Actor
Goal

In addition to Medium and Agent, we can recognize two more generalized participant
roles, viz. those of Range and Beneficiary. The Range role is quite pervasive, as
indicated in Figure 4-7 and Table 4(8) above; it can occur in all types of figure that aie
construed as self-agentive, and also in certain figures of being that are construed as otheragentive. The Range construes the range or domain of the actualization of the Process,
with reference to taxonomic scope (as in play : play tennis/'volley ball), spatial scope (as
in climb : climb mountains/hills), etc.: see further Section 4.4.2 (2) below.

158

4. Figures

The Beneficiary role is more restricted: it occurs in certain subtypes of figures of


doing as Recipient (e.g. they awarded her the Pulitzer Prize : they awarded the Pulitzer
Prize to her) or Client (e.g. she designed them a vacation home : she designed a vacation
home for them), in figures of saying as Receiver (e.g. they told her a story : they told a
story to her) and in a couple of subtypes of figures of being (e.g. he made her a good
husband).
Figure 4-11 summarizes the generalized participant roles in diagrammatic form. It
shows Process and Medium as a complementarity the Process is actualized through the
Medium; and it represents the other participants as external to this nucleus, indicating the
type of role relationship that obtains (x enhancing, + extending and = elaborating: see
further Section 4.5 below).

Fig. 4-11: The generalized participant roles


The two models of participation that we have described in this section thus differ in
degree of generality. The particularistic model construes figures into a small number of
distinct types, sensing, saying, doing and being, with different participants in each; while
the generalized model construes figures as all being alike, having a Process that is
actualized through a Medium. These two models embody complementary perspectives on
participation, the one transitive, the other ergative.' Note that there is no
necessary tie-up between the switch of perspective and the degree of generality: it is a
feature of English that the generalized model is construed in ergative terms.

For further discussion of the transitivity/ ergativity complementarity in the domain of


material clauses, see Davidse (1992c).

159

4.3 Composition: two models of participation

4.3.3 Complementarity of doing and being

A good place to illustrate the complementarity of the two models of participation is in


the area of doing and being. Here the two models also bring out the complementarity of
doing and being as modes of construing a quantum of change in the flow of events. Let
us begin by returning to the examples in Table 4(7) above. Here figures of doing are
(tentatively) differentiated in terms of their outcome: this takes the form either of being or
of doing. We shall focus on the "elaborating" row.
The final stage in the flow of events that constituted the rat's life can be construed as
Process: 'end life' + Medium: 'rat'; that is, some process concerning the termination of
life is actualized through the rat. This may further be construed as having an external
cause: Process: 'end life' + Medium: 'rat' + Agent: 'cat'. Finally, the scope of the Process
may be construed within that element itself or as a separate element, the Range: Process:
'die' or Process: 'happen/ do' + Range: 'dead/ death'. Thus the generalized model allows for
the following possibilities:
Process
happen: die'
*happen'

Medium Range
'rat'1
'rat
'death'

'cat'
'cat'

"happen*
'do: die'
'do'

'rat11
'rat
'rat1

'dead'

'cat'

'be: dead'
"be1
be'

'rat'
'rat'1
'rat

'dead'
'dead'

Agent

'dead'

the ra died
[the ra did a death]', but cf. the
ra underwent/ feigned death
the ra fell dead
the ca kitted the ra
[the ca did the rai dead\\ but cf.
the ca struck the rat dead
the ra was dead
[the ca made the raidead\\ but
cf. the cat made the rat sad

As the table illustrates, the generalized model allows for a range of ways of construing
the termination of the rat's life, where the rat is always construed as Medium and the
termination is construed as Process or as Process + Range; but not all the possibilities
are lexicalized, so there are certain 'gaps'. These gaps seem to be systematic. The
transition from life to death is construed in two phases: (i) as happening/ coining into
being and (ii) as the outcome of happening/ being, (i) The first phase can always be
construed as Process (the process of dying); more restrictedly, it can instead be construed
as Process + Range. Consequently, the Range is 'optional', (ii) The second phase must
always be construed as Process (the process of being) + Range; it cannot be construed as
Process alone. Consequently, the Range is 'obligatory'. In either phases, the Medium +
Process' nucleus can be construed as being self-engendered or as being other-engendered,
with an additional participant the Agent The difference is that the first phase makes

160

4. Figures

explicit bow the Agent engenders the actualization of Process through Medium, whereas
in the second phase only the relation of engendering is specified.9
The two phases of course correspond to two of the modes of construing change in the
particularistic model: the first phase corresponds to the construal of change as doing, the
second phase corresponds to the construal of change as being. Looked at in terms of this
model, the change can be construed as doing: the cat does something, which extends to
impact the rat: Actor: 'cat' + Process: 'strike1 + Goal: 'rat1. The outcome may be specified
as an optional Attribute: Actor: 'cat' + Process: 'strike' + Goal: 'rat' + Attribute: 'dead'.
Alternatively, the change can be construed as being: the rat is something: Carrier: 'rat* +
Process: rbe> + Attribute: 'dead', which may be represented as initiated by the cat:
Attributor: 'cat' + Process: 'be' + Carrier: 'rat' + Attribute: 'dead'.
Figure 4-12 brings the two models together to show (i) how they complement one
another, the generalized one showing how doing and being are based on the same
configuration of Agent + Medium + Range and the particularistic one showing how
doing and being are different configurations of roles; and (ii) how doing and being serve as
complementary perspectives on a quantum of change, construing it either as happening/
coming into being or as outcome of happening/ being.
Doing and being thus focus on different phases of a quantum of change; but either can
be extended in the direction of the other to indicate (with 'being') the source of change or
(with 'doing') the outcome of change. When this happens, the wordings that realize a
figure of doing and a figure of being may come to resemble one another. For example,
take the two wordings he drove his car hot and he drove his friends crazy. They could
both be interpreted as Agent + Process + Medium + Range to show them as related to his
car drove hot and his friends were crazy respectively. But at the same time they arc
differentiated as doing versus being: Actor + Process + Goal + Attribute versus Attributor
+ Process + Carrier + Attribute. This shows that they are related respectively to he drove
his car (without the Attribute) and to his friends were crazy (without the Attributor; but
not to he drove his friends)', and explains why we get his car drove hot but not his friends
drove crazy, and why his car drove hot is agnate with his car ran hot (and his onmoved, rolled, travelled) as another kind of happening but his friends were crazy is agnate
with his friends seemed crazy as another kind of being.
In the simple constructed examples used above, we contrasted doing: happening with
pure being; but being also includes 'coming into being', i.e. 'becoming'. In such cases

In the case of 'dead* as Range, the wording is not very acceptable; but in many cases it is the
regular option: the rain made it very wet, the luggage made it quite heavy, the discount made
it very cheap; the news made him very sad, this experience made her very wise.

161

43 Composition: two models of participation

doing and being both construe change leading up to an outcome, but they use different
models: Process and Process + Range respectively, For example:
Th*
1 avm

cool

Process
Actor/
Medium
process
part.:
thing

and [it]

becomes very
hard .

Carrier/ Process
Medium
process
part:
thing

It

becomes igneous
rock .

Attribute/
Carrier/ Process
Range
Medium
participant: part:
process
quality
thing

Attr./
Range
part.:
thing

Fig. 4-12: Doing and being as complementary perspectives on change


Here the first transition in the state of the lava is represented as doing: happening
Medium/ Actor: lava1 + Process: 'cool*. Coolness is construed as a process; consequently,
it is something that is unfolding and which cannot readily be intensified. In contrast, the
second transition is construed as being: becoming Medium/ Carrier: 'lava' + Process:
'become' + Range/ Attribute: Very hard*. Coolness is construed as a quality rather than as
a process; consequentiy, it is something which can come into being which can be
attained. The second transition is more closely related to the third: 'it becomes very hard :
igneous rock1. However, the two options in construing coolness, as process and as

4. Figures

162

quality, are very close,10 which is shown by a parallel text associated with an
accompanying picture of a volcano:
Tha magma

cool and bardan* and

becomes

Actor/
Medium
participant:
thing

Process

Process

additive sequence of
processes

part.:
thing

igneous
rock.
Attribute/
Range
procs s s

Here the transition in hardness is construed as a process within a figure of doing rather
than as a process + quality within a figure of being. Consequently, the hardness is not
represented as an intensified "destination" coming into being. Further it can be construed
as part of a complex process of cooling-&-hardening within a single figure, as indicated
hi the analysis above.
We started our discussion with the 'elaborating' row of Table 4(7) and all our
subsequent examples have also been elaborating, i.e. elaborating outcomes of doing and
elaborating relations of being. However, doing and being can also be compared and
contrasted with respect to extension and enhancement Hiere are various interesting issues
here; but we shall confine ourselves to an observation about the place where the boundary
is drawn between doing and being.
In the case of elaboration, both 'coming into being' and 'causing to be' can be
construed either as forms of doing or as forms of being. However, in the case of
extension and enhancement, there is no comparable multiplicity of perspectives: they can
only be construed as forms of doing. Table 4(9) sets out the patterns for the three types of
expansion.
Table 4(9): Types of expansion in relation to doing vs. being
[a]
doing

being

happening

phasal: inceptive

non-phasal

elab.

the lava hardened

the lava became hard

the lava was hard

ext.

Jenny got an award

Jenny has an award

enb.

Jenny moved to Canberra

Jenny was in Canberra

happening/ coming into being

being

That is, they are very close in the semantic space of figures: this is a good example of a case
where the topo logical perspective discussed above in Chapter 2, Section 2.11 is helpful
since 'doing: happening* and 'being: becoming* seem to be typologically quite distinct.

163

4.3 Composition: two models of participation

[b]

elab.

doing

being

doing to/with

assigned

the pressure hardened the

the pressure made the lava the lava was hard

lava

bard

non-assigned

ext.

they gave Jenny an award

Jenny has an award

enh.

they moved Jenny to

Jenny was in Canberra

Canberra
being

doing/ making be

To round off this discussion of complementarity, let us reproduce the short text from
which we cited the lava example above. This text illustrates some of the main points
made in this section. We present this text twice, first showing the distinction between
clauses realizing figures of doing and clauses realizing figures of being11 and then
showing the distinction between clauses realizing self-agentive figures and clauses
realizing other-agentive figures. These two versions of the text aie followed by the
diagram that elaborates the text. Note that where one clause contains another that is
rankshif ted the features of both will be shown.
(i) Particularistic model: doing/ being
Tfoe rocks that cover the syrifrCQ of the SflTtll flrs called

the

^flifUl ' P crust. Most of the crust is made of icyngoug rock. Iqmeous
means limada by haatJ.
Inside the flffirttl it ifi vo^v hot hot enough Bto It

rockil.

The melte^ irp^ ^ CflliLgd P^gnia.


Soatim* th magma puh through crack* in tha cruat.
When magma comaa to tha aurfaca it is called lava. Tha lava
cools and becomes veyy hia,r^

It becomes igneous rock.

(ii) Generalized model: sclf-agentive/ ocbcr-agentive


gie_rpcks I that cover the surface of the earthj are called the
earth's crust. Most of the crust is nade of igneous rock. Igneous
means Ijmade by heatl.

11

For another illustration of this distinction in discourse, see the "duck" text in Chapter 11
Section 11.3.1 below.

164

4. Figures

Inside the earth it is very hot hot enough to melt rock. The
melted rock is called magma.

Sometimes the magma pushes through cracks in the crust. When magma
comes to the surface it is called lava. The lava cools and becomes
very hard. It becomes igneous rock.

Commentary: (i) Particularistic model: doing/ being. Figures of doing and being
complement one another in the construction of geological knowledge. Figures of being
construe geological conditions ('rocks cover [= are all over] the surface1, 'the crust is made
of [= is] igneous rock' etc.) and technical categories of geology ('igneous means [= is]
made by heat', 'the melted rock is called [= is] magma', etc.). Figures of doing construe
geological activities ('magma comes to the surface', 'the lava cools', etc.). (ii) Generalized
model: self-agentive/ other-agentive. The geological conditions and activities ate
construed as spontaneously engendered with the exception of 'to melt rock' and 'made
by heat: there are essentially no causes external to the Process + Medium nucleus of
figures. In contrast, figures representing technical categories are construed as otheragentive ('it is called lava1, 'the melted rock is called magma').
The short text above has been extracted from a book for children on collecting rocks.
In academic discourses in geology, the generalized model of participation seems to play
an important role: geological activities are construed on the model of Process + Medium,
either as happening spontaneously or as being brought about by an external cause. This
model seems much more relevant than the impact model where an Actor initiates a

4.3 Composition: two models of participation

165

process, which may then extend to affect (impact) another participant, the Goal. For
example, the configuration 'form* + 'limestone* is construed first as self-agentive and then
as other-agentive:
Lim*ton* can form in many ways as shown in Table 4-4. Mast
limestone probably originates from organisms that remove calcium
carbonate from sea water. The remains of
accumulate to

form

the

these animals may

liaton* directly, or they may be

broken and redeposited.

4.4 Degree of participation


4.4.1 Construing out participants
A figure embodies both analysis and synthesis of our experience of the world: an analysis
into component parts, and a synthesis of these parts into a configuration. That is,
process, participants and circumstances are separated out analytically and are thus given
independent phenomenal statuses. This is a creative act of construal. The world is not
seamless and amorphous; it is highly variable in the way it presents itself to us as
experience in its perceptual salience, physical impact and material & psychological
benefit. But it is not "given" to us as an established order, we have to construe it. Not
surprisingly, there is a great deal of variation in the way that different languages do this.
To cite just one example, the phenomenon of precipitation from clouds is typically
construed in Italian as a fgure with a process alone piove 'rains', in Akan as figure witb
process + one participant nsuo ret0 'water + fall', and in one local variety of Cantonese
as a figure with process + two participants tin lok sut 'sky + drop + water'.
What is relevant to us here, however, is that there is also considerable variation
within one and the same language. Thus in English we have sets of agnate expressions
such as the following:
it started to rain : the rain started
there's a wind : it's windy : the wind's blowing
there was a fog : it was foggy

Such variation is not restricted to the weather; cf.


I fear the consequences : I'm afraid of the consequences : the
consequences scare me
he rejoiced : he felt happy
the shades darkened the room : the shades made the room dark
she was liirping : she was walking with a limp

166

4. Figures

As the examples illustrate, there are numerous points at which the system allows for
alternative semanticizations of the flux of expenence. These may differ in the extent to
which the 'quantum* of experience is analysed into separate components. There is a cline
from unanalysed and continuous to analysed and discrete; from example, from 'it's raining'
(one phenomenon) to 'the sky's dropping water' (a configuration of three phenomena): see
Figure 4-13.

Fig. 4-13: Cline from compositionally unanalysed to analysed experience


While there is potential for variation, however, the variation is not arbitrary or
random. The semantic system embodies certain general principles which guide the choice
of one or other pattern of constatai. These include:12
process: (a) whether the process is non-actualized ('irrealis') or actualized ('realis');
(b) how the process unfolds in time (its eventuation profile);
participants & circumstances: (a) whether they are (i) general class, (ii) nonspecific representatives of a class or (iii) specific representatives; (b) how far, and
in what ways, they are involved in the process.
We shall take the notion of degree of participanthood as an example of the general
principle that the phenomena of experience may be construed as having more, or less,
independent status within the semantic system.

12

Cf. Hopper & Thompson (1980) on degrees of transitivity and Taylor (1989) on
categorization in general.

4.4 Degree of participation

167

4*4.2 Degrees of participanthood

We are using the term participanthood to suggest that the status of participant in the
grammar is not absolute, but rather is a matter of degree. Among the various functional
roles that the grammar construes as participants, we will discuss here two that are at
opposite ends of the scale: Goal, which is has a clear status as prototypical participant,
and Range, whose status as participant is much less clearly established.
(1) Goal. As we have seen, the Goal in the particularistic model corresponds to the
Medium in the generalized model wherever the figure is one of doing to or doing with.
The Goal is impacted in some way by its participation in the Process; the "impact" either
(i) brings a participant into existence or else (ii) manipulates one that already exists.
(i) Prepare th* MUG* according to your favourite recipe

(ii) Fry th aubftrgin* for 5-10 minutes


Skin th* tomato**
Heat th ollv* oil
Boil th *99B hard
Beat all tha it MI together
Shape th* lentil pure* into cakes

If the Goal is something that already exists, the result of the Process is to bring about
some change in its location, make-up, temperature, shape, fee.; and tbe result may be
construed as a separate element, with the function Attribute (hard) or Role (into cakes).
These examples highlight the participant status of the Goal, showing the senses in
which the Goal can be said to be impacted. There is no such impacting in the case of the
Range (see (2) below). There is a further contrast between Goal and Range in the degree
of individuacin that is typical of each. In the examples above, what is impacted is a
specific representative of a class, or specific set of representatives; and this is typical of
the degree of individuation of the Goal. Compare in this respect the contrast between
move the piano, where the piano is Goal, and play the piano, where the piano is
Range:
move the piano Process + Goal: specific representative of class
play the piano Process + Range: general class of instrument

Representatives of a class can be impacted (regardless of whether they are specific or


non-specific at the point in the discourse at which they occur); but it is harder to impact
the general class itself. Consequently, if the Goal is a general class rather than a set of
specifiable representatives, it has a lower degree of participanthood. This appears
iconically in the grammar in the limiting case of a clause where the Goal is simply that
class of phenomenon that can serve as Goal of that particular type of figure: the grammar

168

4. Figures

allows us to select 'goal-intransitive', which means that the Goal is simply not specified
for example:13
he drinks
he eats

heavily [alcohol]
all the time [food]

they've gone to the hills to hunt


he buys and sells

[game]

[any commodity]

Such examples are typically either habitual (the process unfolds repeatedly) or
durational (the process unfolds over time): this generalization across time correlates with
the generalization across potential participants both are ways of generalizing from
experience. In some special cases the generalization of the Goal across a class of entities
is shown by treating it as a mass, dispensing with the plural marker:
They often shoot duck during the winter months.
Such a Goal may even be incorporated into the Process, as in he is baby-sitting (and
even who's baby-sitting me this evening?)', this is a restricted option with figures realized
as ranking clauses, but not uncommon where the figure is used to qualify an element and
is realized by a pre-modifying clause, e.g. a fun-loving colleague, a wood-burning stove.
(2) Range. Like other participants Range is realized grammatically by a nominal
group, but it does not participate in the process operationally: it does not bring about or
act out the process, nor is it affected by it materially or mentally. It specifies the domain
over which the process is actualized. For instance, if a process of walking ranges over
Manhattan, it can be represented as Process + Range: They walked the streets of
Manhattan. There are three respects in which the Range is not a prototypical participant:
(I) its relationship to the Process, () its degree of individuation, (DI) its interpersonal
potential.
(I) Relationship to the Process. The Range is not some entity that is impacted
by the Process; it either (i) expands the Process, or (ii) is projected by it

13

Here, because of its generality, the Goal is predictable expeentially. The Goal may of
course be predictable textually, which is the reverse case: so specific at that point in the
discourse that it can be anaphoricaliy presumed. Typically in such cases an explicit pronoun
is used to refer back; but it can be omitted in certain registers, especially instructional ones
such as recipes: when all the pancakes are made, garnish the dish and serve
with cheese
and egg sauce.

4.4 Degree of participation

169

(i) Where the relationship is one of expansion, this take one of two forms: the Range
either (a) elaborates the Process in an objectified form, or (b) enhances it by delimiting its
scope.14
(i.a) In the first case, where the relation is elaborating, the Range simply restates
the Process or else further specifies it in terms of its class, quality or quantity.15 Here,
we often find related pairs of 'Process : Process + Range'; the latter may involve
nominalizing the process (a form of grammatical metaphor). Examples:
Process

Process + Range

(1) doing
sing

play well

play twice

err
clean regularly

sing a song

play a good gama

play two games

play tennis
make a mistake
do the regular cleaning

(2) saying
ask

ask politely

ask a question

ask a polite question

tell a story

(3) being
matter
suffice
attend
--

be
be
be
be

important
sufficient
attentive
content

In type (3), being (more particularly, ascriptive being), the Range is the Attribute that
is ascribed. This construction, Process + Range/ Attribute, is much more common that
the agnate form with Process only (that is, T>e + important' is the preferred model rather
than 'matter1). The Process just embodies the category meaning of ascriptive being *be
a member of and the Range carries the specific information about the experiential
class. It is interesting to note that the ranged construction sorts out the ambiguity of the

14

The Range elaborates the Process, whereas the circumstantial element Role (see below)
elaborates a participant in its particular participation in the process.
This restatement may involve nominalizing the process itself, as some of the examples
given below do.

170

4. Figures

simple present tense between habituai (doing) and occupational (being): she dances/does a
dance every night, she dances/ is a dancer (by profession).
(i.b) In the second case, where the relation is enhancing, the Range specifies some
entity that delimits the scope of the Process; here, therefore, there is often an agnate form
where the scope is construed as a circumstantial element. For example:
Process + Range

Process + circumstantial element

(1) doing
cross the street
climb the mountain
enter an agreement

cross over the street


cliirb up the mountain
enter into an agreement

(2) being

be a witness
become a prince

act as a witness
turn into a prince

(ii) Where the relation is one of projection, the Range represents the subject matter
(either as a general term, e.g. issue, matter, or as the specific domain of the Process, e.g.
politics, your holiday). As with enhancement, there is often an agnate circumstantial
form.
Process + Range

Process -I- circumstantial element

(1) saying
discuss the issue
talk politics
describe your holiday

talk about the issue


talk about politics
write about your holiday

(2) sensing

ponder the problem


investigate the crime

think about the problem


find out about the crime

Type (ii) is like type (i.b) in that in both types the agnate expression takes the form
of a circumstantial element (grammatically, a prepositional phrase). But in the projected
type the circumstance is one of Matter, whereas in the enhancing type it is typically one
of Extent or Location. This corresponds in process type to the distinction between saying
and sensing on the one hand and doing and being on the other.
(II) Degree of individuation. We shall see below (Chapter 5) that participants
are located at some point along a scale of individuation, ranging from most generalized
(e.g., diamonds are forever) to most individuated (e.g., Elizabeth's diamonds were stolen).

4.4 Degree of participation

171

The Range element tends towards the generalized end of the scale. This is especially the
case with those of the elaborating type (i.a above), where the Range usually represents a
general class; and it is always the case if the figure is one of being, with Range as
Attribute. For example:
Peter plays tennis (cf. is a tennis-player)
Peter plays the piano (cf. is a pianist)
His opinion is not important (cf. does not matter)

As a corollary to this, when some element that has functioned as Range is carried
through the discourse, being picked up either by a lexical repetition or by a pronominal
reference, it is more likely to be being picked up as a class, rather than as individuated:
Sharon plays tennis at the same time every other day . . . . Tennis
is a wonderful game, but tennis-players tend to be very obsessive.
Peter spends a lot of time at the piano ... It is a difficult
instrument.

Hence a form of reference such as the following is somewhat improbable:


Peter used to play the piano; but he sold it.

(Ill) Interpersonal potential. The Range element is not very likely to function
as Subject in the clause: that is, to be entrusted with the interpersonal function of
carrying the burden of the argument (cf. Chapter 1, Section 1.4). This means that passive
clauses with Range as Subject are very much rarer than those where Subject is Goal; and
where they do occur, the participant that is functioning as Medium (Actor, Senser or
Sayer) also tends to be of a generalized kind. Thus tennis is played by everyone is not
uncommon, whereas tennis is played by Sharon is a highly marked construction.
Again, the category of Range/ Attribute provides the limiting case. An Attribute can
never serve as Subject in the clause.
We have summarized the features of two participant roles, Goal and Range, which
vary considerably in their degree of participanthood, lying as it were at the two ends of
this continuum. We saw earlier that the Medium is the element that is most closely
bonded to the Process, the two together forming the nucleus of the figure. Thus the
highest degree of participanthood is that of whichever element, in each particular type of
figure, is conflated with the generalized function of Medium; in the case of a figure of
doing, this is the Goal, the element that is impacted (moved, changed, created or
destroyed) by the Process.

172

4. Figures

At the other end of the cline are those elements whose status as participant is highly
precarious, those which conflate with the generalized function of Range. These, as we
have seen, are closely agnate to other types of figure, either those consisting of Process
alone or those with Process + circumstantial element. We can thus extend the continuum
further, outside the status of participant altogether, into the realm of circumstances. In the
next subsection we discuss the circumstantial roles; and we can order these also in terms
of their degree of involvement in the process (Section 4.5 below),

4.5 Degree of involvement; circumstantial roles


As we have already noted (cf. Chapter 2, Section 2.3), a figure consists of a process,
participants involved in the process, and associated or attendant circumstances. Of
these, the process can be seen as the organizational centre the element that reflects the
relative arrangements of the other parts in the configuration. These other parts
(participants and circumstances) are more or less closely involved in the actualization of
the process. Broadly speaking, participants are directly involved in the process;
circumstances are more peripherally attendant on it.
The different degrees of involvement are reflected in the way the figure, and its
elements, are realized in the grammar of the clause. A participant is realized as a nominal
group, and is typically placed next before or next after the verbal group realizing the
process. Circumstances typically occur further away from the process, and are of two
distinct types. Type 1, simple circumstance, represents a quality; this type is realized as
an adverbial group, and will be discussed under elements below (Chapter 5, Section 5.5).
Type 2, macro circumstance, is realized as a prepositional phrase, which in turn consists
of preposition + nominal group. The nominal group, as we have seen, construes an
entity something that could function directly as a participant Here however the entity
is functioning only as a circumstantial element in the process: a location, or an
instrument, or an accompanying entity, or so on (e.g. don't walk on the grass, I
washed it with sugar soap, she came with her children)', it enters into the clause
by courtesy of the preposition, only indirectly so to speak.
What is the status of these "macro circumstances"? They are really reduced or minor
figures, functioning as elements inside other figures. The preposition is a kind of miniverb; the line between circumstances and figures is a very fuzzy one, and we often find
agnate expressions where one is a prepositional phrase and the other a non-finite clause:
cf. / washed it using sugar soap, she came accompanied by her children. The entity that
occurs inside the macro circumstance is therefore already entering into a relationship with
a reduced form of a process; its participation in the main process is thus mediated and
oblique. We can thus contrast the different statuses of two entities where one is a direct
participant and the other enters in circumstantially; e.g. this dictionary was published in

4.5 Degree of involvement; circumstantial roles

two volumes,

173

where this dictionary is Goal while two volumes

enters into the

publishing process indirectly in a circumstance of Role.


Sometimes some entity can be construed in a figure either as a participant or as (an
element in) a circumstance; in that case, construing it as participant means that it is
being treated as more directly involved. Compare pairs of examples such as:
entity as participant :

entity as circumstance:

(1) doing
shoot the pianist :
grab somebody :
paint the wall :
buy mother a present :

shoot at the pianist


grab at somebody
spread paint on the wall
buy a present for mother

(2) sensing
guess the answer :

find something out :

guess at the answer

find out about something

Hie pianist is more likely to escape unscathed as a circumstance than as a participant;


likewise, the answer seems more inpenetrable if guessed at than if guessed.
The difference in the degree of involvement is also reflected in the extent to which an
element is available for a critical role in the interpersonal metafunction. Prototypically,
as we have noted, participants can be assigned the status of Subject, being made to carry
the burden of the argument, whereas circumstances cannot (cf. Halliday, 1985: Ch. 5).
This distinction is however being obscured in Modern English, where although the
prepositional phrase as a whole cannot function as Subject, the nominal group inside a
prepositional phrase often can; e.g. the grass shouldn't be walked on.
The "degree of involvement", in the sense of how deeply some element is involved in
actualizing the process that is construed by the figure, can thus be represented as a cline:
the difference appears not only between participants and circumstances as a whole, but
also within each of these primary categories, so that there is a continuum from one to the
other along this scale.
At the same time, and cutting across this cline of involvement, we find that like
the participants themselves the circumstantial elements fall into distinct types
according to their relationship to the Process + Medium nucleus. These types correspond
to the four transphenomenal categories of logico-semantic relations that are now familiar:
the circumstance is either a circumstance of projection or a circumstance of expansion
and, if the latter, then either elaborating, extending or enhancing. If we combine the

4. Figures

174

degree of involvement with the logico-semantic categories, we can represent the elements
of a figure in the form of a helix (Figure 4-14).

enhancing
Fig. 4-14: The spectrum of circumstances according to the different types of expansion
and projection
Notes to Figure 4-14
(i) Accompaniment. The circumstance of Accompaniment does not correspond to
any one particular participant role. Rather, it corresponds to an extending of the
participant itself, by addition or variation: John came with Mary is agnate to both John
and Mary came; Mary came without John is agnate to Mary but not John came; Mary
came instead of John is agnate to not John but Mary came.

Grammatically, the

analogous type of participant is one represented by a nominal group complex.


(ii) Manner: comparison. This category lies on the borderline of elaborating and
extending: compare he spoke like an expert 'in the manner of, he spoke as an expert 'in
the role of. The analogous participant is that of Attribute in an ascriptive clause as in he
was/seemed an expert, which is construed as elaborating; but the analogous sequence is
an enhancing clause complex he spoke as if he was an expert.
(iii) Angle. This type of circumstance relates to projection rather than expansion,
and specifically to the projecting not the projected component Hence there is no agnate

175

4.5 Degree of involvement; circumstantial roles

participant; instead, the Angle corresponds to the process of saying (grammatically, the
projecting verbal clause in a locution1 nexus) or the process of sensing (grammatically,
the projecting mental clause in an 'idea' nexus). Thus according to the newspaper
corresponds to the newspaper says; and to her students corresponds to her students think.
The examples in Table 4(10) illustrate both the different degrees and the different kinds
of involvement of elements falling outside the Process + Medium nucleus.
Table 4(10): Expansion and projection and types of participant/ circumstance
elab.

enb.

ext.
proj.

participant
Attribute: quality
they lived happy
Attribute: class
he died a hero;
be became a miser
Range: process
he did/ sang a song
Range: entity
she'll swim the river

circumstance: inner
Manner: quality
they lived happily
Role
he died as a hero;
he turned into a miser

Range: measure
she'll swim a mile
Agent
he opened the door :
the door was opened by
him
Client
he cooked her a dinner : a
dinner for her
Recipient
he sent her an invitation :
an invitation to her

Extent: measure
she'll swim for a mile
Manner: means
the door was opened by/
with a key

Receiver
she told him a story :
a story to him
complexes:
she and her aunt are
travelling

he postponed the meeting


until 4
Accompaniment
she is travelling with her
aunt

Range: entity
the board discussed the
financial situation

Matter
the board talked about the
financial situation

circumstance: outer

Extent: entity
shell swim across the river

Cause: behalf
he accepted the invitation
on her behalf/ for her
Location: directional
he sent a parcel to NY

Cause: reason &c


he accepted the
invitation because of her
Location:
positional
in LA, he sent a parcel
in the morning, he
postponed the meeting

Angle
according to the board,
the financial situation ...

What this brings out is that there are a small number of very general domains within
this overall semantic space, which may be construed in different ways according to the
status they are assigned within the figure. For example, there is one area that is concerned
with the spatial orientation of the process. Construed as an outer circumstance, this
appears as the position within which the process unfolds; construed as an inner
circumstance, it means the direction towards which the process is oriented; construed as a
participant, it shows up as receiver or recipient in the process. Thus this general motif is
manifested in a form which corresponds to its ecological niche at that location. Note that

176

4. Figures

the boundaries do not exactly coincide across the different bands of the helix; in any case,
they are fuzzy, and they tend to become more fuzzy with increasing distance from the
centre. Our characterization here is inevitably somewhat overdeterminate. The outer
circumstances, in turn, are typically agnate to clauses (e.g. while he was in LA, he sent a
parce)\ thus we could construe the same general relationships over again in the form of a
sequence grammatically, as one nexus in a clause complex.

5. Elements
5.1 The primary types of element
Elements serve as component parts of figures. Three primary types of element may be
differentiated according to the generalized categories of configurational roles: process,
participant, and circumstance. In addition to these three, we need to recognize a fourth
category of element, the relator; this is the element which forms figures into sequences.

The congruent (prototypical) grammatical realizations of these types of element in


English are as follows:
process
participant
circumstance
relator

verbal group
nominal group
adverbial group; prepositional phrase
conjunction group

These phenomena are "elemental" in relation to figures (just as chemical elements


such as H and O are elemental in relation to compounds such as 2U, CU2>. But the
elements themselves may be internally complex, as in the following examples of
processes and participants:
processes:
verbal group

verbal group complex

will have boiled


had been going to chop

cool and store


is continuing to fail to meet

participants:
nominal group
a shallow tin
invaluable advice
a tasteless vegetable
powdered white sugar
thin slices of lemon
ballpoint pen remover

nominal group complex


lunch or dinner
some boneless chicken pieces and a few
rashers of bacon

178

5. Elements

However, there is a significant difference between elements and figures in the nature of
their internal organization. While figures consist of phenomena that are ontologically of
different types participants, processes and circumstances, the components of an element
belong in principle to the same type. That is, the components of a participant are
themselves potential participants, and the components of a process are themselves
potential processes. Grammatically speaking, participants are realized by nominal groups,
which are groups of nouns; and processes are realized by verbal groups, which are groups
of verbs. The limiting case of a group is a single word.
The situation with circumstances and relators is a little more complicated, (i)
Circumstances of the "macro" type are realized by prepositional phrases, which as we have
seen are like miniaturized clauses; their components aie thus of different types a
(minor) process plus a participant. Circumstances of the simple type, on the other hand,
are realized by adverbial groups; these are groups of adverbs, like more soundly, not so
very fast, with the single adverb again as the limiting case, (ii) Relators are typically
realized by conjunctions, like and, so, if, thai, because, however; these can form groups,
such as as if, and yet, but conjunctions are more often expanded by adverbs (just because,
even if). In addition there are numerous other types of relator prepositional phrases (in
addition, in the event (that), for fear that), nominal groups remaining from earlier
prepositional phrases ( [at] the moment (that), [on] the day (that)), and various
expressions involving non-finite verbs (supposing (that), provided (that)). The relator
construes a logico-semanuc relation between the clauses in a clause nexus (realizing a
sequence), but it is itself an element in the structure of one or other of the two clauses
concerned; e.g. if you have some ink fish preserved in oil, add a few slices at the same
time as the halibut (Elizabeth David, Italian Food).

5.2 Similarities and differences


processes

between participants and

Being elements, participants and processes occupy roles in figures; but whereas processes
only serve in the single role of Process, participants (as we have seen) range over a much
wider experiential spectrum the direct participant roles of Actor, Goal, Senser,
Phenomenon, and so on, and also the indirect participant roles within circumstances such
as Location and Cause. Thus, seen from the point of view of figures, participants ace
construed as being experientially more complex, in the sense that they can take on a
variety of configurational roles: see Figure 5-1. This difference between participants and
processes is also reflected in differences in their internal organizations, as we shall suggest
below.
In our discussion of figures, we pointed out that participants and processes form a
temporal complementarity: participants persist, whereas processes unfold, through time.
(For a functional-typological comparison of nouns and verbs, stated in terms of

179

5.2 Similarities and differences between participants and processes

prototypes, see Hopper & Thompson, 1985.)This complementarity is reflected both in the
similarities and in the differences between the two.

Fig. 5-1: The plurifunctionality of participants


(i) Both types of element begin with a location in the 'here & now': they construct a
path from the spatio-temporal moment defined instantially by the interaction base the
here & now* of the act of speaking to a primary category of ideational phenomena: see
Figure 5-2. For example:

these these particular two old red brick houses


would

have

been going to

collapse

In other words, both types of group include deixis. But toe deixis is of two different
kinds: nominal deixis (such as near/ remote) and verbal deixis (such as past/ present/
future), structurally realized as Deictic and Finite respectively. What this suggests is that,
since processes occur in time their mode of existence is temporal that is how they
are tied to the speech situation; whereas participants exist in some kind of referential
space, which may be grounded concretely in the speech interaction (this = 'near me'; that ~
'away from me') but may also be a more abstract, discoursed space. Hie latter is the space
where we 'record1 discourse referents as we work our way through a text (this = 'about to
be mentioned (by me)'; that = 'mentioned earlier').

5. Elments

180

Fig. 5-2: Comparison of nominal and verbal group structures


(ii) Secondly, both types of group include a specification of a primary experiential
category Thing in the nominal group and Event in the verbal group. But in the
simple verbal group, this is the only semantic category that is lexicalized other
categories are represented grammatically by auxiliaries; whereas in the nominal group,
there is a large amount of other lexical material.
(ill) Finally, the structures of both types of group constitute a kind of path between
the interpersonal reference point, reflected in the Deictic or Finite, and the
experiential one, reflected in the Thing or Event But in the case of the verbal group,
the path is made up of one or more temporal relations: past/ present/ future in relation to
a moment in some dimension of time construed between 'now' and the time of the
occurrence of the event; for example, see Figure 5-3. In the nominal group, on the other
hand, the path from Deictic to Thing is not a chain in time altough it does reflect time
in another way. The path goes through qualities of various kinds, beginning with qualities
that are textual and transitory (unstable in time) and moving towards increasing
permanence (time-stability) and experiential complexity.
Nominal groups have, in fact, far greater potential than verbal groups for creating
experientially complex categories; and this reflects a fundamental difference between
participants and processes. The nominal group has the potential for intersecting any
number of qualities in the representation of a participant; and this makes it possible for

5.2 Similarities and differences between participants and processes

181

the taxonomic ordering of participants to be considerably more elaborated than that of


processes.

Fig. 5-3: The temporal path in the verbal group from Finite to Event
Both the difference in temporal permanence and the difference in experiential
complexity are reflected logogenetically. Participants tend to persist in the unfolding of a
text; and since they do, they can accrue various qualities. In contrast, processes cannot
persist in text: unlike the deictic system of the nominal group, the deictic system of the
verbal group, the tense system, is not a system for tracking textual instances of processes
as a text unfolds. To achieve persistence in text, processes have to be reconstrued
metaphorically as participants (see Chapter 6). When processes are construed as if they
were participants, they can be established and maintained as referents in a text; hence under
these conditions they also can accrue various qualities.
A different kind of parallelism between participants and processes, having to do with
how they are manifested over space-time, has been suggested by other grammarians (see
e.g. Quirk et al, 1985). Taking the traditional distinction between 'mass' nouns and 'count1
nouns (explained as a contrast between things that are unbounded and things that are
bounded and discrete), they have mapped this distinction onto that between 'states' and
'non-states1, in the realm of processes. It then becomes possible to recognize further
similarities, e.g. between plurality in the realm of participants and repeated or iterated
occurrence in the realm of processes. Such analogies may serve as the basis for
metaphorical reconstrual (cf. Chapter 6 below); for example: they have demolished many
buildings : many demolitions have taken place : : he knew a great deed : he had a great
deal of knowledge; the ball bounced again and again : the repeated bouncing of the ball ::
the train stops three times : the three stops of the train. Jackendoff (1991) presents a
recent treatment of some aspects of the same issue. We shall not pursue the point further
here; it needs to be explored in terms of the transphenomenal types of logico-semantic

5. Elements

182

relation that we bave already referred to in various places in particular the question of
how participants and processes are elaborated and extended in space-time.
5.3 Participants
The preliminary taxonomic cut was shown in Chapter 2, Section 2.5; we repeat the figure
here (Figure 5-4). The discussion of macro-participants will be held over until Chapter 6
below; the present section deals with simple things and qualities.

Fig. 5-4: First step in taxonomy of participants


We have already set up the general theory of participants, defining them in relation to
ascriptive processes of being, being at, and having, in a process-participant configuration.
There are three elements in such a configuration: the Process, intensive, possessive or
circumstantial; the Attribute, which is being ascribed by one or other of these processes;
and the Carrier. It is the role of Carrier which defines the concept of a participant. A
participant, according to this theory, is that which may have assigned to it, in the
discourse, properties, parts or circumstantial features. For example:
(a) properties (Process of 'being')

some dishes are very tolerant


the grain looks orange and full
a kitchen should be a cheerful place
the swede is more nutritious than the turnip
(b) parts (Process of 'having')

it has branching stems covered with a green succulent flesh

183

5.3 Participants

spinach has a decided flavour which some people dislike


they have a pleasant fresh flavour
(c) circumstantial features (Process of 'being at')
this plant is like chicory
these mangoes are from Mexico

the seeds will be inside long coffee-coloured pods


this effect might be because of over-heating

This analysis reveals two important aspects of a participant: (i) that it is a thing that can
'carry1 or be ascribed attributes, and (ii) that the ascription may be of different kinds
intensive (elaborating), possessive (extending), circumstantial (enhancing). We comment
on these in the next two paragraphs.
(i) Experientially, there is a 'carrier1 the Thing and there are 'attributes'
Epithets and other modifiers. However, participants are construed not only experientially
but also logically, which means that the Thing (typically) saves as a Head that can be
modified by successive attributes and that this modifying relation is inherently ascnptive.
There is thus no equivalent, in the nominal group, of the Process in an ascriptive figure;
this is construed instead as the logical relation of modification, indefinitely repeatable. For
instance, corresponding to the figure the swede is nutritious we have the participant the
nutritious swede, which, unlike the figure with its experiential, multivariate organization
(Carrier + Process + Attribute), can be logically expanded through further modification:
the tasty tolerant orange nutritious swede. That is, participants aie construed as things
that can accrue attributes.
(ii) There is the same range of types of ascription as are found in ascriptive figures,
and these, as we have seen in Chapter 3, Section 3.2.3.2, can be interpreted in terms of
the different categories of expansion. These are exemplified in the following table (see
below for discussion of the structural roles):
expansion
[of Hung]
elaboration
(intensive)
extension
(possessive)
enhancement
(circumstantial)

Deictic

Epithet

these
these
a
my aunt's
the
the
these
some
a

nutritious
cheerful

Classifier
Mexican

table
18th c
chicorylike

Thing

Qualifier

swedes
mangoes
kitchen
teapot
stems
leg
vases
mangoes
plant

of the leek
from
Mexico

The table illustrates how participants are interprtable as expansions of things they arc
things, with added qualities. It also shows how things can be construed into highly

184

5. Elements

elaborate taxonomies (e.g. this extremely desirable two-storey double-brick executive


residence) which are categorized by ascriptive figures of these different types
elaborating, extending and enhancing. The Thing in these examples corresponds to the
Carrier of the agnate figure of ascription: these nutritious swedes : these swedes are
nutritious :: these vases from the 18th century : these vases are from the 18th century ::
my aunt's teapot : the teapot is my aunt's/belongs to my aunt. Note that with extension,
there may be a reversal of the Carrier-Attribute correspondence: the Thing may correspond
most naturally to the Attribute, as in my aunt has a teapot : the leek has a stem.
The theoretical principle that a participant can be defined as the potential carrier of an
attribute holds for participants of all kinds. We need however to make one proviso:
namely, that more or less any figure can be construed metaphorically "as if* it was a
participant. This is a central feature of grammatical metaphor, discussed in detail in
Chapter 6 below. An example would be Our earlier encounter with this species [had led us
to believe that...] where 'encounter' is semantically a process: cf. We had encountered
this species earlier [and as a consequence we believed that...]. If 'encounter* is construed
congruently in the grammar as a process, it cannot enter into an ascriptive figure. But
once it is metaphorized into a participant, it can: our earlier encounter with this species
had been almost disastrous. This in fact is one of the discursive contexts favouring this
type of grammatical metaphor, and hence serves indirectly as a further illustration of the
general principle we have outlined.
5.3.1 Things and qualities

Grammatically speaking, (simple) participants are realized by nominal groups, which aie
made up of both things and qualities. In terms of the structure of the nominal group, the
cut-off point between things and qualities is between the Classifier and the Thing: see
Figure 5-5.

Fig. 5-5: Qualities and things in the nominal group


Semantically, Classifiers are qualities of the 'class' type (cf. Table 5(4) in Section 5.3.3.3
below): they are like things and may be derived from things, but unlike things they do not
have independent existence they cannot be established in referential space and reidentified in running discourse. So for example although a 'passenger' is undoubtedly a
thing, in a passenger train, where passenger functions as Classifier, it is being construed

185

5.3 Participants

as a quality; hence it cannot be picked up by anaphoric reference we cannot say this is


a passenger train; they must have valid tickets. (Contrast this train is for passengers; they
must have valid tickets,

where passengers

is functioning as Thing.) Grammatically,

Classifiers are realized by 'substantives' or by 'adjectives' (cf. Section 5.3.3.1 below), art
this indeterminacy in grammatical class is symbolic of their status as qualities which are
like things.
This distribution of qualities and things across the nominal group indicates two related
points:
(i) things are more time-stable than qualities; and
(ii) things are more experientially complex than qualities.
This second point has been brought out in discussions of the question by scholars writing
from very different points of departure:
Boole's (1854: 27) The Laws of Thought:
Appellative or descriptive signs, expressing either the name of a thing, or circumstance
belonging to it. To this class we may obviously refer the substantive proper or
common, and the adjective. These may indeed be regarded as differing only in this
respect, that the former expresses the substantive existence of the individual thing or
things to which it refers; the latter implies that existence. If we attach to the adjective
the universally understood subject "being" or "thing", it becomes virtually a
substantive, and may for all essential purposes of reasoning be replaced by the
substantive.
Paul (1909: 251).
Das Adj. bezeichnet eine einfache oder als einfach vorgestellte Eigenschaft, das Subst.
schliesst einen [Complex von Eigenschaften in sich.
Jespersen (1924: 75):
Apart from "abstracts," then, I find the solution of our problem in the view that on the
whole substantives are more special than adjectives, they are applicable to fewer
objects than adjectives, in the parlance of logicians, the extension of a substantive is
less, and its intension is greater than that of an adjective. The adjective indicates and
singles out one quality, one distinguishing mark, but each substantive suggests, to
whoever understands it, many distinguishing features by which he recognizes the
person or thing in question.
Although Jespersen rejects 'substance1 as the characteristic property of substantives, he
finds an "element of truth" in it (1924: 79-80):
... I am inclined to lay more stress on the greater complexity of qualities denoted by
substantives, as against the singling out of one quality in the case of an adjective. This
complexity is so essential that only in rare cases will it be possible by heaping
adjective upon adjective to arrive at a complete definition of the notion evoked by the
naming of a substantive: there will always, as Bertelsen remarks, remain an undefinable
x, a kernel which may be thought of as "bearer" of the qualities which we may have
specified. This again is what underlies the old definition by means of "substance."
which is thus seen to contain one element of truth though not the whole truth. If one

186

5. Elements

wants a metaphorical figure, substantives may be compared to crystallisations of


qualities which in adjectives are found only in the liquid state.

But this difference in experiential complexity is in turn related to the first of our two
points, in that whatever is being construed as stable, as having persistence through time,
is essentially a construct, an assemblage of different qualities, that (to borrow Jespersen's
metaphor) can be crystallized only as an organic whole. The nominal group embodies this
essential association between complexity and permanence.
Qualities tend to be experientially simple, specifying values along a single dimension
or scale such as age, size, weight, loudness, colour, according to either scalar or binary
distinctions (e.g. scalar: large1 'small1, 'tall1 'short1; binary: 'male' 'female', 'deaJ
'alive'). Things, on the other hand, tend to be experientially more complex than
qualities. They are often definable in terms of an elaborate taxonomy where several
dimensions (parameters) are needed to distinguish them (cf. Section 5.3.2.5 and the
example of clothing in Figure 5-8). Consider for example the scale of size. The various
qualities named by large, big, giant, small, tiny, and so on indicate a region on the scale,
but do not specify the "substance*1 of whatever it is that size is being ascribed to. If we
look for objects that are characterized in terms of size, we will find e.g. a giant, a morsel,
and a mini. These, however, involve far more than the single dimension of size. A giant
is 'any imaginary being of human form but of superhuman size and strength'. A morsel is
'a small bite or portion of food'. A mini is usually understood as 'a small car capable of
holding a normal complement of passengers'. Small objects are typically objects of some
particular kind, e.g. droplet, booklet, and kitchenette.
Even with qualities that form binary and taxonomic oppositions rather than scalar
ones, there seem to be differences in experiential complexity between adjective (quality)
and noun (thing) pairs. A standard example of this contrast between qualities and things
involves the two dimensions of maturity and sex: see Figure 5-6. Taxonomically
organized qualities are often named by denominal adjectives, but they are still
experientially simpler than the corresponding nouns. Consider e.g. nationality,
philosophical persuasion, and biological kind: as qualities, these are classes and therefore
closest to things, but their extension is still greater than that of the things that correspond
to them: see Table 5(1).
Table 5(1): Taxonomic things and qualities
Ex&mple

E los S

nymiino gs Mlfflg

mgflnjri gg ffllflllty

Albanian
Aristotelian

'of Albania*
'of Aristotle'

'human: citizen*
'human: philosopher*

canine

'of dog*

'animal: dog'

(any thing) 'originating in Abania'


(any thing) 'reminiscent of
Aristotle*
(anything) 'doglike'

187

5.3 Participants

Fig. 5-6: Complex objects in terms of simple qualities


Thus an Albanian always means 'a human citizen of Albania', whereas the adjective
Albanian could refer to any of a large number of concrete or abstract entities: Albanian
wine, Albanian literature, Albanian economy.
5.3.2

Things

We have observed that things tend to be relatively stable in space and time, and relatively
complex in their semantic make-up and in their interrelationships; and that these special
characteristics of things are construed linguistically in various ways. By comparison with
other elements, things tend to stand out (i) by their varied roles as elements in figures, (ii)
by the overall weight and discursive force of their primary categorization of experience,
(ui) by their tendency to be elaborated into numerous micro-categories, (iv) by their
complex internal organization, and (v) by their highly systematic relationship one with
another. In this section we explore these characteristics a little further. In the next chapter,
we will take account of a sixth characteristic of things, their power to function as model
for other realms of experience.

188

5. Elements

5.3.2.1 Configurational roles of things

We were able to define a thing by reference to one particular role within a figure: namely,
as Carrier in an ascriptive figure construed grammatically as Carrier + Process +
Attribute. But it is characteristic of things that they can take on a variety of roles in
figures of all kinds (cf. Figure 5-1). Here is a set of examples to illustrate this functional
range:
type of figure participant role
Actor
doing &
happening
Goal
sensing

Phenomenon

saying

Phenomenon
Sayer

being &
having

Attribute

example
The book is losing its cover
They burned 1000 books
They printed 1000 books
She bound the book
She remembered the book
She enjoyed the book
The book says that winter is
the best time to visit
This is a very heavy/
interesting book

perspective on book
book as material object

book as
content/
book as
book as

semiotic
material object
semiotic content
semiotic content

books as class of material


object/ semiotic content

Here we find one and the same class of thing functioning (a) as Actor, as Goal, and as
Beneficiary in a material process; (b) as Phenomenon in a mental process; (c) as Sayer, as
Verbiage, and as Target in a verbal process; (d) as Carrier, as Token, and as Value in a
relational process, and (e) as Minirange in a circumstance of Location.
Even if we generalize across these configurational types in ergative terms, these
examples still cover the full participant spectrum: the 'book* occurs as Medium, as Agent,
as Beneficiary and as Range. There is thus a marked contrast between things and all other
elements. A quality enters into a figure only as Attribute; a process only as Process; and a
circumstance only in some particular circumstantial role. Other than this, the only
functional environment for qualities, processes and circumstances is that where they form
parts of things that is, grammatically, where they enter into the structure of the
nominal group.
in figure (clause)
quality
process
circumstance (loc)
circumstance (matter)

in participant (nom. gp.)

your cold hands


your hands are cold
a falling star
a star is falling
the cat on the roof
the cat sat on the roof
the story concerns two cities a tale of two cities

The fact that these other elements can themselves enter into the specification of a thing is
another indication of the relative complexity of things.

189

5.3 Participants

5.3.2.2 Weight and discursive force of primary categorization

The major classes of thing recognized in our school grammars have traditionally been
presented as a list of category meanings of the word class 'noun1: something like "persons,
other living beings, objects, institutions, and abstractions". We will start with this
categorization, modifying it to take account of the point made in Chapter 4, Section 4.2,
to the effect that the primary distinction within figures is that between conscious
processing and other forms of experience (cf. Figure 4-1): the key participant in a
conscious process, the Senser, is restricted to things that are construed as being endowed
with consciousness, so we take conscious/ non-conscious as the primary distinction. It is
also helpful, in the case of English, to make an initial distinction between objects, which
are treated as bounded, and substances, which are not. This gives us an initial
categorization in the form 'conscious/ non-conscious: animals/ institutions/ objects/
substances/ abstractions'.
The most prominent reflex of the conscious/ non-conscious distinction in English is
that it is built into the system of pronouns:
personal
interrogative

conscious
conscious

he/ she
who

non-conscious
non-conscious

' I
what

This distinction is all-pervasive, since third-person pronouns provide one of the main
resources for constructing discourse through anaphora. The boundary between conscious
and non-conscious, of course, is fluid and negotiable: different systems, and different
speakers (or the same speaker on different occasions), may draw it in different places. But
the guiding principle is that 'conscious* means prototypically adult human and may be
extended outwards (a) to babies, (b) to pets, and (c) to higher animals as well as by
rhetorical strategies of various kinds.
The further categories introduced above are distinguished in the grammar by the class
of general nouns, which are used discursively to refer to instances of the category in
question:1
conscious (human): person, people, man, woman, child, boy, girl; and numerous
terms of endearment and abuse
animal:

creature, animal

institution: place, show, set-up

object:

thing, object

substance:

stuff

General nouns are used anaphorically and are typically unstressed: see Halliday & Hasan
(1976: 274 if). For example: I don't know who that cat belongs to. But I've often seen the
creature around

190

5. Elements

abstraction: business, affair, matter; move, event; fact


These primary categories of things represent an ordering of the phenomena of experience.
This ordering has to do with their inherent potential for bringing about change: that is,
their ability to initiate processes and to affect other participants.
One way of exploring this is by noting which participant roles each category of
participant is typically associated with. The critical roles, in this respect, are those of
Senser, Sayer and Actor, operating respectively in figures of sensing, saying and doing &
happening. When we investigate these, however, we find that the overall categorization of
phenomena that is revealed in this way displays a further dimension of complexity: at the
highest level, all phenomena aie distributed into two bioad experiential realms, the
material and the semiotic. This suggests that we should further modify the schema of
primary categorization by splitting it into these two realms as shown in Figure 5-7. This
figure also splits the categories of object and abstraction between the two realms.

Fig. 5-7: Tbe first few steps in the taxonomy of simple things
We give a brief description of each of these categories in turn.
conscious (prototypically adult human)
typifying roles: active participant in figure of sensing [Senser, e.g. do you think
so?], of saying [Sayer, e.g. the teacher said.,.] and of doing [Actor: middle, e.g.
Pat skipped, or effective, e.g. Chris held the rope].
pronoun he/she/they (also I/you); general noun person etc.; number category:
count (singular/ plural).

5.3 Participants

191

has potential for voluntary action [material: doing, including doing to another
participant; verbal (semiotic): saying] and conscious processing of all kinds
[mental: sensing, including feeling, thinking, intending as well as perceiving].
animal
typifying role: active participant in figure of doing [Actor, typically middle, e.g.
birds fly].
pronoun it/they; general noun creature; number category: count (singular/ plural).
has potential for self-initiated action and movement [processes in which animal
occurs as Actor are (unconscious but) voluntary; and may also be effective, e.g.
... was bitten by a snake]', also for perception [Senser hi process of seeing &
hearing],
object (material)
typifying role: impacted participant in figure of doing [Goal, e.g. build a house,
pick up sticks].
pronoun it/ them; general noun thing; number category: count (singular/ plural).
has extension in space, bounded so participates hi figures as unit whole; if acting,
then in figure of happening [Actor in involuntary process, e.g. the button fell

off}.
substance
typifying role: thing as pan of circumstance [Range in prepositional phrase,
especially Location, e.g. on the ground], rather than having direct role as
participant in figure.
pronoun it; general noun stuff; number category: mass.
has extension in space, but unbounded; can be manipulated and measured; if
participant in figure, is typically being distributed [Goal, e.g, cut the string, keep
rain out].
abstraction (material)
typifying role: as Phenomenon in figure of sensing [e.g. estimate the depth], as
participant in figure of being [Carrier in ascriptive figure, e.g. the colours were
too bright; Value in identifying figure, e.g. the score was 2 !].

192

5. Elements

pronoun it; general noun: none; number category: mass.


has no extension in space and is unbounded; typically some parameter of a material
quality or process.

institution
typifying roles: active participant in figure of saying [Sayer, e.g. the ministry
announced...], of doing [Actor, e.g. the school is closing down]; also of
sensing, typically thinking and intending [Senser, e.g. the class decided that...].
pronoun it ~ they;

general noun people, place, set-up etc.; number category:

count (singular).

has potential for voluntary action, typically semiotic with authority of a collective
[verbal process of ordering, mental process of deciding or judging]; also material
[material processes, middle and effective].
object (semiotic)
typifying role: scope-defining participant in figure of saying [Range in verbal
process, e.g. read the notice, tell me a story]; also active participant [Sayer, e.g.
the book says.,., the regulations require... ].
pronoun it/them, they; general noun (none); number category: count (singular/
plural).
may also exist as material object, e.g. book, clock; has potential for being symbol
source [hence Sayer in projecting clause].
abstraction (semiotic)

typifying role: scope-defining participant in figure of sensing or saying [Range in


mental process, t.g.find oui more information; in verbal process, e.g. tell the
truth]; also participant in figures of being & having [e.g. possessed Attribute,
have you any evidence].
pronoun it; general noun: some attitudinally loaded ones such as nonsense [nonattitudinal ones for some, e.g. idea, fact]; number category: mass.

Institutions of course do appear in the plural although relatively infrequently: the plural
pronoun they typically refers to a single institution.

5.3 Participants

*93

unbounded semiotic substance; may be qualified by projection [as Thing +


Qualifier in nominal group, e.g. the knowledge that they had failed]', no material
existence.
As is to be expected with a very general schematic framework of this kind, the
categories outlined above are not at all sharply defined. Their boundaries are indeterminate,
and we will expect to find things of mixed, overlapping and intermediate kinds at every
step along the cline. Among the more prominent of these intermediate categories we can
recognize three as follow.
(i) Natural forces (tides, hurricanes, etc.); instruments (as extended body parts);
powered artefacts (locomotives, industrial machines, etc.): intermediate between
animals and material objects. These are typically active (including effective
action, moving other objects), but non-volitional; hence when the Actor is of
this category, the process does not admit phases that construe intentionality (e.g.
the hammer tried to force the lock).
(ii) Human collectives: intermediate between conscious beings and institutions.
These can function as Senser in figures of sensing of all kinds, including those
embodying desideration; but they accept either singular or plural pronouns, and if
singular pronominalize with it (e.g. the family says it is united/ the family say
they are united).
(ni) Discrete semiotic abstractions: intermediate between semiotic objects and (nondiscrete) semiotic abstractions. These include non-personalized 'facts' and 'cases',
mental entities like 'thoughts' and 'fears', and speech functions 'questions' and
'orders'; they are bounded, cannot function as Sayer but can accept a projection as
Qualifier (e.g., the order to retreat, her anxiety that she might be disqualified).
In Table 5(2) we have included these intermediate categories.
The general picture we are suggesting, then, is that it is in the category of thing that
the grammar captures to the greatest measure the complexity of the elemental phenomena
of human experience. Put together with the different types of figure, which construe the
complexity of goings-on upon the broad foundational categories of doing, sensing, saying
and being, the different types of participant we have sketched in here foreground the
dual nature of experience as being at once both material and semiotic - a world that is
constituted out of the interaction between entities and meanings. On each of these
dimensions there is a progression from things that are most like to things that are least
like ourselves. The grammar imposes a categorization that is compromising, fluid,
indeterminate and constantly in process of change, along with changes in the human
condition and in the interaction of humans with their environment. Yet it is also strong

5. Elements

194

enough to bear and carry forward this wealth of often conflicting experience, and transmit
it over and over again from one generation of human beings to the next
Table 5(2): Ordering of things according to different entena
(i) role potential in figure:
Senser:

Sayer:

Actor
(in
effective):

conscious
materanimal
nonconscious ial realm
natural
force
object
(material)
substance
abstraction
(material)
semhuman
iotic
collectives
realm
institution
objects
(semiotic)
abstraction
(discrete)
abstraction
(nondiscrete)

(ii) internal organization of


participant:
pron:

s/he/
they
it/ they

general
noun:

person &c
creature,
animal

number:

count/
mass
count
count
count

it/ they
it/ they

thing,
object

count

it
it

stuff

mass
mass
count

it~they/
they
it-they/
they

place,
show, setup

count

it/ they
it/ they

it

count

(see note)

count
mass

5.3.2.3 Elaboration of things into micro categories


What we have suggested in the last section is that the grammar, in its role as a theory of
human experience, categorizes those phenomena that it construes as participants by
locating them in a spectrum based on a scale of distance from the human at one end
humans themselves, and things most similar to (i.e. categonzable as) humans, at the
other end things that are farthest away from being human: concrete substances in the
material world and abstract "substances" in the semiotic world. By reference to the
grammar of the clause on the one hand, and of the nominal group on the other, certain
broad categories are set up such that some things will fall squarely into one category,
while others will lie on the borderline, showing certain features of one category and
certain features of another, or finding themselves equally at home in both.
But there is no limit to the differentiation that may be drawn between one class of
things and another; each of these very broad categories comprises numerous micro

5.3 Participants

195

categories within which (as a glance at Roget's Thesaurus quickly reveals) relatively small
sets of closely related things are grouped together. We naturally think of these as being
semantic groupings, for which we can find general labels by moving a little way up in the
taxonomy: parts of the body, household appliances, edible grains, spectator sports,
emotional disorders, and so on, and so on. Such groupings are most readily presented as
lists of words and word compounds; but they are not simply lexical, they ate
lexicogrammatical, displaying some characteristic combination of grammatical properties
or preferences.
Here we shall do no more than refer to a small selection of such micro-domains,
making a brief comment about each with regard to its special characteristics and its place
in the overall spectrum. For a further account of the organization of one such domain, see
Section 5.3.2.5 below.
Professional associates: doctor, dentist, hairdresser, lawyer, teacher, butcher
General category: 'conscious*.
Particular features: may be "possessed" by clients (my doctor, Jane's music teacher
on model of kin: my daughter, Fred's first cousin)', formerly masculine,
gender now (redesigned as) common (he/she)', membership very variable: butcher
etc. probably leaving this class with change in shopping practices.
Domestic pets: cats, dogs

General category: between 'conscious* and 'animal'.


Particular features: alternation between he/she and it; general nouns as 'animal'
(she's a stupid creature)', individuated by proper names, with attitudinal variants;
addressed as if conscious (what do you want?); expanded in talking to children
(pussy-cat, puppy-dog).
Small human collectives: family, household, class (at school)
General category: between 'conscious1 and 'institution'.

Particular features: common number (as institution: the class is ~are writing a
report)-, can be Senser (the family seemed to think that... ); pronominalized as it
- they, not he/she, but self-referenced as we and addressed either as you or in
third person (Do you eat together?/Does ~ do the household eat together?).
Musical instruments: piano, cello, flute, drum
General category: 'object (material)'

196

5. Elements

Particular features: represented as general class with definite article (play the flute,
study the cello); alternate between Range and Goal (play the piano/ shift the
piano); performer construed as derivative in -ist

or -er

(pianist,

flautist,

drummer).
Two-pronged implements: scissors, pliers, tweezers, tongs, shears, clippers
General category: 'object (material)1
Particular features: inherent plural (hence pronominalized as they: the scissors/ they
are in the drawer)', counted as a pair of... (two pairs of shears, not two shears),
but not referred to as both (contrast a pair of shoes, both shoes).
Drinks: coffee, beer, whisky
General category: 'substance'
Particular features: whereas in general a substance, if counted, means 'a kind of
(these soils are less fertile), the counted form of this category may mean either 'a
kind of (/ like this coffee) or 'a measure of (Would you like a beer? Two coffees
pleasef).
Types of enhancement: reason, time, place, way
General category: between 'semiotic object' and 'abstraction'
Particular features: countable, though typically occurring in singular; are names for
major circumstantial classes (reason : why; time : when; place : where; way :
how); can be qualified by relative clause (without the need for a circumstantial
marker did y ou see the way [that] they glared at us?), varying with clause having
corresponding relative adverb (didyou see how they glared at us?), and sometimes
both together (do you know the reason why they glared at us?).
Nuisances: nuisance, mess, disaster, shambles
General category: 'abstraction'
Particular features: formed as countable (a nuisance) but largely restricted to
singular indefinite (forms such as nuisances, the nuisance, your nuisance are rare
and outside this category); typically exclamative and/or ascriptive (that's a mess;
what a mess), and accompanied by interpersonal Epithets (a horrible shambles,
an utter disaster).

197

5.3 Participants
5.3.2.4 Internal structure of things

We have referred above to the syntagmatic potential of the nominal group: how the
grammar builds up the representation of a thing, expanding outwards by modification. For
example:

latest

European
European

tenspeed
tenspeed
tenspeed

reinforced
reinforced
reinforced
reinforced

touring
touring
touring
touring
touring

bike
bike
bike
bike
bike
bike

The expansion proceeds by adding qualities; these qualities, as we have seen, are typically
ordered in English from right to left according to the degree of systemicity, with the most
systemic (most permanent, least particularized) at the right, the most instantial (least
permanent, most particularized) at the left. Broadly speaking these aie distributed by the
grammar into distinct functions as Classifier, Epithet, Numerativa, and Deictic (for the
different types of quality these represent, see Section 5.3.3 below).
This syntagmatic resource serves prototypically to construe things into strict
taxonomies, based on the principle of hyponymy (see Chapter 2, Section 2.11.3 above).
Thus a touring bike is a kind of bike, a reinforced touring bike is a kind of touring bike,
and so on. Since touring is a Classifier, this means that it is one of a defined set: perhaps
touring/ racing/mountain/exercise/ trail. Epithets do not assign classes, but they specify
a particular dimension of taxonomic space: e.g. source European/ American/ Japanese...'.
One type of classification is meronymic (part-whole), where the thing is classified by
the whole of which it forms a part, e.g. bicycle wheel 'wheel (that forms part) of a
bicycle1. These are often indeterminate in meaning: if definite, they tend to be strictly
meronymic (somebody stole my bicycle wheel), whereas if indefinite, they are often
classifying by type (the kids were playing with a bicycle wheel 'wheel for/ from a
bicycle').
5.3.2.5 Systematic relationships among

things

The expanded form of the nominal group makes explicit the systematic taxonomic
relationship which links one thing with another: it is clear that (theoretically at least) a
touring bike is a kind of bike, a bicycle wheel is a kind of wheel. It is always possible,
of course, to expand in a metaphorical way, as with cart wheel (in turning can wheels) and
Catherine wheel (a kind of firewofk); these are not, strictly speaking, kinds of wheel, so
we have to recognize either that these fall outside the taxonomic organization or that the
taxonomy itself is being extended metaphorically. But this does not affect the general
principle at work; indeed, it is the taxonomic principle which makes such divergence
possible.

198

5. Elements

But we also have to recognize that there is a systematic relationship among a set of
terms such as bicycle, tandem, scooter, car, van, truck and bus. Here the form of wording
gives no clue to any such relationship; all we have is an inventory of different words. Yet
these also form some kind of a taxonomic set: they are all wheeled vehicles. We can have
taxonomies of things without using the syntagmatic resources of the nominal group,
simply by the device of naming the organizing principle is not syntagmatic but
paradigmatic.
But there is a difference. The paradigmatic strategy, that of inventing new names,
typically construes sets of things which are systemically related but not in a relationship
of strict taxonomy. This resource is typically associated with feature networks: that
is, networks made up of systems of features, such that each lexical item (as the name of a
thing) realizes a certain combination of these features selected from different systems
within the network a particular clustering of values of systemic variables.
Figure 5-8 gives an example from an everyday commonsense domain, that of clothing.
Obviously a set such as this does not form a strict taxonomy; and this has certain
consequences. The division into things of this particular semantic space is highly
variable, both synchronically (regional and social dialect, and even idiolect) and
diachronically. Thus what were previously non-occurring combinations may come to
occur (e.g. pantyhose, covering both extremities and torso of lower body), systems may
shin their entry conditions (e.g. trousers previously had 'male* as an entry condition), and
so on. Table 5(3) shows how each clothing item realizes a selection of features from
related systemic sets.
Furthermore, although the characterization is essentially experiential, this kind of
network may easily be intruded by interpersonal systems. There is trace of this here in the
opposition of 'casual' and 'formal'; but probably the most widespread of all such
interpersonal systems is that of 'desirable'/ 'undesirable', often referred to as an opposition
of "purr"/ "snarl". Such systems are much more readily admitted into domains of things
which are not governed by strict taxonomy. There is a host of humorous examples
associated with the person system, such as / have views/you have opinions/he or she has
prejudices.
This resource, the construal of systematically related lexico-semantic sets, illustrates
well the principle of "lexis as most delicate grammar" (Halliday, 1961; Hasan, 1987;
Matthiessen, 1991b; Cross, 1993). We have discussed above the principle that categories
in the experiential grammar are ordered in delicacy, so that starting from the very general
types of process that are construed into figures, we can differentiate both processes and
participants into finer and finer subcategories, until we reach the degree of differentiation
that is associated with the choice of words (lexical items). Note that it is not (usually) the
lexical items themselves that figure as terms of the systems in the network. Rather, the

199

5.3 Participants

systems are systems of features, and the lexical items come in as the synthetic realization
of particular feature combinations. Thus lexis (vocabulary) is part of a unified
lexicogrammar; there is no need to postulate a separate "lexicon" as a pre-existing entity
on which the grammar is made to operate.

Fig. 5-8: Feature network for some items of clothing

5. Elements

200

Table 5(3): Items of clothing defined by intersections from feature network


thing ^
lexical
item
shoe

sex

formality

part

lower & torso


11
"

supporting
: firm
supporting
: soft
nonsupporting
[soft]
"
inner &
"

"

"

body

lower & extrem.

slipper

body

"

"

sock

body

"

"

body
body
body
body

"

"

body

"

stocking
panties
briefs
underpants
trousers

female
female
male

shorts

female

female
female

shirt

dress

casual body

vest
bra
blouse

outer &

body

jeans
skirt

"

outer &

body
body

upper & torso

body
body

"

"
outer &

body

female

inner &

partitioned
"

partitioned:
tubes: fulllength
partitioned:
tubes: halflength
partitioned:
tubes: fulllength
nonpartitioned
partitioned: holes &
short &
closed
"
holes
holes tubes: f~h
& short &
closed ~
opening
tubes: f h & short
& closed
opening
holes tubes: f h & long
& closed

body

coat

body

jacket

body

cloak

body

nonpartitioned

body

partitioned

jumper

female

boles]
"]
["]

opening
tubes: f &
long &
opening
tubes: f (~
h)&
short &
opening
long ~
short &
opening
tubes: f ~
h & short
& closed

201

5.3 Participants

pullover

male

sweater

body

cardigan

body

waistcoat

body

glove

body

cap
bat
scarf

(male)

holes tubes:
fulllength
& short &
opening
holes ~
tubes:
fulllength
& short &
opening
tubes:
fulllength
& short &
opening
holes &
short &
opening

body

casual head
head
head

extrem.

[nonsupporting:] [soft]

on head
on bead
round
neck

'partitioned & upper' = having sleeves' (or 'sleeveless')


'partitioned & lower' = 'having legs' (or 'legholes')
T = 'full-length'
'h' = 'half-length'
optional alternative features: X~Y
features assigned by default: [X]
'unmarked' formality, 'non-specific' sex: not shown
It will be clear that both these strategies, the expansion of word to group and the
accumulation of contrasting words, are strategies for assigning qualities to things: we can
gloss a slipper, for example, as a soft supporting garment worn by either sex at the lower
extremities of the body. In general the qualities involved will come from the same
semantic domains in either case; and there is often fluctuation between the two strategies
within the language we sometimes talk of slippers

and sometimes of soft shoes.

Compare Chapter 7 below, where we point out that there are many realms of things
which are construed in English as different words but in Chinese as syntagmatic
constructs: e,g. bicycle 'self-propelled wheeler', car 'gas wheeler', lorry 'goods wheeler' all
as explicit subcategories of 'wheeled-vehicle'. And although we have presented the two
strategies as discrete, there are of course intermediate modes of construing that form a
continuum between the two: noun-compounding (more syntagmatic) and morphological
derivation (more paradigmatic). So we find compounds such as pushbike, motorbike; ati
derivational series like cycle : bicycle, monocycle, tricycle. It is not difficult to invent
new categories if we need them such as bikelet or megabike. Note also the morphological
strategy for deriving casual terms from formal ones: bicycle > bike (cf. omnibus > bus).

202

5. Elements

The son of strict taxonomy that is typically associated with related sees of nominal
groups is often a feature of special registers of the language; compare our examples from
the domains of cooking and weather in Chapter 8 below. The limiting cases of such
taxonomies are those found in the specialized technical registers of science and technology
(cf. HaUiday & Martin, 1993); these include some which are partially or even wholly
designed in a conscious exploitation of the grammatical resources involved. The "things"
that are construed in this way include the more abstracts concepts of a scientific theory,
the virtual objects that are postulated to explain the more arcane phenomena that impinge
on human experience.
5.3.2.6 Recurrent semantic

principles

In the various categories of thing we find a manifestation of the same very general
semantic principles that we have established before, the principles of projection and
expansion. Certain categories of thing have grammatical properties that relate them* to
one or the other, see Figure 5-9.

Fig. 5-9: Projections and expansions construed as things

203

5.3 Participants

(i) Projections as things


Things of this category are the names of types of projection; in the taxonomy of
things, they are semiotic abstractions, some discrete and some non-discrete, and as a
grammatical class they are referred to under the heading of fact nouns. They aie
grammatically distinct because they can function as Thing in a nominal group with a fact
clause as Qualifier, e.g. the notion thai pigs can fly

(see Halliday, 1985: Section 7.5).

Such clauses have the property that they can stand alone as participants in a figure, e.g.
that pigs can fly is an intriguing notion.
Things of this type fall into four subcategories, defined by their interaction with
modality: cases, chance, proofs and needs. Cases represent simple facts; chances
represent facts to which some degree of modalization is attached; proofs are
demonstrations of facts. The fourth category, that of needs, are facts accompanied by
modulation, that is, where the projection is that of a proposal rather than a proposition.
type of
projection
proposition

proposal

subtype modality

lexical items realizing Thing

case

(simple fact)
'it is the case that'

chance

modalization
'it may be the case that'

proof

caused modalization
'this makes it certain
that'
modulation

accident, case, fact, grounds, idea,


lesson, news, notion, observation,
point, principle, rule,
ability, certainty, chance,
hypothesis, likelihood,
impossibility, possibility,
probability, theory
confirmation, demonstration,
disproof, evidence, implication,
indication, proof,
duty, expectation, necessity, need,
obligation, onus, requirement, rule,

need

Things of this type have an important role in discourse, because they function
anaphorically to refer to (and at the same time to classify) previous sections of text
interpreted as projection. For example:
[Graham's] simulated atmosphere evolved in much the same way as
the real atmosphere had - the temperature rose, and evaporation
and rainfall over the tropical ocean increased closely matching
actual records.
Based on thmmm finding*, Graham concludes that the increases in
sea

surface

tenperatures

could

well

have

caused

the

intensification of the hydrological cycle, explaining the wanning.


(New Scientist, 4 February 1995)

5. Elements

204

(ii) Expansions as things

Certain categories of thing denote an element which is an expansion of something


else, either elaborating it (land of, 'instance of, or 'symbol of) extending it ('pan of,
'amount of, 'collection of, or 'extension of), or enhancing it ('time of, 'place of, 'cause
of, 'manner of). The following are some typical examples:
type of expansion
elaboration

extension

subtype
symbol
kind
instance
part
amount
collection
extension

enhancement

time
place
cause
manner

lexical items realizine Thine


picture, drawing, portrait, painting, photo
type, sort, kind, mode, species, genus, brand,
model
example, illustration, instance
part, element, portion, top, bottom, side; arm,
leg, trunk
unit, cup, glass, jar
herd, crowd, bunch, list
combination, conjunction, addition,
contrast
time (Jthat/wheniwe ran away), era^ period
place ([that/ where] we ran away)
reason ([that/ why) we ran away), result, cause,
purpose
way ([that/ in which] we ran away), manner

Those of the elaborating type may function in the nominal group either as Thing or
as Facet. As Facet (always constructed with of), they serve to construe the element
functioning as Thing in some particular guise or perspective; for example, picture in this
will give you a general picture of the situation, kind in a jet cat is a kind of passenger
vessel example in there "were no examples of successful integration. As Thing, they aie
participants in their own right, either 'objects' (e.g. is that picture for sale) or
'abstractions' (e.g. Darwin showed how species first evolved).
Similarly, those of the extending type may also function in either of these two
grammatical roles. As Facet, they specify some quantity (either by container, e.g. ajar of
jam, or by division, e.g. a piece of cake), some aggregate (e.g. a crowd of onlookers),
some aspect or component (e.g. the other side of the argument, the top of the mountain,
the trunk of the tree), or something added or substituted (e.g. an extension of your ideas,
the latest addition to the family, an alternative to this proposal). As Thing, again, they are
participants, typically concrete objects (e.g. a glass jar, the top [= lid] of the canister,
build an extension on the property).
Things of the enhancing type have already been cited as a micro category in the
preceding Section 5.3.2.5. They construe one or other of the general logical semantic
relations of cause, manner, time and place. As Facet, they specify some circumstance of
an element functioning as Thing: the cause/result/purpose of the breakdown, the manner
of the breakdown, the time/ occasion of the breakdown, the place/ location of the
breakdown; cf. also the circumstances of the breakdown. As Thing, they give the lgico-

5.3 Participants

205

semantic relation itself the status of participant; here the usual words are reason, way,
time, place, and the phenomenon becomes a figure realized by a qualifying clause: the
reason/ way/ time/ place we broke down. These are closest to the borderline with
projections: we may have either expansion the reason for which we broke down, or
projection (the reason) why we broke down. Those with 'reason' and 'way' often enter into
an identifying relation with some other figure; for example, the reason we broke down
was that/because the engine overheated.
Since things of these various kinds are the names of semantic relationships, it is not
surprising that their status is somewhat ambivalent. Consider an expression such as a
volume of poetry: this may be volume as Thing, qualified as being concerned with
poetry, e.g. she picked up an old volume of poetry from the shelf,

or Facet (extending:

amount), e.g. she has just published a new volume of poetry. The latter example shows
up the nature of the ambivalence: the quality 'new' is presumably a quality of the poetry
(cf. I'd like a strong cup of tea). Similarly we have the ambivalent relationship between
parts and wholes (the basis of synecdoche), and between symbols and what they represent,
regarding which to construe as the participant in particular figure. In general, the
uncertainty that arises is whether the expansion is to be construed as an independent thing
or as a facet of something other than itself. This uncertainty is sometimes foregrounded
under pressure from the textual metafunction: as Theme, for example, do we say ihe end
of that story you're never going to hear! or that story you're never going to hear the end

ofn

As with names of projections, names of expansions often serve a role in creating


cohesion in discourse: e.g. Instances such as these ..., Another way of approaching the
situation ..., That aspect hadn't occurred to me. Such expressions construe preceding
figures and sequences as participants in their turn, and so enable the speaker or writer to
make explicit the organization of the discourse itself.
5.3.3 Qualities
The discussion of 'facets' illustrates the difference between things and qualities of things;
it also illustrates the indeterminacy of the distinction. Let us begin our discussion of
qualities by relating them to things.
5.3.3.1 The status of qualities
In Section 5.3.1, we discussed the complementary contributions made by qualities and
things in the construal of participants. We showed that qualities and things differ in two
related respects, temporal stability and experiential complexity: things tend to persist
through time and to represent intersections of many dimensions, whereas qualities tend to
be less stable through time and tend to represent values on single dimensions. Qualities

206

5. Elements

thus construe values on dimensions such as size (e.g., 'big/ small'), weight ('heavy/
light1), and shape ('round1/ 'square'/ 'rectangular1/ 'oval*...).
These values may be of three kinds, according to the type of contrast they set up: (i)
binary, e.g. ('dead'/ 'alive'); (ii) scalar, e.g. ('happy/ 'sad'); (iii) taxonomic, e.g. 'wooden'/
'plastic'/ 'stone'/... . Of the three, the most complex, experientially, ait the taxonomic
qualities; for example, the taxonomy of materials (whether a folk taxonomy or one that is
more scientifically informed) is based on a variety of different features, such as (in a folk
taxonomy) its appearance, its texture, its range of functions, its relative value in different
contexts, and the like. Taxonomic qualities are thus the closest to things; they aie often
realized as denomina! adjectives, or even as nouns, and they tend to function as Classifier
rather than Epithet (i.e. they son things into classes rather than describing them).
Since qualities are assigned to things, they are construed with things as their frame of
reference: in the first instance, a quality characterizes a thing relative to other things in the
same (primary) class. Thus a thick book is not a 'thick thing1; rather, it is a 'thick book'
as opposed to a 'slim book': the scale of 'thick' to 'slim' is relative to book and a thick
book would be much thicker than a thick envelope. This characteristic is particularly
noticeable with scalar qualities, which have received particular attention in semantic
studies; but it is also, in principle, a feature of taxonomic adjectives even those
construing complex classes. For example, the criteria for assigning 'wooden* to a 'spoon',
a 'house' and a 'carnage* are fairly different in tarns of the actual material make-up of these
things.
As a category, qualities lie somewhere along a cline between things and processes, and
their status varies considerably among different languages. In English, qualities belong
more closely with things, since they contribute primarily to the construction of
participants: grammatically, English favours construing a quality as Epithet in a nominal
group, and the class of adjective is clearly related to that of noun. (By contrast, in
Chinese, where qualities are typically construed clausally, as Attribute, rather than
nominally, as Epithet, the adjective is equally clearly related to the verb.) Here English is
similar to Latin, where in the traditional grammars a general class "noun" was classified
into "noun substantive" and "noun adjective":
participant

noun

thing

noun substantive

quality

noun adjective

As an example, consider the following from R.R. (1641), based on Lily's grammar:
A Noun is a part of speech, which signifies a thing, without any difference of time or
person. A Noun is the name of a thing that may be seen, felt, heard, or understood: as,

5.3 Participants

207

the name of my hand in Latine is Mantis, the name of bous is Domus, the name of
goodnesse is Bonitas. Of Nouns, some be Substantives, and some be Adjectives.
A Noun Substantive, is mat which standeth by himself, and requireth not another word
to be joyned with him, to declare his signification: as, Homo, a man. And it is declined
with one article: as, hie Magister, a master: or else with two at the most: as hie & haec
Parent, a father or mother. ...
An Adjective is that, which in speech needeth a Substantive to cleave unto ... A Noun
Adjective is that which cannot stand by himself, in reason or signification, but
requireth to be joyned with another word. ...
A Verb is a part of speach declined with Mood and Tense, and betokeneth doing; as,
Amo, I love: or suffring; as, Amor, I am loved: or being; as, Sum, I am.

Here, nouns and verbs are distinguished as primary classes, whereas substantives and
adjectives are distinguished only as secondary classes. (It is important to keep the degree
of delicacy (primary, secondary, etc.) in mind when interpreting statements to the effect
that a particular language has only nouns and verbs, but no adjectives.) More recently, it
has often been asserted that adjectives are really stative verbs, or "stative predicates". This
is the position in predicate logic, where both qualities and processes are simply predicates
but so also are things. It complicates the description of English grammar; but it is a
reminder of the intermediate status of qualities as elements of figures.
We said that both things and qualities are construed as participants. This does not
mean that they are different kinds of participant. Rather, it means that within the structure
of a participant, they serve different kinds of role.
Typically, a quality combines with a thing to make up a participant in a figure: a dry
plate, where the quality 'dry' is Epithet in the nominal group. The only context in which
a quality serves on its own in a participant role is as Attribute; here it stands in intensive
relation to a participant, either (i) in a figure of being, where the participant is Medium/
Carrier (i.e. its sole function is to have the quality ascribed to it) or (ii) in a figure of
doing, where the participant is Medium/ Actor or Medium/ Goal and quality results from
the doing; for example:
being:

The plate's dry I've made the plate dry

doing:

I've wiped the plate dry

Note that in the doing figure, where / is Actor, the Attribute could be omitted: I've wiped
the plate; in the being figure it cannot, since the process itself is one of ascription and
the other participant, if present, is merely the ascriber.
The Attribute is not a prototypical participant We have already noted that as it stands
it cannot function interpersonally as Subject. On the other hand, it can easily be instated
as a participant by adding the noun or the noun substitute one(s) to the nominal group
which realizes it: this is a dry plaie, this is a dry one. The fact that the thing can be
instated as the Head of a nominal group serving as Attribute illustrates the point already
made: the quality does not construe a separate class of thing, it presumes this class from

208

5. Elements

the environment. Thus this is heavy means that it is heavy relatively to whatever class of
thing it has been assigned to; compare the truck was very heavy/ a very heavy one (i.e.
"heavy for a truck') with the chair was very heavy/ a very heavy one (i.e. 'heavy for a
chair'). Hence it is not possible to re-instate the thing where the Attribute occurs as
resultative in a process of doing: we do not say I've wiped it a dry plate or I've wiped the
plate a dry one.
Thus in its typical construal, as Epithet or as Attribute, a quality is clearly
"participant-like"; we might also note that realized as an adjective in superlative form it
does appear as a participant (these are the driest, pass me the driest; the smallest will fall
through the holes). But there are also environments where a quality resembles a process or
a circumstance.
(i) Some qualities can be construed as processes of doing; here for example there is an
agnate form I've dried the plate, with 'dry' worded as a verb. In these cases, the quality is
not repeated as an Attribute we do not usually say I've dried the plate dry; it may
however reappear in an intensified form, e.g. I've dried it very dry. Very many qualities
may be construed as verbs in this way.

Fig. 5-11: The intermediate status of qualities


(ii) Some qualities may occur as depictive (as opposed to resultative) Attribute in a
figure of doing & happening; in such instances the quality is very close to a circumstance
of manner, as is shown by agnate pairs such as the following:
quality as Attribute:

quality as Manner:

she came home cheerful

she came home cheerfully

he walked in drunk

he walked in drunkenly

5.3 Participants

209

As always with such closely agnate expressions, while they are semantically related they
are not synonymous; we could even imagine a figure such as he walked in drunk quite
soberly.

But they make the point that a quality, when attached to the figure as an

Attribute (rather than to a participant as in the drunken man walked in/ the man who
walked in was drunk), is construed as being more like a circumstance. The fact that
manner circumstances are typically realized by adverbs that are simply derived from (and in
some cases identical with) adjectives is a further symptom of the way a quality may
resemble a circumstance.
Figure 5-11 above summarizes the multivalent status of quality.
5,3.3,2 Types of quality
We suggested in Sections 5.2-3 above that the nominal group is organized as a move
along two semantic dimensions: the elements become increasingly stable in time, and
increasingly complex in their taxonomy of features. Lexicogrammatically, this
corresponds to a move from grammatical items (determiners, determinative adjectives such
as usual, same, typical, cardinal and ordinal numerals) to lexical items (adjectives [in
general], and nouns); that is, a move from closed systems to open sets. The former aie
taxonomically simple (although they are notoriously difficult to interpret in lexical
glosses); they include specific/ non-specific; personal/ demonstrative; near/ far, total/
partial &c (see Haliiday, 1976:131-5, for the systems). In contrast, elements at the latter
end tend to be construed in complex taxonomies; see for example the multiple taxonomy
involved in the lexical construal of clothing (Section 5.3.2.5 above). That is, greater
experiential complexity is handled by means of greater taxonomic complexity. The
semantic movement in the nominal group is summarized in Figure 5-12. Qualities lie at
differences places along these various dimensions; hence they vary in their potential for
taking on roles in different types of figure.3 Table 5(4) suggests a very tentative
classification.
Qualities can be distinguished according to the transphenomenal types of projection
and expansion. Qualities of projection and qualities of expansion differ in a number of
respects. Most fundamentally, they differ in their patterns of agnation. Qualities of
projection are agnate with processes in figures of sensing; for example, happy

in the

happy child (or the child is happy) is agnate with rejoice in the child rejoices.

In

contrast, qualities of expansion display patterns of agnation within figures of being &
having, with variation according to subtype (see below). This fundamental difference
explains other differences; for example, qualities of projection tend to occur in agnate pairs

Thompson (1988) offers a complementary account of 'adjectives' based on their use in


discourse.

5. Elements

210

of the like* and 'please1 type that we find with figures of sensing (e.g. afraid/ scary,
suspicious/suspect, bored/boring), whereas qualities of expansion do not. We will review
qualities of projection first and then turn to qualities of expansion.

Fig. 5-12: Clines in the construal and realization of participants


(i) Qualities of projection are all scalar and they are, as we just noted, agnate
with processes of sensing in figures of sensing; they are often realized by participial verb
forms used as adjectives. The thing they are assigned to, as Attribute or as Epithet, is
agnate either with the Senser (the 'like* type, realized by v-en if verbal in origin: happy,
sad, angry, afraid, frightened, certain, sure) or with the Phenomenon (the 'please' type,
realized by v-ing if verbal in origin: sad(dening), tragic, irritating, scary, certain). From
this it follows that when they are agnate with the Senser, they are ascribed to conscious
beings (as in an angry child/ boss/ cat), whereas when they are agnate with the
Phenomenon, they can be ascribed not only to things but also to metathings, i.e. to
projections (as in it's sad(dening)/tragic/irritating/ scary/certain that they ignore world
opinion). In the former case, these qualities may be 'transferred' to tokens of the senser's
sensing, as in on angry face/ look/ letter/reaction. As we have seen, sensing is agnate
with modality and, in the case of emotive sensing, with attitude. Similarly, qualities of
projection extend to include modalities and attitudes; when the thing (or metathing) they
are assigned to is agnate with the Phenomenon, these qualities aie construed as objective,
impersonal ones: it is certain/ likely/ possible that the moon's a balloon. (Contrast: I'm
certain the moon's a balloon.) With qualities of usuality, this is the only possible
orientation. Qualities of projection are unlikely or impossible as the Attribute of figures
of doing (thus it is hard to interpret she polished it certain [contrast she polished it clean]
and while we can say he drove her mad [caused figure of being] and he drove the car hot
[figure of doing with Attribute], we cannot say he drove the car mad).
(ii) Qualities of expansion cannot normally be assigned to metathings. They
expand the thing they are related to by elaboration, extension or enhancement. These
different subtypes display different patterns of agnation, but as Epithet (or Classifier) +
Thing configurations they are all agnate with figures of being & having.

211

5.3 Participants

Table 5(4): Tentative classification of qualities


taxonomic
type

qualities of:

f*J

agnate
figure
Epithet
(Classif.)
figure of
sensing:
Process

(as souing:)

(as assessment:)

emotive

attitude
(evaluation)

scalar

rejoice,
grieve

cognitive

probability

scalar

desiderative

modulation

scalar

suppose,
believe,
know
ike, want,
desire

usuality

scalar

exp.
elaboration

attribution

class
status

sensemeasur
e

national.
material
etc.
life

sex

marital
etc.
visual:
colour A;
shape
weight
texture

taxonomic

figure of
being:
Attribute

examples
Thing

Senser/
Phen.
happy, sod',
delightful, tragic;
good, bad
doubtful, sure;
likely, certain
willing, keen;
desirable,
necessary
usual, common
Carrier

binary

alive, dead
female, male
single, married

taxonomic

red, blue, pink;


round, oval,
square
heavy, light
rough, smooth
old, young

scalar

ape

propensity
quantity

identity

taxonomic
scalar

temporal

taxonomic

spatial

taxonomic

causal

skilful, naughty

scalar
[dynamic]
scalar

taxonomic

extension

enhancement

etc.
(behav.
qual.)
inexact
exact

That, Burmese
plastic, wooaen

Process:
be,
resemble
Process:
accompany,
replace, be
instead of
Process:
be before,
after, at
be above,
within,
ourside

Carrier,
Token

few, many
one, two, three
same, similar,
analogous,
different
alternative,
contrasting
previous,
preceding,
subsequent
interior, external,
anterior,
posterior
consequent,
resultant;
conditional,
contingent

The most prototypical qualities are those of elaboration, subtype attribution:


they stand in a purely intensive relation to the thing they are assigned to, being construed
as inherent qualities. They are typically scalar, but certain types aie taxonomic or binary.
As Epithets (or Classifiers) they are agnate with the Attribute of a figure of being. Many
of them can serve as the Attribute of a figure of doing (as in he squashed it flat, she
painted it blue) and are related to the outcome of such figures (cf. she heated/ widened/
enlarged it : it was hot/ wide/ large). The limiting case of qualities of attribution are
quantities; they are assigned to a thing as a discoursal instance rather than as a general

212

5. Elements

experiential class. Consequently, they do not serve as Epithets; they have a special role,
that of Numerative (as in two/ many brave volunteers).
Qualities of elaboration, subtype identity are not inherent properties, but rather ate
comparative. Thus the standard of comparison can always be construed: their car is the
same/similar/different : their car is the same as/similar to/different from ours; and they
are agnate with processes: their car is/ resembles/ differs from ours. The line between
elaboration: identity and enhancement: manner, comparison is a fuzzy one; and processual
agnates such as their car resembles ours are within the domain of enhancement
Qualities of extension resemble those of 'elaboration: identity1 in that they construe
a relation between the thing they are assigned to and some other thing. The relation
typically obtains between the thing as a discoursal instance rather than the thing as a
general experiential class. Thus an alternative solution is a solution that can replace the
one we have just been talking about Consequently, properties of extension tend to serve
as deictic elements (Post-Deictic) rather than as Epithets: they indicate how an instance (or
instances) of the general class of the thing they are assigned to is selected from that class:
we need an additional two volunteers means 'two further instances of the general class of
volunteer'. Like other properties serving as Post-Deictic, they may precede the Numerative
(additionalA two); but they may also follow, with no strong contrast in meaning: we need
two additional volunteers.
Qualities of enhancement also resemble those of 'elaboration: identity1 and those of
extension in that they construe a relation between the thing they are assigned to and
another thing. This is brought out by the fact that they are agnate with processes relating
participants circumstantially: subsequent : be (come) after, preceding : be (come) before/
precede, interior : be within, exterior : be outside, as in previous occasions : occasions
coming before this one, interior design : design of what is within a house. This relation is
typically a temporal or spatial one involving the thing as a discoursal instance rather than
the thing as a general experiential class. Consequently, like qualities of extension,
qualities of enhancement tend to serve as deictic elements (Post-Deictic) rather than as
Epithets; as Post-Deictics, they come before Numeratives in the structure of the nominal
group: the preceding/ subsequent two meetings. Here they relate a specific referent,
recoverable in the current situation: the subsequent two meetings : the two meetings that
followed this one. But the spatial qualities can also, being taxonomic, serve to subclassify
the thing they are assigned to: interior monologue, external pipes. Here they relate to a
general class of thing, inferrable from the experiential system: interior monologue :
monologue that is within a person, external pipes : pipes that are outside a house.

5.4 Processes

213

5.4 Processes
Much of what we have to say about processes as elements of figures has already been
brought into the discussion, either under "figures" (Chapter 4) or in the section on
elements in general (Sections 5.1 and 5.2). This was inevitable (i) because a figure is the
semantic construction of a process, having the structural element Process (realized as
verbal group) at its core, and (ii) because in understanding the concept of participant we
are naturally led to contrast participants with processes.
Let us summarize the essential points. The key to the construal of experience is the
perception of change; the grammar construes a quantum of change as a figure (typically
one clause) and sorts out figures in the first instance into those of consciousness (sensing
and saying), those of the material world (doing & happening) and those of logical
relations (being & having). The central element of a figure is the process; 'things1 aie
construed as entities participating in processes, having different roles, of which one is
'that participant in which the process is actualized' (if there is 'flying1, there has to be
something that flies or is flown: birds fly, people fly kites)', hence the grammatical
nucleus of the clause is the configuration of Process with Medium.
While participants aie located in referential space, proceses are located in time. The
verbal group realizing a process constructs a "moment" in time beginning with the 'now*
(the time of speaking) leading up to a categorization of the Event; this is analogous to the
way the nominal group, realizing a participant, constructs a "body" in space beginning
with the 'here* and leading up to a categorization of the Thing. But while the Thing is
enmeshed in a elaborate taxonomy of things, the Event is taxonomically rather simple and
its complexity lies in the construal of time itself. Hence the verbal group is lexically
sparse typically the Event is the only lexicalized part; whereas nominal groups can be
lexically extremely dense (cf. Figures 5-2 and 5-3 above).
Thus from the point of view of the figure, a process is the central element, forming a
nucleus around which participants and circumstantial elements are organized into a
meaningful pattern (Figure 4-13). From the point of view of its own internal
organization, a process is the construal of 'eventing* a phenomenon perceived as having
extension in time. In the rest of this section we shall look briefly at the parameters along
which time is construed, with special reference as throughout to English.
5.4.1 Making sense of time
It is not easy to construe experience of time, and different languages vary considerably in
the way they do it: there are differences from one language to another, and differences
within the same language over the course of time. Like everything else we aie exploring
here, the grammar's model of time has been evolving unconsciously in the context of

214

5. Elements

human survival; it is part of the selective and collective wisdom that the species has
accumulated in the understanding of its relationship to its environment and in the
interaction of its members one with another. And again like everything else in the
construal of experience it is the product of continual compromise, whereby divergent ad
often conflicting aspects of experience are adjusted and accommodated in such a way that
all of them have some place in the total picture.
In transforming experience of time into meaning, human communities have evolved a
number of basic parameters. We can identify four of these that are relevant in the present
context:
(1) the temporal staging of a process: it may be beginning, taking place or
ending.
(underlying concept: a process occupies a certain measure of time)
(2) the temporal perspective on a process: we may frame it in or out of temporal
focus. This takes many different guises in different languages, and even within
the same language; such as (a) in focus: ongoing, out of focus: terminated; (b) in
focus: significant in itself, out of focus: significant for what follows; (c) in
focus: actualized, out of focus: visualized. (It is the last of these that is relevant
to English.)
(underlying concept: a process relates to the flow of experience as a whole,
including other processes)
(3) the temporal profile of a process: it is either unbounded or bounded,
(underlying concept: a process has the potential for being extended in time)
(4) the temporal location of a process: it can be related to 'now' as past, present
or future.
(underlying concept: a process takes place within a linear flow or current of time)
Each of these variables differs from all the others; but at the same time, each is related
to all the others, so that there are certain patterns of association among them. For
example, a process that is unbounded (e.g. travel) is more likely than one that is bounded
(e.g. arrive) to be put under temporal focus (e.g. while travelling is more likely than
while arriving); a process located in the future is more likely to be beginning than ending
(e.g. it will start warming up is more common than it will finish warming up). Some
combinations may be more or less excluded: for example, a process that is beginning can
vary in its perspective (e.g. the sun started to shine/ started shining), whereas one that is

5.4 Processes

215

ending is always actualized (e.g. the sun stopped shining; but not the sun stopped to
shine). Thus, in any given language, (i) one or other parameter may be given prominence,
(ii) two or more parameters may be combined into a single semantic system, (iii) any
parameter may be construed either more grammatically or more lexically, and (iv) a
number of features that are not strictly temporal may be incorporated into the picture,
both ideational ones like attempting/ succeeding and interpersonal ones like the speaker's
angle on the process judgement of its likelihood, desirability, and so on.
When these parameters are grammaticized, they are referred to respectively as (1) phase,
(2) aspect, (3) aktionsart, (4) tense.
5.4.2 Patterns of time in English

English foregrounds location in the flow of time (tense), and construes this not only as
past/ present/ future relative to 'now1 [they paid me/they pay me/ they mil pay me], but
also as past/ present/ future relative to some moment that is relative to now [they am
going to pay me (future in present), they've been paying me (present in past in present)],
with the possibility of up to five shifts of reference point, as in
They said they'd been going to've been paying me all this time . ..

(present in past in future in past in past). This system is fully grammaticized, and is
unusual in that it construes location in time as a logical relation rather than as an
experiential taxonomy; it thus becomes a form of serial time reference. The tense
categories also combine with time adverbs such as already, just, soon [they'd already paid
me, they've just paid me, they soon paid me]. Interestingly, the deictic time reference (that
appealing to 'now') can be switched off; either there is no deixis (the clause is non-finite,
e.g. not having paid me yett...) or the deixis takes the form of modality (speaker's angle
on the process, e.g. they should have paid me).
By comparison with temporal location, temporal perspective (aspect) is relatively
backgrounded in English. (Some 20th century grammarians have interpreted the serial
location as a kind of aspect, with "present in ..." as continuous, "past in ..." as perfect;
but the earner description as we have presented it here accounts more richly for the
semantic patterning.) The temporal perspective takes over, however, when there is no
deictic location (the clause is non-finite); in such cases, instead of making reference to
'now1 the process is construed as either actualized, as in (on) reaching the gallery, turn left,
or visualized, e.g. to reach the gallery, turn left. Sometimes the difference in meaning is
very slight (e.g. a way of doing it/the way to do it)', but it is always there. (We will see
in Chapter 7 that in Chinese, while temporal location is relatively backgrounded,
temporal perspective is foregrounded though the shift of focus is of a diffrait kind
from that in English.)

216

5. Elements

Temporal staging is explicit and lexicalized, with a basic system of categories as


shown in Figure 5-13; for example (this was the continuation of the five-term tense
example above)
. .. but the money kept on not coming through

The combination of staging with perspective has already been referred to; note in this
connection sequences such as
started doing

went on doing

stopped doing

s tau: ted to do

went on to do

stopped to do

where the move in staging creates an increasing semantic distance between the two
perspectives. Staging also extends to other categories that are not strictly temporal, which
the grammar however construes as analogous: especially conation (tries to do/ succeeds in
doing), and appearance (seems to be/turns out to be).

Fig. 5-13: Different types of phase


The temporal profile, of unbounded/ bounded, is not an independent option available to
processes in general. Boundedness is a feature which may accrue to certain classes of
process, typically processes of doing: contrast for example unbounded use with bounded
use up (use some salt but don't use it up!). It is on the borderline of lexis and
grammar: the verb is extended by up or some other adverb of the locative-directional class
(drink up, eat up, load up, pour out, melt down, fly away). The boundedness is not in fact
temporally defined; its degree and often its exact nature is specific to the kind of
process concerned, and may depend on the total figure and on the context; e.g.
pour out the water

'until the bowl is empty/ until everyone has a drink'

write up the results

'in a publishable form'

But there is a clear proportionality involved, which allows us to treat this as a systemic
feature of a significant subset of processes in English.

5.4 Processes

217

Obviously, choices made in all these parameters relate to the figure as a whole (at
least; they may have implications for longer stretches of discourse). The major process
types the primary categories of sensing, saying, doing & happening, being & having
tend to have inherently different temporal characteristics, affecting all aspects of the
way time is construed: in terms of what temporal categories are possible, what the relative
probabilities are, and what the different choices mean (note for example the total lack of
proportionality in pairs such as / go : I'm going * I know : I'm knowing). Part of the
meaning of any fragment of experience is its potentiality for being variably construed in
time.
From an ideational point of view, the temporality of processes is their salient
characteristic. As we saw, there are no semantic or grammatical systems for construing
processes into elaborate taxonomies, as there are for construing things. Apart from
considerations of time we have also shown that processes can be classified according to
their potential for serving in figures of different types (see Chapter 4 above). We shall
raise again the question of differentiating between processes and figures when we consider
alternative treatments in Chapter 11, Section 11.3.2 below.

5.5 Circumstances
We noted in Chapter 4, Section 4.5 that circumstantial elements can be realized
grammatically either as adverbial groups or as prepositional phrases. These represent
circumstances of two different types.
Type (1) are qualities ("simple circumstances" in Chapter 2, Section 2.8) but
construed not as qualities of a particular participant but as qualities of a figure as a whole;
for example, steadily, extremely loudly, perfectly in it rained steadily all night, they were
shouting extremely loudly, it suits your complexion perfectly. Typically such adverbs are
derived in the grammar from adjectives, with the added suffix -/y; a few have special
forms, like well (from good), and sometimes the same form is both adjective and adverb,
e.g. fast in a fast car, she drives fast.
The usual function is as circumstance of Manner, with the meaning 'in such a way',
'to such a degree1; and if the manner of doing determines the quality of the outcome there
may be very little difference between a circumstance of this kind and a resultative
Attribute: cf. don't chop the parsley too fine/ too finely. There are some adverbial
expressions realizing other types of circumstance, e.g. everywhere (Location: space),
recently (Location: time); as well as others which might be interpreted differently because
of the nature of the quality itself, e.g. pointlessly 'in a pointless manner* or 'for no good
reason' (Cause).

218

5. Elements

Sometimes the Manner element is a quality of the process itself, rather than the
manner in which it unfolds, for example he was falsely accused/wrongly dismissed; these
tend to be bonded rather closely to the Process element in the clause. (The adverbial form
also functions as a quality of a quality, like frostily, superficially in a frostily polite
receptionist, what he says is superficially correct; here it is not a circumstantial dement
but is part of the Epithet in the nominal group.)
Finally, some adverbial groups are interpersonal in function the speaker's comment
on the figure, like sensibly in sensibly, they didn't argue 'I consider their behaviour
sensible* (contrast they didn't argue sensibly 'in a sensible manner'). These lie outside the
ideational structure of the clause: they are not serving as qualities in the figure realized by
the clause, but rather derive from features within the interaction base.
Type (2), those circumstantial elements that are realized by prepositional phrases, aie
rather more complex, since they include another element a participant in their
makeup ("macro circumstances" in Chapter 2, Section 2.8): e.g. the table, my knowledge,
peace and quiet in right under the table, without my knowledge, for the sake of peace end
quiet. In such cases the element realized by the nominal group is still functioning as a
participant in the process but indirectly, being implicated only through the mediation
of a preposition. That this is possible is because the preposition itself constitutes a
subsidiary kind of 'process'; one that does not function as Process in the main figure but
is nevertheless related systematically to the spectrum of process types mainly, though
not exclusively, to processes of being.
In Chapter 4, Section 4.5, above we tried to suggest how the circumstantial elements
were related to participants. There, we were looking at them as oblique "cases", from
the point of view of their function in the larger figure. Two points emerged: one, that
participants and circumstances taken together formed a cline, rather than being separated
by a clear boundary; the second, that some of the circumstantial elements could be 'paired
off with participants, being seen as a more oblique manifestation of a similar role. We
were able to incorporate these two points in a helical form of presentation in Figure 4-13.
We are now looking at these same circumstantial elements from a complementary
standpoint, from the point of view of their own internal composition. As we remarked, a
prepositional phrase represents a figure in miniature, with a structure analogous to one
component of a figure closest, perhaps, to Process + Range (so we refer to the
participant in the circumstantial phrase as a "Minirange"). This means that, shifting our
perspective, we can also suggest how the circumstantial elements are related to different
figures. Table 5(5) combines the two perspectives, showing their relationship both to
participants and to figures. We referred to the cline from participants to circumstances as
the "degree of involvement" in the actualization of the process. This degree of
involvement ranged from the closest, the Medium, which is pan of the nucleus of the

5.5 Circumstances

219

figure, to those that appear most remote, circumstances such as Matter (e.g. concerning
your request) and Angle (e.g. in my own opinion). Somewhere in the middle was an area
of overlap, where participants and circumstances are very closely akin. The two overlap
because there are some functions which can be construed either as a form of participant in
the process or as a circumstance attendant on it
Table 5(5): Agnation between participants, circumstances and figures

220

5. Elements

The grammar does draw a line between the two: participant as nominal group,
circumstance as prepositional phrase. But because of the continuous nature of the
distinction, we find three kinds of mixed categories: (i) participants that may look like
circumstances (being introduced by prepositions); (ii) circumstances that may look like
participants (being introduced without prepositions); and (iii) pairs where one is
circumstance, the other participant, but with very little difference in their meaning. These
are illustrated in turn.
(i) Participants introduced by prepositions:
(Recipient) give milk to tha cat

cf. give tha eat milk

(Range)

play all night at pokar

cf. play pokar all night

(Actor)

a bua ran over them

cf. they were run over by a bua

(ii) Circumstantial elements introduced without prepositions:


(Extent)

she jogs 30 minuta

cf. she jogs for 30 minutas

(Location) we're meeting naxt Sunday cf.we're meeting on naxt


Sunday

That those in (i) are being construed primarily as participants, whereas those in (ii) aie
being construed primarily as circumstances, is shown by a variety of other grammatical
factors; to give just one example, the question equivalent to (i) is who? or what? (who
did you give milk to?, not where did you give milk?), whereas the question equivalent to
those in (ii) is how long?, when? (when are we meeting?, not what are we meeting on?).

5.5 Circumstances

221

By the same token, in (iii) we have two different elements, one participant and one
circumstance, but with hardly any difference between the two:
(iii) Participant/ circumstance pairs with little difference in meaning:
(Range) cross the canal (Location) cross over the canal [cross what?]
[cross where?]
(Client) bring the children presents

(Cause) bring presents for the


children

[bring who presents?]

[why bring presents?]

Coming back now to the distinction between type (1) circumstances, those realized by
adverbial groups, and type (2), those realized by prepositional phrases: as typically
happens in language (since grammar abhors detenninacy), we find a crossing between the
two types. This happens mainly in one direction: there are prepositional phrases which
construe qualities of figures, and hence function as circumstances of Manner; for example
in a hurry, without proper care (cf. hurriedly, carelessly). Typically these may also appear
as qualities of participants, as in he was in a hurry (realized as Attribute in a clause), a
man in a hurry (Qualifier in a nominal group not Epithet, since English does not like
phrases and clauses before the Thing). These usually involve some kind of metaphor,
either (as here) grammatical metaphor, where a quality or process is made to look like a
participant (see the next chapter), or lexical metaphor (metaphor in its traditional sense) as
in they left the matter up in the air. Less commonly, we find a cross-over in the other
direction; an example would be the adverb microscopically where this has the sense of
'using a microscope1 (we examined the tissue microscopically),
It is interesting to note how this two-faced character of circumstances that they aie
on the one hand like participants and on the other hand like figures is reflected in their
treatment in languages other than English. In the way they are construed in the grammar,
in a language such as Finnish, where what corresponds to the English preposition is often
a "case" in the nominal group, they appear to be (relatively to English) closer to
participants; while in a language such as Chinese, where what corresponds to the English
preposition is typically a class of verb, they appear (again, relatively to English) closer to
figures. But they typically seem to have a status that lies somewhere intermediate between
the two.
But even within one language, however they are construed grammatically, the status of
circumstantial elements is variable. In English, those that are nearer the centre of our
helix are more like participants, as shown by the examples above; while those that are on
the periphery are more like figures. We can give some examples of agnate pairs of this
kind.

222

5. Elements

circumstance
Role

figure

Accompaniment

as a child (he couldn't tell the


difference)
without enough money (he
was unable to complete his
education)

Cause: condition

in case of fire (call 9000)

when he was a child (he couldn't


tell the difference)
not having enough money /since
he hadn't got enough money (he
was unable to complete his
education)
if there's afire (call 9000)

5.6 Major motifs in the ideation base


The ideation base construes experience as a vast, multidimensional and highly elastic
semantic space. We have only been able to survey a small fraction of this space. But we
have adopted the perspective that we believe most clearly reveals the principles on which
the ideation base is organized, highlighting the least delicate end of the system so as to
produce a map of the system as a whole rather than detailed maps of isolated regions. This
means that, as seen from below, we have approached the ideation base from the
grammatical pole of lexicogrammar rather than from the lexical pole.
In addition to setting out the main dimensions and systems of ideational meaning, we
have tried with this approach to bring out a small number of very general motifs that run
throughout the grammar's construal of experience.
The first of these motifs is that of meaning as expansion: the way regions of
semantic space are opened up and defined by the three vectors of elaboration, extension and
enhancement elaborating a region that is already as it were staked out, extending the
regions boundaries to take in more, and enhancing the region's potential by enrichment
from its environment.
The second motif is that of meaning by projection: the way new dimensions of
semantic space are created by the orders of human consciousness, sensing and saying
by projecting into existence another order of reality, one that is constituted by language
itself.
These two motifs together have made it possible for human beings to transform
experience into meaning, taking the experience of meaning itself the "inner" processes
of consciousness as the central figures, and those with the ability to mean
prototypically humans themselves as the central participants. These then serve as the
point of reference for construing "outer" experience, the complementary experiences of the
processes of doing and of being. Figure 5*14 shows the two motifs manifested in the
environments of sequences, figures and elements.

5.6 Major motifs in the ideation base

223

Fractal types:

Fig. 5-14: Two fundamental semantic motifs


Since projection and expansion operate across the various categories of phenomena, we
referred to them as transphenomenal categories in Chapter 3, Section 3.8 above. As
transphenomenal categories, they are meaning types that are in some sense "meta" to the
organization of the ideation base: they are principles of construing our experience of the
world that generate identical patterns of semantic organization which are of variable
magnitude and which occur in variable semantic environments. Such patterns therefore
constitute fractal types. Projection and expansion are manifested at three levels of
organization: as sequences, as figures and as elements. Table 5(6) presents a partial
summary of the cases we have discussed in Chapters 3 through 5.
Sequences, figures and elements, and their subclasses, constitute different
environments within the ideation base. If we construe some phenomenon of experience as
a sequence of qualification, we give it a location in the overall ideational system that
is quite distinct from the location it would be assigned if it was construed as a figure of
circumstantial cause. These two types are thus not particularly close agnates in the
system, However, the fractal types constitute an additional order of agnation that is

5. Elements

224

projected onto the ideational system as a whole. We can refer to this as fractal
agnation. Because of this, a qualifying sequence and a figure of circumstantial being,
such as cause, aie agnate: they are both manifestations of the fractal type of
enhancement For example, the sequence ebola broke out so 52 people died is agnate with
the figure the outbreak of ebola caused 52 deaths even though they are quite far apart in
their semantic structure.
Table 5(6): Projection and expansion manifested throughout the ideation base

sequences
figures

elements

process +
participants

projection

expansion

sequences
of projection
figures of
projection

sequences of
expansion:
figures of
expansion:

sensing
saying

circumstances

Angle
Matter

participants:
things

conscious/
(nonconscious)
names of
projections
(fact, idea,
possibility,
chance etc.)
qualities of
projection
(attitude,
probability,
modulation)

participants:
qualities

elaboration extension
restating
adding

enhancement
qualifying

doing:

elaborating
outcome

being:

intensive

enhancing
outcome
(motion etc.)
circumstantial
Location
Extent
Frequency
Manner
Cause

extending
outcome
(transfer)
possessive

Role

Accompani
-ment

names of
expansions:

type, kind,
example
etc.

part,
element;
unit etc.

qualities of
expansion:

attribution,
identity

temporal,
addition,
alternation, spatial,
contrast
causal

time, place,
way etc.

The fractal types of projection and expansion aie also a primary resource by which the
semantic system creates new meanings: we illustrated this auto-genetic potential in
Chapter 2, Section 2.11 above. The ideation base thus itself embodies, auto-genetically,
the principles on which it is organized and enabled to develop further, such that the
primary systems of ideational meaning then serve as a grid within which more delicate
categories are construed. Here we have foregrounded especially the motif of elaborating,
with particular reference to its manifestation in the identifying and ascriptive figures of
being. We have tried to show how elaboration makes it possible to "impon" extralinguistic experience into the meaning base by actively construing it (as in 'that [thing
there] is a circle'); and also to "transpon*1 meanings internally from one region of the
ideation base in order to construe new meanings in another (as in 'balance means you hold
it on your fingers and it does not go'). Hie extension of meaning in delicacy not

225

5.6 Major motifs in the ideation base

merely generalizing across different types but construing such types into dimensional and
open-ended taxonomies is a function of the elaborating potential, exploiting the basic
dimensions of the system itself.
There are other "transphenomenaT motifs, often related to these, which are more
specific in their scope; for example, the foregrounding of perceptual space, and of the
concrete having extension in space, so that these serve as models for construing more
abstract, non-spatial realms; and, more specifically, the spatial construction of the human
body as an orientational framework.
Thus our concept of "construing experience through meaning" refers to the construal
in human consciousness of an ideational system in which such motifs play a critical part.
Expansion and projection are, as we put it earlier, fractal principles; they generate
organization within many environments in the ideation base, at different strata and at
different ranks within one stratum. These environments are thus related to one another
through the local manifestations of these different motifs; and this opens up the system's
potential for alternative consonis of experience: for example, the types of expansion
create new meaning potential through "figures of speech" (see Figure 5-15). Specifically
each type of expansion contributes one figurative mode, as in the following examples:

elaboration
intensive
extension
possessive

enhancement
circumstantial

restatement: a resemblance metaphor


is construed between A and
B so that A can be restated
in terms of B
synecdoche
metonymy: a part-whole
relationship is construed
between M and N so that so
that part M can stand for
the whole N
metonymy
a circumstantial
relationship is construed
between O and P so that
circumstance can stand
for that which it is
circumstantial to, P

Example
INTENSITY => LOCATION
IN VERTICAL SPACE, e.g.
He loved her very much =>
He loved her deeply
BODY => BODY PART,
e.g. She has many people to
feed ==> She has many
mouths to feed
OPERATION => MANNER
OF OPERATION, e.g. Think!
=> Use your brains!

In these examples the figures of speech are largely lexical; but the principle they
illustrate that because of the "play" that occurs between diffrait strata the system has
the potential for construing figurative meanings extends throughout the grammar, as
we shall see in the next chapter. What this means is, that whatever is construed can also
be reconstrued, giving yet another dimension to the topology of semantic space.
The ideation base as we have presented it so far, with its framework of sequences,
figures and elements, serves well enough for construing the experience of daily life, and
for organizing and exchanging commonsense knowledge. But it proves inadequate to meet
the semiotic demands of advanced technology and theoretical science. In the construction

5. Elements

226

of scientific knowledge, the system needs to invoke the power of metaphor on a more
global scale.

Fig* 5-15: Figurative expansion of the ideation base

6. Grammatical metaphor
In Chapter 2, we presented an overview of the meaning base. In this overview, we
assumed that phenomena were construed in a 'congruent' form, and the details were then
discussed in Chapters 3 (sequences), 4 (figures) and 5 (elements). To conclude this
summary account of the meaning base of English, we shall consider how these resources
may be expanded, through the deployment of grammaucal metaphor (see also Halliday,
1985: Ch, 10; in press a, b; Halliday & Martin, 1993; Matthiessen, 1995b: Sections
1.4.3, 2.3.2, 3.1.6.1 and 4.12; Ravelli, 1985, 1988).
6.1 Congruent and metaphorical variants
We have seen that sequences, figures, and elements are congruentiy realized in the
grammar as follows:
clause complex
sequence
figure
element

clause
V

group
V

But these resources may be expanded by taking up further options in realization; for
example, sequences may alternatively be realized by clauses and even groups. This is
what we refer to as grammatical metaphor. Grammatical metaphor expands the semanuc
potential of the system.
Suppose that in the course of a text we came across the following portion of wording:
Smith et al. have shown that if one takes alcohol one's brain
rapidly becomes dull. Alcohol's rapid dulling effect on the brain
has also been observed by other researchers in the field.

What is the semantic relationship here between if one takes alcohol one's brain rapidly
becomes dull and alcohol's rapid dulling effect on the brain!
Grammatically, of course, the two have different status in rank: the former is a clause
complex, hypotactic with rising dependency xp A a, whereas the latter is a nominal
group, structure Deictic A EpithetA Classifier A Thing A Qualifier. If each is functioning in
its unmarked role in the realization of a semantic unit, then the first represents a
sequence, the second represents an element. We can also find a third variant that would
represent a figure, intermediate between the two; e.g. the effect of alcohol is a rapa
dulling of the brain.
Here are some other pairs of wordings displaying similarly agnate relationships:

228

6. Grammatical metaphor

strength was needed to meet


driver safety requirements in
the event of missile inpact

it had to be strong because


the driver needed to be safe
if it was inpacted by a
missile

he also credits his former big


he also believes he was
size with much of his career
successful in his career
success
mainly because he was formerly
big
some animals rely on their
great speed to escape from
danger

some animals are able to


escape from danger by moving
very fast

In each pair of examples the grammatical status vanes between clause and clause
complex; and it would not be difficult also to construct a third variant which is a nominal
group, e.g. the strength needed to..., his crediting of..., the reliance of some animals on
... . This would then function as an element in some other figure; for example, the
strength needed to meet driver safety requirements in the event of missile impact was
provided by the use of specially toughened glass.
In order to get a sense of the grammatical relationship between the members of pairs
such as these, let us model one such pair using a simpler example in terms of a
systemic-functional grammar (but taking account only of the ideauonal metafunction):
see Figure 6-1.
If we take as our startingpoint the wording the cast acted brilliantly so the audience
applauded for a long time and derive from it the agnate wording the cast's brilliant acting
drew lengthy applause from the audience, we find the rewoidings as set out in Table 6(1).
Let us summarize some of these in a prose commentary, (i) The verbs acted and
applauded, each functioning as Event in a verbal group functioning as Process, each in a
different clause, have been replaced by verb acting, noun applause, each functioning as
Thing in a nominal group, the two nominal groups functioning respectively as Actor and
Goal in the same clause, (ii) The conjunction so, functioning as logical-semantic relation
between two clauses, has been replaced by the verb drew functioning as Event in a verbal
group functioning as Process in a single clause; thus:
. . . acted ... so ... applauded => . , . acting drew ... applause
(iii) The adverb brilliantly, functioning as Quality in an adverbial group functioning
as Manner in clause (1), has been replaced by the adjective brilliant, functioning as
Epithet in nominal group (1); while the expression for a long time (in origin a
prepositional phrase but now codified as a single item) functioning as Duration in clause
(2) has been replaced by an adjective lengthy functioning as Epithet in nominal group
(2).

229

6.1 Congruent and metaphorical variants

Fig. 6-1: Correspondence between congruent and metaphorical wordings


Table 6(1): Congruent wording and metaphoncal rewording
metaphorical

congruent
group/
phrase
class
nom.gp.

function in
clause

word class

function in group/
group/
phrase
phrase
class

function in
clause

Actor

nom.gp.
(poss.)
rankfhifted
verb

Deictic

nom.gp.

Actor

adjective

Epithet

Event

vbl.gp.

Tocis

Range

prep,
phr.
nom.gp.

hace

'cast'

noun

function in
group/
phrase
Thing

'act'

verb

Event

vbl.gp.

Process

'brilliant'
'so'

adverb
conjunction

Quality

advbl. gp.

Manner
'">

wordclass

'audience'
'applaud'

long time'

noun

Thing

nom.gp.

Actor

verb

Event

vbl.gp.
prep,
phrase

Process

Duration

vero

nom.gorandshifted
noun
adjective

I'kmg

Thing
Epithet

Goal

230

6. Grammatical metaphor

It takes a lot of space to set out the details of such an analysis; but it will be dear
that we could express the relationship between any pair of agnate expressions in explicit
terms in the same general way. To cite two more examples:
The truest confirmation of the
accuracy of our knowledge is
the effectiveness of our
actions.

If we act effectively this


most truly confirms that we
know [things] accurately,

Griffith's energy balance


approach to strength and
fracture also suggested the
importance of surface chemistry
in the mechanical behaviour
of brittle materials. chemistry was
[determining]

Because Griffith approached


[the study of ] strength and
fracture [using the concept
of] balance of energy, we
realized that surface
inportant in
how brittle
materials behaved [under]
mechanical [conditions].

We have set these examples out in pairs; but it will be clear that given any one such
example, there will be a number of different agnate expressions corresponding to it.
Furthermore, between the two variants in each of these pairs it would be possible to
construct a number of intermediate steps: intermediate in the sense that there would be
minimal distance between any one variant and those on either side of it. Thus an example
of an intermediate variant in the first pair above would be
The fact that our knowledge is accurate is nvost truly confirmed by
the fact that our actions are effective.

Compare the original example constructed in Chapter 1, Section 1.2, above.


Before coming to a more theoretical discussion of the issues raised by these examples,
let us note two related general properties that they display. One is that at some point in
the set of rewordings there is typically a shift on the grammatical rank scale: from clause
complex to clause, and/or from clause to nominal group. A generalized form of such a
shift would be:
(clause complex) a happens, so x happens
(clause)

happening a causes happening x

(nominal group)

the cause of happening x by happening a

The second point is that, if we plot the shift in this direction (i.e. 'downwards* on the
rank scale), there is typically some loss of information. We can attest this if we reword
in the reverse order, we are often uncertain how to construct the higher rank variant. For
example: for the truest confirmation of the accuracy of our knowledge is the effectiveness

6.1 Congruent and metaphorical variants

231

of our actions, should we lewad as if we act effectively this confirms that we know
things accurately or as how effectively we act confirms how accurately we know thingsl
Other examples:
Increased responsiveness may
be reflected in feeding
behaviour.

The way [the creature] behaves


when it feeds shows that it
has become more responsive.
-or:

Because [the creature] has


become more responsive it
behaves in a different way
when it feeds.
Higher productivity means
more supporting services.

If more goods are produced,


more supporting services will
be provided. or
More goods cannot be produced
unless more supporting
services are provided.

Here the ambiguity is mainly in the verb. When a logical relationship of this kind is
realized as a verb, then one or both of two ambiguities may arise: (i) the relationship
may be one of cause or proof, either 'x is the result of a* or 'x is the outward sign of a1;
(ii) the relationship may go in either direction, either 'a causes x' or 'a is caused by x'.
But while this is one major source of indeterminacy it is by no means the only one. An
example such as the next shows the extent to which information may be lost when a
clause complex is replaced by a clause, or a clause by a nominal group:

In other words, a nominal group (such as lung cancer death rates) is less explicit than
the corresponding clause (how fast people die when they have lung cancer, how many
people die of lung cancer, &c.), so one such group will correspond to many clauses; aid
a clause (such as higher productivity means more supporting services) is less explicit
than the corresponding clause complex (in order to produce more, you need... or if you
produce more, you will get ...&c.), so one such clause will correspond to many clause
complexes. The principle would seem to be that, where the members of a pair of agnate
wordings differ in rank, the wording that is lower in rank will contain less
information.

232

6. Grammatical metaphor

This suggests that the pattern displayed in the (constructed) example at the beginning
of this section, in which the lower rank wording occurs as a repetition of the wording at a
higher rank (where information has previously been made explicit), may be a typical
context for these more condensed variants. This leads into the general theoretical issue of
what the semantic status of such agnate sets actually is.
6.2 The nature of grammatical metaphor
We refer to this phenomenon under the general rubric of grammatical

metaphor

(following Halliday, 198S). The relationship among a set of agnates such as those just
illustrated is a relationship of metaphor; but it is grammatical, not lexical as metaphor is
in its classical sense. In order to make this clear, let us reverse the direction from which
metaphor is traditionally approached.
6.2.1 Lexical and grammatical metaphor
Hie traditional approach to metaphor is to look at it 'from below' and ask what does a
certain expression mean. For example the lexical metaphor/7<wd means either, literally,
an inundation of water or, metaphorically, an intense emotion as in she fell a flood of
relief. But we could look 'from above* and ask how is intense emotion expressed. Then
we would say it is expressed either, literally, as she felt very relieved or, metaphorically,
as she felt a flood of relief.
Once we look from above in this way, we can see that the phenomenon under
discussion is the same as metaphor in its traditional sense except that what is varied is
not the lexis but the grammar. Thus:

Here in (a) the lexico-semantic domain of Volume' has been mapped onto the lxicosemantic domain of 'meteorological commotion'; while in (b) the grammatico-semantic
domain of 'figures' has been mapped onto the grammatico-semantic domain of
'participants'. The metaphoric principle is the same in both cases; they differ only in
generality.

233

6.2 The nature of grammatical metaphor

Lexical and grammatical metaphor aie not two different phenomena; they are both
aspects of the same general metaphorical strategy by which we expand our semantic
resources for construing experience. The main distinction between them is one of
delicacy. Grammatical metaphor involves the reconstnial of one domain in terms of
another domain, where both are of a very general kind; for example:
co ne ruent domain
sequences
figures
elements: participants: qualities, processes

metaphorical domain

figures
elements: participants
elements: participants: things

Lexical metaphor also involves the reconstnial of one domain in terms of another
domain; but these domains are more delicate in the overall semantic system. For
example:
congruent domain
participant: thing: abstraction ... bright + idea
participant: thing: abstraction ... fanciful,
unrealistic + idea
participant: thing: collective: human ...
congregation in the charge of religious official

metaphorical domain
participant: thing: concrete object ...
brainwave
participant: thing: abstraction ...
pipedream
participant: thing: collective: animal: ...
flock

There are two other characteristics of lexical metaphors which are also central to
metaphor in its grammatical sense. The first is syntagmatic: lexical metaphors tend to
occur in regular clusters, which we shall refer to here as "syndromes"; for example, the
metaphor congregation > flock forms a syndrome together with religious official =>
shepherd, group of believers => fold and so on. The second is paradigmatic: lexical
metaphors typically involve a shift towards the concrete, a move in the direction of
"objectifying" ('making like an object1, not 'making objective'), as the same examples
show.1 For these properties of grammatical metaphor, see Sections 6.4 and 6.5 below.
More than half a century ago, Whorf (1956. 145-6) provided a revealing account of
the metaphorical construction of the domains of duration, intensity, and tendency in
English and other standard languages of Europe:

Because of the vastness of lexis, we do not yet have a general description of lexical
metaphorical syndromes or of the location of metaphorical domains within the overall
ideation base. But it is possible to discern that a central resource for metaphor is human
bodily experience; and that the human body itself, concrete phenomena located in spacetime, and features of daily social Ufe are the most favoured metaphorical motifs. Renton
(1990: 513-514) lists 37 such motifs, which account for 87 %of the 4215 metaphorical
items in his dictionary of metaphor. The most common are human body (23%), animals
(9%), sport (4%), food & drink (4%), war A military (4%), buildings (4%), geography (4%),
clothes (3%), nautical (3%), religion & biblical (3%), transport (2%), plants (2%),
meteorology (2%), science & medicine (2%), colours (2%), commerce (2%), manufacture
(1%), and the remaining types 1% or less. The descriptive challenge is to systematize the
domains of lexical metaphor, as Lakoff & Johnson (1980) and researchers building on their
framework have started to do.

234

6. Grammatical metaphor

To fit discourse to manifold actual situations, all languages need to express durations,
intensities, and tendencies. It is characteristic of SAE and perhaps many other language
types to express them metaphorically, The metaphors are those of spatial extension,
i.e. of size, number (plurality), position, shape, and motion. We express duration by
'long, short, great, much, quick, slow,' etc.; intensity by large, great, much, heavy,
light, high, low, sharp, faint/ etc.; tendency by 'more, increase, grow, turn, get,
approach, go, come, rise, fall, stop, smooth, even, rapid, slow'; and so on through an
almost inexhaustible list of metaphors that we hardly recognize as such, since they are
virtually the only linguistic media available. The nonmetaphorical terms in this field,
like 'early, late, soon, lasting, intense, very, tending,' are a mere handful quite
inadequate to the needs.
It is quite clear how this condition "fits in". It is part of our whole scheme of
OBJECTIFYING imaginatively spatializing qualities and potentials that are quite
nonspatial (so far as any spatially perceptive senses can tell us). Noun-meaning (with
us) proceeds from physical bodies to referents of far other sorts. Since physical bodies
and their outlines in PERCEIVED SPACE are denoted by size and shape terms and reckoned
by cardinal numbers and plurals, these patterns of denotation and reckoning extend to
the symbols of nonspatial meanings, and so suggest an IMAGINARY SPACE. Physical
shapes 'move, stop, rise, sink, approach,' etc. in perceived space; why not these other
referents in their imaginary space? This has gone so far that we can hardly refer to the
simplest nonspatial situation without constant resort to physical metaphors. I "grasp"
the "thread" of another's arguments, but if its "level" is "over my head" my attention
may "wander" and "lose touch" with the "drift" of it, so that when he "comes" to his
"point" we differ "widely," our "views" being indeed so "far apart" that the "things" he
says "appear" "much" too arbitrary, or even "a lot" of nonsense!

As Whorf s examples illustrate, lexical metaphors have grammatical implications:


they occur at a lexical degree of delicacy in the overall system, but precisely because
grammar and lexis form a continuum related by delicacy, lexical domains are in fact more
delicate elaborations of grammatical ones. So for example, if understanding is construed
metaphorically as grasping, it follows that a high degree of understanding can be also
construed according to the same material model: understand very well ==> grasp firmly,
as in she grasped the principles firmly. Similarly, if intensity is construed
metaphorically in terms of location or movement in abstract space, this lexical
reconstrual also has grammatical consequences, e.g. in terms of circumstantial elements
within metaphorical figures: prices fell sharply, prices rose to a new high, costs hit the
ceiling, and so on.
Thus in many instances, not unexpectedly, lexical and grammatical metaphor go
together. What we are most likely to be told, after the performance, is that the audience
gave thunderous applause. In she felt a flood of relief there is not only the lexical
metaphor of flood but also the grammatical metaphor of a flood of relief where intensity
is represented as a Tiling and the emotion as its Qualifier, contrasting with very relieved,
with intensity brought in as Submodifier very to the Epithet relieved. Similarly, in the
example of grammatical metaphor increased responsiveness may be reflected in feeding
behaviour there was also the lexical metaphor reflected But they are not automatically
associated, and in most instances of grammatical metaphor, if we reword in a less
metaphorical direction, we can retain the same lexical items, merely changing their word
class (often with morphological variation, e.g. we act effectively / the effectiveness of
our actions).

6.2 The nature of grammatical metaphor

235

6.2.2 Semogenetic priority of the congruent mode


Usually, as we bave seen, we aie confronted not just with a pair of wordings but with a
larger set; and the first question that arises is whether there is any priority among these
different wordings: why do we regard one or the other as being non-metaphorical, or at
least less metaphorical than the rest? We would not, in considering grammatical
metaphor, maintain a simple dichotomy between literal1 and 'metaphorical1; rather, we
would propose that there is a continuum whose poles are 'least metaphorical* and 'most
metaphorical'. The immediate evidence for this is historical, in the sense of the three axes
of semohistory referred to in Chapter 1, Section 1.5.3 above.
If we view a set of metaphorically agnate wordings synoptically, any member of such
a set appears metaphorical by reference to all the others. Given the announcement [was
made] of his probable resignation and he announced that he would probably resign there
is no reason to say that either is the less metaphorical. But if we view them dynamically,
taking account of their relation in time, then in all three histories the same one precedes:
he announced thai he would probably resign comes before the announcement of his
probable resignation. It evolved earlier in the language (phylogenesis); it is learnt earlier
by children (ontogenesis); and it typically comes earlier in the text (logogenesis). We
have also seen that there is a derivational priority because of the loss of information:
given she announced that she was accepting we can derive the announcement of her
acceptance, but given the announcement of her acceptance we do not know who made the
announcement, she or someone else ('they'); whether she had accepted, was accepting or
would accept; or whether it was a case not of her accepting but of her being accepted
twelve possible rewordings in all. On all these grounds we have to acknowledge that the
metaphorical relationship is not a symmetrical one: there is a definite directionality to it.
such that one end of the continuum is metaphorical and the other is what we shall call
congruent. Thus given the pair
Engines of the 36 class only
appeared on this train ...

congruent
when the load was
reduced
or an engine
failed.

metaphorical
in times of reduced
loading, or engine
failure.

we shall locate the two with respect to each other on a metaphor scale as above. The
expression engine failure evolved after the expression the engine failed in the evolution
of industrial discourse; to explain in times of engine failure to a child you gloss it as
whenever an engine failed (as one of the authors had to do to his 7-year-old son); the text

236

6. Grammatical metaphor

would be likely to progress from loads were reduced, engines failed to reduced loading,
engine failure rather than the other way round. And when we derive one from the other,
we find ambiguity in one direction only: reduced loading might be agnate to loads were
reduced, had been reduced or were lighter than usual, engine failure might be agnate to an
engine failed, the engine failed or engines failed, and to ... failed or had failed in each
case.

6.2.3 Location of grammatical metaphor in the content system

Let us now interpret a prototypical instance of grammatical metaphor in terms of our


semantic model. Consider the example in the second sentence of the following:
The atomic nucleus absorbs and emits energy in quanta, or discrete
units. Each absorption marks its transition to a state of higher
energy and each emission marks its transition to a state of lower
energy.

We will focus simply on each absorption marks its transition (to ...). A more
congruent variant of this would be each time (whenever) it absorbs [energy], it changes/
moves to .... Semantically we have a sequence consisting of two figures with a logical
relationship between them. The first figure consists of two participants, 'the atomic
nucleus* and 'energy', related to a process, 'absorbs'. In the congruent form the pattern of
realization is as follows:
sequence
figure
process
participant
logical relation

clause complex
clause
verbal group
nominal group
conjunction

The second figure is similar to the first except that (assuming we interpret it as 'moves
to') it has a circumstance instead of a second participant; we can add one more piece to the
pattern:
circumstance

prepositional phrase

Looking at these from the standpoint of the evolution of language, when we say they
are the congruent forms we are claiming not merely that they evolved first but that this is
why they evolved. One of the contexts in which grammar came into being one of its
metafunctions was that of construing human experience; and, as we have seen, the
model that emerged was one which construed the continuum of goings-on into
taxonomies: taxonomies of parts (meronymic) and taxonomies of kinds (hyponymic).
The central construct was that of the 'figure'; figures could be further constructed into

6.2 The nature of grammatical metaphor

237

'sequences' and also deconstructed into 'elements'. How did the grammar construe this
hierarchy of phenomena? as clauses, clause complexes, and elements in the structure
of the clause:
sequence
figure
element

clause complex
clause

element of clause structure

The elements making up a figure were of three kinds: a process, participants in that
process, and circumstantial features. How did the grammar construe this classification?
as verbs, nouns, and other things:
process
participant
circumstance

verb (expanding to verbal group)


noun (expanding to nominal group)
(others)

The circumstance could be either some quality of the process or some participant that was
indirectly involved:
circumstance (quality)
circumstance (indirect
participant)

adverb (expanding to adverbial group)


prepositional phrase
(preposition (= minor verb) + nom. gp.)

This pattern of construal was described above, in Chapters 3, 4, and 5; we now draw
on it to model grammatical metaphor. What was described above was the congruent
pattern: 'congruent' in the sense that is the way language evolved. Of course, what we are
recognizing here as two distinct constructions, the semantic and the grammatical, never
had or could have had any existence the one prior to the other; they are our analytic
representation of the overall semioticizing of experience how experience is construed
into meaning. If the congruent pattern had been the only form of construal, we would
probably not have needed to think of semantics and grammar as two separate strata: they
would be merely two facets of the content plane, interpreted on the one hand as function
and on the other as form.
Why then in our present interpretation have we to recognize two parts, one a
lexicogrammar and one a semantics? Because the system continued to evolve beyond that
point, enriching itself (i.e. engendering a richer model of experience) by forcing apart the
two 'facets' of the sign so that each could take on a new partner sequences could be
realized by other things than clause complexes, processes could be realized by other
things than verbs, and so on. We described this step briefly in the introduction, depicting
it as the dissociation of the two halves of the Chinese representation of the sign (see
Chapter 1, Section 1.5.4).

238

6. Grammatical metaphor

It is this step that gives rise to grammatical metaphor. When a sequence is realized as
a clause complex, or a process as a verb, this is congruent: it is the clause complex, and
the verb, in the function in which it evolved. When a sequence is realized as something
other than a clause complex, or a process as something other than a verb, this is
metaphorical. Some other grammatical unit is supplanting them in these functions.
6.2.4 Metafunctional effect of metaphoric shift

What is the effect of this shift, in relation to the construction of experience? Suppose
that a sequence is now realized not as a clause complex but as a clause; anda process not
as a verb but as a noun as in many of the examples above, but let us construct a new
one for the sake of brevity: say his defeat led to his resignation instead of because he had
been defeated he resigned. Are these two simply synonymous, just different ways of
saying the same thing? In principle, of course, they might be; but it is well-known that
idle variants are highly unstable, at least in language and probably in any semiotic
system. Even if through some confluence of historical processes a set of 'free variants'
does emerge, as can happen (e.g. in various types of language contact situation), it will
not be long before those variants come to construe different meanings at which point
they are no longer 'free*.
It seems clear that these sets of agnate forms are not and never have been free variants;
they have always been, at the very least, to some extent context-specific. For example,
the metaphorical forms tend to occur more in writing, the congruent ones tend to occur
more in speech. Hie question is not whether they differ but how, and why, they differ.
What kind of meaning is being construed by the systemic contrast between them?
To explore this question we formulate it in terms of the metafunctions: do they differ
ideationally, or interpersonally, or textually or, of course, in any combination of
these? We have already had clear evidence that they differ in textual meaning: e.g. The
atomic nucleus absorbs energy... Each absorption marks its transition ... Another such
example would be Cracks in glass can grow at speeds of less than one-trillionth of an
inch per hour,... On an atomic scale the slow growth of cracks corresponds to ... Here
the move from absorbs to absorption, from cracks in glass can grow at speeds less than
... to the slow growth of cracks has a very clear status in the rhetorical construction of
the discourse. At the first occurrence, it is presented to us as new information; we had not
been told it before. At the second occurrence, we already know it; it is to be taken for
granted and used as the point of departure for a further item of information (marks its
transition to a state of higher energy; corresponds to the sequential rupturing of
interatomic bonds at rates as low as one bond rupture per hour). The grammatical
resources whereby this pattern is constructed are the textual systems of 'theme' and
'information', which organize the discourse in terms of Theme-Rhente and Given-New
(Halliday, 1967/8; 1985: Ch. 3 & 8; 1988; Fries, 1981, 1992, 1995, in press; Halliday

6.2 The nature of grammatical metaphor

239

& Marn, 1993; Bateman & Mattbiessen, 1993; Matthiessen, 1992, 1995b: Chapter 6).
If the Theme of the clause (realized as the element in first position) is also explicitly
located as Given, this has a very strong 'backgrounding1 effect the message is 'you
already know this; now use it as stepping-off point for a further move, to something you
don't know*. By the same token, the remainder of the clause (either the whole of it, or at
least the culminating element) is strongly 'foregrounded1. The total construction is
obviously a powerful device for reasoning and argumentation.
What is the relation of this to grammatical metaphor? The significant point is that
the Theme of an English clause has to be nominal. Not that there cannot exist other
kinds of Theme adverbials, prepositional phrases or even verbs; but these construe
Themes which are highly maiked (verbal Themes in particular), embodying features of
contrast which are not appropriate in these contexts. The only kind of grammatical entity
that construes the message in precisely the way required without special effects, is a
nominal which may be a nominal group or else a nominalized clause or clause
complex.
If therefore an experiential meaning such as 'nucleus + absorb + energy', or 'crack +
grow + slowly', is to be mapped on to the textual meaning of 'backgrounded as point of
departure', it has to be constructed nominally, in some form such as
[the fact) that the nucleus absorbs energy
the absorbing of energy by the nucleus
the nucleus' absorption of energy
In this way it comes to constitute a textual 'package', a packed and compacted quantum of
information ready to take on its role in the unfolding of the argument. Such packages are
also favoured as foregroundings: the culmination of the clause, as the New information
the listener/ reader is explicitly invited to attend to, often also takes the form of a
metaphohc nominal, e.g. its transition to a state of higher energy; the sequential
rupturing of interatomic bonds in the examples above. We have discussed the thematic
nominal first because that is the one that is easiest to explain in its own terms; but it is
likely that the whole syndrome of features we have been illustrating evolved
together. This syndrome is made up of:
semantic unit
sequence
figure
logical relation

congruently
clause complex
clause
conjunction (relating clauses in
complex)

metaphorically
clause
nominal group
verbal group

If we show this in terms of the table we presented at the beginning of Section 6.1, we
see how the metaphorical expansion of the system fills out previously empty cells:

240

6. Grammatical metaphor

sequence
figure
element

clause complex
con^r.:

clause
metaph:
con^r.:

group
metaph:
metaph:
congr.:
metaph:

Thus each absorption marks its transition ... displays four metaphorical properties:
(1) sequence as clause (the whole); (2), (3) figures as nominal groups: each absorption
and its transition to ... ; (4) logical relation as verb: marks. (Contrast the congruent
representation given earlier.) One of the earliest writers to use this type of metaphor in
English was Isaac Newton, and already in his writing the various features co-occur in this
same syndrome, e.g.
The explosion of gunpowder arises therefore from the violent
action whereby all the Mixture ... is converted into Fume and
Vapour.

Here too we have (1) a sequence realized as a clause; (2), (3) figures as nominal groups:
the explosion of gunpowder, the violent action whereby ...; (4) logical relation as% verb:
arises. The same syndrome is found in the Italian of Galileo, written some decades earlier
(cf. Biagi, 1995); it may well be that English was directly influenced by Italian, since
many 17th century English scientists studied and worked in Italy.
We could take the view, then, at least on the evidence presented up to this point (we
shall consider a greater range of types of grammatical metaphor below), that a pair of
agnate wordings related to each other as metaphorical to congruent differ in their textual
meaning but are identical in their ideational meaning. This would be a plausible
hypothesis, and one that might be adopted in text generation. But there are reasons for
thinking that it is not, in fact, the whole story.
Let us enumerate some of these reasons here, (i) We may wonder why the
backgrounding-foregrounding pattern, which itself is clearly motivated, should then
depend on this tactic of nominal "packaging". Why would there not be other, equally
effective ways of codifying this rhetorical structure? (ii) Secondly, we might note that the
use of this kind of packaging, while it seems to have originated in the context described
for it above, now extends far beyond these requirements. Most scientific writing, most
bureaucratic and technocratic writing, and many other kinds of writing as well (not to
mention many instances of spoken discourse), use metaphorical representations as the
norm, whether or not they are demanded by the rhetorical structure of the discourse, (in)
Thirdly, we have seen that ideational meanings stand in a natural relationship to
ideational wordings. It seems unlikely that this powerful principle relating the two strata
of the content plane would be destroyed by the emergence of grammatical metaphor. On

6.2 The nature of grammatical metaphor

241

the contrary, it would seem that metaphor is possible precisely because the principle of
conguence is still operative.
To explain these observations, we can consider the other two metafunctions the
ideational and interpersonal metafunctions. The metaphors we are considering here are in
fact all shifts within the ideational realm from sequence to figure, from figure to
participant, and so on and their primary effect is ideational. They constitute a resource
for reconsiruing experience along certain lines, redeploying the same categories that
have evolved in the congruent mode of construing experience. Thus when experience of a
quantum of change has been construed as a figure consisting of 'atomic nucleus 4- absorb
+ energy1, it can be reconstrued as if it was a participant* 'absorption (+ of energy) (+ by
atomic nucleus)'. Here the process element of the figure is reconstrued as a thing; and the
participants involved in that process are reconstrued as qualities of that thing. Since they
are qualities, they are no longer "obligatory"; like any other thing, 'absorption' need not
be further specified by reference to qualities. The metaphoric shift does not mean that the
natural relationship between meaning and wording is destroyed; rather, this relationship is
extended further when new domains of realization are opened up to semantic categories
through metaphor. The shift does however create a greater distance from the everyday
experience; the metaphorical mode of construal makes it possible to recast that everyday
experience, retaining only certain features from the congruent wording but adding others
that it did not include.
We shall explore the ideational effect of grammatical metaphor in considerably more
detail below (see in particular Sections 6.7 - 6.8). Here we only need to complete the
metafunctional circuit We have already noted that grammatical metaphor is textually
significant; and we can now relate this textual significance to the ideational significance
we have just mentioned. Ideationally, grammatical metaphor is a resource for
reconstruing experience so that, alongside congruent configurations, we also have
alternative metaphorical ones. At the same time, these different configurations map onto
different textual patterns. For example, a figure maps onto a message; but a participant
maps onto part of a message, so that a figure construed as if it was a participant can be
given a textual status within that message.
In a similar way, phenomena in the ideation base also map onto constructs of
interpersonal meaning. For example, a congruent figure maps onto a move in a
dialogic exchange; it is enacted interpersonally as a proposition or a proposal. It follows
that when phenomena are reconstrued metaphorically within the ideation base, there ae
also interpersonal consequences. For instance, the figure 'atomic nucleus + absorb +
energy' can be enacted interpersonally as a proposition that is open to negotiation: The
atomic nucleus absorbs energy Does it? Yes, it does No, it can't. However,
when this figure is reconstrued as the participant 'absorption (+ of energy) (+ by atomic
nucleus)', it no longer has the potential for being enacted interpersonally as proposition;

242

6. Grammatical metaphor

rather, it would be taken for granted in discourse. You can't argue with the absorption of
energy by the nucleus since it is not enacted as an arguable proposition. Such
interpersonal differences can have a powerful rhetorical effect in persuasive discourse.
(There is an analogous effect with respect to proposals in regulatory discourse.)
But the interpersonal significance of grammatical metaphor is likely to be felt most
clearly at the macroscopic level in the overall pattern of interpersonal relationships, and
the ideological orientation, that emerge over the course of the text. We shall have more
to say about this in a later section (Section 6.7.3).
6.3 How grammatical metaphor evolves: transcategorization

Some form of grammatical metaphor is found in all languages and in all uses of
language. Like lexical metaphor, it is not something odd or exceptional; it is pan of the
inherent nature of language as a social-semiouc system, a natural process by which the
meaning potential is expanded and enriched. Even in the language of small children there
is some grammatical metaphor present almost from the start.
The phenomenon of transcategorizing elements would seem to be a feature of the
grammar of every language. This implies two things: (i) that each etymon belongs
inherently to a major class; (ii) that at least some etymons can be transferred to another
class by some grammatical means, syntactic and/or morphological. Thus in IndoEuropean languages there is typically a battery of derivational morphemes whereby a root
can be transcategorized; for example in English,
flake noun :
shake verb :
awake adj. :

(flaky adj. \flakiness noun) I flake verb


(shaky adj. : shakiness noun / shakily adv.) / shake noun
awaken verb : awakening noun

These Anglo-Saxon resources were reinforced and largely overtaken as productive devices
by those borrowed from Greek and Latin, e.g.
analyse verb:

(analyst n. / analysis n. / analytic adj. : analytically adv.)

nation noun:

(national adj. : nationally adv. / nationalism n. : nationalist


n. : nationalistic adj. / nationalize verb)

behave verb:

behaviour n. : (behavioural adj.: behaviourally adv. /

develop verb:

behaviourist n.)
development noun ... &c.

All these are means of shifting a lexeme from one class to another. If we now relate
them to the types of element, we find that in some instances the semantic nature of the
transcategorization is clear. For example, flake - thing: 'turn into flakes' - process; shake
- process: shaky 'tending to shake1 quality, shaker 'that which shakes (= vessel in which

6.3 How grammatical metaphor evolves: transcategorization

243

dice is shaken)' - thing; awake - quality, awaken 'cause to become awake' - process;
analyse - process, analysi 'one who analyses' - thing. We can gloss these in everyday
terms, without recourse to technicality. In other instances, however, the nature of the
change is less clear. What for example would be the semantic interpretation of shakiness,
awakening, analysis, development? Here we find ourselves using precisely the terms of
our own metalanguage in the definition: 'quality of being shaky1, 'process of being
awake, or causing to become awake', 'process of analysing, developing'.
When this happens, it is a signal that a phenomenon of this other kind quality, or
process is being treated as if it was a thing. The grammar has constructed an
imaginary or fictitious object, called shakiness, by transcategorizing the quality shaky;
similarly by transcategorizing the process develop it has created a pseudo-thing called
development. What is the status of such fictitious objects or pseudo-things? Unlike the
other elements, which lose their original status

in being transcategorized (for

example, shaker is no longer a process, even though it derives from shake), these
elements do not shakiness is still a quality, development is still a process only they
have been construed into things. They are thus a fusion, or 'junction*, of two
semantic elemental categories: shakiness is a 'quality thing1, development is a 'process
thing'. All such junctional elements involve grammatical metaphor.
How do such junctional elements evolve? Probably through the extension of what are
originally transcategorizing derivations. Much of the technical terminology that
developed in ancient Greek was based on four nominalizing processes: (i) -TTI -tes,
which originally meant 'one who [4- process]1, as in KOiT|Tr|c poietes 'one who makes',
and then evolved further as the nominalization of a quality (e.g. navotnc monotes
'porousness'); (ii) -ta -ia, originally 'that which is [+ quality]1, e.g. e\)9eia euthia 'that
which is straight - a line', likewise becoming a nominalized quality, e.g. <(KXVtocata
phantasia 'imagination'; (iii) -^a -ma, originally the product or goal of a process, e.g.
KOtrina poiema 'that which is made1, via an abstract product e.g. Kpayjiot pragma
'that which is done, deed' to a nominalized process e.g. TcXrp^iaplroma 'that which is
made by filling, the sum', Ktvi^a kinema 'movement (a being moved)'; (iv) -ai -sis
'an act of ... [+ process]', e.g. icpai praxis 'doing', rcouimc poiesis

'making;

creation', becoming generalized to other processes e.g. jn^on mimesis 'copying,


imitation', KtVTjaic kinesis 'moving, movement, motion', TcapocAAai pardllaxis
'alternating motion, alternation'. In all these instances it seems that meaning-making by
transcategorizing has evolved, through intermediate stages, into meaning-making by
metaphorizing; there is no sharp une between deriving a thing from a process, as 'one
who makes', 'that which is made', and construing a process as a thing 'making, creation';
with 'action of making, act of making' somewhere in between.
However these 'things' evolved, it is virtually certain that no language as we know it
today is without some metaphoric shifting of this kind. But it is only under certain

244

6. Grammatical metaphor

historical conditions that it comes to be a dominant feature of the semantic system, as it


is in English today. Here is a sentence from a local newspaper sports page:
Dorrigo cruised to an effortless win by virtue of a strong batting
display which saw them conpile 4/194 from 38 overs. (Bellingen
Courier-Sun, 14/ii 90)
There is no obvious motivation for such a high degree of grammatical metaphor; yet it
does not strike us as in any way unusual.
6.4 Types of grammatical metaphor: elemental
It seems to us necessary to identify the types of grammatical metaphor and characterize
them explicitly in relation to the semantics as a whole. We therefore introduce a general
distinction between metaphoric (elements or features) and others. Metaphoric elements, as
we said above, are junctional in that they embody a junction of two semantic categories.
In the previous chapters, 3-5, we dealt just with elements that could be assigned to a
single category: process, thing, quality &c. We shall refer to these as 'ordinary' elements,
and contrast them with 'junctionaT elements which are those that embody grammatical
metaphor. Junctional elements will always have two categories in their description, e.g.
'process thing', 'circumstantial quality', 'relator process'.
Let us now enumerate and classify the principal types of grammatical metaphor in
English. Drawing on our discussion of transcategorization, we shall base our
classification on metaphoric shifts from one elemental class to another. In doing so we
shall treat each type as a phenomenon on its own; we should therefore give a reminder
here that instances of grammatical metaphor do not typically occur in isolation. When we
find grammatical metaphors in discourse, they nearly always cluster into what we are
calling syndromes (see further Section 6.5 below); these are typical clusterings of
metaphorical effects among which there is some kind of interdependence (for example, in
the government decided => the government's decision, there is an obvious relationship
between decide => decision, process as thing, and the government => the government's,
participant as prossessor of the thing). Nevertheless there are two metaphorical effects
here, not just one; and they have been treated separately in the taxonomy that follows.
The elemental metaphors are mappings from a congruent categorial domain to a
metaphorical one. The primary types are set out and exemplified in Table 6(2) below. For
example, the categorial domain of 'process' can be reconstrued metaphorically in terms of
the domains of (i) thing and (ii) quality. (We have added "0" under "congruent domains":
this signals that a metaphorical process may be added, to which there is no corresponding
congruent form, as part of a syndrome in which the original congruent process has been
metaphorized as a thing.)

245

6.4 Types of grammatical metaphor: elemental

Table 6(2): Domains of elemental metaphors


congruent:

metaphorical:

thing,
circumstance
driver [be safe]
decided [today]

quality

process

circumstance
quality
unstable
process
absorb
circumstance
instead of;
on the surface
relator
for/ because [b,
for/ because a]
so [a, so b]

thing

1
3
6
10
because of;
as a result

replaces

causes,
proves;
ensues,
follows from
12
occurs;
imposes;
does, has

5
8

absorptive

instability
absorption

alternative;
superficial

replacement;
surface
1
cause,
proof;
result

causal;
consequent

11
phenomenon,
fact

13 expansion of thing <in environment of 1 or 2>


driver safety], driver's [safety], [safety] of the driver
today's [decision], [decision] of today

A sequence is construed congniently by the grammar as a clause nexus joined by a


conjunction (cf. Chapter 3 above). A nexus may be either paratactic or hypotactic. Where
the sequence is construed paratactically, the preferred order is the iconic one; thus, in the
case of'time' and 'cause', "precedent, then subsequent", "cause, 50 effect". The alternative
causal sequence "effect, for cause" (as in / strove with none, for none was worth my strife)
is rather infrequent; while the alternative temporal sequence apparently does not occur.
(These relationships can of course be expressed cohesively that is, without being
construed as grammatical structures at all: I strove with none. The reason was that ... .)
Where the sequence is construed hypotacticaiiy, either order is possible: "after precedent,
subsequent"/ "subsequent, after precedent"; "because cause, effect"/ "effect, because cause".
Note also "precedent, before subsequent"/ "before subsequent, precedent"; "cause, so that
effect"/ "so that effect, cause" where tying the relator to the 'effect typically implies
intentionality. For example:
[P A a: 'purpose'] So that they could get there in time, they broke the
door down.

[aA P:'purpose']They broke the door down so that ('in order that') they

could ('would be able to1) get there in time.

[a A P: 'result'] They broke door down, so that ('with the result that1)
they could ('were able to') get there in time.
The reason for the disparity is that hypotaxis construes an order of its own ordering in
dependence; whereas in parataxis the only ordering is that being imposed by the grammar
on the experiential phenomena themselves.

Table 6(2) shows that there are clear patterns in the metaphoric shift. For example,
the 'relator1 can be leconstnied metaphorically in terms of any of the other types of
element; but it cannot itself be a target domain in metaphors. Such particular patterns ate
part of more general metaphoric motifs. We shall identify and interpret these patterns and

6. Grammatical metaphor

246

motifs in Section 6.7 below. But first we need to consider the types of metaphoric shift
in a little greated detail.
Table 6(3) gives a more detailed description of the types shown in Table 6(2) above,
together with an example of each. Several of the numbered types identified in the
previous table have been differentiated further into subtypes, represented by Roman
numerals. The first two columns present the metaphoric shift as a grammatical
phenomenon: (1) as a shift of (word) class and (2) as a shift of function (in clause, phrase
or group, as appropriate). The third column gives examples of each type. The last two
columns show the metaphor as a semantic relationship between types of element: (4) the
domain of the congruent variant, then finally (5) that of the metaphorical variant. It
should be remembered that almost every one of the metaphoric categories is immensely
variable. Wherever possible, examples have been drawn from texts cited in the discussion
in the present chapter; but they are just examples, and should not be read as glosses
describing the category as a whole.
Table 6(3): Types of grammatical metaphor
TYPE:

Grammatical shift
(1) grammatical
class

(2) grammatical
functions

adjective => noun

Epithet/ Attribute
Thing

unstable

Event

ii

Auxiliary
Thing

Thing

metaphorical

quality

thing

speed

noun:

congruent

instability;
quickfty)

verb

Semantic element

Example

transform

process:
event

transformation
mil/ going to >

tense; modality

prospect; can/
could =>
possibility,
potential

Catenative
Thing

iii
3
i

preposition(al
phrase) => noun
preposition

Minor Process =>


Thing

try to > attempt;


want to => desire

with

phase;
contingency
circumstance:
minor process

accompaniment; to
destination

ii

prepositional
phrase

Location, Extent
&c => Classifier

conjunction
noun
verb
adjective

Conjunctive
Thing

(dust is] on the


surface s> surface

minor process +
thing

dust

4
5

so *> cause, proof;

relator

tfsf> condition

process:

quality

247

6.4 Types of grammatical metaphor: elemental

Event => Epithet/


Classifier

[poverty J increases

event

s> increasing
[poverty]

Auxiliary
Epithet/ Classifier

ii

was/ used to

tense; modality

previous; must/ will


*> conitant

Catenative
Epither/ Classifier

iii
6

i
ii

prepositional
phrase)
adjective
preposition
prepositional
phrase
conjunction
adjective

Minor Process
Epithet/ Classifier
Location, Extent
&c => Epithet/
Classifier
Conjunctive
Epithet/ Classifier

phase;
contingency
circumstance:

begin (to)
initial

minor process

with
accompanying
[marks are] on the
surface
superficial

minor process +
thing

[marks]

relator

before
previous; so
resultant

8
i

preposition(al
preposition(al
phrase)
preposition

circumstance:
Minor Process
Process

process

minor process

(be) about
concern; (be)
instead of
replace; (go)

traverse

across

ii

prepositional
phrase

Location, Extent
&c => Process

(put) in a box/ in

house %> ox/

minor process +
thing

house

conjunction
verb

Conjunctive
Process

then

follow; so

relator

cause; and
complement

10

conjunction
preposition(al
phrase)
=> preposition

relator

Conjunctive
Minor Process

when -> in times

Conjunctive ->
Location, Extent
&c

so 3*> as a result,

circumstance:

minor process

of; because =>


because of

ii

=> prepositional
phrase

minor process
+ thing

in consequence; if
[it snows]
under/ in [snow(y)J
conditions

11

+ noun

+ Thing

[x]

the fact/

(none)

thing

phenomenon of [x]

12
i

+ verb
+ verb

+ Process

process
[x]
(x] as> [x]
[xj occurs/
occurs/
exists; [x] SB> have,
do M (eg- impact
have an impact)

(none)

6. Grammatical metaphor

248

ii

4- (causative &c)
verb

make [x : y)

(agency &c)

impose [y on x];
think x y]
credit [x with y]

iii

+ (phasal &c)
verb2

started/ wanted [to

(phase &c)

survey] > started/


wanted [a survey]

13

noun => (various)

Thing

the government

thing

[decided]

(a) Qualifier

[decision] of/ by

expansion of
thing:
(qualifying)

the government

(b) Possessive
Deictic
(c) Classifier

(possessive)

the government's
[decision]

(classifying)

governmental )
[decision]

ii

adverb =>
adjective
prepositional
phrase =>
adjective
adverb =>
(various)

Manner

Epithet

[decided] hasttty

circumstance

hasty [decision]

Location, Extent
&c => Epithet

[argued] for a

expansion of
thing:
(descriptive)

long time
lengthy [argument]

Location, Extent
&c => Possessive
Deictic

[announced]
yesterday
yesterday's
[announcement]

prepositional
phrase
(various)

Location, Extent
&c Qualifier

[departed] for the


airport
[departure] for the
airport

Here is a typical passage of written English analysed for the types of grammatical
metaphor listed in the table above (from an editorial in the Sydney Morning Herald ):
The Federal Government's decision to ask the Arbitration
13.i(b)
2.i
Commission to determine whether the BLF has engaged in

12.i

serious industrial misconduct, as part of its


move to
6.
2.i
13.i(b) 2.iii
deregister the BLF in certain states, is one of the weakest
actions ever taken by a government in the face of industrial
2.
12.i
10.
6.ii
thuggery.
1

Note that where the grammar has a simple verbal group (types 12i and 12ii), this construes a
single 'process' element in the figure. Where the clause contains a verbal group complex
(type 12iii), while this (congruently) still construes a single 'process* it is now somewhere
along the cline towards a sequence of two figures. Thus: the government deregistered the
union > the government moved to deregister the union > the government moved/ acted
(in order) to deregister the union. Cf. Chapter 3, Section 3.4.

6.4 Types of grammatical metaphor: elemental

"Their
13.i(b)

249

guerilla tactics and use of thuggery, violence and


3.ii
2.i 13.i(a) 1
1

intimidation have had a disastrous inpact not only on


2.i
12.i
13.
2.i
building employers but also on fellow workers in the industry. "
3.ii
6.ii
Obviously the Government is frightened of union
13.i(c)

reaction to
2.i

its
move to inpose proper behaviour on unions.
13.i(b) 2.ii
12.ii
13.ii
2.i
13.i(a)
[The text "unpacked" according to this interpretation:]
Ihe Federal

Government

have decided

to ask the Arbitration

Commission to determine whether the BLF have seriously misbehaved


in the industry, as they are intending to deregister the BLF in
certain states; no government has ever acted more weakly when
people have behaved like thugs .
"Because they have used tactics like guerillas, and behaved like
thugs , been violent and intimidated people, this has disastrously
affected not only employers in building but also those who work
with them. "
Obviously the Government is frightened how the unions will react
when they begin to make them behave properly.
6.5 Syntagmatic and paradigmatic dimensions of grammatical metaphor
In Section 6.4 we identified individual types of grammatical metaphor, such as "process
=> thing". For analytic purposes we treated these as isolates having just the two values
"congruent/ metaphorical". We now need to consider two respects in which this is an
idealized, oversimplified account of what actually happens. We need to add a dimension of
complexity on both syntagmatic and paradigmatic axes.
Syntagmatically, instances of grammatical metaphor typically occur not in isolation
but in organic clusters or "syndromes". Paradigmatically, there will typically be other
wordings intermediate between an instance of grammatical metaphor and its "most
congruent" agnate variant. These two aspects of complexity are taken up in the next two
subsections.

250

6. Grammatical metaphor

6.5.1 Syntagmatic complexity: syndromes of elemental metaphors


We noted above that the individual types of grammatical metaphor set out in Table 6(3)
tend to occur not one at a time but together in syntagmatic clusters, or "syndromes". We
can enumerate some of the common syndromes as follows,
(1)

2i/l + 13i(a)

the fracture of glass/ the


instability of diamond

(2) 13i(c) 4 2i/l

engine failure/ union intransigence

(3)

the government's decision/

13i(b) + 2i/l

indecision (indecisiveness)
(4)

6ii + 2i/l

interatomic bonding/ industrial


thuggery

(5)

13i(b) + 2i + 13i(a)

his arrest by the police

(6)

13ii + 2i + 12i

rapid bonding occurs

(7)

13ii + 2i + 13i(a)

(8) 5ii + 1 + 13i(a)

yesterday's decision by the group


the apparent innocence of the

accused
(9)

1 + 13i(a) + 13i(b) + 2i the cogency of his argument

The most pervasive of the types of grammatical metaphor listed in Table 6(3) are
types 1 and 2: "quality => thing" and "process => thing". We can see here that these
typically occur as elements of syndromes, such as those just illustrated; these are
syndromes formed of metaphoric shifts from one class of element to another. Some of
these shifts can occur indpendently, while others those we have called type 13
occur only under the driving force of the shift of type 1 or type 2.
These syndromes of elemental metaphors fall into three general types, not very
sharply distinct but worth using as a conceptual framework. The distinction relates to the
rank where the metaphoric reconstnial takes place: (I) from figure to element, (H) from
sequence to figure, (III) from figure with process to figure with process as thing.
I. Figure ==> element
Here a figure is being construed metaphorically on the model of a participant:
grammatically, the figure is construed not as a clause but as a nominal group. There is a
shift in rank from figure to element, and concomitantly a shift in status among the
elements making up the construction. The way these syndromes are construed is shown
in the following displays (omitting numbers (1) to (4), since these are included in the
others).

251

6.5 Syntagmatic and paradigmatic dimensions of grammatical metaphor

(5) he was arrested by the police ==> his arrest by the police 13i(b) + 2i + 13i(a)
figure:

elements:
participant

thing
quality

participant
he

process
arrest
2i (process
thing)
arrest

13i(b) (thing
possessive
expansion)
his

participant
the police

13i(a) (thing
qualifying
expansion)
by the police

process
circumstance

Similarly (from above): the union's use of thuggery


(6) (...) bond rapidly > rapid bonding occurs 13ii + 2i + 12i
elements:
participant

thing
quality

process

figure:
process
bond
2i (process
thing )0/idiR

participants

circumstances
rapidly
13ii (circumstance
expansion of
thing: descriptive)
rapid

12i (+ process)
occur

circumstance

(7) i/re group decided yesterday

yesterday's decision by the group 13ii + 2i +

13i(a)

elements:
participant

thing
quality

process
circumstance

figure:
process
decide
2i (process
thing)
decision

participants
the group

circumstances
yesterday

13i(a) (thing
expansion of
thing: qualifying)
by the group

13ii (circumstance
expansion of
thing: possessive)
yesterday's

6. Grammatical metaphor

252

(8) the accused appeared to be innocent


+ 1 + 13i(a)
figure (being):
process
appear

elements:
participant

the apparent innocence of the accused 5iii

participant
the accused

thing
quality

5iii (phase process


quality)
apparent

participant
(Attribute)
innocent
1 (quality
thing)
innocence

13i(a) (thing
expansion of
thing: qualifying)
of the accused

process
circumstance

(9) he argues cogently ==> the cogency of his argument 1 + 13i(a) + 13i(c) + 2i
figure:
process
argue

elements:
participant

participants
he

thing

quality

process
circumstance

II. Sequence

2i (process
thing) argument
then:
13i(a) (thing
expansion of
thing: qualifying)
of argument

13i(b) (thing
expansion of
thing: possessive)
his

circumstances
cogently
13ii (circumstance
expansion of
thing: descriptive),
cogent
then:
1 (quality
thing)
cogency

figure

Here the syndrome involving the shift of type 1 or 2 occurs in a more general
environment, that of construing a sequence on the model of a gure grammatically,
the sequence is construed not as a clause complex but as a clause.

253

6.5 Syntagmatic and paradigmatic dimensions of grammatical metaphor

(1) They shredded the documents before they departed for the airport => (They
shredded the documents) before their departure for the airport lOi + 13i(b) + 2i + 13ii
figures:

elements:

figure

participant

sequence
figure
(They
shredded the
documents)

relator
before

figure
they departed for
the airport

thing
quality

process
circumstance

circumstance)
(relator + figure
before their departure for the airport
(figure
lOi (relator
participant)
minor process)
their departure for
before
the airport

(2) They shredded the documents before they departed for the airport => Their
shredding of the documents preceded their departure for the airport 13i(b)+ 2i + 13i(a) +
9 + ...
figures:

elements:

figure

participant

thing

sequence
figure
They shredded the
documents
[process +
participant +
participant]
2i (process
thing) shredding

quality

13i(b) (thing
expansion of
thing: possessive)
their

quality

13i(a) (thing
expansion of
thing: qualifying)
of the documents

process

relator
before

figure
they departed for
the airport
[process +
participant +
circumstance]
2i (process
thing)
departure
13i(b) (thing
expansion
of thing:
possessive)
their
13ii
(circumstance
expansion
of thing:
qualifying)
for the airport

9 (relator
process) precede

circumstance

These examples of a 'sequence

figure' metaphor both involved a sequence of the

expansion type. However, this type of metaphoric shift also occurs with projection
sequences; for example:
The colonel declared his innocence.
where the congruent form would be a projection, either hypotactic or paratactic:

6. Grammatical metaphor

254

The colonel declared that he was innocent.


The colonel declared, "I am innocent*.
(lu) Figure with process ==> Figure with process as thing
The kinds of syndrome illustrated under (I) and (u) represent the main motifs in
ideational metaphor (to be explored further in Section 6.7 below). But there are also
syndromes of various other kinds, not clearly belonging to either of these categories.
Those illustrated here are variants of the figure => element syndrome, where only pan of
the figure is reconstmed as a participant and the syndrome often involves an elemental
metaphor of type 12i (+process).
(1) They surveyed the property => (They) did a survey of the property 12i + 2i +
13i(a)
figure
process
surveyed

elements:
participant

thing

participant
(they)

2i (process
thing)
a survey

quality

process

participant
the property

13i(a) (thing
expansion of
thing: qualifying)
of the property
12i (+ process)
did

circumstance

(2) They started to survey the property


12iii + 2i + 13i(a)
elements:
participant

thing

figure
process
start to survey
2i (process
thing)
a survey

circumstance

participant
the property

13i(a) (thing
expansion of
thing: qualifying)
of the property

quality

process

(They) started a survey of the property

12iii (phase
process)
start

participant
(they)

6.5 Syntagmatic and paradigmatic dimensions of grammatical metaphor

(3) They discussed in the early afternoona


afternoon) 13i(b) + 2i 4- 12i
figure
process
discussed

elements:
participant

thing
quality

process
circumstance
6.5.2

Paradigmatic

255

Their discussion took place (in the early

participant
they

circumstance
(in the early
afternoon)

2i (process
thing)
discussion

12i (+ process)
took place

13i(b) (thing
expansion of
thing: possessive)
their

complexity: degree of metaphoricity

and steps

in

"unpacking"

In selecting examples to illustrate typical syndromes of grammatical metaphor we chose


as far as possible instances where the metaphorical derivation, or re-wording, of each
element could be 'unpacked1 in a single step. Thus in the fracture of glass, fracture has
been shifted from process to participant (thing) and glass from participant (thing) to
qualifying expansion of thing. When we reword as a clause glass fractures we have
reached a congruent form.
Very often, however, in deriving any one element we should have to take two or more
steps in the course of unpacking. For example, in the development of our understanding,
which has the same immediate history, we can similarly unpack to our understanding
develops. However, our understanding is itself a metaphorical entity consisting of a
'thing' derived from a process; and a 'possessor' derived from a participant; so it can be
unpacked further to we understand, giving something like [the way] we understand
develops. We then might want to consider develops as (be/ become) quality => process,
so [the extent to which] we understand becomes greater, i.e. we understand more and
more. One or two of the examples treated above involved more than one step in this
way.
It would take many pages of discussion to follow through such examples discursively
in any detail. Instead we will present one brief example with step-by-step unpacking of
all the grammatical metaphors simultaneously, using a form of graphic display (with
labelling of grammatical functions) which will give a sense of the overall metaphorical
space that may be covered in a single instance of this kind: see Figure 6-2.

256

6. Grammatical metaphor

Fig. 6-2: Step by step "unpacking" of metaphorical wording

It will also (see the discussion below) bring out another equally important issue,
which takes us bade to the point made in Section 6.1 above: that metaphoricity is a
relative matter. We can take a pair of agnate expressions and order them with respect to
each other, showing that one of the two is more congruent But if we are rewording in
either direction 'packing1 or 'unpacking', to use the (lexical) metaphor that was adopted
for this by the students to whom it was originally presented , there is no clearly
definable point where we say 'now we have reached the end*. Obviously we cannot go on
for very long; the actual number of steps taken will in fact be extremely limited. But any

6.5 Syntagmatic and paradigmatic dimensions of grammatical metaphor

257

sequence that is reasonably complex in its semantic patterning will be likely to show
considerable elasticity at both ends of the continuum.
One interesting strategy that can be used for 'unpacking' is the following. Take a
sentence that is highly metaphorical but reasonably non-technical, and reword it so that
you think it would become intelligible to a 15-year-old. Then re-word it again to make it
accessible to a 12-year-old; then to a 9-year-old, and then to a child of 6. (It is probably
not worth trying to reduce the intervals to less than three years or the age to less than
six!) Ideally this should be done in dialogue with some member, or members, of the age
group in question; but if they are not available you can construct them in your mind's
eye and attempt it monologically. You may not succeed in enlightening any small
children in this way; but you will gain a remarkable insight into the workings of
grammatical metaphor. Here is an example from a philosophical text, unpacked step by
step in this way.
"The truest confirmation of the accuracy of our knowledge is the
effectiveness of our actions."3
15 The fact that our knowledge is accurate is best confirmed by
the fact that our actions are effective.
12 What proves that we know things accurately is the fact that we
can act effectively.
9 Ttie best way of telling that we know what ' s happening is to see
that what we do is working.
6 You know you've got the right idea because you can do something
and it works. Like watering plants; you water them, and they grow.
3 Look - wasn't it good that we watered that philodendron? See how
well it's growing!

Let us know construct a piece of conversation that might be addressed to a three-year-old,


and repack it along similar lines:
3 Look - it must be raining! People have got their umbrellas open.

From Leszek Koiakowski, Main Currents of Marxism, trans. P.S. Falla (Oxford University
Press, 1981). There would of course be very many ways of unpacking this particular
syndrome of metaphors. We have retained the thematic ordering ('know + accurately' before
'act -i- effectively'), and the internal form of the logical-semantic relation ('actions prove
knowledge1, not 'knowledge leads to actions'). We assume that the younger versions would
be part of an ongoing dialogue.

6. Grammatical metaphor

258

6 You can tell it's raining because people have got their
umbrellas open.
9 Hie best way of telling that it's raining is that people's
umbrellas are open.
12 What proves that it's rainy weather is the fact
umbrellas have been extended.

that the

15 The fact that the weather is pluvious is best confirmed by the


fact that the umbrellas are extended.
"The truest confirmation of the pluviosity of the weather is the
extendedness of the umbrellas.*

Like lexical metaphors, so also grammatical metaphors may become codified in the
language and 'take over' as the normal mode of expression. We shall refer below (Section
6.7.1) to two instances of this type in English: those like have a look, take a step aid
those like span, cover, accompany. There aie other features of the grammar that are in
origin metaphorical (many of the uses of the possessive form, such as [he didn't approve
of] my leaving home - cf. the syndromes listed above), but whose origin is so obscured
by the natural evolution of the language that all sense of their metaphoric nature has long
since been lost.
The overall effect of the grammatical metaphor is that semantic relations between one
element and another, and between one figure and another, become progressively less
explicitas the degree of metaphoricity increases. We can illustrate this by taking a text
example and relating it to more congruent and more metaphorical variants:
[i] (most congruent)
glass
participant a

cracks
does

process b

ace. to

the harder
you
I ace, to. I circumstance I participant w | doe

more quickly
circumstance c

press
I process y

(lineally
determined by
how far)
on it
I to participant I

[U]
cracks in glass
participant b in a does

grow
process d I ace. to

faster
circumstance c I (lineally determined by
1 how far)

6.5 Syntagmatic and paradigmatic dimensions of grammatical metaphor

the more pressure


I participant y
| has

is put on
I process z

259

I done to

[iii]
glass crack growth
participant abd

is
| has

if
I (under condition that)

faster
[attribute c
greater stress
| participant ly

is applied
| process z

| has

| donetoi

[iv]
the rate of glass crack growth
participant c of abd

depends on
is caused by

magnitude of the applied stress*


participant x of zy

(he

[v] (most metaphorical)


glass crack growth rate
| participant abdc

is associated with
causes/ is caused by

applied stress magnitude


I participant zyx

Notice how the semantic information construed by the grammar in the most congruent
version is gradually lost at each step in the course of metaphoric rewording.
Original version the rate of crack growth depends on the magnitude of the applied stress
(from Michalske & Bunker, The fracturing of glass', in Scientific American, December
1987).

6.6 Metaphor, transcategorization

& rankshift: semogenic resources

We noted in our discussion of the grammar (Chapter 1, Section 1.4 above) that the
grammars of natural languages include among their resources the potential for rank
shift, whereby one grammatical unit functions in the place of another: typically, a
clause functions in an environment whose original defining occupant is a nominal group.
This is one way in which the resources of grammar have expanded as language evolved.
Rank shift is not inherently metaphorical. There is a parallel here between rank shift
and class shift. In origin, both these could be described as metaphorical semogenic
processes: a verb or adjective is meiamoiphosed into a noun (a shift of class, e.g.
strong : strength, lose : loss ), a clause is metamorphosed into a group (a shift of rank,
e.g. they went bankrupt : their bankruptcy). But as a synchronie relation neither of
these necessarily involves metaphor; there may be no systematic alternation such as there
is between a metaphoric and a congruent form. We have already discussed nonmetaphorical forms of class shift, under the heading of transcategorization.
Similarly, in the following examples of rank shift, where a clause is rankshifted to
function either as Head (1,2) or as Qualifier (3, 4) of a nominal group, no grammatical
metaphor is involved:

260

6. Grammatical metaphor

1. Not having a proper job] made my life unbearable


(semantically) non-projected figure as participant
(grammatically) clause (as Head of nominal group) functioning (as Agent) in clause
structure: "act" type
2. IHOW they escaped] was a mystery
(semantically) projected figure as participant
(grammatically) clause (as Head of nominal group) functioning (as Carrier) in clause
structure: "fact" type
3. That woman U(who was) sitting behind the desk] reminded me of
Tracy
(semantically) non-projected figure as quality
(grammatically) clause functioning as Qualifier of nominal group ("defining relative"
clause)
4. The idea ([that anyone would visit/ of anyone visiting] seems
incredible
(semantically) projected figure as quality
(grammatically) clause functioning as Qualifier of nominal group

Class shift becomes metaphorical when the "shifted" term creates a semantic
junction with the original. A good way of illustrating this is to bring together two
instances of the same lexical item, used once as (non-metaphorical) transcaiegonzaion
and once as grammatical metaphor. Let us return to a previous example (cf. Section 6.1
above):
many failures are preceded by the slow extension of existing cracks
thing relator => process
thing
process
process

More congruently, this would be many (pieces of glass) fail after the cracks have slowly
extended, or often the cracks slowly extend and then the glass fails. The congruent form
is a sequence of two figures linked by a relator; in the metaphoric form, each figure
becomes a participant and the relator becomes a (relational) process to which two
participants subscnbe a syndrome of 2i + 9 + 2i. Here, then, failure is an instance of
metaphorical class shift; there is a semantic junction between two features:

6.6 Metaphor, transcategorization & rankshift: semogenic resources

261

(1) 'process' (class meaning of verb fail),


(2) 'thing/ participant' (class meaning of noun failure).
Note that there is an asymmetry between the two a time line, such that the feature
'thing' is as it were a reconstrual of the original feature 'process1: we could gloss it as "a
process reconstrued as a participant". Contrast this now with failure in a technical
expression such as heart failure; in origin this was no doubt a grammatical metaphor for
the heart fails, but the metaphorical quality has since been lost, or at least significantly
weakened (the metaphor is "dead"), and heart failure is now the only congruent form.
Likewise contrast he regretted his failure to act, agnate to that he had failed to act (or that
he had not acted ), where failure is a grammatical metaphor, with he always felt that he
was a failure, where failure is now the congruent fonn and this is not a metaphorical
agnate of he always felt that he had failed
When we reach this point, we find that rank shift sometimes operates as an alternative
to class shift of this metaphorical kind. In examples A-l and A-2, there is metaphorical
class shift:
A-l a cow is a ruad. nm. ting quadruped [verb: adjective; process
A-2 your cap* was a miracle [verb: noun; process

quality

thing]

In B-l and B-2, on the other hand, there is rank shift but no class shift; these are the
agnate congruent forms:
B-l a cow is a quadruped that ch*w thm cud/ that rumint**
B-2 it was a miracle how/ tliat you ffot tray/ you mcmpmd
Here 'ruminate' and 'escape' remain as process, without shifting in class. This now helps
to explain the meaning of forms such as C-l and C-2:
C-l a cow is a cud-chewing

quadruped

C-2 it was a miracle you/ your getting away


These represent a kind of semantic compromise, a means of having it both ways. The
junction is now as it were symmetrical, with the time une removed, so that 'chew + cud',
'you + escape1 are simultaneously both figures and (parts of) elements. What the
grammar is construing here is an exchange of functions without shifting class; the figure
takes on a special kind of clausal structure which conserves the transitivity relations.

262

6. Grammatical metaphor

We have shown that the typical manifestation of grammatical metaphor in discourse


is as a 'syndrome' of features including both class shift and rank shift. Let us now recall
two of the frequently occurring syndromes described earlier, noting where these features
occur (we ignore here shifts of type 13, since these follow automatically from types 1

and 2):
(1)Their frequent dismissal of personnel does not inspire people's
confidence
[class shin] 2 + 9 + 1 (process 'dismiss* as thing, relator 'cause* as process ['inspire'],
quality 'confident1 as thing)
[rankshift] sequence as figure
[congruent variant]
Because

they

frequently

dismiss

personnel,

people

are

not

confident [in them]


(2) Rapid bonding resulted
[class shift] 2 + 9 (process 'bond' as thing, relator 'cause* as process)
[rankshift] none (figure as figure)
[congruent variant]
As a result [the substances] rapidly bonded.4
Note that in (2) the other term in the relation is to be presumed from the preceding text.
These two syndromes really variants of a single syndrome define the type of
clause that might well be considered as the favourite type of modern scientific English.
The metaphorical 'things' are packaged processes or qualities that have thematic or
informational value in the text, construed as Theme in the clause or New in the
information unit (N-Rheme in the clause: see Fries, 1992). The metaphorical 'process* is
one of the typical semantic relations that link figures into a sequence: 'cause* & 'time' are
the prototypical relations in question, but in modem writing 'identity* has tended to take
over as the central type often by a further metaphorical step whereby causes becomes
is the cause of, results becomes is the result of. At the same time, a further distinction
has evolved between two kinds of logical-semantic relation: (i) a relation 'in rebus*

Note the distinction between bonding occurred, type llii, congruently [the substances]
bonded, and bonding resulted, type 9, congruently so [the substances] bonded. The
wording bonding ensued is probably also type 9, congruently then [the substances] bonded
although, as is often the case, the metaphorical version is somewhat less explicit, and
may suggest more than one congruent unpacking. There is a similar uncertainty with impose
in the text at the end of Section 6.4 above; we have interpreted impose proper behaviour on
as agnate to make ... behave properly, but it could be unpacked as two figures, act so that
... behave properly.

6.6 Metaphor, transcategorizaton & rankshift: semogenic resources

263

between two experiential events, (it) a relation 'in verbis between two stages in the
discourse; e.g.
(i) Political pressures brought about his downfall. Major changes
ensued.

(ii) This section


discussion follows.

illustrates

the

main

argument.

Further

These correspond respectively to the 'external' and 'internal' types of conjunctive relations
in cohesion (Halliday & Hasan, 1976: Ch. 5; Martin, 1992: Section 4.2.3).
6.7 Interpretation of grammatical metaphor
Metaphorical syndromes exploit very general semantic resources which have always been
present in the language, but have come to be foregrounded by the new demands made on
the language as a result of changing historical conditions during the last half millennium.
Grammatical metaphor has played a central part in the construction of new meanings
through this period in our history. Let us try to interpret the broad outlines of this
development.
In the semantic construction of experience, 'process' and 'participant' emerge as
prototypical categories; and there is a broad agreement among different languages both
about the nature of this distinction and about which particular phenomena should be
assigned to which category. But as in any semiotic endeavour there are always some
domains of uncertainty: are rain, wind, thunder processes or things? are fear, worry, regret
processes or qualities? Examples like these prevent the categories from being too reified
and rigid, and provide a kind of gateway of analogy through which a phenomenon
can drift or be propelled from one category to another. In transcategorization some other
semantic feature triggers the propulsion; e.g. dark + make/ become = darken, flake + like/
composed of = flaky. In metaphor, however, the phenomenon is reconstrued as another
category; what is being exploited is the potential that arises but only after the
categories have first been construed as distinct; not otherwise of treating every
phenomenon in more ways than one. In this process the original interpretation is not
supplanted; it is combined with the new one into a more complex whole.
6.7.1 Motifs in grammatical metaphor
It is possible to distinguish two predominant motifs in the phenomena characterized
here: one major or primary and one minor or secondary one.
(i) The primary motif is clearly the drift towards 'thing*.

264

6. Grammatical metaphor

() The secondary motif is what appears as a tendency in the opposite direction:


the move from 'thing' into what might be interpreted as a manifestation of
'quality' (qualifying, possessive or classifying expansions of the 'thing').
We will consider the explanation for this secondary motif below; meanwhile, we can
summarize the principle of metaphoric shift as in Figure 6-3 (cf. Table 6(2) above).

Fig. 6-3: Direction of metaphorization


We can see from this figure that the drift towards 'thinginess' is the culminating and
most clearly articulated form of a shin which can be characterized in more general terms
as a shift towards the experiential towards that mode of construing experience
that is most readily organized into paradigmatic sets and contrasts. Things are more easily
taxonomized than qualities, qualities than processes, and processes more easily than
circumstances or relations. Since the 'noun-ness' is being used to construe phenomena
that stan out as something else than a noun, metaphors will inevitably be abstract. If the
surgeon makes an incision, instead of cutting, the cut is being presented as a more
abstract version of the experience. This is further reinforced if, as is often the case, the
metaphorical term comes from the more highly valued lexical stockpile of Latin and
Greek roots.

6.7 Interpretation of grammatical metaphor

265

We can explore the implications of the "drift towards thinginess" by reviewing the
different types of element in terms of the potential that they embody for construing
experience (cf. Chapter 5 above).
Participants are realized by nominal

groups, which allow more or less

indefinite expansion (through the univariate structure of modification). This expansion is


the grammar's way of constructing taxonomies of things: grouping them into classes,
assigning properties to them, quantifying them and then uniquely identifying any
individual thing, or any number, set or class of things, in relation to the 'here-&-now' of
the speech event. The expansion involves open sets of things and qualities, realized by
lexical items; but it can also capture a circumstance, realized by a prepositional phrase, or
an entire figure, realized by a clause, and put it to use as a quality in describing or
identifying such a thing or set of things, e.g. this unique 20-piece handprinted china
dinner service with optional accessories never before offered for sale at such a bargain
pnce.
Qualities are attached to things, and so contribute to this overall expansion. They also
have possibilities of expansion of their own, by submodification (at least for intensity,
but sometimes along other lines as well: very long, longest; dark blue, red hot).
The grammatical potential of the nominal group for construing taxonomies by
syntagmatic expansion was described in Chapter 5, Section 5.3.2.4 above. This
grammatical potential for taxonomizing is complemented by the lexical potential of the
nominal group for construing feature networks: see Chapter 5, Section 5.3.2.5 above. So
by construing any phenomenon of experience as a thing, we give it the maximum
potential for semantic elaboration.
Processes are realized by verbal groups, where, typically, the only lexical
material is the verb itself, functioning as Event (material action or event, conscious or
verbal process, or relation). Apart from the adverbial complement of a phrasal verb,
which may serve to construct a distinct lexical item e.g. make out (I can't make out the
difference), come to (she'll come to in a minute), let on (don't let on about this), &c.,
all contrasts made by the verbal group are grammatical ones tense and other quasitemporal systems, and modality. There is no lexical expansion classifying processes into
taxonomies or assigning them sets of contrasting qualities.
Processes do resemble other processes, but they share different features with different
others; no single line-up is dominant enough to form the basis for permanent hyponymy.
For example, if we consider a small subset of the words expressing verbal processes
offer, tell, promise, threaten, recommend, warn:

266

6. Grammatical metaphor

(1) offer, promise, threaten have the feature 'offer'; tell, recommend, warn have the
feature 'command'
(2) offer, tell are neutral in orientation; promise, threaten, recommend, warn
have the feature 'oriented to addressee*
(3) within the addressee-oriented, promue, recommend have the feature 'desirable1,
threaten, warn have the feature 'undesirable'.
(4) offer, promise, recommend take direct participant ('propose to give ... to
Receiver1; 'propose that Receiver should obtain ...').
(5) tell, warn take circumstance of Matter 'about...'.
Processes thus have much less potential than participants for being characterized and
taxonomized. For example, with a process like decide we can add a circumstance to it,
saying he decided quickly or he decided on the spur of the moment; but if we want to
identify the occasion as unique we have to say this decision, the previous decision, the
only good decision he ever made. We can say his absurd decision but not he decided
absurdly at least not in the same sense, since absurdly could only characterize the
figure (how he carried out the act of deciding), not the quality of the process of deciding
as such.
The minor processes that form circumstances (realized as prepositions in English)
are even less taxonomizeable; they are intermediate between processes and relators, and
only the spatial ones (spatio-temporal) display any real paradigmatic organization (to/
from// towards/ away from; inside/ outside// into/ out of; before/ after// in front of/
behind &c.).
Relators show the least organization of any, since they are experienced only
indirectly in the form of logical relations between other configurations; they share some
of the systematic features of minor processes, but other than that they display only the
contrast between the two relative statuses they assign to these configurations, as being
equal or unequal (paratactic/ hypotactic, in the grammar) a then x I x after all b so v
/ v because b &c. Thus a relator can be metaphorically reconstrued into any other
category:

6.7 Interpretation of grammatical metaphor

relator:

(equal) so
(unequal) because

267

a happened; so x happened
* happened, because a happened

minor process: because of x happened because of a


process: cause

that a happened caused x to happen

quality: causal

happening a was (in a) causal (relation)


to happening x

thing: cause

happening a was the cause of happening x

whereas a process can be reconstruct! only as a participant (quality or thing), and a


quality only as a thing.
Thus in English the more structure that is to be imposed on experience the more
pressure there is to construe it in the form of things. But things are merely the end-point
of the metaphonc scale, as set out in Figure 6-3. Processes, though more constrained
tban things, still bave more semantic potential than relators: they accommodate
categories of time and phase, among others, and are construed in open lexical sets,
whereas relators form closed systems. So there is also a pressure mere too, to
metaphorize conjunctions into verbs: then, so, because, before, therefore becoming
follow, result, cause, anticipate, prove. (Circumstances are something of a special case
because most of them already contain participants in minor, subsidiary processes
prepositional phrases in the grammar.) But it remains true that things are the most
susceptible of being classified and organized into taxonomies; hence the primary motif of
grammatical metaphor is that of construing a world in the form of things.
6.7.2 Metaphoric instability of relators
Let us now look at this phenomenon of meiaphoric drift from the other end, asking not
what are the special properties of participants that make them the preferred destination of
metaphonc movement, but what are the special properties of relators that make them the
most volatile and easily displaced. It is clear that, among the various elements involved,
the relators are the most unstable in terms of their susceptibility to metaphonc
transformation. They are, as it were, the first to leave; and they travel farther than the
rest.
We can perhaps link this property of relators to their status in the overall ideation
base. Relators construe the highly generalized logico-semantic relations of expansion that
join figures into sequences: elaborating, extending, enhancing (Chapter 3, Section 3.2).
We have remarked already on the fact that these relationships of expansion pervade very
many regions of the semantic system: they are manifested in the organization of figures
of being, in the types of circumstantial element that occur within a figure, in the
taxonomy of 'things', and elsewhere, as well as of course in their 'home' region of the

268

6. Grammatical metaphor

constatai of sequences, as links between one figure and another. This led us to
characterize the categories of expansion as "transphenomenar and as "fractal" (Chapter 5,
Section 5.6): transphenomenal in the sense that they re-appear across the spectrum of
different types of phenomena construed by the ideational system; and fractal in the sense
that they serve as general principles of the constatai of experience, generating identical
patterns of organization of variable magnitude and in variable semantic environments.
It is these characteristics of relators that make them particularly liable to migrate: to
be displaced metaphorically from their congruent status (as paratactic and hypotactic
conjunctions) and to appear in other guises in other locations as minor processes (in
circumstantial elements), as processes, as qualities and as things. Thanks to this
metaphoric instability, relators are able to play a central part in the re-constnial of
experience that is a feature of the discourse of the sciences that makes these discourses
possible, in fact, and hence provides the semiotic foundation for the construction of
scientific knowledge. We return to this enquiry in Chapter 14 below.
6.7.3

Grammatical metaphor and experiential meaning

The "secondary motif* referred to in Figure 6-3 above, numbered 13i in Table 6(3), is
that whereby a 'thing1 (congruently construed as a noun functioning as Thing in a
nommai group) is metaphorized on the model of some quality qualifying, possessive
or classifying. This represents a shin one step backwards* along the logical-experiential
scale, It is thus contrary to the prevailing general tendency, since something that is
congruently a participant on its own terms is now treated as existing only by virtue of
some other participant.
This type of shift occurs only in syndromes, where it accompanies a metaphor of
either type 1 or type 2. In such a syndrome, the process is reconstnied as a participant;
and as a corollary the participants in that process become its 'qualities'. For example,
Griffith's energy balance approach to strength and fracture, where the participants strength
and fracture and energy balance have become 'qualities' expanding the metaphoric 'thing*
approach- compare the more congruent Griffith approached strength and fracture in
terms of (the concept of) energy balance. The process itself may of course already be
metaphorized from something else, e.g. replace in
The Council's proposed replacement of subsidies by a loan
The Council proposed to replace subsidies by a loan
The Council proposed to lend [money] instead of subsidizing
This perturbation of the dominant pattern has the effect of making a participant more
abstract. In the engine failed, the engine is set up as a thing. In engine failure, it is as it
were deconstructed into a mere characteristic of some other 'thing', a way of classifying

6.7 Interpretation of grammatical metaphor

269

failure into its various contrasting kinds, such as crop failure, power failure and heart
failure. The engine has lost its identity it has no Deictic (note that it cannot be
individuated any longer only the failure can: this engine failure, the earlier engine
failure, any future engine failures, etc.); and it has exchanged 'thingness' with an
ephemeral process, that of failing. But it is still within the compass of a participant in
the figure; grammatically, it is within the nominal group.
We can say therefore that grammatical metaphor is predominantly a 'nominalizing'
tendency. But if we look at it semantically we can see that it is a shift from the logical
towards the experiential: that is, making maximum use of the potential that the system
has evolved for classifying experience, by turning all phenomena into the most
classifiable form or at least into a form that is more classifiable than that in which
they have been congruently construed. We saw in discussing have a look, make a mistake
&c. that if you make look, mistake into nouns you can expand them within nominal
groups: have another good long look, don't make the same silly spelling mistake again!
We now have classes of mistake (spelling mistake); properties, both experiential (long
look) and interpersonal (silly mistake), quantities, and identities (thai same mistake,
another look, three mistakes). The same principle holds when any process is reconstrued
metaphorically as a thing, as in the many examples cited throughout this section.
It should be remembered that the account we are offering here is always an
interpretation of the semantics of English. (We shall refer briefly to metaphor in Chinese
in Chapter 7.) In English, then, the metaphoric movement is from the logical towards
the experiential and, within the experiential, from processes to things. Hence when we
find, in the evolution of scientific discourse in the six centuries since Chaucer's Treatise
on the Astrolabe, a historical progression in the favoured clause type (along the lines of
the implicational scales in Figure 6-3 above):

we recognize this as a shift towards a more highly taxonomized way of meaning. But the
basis for such a shift is found in the ordinary spoken language of everyday discourse.
Consider the following series of examples:
You made three mistakes.
That was his biggest mistake.
Give it another big push.

270

6. Grammatical metaphor

She gave him one of her most heart-warming smiles.


Can't I have just twD little bites of your cake?
That last dive was the best dive I've ever done.

All these are instances of grammatical metaphor, with 'mistake' (verb err), push,
smile, bite, dive turned into things (nouns) and the 'process' taking the form of a
lexically very general verb give, have, do, take, make which retains the full semantic
potential of a figure (tense, modality, &c.). The effect of nominalizing these processes is
to open them up to all the 'quality' potential that is associated with things: they can be
classified, qualified, quantified, identified and described. This range of grammatical
metaphors has become fully codified in English and is, in fact, used by children almost
from the start Another type that is also not specifically associated with learned discourse
with the registers of educational knowledge is that of be/ go + minor process =>
process (type 8i in Table 6(3)): for example
her speech covered five points ('was about')
the road skirts the lake ('goes alongside')
shall I accompany you? ('go with')
this replaces the one you had before ('is instead of)
who does she resemble most? ('is like')

These are not in the language of a pre-school child; they are learnt as the written language
of the primary school, intermediate between the commonsense language of daily life in
the home and the technicaiized educational discourse of the secondary school. We shall
return to the difference and complementarity between commonsense and educational
knowledge in Chapters 14 and 15 .
6.7.4 The significance of grammatical metaphor

We can then summarize our interpretation of grammatical metaphor as follows.


(1) There is an increase in textual meaning, since participants have the most clearly
defined status as information: in particular, they can be construed (by the thematic and
information systems) into a 'backgrounded + foregrounded1 pattern which maximizes the
information potential of the figure.
(2) There is a loss of experiential meaning, since the configurational relations are
inexplicit and so are many of the semantic features of the elements (e.g. engine failure :
an engine / engines//the engine/the engines; failed/fail/mil fail &c.)
(3) There is a further loss of experiential meaning, since the categories of experience
become blurred (failure is not most obviously felt as a 'thing', otherwise it would have
been construed as such in the first place); the construction of reality becomes a

271

6.7 Interpretation of grammatical metaphor

construction of unreality, detached from ordinary experience and hence inaccessible and
remote.
(4) There is however a gain in the potential for experiential information, because
the participant, more than any other element, can be expanded in respect of a wide range
of semantic features; this enables anything construed as a thing to become part of an
experiential taxonomy which embodies far greater generalization about the overall nature
of experience. Martin (Halliday & Martin, 1993: Chapters 9 and 10) has shown that it
would in fact be impossible to construct technical knowledge without grammatical
metaphor of this kind.
Thus grammatical metaphor is a means of having things both ways. An element that
is transcategorized loses its original status because of the nature of the semantic feature(s)
with which it comes to be combined (e.g. 'like ...' is a quality; so when we say mousy
like a mouse* this is only a quality it has none of the thing-ness of the original
mouse). A element that is metaphorized does not lose its original status. Its construction
is not triggered by its being associated with any new semantic feature. If it has a new
semantic feature this is as a result of the metaphorizing process. So failure

is both

process and thing: it is a process construed as a thing (or rather, a phenomenon construed
as a process and reconstrued as a thing); its initial status as process remains, but because
it has been nominalized, and the prototypical meaning of a noun is a thing, it also
acquires a semantic status as something that participates

in processes: see Figure 6-

4. It has become a functional' construct, combining two of the basic properties that the
grammar evolved as it grew into a theory of experience.
Metaphors are dangerous, however, they have too much power, and grammatical
metaphor is no different in this respect. Because it leaves the relations within a gure
almost totally inexplicit, this demands that they should be in some sense already in
place. In the typical rhetorical context for the highly favoured 'backgrounding' type, as we
have seen, the configurational relations have been established in the preceding discourse:
cf. the first example in Section 6.1 above ... if one takes alcohol one's brain rapidly
becomes dull Alcohol's rapid dulling effect on the brain... Here by the time we reach the
metaphor they arc already in place: we know that rapid dulling effect means 'causes ...
rapidly to become dull', not any of the other things it might mean such as 'has an effect
which soon becomes dull, or blunted*. But this is an idealized example, constructed for
the purpose. Usually the configurational pattern will have been built up over long
stretches of the text, or (especially if it is a technical form of discourse) over a great
variety of different texts for example, a series of textbooks used in teaching a science
subject throughout a school. Very often the learner has to construct the configurational
relations from various sources without their being made fully explicit in any one place;
and in the limiting (but by no means unusual) case they have never been made explicit at
all, so that the figure has to be construed from the metaphor a very difficult

272

6. Grammatical metaphor

task indeed. So the more the extent of grammatical metaphor in a text, the more that text
is loaded against the learner, and against anyone who is an outsider to the register in
question. It becomes elitist discourse, in which the function of constructing knowledge
goes together with the function of restricting access to that knowledge, making it
impenetrable to all except those who have the means of admission to the inside, or the
select group of those who are already there.
It is this other potential that grammatical metaphor has, for making meaning that is
obscure, arcane and exclusive, that makes it ideal as a mode of discourse for establishing
and maintaining status, prestige and hierarchy, and to establish the paternalistic authority
of a technocratic elite whose message is 'this is all too hard for you to understand; so
leave the decision-making to us1 (see Lemke, 1990b). Even those who most exploit its
potential for organizing and constructing knowledge theoretical physicists and other
specialists in the natural sciences - are now finding that they have had 'too much of a
good thing' and are seeking ways of overcoming it and carrying it to less extreme
manifestations. But this involves the whole consideration of language in the construal of
experience, to which we return in the culminating sections of the book.

Fig. 6-4: Congruent construal and metaphorical reconstrual junctional constructs


6.7.5 Metaphor as interpersonal

reconstruction

It is many years now since Whorf first drew attention to some of the prevailing
metaphors in what he referred to as "Standard Average European" languages: such things
as the way cognitive processes are typically construed in terms of concrete actions and
movements in physical space: e.g. grasp, follow = 'understand1, the line or direction of
an argument, and so on (cf. Section 6.2.1 above). In a well-known paper, Reddy (1979)

6.7 Interpretation of grammatical metaphor

273

explored this particular domain in greater depth and showed how in English the entire
semantic field of saying and sensing is permeated by what he called the "conduit
metaphor", according to which meaning is "contained" in thoughts or words and may be
"conveyed" along some "channel" from a speaker to a listener.
Lakoff & Johnson (1980), in Metaphors We Live By, showed just how many of our
basic conceptual schemata and reasoning strategies aie shaped by their metaphorical
make-up in the everyday language. To continue with this same domain: using examples
such as you're going around in circles, their argument has holes in it, if we keep going
the way we're going we'll fit all our facts in,

they demonstrate that the motif of

'argument' is construed by a combination of two metaphors, the 'journey1 and the


'container' (Chapter 16). Such patterns of coherence (CM* "frames of consistency", in
Whorf s term) across metaphoric regions are typical of the "structuring of concepts" (p.
96) which determine (to quote their own use of this same metaphor) "how human beings
get a handle on the concept" (p. 116) and function with it hi daily Ufe.
Although they make some reference to particular grammatical categories (e.g. "with
few exceptions,... in all the languages of the world the word or grammatical device that
indicates ACCOMPANIMENT also indicates INSTRUMENTALITY" (p. 135) with
in English), Lakoff & Johnson's "metaphors we live by" are largely presented as lexical
metaphors: that is, in terms of individual words, and sets of words that are semantical!y
related. Sometimes however the metaphors of daily life arise rather from metaphoric
movement in the grammar: for example, many of our everyday expressions for
behavioural processes, like have a bath, take a look, give a smile, do a turn, involve
construing the process (congruently a verb) in the form of a noun. Such metaphors are
even less accessible to conscious reflection than the lexical ones, and so readily diffuse
throughout the system and become the norm.
Since we are suggesting that grammatical metaphor is not simply an alternative
realization of the same meanings, but a distinct construing of experience in which there
is junction of semantic features (have a bath, do a turn may be "dead metaphors", but in
bring about a conclusion there is clearly a junction between the 'process' meaning of
'conclude1 and the 'thing1 meaning of noun), we could expect to find this same
phenomenon in lexical metaphor: that the 'literal' meaning of the transferred term would
remain, injunction with the features acquired in its metaphorical environment. Just as, in
bring about a conclusion, conclusion combines the category (word class) meanings of
verb and noun, so in get a handle on the concept, handle combines the item (word)
meanings of handle and idea. Just one instance, or a few instances, would have little or
no effect; but when there is a rather massive frame of consistency whereby the same
metaphor, or metaphoric syndrome, extends across a major region of semantic space this
must play a significant part in our overall construction of reality. So, to cite another
example from Lakoff s work, in his (1992) study of the metaphors "used to justify [the

274

6. Grammatical metaphor

1990] war in the [Persian] Gulf \ he identifies a number of dominant motifs he refers
to these as "metaphoric systems" such as state-as-person, fairy tale of the just war
(with "self-defence" and "rescue" scenarios), ruler-for-state metonymy; war as, selectively,
violent crime, competitive game or medicine all of which he finds to have been
applied in portraying Saddam Hussein as villain, Kuwait as victim, and in constructing
the concept of "victory" ('the game is over'), of the "costs" of war and so on. Lakoff
comments "What metaphor does is limit what we notice, highlight what we do see, and
provide part of the inferential structure that we reason with".
Lakoff s conclusion is that, while we cannot avoid such an all-pervasive outbreak of
metaphor, we can learn to recognize it and to understand the harmful effects it may have
("that it can kill", in his formulation); we may also be able to sede more benign forms of
metaphor to replace it Looking at such examples in our own terms, we would want to
add another dimension to the interprtation, by seeing the metaphoric process as
essentially a lexicogrammatical one and pointing to the grammatical element in the
overall construct This enables us to do two things. On the one hand, we can bring out a
further aspect of the semantic picture by pointing to the conjunction of category
meanings an aspect of grammatical semantics that is involved; and on the other
hand, we can relate this particular metaphoric phenomenon to the overall semantic
potential of the system the construal of experience as a generalized ideation base. And
we should stress once again that to describe the "reality" that is construed in this way as
being generalized does not imply that it is "coherent", in the sense that it is internally
consistent and unselfcontradictory. On the contrary: much of the power of metaphor
derives from the tensions and contradictions set up (a) within the metaphor itself, (b)
between one metaphor and another, and (c) between the metaphor and other regions of the
ideation base. Thus to follow through the metaphor of war as game, we can recognize a
number of other factors relevant to its interpretation: both are typically construed as
behaviour + Range (fight a battle/war, play a game/set... ); each can stand as metaphor
for the other (cf. the language of sports commentaries); the war-as-game metaphor
conflicts with their different semantic loadings in affective clauses (e.g. enjoy the game!)
and, even more, in impersonal existentials (there's a war on, there was a fierce battle last
night see below on the depersonalizing of war discourse); and so on. This is not to
say that every instance of a lexical metaphor resonates powerfully throughout the
grammar. But the general phenomenon of metaphor, as an inherent property of language
as a stratified semiotic, is a feature of the system as a whole of the construal of
meaning in lexicogrammatical terms.
One of the instances of grammatical metaphor we will use to illustrate the principle
of (lexico-)grammatical unpacking in Section 6.8.3 below was in fact taken from a
military training document; so it is interesting to review it in this light. One of the
features that stands out is that, with the syndrome of grammatical metaphors, all

6,7 Interpretation of grammatical metaphor

275

personalized participants disappear, and it is noticeable that this is not just a feature of
this particular clause there are no human participants throughout the text, with the
single exception of the enemy, which is the nearest thing to a person that occurs. The
following is a typical extract:
The Airland Battle Concept outlines an approach to military
operations which realizes the full potential of US forces. Two
notions - extending the battlefield and integrating conventional
nuclear chemical and electronic means - are blended to describe a
battlefield vfriere the enemy is attacked to the full depth of his
formations. What we seek is a capability for early initiative of
offensive action by air and land forces to bring about the
conclusion of the battle on our terms. [ . . . 3
This concept does not propose new and radical ways to fight.
Rather it describes conflict in terms of an environment which
considers not only conventional systems, but also chemical/
nuclear, and electronic. It also forces consideration of this
conflict in terms of reaching the anamy's follow-on echelons.
Consideration of such a battlefield is necessary if we are to
reinforce the prospects of winning.
What the metaphor does is on the one band to construct an entirely abstract world of
virtual objects such as concept, approach, capability, environment, considerations,
prospects, and potential', and on the other hand to set this up in stark contradiction to the
highly concrete processes that begin to emerge as the metaphors are unpacked. This in
turn is a kind of hyper-metaphor for modern war, in which the only "military" action
taken may be the inherently benign one of keying a message into a computer the
outcome of which is that thousands of people, thousands of miles away, die a violent and
wholly unmetaphorical death. Granted that this is an extreme and perhaps sensational
example; but the point we are trying to make is one that is central to language itself
namely, that it is the potential for (grammatical) metaphor (itself a product of the
stratified ideational resource system) that makes it possible to construe experience in
terms of such complementarities and contradictions.
6.7.6 Metaphor as ideational

reconstruction

We have suggested that grammatical metaphor has been central in the construction of
scientific, or uncommonsense, experience. This is equally true of lexical metaphor:
experience construed within various disciplinary frames depends on metaphorical
syndromes that are lexicogrammatical in nature i.e., that combine lexical and
grammatical metaphorical syndromes. We will return to this point in Part V when we

276

6. Grammatical metaphor

examine the metaphorical construction of the domain of study in mainstream cognitive


science in Chapter 14. Here we will illustrate briefly with reference to the discourse of
economics (recall here Whorf s account of intensity in Section 6.2.1 above). Consider the
following repon from the Federal Reserve Board in the U.S.:
St*p

dclins in capital spending comnitments and building

permits, along with a drop in the money stock puntad the leading
composite down for the fifth time in the past 11 months to a
lvl

of 0.5% bftlow

its

high

in

[month]

[year].

Such a

clin* is highly unusual at this stage in an xptnslon;

for

example, in the three most recent xp*nmionm, the leaders were


rising, on average, at about a 7% clip at comparable ph* in
the cycle. While not signalling an outright r*caaion,
current protracted

the

of the leading indicators

mlugglmhu+mm

appears consistent with our prognosis of luggiah

real GNP

growth over the next few quarters.


Economics construes quantities in terms of economic things: either their location and
movement in space (shown above in bold) or their size and growth/ shrinkage (shown
above in bold italics). These lexicogrammatical metaphors may be further metaphorized
'steep + decline1. These
by grammatical metaphor, as with 'decline + steeply'
metaphorical movements can then take on participant roles in other movements, as in
'steep declines... pushed the leading composite down*. The report is highly metaphorical,
as the following chain suggests: 'people were committed to spending capital less and less
quickly'

'[people's] commitments to spending capital declined sharply'

'commitments to capital spending declined sharply'

'capital spending commitments

declined sharply'
'sharp decline in capital spending commitments'. We could imagine
one step further towards the metaphorical pole: 'sharp capital spending commitment
decline'!
The spatial metaphor actually makes it possible to express the domain graphically as
well as verbally. The following passage illustrates how the spatial metaphor serves as the
cross-over point between the two semiotic systems:5
With these considerations in mind, we can construct the marginal
cost of funds schedule shown in Figure 4.6. Region A represents
financing done by the
depreciation

firm from

(D). Ihere is

retaining earnings

no risk

factor

involved

(RE) or
in this

From Michael K. Evans, 1969. Macroeconomic Activity. Theory, forecasting and control.
New York: Harper & Row, Publishers, p. 89.

277

6.7 Interpretation of grammatical metaphor

r0ion; the only cost of funds is the foregone interest that


could be earned by investing funds lwhr. Thus true cost of
borrowing in region A is equal to market interest rate. Region
B represents financing done by borrowing from banks or bonds. The
harp ri in the true cost of borrowing is not primarily due to
a rim* in the market interest rate at which firms must borrow.
It is due instead to the imputed risk factor which occurs with
increased debt servicing. Region C represents

financing done

through equity capital. Here again there is no imputed risk,


because the firm does not have to pay dividends. The gradual
upward ftlop* is due to the fact that as a firm offers more and
more of its stock on the market, this will invariably dpre
its price and raise the yield that is paid.

Pig. 4.6

The correlation between the two semiotics is constructed verbally by means of figures of
identifying, e.g. region A represents financing (shown in italics). Quantities and
quantitative changes are construed verbally in an abstract space, e.g. gradual upward slope
(shown in bold); and also diagrammatically in a symbolic space.
In the construction of meaning in economics, economic things such as prices,
salaries, yields, financing are quantities. At the same time, they are also elastic bodies
that grow and shrink; and also 'mountaineers', climbing up and down slopes. This is at
one level an impossible conjunction: things cannot be both quantities and mobile elastic
bodies but the semantic system of English

makes this possible through

lexicogrammaticaJ metaphor. This proliferation of perspectives is pan of the semogenic


power of metaphor.

278

6. Grammatical metaphor

We noted above that the deployment of spatial metaphor in the construction of


economic "knowledge1 serves, among other things, as an interface to another form of
semiotic that of using space symbolically in a diagrammatic representation of
quantities. The same kind of principle of cross-over between semiotic modes applies to
the construal of meaning in linguistics. We have relied in our own discussion on
metaphors of abstract space for construing meaning most centrally, metaphors of
semantic networks and semantic space. These metaphors allow us to cross over to
diagrams of symbolic space, viz. system networks (as a kind of acyclic directed graph)
and topological representations. The system networks can, in turn, be restated
algebraically, as in the various computational implementations of systemic accounts.
The question of how to restate the topological representations so that they can be
manipulated computationally remains very much more open (see Matthiessen, 1995a, on
fuzzy representation).
When we now come to consider the question representing our account of
(grammatical) metaphor in analytic terms, we will naturally be asking, among other
things, how to represent metaphor as we ourselves metaphorically construe it as an
expansion of semantic space by means of junctions.
6.8 Representing grammatical metaphor

We have sketched a theoretical account of grammatical metaphor an account that


relates it to our overall conception of the ideation base. This still leaves open the
question of how to model grammatical metaphor at the lower level of formal
representation; and in particular, how to represent junctional categories.
6.8.1 Recapitulation for the purpose of representation

Let us stan with a quick recapitulation of the central features of grammatical metaphor.
Categories in the ideation base are realized by those categories in the ideational grammar
with which they have co-evolved. These are the congruent realizations that developed first
in the language, are learnt first by children and tend to occur first in a text For instance,
a sequence is realized by a clause complex, and the figures related in the sequence are
realized by the clauses strung together in the clause complex: see Figure 6-5. Thus a
sequence consisting of four temporally related figures such as the following:

279

6.8 Representing grammatical metaphor

She rose to power and popularity in a phenomenal way

and

then

she fell in a fantastic way

and

then

she was resurrected politically even more rapidly and


remarkably
less than

four years before

she was assassinated in the autumn of 19&4.

would be realized congruenuy by a clause complex consisting of four clauses related by


enhancement, the first three paratactic and the final pair forming a hypotactic
subcomplex:
1 She rose to power and popularity in a phenomenal way, X 2 she
fell [from power and popularity] in a fantastic way, X3ct and then
she was resurrected politically even more rapidly and remarkably
3*pless than four years before she was assassinated in the autumn
of 1984.

Fig. 6-5: Congruent realizations

280

6. Grammatical metaphor

(There are metaphors internal to the figures e.g. phase construed as a vertical motion
relative to a quality dimension (rise to power = 'become powerful') but we are
concerned with the realization of the sequence as a whole. Deconstructing the metaphors
internal to the figures, we would get something like the following: She became powerful
and popular in a phenomenal way, she became powerless and unpopular in a fantastic
way, and she was made politically powerful and popular again even more rapidly and
remarkably less than four years before she was assassinated in the autumn of 1984.)
The system then comes to be expanded through shift in rank and in class. Sequences
come to be realized not only by clause complexes but also by clauses, and figures come
to be realized not only by clauses but also by groups/ phrases, as shown in Figure 6-6.
They are pushed downwards in complexity and rank relative to their congruent
realizations in the grammar.

Fig. 6-6: Metaphorical realization of sequences and figures


Thus our earlier sequence can be realized metaphorically by a single circumstantial
relational clause:
Her phenomenal rise to power and popularity was followed by a
fantastic fall and then by an even more rapid and remarkable
political resurrection less than four years before her
assassination in the autumn of 1984. (Inder Malhotra, Indira Gandhi. A
personal and political biography, p. 11.)

The clause structure is given in Figure 6-7. The two alternative realizations are shown
in Figure 6-8 as a shift from the congruent to the metaphorical within the
lexicogrammar. Such a representation suggests that the congruent and the metaphorical

281

6.8 Representing grammatical metaphor

are simply realization^ variants. But is this kind of interprtation rich enough? At this
point, let us consider the issue of representation with respect to structural configurations
in text such as the one shown in Figure 6-8. We will then turn to the question of how
the metaphorical expansion of the semantic system might be represented.

U) Her phenomenal
rise to power and
popularity

Value:
nom. gp.
by [2] a fantastic fall

was followed

Process:
verbal gp.

even more rapid


and remarkable

Deictic Epithet

Token:
prep, phrase complex

and then by [3] an even more rapid and


remarkable political
resurrection

prep, phrase

[3] an

by 12] a fantastic fall


and thn
by [3] an even more rapid and remarkable political resurrection
[less than four years before [4] her assassination in the autumn of 1984].

[less than four years before


[4] her assassination
in the autumn of 19841-

prep, phrase

political

Classifier

resurrection

Thing

[4] less than four years before her assassination in


the autumn of 1984
Qualifier

Fig. 6-1; Metaphorical realization of sequence as circumstantial relational clause


6.8.2 Representing text examples
The following are some of the factors to be taken into account when we try to represent
examples of semantic structural configurations:
(i) Do we represent the text as it is (metaphorically)?
(u) Do we represent the text in "unpacked" form (congruently)?
(iii) How far do we unpack it (move towards congruence)?
(iv) If we give two (or more) representations, what relationship do we show
between them (how do we show the agnation between congruent and
metaphorical)?
For example, given [engines of the 36 class only appeared on this train] in times of
reduced loading or engine failure:

6. Grammatical metaphor

282

(i) Do we analyse in times of reduced loading ?


() Do we analyse when the load was reduced?
(iii) Do we unpack to when less freight was being carried?
(iv) Given the analyses in (i) and (ii) (and (iii) if applicable), how do we show
them to be related?

Fig. 6-8: From congruent to metaphorical realization of sequence


(i) Do we represent the text as it is (metaphorically)?
In the account of the introductory example above concerning the fall and rise of Indira
Gandhi, the burden of the representation would be placed entirely on the inter-stratal
relation between semantics and lexicogrammar. The metaphorical variant is introduced
merely as a realizational variant. In a text-generation system, this would imply
that, instead of a sequence being realized as a clause complex from the start, the

6.8 Representing grammatical metaphor

283

realization is 'delayed' until the grammatical domain is that of the clause rather than the
clause complex, and the clausal resources of TRANSriWITY are used to realize the
sequence, as with the example in Figure 6-7, This means that the sequence is realized as a
circumstantial relational clause. The first gure is realized as a nominal group serving as
the Value and the remaining three figures of the sequence are grouped together as the
Token, internally organized as a nominal group complex (if we treat the preposition by
simply as a structure marker). The last figure is further downgraded as a Qualifier in the
second nominal group of the complex. These realizational relations are shown in Figure

6-9.

Fig. 6-9: Sequence realized by clause


Such an approach, where grammatical metaphor is a free realizational variant could
be made to work; and there may be good practical reasons for adopting it for certain
purposes. However, we have shown that the metaphorical version is not simply a
meaningless (i.e. synonymous) variant of some more congruent form; it is 'junctional'
that is, it embodies semantic features deriving from its own lexicogrammatical
properties. We therefore analyse it in the first instance just as it stands see Figure 610 for the engine failure example. Here we have the logico-semantic relation of time
'when' (congnienuy construed as an expanding sequence) construed as if it was a minor
process 'during1; and the figures load being reduced' and 'engine failing' construed as
elements participating in this minor process.6

We might note here that in analysing in times of as a preposition group we have already
gone beyond analysing it strictly 'as it stands', since the Head of the nominal group
construction is times: that is, the logico-semantic relation is 'first' construed nominally, as
if it was itself a participating element (a minor process of location). We can bring this out

284

6. Grammatical metaphor

Fig. 6-10: Analysis of metaphorical text 'as it stands'

Fig. 6-11: The same represented as modification structure


(ii) Do we represent the text in "unpacked" form (congruently)?
If the metaphor can be unpacked demetaphorized to yield a more congruent
wording such as whenever) the load was reduced, or an engine failed, then we can
represent the text over again in this congruent form (see Figure 6-12). Here the sequential
logicosemantic relation of time is construed as a hypotactic conjunction when (or
whenever), and *being reduced1 and 'failing* are construed as processes. What we are
saying, with this analysis, is that in times of... represents a separate figure, rather than a
circumstance within a figure; and that this in turn is made up of a paratactic sequence of
two figures, one ascriptive 'the load was reduced' (or possibly one of doing-&-happening,
'the load had been reduced1), the other one of doing & happening 'the engine failed'. The

by representing the nominal structure logically in terms of modification (see Figure 6-11).
It would be possible to go one stage further and interpret times experientially as Thing; if
we do this we are saying that it is actively metaphorical, retaining the semantic feature of
participant from its realization as a noun. Usually in analysing contemporary texts we do
not treat forms such as these as active metaphors (cf. facet expressions such as in front of),
since they appear semantically and grammatically dormant Historically all such
expressions embody a nominalizing metaphor.

285

6.8 Representing grammatical metaphor

text as it stands is metaphorical; a more congruent constnial would be as reworded here;


and the analysis now reveals the difference between the two.
...

when the load was

reduced

or an engine

failed

Fig. 6-12: Analysis of metaphorical text reworded in more congruent form


(iii) How far do we unpack it (move towards congruence)?
Regarding question (iii), we could continue to 'unpack1 the metaphor further at one
point, for example from (when) the load was reduced to (when) less freight was being
carried. This does not change its status as a figure; it merely redistributes the semantic
features among the elements of the figure (see Figure 6-13).

Fig. 6-13: Further unpacking of first figure in paratactic expansion


The argument against taking this further step is that it introduces another element of
metaphor. In (1), the bracketing is ((carry + goods): fewer), and this construes the
message congrueny with 'the goods carried1 as point of departure (the load as Subject/
Theme) and fewer1 in culminative position; whereas in (2) the bracketing is ((goods +
fewer): carry), which distributes the information in a rather less congruent form, having
'cany' instead of 'fewer' in the position typically associated with the 'news'. In other
words, the cost of removing the experiential metaphor is to introduce a new textual
metaphor; this provides an important insight into why experiential metaphors evolve in
the first place, namely in order to get the texture right (cf. Section 6.3 above; Halliday,
in press a, b), but is also a useful criterion for stopping short in the process of
unpacking. Of course, this is a feature that applies only to certain classes of example, not

286

6. Grammatical metaphor

to all; but it is typical of the son of consideration that would be relevant to answering
the third of our four questions.
(iv) How do we show the agnation between congruent and metaphorical?

Coming now to the fourth question: how do we show the analyses under (i) and (ii) to
be related? There are really two distinct parts to this question. One is: do we need two
separate phases in our representation? And the second is: if we do need both, do we need
a third phase whereby the two are shown to be related? The first raises the issue of
metaphorical junction. In our interpretation, the text as it stands, with the grammatical
metaphor left in, embodies semantic junction: it is not just a variant form, identical in
meaning with its congruent agnate it also incorporates semantic features from the
categories that its own form would congruently construe. Thus engine failure is not
synonymous with engines fail; it is both a figure consisting of participant ('engine1)
and process ('fail') am/ an element (participant) consisting of thing ('failure') + classifier
('engine'). In other words, we need both analyses in order to represent it adequately.
This will always be true whenever the metaphor can be unpacked to yield a plausible
more congruent form. And this is what distinguishes a grammatical metaphor from a
technical term. Almost all technical terms start out as grammatical metaphors; but they
are grammatical metaphors which can no longer be unpacked When a wording becomes
technicalized, a new meaning has been construed almost always, in our present-day
construction of knowledge, a new thing (participating entity); and the junction with any
more congruent agnates is (more or less quickly) dissolved If for example we said that
engine failure had now become a technical term, what would we mean by this? We would
mean that the semantic bond with a figure an/the engine fails had been ruptured (it could
no longer be 'unpacked'); and that a new meaning, an abstract participating entity or thing
'engine failure' had come into being which had the full semantic freedom to participate
in figures, to admit of classes and properties, and the like.
But let us return to grammatical metaphor that has not become technicalized and
retains its character as semantic junction. Here we might want not merely to retain the
two phases of representation, that of the metaphorical wording as it stands and that of the
congruent wording as it is unpacked, but also to build in some representation of the
agnate relationship between them. This can be done with some kind of composite
representation as hi Figure 6-14 (and cf. Figures 6-1, 6-2 above). (Such diagrams are
considerably more effective if they can be colour-coded)

287

6.8 Representing grammatical metaphor

in times of

reduced

minor Process

minor Range:
1

Epithet

Relator

when

Carrier

the load

Process

was

loading

or

engine

failure

Thing

Classifier

Attribute

reduced

or

Thing

Actor

Process

an engine

failed

Fig. 6-14: Composite representation of metaphorical and congruent wordings


The alternative is to try to conflate the two into a single representational phase. It is
difficult to do this without making the result far too complicated; but an attempt is made
in Figure 6-15. What this does is set up distinct junctional categories based on the types
of grammatical metaphor shown in Table 6(3) above. This kind of approach is perhaps
the one to be developed further, since it brings out more clearly the potential of
grammatical metaphor as a semogenic resource. But we shall not attempt to pursue it
further here.

6. Grammatical metaphor

288

Fig* 6-15: Single phase representation of text with metaphorical wording


6.8.3

Grammatical metaphor as elaborating relation between semantic

configurations
The type of representation used in Figure 6-14 (and also in the representation of the
'prolonged exposure' example in Section 6.5 above) shows metaphor as a correspondence
between two semantic configurations. The correspondence that is construed through
grammatical metaphor is an elaborating relationship: an identity is set up between
two patterns, a sequence and a figure, a figure and a participant, and so on. In this
identity, the metaphorical term is the Token' and the congruent term is the 'Value* (cf.
Figure 6-4 above): 'engine failure' stands for (means, represents) 'engines fail'. This is the
'core' meaning of the elaborating relation; but it also covers the senses of 'summarize',
'distil' the metaphor may 'distil1 congruent meanings that have accumulated in the
text. The identity holds between the two configurations as a whole; but, as our
representations indicate, the components of the configurations are also mapped one on to
another: see Figure 6-16.
The metaphorical relation is thus similar to inter-stratal realization in that it
construes a token-value type of relation. Here, however, the relation is intra-stratal:
the identity holds between different meanings, not between meanings and wordings. The
metaphor consists in relating different semantic domains of experience (cf. Section
6.2.1): the domain of figures is construed in terms of the domain of participants, and so
on (just as in a familiar lexical metaphor the domain of intensity is construed in terms of
the domain of vertical space). It is the fact that metaphor multiplies meanings within the
semantic system that opens up the possibility of metaphorical chains, with one
congruent starting-point and another highly metaphorical end-point (A"' stands for A"
stands for A' stands for A; e.g. 'engine failure* stands for 'the failing of an engine' stands
for 'an engine failed'). The semantic system is being expanded along the dimension of the

6.8 Representing grammatical metaphor

289

metaphorical token-value relation; but tbe expansion is still within the semantic system
itself.

Fig. 6-16: Metaphorical Token representing congruent Value


Another possible form of representation would be in terms of the VIEW relation
introduced by Jacobs (1987) to handle metaphorical pairs. Jacobs suggests that the
relationship between pairs such as Ali punched Frazier and All gave Frazier a punch can
be handled at tbe level of conceptual structure by means of a metaphorical relation of
VIEW set up between the two types of process. The VIEW represents the metaphor of
action as transfer-event, in Jacobs1 terms: see Figure 6-17. According to this
representation, an action is viewed as a transfer-event and the object of the action as the
recipient of the transfer. The VIEW relation reflects the same type of correspondence that
is embodied in the box diagram interpretation of a grammatical metaphor used in Halliday
(1985); cf. Figure 6-14 above. For instance, the metaphorical example Ali gave Frazier a
punch can be diagrammed as in Figure 6-18. Note that the metaphorical function of a
punch is modelled not on that of Goal but on that of Range in traditional terms, it is
a "cognate object" rather than a "direct object" (Halliday, 1985: Section 5.6).

6. Grammatical metaphor

290

Fig. 6-17: Metaphorical VIEW adapted from Jacobs (1987)

Fig. 6-18: Metaphorical correspondence in box diagram


For most types of metaphor, it is important to take account of the fact that the
metaphor represents more than one construction of goings-on; the metaphor adds a further
perspective on the phenomenon being represented, without displacing the perspective that
is congruent Thus, Ali gave Frazier a punch is like Ali gave Frazier a rose; but it is
also unlike it precisely because of its metaphorical status: it evokes a non-metaphorical
agnate Ali punched Frazier. There are also grammatical distinctions between the two; for
instance, there is no systematic proportion
Ali gave Frazier a rose : Ali gave a rose to Frazier

::

Ali gave Frazier a punch : Ali gave a punch to Frazier


The grammatical metaphor will typically show features of the congruent perspective as
well as features of the metaphorical one.
The approach outlined by Jacobs may account for certain types of grammatical
metaphor. However, it leaves a number of issues still to be addressed, (i) The VIEW
relation represents the coirespondence between two concepts and their roles, but it does
not adequately represent the meaning of the metaphorical view. For instance, what does it

6.8 Representing grammatical metaphor

291

mean to view an 'action* as a 'transfer-event'? (ii) Secondly, it does not take account of
the conditions under which the metaphorical view is chosen over the congruent one. For
instance, with an example such as 'Ali gave Frazier a punch', the crucial factor is
probably not that it is being viewed as a transfer-event; rather, a more complex twofold
"remeaning" has taken place. The process 'punched' is being represented as a participant 'a
punch': it thus takes on the experiential status of a thing it is objectified, and so can
be disposed of (like a gin); and by the same token, it takes on the textual potential
whereby it can appear in culminative position in the clause the unmarked locus of
new information. But it does not thereby lose its experiential standing as a process.
As far as representing text instances of grammatical metaphor is concerned, the two
forms of representation that we have used in our own discussion in this section the
composite box diagram above, and the arrow-field diagram (Figure 6-8 in Section 6.8.1)
could be seen as variants of Jacobs' view approach. But they do perhaps constitute a
third type, since they show the construction of metaphorical space; or rather, the
metaphorical construction of space, the way the grammatical metaphor adds a further
dimension of depth to semantic construal of experience. The box diagram shows this in
synoptic form, presenting each element in the grammatical structure as semantically
complex. The arrow-field diagram shows it in dynamic form, the motifs of the congruent
wording being tracked through one by one to their metaphorical values.
As we presented these, they both portrayed a simple binary relation between two
forms of wording, one metaphorical the other congruent But we have pointed out more
than once that the metaphorical dimension in grammar is in reality a cline. There are
often numerous intermediate steps between the "most congruent" and the "most
metaphorical" wordings; indeed it is the scale of metaphoncity that is reasonably clearly
defined, not its end points. Given two agnate wordings that are positioned along this
scale, we seldom have any difficulty in locating them relative to each other: we know
which of the two is the more metaphorical one. But we hard put to it to specify a point
at either end where we feel we could go no further.
It is not difficult to embody these intermediate steps in a form of representation such
as the following in Fig. 6-19. This does not incorporate any new features; it merely
combines the direction of derivation with the functional labelling. What it does, however,
is to model the history of the particular instance, showing what semantic features it has,
as it were, picked up along the way. It would be possible to explore many other modes of
visual presentation; there is no single ideal form, and the approach will vary according to
the purpose of the teacher or researcher. Any picture that brings out the two properties of
elasticity and directionality would meet the basic requirements.

292

6. Grammatical metaphor

Il what I we I seek 11 is I a capability [for [early initiative [of [offensive action [by
[air and land forces J ] I to bring about i
the conclusion [ of [battle] ] [on [our terms ] ] I ] ] ] ] II
Fig. 6-19: From metaphorical to congruent with intermediate steps

6.8 Representing grammatical metaphor

293

6.8.4 The representation of metaphor in the system


We have considered how metaphorical examples might be represented as semantic
configurations related to congruent configurations, very likely with several intermediate
steps. We now have to take one step further and explore how semantic types can be
represented as metaphorically related to congruent types within the overall semantic
system. Such a representation has to meet the kinds of ideational and textual demands
that we have already considered in our representation of examples:
(i) ideational: the representation has to show that metaphor constitutes an
expansion of the semantic system. The expansion is an elaborating one, creating
chains of token-value relations; and it increases the semantic potential for
construing experience.
(ii) textual: the representation has to make it possible to show how metaphorical
and congruent variants aie given different values in the text base in the
textual semantics.
These considerations suggest to us that metaphor is best construed as an opening up
of a new dimension of the semantic system that allows the whole system to be elaborated
by itself along a token-value chain. In other words, what we identified as a simplified
way of dealing with grammatical metaphor in the relation between semantics and
lexicogrammar in Figure 6-8 we can now recognize as a dimension internal to the
semantic system. Figure 6-20 shows this opening up of a metaphorical dimension within
the semantic system.
This diagrammatic sketch is also intended to show that metaphor involves a mapping
from one ideational domain to another, construing a value-token relation between the two
(cf. Figure 6-16 above). For instance, the whole semantic domain of sequences may be
mapped onto the domain of figures. Hie token domain may in turn be the value in a
further metaphorical move, such as the move from figures to participants. Hie
metaphorical expansion thus can thus involve multiple planes. As we noted above in
Section 6.7, the general tendency in the metaphorical move away from the congruent is
away from the logical towards the experiential; and within the experiential towards the
domain of participants in figures of being & having. We shall return at the end to a
related aspect of this move a move from commonsense or folk models of our
experience towards uncommonsense or scientific onesand illustrate it from the field of
cognitive science with respect to the construal of 'the mind' (see Chapter 14). The
proliferation of ideational meaning through metaphor thus also means that a person gains
access to a wider repertoire of communicative roles.

294

6. Grammatical metaphor

Fig. 6-20: Metaphorical elaboration of the semantic system


Each 'plane* has the same kind of interface to the textual metafunction. In choosing
whether to construe some phenomenon as a sequence of figures or as a sequence
reconstrued as a figure, we can thus also compare the implications for the textual
presentation of these alternative ideational articulations as quanta of information.
Each plane also has the same realizational potential in the lexicogrammar. Thus if the
domain of sequences is construed metaphorically within the domain of figures, the
realizational domain in the grammar will automatically be that of the metaphor: it will
be the domain of clauses rather than that of clause complexes.
The diagram in Figure 6-20 illustrates global mappings between sequences and
figures, and between figures and participants. We also need to take account of local
mappings: some more delicate type of figure must be able to be represented by some
more delicate type of participant. Thus figures of quality ascription will typically be
represented by participants with the quality rather than the process reified as the thing:
lier speech was brilliant' => 'the brilliance of her speech*.
Our representational sketch is just that a sketch. Many details have to be filled in;
and the representation has to be made fully explicit There is one further consideration we
have not mentioned yet that we believe is part of the key to a general solution of the
representational problem. The whole metaphorical elaboration is made possible by a
fractal pattern that runs through the whole system. We have suggested that the
metaphorical elaboration is a token-value relation; but in order for it to be a token-value
relation within the semantic system, it has to be natural in the sense that the token and

6.8 Representing grammatical metaphor

295

value domains have to be similar enough to allow the token to stand for the value. For
instance, a sequence has to be similar enough to a figure to allow it to stand as a
metaphorical token for this congruent value. The principle behind this similarity is the
fractal pattern of projection/ expansion that we met throughout the semantic system of
sequences, figures, and elements in Chapters 3,4, and 5 above and that was summarized
in Section 5.6.
That is, while grammatical metaphor constitutes a move from one "phenomenal
domain" to another from sequence to figure, and from figure to element, this move is
made possible because the fractal types engender continuity across these domains: the
metaphorical move from one phenomenal domain to another takes place within one and
the same transphenomenal domain. For example, the metaphorical shift from he added
and smiled to he added with a smile is a shift from the phenomenal domain of sequence
to the phenomenal domain of figure (accompanied by the shift from figure and smiled to
element with a smile)', but the transphenomenal domain of 'extension' remains constant:
the extending sequence he added and smiled is metaphorically agnate with the figure he
added with a smile with an extending circumstance.
The detailed example of 'cause* as a grammatical motif given in Appendix 3 of
Halliday (1994) can be interpreted along these Unes: 'cause* is a transphenomenal type
within the more general type of 'enhancement*. It is manifested within sequences, figures
and elements, so like other transphenomenal types it engenders transphenomenal relations
of agnation involving grammatical metaphor (cf. the schematic example on p. 267
above); for example:
[sequence (congruent):]
she died, b*cau* she didn't know the rules
[figure (metaphorical):]
she died through ignorance of the rules
her death was du to ignorance of the rules
her death ** c*ud hy ignorance of the rules
ignorance of the rules was the c*u* of her death
[element (metaphorical):] her death du to

ignorance of the rules (was

reported in the paper)

In conclusion, we will give one set of examples for each transphenomenal type: see
Table 6(4). (For ftirther examples, see Matthiessen, 1995b: 163-8.)

296

6. Grammatical metapbor

Table 6(4): Transphenomenal agnation

projection

expansion
elaboration

extension

enhancement

sequence

He ordered that the


chief be dismissed

The treaty was


concluded, which
was important

He added and
smiled

Matthew sloped off


to chat with Kate

figure

He ordered the
dismissal of the
chief.

The conclusion of
the treaty was
important

He added witha
smile

Matthew sloped off


for a chat with Kate

1. Comparison with Chinese


In our descriptive account so far we have focussed almost entirely on English; but having
established a fairly comprehensive (though not very delicate) map of the resources of
English ideational semantics, we should now be able to compare and contrast with other
languages. Here we shall provide a very brief sketch of Chinese ideational semantics
against the background of our previous account of English; and in Section 7.8 below, we
shall conclude with a few remarks about typological variation across languages in general.
Comparative description of this kind is important not only for theoretical reasons but also
in relation to tasks in natural language processing such as machine translation and
multilingual text generation and understanding.

7.1 Historical background


In the perspective of world languages as a whole, Chinese and English share many
common features in their histories. Both are languages that have been associated with a
long period of settlement. In their modem form, both have been shaped by the extended
period over which their speakers were first and foremost agricultural producers, living in
villages, staying most of the time in one place, and very gradually developing on the one
hand the technology and on the other hand the forms of social organization that went with
their agricultural life style. Both languages were written down; some members of the
community learnt to make meaning in the written mode, by reading and writing, while
others did not And both languages engendered forms of verbal art and other highly valued
discourse such as religious texts, which were on the one hand codified in writing and on
the other hand also committed to memory. In all these respects Chinese and English have
much more in common, in their historical conditions, than either has with languages that
developed outside the Eurasian culture band (or with many that developed within it, such
as the languages of the Caucasus).
There are also some far-reaching differences. The two belong, obviously, to different
ends of that cultural continuum, and were located within very different material
environments. In particular, English emerged out of the westward movement from the
centre of the continent, while Chinese evolved in the migration towards the east; the trek
to the west, into and across Europe, was much more uneven, sporadic, and disruptive than
the relatively steady and homogeneous eastward movement (only relatively steady, of
course, but significantly more so than the shifts of population into Europe). More
importantly, perhaps, while Chinese was always at the centre of the eastern cultural
sphere, English at least for the time during which it has had a distinct history has
mostly been at the periphery. In that respect, English has evolved under conditions more
like those of Japanese. The populations of Britain and Japan were both regularly overrun
from outside: Britain by Celts, Romans, Anglo-Saxons and Norsemen/ Normans; Japan
by Austronesians, Altaic peoples and Chinese, each following the previous arrivals within
some five to seven centuries; their modes of production and social structures, including

298

7. Comparison with Chinese

the structures of feudalism, were evolved with frequent intrusions from outside. The time
depth was fairly similar (even the present hereditary monarchies in Japan and Britain
probably originated in the same century). Similarly, the Japanese and English languages
were first written down at roughly the same time each in a script that was imported
from another language (in the case of English, a distantly related language; in the case of
Japanese, a language that was not related at all, namely Chinese).
Both Japanese and English evolved through various perturbations and discontinuities,
and probably both were equally creolized in the process (whereas Chinese remained very
largely unmixed). Japanese borrowed its learned vocabulary from (ultimately classical)
Chinese, English from (ultimately classical) Latin; both languages of learning came in
more than one wave, and both were mediated through spoken languages descended from
these classical tongues (early forms of Northern (Mandarin) and Central (Wu) Chinese;
Norman and other early forms of French). In the modern forms of English and Japanese,
the relative frequency, the functions and the domains of the borrowed vocabulary are
remarkably similar. The effect, in both, has been to develop two co-existent phonological
systems, which are only partly homogenized, and are kept apart (more or less clearly) by
their different representation in the script (in English, the spellings of Graeco-Romance
and Anglo-Saxon words follow distinct conventions; in Japanese, words of Chinese origin
are written in kanji, those of Japanese origin in a mixture of kanji and kana). There is an
interesting reversal at this point: in English, the learned words tend to be long and
morphologically complex, the native words short and simple, whereas in Japanese it is
the other way round.
Both English and Japanese have been heavily technologized, as their countries became,
in turn, the 'workshop of the world' Britain in the nineteenth century, followed by
America after two world wars; and now Japan. Chinese technology led the world for about
two thousand years, till about 1500; its technical forms of discourse were then comparable
to those of late medieval Latin. But there was no effective development of scientific
discourse in Chinese till the beginning of the twentieth century. Now, 100 years later,
scientific Chinese is strikingly similar to scientific English, partly because it has
borrowed some of the semantic styles, through translation, and partly because scientific
registers tend to call for the same kind of organization anyway. But there are no word
borrowings in Chinese, or hardly any; unlike English and Japanese, both of which are
'borrowing' languages, Chinese creates new word meanings by 'calquing' constructing
new forms out of its own lexical stock by compounding, often on the model of some
outside source. All three are now 'Pacific Rim1 languages, with English in the United
States, Canada, Singapore, Australia and New Zealand and Chinese not only in mainland
China, Hong Kong (soon to be part of mainland China), Taiwan and Singapore, but also
wherever there are overseas Chinese communities; Japanese remains, up to now, relatively
unexported, And all three are major languages of information, likely to be the centre of

7.1 Historical background

299

information technology, from word processing to machine translation, for as far as we can
see into the future (but it would be rash to predict beyond about twenty-five years).
Following this thumbnail sketch of the historical context, we shall attempt to make a
few points suggesting how Chinese would be represented, by comparison with English,
in terms of the model we have outlined.

7.2 Some general features compared


It goes without saying that Chinese shares with English all those properties which are the
properties of human language as such. But despite the many assertions that are made
about "language uni versais", these are from being clearly defined largely because, in
our view, they have usually been sought at rather too concrete a level. We shall not
attempt to explore this issue here. In what follows we shall simply present certain
features in respect of which Chinese may be contrasted with English, in its construal of
ideational meaning; beginning with a few general observations and then proceeding to
some more specific points of comparison (cf. Halliday, 1956; 1959).
(i) As we pointed out right at the start, one property of language which is universal is
the stratal organization of the content plane into semantics and lexicogrammar, with the
lexicogrammatical stratum forming a continuum: at one pole are the most "grammatical"
features, closed systems of just two or three terms, mutually defining along a single
dimension and with very general meanings and contexts of use; at the other pole are the
most lexical features, open sets of an indefinite number of items, taxonomically arranged
along various dimensions and with highly specific meanings and contexts of use.
In most cases, the two languages locate the various semantic domains at roughly
equivalent points along the lexicogrammatical continuum. Thus, in both, polarity, person
and nominal deixis are clearly grammatical, while the vast thesaurus of experiential
categories of processes, things and qualities are clearly lexical. But there are one or two
significant disparities. For example, Chinese treats the "phase" of a process more
grammatically than English: there is in the Chinese verbal group a network of systems of
completive phase and directional phase realized by a large but closed set of "post-verbs"
(postpositive verbs), e.g.:

completive:

dun 'apart, into two pieces'

zh

break

zhdun

break in two

jin

cut (with scissors)

jindun

cut in two

qi

cut (with knife)

qidun

cut in two

kn

chop (e.g. tree)

kndun

chop down

shui

fall (of person)

shuidun

break (limb) by falling

300

7. Comparison with Chnese

directional:

chu 'exit1 (often in compound

chli 'come out1,

chuqu

'go out')

kan

look

Z.U

walk

zuch

walkout

fen

separate

fench

distinguish

gn

chase, drive

gnch

drive out

kanchli

make out by looking

On the other hand, "tense" (linear time) in Chinese is considerably less grammaticalized
than in English, being construed through definite and indefinite time adverbs like 'already1,
'soon1, 'yesterday', last year'; and there is no necessary representation of linear time in the
clause (see next point).
(ii) In English, while lexicalized meanings are optional, grammaticalized meanings are
typically obligatory in their functional environments. For example, the English nominal
group the train is entirely neutral as between 'passenger1 and 'goods1, and among 'electric1,
'steam1 and 'diesel', because these meanings are construed lexically; but it is explicitly
'singular' (and therefore 'not plural'), and it is explicitly marked as 'definite identity
known', because number and deixis are construed grammatically. Thus train does not
normally occur as a nominal group by itself; the deixis is always specified (note that the
plural trains is selecting explicitly in the deictic system).
In Chinese, on the other hand, not only lexicalized meanings but also many of those
that are construed grammatically have this characteristic of being optional. So
huch 'train' is neutral not only as regards consciousness of cargo and type of fuel but
also as regards both number and deixis. It regularly occurs as a nominal group by itself,
and may be equivalent to English 'a train', 'trains', 'the train' or 1the trains'. (Number is
obligatory in Chinese only in the personal pronouns; deixis not at all.) Similarly, aspect
in Chinese, while unlike tense it is fully grammaticalized, includes a "neutral"
(unmarked) term in which neither perfective nor imperfective is selected.
(iii) Thirdly, and related to the last point, if we look at Chinese from the point of view
of English, it appears that in its lexicogrammar as a whole there is some tendency for
avoiding unnecessary specificity. As we have seen, grammatical systems tend to have
unmarked, neutral terms; the neutral term "opts out" from choosing in the system
either because the systemic choice will be irrelevant in the given context, or because,
while it is not irrelevant, the meaning is construed elsewhere (in the text or in the
situation). Likewise lexically, at least in construing participants, it is typical for Chinese
to use a general term where the more specific one, even if relevant, is rendered unnecessary
by the context. We can illustrate both of these from a single example: imagine parent and
child standing at a bus stop, and the bus arriving in front of them. The English speaking
parent says Hurry up and get on the bus! The Chinese speaking parent says

7.2 Some general features compared

301

Ni kui shng ch bal 'you + quickly + go up + wheeled vehicle [+ mood


particle]'. English specifies both

(grammatically) the deixis that the bus is

identifiable (in this case by the situation; they may also have been talking about it)
and (lexically) the type of vehicle. Chinese specifies neither; there is no need for a deictic
(such as

nage 'that'), and no need to specify that the vehicle is a bus

gnggng qch ) as distinct from a train

huoch ) or a horse drawn cart

mache ). It would seem a little odd to bring in any further detail!


(iv) Fourthly, while in technical and other elaborated forms of Chinese the
nominalizing tendency of grammatical metaphor is every bit as prominent a feature as it
is in English, it perhaps does not penetrate as deeply into everyday discourse. To give a
simple example: in English, a typical, and entirely colloquial, answer to an enquiry about
some commodity in a shop or market would be they come in all sizes. The more likely
answer in Chinese would be

dde xiode du you 'there are both big ones

and little ones'. This is not from any lack of the relevant abstract terms such as 'size'; in
some cases Chinese is more endowed with these than English is, because Chinese
distinguishes between two different senses of such terms: between size meaning 'how
big?' (as in / need to know the size) and size meaning 'the fact that ... is big' (as in I'm
impressed by the size). But there are numerous kinds of expression like these where the
more nominalized form has not taken over in the encounters of daily life.
All these points will be illustrated in more specific contexts in the course of the next
few sections.

7.3 Sequences
Sequences, figures and elements are distinguished, and related to each other, in Chinese as
they are in English, with the same congruent realizations in the grammar:
sequence

clause complex

figure

clause

element

element of clause structure

The degree of overall indeterminacy is about the same as that in English.


In the written language, the clause complex is codified as a sentence. In spoken Chinese
(Mandarin), as in English, decisions have to be take about where clause complexes begin
and end; in both languages it is necessary to recognize some grammatical structure even
where it is not explicitly marked by conjunction. Semantically, the same range of logical
relations is involved in constructing sequences; the relationship within any one nexus
may be one of expansion or projection, and may be either equal or unequal (paratactic or
hypotactic, in the grammar). The main differences are found in the grammar itself: in
expansion, in Chinese, (i) the unmarked paratactic extending relation, 'and', is typically

302

7. Comparison with Chinese

not marked by any conjunction; (ii) in a hypotactic nexus the dependent clause almost
always precedes the one on which it depends, and (iii) the dependency relation is signalled
obligatorily in the primary clause and optionally in the dependent one, instead of the other
way round as in English. Where the dependent clause is marked, the conjunction may
occur in various places (at the beginning, following the Subject, or at the end), or even in
two and (with an effort) all three, * instead of being just confined to the beginning of the
clause. Thus the pattern of preferences in construing dependent sequences in Chinese is as
follows:

In projection, the same distinction is made between ideas and sayings, and between
direct (paratactic) and indirect (hypotactic) modes of projecting them. Since there is no
system of tense in Chinese, there is no tense sequence; but the shift of deixis takes place
in the same way as in English (cf. e.g.

mingtin 'tomorrow',

dr

tin

'the next day , as well as of course the shift of pronouns and demonstratives), and there are
minor adjustments for mood, for example between direct and projected questions.
The (paradigmatic) range of different logical relations through which figures are
construed into sequences is very much the same in Chinese as in English; and, as a
corollary to this, the (syntagmauc) extent of the sequences is very similar. It should be
remembered, of course, that our concern here is with those registers which are likely to be
encountered in information processing automatic documentation and retrieval, text
generation and parsing, machine translation and the like. Not all of these are technical
registers; but they are all to some extent multinational, and so permeable to semantic
influences from one to another. In these registers, in translating between Chinese and

jir ta rugu
One of the authors noted, in conversation, the clause
hot bu znchng de hua 'If he still doesn't agree', in which the meaning 'if is expressed three
times over (
de hua}

7.3 Sequences

303

English it is possible to match the sequences syntagmatically one to one: what works as a
construction of figures in the one language will also work in the other.
As regards the individual clause nexus, two points arise from the fact that in Chinese
the grammar constructs dependency relations by marking the primary rather than the
dependent clause. One is that there is more indeterminacy between parataxis and hypotaxis
than in English (unless one construes as hypotactic only those instances where the
dependent clause is marked as such, which is contrary to one's sense of their
proportionalities as agnate sets).^ The other is that some English figures have to be
reconstructed in order to go into Chinese, particularly the notorious 'unless1 and 'until': the
Chinese parallel to 'a not until/ unless jc ' is '(if/ when) x only then a '. There is thus a
subsystem of relators
English:

jia 'then1/

ci 'only then' which has no direct parallel in

rii daylng wo ji gos r Til tell you if you promise',


rii daylng w ci gos r 'I won't tell you unless you promise'.

7.4 Figures
Figures are also clearly comparable between the two languages. Like that of English, the
figure in Chinese is typically a configuration of a process with one, two or three
participants (even meteorological processes require one: 'rain' is construed as a thing,
which 'falls') and optional circumstantial elements. The main differences in the way
figures are constructed are the following,
(1) The overall construction of process types is the same as in English: there arc
processes of doing, sensing, saying, and being, accommodating the same participant types
and participant roles. In processes of doing, the transitivity potential is also very similar,
with two-participant processes typically constructed ergatively:
wo ki chzi 'the car drives : I drive the car'. Sensings however
are not normally construed as action by some phenomenon on the conscious being; there
is no regular configuration of the type 'it pleases me', 'it reminds me1. Analogous patterns
are construed analytically with a semantic feature of causative:
shi wo gaoxing 'makes me happy, pleases me'
si w xingdo 'makes me think of, reminds me'

That is, if a nexus of two clauses neither of which is marked for dependency is always taken
as paratactic. But one of the strong arguments for recognizing the system of 'taxis'
(hypotaxis/ parataxis) in Chinese is precisely that a nexus with both clauses unmarked in
most likely to be interpreted as b A a (cf. the diagram above). Thus x
goo, y di ( f x high,
y low') could mean either (paratactic) 'x is high and y is low' or (hypotactic) 'when x is high,
y is low1. (The third interpretation, 'x is high when y is low', is almost though not quite
impossible.)

304

7. Comparison with Chinese

(2) As in English, the process is the category of experience that is located in time. In
English, time is construed grammatically as tense: as a flow, with a more or less extended
'present' forming a moving but impermeable barrier between 'past1 and 'future'; and each
instance of a process is located somewhere in the flow. In Chinese, time is construed
grammatically

as aspect: and specifically as an opposition of unfolding versus

culminating; and each instance of a process may

be given a value in this opposition,


1

either 'significant as unfolding in its own right or as 'significant as culminating


perhaps by virtue of its consequences'. For example:

(i)

wo knzhe bo T (am, was) reading the newspaper' (imperfective)

(ii-a)

wo knle bo 'when I (have, had) read the newspaper' (perfective 1)

(ii.b)

wo kn bo le T (have, had) read the newspaper' (perfective 2)

There is a further difference, however, in that the system of aspect accommodates an


unmarked option:

(ui)

wo kn bo 'I + read-f newspaper' (unmarked)

So whereas in English each process in a figure (provided the figure is arguable


grammatically, one where the clause is finite) must be located somewhere in the
construction of tense, in Chinese the process may simply be left as neutral, without being
assigned to either category of aspect.
It might be suggested that the last type, though formally unmarked, actually represented
a third marked term in the system as happens in English where formally unmarked
terms of positive and singular are just as marked semantically as the formally marked
alternatives (thus singular excludes plural as categorically as plural excludes singular).
This is certainly true of some categories in Chinese, e.g. positive polarity. But with
aspect this is not the case. Neutral aspect, as in

wo kn bo, does not exclude the

meanings of the other terms; it is simply not selecting in the perfective/ imperfective
system.
(3) Qualities in Chinese include within themselves the 'be' (the relation of
ascription); hence ascriptive figures (e.g. 'I am busy1) are construed as two elements ('I +
be busy'

wd mang ). Syntactically therefore they belong to the class of verbal

elements rather than the class of nominal elements i.e. in Chinese "adjectives" are
verbs, whereas in English they are a kind of noun. As in English, ascriptive figures are
located in time (but by aspect, of course; e.g.
1

png le 'culminate in being fat, get

fat ). Qualities can be used to construct participants (e.g.

png rn 'a fat man'), but,

relative to English, this is much less favoured. Where English prefers to introduce
qualities in a nominal group, as Epithet + Thing fat man, Chinese prefers to introduce
qualities in a clause, as Carrier + Attribute 'man is-fat'. Thus while an English speaker

7.4 Figures

305

will say That's a very fat man, a Chinese speaker will say IP^AIff Nage rn hnpng
'that man (is) very fat'. If the quality is to be construed in a nominal group in Chinese,
there is a tendency for it first to be constructed into a figure and then deconstructed again
hn png de rn 'a man who is very fat'. In a simple
in a formofrankshift:
sequence of quality + thing the quality is often a means of assigning to a class: e.g.
man ch 'slow (= stopping) train*.
Figures such as 'she has long hair', 'I have a headache1 are construed ascriptively
wo tau tng 'I +
(predicad vely):
ta tufa chng 'she + hair + be long',
head -f be painful1. Note that for both Chinese and English the problem with such
constructions is to make the person rather than the part of the body thematic (which in
both languages means making the person the first element in a clausal figure). English
achieves this by constructing the quality 'into' the participant and ascribing this to the
person by possession I + have + a headache/ a sore throat'; Chinese does it by detaching
the person as a syntactic 'absolute' located at the beginning of the clause and constructing
the quality predicatively '[as for] me -f head + aches/ throat -f is sore'. The English
construction is mildly metaphorical. (On Theme in Chinese, see Fang, McDonald &
Cheng, 1995; on the grammar of pain, see Halliday, in press c.)
(4) The accommodation of elements in a figure is very similar to that of English;
as mentioned above, processes may have one, two or three participants, and the
distribution is closely parallel. As in English, 'indirect1 elements are introduced
circumstantially, with a 'mini-process' locating them with respect to the main process. In
Chinese, such circumstantials retain more of a 'process1 flavour (see below, however, for
their internal construction); the Chinese equivalents of English prepositions are clearly
verbs and can construct figures on their own (e.g.
do 'reach; to':
ta do
Beijing qii 'he's going to Peking1;

ta dole Beijing '(when) he has reached

Peking'). Thus the interpolation of such circumstances as downranked figures is relatively


more foregrounded in Chinese.

7.5 Elements
But it is in the construction of elements that Chinese and English differ most although
even here the similarities

are more striking than the differences. There is no

morphological variation or marking in Chinese; but the primary categories of participant


and process are clearly construed in the forms of word classes nouns and verbs by
the syntax, both the syntax of the group and the syntax of the clause. The nominal and
verbal groups are again rather similar to those in English.
7.5.1 Nominal

groups

The nominal group is constructed out of much the same materials, arranged more or
less in the same order, but with two significant differences. (1) A thing, with or without

306

7. Comparison with Chinese

associated quality, may stand by itself:

sh

'tree, the tree(s)1,

po

'ftir

jacket(s)'; if however it is either identified (made definite) (other than by possession) or


quantified (or both), it typically also gets assigned to a measure-type, either collective,
partitive, quantitative or individuative, e.g.
(a)

son bei cha three cups of tea


nkuicodi that (piece of) lawn

(b)

zhe liangbao ylfu these two parcels of clothing

(c)
(4)

yl su fangzi

one (individuated) house

vi k sh one (individuated) tree


The last has no equivalent in English; these are the unit- or measure-words of Chinese,
which are sometimes called 'noun classifiers' because they group things into classes
according to a mixture of criteria based on shape (long, fiat &c.), natural class (birds and
mammals, trees <fcc.) and functional domain (books, letters, living quarters &c.). Now,
unlike English, where the grammar makes a categorial distinction between objects
(bounded) and substances (unbounded), in Chinese 'things' are not inherently construed
into these two types; that is to say, there is no clear opposition of count/ mass in the
grammar. However, this distinction is in fact construed cryptotypically (in Mandarin,
though not in Cantonese), because corresponding to type (d), the individuating measure
words (and corresponding only to this set), is an unmarked term

ge (the original of the

pidgin piece, piecee) which in certain contexts replaces the specific individuating term.
Hence the 'entity1 is anything that can be measured as

g and the set of things that are

defined in this way is similar to the set of things which are construed as count in English.
(2) In the syntactic construction of the Chinese nominal group the Thing comes last;
hence while English distributes qualities on either side of the Thing (simple ones before,
macro-qualities after), Chinese lines them all up in front. This has a minor effect,
perhaps, in making a sharper distinction between two kinds of quality roughly, simple
versus macro in English than in Chinese; but it becomes significant in the Chinese
construction of complex taxonomies, and we shall return to it below.
7.5.2 Verbal groups
The verbal group is also not unlike that of English, with a lexical verb and various
closed classes of 'auxiliaries'; there are no tense forms, of course, but there are modal
forms forming a system of modality, codified grammatically although not to the same
degree as in English. But there is one major difference in the semantic construction of
processes.

7.5 Elements

307

Given that a process takes place in time, there will be some sense in which it has a
beginning, a middle and an end. These may be observable as distinct constituents, if it is a
process having duration; even if it is instantaneous, however, they represent possible
facets, points of view from which it can be considered. In English, the meaning of a
process typically includes its completion: if I 'cut' a piece of string, I cut it in two.
To construe a process as non-completed, English uses conative or inceptive phase: 'try to
cut it', 'start to cut it1. In Chinese the meaning of a process does not imply

its

completion; if it is to be construed as completed this is made explicit by a second verb


marking completive phase, either resultative or directional. Thus 'cut

jin, 'cut in two'

laidao 'come so as to arrive, reach1. In English no

jindun; 3f loi 'corne',

such systematic distinction is made; there are a few processes that require completion,
such as 'seek', but there the distinction is lexicalized, look for/ find. In Chinese it is
codified grammatically; there is a small set of seven directionals and a larger set of 60 - 80
rsultat!ves, some of which are specialized to just one or two processes. This system is of
course related to the temporal category of aspect: a process + completive is typically
culminative in aspect. See Figure 7-1.

Fig. 7-1: Process as 'phased' in English and Chinese


It is difficult to exemplify cryptogrammatic features of this kind because they depend on
myriads of small semantic encounters none of which by itself is likely to be particularly
striking. And it is impossible to investigate them experimentally, because conscious
reflection distorts the semantic process: what speakers say they would say is very different
from what they really do say. But we could contrast the following two wordings, both
taken from spontaneous dialogue:
I tried to phone Sam but didn't get him; he was out.

W5

gngcai

gi

just now

to

ta

da

dinhu,

dbtng

him

call

phone

call-not-through

(I tried to telephone him just now but couldn't get through.)

308

7. Comparison with Chinese

It would be possible in English to say I phoned Sam but didn't get him, but it is less
likely and on the occasion recorded above, the speaker had actually started to say that
but then corrected herself: I phoned / tried to phone Sam, ....
unlikely that anyone would say I tried to phone him'

In Chinese it is very
wo shlshl gi

t5 da dinhu would mean something like 'I tried out the experiment of phoning him1).
Thus whereas in English the inceptive/ conative contrasts with an unmarked phase which
is (by implication) completive, in Chinese the completive contrasts with an unmarked
phase which is (by implication) inceptive/ conative.
There is the further consequence that the negation of a process in Chinese may be
construed in four different ways:
aspect

phase

bjin

neutral

neutral

not -t- cut

mi jin

culminative

neutral

not yet + cut

jin bu. dun

neutral

completive

cannot + cut (through)

mijindun

culminative

completive

not yet + succeed + cut

This contrasts with the English construction of polarity in association with primary and
secondary tense.

7.5.3 Prepositional phrases


We said earlier that circumstantial elements were constructed in Chinese, as in English,
by associating a participant with an ancillary or minor process -- one that is not however
available for constructing an entire figure, since it accepts only the one participant. In
English, this type of circumstance is the prepositional phrase, and the class of
prepositions is largely distinct from that of lexical verbs, although some prepositions are
clearly non-finite forms of verbs in origin (e.g. concerning, during). In Chinese these
'minor processes' are related more explicitly to figures, since all words of the class of "preverb" (prepositive verb, analogous to prepositions) are words which function also as full
processes. We should add one further point about these circumstantial elements: that the
relation by which they are 'attached' to the figure as a whole is often construed not just by
the minor process but by the minor process plus a generalized ('thing' that is a) facet of
the participant in question. The model for these is of course a locative, e.g.:
zi

huyuan

at

garden

'in the garden'


inside

but other, more abstract locations (including temporals) are construed in the same way:
zi

bozhi

shng

'in the newspaper'

7.5 Elements
at newspaper
zi zh zhng

309
top
qngxng zhxi

under these circumstances'

at this kind (of) circumstances bottom


Thus the relation constituting the minor process is at is were deconstructed into a
'relation1 (typically 'at', 'to', 'from', 'around', 'along') and a 'facet' (back, front, top, bottom,
left, right, inside, outside, before, after, middle).
These 'facet* nouns form an exception to the principle that the lexicalized 'thing* is the
final component of the nominal group, since they follow it in every case. On the other
hand they can in turn be construed as Head, by having the 'thing' made to modify them
shgui de qinmian 'in front of the bookcase1. They
with a f de 'of:
thus have the same kind of intermediate status that nouns like 'front' (cf. in front of vs.
in/from the front of) have in English. There is a sense in which, in Chinese
circumstantial 'pre-verb' (relation) : full verb
::

circumstantial 'post-noun' (facet) : full noun


both as it were representing the relational rather than the substantive aspect of
processes and things. In English the proportion is lacking because prepositions are less
clearly related to verbs.
We can now return to the construction of processes and of things, as the major classes
of elements. We have already seen that the nominal and verbal groups in Chinese are in
significant respects similar to those of English. We now need to note an important
difference between them in the way they construe experience.

7.6 Processes and things


We saw that, in English, things were much more richly taxonomized than processes. One
of the principles behind grammatical metaphor was that, by reconstituting processes as
things, we are able to construct them systematically into classes and hence, ultimately,
into the technical abstractions of a scientific theory. Processes are much less constructible
into taxonomies; they lack the necessary stability through time.
This distinction appears considerably heightened in Chinese, in various respects. (1)
Processes are construed as very specific: we may note, for example, twelve verbs
corresponding to 'cut1, twenty or more corresponding to 'carry', and so on:

310

7. Comparison with Chinese

qi

eut

duo

cut (meat), chop up

jin

cut with scissors

&
cM

cut (grain), mow


cut (cloth)

xio

cut (skin off fruit), pare

xi

cut (nails, small branches), trim, prune

li

cut (logs), split

P*
kan

cut (firewood chips), chop

chai

cut (paper), slit

cut (skin, flesh), slash, gash

gu

cut (flesh), nick

ji

cut off, sever

na

take, bring, fetch, hold, carry

bo

in both arms, against body

cut (tree), chop

hui

in the crook of one arm

png

in cupped hands

tio

on pole across both shoulders

dan

on pole across one shoulder

tai

supported from underneath

by handle

kng

on one person's shoulder

gong
duan

on two people's shoulders


in hands with arms outstretched

ji

underarm

han

inside mouth

dido

protruding from mouth

ding

on head

bi

on back

di

(take) along with

qi are the most general terms, typically used in abstract and metaphoric
senses; but they are not strictly superordinates and in concrete contexts the preferred
choice is the specific one. This contrasts not merely with English (which prefers general
verbs) but also, more importantly, with the way things are construed in Chinese. While
processes are construed as specifically as possible, things are construed as generally as
possible. The way this works is as follows.

7,6 Processes and things

311

Things are typically construed in sets consisting of a superordinate plus a group of


hyponyms, the relationship being explicitly signalled in the structure of the words
themselves: the hyponym is a compound, formed of the superordinate preceded by another
lexical element. For example,

writing implement

qinbi

pencil (lead )

mob

brush (hair )

moshuWi

pen (ink-water )

&c

che

wheeled vehicle

dinch

tram (electric )

huch

train (fire )

zxingche

bicycle (self-go )

mch

horse cart (horse )

qch

car (automobile) (gas )

&c

in

fl.

nqi 'machine'

dz.ij

typewriter (strike-character )

jlsunjl

computer (calculate )

fiij

aeroplane (fly )

zhbj

loom (weave-cloth )

gcoj

lawnmower (cut-grass )

xyi

washing machine (wash-clothes )

ynshuj

printing press (print )

zhoxingj

camera (photograph )

&c
Thus the taxonomic organization is made fully explicit in the naming: a
is a kind of

b, a

huch is a kind of

qinbi

che, and so on. This happens

sporadically in English, for example with (some of) the names of fishes, trees and birds;
but it is by no means regular and it can be tricksy: a shoetree, for example, is not a
kind of tree (the word shoetree is not a hyponym of tree), nor is a foxglove a kind of
glove, or a saucebocu a kind of boat. In Chinese there are very few such metaphorical
compounds; moreover, large parts of the everyday noun stock are constructed in this way,

312

7. Comparison with Chinese

in the form of explicit taxonomies. These can sometimes be extended to three or even
more levels.
Furthermore, in contrast to processes, where the principle is 'be as specific as you can1,
in referring to things the principle is 'be only as specific as you need'. So when asking
your friend to pass you your pencil, which both of you can see in front of you, you do
not specify what kind of a
b
(writing implement) it is; you say
wo nag bi qng n dlgulai 'please pass my
more likely just

b\ across' or

bi. Thus, in discourse, things are construed at the most general level

that their naming and the context permits.


What this means is that the dichotomy of experience into processes and things is rather
more explicitly semanticized in Chinese than in English. While in both languages the
same lexeme may often be either noun or verb, the two are clearly distinct in the syntax;
in Chinese, in addition, there are some systematic differences in the semogenic principles
by which processes and things are constructed. One interesting consequence of this is that
while processes tend to be differently named from one dialect to the next, things
usually have the same names throughout the whole of China (with different
pronunciations, of course; but the same lexemes). More important for our present
purposes, however, is the significance for grammatical metaphor.

7.7 Grammatical metaphor


We will limit these brief comments to grammatical metaphor in its more scientific and
technical aspects. As in English, so in Chinese technical terms are typically.nouns; and
we have seen how the typical structures of the everyday nominal vocabulary lend
themselves to the creation of technical taxonomies. Sets of abstract and theoretical terms
are readily constructed on this model, using entirely Chinese lexical resources; and the
taxonomic relationships among them are fully transparent. Compare the Chinese and
English terms in a set such as the following:

lu

rate
piril ('repeat')

frequency

slii ('fast')

speed

ylnl ('sound')

musical pitch

hxl (breathe)

respiration rate

zhduzhuanl ('turn-over ')

turnover

zhshu ('refract')

index of refraction

7.7 Grammatical metaphor

313

Thus the semiotic transition from commonsense knowledge and experience on the one
hand to educational experience and the technical knowledge of the disciplines on the other
may appear somewhat less abrupt in Chinese than in English.
Many such creations are, of course, metaphorical in the sense in which we used the
term with respect to English. We have not undertaken a systematic survey of types of
grammatical metaphor in Chinese; but from a study of scientific writing in various fields
we have the impression that the range of types of metaphor is not very different (Halliday
& Martin, 1993: Chapter 7). The prevailing tendency is to reconstrue other phenomena as
things, which means as in English that whole figures are reconstrued as qualities of
those things.
Again as in English, the nominal group in Chinese stretches to accommodate this kind
of reconstruction. But its resources are somewhat different. On the one hand, as we have
seen, the Head of the nominal group always comes at the end, an instance of the very
dominant principle in Chinese whereby modification is regressive all modifiers precede
what they modify. So everything that is construed as a quality has to precede the thing to
which it is assigned.
This can lead to considerable syntactic ambiguity rather more, in fact, than in
English, because in English such meanings can be construed as prepositional phrases and
clauses coming after the Head. In Chinese, too, both circumstances and figures can
become qualities of a thing, represented as phrases and clauses in the nominal group; but
these always come before the Head, along with the 'ordinary' qualities that are construed
as single words. So for example in the following clause everything up to ffirife dun kui
'fragment' forms a single nominal group, in which $r^ dun kui is the Head noun.

you

duncng

min

qngxi

xingxing

de

ling tio

from

fault

plane

inclined

to wards (face to face)

**

two

zhng

duncng

zchng,

ling

duncng

zhngjin

xingdui

straight

fault

formed

two

faults

between

relatively

xinlu

de

dunkui

[jio

diqin

hu

dunxin pndi]

depressed

**

fragment

[is called

"graben"

or

"depression trough" ]

314

7. Comparison with Chinese


** Note: The structure marker

signals that what precedes it modifies (is

de

dependent on) what follows.


We have referred above to how information is lost, in English, when figures are
metaphorically construed as single elements: between, say, how many people die because
their lungs develop cancer and lung cancer death rates. In Chinese there is even greater
loss of information, for two reasons: first, because more items (words, phrases and
clauses) are strung together (since all precede the Head), and so the number of possible
bracketings is higher; and secondly, because most of the semantic relations among downranked elements within the nominal group are left implicit in Chinese, whereas in
English at least some are preserved. We can compare the above Chinese example with its
English translation:
A relatively lowered
formed

by

two

fragment

straight

between

faults

whose

two faults,
planes

are

inclined towards each other, [is called a "graben" or


"depression trough"]
The reader has to supply rather more of the linking structure in the Chinese than in a
typical English equivalent.
On the whole, however, the scope of grammatical metaphor in technical and other
formal written varieties of modern Chinese is about the same as it is in English. As far as
its effect on the construction of meaning is concerned, grammatical metaphor in the form
of nominalization (turning processes and qualities into things) engenders considerable
syntactic complexity and ambiguity, perhaps more in Chinese than in English; at the
same time, the construction of technical vocabulary and technical taxonomies is rather
more accessible in Chinese than it is in English. Overall, however, as far as technical
discourse is concerned the two languages are not excessively different in their creation of
meanings; and this has implications for any kind of shared information processing, such
as multilingual text generation, in which both English and Chinese are involved.

7*8 The meaning base: concluding remarks


Chinese and English share many common features in the construction of the ideation
base. Thus, if we consider how they construe happenings as 'quanta of experience1,
the two languages are to a large extent congruent: Chinese sequences tend to correspond to
English sequences, and Chinese figures to English figures. At the same time, both
languages allow a comparable degree of elasticity at this point in the system, whereby one
and the same happening in the phenomenal world may be construed either as a sequence or
as a figure or as something indeterminate between the two. In English, we may say
either she mended it II by using string or she mended it with string,

construing the

means either as a full process in a separate figure or as a minor one in the form of a

7.8 The meaning base: concluding remarks

315

circumstance. In Chinese, on the other hand, these two overlap, so that


ta yong shngzi lai xitt could be interpreted either way.
One major source of this kind of elasticity is the manifestation of the very general
semantic types of projection & expansion throughout the ideation base. In English, as we
have shown, we find them manifested both logically as relations and experientially as
elements within figures:
[i] logical manifestation:
as relations between figures in the construal of sequences (realized by clause
complexes);
as relations between processes in the construal of figures (realized by verbal
group complexes);
[ii] experiential manifestation:
as circumstances within figures (realized by prepositional phrases or adverbial
groups);
as circumstantial processes within figures of being & having (realized by
relational clauses);
as non-nuclear participants within figures, i.e., participants other than Medium
Agent, Beneficiary, Range (realized by nominal groups with or without a
preposition).
This elasticity of construal is further increased through grammatical metaphor, which
also relies on the manifestation of projection & expansion within participants (realized by
the different types of modification in nominal groups). We have illustrated the range of
options throughout the system at various points, and a schematic example will be enough
here:
[i] logical
A happened, so B happened;
A caused B to happen;
[ii] experiential
A happened causing B/ B happened because of A;
A happening caused B happening
A affected B ['cause-happened1]
with additional metaphorical variants:

316

7. Comparison with Chinese


B happened because of the happening of A
the happening of A caused the happening of B
the happening of A was the cause of the happening of B

As we have shown in the context of grammatical metaphor, the choice among


alternative construals is made on the basis of both ideational and textual factors. These
factors 'conspire1 together so that different strategies are favoured in different registers: the
congruent form (sequences) in casual speech, the metaphoric form (figures of being &
having) in elaborated forms of writing.
Where there is variation of this kind within one language, we may expect to find
typological variation across different languages. Consider for example the construal of a
speech event. Here English favours a single figure with a configuration of Sayer +
Process + Receiver: A said to B. Chinese favours an intermediate, borderline type:
B shuo
[A facing B said], (Cantonese) A wa bi
(Mandarin) A dul B
shuo
B tng
!/f [A said giving B to hear]. In Trique, a language spoken in Mexico, a speech event is
construed as a sequence of two figures (Longacre, 1985: 262-3), as in Gatah Juan
guni Maria, which we can gloss adhocly as Process ['said'] + Sayer

Process

[heard'] + Senser. Akan, a language spoken in West Africa, is in a loose sense


intermediate between the English model and the Trique model: a speech event is construed
as one figure, but to bring in a 'receiver', the Process is construed as a sequence of
processes (realized by a 'serial-verb construction' within a simple clause), as in ka
kyert 'tell: say 4 show' (e.g., ofaz aszm fcyerce Kofi. 'Process a: he: say + Verbiage:
something + Process (i: show 4- Receiver: Kofi', i.e. 'he told Kofi something'). In general,
Akan tends to construe the equivalent of English beneficiaries on this model; and it also
construes causal, temporal, and spatial enhancements along these lines by means of
sequences of processes.
Such examples illustrate the kind of typological variation in the construal of quanta of
experience that we suggested above, but they are only fragments: we need more
comparative studies based on comprehensive accounts of particular languages. Pawley
(1987) notes the general lack of typological studies in this area outside of lexis, and
contributes an important investigation of differences between Kalam, a language of Papua
New Guinea, and English in construing experience of change ("encoding events", in
Pawley's terms). Kalam and English may, he suggests, "indicate the outer limits of
variation among languages in resources and conventions encoding event-like phenomena";
"Kalam may be as different from English as any language on earth" (p. 335).
The Kalam clause usually construes what Pawley calls a "simple event" (a "conceptual
event which comprises a single action, unrepeated", typically "within the space of a few
seconds or less"), represented minimally as a Process. This Process is realized by a single

317

7.8 The meaning base: concluding remarks

inflected verb (showing "subject" reference and various other categories such as
tense/aspect and mood) or by a series of verbs "up to five or six bare verbs plus one
inflected verb". The class of verb is different from that of Chinese or English in one
critical respect: the set of verb stems is closed, with about 90 members, and out of these
fewer than 30 make up 90% of all instances in text.^ These few verb stems all have "very
broad or abstract meanings", which are naturally fairly ineffable in English; examples
include a- 'control, constrain, get, hold, touch ...', g- 'do, act, make, work, occur, happen,
...', md- 'exist, be alive, dwell, stay, remain, ...', mrj 'perceive, sense, be aware*. More
specific meanings are achieved not by an increase in experiential delicacy in the taxonomy
of processes realized by verbs, but by complexing verbs or clauses.
Both Kalam and English can construe the following roles within one clause: Process
(realized as verb), "actor/ agent, patient", "time and beneficiary" (p. 353). However,
English can expand the clause nucleus considerably further. For example, the happening
construed as one figure in English realized by the simple clause The man threw a stick
over the fence into the garden (Actor: 'the man1 + Process: 'threw' + Goal: 'a stick1 +
Location: 'over the fence into the garden') would be construed as three figures realized by
three clauses in Kalam (p. 354; our adhoc functional interpretation):

mon- d

yokek

waty

at

amb

wog-mgan

yowp

he-

fence

above

it-went

garden-inside it-fell

Process

Location

day
man

stick

hold

displaceddifferent
subject
Actor

Goal Process

Location

Process

In English, the path of motion could also be construed as a sequence (realized by a clause
complex), as in The man threw the stick, so it went over the fence and then fell into the
garden.

This is not only ideationally different but also textually, allowing for an

unmarked mapping of the ideational meanings onto three information units instead of
only one.
Pawley identifies the following enhancing relations as points of contrast between
English and Kalam. English may construe them within a single clause; Kalam construes
them by means of separate clauses: instrument, direction, location/ source, cause,
beneficiary (pp. 354-5). Pawley summarizes the differences as follows (p. 356):

Having a fairly small stock of verbs is a feature of "a good number of Papuan languages"
according to Foley (1986), who discusses "verbal semantics" in Kalam and other Papuan
languages.

318

7. Comparison with Chinese


... clauses do rather different jobs in the two languages. English clause structure is a

syntactic Procrustean Bed into which a wide range of diverse conceptual structures are
squeezed. For example, unlike Kalam, English allows several conceptual situations/
events to be fused into a single clause. This result is largely achieved by reducing
certain situations/ events to the status of peripheral or backgrounded elements in the
clause, expressed as arguments of the verb. [...] ... verbs do rather different jobs in the
two languages. In the case of English, it is convenient to speak of a division of labor
between full verbs, which do lexical-referential work, and grammatical functors
(auxiliaries, prepositions, etc.) which do grammatical work. In Kalam verbs do both
kinds of work, but always as full verbs. But Kalam is more restrictive than English in
the amount of information it allows to be compressed into a verb stem, and the kinds
of case relations which may be associated with a single verb.
It is perhaps possible to interpret the kind of variation across languages we have just
discussed with reference to Trique, Akan, and Kalam as variation in the division of labour
between the two modes of the ideational metafunction, the experiential and the logical. In
construing some flux of experience, Chinese and English have considerable experiential
resources for construing figures as configurations of a nucleus (Process + Medium), plus
up to 2 additional participants, and a range of different circumstantial elements; with large
stocks of verbs organized into fairly extensive taxonomies. In writing, Chinese and
English experientialize experience even further through grammatical metaphor (see Figure
6-3 in Chapter 6, Section 6.7.1). Other languages, such as Kalam, may foreground the
logical mode of construal over the experiential mode, with considerable logical resources
for construing sequences of processes and sequences of figures.
In general, wherever there is indeterminacy within a language, we may expect to find
this reflected in typological variation. Let us briefly cite three further examples. (1) We
saw that in the experiential grammar of the English clause there was a complementarity of
perspective between the transitive and the ergative: processes may be construed either as
'one participant is doing something, which may or may not extend to another participant
that it is being done to1, or as 'one participant is involved in something, which may or
may not be brought about by another participant that is the agent of it'. Probably all
languages display this transitive/ ergative complementarity in their transitivity systems;
but at the same time it appears at different depths and in different proportions. (2)
Secondly, we referred to projection as something that overlaps the 'boundary' between
interpersonal and ideational metafunctional space; in English it is typically construed
ideationally, though with a close relationship to the interpersonal systems of modality and
mood. Other languages locate projection rather differently in relation to this boundary,
sometimes foregrounding its interpersonal aspects, for example through a special category
of 'reporting' mood. (3) Some process types tend to lie on the borderline between major
categories, forming mixed and overlapping categories; typical of these are the behavioural
and existential processes in English. It is likely that equivalent types of process will be

7.8 The meaning base: concluding remarks

319

liable to greater typological variation than those that fall squarely within the core
categories of material, mental, and relational. (See Chapter 13, Section 13.4 below for a
note on the kinds of indeterminacy in language.)
We are not attempting to take up such typological issues here. But we do want to stress
the extent of the variation that is possible in the construal of experience among different
human groups; and it is helpful to consider languages that are culturally very disparate,
and remote from well-researched languages like Chinese and English. There can be
striking variation also in the semantic construction of elements. In the Reefs - Santa Cruz
family of languages (yiwo; Nanggo, Londa, South-western Santa Cruz) in the southwestern Pacific, there is a highly complex classification of phenomena involving
individual lexical items, often with very specialized semantic ranges, which combine (i)
with each other, and (ii) with verbal or nominal affixes, including three cross-cutting noun
classes and a large inventory of verbal categories, to produce complex words whose
meaning is often (from our point of view) entirely opaque and unpredictable (see Wurm,
1987). The following example of derivatives from a noun base illustrates this point:
*-modyi indicates the concept of the right hand and strength. It also occurs only
with noun class prefixes, such as lo- acquisition through labour and effort class
prefix, resulting in lo-modyi

"small adze" (something used for acquiring

something else through labour and effort by one's strong right hand"); moextending far class prefix, resulting in mo-modyi

"small outrigger canoe"

("something that goes afar when using one's strong right hand for paddling"); ovamangrove class, resulting in oya-modyi

"a type of mangrove with very strong

wood". (Wurm, 1987: 445-6)


The phenomena of human experience are held in tension by so many intersecting
analogical lines that, while all of us have the same brains and live on the surface of the
same planet, such diverse ways of semiotic mapping are not only possible but inevitable.

This brief excursion into Chinese and other languages concludes our survey and
discussion of the meaning base. In this part, we have only hinted at the question of what
role the meaning base plays in text processing. We shall now explore this issue further in
Pan in. Then we shall turn to another issue we have not discussed in this part, viz. how
our approach resembles and differs from other approaches, both theoretically and
descriptively. This will be the topic of Part FV.

Part III:
The meaning base as a resource
in language processing systems
We shall now explore how the meaning base we sketched above in Part II can serve as
a resource in language processing systems, in particular in text generation systems. We
shall start with the question of how specific 'domains1 can be modelled, given the general
meaning base outlined above. We shall then move on to the role of the meaning base in
the global architecture of a text generation system.

8. Building an ideation base


8.1 General
In any given language processing task, there will be one or more domains to be modelled
as part of our general taxonomy. To illustrate how the taxonomy is used, we have
selected the following two specific domains:
(a) meteorological: the ideational meanings of weather reports (Section 8.2)
(b) culinary: the ideational meanings of recipes (Section 8.3)
First we need to say a few words about extending the meaning base to cover such
specific domains. There are two central points (i) the relationship between context and
language, and (ii) the relationship between the general model presented in Part II and
specific domains such as the meteorological and culinary ones.
8.1.1 Field and ideation base

Just as the semantic system is functionally diversified (into the ideational, interpersonal
and textual metafunctions), so the context in which language is 'embedded* is also
diversified. The context encompasses both the field of activity and subject matter with
which the text is concerned ('what's going on, and what is it about?') and the tenor of the
relationship between the interactants, between speaker and listener, in terms of social
roles in general and those created through language in particular ('who are taking pan?').
The field is thus the culturally recognized repertoires of social practices and concerns, and
the tenor the culturally recognized repertoires of role relationships and interactive patterns.
Now, both these contextual variables are, in some sense, independent of language, even

321

8.1 General

though they are constituted in language and the other semiotic systems of a culture. That
is, they concern realities that exist alongside the reality created by language itself,
semiotic reality. However, there is a third contextual variable that is specifically
concerned with the part language is playing in any given context the symbolic mode,
how the linguistic resources are deployed. This covers both the medium (spoken, written,
and various subtypes such as written in order to be spoken) and the rhetorical function
persuasive, didactic, informative, etc..
Together, field, tenor and mode define the 'ecological matrix* hi which particular types
of text are processed: there is a systematic relationship between such matrices (particular
combinations of field, tenor and mode values) and particular types of text. We can see this
clearly with actual instances of text for example, an individual recipe or weather
forecast; but these aie instances of general classes, to be characterized in terms of the
systemic potential that is instantiated in them. That is, there is a correlation not only
between a contextual matrix and a given instance of a recipe but also between that matrix
and the linguistic potential that is deployed in recipes in general. This latter correlation is
known as a functional variety or register of the general systemic potential. This much
has been known for a long time the notion of functional dialect was worked out by the
Prague School in the 1930s (see e.g. Havranek, 1932, and Vachek, 1964), and systemic
register theory has its roots in Firth's (e.g., 1957) work on restricted languages. The
concept of register is also recognized in computational linguistics under the heading of
sublanguage,

particularly in work on machine translation (see e.g. Kittredge &

Lehrberger, 1982; Kittredge, 1987). However, we can take one further step, as systemic
theory did in the 1970s (see Halliday, 1978a), and recognize that the co-variation between
context and language is not imdifferentiated it is differentiated according to the
functional diversification of each of the two strata: the variables of context correlate
respectively with the metafunctions of the language, field with ideational, tenor with
interpersonal, and mode with textual. These pairs are mutually predictive. Thus of the
three contextual variables it is field that is implicated in variation within our ideational
meaning base.
From this point of view, field can be characterized as the deployment and organization
of the ideation base. As we have already implied, there are two aspects to this category. In
most contexts, there is both a first order field and a second order field the first crier
field is the social activity being pursued (e.g., instructing somebody in how to prepare a
dish, predicting tomorrow's weather, informing somebody about yellow-pages
information over the phone) and the second order field is the 'subject matter1 the activity is
concerned with (e.g., the ingredients and methods of cooking, meteorology, construction
businesses, international travel). So for instance, in a context where a telephone operator
provides a caller with information there is (i) the social activity of exchanging
information as a service and (ii) the area of information, e.g. copying & printing services.

322

8. Building an ideation base

And both these guide the way the ideation base is deployed Fields vary in both these
respects; and a full account of field would include a typology of the possible first and
second order values that occur in a culture. Such a typology would show how closely
various fields are related how they form families. So given three different fields, 1, 2,
and 3, the typology will show three different ways of deploying the ideation base. Each
field is projected onto one variety of the ideation base: that is, it can be thought of as
activating some portion of the total semantic resources see Figure 8-1. This
projection of field onto the ideation base involves both the particular domain and the
general types under which this domain is classified.

Fig. 8-1: Types of field within context projected as organization onto the ideation base
A given field is thus constituted as a principled selection of types from within the
ideation base; this is so to speak its semantic image projected from context. If the field is
defined in relatively broad terms, it may activate most of the general part of the ideation
base; but a more restricted field may call on only one particular pan of it. In such cases, it
may be useful to reduce the whole ideation base to only those parts that are implicated for
the particular field; that is, to set up a field-specific ideational semantics. (More generally,
this might be a context-specific semantics, including interpersonal and textual as well as
ideational meanings).1 However, the price of this would be to isolate such a field from
others to which it is related.

The latter option was explored in Halliday (1973) and used as the basis for the SLANG
generator developed by Patten (1988) see also Mattbiessen (1990a) and Mattbiessen &
Bateman (1991) for discussion relative to text generation, and Caffarel (1990, 1992) for
discussion pertaining to the semantic system or systems of tense in French.

8.1 General

323

8.1.2 Domain models and the general model


As we have said, the semantic correlate of a contextual field is a domain. When we
model the ideational semantics of a particular field, we create a domain model. What is
the relation of this to the general model outlined in Pan II?
Domain models are variants of the general model. A particular domain model
specifies which of the semantic systems in the overall model are activated in a particular
contextual field: the ideational meanings that are "at risk" (Halliday, 1978a; cf.
ODonnell, 1990). Each field thus has its own semantic profile, which can be seen
against the background of the overall semantic potential. For instance, in the culinary
field of recipes (cf. Section 8.3 below), the domain model occupies a fairly narrow band
within the overall model of figures of doing and being, including various particular
subtypes such as cutting, being crisp. If we then switch our point of vantage, the overall
model appears as a generalization across the full range of such field-specific varieties.
The relationship which links the domain models with the general model is
instantiation. Instantiation relates the system to the instance, at any given stratum;
thus, at the semantic stratum, instantiation is as shown in Figure 8*2. We can choose to
model at any point along this instantiation cline. If we model the semantic system, we
have the "general model11 outlined in Part . We could choose, at the other extreme, to
model a particular instance: the semantics of just one text. But we are more likely to
model some intermediate region, a semantic space that is less than the overall meaning
potential of the language, but greater than that occupied by any single text. In other
words, we expect typically to be modelling a domain.
A domain is located somewhere between the system and the instance. Hence it can be
looked at from either end. Seen from the 'system* end, a domain is regular and repeated
patterning in the way the potential is deployed, over the course of time. Seen from the
'instance' end, it is generalization across particular instances of text.
Instances do not usually repeat each other word for word; there is some variation, for
example between one cooking recipe and another. It is this variation that constructs the
domain. With each instance of the field, the overall meaning potential is instantiated in
slightly varying ways; this variation both confirms the earlier patterns and nudges them
along some novel path.

324

8. Building an ideation base

Fig. 8-2: The cline of instantiation at the semantic stratum


Thus when we model a domain, we aie still defining a meaning potential, a
semantic space in which meanings are constantly being realigned and new meanings
created. In other words, we are describing a register; or, more accurately, a region of
registerial variation (sometimes called 'diatypic' variation to suggest the analogy with
dialectal variation). Register is systematic variation within the overall meaning potential;
from this point of view, a domain is a region of registerial variation corresponding to the
contextual variable of field the meaning potential determined by, and determining,
'what's going on1.
Figure 8-3 represents the location of registerial domains within the ideation base along
the dimensions of instantiation and variation. The overall semantic potential is diversified
into registerial varieties that emerge as patterns of instantiation. In the figure, we have
had to show the registerial domains as discrete locations within the semiotic space of
instantiation & variation. This is, of course, only a convenient representational fiction.
(i) On the one hand, registerial domains form families within which there is variation,
such as the family of culinary procedures, which encompasses a variety of different kinds
of recipes, or the more extended family of procedures for creating artefacts, (ii) On the
other hand, the variation is multidimensional.
Since domains are instantiations of the overall ideational potential, any properties
associated with types in this potential are also present in the domains. Thus 'cooking' in
the recipe domain carries with it whatever information is associated with 'cooking* in the
ideational potential in particular, information concerning the participant roles and the
nature of the phenomena that take on these roles. At the same time, it is likely that

325

8.1 General

further constraints will be attached that are specific to a given domain. For instance,
'cooking' in recipes may be distinguished from 'cooking1 in the overall ideational potential
(i) in that it has to have an Agent, and (ii) this Agent is restricted to being the addressee.

Fig. 8-3: Domain models in relation to the overall model


The diagram in Figure 8-3 shows that delicacy is distinct from instantiation. Delicacy
is the degree of detail, or specificity, to which the description is taken. A contextual field
may be specified to varying degrees of delicacy, e.g. see Figure 8-4. We may compare the
category of 'semanc field1 in the field theory associated with the name of Trier; these are
lexicosemantic (that is, the contrasts in meaning aie those realized by lexical choices),
which indicates that they are located towards the more delicate end of the content
continuum. Our domain models are likely to occupy more of a middle ground; for most
purposes, we will be more inclined to model, say, 'team & ball games', than either
'recreation'at the very general extreme or hockey' at the very specific. Such a domain is
general enough to cover a reasonably expansive semantic space, both instantially (large
number of instances) and potentially (capable of engendering further meanings), while

326

8. Building an ideation base

still being specific enough to allow us to construe particular semantic features (types of
figure, elements, etc.).

Fig. 8-4: Example of delicacy in field


Thus in selecting the degree of delicacy of the field description we are locating
ourselves in the semiotic space defined by instantiation and variation. Any properties
associated with less delicate categories are inherited by those that are more delicate. For
instance, 'simmering' is a kind of 'cooking', both of which are culinary phenomena; but
'cooking* is also, via some intermediate steps, a kind of 'dispositive doing'. Consequently,
simmering will inherit the participant roles of (i) Actor, (ii) Goal, and (iii) resultative

8.1 General

327

Attribute (the construal of the resulting change affecting the Goal); and, like figures in
general, it can cooccur with a number of circumstantial roles such as Manner, Time, and
Duration. These properties do not have to be specified ova- again for simmering. For
instance, Highton & Highton (1964) define simmering as 'cooking very slowly in water
which is just below boiling point'. This suggests that it can be modelled as a subtype of
'cooking1 where the Manner is restricted to Very slowly in water just below boiling
point1. At the same time, 'simmering' will inherit from 'cooking* the value restriction on
the Goal: that is, only that which can be cooked can be simmered.
It is important perhaps to make quite explicit the distinction among the three concepts
of realization, instantiation, and delicacy, since each of these is a distinct scale of
abstraction. It is easiest to describe them in terms of metalanguage dynamics: what we are
doing when we move along these different scales. (1) Realization is the relation of one
stratum to other strata (in any stratified system, with language as prototypical); when we
shift attention from semantics 'upwards* into context or 'downwards* to lexicogrammar and
phonology/ graphology, we are moving in realization. We can do this at any degree of
delicacy, from most general to most specific; and we can do it at any point along the
instantiation scale, from system to text. (2) Instantiation is the relation between the
system and the instance. When we shin attention along this scale, we are moving
between the potential that is embodied in any stratum and the deployment of that
potential in instances on the same stratum (between the climate and the weather, to use
the analogy from our illustration below). Again, this move can be made at any degree of
delicacy. (3) Delicacy is the relation between the most general features and the most
specific. When we shift attention from, say, 'recreation' to iockey' at the level (stratum)
of context, or from 'syllable' to long open nasal syllable' to /p:/ in phonology, we are
moving in delicacy. Again, we can do this at any point along the instantiation scale.
It is common in natural language processing to approach the problem of modelling in
a philosophical mode: one reasons about what the domain is or could be like, perhaps
starting with some central 'concepts'. But a more powerful alternative is to derive a
tentative sketch of the domain model and its relationship to the general model through
lexicogrammatical analysis of a corpus of texts in the appropriate register, e.g. a corpus
of weather forecasts or a corpus of recces. This is in fact how we have approached the
culinary and meteorological domains: exploring the lexicogrammatical patterns in sample
texts and relating them to the semantic system.
We now pass to the illustration of these two domain models, meteorological aid
culinary.

328

8. Building an ideation base

8.2 The language of the weather


8.2.1

Introduction

A considerable amount of discourse, in English, and no doubt in every language, is


concerned with the weather. As far as English is concerned this ranges from casual
greetings and observations that fall within what Malinowski (e.g., 1923) called 'phatic
communion' communing with people by 'phasis', i.e. togetherness through talk to
theories of climate formulated at a highly technical and abstract level in scientific papers
and reports.
The terms climate and weather are related to each other as langue to parole: as the
system of language is to its instantiation in text. Weather is the instantiation of climate;
climate is the system 'behind' the weather. As with language, there is, of course, only one
set of phenomena here not two; when we refer to climate we are construing general
principles and tendencies that 'explain' the multidimensional microvariation that is what
we actually have to live with when gardening or planning an outing and what farmers
and sailors have had to struggle with wherever, as a result of settlement, weather became
the dominant factor in shaping the human condition. This variation is represented in
countless folk sayings both old and new, staid and humorous, many of which are attempts
at predicting what will come next, e.g.
Red sky at night is the shepherd's delight.
Red sky in the morning is the shepherd's warning.

Some suggest the futility of making any predictions at all, a point of view naturally
voiced by the Irish:
If you can see the hills it's going to rain. If you can't see the
hills, it's raining.

In colonial days the British who bad served in India used to explain what it was like to
live there by saying
In England we have no climate only weather. In India they have
no weather only climate.
And now that we are less 'at the mercy of the weather1 than our ancestors used to be we
can afford to take a more positive view of variation:
If you don't like the weather, just wait five minutes.

8.2 The language of the weather

329

Not surprisingly, the weather is a good domain for applying chaos theory, from the
most sophisticated to the most popular level. The sheer number of different variables, and
the immense variation in scale across which any one of them has to be observed and
measured, ensures that there are no simple linearities relating weather to features of
climate; just as in acoustics, where minor fluctuations in the phasing of different
variables produce what appear as totally different wave shapes (causing major problems
for automatic speech recognition), so in the weather small perturbations in the timing of
different components cause local weather patterns to vary in apparently unpredictable
ways. Thus predicting weather is rather like predicting text: one can make certain
predictions about what people are going to say or write with a certain probability of being
right a probability that is significantly greater than chance, but not great enough to be
easily used in (for example) parsing programs, because of the great number of variables
that play a part in conditioning choices in the text. Similarly, predicting the weather
appeared to have reached a similar level of probability, but then to have struck a plateau
from which it proved immensely difficult to proceed any higher, We understand that what
finally enabled the forecasting to move beyond this level of success was a shift in the
relative perspective that is given to chaos and order, instead of saying 'the unmarked
state of a system is one of order, our task is to see through the apparent chaos behind
which the order is hidden', if you say rather 'the unmarked state of a system is one of
chaos; our task is to impose some kind of order upon it' you open the way to improving
the accuracy of your predictions about one level up (say, from around 0.7 to around 0.8,
which would be very significant in a number of practical contexts). Since linguistic
patterns are very clearly fractal, with patterns repeated not merely across differences of
scale but also across differences of level of realization (stratum), it seems possible that a
similar shift in our thinking will be helpful in text forecasting also. (The starting point
in linguistics is however very different from that in physical science, the received wisdom
in linguistics being that no prediction is possible at all. We should stress, perhaps, that
we do not subscribe to this view.)
Wignell, Martin & Eggins (1990) have shown that in a high school geography text
book which they analysed in lexicogrammatical terms (using systemic grammar) one of
the critical steps that the learner had to take was that of constructing from the text a
taxonomy of climate. This taxonomy was both compositional (meronymic pans of
climate) and hyponymic (kinds of climate), yielding a structure that is summarized in
Figure 8-5. It was not set out in the form of tables or diagrams; but it could be recovered
from the text, and Wignell et al. showed what discourse-semantic patterns (types of
process and so on) were characteristically built up in the grammar in the course of
presenting the information on the basis of which this taxonomy could be construed.
Given that, as part of mastering scientific knowledge in the course of education, the
learner has to construct such taxonomies, we can interpret this task as the learner's

330

8. Building an ideation base

construction of meanings which in turn are construed in specific lexicogrammatical


forms. The general principles behind this meaning-making or semogenic approach to
learning, and more specifically to learning science, will be found in the work of Lemke
(1984; 1990a). Lemke's 'thematic systems' are the systems of meanings that the pupils
are required (by the teacher, text book, syllabus and educational ideology) to build up into
their ideational meaning base. In this process, the children are simultaneously working at
both ends of the instantaoo cune, constructing both system and text They have to
build up the systems of 'geographical meanings' (such as the two taxonomies shown
above) from the written and spoken texts they engage with; and these systems are their
potential for creating their own geography texts.

Fig. 8-5: Geographical taxonomy of climate (meronymic & hyponymic)


Here we are focussing not on the language of the geography text book, or of climatic
theory as scientific construct, but on another register having to do with the weather,
namely the language of weather forecasting. We have used as our sources those weather
forecasts that are presented in print in a special section of a daily newspaper, the
specimens that we used for detailed analysis were taken from the Chicago Tribune and the
New York Times, but we have not limited ourselves exclusively to the forms of text
employed in these two organs. We use our outline of the meaning base, as presented in
Part II, to characterize the construction of ideational meanings in this register, in

8.2 The language of the weather

33 1

illustration of the principies of this approach as they would apply to a form of discourse
that is semi-specialized and semi-technical. Since much of the text that needs to be
processed in information systems of various kinds is of roughly this degree of
specialization and technicality it seemed an appropriate choice for purposes of illustration.
It is not a 'closed1 register, i.e. one in which the set of possible messages can be listed;
but neither is it a random sample of the English language. Anyone having mastered
English beyond a certain degree of proficiency could construct the text from the situation
or, equally, construct the situation from the text Here is a typical sentence:
Tonight will be cloudy, with a chance of rain or thunderstorms.
(NYT)

There is little doubt about the source of an instance such as this.


One final introductory point concerns grammatical metaphor. This register is not
highly metaphorical; but it does contain instances of grammatical metaphor, e.g. in the
woke of recen storminess. These are part of the language of weather forecasting and have
to be treated as such storminess is construed as a participant (thing) and thus enters
into the paradigmatic construction of phenomena as systems or sets of things. Weather is
one of the domains of experience that is particularly problematic in this respect; we
referred above to the example of rain as indeterminately process or thing, with neither
appearing as obviously the more congruent form. In the overall semantics of weather
forecasting, it is not so much the metaphorical nature of particular instances that is
significant; what is of interest is whether, and if so how far, grammatical metaphor is
systematically involved in construing this aspect of experience into information into a
commodity that can be stored, disseminated, withheld, used for policy-making and
planning, and so on.
Let us first attempt to describe the context of weather forecasting informally in terms
of field, tenor and mode: its 'contextual configuration* as defined by Hasan in Halliday &
Hasan (1985/9).
I. Field
Keeping people informed: disseminating information by mass media* print
newspaper: daily.
State of natural environment: natural phenomena: weather present conditions
and prognostications for immediate future (0-2 days), specifying locality
(varying focus, from district to world, with emphasis on 'region').
Special features: (1) explanations of events; (2) warnings of disaster; (3)
information for special interest groups: those engaged in agriculture and
gardening, boating and fishing, aviation, sport

332

8. Building an ideation base

II. Tenor

Expert to laity, but with only moderate distance.


Expert standpoint: impersonal, with acknowledgment of uncertainty; attitude to
subject-matter neutral.
Audience: interested general public, plus some groups of special concern (whose
actions may be affected by the information).
Monologue.
III. Mode
Written: print: prose format
Rhetorical mode: informative; semi-technical.
Accompanied by other visual modes: maps, with special (technical) symbols.
Day-to-day continuity, with date and place of origin specified.
Since we are concerned here with the ideation base, the relevant aspects of the
situation are mainly those specified under 'field*. If we are generating texts, a
description of the field will show which areas of ideational meaning are likely to be
activated, and which options within those areas foregrounded. In other words, it will point
to semantic features incorporated in the domain model. If we are parsing, we shall be
able to construct the field from the meanings that are foregrounded in the text; and as the
field emerges (or if it is known beforehand) it can be used to guide the parsing, for
instance by ranking alternative interpretations and by managing the complexity of the
parsing process.
For example: in the sample of texts used for this study, two thirds of all finite clauses
had finite verbal operator will, and another 30% had finite verbal or some other form of
modality: could, is/ are likely &c. Semantically, these represent future time aid
probability, which in turn are meanings we expect to find in a context of prognostication
and prediction and torn which, therefore, we are able to construe such a context Note
that what is significant is not that there are one or two odd instances of future time
reference, but that future has taken over as the norm; globally (across contexts) it is a
highly marked choice of time reference (cf. Halliday & James, 1993), but locally in this
context it has become the unmarked option (see further Chapter 13, Section 13.3 below).
In what follows we shall summarize the main features of weather forecasting texts in
terms of the semantic model we sketched in Part u - recognizing sequences (Section
8.2.2), figures (Section 8.2.3) and elements (Section 8.2.4). Since this is not a
quantitative study we have included features from other weather forecast texts which were
not included in the original sample. The aim is to show how discourse in a specific

8.2 The language of the weather

333

register can be related to, and derived from, the meaning potential of the linguistic system
as a whole.
8.2.2

Sequences

The meanings that are construed as sequences in weather forecasting texts are fairly
restricted ones. There is very little projection (none in the sample studied), and within
expansion only the logical relation of extending is exploited to any extent, though there
is a certain amount of elaborating as well (no enhancing, apparently; cf. below on
causality in faz figures}.
Extending relations, in turn, are either additive or adversative (contrastive). The
additive nexus joins two related figures either on an equal basis with and:
The chance of showers will end by Sunday night, and winds will
shift to north (d)
Skies are expected to clear Wednesday, and afternoon highs will
approach 70 (CT)
or unequally, the second being non-finite:
Winds will shift to north, pushing {'and this will push 1 ) lows
into the low 50s (CT)
Tonight will be cloudy/ with ('and there will be') a chance of
rain or thunderstorms (NYT)
An additive nexus can then be extended further, like the following with one unequal
and one equal:
Warmer tempratures are likely Thursday, with ('and there will
be') partly cloudy skies and highs ('will be') in the mid- to
upper 70s. (CT)
No sequence of more than three figures occurred in the sample, and very few sequences
consisted of more than two (i.e. a single nexus).
Adversative nexuses contrasted two figures, always in terms of either place ('here rain,
elsewhere dry') or time ('morning sunny, afternoon cloudy'); they could be equal or
unequal, in varying degrees. Note that in the first example the sequence containing the
adversative relation is construed in the grammar not as a clause complex, but as a
cohesive tie between complexes:

334

8. Building an ideation base

High pressure will keep skies sunny and dry from New England south
to Maryland. However, the Virginias and Carolinas will be hot &
humid .. . (CI)
Skies will be sunny tomorrow morning, but will become increasingly
cloudy during the afternoon. (NYT)
Skies will be clear to partly cloudy over the rest of California,
though widely scattered thunderstorms could develop in southern
sections. (CI)
The heatwave in the Southeast will weaken slightly,
northern Florida will remain hot. (CT)
Morning skies will be partly cloudy today, becoming
become1) partly sunny by afternoon. (NYT)

although

( 'but will

by which time, of course, they will no longer be morning skies. Less common is a
third type of extending, the subtract!ve:
The rest of the south will be mostly dry and sunny, with only
{'except that there will be') isolated showers in Florida. (CT)
Elaborating nexuses are less common than extending, and seem to be always unequal:
High tenperatures will be 70s to low 80s, warmest
Carolinas. (CT)

in

the

The second figure adds specificity to the first. The following might also be
elaborating, though it is probably to be reinterpreted as temporal, 'while at the same
time':
Skies will be partly sunny in Alabama and Georgia as scattered
thunderstorms diminish later in the day. (CT)
We could derive from these the general principles that, given any two figures, (1) if
they represent different but related weather features staying constant in one location (time
and place), they form an additive nexus; (2) if they represent the same weather features
varying across dnrent locations (time or place), they form an adversative nexus; and
peihaps (3) if they represent the same weather features varying across locations where the
second location is part of the first, they form a subtractive nexus. One instance not
covered by these principles is

8.2 The language of the weather

335

Highs will be nostly 70s and 80s; a few 60s are likely near the
Canadian border, and parts of the southern plains will reach the
90s. (CT)

where the second nexus would be expected to be adversative; the reason it is additive is
that the last two figures are alike in both being exceptions (subtracve) to the first.

Fig. 8-6: Meteorological sequences


Any such nexus may assign equal or unequal status to the figures; where it is unequal,
the sequence seems to be typically descending (grammatically, the dependent clause
follows). It is difficult to find a general principle governing the choice of status, but in
the adversative there is perhaps a prosody of evaluation bere: if the second condition is felt
to be 'worse1 (less desirable) than the first, the adversative relation is more likely to be
equal, because of the somewhat negative connotation of an explicit 'but*. Thus compare:
Morning skies will be partly cloudy today, becoming partly sunny
by afternoon. (NYT)
Skies will be sunny tomorrow morning, but will become increasingly
cloudy during the afternoon. (NYT)

In New York, at least, sunny skies are to be preferred ova* cloudy ones.
We can represent the postulated semantic options for sequences in a network as shown
in Figure 8-6 above. We have included the distinction, within the unequal, between those
where the second, minor figure is arguable and those where it is inarguable
(grammatically, finite versus non-finite hypotactic clause), although in the main sample
the only clear example of a finite dependent clause is skies will be partly sunny ... ay

336

8. Building an ideation base

thunderstorms diminish (adversativos with although, though are hardly dependent). It


seems likely that the distinction will be significant in the register as a whole.
8.2.3 Figures
All figures in the sample are figures of being; there is no doing, sensing or saying. Of the
three modes of being existing, ascribing and identifying those that occur are the
first two.
8.2.3.1 Figures of 'existence1
Existing includes being in existence, coming into existence (emerging) and staying in
existing (persisting); with or without some external cause. We will exemplify those
without a causer first
(1) Being

in existence is typically associated with some assessment of

probability; this is then construed metaphorically as an ascnptive figure with the


probability as attribute:
rain is forecast today for the Middle Atlantic States and the
Missouri Valley (NYT)
clouds and showers are likely in Idaho and parts of Montana (CT)
rain is likely in the Ohio River Valley and Kentucky (CT)
scattered showers are also expected in the Carolinas (NYT)

(Note that some such examples are, however, still ascnptive in their congruent mode; e.g.
clear skies are forecast for tonight, high temperatures are expected, where the sense is
'skies will be clear', 'temperature will be high'. See Section 8.2.3.2 below.)
It seems that figures of 'being in existence' without attached probability occur mainly
in nonfmite clauses; e.g.

with only isolated showers in Florida (CT)


(For the subregister where 'figures' consist of one element only, such as Variable
cloudiness, see below, the final paragraph of the present section.) The simple existential
there mil be... is not generally favoured, although related formulations sometimes occur
with the sense of 'there is a warning', e.g.

an air pollution advisory is also in effect for Cook and Lake


Counties until 4 PM Saturday (CT)

337

8.2 The language of the weather

(2) Coming into existence, on the other hand, is commonly asserted either with
or without probabilities attached, e.g.
showers and thunderstorms will gather over the northern Rockies

(CD

widely scattered showers will develop in the Balkans (CT)


scattered showers may develop (CT)
Verbs such as gather and develop typically realize figures of doing; here, however, they
represent figures of being, in a metaphorical syndrome in which the weather processes ate
construed as participants, e.g. showers. These examples are different from congruent
figures of doing, such as birds will gather over the northern Rockies, in both aspects of
the participant-process configuration: (i) showers, thunderstorms, and other such
metaphorical participants are not like Actors such as birds that are involved in figures of
doing; and (ii) the processes of gathering, developing, etc. are metaphoric expressions of
phase they are inceptive, with the sense of 'beginning to be', whereas there is no
inceptive feature in such processes in figures of doing.
Since it is entirely possible to create a semantic proportion of the kind:
modality \ phase
\
(a) absolute
(b) probable

(1) be in existence

(2) come into existence

ntinwitioctur

rain will develop


rain may develop

rain may occur

it is not obvious why type (la) should be disfavoured, as it seems to be although


possibly it is felt to be more absolute than (2a) on the grounds that the inceptive phase
already bas a slight prosody of uncertainty about it.
(3) Staying in existence and (4) ceasing to exist should also be included as
options (e.g. rain will/ may persist; rain will/ may cease), although much rarer in the
sample studied.
What are the 'things' that are being said to 'exist', in a weather forecast? Typically they
are forms of precipitation: rain, snow, showers &c. Sometimes, by a further grammatical
metaphor, it is the probability that is said to exist, e.g.:
with (= 'there being 1 ) a chance of rain or thunderstorms (NYT)
the chance of showers will end by Sunday night (CT)

the latter suggesting that an opportunity is about to be missed!

338

S. Building an ideation base

Existing may be construed as the effect of some external cause, with or without a
probability being attached. Typical 'causes' aie weather systems, fronts, and high or low
measurements of pressure:
a warm front may bring scattered showers or thunderstorms to the
northern Tennessee Valley (CI)
a separate weather system will bring thunderstorms to the eastern
Dakotas and Minnesota in

the morning and into the Great Lakes

later in the day (CI)


low pressure will produce scattered showers from the Pacific (CT)
a col front

s, talked ji^ the

^groJxnflS

may

set

off

scattered

afternoon thunderstorms (CT)

There is a systemic relation between produce in this context, as 'causing to be', and
produce as 'doing' with a doer involved; analogous to that which we saw with develop
and gather above (as 'beginning to be' and as 'doing'). These provide interesting
illustrations of the complementarity of 'being' and 'doing', which we discussed in Chapter
4, Section 4.3.3 above.
8.2.3.2 Figures of 'ascription* (attribution)

Ascribing Cbeing' in the sense of carrying an attribute), like existing, distinguishes a


neutral phase 'be' and marked phases of becoming, staying or ceasing to be. The carrier of
the attribute may be (a) a feature (a component thing or quality) of the weather, e.g. skies,
temperatures, storms, or (b) a time or place; and again, some external cause may be
involved.
(a) Carrier as weather feature. Those where the carrier is a weather feature
typically consist of a configuration of this feature with a relational process 'be', an
attribute, and a circumstance of time or place:
skies will be clear tonight (NYT)
some storms could be heavy with hail and strong winds (CT)
skies will be mostly sunny over the southern plains (CT)
high temperatures will be in the middle 70s2 (NYT)
lows will be near 60 degrees (NYT)

^ Note that high in this example is a class of temperature not a property the meaning is
'the top temperature in the 24 hour period'; cf. highs, lows.

8.2 The language of the weather

339

(i) Sometimes the circumstance of time is construed as a class of the weather feature
in question, e.g.
morning skies will be partly cloudy today (NYI)
afternoon highs will range from 50s and 60s in the northern
Rockies to 100s in Arizona deserts (CI)
afternoon highs will approach 70 (CT)
In the last of these the process is metaphonc for te' + minor process (type 8.i in Chapter
6, Table 6(3)): congniently will be near 70.
(ii) Secondly, the attribute itself may be a circumstance of place, of the 'extent1 type;
the feature 'extent' is also present in the process:
rain will be scattered across the central lake region (NYT)
rain showers will be scattered from New England to the Virginias
today (NYT)
showers will also extend from MDntana, across northern Idaho, to
Washington (NYT)
There are also odd occurrences of a related figure in which the place is not being
ascribed to the weather feature as its attribute but rather set up as something to be
defined by it, e.g.
sunny skies will dominate most of the region
Grammatically such examples aie identifying clauses, as shown by the fact that they
select for voice although here the sense of the agnate passive clause most of the region
will be dominated by sunny skies suggests that dominate should be interpreted as an
ascriptive, i.e. as 'prevail1 + circumstance of extent 'over most of the region', rather than
as 'prevail over' + 'most of the region'.
(iii) Thirdly, while there were no examples of become in the texts under study, a
figure such as
skies will become cloudy in the late afternoon
would presumably be a possible sentence in a weather forecast. There are however many
instances where an attribute is transcategorized as a process of becoming, e.g.
the heat in the Southeast will weaken slightly {' become weaker ' )
(CT)

340

8. Building an ideation base

scattered thunderstorms will diminish ('become less 1 ) later in the


day (CT)
skies are expected to clear ('become clear') Wednesday (CI)
winds will shift to north ('become northerly1) (CI)
In the last example, shift is relational not material; the sense is 'change so as to become
northerly', not 'move to the north' (cf. the sand dunes shifted to the north).
(iv) Fourthly, the configuration may include an extrinsic cause, as in the
following example with keep ('cause to remain'):
high pressure will keep skies sunny and dry from New England south
to Maryland (CI)
(b) Carrier as time/place. Those where the carrier is a time or place form
configurations consisting simply of this feature with 'be' plus attribute:
the Plains States will be sunny, warm and dry (NY1)
tonight will be mostly fair (NYT)
most of the South West will be clear to partly cloudy and dry (CT)
northern Florida will remain hot (CT)
Occasionally this relationship is construed as a process of acquiring, with the acquired
possession as Range, as in:3
some areas could get 1-3 inches of rain (CT)
We saw above that some existential figures are construed metaphorically as ascriptive:
namely those where a probability is represented in the form of an ascribed quality, as in:
rain is likely in the Ohio River Valley and Kentucky (CI)
scattered showers are also expected in the Carolinas (NYT)
widely scattered showers and storms are possible in the northern
Tennessee Valley (CT)
These are congruently existential: 'there may be rain', 'there may be scattered showers'
(not 'showers may be scattered1). Similarly in an ascnptive figure the probability may be

Such instances are on the borderline of being and doing: the verb get suggests a process of
doing, with obtain and acquire as agnates (cf. Halliday, 1985: 125), but here on the other
hand the inore likely agnate seems to be a process of being with have. Cf. again Chapter 4,
Section 4.3.3 above.

341

8.2 The language of the weather

mapped onto an ascribed quality; the attribute is then construed as a quality of the 'carrier'
participant (Epithet + Thing construing Carrier + Attribute); e.g.
warmer temperatures are likely Thursday (CT)
clear skies are forecast for tonight (NYT)
high tempratures in the 100s are forecast for Texas and the
Southwest (NYT)
These are ascriptive: the congruent sense is 'temperatures may be warmer', 'skies may be
clear'. The relationship is diagrammed in Figure 8-7.

Fig. 8-7: Modalized ascription reconstrued as ascription of modality


Some instances seem indeterminate, e.g.
heavy to

severe thunderstorms

are possible

in

Colorado and

northern sections of Utah and Nevada during the afternoon (CI)


either 'there may be heavy to severe thunderstorms' or 'thunderstorms may be heavy to
severe'.

The carrier may be a metaphoric nominalization of a process or property as thing:


variable cloudiness is expected tomorrow (NYT)
gradual clearing is expected tonight (NYT)

S. Building an ideation base

342

8.2.3.3 Summary of gures

Our network of figures would then be as shown in Figure 8-8. This may be too
constrained, in not generating caused ascriptives with location as carrier, e.g. mil keep
tomorrow fine; but no instances of this type were found. On the other hand it may be too
generous, in allowing for all phases with all types of existence and ascription, whereas
only a subset were found to be represented in the texts studied; here however there seems
no clear reason for excluding the remaining combinations (e.g. rain will persist; a cool
front will reduce temperatures). It would also be possible to treat as a systematic option
the mapping of probability on to attribute or existent: see Figure 8-9.

Fig. 8-8: Types of figure in domain of weather forecasts


Let us enumerate the elements and configurations that are generated by this network.
We specify paths through the network (selection expressions) together with the
configurations of elements by which they are realized:
(1) 'Existing' figures
exist uncaused
Existent + Bel (+ Place) (+ Time)
exist: cause
Agent + Be2 + Existent (+ Place) (+ Time)

8.2 The language of the weather

Fig. 8-9: Types of fgure, including construal of probability


(2) 'Ascribing' figures
ascribe: weather as Carrier uncaused
Carrier + Be3 + Attribute (+ Location) (+ Extent) (+ Time)
Carrier + Be5 + Extent (+ Location) (+ Time)
Carrier + [Attribute = Process] (+ Time)
ascribe: weather as Carrier: caused
Agent + Be4 + Attribute (+ Location) (+ Extent) (+ Time)
ascribe: location as Carrier: uncaused*
[Cama- = Location] + Be3 + Attribute (-1- Time)
[Carrier = Time] + Be3 + Attribute (+ Location) (+ Extent)
[Carrier Location] + Be6 + Attribute
Notes:

(1)
Bel = 'exist, come to exist, continue to exist, cease to exist'

Be2 ='cause to exist1


Be3 = 'have attribute ascribed'
Be4 s'cause to have attribute ascribed'

343

344

8. Building an ideation base

Be5 = Tiave extension*


Be6 = 'hve possession'
For examples, see Table 8(3) below.
Location = location in space1
Extent = 'extent in space*
For examples, see Section 8.2.4.3 below.
(2) Certain further distinctions are not introduced in the network, viz.:
(i) have attribute ascribed ^ Be3 / have extension ^ Be5
(ii) have attribute ascribed: 'become weak' ^ Be3 / 'weaken' ^ Attribute - Process
(iii) location as carrier spatial location as carrier ^ Location s Carrier / temporal location as
carrier ^ Times Carrier
(iv) spatial location as carrier have attribute ascribed ^ Be3 / have possession ^ Be6
(3) Features of time and place are not included in the network; in principle, any
selection expression may be accompanied by either or both of these.
Tables 8(1) and 8(2) display the features in the network (rows) and show how they are
realized as the presence of functions or constraints on functions (columns).
Finally, we note transcategorizations and junctions, showing the type of element
involved:
Transcategorizations

Attribute => Being

Junctions

Probability Attribute
Probability = Existent
Tune = Carrier
Attribute Existent

process (weaken,

diminish, end,
increase, clear)
participant: quality
participant: thing
participant: thing
participant: thing

345

8.2 The language of the weather

Table 8(1): Features of 'existing' figures


Be
Be

exist

uncaused

Bel:
process,
'exist' (be/
come to be/
stay being)
Be2:
process,
'cause to
exist'
(bring,
produce, set
off)

caused

Agent

Existent

Place
(Place)

Time
(Time)

Agent

probable
determined
neutral
phase be
marked
phase

Attribute

Existent:
participant:
(quality:)
thing

Probability
= Attribute

become
stay
cease

Bel:
be
Bel: gather,
develop
Bel: persist '

There is one sub-register within the "weather" register where the figures appear to
be constructed not as configurations but as single elements; for example
Saturday: Mostly sunny; highs 78 to 83.
Saturday night: Clear; lows 58-63.
Sunday: Increasing cloudiness, chance of thundershowers
north; highs 80s. (CT)
However, when they are scrutinized more closely such figures turn out to be identical
in their semantic construction with some of those in the primary register. What is
different is the type of grammatical structure through which they are realized. Those
just cited are in telegraphic form (cf. Halliday, 1985: Appendix 2), and it is not
difficult to match them with their expanded agnates:
Saturday: Mostly sunny;

Saturday will be mostly sunny.

Sunday: Increasing cloudiness,

Cloudiness will increase

chance of thunderstorms north;

with a chance of thunderstorms

Sunday,

in the north.
highs 80s.

High temperatures will be in


the 80s.

Semantically however they do not need any special provision.

8. Building an ideation base

346

Table 8(2): Features of 'ascribing* figures


Process:
Be

Agent

Be

ascribe
weather
feature as
carrier
location as
carrier

Be3:

caused

Bc4:
process,
causing to
have
ascribed'
Be:
modalized

probability
as process
probability
as attribute

determined
neutral
phase be
oecome
marked
phase
stay
cease

8.2.4

Attribute

Location
(location
in space)

Extent Time
(extent
in
space)

Carrier

Attribute

(Location)

(Extent)

Gamer
Location
OR
Carrier
Time

uncaused

probable

Carrier

Location

Time

Causer

Probability
s Attribute

Be3:
be
Be3:
become,
turn, shift
Be3:
remain,
stay

Elements

The elements occurring in the weather forecast texts can be specified according to their
potential functions within the figures just described. In the most general terms, they can
be categorized as processes, participants, and circumstances:
(1) processes:

existing, having ascribed, extending, possessing (= Be 16 above)

(2) participants:

precipitation, and other weather phenomena; weather


systems;
temperature and pressure; times and places; probabilities

(3) circumstances:

locations & extents in time & space (= Place 1-2 above)

We will consider each of these in turn.

(Time)

347

8.2 Tbe language of the weather

8.2.4.1

Processes

All tbe processes in these texts aie types of being: existing (including causing to exist),
having some attribute (including causing to have some attribute), or having some
extension or possession. These are shown together with examples in Table 8(3).
Table 8(3): Processes: types of T>eing'
FUNCTION IN
FIGURE

TYPE

EXAMPLES

Bel

exist:

be
: come to be

Rather, develop

: keep being

persist

Be2

cause to exist

bring, produce, set off

Be3

be ... (= have attribute ascribed)

be, range, approach

become ...

become, turn, shift

keep ...

remain, stay

become ...

weaken, diminish, end, increase, clear

cause to be ... (= have attribute

make, keep

Attribute =>
Process
Be4

ascribed)
BeS

be ... (~ have extension)

extend, be scattered, dominate

Be6

be ... (= have possession)

have, get

In addition, processes may have attached to them an expression of probability: may,


could, should, etc.. (Probability may also be construed metaphorically as a participant
serving as Existent or Attribute.)

8.2.4.2 Participants
Participants of the subtype 'thing' include various types of precipitation; skies, wind,
frost and cloud; conditions; temperature, pressure; weather system, front; particular times
and places. They also include two sets of qualities construed as things: names of weather
conditions, and names of probabilities. These are differentiated (i) according to which
functional roles they take on in figures and (ii) according to which qualities are ascribed to
them. Table 8(4) shows which types of participant take on which functions in a figure.
We have separated out a function of Possession (e.g. showers, in some parts will have
showers) from that of Attribute, of which it is a subtype, because in this register such
"possessed" Attributes pattern more like Existants than like Attributes of other types.

8. Building an ideation base

348

Table 8(4): Participants according to roles in figures


TYPE OF
PARTICIPANT:

things

FUNCTION IN FIGURE:
1. Existent
or Attribute

2. Carrier

3. Agent

4.
Attribute

precipitation
other: wind,
frost, cloud
quality:
quality of
expansion:
cloudiness,
storminess

Attribute =
Existent

(see below)

quality:
quality of
projection
(chance,
possibility,
likelihood)

Probability =
Existent

(see below)

temperatures
skies
conditions
places

Location =
Carrier

times

Time =
Carrier

front
weather
system
pressure
qualities (not construed as
things)
We can now group the things into sets according to their functional roles: (1) Existent
or Possession, (2) Carrier, (3) Agent, (4) Attribute; and specify the various qualities that
are assigned to the members of each set.
(1) Existent or Possession

Table 8(5) shows the qualities ascribed to participants functioning as Existent or


Possession.

8.2 The language of the weather

349

Table 8(5): Qualities ascribed to things as Existent or Possession

(2) Carrier
All the types of 'thing' that function as Existent or Possession can also function as
Carrier in a figure of ascription; see Table 8(5) for the qualities assigned to these. In

350

8. Building an ideation base

addition, as Carrier we find temperatures (classified as high/ low; also highs}, skies
(classified as morning, afternoon, evening or nighi) and conditions.
There are also circumstantial elements in this function: Location as Carrier and Tune
as Carrier.
Location = Carrier

[any expression of location, e.g. the Plains States, the northern plateau region,
Texas, inland, the northern half of the country, the Carolinas, the Rockies, the
metropolitan area, the Northwest, the Mexican border, &c&c]
Time s Carrier

today; this morning, this afternoon, this evening, tonight


tomorrow; tomorrow morning/ afternoon/ evening/ night
mornings, afternoons, evenings, nights
the morning hours, the early/ middle/ later part of the day
(3) Agent

The 'things' that function as Agent in figures of existence and ascription are front,
weather system and pressure; the qualities that are assigned to these have to do with degree
or with position and movement:
TYPE OF
THING:
front
weather system
pressure

TYPE OF QUALITY:
temperature
warm, cool, cold
occluded

space
stationary,
moving
separate,
departing

intensity

high, low

(4) Attribute

All the types of quality that can occur as Epithet with the things functioning as
Existent or Possession (see Table 8(5)) can also occur as Attribute in a figure of
ascription (e.g. heavy showers/ showers will be heavy), though some are less likely in
this latter function (e.g. occasional). In addition, other qualities occur as Attribute in
configuration with a specific Carrier; these are shown in Table 8(6).

8.2 The language of the weather

351

8.2.4.3 Circumstances
Circumstances are almost exclusively spatio-temporal locations and extents in spacetime; the only others are a small number of circumstances of manner. Those occurring in
our sample can be listed as follows.
1. Place 1: location
in (on)/ to/ from/ near [+ place expression]
north/ south/ east/ west of [+ place expression]
2. Place 2: extent
over/ across [+ place expression]
as far as [+ place expression]
from [+ place expression] to [+ place expression]
generally
3. Time
today, this morning, this afternoon, this evening, tonight
tomorrow, tomorrow morning/ afternoon/ evening/ night
during/ in + the day/ morning/ afternoon/ evening/ night
during/ in + the morning hours
during/ in + the early/ middle/ later part of the day
4. Manner
slightly, gradually
8.2.5 Commentary on weather forecasting domain
What we have set out to do in this section is to use our general semantic characterization
of the ideation base to construct a semantic representation of the ideational band within a
particular register of English a "domain", that of weather forecasts as printed in the
daily press. The examples were drawn from a small collection of such texts; but the
sample was used as a means of entry to show what were the typical meaning formations
within this register, not as a corpus covering the entire range of possibilities. So both the
networks and the constructional categories go beyond what was present in the sample.
The words used to exemplify the elements were taken from the sample texts, but it should

352

8. Building an ideation base

not be difficult to assign otber words to the appropriate categories when extending the
coverage beyond what we have been able to include here (cf. Rgure 8-3),
Table 8(6): Other Attributes combining with specific Carriers
Carrier

temperatures

conditions,
locations,
times
skies

Attribute
of expansion:
elaborating
temperature

of projection
humidity

hot, warm,
mild, cool,
cold, [all + er]
freezing;
high, low [4
er];
in the low/ mid/
upper [decile,
e.g. 50s]
hot, warm,
dry, humid
mild, cool, cold

front,
weather
system
pressure

[sky]

[wind]

cloudy

stormy

(partly/
mainly/
mostly +)
clear,
sunny,
cloudy,
overcast

space

stationary,
stalled,
moving

intensity

high,
low

probability;
expectation;
prediction

Probability =
Attribute
likely/
probably,
possible;
expected,
forecast,
projected,
predicted

In texts of any register, unless it is totally closed and listable, there will always be
maverick instances which aie more or less outside the range of typical forms. Here for
example we found [most of the region will be sunny and dry] in the wake of recen
storminess and/0 departing weather system]may still exert enough influence to [produce
showers in the Pacific Northwest]; neither in the wake of nor exert enough influence to
bave been included in our coverage. They could be. On the other hand, the patterns they
represent have little generality in this register, so the pay off would be rather small. In
generating weather forecasts one could do without them, and not much would be lost if
one failed to interpret them in parsing. We have concentrated on patterns which are
typical, frequent and productive. (We have also ignored those passages which refer to the
weather in the past. But the same principles applying to tomorrow will be ... will also
apply to yesterday was....)
A description of this kind can be taken over by the grammar and turned into text.
Here we have not attempted to take it the whole way (see further Chapter 9 below); the

8.2 The language of the weather

353

specification of future time reference in the process, for example, would be taken into the
grammar as a preselection of tense: future, which in turn would be realized as mil +
the verb representing the process. We have indicated some of the wordings, both lexical
and grammatical, in order to make our account clear; but these are not necessarily in
fact not usually the only output forms. For example, we have specified clear skies are
expected and skies should become clear, because these are construed differently in

the

semantics: they are agnate but they are not synonymous. On the other hand we have
not described skies are expected to clear as something separate from skies should become
clear; not because we regard them as completely identical in meaning, but because the
distinction is a more delicate one and we consider it would be out of place in the present
context. The choice between the two would be left to the grammar.
There aie two important principles at work, (i) One is that of delicacy. In a
paradigmatic model such as the present one, both the lexicogrammar and the semantics
accommodate variable delicacy: two agnate constructions may be both 'the same* and
'different1 according to the delicacy of the focus adopted, (ii) Secondly, although we are
presenting the semantics as input (in the case of a text generation system), the
assumption being that the grammar then does what it is told, our metatheoretic position
is that the construction of meaning is both a discourse-semantic and

lexicogrammatical process, so that the particular way in which this is modelled in any
designed information processing system (e.g. a text generation system) is an artefact of
that system, reflecting on the one hand the state of the technology (hardware and software)
and on the other hand the aims of its designers and the level of their understanding of the
language and language variety they are working with.
In the present instance, the system might be that of Penman, in which the text
generation is driven by the grammar, but at every choice point there is a 'chooser' which
consults the meaning base for instructions on which way to go. Or one might adopt,
instead of a 'chooser', a principle that we might call a 'charger1, whereby the text is
generated in the semantics which then 'charges' the grammar with the task of realizing the
sequences, figures and elements that have been construed. In both these models (see
further Chapter 9 below), some discretion will be left to the grammar. Partly this is for
metafunctional reasons; we should recall here that our present semantics is ideational only
we are describing only the 'ideation1 base, not the interaction base and the text base,
and these also provide input to the grammar. But even with this input there will still be
underdetennination; the grammar will never be exhaustively controlled 'from above'. It
would take us beyond our present scope to try to discuss principles for selection within
the grammar itselfdefault choices, favouring or avoiding repetition, and the like. What
we do want to stress is that an information system needs to be flexible enough to allow
for various ways of modelling the relationship between the semantics and the grammar
(including matures of various ways); and that this is likely to be more readily achieved if

354

8. Building an ideation base

the level 'above* the grammar is construed as a meaning base that is, in terms of
linguistic concepts derived from the grammar itself than if it is construed in the form
of knowledge that in some sense 'exists* independently of the language that encodes it
Furthermore, the meaning base in turn can be related to a further level of environment,
the context of situation (cf. Steiner, 1988a). We have given only a very general
sketch of the situation type from which these weather forecast texts derive. But the same
point could be made here too: we may model the relationship in one direction or the
other, but text and context are construed together. Itere is an important variable here, of
course, that we have already referai to as the cline from language in reflection1 to
'language in action*. In situations of the 'language in action* kind, where the discourse is a
relatively minor component of the total activity, the grammar and semantics are
obviously less constructive of the whole than in a 'reflection1 context such as the present
one. In weather forecasting, while the weather itself is not constituted of language, the
activity of forecasting is; the entire situation is built up out of black marks on large
white pieces of paper towards which is directed the attention of some human
consciousness, Out of this is construed a particular context or situation-type; since the
semantics is specific to that situation, the leader knows where he is, construes the
situation in the particular instantial form ('oh so it's going to rain this afternoon1), and
perhaps varies his actions in accordance with this construction.
How is the non-discursive component the weather made to impact on the
discourse? In the old days, I went out and sniffed the air; there was no institutionalized
register of weather forecasting, only some bodily activities, physical and conscious,
forming part of the total continuum of existence. Now, we may envisage something more
along these lines: that the weather satellites record pressures, temperatures, air movements
and so on and transmit a large number of measurements with indication of their place and
time of origin. The weather has now become information; but in numerical, not yet
discursive form. Using guiding principles derived from some representation of the
situation, the system construes these measurements into meanings, in the form of
sequences, figures and elements. The knowledge that the system needs in order to take this
step is itself represented in this form: that is, in the form of meanings, such as 'low
pressure will produce rain1, which, although they may be considerably more complex, are
constructed out of the same resources. The resulting semantic formations are then
construed by the lexicogrammar into wordings, and output in the form of text.
The system may also be designed to produce output in the form of other, nondiscursive graphic modes, most typically maps. One of the interesting questions to be
explored is bow far these other modes of meaning can be built into the same overall
model, and at what points for example, as alternative 'wordings' of a given semantic
construction. In this particular register it seems that much of what it put out in the form
of a map could be interpreted in this way; that, for example, isotherms on the map and

355

8.2 The language of the weather

clauses such as high temperatures will range from 60s in the northern Rockies to 100s in
Arizona could be construed as alternative realizations of the same semantic figure, or
sequence of figures. Many visual semiotic systems are 'paraphrastic' in this sense,
although certainly not all.
Finally we should note that not all discourse about the weather, even if it is printed
under the same rubric in the same newspaper, falls within the register described above.
Here is a much more heteroglossic meteorological text
South in 6th day of deadly heat
From Chicago Tribune wires
ATLANTA Churches offered air-conditioned refuge for poor people,
Alabama provided fans to poultry fanners and some merchants ran
out of ice Thursday as Southerners baked and crops wilted under a
killer heat wave for a sixth straight day.
But forecasters promised that relief

from the 100-degree-plus

temperatures was just a few days away.

Characterized in terms of the general domains of experience construed by figures,


weather forecasts fall within being & having and this is one respect in which they
differ from the last text quoted above (e.g. forecasters promised ...). This does not mean
that they are completely static there are phases of being (coming to be, etc.); nor that
there is no sense of causation fronts, weather systems, and pressures can serve as
Agent. However, the basic motif is one of states of being without any external causes;
put in terms of scale, a weather forecast construes a macro-being the meteorological
state of the environment consisting of a number of atomic micro-beings which move
in and out of existence of their own accord. The domain construed in recipes, which we
turn to now, is at another pole of figures one of doing, where a human acts to impact
on the world (within the confines of the kitchen).
8.3. The language of recipes
Having satisfied ourselves that today's skies will be sunny, we can worry about today's
food. We need a plan, a recipe. Breakfast seems the natural place to start
1. Put nutritious Kellogg's Sultana Bran into a clean bowl.
2. Pour fresh milk.
3 . Ready to serve,
4. Eat and enjoy Kellogg's Sultana Bran.

356

8. Building an ideation base

A recipe is a kind of procedural text a sequence of operations for arriving at some


well-defined end result, such as an assembled piece of furniture, a smoothly running car, a
well-kept garden or a dish (e.g. Martin, 1985: 5-6; Longacre, 1974, 1976). The notion of
procedure can be taken as a summary gloss of the domain; it is a macro-operation,
consisting of a number of atomic micro-operations. It is dominated by procedures
(algorithms, figures of doing) that lead to some specific goal: the dominant cause is
purposive. An agent or agents try to produce, assemble, repair, etc..
From the interpersonal point of view, the notion of procedure can also be taken as a
gloss on the interaction between the writer and the reader, it is a macro-proposal,
consisting of a number of instructions or directions to the reader.
8.3.1 The specific domain of a recipe
Before interpreting the culinary field in terms of our general taxonomy, we will present it
first in terms of its own register-specific organization. In a recipe, we find representations
of the cook, food items (including both dishes and their ingredient parts: the butter, the
onion, the stock ...)), utensils (with a knife, ...) and other tools, time units (20-30
minutes, ...), culinary happenings that occur without the cook's intervention, culinary
operations where the cook acts on food items; and culinary ascriptions where a food item
is assigned to a class of dish or to a state of consistency, colour, etc. (tender, brown, ...).
There is also a component of judgement, either personal, the cook's preference, or
impersonal, the desirability or necessity of an act. Consider the following example:4
Hungarian Potatoes
[ingredients]
Melt the butter in a saucepan and add the onion; cook gently until
soft. Add the stock and the pure. Bring to the boil and add the
potato cut into 1/2 - 3/4 inch dice. Add seasoning and paprika,
cover and cook gently until tender, about 20-30 minutes. Taste and
adjust the seasoning if necessary. Serve with garnish. (H & H, p.

192)
Roast Artichokes
Place some oil or vegetable/ nut fat in a roas ting-tin and heat to
about 400 F. Place peeled artichokes in tins and roast until

For our examples, we use N.B. Highton & R.B. Highton, 1964, The Home Book of
Vegetarian Cookery. London: Faber and Faber [H & H].

8.3 The language of recipes

357

cooked. They will become discoloured but the flavour is very good.
You nay prefer to roast them around a nut or similar savoury. If
they are roasted in their skins, the skins are uneatable as they
become bitter but the flesh is good. (H&H,p. 138)
Artichoke Chips
Peel artichokes, cut into thin slices and place immediately in
acidulated water - leave there for 20-30 minutes and then pat dry
with a clean absorbent cloth (a clean drying-up cloth is good but
should be cleaned after use) . Fry immediately in deep fat or oil.
Serve at once. (H & H, p. 138)
Asparagus Rolls
Prepare some small rolls (from your favourite bread recipe) well
cooked and crusty, remove from the oven and slice off the tops.
Scoop out the inside and fill with asparagus tips and mousseline
sauce. Replace the lids of the rolls and warm in the oven for 1015 minutes. This is a good savoury, either for the beginning or
end of a meal, or as a supper dish. (H & H, p. 139)
As far as figures are concerned, there aie some states of culinary being until tender,
until cooked, but the flavour is very good. Here an Attribute of some material quality is
ascribed to an ingredient or whole dish serving as Carrier. (There is also a minor type
with a modality as Attribute and a typically implicit figure as Carrier if necessary. )
However, operations changing the state of what is being prepared tend to dominate
then pat dry with a clean absorbent cloth, cook gently, leave there for 20-30 minutes.
Here the Medium is food, an ingredient or the emerging dish, and the Agent is the cook.
The result of the change may be construed as an Attribute characterizing the Medium
then pat dry, cut into thin slices; or the Attribute may construe a condition of the
Medium Serve hot

(not illustrated in the recipes quoted above). The likely

circumstances are Place (in tins, in the oven), Manner-quality (gently), Manner-means
(either equipment: with a clean absorbent cloth, or ingredient: with asparagus tips d
mousseline sauce\ Duration (for 20-30 minutes). The groupings of the classes just
mentioned and illustrated are shown in Figure 8-10. It also shows the value restrictions
on Medium and Agent as examples of such restrictions in the culinary domain.

8. Building an ideation base

358

Fig. 8-10: Taxonomy specific to the culinary domain (in recipes)


Interestingly, the culinary domain seems to be largely congruent in recipes; there is
very little grammatical metaphor, if any. If we consider recipes written for children where
grammatical metaphor is likely to be avoided (cf. Chapter 6, Section 6.8), we find that
they are quite similar to adult recipes in this respect; they are also congruent for
example:
Grilled Cheese and Tomato Sandwich
Preheat broiler.
On a piece of Whole-Wheat Bread (page 51), spread slices of any
cheese you like. Sprinkle a little dill weed, salt, and onion
powder on the cheese. Place slices of tomato on top of the cheese
and cover with another piece of bread.
Put sandwich in the broiler, 2-3 inches from the flames or coil.
Watch carefully and toast until bread turns golden brown and
cheese begins to melt. Turn sandwich over and repeat process.
(Sharon Cadwallader, Cooking Adventures for Kids. Boston: Houghton Mifflin
Company.)
There may be differences in the sophistication of the dish and type of instructions
(controlling behaviour, e.g., Watch carefully ); but there is no perceptible difference in the
degree of metaphor between the recipes for kids exemplified here and adult recipes both

8.3 The language of recipes

359

types are quite congruent In this respect, they contrast with texts concerned with
scientific exposition for children and for adults.
The domain-specific taxonomy in Figure 8-10 can be contrasted with other domainspecific taxonomies (cf. Section 8.2 above); we will return to this comparison. The point
is that the domain taxonomy above is incomplete in relation to the overall ideational
potential and that it foregrounds certain distinctions. We will now turn to its relationship
to the general taxonomy.
8.3.2 Relation to general taxonomy: overview
How does our domain-specific taxonomy differ nom the general one presented in Pan H?
We shall approach this question from the point of view of the breadth of the culinary
taxonomy, its depth, its partialness, and its favoured types. Obviously, the culinary
taxonomy differs from the general one in terms of breadth; there are general categories we
simply do not find in the culinary taxonomy. For example, among the figures, there are
no figures of saying at all. But, more important than the breadth per se is the nature of
the categories we do not find. For example, there are no metathings no facts, ideas, or
locutions. Things are not meta-; they are predominantly concrete objects.
The narrowed scope in the culinary taxonomy also means that very general categories
such as 'doing', 'being', and 'object1 can be given correspondingly narrower interpretations.
For example, in this register, 'being* means 'intensive & ascriptive being', since there are
no other subtypes; and 'object1 means 'concrete object1, since there are no abstract ones.
Similarly, 'doing' means 'concrete doing'. The restricted extension of subtypes is
illustrated in Figure 8-11 for figures of being.

Fig. 8-11: Restricted extension of subtypes of being

360

8. Building an ideation base

As we can see from the illustration above, another aspect of the narrowed breadth is
that the taxonomy is telescoped or truncated in intermediate delicacy. Thus 'being' can be
differentiated as "be1 and 'make1 without the intermediate steps 'intensive', 'asoiptive',
'non-phased1, and 'real'. (There is a price to be paid for this reduction, of course, since
information about >e' in the paradigm of figures of being is lost.) Similarly, we can go
from 'object' to 'cook' in one step; since there is only one kind of person, it is not
necessary to include intermediate steps in a taxonomy of persons.
There is another important consequence of the absence of certain subtypes: not
surprisingly, it affects the organization of the (field-specific) taxonomy. In the general
case, agency and process type are parallel, but in the culinary taxonomy distinctions in
agency are only evident in figures of doing; sensing and being are both middle (that is, no
agency feature can be present). As a result, two field-specific alternatives seem justified,
one with process type as the primary distinction and the other with agency as the primary
distinction: see Figure 8-12.

Fig. 8-12: Alternative organizations in restricted field


Taxonomies (A) and (B) differ but they constitute partial views of our general
taxonomy: they give a glimpse of the whole system, but not enough to indicate that
agency and process types are independent variables. Both models look as if sensing 2nd
being & having processes cannot be effected.
In (A), processes are of the physical world (= culinary world; the physical universe
the kitchen), of the world of consciousness (= preference/desire), of the world

361

8.3 The language of recipes

of being (~ membership); and if they are of the physical world, then either
spontaneous (happening) or brought about by an agent (doing to/with).
In (B), processes are either caused physical (= culinary) operations or else
spontaneous, in which case they occur in the world of physical (= culinary)
reality, consciousness, or abstract relations.
In either case, the notion of agency is quite narrow. While, in the overall ideational
potential, agents of processes can be metathings as well as things and, within things,
abstractions as well as physical objects, tools, animals, persons, and natural forces, the
culinary notion of agent is restricted

(Value restricted1) to be a person, more

specifically, a(n apprentice) cook: see Figure 8-13.


So, for example, tools do not as a rule figure as agents in the register of recipes, only
as means, enabling the occurrence of the process:
[Means:] With a sharp knife, [Agent] 'you 1 cut . . .
cf. the general case:
The knife cut my hand [BUT, Highton & Highton, p. 54: The point of

the

knife must pierce the centre of the grapefruit on each insertion]

Fig. 8-13: The culinary notion of agent

362

8. Building an ideation base

But th notion of agency in the culinary world is nairow not only in terms of the
spectrum of possible Agents, but also in terms of the directness and nature of the Agent's
involvement

in the process. While in the general case, agency is not inherently

intentional (volitional), here it is: 'doing to/with' means 'doing to/with intentionally'.
This is reinforced by the range of interpersonal options: volitional Agent = addressee
responsible for carrying out instruction. The Agent of chopping, frying, adding,
mixing, mashing, &c. brings about the occurrence of these processes intentionally and is,
moreover, assigned the responsibility by the writer of the recipe. (The writer and the
reader are far apart in time & space; a written recipe is very different in this respect from
spoken instructions where joint participation is possible, as in let's 'you +11.)
Related to the intentionality of the Agent is the directness of its participation in the
process; in the culinary world we have a kind of 'hands-on* agency. In contrast, agency in
general can be quite indirect for example: we dress our dancers in white.
When we compared the breadth of the culinary taxonomy with that of the general one,
we saw that there are certain semantic types missing e.g. figures of saying and
metathings. But among those that are present, there may be considerable variation in
frequency: some types are marginal they are almost missing, while other types are
favoured and predominate throughout the text. We can see this difference both in the
number of subtypes in the registerial domain and in the number of tokens in the recipes
themselves. For example, consider the first cuts in process types, shown again in Figure
8-14.

Fig. 8-14: The most general types of figure


Among these, doings dominate both in terms of tokens in a recipe (they are
instantiated in every step in the method stage) and in terms of the elaboration of the
subtaxonomy: frying, baking, mashing, mixing, adding, peeling, scooping out, rinsing,
etc. applied to the various ingredients. There are very few tokens of figures of sensing,
and their subtaxonomy is not elaborated (the only subtype is that of 'liking'). Figures of
being are more common, but their subtaxonomy is not elaborated. We can summarize
these facts graphically as in Figure 8-15.

363

8.3 The language of recipes

Fig. 8-15: Generality of figure types


Table 8(7): Types of phenomena in culinary domain
EXAMPLE
chopping, mashing,
adding, removing ...
simmering, boiling ...

TYPE
figures

doing

sensing

saying
being

elements

non-projected
projected
(fact// idea/
locution)
participants

doing (to/with)
happening
perceiving
thinking
desiring
feeling

wanting, liking

being, making

intensive
possessive
circumstantial

things

qualities

circumstances

conscious
nonconscious

of
projection
of
expansion

'cook'
animals
objects
[material]
substances
abstractions
[material]
institutions
objects
[sernioticl
abstractions
[semiotic]

onion, leek, fillet,


cutlet ...
stock, flower, ham ...
temperature, colour ...

sensemeasure

brown, hot ...


in an oven, on a plate,
for 30 minutes

The diagram serves to correct the earlier categorial picture by bringing out the relative
prominence of the different types (see further Chapter 13, Section 13.3 below on the
probabisc interpretation of the system). By introducing the concept of probability, we

364

8. Building an ideation base

are able to order the different types according to their contribution to the domain. By the
same token, a type that is not represented may be not totally impossible but just highly
improbable. For example, figures of saying are not inconceivable in the world of recipes;
they are improbable, but could occasionally occur:
may we suggest that you add a clove of garlic . ..

When we show doing as the dominant process type, we have given substance to our
initial observation that recipes are macro-doings (macro-operations).
Among the simple things, material objects dominate. There is only one kind of
person, the cook; this is a world with only one Agent. There are virtually no abstractions
(e.g. 'flavour1, 'colour', 'temperature'), only concrete material objects and substances.
There are few qualities and circumstances. Qualities are either "material" (object qualities:
texture, colour, and consistency) or "modal" (act qualities: necessity and desirability).
Table 8(7) above presents a rough summary of the observations made so far about
restrictions on the general taxonomy.
Having presented this general overview, we can now discuss the culinary field in some
more detail. Sequences will serve as our starting point.
8.3.3 Sequences in the culinary domain

In the semantic system in general, sequences are either projections or expansions; but in
the culinary domain only expansions occur.5 Expansions are themselves fairly restricted.
The favourite type is temporal enhancement with equal figures, which construes
(sub)procedures as sequences of culinary operations. Enhancements may also be unequal,
expressing either temporal limit or condition, in which case the conditioning figure is
essentially restricted to being a non-operation, i.e. a culinary happening, a sensing or a
being.
Examples:
Elaboration:
This vegetable is not very popular;
> it has a distinctive sweetish flavour, and
like brussels sprouts and celery, it is the betterfor having been frosted.

There are occasional instances which could be interpreted as projection within the Process of
a figure (cf. Halliday, 1985: Ch. 7 Additional, Section 6). They serve to indicate a possible
alternative that may be preferred, e.g. may prefer to roast in the example You may prefer to
roast them around a nut or similar savoury.

365

8.3 The language of recipes

Extension:
adversative: They mil become discoloured

bul the flavour is very good.

Enhancements:
and ['then'] warm in the

temporal: succession & equal: Replace the lids


oven for 10-25 minutes.

temporal: point & unequal: When the pan and fat are hot,

pour in the

egg mixture, stir slowly with a fork...


temporal: limit & unequal:
qualitative limit: roast

until cooked; cook gently

bottom is set and light brown; cook

until the

until nearly tender but not

brown.
happening limit: (i) 'until1 place in a hot oven 375 F. 425 F.

until

until the oil is just smoking,


simmering well; heat
add 2-3 tablespoons of cooked sweet corn to the egg mixture

before cooking.
causal-conditional & unequal:
qualitative limit: If the pastry tends to brown too much on top,
cover with a greased paper.
preference limit toss the artichokes in butter, chopped parsley and seasoning
if so desired; if you prefer not to mix proteins,
ifsodesired;ifyoupref rnot omixproteins,
souffls flavoured with vegetables can easily be made.
causal-purposive & unequal: stir slowly with a fork

to allow the liquid

to run on to the bottom and set; the omelette should be cooked as quickly as
so that the egg does not become tough and leathery.
possible
8.3.4

Figures in the culinary domain

The dominant type of culinary figure is an operation by the cook on a food item,
potentially with a means and a duration it is a type of doing to/with (as opposed to
happening). The Means is some kind of equipment and the duration is a quantified unit of
time. The roles of this figure serve to differentiate a number of different kinds of objects,
viz. [person:] cook (filling the Agent/Actor role), [concrete:! food item (Medium/Goal),

366

8. Building an ideation base

[concrete:] equipment (Means), and [abstract] time-unit (Duration). The mterdependence


between the figure taxonomy and the element taxonomy is a specific example of what
was observed to be a quite general tendency above in Pan . The diagrammatic
representation is given in Figure 8-16 below.
Examples of lexicogrammatically represented subtypes of this culinary operation
include:
wash and peel artichokes
cook the washed apricots for 5-10 minutes
with a knife make a cut
mash the potatoes with a potato-masher
add the tomatoes with a little seasoning
garnish with plenty of onion rings
This culinary operation gives us a sense of what the culinary world is like. Hiere is
one Agent in control of the world, the cook;6 he or she only participates in an operation
as the agent In other words, the cook is always the doer, never the done to or with.
Similarly, there is only one kind of Medium, food items. These two kinds of things are
thus sharply distinct from the point of view of what participant roles they take on. There
are two other classes of things, items of equipment and time-units. The former are almost
always the Means in a process rather than serving a participant role; they never compete
with the cook for the Agent role, as tools and instruments may do in other registers.
Time-units in the world of food preparation never serve as participants in a figure; they
are only involved indirectly as circumstances of Duration.
Now, it is perfectly possible to conceive of a world where the cook is not in sole
control of culinary processes; but this would not be the world of recipes (it may be more

On occasion, an implement may serve as Agent, as in the point of the knife must pierce the
centre of the grapefruit on each insertion. This is agnate with You must pierce the centre of
the grapefruit with the point of the knife on each insertion, with the cook as Agent and the
implement as Manner-means.

8.3 The language of recipes

367

Fig. 8-16: Network showing typical culinary operation


Although culinary operations dominate, there are figures of other kinds also: culinary
happenings, liking/wanting, and ascriptions. To give a sense of the range of possible
figures, we list the total set of primary types below, including both the subtypes of

8. Building an ideation base

368

culinary operation and other types of doing, sensing, and being & having with examples
of processes tilling the Process role (taken from Highton & Highton, 1964: 55-7).
1. doing & happening:
prepare, finish,
use,
cut, halve, cube, slice, dice, chop, serrate, pierce, break, chew
grate,
remove, free, take away, leave, drain,
move, put, return,
add, mix, join
sprinkle, spread, pour, pack,
garnish, dress, fill, stuff, coat
grill, cook,
chill,
serve,
allow
(cut... into cubes : cube; cut... in half : halve)
2. sensing:
desire
3. being & having:
intensive: be (attractive, underdone, soft, better), make
circumstantial: replace
Next, we will also list examples of figures as they are realized in running text with
fillers of process, participant and circumstance roles (H & H, pp. 54, 55, 138): see Table
8(8).

Table 8(8): Examples of processes, participants and circumstances

in

figures
Process

halve
allow
remove

cut

participant roles

drcnmsUnce roles

Mejora

Manner

other
partkip.
roles

quality

the grapefruit
Benef.:
one naif
for each
person
raps
ttne
grapefruit]

remove
move

knife

free
remove

each section
theskio

Place
means

around
a
grapefr the
nit-knife centre
up and
down with a
sawing
motion

Purpose

Accomp.

Role

369

8.3 The language of recipes

sprinkle

grapefruit

add

chill

kirsch
grapefruit
a half cherry

chill

cut

aie melon
[the melon]

remove
serve

the seeds
[the melon]

may be
used
to replace
serve

lemon
wedges
the ginger
[the melon]

wash and
peel
place

artichokes

add

herbs or a
bouquet garni
[dish]

put

place
until
simmering
reduce
remove
drain
toss
drain
keep

use
cover
serve

with a
little
brown
sugar
in the
centre

into
wedge
shapes
with
powdered
or chopped
ginger and
brown
sugar

on cold
plates

[artichokes]

in a hot
oven ...

well

[it]
heat
bouquet ganu
the
artichokes
the
artichokes
the
artichokes
[the
artichokes]
the stock

with a
squeeze of
lemon
juice and
some fried
onions

ma
boiling
white
stock

to 300F.350F.

in butter,

Range:
warm

the
artichokes
[the dish]

for a
sauce

with it
in the
casserole

Culinary happenings are processes that involve food items as mediums; they go
on in time without the cook's intervention, primarily cooking, boiling, simmering,
evaporating, smoking, and coming away. In recipes they are temporally or causally related
as conditions on processes that are controlled by the cook; e.g.
... until (it) has evaporated
... if (they) come away
... until (it) begins to smoke
.. . until the mixture comes away from the sides
... so that (it) does not boil

370

8. Building an ideation base

Sensings are mainly of the desire type, more particularly 'liking1, with the cook as
the Senser and some kind or aspect of a dish as the Phenomenon. They are related to
culinary operations in conditional enhancing sequences and can save to condition an
alternative method:
... if you want a more substantial stuffing
... if you would like a mere hint instead
(the rings can be egg-and-breadcnnribed) if you prefer

Figures of being & having are ascriptive, either ascribing a quality to an


ingredient, an act or a dish, or assigning a dish to a particular class.
... but the flavour is very good
... until ( they) are brown
... until the pastry is snooth and pliable
8.3.5 Simple things in the culinary domain

Simple things can be differentiated according to the roles they may play in figures, as we
have already seen in Figure 8-16 above. Thus we have:
role type
participant roles
circumstantial roles

role
Agent/ Actor
Medium/ Goal or Actor
Range/ Attribute
Manner-means (of doing)
Manner-means (of doing)
Duration (of doing)

filler
cook
food
class [of dish]
item of equipment (i.e.
implement)
speed, degree, etc.
time

These correspondences indicate how the culinary world is constrained in terms of the
nature of participants and circumstances; in terms of who can be expected to do what to
whom, and how. When we check whether humans and non-humans can participate as
Medium and/or Agent in figures of doing, these constraints are brought out quite clearly:
see Table 8(9).
This concludes our illustration of domain models and their relation to the meaning
base we presented in Part II. Adding such domain models is a necessary step towards
having a functional text processing system. But their importance goes beyond their role
in text generation and parsing. On the one hand, it is likely that in most task-oriented
descriptions of language, whether in natural language processing or in other fields such as
education, language disorders, legal practice and the like, domains of this intermediate
scope will typically be the most relevant; while on the other hand, in our view, the
modelling of specific domains of ideational meaning is an essential component of our
theoretical understanding of the nature of human cognition.

8.3 The language of recipes

371

Table 8(9): Figures of doing, showing types of Agent and Medium found in culinary
domain (italicized)

9. Using the ideation base in text processing


In this chapter we shall go outside the ideation base and locate it in relation to other parts
of a model of the linguistic system as a resource for meaning in context. We shall look at
the kind of generation system that seems technically manageable at the present stage of
development.

9.1 Organization of a generation system


9.1.1 Relationship to a client system
The need to present information in the form of text (or indeed in the form of graphs,
tables, maps, etc.) usually arises in a computational system that was not developed for
the purposes of generating text in the first place. Most typically, this is a database
system or an expert system; we can call it the client system since it will engage the
services of, and define the required output of, the generation system to be developed:1 see
the Figure 9-1 below. The client system will provide some 'raw1 information source
the data base of a data base system (e.g., containing information about the weather) or the
reasoning of an expert system (in some form, e.g. as the rules of the system and traces of
the execution of these rules).
The central 'module* to be built is the model of the linguistic system itself; this is
where the ideation base is located. In addition, we shall need to construct a model of the
context in which the linguistic system and the client system are embedded. The context
will serve to co-ordinate the client system and the linguistic system and it will support
the generation process.
To accommodate the client system, we have to undertake both contextual modelling and
linguistic modelling. The need for text and the raw information source have to be
interpreted in this context in terms of field, tenor and mode (see Chapter 8, Section 8.1.1
above). Furthermore, the raw information source has to be construed in terms of a domain
model capable of being verbalized. That is, whatever the demands on the information
source are in the client system, they do not include the demands placed on it by a
linguistic generator; it has to be transformed from a data base to a meaning base.2 Once

Or if such a generation system already exists, the task is to slot the client system into its
context and to 'customize' the generation system to allow it to interact with the client
system.

We assume that the client system has not been conceived of, and modelled as, a languagebased system. However, it follows from our approach that we would see them as languagebased. Thus reasoning in the expert system would be modelled as semantic processes,
which would make it possible to build in more of the power of language in an explicit
way: see Halliday (1995). This is also the approach being developed by Michio Sugeno
and his researchers under the heading of intelligent computing.

9.1 Organization of a generation system

373

information from the database has been transformed into meaning, it can be fused with
the meanings of the ideation base, more specifically the meanings of the relevant domain
model (cf. Sugeno, 1993; Kobayashi, 1995). This transformation involves classifying the
objects and relations of the data base within a domain model and then locating this
domain model on the cline of instantiation relative to the overall ideation base as
described in Chapter 8, Section 8.1.2 above (see Figure 8-3). The transformation may
also involve adding temporal features where these are not part of the information encoded
in the data base.

Fig. 9-1: A text-generation system embedded in context together with a client


system providing the raw information source
The development of the overall generation system will thus include 'growing1 a
linguistic component as well as revising and expanding what has been imported from the
client system into its ideation base. This expansion will thus take place stratally within
the semantics and metafunctionally within the ideational metafunction. In addition, the
generation system will have to include subsystems that are not likely to have any seeds in
the client system the lower strata of lexicogrammar and phonology (graphology), of
course, but also the non-ideational parts of the overall meaning base the interaction

374

9. Using the ideation base in text processing

base and the text base at the semantic stratum. The system will have to have a model of
how to interact with its addressee (the interaction base, relating to tenor within context)
and also a model of how to present ideational and interpersonal meanings as text in
context (the text base, relating to mode).
If the task is to produce spoken text, the text generator has to be miked up with a
speech synthesizer through an interface, as shown in Figure 9-1 above. We can then see
text generation as a bridge between two independently existing systems, the client system
a database, expert system, etc. developed within computer science and the 'output
system* the speech synthesizer developed within electrical engineering. These two
systems interface with the linguistic system at its two outer strata: the client system
interfaces with semantic stratum and the speech synthesizer interfaces with the
phonological stratum. In this case, a text generation system is doubly accountable
both upwards and downwards.
9.1.2

Generation system

Let's now turn to the question of how a generation system is organized so as to generate
text We will discuss this basically at the level of theoretical specification (cf. Chapter 1,
Section 1.9.1). The implementation with which we are most familiar is that of the
Penman system, which realizes a number of the design properties to be mentioned. We
will draw on this system in various parts of our discussion; but our focus is on the
theoretical considerations that have guided the linguistic aspects of its development, and
can serve to guide developments in the future. The Penman system has been discussed in
various places, e.g. Mann (1982), which is an early design; Mann & Matthiessen (1985),
Matthiessen & Bateman (1991); Hovy (1988a, 1991); and a multilingual extension
developed by a collaboration between Sydney and Darmstadt is discussed in e.g. Bateman,
Matthiessen, Nanri & Zeng (1991), Matthiessen, Nanri & Zeng (1991), and Bateman,
Matthiessen & Zeng (forthc.). Other current systemic text-generation systems are
COMMUNAL, developed by Robin Fawcett and his team (see e.g. Fawcett (1988b) and
Fawcett, Tucker & Lin (1992), and Fawcett (1981) for an early model of systemic
generation), KPML (Komet-Penman multilingual), as a develpment of Penman together
with the Komet system, developed by Bateman (1996) and his team (see also Teich,
1995; Teich & Bateman, 1994); MULTEX, a multilingual systemic generator developed
by Zeng (e.g. 1993,1996) as an alternative to the multilingual version of Penman within
the multilingual project at Macquarie University in Sydney, and ODonnell's (1994)
WAG system for sentence analysis and generation.

9.1 Organization of a generation system

375

A text-generation system can be modelled in terms of the overall theoretical framework


we sketched in Chapter 8, Section 8.1. Most fundamentally, it can be modelled as
stratified resources extended along tbe cune of instantiation from overall potential to
particular instance: see Figure 8-3 above.
9.1.2.1 Stratification of resources
Let us first review the strata! organization of the resources. The resources are stratified
into context and language; language is in turn stratified into semantics, lexicogrammar
and phonology/ graphology (see Chapter 1, Section 1.2, Figure 1-1); and as we shall see,
the processes of instantiation are guided by this stratification. Stratification simplifies the
global organization of a generation system since it introduces the strata as more local
organizational domains; it is a way of managing complexity. For instance, instead of one
large unwieldy semanticogrammatical system, there are two strata semantics and
lexigrammar.
At the same time, this strata] organization means that it is crucial to specify the
realizational relations between strata inter-stratal realization. In systemic theory,
this relationship is stated in terms of the organization of the higher stratum for a
simple reason: a higher stratum provides a more comprehensive environment than a lower
one (as our strata! figures with concentric circles suggest). For instance, a certain semantic
feature such as 'probability* may be dispersed in realization throughout various regions of
the grammar (/ think; in my opinion; probably; will). More specifically, inter-stratal
realization is specified by means of inter-stratal preselection: contextual features arc
realized by preselection within the semantic system, semantic features are realized by
preselection within the lexicogrammatical system, and lexicogrammatical features aie
realized by preselection within the phonological/ graphological system. This type of
preselection may take different forms between different strata! boundaries, but the principle
is quite general.
[1] Context. Context is the 'semiotic environment1 of language (and other sociosemiotic systems such as image systems [maps, diagrams, etc.]); its systems specify what
demands may be placed on language and what role it may play in responding to those
demands. There are three sets of contextual systems field, tenor and mode; we gave an
example of a fragment from field in Figure 8-4 above. Recurrent combinations of (ranges
of) field, tenor and mode values define regions within the overall system of the context of
culture. In most cases, text generation will take place within such a contextual region

376

9. Using the ideation base in text processing

within a range of field, tenor and mode values. For example, weather forecasts ate
generated within the contextual region characterized in Chapter 8, Section 8.2.1 above.
[1/2]

Realization of context in semantics. Such contextual regions evolve

together with special functional varieties of the language or "registers". These are, in the
first instance, registerial subsystems of the semantic system. A register is a semantic
region within the overall semantic space. It is made up of contributions from all three
metafunctions: a domain within the ideation base and similar regions within the
interpersonal and textual parts of the overall semantic space. Our specifications of the
meteorological and culinary domains constitute the ideational aspects of the registers of
weather forecasting and recipes.
A given contextual specification of field, tenor and mode is thus realized by
"preselecting" a register within the semantic system; Figure 8-1 above shows ideational
domains being preselected by different fields.3 This contextual preselection within the
semantics narrows down the overall potential to a registerial subpotential; a field
specification narrows down the ideational potential from that of the overall ideation base
to that of a domain model.
[2] Semantics meaning base. The semantic system, or "meaning base", thus
consists of the three familiar metafunctional contributions, first introduced in Chapter 1,
Section 1.3: the ideation base, the pan of the meaning base that we have focussed on, but
also the other two bases, the interaction base and the text base. All three bases ate
extended along the cline of instantiation from potential to instance. As we have seen
(Chapter 8), the ideation base includes both the most general ideational meaning potential
and a repertoire of domain models located along the cline of instantiation midway between
potential and instance and associated with particular ranges of values within field. Let us
say a few more words about (i) the interaction base and (ii) the text base. Table 9(1)
summarizes those aspects of the three bases to be discussed.
(i) The interaction base provides the resources for enacting social roles and
relations as meaning, (prototypically) in dialogue. The interaction base includes the
strategies for adopting and assigning speech roles, for giving and demanding assessments,
and the like. The locus of these strategies is a unit of interaction or move. A move is

If we introduce time, we can think of the context as activating the register: cf. O'Donnell,
1990, on the notion of activation.

377

9.1 Organization of a generation system

typically mapped onto a gure from the ideation base: a speaker construes a quantum of
experience as a fgure and enacts this figure as a move in dialogue, either as a proposition
or as a proposal. This mapping between fgure and move is a central feature of the way we
jointly construct and negotiate experience.
Table 9(1): The metafunctional diversification of the meaning base

upwards: context
mode of meaning

semantic units

register variant

ideation base
field
construing
(experience)
sequence
figure
[quantum of change]
element
domain models

text base
mode
creating (ideational
and interpersonal
meanings) as
information

interaction base
tenor
enacting (social roles &
relations)

message
[quantum of
information]

move
[quantum of
interaction]
exchange relationships

The interaction base extends along the cune of instantiation. (1) At the potential end
of the cline of instantiation, the interpersonal strategies that have the move as their
domain are defined by all the options persons have in exchanging meanings with one
another, adopting speech roles of giving/ demanding information or goods-&-services and
assigning complementary roles of accepting or giving on demand to the addressee. These
constitute all the patterns of interaction within a culture. (2) Midway between potential
and instance, sets of such strategies cluster within ranges of tenor values. Such a cluster
is the interpersonal analogue of a domain in the ideation base: it is a region within the
overall interpersonal space of meaning, selected according to tenor, just as a domain is a
region within the overall ideational space of meaning, selected according to field. The
options in interpersonal meaning that make up the cluster together enact a tenor
relationship, such as that between a client and a server (unequal in power, low in
familiarity, neutral in affect) where the client initiates demands for goods-&-services and
information about them and the server responds, such as that between two friends (equal in
power, high in familiarity, positive in affect) where the interpersonal options are wideranging, or such as that between the writer of a weather report and the readers (unequal in
expertise, no familiarity, neutral in affect) where the writer gives information and the
readers accept We might call such a cluster an exchange relationship to foreground
that it is semantic (i.e. constituted in meaning through exchanges of meaning) and that it
is interpersonal (rather than one-sidedly personal). To indicate that it is analogous to a
domain model, we might have called it an exchange or interaction "model"; but we have

378

9. Using the ideation base in text processing

avoided that term because it suggests a coastrual of something and construal is the
ideational mode of meaning it is more like a protocol than like a model.
A particular domain model within the ideation base correlates with a particular
exchange relationship within the interaction base and together they form the ideational and
interpersonal aspects of a register. An exchange relationship thus has ideational
implications: it involves the exchange of some ideational meanings rather than others and
it embodies a division of labour between the interactants in the exchange relationship. For
example, in a service encounter in a local shop (cf. Halliday & Hasan, 1985; Ventola,
1987), the customer may demand goods and information about goods within the relevant
domain and the server supplies these on demand and demands goods (payment) in return;
and this involves the general domain of business transaction and the particular domain of
the business (e.g. hardware). An exchange relationship thus gives interpersonal values to
meanings within the domain model it is associated with. For example, as we have seen,
the ideation base embodies both congruent and metaphorical constatais of experience; and
these variants will be selected partly according to differences in interpersonal distance
along dimensions such as age and expertise (cf. Chapter 6, Section 6.5.2 above).
The text base is oriented towards the ideation base and the interaction base. It
provides the resources for constructing meanings from these two bases as information of a
kind that can be shared as text. An ideational figure and an interpersonal move ate
constructed as information in the form of a message. Such a message is related to the
preceding discourse and differentiates informational statuses in terms of thematicity and
newsworthiness. From the speaker's point of view, the text base is a resource for
developing a text, message by message, and for guiding the listener in his/ her
interpretation of the text; and from the listener's point of view, it is a resource for
constructing such an interpretation (for building up an instantial system; see below). We
shall return to the text base and its relationship to the ideation base in some more detail
below, in Section 9.3.
[2/3] Realization of semantics in lexicogrammar. Semantic features are
realized by lexicogrammatical ones; we have illustrated this relationship within the
ideational metafunction at various points in our discussion (e.g. sequence ^ clause
complex; figure of doing * material clause). The realizational relationship between
semantics and lexicogrammar is one of preselection: semantic features such as 'sequence',
'figure', and 'doing1 are realized in lexicogrammar by means of prespecification of
lexicogrammatical information, most centrally preselection of lexicogrammatical features.

9.1 Organization of a generation system

379

For instance, 'doing' is realized by the preselection of the clause feature 'material1, which
means that the clause that realizes a figure of doing is constrained to be a material clause.
This type of approach was adopted by Patten (1988) in his generator SLANG, and has
also been developed for the multilingual version of Penman at Macquarie University (see
Zeng, 1996). Here the realization of semantics in lexicogrammar is modelled as a direct
one. However, researchers in text generation have developed an alternative way of
modelling the realization of semantics in lexicogrammar (see Matthiessen, 1990a, for a
comparison of the two approaches).4
In this approach, realization is mediated by an interface between semantics aid
lexicogrammar. Semantic features such as 'sequence1, 'figure', and 'doing' are first
expressed as specifications of various fields of information in a local plan for a clause
(complex). For, instance, 'doing* would be expressed as a particular type-specification for
the local dan field FIGURE (e.g., FIGURE = x/ doing, meaning the value is the
instantial figure 'x1 of type doing). This local plan spcification then supports the
computation of responses to inquiries presented by a special interface between the
lexicogrammar and the semantics, the chooser-&-inquiry interface. This approach was
developed for the Penman generator (e.g., Mann, 1983a,b; Matthiessen,

1988b;

Matthiessen & Bateman, 1991: Section 6.4).


These alternatives are contrasted diagrammatically in Figure 9-2. While direct
realization clearly is simpler in that it does not involve the addition of an interstratal
interface, mediated realization makes it easier to co-ordinate and integrate information from
various sources (ideational, interpersonal and textual). In particular, during that phase of
the development of the Penman system when the semantic environment of the
lexicogrammar was still being designed and specified, the mediated approach provided a
detailed set of demands for information to be accommodated within the semantic resources.
In our example in Section 9.2 below, we shall use mediated realization since it will be
most familiar from discussions of the Penman system.

Other mechanisms have also been used, for instance relations that can be interpreted as
inter-stratal pairings (Jacobs, 1985), semantic conditions on nik application (as in
augmented phrase structure grammar), unification of semantic and grammatical
representations (McKeown, 1982).

380

9. Using the ideation base in text processing

(i) direct inter-strata! realization


through interface

(ii) realization mediated

Fig. 9-2: Direct and mediated inter-stratal realization


[3] Lexicogrammar is the resource for construing meaning as wording. It is
organized as a set of system networks representing options in wording. Systemic options
may have realization statements associated with them; these statements specify how the
options are realized in wording (structure & grammatical/ lexical items). Figure 9-3 shows
a simplified version of the ideational and interpersonal systems of the clause.
The lexicogrammatical system networks are distributed by metafunction (ideational:
experiential & logical interpersonal textual) and by rank (clause group/ phrase
word morpheme) and extend in delicacy from grammar to lexis.5 As we have noted,

In systemic functional theory, lexis is thus interpreted as most delicate or specific


grammar (on the implications of this perspective for text generation, see Matthiessen,
1991b, and Cross, 1991, 1992, 1993). Grammar comprises syntax and morphology;
there is no strata! boundary between the two, but merely a move down the rank scale:
"syntax" is simply the grammar of clauses and groups/ phrases and morphology is the
grammar of words and morphemes. Systemic functional grammatical theory can be seen as
part of a (typological) family of grammars that have been used in computational
linguistics for various tasks. Winograd (1983) calls this family 'feature and function*
grammars since most members rely heavily on (declarative) representations involving
both features and functions. Another salient characteristic is the clear separation of
grammatical representations as declarative data structures and grammatical processes,
with unification as a central operation (see Shieber, 1986, on unification based grammars
and Kasper, 1988, on the inter-translatability of Systemic Functional Grammar and Kay's,
e.g. 1979, 1985, Functional Unification Grammar). Other members include Lexical
Functional Grammar, Functional Unification Grammar, Generalized Phrase Structure
Grammar, and Head-Driven Phrase Structure Grammar (HPSG). HPSG as developed by
Pollard & Sag (1987, 1993) is of particular comparative interest, since it includes typing
of feature structures and organization of types into subsumption lattices: this means that
the paradigmatic organization of grammar is foregrounded more explicitly than in
previous approaches within the generative tradition. Systemic Functional Grammar is

9.1 Organization of a generation system

381

the ideation base is realized by the ideational resources: sequences and figures are realized
at clause rank by clause complexes and (simple) clauses respectively; and elements aie
realized at group/ phrase rank. In a similar way, interpersonal meanings are realized by
interpersonal features in the lexicogrammar, and textual meanings by textual ones. The
grammar unifies the different metafunctional contributions; for example, a figure, a move
and a message are unified in their realization as a clause. It achieves this unification by
realizing combinations of ideational, interpersonal and textual features in the same
wording. For example, the wording well unfortunately they must have missed the train is
a realizational unification of a figure (of doing & happening), a move (of giving
information, assessed as certain and undesirable) and a message.
The wordings that are constructed by the lexicogrammatical system are in turn realized
in one or other of the two expression strata, phonology or graphology. Figure 94
represents the stratal organization of the resources, diversified into field, tenor and mode
(within context) and into ideational, interpersonal and textual (within the content strata of
language). As we have noted, these resources are extended along the cline of instantiation
from potential (language in context of culture) via subpotentials (registers in situation
types) to instances (texts in contexts of situation). Figure 9-4 shows the stratified
resources as being extendable along the cline of instantiation. Let us now map out the
overall semiotic space defined by stratification and instantiation, and then discuss how the
processes of instantiation move through this space. Table 9(2) represents the intersection
of instantiation and stratification, providing a schematic map of the overall semiotic space
(the table has been adapted from a fuller version presented in Halliday, 1995), We have
shown the ideational 'slice' of the overall picture in bold italics. (We have left the
phonological cells unspecified since phonology falls outside the scope of our discussion
of the ideation base in text generation; but the cline of instantiation is equally important
at this lowest stratum of language.)

unique in organizing features into choice networks, as the central organizing principle,
and in being multifunctional; the combination of these properties make it particularly
suited to text generation.

382

9. Using the ideation base in text processing

Fig. 9-3: Fragment of the clause grammar


9.1.2.2 Processes of instantiation
To the model summarized in Table 9(2), we can now add processes of instantiation: in
generation, they have to move from potential to instance (from system to text);
and in analysis, they have to move from instance to potential (from text to
system). These processes are explicitly located in time; they unfold in time, so the model
has to specify how they are sequenced relative to one another. (As we will suggest below,
in generation, options and realizations trad to be instantiated first at higher strata and later
at lower strata; but there is no strict temporal linearity.)

9.1 Organization of a generation system

383

Fig. 9-4: The stratal and metafunctional 'address' of the ideation base in the overall
resources
In generation, instantiation can in principle be initiated at any point along the cline of
instantiation. This is precisely the significance of the fact that instantiation is a cune.
Most likely a given generation task can be addressed in terms of an existing registerial
system, so instantiation can be initiated midway along the cline of instantiation. That is,
it is not necessary to select options from within the overall meaning potential, but only
from within a much more restricted registerial subpotential (cf. Patten, 1988, on the
significance of this restriction in a problem-solving approach to text generation).
Generation may also operate further down the cline of instantiation, with more local
subpotential within a registerial potential.

384

9. Using the ideation base in text processing

Table 9(2): Stratified resources extending along cline of instantiation

A text is generated within the logogenetic

time-frame (see Chapter 1, Section

1.5.3 above). In fact, generation is a logogenetic process: it creates meaning in the course
of instantiation as the text unfolds see Figure 9-5. If we look at logogenesis from the
point of view of the system (rather than from the point of view of each instance), we can
see that logogenesis builds up a version of the system that is particular to the text being
generated (cf. Butt, 1983, on "semantic drift" in unfolding text and Butt, 1987, on "latent
patterning"): the speaker/ writer uses this changing system as a resource hi creating the
text; and the listener/ reader has to reconstruct something like that system in the process
of interpreting the text with the changing system as a resource for the process of
interpretation. We can call this an instantial system (see Matthiessen, 1993c). For

9.1 Organization of a generation system

385

example, in the course of logogenesis, a recipe is built up as a series of ordered loops


through the system of sequences, whereas an encyclopaedic entry may be built up as a
systemic taxonomy, developed step by step in delicacy. An instantial system may fall
entirely within the registerial system it instantiates; in other words, the meanings created
within it may all have been created before. However, it may also create new meanings
new to the speaker and/or listener. In either case, the instantial system is built up
successively by the generation process; but as it is developed, it in turn becomes a
resource for further instantiation. We will illustrate this process in some detail below.

Fig, 9-5: Logogenesis


We had indicated how the process of instantiation relates to the cline of instantiation.
How does this process relate to the other global dimension of organization shown in
Figure 9-4 and set out in Table 9(2) above i.e. the dimension of stratification? The
strata are ordered in symbolic abstraction, but they are not ordered in instantiation time.

386

9. Using the ideation base in text processing

The process of instantiation can shunt up and down the stratal hierarchy. However, the
general tendency in instantiation is one of stratal desce: see Figure 9-6. First
systemic features are mgtimfiate^ (selected) at the highest stratum and their associated
realization statements are also instantiated (executed). Then the instantial specifications at
this stratum are realized at the stratum below. Within this overall stratal descent, there is
interleaving: higher-strata! systems need not be fully instantiated until lower-stratal ones
have been instantiated. This means that selections at higher strata can be made in the
logogenetic environment of preceding selections at lower strata (see Zeng, 1996, for more
detail on systemic theoretical considerations and their implementation in the MULTEX
generation system). This reflects the general theoretical principle that the relationship
between two strata is a solidary one, with instantiation proceeding "dialogically".

Fig. 9-6: Organization of system into resources and processes


For present purposes, it will be helpful to differentiate three major phases in the
process of generation, viz. (i) situatonal specification, where the text to be generated is
'situated* in an instantial context specified (up to a point) in terms of field, tenor and
mode, (ii) semantic generation, where the text as a whole is planned globally and more

9.1 Organization of a generation system

387

local text plans are 'spawned* [text as meaning], and (iii) lexicogrammatical generation,
where the semantic plans are realized in lexicogrammatical wordings [text as wording].
The two approaches to the mapping between semantics and lexicogrammar mentioned
above, direct and mediated realization, have different implications for the realization of
local semantic plans.
(i) If the realization is direct, the process will essentially be a backward chaining
traversal of the lexicogrammatical system network: preselections specify
delicate lexicogrammatical features and the main task of the generation
process is to infer the paths that lead to these features and to execute any
structural realization statements on that path.
(ii) If the realization is mediated, the process will essentially be one of forward
chaining traversal of the lexicogrammatical system network, with activation
of the chooser of each system for considtating the local plan to obtain the
semantic information needed to make the choice.
We will now illustrate how a generation system may operate by looking at these three
major phases in the generation process.

9.2 An example from recipe generation


As our illustrative generation task, we shall take the following recipe for Almond and
Apricot Stuffing (Highton & Highton, p. 127):
ALMOND AND APRICOT STUFFING
[Ingredient* :]

1 02. butter or suenut


2 large chopped onions
4 oz. coarsely grated or chopped almonds
6 fresh or about 12 soaked apricots, diced
finely
1 tablespoon chopped fresh herbs
Grated zest and juice of 1/2 lemon
Sea salt and brown sugar to taste
[Preparation:] (1) Melt the fat and (2) fry the onions (3)
until slightly brown, (4) add the alitonds and

(5) continue

388

9. Using the ideation base in text processing

cooking for 3-5 minutes. (6) Remove from the heat and (7) add
the remainder of the ingredients.
[U***] (8) This makes a very good stuffing for cucumbers.
[Variation:]

(9)If you want a more substantial

stuffing

(10) add a little mashed potato and [Alternative ut] (11)


use for stuffing marrows.
9.2.1 Contextual specification
The recipe generator has the contextual resources to plan and organize a recipe,
giving it the text structure
Preparation

Use

Variation

indicated in bold face in the text above: Ingredients


A

Alternative use. This structure is projected from the

context of situation. The text is largely organized on the model of the food preparation
activity itself, or, more generally, on the model of purposeful behaviour aimed at solving
a problem. The problem is the making of the dish. The text starts out by specifying the
preconditions for solving the problem (Ingredients). It then specifies a step by step
algorithm for the purpose (the steps in Preparation). This algorithm is an abstract or
summary of the fully detailed procedures represented in the culinary domain. For example,
the text reads Melt the fat and fry the onions until slightly brown, leaving out the
specification of where the fat is melted, the step of placing the onions in the frying pan
with the melted fat, and so on, since an adult reader can be expected to Till in' these steps
and specifications. Once the basic algorithm has been introduced, a variation of it can be
given (Variation). In addition, there is a specification of the use of the dish created by the
method (and then again of the alternative use of the version of the dish created by the
alternative method). There is no strong general reason why Use should follow Preparation
in the recipe, since it is not pan of or dependent on the procedure; and in many recipes it
precedes Ingredients.
The generic structure potential is diagrammed below (we have used Hasan's,
conventions first (Hasan, 1978; Hasan in Halliday & Hasan, 1985): optionality is
indicated by parentheses and sequence by A; and then restated as simple transition network
(cf. McKeown, 1985), fixing the placement of Use at the end): see Figure 9-7.
Preparation is the only obligatory element of structure, which is the
sequence of directions taking the reader through the steps of making a
particular dish. Depending on various factors, one or more additional
elements may be present.

9.2 An example from recipe generation

Fig. 9-7: Generic structure of recipe

Use may occur either at the beginning of the recipe or at the end. It is a
specification of how the dish can be used, e.g. as a starter or as the main
dish.
The Ingredients element specifies the ingredients used in the Preparation
phase; it is usually just a list. Mention of ingredients in the Preparation
phase of the recipe may satisfy the need to specify them, in which case there
is no Ingredients element.
Serving specifies how the dish is to be served - hot, immediately, garnished
with parsley, etc.
Finally, Variations may be present to give alternative methods of
preparation and/or substitute ingredients. Alternative methods or ingredients
may also be introduced during Preparation or Serving by means of an
alternative complex (Serve with onion or brown sauce or a green vegetable;
Serve immediately with tomato or herb sauce), in which case there is no need
for a separate Variations element
9.2.2 Semantic generation

The contextual specification guides the process of semantic generation, which instantiates
information from the ideation base: the generic elements shown above are indexed into
the meaning base. (See Hasan, 1984b, for a detailed discussion of the semantic realization
of generic elements from the contextual specification.) Since procedures in general and
recipes in particular are ideationally oriented (in contrast to persuasive texts such as
advertisements, which are interpersonally oriented), each generic element corresponds to

390

9. Using the ideation base in text processing

one or more sequences of certain types. So Preparation, for example, is indexed into
culinary sequences sequences of operations like chopping and frying ingredients. This
is the syntagmatic aspect of the way in which different context types project particular
organizations onto the ideation base.
Those features of the ideation base that are instantiated in this text are set out
informally in Figure 9-8. The diagram specifies (i) the sequence into which the text is
organized, and (ii) the figures comprising each sequence. We have included in square
brackets the full specification of the participants that are ingredients in the recipe together
with some other implicit information in angle brackets. This is obviously still only a
partial specification: the categories are labelled in rather general terms (i.e., they are not
very delicate); and it is likely there are additional operations to be performed in the kitchen
that get left out in the actual text because they are 'obvious' to an (adult) leader before
you can melt the fat, you have to light the stove, etc.. as well as other pieces of
information such as where the fat is melted.
While the structure of a recipe essentially follows the organization of a culinary
sequence, there seem to be two kinds of perturbations in the way the information is
presented in the discourse:
(i) all the referents which are of the type 'ingredient' aie extracted and presented
first as Ingredients. They are then picked up later by means of anaphoric
reference. (Note that this presupposes familiarity with the domain taxonomy;
for example: / oz. blatter or suenut
the fat. ) Some cookery books do
not present the referents as a separate generic element but may print them in
a special font for easy identification.
(ii) the alternative method need not be presented as Variation at the point at
which it constitutes a variation on the main method, but may be separated by
Use: Procedure A Use A Variation.
This is one simple illustration of the need to separate the organization of the text,
which is likely to reflect considerations arising from all three metafunctions and all three
contextual variables, from the organization of figures into sequences in the ideation base.
Both domain specifications and rhetorical specifications are part of the local plans that
guide micro-generation. They are thus available as sources of information for responding
to inquiries in the 'chooser and inquiry' interface between the grammar and the semantics.

391

9.2 An example from recipe generation

Fig. 9-8: Instantial sequence Supporting the generation of a recipe


Let's assume that the information

in the local plan for Unit (1) has three

metafunctional regions (ideation base, interaction base, text base), as shown in Figure 9-

9:

392

9. Using the ideation base in text processing

(i) there is a particular figure of melting in a frying pan, executed by the cook
and operating CHI the fat (from the list of ingredients) [ideation base];
() there is a speech function of instruction assigning modal responsibility to
the reader [interaction base], and
(iii) the process is specified as having thematic status in the plan [text base].

Hie meiafunctional fields in the local plan for unit (1) are as follows:6

In the Penman system, local plans are specified in a special notation, SPL ('sentence plan
language'), described in e.g. Kasper (1989); we have used a similar form.

9.2 An example from recipe generation

393

PLA:
figure: melting-figure 1
process: melting 1
actor: 'you1
goal: fat 1
time: tl
move: (speech function:) instruction 1
speaker: system
addressee: 'you1
speaking-time: 'now'
message:
theme: melting 1
recoverable (identifiable): fail
The ideational part of the local plan has been drawn from the instantial ideation base
(as indicated in Figure 9-9), where the procedure for making the stuffing is stored as a
sequence of figures constituting a particular culinary procedure diagrammatically, see
Figure 9-10.

Fig. 9-10: First step in Prparation, the figure of melting


The location of the melting can be assumed to be part of the ideation base; but it is
not instantiated in the text since it is predictable, given that the reader is assumed to be
familiar with processes of melting, frying, heating, and so on.

394

9. Using the ideation base in text processing

9.2.3 Lexicogrammatical generation


The grammar provides the recipe generator with the resources to represent the
information in the figure above grammatically and to integrate it with interpersonal and
textual considerations. The grammatical fragment we need for our illustration was given
in Figure 9-3 above.
Since lexicogrammatical choice in the Penman system is guided by choosers (see
Section 9.1.2 above), the traversal of the system network is essentially forward chaining
from least delicate to most delicate (left to right in graphic representations of the
system network). If there are simultaneous systems available for entry, each one has to be
entered ideally in parallel (see Tung, Matthiessen & Sondheimer, 1988) but in current
practice in sequence. For any enterable system (i.e., a system whose entry condition has
been satisfied), the traversal algorithm activates the system's chooser, steps through the
chooser-decision tree until a selection of one of the system's output features has been
reached, selects that feature, and executes any realization statements associated with it. At
this point, the cycle has been completed and can start again for a new system except
for some record keeping: given that a new feature has been selected, the traversal
algorithm has first to check whether any new systems have become enterable. For more
detail on the traversal, see Matthiessen & Bateman (1991: 100-9,236-40).
Moving from left to right in the network in Figure 9-3, we come to the PROCESS
TYPE system (material / mental / verbal / relational). The chooser presents a number of
inquines to the semantics to determine what kind of process 'melt#' is. First it presents an
inquiry that can be waded informally as follows: Is the figure 'melt-figure^' one that
involves a conscious being in some kind of sensing; i.e. processing in the consciousness?
Clearly, melting is not, but how does the generation system figure out the response? All
it has to do is check the taxonomy of figures. First, it finds that 'melt-figure^' is an
instantiation of the figure 'melt* in the ideation base potential. Then it moves up the
delicacy hierarchy to try to find 'sensing'. It fails, since the figure 'melt* is classified under
'doing* rather than 'sensing* and concludes that the answer is "no, it's not sensing". The
taxonomic information used is diagrammed hi Figure 9-11.

9.2 An example from recipe generation

395

Fig. 9-11: 'mek' in toe process taxonomy


After the negative response, the chooser proceeds to ask the next inquiry: Is the figure
'melt"figure#' a saying, i.e. a symbolic one involving a Sayer and potentially a Receiver?
Again, the response is a negative one, since the figure 'melt* is not classified under
'saying* in the taxonomy. The chooser then asks a similar kind of question to establish
whether 'melt-figure #' is a being & having figure, finds out that it is not, and finally
asks: Is the figure 'melt-figure #' a doing one involving an Actor? Since 'melt1 is classified
under 'doing' in the model, the response is in the affirmative and the chooser acts on it by
choosing the grammatical feature 'material'. As a result, there is potentially an Actor
present in the TRANSITIVITY structure of the clause (represented by [+Actor] in the
network); it would be present if the MOOD component of the grammar inserted a Subject
(in an operative clause), but there are good interpersonal reasons for leaving the Subject
implicit, namely that in questions and commands the unmarked interpretation is 'you*.
The chooser of the AGENCY system (effective / middle) asks a couple of inquiries and
chooses the grammatical feature 'effective'. As a result, the function Agent is potentially
present and, if present, conflated with Actor. In a material clause, there is a Goal, which is
conflated with Medium. The choices of the features 'effective1 and 'material' satisfy the
entry condition to the system DOING TYPE (creative / dispositive) and the traversal of
the TRANSmVITY part of the network continues.
At the same time (in principle, although in the current model the traversal is
sequential), the MOOD part of the clause network is explored. The choosers of the

9. Using the ideation base in text processing

396

systems in this region make inquiries that are not answered by consulting the domain
model. Rather, the responses to MOOD inquiries are based on a taxonomy of speech
functions (cf. Halliday, 1984b) in the interaction base. The choosers find out that the
melting of the fat is an instruction, that the addressee (reader) is the one responsible for
the success of the instruction, and that there is no good reason to specify the reader
explicitly. The choosers decide on an 'imperative' clause with an implicit Subject (which
is the default option when the Subject is the addressee). Thus, as already mentioned, the
potential constituent Subject/Agent/Actor is left implicit. The combined MOOD and
TRANSITIVITY structure is diagrammed in Figure 9-12.
Melt

the butter

Predicates

Complement

Process

Medium/ Goal

[verbal group: nonfinite]

[nominal group]

Fig. 9-12: MOOD and TRANSITIVITY structures combined


Predicator/ Process is then developed as a non-finite verbal group and Complement/
Medium/ Goal as a nominal group.
Let's now consider the generation of two further units, Unit (8) and Unit (9).
Unit (8)
To express unit (8), the generator has to make a different set of TRANSITIVITY and
MOOD selections. The choosers discover that the process to be represented is a relation
rather than a culinary operation such as melting: the dish is to be represented as belonging
to a certain class of stuffings, viz. stuffings for cucumbers; the features 'relational &
middle' are chosen, and then 'ascriptive'. The resulting TRANSITIVITY structure is
Attribuend/Medium (this ) Process (makes ) Attribute/Range (a good stuffing

far

cucumbers). At the same time, the generator is informing rather than instructing the
leader, and the chooser of the system INDICATIVE TYPE Cdeclarative / interrogative1)
discovers the current speech function is a statement and chooses 'declarative'. As a result
Subject is ordered before Finite. The combined TRANSITIVITY and MOOD structure is
shown in Figure 9-13.

This

makes

a good stuffing for cucumbers

397

9.2 An example from recipe generation

Subject

Finite/ Predicate*

Complement

Medium/ Carrier

Process

Range/ Attribute

[verbal group: finite:


present]

[nominal group]

Fig. 9-13: Combined TRANSITIVITY and MOOD structure for (8)


Unit (9)
As a final example from this text, let's consider unit (9) from the Variation stage of
the recipe. The figure of wanting is classified under 'sensing' with 'reaction' as one of the
intermediate classes. Based on this taxonomy, the grammatical features 'mental' aid
'reactive' are chosen. The entry condition to the system PHENOMENON TYPE
('phenomenal / metaphenomenal') is satisned and its chooser determines what the nature of
the Phenomenon participant is; it asks the environment whether the Phenomenon desired,
a stuffing of more substance, is a metaphenomenon (an idea, fact, or saying) or an
ordinary phenomenon. When the taxonomy is consulted, it is found that stuffings are
things rather than metaphenomena and the grammatical feature 'phenomenal' is chosen. As
a result, the function Phenomenon is preselected to be a nominal group rather than a
clause. In terms of speech function, the generator is hypothesizing rather than stating or
instructing and the grammatical feature 'dependent' is chosen.
From an experiential point of view, the Preparation part of the recipe is a macroprocess, more particularly a macro-operation (dispositive doing): an orciered series of
operations on food items. From an interpersonal point of view, it is a macro-proposal,
more specifically a macro-instruction: an ordered series of instructions to the reader.
In the generation examples we have discussed in this section, we have focussed on the
use of ideation base information and made a few observations about the interaction base
and interpersonal clause generation. However, we have said very little about the text base
apart from noting that it also contributes to local plans. The reason is that we wanted to
explore, in a separate step, the way in which the text base relies on the ideation base in a
text processing system. We are also moving to territory where less work has been done in
NLP and our discussion will be more exploratory.

398

9. Using the ideation base in text processing

9.3 Ideation base supporting text base


The textual metafunction differs from the ideational one in a number of fundamental
respects its mode of syntagmatic progression is wave-like, with periodic prominence;
it is inherently dynamic in that it organizes text as process; and it is a second-order mode
of meaning. Here we shall focus on the last property (mentioned in Chapter 1, Section
1.3) since this leads into the question of what the nature of the text base is and how it
might be modelled given current techniques (see Matthiessen, 1992, from which the
discussion here is drawn; see also Matthiessen, 1995c; Matthiessen & Bateman, 1991:
219-230; Bateman & Matthiessen, 1993).
9.3.1 Second order nature of textual metafunction

The textual metafunction is second-order in the sense that it is concerned with semiotic
reality: that is, reality in the form of meaning. This dimension of reality is itself
constructed by other two metafunctions: the ideational which construes a natural
reality, and the interpersonal, which enacts an intersubjective reality. (Cf. Chapter 1,
Section 1.3, and Figure 1-4 for diagrammatic interpretation). The function of the textual
metafunction is thus an enabling one with respect to the rest; it takes over the semiotic
resources brought into being by the other two metafunctions and as it were
operationalizes them:
All the categories under this third heading [i.e., mode, MAKH & CM] are secondorder categories, in that they are denned by reference to language and depend for
their existence on the prior phenomenon of text. It is in this sense that the textual
component in the semantic system was said to have an 'enabling' function vis--vis
the other two: it is only through the encoding of semiotic interaction as text that the
ideational and interpersonal components of meaning can become operational in an
environment. (Haliiday, 1978b: 145)

This second-order, enabling nature of the textual metafunction is seen both at the level
of context, where mode (the functions assigned to language in the situation) is secondorder in relation to field and tenor (the ongoing social processes and interactant roles),
and at the levels of content the semantics and the lexicogrammar, where the systems of
THEME and INFORMATION, and the various types of cohesion, are second-aider in
relation to ideational and interpersonal systems of TRANSITIVITY, MOOD, and the rest
See Matthiessen (1992) for a general discussion of the textual metafunction.
One manifestation of the second-order nature of the textual metafunction that is
important for our purposes is grammatical metaphor (cf. Chapter 6 above). Grammatical
metaphor is a 'second-order* use of the grammatical resources: one grammatical feature or

9.3 Ideation base supporting text base

399

set of features is used as a metaphor for another feature or set of features; and, since
features are realized by structures, one grammatical structure comes to stand for another
with the semantic effects discussed in Chapter 6. For instance, an identifying clause may
be used to represent a non-identifying one, thereby providing an alternative construal of
that other clause as a configuration of Identified + Process + Identifier. Thus, the clause
you want this may be reconfigured as an identifying clause by nominalizing 'the thing
that you want', what you want, and identifying it with this, either as what you want is
this or as this is what you want:
and I said "I am not conpetent to do it and I wouldn't have
nty name on the title page to do it" and I said "I'm bloody sure
that Hilary and Gavin aren't coupe tent to do it either" and I
said "if this is uiiat you want, I would put maximum pressure
upon somebody like Derek Brainback to do it" but I said .. .
(Svartvik & Quirk, 1980: 802-3)
The two versions of the clause are related in Figure 9-14.

Fig. 9-14: Clause construed as identifying clause (1) THEME


IDENTIFICATION
The motivation behind the identifying metaphor is actually textual: the alternative
configuration in the identifying clause constitutes a textual alternative for distributing
information in the clause, in which the message is structured as an equation between two
terms. This system is known as THEME IDENTIFICATION. What is significant here is
that the textual organization is realized by the second-order resource of grammatical
metaphor. That is, the grammar is as it were turned back on itself: it reconstrues itself
with a particular effect in the discourse, as diagrammed in Figure 9-15. We can see, then,

400

9. Using the ideation base in text processing

that experiential grammatical metaphor is a strategy for creating a 'carrier' of textual


meanings, as noted in Chapter 6.

Fig. 9-15: Textual organization carried by experiential metaphor


A related textual system is that of THEME PREDICATION (the so-called "cleft
construction" of formal grammar), which similarly employs the identifying type of clause
to achieve a particular balance of meaning in the discourse.7 We again give one example
taken from Svartvik & Quirk (1980: 255):
A: There's a lot more in grammar than people notice. People
always notice the lexis.
B: Yes.

A: Lots has been done about that


get so far and so much

fun out

of

but I mean you can only


'pavement',

'sidewalk1,

etcetera .
B: ton.

A: It's the granroar where the fun is.


B: / / I Y // 4 it's the grannir which is interesting //

Both these systems, THEME IDENTIFICATION and THEME PREDICATION, assign


exclusive identity to the identified term of the equation. They differ, however, in the kind
of prominence this identity carries with it: in THEME IDENTIFICATION, in which the
original figure is restructured, the effect is one of second-order (semiotic) prominence,
whereas in THEME PREDICATION, where the figure is not restructured, but one element
within it is explicitly predicated, the effect is that of assigning first-order (natural or
inter-subjective) prominence to the predicated element.

9.3 Ideation base supporting text base

401

Here 'the thing which is un tere sting1 is identified as grammar rather than lexis; the
relationship between the grammar is interesting and it's the grammar which is interesting
is shown in Figure 9-16.

Fig. 9-16: Clause construed as identifying clause (2) THEME


PREDICATION
9*3.2 Grammatical metaphor as second-order semantic resource
These two clause systems, THEME IDENTIFICATION and THEME PREDICATION,
are components of the overall system of THEME in English grammar. They produce
structures in which identifying relational processes are used to reconstrue figures as
equations (the two together are referred to as "thematic equatives" in Halliday, 1985: 414). But ideational grammatical metaphors typically have a discourse function of this kind;
they are as it were pressed into service by the textual metafunction, to provide
alternative groupings of quanta of information. Here are two further examples
of this deployment of ideational metaphors.
(1) Metaphorical mental clauses of perception, such as the fifth day saw them ai the
summit

(cf. Halliday, 1985: 324-5), may serve to create two quanta of information,

Senser (Theme) + Phenomenon (New); consider the following example:


[Theme underlined in ranking clauses, paragraph 2; seasonal Themes in italics;
relevant culmination of the New in bold.]

One of Australia's most majestic irountain ranges is one of


Sydney's most popular year-round playgrounds. The Blue Mountains
to the west of the city have beckoned Sydneysiders to its
resorts since the last century, but only after World War I have
the pleasures of the Blue Mountains been developed to attract

402

9. Using the ideation base in text processing

foreign visitors as well. The main town in the mountains is


Katoomba, 104 km (62 mi) west of Sydney.
Spring and fall are the most beautiful times of year here. Jfc
springtime. millions of wildflowers and trees bud, and the many
planned qarflens in the region start to flourish, fn f-flj /

the

North American species of trees introduced long ago to

the

region oak, elm, chestnut, beech, and birch do the same in


the

Blue Mountains

as

they would

in

the

Catskills:

turn

brilliant reds, oranges, and yellows. 2aaa&r finds c*np*r


and hitara d0cnding on thm mountain in throngs, and
winter is the time the mountains are at their quietest and most
peaceful, offering perfect solitude for city escapees.
If you're looking for a swinging resort, the Blue Mountains
may not be for you. . . . (Fodor'sSydney,p. 115)

Here the metaphoncal clause summer finds campers and hikers descending on the
mountains in throngs

is textually motivated in terms of both THEME and

INFORMATION. From the point of view of THEME, it provides summer as the


unmarked Theme of the clause one instalment in the seasonal method of development
From the point of view of INFORMATION, it groups campers and hikers descending on
the mountains in throngs as one quantum of information (the Phenomenon of the
metaphorical mental clause); contrast the congruent version, which has the same Theme
(marked) but a different culmination of the New: In summer, campers and hikers descend
on the mountains in throngs.
(2) In the second type to be exemplified, a clause or clause complex is nominalized as
a Medium participant in a clause whose Process simply means 'happen* (i.e., existence of
a process: happen, occur, take place; begin, continue, stop). The Medium constitutes a
thematic quantum of information, as in the following example:
(Information involved in progression in italics; relevant culmination of New in
bold.]
The speed of light, and of all electromagnetic

waves, was

given as a constant toy Maxwell's equations, and this speed and


the existence of the waves themselves was independent of any
outside effect. However, Einstein realized that if an observer
was travelling alongside a light wave at the same speed as the

9.3 Ideation base supporting text base

403

light wave, the wave would essentially disappear, as no wave


peaks or troughs would pass by the observer. But
_ug
riipmpfffflfvinr?g of light: waves because of the motion of an
observer should not happen according to ltajnrll, so Einstein
concluded that either Maxwell's equations were wrong or that no
observer could move at the speed of light. He preferred the
latter explanation for a particular reason. (M Shallis, On time: An
investigation into scientific knowledge and human experience, p. 38.)
Here the metaphorical clause makes it possible to summarize the preceding clause
complex the wave would essentially disappear, as no wave peaks or troughs would pass
by the observer as a thematic nominalized participant, the disappearance of light waves
because of the motion of an observer.
9.3.3 Text base stated in terms of ideation base
The principle of the deployment of ideational metaphor by the textual component,
illustrated in Figure 9-15, also applies to the bases of the semantic stratum. The text base
both sets up information states within figures (presents them as messages organized
around quanta of information) and guides the movement from one figure to another. In the
former capacity, the text base can be interpreted as patterns stated over the ideation
base. That is, the representation of ideational meaning in the ideation base constitutes the
first order of representation in terms of which second-order, textual meaning can be
specified.
We have seen that ideational meanings are represented as a large network of nodes
linked by various kinds of relations (such as hyponymy and participant relations). When a
text is produced, certain nodes in this network aie selected for presentation (cf. Sigurd,
1977). As an example, Figure 9-17 presents a semantic network of the information
contained in a biographical note, the short biography of Joseph Conrad in the Penguin
edition of Nostramo (for a discussion of the representation and generation of this text in
Chinese, see Bateman, Matthiessen & Zeng, forthc; for the representation of texts as
networks of meaning, see de Beaugrande, 1980, in press.).

404

9. Using the ideation base in text processing

Fig. 9-17: Example of informal ideational semantic network


We have drawn this network as a succession of figures, with processes (be born, pass,
be, etc.) as nodes accompanied by participant roles (Medium, Range) and circumstantial
roles (Time, Place, Destination, etc.). We have only included infonnation that is present
in the source text;8 any actual semantic network would include a good deal of infonnation

The text presents instantial meanings concerning Conrad; we have not represented the
semantic systemic potential that these meanings instantiate. This potential would
include, among other things, a hyponymic taxonomy.

9.3 Ideation base supporting text base

405

that is not presented in any particular text: producing a text means selecting from the
information given in the network (just as with our recipe example earlier in Chapter 8).
We have organized the diagram in such a way that it is fairly easy to see how one can read
it as a text. The vertical axis represents process time, from 1857 onwards. Processes on
the same horizontal line are related by some type of logico-semantic relation within a
clause complex in the text. The actual Conrad biography reads as follows (clauses not
diagrammed in Figure 9-17 are in italics; ideational Themes are underlined and will be
discussed presently); note that this is only one possible 'textual rendering1 of the ideational
meanings represented in Figure 9-17:
Joseph Conrad (originally Konrad Korzenj.owski) was born in
Russian Poland in 1857, and
passed his childhood in the
shadow of revolution. His parents died when fce, was quite young.
At the aoe of seventeen he went to Marseille to become an
apprentice in the merchant marine. lilis began a long period of
adventure

at

sea,

Conrad

having

his

share

of

hardship,

shipwreck, and other accidents.


He became a British subject in 1886. In 1889, at the aore of
thirtv-one. he came to London for a rest after fifteen years at
sea. On this short London holiday, he began writing a sea novel,
which, after

surviving subsequent

jungle travel, shipwreck on

the Congo, and a railway cloakroom in Berlin, came into the


hands of Edward Garnett and through him to a London publisher.
The book was 'Almayer's Folly',

destined to Jbe the first of a

long series of novels and stories,


mostly inspired by his
experiences of life at sea, which have placed him in the front
rank of English literature. He died in 1924.
Once the semantic network has been defined, it is then possible to state textual
patterns over the network, such as those imposed on it by grammatical metaphor (cf,
again Figure 9-15). One way of representing such patterns is to partition the network
to divide the network into domains or spaces consisting of collections of related nodes
and links (see Hendrix, 1978; 1979).9 Partitioned semantic networks are characterized
succinctly in Bundy (1986: 110) as follows:

Alternatively, we can define processes in terms of the semantic network processes for

406

9. Using the ideation base in text processing

Means of enhancing the organizational and expressive power of semantic nets


through the grouping of nodes and links, associated with Hendrix, Nodes and links
may figure in one or more * spaces', which may themselves be bundled into higherlevel 'yiftac*, which can be exploited autonomously and structured hierarchically.
The effective encoding of logical statements involving connectives and
quantifiers was an important motivation for partitioning, but the partitioning
mechanisms involved are sufficiently well-founded, general and powerful to support
the dynamic representation of a wide range of language and world knowledge.

As noted in this characterization, such partitions have been used to represent various
kinds of information; of particular interest here is the use of partitioned spaces to represent
textual states of prominence. Grosz (1978) presents the idea of representing focus as a
partitioned space in a semantic network what she calls a focus space. A focus space
"contains those items that are in the focus of attention of the dialog participants during a
particular part of the dialog" (op cit., p. 233). Since focus is really only one kind of
textual status of prominence, we need to generalize Grosz's original proposal: textual
states of prominence in general can be modelled as partitioned spaces in a semantic
network. Against the background of this, it is now possible to suggest very briefly how
information in the text base can be modelled:
(i) The second-order character of textual information is captured by defining it in
terms of the already existing semantic network in the ideation base (the firstorder representation, such as the fragment shown in Figure 9-17). This is
clearly only a first approximation: as we snowed in Chapter 6 and noted
again here, the textual metafunction may in fact motivate ideational metaphor
as a means of 'carrying' textual organization.
(ii) Textual prominences constituting textual statuses can then be modelled as
partitioned textual spaces of the semantic network. As already noted, this is a
generalization of Grosz's notion of focus spaces to include thematic spaces,
new spaces, identifiability spaces and so on. This is also only a first
approximation: textual prominence is a matter of degree and we need to think
of a textual space not as a clearly bounded region but rather as a central
region, the peak of prominence, from which one can move to more peripheral
regions, the troughs of non-prominence. Such gradience is necessary not only
to deal with degrees of thematicity and newsworthiness but also to handle

moving through it and linearizing' it as text; textual states such as Theme could then be
modelled by making the process visit nodes in the network in a certain sequence (cf.
Sowa, 1983b). For a general discussion of the representation of textual and interpersonal
meaning in a lattice, see Parker-Rhodes (1978).

9.3 Ideation base supporting text base

407

identifiability by Taidging' (cf. Clark, 1975; for example, bridging from a


centrally identifiable whole to more peripherally identifiable parts).10
The use of textual spaces defined on an ideational semantic network in the ideation
base can be illustrated by adding thematic spaces to Figure 9-17 above: see Figure 9-18.
With the exception of Conrad's parents, the thematic spaces are either Conrad or a time.
Thematic spaces in an ideational semantic network can be seen as a model of the
systemic understanding of Theme and method of development articulated by Martin, where
"field" corresponds to what has been discussed here in terms of ideational semantic
networks in the ideation base:
Method of development ... establishes an angle on the field. This angle will be
sensitive to a text's generic structure where this is realised in stages. Method of
development is the lens through which a field is constructed; of all the experiential
meanings available in a given field, it will pick on just a few, and weave them
through Theme time and again to ground the text to give interlocutors something
to hang onto, something to come back to an orientation, a perspective, a point
of view, a perch, a purchase. (Halliday & Martin, 1993)

The speaker thus selects 'thematic spaces' as points of entry into larger regions of the
ideational semantic network. From the listener's point of view, these thematic spaces
constitute indications of where to integrate the new information being presented in the
text cf. Reinhart (1982). If we think of the listener's processing of a text as being
partly a matter of expanding his or her current semantic network with new information,
the thematic spaces guide him/ her to appropriate expansion points.
In the Conrad biography, there are essentially two competing types of node to be
partitioned into thematic spaces: (i) Conrad himself and (ii) significant times in his life.
(Conrad's works come in as another type of Theme in the second half of the text not
diagrammed in Figures 9-17 and 9-18.) The writer has resolved the conflict by alternating
between these two candidates; in the segment shown in Figure 9-18, temporal Themes are
selected when there is a transition in Conrad's Ufe associated with Conrad's move to a new
destination (Marseille, London). (One could, of course, write the text in such a way that
times consistently fall within the thematic spaces: compare the different thematic versions
of texts provided in Fries, 1981, and Martin, 1992.) Here Conrad remains the Subject of

10

Orosz (1978: 273) operates with both explicit focus and implicit focus. When a physical
object in the semantic network is in explicit focus, its subparts are in implicit focus.

9. Using the ideation base in text processing

408

the clause. But even when Conrad is thematic, the discoursal movement through the
semantic network is still essentially a chronological one (with certain elaborating
excursions realized in dependent clauses).

Fig. 9-18: Semantic network partitioned into textual spaces


So far we have hinted at the possibility of taking a 'snapshot' of textual information
by

partitioning a semantic network into textual spaces representing peaks of textual

prominence11 (crudely, since the spaces are discretely bounded regions at this stage). This

11

Since at this stage the spaces are discretely bounded regions, the effect of partitioning is

9.3 Ideation base supporting text base

409

still leaves us with the task of modelling the guided transitions between different
textual states (again not unrelated to organization in the ideation base: cf. Chapter 3,
Sections 3.4 and 3.6 above; we shall return briefly to this point in our last summary
example). To do this, to deal with what Grosz (1978: 233) called the dynamic requirement
on focus representation, we can use one of the mechanisms developed in computational
linguistics for dealing with 'discourse history1. Grosz used the computational notion of a
stack to model transitions or shifts from one focus space to another, see Figure 9-19.12
As a discourse develops, focus spaces are stacked one on top of another so that the most
recent is always on top of the stack. The stack itself can thus be used a record of
progression through discourse time. Now, the stack is always manipulated from the top:
if a new focus space is to be added to the stack, it is pushed onto the top of the stack; and
if an old one is to be removed, it is popped off the top of the stack.
focus space

Fig. 9-19: The stack as a model of successive textual states


For instance, the succession of thematic spaces might include from bottom to top: 'at
the age of seventeen', 'in 1889, at the age of thirty-one1, 'on this short London holiday'.
While the account based on the stack mechanism is a very explicit model of how textual
states of prominence might get manipulated in the development of discourse, it is very
clear from a functional-linguistic point of view that the stack is too simple a record of
discourse history. In particular, it represents a discourse as a flat structure; it does not

to make prominence appear as categorical feature. A more theoretically informed


representation would show prominence as a gradient (cf. Pattabhiraman's, 1992,
modelling of salience).
12

As an ordering principle, the stack means last first out'; this contrasts with the queue,
where the principle is 'first in first out1. For a discussion of the stack in the modelling of
exchange, see O'Donnell (1990).

410

9. Using the ideation base in text processing

represent the kind of hierarchic constituency organization or internal interdependency


nesting that various approaches to discourse have revealed: see for example Fox (1987) on
reference and rhetorical organization (modelled in terms of Rhetorical Structure Theory). A
typical guide-book text would illustrate the issue with respect to Theme: while the global
principle of development is a spatial one (say, that of the walking tour) and is constructed
in spatial Themes, this type of text often changes locally to other principles, such as the
temporal one reflected in As the battle ended, thf lM .Yfl fidirfl are scud to have mapped
themselves in the Mexican flag and

jumped from the hill to their deaths rather .than

surrender to the U.S. forces, which occurs on a walking tour of a park in Mexico City.
After this temporal detour, the text returns to the spatial development, where it left off:
see Figure 9-19. This is quite typical of Theme selection in discourse (cf. suspended
exchanges in conversations). There is no problem with returning to the earlier principle
for moving from one Theme to another even though it has been interrupted; it may be
necessary to use some strategy for marking the return Theme, such as as for, as to,
regarding + nominal group.
To model the kind of situation that we have just illustrated, McCoy & Cheng (1991)
propose hierarchic trees as a mechanism for controlling focus shifts, and Hovy & McCoy
(1989) relate this work to Rhetorical Structure Theory. Batanan & Mattbiessen (1993)
suggest that Rhetorical Structure Theory can be used to model transitions in a text from
one textual state to another. These rhetorical transitions constitute the method of
development of a text (see Fries, 1981, 1995 for the relationship between Theme and
method of development). The Conrad biography illustrates this principle, but in a slightly
complex way: the main type of transition is temporal sequence, which means that
thematic spaces are likely to included reference to time, and that shifts from one space to
another will occur along with the sequence in time; but sequence often also involves a
constant participant (Conrad in this text), which is then also a likely thematic candidate.

9.3 Ideation base supporting text base

411

Fig. 9-19: Detour through temporal stack, with return to spatial one
9.3.4 Summary example of text base

We can now use an example to summarize what we have said about the text base in
relation to the ideation base. Assume the generation system is at the point at which it
will generate the first clause in the clause complex At the age of seventeen he went to
Marseille to become an apprentice in the merchant marine. The local plan includes the
following information:
Plan 51 going

412

9. Using the ideation base in text processing

message:

: theme (age 17)


: JdentiFiability identifiable (Conrad)

This plan is shown as instantial meaning in the ideation base in Figure 9-20.
The ideational infonnation is of a familiar kind. "Going 23" instantiates the general
class of going, which is a subclass of moving; that is, the potential specifies a taxonomy
(which is not a strict one) of what can be meant. For instance, it is possible to find out
that 'going1 is a type of 'figure' which in turn is a type of 'phenomenon1 (the most general
concept in the taxonomy). Going 23 includes one participant, Conrad, and two
circumstances, Marseille and age 17; it has been selected for verbalization and is par of
a local plan to be handed over to the process of local (micro-) generation. But that local
plan also contains textual (and interpersonal) information, viz. that the Theme has been
planned to be 'age 17'. This is a specification of a textual partition in the ideation base,
representing (as we have seen) a textual state of information. It can be motivated by
reference to the other major phenomenon the text base hqs to deal with textual
transitions. As we suggested already in Chapter 3, Section 3.4 (cf. also Figure 3-10), the
relational organization of text (the rhetorical transitions) that the text base manages may
be supported by sequences in the ideation base. This is likely to happen both in narrative
text (such as the biography discussed in this section) and in procedural text (such as the
recipes discussed earlier). In any case, at that point in the global text plan where there is a
local plan covering the clause complex At the a$e of seventeen he went to Marseille to
become an apprentice in the merchant marine , a purposive relation is introduced within
the sequence; but that sequence is itself in a relation of temporal succession to an earlier
sequence: see Figure 9-21. That is, the going to Marseille expands the previous text in
terms of temporal succession; and there is every reason to select a Theme that indicates
the point of temporal expansion Conrad's age (realized as ai the age of seventeen in
initial position in the clause).
It is clear, then, that the ideation base supports the text base: textual infonnation can
be stated as patterns over the text base partitions, and moves between partitions. At the
same time, it is also very clear that the way the ideation base is organized the various
configurations that have evolved must have been shaped by textual pressure. So, for
instance, given that a relationship could be construed either as 'that boy's hair is green1 or
'that boy has green hair1, the preference in English for the construal where the whole is

9.3 Ideation base supporting text base

413

Carrier and the part is Attribute can be understood in textual terms: it means that the
whole can serve as Theme. We have already seen that the textual metafunction is a
powerful pan of the explanation of ideational metaphor ideational meaning is leconstrued
in such a way that it suits textual organization when meanings are being distributed in
text. This is an area where the evolved system of English and designed systems such as
logics differ sharply: the latter are not designed to construe textual differences and instead
place a high value on canonical forms.

414

9. Using the ideation base in text processing

Fig. 9-21: Rhetorical transitions in the text plans

Part IV:
Theoretical and descriptive alternatives
10, Alternative approaches to meaning
We began our book with some general observations about the approach we have adopted
seeing knowledge as meaning, approaching it through the grammar, taking account of
different kinds of seme-histories, and so on (Part I). We then introduced the ideational
meaning base and discussed the different kinds of phenomena in some detail (Part ). We
then went on to show how domain models are located within the ideation base and how a
language processing system can draw upon these ideational resources (Part lu). We arc
now in a position to compare and contrast our approach with alternatives both
theoretical alternatives and global descriptive ones (this chapter) and alternatives in details
of description (Chapters 11 and 12).

10.1 Logico-philosophical vs. rhetorical-ethnographic


orientations
It is impossible to offer anything more than the barest outline of the historical
background to work on the ideation base. We can identify two main traditions in Western
thinking about meaning (see Halliday, 1977):
(i) one oriented towards logic and philosophy, with language seen as a system of
rules;
(ii) one oriented towards rhetoric and ethnography, with language seen as resource.
It is typically the logical-philosophical tradition that provides the background for work on
knowledge representation and proposals for the knowledge base. Since the 50s, a link has
been forged between this tradition and cognitivism under the general rubric of cognitive
science. However, although it is less often referred to, the rhetorical-ethnographic
tradition is equally relevant to work on the modelling and representation of knowledge
for example, in recent times, the work on folk taxonomy carried out by anthropologists
and anthropological linguists (often in the framework of ethnoscience) and the work on
intellectualization in the Prague School framework are central to understanding the
organization of the ideation base. Our own work here falls mainly within the second
tradition but we have taken account of the first tradition, and the general intellectual
environment in which versions of our meaning base are being used also derives primarily
from the first tradition. Indeed, the two traditions can in many respects be seen as

416

10. Alternative approaches to meaning

complementary, as contributing different aspects of the overall picture. Our own


foundation, however, is functional, as noted in Part I.
Mainstream work in psycholinguistics, e.g. Miller & Johnson-Laird (1976), is in
general located within the logico-philosophical tradition. Two other distinctive
contributions to the study of meaning within linguistics, those of Lamb (1964, 1992) and
of Wierzbicka (1980, 1988), derive in more or less equal measure from both the major
intellectual movements we have referred to.
The two orientations towards meaning thus differ externally in what disciplines they
recognize as models. These external differences are asscociated with internal
differences as well.
(i) First, the orientations differ with respect to where they locate meaning in
relation to the stratal interpretation of language:
(a) intra-stratal: meaning is seen as immanent something that is constructed
in, and so is part of, language itself. The immanent interpretation of meaning is
characteristic of the rhetorical-ethnographic orientation, including our own
approach.
(b) extra-stratal: meaning is seen as transcendent something that lies outside
the limits of language. The transcendent interpretation of meaning is
characteristic of the logico-philosophical orientation.
Many traditional notions of meaning are of the second kind meaning as reference,
meaning as idea or concept, meaning as image. These notions have in common that they
are 'external* conceptions of meaning; instead of accounting for meaning in terms of a
stratum within language, they interpret it in terms of some system outside of language,
either the 'real world* or another semiotic system such as that of imagery. The modem
split within the "transcendent" view is between what Barwise (1988: 23) calls the worldoriented tradition and the mind-oriented tradition, which he interprets as public vs.
private accounts of meaning:
The world oriented tradition in semantics, from Tarski on, has focussed on the public
aspect of meaning by trying to identify the meaning of a sentence or text with its truth
conditions, the conditions on the actual world that are needed to insure its truth
(Davidson 1967 and Montague 1974). The psychological tradition, by contrast, has
focussed on the private aspect by trying to identify meaning with an intrinsically
meaningful mental representation (Fodor 1975 or Jackendoff 1983 e.g.).

The world-oriented tradition interprets meaning by reference to (models of) the world; for
example, the meaning of a proper noun would be an individual in the world, whereas the
meaning of an intransitive verb such as run would be a set of individuals (e.g. the set of
individuals engaged in the act of running). The mind-oriented tradition interprets meaning

10.1 Logico-philosophical vs. rhetorical-ethnographic orientations

417

by reference to the mind; typically semantics is interpreted as that part of the cognitive
system that can be "verbalized".
(ii) Secondly, the approaches differ with respect to what they take as the basic unit
of meaning. In the logico-philosophical orientation, the basic unit tends to be
determined "from below", from grammar: since sentences are seen as encoding
propositions, the basic unit of semantics is the proposition (as in prepositional
calculus).1 In contrast, in the rhetorical-ethnographic orientation, the basic unit tends to
be determined "from above", from context: since language is seen as functioning in
context, the basic unit of semantic is the text (see Halliday & Hasan, 1976; Halliday,
1978a). So in the logico-philosophical orientation, semantics means in the first instance
prepositional semantics,2 whereas in the other orientation it means text or discourse
semantics. (We have not attempted to extend our present coverage to include the semantics
of discourse; see Martin, 1992, for a comprehensive model of discourse semantics in
systemic functional terms. A rich meaning-oriented approach to text linguistics will be
found in de Beaugrande, 1980, and in a new framework in de Beaugrande, in press.)
(iii) The two orientations differ in the metafunctional scope of their models of
semantics. In the logico-philosophical orientation, meaning is closely associated with
representation, reference, denotation, extension or 'aboutness', so the metafunctional scope
is restricted to the ideational metafunction: semantics means ideational semantics. In the
rhetorical-ethnographic orientation, meaning is closely associated with rhetorical concerns,
so the metafunctional scope involves all three metafunctions: semantics means ideational,
interpersonal and textual semantics; it is multifunctional. If interpersonal and textual
meanings ate dealt with by logico-philosophical accounts (they are often outside their
scope), they are handled tinder the heading of pragmatics rather than the heading of
semantics. For example, speech act theory was developed as a logico-philosophical
interpretation of speech function (or rather of its ideational constmal) and has come to be
included within pragmatics.
(iv) Finally, we can note that the two orientations differ in what kind of semantic
organization they focus on. In the logico-philosophical orientation scholars have
focussed on syntagmatic organization: they have been concerned with semantic structure
including principles relating to structure such as those of compositionality and

Note that proposition has a different sense here from proposition


semantics, where it contrasts with proposal
2

in our interpersonal

This is changing to some extent, e.g. with the formal semantic work by Hans Kamp and his
followers. In computational linguistics it was necessary to model discourse patterns from a
fairly early stage, so approaches to discourse semantics originating in the logicophilosophical tradition began to show up in the 1970s.

418

10. Alternative approaches to meaning

semantic decomposition. For example, in their analysis of the senses of "words", they
have tended to analyse these as being composed of semantic components, semantic
markers, semantic primitives or the like (following Katz & Fodor, 1963, in the generative
tradition). In the rhetorical-ethnographic orientation scholars have focussed on both
paradigmatic and syntagmatic organization, often foregrounding the paradigmatic: they
have been concerned with the semantic system, with the meaning potential including
principles of taxonomy and the metafunctional simultaneity of systems. For example,
scholars in "ethnoscience" have studied folk taxonomies of animals, plants, diseases and
the like (cf. Chapter 2, Section 2.11.3 above); and systemic-functional scholars have tried
to map out semantic systems such as those of speech function and conjunction. Another
example is our own work on the ideation base presented here.
The two different orientations are summarized with the internal differences we have
just reviewed in Figure 10-1. We now turn to the two variants within the logicophilosophical tradition, the "world-oriented" one of formal semantics (Section 10.2) and
the "mind-oriented" one of cognitive semantics (Section 10.3).

Fig. 10-1: The manifestation of the two traditions in the study of meaning

10.2 Formal semantics

10.2 Formal semantics


Relevant discussions primarily illuminating the logical-philosophical

tradition

include Kneale & Kneale (1962) on the history of logic; Large (1985), Knowlson (1975),
Slaughter (1986), Salmon (1966; 1979) on the history of 'artificial languages' and
taxonomy in the period of early Western science; Sowa (1983a, 1991), Brachman (1979)
and Brachman & Levesque (1985) on modem knowledge representation in AI, with notes
on precursors; Eco et al (1988) on different contemporary views of the status of meaning
and cognition; Bursill-Hall (1971) on medieval theories of meaning. Here we are merely
noting some particularly important developments. These aie charted in Figure 10-2.
The development of logic is of interest in itself in the present context as a tradition of
'semiotic design* whereby logical systems were developed out of the resources that have
evolved in language. This design of logical systems was earned out to meet certain
restricted purposes having to do with reasoning; but once such systems had been
developed, philosophers and linguists turned them back on language and used them as
systems for representing and theorizing meaning. In the present century, when formal
semantics was developed for underpinning logical systems, this was then taken as a model
for work on the semantics of natural languages.
As we have shown, natural language includes a "natural logic" (Chapter 3); and in the
discourses of philosophy this was gradually transformed into models of logic and
reasoning. Such philsophical models were first constituted in language, but later
developed into the designed systems of symbolic logic. Leibniz had a vision of an
"algebra of thought" for correct reasoning. He took note of the work by Wilkins and
others on "artificial languages"; but these were concerned in the first instance with
taxonomic organization (within the domain of the experiential mode of meaning), whereas
Leibniz' primary concern was with reasoning (within the domain of the logical mode of
meaning). His own work was left unfinished, but symbolic logic can be seen as a
continuation of his attempt to design a special semiotics for reasoning with. In the 19th
century, de Morgan codified the "laws of reasoning" (modus ponens, etc.); and
subsequently Boole (1854) formulated an algebra of the relations of conjunction and
disjunction and of negation, "Boolean algebra", which he thought of as constituting the
"laws of thought". In 1879, Frege published his "Begriffsschrift", subtitled "eine der
Arithmetischen nachgebildete Formelsprache des reinen Denkens". Frege's work laid the
foundation for predicate logic, and for modem symbolic or mathematical logic in general.
One of his contributions to the emergence of formal semantics was his distinction
between Sinn

and Bedeutwg (Frege, 1892),

translated into English as sense and

reference. This distinction was later developed by Camap as the contrast between intension
and extension, and is Dart of the foundation of semantic.* nf th* mnHpi-thprti^ I

420

10. Alternative approaches to meaning

Fig. 10-2: Development of approaches to meaning


In the 20th century, as the scope of logical systems grew, it became possible to use
them as a representational resource in the description of meaning in language: cf. Chapter
1, Section 1.9.2 above. In the 1930s, Alfred Tarski developed the truth conditional
approach to semantics for the designed semiotic system of predicate logic. He criticized
the current correspondence theory of truth, according to which a proposition is true if it
corresponds to a fact (proposed e.g. by Russell and Wittgenstein but in fact going back to
Aristotle), adopting another version whereby truth is defined in terms of truth

10.2 Formal semantics

421

conditions:3 a proposition is true if and only if it refers to an actual state of affairs in


the world. The example that is usually cited is: snow is white [proposition] is true if and
only if snow is white [state of affairs in the world]. In this way, states of affairs came to
be represented as models of the world, and the 'meaning* of the constructs of logic as
either extensions in these models or as truth values. For example, the extension of a
constant is an individual, the extension of one-place predicate is a set of individuals. The
semantics of logical systems was thus stated in terms of model theory. This approach
to meaning is objectivist: (extensional) meanings are phenomena existing in the world.
There are no meaners construing meanings, and there is no perceptual system mediating
between semiotic expressions and their extensions. Truth is a matter of correspondence,
not (as it is in everyday discourse) a question of consensus among people.
In the 1950s and 1960s, the philosopher Richard Montague took the semantics of
designed logics as a model for describing the meaning of natural language: "having worked
with formal languages of increasing complexity, he came to realize that by having a
lexicon of English words and a list of structural operations consisting of English
constructions, English itself could be regarded as one of these formal Languages"
(Thomason in his introduction to Montague, 1974: 16). Bach (1989: 8) calls this
"Montague's Thesis": "natural languages can be described as interpreted formal systems".
That is, a natural, evolved language such as English was treated as if it was a formal,
designed "language" such as predicate logic. In order for it to be treated in this way, it bad
to be disambiguated so that its expressions would be unambigous just as those of formal
logic are. But Montague himself did not see this approach as a metaphorical treatment
of English. In a paper entitled "English as a formal language", he writes:
I reject the contention that an important theoretical difference exists between formal
and natural languages. ... I regard the construction of a theory of truth or rather, the
more general notion of truth under an arbitrary interpretation as the basic goal of
serious syntax and semantics; and the developments emanating from the Massachusetts
Institute of Technology offer little promise towards that end. ... I shall accordingly
present a precise treatment, culminating in a theory of truth, of a formal language that I
believe may be reasonably regarded as a fragment of ordinary English. ... The treatment
given here will be found to resemble the usual syntax and model theory (or semantics) of
the predicate calculus [footnote: The model theory of the predicate calculus is of course
due to Alfred Tarski].

Formal semantics, or "Montague semantics" as the approach has also come to be


called, was taken up in linguistics by e.g. Dowty et al (1981), Bach (1989), Cann (1993),
Partee (1975), and carried further by Keenan & Faltz (1985). In philosophy, Barwise &
Perry (1983) introduced new ideas into formal semantics in the form of their "situation
semantics". For a recent critique of formal semantics, see Bickhart & Campbell (1992).

Haack (1978: 99) calls Taiski's theory of truth a semantic theory; on the question whether
it is a correspondence theory, see pp. 112-114.

422

10. Alternative approaches to meaning

In its relation to syntax, the formal approach to semantics is compositional, being


formulated in terms of syntagmatic organization: the meaning of larger syntagmata units
such as sentences is built up as functions of smaller units such as nouns and verbs. This
semantic composition is in tandem with the syntax, so what becomes important is the
nature of the syntactic rules themselves, not just the result of their application. Montague
used a form of categorial grammar in formalizing the rules of syntax for this purpose.
In modelling meaning, formal semantics models it as both extensional and
intensional: extensional meaning is cast in terms of sets of entities in a model of some
world, and intensional meaning is the relation between a linguistic expression and its
extension in any possible world Formal semantics has thus been concerned with how
semantic representations can be related to some extra-linguistic world model. Dowty et ai
(1981: 5) emphasize the relationship between language and the world in their introductory
characterization of truth conditional semantics:
"... truth conditional semantics, in contrast to other approaches mentioned [including
Katz & Fodor, Jackendoff, and generative semantics, MAKH & CM], is based squarely
on the assumption that the proper business of semantics is to specify how language
connects with the world in other words, to explicate the inherent "aboutness" of
language."

This is, of course, an ideational orientation: the focus here is on representational meaning
only. One central method has been to build models of the world in set theoretic terms, and
to relate these to linguistic expressions the model theoretic approach mentioned above.
The model theoretic approach to meaning is a transcendent one: meanings are located
in the world outside of language. This is seen as a strength by the proponents of the
model theoretic approach. Dowty et al (1981) criticize various other theories of meaning
for ignoring the "inherent aboutness of language", its relationship to the real world:
Any theory which ignores this central property, it is argued, cannot be an adequate
theory of natural language. Examples would be theories which, in effect, give the
meaning of a sentence by translating it into another language, such as a system of
semantic markers or some son of formal logic, where this language is not further
interpreted by specifying its connection to the world. The approach of Katz and his coworkers seems to be of this sort (Katz and Fodor, 1963; Katz & Postal, 1964), as is that
of Jackendoff (1972) and of the framework of Generative Semantics (Lakoff, 1972;
McCawley, 1973; Postal, 1970).

It would seem that formal semantics is quite far from the concerns of cognitive science;
but formal semantics is often carried out within the broad program that cognitive
scientists adopt. Writing from a cognitive standpoint, Johnson-Laird (1983) assesses
model-theoretic semantics as follows:
The power of model-theoretic semantics resides in its explicit and rigorous approach to
the composition of meanings. ... Some idealizations definitely complicate the
'ecumenical' use of model-theoretic semantics as a guide to psychological semantics.
One such idealization is the mapping of language to model without any reference to the
human mind, and this omission gives rise to certain intractable difficulties with the

10.2 Formal semantics

423

semantics of sentences about beliefs and other such prepositional attitudes. These
difficulties are readily resolved within the framework of mental models, (p. 180)
... model-theoretic semantics should specify what is computed in understanding a
sentence, and psychological semantics should specify how it is computed, (p. 167)
Johnson-Laird's own psychological approach is one based on mental

models. In

contrast to this approach, Jackendoff (1983: Ch. 2), working within cognitive semantics,
argues that the account of 'aboutness' in formal semantics is simplistic:
... I will take issue with the naive (and nearly universally accepted) answer that the
information language conveys is about the real world, (p. 24) ... If indeed the world as
experienced owes so much to mental processes of organization, it is crucial for a
psychological theory to distinguish carefully between the source of environmental
input and the world as experienced. For convenience, I will call the former the real world
and the latter tbt projected world (experience world or phenomenal world would also be
appropriate), (p. 28) ... It should now be clear why we must take issue with the naive
position that the information conveyed by language is about the real world. We have
conscious access only to the projected world the world as unconsciously organized
by the mind; and we can talk about things only insofar as they have achieved mental
representation through these processes of organization. Hence the information
conveyed by language must be about the projected world. We must explain the naive
position as a consequence of our being constituted to treat the projected world as reality.
According to this view, the real world plays only an indirect role in language: it serves
as one kind of fodder for the organizing processes that give rise to the projected world.
If this is the case, we must question the centrality to natural language semantics of the
notions of truth and reference as traditionally conceived. Truth is generally regarded as a
relationship between a certain subset of sentences (the true ones) and the real world;
reference is regarded as a relationship between expressions in a language and things in
the real world that these expressions refer to. Having rejected the direct connection of
the real world to language, we should not take these notions as starting points for a
theory of meaning. Thus an approach such as that of Davidson (1970), which attempts
to explicate natural language semantics in terms of Tarskian recursive theory of truth, is
antithetical to our own inquiry, (pp. 29-30)
The projected world in Jackendoff s account is the result of creative acts of perception:
it is constructed as a model of sensory input, but with the significant addition of
information from the perceptual system itself. Writing from a different orientation within
cognitive semantics (see below), Lakoff (e.g. 1988) is also very critical of what he calls
"objectivist metaphysics11 (metaphysical realism). He presents a detailed critique of this
position, also noting the problem that arises if meanings are located in the world:
To view meaning as residing only in the relationship between symbols and external
reality is to make the implicit claim that neither color categories, nor any other
secondary category, should exist as meaningful cognitive categories. Yet color
categories are real categories of the mind. They are meaningful, they are used in reason,
and their meaning must be accounted for. But the mechanism of objectivist cognition
cannot be changed to accommodate them without giving up on the symbolic category of
meaning. But to do that is to abandon the heart of the objectivist program, (p. 132)
We also have emphasized that reality is not something that is given to us; we have to
construct an interpretation of it or, as we prefer to put it, we have to construe our
experience. Interpretation is a semiotic process, and our interpretation takes into
account not only the concrete natural world but also the socio-cultural realm that is

424

10. Alternative approaches to meaning

brought into existence as a semiotic construct (see Hasan, 1984a, for discussion, with
reference to Wharf). The constructive power of meaning is perhaps

most easily

observable in scientific and technical fields where the resources of language are used in
metaphorical ways; for example, the technical concept of financial markets is constructed
largely by means of the notion of spatial movement (see also Chapter 6, Section 6.7.4
above):
With exports falling

to their

lowest

level

in a year,

the

trade deficit widened sharply relative to the average hortfall


in the first quarter. The fact that export demand in April stands
at a 16% annual rate below the first quarter average underscores
the

combination

of

adverse

currency

effects

and

eluggish

activity abroad. Iitport demand rose fractionally, with all

of the

gain reflecting an upturn in the energy area.

But in everyday language also we can see the process of interpretation at work; here too
quantity and intensity are constructed metaphorically, by means of spatial resources.

10.3 Cognitive semantics


Cognitive semantics has emerged out of the generative tradition in the last ten to fifteen
years. It is a natural development of the general cognitivist research program that began in
the 1950s with the birth of cognitive science, Artificial Intelligence, and Chomsky's
cognitively oriented linguistics.
10*3.1 Two approaches within cognitive semantics
Those adopting the cognitive approach to semantics share certain assumptions about
semantic organization as part of conceptual organization, and tend to reject formal,
Montague-style semantics, as indicated by the passages from Jackendoff and Lakotf quoted
above. However, they seem to fall into two groupings, which can be conveniently
described in terms of the US coastline.
(i) On the East Coast, Jackendoff (e.g., 1983), drawing on his earlier work on
interpretive semantics (Jackendoff, 1972), has developed a generativist type of cognitive
semantics,4 which he calls conceptual semantics. This is integrated with a number of
aspects of modem generative theory, e.g. the X-bar syntactic subtheory. Jackendoff puts

Jackendoff calls his approach conceptual semantics and Lakoff calls his cognitive semantics.
We use the tenu cognitive semantics as the generic term for cognitively oriented approaches
to semantics.

425

10.3 Cognitive semantics

forward a conceptual ontology and suggestions for conceptual structure that we shall
return to in Section 10.3.2 below. This version of cognitive semantics is arguably the
more closely associated with the logical and philosophical tradition (cf. Jackendoff, 1988:
81-2).
(ii) On the West Coast, a number of linguists have developed a "cognitive" alternative
to generative linguistics. Some of them (e.g., Lakoff, Langacker) come from a generative
background (Lakoff s starting point was generative semantics), but have made a radical
departure from this tradition. They have widened the scope of study relative to the
generativist research agenda so as to include metaphor as a prominent feature (Lakoff &
Johnson, 1980, and other subsequent writings, such as Lakoff, 1987, 1988) and a detailed
theoretical model of the relationship of language to cognition and perception (Langacker,
1987). A few have also oriented their work towards discourse (notably Chafe, e.g. 1979;
1987; cf. also Tomlin's, 1987a, discussion of the linguistic reflection of cognitive
events). This version of cognitive semantics is arguably more closely associated with the
rhetorical and ethnographic tradition (perhaps not so much in terms of its roots, but in
terms of where it is headed); cognitive anthropology, with its interest in folk taxonomy
and more recently in cultural models, provides a meeting point between the two.
Various aspects of the West Coast work in cognitive semantics are relevant to the
organization of the ideation base; for example, the work on metaphorical systems already
mentioned, Talmy's (e.g. 1985) work on lexicalization, and Chafe's (1970) early work on
the organization of meaning. We shall return to a brief discussion of Chafe's typology of
'semantic verbs' in Chapter 12, Section 12,3.2 below,
10.3.2 Jackendoffs conceptual

semantics

Jackendoff (1983) presents a major study of semantics and cognition from a generativist
point of view. As part of that study, he presents an ontology of conceptual types that are
linguistically motivated. While his purpose is theoretical rather than descriptive, and the
ontology is not very extensive, it has become a frame of reference for work in this area; it
will be useful to characterize this ontology briefly and compare it with the organization of
our ideation base.
Jackendoff sees semantic organization as part of conceptual organization that pan
which can be verbalized; this is a position that distinguishes him from a number of other
generativists. He identifies two possible positions on the relationship between semantic
organization and conceptual organization (1983: Section 1.7; interpreted by us in Figure
10-3):

426

10. Alternative approaches to meaning

(1) "conceptual structure could be a further level beyond semantic structure, related
to it by a rule component, often called pragmatics, that specifies the relation of
linguistic meaning to discourse and to extralinguistic setting."
(2) "semantic structures could be simply a subset of conceptual structures just
those conceptual structures that happen to be verbally expressible/'
(1) Conceptual organization as separate level

(2) Semantic organization as verbalized


subpart of conceptual organization

Fig. 10-3: Two possibles relationships between conceptual and semantic organization
Jackendoff and a number of others now prefer the second position. It is also shared by
e.g. Langacker (1987), representing cognitive semantics from the other US coastline:
Meaning is a mental phenomenon that must eventually be described with reference to
cognitive processing. I therefore side with Chafe (1970, p. 74-76) by adopting a
"conceptual" or "ideational" view of meaning ... I assume it is possible at least in
theory (if not yet in practice) to describe in a principled, coherent, and explicit manner
the internal structure of such phenomena as thoughts, concepts, perceptions, images
and mental experience in general. The term conceptual structure will be applied
indiscriminately to any such entity, whether linguistic or nonlinguistic. A semantic
structure is then defined as a conceptual structure that functions as the semantic pole of a
linguistic expression. Hence semantic structures are regarded as conceptualizations
shaped for symbolic purposes according to the dictates of linguistic convention, (pp.
97-8)

From our standpoint, this appears as a transcendent interpretation of meaning: we on the


other hand prefer an immanent approach to meaning, where "conceptual organization" is
interpreted as meaning that is created by various semiotic systems, among which language
is the primary one (see Chapter 1, Section 1.1 above). We shall review the implications
of this approach below; see also Chapter 14 on construing "the mind". Jackendoff presents
evidence suggesting that there are interesting correspondences between the conceptual
categories that are expressed linguistically and the categories that have to be posited in an
account of vision, derived from studies of perception. This is complementary to recent

427

10.3 Cognitive semantics

systemic research investigating semiotic systems other than language, such as visual art
and diagrams (Kress & van Leeuwen, 1990,1996; OTooie, 1989,1992,1994).
Jackendoff takes seriously the relation between conceptual organization and syntactic
organization; this might be challenged from a classical formalist point of view, but from
a functional point of view it is quite natural (cf. Chapter 1, Section 1.1). In particular, he
identifies correspondences between syntactic classes (i.e., categories in generative tarns)
and conceptual ones. Such correspondence is in fact a major source of evidence for the
conceptual ontology. In particular, Jackendoff uses wh-items and non-interrogative
reference items to support the ontology; he recognizes things, amounts, places, directions,
manners, events and actions. For instance, both things and places have to be recognized
because English has both the forms what did you buy? and where

is my coat? The

ontology is tabulated in Table 10(1) below together with the grammatical evidence for
each type. The left-most column provides a rough translation into our ideation base
ontology.
Table 10(1): Jackendoff s (1983) ontology
the ideation base
ontology
element:
participant
element:
circumstance

Jackendoffs
ontological category
thing
amount
place
direction
manner

element:
process

event
action

what (did you buy?)


how long (was the fish?)
where (is my coat?)
where (did they go?)
how (did you cook the
eggs?)
what happened (next?)
what (did you) do?

this, that, yay long


here &. there
thataway
thus, so, this way
that ... happen
do it

Jackendoff contrasts his ontology with that of "standard first-order logic", and makes
the important point that the ontological classes of logic are vastly underdifferentiated from
a linguistic point of view that this type of logic is not an adequate theory of the
semantic structure of natural language (but cf. Chapter 1, Section 1.9.2 above, where we
raised the issue of the status of logic as a system of representation). Compared to this
type of logic, Jackendoffs ontology is much more highly differentiated However, it is
not very rich compared to what we believe is needed in an account of the ideation base.
The classes recognized are roughly a list of semantic correlates of word classes at primary
or secondary delicacy (such as one finds in traditional grammars). The list is not
exhaustive, it does not include any significant paradigmatic organization (i.e., it contains
no organization showing how types are arranged into a subsumption lattice) and some of
the most revealing distinctions of the ideational semantics of English are absent e.g.
the distinction between phenomena and metaphenomena, the recognition of the role
played by projection, and the expansion of the system through grammatical metaphor.

428

10. Alternative approaches to meaning

These are general observations. Since Jackendoff relies on reference as a source of


evidence for the ontological distinctions, he might in fact have taken note of the
semantically ciucial phenomena of 'extended reference* (the move to 'macro') and reference
to fact (the move to 'meta') discussed in Halliday & Hasan (1976) cf. Chapter 2,
Section 2.11.3.2.
Returning to Barwise's contrast between the mind-oriented view of meaning and the
world-oriented view discussed in Section 10.1, we can note Barwise's general argument
against the mind-oriented view:
... representational mental states have meaning in exactly the same way that sentences
and texts have meaning, and saying what one means is a complicated matter. This
makes attempts to explicate linguistic meaning in terms of mental representations an
evasion of the main issue: How do meaningful representations of all kinds, sentences
and states, mean what they do? (Barwise, 1988: 38)

We acknowledge this problem, but we believe the solution lies in a socio-semiotic view
of meaning such as the one we are presenting here. Jackendoff views information about
the projected world in conceptual terms; hence reality construction is seen as a process
taking place within the consciousness of the individual. Our own view, that the projected
world is a semantic construction, foregrounds the interpersonal perspective: meaning is
construed in collaboration. Meanings are exchanged; and the "projected world" is
constantly calibrated against the interpersonal negotiation of meaning. This means that
consensus and conflict take over much of the domain that is usually conceptualized in
terms of truth and falsehood (cf. Eggins, 1990). The semantic system (as pan of the
linguistic system) is shared; it is part of our social being. Thus while our view that the
ideational semantic system construes human experience is similar to what Lakoff (1987;
1988) calls the position of "experientialist cognition" (the position he has himself
espoused, in contrast with what he calls "objectivist cognition'1), it differs in that for us
construing experience is an intersubjective process. It is at once both setniotic (the
construction of meaning) and social (as in Peter Berger's "social construction of reality":
cf. Berger & Luckmann, 1966, Wuthnow et al, 1984). It is the intersection of these two
perspectives that characterizes the social semiotic we are attempting to present in this
book (cf. Lemke, 1995; Thibault, 1993).
10.3.3 Fawcett's cognitive model of an interactive mind

Within systemic-functional linguistics, Fawceu (e.g. 1980) has pioneered a "cognitive


model of an interactive mind". There are many fundamental similarities with tbe approach
we are taking here, e.g. in construing an experiential system of process configuration
within the content plane. However, there are two related differences of particular interest in
the context of our present discussion:

10.3 Cognitive semantics

429

(i) in Fawcett's model, there is only one system-structure cycle within the content
plane: systems aie interpreted as the semantics, linked through a "realisation
component" to [content] form, which includes items and syntax, the latter being
modelled structurally but not systemically;
(ii) in Fawcett's model, the semantics is separate from the "knowledge of the
universe", with the latter as a "component" outside the linguistic system
including "long term memory" and "short term sort of knowledge".
With respect to (i), in our model there are two system-structure cycles, one in the
semantics and one in the lexicogrammar. Terms in semantic systems are realized in
semantic structures; and semantic systems and structures are in turn realized in
lexicogrammatical ones. As we saw in Chapter 6 in particular, grammatical metaphor is a
central reason in our account for treating axis and stratification as independent dimensions,
so that we have both semantic systems and structures and lexicogrammatical systems and
structures. Since we allow for a stratification of content systems into semantics and
lexicogrammar, we are in a stronger position to construe knowledge in terms of meaning.
That is, the semantics can become more powerful and extensive if the lexicogrammar
includes systems. It follows then with respect to (ii) that for us "knowledge of the
universe" is construed as meaning rather than as knowledge. This meaning is in the first
instance created in language; but we have noted that meaning is created in other semiotic
systems as well, both other social-semiotic systems and other semiotic systems such as
perception (cf. Chapter 15 below). Our account gives language more of a central
integrative role in the overall system. It is the one semiotic system which is able to
construe meanings from semiotic systems in general.
Fawcett's model, although in certain ways closer to mainstream cognitive science than
ours, is also a systemic-functional model. In other words it is within the same general
theoretical framework as that within which our own work is located.

10.4 Work on meaning in NLP


The work we have discussed so far has its roots in linguistics, philosophy, or
anthropology. Some of it has made contact with natural language processing (NLP); for
example, Rosner & Johnson (1992) contains contributions on Montague semantics in
computational linguistics, and Lamb's work has been computationally oriented from the
start. However, there is also a considerable body of linguistic work that is more internal
to NLP, and to certain aspects of AI. Arguably most of this work in the area of semantics
has been concerned with issues of representation consider, for example, the collection
of articles in Brachman & Levesque's (1985) reader in knowledge representation, and that
in the more recent volume edited by Sowa (1991). Schank's work from the 1970s onwards

430

10. Alternative approaches to meaning

is an exception: he and his associates have been concerned not only with forms of
representation (scripts, conceptual dependency) but also with an ontology of primitive acts
and states (with emphasis on the former) out of which word senses can be built. We shall
discuss this work briefly in Section 10.4.1 below.
In the last decade or so, interest in comprehensive ontologies has grown, and the work
on the Penman "upper model" can be seen as an early and continuing move in this
direction; cf. also Klose et al's (1992) work on the ULOG project Other NLP research
has usually been lexically oriented and thus complementary to our own work. This
includes research on the semantic organization of a dictionary as embodied in its denning
vocabulary, e.g. Amsler (1981); the thesaurus-like organization of 'concepts' in the
"Concept Dictionary" of the Japan Electronic Dictionary Research Institute (e.g. Technical
Report 009, 1988); the WordNet system of Princeton University, developed to reveal
lexical semantic relations for an extensive set of lexical items (see Miller & Fellbaum,
1991); the Lexicon Project at MIT, by B. Levin and others (cf. Levin, 1993); research by
Okada and associates at Kyushu Institute of Technology (e.g. Nakamura & Okada, 1991);
and Hobbs' (e.g. 1984,1987) Tacitus project, where the aim was to derive a naive world
view from careful and detailed study of word senses. One approach with roots in both
linguistics and computational linguistics is Dahlgren's (1988) naive semantics, which
includes an ontology (see Section 10.4.2 below).
10.4.1 Schank's conceptual dependency

Schank's basic approach was developed in the 1970s (Schank, 1972; Schank & Abelson,
1977) and has been tested and refined since then in numerous systems and accounts
(Schank, 1982; Schank & Kass, 1988). The current approach includes (i) an account of
macro-organization (scripts and, later also, memory organization packages), and (ii) an
account of micro-organization (primitive acts and states entering into conceptual
dependency (CD) relations).
Micro-semantic organization is handled by CD, which was the first part of the model
to be developed, and is characterized by Schank and Kass (1988: 182) as "a theory about
the representation of the meaning of sentences". CD adheres to a number of principles: the
representation should be canonical (only one representation for sentences with identical
meaning), unambiguous (a given CD structure should have only one meaning), and
explicit (grammatically implicit information should be made explicit). Further, CD
representation is intended to be "language free" an early motivation for this was the
need for a "language free" representation in machine translation work
CD is very much a theory built on what we interpret as the extending subtype of
expansion (see Chapter 3, Section 3.4 above) it is a decomposition^, syntagmatically

10.4 Work on meaning in NLP

431

oriented theory, and in this respect it is similar to Jackendoff s theory discussed in Chapter
4, Section 4.1. Schank & Kass (1988:184) use a metaphor of matter:
The basic pro positional molecules of CD are called conceptualizations. A
conceptualization can represent either an action or a state. Conceptualizations consist
of a main predicate and some number of case-slots. For action conceptualizations the
main predicate is one of the primitive actions (called ACTs). For state
conceptualizations it is one of the primitive states. Each slot can be filled by either a
symbol or another conceptualization. Action conceptualizations have an actor, object,
direction and sometimes an instrument case. The instrument is another action, which
was performed in order to accomplish the main action. State conceptualizations specify
an object and the value (along some arbitrary scale) of some state the object is in.

The distinction between actions and states is familiar, we meet it in many approaches in
one form or another. Composition in CD is clearly not rank-based: relationship between
different (non-primitive) actions is shown in terms of decomposition rather than systemic
agnation. ACTS correspond to our figures, with the exception that their complexity (i.e.
their potential for expansion) is interpreted in terms of extensions in composition rather
than in terms of elaboration in delicacy. This means that ACTS often consist of other
acts, whereas figures only do so only under restricted conditions (e.g. as the Phenomenon
of a figure of sensing, for example the eagle landing in Can you see the eagle landing?).
Let us consider an example provided by Schank & Kass (1988: 186); we have indicated
ACTS in bold:
English: Mary gave John her car
CD: (ATRANS (ACTOR (PERSON (NAME MARY)))
(OBJECT (PHYS-OBJECT (TYPE CAR)))
(DIRECTION (FROM (PERSON (NAME (MARY))))
(TO (PERSON (NAME (JOHN))))))
English: Mary read "Fine Dining in New Haven"
CD: (MTRANS (ACTOR (MARY))
(OBJECT (INFORMATION)
(DIRECTION (FROM (BOOK (NAME "Fine Dining ...")))
(TO (CP (PART-OF (MARY)))))
(INSTRUMENT
(ATTEND
(ACTOR (MARY))
(OBJECT (EYES (PART-OF (MARY))))
(DIRECTION
(TO (BOOK (NAME "Fine Dining ...")))))))

The first example corresponds to a figure of doing and, since 'giving1 is quite general
in the typology of acts of transfer, it is represented by the simple ACT ATRANS ("the
transfer of an abstract relationship such as possession, ownership, or control"). Tlie
Recipient role is treated as the 'destination' of DIRECTION. The second example
corresponds to a figure of sensing. The top-level ACT is MTRANS ("the transfer of
information between animals or within an animal"). The roles are the same as for
ATRANS except that there is also an INSTRUMENT role filled by another

432

10. Alternative approaches to meaning

conceptualization whose ACT is ATTEND ("the action of attending or focussing a sense


organ towards a stimulus"). So the interpretation is roughly Mary transfers information
mentally from "Fine Dining ..." to part of her by directing her eyes to "Fine Dining ..." '.
It is intriguing that this analysis of 'read* treats it as a form of expansion specifically
an enhancing sequence.
ATRANS, MTRANS, and ATTEND are drawn from a set of eleven primitive acts.
We list them here grouped accotding to our primary differentiation among figures; the
grouping is based on our interpretation of the definitions and examples given by Schank
& Kass (1988: 184-5): see Table 10(2).
Table 10(2): Schank's primitive acts and our types of figure
subdomain of figure in the ideation base
doing

sensing (or as behaviour)


saying (but also other)
being & having

primitive act
ATRANS
PIKANS
PROPEL
MOVE
GRASP
INGEST
EXPEL
MTRANS
MBUILD
ATTEND
SPEAK
(primitive states such as JOY, ANGER,
HEALTH, ...)*

* It seems that states correspond to figures of being & having of the ascriptive intensive type.

10.4 Work on meaning in NLP

433

Fig. 10-4: Lexical system network adapted from Goldman (1974)


We see how Schank's Conceptual Dependency can be used to differentiate among nonprimitive acts if we consider an example from Goldman's (1974) account of lexical
differentiation and choice in his BABEL system. Goldman differentiates among word
senses in a discrimination network by using ACTS, and restrictions on the fillers of their
roles. Thus, for example, given a conceptual representation 'henry + ingest + beer' (where
INGEST is one of the primitive ACTS from the listed quoted above), the lexical network
will allow us to select 'drink' for 'ingest', since "beer' is liquid.
In generation, BABEL is presented with a conceptual structure. As BABEL traverses
the discrimination net, it checks this conceptual structure against defining characteristics
each time it comes to a node in the net. The outcome is the selection of a word sense.
Goldman developed fifteen discrimination nets for BABEL. We have adapted one of them
freely and present it here as a system network with value restrictions on roles; it covers
some processes of ingesting: see Figure 10-4 above, (We have maintained the binary
splits of Goldman's discrimination network and also its arrangement as a strict
taxonomy.) The discriminations are based on the reason for ingesting and the nature of

434

10. Alternative approaches to meaning

what is ingested. Hie first inquiry ascertains whether the actor undertakes the ingesting to
improve his or her health; if so, the feature 'for health1 is chosen and the lexical item is
take (as in take cough medicine).
The example illustrates how primitive ACTS can be incorporated into a typology of
(non-primitive) acts; and shows that, when they are interpreted in this way, the 'primitive*
seems to mean 'indelicate*. This is an elaborating rather than an extending notion of
primitiveness; but it leaves unclear the relation of primitive ACTS to other ACTS in
conceptual structures like that with 'read' above. The systemic interpretation given in 10-4
raises various questions. For instance, can the 'primitive* ACTS be organized systemicaily
themselves? What are the implications for a typology of the entities (objects, etc.) that
can serve as fillers of the ACT roles?
CD is intended to be "language free". CD representations cannot support
grammaucalization in particular languages in the way our ideation base does. Thus (i) CD
theory cannot explain why grammar is organized as it is (perhaps less of concern in AI
than in linguistics), since it has no natural relationship with particular grammars; and (ii)
it cannot show how grammar construes experience the contribution made by the
grammar of a particular language would be neutralized, as would, presumably, the
difference between congruent and metaphorical constmals. All other things being equal,
the fact that CD representations do not support grammaticalization would seem to be a
serious drawback. One could argue, as the proponents of CD have done, that this is in fact
part of its virtue it opens up the possibility of a language neutral representation.
However, that is not the only path to take (cf. Chapter 7): it is possible to construct a
multilingual ideation base that is an assemblage of perspectives from different languages
(Bateman, Matthiessen & Zeng, forthc.; Zeng, 1996).
CD was the first pan to be developed of a more extensive account While the domain
of CD was the sentence, Schank & Abelson (1977) proposed a form of organization that
we can recognize as being based on context of situation their script, "essentially a
pre-packaged inference chain relating to a specific routine situation" (Schank & Kass,
1988:190). The script that has become most famous is the restaurant script roughly a
sequence of conceptualizations. Scripts can be used to support processes in text generation
and understanding. In systemic theory and in Firthian system-structure theory, there has
been comparable work on the organization of situation types into stages: first the
pioneering work by Mitchell (1957), and later the work by Hasan (1978; 1984b), followed
up in various studies, and applied to text generation by Cross (1991). One difference is
that in the systemic work situation types were always seen as part of a family or typology
of contexts of situation, variable according to field, tenor and mode (see Martin, 1992).
Schank (1982) introduced a new concept, that of Memory Organization Packages (MOPs)
as successor to the earlier scripts; these are linked to one another in an abstraction
hierarchy, and they are also organized into 'scenes' such that one scene may be shared

10.4 Work on meaning in NLP

435

across several MOPs. This strongly suggests a rank-like organization projected onto the
ideation base, in the form of an ascending sequence of acts/ conceptualizations, scenes,
MOPs and sequences of MOPs. A script might be interpreted as a conventionalized
sequence projected onto the ideation base from a particular context of situation.
10.4.2 Dahlgren's naive semantics

Dahlgren (1988) and colleagues have developed a valuable, fairly extensive account of
semantic organization comparable to our ideation base, one of whose purposes is to
support text processing in a computational system. Dahlgren calls the approach naive
semantics, because it is intended to model "the detailed naive theory associated with
words" (p. 28) that is, in contrast to our approach through the clause, based on
cryptogrammar, the move into the system in naive semantics is through words. (Since it
is concerned with word senses, its task is obviously not to support the grammaticalization
of ideation base information; rather it is to be used for inference based on word senses in
text understanding.) It is proposed as an alternative to theories based on primitives of
meaning, and draws on prototype theory instead. It is intended to reflect both linguistic
and psychological evidence.
Concepts are organized into an ontological hierarchy called ontological rather than
taxonomic because naive semantics is a realist theory. Each node in the hierarchy may
have generic knowledge associated with it (p. 33). This knowledge is organized according
to types of feature such as age, colour, sex, location, has pan; for noun concepts there are
54 such features in all, distinguished into "typical features" and "inherent features" (p. 59).
So, for instance, office, in the sense of a work place, is characterized as follows (p. 60):
{PLACE
<typical:> (haspart (*, chairs),
haspart (*, desks),
haspart (*, typewriters),
location (downtown)},

<inherent:> {haspart (*, doors),


haspart (*, windows),
location (building)}}

That is, to use an identifying clause: an office is a place (typically) with chairs, desks, and
typewriters located in downtown and (inherently) with doors and windows in a building.
It is construed as a participant with extending ('with') and enhancing ('in') roles: the
features are comparable to the roles we have used in characterizing types in the ideation
base. However, the roles we posit are grammatically motivated they are construed in
the grammar as those 'features' associated with participants, figures, and so on. Dahlgren
identifies a rich but quite restricted set of features 54 for nouns (as already noted; p. 74)
and another, smaller, set for relational types (location, time, cause, enablement, p. 99).
The former fall within the domain of participant and, as we would predict, we find e.g.
features corresponding to different kinds of Epithet colour, size, age, etc.. The latter

43 6

10. Alternative approaches to meaning

fall within the domain of figures and, as we would predict, we find features concerned with
enhancing (circumstantial) roles location, time, cause, enablement, consequence,
sequel, manner, and purpose. This is a useful confirmation of the value of approaching the
semantics of verbs from figures, construed in the grammar as clauses as we have done.5
Now, even if the set of features is not arbitrarily large, the possible number of
combinations of sets of features associated with nodes in the ontology is, of course, very
extensive. Dahlgren demonstrates that the combinations are constrained through kind
types (for nouns) and relational types (for verbs), She characterizes these "correlational
constraints'1 as "commonsense reflections of the actual world" (p. 69): "kind types are
types of kind terms with predictable feature types in their descriptions". For instance, the
type ENTITY from the noun ontology has the feature types liaspait' and 'partof associated
with it, and the more delicate type PHYSICAL adds 'color1, 'size', and 'texture'. From the
point of view of our ideation base, these then are types with associated realization
statements; and one central reason for recognizing them as types in the system network is
precisely the 'features' they have associated with them. More delicate types simply inherit
the 'features' from less delicate types.
The typology of entities proposed is set out in Figure 10-5 (based on Dahlgren, 1988:
Chapters 2 and 4). The nominal kind types in this ontology are the following (p. 74):
entity, sentient, living, animal, physical, social, role, person, prepositional, artefact, and
institution.
The relational subontology reflects Vendler's (1967) typology, which we shall present
and comment on in Chapter 12, Section 12.2; it adds the distinctions 'mental/ emotional/
nonmentaT and (for events) 'goal/ nongoaT. The evidence for these is partly psychological;
but there is not enough detailed discussion of the individual types, or linguistic evidence
for the distinctions, to enable us to compare the overall typology with our own.

It would be helpful to be able to review the full set of features used in Naive Semantics and
interpret them systematically in terms of our ideation base. But most of the features are only
listed, and it is difficult to read off their meaning with enough certainty to support the
interpretation.

10.5 Stratal perspectives on meaning

43 7

10.5 Stratal perspectives on meaning


To conclude this brief survey of alternative approaches to meaning, let us raise the
question of how the system of meaning, the semantic system, relates to other systems
within and outside language. Given our stratal interpretation of the semantic system as an
interlevei, we can characterize the perspectives on meaning that may be taken by a
semantic theory along the following lines:
(i) intra-stratal (internal organization of semantics): a theory of
meaning may focus on the organization of meaning itself using some framework
for describing sense relations e.g. the paradigmatic organization of the
meaning potential, as in our present work; the dimensions of semantic agnation,
as in (classical) componential analysis; the syntagmatic decomposition of senses,
as in Katz & Fodor's theory, in interpretive semantics and in generative
semantics.
(ii) inter-stratal: a theory may focus on the stratal interfaces of the semantic
stratum how meaning relates upwards to context (according to our approach)
and how it relates downwards to lexicogrammar:
(11.1) inter-stratal: upwards (semantics in relation to context): the
question of how meaning relates upwards to context is foregrounded in rhetoricalethnographic approaches rather than in logical-philosophical ones. We discussed
the contextual perspective on meaning in Chapters 8 and 9 (see e.g. Figure 8-1),
focussing on the relationship between differences in contextual settings of field,
tenor and mode, and on registerial variation in meaning.
(11.2)
inter-stratal: downwards (semantics
in relation
to
lexicogrammar): the interface between semantics and lexicogrammar is
internal to language and has received far more attention in studies of meaning
from all standpoints than has the interface between semantics and context. In the
logical-philosophical approach, within generative linguistics, inteipretive
semantics has focussed on the question of how semantic representations can be
derived from below, from syntactic ones; and an important aspect of the debate in
the late 60s and early 70s was precisely concerned with the directionality of interstratal mappings and the nature of the inter-stratal boundary. One key question
that emerged, particularly in the 1970s and early 1980s, was whether syntax is
autonomous or not. In the standard Chomskyan theory it was; but this was
rejected by Montague and those who were influenced by his idea of building
syntactic and semantic specifications "in tandem" (as in the successive
developments of GPSG and HPSG). Within the ihetorical-ethnographic

438

10. Alternative approaches to meaning

approach, we have taken the position that not only is lexicogrammar not
autonomous, but it is natural in relation to semantics: our approach to the
ideation base rests on this theoretical assumption (cf. Chapter 1, Section 1.2
above). This is what explains the further possibility of grammatical metaphor,
opened up at the interface between semantics and lexicogrammar.

Fig. 10-5: Dahlgren's typology of entities


(iii) extra-straUl (relation between language and non-semiotic
systems): the theory may focus on how language in general and semantics in
particular relate to systems of other kinds, i.e. to non-semiotic systems. We have
already discussed the way in which this focus has been built into transcendent

10.5 Stratal perspectives on meaning

439

approaches to meaning, where meaning is 'exported' into the realms of the


material world or the mind.
The strata! perspectives on meaning are represented in Figure 10-6. We are not
suggesting that a given theory deals only with one of the three areas of strata! concern.
Interpretive semantics is clearly concerned both with the mapping between syntax and
semantics and with the internal organization of meaning in terms of decomposition (as in
Jackendoff, 1972). Looked at from the standpoint of a strata! interpretation of language,
these different approaches are not so much mutually exclusive alternatives as they have
often been thought to be as complementary perspectives on meaning: they focus on
different aspects of the stratum of semantics its internal organization or its interfaces to
other systems, linguistic, conceptual or physical. This is not to say, of course, that they
could all be put together into one internally consistent theory of meaning! For example, it
is hard to reconcile a model-theoretic approach with an experientialist cognitive one or
with our own immanent approach. However, this does not mean that, from our
standpoint, it does not make sense to model the steps that relate semiotic systems to
material ones. It does; but the relationship has to be modelled in such a way that we can
show how people as biological organisms and socio-semiotic persons interact with their
material environment. We leave these issues here; but we shall return to them in the final
part of this book. After problematizing the construction of "the mind" in Chapter 14, we
shall discuss how the semantic system relates to other systems in Chapter 15, Section
15.1.

440

10. Alternative approaches to meaning

Fig. 10-6: Approaches to different aspects of semantics

11. Distortion and transformation


In the preceding chapter, we compared alternative interpretations of the semantic system in
terms of how it is located relative to a global model of language (and of whatever is
conceived of as its environment). But if the task is to build a knowledge base, then the
question arises to what extent language is involved at all. The interpretation of the
ideational semantics of English could obviously be different from what we have proposed
here. Leaving aside approaches to meaning that do not proceed from linguistic
interpretation (cf. Chapter 10, Section 10.1 above), there are others that contrast with ours
but are still grounded in language, with reference points either in linguistics or in
philosophy. Others again fall somewhere in between a linguistically motivated model and
a model supported on other grounds: for example, theories of deep cases in case grammar
(e.g. instruments and beneficiaries) and theories with semantic terms derived from
predicate logic. The latter can lead to the argument that, since predicate logic is being used
to represent the semantic structure of English, this structure must be like that of predicate
logic; that is, they can lead to arguing not from the object language but from the
metalanguage. We shall begin by examining briefly the possible frames of reference that
have been adopted in work on ontologies and on ontological categories (Section 11.1). We
then turn to the notion that language cannot be trusted as a frame of reference because it
'distorts1 thought or reality, and to the related notion that syntax cannot be trusted because
it 'distorts' semantics (Section 11.2). Finally, we discuss the typological status of
qualities, as an illustrative example of the tendency to interpret one class of phenomenon
on the model of another (Section 11.3).

11.1 Frames of reference


In Figure 10-6 above we summarized different perspectives on meaning intra-stratal,
inter-stratal and extra-strata! ones. If we set ourselves the task of creating an ontology for
a "knowledge base", each of these perspectives might serve as a frame of reference. (There
are others as well, to be mentioned below.)
From our standpoint, the choice of frames of reference follows from our conception of
the "knowledge base" as a "meaning base" (see Chapter 1, Section 1.1 above), where our
conception of meaning is immanent (see Chapter 10, Section 10.1 above). Throughout
our discussion of the organization of the meaning base, we have made reference both to
intra-stratal considerations (such as patterns of agnation and the transphenomenal types
that emerged in the course of our exploration) and to inter-stratal considerations. With
respect to the latter, we have foregrounded considerations 'from below1, from the stratum
of lexicogrammar (see Chapter 1, Section 1.2 and Chapter 2, Section 2.2). There were
two main reasons for this: on the one hand, the meaning base has to be realized in waded
texts and the statements of realization will be simpler if the resources of wording are part
of the picture from the start; on the other hand, the relationship between meaning and
wording, between the system of semantics and that of lexicogrammar, is a natural one:

442

11. Distortion and transformation

they are both strata of the content. Although it has not been part of the central argument
about particular distinctions in the meaning base, we have also referred to the relationship
between semantics and context (Chapter 8, Section 8.1 and Chapter 9, Section 9.1). Just
as the meaning base has to be accountable lexicogrammatically, it also has to be
accountable contextually. In our discussion, we emphasized register or functional variation
as one of the keys to the relationship between semantics and context.
If a different standpoint is adopted, the frame of reference may be an extra-semantic
one: either because the approach to meaning is transcendent rather than immanent or
because the object of modelling is taken to be knowledge rather than meaning. Work on
participant roles and circumstance roles in figures, cast in terms of "case relations" or
"deep cases'1, provides a good example of constrasting frames of reference. Starosta (1988:
115-7) contrasts two approaches to such roles:
Case relations (CRs) are syntactic-semantic relations obtaining between (nonpredicate) nouns and their regents, which can be verbs, prepositions, or other nouns.
CRs include the kind of intuitive categories characterized in traditional and modem
generative grammar by terms such as 'Agent', Patient', etc. Their function is mediating
between overt grammatical configurations (word order, case inflections, etc.) and
semantic role concepts and scope phenomena, and facilitating discourse cohesion in
languages such as Dyirbal.
Case grammar in the Fillmorean tradition is based on a not necessarily linguistic
intuition about actions or processes or states, participants in those actions, processes,
and states, and the roles played by the obligatory or optionally present participants in
these actions, processes and states. These roles are the 'case relations1 or 'deep cases1 in
classical case grammars and many computational applications, and the 'thematic
relations' in the Chomsky an tradition, which is based on Gruber's work (Gruber, 1965).
In practice, such roles are established without any necessary connection with language
at all: one simply envisions a kind of silent movie of an action, for example a person
loading hay onto a wagon with a pitchfork, identifies the elements necessary for such
an action to take place (someone to do the loading, something to be loaded, someplace
to load onto or into, and something to load with) and assigns a 'case relation' (or
'thematic relation') to each one. Then the case relations in any sentence referring to that
situation must correspond to the previously established silent movie roles. Because this
determination is independent of the particular way in which the action is described
linguistically, case relations in this approach are constant across sentences and across
languages.
The situation-oriented procedure is reflected in the use of paraphrase to identify case
relations in both the Fillmorean case grammar and the Gruberian thematic relations
traditions: two sentences are paraphrases if they have the same troth values, which is a
more precise way of saying that they characterize the same external situations. ... In
effect, what this practice amounts to is constructing a typology of situations, which is
independent of language, so that case relations are necessarily universal: ...

As Starosta describes them, the considerations brought to bear on "case relations" ate
thus derived from an extra-semantic perspective. In contrast, Starosta's own approach is
language-based (op cit., p. 119):
The lexicase approach to case relations differs from the usual Fillmorean case relations
and Cbomskyan thematic relation analyses in that its definitions are stated not in terms
of external situations directly, but rather in terms of sentence-specific PERSPECTIVES
of those situations ... Thus if two different sentences refer to the same situation but
portray if from different viewpoints, they may contain quite different arrays of case
relations. ... the redefinition of case roles in terms of perspective rather than situation

11.1 Frames of reference

443

results in a dramatic increase in the language-specific and cross-typological


explanatory power of the theory, as reflected in the large number of generalizations that
can be stated assuming this sort of definition but not assuming a Fillmorcan situational
one.

Our own approach is also language-based- participant roles and circumstance roles in
the figures are based on ultra-semantic considerations (e.g. the transphenomenal types of
projection and expansion) and on inter-stratal considerations from below, from
lexicogrammar. The original systemic work on transitivity roles in the clause (Halliday,
1967/8), which appeared at about the same time as Fillmore's (1968) Case for Case, had
the language-based frame of reference advocated by Starosta.1 For further discussion of the
issues involved from a systemic point of view, see Martin (1996 a, b). We will compare
and contrast certain proposals in this area in Chapter 12 below.
Similar considerations to those just exemplified with respect to "case relations" apply
to all areas of the organization of a "knowledge base". In a discussion of a particular
ontology for a knowledge base (Wissensbasis), the LILOG ontology, Lang (1992, in
Klose et al, 1992) differentiates between ontological types that have a conceptual base and
ontological types that have a linguistic base:
Wenn eine Sorte nient bloB ein etikettierter Knoten in der Verbansstruktur sein soil,
sondera legitimiertes Element einer Wissensbasis, die (i) auBersprachliches
Alltagswissen reprasentiert und zwar so, dafi sie (ii) ber (deutsche) Texte zugreifbar ist
und (deutsche) Texte aus ihr generierbar sein so 11 en, dann stellt sich die keineswegs nur
sophistische Frage (F):
(F) 1st die gegebene Sorte X
(a) eine konzeptuelle (d.h. sprachunabhngige) {Catgorie, deren Bezug zu
sprachlichen Einheiten zwar regular, aber prinzipiell arbitrar ist und auch
einzelsprachlich variiert,
oder
(b) verdankt die Sorte X ihre Aufnahme in die Ontologie dem Vorhandensein under
haufigen Okkurenz eines gleichnamigen Ausdrucks in den (deutschen) Texten, d.h.
in der Textbasis, die der Ontologie-Konstruktion als Ausgangsmaterial zugrunde
liegt?

As examples of conceptually based types he cites *EmitaV, 'Zustand', 'Ereignis', and


'Situation*.

We should note that FiUmore (1968) was carefully worked out, taking grammatical criteria
into consideration, including Whorfian cryptotypes.

444

11. Distortion and transformation

From our standpoint types motivated in terms of Lang's category (b) clearly belong to
the ideation base; and they aie motivated intra-straially and inter-stratally. Types motivated
in terms of category (a) would (by definition) be extra-semantic. Consequently they would
fall outside the ideation base. However, if we reconstme knowledge as meaning, it follows
that we cannot simply refer such types to an extra-semantic conceptual system (as could
be done in a cognitive approach to semantics; cf. Chapter 10, Section 10.3 above): we
would have to explore the possibility that these types are construed as meaning by some
semiotic system other than language. We shall return to the relationship between
language and other semiotic systems in Chapter 15, Section 15.1. Here let us just
reiterate our view that all of experience is construed as meaning. Language is the primary
semiotic system for transforming experience into meaning; and it is the only semiotic
system whose meaning base can serve to transform meanings construed in other systems
(including perceptual ones) and thus integrate our experience from all its various sources
(cf. Sugeno's 1993 application of the notion of data fusion).
It might be objected that this view leaves no room for scientific or metaphysical
models for example, that we do not allow for the possibility that science has advanced
our understanding of the world. This objection would be misplaced: such models ate
construed in the ideation base as domain models within the overall meaning potential (cf.
Chapter 8, Section 8.1). We shall return to the question of different models in Chapter 14.

11.2 The motif of distortion


Attempts to create an ontology without reference to language in general, or without
reference to lexicogrammar in particular, seem to derive from the belief, familiar in the
history of Western thinking, that language comes between us and a 'real* or scientific
understanding of the world, that it somehow somehow distorts our awareness of reality.
There are two somewhat distinct versions of this belief. The first is the notion that
language distorts reality or, as a variant of this, that language distorts our thinking
(which includes our thinking about reality). This is extra-linguistic deception:
language is deceiving us by the way it represents something else. The second is the
notion that syntax distorts semantics. This is intra-lingiiistic deception: language is
deceiving us by the way one part of it represents another part. We will say a little about
each of these in turn.
11.2.1 The notion that language distorts (our thinking about) reality
The view that language distorts the picture of reality, and that there is a mismatch
between language and thought, is reflected the opposition of "deep" and "surface" as these
figured in a prominent approach to text generation. In this model, "deep generation" is
concerned with thinking what to say (the thoughts "behind" the words) and "surface
generation" with how to say it (see e.g. McKeown & Swartout, 1987). This view was

11.2 The motif of distortion

445

discussed many years ago by Whorf, in tenns which aie still relevant today; Whorf refers
to it as the "natural logic" view:
According to natural logic, the fact that every person has talked fluently since infancy
makes every man his own authority on the process by which he formulates and
communicates. He has merely to consult a common substratum of logic or reason which
he and everyone else are supposed to possess. Natural logic says that talking is merely
an incidental process concerned with communication, not with formulation of ideas.
Talking, or the use of language, is supposed only to "express" what is essentially
already formulated nonlinguistically. Formulation is an independent process, called
thought or thinking, and is supposed to be largely indifferent to the nature of particular
languages. Languages have grammars, which are assumed to be merely norms of
conventional and social correctness, but the use of language is supposed to be guided
not so much by them as by correct, rational, or intelligent THINKING.
Thought, in this view, does not depend on grammar but on laws of logic or reason which
are supposed to be the same for all observers of the universe to represent a rationale
in the universe that can be "found" independently by all intelligent observers, whether
they speak Chinese or Choc taw. In our own culture, the formulations of mathematics
and of formal logic have acquired the reputation of dealing with this order of things:
i.e., with the realm and laws of pure thought. Natural logic holds that different
languages are essentially parallel methods for expressing this one-and-the-same
rationale of thought and, hence, differ really in but minor ways which may seem
important only because they are seen at close range. It holds that mathematics,
symbolic logic, philosophy, and so on are systems contrasted with language which deal
directly with this realm of thought, not that they are themselves specialized extensions
of language. (Whorf, 1956: 207-8)
After characterising the position of 'natural logic' in this way, Whorf (op cit.: 211)
goes on to identify two problems with it:
Natural logic contains two fallacies: First, it does not see that the phenomena of a
language are to its own speakers largely of a background character and so are outside the
critical consciousness and control of the speaker who is expounding natural logic.
Hence, when anyone, as a natural logician, is talking about reason, logic, and the laws
of correct thinking, he is apt to be simply marching in step with purely grammatical
facts that have somewhat of a background character in his own language or family of
languages but are by no means universal in all languages and in no sense a common
substratum of reason.
Second, natural logic confuses agreement about subject matter, attained through use of
language, with knowledge of the linguistic processes by which agreement is attained:
i.e., with the province of the despised (and to its notion superfluous) grammarian.
This belief in the distorting effect of language was propounded by the early European
humanists, who held that medieval scholars had focussed too strongly on language,
whereas the real task of the scientist was to see through the verbal disguise and penetrate
to the reality underneath (for example, Francis Bacon's well-known warnings against the
seductive power of natural language). Natural languages were considered to be inadequate
vehicles for the new scientific knowledge; hence it was necessary to construct artificial
languages to record, transmit and extend it. These artificial languages would, it was
thought, be more in harmony with the objective world of experience.
This same attitude continues to prevail in the ways people talk about language in our
own time. It dominated much of the early work on machine translation in the 1950s aid

446

11. Distortion and transformation

1960s; the task of the analysis was seen to be that of stripping the underlying ideas of
their linguistic disguise (Firth, 1956, referred scathingly to current formulations according
to which language was a "clothing" for "naked ideas"). When the "interlingua" model was
proposed, many of those working in the field regarded it not as a. construction of meaning
that would be a compromise among different linguistic systems but as a language-free
representation of concepts and conceptual structures (cf. Schank's conceptual dependency
in Chapter 10, Section 10.4.1 above), very much in the 17th century tradition.
The same view of language as distortion is frequently to be met with when language is
contained within a model of communication, where a common motif is that language is a
vehicle for lying, or at least for concealing the truth. One might surmise that, if language
is defined in communicational terms, as a means of transmitting information
especially if this is combined with a semantics based on considerations of truth then
this is how it is likely to appear.
11.2.2 The notion that lexicogrammar [syntax] distorts semantics

The view that syntax distorts semantics implies that the relation of grammar to meaning
is indirect and arbitrary. This view became tenable in modern linguistics, where meaning
was either excluded from its scope altogether, as among structuralist linguists in the U.S.,
or, with Chomsky, kept at a distance by the metaphor of deep and surface structure in the
syntax, only the former being semantically responsible. This paved the way for a number
of analyses on the model of 'surface x is really deep y'. We find suggestions such as the
following: adjectives are really verbs (e.g. Chafe, 1970), nouns are really verbs (cf. Bach,
1968), pronouns are really articles (Postal, 1966), negation is really a [higher] verb, tense
is really a [higher] verb (cf. Huddleston, 1969), auxiliaries aie really full verbs, verbal
group complexes are really reductions of embedded clauses, moods are really separate
clauses of saying, and so on.
For example, Bach (1968) argued that the distinction between nouns, verbs and
adjectives is a superficial one and that nouns are really something else:
Grammars of English traditionally maintain a sharp division between nouns, verbs, and
adjectives. It is my purpose here to demonstrate that the differences between these 'parts
of speech* exist only on a relatively superficial level and that the fundamental
dichotomy underlying the distinctions is of quite a different sort. By saying 'sentences'
rather than 'English sentences', I intend to suggest that the deep structures of sentences
in different languages are identical; that is, I am subscribing to the idea of a universal
set of base rules, (p. 91)
To summarize, I have argued on the basis of many pieces of evidence that it is
reasonable to suppose that all nouns come from relative clauses based on the predicate
nominal constituent. ... I have tried to show that the distinctions between such parts of
speech as nouns, adjectives, and verbs have no direct representation as such in the base,
but are results of transformational developments in one or another language. ... The
base component here looks in some ways very much like the logical systems familiar
from the work of modern logicians like Rudolf Catnap, Hans Reicbenbach, and others.
In particular such systems do not have any subdivision of lexical items' into nouns,

11.2 The motif of distortion

447

verbs, and adjectives. Much more basic is the distinction between variables, names, and
general 'predicates'... It should not be surprising that a system of universal base rules
should turn out to be very close to such systems, which are after all the result of
analyzing the most basic conceptual relationships that exist in natural languages. Such
a system expresses directly the idea that it is possible to convey any conceptual content
in any language, even though the particular lexical items available will vary widely
from one language to another a direct denial of the Humboldt-Sapir-Whorf
hypothesis in its strongest form. (pp. 121-2)

Such analyses were often supported by universalist arguments such as "negation is a


verb in certain languages, so it is reasonable to claim that it is really a verb in all". As
the quote from Bach illustrates, they tended to make deep structure, the 'real1 structure,
look like predicate logic. Surface structure came out looking like a (transformationally)
twisted version of logical structures. But predicate logic had been derived from one
particular area of the grammar, a simplified version of the experiential aspect of the
clause; it could be used as an idealized model of certain types of figure, for the purpose of
explicit rule-based reasoning, but it was not intended to be a tool for analysing the entire
semantic structure of a natural language.2 This view has largely been abandoned and the
notion of a semantically irresponsible surface structure is no longer generally held.
In rejecting both these views of language as distortion, we ate not propounding an
alternative version according to which language is a perfect match. What is wrong with all
such conceptions is that they misconstrue the nature of a semiotic system the
fundamental relation of realization to which we are always having to return. A semiotic
system is not some kind of outer garment which may either reveal or conceal what is
beneath. Rather, it is a transformation of experience into meanings, and each stratum
within the system is construed by, and construes, all the rest. A "language", in this sense,
may be artificially constnicted or engineered, like a scientific theory or a logic; but all
such semiotics are ultimately related to natural language, and natural language is still an
accomplice in their overall construction of reality.
But whether or not it is engineered in this way, natural language will continue to
evolve. The artificial languages of the 17th century were never actually used; but this d
not mean that the forms of natural language persisted without change. On the contrary,

In tenns of our model of a stratified metalanguage (see Chapter 1, Section 1.9.1 above), we
can see that taking the categories of predicate logic (or any other logical systems) to be
linguistic ones constitutes a strata! slippage: categories from the level of representation in
the metalanguage are imported into the theoretical account of the object language.

448

11. Distortion and transformation

new registers were always evolving, some of them as pan of the ongoing reconstruction
of experience in the form of systematic knowledge and experimental science.

11.3 The status of qualities: an illustration


We shall not attempt to document the views referred to above in any greater detail. But it
seems important to take them into account as part of the overall picture; not just because
such views, in their turn, distort the nature of language, but also because they can
obstruct progress in particular language-based activities for example, the practice of
"intelligent computing", which is computing carried out entirely in natural language
(Sugeno, 1993). In the remainder of this chapter, we will illustrate these notions by
considering just one of the types of element that make up a figure, namely qualities, and
showing how the way in which these are described can be influenced by such thinking.
This will refer mainly to the idea of tntra-linguistic distortion a mismatch between
syntax and semantics; but one can often recognize an implication that behind this lies a
distortion of reality. In Chapter 12, when we come to look at a number of different
descriptions that stand as alternatives to our own, we will also find certain traces of this
general perspective on language.
The problem of classifying qualities is often approached grammatically as a question
of word classification: are adjectives funny verbs, funny nouns, or a completely distinct
word class? All three questions have been asked, and all three answered in the affirmative.
The questions themselves reflect one of the biases of traditional grammar: it is based on
words. Our approach to these questions is a functional one, which does not start with
word classes taken out of context; so here we will begin our enquiry with the clause.
Semantically, the relevant question is how a figure is organized in English. This allows
us to look at qualities 'from above', keeping their semantic environment in view.
11.3*1 Some significant transitivity types
There are two transitivity types that will be significant for arguments concerning the
location of qualities in relation to other semantic classes: (1) middle material clauses, and
(2) middle relational clauses. These types were described in Chapters 2 and 4; they are
distinguished in terms of both process type and agency. Some culinary examples are:
middle material (happening):
Add the mushrooms, washed and chopped, and cook until all
liquid from the mushrooms [Medium/Ac tor] has evaporated

the

[Process]

11.3 The status of qualities: an illustration

449

middle relational (ascriptive):


Use a little stock or apple juice, if the stuffing
[Medium/Carrier] is [Process] too firm [ Range/At tribute 3 .
This [Medium/Carrier] makes [Process]a good stuffing for a whole
cabbage [Range/Attribute].
[Medium/Carrier]

They

will

become

[Process]

discoloured

[Process]

very good

[Range/Attribute],
but

the flavour

[Medium/Carrier]

is

[Range/Attribute].
These two types realize figures of (i) doing & happening and (ii) being & having
respectively. Let us review their structures and discuss their distinct discourse
contributions.
(1) The material clause above represents a happening in which the Medium (all the
liquid from the mushrooms) participates; the happening is a change from liquid to gaseous
state. The clause indicates that the cooking should continue as long as the liquid keeps
evaporating. We can represent the semantic structure that is construed in this way by
Figure 11-1.

Fig. 11-1: Figure of happening

450

11. Distortion and transformation

In recipes, material clauses form the backbone of the procedure. This is something these
texts have in common with narratives: here is a traditional tale that is moved forward by
material clauses alone.
Jack and Jill went up the hill to fetch a pail of water. Jack fell
down and broke his crown, and Jill came tumbling after. Up Jack
got and home he trot as fast as he could caper; went to bed to
mend his head with vinegar and brown paper.

Material processes lend themselves naturally to sequential ordering in time; this is much
less a feature of other process types.
(2) The relational clauses above represent descriptions in which the Medium is
related to a quality or a class. More generally, the Medium is related as a member of a set,
which is defined either by a quality or by a class. The relation is a composite of the
participants (the Carrier and the Attribute) and the nuclear relation. The nuclear relation is
not necessarily a state; it can be either a being Cbe' and 'make* in the examples above) or a
becoming ("become" above), both of which are located in time, as are processes in general.
But its participants are static things; the Carrier is an individual or class, and the Attribute
is a lasting quality or a wider class. The Attribute of the relation of becoming applies to
the Carrier in the final state of the becoming; it is a resultative Attribute. We can
represent the semantic structure of the last relational clause as in Figure 11-2.

Fig. 11-2: Figure of ascripuve being


Descriptions and characterizations draw heavily on the grammatical resources of
transitivity to represent relational processes. Consider for example the following entry for
"duck" in The New International Encyclopaedia:

11.3 The status of qualities: an illustration

451

DOCK. (...) The ducks are a large and well-defined group of birds
of

the

order

Anseres

and

the

family

Anatidae.

distinguished from the swans and geese by having

They

are

the tarsi

scutellate in front, and the sexes are unlike in color; and they
are separated from the mergansers by the shape of the bill, which
is broad and flattened .
The term "duck" is generally restricted to designate the female,
while "drake" is the term applied to the male.
The ducks are largely animal feeders (insects, snails, frogs,
fish, etc.), although some species are noted for their fondness
for wild celery and other vegetables.
The legs are short and placed far back, so that ducks move with
difficulty on land, and with the characteristic gait known as a
waddle, but they are splendid

swimmers, and are noted for their

powers of diving and winning under water.


The neck is short as conpared with geese or swans. There is a
peculiar anatomical feature of the windpipe, consisting of a large
dilatation of the trachea on each side at its bifurcation.
The tongue is large and fleshy and very sensitive.
The plumage is remarkably thick, soft, and conpact.
The wings are stiff, strong, and pointed, giving the power of
rapid and vigorous flight, though the speed of wild ducks has
probably been exaggerated.
Ihe tail is of variable shape and made up of usually 14 or 16
feathers.
Ihe oil gland is always present and well developed, with two
openings, and crowned with a tuft of feathers.
Ihe ducks hav been easily tamed, and many breeds are known
which will be considered later by themselves. (...)

Here the ducks are ascribed to classes (group of a particular order, animal feeders,
swimmers, etc.); distinguishing subparts are specified, and the various parts of a duck

45 2

11. Distortion and transformation

(legs, plumage, etc.) are characterized in terms of their qualitative attributes. Note that the
text is itself organized as a series of elaborating relations (constituency, attribution, etc.).
The functional differentiation between material clauses and relational ones becomes
very dear when we look at texts that draw on both: they make different kinds of
contribution to the creation of the text For example, in a procedural text such as a recipe,
material clauses serve to express the steps in the procedure to be followed. In contrast,
relational clauses occur at a different stage in the text, the stage concerned with the use of
the dish that is the outcome of this procedure (cf. Chapter 9, Section 9.2.1 above):
SWEET CORN STUFFING.
[Ingredients: ...]
[Procdure :] Mlt

the butter and cook the onion in

it

without

browning. Strip the cobs and add the sweet corn to the onion.
Cook for a further 2-3 minutes. R*aov from the heat and add the
remainder of the ingredients.
[Use:] This makes a good stuffing for tomatoes.

Other examples of TJse':


These are lovely when fresh but after a day or two they should be
warmed through before being eaten. They are very pleasant served
with soup or salad.
Young beetroots are very good on their own.
This dish is very attractive in appearance.

11.3.2 Three different transitivity interpretations

The previous discussion shows the importance of recognizing material and relational
clauses as distinct types having complementary roles in the discourse. In our
interpretation, neither type is accorded priority. There are, however, descriptions in which
one of the two is interpreted on the model of the other. Middle material clauses have
served as a model for interpreting middle relational (ascriptive) clauses; likewise middle
relational clauses have served as a model for interpreting material clauses. For the
purposes of our discussion, we can identify three different positions.
[i] According to the first position, a material nuclear process is interpreted as if it
was a participant in a relational process. This gives a logical analysis of the form
"Subject "be1 Predicate": Socrates ran is analysed as "Subject: 'Socrates' 'be'

11.3 The status of qualities: an illustration

453

Predicate: 'running'". In terms of grammatical classes, a verb is a copula plus an


adjective.
[ii] The second position is the reverse of the first in terms of the distribution of the
nuclear process and participants. A relational process is interpreted on the model
of a material one; the attribute (participant) is interpreted as if it was a nuclear
process. Thus, to continue with the same artificial examples, Socrates was white
is analysed as 'Socrates whited1. The copula is not interpreted as representing a
process but is thought to be only a "bearer" of tense, person, and number.
[iii] The third position accepts both the material process model and the relational
process model without trying to interpret one in terms of the other. According to
this model, Socrates was white is not primarily like Socrates ran but is more
directly related to Socrates was a philosopher and, by an additional step, to
Socrates was the teacher of Plato. In this model, nuclear processes are verbs and
participants are nominis (substantival or adjectival).
The three positions are summarized diagrammatically in Figure 11-3.

Fig. 11-3: Transitivity models three positions


The three approaches have different consequences for word class assignments.
According to the first approach, verbs are really adjectives (their true shape is revealed in
the adjectival participle), while the second approach assumes that adjectives are really
verbs; they just happen to be superficially defective (in English) in that they cannot be

454

11. Distortion and transformation

inflected The third approach aligns adjectives with 'substantives', but it still allows for
two possibilities. They can either be treated as independent classes or they can be grouped
together as nouns. The alternatives are summarized in Figure 11-4.

Fig. 11-4: The alternative models in terms of class assignments


The three positions are summarized in Table 11(1) below in terms of both transitivity
model and class assignments. We shall start with interpretations (i) and (ii), where one
configuration is transformed into the other one.
Table 11(1): Transitivity models and approaches to classes

Position

Transitivity model

Classes

[il

relational (ascriptive)

verb is adjective (adjectival participle)

[ii]

material

adjective is (stative) verb

[iii]

relational & material

adjective & verb


[1] adjective is a noun
[2] all distinct

11.3.3 The transforming interpretations: adjectives & verbs vs. nouns

The first two positions identified above transform one configuration into another so that
the first can be interpreted on the model of the second. Similarly, there is transformation
of the classes: adjectives are really verbs or verbs are really adjectives.

11.3 The status of qualities: an illustration

455

Both transforming interpretations can be seen as attempts to find a canonical


logical form, either Subject - Copula - Predicate or Subject - Predicate, rather than
have to operate with different forms. These two logical structurings correspond to
relational and material models respectively:
material
Subject
Actor

relational
- Predicate
Process

Subject - Copula Carrier


Process

Predicate
Attribute

11.3.3.1 Subject - Copula - Predicate (relational model)


As we have seen, of the two approaches that group adjectives and verbs together, one is
based on the model of an ascriptive relational clause: Socrates was white. Interpreted
according to this model, the verb in Socrates ran is really a copula plus an adjective,
'Socrates was running1 (on the possibility of the expansion, cf. Aristotle [On
Interpretation]). This kind of approach was pursued in logic and philosophical grammar in
the Middle Ages (cf. Padley, 1985: 374; Kneale & Kneale, 1962: 206-7, on Abelard); we
also find it in the Port-Royal philosophical grammarians, and in Wilkins' (1668) work.
As late as 1924, Jespersen could note "Logicians are fond of analysing all sentences into
the three elements, subject, copula, and predicate" and observed that linguists "must find
this analysis unsatisfactory". It seems that they do; the approach probably does not have
any adherents any longer (except perhaps as an analysis of the so-called progressive aspect
and the passive voice on a relational model; cf. Langacker, 1978). It might be argued that
the copula is semantically motivated as the locus of (primary) tense or modality and
polarity. However, this is an interpersonal function, not an ideational one: in the grammar
of the English clause, there is indeed an element which serves as the locus of these
categories, viz. the Finite, which is separate, or separable, from the Predicator, and
combines with another interpersonal function, the Subject, to form the Mood element of
the clause (as in [Subject:] they [Finite:] will [PredicatorJ come, [TagFinite:] won't
[TagSubjecu] they?: see Halliday (1985: Ch. 4). This element is present in the verbal
group anyway; there is no need to invent an artificial copula to serve its functions.
While the medieval Subject + Copula + Predicate analysis is no longer favoured, it is
no more far-fetched than the opposite position which interprets relational processes on the
model of material processes. Relationally interpreted, Aristotle runs/jogs /swims

(well)

means that Aristotle is a member of the set of running/ jogging/ swimming entities; and
this is close enough to Aristotle is a (good) runner/jogger/swimmer. But notice how we
have arrived at this apparent correspondence between the material model and the relational
model. We have taken one special case of the material type, where the occurrence of the
material process is not a specific instance but rather a potential that depends on the actor's

11. Distortion and transformation

456

ability (it relates specifically to the simple present tense, whereas in material clauses the
unmarked present tense is the present-in-present Aristotle is running). It is this 'behaviour
potential* that can be taken as the basis for ascribing the Behaver to a class, using the
relational model (Of. the pairs Socrates speaks Greek/Socrates can speak Greek [verbal]
and Socrates is a speaker of Greek [relational]; and the door opens easily 'can easily be
opened* [material] and the door is easy Ufo open 'is an easy one to open1 [relational].)
However, if we turn to cases where the occurrence of the material process is a specific
instance, as in Jack is running up the hill there is no obvious relational alternative.
Similarly, repeated occurrences that constitute habits, as in Socrates swims twice a week,
cannot be represented relaonally (except by assigning Socrates to the class of bi-weekly
swimmers: Socrates is a bi-weekly swimmer).
The correspondences are brought out in texts or segments of texts that describe classes,
i.e. that are essentially relational. Consider the following excerpts from our duck text and
from a geology text. (Agnate variants of the original are given in italics.)
Relational

Material

The ducks are largely animal


feeders (insects, snails, frogs,
fish, etc.), [...]

The ducks feed largely on animals ....

The legs are short and placed far back,

so that ducks move with difficulty on


land, and with the characteristic gait
known as a waddle,

but they are splendid swimmers,

but they swim


[...]

[...]

splendidly,

The tongue is large and fleshy


and very sensitive.
The plumage is remarkably thick,
soft, and compact.

t...]
The ducks are easy to tame or
The ducks are easily tameable

Rivers are agents of erosion,


transportation, and deposition.

The ducks have been easily tamed.

i...]

That is, they carve their own valleys


and carry the eroded material downstream, where it is either deposited
by the river or delivered into a lake
or ocean.

457

11.3 The status of qualities: an illustration

Rivers are the most important agents


in transporting the products of

erosion:

they drain glaciers, and, even in their

infrequent times of flowage on deserts,


they probably are able to carry more
material than the wind can.

(R.J. Foster, Physical Geology, p. 149)


In the geology text, the writer specifies a restatement relation ('that is1, V) between the
relational and material models, which is possible since the text deals with generic classes
of rivers, etc.. Table 11(2) sums up the difference between material and relational
processes depending on the type of occurrence.
Table 11(2): Transitivity and potentiality of occurrence
PROCESS TYPE

TYPE OF OCCURRENCE
habitual
potential
instantiation

material

Socrates swims well

Socrates swims in the


morning

relational

Socrates is a good
swimmer

Socrates is a
matutinal swimmer

instance
Socrates swam across
the river; Socrates is
swimmmg

11.3.3.2 Subject - Predicate (middle material model)


In modern predicate logic, a different view is taken, whereby nouns, adjectives, and verbs
are ail predicates. The second approach we mentioned above brings two of these, adjectives
and verbs, together as a single class of verbs. Here the model is that of as a material
clause, exemplified by Socrates ran ('run (Socrates)1). According to this model, Socrates
was white is really Scrates whited: 'white (Socrates)1. In this view, the adjective is a
surface class, and the need for a copula to cany tense, number, and person is also a
property of surface structure.
One example of this approach is Chafe (1970), Chafe treats adjectives and prepositions
as "semantic verbs". He comments on his position in relation to the traditional analysis
(pp. 159-160):
It runs counter to traditions based on the study of surface structure to consider in a verb
root, as in fact it does to consider wide a verb root, as we did earlier. In the surface
structure of Engk'sh and many other languages, in is traditionally regarded as a
preposition and wide as an adjective. Both are typically preceded by forms of be, which
plays the surface role of verb: The knife is in the box, The road is wide. This be,
however, is surely not the semantic verb root in such sentences. Its function is to carry
the tense and other inflectional units originally attached to roots like m and wide
because the latter, as surface items in English, are incapable of being so inflected. In
addition, we find in English surface phrases like the knife in the box and the wide road
where be is not present and in and be appear to be outside the context of verbs
altogether. As many readers may realize, however, phrases like these are readily
explainable on the basis of semantic structures in which in and wide again are verb

4-58

11. Distortion and transformation

roots. ... It is significant to observe that there are languages in which such items do
appear in surface structures in the form of verbs, not prepositions or adjectives.

In another context, Chafe (1987:34) restates his position as follows:


One clear type of word sequence that may express a single new concept is the
construction that consists of a copula followed by either an adjective, a prepositional
phrase, or a noun phrase. Examples from this narrative include "is funny" (1), "was
interesting" (38), "was at Wesleyan" (19), or "was still a men's school" (20). In many
such cases one can imagine the concept in question being expressed by a single word in
some other language, where it has become institutionalized.

There are three arguments given by Chafe in favour of this position:


1. "be... is surely not the semantic verb root in such sentences. Its function is to
carry the tense and other inflectional units" ... "what I hope is a noncontroversial case of a single quality being expressed by a word sequence"
2. "we find in English surface phrases like the knife in the box and the wide road
where be is not present and in and be appear to be outside the context of verbs
altogether"
3. "It is significant to observe that there are languages in which such items do
appear in surface structures in the form of verbs, not prepositions or adjectives."
The first point (1) is not an argument, of course, and may not be intended as one. It is
merely an assumption, that be is not a semantic verb (a process, in our terms); as such it
is just an alternative to the assumption that be represents a process.
It is not unusual for philosophers, and even for linguists, to claim of items that are
very general in meaning that they have no meaning at all but are mere structural placeholders, prop-words, or dummies; this has been the fate of words such as it, there, one,
be, and do in English. In relation to be, it is pertinent to observe that be is only the
most general member of an ascriptive set that includes also seem, appear, look, sound,
taste, feel; weigh, measure, number; become, turn, fall, and grow. These can be contrasted
with be as in the following made-up examples, where the contrast is between real (be )
and apparent (seem ) ascription:
He's very nice. - No, he only seems nice.
She seems very nice. - She is

very nice.

Such words may also occur as more specific variants of be; e.g. weighing is a specific
mode of being and weigh is used to indicate T>e + in weight':
It ' s five pounds. / It weighs five pounds.

11.3 The status of qualities: an illustration

459

It does not make sense to exclude the most general member of this set from having
relational meaning simply because its meaning is general. It contrasts with more specific
members, just as e.g. do contrasts with a number of more specific verbs that serve in
material clauses.
Chafe's second point (2) argues from the structure of the nominal group for a particular
interpretation of clause structure. If this argument is to have any force, we must be able to
show that the following relational proportion holds: wide : road [in 'the wide road'] :: wide
: road [in 'the road is wide']. In other words, wide and road must be related functionally in
comparable ways. We do not think they are: rather, in the nominal group they are related
non-contrastively as modifier to modified, but in the clause they are related as participant
1 (Attribute) to participant 2 (Carrier), through a process which embodies a contrast with
other possible relationships.
Furthermore, the fact that an item may be absent in environment x, and is always
absent in environment y, does not mean that it is meaningless if it is present in
environment x. For example, for as a marker of duration may be absent, as in we worked
seven hours, but it does not follow that/<?r does not mark duration when it is present.
Table 11(3): Transitivity models and transformations

Chafe's third argument (3) may again not really be intended as an argument but rather
as an observation about the plausibility of the analysis, based on languages other than
English. In any case, it is not directly applicable to English, unless we do adopt the
assumption that all languages categorize our experience in the same way. But this
assumption is not justified. As we saw in our brief characterization of Chinese (Chapter 7
above), there is a difference between Chinese and English precisely at this point in the
system, in that qualities in Chinese are realized by a class that is more like a type of verb

460

11. Distortion and transformation

than the English adjectival realization. But this diffrence in class assignment is not an
argument in favour of English really being like Chinese in the construal of qualities.
Rather, it is an indication that qualities are ontologically unstable across languages (cf.
Chapter 5, Section 5.3.3).
To summarize the discussion so far, we have identified the following transformational
interpretations in the paradigm of material and relational clauses: see Table 11(3) above.
11.3.3*3 Filling out the paradigm
We have set this out as a paradigm with two associated variables (i) PROCESS TYPE
(relational/ material) and (ii) RANGING (the possible addition of a Range participant,
ranged/ non-ranged). The table makes it appear that material and non-ranged are always
associated, and that relational and ranged are always associated. These are the unmarked
combinations; but ranging and process type are, in fact, independent variables, and in
English the remainder of the paradigm is filled out. Consequently, the transformational
interpretations raise a problem.
Many non-ranged material processes can be construed as ranged, with the nuclear
process as the range participant: do [Process] + a dance [Range] instead of dance
[Process]. The ranged clause is one of a small set of very general Processes: do, make,
take, have, give. So where position 1 in Section 11.3.1.2 above postulated a hypothetical
or virtual form Aristotle ['Subject'] was [copula] dancing ['Predicate'], with a relational
process *be', English has evolved an actual clause type Aristotle [Medium] did [Process]
a dance [Range], with a material process 'do* (cf. the discussion in Chapter 4, Section
4.4.2 and Chapter 6). The two processes *be' and 'do* are comparable in terms of
generality; both are very general. But they are clearly distinct in the same way as material
and relational processes in general are distinct, as is shown by the unmarked tense
selection in each case, for example Aristotle is dancing / Aristotle is doing a dance, by
contrast with Aristotle is wise (see Chapter 2, Section 2.11.3.1 and Chapter 4, Section
4.2.2 on the tense selection).
Similarly, although ascriptive relational processes are typically ranged, with the
quality ascribed to the Carrier as the Attribute/Range, we find ascriptive non-ranged
processes as well. For example, if we start with The problem is
important

[Range/Attribute] (to me),

[Process] (very)

we can relate this to The problem matters

[Processif great deal) (to me). In a sense, the Attribute is incorporated into the Process
'matter' in the same way as the Means is incorporated into a verb such as hammer. This
class of ascriptive processes includes matter, count (as in your past achievements don't
count), suffice w figure, and also an evaluative set, such as stink, suck, drip, reek. But
in English this is the marked way of ascribing a quality to a thing, and these processes are
clearly not material, as is shown by the unmarked tense selection, for example: it doesn't

11.3 The status of qualities : an illustration

461

matter 'it is not important1 rather than it's not mattering, thus parallel to it's not
important rather than it's not being important,
The paradigm is summarized in 11(4) below.
Table 11(4): The paradigm a wider focus

iaterial
relational

| non-ranged
| A. danced
I A, 's ideas mattered

| ranged
j A. did a dance
tA.'s ideas were important

Thus both the approaches discussed in this section are problematic. Let us now look
in more detail at the third position.
11.3.4 The non-transformed relational interpretation
The third approach, which does not involve any transformations of transitivity patterns,
maintains the alignment of ascriptive clauses with relational clauses in general rather than
moving them into the material domain. Furthermore, it unifies the notion of participant,
since Attributes are interpreted as participants rather than as processes. We will explore
these two consequences in turn.
11.3.4.1 The intensive relation: ascription and identification
Let us return to the paradigm of relational processes, starting with three examples. Under
our interpretation they are all brought together as intensive relational processes:
(la) This is very savoury
(Ib) This is a good savoury
(2) This is the best savoury

The best savoury is this

Examples (la and Ib) are ascriptive intensive relations. Tliey represent a relation of
inclusion: the referent is included in the class of savoury things.
Example (2) above is an equative intensive relation; it is a relation of identity, where
'this' is identified as 'the best savoury'. The identification can be reversed: The best
savoury is this. Identity and inclusion are two versions of the intensive relation: identity
is the limiting case of inclusion and inclusion is partial identity. Thus while habits are
complicated acts represents a relation of inclusion 'habits belong to the class of
complicated acts', the following dfinition is an identity; the specification 'complicated
acts' has been qualified so that it describes a one-member class:

462

11. Distortion and transformation

Habits are conplicated acts which, when learned, become automatic


through constant repetition. Habits may be good or bad. Good
habits are useful, for they nake us more efficient by saving time
and energy. (Kraus, Concepts in Modem Biology, p. 193)

Being an identity, it is reversible Habits are complicated acts which, when earned
become automatic through constant repetition : Complicated acts which, when learned,
become automatic through constant repetition are habits. In contrast, Habits are
complicated acts is not reversible. The similarity between identity and ascription is
indicated by the fact that both can be expressed by be, remain or become; the difference
between them by the fact that they belong to different sets: see Figure 11-5.

Fig. 11-5: The markers of intensive relations


Here is an example where an equative is followed by an ascriptive intensive relation;
the first identifies a particle and the second characterizes it:
To complete our understanding of natural radiation, it is
necessary to consider another particle. This particle is. the
neutron, and it 2. particularly interesting because of its central
role in atomic and nuclear bombs. (R.J. Foster, Physical Geology)
Another relational type is (generalized) possession, an important subtype of which is
the part-whole relation. If a quality is represented as if it was an abstract thing, it can be
possessed; people can have beauty, virtue, and grace. The potential paraphrase relationship
between examples like She has great beauty and She is very beautiful points to the
closeness between having and being. The two typically go together in descriptions, of
course. In the first example below the two are in a restatement relation:
Minerals are crystalline substructures; that is, they have an
orderly internal structure (arrangement of atoms) . ...

11.3 The status of qualities: an illustration

463

Protons are heavy and have a positive electrical charge. ...


In addition, an atom may have neutrons in its nucleus. Neutrons
have a mass slightly slightly lighter than that of a proton plus
an electron, and are electrically neutral.
(RJ. Foster, Physical Geology)
To sum up, when ascription is given an interpretation in its own right, it has the
following relational neighbours: see Table 11(5).
Table 11(5): Pan of the relational paradigm
possessive
intensive

middle
This bas a good savour
This is a good savoury
This is very savoury

effective
This is the best savoury
This is the most savoury

We will now turn to the second consequence, the implications for the contrast between
participants and processes.
11.3.4.2

Participants (nouns:

substantives

& adjectives)

vs. processes

(verbs)
If the material model is used to interpret ascripuve processes, like the one represented by
the clause This is very savoury, the notion of process will be widened to include a
realization like savoury

and the notion of participant will be narrowed to exclude it

(Position [ii]). In contrast, with our own interpretation participant is a more general
notion that includes the realization of qualities: see Figure 11-6.
With the unified notion of participant, the class realizations of participants and
processes in the semantics are as follows:
semantics

grammar (class)

participant:

nominal group;
noun * substantive / adjective

process:

verbal group

That is, a participant is realized by a nominal group and a process is realized by a verbal
group. The significance of this generalization is that exceptions to it are always
metaphorical. In particular, it is only when a process is being construed as if it was a
participant that it will be realized in nominal form (see Chapter 6).

464

11. Distortion and transformation

Fig. 11-6: Non-transforming [iii] vs. transforming, material model [ii] notions of
participant
We have seen that where the process is of a material type, there is often an agnate
form, originating as a grammatical metaphor, whereby the happening is construed as the
Range participant configured with a generalized Process such as do, make. Where the
process is material, the metaphor involves no change in process type.
We can contrast this with the situation in a mental clause. Here, in the metaphorical
version, the sensing is construed as the Range/ Attribute of an possessive ascriptive
relational process, with a generalized ascriptive verb such as have. There is thus a shin in
process type from mental to relational. Examples are shown in Table 11(6) below.
Table 11(6): Representation of process congruent & metaphorical

congruent process

[il

material

[]

mental
relational

[verbal group]
HO danced
He hated dogs

metaphorical process as
participant (Range)
[nominal group]
He did a dance
He had a hatred of dogs

As always in the case of grammatical metaphor, we need the metaphorical


representation as well as the congruent one to account for the organization of the
metaphor. For example, the Qualifier in the attributive nominal group of he hada hatred
of dogs corresponds to the Phenomenon of the congruent version: see Figure 11-7.

He

had

hatred

of dogs

11.3 The status of qualities: an illustration

metaphorical

Carrier

Process

Attribute
Thing

Qualifier

Senser

Process

Phenomenon

He

hated

dogs

Deictic
congruent

465

Fig. 11-7: Metaphorical and congruent representations


It seems therefore that there is no real advantage, theoretical or descriptive, in
analysing the grammatical realizations of qualities as if they were something other than
what they seem. On the contrary: not only does this complicate the description, but more
importantly it fails to bring out systematic patterns of agnation. This is a small
indication of something we believe to be a general principle: if you treat language as
distortion, you end up by distorting language.

12. Figures and processes


A characteristic of work on grammatical semantics, where this is based on linguistics or
on natural language philosophy, is to move in at the lower ranks of the grammar rafter
than the higher ones and to start with classes rather than with functions. This is a
continuation of the method of traditional grammar, which (because it originated with the
study of observable features in language, cf. Halliday, 1984a) was wrf-oriented and leant
heavily on word classes in its descriptive statements, as we noted in Chapter 10, Section
10.1. We find this tendency in discussions of word classes and their semantic values
the issue of the proper interpretation of adjectives, the exploration of various verb types,
and so on.

It would take too long to contrast every aspect of our lattice with other proposals,
either existing or conceivable ones, so we shall only highlight a few key areas. In doing
this, we have to keep in mind that we are dealing with a semantic system and not with
a collection of unrelated items. Consequently, it is of little value to argue about an
isolated node in the system; we have to operate with a broader focus. For example, if we
recognize figures of saying (usually overlooked as a distinct type), this goes together
with certain other features: with the distinction between phenomena and metaphenomena,
and between ideas and sayings, with the organization of projections as sequences rather
than as figures, with the identification of symbol sources as a kind of participant, and
with the recognition of circumstances of matter. In other words, we have to consider
syndromes of features that occupy a region of semantic space.
In this chapter we shall be concerned with alternative approaches to a typology of
configurational phenomena of figures, as composites of processes, participants, and
circumstances. Various classifications have been proposed not only in linguistics but also
in philosophy and sociology. What we are looking for are the distinctions that are
systematic in English semantics, rather than incidental distinctions or distinctions that
follow automatically from others.

12.1 Criteria for process typology


In Chapter 5, Section 5.4, we discussed two perspectives on a process: a process is both
an organizer of participants and an event that is instantiate in time. These two
perspectives lead to different criteria for establishing process typologies. For example,
mil have left

is the Process part of they will have left

the house by now, where it

organizes the participants they and the house; but it also has internal organization as an
have
left: see Figure 12-1.
event will
The two perspectives are associated with different grammatical units, the clause (for the
participant organizing perspective) and the verbal group (for the temporal instantiation).
The difference between the two perspectives has been obscured for two reasons. In the first
place, syntacticians have continued to tend to fail to recognize the unity of the verbal

12.1 Criteris for process typology

467

group, favouring instead a Predicate-based constituent, the Verb Phrase, with the
Auxiliary then detached within it or located outside it altogether; furthermore, parts of the
verbal group complex have been exported to a higher node (e.g. seem, continue). In the
second place, the preference for immediate constituency rather than rank-based models of
constituency effectively masks the generalizations that can be made about any one rank.

Fig. 12-1: The two perspectives on a process


(i) In the participant-organizing perspective (see Chapter 5, Section 5.3), it is the
nature of the participants involved in the process that determines the different process
types. Relevant variables include the following:
1. Is some participant created, brought into existence, by the process?
2. Is some participant restricted to conscious being?
3. Can some participant be a metathing as well as a thing?
4. Is the process directed towards some participant?
5. Does the process benefit some participant?
6. Does the process occur spontaneously or does it need an input of energy?
7. Does the process affect some participant materially or does it impinge on its
consciousness?
8. Is the process symmetric?
9. Is the process reflexive?
Questions such as these lead to typologies with terms like action, transaction,
happening, sensing, saying, relation, existential, ambient, and so on. In the typology
presented here as pan of the ideation base, there are two very general considerations:

468

12. Figures and processes

1. process type: what kind of reality does the figure or process configuration
pertain to (the material world, the world of consciousness, the world of
symbolization, the world of abstract relations)?
2. agency: is the occurrence of the process (in conjunction with the medium)
caused by an entity that is external to it (an agent)?
(ii) In the temporal instantiation perspective, the issue is the occurrence of the process
as an event located in time (cf. Ch. 5, Section 5.4.1): how does it unfold through time,
and what is its temporal profile? Here the relevant variables are such as the following:
1. Is the process homogeneous during its occurrence or does it decompose into a
sequence of distinct phases (stages); is it a mini-tableau or a mini-drama?
2. Is the period of occurrence a relatively short interval, or is it an extended
interval?
3. Is the whole period of occurrence in view or only a phase of it (e.g. beginning
or end)?
4. Does the process tend to occur once or repeatedly?
These questions lead to typologies that include terms like stative, dynamic, perfective,
imperfective, punctiliar, iterative, durative, and so on.
The distinction most commonly drawn here is based on change. Is there change over
time or not; i.e. is there a change in the course of the occurrence of the process? The most
common dichotomy is state vs. non-state (with terminological oppositions such as
stative/ dynamic); this has been favoured both by philosophers (see e.g. Nordenfelt, 1977,
and his references) and by linguists (see e.g. Quirk et al, 1985 and their references). States
and non-states have different temporal profiles. States are homogeneous; any time we
check a process whose occurrence is a state, it will be the same. Non-states, or changes,
are not homogeneous; during the course of the occurrence of a process something will
have changed, for example the spatial location of a participant (as with processes of
movement) or parts of a participant, or some other attribute of a participant (e.g.
possession or location in a 'quality space' such as colour or temperature).
As we have seen, the two perspectives represent two different kinds of profile. One is
the configuration of process, participants, and circumstances the transitivity profile;
and the other is the occurrence or unfolding of an event through time the temporal
profile.
The two perspectives are not unrelated, of course. As we shall see, the specification of
an element of transitivity structure may determine the temporal profile: the created Goal
that constitutes the completion of the performance of the process (as in Mr. Standings
built a house), a Range that constitutes its finite scope (They sang two Hungarian folk

12.1 Criteris for process typology

469

songs), a destination that gives its spatio-temporal endpoint (He walked to he store), a
rsolutive Attribute that constitutes the (qualitative) endstate (He was shot dead), and so
on. But the mere presence of such an element is not sufficient to determine the temporal
profile; it is also influenced by the Ixmndedness' of the elements: are they in infinite
supply or not a definite number of units, or an indefinite number? In general, then, the
temporal profile is determined by other factors such as the presence, and the boundedness
in quantity, of participants and circumstances.
We shall start by looking at one particular temporal typology of processes that a
number of researchers have based their work on, Vendler's (1967) activity /
accomplishment / achievement / state typology; this involves consideration of both
perspectives, but we shall intersect it with our process types in order to compare the two.
Then, we explore a typology based on control (intentionality, volitionality, etc.), and
look at proposals using this distinction together with the distinction between states and
non-states (events: activities, accomplishment, achievements).

12.2
Temporal
profile:
achievements, and states

activities,

accomplishments,

There have been numerous treatments of temporal categories such as aspect, "lexical
aspect" (Aktionsart) and phase, and various typologies have been proposed, Vendler's
(1967) classic philosophical discussion does not represent the most recent work, but it has
been used in a number of later studies (cf. Verkuyl, 1972, Dowty, 1979, Nordenfelt,
1977, Platzack, 1979, Foley & Van Valin 1984, Dahlgren, 1988) and will serve as a
starting point for a comparison, together with other proposals based on Vendler's ideas.1
Vendler focuses "primarily upon the time schemata presupposed by various verbs", while
recognizing that other factors such as "the presence or absence of an object ... also enter
the picture".
Vendler recognizes two groups of verbs based on whether they "possess continuous
tenses" or not2
(1) Group 1.

Verbs from the first group can be used to answer the question

"What are you doing?". They fall into two "species", activities and
accomplishments. Activities are homogeneous through time (drawing,
running, eating, etc.), whereas accomplishments work up towards a climax

For a systemic functional treatment of "Aktionsarten", see Steiner (1991: Section 3.10).

Vendler's characterization in tenns of 'continuous tense1 in fact breaks down: his


achievements (see below) may also go with secondary present ('continuous present').

12. Figures and processes

470

(drawing a circle, running a mile, eating a steak, etc.). Platzack (1979) calls them
unbounded processes (activities) and bounded processes (accomplishments), in
order to bring out the character of the temporal profile.
(2) Group 2. Verbs in the second group can be used to answer the question "Do
you [verb: know/love ...]". Vendter distinguishes two subtypes, achievements
and states, which are called punctual events and states in Platzack's (1979) terms.
Achievements include a short (or instantaneous) transition to an endstate, e.g.
reaching a peak, spotting, and realizing. States involve no change: thinking
that, knowing, having, etc..
The four types and their temporal characteristics as discussed by Vendler are given in
Table 12(1).
Table 12(1): Vendler's four event types
Type
events

Admit
"continuous"
tense

Temporal
profile

activities
(unbounded)

yes

accomplishments
(bounded)

yes

achievements
(punctual)

no

running,
pushing a
cart
climax
running a
mile,
drawing a
circle
occur at single reaching the
top, winning
moment
a race
last for a period having,
believing
of time

states

no

Example

homogeneous

Platzack (op cit.) uses three dimensions to distinguish Vendler's four event types, viz.
change vs. no change, bounded vs. unbounded (in time), and duration vs. no duration. If
we try to tabulate these as independent dimensions, we find that there are certain
dependencies. 'No change* (state) implies duration and unbounded, for example. Using
graphic representations from Schneider (1977: 33) cited in Platzack, we can summarize the
parametric analysis of the four event types as in Table 12(2).
The dependencies that emerge from the table suggest the systemic arrangement of
Platzack's three parameters shown in Figure 12-2 as one possible systemic interpretation.
Dahlgren (1988: 85) arrives at a similar arrangement, represented in the form of a decision
tree (her Figure 10).

12.2 Temporal profile: activities, accomplishments, achievements, and states

471

Table 12(2): Parametric and graphic analysis of Vendler's event types

Fig. 12-2: Systemic organization of the temporal parameters


Having presented Vendler's temporally based taxonomy of events, we can now
intersect it with the typology of figures we introduced earlier, viz. doing (& happening),
sensing, saying, and being & having. Using only examples given by Vendler, we find the
correspondences as in Table 12(3).
According to this table, activities are always processes in figures of doing.
Processes of motion, either middle or effective, form an important subclass (walk, run,
push, etc.), unless there is a single destination (fly to Nairobi, walk home, etc.). We also
find a behavioural process like think about (Vendler's example) and we might aid other
behaviourals such as ponder, listen to, watch, look at, and smile; but behavioural
processes with a delimited range would not presumably be classified as activities, because
they are bounded: watch the film Brief Encounter.

472

12. Figures and processes

Table 12(3): Vendler's event types and the typology of figures

doing
activities

accomplishments
achievements
states

run, walk,
swim, push/
pull sth,
think about
paint a
picture, make
a chair, grow
up
reach a peak,
win a race,
die, start
[habits, e.g.: he
smokes]

sensing

saying

recognize,
realize,
spot
desire,
want, like,
dislike,
love , hate ,
know,
believe,
think that

being & having

have, possess,
dominate, rule

Accomplishments are similarly always processes in figures of doing. Resultative


or creative actions seem to be an important subclass of these: the accomplishment
(temporal bounding) lies in the creation of the Goal (make a chair, build a gazebo, etc.),
but we also find doings of a middle type such as grow up where there is a definite end
state.
The difference between an activity and an accomplishment may in some cases depend
on the presence of a Range. Thus non-ranged climbing is (presumably) an activity,
wbereas ranged climbing like climbing a mountain or a hill is an accomplishment. In the
first case the process may go on indefinitely in time, but in the second case it is bounded
by the Range. The following pairs are similar: sing vs. sing a song, smoke vs. smoke a
cigar, play vs. play a game of tennis, and run vs. run a mile

(Vendler's example).

However, what is important is clearly not just the presence of a Range, but rather the
nature of the Range: is it a unit or a set of units? For example, contrast play a game of
tennis (an accomplishment: a particular quantity of playing) with play tennis (an
activity: a particular kind of playing activity).
The same is also true of a Goal created by the process: if it is a single unit, we have
an accomplishment, as in Henry is making a chair; but if we have an indefinite number
of units, we have an activity, as in Henry makes chairs for a living. If the Goal is a mass,
the interpretation seems to depend on whether the mass is conventionally produced in
finite units or not; contrast make tea (an accomplishment) with make gold (a neverending activity?).
Again, the presence or absence of a resultative Attribute may determine whether an
effective material process is an activity or an accomplishment. Thus, beating somebody

12.2 Temporal profile: activities, accomplishments, achievements, and states

473

may be an activity (as in He kept beating him), but beating somebody dead is an
accomplishment (He beat him dead).
In a similar way, it seems that the presence or absence of a circumstantial element
may determine whether the process in a figure of doing is interpreted as an activity or an
accomplishment For example, contrast walk (an activity) with walk to the market (an
accomplishment; the spatial destination correlates with a temporal bound).
Further, if the process is expanded the interpretation may change from activity to
achievement. Thus while running is an activity, stop running

is presumably an

achievement just as stopping is.


Vendler's achievements are transitional: accomplishments occur over a stretch of
time, while achievements are more instantaneous. According to the examples in the table,
the transition can be either one of doing (die, win a race ) or one of sensing (recognize,
realize, spot ). In the doing type, the achievement may lie in the Range specified for the
Process. Thus die is an achievement; the nature of the achievement may be specified, as
in die the death of a hero.
To judge from Vendler's examples, states are either processes in figures of sensing
(desire, like, hate, know, believe ) or processes in figures of being & having (have,
possess ). However, Do you smoke? is a question about a state, according to Vendler
(contrasting with the activity in Are you smoking?)', it would appear, therefore, that
habits of doing aie states. If we expand the process in a figure of sensing or being &
having, we may no longer have a state. For example, she has stopped loving him

is

presumably an achievement. And what about: I'm beginning to believe that he knew all
along; I'm understanding the problem better and better; I'm liking this album more and
more (said by a DJ playing it repeatedly)?
When we take the four types of figure as our starting point, we see that processes in
figures of doing can be activities, accomplishments, or achievements, and, in a habitual
interpretation (and perhaps also in so-called eternal truths), states. Processes in figures of
sensing can be achievements or, predominantly, states. (There are sensing activities
(ponder, listen, watch, smell, etc.), but we interpret these as behavioural rather than as
inert sensing.)
There are no processes in figures of saying exemplified in the table of comparison
above. When we intersect such processes with the various event types, we find examples
of all classes: saying states, saying achievements, saying accomplishments, and saying
activities. Examples like the sign says that visitors have to take off their shoes and the
law states unambiguously that this is a felony would appear to be states. Sayings that
can be interpreted as causative sensings such as remind sb that 'cause to remember' and

474

12. Figures and processes

convince sb that 'cause to believe* seem to be achievements. Next, ranged figures of


saying where the Range specifies a genre or speech-functional 'unit* of verbalization are
accomplishments: tell (sb) a story, ask a question, etc.. Finally, there are saying
activities: speak and talk.
Processes in figures of being & having, finally, are states, according to the
examples in the table. However, we have to add inchoatives become, grow, turn, etc.
and they seem to be accomplishments, become an adult, grow dark, turn sour, etc.
(Range: result of change); or activities, become larger, grow heavier, etc. (Range:
unbounded). (Incidentally, notice the difference between doing widen, grow, expand, etc.,
which presumably are activities, since there is no end state, and being & having become
wide, big, large, etc., which would be accomplishments.)
The result of our brief exploration of the intersection of Vendter's typology of
processes with our typology of figures is summarized in Table 12(4).
We can draw some conclusions from our comparison of Vendler's event taxonomy and
our typology of figures. First, Vendler presents his distinctions as pertaining to verbs;
but, as we have seen, the configuration and nature of participants and circumstances are
just as important as the verb realizing the nuclear Process of a figure (cf. Ehrich, 1987),
particularly in deciding whether a process is an activity, an accomplishment, or an
achievement
Second, the correlations are not particularly systematic. For example: sing (activity),
sing a ditty (accomplishment), sing folk songs (activity), sing a ditty every evening
(habitual accomplishment => state). In some sense, the particular type of event is an
automatic consequence of the particular configuration of the process and the nature
of its participants and circumstances.
The distinction between state and non-state (event) seems more stable in relation to
our process types than the subtypes of event (activities, accomplishments, and
achievements). We can state the relations as implications: see Figure 12-3.
These correlations are perhaps not surprising. Given that there are two aspects to a
process configurational (the composition of a figure into process, participants, and
circumstances) and temporal (its profile in time) processes without participants
(meteorological ones) have to involve change in order to be construed as processes at all
(consequently they are always events); likewise one-participant processes, with
existentials as the limiting case. Conversely, in order for a state to be construed as a
process, it has to rate high as a participant organizer; it has to be interpreted as a relation
between two participants.

12.2 Temporal profile: activities, accomplishments, achievements, and states

475

Table 12(4): Elaborated table of correspondences

activities

accomplishments
achievements
states

doing

sensing

saying

being & having

run, walk,
swim, push/
pull sth,
ponder
paint a
picture, make
a chair, grow
up
reach a peak,
win a race,
die, start
[habits, e.g.: he
smokes]

ponder

speak,
talk,
chat

be [naughty/
difficult]

recognize,
realize,
spot
desire,
want, like,
dislike,
love, hate,
know,
believe,
think that

become, grow,
tell a
go
story,
ask a
question
convince, turn, fall
remind

say, ask, have, possess,


tell
dominate, rule

Fig. 12-3: Figure and stativity implicalional relations


As noted earlier, Vendler's typology has fonned the basis of various other frameworks.
We will return in Section 12.4 below to Dowty's (1979) development of Vendier's
typology, and the use Foley & Van Valin (1984) make of Dowty's work in their typology
events.
Our general conclusion is that as a systemic specification of the potential of processes
of the options that are open in their construal Vendler's typology is essentially
epiphenomenal. The temporal profile of a process is an interpretation that results from
various factors that are systemic, such as the configurational profile of the figure in

12. Figures and processes

476

which the process serves, the nature and quantity of participants involved in it, and the
phase of the process itself; but it does not define an area of semantic potential.

12.3 Control (and stativity)


We have been considering a typology of processes based on temporal distinctions. Next,
we turn to typologies that incorporate a notion of control. These aie concerned with the
nature of participation and process in the figure as a whole rather than the Process
element alone. But such typologies often embody both control and stativity: the general
question is whether one of the participants involved in a process controls the occurrence of
it or not. The control parameter has been used to distinguish pairs of participant roles
such as the following:
control
agent
actor
agent

no control
patient
undergoer
force

There are different possible kinds of control, and scholars vary in the way they use the
notion: see Figure 12-4.

Fig. 12-4: Types of control

We can see the difference among these notions of control in the classification of
'natural force* agents: Chafe (1970) argues that they are potent entities and hence agents
(in his sense), while Dik (1978) treats them as forces rather than agents, and Nordenfelt
(1977) also takes them to be non-agentive (specifically, non-intentional): see Figure 12-5.

477

12.3 Control (and stativity)

Fig. 12-5: Breaking and control


This difference in the treatment of natural forces suggests that control by will is a stronger
and more restrictive notion of control than control by power. We will consider these three
notions of control, each in turn:
1. Intentionality: Nordenfelt's (1977) notion of control, which is defined by
intentionality.
2. Potency: Chafe's (1970) notion of control, which is restricted to nonstates,
where an action is carried out by a potent doer.
3. Power to determine: Dik's (1978) more generalized notion of control, which
is applicable to states as well as actions ("a state of affairs is controlled if one of
the entities involved in it, the controller, has the power to determine whether or
not the state will obtain.").
12.3.1

Intentionality

Nordenfelt (1977) proposes a philosophical account of episodes drawing on G. von Wright


and D. Dowty. It is "concerned with episodes in the world and not with words or any other
linguistic elements" (p. 41). However, the "limits and contents of the episodes will be
determined by ordinary English sentences. Thus, John runs signifies an episode; so do the
wind breaks the window and Peter kills John by shooting him with a gun". It is not
clear that these two positions are compatible: it would seem that they build an
ambivalence into the foundation of the account, between accounting for (physical) action
etc. (a philosophical construal of action) and accounting for the constmal of action etc. in
English (a philosophical construal of the construal of action in English). But this kind of
ambivalence seems to us to characterize a good deal of philosophically oriented work on
categories such as action and intention.

478

12. Figures and processes

Nordenfelt distinguishes states, processes, and events (the coming about either of a
state or of a process); and he intersects these with 'causative' vs. 'non-causative' and
'agentive' vs. 'non-agentive1. Agency is really intentionality; it is independent of causality.
For example, A holds the book is 'causative' and 'agentive', whereas The pillar supports
the bridge is 'causative' and 'non-agentive'. Nordenfelt characterizes agency as follows (p.
54):
... p shall] adopt the strong notion of action which is completely tied to the notion of
intention. According to this line of thought an action is an episode, which is either
directly intended by a human agent or believed by the agent to be a necessary means for
the realisation of an intention of his.

To the extent that intentionality enters into the semantic system of a language, it could in
principle be construed ideationally as some aspect of the overall system of figures or be
enacted interpersonally as a modality of inclination. If it is construed ideationally, it could
in principle be a global property of a figure as a whole or a local property of the process
or a participant
In English, intentionality can be enacted within the interpersonal system of imperative
modality (modulation) in the form of inclination, as in he won't be ordered around
('refuses to be'). It is independent of ideational agency; inclination is oriented towards the
Subject, i.e. the modally responsible element of the clause, rather than (say) the Actor or
Agent (as our passive example illustrates).
Intentionality can be construed in the clause as circumstances of Manner ('according to
plan', e.g. he found the book by chance; he turned left by mistake; he turned left
intentionally) and as verbal group complexes of projection or enhancement realizing the
Process (e.g., she intended to leave at 4; she happened to be in the neighbourhood ). It
may also be a factor in certain lexical contrasts within delicate process types; but we have
no evidence that intentionality is a primary variable in the semantic system of figures and
its grammatical constmal in transitivity. Rather, whether an example is read as construing
something as intentional or non-intentional will depend on a variety of factors, such as
the consciousness of the most active participant. Thus all Nordenfelt's examples
contrasting 'agentive* and *non-agentive' in his summary (p. 57) involve A (presumably a
person) contrasting with non-persons (either inanimate objects and forces or lower
animates), e.g. A keeps water running : The pump keeps the water running :: A shuts the
door : The wind breaks the window :: A starts the engine : The germs initiate the
deterioration of the organism :: A lets the door close : The window breaks.
12.3.2 Potency (Chafe)

Chafe (1970) draws a basic distinction between states (e.g. The wood is dry ) and nonstates. Non-states involve either one participant undergoing the process (a "process"
component involving a patient, e.g. The wood dried; Henry died ), or one participant in

12.3 Control (and stativity)

479

control of the process as a doer (an action component done by an agent, e.g. The boy
sang). The main probe Chafe uses to differentiate processes from actions is "What did N
do?" (action) vs, "What happened to N?" (process):
What happened to the wood? It dried.
What did the boy do? He sang.
In examples like these, "process" (in Chafe's sense) and "action" can be interpreted as
subtypes of middle figures of doing, and "Patient" and "Agent" as further differentiation of
our Actor: see Figure 12-6.

Fig. 12-6: Two types of Actor according to potency


Both "What happened to the butter?" and "What did Henry do?" can be answered by Henry
melted the butter: a process can be earned out by an Agent; and it can combine with an
action to give an "action process": Henry (Agent) melted the butter (Patient). In Chafe's
analysis, there is an oscillation between an ergative interpretation and a transitive
interpretation, with "Agent" used in both middle and effective as in a transitive model, but
with a grouping of Process and Patient in the effective as in an ergative model: see Figure
12-7. In contrast, our interpretation (based on Halliday, 1967/8; 1985) accommodates both
the transitive (Actor + Process [-fGoal]) and the ergative model (Process + Medium
[+Agent]) as complementary perspectives: see Figure 12-8.
Chafe's notion of agent comes closer to our notion of actor (a transitive function) than
to our notion of agent (an ergative function); but it is more restricted than our actor, since
our actor may also correspond to Chafe's patient. The agent of an action is "a thing which
has the power to do something, a thing which has a force of its own, which is self-

480

12. Figures and processes

motivated"; agents are "largely animate beings" but also include e.g. heat, wind, ship.3
Chafe characterizes these things as 'potent*.

Fig, 12-7: Oscillation between ergative and transitive in Chafe's model

Fig. 12-8: Representing both transitive and ergative models

Contrast Dik's (1978) treatment (discussed below) where The wind opened the door is
interpreted as an uncontrolled process rather than a controlled action. For Dik, control is not
only power but power to determine.

12.3 Control (and stativity)

481

It would seem that Chafe's paradigm is incomplete; we can find action-actions and
action-states in addition to action-processes. Compare:
process
action
state

non-action
The butter melted
The soldiers marched/ ate
We were warm

action
The sun melted the butter
The officer marched/ fed the soldiers
The sun kept us warm

Chafe links his notion of control to a typology of things ("semantic nouns1'); agents
are things which are potent and typically animate:
figure type:

participant role:

thing filler:

action

agent

potent
(animate / inanimate:
heat, wind, ship,...)

(Note that this model treats potency as a property that is inherent in a certain class of
things: inherent ability to act.) This correlation between figure type and thing type
contrasts with our correlation between (i) figure & participant type (the senser of a
sensing) and (ii) consciousness (inherent ability to sense):
figure type:

participant role:

sensing Senser

conscious being

thing filler:

contrastsaying

Sayer

symbol source

The question arises: should we add a category of potent (ability to act) to our general
taxonomy and integrate it with the distinction based on consciousness (ability to sense)?
The notion of consciousness is very clearly reflected in the grammar in addition to the
value restrictions just mentioned, we find it reflected in pronominal choices (he/she vs. it;
who vs. which/that; etc.). Further, metaphorical Sensers are quite clearly discernible: My
car dislikes inexperienced drivers. But is potency equally clearly defined? Let us consider
first whether or not all agents are potent, starting with figures of doing.
Are all Actors potent? Obviously, the answer depends on how we define potency. If
the figure is middle, it can be interpreted fairly narrowly, since the agent of an action is in
opposition to the patient of a process (in Chafe's sense): y in (y process) can be either
agent or patient However, if the figure is effective, Chafe's paradigm only allows for one
alternative: x in x (y process) can only be agent. Since an agent is supposed to be potent,
it seems we have to allow for a fairly wide interpretation of the notion of potency.

482

12. Figures and processes

Although, according to Chafe, most agents are animate as well as potent, he notes (p.
109) that "there seem to be some nouns ... which aie not animate but which may
nevertheless occur as agents"; his examples are:
The heat malted the butter
The wind opened the door
The ship destroyed the pier

(Note the emergence of potency in the following sequence: John (Actor / Patient) fell and
[he] (Actor / Agent) hit the rock (Patient), where John's falling impacts on the rock.)
However, in addition to persons, animals, natural forces, and self-propelled objects like
ships, we find tools, abstractions & conditions, and events as agents. It is important to
draw examples from fairly abstract fields to counter-balance the concrete examples given
by Chafe (breaking dishes, drying wood, melting butter, opening doors, destroying piers,
etc.). Here are some examples, beginning with three that are constructed as illustration:
Lack of sleep
The key

killed him in the end

opened the door

Circumstances

scattered the family

those proposals which

do not affect the basic structure of the

Constitution and those which do (Quadrant, Sept 86)


Quantitative proposals

set fixed limits on fiscal aggregates, (op

cit.)

Politically committed art took

over one wing of the modernist

movement. (David Harvey, The condition of postmodern!ty, p. 33)


The myth either had to redeem us from 'the formless universe of
contingency1 or, more programmatically, to provide the impetus for
a new project for human endeavour, (opcit.,p.31)
When

the

sense of

progress

10

checked

by

depression

or

recession, by war or social disruption, . . . (opcit,p.202)


Ohe

new

systems

of

transportation

tightened the skein of internationalism

and

corntuni cation

. . . (opcit, 278)

Through these mechanisms capitalism create* its


historical geography, (opcit,p.343)

own distinctive

483

123 Control (and stativity)

AU these examples have Agents of doing (effective: Actor/Agent). Some of them


extend the notion of Agent from concrete doing to abstract doing, but they would
presumably still be action-processes in Chafe's model. Our own notion of Agent is not
confined to doings; it extends to the other process types as well (see Chapter 3, Section
3.3). There are Agents of sensing (phenomena impinging on a Senser's consciousness)
and Agents of being & having, as well as Agents of doing. When we move from figures
of doing to the other types (which Chafe does not do, of course, since his Agent is
restricted to actions), there are many cases where the notion of potency seems very had
to apply:
Brilliance in other people scares me (Agent of sensing = Phenomenon)
The fact that you lied didn ' t surprise me at all (Agent of sensing =
Phenomenon)
The solution to our problem is this: (Agent of being & having = Token)
Anne ' s my sis ter-in-law (Agent of being & having = Token)
The schedule kept us busy (Agent of being & having = Attributor)
We can either opt for a very wide interpretation of "potent" or allow for the possibility
of non-potent Agents; the latter would extend the "action" paradigm even further. Table
12(5) summarizes possible Agents, clearly potent or possibly non-potent, in the
environment of Chafe's different process types.
Table 12(5): Extending the paradigm to non-potent agents
first participant

outer participant =

state

The wood is dry


[Patient]
Henry died
The wood dried
[Patieot]
Henry sang
The soldiers
inarched
[Agent: potent]

[Agent: potent]
The sun made/ kept
the vrood dry
They killed Henry
They dried the
rood
(They made him
sing)
He inarched the
soldiers

non- process
state
action

[Agent: ? potent]
The lack of rain
kept the wood dry
The lack of
medicine/ Greed
killed Henry

Chafe's typology has since been adopted by Cook (e.g. 1977), who retains all of
Chafe's distinctions. Cook tabulates those we have already discussed (state vs. nonstate:
process/ action/ action-process) against another set of his own (basic vs. nonbasic:
experiential/ benefactive/ locative). We can represent this systemically as in Figure 12-9.

12. Figures and processes

484

Fig. 12-9: Systemic representation of the Chafe/Cook typology


The option 'nonbasic' means the addition of a further participant, a Benefactive,
Experiencer, or Locative, as shown in Table 12(6) together with the basic types.
Table 12(6): The Chafe/Cook typology and process types (material unless otherwise
marked)

process
action

action-process

experiential

benefactive

OsProc
broken, dry, dead,
tight
[relational]
Obj Proc
break, die, dry,
tighten
AgProc
dance, laugh, play,
sing

Ben Obj Proc


have (got), own

Obj Loc Proc


(be) in, be (on)

Exp Obj Proc


know, like, want

[relational]
Ben Obj Proc
find, lose, win

[relational]
Obj Loc Proc
come, go, move

Ag Ben Proc
arm, bribe, help,
supply

Ag Loc Proc
come, go, run,
walk

Ag Ben Obj Proc


buy, give, sell,
accept

Ag Obj Loc Proc


bring, place, put,
take

[mental]
Exp Obj Proc
feel, hear, see
[mental]
Ag Exp Proc
frighten, please
[mental]
ask, question
[verbal]
Ag Exp Obj Proc
ask, say, tell,
speak
[verbal]

+ Ben

state

locative

basic

Ag Obj Proc
break, dry, kill,
tighten

+ Loc

+ Exp

485

12.3 Control (and stativity)

12.3.3 Power to determine (Dik)


Like Chafe (1970), Dik (1978) uses the parameters of control and stativity (dynamicity) to
make the most general distinctions in the typology of figures (or states of affairs, to use
his term). One difference is that Dik intersects the two parameters so that there are
controlled states in his typology rather than only controlled processes ("action processes"
in Chafe's terms); in all, there are four basic types of state of affairs: actions, positions,
processes, and states. They are tabulated in Table 12(7) with a simple example of each
type.
Table 12(7): Dik's basic types of state of affairs

dynamic

static

controlled

uncontrolled

action

process

John ran away

John fell down

position

state

John stayed motionless

Roses are red

Each type of state of affairs, action, process, etc., can have one or more arguments. For
example, John ran away, John read a book, John gave Peter a book, John sent a book to
London, and John took a book from the shelf are all actions.
Dik (op cit.: 35) argues that the two parameters "enable us to define a number of
selection restrictions pertaining to nuclear predications taken as a whole in a simple way".
The examples he gives are (i) orders and requests, and promises, (ii) Manner, Beneficiary
(Behalf in our terms), and Instrument (a kind of Means in our terms). These fall into two
groups; the first is concerned with MOOD and the second with CIRCUMSTANCES.
According to Dik, their distributions correlate with the distinction between controlled and
not controlled as shown in Table 12(8).
We comment on each of these two groups in turn.
(i) Mood is an interpersonal system rather than an experiential one. What is at issue
in the direct speech imperatives is whether a participant can function as Subject or not;
according to Dik's claim, it can only be the Subject if the state of affairs is controlled.
However, we would suggest that there are two different notions here. The first is the
experiential notion of a participant's control of (the execution of) a process, while the
second is the interpersonal notion of modal responsibility (see HaUiday, 1985: Ch. 4).

486

12. Figures and processes

Thus, in an imperative clause the Subject is the participant that is made responsible for
the command. In examples such as
Fall asleep, will you.
Don't be stupid; be intelligent instead!

(neither of which is at all problematic) one participant is given modal responsibility, even
though the participant assigned this responsibility may not normally be thought of as
being in control of the process from an experiential point of view. Such examples are
possible precisely because modal responsibility and "control" are metafunctionally distinct
variables; there is merely an unmarked association between them. Thus the main
constraint is an interpersonal one: the modally responsible participant, the Subject, of an
imperative clause has to be an interactant the speaker, the addressee or the combination
of the two: Be quiet; go to sleep! Be patient! Be guided by your parents! Don't be scared!
Don't get mad, get even! are all clauses with an implicit addressee Subject, assigned the
responsibility of making the command a success, regardless of the participant role.
Table 12(8): States of affairs and "selection restrictions"; shaded areas ruled out by
Dik's typology

MOOD
orders & requests
+ control

promises

John, come
here!
Bill ordered
John to be
polite

- control

CIRCUMSTANT1ATON

John, t*U
a3lap!
MU 0rt*r*d
John to ba
ifctalligent

John promised
Bill to be
polite

behalf

instrument

John cur down


the tree for
my sake

John cut down


the tree with
an axe

John remained
in the hotel
for my sake

John kept
himself in
balance with a
counter -weight

The t*:* *U
dovm for sty
****

The tree ieU


<S0wn with n
axe

John proRdsod Tftfi tr#6 *aa


Bill to be
r*i for w
sake
intelligent

ohfl knew the


answer with
Ida
intJlliffafje

(ii) The second set of considerations have to do with circumstances. It is important


to note that Dik has chosen to focus on circumstances rather than on participants: given
that participants are more nuclear to (more directly involved in) a process than
circumstances, one would expect them to rank higher in their influence on the typology of
figures than the less central circumstances. Since we have taken participant organization
as the starting point for our figure typology, our typology and Dik's noon of control

12.3 Control (and stativity)

487

may complement each other: see Figure 12-10. Moreover, if we take the more nuclear part
of a figure as the primary basis for a typology, it is clear which elements have to be
considered the Process and the participants involved in it. If we take circumstances as a
primary basis, we have to decide which circumstances should be considered. Why the ones
selected by Dik and not, for example, Matter? If we select Matter as a central criterion, the
parameter of projection would be more important than that of control since it seems only
figures that can project (essentially sensing and saying) can have a circumstance of
Matter4

Fig. 12-10: Parameters and nuclearity


Agent is related to Means (see Halliday, 1985: 139) and it seems reasonable that the
presence of the Means depends on the control exercised by the Agent. But what about
Behalf: does it really depend on the control of an explicit Agent controlling the process?
Consider for example:
Ihe door was ajar for the sake of the cat.
The air-conditioning was off for our sake.

Dik characterizes the distribution of various types of Manner (-quality) in terms of


dynamicity as well as control. He suggests that states (-control, -dynamic) are not
compatible with Manner (-quality), while other actions, positions, and processes are
compatible with various subtypes. But consider the following example of the state of

This is borne out by text counts. For example, in a roughly 90,000 woid sample from Time
magazine, circumstances of Matter marked by about occur with 'sensing' and 'saying' and not
with doing and being except where being involves a mental Attribute.

488

12. Figures and processes

existence, taken from a discussion about the existence of acts (our holding of the
expression of manner):5
Short arguments for their existence ['the existence of acts', MAKH & CM]
are easy enough to coin by. Here is one I rather like.
(I) Sirhan killed Kennedy
is true; and surely it is equivalent to
( I I ) There was a killing of Kennedy by Sirhan,
vrtiich is therefore also true. But if

there was at any time an

entity that was a killing, there was an act at that time. So there
(tnl*ly) are acts. (Thomson, 1977: 13)

Based on Due's examples only, it seems that dynamic states of affairs correspond either
to figures of doing or to those of effective sensing, while the static ones correspond either
to being & having or to middle sensing. The distinction between controlled and not
controlled states of affairs does not correspond to any distinction of comparable generality.
Thus both John ran away (action) and John fell down (process) are figures of doing.
Similarly, both John remained in the hotel (position) and The substance is red (state)
are figures of being & having.
On the other hand, if we take our own types of figure as point of departure, in this
survey of his typology, certain other discrepancies appear. First, Dik does not give any
examples of figures of saying. Secondly, doings are all dynamic, and being &
having figures are all static. Thirdly, as far as sensing is concerned, there are a few
examples of middle sensing figures, which are states; and diere is one example of an
effective sensing, The dog frightened John, which is interpreted as a process (in Dik1 s
sense). Sensings do not form a homogeneous class in Dik's typology; nor does he accept
Experiencer (cf. Senser) and Experienced (cf. Phenomenon) as functions (1978:41-43):
... experiences are conceptualized within the model of non-experiences, and different
analogies are operative both within and across languages. There is no single underlying
semantic representation of experiences, and no special functions need to be assigned to
the Experiencer and the Experienced. That some experience is involved will result from
the interpretation of the meaning of the predicate, and of the properties of the terms
related by the predicate.6

Notice how the argument hinges on the metaphorical reconstrual of (1) as (!')!

' Incidentally, we may note that this last observation could be applied to the distinction
between controlled and uncontrolled (e.g. action vs. process).

12.3 Control (and stativity)

489

The diffrait expressions of experiences cited by Dik are


Dik's type
Example
state
John was afraid of the dog
John had a great fear of the dog state
process
The dog frightened John

Our type
being & having: intensive
being & having: possessive
sensing

The first two examples are figures of being & having, in our terms; further, they are
metaphorical versions of John feared the dog (cf. Chapter 6, Sections 6.8 and Chapter 11,
Section 11.3 above). Since they are metaphorical, they do not constitute arguments
against recognizing a congruent representation of a class of mental processes any more
than an example like Henry began his walk is an argument against recognizing Henry
[Actor] began to walk [Process) as a figure of doing. Grammatical metaphors do
reorganize experience; that is one reason why they are identified as metaphors, rather than
simply being incorporated into the congruent options. Since they are sensings of the
affective type, the first two examples have in common with their congruent equivalents
the possibility of a fact as Phenomenon: John feared that the others had lost their way,
John was afraid that the others had lost their way ["because of the (possible) fact that'],
John had a great fear the others had lost their way. (All in the senses of 'was distressed by
the (possible) fact that'.)
In the third example, The dog frightened John, the experience is represented "as if
some Force works on the experiencer". In our view, this is not an argument against the
interpretation of the example as a figure of sensing; rather, it simply reflects the ergative
interpretation of it with the dog as Agent, comparable to Dik's Force in this particular
case, and John as Medium. In other words, in our model both interpretations of the
example are captured simultaneously: The dog [Agent/Phenomenon] frightened [Process]
John [Medium/Senser]. The ergative interpretation brings out its agnate relation to
examples like The wind [Agent/Actor] opened [Process]//*? door [Medium/Goal]; the
transitive interpretation preserves its relation to John [Medium/Senser] feared
[Process]/^ dog [Range/Phenomenon], as does the identity of John as Medium in both
John feared the dog and The dog frightened John. In other words, a figure of sensing,
configured as Senser + Process + Phenomenon, can be represented either as the
Phenomenon impinging on the Sensor's mental processing (effective: Phenomenon as
Agent) or as the Senser's mental processing ranging over the Phenomenon (middle:
Phenomenon as Range) see Chapter 4, Section 4.3.2.3.1 above. This kind of a reversal
of perspective on the figure (like/pleasef fear/frighten, believe/convince, etc.) is a special
feature of sensing and one important reason for recognizing it as a distinct type. (This is
also true of the metaphorical variant; cf. John was afraid of the dog : The dog was scary
[to John]. )

490

12. Figures and processes

The sensing example discussed by Dik, The dog frightened John, is phenomenal rather
than metaphenomenaL Consequently, it does not illustrate one of the other reasons for
recognizing sensing as a distinct type: as we have seen (reference as above), a figure of
sensing may combine with a metaphenomenon or fact (It frightened/surprised/pleased
John thai there was no reply to his letter. ) They are, of course, quite different from
figures of doing in this respect For example, while metaphenomena can please and
frighten conscious beings (sensing), they cannot open doors (doing).
Table 12(9) shows the comparison of Dik's typology of states of affairs with our
typology of figures.
Table 12(9): Dik's typology and the process types

doing
action

John ran away

process

John fell
down

sensing

being &
having

The dog
frightened
John [effective]

position
state

saying

John stayed
motionless
John saw a
beautiful bird
[middle]

Roses are red

As the table shows clearly, Dik's account does not embody any generalizations about
figures of sensing; it takes no account of projection, reversibility, the wide range of
possible entities serving as Phenomenon, or the constraint on the Senser requiring it to
be conscious. It also fails to recognize saying as a distinct type, where again projection is
a major consideration. Dik's account thus does not recognize what, according to our
interpretation, the grammar construes as the symbolic centre of the universe. Being &
having is split between positions and states; and examples such as John became the
leader; the milk turned sour would presumably be categorized as process. Since being &
having is fragmented, the generalization that the different types of expansion (elaboration,
extension, and enhancement) are manifested as different subtypes of being & having
intensive, possessive, and circumstantial would be difficult to make. Yet such
generalizations are needed to explain the nucleus of these figures (process + participants)
and the role they play in sequences, especially those involving projection.

1 2.4 Foley & Van Valin

491

12.4 Foley & Van Valin


12.4.1 Two sets of semantic roles according to specificity/ subsumption
Foley & Van Valin (1984: Ch. 2) characterize their approach to transitivity within 'Role
and Reference Grammar* as follows:
One of the most fundamental problems in the analysis of clause structure is the
characterization of predicates and the semantic relations which obtain between them and
their arguments. ... In this chapter we will develop a system for capturing the semantic
role structure of the clause. This system is based on an opposition between the notions
of actor and undergoer, on the one hand, and a program of lexical decomposition of
predicates into a set of primitive predicates and operators, on the other.

As the quote indicates, they propose two sets of semantic roles:


macro-roles: These are "generalized semantic relations between a predicate and
its argument": actor, undergoer.
token-specific semantic relations: Verbs are decomposed into predicates,
general operators (DO, BECOME) and connectives (CAUSE); and token-specific
semantic relations (thematic relations) are assigned according to this semantic
representation of verbs: agent, effector, locative, theme, patient.
It seems to us that there are two advances here over 'classical* case grammar:
(i) macro-roles aie introduced, making a number of generalizations possible, in
particular in the 'inter-face' to syntactic functions, for example in the statement
of voice (Actor vs. Undergoer as Subject)
(ii) specific 'case roles' are identified on a more principled basis, with reference to
verb classification.
But, interpreted in systemic terms, the relationship between the two sets of roles, macroroles and token-specific semantic relations, is one of delicacy. This does not capture the
systemic insight that a transitivity system in one language may embody complementary
perspectives (e.g., ergative + transitive); and that different languages use different
transitivity models (contrast English and Tagalog). The differences between RRG and the
systemic-functional account of grammar (SFG) in this respect can be tabulated as follows.

12. Figures and processes

492

Relationship between

RRG

SFG

two sets of roles:


difference of perspective (New)

ergative model and transitive


model

difference of delicacy

macro-roles and token-specific generalized ergative roles and


semantic relations

specific transitive roles

ID contrast to the descriptive transitivity categories used in SFG, the RRG categories are
universal in the eue sense (p. 28):
These constructs have the same status as Jakobsonian distinctive features; languages
employ different sets of them in a variety of ways, but they provide the basis for a
unified description of aspects of phonological systems. Thus the concepts to be
developed in this chapter should be considered etic categories out of which the analyst
constructs emic analyses of the verb systems of particular languages.
12.4.2 Macro-roles
The macro-roles, actor and undergoer, are characterized as follows (p. 29):
actor: "the argument of a predicate which expresses the participant which
performs, effects, instigates, or controls the situation denoted by the predicate"
undergoer: "the argument which expresses the participant which does not
perform, initiate, or control any situation but rather is affected by it in some

way"
Foley and Van Valin give the following introductory example:
Ihe hunter shot the bear
Actor
Undergoer

Ihe beau: was shot by the hunter


Undergoer
Actor

The example illustrates that Actor and Undergoer may be mapped onto "syntactic
subject" and "syntactic object" in different ways. Foley & Van Valin point out that this is
true also of "single-argument predicates":
John ran down the street
Actor

Ihe janitor suddenly became ill


Undergoer

Mary swam for an hour


Actor

The door opened


Undergoer

The boy went to the store


Actor

Fritz was very unhappy


Undergoer

Since they are generalized roles, Actor and Undergoer are not equivalent to case
roles or thematic relations. Both Actor and Undergoer can range over a number of more

493

12.4 Foley & Van Valin

specific case roles. Foley and Van Valin give the following examples to illustrate this
point:
Colin killed the taipan
Actor/
Agent

Phil threw the ball to the umpire


Undergoer/
Theme

The rock shattered the mirror


Actor/
Instrument

The avalanche crushed the cottage


Undergoer/
Patient

The lawyer received a telegram


Actor/
Recipient, Goal

The arrow hit the target


Undergoer/
Locative

The dog senses the earthquake


Actor/
Experience!

The mugger robbed Fred of $ 50.00


Undergoer/
Source

The sun emits radiation


Actor/
Source

The announcer presented Mary


Undergoer/
Recipient/
Goal
with the award.

The characterization of Actor and Undergoer does not differentiate the ergative and
transitive perspectives on transitivity. Even though the characterization of Actor and
Undergoer suggests that they correspond to the transitive pair Actor + Goal, Actor and
Undergoer are not in fact applied in this way. For instance, the Undergoer corresponds to
Actor [SFG] as well as to Goal [SFG], and in this respect, it resembles Medium [SFG];
but Actor [RRG] may also correspond to Medium, as the examples in Table 12(10)
show.
Table 12(10): Comparison of Actor & Undergoer with Agent & Medium
Agent
[material & effective:]
Colin killed the taipan

Actor

Undergoer

Medium
[material & middle:]

The sun emits radiation

[mental & middle:]


The dos senses the earthquake
[material & effective:]
Phil threw the ball to the umpire
[material & middle:]
The door opened
[relational & middle:]
Fritz was very unhappy

Contrast the characterizations of Actor and Undergoer in RRG given above with those
of Medium and Agent (Halliday, 1985:146-7):
Medium: "Every process has associated with it one participant that is the key
figure in that process; this is the one through which the process is actualized, and

494

12. Figures and processes

without which there would be no process at all. ... It is the entity through the
medium of which the process comes into existence."
Agent: "... in addition to the Medium, there may be another participant functioning
as an external cause.... Either the process is represented as self-engendering, in
which case there is no separate Agent; or it is represented as engendered from
outside, in which case there is another participant functioning as Agent"
The Actor-Undergoer model in RRG, when applied to English, does not capture the
significant generalizations embodied in this notion of Medium, such as
the aMedium is restricted relative to the Process in the same way regardless of
whether the clause is effective (+ Agent) or middle: the door opened : she
opened the door :: the ice melted : she melted the ice :: the milk spilt :
she spilt the milk :: she grieved (at the news) : the news grieved her
the Medium is the participant that is most closely bonded to the Process by
lexical collocation.
the Medium is the element which combines with the Process to form the clause
nucleus; it is this nucleus that determines how the process is subclassified and
interpreted (cf. run + dog; run + factory; run + nose; cut + hair; cut + grass; cut
-i- meat; kill/ die + animal; kill/ die + light; kill/ die + motion) and which carries
the main systemic potential (contrast open + door : open + bank account; run +
factory : run + jogger).
the Medium is the one participant that cannot be treated circumstantially; hence
(other than in one special kind of clause, the medio-passive) it is never mediated
by a preposition.
12.4.3 Token-specific roles and verb classification
Foley & Van Valin (1984: 33-4) note that there are two approaches to token-specific roles
(relations):
(i) Fillmore's (1968) case grammar Here a number of case roles is identified as a
universal set. These roles exist independently of (classes of) predicates, though
predicates (verbs) may be classified according to case frames.
(ii) Gniber's (1965, 1976) & Jackendoffs (1972, 1976) thematic relations: Tbese
aie explicitly derived from the semantic decomposition of predicates. Thematic
relations "aie a function of the argument positions of abstract predicates" such as

495

12.4 Foley & Van Valin

CAUSE, and GO; for instance, GO is a three-place predicate: GO (x, y, z), where
x =* theme, y source and z = goal [in the sense of destination].
Foley & Van Valin follow the second approach, but they base their decomposition not
on Jackendoff s work but on Dowty's (1979) verb classification, which is in turn based on
Vendter's (1967) verb classes. As we saw in Chapter 12, Section 12.2 above, Vendler's
typology is based in the first instance on considerations of how the Process unfolds in
time its temporal profile. In contrast, the process types used in systemic
interpretations are based on considerations of the nature of nuclear transitivity how
participants interact, what kinds of entities they are, whether projection is possible, etc.
Note that semantic decomposition is a way of handling paradigmatic, systemic relations
as if they were syntagmatic: general features such as 'go1 and 'cause* are actually posited as
constituent elements of the abstract semantic decompositions of verbs/ predicates. See
further the next section below.
12.4.4 Dowty's (1979) development of Vendler's verb classes
Dowty bases his account on Vendler's four classes of verb states, activities,
accomplishments and achievements. The criteria used by Dowty (1979) for distinguishing
states, activities, accomplishments and achievements, ate summarized by Foley & Van
Valin (1984: 37); they are essentially concerned with establishing the temporal profile of
the process unfolding in time (see Table 12(11)).
The basic program quoted by Foley & Van Valin is described by Dowty (1979: 71) as
follows:
Tbe idea is that the different aspectual properties of the various kinds of verbs can be
explained by postulating a single homogeneous class of predicates - stative predicates
- plus three or four sentential operatives and connectives. English stative verbs are
supposed to correspond directly to these stative predicates in logical structure, while
verbs of the other categories have logical structures that consist of one or more stative
predicates embedded in complex sentences formed with these 'aspectual* connectives and
operators.

It is important to note that Dowty is here concerned with explaining aspectual


properties of verbs i.e.f properties that have to do with the unfolding of the Process in
time and not, in the first instance, with the Process as part of a transitivity configuration
with participants and circumstances.
The operators and connectives are as follows:
DO "under the unmediated control of the agent" (Dowty, 1979:76)
BECOME "where 0 is any predicate formula and t is any time, BECOME 0 is
true at t if ^ is true at t and false at I-1 ".

(Dowty, 1979: 118)

CAUSE a logical connective between events (cf. our enhancing sequences).

12. Figures and processes

496

Table 12(11): Criteria for distinguishing verb classes


accomplishments
paint a
run, walk,
swim, push a picture,
cart, drive a make a
chair,
car
deliver a
sermon,
drawa
circle,
recover
from illness

achievements
find, lose,

OK

OK

OK

bad

bad

bad

OK

OK

N.A.

N.A.

states
know,
believe,
desire,
love;
have

examples

temporal
profile

other

meets non-stative tests


has habitual
interpretation in simple
present tense
V for an hour, spend an
hour V-ing
V in an hour, take an
hour to V
V for an hour entails V
at all times in the hour
x is V-ing entails x
hasV-ed
complement of 'stop*
[i.e., in phase with
'stop']
complement of 'finish*
[i.e., in phase with
'finish']
x V-ed in an hour
entails x was V-ing
during that hour
ambiguity with 'almost*
occurs with 'studiously,
attentively, carefully'
etc.

..

activities

N.A.

die;

recognize,
spot

OK

OK

OK

bad

bad

bad

OK

bad

N.A.

N.A.

bad

..
OK

"

OK

bad

The logical structures of the semantics of the four verb classes are given in Table
12(12) (Foley & Van Valin's Table 2, p. 39).
Table 12(12): Verb classes and logical structure
verb class

logical structure

state
achievement
activity
accomplishment

predicate' (x)
BECOME predicate' (x)
DO (x, (predicate* (x) )
CAUSE (where is normally an activity verb and
an achievement verb)*

NOTE: * This statement is taken from Foley & Van Valin; in fact, however,
would not be verbs but logical structures involving interpretations of verbs.

is

and

497

12.4Foley & Van Valin

12.4.5 Verb classes in RRG


As noted earlier, the semantic decomposition illustrated here is really a way of handling
paradigmatic, systemic relations in syntagmatic terms. However, the 'operators' and
'connectives', instead of being interpreted concretely as components of the logical structure
of predicate (syntagmatic interpretation), can be construed paradigmatically, as systemic
features: see Figure 12-11.
This systemic interpretation raises interesting questions which are not brought out in
Foley & Van Valin's decompositional approach: whether the systems are strictly
experiential or quasi-logical, and whether the set of features is exhaustive or should be
increased to include other types of expansion (cf. Chapter 3, Section 3.2 where these are
shown to be pervasive throughout the semantic system).
The status of the Vendlerian verb classes has to be questioned. Foley and Van Valin
(p. 39) note:
there are a number of interrelations among the verb classes which are not easily
expressed in the Vendler foui-way classification ... For example, activity verbs may
become accomplishment verbs if a definite goal is added, e.g. walk (activity) versus
walk to the store (accomplishment).

Table 12(13): The RRG verb classes


TYPES
state

achievement
activity

locational
non-locational

potential
controllable

state/ condition
of being
possession
perception
cognition
controlled
uncontrolled

motional
accomplish
-ment

EXAMPLES
be in, at, on, under
be tall, sick, dead,
nappy, a/raid, fat
have
saw, smell
know, think, believe

ROLES
theme, locative
patient

smile intentionally;
walk, swim, talk;
ignore
smile instinctively
fall, roll, rotate
break glass

agent

locative, theme
locative, theme
locative, theme
BECOME + state

effector
theme
x CAUSE (BECOME
state)

This is one important reason why these classes are not verb classes the classification
actually derives not only from the verb but also from participants and circumstances:
whether or not they are present, and if so, whether or not they are bounded, specific, etc..
This again suggests (cf. Section 3.1) that the Vendlerian classification is not a system
that is inherent in the general system of the grammar or the semantics: it is merely a
possible interpretation of the profile of the unfolding of a process in time, one that is
determined by many factors. Foley and Van Valin (Section 2.5, pp. 47-53) suggest how

498

12. Figures and processes

Dowty's typology and logical decomposition can be used to derive case roles from
predicate structures in a principled way. They recognize some further subclasses of states
and activities. The different types and semantic relations they establish are tabulated in
Table 12(13) above.

Fig* 12-11: Systemic interpretation of Dowty's verb classes


The types recognized here are interpreted in system network form (states and activities)
in Figure 12-12. The semantic roles can then be introduced by means of realization
statements: see Figure 12-13. As soon as they are made explicit in this way, a number of
questions arise; for example:
(i) why are 'possessive1, 'perception' and 'cognition' not grouped together, since
they have the same structural consequence in terms of semantic roles: +
Locative, +Tbeme?
(ii) how are the semantic roles restricted what kinds of entities can serve as
Theme, Patient, etc.? (Note that this has to be made explicit in a system network
with realization statements.)
(iii) what is the entry condition to the system network fragment shown above?

12.4 Foley & Van Valin

499

(iv) what other predictions about transitivity can be made, e.g. concerning
projection? (AU that the account makes explicit is the distribution of roles such
as Theme Locative, and Effector.)

Fig. 12-13: Addition of realization statements

500

12. Figures and processes

In their reasoning about how some of the verb types should be interpreted in terms of
semantic roles, Foley & Van Valin's account differs from our systemic account of types
in the ideation base in two respects:
(i) Foley & Van Valin reason from accounts of the 'real world';
(ii) Foley & Van Valin argue from paraphrases and metaphorical variants.
(i) Their interpretation of perception seems to be based on a physical interpretation
rather than an exploration of reactances in the grammar (pp. 48-9):
A consideration of what happens in a perceptual event yields a clue to the logical
structure of perception verbs. In physical terms, a stimulus of some kind, e.g. visual,
auditory, or tactile, comes into contact with a sense organ of the perceiver, and this sets
off a complex chain of events in the nervous system of the perceiver. Since the crucial
feature of this process is the contact between the stimulus and the sense organ,
perception may be viewed as having an essential locational aspect, and accordingly we
will analyze perception verbs as having a locative component to tHHT meaning. (See
Wierzbicka 1980 for detailed arguments in support of a locative analysis of perception
verbs.)

By themselves, these observations would seem to have no implications for a semiotic


interpretation of perception: the grammar and semantics may construe our experience of
the world in the same way as (modem) science does, but there is no reason why this
should necessarily be the case. What is relevant is how it is construed in the grammar.
(ii) In their interpretation of verbs of cognition, Foley and Van Valin appeal to
paraphrases and metaphorical renderings (pp. 49-50):
Cognition and prepositional attitude verbs such as know, think and believe
explicitly stative paraphrases which reveal features of their logical structure.

have

(2.29) a. Fred believes/ thinks that Ronald is a fool.


b. Fred is of the opinion that Ronald is a fool.
c. Fred holds the belief that Ronald is a fool.
(2.30) a. Barbara knows French cooking thoroughly.
b. Barbara has a thorough knowledge of French cooking.
(2.31) a. Max knows that the world is round.
b. It is known to Max that the world is round.
The paraphrases of the (a) sentences in these sets of examples reveal the stative nature
of these verbs, especially (2.29b) and (2.3 Ib), and the possessive or 'having' aspect of
their meaning, as in (2.29c) and (2.30b). We argued above that alienable possession is
at least in part a locational relationship, and consequently it appears that knowing and
believing also have a locative component to their meaning. The location aspect of
cognition is brought out even more clearly in the various metaphors we commonly use.
... In these metaphors and many others thinking is characterized as having something
in or on one's mind. Thus an important part of the meaning of cognition verbs is
locational, and from the examples in (2.29) - (232) we may conclude that the thought,
belief, or knowledge is located with respect to the thinker, believer, or knower's mind.

501

12.4 Foley & Van Valin

This line of argumentation seems very similar to the strategy that was used in classical
transformational grammar: one constniction is observed to have a paraphrase (often
metaphorical, as we can now see), and the constniction is then analysed as if it was in this
paraphrased form derived from it, in the case of transformational grammar. The fact
that we find systematic pairs of process : process + range such as (doing) dance : do a
dance :: (sensing) know : (being & having) have knowledge is significant; but it does
not mean that dance is semantically "really" 'do a dance1 or that know is semantically
"really" 'have knowledge'. If we argued along these lines, most of grammar could be
reconstmed metaphorically as forms of being & having, realized by relational clauses (as
we saw in Chapter 6); compare:
She is exploring the other alternative :
She is in engaged in the exploration of the other alternative
He is patrolling :
He is on patrol
They are fighting :
They are at war
It hasn't rained for seven years so forests burn frequently :
Ihe seven-year lack of rain leads to frequent forest fires

Such interpretative reductions tend to obscure the multiplicity of perspectives built into
the transitivity system.
Before leaving the RRG verb classes, we can intersect them with the primary types of
figure we have posited here. The comparison is based on the examples discussed in Foley
& Van Valin (1984): see Table 12(14).
Table 12(14): RRG verb classes and our types of figure
being
&
having
state

achievement
activity

locational
nonlocational

potentially
controllable

state/
condition of
being
possession
perception
cognition
controlled

uncontrolled
motional
accomplish
-ment

sensing

affection
??

recognize
realize
ignore

saying

doing

502

12. Figures and processes

There are some types within the ideation base where it is uncertain what the equivalent
RRG classes are: meteorological, effective sensing (the 'please1 type), affective sensing,
saying, identifying, and causative ascriptive being & having. The absence of saying from
the list conforms to the pattern we have found with other similar typologies; yet saying is
signicant as the key to the construal of projection and non-material experience. Let us
now turn to the account of participant roles.
12.4.6 Relationship between the two sets of roles

Foley & Van Valin characterize the mapping between macro-roles and 'token-specific1
ones as follows (pp. 54-5):
Actors may be agents, effectors, or locatives, depending primarily upon the verb it (sic)
cooccurs with. In any given clause, although arguments bearing more than one of these
semantics relations may be present, the actor may be interpreted as only one of them,
and the possibilities for interpreting the actor, i.e. the linking of NPs with argument
positions in the logical structure of the verb, always follow a strict hierarchy.
... Tfous the hierarchy of accessibility to (or interpretation of) actorhood is agent >
effector > locative, when more than one of the relations occurs in the clause.
The situation regarding the undergoer is more complex, because a given verb with a
single logical structure may allow two arguments bearing different semantic relations to
be undergoer. There are no such variable choices for the actor.
An example of variable undergoer assignment is ... a. John gave the book to Bill b.
John gave Bill the book.

Foley & Van Valin summarize their observations concerning the relationship between
macro-roles and 'token-specific1 semantic roles: see Figure 12-14.

Fig. 12-14: Foley & Van Valin's mapping between the two sets of roles

12.4 Foley & Van Valin

503

As noted earlier, the macro-roles are seen as an 'interface* to syntactic functions such as
Subject and Object They facilitate the specification of a number of syntactic
constructions such as voice (active/ passive).
As noted at the beginning, the RRG account at this point rests on the notion of
generality. However, the systemic accounts of transitivity in e.g. English and Chinese
rest on the notion of alternative perspectives. Hie participant roles Medium, Agent,
Range and Beneficiary embody not just a different degree of generality from the processspecific participant roles Actor, Goal, Recipient; Senser, Phenomenon; etc. but, more
importantly, a different perspective. They can capture different kinds of facts about the
transitivity system of a language. Moreover, as Martin's (1996a) work on Tagalog
transitivity shows, it is important to allow for variation in transitivity models across
languages: Tagalog embodies a model that is significantly different from the models
English transitivity is based on.
In RRG, transitivity is treated within the semantics: macro-roles and 'token-specific'
roles are semantic roles, and the 'token-specific' ones are introduced with the logical
structure of semantic predicates (together with the operators DO and BECOME, and the
connective CAUSE). In contrast, we treat transitivity both within semantics (the
paradigmatic and syntagmatic organization of figures) and within lexicogrammar (the
grammar of transitivity): it is a system construed within the content plane of language
both in the ideational component in the lexicogrammar and in the ideation base. This twostrata! approach to transitivity makes it possible to model the resource of grammatical
metaphor and is fundamental to work on multilingual systems for generating text.

12.5 Inherent indeterminacy, and variation


interpretations

in systemic

We have drawn attention at various points to the overall indeterminacy of language,


something that we see as a necessary condition of its ability to function in the constmal
of experience. Our experience of the 'goings-on1 in and around ourselves is so rich and
many-faceted that no semiotic system that attempted to impose on it a rule-bound,
determinate frame of reference would be 'functional' as a resource for survival.
We have tried to show such indeterminacy as a positive feature, and build it into our
suggested meta-construal not as some wayward or exceptional extravagance but as an
unmarked state of affairs that is recognized to be the norm. To this extent, our grammatics
becomes itself a metaphor for the grammar that is, to the extent that we are able to
enact this indeterminacy in our own representations.
In our treatment of figures in the ideation base, we stressed the fluidity of the
boundaries among the various types of figure: doing-&-happening, sensing, saying,

504

12. Figures and processes

being-&-having. We construe these in the grammar as a system of process types: at


primary delicacy, material, mental, verbal, relational. These are sections on a continuum
or, better, regions in an n-dimensional semantic space; but they are not demarcated by
any uniquely self-selecting set of criteria. A stratified semiotic defines three perspectives,
which (following the most familiar metaphor) we refer to as 'from above', 'from
roundabout', and 'from below': looking at a given stratum from above means treating it as
the expression of some content, looking at it from below means treating it as the content
of some expression, while looking at it from roundabout means treating it in the context
of (i.e. in relation to other features of) its own stratum. Adopting this trinocular
perspective we have identified the various process types in terms of their nuclear
transitivity; this seems to us the most operationally useful approach, since it takes
account of what they mean, how they are expressed, and what their systemic potential is
without privileging any one perspective at the expense of the other two.
However, even if we adopt this approach, this still does not impose a single
determinate form on the description. As we have already noted, Martin, in his systemic
treatment of processes in Tagalog (1996a), offers a different interpretation of nuclear
transitivity: he defines it in terms of orientation, rather than configuration, and hence
operates with a significantly different concept of participant function. And even within our
own configuration^ model of English there is often considerable 'play'. A good example
of this is Fawcett (1987). We have referred earlier to Fawcett's work in systemic theory,
especially his COMMUNAL natural language processing system; the overall architecture
of Fawcett's text generation model is somewhat different from Penman, but the basic
design of the grammar is the same. It differs, however, in the treatment of relational
processes, and in the distinction made between relational and material.
The frontier between relational and material processes is typically fuzzy, and as
Fawcett points out the grammar's construal of possession and circumstantiation (the latter
prototypified by location in space and time) may be interpreted in more than one way.
Fawcett incorporates into the "relational: possessive" category, processes of giving and
acquiring; reduces the circumstantial to locational processes only; and includes within
these, processes of going and sending (for discussion of this area, see Davidse, 1996b). As
is to be expected, this alternative analysis embodies certain generalizations that are not
made in our account of figures, and ignores certain others which are. (His abandonment of
the distinction between attributive and identifying seems harder to motivate, since this
cannot in fact be explained as a textual (thematic) system in the way that Fawcett
proposes; cf. Davidse, 1996a.) The point to be brought out here is that there is no unique
'correct description1 of a semiotic system, or of any phenomena that have to do with
meaning. Strata! relations are not relations of cause & effect, and a stratified system
cannot be reduced to constructions of content-expression pairs.

12.5 Inherent indeterminacy, and variation in systemic interpretations

505

This is not to say, of course, when a choice is made among a set of alternative
descriptions for representing features in grammar that are inherently indeterminate, luce the
types of process in a transitivity system, that the choice is insignificant, or merely
random. On the contrary: it resonates throughout the grammatics as a whole (or should
do, if the description has any claim to be comprehensive). Again there is the analogy
between the grammatics and the grammar: just as no region of the grammar is isolated
from the rest, so every descriptive statement has consequences throughout the description.
Fawcett's alternative model for relational processes, with its particular features such as
treating 'giving' and "placing1 as agentive possessives and locatives ('make ... have', 'make
... be at') rather than as material dispositives, has to be understood in its total explanatory
context: (i) in relation to its repercussions within the transitivity system, both the
trinocular perspective on transitivity itself (from above, as generalizations about meaning;
from roundabout, its consequences for agnateness, delicacy and the move towards lexis;
from below, as regularities in the realization) and the overall topology of the content
transitivity in relation to the semantic construal of causality, agency, disposal, and so on;
(ii) in relation to Fawcett's architectural design, which differs from ours in having a single
system-structure cycle for the two strata of semantics and lexicogrammar (his "syntax")
and then adding a further level of description that is expressed in cognitive terms.

12.6 Conclusion
In this part, we have engaged in metalinguistic typology and comparison, trying to bring
out theoretical and descriptive similarities between other approaches and our own. We
have focussed on approaches that were developed in the context of cognitive science and
natural language processing. Hence we have not taken up theoretical issues in social
semiotics and functional linguistics within the European tradition, nor have we explored
descriptive contributions such as Ernst Leisi (1955), Eugene Nida (1975), Adolf Noreen's
(1904-12) monumental Van Sprdk, or Roget's Thesaurus. Among approaches that have
been taken up in cognitive science and natural language processing, we have not discussed
work located within a lexical perspective, such as those of Igor Mel'chuk and others
(including lexical functions) and of Levin (e.g. 1993), or the various research projects in
NLP of an encyclopaedic nature. We have also not been able to deal with important
studies of particular topics within semantics, such as Kiparsky & Kiparsky's (1970)
classic treatment of facts, or Givn's (1980) and Ransom's (1986) approaches ,to the
semantics of so-called complementation. There are many similarities among all these
approaches; but there are also significant differences, and it seems worth pointing out that
these can often be traced back to the nature of the semantic system itself. There seem to
be two main reasons for this.
(i) The system often embodies complementarities providing more than one way of
construing similar types of experience for example, construing perception either as

506

12. Figures and processes

inert sensing or as active behaviour, or construing phenomena of all kinds either


congruently or metaphorically. If one is concerned with a restricted, designed semiotic
system, then the appropriate strategy, when one is faced with more than one option in
natural language, may be to select one canonical form; but if one's concern is with
language as a whole, the only response that will hold up in the long term is to
acknowledge the variation and explore its function.
(ii) The system is polysystemic from the point of view of contextual
diversification: given a particular context type or a set of related types, some parts of the
overall semantic system will be deployed in a systematic way whereas others will remain
inactive. For instance, if contexts where the goal is to instruct somebody in a procedure
were the only source of evidence fa- the semantic system, we could safely ignore figures
of sensing and saying and the logical relationship of projection. And yet in the general
system, which is the overall meaning-making resource deployed across all the different
types of context, these figures and sequences turn out to be essential both to the
functioning of the semantic system and to our own understanding of its fundamental
principles.

Part V:
Language and the construal of experience
13. Language as a multifunctional system-&-process
13.1 Types of system-&-process
We are treating language as a semiotic system, and it may be helpful to locate this
concept within the context of the history of ideas, albeit in a very sketchy fashion. As we
conceive of it, the term "semiotic" is framed within a linear taxonomy of "physical
biological social semiotic"; and the term "system" is a shortened form of "system-process", there being no single word that encapsulates both the synoptic and the
dynamic perspectives (we have referred to the term "climatic system" with the same
observation on how it is to be understood).
Why a "linear" taxonomy? There is an ordering among these four types of system,
which we can appreciate most readily, perhaps, at the "meta" level: this is the order in
which they have come to be studied and interpreted, in the past five hundred years of
human scholarship. We should add here, perhaps, that we emphatically do not equate
history of scholarship with the history of western scholarship; in our own interpretation
of language we have in fact dawn explicitly on other scholariy traditions besides the
western one. But it happens that, for complex historical reasons, it was European
thinkers who first cracked the codes that is, who took our understanding of phenomena
up one level, on to what came to be recognized as the "scientific" plane; and while this is
certainly not the end of the story there are no doubt many higher levels of
understanding still to be attained it is as far as we have got up to the present And the
codes were cracked in a certain historical order the first to be understood, at this level,
were the physical systems; then the biological, and after these the social. There was a gap
of a few generations between each step: the son of interval that separated Galileo and
Newton from Lamarck and Darwin, and, somewhat shortened, the latter from Marx,
Weber and Durkheim. In other words, there was a certain intellectual distance to be
covered in bringing a comparable kind of insight into these different types of system-&process.

508

13. Language as a multifunctional system-&-process

This seems quite understandable when we reflect that there is an ordering among the
phenomena themselves, in the way these systems are each made up. Physical systems are
just physical systems. Biological systems, however, are not just biological systems; they
are at once both biological and physical. Social systems are all three: social, biological
and physical. This makes them increasingly difficult for us to comprehend. This is not
the same as saying that social systems are more complex than biological ones, or
biological than physical; there are too many different ways in which things can be
complex, for any such observation to make much sense. But they are increasingly
complex in this particular respect; and this means that it is increasingly difficult to
recognize the essential nature of the phenomena concerned. What is problematic is the
relationship between the system and the instance; or, to put it another way, what is the
nature of a "fact0 in these different realms of experience? A biological fact is different
from a physical fact, and a social fact is different again; the relationship between that
which can be observed, and the system-&-process lying behind what is observed, is
significantly harder to establish when the system is a social system, because the
phenomena involved are simultaneously of all three kinds.
One consequence of this (and no doubt one reason for the further time it took to crack
the codes) was that each new step required a shift of perspective. For understanding
physical systems, the critical approach was that of measurement; the dominant theme was
mathematics, and the perspective essentially a synoptic one. But this did not serve well
for interpreting biological systems; these are better understood in terms of change, so the
perspective bad to be altered, to become dynamic, with evolution as the dominant theme.
For social systems, however, the dynamic perspective by itself lacked explanatory power,
and in the present century it was overtaken by another synoptic approach, the theme of
structuralism. Our conception of the nature of social systems has been largely moulded in
structuralist terms,
In explaining these thematic shifts in terms of intellectual movements, the challenges
presented by these different domains of human experience, we are not ignoring or denying
the significance of the social and political processes that were taking place throughout
these centuries in Europe. To understand the development of structuralism, like that of
evolutionary theory before it, it is of course necessary to place these thematic shifts in
their general historical contexts. But it would be equally onesided to ignore or deny the
relevance of the intellectual agenda. These make up an essential ingredient in the overall

13.1 Types of system-&-process

509

historical picture, no part of which can he detached as a single ultimate cause of all the
rest
What then of semiotic systems? Once again with apologies for the inevitable
oversimplifying, let us try and identify what it is that is added with each step in the
systemic progression. A biological system is a physical system with the added
component of "life"; it is a living physical system. In comparable terms, a social system
is a biological system with the added component of "value" (which explains the need for a
synoptic approach, since value is something that is manifested in forms of structure). A
semiotic system, then, is a social system with the added component of "meaning".
Meaning can be thought of (and was thought of by Saussure) as just a kind of social
value; but it is value in a significantly different sense value that is construed
symbolically. Meaning can only be construed symbolically, because it is intrinsically
paradigmatic, as Saussure understood and built in to his own definition of valeur.
Semiotic systems are social systems where value has been further transformed into
meaning.
In their earnest lectures in linguistics, our students are encouraged to think of
language in precisely this multimodal perspective. It can be studied as a physical system,
in acoustics and in the physical aspect of articulation (air pressure measurements and so
on). It can be studied as a biological system, in the physiological aspect of articulation
and in the neurophysiology of the brain. It can be studied as a social system, as the
primary mode of human interaction. And of course it can be studied as a semiotic system,
in the core areas of lexicogrammar, phonology and semantics. If linguistics is conceived
of as a discipline that is, as denned by its object of study (in this case, language)
then it must encompass within itself theories and methods of all four different kinds.
The students are also encouraged, of course, to acknowledge semiotic systems other
than that of language: forms of art, such as painting and architecture and music, ritual and
other behaviour patterns; as well as ways of presenting the self in make-up and dress.
Language is set apart, however, as the prototypical semiotic system, on a variety of
different grounds: it is the only one that evolved specifically as a semiotic system; it is
the one semiotic into which all others can be "translated"; and (the least questionable, in
our view) it is the one whereby the human species as a whole, and each individual
member of that species, construes experience and constructs a social order. In this last

510

13. Language as a multifunctional system-&-process

respect, ail other semiotic systems aie derivative: they have meaning potential only by
reference to models of experience, and forms of social relationship, that have already been
established in language. It is this that justifies us in taking language as the prototype of
systems of meaning.
But there is one problem with this intellectual strategy: that the code for systems of
this fourth type has not yet been cracked There ate two aspects to this problem. One is
that we do not yet fully understand the nature of a linguistic fact: this is the problem of
instantiation. The other is that we do not yet fully understand the nature of the
relationship that is the semiotic analogue of the "cause : effect" of classical physics: this
is the problem of realization. It is true that Saussure, and even more Hjelmslev, took
important strides towards an understanding; but we aie still arguing about what Saussure
really meant (to us it seems that he had not clearly separated the two concepts of
instantiation and realization), and Hjelmslev has largely been ignored Sydney Lamb
(e.g., 1966a,b) is almost the only person who has tried to follow through his
achievements. Probably it will be well into the next century before the picture comes to
be clear.
Meanwhile what has become clear is that there is the (by now) familiar interplay
between phenomenon and observer. We have talked of physical, biological, social and
semiotic systems as being categories of phenomena which in an important sense they
are. But they may also be thought of as different stances taken by the observer; thus we
find physical and biological systems being interpreted as semiotic systems, in a kind
of intellectual game which turns out to reveal new aspects of physical and biological
processes. It is obviously beyond our scope and indeed beyond our capabilities to
pursue these matters here. But they add a whole new dimension to our grammatics, to the
concept of a theory of grammar as a metalheory of human experience.
Here we shall simply try to summarize what we mean by interpreting a natural
language in this way, as the form in which human experience is construed. But in doing
so we shall also say something about its other semiotic functions. We have emphasized
the ideational, because this is what the present book is about. But there is a danger that
the equally important, complementary function of language as the means whereby human
societies are constructed may be relegated into the background; so the present chapter will

13.1 Types of system-&-process

511

provide an opportunity to put the ideation base hack into its wider social-semiotic
environment.

13.2 Review of modes of meaning metafunctional and


metaphorical
13.2.1 The metafunctions
We have stressed all along that a language is a system for creating meaning; and that its
meaning potential has evolved around three motifs what we refer to as the
"metafunctions" of ideational, interpersonal and textual, with the ideational in turn
comprising an experiential component and a logical component. These are the multiple
aspects of the content plane the grammar (in its usual sense of lexicogrammar) and the
semantics. Since the powerhouse of language lies in the grammar, we shall refer to them
here as aspects of the grammar; hut it is important to insist that they could not be "in"
the one without also being "in" the other. It makes no sense to ask whether the
metafunctions are grammatical or semantic; the only possible answer would be "yes".
Ideationally, the grammar is a theory of human experience; it is our interpretation
of all that goes on around us, and also inside ourselves. There are two parts to this: one
the representation of the processes themselves, which we refer to as the "experiential"; the
other the representation of the relations between one process and another, and it is this
that we refer to as the "logical". The two together constitute the "ideational"
metafunction, whereby language construes our experiential world. The word "construe" is
used to suggest an intellectual construction though one that, of course, we then use as
a guide to action.
In ter personally, the grammar is not a theory but a way of doing; it is our
construction of social relationships, both those that define society and our own place in
it, and those that pertain to the immediate dialogic situation. This constitutes the
"interpersonal" metafunction, whereby language constructs our social collective and,
thereby, our personal being. The word "construct" is used to suggest a form of enactment
though something on which we inevitably build a theory, of ourself and the various
"others" to whom we relate.

512

13. Language as a multifunctional system-&-process

Textually, the grammar is the creating of information; it engenders discourse, the


patterned forms of wording that constitute meaningful semiotic contexts. From one point
of view, therefore, this "textual" metafunction has an enabling force, since it is this that
allows the other two to operate at all. But at the same time it brings into being a world
of its own, a world that is constituted semiotically. With the textual metafunction
language not only construes and enacts our reality but also becomes part of the reality
that it is construing and enacting.
In all these metafunctions, the language does not take over and reproduce some
readymade semantic space. There is no such space until the grammar comes along to
construe it.
13.2.2 Ideational metafunction:

experiential

The basic component of all experience is change: when something changes from one state
to another, it projects itself on to our consciousness. This may be something in the
external environment; we can see this happening with small babies, who are first jerked
into semiosis by dramatic perturbations such as a loud noise or a flashing light. The
grammar construes this experience of change in the form of a process configuration:
the fundamental element of grammar is a clause, and the clause presents the parameters
within which processes may unfold.
The grammar does this by deconstructing the process into component parts.
Typically, as in English and many, perhaps all, other languages, these are of three kinds:
first the process itself, secondly certain phenomena construed as participants in the
process, and thirdly, other phenomena that are associated with the process
circumstantially.
Suppose we are standing on the shore, and there is a rapid movement across our line
of vision. We construe this grammatically as
birds + are flying + across the sea

This is obviously not the only way such an experience could be "semanticized": it might
be construed as a single unanalysed phenomenon, e.g. it's winging. Some processes are
in fact construed in this way: in English, for example, meteorological processes such as

13.2 Review of modes of meaning metafnctional and metaphorical

513

it's raining. But in most instances the theory propounded by the grammar is that this is a
composite phenomenon, an organic construction of functionally distinct parts: here, a
process are flying, something participating in this process, namely birds, and a relevant
circumstance across the sea. This allows for other things to fly across the sea, such as
insects and aeroplanes; for birds to fly in other locations, such as over the trees, and to do
other things than flying, such as singing or quarrelling. The meaning potential here is
clearly far greater than if a different lexical item was used to construe every possible
configuration.
The significant step that took place in human grammars in this context was,
obviously, the evolution of common nouns or rather, of common words, since verbs
are also "common" in this sense: that is, words denoting classes rather than individuals. It
is usually assumed that these evolved out of "proper" words, prototypically the names of
individual persons; the ontogenetic evidence suggests that this is one source but not the
only one, another source being rather in the interpersonal function. Be that as it may,
construing processes in this way clearly depends on generalizing whole classes of
phenomena; the grammar sets up classes of process, of participant and of circumstance.
There are various ways of doing this; one that is familiar in many languages is by means
of a taxonomy of different kinds of word, as in Figure 13-1. The classes of word may be
distinguished by their internal form, or by the way they are able to enter into larger
constructions (or both). Typically the most complex is the class of circumstantial
elements, because these are themselves often formed as complex constructions; there may
be simple words (a class of adverbs), but there may also be constructs like English
prepositional phrases, the function of which is to bring in other potential participants but
to bring them in indirectly, like the sea in across the sea. The theory behind this is that
there are two ways in which an entity can be involved: either directly as a participant in
the process, or indirectly in a circumstantial role, such as the place where the process
happens. This indirect participant is often construed as participating in a kind of
secondary process tangential to the main one (grammatically, a prepositional phrase is a
reduced variant of a clause).
At the same time, while recognizing a general category of "process" to construe our
experience of change, the grammar also recognizes that not all processes are alike. As
human beings we become aware (and again we can see this in the actions of tiny infants)
that phenomena fall into two distinct types: those happening outside ourselves, which we

S14

13. Language as a multifunctional system-&.-process

can see and hear, and those happening within our own consciousness thoughts and
feelings, and also the sensations of seeing and hearing, as distinct from whatever is seen
and heard. The grammar construes this as a distinction between "material processes" and
"mental processes". Mental processes aie specifically attributed to conscious beings:
humans, and some of our more intimate animal consorts. Languages construe this pattern
in many different ways, and draw the line at different points; as always we are relating our
account to the particulars of English. Here the grammar postulates a third type of process
intermediate between these two: "behavioural" processes, in which inner events ate
externalized as bodily behaviour, like staring, thinking (in the sense of pondering) or
crying.
Language itself, of course, is a form of human behaviour, but "languaging"
constitutes, for the grammar, another distinct type of process, that of "verbal" (or, better,
"symbolic") processes. An act of saying is not simply externalizing inner events; it is
actively transforming them, into an event of a different kind. It then resembles other
semiotic events, many of which do not require a conscious information source (your diary
says you have a denial appointment, the light says stop). And these in turn shade into
something else, which the grammar again construes as phenomenally distinct relations
of identity (including symbolic identity, like red means stop) and attribution. Expressions
of this kind, which in English often have the verb be, hardly seem to fit the label
"process" at all; but the grammar firmly represents them as such, so we call them
"relational processes". They are modelled, in fact, on the two basic relationships that
characterize semiotic systems: realization (identifying processes such as this is ('realizes')
my sister), and instantiation (attributive processes such as she is ('instantiates') a student
of law). Finally, there is the phenomenon of existing still construed, grammatically,
as a type of process. What is said to exist may be an entity, something that persists
through time, like there's a letter for you; but it may also, in many languages, be a
happening, as in there was a fight. Here we have something that could alternatively be
construed as a material process (people were fighting ), which suggests that "existential"
processes are another intermediate type, something between the relational and the
material.
This part of the grammar, then the grammar of clauses , constitutes a theory
about the types of process that make up human experience. In English (which is probably
fairly typical), the three principal categories that we are calling the material, mental and

13.2 Review of modes of meaning metafunctional and metaphorical

515

relational aie rather clearly distinct on a number of formal grounds; the other three appear
as mixed or intermediate types lying on the borderlines. (In fact the category of verbal
process is more clearly distinct than the other two; and in view of its central place in the
semantic system, we have treated it throughout the present study as a primary category.)
The total picture is of a continuum; but not between two poles rather something that
we would represent in the form of a circle. Figure 13-2 shows this in diagrammatic form.

Fig. 13-1: Taxonomy of word classes in English

516

13. Language as a multifunctional system-&-process

Fig. 13-2: Circle of process types (from Halliday, 1985 (1994 edition))
In construing experience in this way, the grammar is providing a resource for thinking
with. A strict taxonomy of separate process types would impose too much discontinuity,
while a bipolar continuum would precisely be too much polarized. What the grammar
offers is, rather, a flexible semantic space, continuous and elastic, which can be contorted
and expanded without losing its topological order. Since it evolved with the human
species, it is full of anomalies, contradictions and compromises; precisely the properties
which make it possible for a child to learn, because only a system of this kind could
accommodate the disorder that is inherent in experience itself.
With each type of process, the grammar associates one or two favourite configurations
of participants. These usually vary considerably from one process type to another, in
English, there is a strong link between the role of "Senser" (the one who knows, thinks

13.2 Review of modes of meaning metafunctional and metaphorical

517

&c.) in a mental process and the personal pronouns he and she, such that putting it in
this role (it didn't believe me, for example) creates an anomaly we wonder what this
"it" could be? At the same time, all processes are interpreted as having something in
common, in that typically there is one participant that is inherently associated with the
process without which the process could not take place at all, like birds in birds are
flying. That may be the only one, in which case that participant is held accountable (even
if involuntarily!) and the process is said to stop there. Alternatively, another participant
may be involved; either as an external agent, like children in children are flying kites, or
as a goal, like a letter in I'm writing a letter. When it comes to mental processes (and
here is one of the contradictions referred to above), the grammar is uncertain whether the
participant other than the Senser is doing duty as agent or not; if I'm doubtful about
something, for example, I may say your story doesn't convince me, which makes your
story look like an agent, or I may say I don't believe your story, which makes the role of
your story very different not exactly a goal, but like an expression of scope. Many
languages display some such ambivalence about mental processes, which do not match
up neatly with material processes in our experience in any obvious way.
As for the elements that make up the process configuration, we have seen that the
foundation of the grammar's theory of experience was laid down in the simplest terms in
the evolution of word classes: verbs, nouns, and others. But the constituents of a clause
are not, in fact, verbs and nouns; they are more complex expressions that have expanded
from verbs and nouns, which we call "verbal groups" and "nominal groups". At this
point, we find a considerable difference between the two, in the kind of expansion that
they engender. The formal patterns vary, as always, among different languages; but the
underlying principles seem to be fairly constant.
Processes take place in space-time which the grammar may model in a variety of
different ways: the happening is upstream or downstream, past or future, real or
imaginary. Typically the point of reference is the speech situation: there is some deictic
feature relating what is being said to the current "moment" in time and space. In English
the deixis is achieved by locating the process on a linear time-scale with 'present1 as a
fulcrum between 'past' and 'future1, or else by locating it on one or other of a cluster of
scales whereby the speaker intrudes his or her own judgment on it, There may be a wide
variety of other attachments to the verb modalities, aspects, phases and the like, which
the grammar construes as features of the process; consider English examples like wasn't

518

13. Language as a multifunctional system-&-process

going to stan trying to help. On the other hand, the processes themselves are not, in
general, construed into systematic taxonomies, and the verb is expanded by grammatical
rather than lexical means.
The participants, on the other hand, which represent the prototype of entities
persisting in time and space, are usually not subject to this kind of modification; but they
are organized in fairly elaborate taxonomies. These may be construed as systematic
relations among different lexical items: thus eyes, nose, mouth, chin are all different parts
of face, and lamb, pork , mutton, beef arc all different kinds of meat. Something of the
same sort happens with verbs, but to a much lesser extent. The other resource for
constructing taxonomies of things is the expansion of the nominal group, and here the
picture is very different from that with verbs. Nouns are expanded lexically as well as
grammatically, so that, while entities (tike processes) are located deictically relative to the
>

here~&- now1, they are also (unlike processes) extensively classified and described. An

example such as those two nice colourful picture postcards of Honolulu that Sandy sent
us shows these resources at work: cards are classified as postcards rather than, say,
playing cards; postcards as picture postcards not plain postcards; picture postcards are
described as colourful, and also (signalling the speaker's attitude to them) as nice; they are
quantified, as two, and specified deictically as those. Further than that, both a
circumstantial feature (of Honolulu) and even an entire process (that Sandy sent us) can be
brought in as characteristics which specify more exactly the particular cards in question.
Thus the grammar has the potential for construing a complex arrangement of classes and
subclasses for any entity which participates in a process; or, on the other hand, of naming
it as an individual, by using a "proper" noun instead of a common one. Proper nouns are
already fully specific, and hence seldom expanded experientially (they are often expanded
interpersonally!); but common nouns are almost indefinitely expandable, and it is this
resource which organizes our universe into its elaborate taxonomies of things.
The third type of constituent of the clause, referred to above as "others" (that is,
elements that are neither verbal nor nominal groups), evolved as the representation of a
kind of "third party" to the process. This may be some qualification of the process in
terms of its manner of occurrence (an adverbial group, in the case of English); or it may
be an entity that is involved in the process but only indirectly (in English, a
prepositional phrase, consisting of preposition plus nominal group). The Une between

13.2 Review of modes of meaning metafunctional and metaphorical

519

direct and indirect participants is a fuzzy one, and sometimes what seems to be the same
role in the process can be construed in either way, for example pairs like:
they gave the winner a prize

they gave a prize to the


winner

he acted Hamlet brilliantly

he acted brilliantly as
Hamlet

she rides her bicycle to work she rides to work on her


bicycle
the fanners chop dovni the trees

the trees are chopped down


by the fanners

where the winner, Hamlet, her bicycle and the farmers appear first as direct and then as
indirect participants. Here the grammar is in fact using the structural resource of plus or
minus preposition to construe a different kind of contrast, having to do with status in the
message. But the distinction is significant because, as we saw above, such
"circumstantial11 elements tend also to function as qualifications not of the process but of
some entity that is itself a participant: as well as the ice lies thinly on the water, with
thinly and on the water as circumstantial elements of the clause, we have the thin ice
(lying) on the water, where these have now become modifiers of ice and then the thin
layer of ice on the water. And since the prepositional phrase has a nominal group inside
it, this opens up the possibility of further expansion, like the ice on the water in the
pond by the oak trees in the corner of the wood. Thus incorporating the circumstantial
element into the representation of a participant does not merely add one feature to the
specification; it allows more or less indefinite scope, particularly in combination with the
incorporation of an entire process. (We have already pointed out that the prepositional
phrase is in fact a miniaturized clause; so the two really constitute a single resource, that
of using a process to specify a particular class of entity.) This potential was crucial to the
development of science and mathematics (see further on this below).
13.2.3 Ideational me ta function: logical
Turning now to the logical part of the grammar's ideational resources: this is the part that
is concerned not with individual processes but with the relation between one process aid
another. In calling this "logical" we are using the term in the sense of natural language

520

13. Language as a multifunctional system-&~process

logic: that is, grammatical logic, not formal logic although, of course, this is the
source from which formal logic is ultimately derived.
The changes that constitute our experience are not all independent of one another. On
the contrary; our experience is that one thing leads to another, and there is in principle no
limit to an experiential chain. But the exact nature of the relationship may vary from one
transition to another; so the grammar construes the relationship between processes
dyadically, in the form of a nexus between a pair of clauses. The first process may have a
second process related to it, by a relationship such as sequence in time, or cause and
effect; this in turn may have another one related to it, either by the same relationship or
by a different one in either case, the relationship is construed as holding between the
members of a pair. So the logical system, within the ideational metafunction, engenders a
different kind of linguistic structure from that of the experiential system. In the logical
world, the parts are not constituents of an organic configuration, like the process,
participants and circumstances of the clause. They are elements standing to each other in a
potentially iterative relationship; and each element represents an entire process.
Two kinds of logical relationship are construed by the grammar in this way. One is
that of "expansion", in which the two processes are of the same order of experience and
the second one is interpreted as in some respect expanding on the first. The other is that
of "projection", in which the second process is construed as belonging to a different Oder
of experience: it is projected, by the first one, on to the semiotic plane. Each of these
defines a complex region of semantic space.
Conceptually perhaps the simplest way of expanding a process is by elaborating on it:
saying it over again (or something very like it, with repetition as the limiting case), or
else exemplifying it, or clarifying it in some other way. The grammar represents this
relationship symbolically in English by prosodie means: the same intonation pattern is
repeated, for example we're shut out; they won't lei us in. But since this does not appear
in writing, various purely written symbols are used instead, typically i.e., e.g. and viz.
The second type of expansion consists in extending one process by construing another
one as an addition to it (with 'and* as the limiting case); or as an alternative to it, a
replacement for it, or as some form of reservation or contrast. Here the grammar typically
employs conjunctions, like and, or, but, instead, besides. The third type of expansion is
one of enhancing the first process by another one setting up a specific semantic

13.2 Review of modes of meaning melafunctional and metaphorical

521

relationship, of which the principal ones are time, cause, condition, concession, and
means, Here again the grammar deploys a range of different conjunctions, which mark
either the enhancing clause (when, because, by, though, if and so on) or the one that is
being enhanced (e.g. then 'at that time1, then 'in that case', so, thus, yet ).
In projection, one process is used to construe another one, such that the latter
becomes a representation of what someone says or thinks. The types of process that have
this power of projection are the verbal and the mental processes: he says (that...), he told
(her to ...), she thinks (that...), she wanted (him to ...). Thus the projection operates at
either of the two strata of the content plane: either that of the wording, where the
projection is by a verbal process, or at that of the meaning, where it is by a mental
process. Because the grammar can project in this way, semiotic events, both those which
are externalized as sayings and those which are internalized as thoughts, are brought
within the overall domain of the phenomena of experience.
Probably some such system of logical relationships between processes will be found
in all languages, though as always there is great variation in the formal resources that ace
deployed, and also in the systematic semantic organization of the relationships
themselves. In English, and many other languages, the grammar makes a systematic
distinction in the relative status that is accorded to the two processes entering into such a
logical nexus. Either the two are construed as being equal in status, or one is construed as
being dependent on the other. In principle, any particular type of expansion or projection
can be interpreted in either way, either as "paratactic" or as "hypotactic"; but in fact there
is some degree of partial association: certain combinations are favoured, and others
correspondingly disfavoured. For example, in English, when one process is construed as a
simple restatement of, or addition to, another, the two are likely to have equal status;
whereas where one is seen as enhancing the other they are usually unequal a means is
secondary to what has been achieved by it, a cause is secondary to its effect. (Note that
these are overall quantitative tendencies; in any one instance the choice may go either
way.) Similarly in a projecting relationship, the two elements in a verbal projection are
typically equal in status, while those of a mental projection are unequal. This is not
surprising: since you can hear what a person says, you give the wording the full status of
a direct experience, as in Mary said, "I will wait here for you tomorrow"; whereas you
cannot observe what a person thinks, so this is more likely to be construed as dpendait
on the process that projects it, as in Mary thought/ decided she would wait there for him

522

13. Language as a multifunctional system-&-process

the next day. In the first, the deictic standpoint is that of the sayer, namely Mary; what
she said is quoted as "direct speech". In the second, the deictic standpoint is that of the
person speaking; what Mary thought is reported as "indirect thought". Again, it is always
possible to report speech and even to quote thought, with the speaker acting as an
omniscient narrator; but those combinations are less favoured in everyday English
discourse.
In its ideational metafunction, language construes the human experience the human
capacity for experiencing into a massive powerhouse of meaning. It does so by
creating a multidimensional semantic space, highly elastic, in which each vector forms a
line of tension (the vectors are what are represented in our system networks as
"systems"). Movement within this space sets up complementarities of various kinds:
alternative, sometimes contradictory, constructions of experience, indeterminacies,
ambiguities and blends, so that a grammar, as a general theory of experience, is a bundle
of uneasy compromises. No one dimension of experience is represented in an ideal form,
because this would conflict destructively with all the others; instead, each dimension is
nidged so that it can coexist with those that intersect with it. We can illustrate these
compromises with some very simple examples from English:
between a single process and a sequence of logically related processes, e.g. she
told me / she gave me to know
between related processes and related elements within one process, e.g. they
fight harder than we fight / they fight harder than us
between the process itself (action, event &c.) and a participant, e.g. the day
dawns / the dawn comes
between two processes and one process with a circumstantial element, e.g. use
a spade to dig the ground / dig the ground with a spade
between a participant in a process and a circumstantial element, e.g. she rides
her bicycle to work / she rides to work on her bicycle
between one cluster of participant roles and another, e.g. the mistake didn't
strike me /1 didn't notice the mistake

13.2 Review of modes of meaning metafunctional and metaphorical

523

between one process type and another, e.g. why do you grieve? / why are you
S2d?

Such pairs are not synonymous, because each of the two representations aligns the
experience in question with a different set of other experiences. It is obvious that there is
scope for an enormous amount of variation in the way dnrent languages accommodate
the innumerable possible complementarities of this kind We have outlined the picture
from the point of view of English, but doing so as far as possible in a way that would
enable the relevant questions to be raised for other languages. In the last resort each
language construes experience in its own way has its own "characterology", as the
Prague linguists expressed it. But every language embodies a working, and workable,
schedule of compromises, that taken all together constitute its speakers' construction of
reality.
13.2.4 Interpersonal me ta function

Our concern in this book is primarily with what we have called the "ideation base", the
systems of meaning into which, through language, human beings construe their
collective and individual experience. Usually when we talk about the linguistic
"construction of reality" this is the aspect of reality that comes to mind. But at the same
time as construing experience in the same breath, so to speak they are also,
through language, enacting their interpersonal relationships; and this interpersonal
component of meaning is no less part of what is constituted for us as "reality". If the
ideational component is language as a mode of reflection, the interpersonal component is
language as a mode of action; and reality consists as much in what we do as in what we
think.
When we say that the grammar enacts interpersonal relationships, we mean
relationships of all kinds from the transient exchange of speech roles in temporary
transactional encounters (How are you? Good, thanks; and you? Coming along.
Now what can I do for you?) to the enduring familial and other networks that constitute
the structure of society. We tend to be less aware of this metafunction of language, at
least in more learned contexts; partly because, as adults in a literate culture, we aie
conditioned to thinking of meaning purely in ideational terms (language as a means of
"expressing thought"), and partly because it is less obvious that talking is a way of doing

524

13. Language as a multifunctional system~&-process

of acting on others (and through them, on our shared environment) and in the process,
constructing society. But the interpersonal and the ideational aie the two facets of our
everchanging social semiotic.
The most immediate way in which we act, grammatically, is through our choice of
speech function. One kind of speech function is a command; this is obviously a way of
getting someone to do something, but we tend to think of it as being in this respect
untypical. However all speech functions are modes of action, whether command or offer,
question or statement, or any of their innumerable combinations and subcategones. All
dialogue is a process of exchanging meaning, in which the speaker is enacting, at any one
time, a particular interpersonal relationship, including his own role and the role he is
assigning to the listener (i.e. he is specifying a network of interpretations for his own and
the others' behaviour). Grammatically, each time he says a clause, he is not only
construing a process (as described in the last section) but also, unless he makes it
logically dependent on another clause, acting out a speech function; and this embodies
two simultaneous choices. The speaker is either giving or else requiring the other person
to give that is, demanding. And the commodity being given or demanded may be
either "goods-&-services" or "information". Each of the four combinations defines one of
the primary speech functions:
give + goods-&-services:

offer

[accept goods-&-services given!]

demand + goods-&-services:

command

[give goods-&-services demanded!]

give + information:

statement

[accept information given!]

demand + information:

question

[give information demanded ! ]

If the commodity being exchanged is goods-&-services, then the action that is given
or dmnaiiflpd is typically a non-verbal one: what is being exchanged is something other
than a construction of meanings, and the meanings serve to bring the exchange about. In
principle the listener need not say anything at all; but listeners usually do, typically by
reversing the role, responding to an offer with a command and to a command with an
offer
[offer] Shall I help you find it? - Yes; please do! [command]

13.2 Review of modes of meaning metafunctional and metaphorical

525

[command] Come and help me find it! - All right; I will, [offer]
If on the other hand the commodity being exchanged is information, then this is in fact
made of meaning; the speaker's action, and that of the listener if responding to a question,
is bound to be a verbal one, because here language is not only the means of carrying out
the exchange, it is also the nature of the exchange itself.
Choosing a particular speech function is, obviously, only one step in a dialogue;
what the grammar creates, through the system of "mood", is the potential for arguing, for
an ongoing dynamic exchange of speech roles among the interactants in a conversation.
The mood system, together with other systems associated with it, constructs a great range
of speechfunctional variation; and since in principle any ideational meaning can be
mapped on to any interpersonal meaning, this makes it possible to construe any aspect of
experience in any dialogic form. If the ideational metafunction is language in its "third
person11 guise, the interpersonal is language in its "first and second person" guise; the
interaction of a 'me1 and a 'you'. The 'me' and the 'you' are of course constructed in
language; they have no existence outside the social semiotic. Once constructed, me and
you then become a part of experience and can be referred to alongside the him, the her and
the it; but note that (unlike their interpersonal meaning, which does not change) their
ideational meaning changes every time there is a change of speaker (this is what makes
me and you so difficult for children to learn). Note in this connection that the basic
distinction constructed by language is not, as sometimes claimed, that between 'me* and
'the rest'; it is that between 'me-&-you\ on the one hand, and the rest the 'third party'
on the other. This distinction is coded in the grammar at many places, for example in
the system of modality (see below); there can be no "first person" unless there is a
"second person" with whom these roles can be alternately acted out.
The mood system constructs the clause as a move in an argument: either as a
"proposition" (statement or question) or as a "proposal" (offer or command). The system
provides scope for argument by incorporating an opposition of 'on1 or 'off: each clause
assigns either positive or negative polarity. Every proposal or proposition selects erne or
the other: either that was a snake or thai wasn't a snake, either catch it! or don't catch it!
But at the same time the interpersonal grammar goes much further, it rejects a simple
polarity of 'yes1 and 'no', opening up a broad semantic space in between. This is the area
of "modality", where the interactants present different aspects of their own judgments and

526

13. Language as a multifunctional system-&-process

opinions, exploring the validity of what is being said and typically locating it somewhere
between the positive and negative poles.
Languages differ considerably in their construction of this space, and in the extent to
which they interpret it grammatically. In English, there are four distinct grammatical
traverses between 'yes* and 'no', two deriving from the polarity of propositions ('it is/ it
isn't') and two from the polarity of proposals ('do! / don't!'). The former are along the
dimensions of "probability" and "usuality", the first being the more elaborated of the two
(because more arguable); for example,
That was a snake.
It wasn't. It can't possibly have been a snake.
Couldn't it? Don't you think so? I think it mi#it have been.
- - I t probably wasn't. But snakes can appear round here.
All of these represent probability except the final can which means 'sometimes*
(usuality). Note that it is always the judgment of speaker or listener that is represented as
a choice of modality, not that of any third party (this is one of the boundaries drawn
between 'me-&-you' and 'the rest'). The latter occupy the dimensions of "obligation" and
"readiness", readiness including both inclination and ability; for example,
They ought to clean this place up. People will leave it so untidy.

Ihey can't; they haven't got the equipment. They're not

supposed to clean it anyway.


You mean they won't. But someone must. Can we?
I don't see whey we shouldn't. Will you help?
Although these are derived from the sense of proposals ('you are required / supposed /
allowed; I am able / willing1), they are not restricted to clauses having these speech
functions; obligation and readiness are construed by the grammar propositionally and
hence are used freely with third persons. But they still represent the judgments of speaker
or listener on the obligations or inclinations involved (he ought to help, she will help).
Modality is a rich resource for speakers to intrude their own views into the discourse:
their assessments of what is likely or typical, their judgments of the rights and wrongs of
the situation and of where other people stand in this regard But there are numerous other

13.2 Review of modes of meaning metaftinctional and metaphorical

527

kinds of interpersonal meaning constructed by the resources of the grammar. These


include comments about how desirable or plausible or self-evident something is,
expressions of attitude in referring to persons and objects, sets of words with similar
experiential meaning but distinguished interpersonally by connotation (sometimes called
"purr words" and "snarl words"), and numerous forms of personal address and reference
(kinship terms, personal names, honorifics, endearments, insults and the like). For a
recent theoretical account of the field of "appraisal", see Martin (in press); for an analysis
of appraisal in casual conversation, see Slade (1996).
Unlike ideational meanings, which tend to be located at definable locations in the
grammatical structure, interpersonal meanings tend to be strung throughout the discourse,
by an accumulation of grammatical and lexical features or by other devices such as voice
quality and intonation contours. This signals the fact that interpersonal meanings aie
more diffuse: they relate to the figure as a whole, rather than to one of its elements; or to
a whole turn in the dialogue, or even to some more extended passage of the discourse.
Some particular interpersonal colouring may inform the whole of an individual speaker's
interaction with another person; and because human societies are inherently hierarchical,
the interpersonal component of the grammar in many languages enacts networks of social
relationships with varying degrees of inequality and of distance. Thus there may be
regular lexicogrammaticaJ variants used to maintain different alignments of speaker and
listener, and even of third parties, on vectors of power and familiarity; such forms may be
located at one point in each grammatical structure (for example in the endings of the verb)
or dispersed prosodically throughout the wording of the clause. But even in a language
such as English, where there are no such systematic speech styles institutionalized in the
grammar, there is always some functional variation along these lines: we have no
difficulty in recognizing what are the more formal and what are the more informal
variants among different samples of spoken or written discourse.
Interpersonal meaning is mapped on to ideational meaning at all points from the most
micro to the most macro: from modality and speech function in the clause (or even
features built in to the morphology of the word, like the diminutives characteristic of
many languages) to settings affecting the whole of a particular register, like the aura of
power and distance that we associate with the language of bureaucracy. These are the
various ways in which language functions as a mode of action; and these meanings, no
less than ideational ones, are brought into existence by the grammar.

528

13.2.5 Textual

13. Language as a multifunctional system-&-process

metafunction

There is a third component in the linguistic construction of meaning; this is what we


refer to as the "textual" metafunction. If we were trying to find a term to match the
expressions "language as reflection" and "language as action" that we used to gloss the
ideational and interpersonal metafunctions, we might come up with "language as
information"; but this is itself not very informative. It is a difficult concept because,
unlike the other two, the textual metafunction has no obviously distinct function at the
back of it All uses of language involve the creation of text.
But at the same time this is precisely the context in which the textual metafunction
may be understood. The concept of a metafunction is "meta" in the sense that it refers not
to functions of individual utterances functions of the instance but to functional
components of the system of language. They are, of course, "functional" in origin that
is, they derive from the functions of language as manifested in instances of use; and if we
observe children developing their mother tongue we can see how the ideational and
interpersonal resources of the system gradually emerge from the earliest semiotic
encounters, in a way which may plausibly mimic how the metafunctions originally
evolved. The textual metafunction is different because it does not originate in an extrinsic
context of this kind. Rather, it is intrinsic to language itself. Hie "textual" metafunction
is the name we give to the systematic resources a language must have for creating
discourse: for ensuring that each instance of text makes contact with its environment. The
"environment" includes both the context of situation and other instances of text. Relative
to the other metafunctions, therefore, the textual metafunction appears in an enabling
role; without its resources, neither ideational nor interpersonal constructs would make
sense.
Since these resources are oriented towards discourse, many of the "textual" systems in
any language have a domain potentially higher than the clause and clause complex; they
set up relationships that create semantic cohesion, and these ate not restricted by the
limitations of grammatical structure. But they also contribute a critical dimension to the
overall grammar of the clause. In this guise, the clause functions as a quantum of
information; it is construed as a message, with a range of possible structures providing
for different interpretations according to the discourse environment in which it occurs.
These structures have been less fully described than those of the other metafunctions; they

13.2 Review of modes of meaning metafunctional and metaphorical

5 29

were brought to the notice of grammarians by Mathesius and his colleagues of the Prague
school in the first half of the present century. But it seems that one typical way of
construing the clause as a message is as a combination of two perspectives, that of the
speaker and that of the listener (the latter, of course, being also as modelled for the
listener by the speaker). We can see this pattern clearly present in English.
From the speaker's point of view, a piece of information has a specific point of
departure; the Prague scholars rcfored to this as the "Theme". The Theme, in English,
always includes one element that has an experiential function, typically a participant in
the process; it may include other elements as well, fa* example an interpersonal
expression of modality if the speaker is tbematizing his/ her own point of view. One way
of signalling what is thematic is by putting it first in the clause, as is done in English,
where everything up to and including the first experiential element constitutes the
speaker's chosen point of departure; for example But surely time is defined as that which
you can't turn back?, where the Theme is but surely time. Here the speaker is construing
a message around the theme of 'what I'm saying is contrary to what went before' (but);
'it's my opinion, and I'd like to challenge you with it* (surely); and 'the startingpoim is
the topic of time'. The remainder of the clause constitutes the body of the message,
labelled grammatically as the "Rheme".
But the "Theme + Rheme" configuration becomes a message only when paired with
another one, that of "Given + New". This construes a piece of information from the
complementary point of view, as something having news value something the listener
is being invited to attend to. It may not contain anything the listener has not heard
before; a great deal of "news" is totally familiar, being simply contrasted or even
reiterated. On the other hand, the entire message may consist of unknown information,
for example the first clause of a piece of fiction. But the message is construed along
prototypical lines as an equilibrium of the given and the new, with a climax in the form
of a focal point of information: 'this is to be the focus of your attention1. This focal point
usually comes at the end; but (unlike the Theme + Rheme) the Given + New structure is
not signalled, in English, by word order it is signalled by intonation, and specifically
by pitch prominence, the point of maximum perturbation (falling, rising or complex) in
the intonation contour. The principle behind this is clean if the Theme always came first,
and the New always came last, there would be no possibility of combining them; whereas
one powerful form of a message powerful because highly marked is that in which

530

13. Language as a multifunctional systems-process

the two are mapped on to one another, as in no wonder they were annoyed (where the
focus is on the interpersonal theme no wonder).
All languages display some form of textual organization of the clause. How far this
kind of speaker - listener complementarity, with a quantum of information being
construed out of the tension between the two, is a general or prototypical feature of this
aspect of the grammar is not at all clear. In Austronesian languages, for example, there is
typically a much more complex pattern of interrelationship between the textual and the
ideational structures of the clause (mapping of Theme on to different transitivity roles)
than is found in Indo-European. Even with regard to English, where it is well established
how the flow of information is engendered in the grammar, opinions differ as to how tar
this should be seen as one continuous movement and how far as the intersection of two
different periodicities (as we are inclined to interpret it). It may be a general principle that
thematic status is more closely tied to the clause (as the locus of experiential and
interpersonal choices) than is the listener-oriented pattern of given and new; in English
the "quantum" of information that is defined by this latter construction is not, in fact,
identical with the clause and may be smaller or larger. But all discourse is organized
around these two motifs, which between them "add value" to the clause, enabling it to
'mean1 effectively in the context in which it occurs.
Over and above its contribution to the grammar of the clause, what we are calling the
"textual" metafunctional component comprises a further set of resources, which construe
clauses and clause complexes into longer stretches of discourse without the formality of
further grammatical structure. These are the resources for creating "cohesion". These are
of four kinds: reference (sometimes called "phora", to distinguish it from reference as
denned in the philosophy of language), ellipsis, conjunction, and lexical cohesion.
Reference is a way of referring to things that are already semiotically accessible:
either actually, in the text, or potentially, in the context of situation. The English
reference systems are the personals, especially the third person pronouns and determiners
he/him/his she/her/hers it/its they/them/ theirAheirs, and the demonstratives Ms/these
thai/those

and the maverick the (which emerged as a weakened form of that). Such

systems evolved in a deictic function; when used anaphorically or cataphorically (that is,
in deictic relation to the text), they create cohesion. There is also a third source of

13.2 Review of modes of meaning metafunctional and metaphorical

531

referential cohesion, through the use of comparison, with words such as same, other,
different, less, smaller.
In ellipsis, some features which are present in the semantic construction of the
clause (or other unit) are not realized explicitly in the wording, which cannot then be
interpreted unless these features are retrieved from elsewhere. Here it is not the meaning
that is being referred to; it is the wording that is being retrieved, usually from the
immediately preceding clause (whereas reference can span considerable distances in the
text). Ellipsis is particularly characteristic of dialogue, especially adjacency pairs such as
question and answer. Sometimes in English a substitute element is inserted as a
placeholder; e.g. ones in Which lanes are closed? The northbound ones.
In conjunction, the various logical-semantic relations of expansion that construe
clause-complex structures (discussed above under the "logical" metafunction) are deployed
instead as a source of cohesion. There are a large number of such conjunctive expressions,
ranging from single words like however, moreover, otherwise (many of them originally
composite forms) to prepositional phrases like in that case, in other words, at the same
time (often containing a reference word inside them). They cover more or less the same
range of meanings that we referred to as "elaborating", "extending" and "enhancing'; but
they do not establish any structural relationship in the grammar, and this is recognized in
written English, where they regularly occur following a full stop.
Lexical cohesion refers to cohesion that is brought about by lexical means:
choosing a word that is related in a systematic way to one that has occurred before. The
range of semantic relations that can create cohesion in this way is very wide; but there are
five principal conditions under which it occurs. These are: repetition, where the speaker
simply repeats the same word; synonymy/antonymy, where a word is chosen that is
similar or opposed in meaning; hyponymy/meronymy, where a word is chosen that is
related by Tcind of or 'part of either vertically, like melon... fruit or car... wheel, or
horizontally, like melon ... plum, or wheel... mudguard; and collocation, which does not
necessarily imply any particular semantic relationship but means simply that a word is
chosen which is regularly associated with a previous one, like aim coming shortly after
target, such that a resonance is felt between the two.

532

13. Language as a multifunctional system-&-process

As far as the textual metafunction is concerned, therefore, any one clause will
typically embody two sets of semantic choices. One will be its organization as a
message, a piece of information flowing from speaker to hearer, its limits defined by the
speaker's point of departure and the focus of attention projected by the speaker on to the
listener. The other will be the cohesion it sets up with the preceding moments of the
discourse, as well as with other discourses and with the total semiotic environment.
These enable the clause to function effectively as reflection and as action. But in the
course of serving this enabling role, the textual component opens up a new dimension of
meaning potential, in that it construes a further plane of "reality" that is as it were made
of language meaning not as action or as reflection but as information. In the modem
world, when we increasingly live by exchanging information, rather than by exchanging
goods-&-services as hitherto, this aspect of meaning potential is coming more and more
to be foregrounded But it has always been there; and this is not the first time in history
that it has proved to be an indispensable resource.
13.2.6 Ideational metaphor
We described the clause, earlier, as a construction of experiential meaning: a configuration
of process, participants and circumstances set up by the grammar in the course of its
evolution as a theory of experience. We will come back to this in a moment; meanwhile
we have tried to supplement and add depth to this account by bringing in the other two
planes of meaning, the interpersonal and the textual. For a fuller understanding of the
clause, we have to recognize that it evolved simultaneously as reflection, as action and as
information: that is, not only as a representation of the phenomena of our experience but
also as a means of social action, of moving around in interpersonal space (and so defining
that space and those who occupy it); and as a semiotic construct, whereby language itself
becomes a pan of, and a metaphor for, the reality it has evolved to construe and to
construct.
These three "metafunctions" are interdependent; no one could be developed except in
the context of the other two. When we talk of the clause as a mapping of these three
dimensions of meaning into a single complex grammatical structure, we seem to imply
that each somehow "exists" independently; but they do not. There are or could be
semiotics that are monofunctional in this way; but only very partial ones, dedicated to
specific tasks. A general, all-purpose semiotic system could not evolve except in the

13.2 Review of modes of meaning metafunctional and metaphorical

533

interplay of action and reflection, a mode of understanding and a mode of doing with
itself included within its operational domain. Such a semiotic system is called a
language.
We are accustomed to thinking of a language as being prototypically realized as
speech; justifiably enough, since language first evolved in the spoken medium. Also, it
is the spoken medium that language is first mastered in by a child in the typical case.
But, as we remarked in Part I, side by side with spoken language there has evolved
another form of expression that might equally well be taken to represent language in its
canonical form; namely Sign, the mother tongue in communities of the deaf.
13.2.6.1 Deaf Sign
From one point of view, as we introduced it in the earlier discussion, Sign will appear as
a realizational variant: that is to say, the meanings of language its semantic system
may be construed either in sound or in gesture. There are, in fact, sign systems that are
constructed along these lines: in principle this is what is meant by "signed English". But
if we consider the nature of the two forms of expression, vocal on the one hand sou
gestural on the other, it is clear that they have very different properties. Gesture operates
in a 3-dimensional "signing space" defined by reference to the signer's body and its parts,
and movement within that space is entirely accessible to the receiver, thus in addition to
succession in time (which is common to both), the gestural medium can exploit a
number of parameters of spatial variation: the "articulatory organs" (fingers, hands, arms,
other body parts), their location, orientation, thrust (direction and speed of movement) and
so on.

In the most immediate sense, as regards their potential for construing signifiers
elements of wording and their arrangement in combination both these forms of
expression, vocal and gestural, are open-ended Neither of them imposes a limit on the
inventory of morphemes or their configuration in grammatical structures. Nevertheless
they are significantly different in the kinds of resource they offer for making meaning.
Perhaps the major difference between the two is their potential for iconicity. We have
referred all along to the primarily visual nature of human experience: how much of it is
constituted as location, and especially movement, in space. Now, both gestural and vocal
resources involve positioning and moving the organs of articulation in space; but the

534

13. Language as a multifunctional system-&-process

position and movement of the vocal organs, besides being largely out of sight of the
listener, is very much mote constrained; and while this permits a limited degree of
iconicity (association of close vowels with 'small', open with 'large', for example) this
can never be more than a marginal feature of the system as a whole. Thus even allowing
for the additional iconic potential of loudness and length, the expression systems of
spoken languages must remain prototypically conventional. This is the familiar principle
of the "arbitrariness" (i.e. conventionality) of the linguistic sign.
Gestural systems, by contrast, have a far greater potential for construing experience
iconically. Thus in Johnston (1989: 16) "signs arc roughly graded into four classes of
transparent, translucent, obscure and opaque signs, depending on how iconic a
sign is"; and while most signs fall in between the two extremes Johnston grades them
as obscure or translucent, rather than opaque or transparent many of those he labels
"obscure" have a popular explanation in iconic terms (e.g. 2388: n. CAMERA, v. TAKE
A PICTURE, PHOTOGRAPH. Obscure action. Popular explanation: 'holding a camera
and depressing the shutter button' [p. 301]). This suggests that even if particular
explanations aie "nothing more than deaf folklore" (p. 16), the system as a whole is
perceived as prototypically iconic; and this feature is borne out in two important respects.
One is that many of the signs construing basic categories of experience that would be
learnt very early in childhood, in the transition from protolanguage to mother tongue
examples are 1291 GET; 1479 HOLD; 1824 RUN; 2473 BIRD; 2759 DRINK, CUP; 36
BED; 163 UP are clearly iconic, and so would tend to establish iconicity as the norm.
The other is that individual signs may be modified in a distinctively iconic fashion; e.g.
1471 "v LARGE, BIG, (with amplification) great, (with amplification and stress)
enormous, huge, immense"; see in particular the section on "sign modification" in
Johnston (1989: 494-9). As Johnston comments (p. 513), "A language which is itself
visual and spatial has far more opportunities than an auditory one to map onto itself
those very visual and spatial qualities of the world it wishes to represent".
Signers are also members of another language community, that of the (predominantly
hearing) speakers of English, or whatever language is spoken around them; the two
groups interact, and there is obviously no insulation between the two language systems.
This gives rise to contact phenomena of two kinds: on the one hand, intermediate forms
whereby English is realized in sign expressions (signed English, and finger-spelling),
including a large number of new, "contrived" signs; and on the other hand, constant

13.2 Review of modes of meaning metafunctional and metaphorical

535

intrusion of English forms of expression, and therefore of English modes of meaning,


into the sign language itself. The situation is then further complicated by the low status
accorded to Sign by many members of the hearing community, often including the
educational authorities, who (since they did not understand it) not only refused to see it as
a potential vehicle for education and systematic knowledge but in some cases attempted to
suppress it altogether.
But for children who aie born without hearing, Sign performs all the functions of a
naturally evolving mother tongue. It is their primary means for the constmal of
experience, in precisely the sense in which we are using that expression throughout the
present book the construction of their essential "ideation base". (It serves the other
metarunctions as well, of course; for discussion of this see Johnston, 1992.) In precisely
the same sense, but not in the precisely the same way: the semantic system that is
construed in (say) Australian or British Sign is not the semantic system of English,
despite its being constantly permeated by English in the ways referred to above. It is a
system needing to be described in its own terms, based on detailed studies of its grammar
and discourse such as now being carried out for Auslan by Trevor Johnston. Such a
description is designed first and foremost for the needs of the deaf community; but it will
have general significance as a source of further insight into the semantics of spoken
languages, making it possible to view them comparatively in the light of an alternative
construction of reality.
From Sign, we can get further insight into the construal of experience as ideational
meaning because of its greater potential for iconicity in the expression. Semantic space
can be construed iconically in signing within a continuous space-time, constituted as
bodily experience for the signer and as pan of shared visual experience for signer and
addressee.
13.2.6.2 Metaphoric and congruent wordings
From a comparative standpoint, spoken languages and deaf sign languages stand to each
other in a metaphohc relationship, as alternative consonis of a (largely, though not
totally) shared experience. What appear at first simply as different modes of expression
turn out to have, associated with them, somewhat different constructions of meaning. It
would be beyond our scope, and our competence, to pursue this further. But it brings us

536

13. Language as a multifunctional system~&-process

back to another dimension of metaphor to which it bears a rather striking analogy. Many
"spoken" languages may be realized in two different media: in speech, and in writing. At
first this presents itself just as two modes of expression; but when we look more closely
at discourse in spoken and written language we find regularly associated differences in
grammatical construction. We find written language constructed in nominal groups,
whereas spoken language is typically constructed in clauses. And since it is in the
grammar that our experience is construed into meaning, what we are seeing are different
forms of the construction of experience, one couched primarily in terms of figures, the
other in terms of elements that make up such figures, mainly those that function as
participating entities.
Our basic approach to this is embodied in the term "metaphor", as used in the context
of metaphor in the grammar. We used the expression "grammatical metaphor" to refer to
a complex set of interrelated effects whereby, in English and many other languages, there
have evolved what seem to be alternative representations of processes and properties, hi
terms of word classes, meanings prototypically construed as verbs or adjectives come to
be construed as nouns instead. But, as we saw, this is simply the superficial
manifestation of a wider and deeper phenomenon affecting the entire construal of
experiential meanings in the grammar.
To recapitulate with a very simple example: Given a pair of expressions such as in
times of engine failure and whenever an engine failed the two are related to each other by
grammatical metaphor. A particular phenomenon has been construed (i) as a prepositional
phrase with a nominal group as its Complement, (ii) as a hypotactic clause introduced by
a conjunction; moreover, the lexical content has been construed in two quite different
ways:
'time'
'engine1
fail'

(i) prepositional phrase


Head/Thing in nominal group
Classifier in nominal group in ofphrase qualifying time
Head/Thing in nominal group in of
-phrase qualifying time

(ii) dependent clause


hypotactic conjunction
Head/Thing in nominal group
functioning as Actor
Head/Event in verbal group
functioning as Process

If we take just the question of which elements function as Thing, the two are exactly
complementary: in (i) the Thing nouns are time mo failure, while in (ii) the only Thing

13.2 Review of modes of meaning metafunctional and metaphorical

537

noun is engine. This relationship is analogous to that of metaphor in its usual, lexical
sense; only here the transfer is not between words but between grammatical classes.
From a purely descriptive point of view, each version is metaphorical from the
standpoint of the other; there is no inherent priority accorded to either. Once we bring in
considerations of history, however, a clear priority emerges; and it is the same priority
whichever of the three diachronic dimensions we choose to invoke the phylogenetic
(history of the language), the ontogenetic (history of the individual) or the logogenetic
(history of the text). In all these three histories, version (ii), the clausal comes first. This
form of construal evolved first in the history of English; version (i) emerged only as the
result of a long process of later evolution. It comes first in the Ufe of a child; children
master version (i) only after a long (in terms of their young lives) process of becoming
literate and being educated And it comes first in the unfolding of a text; we are much
more likely to be told first that engines fail and only then to hear about a phenomenon of
engine failure. Once we take note of progression in time, then given a pair of such
expressions we can identify one of the two as the more metaphorical. The process is one
of movement away from what we referred to as a "congruent" form.
"Congruent" is of course a contingent term. What it is saying is that, at the present
moment in human history we can recognize forms of language which seem to represent a
common coding of experience: this is the configuration that we refened to as "process +
participant + circumstance" which is construed in grammars through some version of the
trichotomy of verb, noun and the rest If we relate this to English, it is the form of
English that is learnt as a mother tongue, in which phenomena are interpreted clausally,
in a kind of dynamic equilibrium of happenings and things. The prototypical thing is a
concrete object which can be related by similarity to certain other objects, such that taken
together they form a class, like engines. The prototypical happening is a change in the
environment that is perceptible to the senses, or a change in the senser's own
consciousness. A process is a happening involving one or two such objects, or one
object and a conscious being. When children move from their own constructed
protolanguage into the mother tongue, this provides a theory which they can use to give
a plausible construction to their own individual experience.
Under certain historical conditions, such a theory may come to be modified or
reconstructed. No doubt there have been various more or less catastrophic changes in

538

13. Language as a multifunctional system-&~process

earlier human history which have brought about relatively rapid changes in language
relatively, that is, to the gradual evolution of the system that has taken place all the time.
We have no means of knowing about these. But it seems likely that what we are here
calling grammatical metaphor represents one such partial reconstruction, in which, in the
context of science and technology, a rather different kind of "reality" is being construed.
It might be maintained that a pair of expressions such as in times of engine failure
and whenever an engine failed are simply synonymous, and do not imply any
reconstruction of experience. But there are two problems with this view, One is that of
history, referred to above. If neither had preceded the other, they could simply be free
alternatives (though language is seldom so extravagant with its resources as this would
imply!). But since one form of wording came first, it inevitably acquired a rich semantic
loading. Since nouns evolved as names of classes of things, anything which is
represented as a noun inevitably acquires the status of a thing, with the implication of a
concrete object as the prototype. Thus in engine failure, the grammar has construed a
thing called failure; and the nominal group then accommodates classes of failure (with
another noun as Classifier), such as crop failure, heart failure and engine failure. Thus
engine failure and engines fail are precisely not synonymous, because in engine failure
the happening fair has acquired an additional semantic feature as the name of a class of
things.
The second problem is that of sheer scale. If only odd, more or less random instances
of this kind of metaphor occurred, they could have little effect on the system as a whole.
But given the massive scale of this shift in the grammar, affecting as it does entire
registers of modern English, it cannot simply be dismissed as meaningless variation. As
we saw in Chapter 6, the metaphoric processes themselves are highly systematic;
moreover they occur in typical syndromes, so that it is not just one aspect of the
construction that is affected. Rather, the entire perspective is shifted sideways, so that
each element in the configuration is reconstructed as something else. When this pattern
comes to predominate throughout a large proportion of the discourse of adult life, it
amounts to a fairly major resemanticizing of experience.

13.2 Review of modes of meaning metafunctional and metaphorical

13.2.6.3 Origins of grammatical metaphor: the evolution

539

of scientific

discourse
Why did such a significant development take place? The most important single factor was
undoubtedly the evolution of science and technology. It is possible to trace the emergence
of this pattern of grammatical metaphor back to the origins of western science in ancient
Greece, and to follow its development step by step; each stage in the evolution of the
grammar realizes a stage in the evolution of a world view.
The philosopher-scientists

of the ancient Greek world, Thaes, Pythagoras,

Anaximander and their successors, inherited a language with a grammar of the kind
outlined above, in which experiential meanings were construed in clausal patterns as a
balanced interplay of happenings and things; nouns enjoyed no special privileged status.
In the course of their writings (and no doubt first of all in the course of their sayings,
only we have no access to these) they distilled this into a language of learning. We do not
know how much they reflected on this process; it is unlikely they engaged in any very
explicit language planning. What they did was to exploit the resources of everyday Greek,
its fundamental semogenic potential. In particular, they exploited two of its grammatical
powers: the power of forming new words, and the power of extending grammatical
structures (cf. Chatper 6, Section 6.3 above).
The first of these was their resource for creating technical terms. For systematic
scholarship it is necessary to technicalize some of the words that are used, and this
imposes two requirements: the words must be interpretable in an abstract sense, since
they need to refer not to outward appearances but to the properties and principles that lie
behind them; and they need to relate to one another in a regular and systematic way, so as
to form stable taxonomies. Ancient Greek was a language of seulement, in which the
potential for this kind of development lay predominantly in the nouns; and there existed
already a number of noun-forming suffixes by which words of other classes verbs aid
adjectives were transcategorized. Two of these were particularly potent: the 'active*
ending -OIC-MJ and the 'passive* ending -jia -ma. A number of derivatives with these
suffixes had been around for a long time: for example, from Ttpaaaco prsso 'do' the
nouns rcpai praxis 'a doing, action' andnpay^a pragma 'something done, a deed,
act'; and from noteco poio 'make1 the nouns TCOIIIOI poiesis 'making, creation,
production' and TcoiT^a poterna "something made, a product'. Using these and other

540

13. Language as a multifunctional system-&-process

nominalizing suffixes the Greek scientists created hundreds of new technical terms; and by
combining them with other derivational resources they developed extended series of
semantically related forms, for example the following set deriving ultimately from
los 'other':
(1) with -<nc -sis, aAAoiuxnc alloiosis 'change, alteration';
(2) oAAotpioc alltrios "belonging to another'; then, with -oic -sis,
allotrfosis 'estrangement, alienation';
(3) aXXaaoco alldsso 'make other, change'; TcapaXXaocco paralldsso 'make
alternate, transpose'; then with -oi -sis, napaAAai pardlkuas 'alternation,
alternating motion', and with

-(act

-ma, TtapaAAcrftia

parllagma

'interchange, variation'.
In this way they established the foundations of the lexical component of a technical
discourse, and the principles on which it could be indefinitely extended.
The second of the resources that was brought into play was a syntactic one, the
structure of the nominal group. The nominal group of ancient Greek was very like that of
modern English: it had a similar arrangement of elements around the Head noun, allowing
both prepositional phrases and clauses in modifying function (with some difference of
ordering), and included among its deictic elements one which was very close to the
English the. Thus any noun could accumulate qualifying clauses and phrases which were
explicitly signalled as defining, analogous to English the electrons in an atom, the angles
which make up a triangle. One context which demanded elaborate nominal group
structures of this kind was that of mathematics, as scholars conducted more and more
sophisticated measurements, for example in their attempts to understand planetary
motion. Here is the English translation of a nominal group from the work of Aristarchus
of Sanios, sometimes referred to as "the ancient Copernicus" because he was the first to
propose that the earth revolved around the sun:

the straight line subtending the portion intercepted within the


earth's shadow of the circumference of the circle in vdiich the
extremities of the diameter of the circle dividing the dark and
the bright portions in the moon move

13.2 Review of modes of meaning metarunctional and metaphorical

541

This has 32 words in the original Greek (fewer than the English because the equivalent of
of is the genitive case of the noun); note that it is only the Subject of the clause, which
continues ... is less than twice the diameter of the moon.1
The forms of scientific discourse developed by the Greeks were then taken over into
Latin. Although Latin differed in certain significant respects (for example, it had no
definite article, and it did not readily accept prepositional phrases as Qualifier in the
nominal group), it was close enough to Greek, both linguistically and culturally, for this
to present few problems. Most Greek derivational compounds could be cakjued directly
into Latin (e.g., 7tepi<|>epeia peripheria to circumferens ); Latin had its own stock of
nominalizing suffixes, like -atio(nem) and -mentum; and a reasonably similar potential
for expanding nominal groups. As Latin took over as the language of learning throughout
the greater portion of Europe, it had already developed an equivalent semogenic power. In
the medieval period Latin continued to serve; but by this time, although its morphology
was largely unchanged, it had taken on the semantic patterning of the vernacular European
languages. So when Latin itself was replaced by these languages, the transition was not
unlike that which had taken place earlier into Latin from Greek: first the Greeks developed
new meanings in Greek form, then these meanings were taken over into Latin forms,
then new meanings were developed in Latin, then these new meanings were taken over

The one minor piece of linguistic engineering that had to be undertaken was to ensure that the
prepositional phrases in such constructions were placed after the Head word rather than
before. It was also possible for a qualifying expression to be inserted between the Deictic
and the Head, equivalent to English the in an atom electrons; this has only a limited
potential for expansion, as can be shown by rewriting "The House that Jack Built" in this
format:
This is the that Jack built house.
This is the that lay in the that Jack built house matt.
This is the that ate the that lay in the that Jack built house malt rat.
This is the that killed the that ate the that lay in the that Jack built house malt rat cat.

...

With the qualifying phrase or clause at the end, the structure branches "to the right" (using
the linear metaphor derived from European orthographies) and there is much less restraint
on adding further elements. On the other hand, the fact that it is possible to put the
qualifying element before the Head, which in English it is not, helps to avoid some
ambiguity in the bracketing, as this example shows: here, both (which is) intercepted in the
shadow of the earth and of the circle along which ... are qualifying the word circumference,
and the Greek makes this clear with the intercepted-in-the-shadow-of-the-earth
circumference of the circle along which ... . The English translation cannot follow this
ordering.

542

13. Language as a multifunctional system-&-process

into the modem European languages, then new meanings were developed within these
languages. Thus there was a continuous evolution in the discourse of technology and
science: in each transition, one component of the system was preserved.
This discourse first appears in English in the work of Chaucer, for example his
"Treatise on the Astrolabe", written about 1391. Here we find the same linguistic
resources brought into play: nouns as technical terms, and extended nominal groups. The
former are partly technological (to do with the construction and operation of the
astrolabe), generally Anglo-Saxon or Norman French, like plate, ring, turet (eye, or
swivel), net (from rete 'net', i.e. grid), moder ('mother', body of the instrument); and
partly theoretical (from astronomy, mathematics or general methodology), mainly
borrowings from Latin like altitude, ecliptik, clymat (climatic zone), degree, equation,
conclusioun, evidence. The latter do not attain any spectacular length but involve the
expected mixture of clauses and prepositional phrases, as in the same number of altitude
on the west side of this line meridional as he was caught on the east side. This is dearly
the discourse of organized knowledge; but it is not sharply set off from the language of
everyday life.
It is with the "new learning" of the Renaissance that a distinct language of science
begins to emerge, with a vastly greater dependence on grammatical metaphor. The earlier
exercises in nominalization had been abstract but only minimally metaphorical; there is a
trace of grammatical metaphor in expressions like conclusion and the same number of
altitude, but no more than is found in the language of daily life. When we come to the
writing of Newton, however, we find formulations such as the following:
. . . by these two Experiments it appears, that in equal Incidences
there is a considerable inequality of Refractions.
. . . the cause of Reflexion is not the inpinging of Light on the
solid iirpervious parts of bodies, but .. .
. . . if the thickness of the body be much less than the Interval of
the Fits of easy Reflexion and Transmission of the Rays, the Body
loseth its reflecting power.

13.2 Review of modes of meaning metafunctional and metaphorical

543

These contain a great deal of grammatical metaphor, contrast mem with more congruent
forms of expression such as light is refracted unequally (even) when it falls at the same
angle; light is reflected not because it impinges on the solid, impervious parts of bodies',
if the body is much less thick than the interval between the points where the rays OK
easily reflected and (where they are easily) transmuted, the body is no longer able to
reflect (light). Why has the mode of expression changed along just these lines?
If we look at Newton's "Opticks", from which these are taken, we find that it consists
of three simultaneous discourses interspersed. In one of these phases, Newton describes
his experiments; in another he draws conclusions from the experiments; and in the third
he provides mathematical explanations. The language of the first phase is non-technical
and non-metaphorical; e.g. / looked through the Prism upon the hole, and turning the
Prism to and fro about its Axis, to make the Image of the hole ascend and descend ... /
stopp'd the Prism,... . The third is like the passage from Ahstarchus quoted above, e.g.
The Excesses of the Sines of Refraction of several sorts of Rays above their common
Sine of Incidence when the Refractions are made out of divers denser Mediums
immediately into one and the same rarer Medium, suppose of Air,... . The examples just
cited of grammatical metaphor are typical of the second phase. It is here that Newton is
proceeding by logical steps through a reasoned argument; and he frequently needs to
summarize the argument up to that point, or in anticipation of what is to come. A
typical sequence would be the following:
... vfaen Light goes out of Air through several

contiguous

refracting Mediums as through Water and Glass, . .. that Light . . .


continues ever after to be white. ... ... the permanent whiteness
argues, that . . .
The metaphorical nominalization permanent whiteness summarizes the earlier sequence of
inductions. There are similar contexts for expressions such as an inequality of
Refractions, the impinging of Light on ... Bodies, the thickness of the Body, reflecting
power in the passages cited above.
The nominalized forms inequality, impinging, thickness are not technical terms; or
rather, they are so to speak technicalized for the given instance, but they do not lose their
semantic status as property or event Why then are they reconstrued as nouns? The reason

544

13. Language as a multifunctional system-&-process

is to be found in the grammar of the textual metafunction. In order to function with the
requisite value in the message, which means either as Theme or as focus of information,
they cannot remain as complete clauses; they have to be "packaged" into single elements
of clause structure, and the only available constituent for this purpose is the nominal
group. Instead of being a process in its own right, light impinges on a body, the
phenomenon in question is construed as a participant, the impinging of light on a body.
It can then take on a clearly defined status in the grammatical construction of the
discourse.
What is beginning to emerge here is a grammar for experimental science: a way of
construing experiential meaning so that it can be organized textually into a form of
discourse for the advancement of learning. Again, this was not achieved by any conscious
act of language planning; ironically, while the leaders of the new scholarship paid a great
deal of attention to language, and recognized that their languages needed to be leequipped
if they were to meet these new demands, they saw the problem in lexical rather than
syntactic terms as a problem of making their lexical taxonomies at once explicit and
rigorous. The designed systems they came up with were of great interest for what they
revealed about the nature of language (as embodied some generations later in Roget's
English "Thesaurus"); but played no part in their scientific endeavours. The grammatical
innovations, on the other hand, which were not designed at all, proved invaluable.
When the processes and properties turn into nouns, the verbs do not disappear from
the scene. Scientific discourse is still written in clauses, and these clauses still have verbs
in them. Let us return briefly to the examples from Newton's "Opticks":
those

Colours

argue

diverging

and

separation

of

the

heterogeneous Rays from one another by means of their unequal


Refractions
the variety of Colours depends upon the Conposition of Light
the cause of Reflexion is not the inpinging of Light on the solid,
impervious parts of Bodies

We might suggest a more congruent form of the first and second examples here: colours
vary because light is composed [in this way]\ because those colours (appear] we know the

13.2 Review of modes of meaning metafunctional and metaphorical

545

heterogeneous rays diverge and separate from one another.... The verbs depend upon aid
argue both express a logical-semantic relation between the two ncounalized processes:
either an external cause, 'a happens; so x happens', or an internal cause, 'b happens, so we
know y happens1. This is another grammatical metaphor; the congruent form of
representation of a logical relation is a conjunction. The two types of metaphor work
together, to construe the two processes as one: 'happening a causes happening x\
'happening b proves happening y'. It is not the case, of course, that this type of
construction had never occurred in English before; it had. But it was rare; whereas from
the time of Newton onwards it gradually took over, becoming the most favoured clause
type of scientific language as indeed we find it today.
We have suggested that the immediate context for this change was a discursive one:
the evolution of a register of experimental science, in which certain forms of
argumentation were highly valued. This is usually interpreted simply as the emergence of
a particular genre, the scientific article; but that is only one side of the story no such
genre could have come into being without these changes in the grammar of the clause. At
the same time, they have other significant consequences. We have already pointed out the
fact that one effect of grammatical metaphor is to renda many of the semantic
relationships implicit: if the happening is construed as a clause, the semantic relations are
spelt out in the configuration of grammatical elements, whereas if it is construed as a
nominal group they are not, or only partially so (compare his energy balance approach to
strength and fracture with he investigated how strong [glass] was, and how it fractured,
using [the idea that] the energy [...] balanced out). On the whole, the greater the degree of
metaphor in the grammar, the more the reader needs to know in order to understand the
text.
But to say that the semantic relations have become less explicit is to imply that these
relations themselves have not changed. In one sense, this is true: we can "unpack" the
metaphor, and experts will generally agree on how to do it. But in another sense it is not
true. Scientific discourse began, as we saw, with the creation of technical taxonomies and
mathematical constructs; these were already modulating the semiotic construal of
experience, even if only at the margins, by creating a new realm of abstract things that
had not existed before. But the transformation brought about by the renaissance was a
more fundamental one; not only was this realm of abstract things greatly extended, but,
more significantly, phenomena hitherto construed as processes and properties were now

546

13. Language as a multifunctional system-&-process

transformed into things they were reconstrued, by grammatical metaphor. We have


illustrated this transformation from English; but it took place in all those languages that
took over the semiotic functions of medieval Latin.
We have shown that the motive for leconstming experience in this way was in the
first instance a textual one: in the grammars of these languages, when one is developing a
reasoned chain of logical argument such that complex phenomena have to be given a
clearly defined status in the organization of information (the clause as "message"), such
phenomena have to be constructed in nominal form. But there is no insulation between
one part of the grammar and another, and this inevitably has ideational effects. Any
semantic construct that appears as topical Theme has a function in transitivity; if it is
formed as a nominal group, it is potentially a participant in some process, and therefore
at some level it is an entity, a thing. If we say diamond is transformed into graphite, this
is a process involving two things, diamond and graphite; if we recoostrue this as the
transformation of diamond into graphite it has become one thing, transformation, with
diamond and graphite serving only circumstantially to qualify it as a thing of a certain
kind.
This reconstrual of experience is complex: what were first construed as happenings
have become things, with the original things now serving merely as their appendages;
but at the same time what were first construed as logical relationships between processes
have been reconstrued as processes in their own right. So the transformation of diamond
into graphite is caused by ... . It is also complex in another way: the original status
accorded to the phenomenon is not lost, but enters into a metaphorical nexus with the
new one. So transformation is still a process, as well as being a thing; is caused by, as
well as being a process, is still a logical relation between processes. But, as we have
seen, the overwhelmingly predominating effect of this reconstrual is a nominalizing one,
in which other phenomena are transformed into things. This is a major shift in ideational
terms, and plays a significant part in the historical semiotic.
In the world of classical physics, the flux of experience was held under control: reality
had to be prevented from wriggling, while it could be observed and experimented with.
The control over experience is partly a physical matter; but it is also in part semiotic, and
the semiotic control of experience is achieved by the nominalizing power of the grammar.
Since it is the grammar that has construed it in the first place, the grammar is able to

13.2 Review of modes of meaning metafunctional and metaphorical

547

transform it by reconstruing it in other terms. Grammatical metaphor played an important


role in shaping our humanist world.
But it shaped it in a way which soon came to be felt as decidedly inhuman. Already at
the end of the eighteenth century, within a hundred years of Newton's "Opticks", people
were reacting against the rigidity of the world of physics; what they could not accept were
the ideological constraints set up by scientific discourse, by a grammar which construed
all experience in terms of things. In our own twentieth century the scientists themselves
have become weary of it, finding that it prevents them from engaging with the
Kletenninacy and the flow that they now regard as fundamental let alone with the
concept of the universe as conscious and communicating, as something itself to be
interpreted as a semiotic system-&>process. Once we conceive of reality in semiotic
terms, it can no longer surprise us that language has the power to construe it, maintain
it, and transform it into something else.

13.3 Indeterminacy and probability in language


We have not here explicitly foregrounded the concept of indeterminacy in language. But
there is a reason for this. To foreground indetenninacy is to treat it as something special,
as a marked feature that stands out from, and hence distorts and destabilizes, the
phenomenon under scrutiny. Here however we take indetenninacy for granted, as a normal
and necessary feature of an evolved and functioning semiotic system. Rather than being
something that needs to be especially remarked on, it is something that should be built in
to our ways of representing and interpreting language: part of the background, rather than
the foreground, to our account of the construal of experience.
We should however make a few observations about the significance of indetenninacy
in relation to our present view of the meaning base. What does it mean to say that a
natural language is an indeterminate system? hi the most general terms, it suggests that
the generalized categories that constitute language as a system as "order", rather than
as randomness or "chaos" (let us say randomness rather than chaos, since chaos in its
technical reading is also a form of order) are typically not categorical: that is, they do
not display detenninate boundaries, fixed criteria of membership, or stable relationships
from one stratum to another. We could refer to them as "fuzzy", in the sense in which
this term is used in fuzzy logic, fuzzy computing, etc.; but we prefer to retain the term

548

13. Language as a multifunctional system-&-process

"indeterminate" for the phenomena themselves, since "fuzzy" is usually applied to the
theoretical modelling of the phenomena (it refers to meta-fuzz rather than fuzz).
This issue of order vs. randomness has surfaced at various times throughout the
history of linguistics, as analogy vs. anomaly, theory vs. usage, and the like; the
prevailing ideology has usually been on the side of order. Part of the reason for this is
simply that order is easier to describe: indeed the act of describing typically imposes
order, because it involves naming, classifying and taxonomizing. This is true of any form
of systematic knowledge. But pan of the reason lies in the nature of language itself, or at
least the kind of language that linguistics was designed to account for. The exclusive
object of study in linguistics was written language; and written language gives a much
greater appearance of order than does language in its spoken form.
We say "gives a greater appearance of order" because the actual picture is very much
more complex. The immediate appearance of order in written language the fact that it
is presented to us in neat blocks and rows upon a page (or the equivalent, in other forms
of technology), whereas speech is notorious for its hesitations, false starts, backtracking,
clearing of the throat and whatever is simply a consequence of the fact that we do not
display its history: we leave out the provisional attempts and early drafts, and "publish"
only the finished product When analogous measures are taken with spoken language
there is no significant difference between the two: speech is just as orderly as writing (cf.
Halliday, 1985/9),
It is when we come to look at the ideation base that significant variation does begin
to appear. We see this in its clearest form in grammatical metaphor, which is typically
associated with writing rather than speech. Grammatical metaphor objectifies our
experience, transforming its being and happening into things; in so doing, it privileges
order, since experience can now be categorized into classes and hierarchies of classes,
which are significantly more determinate than the processes and properties favoured by the
grammar in its congruent form. But even non-metaphorical forms of writing construe
with greater detenninacy. We may cite two very pervasive distinctions between spoken
and written discourse. On the one hand, writing construes the text into clear-cut
constituents, marked off by spacing and other forms of punctuation; in spoken language
there are no clear beginnings and endings in the expression (we cannot refer to pauses,
since they tend to occur at transition points before something that is less predictable;

13.3 Indeterminacy and probability in language

549

pauses seldom mark the text's grammatical boundaries). On the other hand, many
interpersonal and textual systems are realized in speech by intonation, and most
intonation contrasts are gradual rather than categorical. Thus both syntagmatically and
paradigmatically written language tends towards greater detenninacy; hence our received
model of language, in the mainstream grammatical tradition, emphasizes clear-cut
constituents and classes. Not that it has no tolerance at all for mixed and intermediate
categories; but it treats them as the exception, not the norm.
13.3.1 Types of indeterminacy
We have tried to make the point that the human condition is such that no singulary,
determinate construction of experience would enable us to survive. We have to be able to
see things in indeterminate ways: now this, now that, partly one thing, partly the other
the transitivity system is a paradigm example, and that lies at the core of the
experiential component of grammar. There are perhaps five basic types of indeterminacy
in the ideation base: ambiguities, blends, overlaps, neutralizations, and complementarities
although it should be recognized from the start that these categories are also somewhat
indeterminate in themselves. What follows is a brief characterization of each in turn:
(1) ambiguities ('either a or x'): one form of wording construes two distinct
meanings, each of which is exclusive of the other.
(2) blends Cboth b and yf): one form of wording construes two different
meanings, both of which are blended into a single whole.
(3) overlaps ('partly c, partly z1): two categories overlap so that certain members
display some features of each.
(4) neutralizations: in certain contexts the difference between two categories
disappears.
(5) complementarities: certain semantic features or domains are construed in
two contradictory ways.
Examples of these types of indeterminacy follow.

550

13. Language as a multifunctional system-<fc-process

(1) Ambiguity
(i) must:
You must be very careful! (when you do that)
obligation 'it is essential that you should be*
You must be very careless! (to have done that)
probability 1 am certain that you were*

Here the listener/ reader adopts either one interpretation or the other usually, of
course, without noticing that there is another possible meaning. In (i), the Attribute
suggests the choice (one does not usually instruct someone to be careless!); but cf. you
must be very sure of yourself ("before you do that1/ 'to have done that1). Compare also (ii),
where the ambiguity is one that is typical of identifying clauses.
(ii) home is vfriere your heart is
Token

Value 'if you live in a place, you love it'

Value Token 'if you love a place, it is home to you'


(2) Blend
they might win tomorrow
ability 'they may be able to'
x

probability 'it is possible they will'

Here, on the other hand, the meaning of the oblique modal might combines the two
senses of 'able* and 'possible', rather than requiring the listener to choose between them. If
the verbal group is 'past', however, this again becomes an ambiguity:
they might have won
ability 'they were capable of winning (but they didn't)'
probability 'it is possible that they won (we don't know)'

551

13.3 Indeterminacy and probability in language

(3) Overlap ('borderline case1)

(behavioural)

mental

listen
unmarked present
present-in-present

bear
unmarked present:
simple

I'm waiting
can't project

I'm listening
can't project

hear
can project

*/'m waiting that


they're away
does not impute
consciousness

*/'m listening that


they 're away
does impute
consciousness

/ hear that they're away

it (the bus) waited


is probed by do

it listened = the cat


?

PROCESS material
TYPE
wait
Example
unmarked present:
1.
present-in-present
:

2.

3.

4.

\
the best thing to do is \
wait
I

does impute
consciousness
ji heard = the cat
is not probed by do
:

*the best thing to do


is hear

Behavioural processes such as listen, watch share some features with material
processes ('present-in-present' as unmarked tense; no projection), other features with
mental processes (the Medium/ Behaver is a conscious being). They lie on the borderline
between 'doing' and 'sensing1 (so can be re-iterated as do in some contexts but not in all).
(4) Neutralization
finite

I get tired...

condition if I run

non-finite
+ preposition

non-finite
- preposition

Source from running

^^SFSSlSS.LlSS..
time when I run

running
manner with running

If the dependent clause in such an environment is finite, it selects one or other type of
enhancing relation: condition, cause or time. If the dependent clause is non-finite, the
distinction is partially or wholly neutralized.
(5) Complementarity
transitivity

(i): transitive perspective

dry = 'make dry' (cf. wipe)

13. Language as a multifunctional system-<fe-process

552

'11 dry/wipe
Process
'11 dry/ wipe

they
Actor
they
transitivity

- Goal
the dishes

What will they dry?'


'the children will dry the dishes'

(ii): ergative perspective

dry = 'become dry' (cf. fade)

+ Agent
the sun

Process
will dry/ fade

they
Medium
them

11 dry/fade
Process

- What will dry them?1


'the sun will dry the clothes'

As discussed in Chapter 4 above, the grammar adopts two complementary


perspectives on agency, the transitive and the ergative. Most processes are oriented
primarily within one perspective or the other; here we have illustrated the
complementarity with a verb that is at home in both.
Local indeterminacies of all these types are found in all regions of the content plane,
either within one stratum or at the interface between one stratum and another (including
of course puns, which are formed at the interface of content and expression). Some of
them involve very general categories, and hence resonate across wide stretches of semantic
space, like the transitive/ergative complementarity. We may call them "local", however,
in contrast to one global form of indeterminacy which is a feature of the entire system of
language, and probably of any evolved semiotic system, namely its probabilistic
character. Since we have not attempted to model this feature explicitly, it is perhaps
important at least to adumbrate it in a few paragraphs.
13.3.2 Probability
We have sometimes refened to the relative frequency of a particular feature of the
grammar. For instance, in our two examples of the meaning base as a resource in
language processing, certain patterns characteristically recurred: future tense in the weather
forecasts, imperative mood in the recipes. In each case this was a special feature
pertaining to the register in question: in weather forecasts, the future tense is especially
frequent relative to the other primary tenses.
To say this means that there is a general expectancy in English discourse that, again
relative to the other primary tenses, future will occur less frequently than it does here. In

13.3 Indeterminacy and probability in language

553

other words, there is some global expectation, in the grammar of English, about the
relative frequency of the different terms in the primary tense system, past, present and
future. Similarly there is some global expectation about the relative frequency of
imperative and indicative mood. Frequency in the text is to be interpreted, therefore, as
the manifestation of underlying probability in the system.
Until recently it has not been possible to establish global probabilities of this kind,
because to do so it is necessary to process very large numbers of instances, far more than
could hope to be achieved given the time it takes to analyse the grammar of a text. We
have been familiar with lexical probabilities for the greater pan of a century, because
words are relatively easy to count; and it has been recognized that the global probability
of a word's occurrence is an integral part of its functioning in the linguistic system (Zipf,
1935). Those who know a language are rather sensitive to lexical frequencies, and will
readily gloss a rare word by a more common one if called upon to explain something not
understood. It seems fairly obvious that the same probabilistic principle will hold for
terms in grammatical systems, although the suggestion has often met with considerable
resistance, sometimes in a quaintly self-contradictory form: on being told that he is more
likely to use this feature than that, the speaker protests that he is perfectly free to deviate
from the norm not recognizing that, in that case, there must be a norm for him to be
departing from.
We now have very large corpuses, of English and some other languages, readily
available and able to be accessed by sophisticated programs of software. It thus becomes
possible to establish grammatical frequencies to match the quantitative patterns already
established in the lexis. It is still not easy, because grammatical categories are typically
not realized in any consistent and unambiguous way; there has to be rather complex
pattern-matching to enable the program to recognize even apparently regular grammatical
categories like those of tense and mood in English. But some studies have been
undertaken; with regard to English primary tense, for example, in a corpus of about a
million and a quarter finite clauses, past and present were found to be very closely equal
whereas future was considerably the least common of the three (Halliday & James, 1993).
This now gives substance to our observation about the future tense in weather
forecasts. What we are saying is that there is a global pattern of probabilities in English,
including a probability profile of the tense system whereby the probability of future is

554

13. Language as a multifunctional system-&-process

(say) 0.1. Hie register of weather forecasting, however, sets up a local pattern in which
the probability of future is (say) 0.5. This changes the meaning of the system of primary
tense, because it reverses the marking and hence sets up a new relationship among the
different terms. It construes a realm of experience in which the future becomes the
familiar dimension of time, the point of reference by which both present and past are
defined.
This illustrates one very powerful feature of a probabilistic system of this kind: that it
acxommodates systematic functional variation. We have discussed the two text types
illustrated in Part HI, weather forecasting and recipes, as examples of variation in register
of the way in which the meaning selections in texts tend to vary systematically with their
contextual function their value in the social process. We pointed out that this
variation is seldom categorical: except in very closed registers, or "restricted languages",
no major options are likely to be totally excluded. What happens is that the probabilities
are reset. This may be a relatively minor skewing affecting a large number of semantic
features; but one or other system may stand out by being particularly clearly realigned, as
happens with tense and with mood, respectively, in our two examples. We can in fact
define register variation as the resetting of probabilities in the lexicogrammatical and
semantic systems, including those in the ideation base.
We observe these probabilities in the form of frequencies in the text; thus in order to
investigate register variation, we need to have very large corpuses, with texts of different
types covering a wide range of functional varieties. (Note that there can be no such thing
as a structured sample of such variation, any more than there can be of dialectal variation;
you cannot quantify the functional spread of a language.) This makes it possible to
establish both the global and the local probabilities. The objection has sometimes been
made that global probabilities in a language are meaningless; that since every text is in
some register or other, there can be no such thing as a quantitative profile of the language
system as a whole. But the two perspectives are merely different standpoints of the
observer; we can vary depth of focus just as we will, subject only to the constraint of the
quantity of text data available it makes no difference to the validity of the results
obtained. (It may affect their usefulness, of course; but that depends on the purpose of any
particular investigation. For some purposes it is precisely the global probabilities that we
need to know. Again it is interesting that no such objection is raised to the concept of
global probabilities in the lexis.)

133 Indeterminacy and probability in language

555

The significance of global probabilities in the grammar emerges in various ways,


both historical and synchronie. Historical change in language is typically a quantitative
process, in which probabilities in systems at every level are gradually nudged in one
direction or another, now and again becoming categorical so that some systemic upheaval
takes place. Each instantiation of a tense form, say, whenever someone is speaking or
writing in English, minutely perturbs the probabilities of the system because what we
call "system" and "instance" ate one and the same phenomenon, being observed from
different depths in time. There are, of course, more catastrophic types of change:
languages become creolized, creoiized systems in turn become decreolized, or a language
ceases to be spoken altogether. At the "instance" end, a single highly-valued instance may
exert a disproportionate effect quotations from the Bible and from Shakespeare are
familiar triggers of this "Hamlet factor" in English (in media discourse today almost
every change is a sea change, which goes into our folk taxonomy of types of change).
But such qualitative effects take place against a background of microscopic quantitative
pressures, the son of nanosemiotic processes by which a language is ongoingly
restructured as potential out of the innumerable instantial encounters of daily life
the "sheer weight of numbers", as we sometimes call it. And in the ontogenetic
dimension of history, the growth and development of language in a human child, an
analogous dialectic can be observed: highly valued instances of text (rhymes, favourite
stories and the like) interact with the quantitative pressures of the talk going on around
(the child's access to the global probabilities note that children do not begin learning
the grammar by sorting out functional variation!) to yield a meaning potential that is a
reasonable copy of the probabilistic system shared among the community.
Synchronically (that is, viewed synoptically in this way as a meaning potential) a
language is, as we have said, a probabilistic system: if we say that, in the grammar, there
is a system of primary tenses past/ present/ future, we assume the rider 'with a certain
probability attached to them*. But we do not, of course, speak or write with one
grammatical system at a time. Systems intersect with each other simultaneously (we
choose tense along with voice, polarity, mood, transitivity and so on), and they follow
each other in linear succession (we choose tense in clause 1, again in clause 2, again in
clause 3 and so on). Each instance has its environment, both of previous instances, and of
simultaneous instances of systems with their own sets of probabilities. We shall not
attempt to discuss this issue here; except to refer briefly to what is one important aspect

556

13. Language as a multifunctional system-&-process

of indeterminacy, namely partial association between systems. We model the


grammar as if each of these choices was independent that the choice of tense, say, is not
affected by the simultaneous choice of mood. This may, in fact, be so; but when choice
is made in two systems simultaneously, such that each serves as environment for the
other, there is often a conditioning effect on the probabilities. This may be an indication
of a change in progress, or it may be a stable feature of the overall system. Nesbitt &
Plum (1988) have shown, on the basis of a quantitative study, that there is mutual
conditioning in English between the systems of projection (locution/ idea) and taxis
(parataxis/ hypotaxis, Le. (in this environment) quoting/ reporting), such that there is a
positive association (1) of quoting and locution, (2) of repotting and idea. In other words,
the favoured combinations are those exemplified in She said "I understand now", She
thought she understood then. This means that, in construing people's sayings and
sensing, we do allow for projecting someone's speech into our own deictic space (She
said she understood then ), and for projecting someone's thought as if we had heard it
expressed (She thought, "I understand now"); but these are marked options, clearly
contrasting with our usual representation of these familiar aspects of experience. This in
turn relates to the different interpretations of thought in these two environments:
unmarked 'held the opinion' (which may be disputed by the speaker, But she didn't)',
marked 'said to herself.
Such partial associations, like probabilities as a whole (and in fact like all forms of
indeterminacy), are difficult to represent in an explicit way in the grammatics. This has
long been recognized in computational work in text generation and parsing2. The fact that
these indetenninacies are difficult to handle computationally is another manifestation of
something we know already: that it is hard for humans to construct a machine which will
perform human-like semiotic tasks. (We should not fall into the familiar trap of saying
"because it cannot be accommodated by our present logic, therefore it does not occur"!)
The relevant question here is, what is the place of semantic indeterminacy in our overall
construal of experience? We have not taken up that issue in this book; but we have tried
to embody a general awareness of indeterminacy in our overall interpretative frame.

Nigel, the systemic-functional generation grammar of the Penman system, had probabilities
attached to terms in systems right from the start in 1980. They were used in random
generation when the grammar was tested. For probabilistic analysis, see e.g. Sampson
(1987, 1992.)

13.3 Indeterminacy and probability in language

557

13.3.3 The significance of indeterminacy


How would we summarize the significance of indeterminacy, from the point of view of
the semantic construal of experience? It lies first and foremost, perhaps, in the general
principle that is being proclaimed, if indeterminacy is a typical and unremarkable feature
of the grammar: that 'this is the way things are1. Our "reality" is inherently messy; it
would be hard to construe experience, in a way that was beneficial to survival, with a
semiotic system whose typical categories were well-defined, clearly bounded, and cxdered
by certainty rather than probability. This is the problem with designed systems, including
semiotic ones: as a rule, they fail to provide adequately for mess.
Secondly, systems have varying probability profiles, so that (in terms of information
theory) they carry differential loads of information: the skewer the probabilities of the
terms in a system, the greater the redundancy that it carries hence the less we need to
attend to its unmarked state. For example: it has been found that, in an English clause,
positive is about ten times as frequent as negative (Halliday & James, 1993). What this
means is that we build in to our sense of a figure the presumption that something is or
something happens, rather than that something is not or does not happen; extra work has
to be done if a process is being construed as negative. The same applies to future tense,
as we saw: extra grammatical energy is required to assign a figure to the future.
But, thirdly, within the overall construction of experience, the diversity of spheres of
social action is realized by variation in the line-up of semantic features that is, by
variation in register. The probabilities are reset; and in some cases one or two "critical
systems" are strongly affected in this way, such that the local norm skews the system, or
perhaps even reverses the skewing set up by the global norm. It is here that we find
future taking over as the unmarked primary tense in weather forecasting. As we said
above, we define register variation in just these terms, as the ongoing resetting of
probabilities in the lexicogrammar, which then functions to construe the ongoing
variation at the level of the social process,
Moving away from consideration of probabilities, there are also, fourthly, regions in
the ideation base whose construal typically involves one or other of the particular kinds
of indeterminacy that we referred to earlier in this section. Let us try to offer some very
brief illustrations of this, building on the few examples that were given above. The first

558

13. Language as a multifunctional system-&-process

is concerned with ambiguities and blends. The illustration comes from the interpersonal
metafunction; here the meaning that is construed in the ideation base is being assessed
(judged probable, desirable &c.) at the same time as it is intersubjectively enacted as
proposition (statement, question) or proposal (offer, command).
Modality, the speaker's angle on what is or what should be, is notoriously fluid and
shifting in its categories, probably in every language. In English, there is a fairly clearly
defined semantic region construed at the intersection of a number of grammatical systems,
including (1) type: probability/ usuality // obligation/ readiness: inclination/ ability; (2)
value: median // high/ low; (3) orientation: objective/ subjective; (4) immediacy:
immediate (neutral) / remote (oblique); (5) polarity: positive/ negative; and one or two
others (for a summary, see Halliday, 1985). These are realized synthetically in various
ways, one of which is by the modal finite operators can, could, may, might, will ,
would, should, must, oughl-to (and one or two other fringe members). There is a great
deal of indeterminacy throughout the region; but it is of more than one kind. At one
"corner", if we combine value with neutral (immediate), then the resulting wordings aie
ambiguous, as to the type of modality expressed: thus must has three clearly distinct
meanings, (a) as probability (e.g. tha must be Mary 'certainly that is Mary'), (b) as
obligation (e.g. you must wear a helmet 'it is essential that you wear a helmet'), (c) as
readiness: inclination (e.g. if y ou must make all that noise 'if you insist on making all
that noise1). That these are truly ambiguous can be gathered from an example such as she
must complain, which has to be interpreted in one or another of these diffrait
meanings the context will of course usually make it clear which; for example:
following
I wonder why they take all
that trouble just for her?'
1 don't think they'll let her
return it!'
'Whatever happens she's
never satisfied'

exempte
She must complain.
She must complain.
She must complain.

systepc fffvrtipft
probability: 'the reason is
certainly that she complains'
obligation: 'it is essential
that she should complain'
readiness: 'she
insists on
complaining1

But at the opposite corner, so to speak, if we combine low value with oblique (remote),
the result is typically blending rather than ambiguity: e.g. it couldn't hurt you to
apologize is a blend of 'it would not be able to hurt you' (readiness: ability), 'it is
unlikely that it would hurt you' (probability) and even perhaps 'it would not be allowed to
hurt you' (obligation). In other words, looking at it from the point of view of blending,

13.3 Indeterminacy and probability in language

559

in the region of 'what I think1/ 'what is wanted', it is easiest to blend the low values 'what
I can conceive of with 'what is permitted', especially in 'remote' conditions (hypothetical,
projected or tentative) [realized as could, might]\ and hardest to blend the high values
'what I am convinced of with 'what is required', especially when 'immediate' [realized as
must]. This is diagrammed in Figure 13-3. This shows that, in modality, a very complex
region where the metafunctions themselves overlap ('what I think is' and 'what ought to
be' blending in 'what I think ought to be'), the indeterminacy even extends to
indeterminacy between the different types of indeterminacy!
The next illustration is concerned with overlaps and complementarities. For this we
can turn to the system of transitivity. As we saw, the grammar distinguishes a number of
types of process, material, mental, verbal and relational; the distinctions are made by a
cluster of syntactic variables the valency of associated participant roles, the class of
entity that takes on each role, the potential for combining with other figures, the
associated tense systems and the like. But since these variables "draw the line" at different
places, there are areas of overlap, with mixed categories that share some characteristics
with one group and some with another. We gave the example of behavioural processes;
these are a mixed category, formed by the overlap of the material, on the one side, and
the mental or verbal on the other. Behaving is construed as a type of figure that (like the
mental) typically has a conscious participant as the central role, and does not extend
beyond this to a second participant; but, on the other hand, it does not project, and it has
a time frame like that of the material. Thus behavioural processes lie squarely athwan a
fuzzy borderline. In those figures where there is a second direct participant, some form of
agency runs through all the different types of process; but agency is such a complex
aspect of human experience that the grammar does not delineate it by a single stroke, but
construes it by means of a fundamental complementarity, that between the transitive
and the ergative perspectives. Thus figures involving two direct participants, such as
Actor + Goal in the material, are aligned along two different axes: the transitive one,
based on the potential extension of force (mechanical energy) from a doer to another
entity; and the ergative one, based on the potential introduction of agency (causal energy)
from another entity as external source. Thus the earthquake shook the house is construed
both as 'earthquake + shake' plus optional Goal 'house1, and as 'house + shake' plus
optional Agent 'earthquake1. As always in cases of complementarity, certain parts of the

560

13. Language as a multifunctional system-&-process

region are more strongly aligned to one perspective, other parts to the other, but the total
picture requires the confrontation of the two.

Fig. 13-3: Ambiguities and blends in the different corners of modality


Thirdly, neutralization. When two figures are linked into a sequence, by some logicalsemantic relation, there is a rather wide range of possible semantic relations between the
two: the relations of time and of cause and condition are particularly elaborated in this
respect, but there are others besides the manner, the matter, and so on. The
distinctions among these relationships, however, may be to a greater or lesser degree
neutralized, where one clause is construed as dependent on the other; this happens as
the dependent clause moves from finite to non-finite status. For example, in they get
caught taking bribes the distinction that would be made in the agnate finite clause,
among, say they get caught if they take bribes, they get caught when they take bribes,
they get caught because they take bribes, is simply neutralized it is not a blend of all
three, nor is there any ambiguity involved. An intermediate degree of specificity, with
partial neutralization, can be seen in the non-finite clause with accompanying

13.3 Indeterminacy and probability in language

561

preposition, as in they get caught for taking bribes. What happens here is that the fact
that there is

a connection between the two figures is unequivocally construed by the

dependency; but the nature of this connection what kind of

logical relationship is

being set up does not enter the picture.


There are of course many different contexts for all these indetenninacies, in different
regions of the total semantic space. Certain types of ambiguity appear to be not so much
artefacts of the realization (not just grammatical puns, so to speak) but rather another
kind of complementarity, where the grammar is as it were "having things both ways"
both interpretations have to be accepted at one and the same time. This is sometimes the
case with Token + Value structures, in figures of being. These clauses are always
ambiguous, if the verb is be, since this verb does not mark the passive; yet some depend
on being interpreted both ways particularly, perhaps, some proverbial sayings, Thus,
one man's meat is another man's poison is both Token A Value 'what one person likes
may displease another' and Value A Token 'what one person dislikes may please another';
contrast what's sauce for the goose is sauce for the gander, which can be interpreted only
as Token

Value. (That these are true potential ambiguities can be seen from the

following example of popular economics, given in answer to the query about personal
wealth:
Total net personal wealth is between tvro and a half

and three

times Gross Domestic Product.

This might mean either 'personal wealth accounts for 2.5 - 3 times GDP' (Token A Value:
that's how it's evaluated), or 'personal wealth is expressed as 2.5 - 3 times GDP' (Value A
Token: that's how it's calculated). See the discussion in Halliday (1985: Ch. 5).)
It seems likely that all these different kinds of indeterminacy are what make it
possible for the grammar to offer a plausible constmal of experience one that is rich
enough, yet fluid enough, for human beings to live with. We should stress once again
that the examples cited here are features of the ideation base of one particular language,
namely English. No other language will be identical. Indeed the distribution of
indeterminacies is likely to be precisely one of the features in which languages differ
most, and even perhaps varieties within one and the same language. But every language

562

13. Language as a multifunctional system-&-process

depends on indeterminacy as a resource for meaning even if our grammatics is not yet
very clever at teasing it out

13.4 Polysystemicness
One feature that emerges from the preceding discussion (and from our treatment
throughout) is the polysystemic nature of the ideation base. The ideauonal meaning
potential embodies not one single semantic system but rather several such systems
coexisting; in Firth's terms, it is a "system of systems11 in two distinct but related
ways.
(i) There are internal complementarities:
in the congruent mode of construing experience, there is a metafunctional
complementarity: the ideauonal potential offers two complementary modes
for construing experience the highly generalized logical

mode, with

projection & expansion as the dominant semantic motifs, and the more
particularized experiential

mode, with its typologies of processes, things,

qualities, and circumstances;


within this congruent mode, there is a fractal complementarity: the highly
generalized semantic types of projection & expansion aie manifested in
complementary domains those of sequences, figures, and participants; so
that, for example, some phenomenon of experience construed as having
temporal expansion might appear either as a sequence or as a configuration;
still in its congruent mode, there aie systemic

complementarities: the

ideauonal potential offers systemic complementarities such as the ergative aid


transitive models of participation in processes, and the mass and count
(singular/ plural) models of quantity;
beyond its congruent mode, there is metaphorical

complementarity: the

ideauonal model offers a complementarity between the congruent mode itself


and the metaphorical mode, making it possible to take some phenomenon as
already construed and then reconstrue it as if it was a phenomenon of a different
kind

13.4 Polysystemicness

563

Such complementarities constitute one form of indeterminacy of the system one that
allows it to be "polysystemic" in the particular sense of embodying more than one way
of construing experience.
(ii) In addition, the ideation base is polysystemic in another sense: registerial
variation. We have seen that such variation can be construed in terms of the
probabilistic nature of the linguistic system, as variation in the probabilities associated
with terms in systems. Seen in this light, a register is a particular probabilistic setting of
the system; and the move from one register to another is a re-setting of these
probabilities. What is globally the 'same* ideational semantic system can thus appear as a
collection of diffrait systems, as one moves along the cline of instantiation from
potential to instance (see Figures 8-1 and 8-3 in Chapter 8, Section 8.1). As we noted
above, the effect is quantitative; but it is also qualitative, in the sense that it provides
different perspectives on experience within the same system. We referred to the change in
the perspective on time that takes place when one moves into the realm of weather
forecasting. In the limiting case, this effect may be qualitative, in that certain options
may simply be absent in a system of a given register, having the probability of zero. Our
accounts of the registers of recipes and weather forecasts contain illustrations of this: we
noted, for example, the highly restricted construal of agency in the world of the kitchen.
The overall ideation base thus comprises many different registerial variants
register-specific systems that we called domain models. Now, just as the overall ideation
base is a theory of our total experience of the world around us and inside us the theory
that is shared by the culture as a whole, so also the different registerial variants constitute
different 'subtheories' of our experience. These 'subtheories' may complement one another
by simply being concerned with different domains of experience the culinary and
meteorological domains in our examples in Part III. This complementarity is purely
additive, although for society as a whole it constitutes the semiotic aspect of the division
of labour, whereby different people construe different facets of the overall cultural
experience. But such subtheories may also be concerned with more or less the same
domain, bringing alternative theoretical perspectives to the construal of experience that is
shared. Halliday (1971) shows how this is achieved in William Golding's novel The
Inheritors, by means of alternative deployments of the resources of transitivity as Golding
presents the world view of two groups of early humans; these different perspectives on
the shared experience are constituted as variants of the same overall transitivity system.

564

13, Language as a multifunctional system-&-process

From an educational point of view, the most fundamental complementarity is the move
from the registers of everyday life to the registers of education: this is a move from folk
or commonsense models to the "uncommonsense" models of systematic and technical
knowledge. To say that the ideational system is polysystemic means that it can support
these different theoretical angles on experience: semantic variation of all kinds is the
manifestation of the different theoretical interpretations that language places on
experience. In the next chapter we shall try to illustrate this point, by a consideration of
the realm of human consciousness as this is construed in daily Ufe and in (mainstream)
cognitive science.

14. Construing ideational models: consciousness in


daily life and in cognitive science
As a resource for making sense of our experience, the ideation base enables us to construe
a range of different theories, commonsense as well as scientific. There is a cline between
folk, or commonsense, theories and scientific, or uncommonsense, ones; and at any point
along the cline alternative theories may be in competition. The ideation base, by dint of
being polysystemic, accommodates variation along this cline, not only from folk to
scientific but also across alternatives: it embodies both congruent and metaphorical
construals of experience, and it provides elasticity within the overall construction space.
We shall examine this variation with particular refeaence to those phenomena to which
we are offering the ideation base as a conceptual alternative: the mind, knowledge,
cognition as the concerns of cognitive science. Drawing on our discussion in Part II,
we shall suggest that the domain of cognitive science is construed ideationally within the
folk model; but that this model is extended metaphorically in cognitive science itself, and
this extension in fact invites the interpretation of knowledge as meaning. First, however,
let us make a few comments on the cline from folk to scientific and the variation in the
degree of conscious awareness that this involves.

14.1 Models embodied in the ideation base


Researchers concerned with culture have distinguished among different kinds of cultural
models folk models, expert models, scientific models (see e.g. Holland & Quinn,
1987). Cognitive science operates with a scientific model of the individual mind; but, we
shall suggest, it is one that is based fairly uncritically on certain aspects of a folk model,
in particular in its selection of, and perspective on, its own domain of enquiry. Our
discussion will relate to the models we have identified within the ideation base the
general potential model, and the more specific domain models.
14.1.1 Models at different levels of abstraction
Any given model of experience exists at different orders of abstraction; see Figure 14-1. It
is a configuration of higher-level meanings within the context of culture; at the same
time, it is also construed semantical!?, in the ideation base. The relationship
between these two orders of abstraction, contextual and semantic, is a stratal one; hence a
model is a cultural construct that is construed in language (together with other, languagedependent semiotic systems such as expository drawings and diagrams).

566

14. Consciousness in daily life and in cognitive science

Fig. 14-1: The different orders of abstraction in a model and the different degrees of
instantiation
At both these strata, models are also located along the cline of instantiation (see
Chapter 8, Section 8.1 and Figure 8-3), running between the potential the overall
resources for making meaning, within the context of culture, and the instance
instantial 'texts' constituted of meanings that have been selected from this potential,
within particular contexts of situation.1 The potential end of the cline of instantiation
embodies all the contextual-semantic models that the culture embraces. The everyday folk
models are likely to be embraced unconsciously by everybody in the culture, because
they are everyday models, instantiated in casual conversation, and because they aie

Text* or 'discourse* here has to be understood not just at the lowest level of abstraction in
language as a realization in speech or writing, but primarily as configurations of linguistic
and higher-level cultural meanings as generally throughout this book (see Halliday,
1978a; Halliday & Hasan, 1985; Martin, 1992).

14.1 Models embodied in the ideation base

567

construed as congruent in the cryptogrammar. The general model of the phenomena of our
experience, including those of our own consciousness seeing, thinking, wanting, and
feeling , is of this highly generalized kind.
In contrast, scientific models are much more contextually constrained: they aie
developed, maintained, changed and transmitted within those situation types that we
associate with scientific language, We can characterize them in terms of field, tenor and
mode as follows (from Halliday & Martin, 1993: 54):
field: (i) extending, transmitting or exploring knowledge (ii) in the physical,
biological or social sciences;
tenor: addressed to specialists, to learners or to laymen, from within the same
group (e.g. specialist to specialist) or across groups (e.g. lecturer to students);
and
mode: phonic or graphic channel, most congruent (e.g., formal 'written language*
with graphic channel) or less so (e.g., formal with phonic channel), and with
variation in rhetorical function expository, hortatory, polemic, imaginative
and so on.
These ranges of field, tenor and mode values define a great variety of situation types
within institutions of higher education, of research and of technological development.
However, these situation types are quite constrained relative to the context of culture as a
whole: only certain members of the culture participate in these situation types and engage
with the scientific models that are developed, maintained, changed and transmitted within
them. In this respect, scientific models are clearly subcultural models: contextually they
are located somewhere between the potential and the instance. If we focus on particular
scientific models, such as those of the mind in cognitive science, we will find that they
are even more contextually constrained. Figure 14-1 shows the dispersal of contextualsemantic models along the cline of instantiation. As the figure indicates, an inherent
property of instantiation is variation; and scientific models (like other subcultural models)
vary one in relation to another. Sometimes they are complementary, sometimes they
conflict.

568

14. Consciousness in daily life and in cognitive science

14.1.2 Models at different levels of awareness


There is thus a range of variation from our everyday folk models to scientific models,
with expert models somewhere coming in between (Linde, 1987). Such models vary
considerably in the degree to which we are consciously aware of them as models (cf.
Whorf s, 1956, notion of critical consciousness). We are more aware of models that 'stand
out* as belonging a particular subculture than of those that are pan of our everyday
repertoire; and we are more aware of scientific models than of folk models.
Whatever the scope and sophistication of a model, however, we are likely to be more
aware of a model as a cultural construct than as a linguistic construct, since language is
typically further from our conscious attention. Certain aspects of language are closer to
conscious awareness than others (see Halliday, 1987): these are the more exposed parts of
language, which are also the parts that tend to get studied first. In Western thinking about
language, the most exposed aspect of language has been the "word": to talk is to "put
things into words". The folk notion of the "word" is really a conflation of two different
abstractions, one lexical and one grammatical.
(i) Vocabulary (lexis): the word as lexical item, or "lexeme". This is construed as
an isolate, a 'thing' that can be counted and sorted in (alphabetical) order. People "look
for" words, they "put thoughts into" them, "put them into" or "take them out of another's
mouth", and nowadays they keep collections of words on their shelves or in their
computers in the form of dictionaries. Specialist knowledge is thought of as a matter of
terminology. The taxonomic organization of vocabulary is less exposed: it is made
explicit in Roget's Thesaurus, but is only implicit in a standard dictionary. Lexical
taxonomy was the first area of language to be systematically studied by anthropologists,
when they began to explore cultural knowledge as it is embodied in folk taxonomies of
plants, animals, diseases and the like.
(ii) Grammar: the word as one of the ranks in the grammatical system. This is, not
surprisingly, where Western linguistic theory as we know it today began in classical
times, with the study of words varying in form according to their case, number, aspect,
person etc.. Word-based systems such as these do provide a way in to studying
grammatical semantics: but the meanings they construe are always more complex than

14.1 Models embodied in the ideation base

569

the categories that appear as formal variants, and grammarians bave had to become aware
of covert patterns.
It is only in more recent times that the more covert areas of grammar have been
systematically studied those that, following Whorf, we have referred to as
"cryptogrammar" (see Chapter 2, Section 2.8 above). Whorf (1956) distinguished between
overt and covert categories and pointed out that covert categories were often also
"cryptotypes" categories whose meanings were complex and difficult to access. Many
aspects of clause grammar, and of the grammar of clause complexes, are essentially
cryptotypic. It is the analysis of some of these more covert features embodied in the
everyday grammar, in particular the theory of mental processes, that throws light on the
domain of cognitive science.
In addition to the different degrees of their awareness of different grammatical units,
such as words and clauses, people are also not equally aware of the different kinds of
functions in which the resources of language are organized. In particular, in
constructing and reasoning about more conscious models, people are readily aware of
those linguistic resources whose function it is to interpret and represent experience, those
of the ideational metafunction; but they are much less aware of those of the other two
metafunctions, the interpersonal and the textual no doubt because these do not
embody representations of experience but reflect our engagement with the world in
different ways. However, although they tend to be overlooked when one comes to build a
'scientific' model of language and the mind, these other metafunctions are no less
important than the ideational. We will take up this issue again below with respect to the
balance between ideational and interpersonal meaning (for related discussion of the textual
metaf unction, see Chapter 9 above).
14.1.3 Emergence of scientific registers

In the early scientific period in the West, new registers evolved as part of the ongoing
reconstruction of experience in the form of systematic knowledge and experimental
science (cf. Chapter 13, Section 13.2.63). Perhaps the earliest to evolve in these new
contexts were those registers associated with the exploration, storage and dissemination of
new knowledge about plants and herbs in the 16th and 17th centuries. These new contexts
put pressure on the linguistic resources, and the meaning-creating power of these

570

14. Consciousness in daily life and in cognitive science

resources correspondingly increased. We can hypothesize a gradual evolution from the


registers of ordinary language, with their folk models of the world, including folk
taxonomies of plants and herbs, to more specialized scientific ones. It was at this point,
as people became aware of the rapid development of new knowledge and the need for
processing and storing it, that conscious design of language began, with nomenclatures
and taxonomies being explicitly discussed and fixed. At the same time but in this case
without conscious design new ways of meaning evolved in the construction of figures
and sequences.
In the evolution of a more scientific approach to information about plants and herbs,
it is possible to trace not only a move away from folk taxonomy towards scientific
taxonomy, but also a move towards an organization of discourse that represents
information succinctly and uniformly, culminating in Linnaeus' descriptions; see
Slaughter (1986) for discussion. The following extracts spanning a period of about 225
years are quoted from Slaughter's account:
Banckes' (1525) Herbal:
Asterion or Lunary groweth among stones and in high places. Ihis
herb showeth by night. This herb hath yellow flowers whole and
round as a cockbell, or else like to foxgloves. The leaves of this
herb be round and blue, and they have the mark of the moon in the
middle, as it were three-leaved grass, but the leaves thereof be
more [larger] and they be round as a penny; and the stalk of this
herb is red, and this herb seemeth as it were musk, and the juice
thereof is yellow. And this herb groweth in the new moon without
leaves, and every day springeth a new leaf to the end of fifteen
days, and after fifteen days it loseth every day a leaf as the
moon waneth, and it springeth and waneth as doth the moon; and
where it groweth, there groweth great quantity. The virtue of this
herb is thus. They that eat of the berries or of the herb in
waning of the moon, vfoen it is in signo virginis, if he have the
falling evil, he shall be whole thereof; or if he bear this about
his neck he shall be helped without doubt.

14,1 Models embodied in the ideation base

571

Gerard (1597), The Herbal or General Historie of Plants:


The first of the Daffodils is that with the purple crowne or
circle, having small narrow leaves, thicke fat, and full of slimie
juice; among the which riseth up a naked stalke smooth and hollow,
of a foot high, bearing at the top a faire milke white floure
growing forth of a hood or thin filme such as the flours of onions
are wrapped in: in the midst of which floure is a round circle or
small coronet of a yellowish colour, purfled or bordered about the
edge of the said ring or circle with a pleasant purple colour;
which being past, there followeth a thicke know or button, wherein
is contained blacke round seed. The root is white, bulbous or
onion-fashion.

Bauhin (1620), Prodomus theatri Botanici:


From a short tapering root, by no means fibrous, spring several
stalks about 18 inches long: they straggle over the ground, and
are cylindrical in shape and furrowed, becoming gradually white
near the root with a slight coating of down, and spreading out
into little sprays. The plant has but few leaves, similar to those
of Beta nigra., except that they are smaller, and supplied with
long petoiles. The flowers are small, and of greenish yellow. Ihe
fruits one can see growing in large numbers close by the root, and
from that point they spread along the stalk, at almost every leaf,
hey are rough and tubercled and separate into three reflexed
points. In their cavity, one grain of the shape of an Adonis seed
is contained; it is slightly rounded and ends in a point, and is
covered with a double layer of reddish membrane, the inner one
eclosing a white, farinaceous core.

5 72

14. Consciousness in daily life and in cognitive science

Carolus Linnaeus (1754), Genera plantantm:


Urtica
Masculi flores
CAL.

Perianthium tetraphyllum:

foliolis

subrotundis,

concavis,

obtus is,
COR. Ptala nulla.
Nectarium

in

centro

f loris,

urceolatum,

integrum,

infern

angustias, minimum.
STAM. Fileonenta quatuor/ subulata, longitudine calycis, patentia,
intra singulum folium calycinum singula. Antherae biloculares.
Here we can see the emergence of a specialist form of discourse, as commonsense
knowledge about plants is gradually turned into uncommonsense, scientific knowledge.
Linnaeus' specification could be turned into a designed frame-based system of the kind
used today in frame-based inheritance networks (cf. Chapter 1, Section 1.9.5).
As designed semiotic systems emerge, both the registers of everyday language and the
original specialist registers continue to exist and to develop; folk models of the world
will co-exist alongside the scientific ones (see e.g. Halliday & Martin, 1993: Ch. 8). A
certain degree of intertranslatablity is likely to be maintained linguistic renderings of
logical or mathematical formulas, for instance; and this constitutes one of the contexts in
which ordinary language is brought into explicit contact with more scientific varieties.
There will always be some complementarity of function between the more designed
varieties and those that are naturally evolving. They may be allocated to different spheres
of activity: for example, the language of bud-watchers vs. the language of ornithologists.
Butin other cases the two are closely integrated as submotifs within a single sphere: for
example, the use of both natural language and mathematical expressions side by side in
the learning and practice of mathematics. This kind of interpntration still entails a
semiotic complementarity, but of a very sensitive kind, requiring a delicate interpretation
of the context in order to bring it out.
14.1.4 The shift from folk models to scientific ones
The shift from folk models to scientific ones thus takes place over a long period of time;
it typically involves several factors: see Figure 14-2.

14.1 Models embodied in the ideation base

573

Fig. 14-2: The shift from folk models to scientific ones


As we noted above, folk models are pan of the unconscious background of thinking in
everyday situation types; they have evolved without any conscious design and are not
associated with academic contexts. Folk models can also be more conscious, of course
these are the models that people talk about, that they believe they believe (cf. Figure 14-3
to be discussed below). Scientific models are consciously designed in more restricted
situation types, usually within academic institutions, to serve as resources in reasoning
about the world.

As we also noted above (cf. Figure 14-1), we can interpret folk models and scientific
ones as co-existing varieties of the same basic system within the ideation base. In the
first instance, we will, of course, be aware of them as differing in particular domains
e.g. as operating with different lexical semantic organizations (see further below); but

14. Consciousness in daily life and in cognitive science

574

they also tend to construe experience differently in general terms scientific models tend
to rely on grammatical metaphor and thus embody a metaphorical construal of experience:
see Figure 14-3.

Fig. 14-3: Folk and scientific models within the ideation base
In the move from folk models to scientific ones, the first aspect of language to which
scholars directed their attention was, as we would expect, the vocabulary. Thus in the
early science of the 16th and 17th centuries, scholars were concerned with the taxonomic
interpretation of the world; they tended to feel that ordinary language let them down, and
started to explore the possibilities of artificial languages, culminating in Bishop Wilkins'

14,1 Models embodied in the ideation base

575

(1668) Universal Language and Real Character. What this meant was that scientists woe
moving away from folk taxonomy as a way of interpreting the world, towards a scientific
taxonomy, in which further taxa were introduced and terms were more explicitly defined
(cf. Chpater 2, Section 2.11.3 above). The solution to their problem thus turned out to
be not, as they thought, the development of an artificial language, but rather the
development of new regularities in the forms of ordinary language.
Within the grammar, there was some awareness of the grammatical construction of
words; for instance, Wilkins introduced 'transcendental particles' on the model of (Latin)
derivational morphology so as to be able to expand the vocabulary in systematic ways.
But these scholars did not discuss the less exposed aspects of grammar the covert
categories of the clause and clause complex. They seem to have remained unaware of the
part that was played by such patterns in constructing their discourse. However, by this
time, the cryptogrammar of scientific English was already evolving along the Unes
discussed in Chapter 6; and these changes were accelerated in the discourses of Newton,
whose work marked the end of purely taxonomic science and laid the foundation of a new
model based on experimentation, general laws and predictions, and in work of later 18th
century and 19th century scholars. The most central aspect of the various changes that
took place was the reification of experience the grammatical metaphor whereby
processes were reconstrued as things. In the language of everyday commonsense, A
attracts B, so B moves a complex of two clauses; in the language of science, attraction
causes (or is the cause of) movement where the everyday clause complex, the sequence
of two processes of action, has been 'compressed' into one clause with two nominalized
elements, and a single process of being (cause *be causally', or be). When they are
reconstrued as things, processes lose their location in time and often also their
participants; for instance, A attracts B is likely to be reconstrued simply as attraction.
Attraction, repulsion, motion, gravity, acceleration, etc. can then be taxonomized in the
same way as ordinary things such as plants and animals; they become part of an explicit
taxonomy of metaphorical things. These basic resources were already in place in ordinary
language the nominal group for representing things and for organizing them into
taxonomies,2 nominalizing suffixes forreconstruing non-things as things, and so on; but

The nominal group offers the resources for taxonomizing things but the verbal group does
not offer any equivalent resources for taxonomizing processes (cf. Chapter 4 above). Thus if

576

14. Consciousness in daily life and in cognitive science

their potential was being exploited to a greater extent and in significantly different ways.
This change in the grammar entailed a change in world view, towards a static, reified
world so much so that Bohm (1979) complains that language makes it hard to
represent the kind of flux that modem physics likes to deal with. Bohm's dissatisfaction
is directed at language in general; but bis real target is or should be the language of
science. The everyday language of casual speech is, by and large, a language of flux,
construing experience in much the way that Bohm seems to demand (see Halliday, 1987).
Over the last decade or so, detailed work on the semantic patterns of the registers of
different disciplines has shown how scientific models are construed not only in physics
but also in other disciplines at secondary and tertiary level. Unsworth (1995) shows how
the resources of the ideation base are deployed in physics to reconstrue commonsense
experience e.g. sounds that we can hear arc construed into microscopic and
macroscopic sequences of processes, and these are then 'condensed' by means of
grammatical metaphor into 'things' such as vibration, contraction, and rarefaction.
Wignell et al (1990) show how taxonomic order is built up in geological models (cf.
Chapter 2, Section 2.11.3 [1] above) and how sequences of geological processes can be
'distilled' into metaphorical 'things' such as lithification. In these models, technicality is a
salient feature. In contrast, Eggins et al (1993) show that models of the past hi secondaryschool text books of history do not rely on technicality (with the exception of a few
technical terms for periods, such as "the Renaissance"), but rather on metaphorical
reconstructions of the past. Instead of people taking part in processes unfolding over
time, these processes are reconstrued as things (such as famine, war, migration) having
temporal location. This makes it possible for historical models to generalize over
particular events that people might observe, and to construe long-term patterns and
tendencies.
We now turn to another area of human experience our experience of taking part in
seeing, thinking, wanting, and feeling, i.e. in processes of consciousness. We begin by

'remember* is reined, it can be construed as a thing which is easily taxonomized


'memory: short term memory/ long term memory; semantic memory', and so on.

14.1 Models embodied in the ideation base

577

reviewing the general folk model of such processes, and then explore how this folk model
has been reconstrued in cognitive science.

14.2 Sensing: the folk model


We have shown how the system of the ideation base construes consciousness: as
conscious processing by a conscious being. Conscious processing can create a higherorder world of ideas (or, as we would say, meanings), comparable in certain respects to
Popper's World 3; this defines the essential distinction between projection and expansion
as ways of relating one figure to another (Chapter 3). Conscious processes themselves
appear as the central figures in the construal of experience (Chapter 4), and they aie
pivotal in differentiating among the various types of participant (Chapter 5). This folk
model is constituted in innumerable encounters in the course of casual conversation; and
it is instantiated again and again in contexts of everyday life.
Conscious processes are of two kinds: sensing, and saying. Since what we aie
exploring here is the modelling of "mind", our concern in the first instance will be with
sensing. Processes of saying will be implicated later on. Let us first review the folk
model by exploring some text instances of how speakers and writers construe
consciousness. In the following passage from a dialogue, one speaker refers to her own
conscious processing:
Text (1): Extract of casual conversation (backchannelling omitted)
When I feel really depressed, I think "what a horrible lot", and
I call them to myself - I think "they're absolutely horrible".
"What other parent ever had such children?" I think; "I've simply
devoted nty life to that lot". But when I ' m feeling in a good irood,
and I them sort of more or less as individuals, I think to
myself "I don't cars whether they're sort of particularly devoted
or not." "They're so lovely" I think. (Svartvik& Quirk, 1980:319.)
Processes of sensing are shown in the extract in bold. The following are some notes
on our analysis.
(i)feel. The two clauses I feel depressed, I'm feeling in a good mood are on the
borderline of sensing and being; we treat them here as processes of being, since

578

14. Consciousness in daily life and in cognitive science

they seem agnate to I am depressed

(relational: ascriptive, with feel as non-

salient syllable); cf. the proportion


I feel depressed : I am depressed :: I don't feel depressed : I am
not depressed

rather than to I feel thai I am depressed (mental: cognitive, with feel as salient),
where the proportionality does not hold:
I feel depressed : I feel that I am depressed * I don't feel
depressed : I don't feel that I am depressed

where the last is not contradicted by but in fact 1 am.


(ii) see. The clause / see them as individuals could be either mental: cognitive +
projection 1 consider that they are individuals' or mental: perceptive + Role 1
see them in their guise as individuals'. The latter interpretation is agnate to /
look at them as individuals, which seems more plausible here. But in either
case the see figure is one of sensing.
(iii) care. The clause/don't care is grammatically mental: emotive. The following
clause whether they're ... devoted or not is a projection; but it is one of the
'fact' type, not projected by the mental clause itself (see the second paragraph
below).
The speaker construes her processes of thinking as (1) a figure of sensing, realized by
a mental clause (/ think), and (2) a projected figure, realized by a clause representing the
'content' of that clause the idea projected by her thinking (they're absolutely horrible; I
don't care whether they're son of particularly devoted or not; they're so lovely; and so
on). The figure of sensing is a configuration of a Process (think) and the participant
engaged in sensing, the Senser (/); that is, consciousness is construed as a
complementarity of change through time and persistence through time as a conscious
participant involved in an unfolding process.
In the grammar, the 'idea* is a separate clause that combines with the 'sensing' clause
in a clause complex, through the relation of projection; see Figure 14-4 for a structural
interpretation. In our analysis (unlike that of the mainstream grammatical tradition), the

14.2 Sensing: folk and scientific

579

projected clause is not a constituent part of the mental or verbal clause by which it is
projected. There are numerous reasons for this; some of them are grammatical for
example, it cannot be the focus of theme-predication [we do not say: it is that they're
absolutely horrible that I think]; it cannot be the Subject of a passive mental clause [we
do not say: they're absolutely horrible is thought by me]\ it is presumed by the substitute
so , which is also used to presume conditional clauses in clause complexes: / think
they're absolutely horrible and my husband thinks so too]. But these, in turn, reflect the
semantic nature of projection: this is a relationship between two figures, not a device
whereby one becomes a participant inside another. We can thus show the difference
between these and 'fact' clauses, those where the idea clause is a projection but it is not
the accompanying mental clause that is doing the projecting; such readymade projections
do function as constituents. An example here is / don't care whether they are devoted or
not; compare it's not whether they are devoted or not that I care about. Figure 14-5
shows how the proposition construed by the idea clause is projected, as the "content of
consciousness", by the Senser involved in the process of sensing. This content is brought
into existence by the sensing process, as actualized through the Senser; and it is construed
as being of a higher order of semiotic abstraction than the process of sensing itself (i.e. it
is always at one further remove from the instantial context).
/
1
mental clause
Senser:
nom. gp.t
conscious

think
Process:
verbal gp.

they're absolutely horrible


'2
projected clause

Fig. 14-4: Clause complex of mental clause projecting an idea clause


In the conversation above, the 'ideas' projected by the figures of sensing are quoted
a kind of interior monologue, Typically, however, figures of sensing project ideas as
reported.3 The potential for projection, and the contrast between these two modes of
projection, indicate a similarity between figures of sensing and figures of saying;
interestingly, one of the figures of sensing in this extract includes a participant that is

A number of these figures, if we consider just their wording, could be interpreted as reported.
However, the context of the discourse, and their realization in phonology, makes it clear
that all are in fact quoted projections.

580

14. Consciousness in daily life and in cognitive science

more characteristic of figures of saying a Receiver: myself (in / think to myself). We


shall return to the relationship between sensing and saying in Section 14.4 below.

Fig. 14-5: The grammar's construal of consciousness as meaning


Speakers in casual conversation are often concerned with construing their own
consciousness. Indeed, Painter (1993) documents how one child rst learned to construe
mental projection: he began with figures in which he himself was the Senser. The system
made it possible for him then to generalize his own experience of consciousness by
construing other persons in the Senser role, as he built up a model in which this role
could be occupied by any conscious (prototypically human) being. Thus in narrative, the
writer often construes his or her characters' consciousness on the same model as in Figure
14-4 above. Projected ideas are typically reported rather than quoted. For example:

14.2 Sensing: folk and scientific

581

Text (2): Extract from A first time for everything


Jessica Steele, A first time for everything, Mills & Boon.

projection text
If she were to be truthful,
"B
Joss
ownmd,
a

-p
a
-p
a

a
'0

a
'3la
PPP
P+2

a
Bl
B+2

a
Ba
Bxpa

she didn't feel very much like going out that Monday
evening .
She could not hmvm aid
vfoy particularly it was that the Beacon Theatre Group had
no appeal that night,
though, as she poured herself a second cup of coffee,
she didn't think
the dull and gloomy weather had very much to do with the
way she was feeling.
Silently she sipped her coffee,
and a few minutes later she carried her used dishes from
her dining-room and through to her smart cream and pale
green kitchen,
knowing
that she would go out
It wasn't in her nature to let anyone down,
and Abby, her closest friend, was at present smitten with
Fergus Perrott
and, for some reason, seemed to need her along
to boost her confidence.
Joss set about tackling her washing up,
raflaeting
that Fergus liad not m*k*d Abby
to go out with him yet,
but that these new and frequent visits to the Beacon
Theatre Group an offshoot of the Beacon Oil Sports and
Social Club seemed to be paying dividends.
She was mur* she had tan a gleam of interest in Fergus's
eyes last Friday
as he ' d watched Abby during rehearsals .
All three of them worked for Beacon Oil at Beacon House,
London she and Abby both on the secretarial side,
vtfiile Fergus worked in Personnel.
Thinking of last Friday,
Joss ponderad
that she must have been feeling a bit like today's weather
then or maybe started to feel in the need of something more
stimulating than her present job,
for she had raliad
that she had been only half joking
when she ' d mmktd Fergus then

582

14. Consciousness in daily life and in cognitive science

'P*P"P

if his department had been notified of any interesting


secretarial vacancies.
...

Leaving her neat and tidy kitchen,


Joss went
to get ready to go and pick up Abby.
They had known each other for three years now.
[] The whole time in fact
since, at twenty years old, Joss, with her feet well and
truly on the secretarial ladder, had started to work at
Beacon oil.
In an attempt to lift her glum spirits, Joss tried to count
her blessings.
Here Joss's conscious processing is construed as part of the sequence of her activities. The
figures of sensing are often accompanied by the content of her thinking, represented by a
projected idea The figures of sensing are also often construed as simultaneous with
figures of doing & happening, involving Joss as Actor, as in as she poured herself a
second cup of coffee, she didn't think the dull and gloomy weather had very much to do
with the way she was feeling. Here enhancing sequences are used to construe figures of
sensing and figures of doing & happening as unfolding in parallel, with the same person
in the pivotal roles of Senser and Actor respectively. In addition to deploying figures of
sensing to construe Joss's consciousness, the writer has also drawn on a lexical
metaphorical motif of (vertical) movement in an abstract mental space: boost her
confidence, lift her glum spirits. We will return to the lexical metaphor below.
Like Text (1) above, Text (3) provides an example of a speaker construing his/ her
own processes of consciousness:
Text (3): Extract of casual conversation
A: Oh dear,

(sighs) one orgt

how time runs

[?].

I think

Malcolm's twenty-seven, twenty-eight, perhaps a bit more. I don't


know.
B: I eventually Btinatt twenty-eight, twenty-nine. I must have
looked at him for some time.
A: He went -

B: He's not, he's not easy to gummm actually.

583

14.2 Sensing: folk and scientific

A: No. He got a brilliant first when he was twenty and it meant he


couldn't graduate till he was twenty-one. They wouldn't give it to
you. And he stayed - did he stay at Oxford to do a postgrad year
or did he come immediately here? I can't rnnb*r

He's working

for a Ph.D. here I think.


B: m

A: But I think he gets so involved in his computer business that I


don't know how his Ph.D. is going.
B: Shouldn't think he had much time left.
A: I shouldn't think so. (Svartvik & Quirk, 1980: 153)

Here also the projected figures are reported rather than quoted (for example, one forgets
how time runs rather than one forgets: how does time run?): as already noted, reporting
is the typical mode of projection for figures of sensing. The figures of sensing are again
realized congruently by mental clauses. However, some of these mental clauses are
metaphorical: they also stand for interpersonal assessments of modality. Thus the
projecting clause complex / think
Malcolm's twenty-seven, twenty -eight,
perhaps a bit more is agnate to a simple clause with either a mood Adjunct (Malcolm is
probably twenty-seven, ...) or a modal Finite within the Mood element (Malcolm will be
twenty-seven, .,.). A projection mental clause such as / (don't) think, since ideationally it
realizes a figure of sensing, construes the speaker as 'Senser at the time of speaking' (it
occurs metaphorically only in simple present tense); at the same time, it enacts the
speaker's own 'intrusion' into the dialogue his or her judgment about how much
validity can be attached to the proposition contained in the projected clause.
Interpersonal metaphor is thus the hinge between the ideational and the interpersonal
modes of constructing the self. In the ideational mode we construe

ourselves as

conscious Sensers, while in the interpersonal mode we enact ourselves as speakers


interacting with addressees; the metaphor brings the two together in such a way that the
ideational construal stands for the interpersonal

enactment (see Figure 14-

6). The grammar of everyday discourse thus clearly points to the significance of
interpersonal meaning in the way that we construct ourselves the self is not only

14. Consciousness in daily life and in cognitive science

584

construed but it is also enacted. Cognitive scientists, however, have derived their object
of study, and their model of this object from the ideational perspective alone, failing to
take the interpersonal perspective that of enacting into account.

Fig. 14-6: Ideational construal standing for interpersonal enactment


The everyday grammar's contribution to the construal of sensing is thus both rich and
varied. Some features of it are particularly significant to the uncommonsense model of
mainstream cognitive science. The grammar separates out consciousness from the rest of
our experience in the form of mental processes, capable of projecting ideas; but in
addition, consciousness may also be 'externalized1 in the form of verbal processes, capable
of projecting locutions. This 'world view1 can be depicted using the conventions of comic
strips: see Figure 14-7. The conventions of comic strips clearly differentiate between
figures of sensing and saying on the one hand and figures of being & having and doing &
happening on the other (cf. Chapter 3, Section 3.3.1 above, esp. Figure 3-5): the latter
are represented graphically, whereas the former are represented linguistically, in terms of
their projected content Comic strips thus codify the higher-order nature of projections and
their constitution in language.

14.2 Sensing: folk and scientific

585

Fig. 14-7: The container model of the mind of lexical metaphors


The grammar thus construes sensing as a bounded semantic domain within our total
experience of change. This picture is further enriched through lexis, prominently through
lexical metaphors (cf. Chapter 6, Section 6.2.1 above). Metaphors relating to space, with
the mind as a container (Reddy, 1979; cf, Lakoff & Kvecses, 1987, on Anger), a finite
space or a physical entity reinforce the grammar's construal of a bounded domain of
sensing. We cited the lexical metaphors boost her confidence, lift her glum spirits from
Text (2) above; other examples of lexical metaphors include:
(i) mind:

mind as space: cross one's mind, broaden the mind, be out of one's mind, to be
driven out of one's mind, get something out of one's mindt search one's mind,
at the back of one's mind, to put at the back of one's mind, at the front of one's
mind

586

14. Consciousness in daily life and in cognitive science

mind as container: occupy the mind, escape/slip one's mind, an open mind, a
closed mind, keep in mind, to have in mind,
mind as physical organ: blow one's mind, to boggle the mind, to have
something on one's mind, the mind recoils
(ii) other mental constructs:
emotion as location in vertical space: be up/ down, be high/ low, depress sb,
lift sb's spirits, spirits soar; fall in love, love deeply, abhor/ detest/ dislike
deeply
emotion as liquid/ gas (contained in body): explode, vent one's anger, blow
one's top, to boil over, to smoulder, to cool down, to keep the lid on,
This mind-space may enter into material processes of storing, searching, crossing,
escaping etc., either as participant or as circumstance, and also into relational processes of
"being + Location". It is interesting to note that in these various lexical metaphors the
Sensers are still very much present; they are not effaced. In fact, a number of these lexical
metaphors constructed on the model of material clauses retain the option of projecting;
see Figure 14-8.

he

kept

in mind

that the moon was a


balloon

Senser

Process

Location

idea clause

'remembered*
Fig. 14-8: Projecting combined with material lexical metaphor
Let's now consider how this folk model informs the models of consciousness in
(mainstream) cognitive science.
14.3 From folks to scientists
What is the nature of the move from our everyday construal of the experience of
consciousness our folk theory of Sensers sensing phenomena or projecting ideas to

587

14.3 From folks to scientists

the way cognitive scientists construe that experience? We can see the essential nature of
this move when the folk theory is reconstmed as if it was a scientific one. When Dennett
(1981) tries to characterize folk psychology, he has in fact already made the move (our
bolding):
What

are

beliefs?

Very roughly

folk

psychology

has

it

that

belief"* are information-bearing states of people that arise from


perception
desire,

and that,

together

with

lead to intelligent action.

uncontroversial,

but does

folk

appropriately

related

That much is relatively

psychology

also

have

it

that

nonhuman animals have beliefs? If so, what is the role of language


in belief? Are belief constructed of parts? If so, what are the
parts? Ideas? Concepts? Words? Pictures? Are belief like speech
acts or maps or instruction manuals or sentences? Is it

iitplicit

in folk psychology that beliefs enter into causal relations, or


that they don't? How do decision
between

belief-deire

introspectible,

and if

conplexes
so,

what

and intention intervene


and actions?

authority

Are

do the

belief
believer's

pronouncement have? (p. 91, 1990 reprint)

Instead of saying people believe that..., people want others to ..., etc., as people do
in everyday discourse, Dennett writes beliefs, desires, etc.. Here mental clauses with a
Senser, a Process of sensing (believing, wanting, desiring, etc.), and either a
Phenomenon entering into (or being created by) the Senser's consciousness or an idea
projected as a separate clause, have been reconstrued as nominal groups with a
nominalized mental process as Thing/ Head and (typically) neither participants nor
projected ideas. This reconstrual is brought about by grammatical metaphor meanings
that are normally construed by clauses are construed as if they were meanings that ate
normally construed by nominal groups. If Dennett had tried to constitute the folk theory
as it is, instead of construing it as if it was a scientific theory, he might have written:
Very roughly, when people believe something, they believe that
something has happened because they have seen or heard it happen;
and if they believe that something has happened and they want
something else to happen, they do something about it.

588

14. Consciousness in daily life and in cognitive science

The two modes of construing our experience of consciousness ate compared in


grammatical terms in Figure 14-9.
(1) Folk model (congruent construal):
they

believe

ota:

that something has


happened

>

mental: cognitive &


metaphenomenal:
idea

(X'P:
clause:
projection
(idea)

because

they

have seen or
heard

it happen

> *3:
mental: perceptive & macrophenomenal

Senser

Process

Senser

Process

nom.
gP-consc.

verbal
group

nom. gp.:
consc.

verbal group

Pbenom
enon
clause:
act

(2) Scientific model (metaphorical reconstrual):


Beliefs (= informationarise from
bearing states)
relational: identifying & circumstantial
Token
Process
nom. gp.
verbal gp.
Event:
Thing:
verbalized conjunctive
nominal ized
process of cognition
relation of cause

perceptions
Value
nom. gp.
Thing:
nominalized
process of
perception

Fig. 14-9: Congruent folk model reconstrued metaphorically as scientific model


The scientific model is metaphorical; and it stands as a metaphor for the congruent folk
model. As the figure makes explicit, there is a considerable loss of ideational
information as one moves from the congruent mode to the metaphorical mode:
grammatically, a clause complex is compressed into a clause, and the clauses that ate
combined in the clause complex are compressed into nominal groups. As a result, the
subtle distinction between the cognitive projection of ideas (sb believing that....) and the
perceptive sensing of acts (sb seeing sth happen) is lost, and participants can be left
implicit. (The mental clause's capacity for projection is echoed in Dennett's definition of
belief:

[Token/ Identified] beliefs

[Process:] are

[Value/ Identifier:] information-

bearing states... .) The possibility of leaving participants implicit means in practice that
Sensers are effaced in the scientific model and, as a result, the consciousness we

589

14.3 From folks to scientists

experience in the living of life is also construed out of the picture, being replaced with
unconscious processes not accessible to our experience.
The move from the congruent mode of the folk theory to the metaphorical mode that
provides the resource for theories within cognitive science can often be found in
introductions to accounts of how people sense the phenomena of their experience, as in
the following passage from Restak (1988: 242), an introductory book on "the mind" (our
bolding, underlining and italics):
to
example,

explain
is

the mind's

operation

in

thinking.

Here,

for

an everyday situation: two people meet on a beach.

Michelle rco0niz*a that she's encountered Michael before, but


can't come up with his name. But she does recall

that he's a

doctor, specializes in paediatrics, and lives in New York City.


Michael r*member Michelle's name but can't dredge up from hi
memory any biographical details about her.
previously

at

party

given

by a

He recall they met

friend

to

celebrate

the

completion of his residency. Michael can bring vividly to mind


what Michelle was wearing and how attractive he found her.
Ordinary xprinc like this raise important questions about
how

thinking

rcogniz

is

organized.

Michael's

him, but can't

face,

Why is

rmmbr

come qp with his

it

that

significant

Michelle can
facts

about

name? In Michael's case/

the

organization would seem to be different: he can r*mmbr names


but specific life details are only a blur. What kind of mental
organization in Michael might account for these differences?
Ihe most popular metaphor for the human mind is that of a huge and
intricate filing system. When Michelle and Michael encounter each
other, facial

r cogni t i on sets off an elaborate search through

"files" stored within billions of neurons. . . . (Restak, 1988: 242)

Restak begins by giving a common-sense account, using the congruent grammar of


mental clauses with projected ideas: Michelle recognizes

> that ..., she does

590

14. Consciousness in daily life and in cognitive science

recall

that.... He then uses a lexical metaphor for remembering that is pan of


our everyday lexicogrammardredge up from memory, bring to mind. Here the 'mind'
is construed as an object with extension in space a container. The mind is also

construed as a circumstance (from memory, to mind), in a mental clause that is partially


constructed on the material model of manipulating an object but which is still mental,
since it is configured with a conscious participant (the Senser) and since it can project
that she
what Michelle was wearing; cf. bring to mind
was intelligent, dredge up from memory
that she was a physician). When he has
finished this commonsense account, he raises general issues for an uncommonsense,

(bring to mind

scientific account by using the metaphorical mode: experiences like this (rather than
people experience similar situations or the like), thinking (rather than how people
think), facial recognition (rather than people recognize faces), and so on. (When later he
harks back to the illustrative account of a particular situation of everyday experience, he
returns to the congruent mode.)
Restak also deploys the lexical metaphor of a mental space that we find in the
commonsense model a space in which 'objects' can be stored, which can be searched,
and so on: facial recognition sets off an elaborate search through "flies" stored within
billions of neurons. The spatial metaphor of the commonsense model is taken over by
cognitive scientists. It serves as the sources of processes in their model of the mind
processes of storing, searching, retrieving etc. within figures of doing & happening and
processes of being located at/in within figures of being & having. That is, processes of
sensing are reified, and processes of doing & happening and of being & having take thenplace. The spatial metaphor also opens up the way for modelling the mind along
computational unes: human memory can be modelled on computer memory.
This shift in the passage quoted from Restak is representative of what we think
happens with respect to how mainstream cognitive scientists construe their object of
study. To illustrate further some features of cognitive models, we have included a short
extract from a book on the psychology of cognition. It is the very first paragraph of the
book. We have analysed each clause in terms of the major types of figure doing,
sensing, saying and being-&-having: see the columns to the right of the text in Table
14(1).

591

14.3 From folks to scientists

Table 14(1): Analysis of passage on cognition4

TEXT
(Processes in bold)

Type of figure
doing sensing

saying being
&
having

[1 ] Semantic memory is concerned with the


structure of knowledge.
with dhow jcnowtedge is stored, cross-referenced and
indexed]):
it is concerned with the organization of everyday
world knowledge, and
with the representation of meaning.
[2] Semantic memory is not just an internal
dictionary
Bin which linguistic terms are listed and
defined!
[3] The elements are concepts.
and although most concepts are denned by their
properties
not all concepts are verbal ones.
[4] Facts or propositions are represented by
concepts
Q linked in particular relationships]!
and sets of propositions combine
to form related areas of knowledge.
[5] Although there is a common core of culturally
shared knowledge,
semantic memory is personal
because each individual's knowledge and experience
differ.
[6] It is not just a static mental encyclopaedia, but a
working system,
in which new facts are constantly being
incorporated,
stored knowledge is being updated and
reclassified,
and particular items of information are being
sought, located, assembled and retrieved.
[7] It represents one of the most important,
interesting and difficult areas of study in cognitive
psychology today.

The excerpt is taken from the introductory paragraph to Chapter 1, Semantic memory and the
structure of knowledge, of Gillian Cohen, 1977. The psychology of cognition. New York:
Academic Press.

14. Consciousness in daily life and in cognitive science

592

[8] Semantic organization is especially important


because it is one of the most powerful and pervasive
determinants of performance in mental tasks.
[9] Plow knowledge is arranged])
determines
Qhow we speak
and how we understand,
how we solve problems
and how we remember])
[10] It is worth
[analysing in some detail the reasons
([why semantic memory is especially intractable to
study] B
because many of these problems are endemic in
cognitive psychology generally,
and because the limitations
[which are inherent in our methodology]]
are quite strikingly illustrated in this area.
[11] Methodological problems will form a
recurring theme throughout the book.
[12] The validity of theories of cognitive function
depends on the power of the research methods,
so that we cannot hope to assess the value of
the evidence, nor the truth of the conclusions
unless we scrutinize the methods,
probing their weaknesses,
and trying to evaluate the advantages and
disadvantages of different approaches.
Note: how we speak in (9) is verbal activity; it is on the border between verbal processes
and material ones: we can interpret as verbal (but it cannot project), or as behavioural (here
simply treated as a subtype of material).
Figures of sensing are realized not as mental clauses but as nominal groups taking on
roles in clauses of various process types (predominantly relational; cf. below). We can
contrast this scientific excerpt with a fragment of ordinary speech: see Table 14(2).
The dominant process type in the psychology text is relational being, representing,
defining, and so on. There is one short passage of doing, sentence number [6]; it describes
possible operations on memory. There are only two congruent mental processes with
Sensers (we), one of understanding (how we understand) and one of remembering (how we
remember); but both of them occur in a rankshifted (embedded) clause complex:
[[HOW knowledge i arranged! dtrmin
we

uhow we

apk and how

undr*tand, how we *olv* problems and how we rmanbrll

593

14.3 From folks to scientists

Moreover, the sentence says that how we sense is determined by how we arrange
knowledge.
Table 14(2): Contrast between folk model of people thinking and scientific model of
people sensing
material
folk
model:
clause:

scientific
model:
nominal
group
serving in
clause:

mental

relational

you would see her ...;


I've known her come in
and say ...; I'd forgotten
I'd been ill; I remember
now when she carried
Walter ...; I think she
deemed I wasn't big
enough ...

in which new facts are


constantly being incorporated,
stored knowledge i s
being updated and
reclassified,
and particular items of
information are being
sought, located,
assembled and
retrieved.

Semantic memory is
concerned with the
structure of knowledge. ...
Facts or propos.itfon.gf are
represented by concepts. . .

In figures of doing, mental constructs serve as the Goal being manipulated in the
mental space:
Location:

in which

Goal:

new facts
stored knowledge
and particular items
of information

Process:

are constantly being incorporated


is being updated and reclassified
are being sought, located,
assembled and retrieved

Actor:
9
?

thashewas physican).When has

These figures are realized by passive clauses, where the Actor is absent. This makes it
possible to construe sensing without any explicit personal agency; the processes are
initiated at a subpersonal level. Edelman (1992: 237-8) comments on the general tendency
that the example above illustrates:
Human memory is not at all like computer memory. ... In whatever form, human
memory involves an apparently open-ended set of connections between subjects and a
rich texture of previous knowledge that cannot be adequately represented by the
impoverished language of computer science "storage", "retrieval", "input",
"output". To have memory, one must be able to repeat a performance, to assert, to relate
matters and categories to one's own position in time and space. To do this, one must
have a self, and a conscious self at that Otherwise, one must postulate a little man to
carry out retrieval (in computers, it is we, the programmers, who are the little men).

594

14. Consciousness in daily life and in cognitive science

The text extract in Table 14(1) above is quite representative. Let us cite just one
additional example. The following extract is from the opening chapter of a recent text
book in cognitive science (Stillings et al, 1987: 1; our holding = processes in congruent
mental clauses, underlining = figures of sensing metaphorized as nominal groups or
names of sensing):
1. What is Cognitive Science?
One of the most important intellectual developments of the past few decades has been
the birth of an exciting new interdisciplinary field called cognitive science.
Researchers in psychology, linguistics, computer science, philosophy, and
neuroscience realized that they were asking many of the same questions about the
nature of the human rnin^ and that they had developed complementary and potentially
synergetic methods of investigation. The word cognitive refers to perceiving and
knowing. Thus, cognitive science is the science of the mind. Cognitive scientists
seek to understand perceiving, thinking, remembering, understanding language,
learning, and other mental phenomena. Their research is remarkably diverse, ranging
from observing children, through programming computers to do complex problem
solving, to analyzing the nature of meaning.
1.1 The Nature of Cognitive Science
Cognitive scientists view the human mind as a complex system that receives, stores,
retrieves, transforms, and transmits information. There are four important assumptions
to this information-processing view.
Formal Information Processes
The first corollary is that information and information processes can be studied as
patterns and manipulation of patterns. To clarify this assumption, let us look at an
example. Consider the following longhand multiplication problem: [...]

As in the examples above, the object of study of cognitive study is constructed by


ideational metaphor, as reified sensing (perceiving, thinking) or as names of sensing (the
mind, mental phenomena). There are also a number of congniently realized figures of
sensing mental clauses with Senser + Process, with the potential for projecting. But
these figures of sensing are not what cognitive science studies; rather they construe the
processes of doing cognitive science: the Sensers are (cognitive) scientists. This is quite a
normal use of congniently realized figures of sensing in scientific discourse; but here they
are rather striking, because of the disjunction between the congruent construal of the
scientists' own mental processes and the incongruent construal of those they are trying to
understand

14.4 The model in mainstream cognitive science

595

14.4 The model in mainstream cognitive science


To sum up the discussion so far; while the domain of scientific theorizing about
cognition is determined by the grammar of processes of sensing, the model is
depersonalized, and sensing is construed metaphorically in terms of abstract "things" such
as knowledge, memory, concepts. This suggests that mainstream cognitive science is
basically an elaborated variety of a folk model, rather than a different scientific alternative
(cf. Matthiessen, 1993a):
(i) The congruent ideational system separates out consciousness from the rest of
our experience and construes it as a domain of sensing, embodying a Medium +
Process complementarity where conscious beings (Medium) perceive, think,
want, feel (Process). Sensing is thus 'mediated' through the Senser; and this
process may project ideas into semiotic existence. This domain of Sensers
sensing (that...), which is construed in the congruent system, is taken over in
cognitive science. However, it is not taken over as being itself a theory of
conscious processing; instead, it is treated as a phenomenon that is, sensing
is turned into the object of study.
() Since it is not taken over as a theory, the fundamental insights of the folk
theory are ignored: figures of "Sensers sensing (that ...)" are re-construed
through grammatical metaphor as participants. In particular, the domain of
sensing is reified as the "mind", so that instead of somebody perceiving things
happening, or somebody thinking that the moon was a balloon, the model of
cognitive science has perception, vision, cognition, learning, memory, ... .
(iii) Since figures of sensing are reified as participants, the processes of sensing are
likewise turned into things, and the participants in sensing, the Sensers, aie
typically effaced. Ibe Senser/ sensing complementarity of the folk model is
thus lost, as is the feature of Sensers projecting ideas into existence.
(iv) Since figures of sensing are reified as participants, they can themselves be
construed in participant roles. Here another feature of the folk model is taken
oven its spatial metaphor is retained and further elaborated Thus the mind is
construed as a space where the metaphorical participants of sensing are involved

596

14. Consciousness in daily life and in cognitive science

in processes of doing & happening and of being & having: thoughts, concepts,
memories, images are stored, located, retrieved, activated and so on.5
(v) Since figures of sensing are reined as participants, the path is opened up to the
taxonomic interpretation of sensing, in the form of scientific taxonomy:
memory long-term/ short-term memory, sensory memory, semantic memory;
recall free recall; learning associative learning/ cognitive learning/
classical conditioning;....
(vi) Since Sensers are effaced, and projection is lost as a feature of the Senser/
sensing complementarity, the gateway to the interpersonal realm where
Sensers aie enacted as interactants in dialogic exchange is closed, and the
interpersonal element in the ideational/ interpersonal complementarity is lost.
Figure 14-10 represents the central motif in the metaphorical reconstrual of sensing:
sensing is 'extracted* from figures of sensing as a domain, and reified to become one of a
variety of participants that take on roles in figures of being & having and doing &
happening, taking place in the mind construed as a container.
The 'scientific model' in mainstream cognitive science is centrally concerned with
information located in the individual's mind.6 This information is organized in
some way as a conceptual system. The definition given in a newsletter produced by the
cognitive science panel of the Australian Research Council, October 1989 (our italics)
was as follows:
Cognitive science is the systematic study of mental processes. Amongst the
disciplines commonly involved are cognitive psychology, linguistics, philosophy,
computer science and neuroscience. Cognitive science seeks to elucidate the
information-acquisition and information-processing mechanisms underlying cognitive

The widespread lexical metaphor of memory as a kind of space predates cognitive


psychology by many hundreds of years; see e.g. Yates, 1966, on medieval notions such as
the "memory theatre" used as aids to remembering.

Alongside this cognitivist approach, there is a material one embodied in formal approaches
to semantics, where the 'aboutness1 of linguistic expressions is taken as central and these
expressions are interpreted in terms of models of possible worlds: see Sections 10.1 and
10.2 above. However, in this respect there is a fonnal-cognitivist alliance: meaning is
interpreted not as something in its own right but as something outside language, either a
mental construct (concepts, ideas etc.) or a material one (referents in the real world or a
formal model of a possible world).

14.4 The model in mainstream cognitive science

597

tasks like perception, recognition, storage of information and its retrieval from
memory, problem-solving, language acquisition, language comprehension and
language production. Cognitive scientists seek to construct and test explicit theories
of the mind, specifying the kinds of information processing that occur in cognitive
activity, and to model the ways in which the ability to perform such tasks are acquired,
changed or impaired. It is common for cognitive scientists to express models of mental
processes in explicitly formal and/or computational terms.

Fig. 14-10: Scientific reconstrual of the mental domain


Other definitions include Simon & Kaplan (1989: 1):
Cognitive science is the study of intelligence and intelligent systems, with particular
reference to intelligent behavior as computation. Although no really satisfactory
intentional definition of intelligence has been proposed, we are ordinarily willing to
judge when intelligence is being exhibited by our fellow human beings. ...

and Wortman & Loftus (1985: 129):


To cognitive psychologists ... the thought processes going on in a person's mind are
just as important as the overt behaviors we can see.

598

14. Consciousness in daily life and in cognitive science

The mainstream cognitive science model is thus basically derived from a variety of the
commonsense model. It creates a metaphorical distance
from experience as
construed in our congruent grammar, so that the conscious processing that we experience
can be reconstrued as a 'subconscious' domain that we do not have access to an abstract
space where figures of doing & happening and of being & having aie the ones that
operate, rather than figures of sensing. This would seem to be at one remove (at least)
from the folk model, which might reasonably be seen as one of experientialist cognition
in Lakoffs (1988) sense one that is in direct contact with the everyday, embodied
experience of Sensers. Thus, the metaphorical reconstrual of mental processes effaces
the Sensers

involved in these processes the conscious beings, prototypically

human, who are thinking, knowing, believing, remembering and so on. This effacement
of the Sensers is of course not accidental: in fact, one central feature of the way in
which cognitivists reconstrue mental processing in metaphorical terms is that the
grammatical metaphor makes it possible to distance the account from our everyday
experience. Restak (1988: 243-4) puts this succinctly as follows (our italics):
How does the mind construct this representational system? What goes on in the brain
so that the word beach evokes a host of associations? Certainly simple introspection
doesn't offer an answer to this question. Although you are subjectively aware of each
association when it arises, you haven't any idea why it has sprung to mind. Many of the
mind's operations remain permanently inaccessible. Every one of us sometimes
"knows" a certain word, yet can't get it out despite its being "on the tip of the tongue."
Even when we believe we know how we think, we may be wrong. An experiment
conducted several years ago shed some light on the degree of access we have to
operations of our own mind. [...]
The disparity between the mind's operations
aspect of trying to discover how we think.

and our access to those operations is an

Here a contrast is construed in the grammar between the congruent we think, we


believe, we know and the metaphorical the mind's operations. This contrast represents the
conflict between our everyday experience of

ourselves seeing, feeling, thinking,

remembering, and so on, and the "scientific" model of cognitive science. Indeed, Dennett
(1988) makes the generalization that "every cognitivist theory currently defended or
envisaged ... is a theory of the sub-personal level". Given this orientation, it would thus
seem that the unified senser existing as a person who "senses" is an illusion construed by
the grammar as part of a folk theory of our own sense of conscious processing.
D'Andrade (1987) discusses two 'scientific' developments of the folk model, academic
psychology and psychoanalysis; he suggests that "though the academic and

14.4 The model in mainstream cognitive science

599

psychoanalytic models modify the folk model, it is clear that these are modifications of
an already existing conception of the mind". These two models move away from the folk
model in two directions, (i) They reinterpret figures of sensing as figures of doing or
being-&-having; that is, they interpret mental phenomena in material terms. With the
growth of cognitive psychology, this situation has changed, of course; it is no longer
disallowed to talk about mental processes, (ii) They emphasize motivation as an
important unconscious psychological factor; thus they introduce unconsciousness in the
account of the workings of the human mind. In the systems of process types in the
grammar, there is no 'unconscious' type of sensing distinct from the conscious ones that
can project ideas. D'Andrade summarizes these differences as follows:
Thus, even though the academic and psychoanalytic models have their origins in the
folk model, both are deeply at variance with the folk model. That is, the folk model
treats the conscious mental states as having central causal powers. In the folk model,
one does what one does primarily because of what one consciously feels and thinks.
The causal center for the academic model is in the various physical states of the
organism in tissue needs, external stimuli, or neural activation. For the
psychoanalytic model, the causal center is in unconscious mental states. Given these
differences in the location of the causal center of the operations of the mind, the three
models are likely to continue to diverge.

14.5 Beyond sensing folk & scientific


In a way, the two directions away from sensing that D'Andrade identifies in the
unconscious folk model material reinterpretation and 'unconsciousness' are
opposites: the first reconstrues sensing in terms that aie more readily observable by
scientific method (i.e., method other than introspection), while the other introduces a
factor that is even less readily observable than conscious sensing: unconscious
motivation. But they share the characteristic that they construct the 'mind1 as remote from
our everyday experience with sensing.
At the same time the 'scientific' models of the mind fail to extend consciousness in
the way it is extended by the grammar of English. There are, in fact, two complementary
perspectives embodied in the semantic and grammatical systems of English; and together
they point towards an alternative interpretation both of 'information' as constructed in
cognitive science and of the individualized 'mind' that is its object of study.
(i) Ideational: Sensers and Sayers. The ideational resources of language aie
primarily a theory of experience, so they are reflected fairly directly in
consciously designed theories such as those of cognitive science. If we stay
within the ideational metafunction, where mental processes are construed, we

600

14. Consciousness in daily life and in cognitive science

also find other processes that are complementary to these: those of saying
(verbal processes) and those of symbolizing (a type of relational process).
(ii) Interpersonal: interjetants. If we move outside the ideational
metafunction to the interpersonal, the resource through which we interact with
other people, we find that here we are acting out our conscious selves
"modelling" consciousness not by construing it but by enacting it. Since this
kind of meaning is non-referential it is not taken account of in scientific
theories at all.
Both these perspectives that of the construal of processes other than the
mental (saying and symbolizing), and that of meaning as enacting as well as
meaning as construing are absent from the cognitive science modelling of mind;
and in our view they could with advantage be brought into the picture when we try to
understand these complex and central areas of human experience. To do so would both
enrich the cognitive model and steer it away from obsessions with information, with
knowledge as a separate 'thing1 divorced from meaning, and with mind as the exclusive
property of an individual organism bounded by its skin.
What is common to these two further sources of insight is that both depend on
projection, (i) The potential for projecting is shared by sensing and saying; and (as we
have seen in Chapter 3), when they are considered together, they reveal a very powerful
principle that is embodied in the folk model: that through projection, we construe the
experience of 'meaning* as a layered, or stratified, phenomenon, with 'meanings'
projected by sensing and 'wordings' projected by saying, (ii) Projection also brings the
ideational and the interpersonal aspects of consciousness together. Ideationally, projection
is an mode of construal in figures of sensing and saying, sensers and sayers construe
meanings and wordings. Interpersonally, projection is an mode of enactment in moves
in dialogue, interactants enact propositions and proposals. Interpersonal metaphors of
mood and modality bring out the relationship between the two: here interactants
simultaneously both enact propositions and proposals interpersonally and construe
this enacting in such a way that the ideational construal comes to stand as a metaphor for
aspects of the interpersonal enactment (see Figure 14-6 above).
The two extensions of scientific and folk models are contrasted in Figure 14-11.

14.5 Beyond sensing folk & scientific

601

Fig. 14-11: Extensions of 'sensing1 in folk and scientific models


Both of these paths leading beyond the figures of sensing embodied in the folk model,
beyond the ideational domain of sensing, are of fundamental importance to our conception
of the individual mind and hence of the domain of cognitive science. By implication,
cognitive science should not only be 'cognitive', it should also be semiotic, because it is
the notion of meaning that enables us to see the connection between sensing and
saying, and between theory and enactment of consciousness. The folk model developed
unconsciously and collectively out of shared experience over hundreds of generations
construes and enacts the complexity of being a person by bringing to it a multiplicity of
perspectives. It extends the 'mind' by refracting it through language, the resource that
makes the "cognitive operations" possible and by the same token ensures that they are
not subjective but intersubjective.

15, Language and the making of meaning


Throughout our discussion of language and meaning (and it may be useful to stress here
once again that we have described only the ideational component of meaning the
ideation base not the interpersonal and textual perspectives), we have tried to
present a comprehensive picture with language at the centre of the stage. Language is not
a second-order code through which meanings created in some higher-order realm of
existence are mysteriously made manifest and brought to light. To borrow the conceit
that Firth was fond of caricaturing, there are no "naked ideas" lurking in the background
waiting to be clothed. It is language that creates meaning, in the sense that meaning has
for us as human beings (which is the only sense of it that we can know). Language is
able to create meaning because it is related to our material being (ourselves, and our
environment) in three distinct and complementary ways. In the first place, it is a part of
the material world: the processes of language take place in physiological (including
neural) and physical space and time. In the second place, it is a theory about

the

material world: language models the space-time environment, including itself (cf.
Matthiessen, 1991a; Matthiessen & Nesbitt, 1996), in a "rich" theoretical mode: that is,
both construing it (our ideation base) and enacting it (our interaction base). In the
third place, it is a metaphor for the material world: the way that language itself is
organized, as a stratified, metafunctional system, recapitulates acts out, so to speak
both

the make-up of this environment in natural (physical-biological), social and

semiotic systems-&-processes (our metafunctions) and the internal contradictions,


complementarities and fractal patterning by which all such systems-&-processes are
characterized (our stratification).
In other words, language has evolved as part of our own evolution. It is not arbitrary;
on the contrary, it is the semiotic refraction of our own existence in the physical,
biological, social and semiotic modes. It is not autonomous; it is itself part of a more
complex semiotic construct which, as we have tried to show (Pan I above), can be
modelled in stratal terms such that language as a whole is related by realization to a
higher level of context (context of situation and of culture). This contextualization of
language, we suggested, was the critical factor which made it possible to relate language
to other systems-&-processes, both other semiotic systems and systems of other kinds.

15.1 Language and other semiotic systems


In the last chapter we reviewed the "scientific" model of the mind that informs cognitive
science, looking at it from our point of view of construing experience through meaning.
We showed that 'mind1 is a construct of the ideation base, owing much to the

15.1 Language and other semiotic systems

603

commonsense picture of the world that is embodied in everyday grammar; but


problematic because it draws on this account onesidedly. The scientific model takes off
from the grammar of mental processes (seeing, feeling, thinking), but ignores verbal
processes (saying) although the two are both processes of consciousness, are closely
related grammatically, and share the critical feature of being able to create meaning by
projection. It takes off from the ideational metafunction, but ignores the interpersonal
although our folk perception of consciousness derives from both. Our sense of ourselves
as conscious beings comes as much from the fact that we talk as from the fact that we
think and feel; and owes as much to the nature of meaning as social action as it does to
the nature of meaning as individual reflection.
Others have also been critical of the established academic view of mind; and some
recent book titles suggest the kinds of alternatives that have been offered: "embodied
mind", "social mind", "discursive mind". These suggest that the concept of 'mind* should
be brought into close relation with other phenomena biological, social, or semiotic.
We will return to these in a moment. But once this has been done, the mind itself tends
to disappear, it is no longer necessary as a construct sui generis. Instead of experience
being construed by the mind, in the form of knowledge, we can say that experience is
construed by the grammar; to 'know* something is to have transformed some portion of
experience into meaning. To adopt this perspective is to theorize "cognitive processes" in
terms of semiotic, social and biological systems; and thus to see them as a natural
concomitant of the processes of evolution.
15.1.1 The representation of meaning in language, in relation to other
semiotic systems
We have located ourselves, throughout the book, in a certain region of metasemiotic
space: that is, we adopted a particular perspective on what we are trying to explain. Our
central concern has obviously been with 'meaning': our interpretation of meaning is
immanent, so that meaning is inside language, not some separate, higher domain of
human experience. (For meaning-making systems other than language, see Section
15.1.2 below.) The central meaning-making resource in language its "content plane"
(within which the ideation base, which has been our focus of attention, is one part) i s

604

15. Language and the making of meaning

stratified into two systems: that of lexicogrammar, and that of semantics.1 The semantic
system is the 'outer' layer, the interface where experience is transformed into meaning.
The 'inner' layer is the grammar, which masterminds the way this transformation takes
place. This deconstrual of the content plane into two strata, referred to in the first chapter,
is a unique feature of the post-infancy human semiotic, corresponding to Edelman's
(1992) "higher-order consciousness" as the distinguishing characteristic of homo sapiens.
Since we are interested in how experience is construed, we have focussed attention on
the semantics: concepts like "figure", "element", "process", "thing" are categories of
semantic theory. But in modelling the semantic system we face a choice: namely, how far
"above" the grammar we should try to push it. Since the decision has to be made with
reference to the grammar, this is equivalent to asking how abstract the theoretical
constructs are going to be. We have chosen to locate ourselves at a low point on the scale
of abstraction, keeping the semantics and the grammar always within hailing distance.
There were various reasons for this. First, we wanted to show the grammar at work in
construing experience; since we are proposing this as an alternative to cognitive theories,
with an "ideation base" rather than a "knowledge base", we need to posit categories such
that their construal in the lexicogrammar is explicit. Secondly, we wanted to present the
grammar as "natural", not arbitrary; this is an essential aspect of the evolution of
language from a primary semiotic such as that of human infants (see Section 15.2
below). Thirdly, we wanted to explain the vast expansion of the meaning potential that
takes place through grammatical metaphor; this depends on the initial congruence
between grammatical and semantic categories.
But in any case, it is not really possible to produce a more abstract model of
semantics until the less abstract model has been developed first. One has to be able to
renew connection with the grammar. Other scholars working in systemic semantics have

Thus semantics, as a field of study, is located within linguistics. We should therefore make it
clear that it is not being used in the traditional sense that it has had within linguistics, of
the study of the meanings of words. It is used in the sense it has always had in systemic
theory, namely the total meaning-making system of a natural language. Semantics thus
relates to the lexicogrammar as a whole. We can talk of "lexical semantics" if we want to
foreground the meanings of words (lexical items functioning in open sets), and of
"grammatical semantics" if we want to foreground the meanings of closed grammatical
systems; but just as the lexicogrammar itself is a continuum, so even more so, in fact
is there continuity between these two aspects of semantics, so we have not found it
necessary, except in one or two instances, to make this terminological distinction.

15.1 Language and other semiotic systems

605

been careful to maintain this connectivity, making clear how the semantic categories are
construed in grammatical terms; for example, Hasan's (1996) semantic networks, Martin's
(1992) discourse semantics, and Fawcett's (e.g. 1994) text-generation model "GENESYS"
(Fawcett describes his model as "cognitive" but it is firmly anchored in the grammar).
Functional linguistics generally is moving in the direction of more abstract
representations of semantics; but there is no comprehensive description available yet.
15.1.2 Socio-semiotic and bio-semiotic systems
We need now to orient our immanent conception of meaning with respect to semiotic
systems other than language. This notion of "other semiotic systems" is a rather complex
one; there are many ways of indexing such systems, but we can perhaps try to locate
them according to their relationship with (the semantic system of) language. This
relationship may be of different kinds. Let us first of all distinguish the broad categories
of "socio-semiotic" and "bio-semiotic", and consider the socio-semiotic first.
(i) Socio-semiotic

systems

(a) Socio-semiotic systems that are realized through language. This category
corresponds to Hjelmslev's (1943) concept of a "connotative semiotic": a higher-level
system that has language as its plane of expression. These include theories: every
theoretical construction, scientific, philosophical, aesthetic, and so on, is a higher-level
semiotic realized in language. They also include the codified aspects of social institutions
such as the law, the financial system, constitutions and codes of practice. Martin
interprets both genre and ideology in this light, as social activity structures and
ideological formations that are realized in language (e.g. 1985, 1992; for a critique of
Martin's view, see Hasan, 1995).
From a semantic point of view, such systems constitute contexts for language; they
can thus be modelled as part of a general linguistic theory, being interpreted as a higher
stratum of language itself. That is to say, we can extend the series:
the system of phonology realizes that of lexicogrammar;
the system of [lexicogrammar
semantics;

realized in phonology] realizes that of

606

15. Language and the making of meaning

the system of [semantics realized in [lexicogrammar realized

in

phonology]] which is what we call "language" realizes the system of


context (i.e. the "culture", considered as a semiotic potential).
Such higher-level systems (theories, institutions, genres), since they are realized in
language, aie realized as subsystems within the semantics and the grammar. These
subsystems are what we have refored to as registers; cf. our examples of cooking and
weather forecasting in Chapter 8.
(b) Socio-semiotic systems that are parasitic on language, in the sense that they
depend on the fact that those who use them are articulate ('linguale') beings. These include
the visual arts, music and dance; modes of dressing, cooking, organizing living space and
other forms of meaning-making behaviour; and also charts, maps, diagrams, figures and
the like. Many socio-semiouc systems are combinations of types (a) and (b); for
example, religious ceremonials and most types of dramatic performance.
These systems enter into relation with language in two ways. On the one hand, they
are mtonymie to language: they are complementary, non-linguistic resources whereby
higher-level systems may be realized (e.g. ideological formations realized through forms
of an; theoretical constructs realized through figures and diagrams). On the other hand,
they relate metaphorically to language: they are constructed, stratally and
metafunctionally, in the image of language itself, and hence can be modelled on language
as prototype, being described "as if they had their own grammar and semantics (Kress &
vanLeeuwen, 1990,1996; van Leeuwen, 1988, 1991; Steiner, 1984, 1988b; Winograd,
1968; OToole, 1994, 1995; Matthiessen, Kobayashi & Zeng, 1995; Lemke, in press).
(See especially in this connection OToole's analysis of painting, architecture aid
sculpture in stralal and metafunctional terms.)
(ii) Bio-semiotic systems
(a) Bio-semiotic systems through which language interfaces with its
biological

environment

(1) Systems that interface with the content plane. These are the systems
and processes of human perception, tactile, auditory, visual, and so on. They ate
themselves semiotic, in that what the organism "sees" is what is construed by the brain

15.1 Language and other semiotic systems

607

into meaning; this then becomes the "input" to the semantic system and is transformed
into higher-order meaning of the linguistic kind.
(2) Systems that interface with the expression plane. These are the
physiological systems and processes of the production and reception of speech: motor
systems of articulation (air stream mechanisms, constrictions and oscillations of the
larynx and other organs, movements of tongue and lips, shaping of the buccal cavity) and
receptor systems of auditory perception in the various regions of the ear. When language
comes to be written, analogous systems come into play for the production and reception
of visual expressions.
(b) The brain as bio-semiotic system. The neural events that constitute the
various interface systems are themselves in the broadest sense semiotic: terms such as
"communication", "exchange of information", that are used to characterize the activities
of the brain are less abstract variants of the concept of "semiotic systems & processes".
At the same time, the neural networks can be thought of as "realizing" the system of
language, in the sense that it is in the brain that language materializes as a process of
the bio-physical world, hi this perspective the relationship between language and the
brain is itself a semiotic one, analogous to that between the content plane and the
expression plane within language itself; and by the same analogy, there is no necessary or
"natural" relationship such that certain parts of the neural network (certain locations
within the brain) are dedicated to language or to any particular subsystem within it. The
analogy is relevant here because it is the fact that language and the perceptual systems
share a common "realization" in neural networks and neural processes that enables
language to function as a dynamic open system, one that persists in time by constantly
being modified through ongoing exchanges with its environment.
15.1.3 The evolution of reality construction

Neurobiologists interpret the evolution of the brain in vertebrate species as the evolution
of the species' potential for constructing reality (Jerison, 1973 (1992); Edelman, 1992).
Evolution brings about a constant change in the organism's relation to its environment;
this relationship becomes increasingly complex, so that the organism's model

of the

environment has to become increasingly complex in its turn. This, it is suggested, is


what "drives" the evolution of more and more complex brain structures.

608

15. Language and the making of meaning

Edelman (1992) relates this evolutionary perspective to the emergence of


consciousness, which he explains in terms of a neurological account of brain structures.
He interprets consciousness in what we can think of as semiotic terms, making a
distinction between primary consciousness, which depends on the constmal of scenes or
images, and higher-order consciousness, which depends on the construal of meaning in
language:
Primary consciousness is the state of being mentally aware of things in the world of
having mental images in the present. But it is not accompanied by any sense of a
person with a past and future. It is what one may presume to be possessed by some
nonlinguistic and nonsemantic animals ... (p. 112) ... Obviously, primary
consciousness must be efficacious if this biological account is correct. ... primary
consciousness helps to abstract and organize complex changes in an environment
involving multiple parallel signals. ... Primary consciousness provides a means of
relating an individual's present input to its acts and past rewards. By presenting a
correlative scene, it provides an adaptive way of directing attention during the
sequencing of complex learning tasks, (p. 121) ... Primary consciousness is required
for the evolution of higher-order consciousness. But it is limited to a small memorial
interval around a time chunk I call the present, (p. 122)
In contrast [to primary consciousness], higher-order consciousness involves the
recognition by a thinking subject of his or her own acts or affections. It embodies a
model of the personal, and of the past and the future as well as the present. It exhibits
direct awareness the noninferential or immediate awareness of mental episodes
without the involvement of sense organs or receptors. It is what we as humans have in
addition to primary consciousness. We are conscious of being conscious, (p. 112) ...
How can the tyranny of this remembered present [i.e. the restrictions of primary
consciousness, MAKH & CM] be broken? The imprecise answer is: By the evolution of
new forms of symbolic memory and new systems serving social communication and
transmission. In its most developed form, this means the evolutionary acquisition of
the capability of language. Inasmuch as human beings are the only species with
language, it also means that higher-order consciousness has flowered in our species, (p.
125) ... Long-term storage of symbolic relations, acquired through interactions with
other individuals of the same species, is critical to the self-concept. The acquisition is
accompanied by the categorization of sentences related to self and nonself and their
connection to events in primary consciousness, (p. 132) ... An inner life, based on the
emergence of language in a speech community, becomes possible. ... Higher-order
consciousness adds socially constructed selfhood to this picture of biological
individuality. The freeing of parts of conscious thought from the constraints of an
immediate present and the increased richness of social communication allow for the
anticipation of future states and for planned behavior, (p. 133-4)
Edelman's account indicates the evolutionary value of the human potential for construing
processes of consciousness, discussed in Chapter 14 above. In contrast with mainstream
cognitive science, Edelman makes language the central resource and relates it to the social
construction of the self.
Our conception of the "meaning base" is fully compatible with this line of
interpretation (cf. Halliday, 1995a): the human brain has evolved in the construction of a
functioning model of "reality". We prefer to conceptualize "reality construction" in terms

15.1 Language and other semiotic systems

609

of construing experience. This is not so much because it avoids metaphysical issues


about the ultimate nature of reality we are prepared to acknowledge a broadly
materialist position; rather, we have three more specific considerations in mind, (i) One is
that what is being construed by the brain is not the environment as such, but the impact
of that environment on the organism and the ongoing material and semiotic exchange
between the two. (ii) The second is that we want to emphasize the evolutionary
perspective, since this allows us to start from what human beings have in common with
other species rather than always insisting on our own uniqueness: when we talk of
"construction of reality" it is almost impossible to avoid taking our own construction as
the norm, whereas parakeets, pythons, and porpoises have very different experiences to
construe different both from each other's and from those of people, (iii) The third point
is that the concept of experience is, or can be, a collective one: experience is something
that is shared by the members of the species construed as a "collective consciousness",
in Durkheim's classic formulation.
This last point needs to be clarified. Edelman's interpretation of higher-order
consciousness referred to above suggests that this form of consciousness (unlike primary
consciousness) is constituted in language. Language is a socio-semiotic system, so it
follows that higher-order consciousness is constituted $0ci0-semiotically; and since
socio-semiotic systems are collective, it follows that higher-order consciousness must
also be collective. Collective consciousness is an attribute of human social groups the
members of a given culture. But we need to distinguish between the consciousness of a
social group and the consciousness of a species, whose collective construal of experience
is codified in the structure of the brain. All human populations have the same brain, and
to that extent all construe experience in the same way. But humans live in social groups,
and their local environments vary one from the other; to that extent, different groups
construe experience in different ways. The significance of this for us is that language is
the resource for both: both what is common to the species as a whole, and what is
specific to the given culture. In the way these two components are construed in the
grammar, we cannot tell them apart. But it is the role of language in the construction of
experience as meaning as shared activity and collaboratively constructed resource
that gives substance to the concept of collective consciousness as an attribute of the
human condition.

610

15. Language and the making of meaning

It has been found that, among the higher primates, those species that live in social
groups have more complex brain structures, other things being equal, than those that live
apart as individuals (Dunbar, 1992, quoted in New Scientist ). This, too, is something
which resonates with our interpretation with language. In our account of the ideation
base, we have stressed the interactive, dialogic nature of the construal of experience. But
we have also stressed that the ideation base is only one component of the total semantic
resource: as well as construing our experience in language, we also use it to enact our
interpersonal relationships. Because we are social animals, there is an added dimension of
meaning for language to cope with (cf. our discussion of the ideation base and the
interaction base in Chapter 9, Section 9.1.2.1). We cannot observe how these resources
have evolved in the history of the human species. But we can observe how they grow in
the development of the human individual. In the final sections (15.2 and 15.3) we will
give a very brief sketch of our topic from an ontogenetic point of view.
15.1.4

Biological,

"individual

social and socio-semiotic

environments

for the

The human individual is at once a biological "individual", a social "individual" and a


socio-semiotic "individual":
as a biological "individual", s/he is an organism, born into a biological
population as a member of the human species.
as a social "individual", s/he is a person, bom into a social group as a member of
society. "Person" is a complex construct; it can be characterized as a
constellation of social roles or personae entering into social networks (see
Argyle et al, 1981, for a discussion from a sociological point of view; Firth,
1950, relates personae to language, and Butt, 1991, develops further his theory
from a systemic-functional point of view, in the light of sociological models of
social networks).
as a socio-semiotic "individual", s/he is a meaner, born into a meaning group as
a member of a speech community. "Meaner" is also a complex construct. For
the socio-semiotic construction of the individual subject, see Thibault (1993);
for the modelling of the individual subject by reference to concept of 'meaning
potential1, see Sefton (1995).

15.1 Language and other semiotic systems

611

These different levels of individuality map onto one another: a meaner is a person, and
a person is a biological organism. But the mappings are complex; and at each level an
individual lives in different environments in different networks of relations. Lemke
(1995, in particular Ch. 5) provides an insightful "dissection" of the "notion of the
individual human subject1':
The biological organism and the social persona are profoundly different social
constructions. The different systems of social practices, including discourse practices,
through which these two notions are constituted, have their meanings and are made use
of, are radically incommensurable. The biological notion of a human organism as an
identifiable individual unit of analysis depends on the specific scientific practices we
use to construct the identity, the boundedness, the integrity, and the continuity across
interactions of this unit. ... The social-biographical person is also an individual in so
far as we construct its identity, boundedness, integrity and continuity, but the social
practices and discourses we deploy in these constructions are quite different. We define
the social person in terms of social interactions, social roles, socially and culturally
meaningful behavior patterns. ... We obtain the common sense notion of a human
individual only by a complex process of conflation: mapping the social-biographical
person onto the physical-biological organism. This, too, is accomplished by our
cultural patterns of discourse, and the associated actional practices. Because the
classical notion of a human individual is constructed in this way, if we no longer make
the traditional metaphysical presumption of a single 'real object to which each of these
discursive systems "refers1 or 'on which' it acts, there is no longer any reason to
suppose that 'the individual1 constructed by each of these systems of practices coincides
with those constructed through the others, (p. 81)

Lemke then goes on to explore material, social and semiotic constructions of "individual"
and to suggest where we need to depart from the received notion of individuality (cf. Ch.
14 for the mental aspect of this notion). Towards the end of his discussion he writes:
We need some latter-day Jean Piaget to write The Child's Construction of the Sense of
Self. It should tell us how the child (and later the adult), enmeshed in semiotically and
materially mediated interactions with other members of a community and with the
material environment, progressively recapitulates (always to some degree
individuating) a trajectory of development that leads to our constructing the sense of a
Self, a Self that looks out through the windows of the eyes, that initiates motor actions
by 'will' and 'intention', that 'feels' the sensations which impinge on a body in which it
sits, but of which it is not truly a physical part. It will tell the story of how we are
taught to think of ourselves as Selves.

This is part of the process whereby children learn how to mean,

15.2 Construing experience: ontogenetic perspective


15.2.1 Protolangtiage and language
A human infant is a social being from birth (cf. Trevarthen, 1987). Newborn children can
exchange attention with their mothers, addressing them and recognizing that they arc
being addressed by them; the infant's whole body is actively involved in the exchange.

612

15. Language and the making of meaning

This is "pre-language" ("pie-meaning", even "pre-text"); but it is not language no


distinction is yet being made between symbolic and non-symbolic acts. Then, as they
become aware of themselves and their environment, children feel a tension building up
between two facets of their experience: between what they perceive as happening "out
there" and what is happening "in here", within their own borders so to speak. We can
watch babies of around 3 - 4 months struggling to reconcile these complex sensations:
they can see a coloured object, reach out, and grasp it and pull it towards them. The inner
and the outer forms of this experience have to be brought into line; in order to achieve
this, children begin to act in a new, distinctively symbolic mode. A typical example of
such an "act of meaning" is the high-pitched squeak a child of around 5 months may
produce when some commotion takes place that has to be assimilated. Adults interpret
these proto-signs as a demand for explanation: "Yes, that's a bus starting up. Isn't it
noisy!" Thus meaning arises out of the impact between the material and the conscious as
the two facets of a child's ongoing experience.
Children gradually build up an inventory of such proto-signs, and towards the end of
the first year the signs begin to form systems, sets of contrasting terms in particular
proto-semantic domains or micro-functions: typically, the instrumental (e.g. 'I want/1
don't want'), regulatory (e.g. Do that!/ Do that!!'), interactional (e.g. Tm here/ where
are you?'), and personal domains (e.g. 'I like that/ I'm curious about that'). These already
foreshadow the semantic motifs of the adult language, the experiential and interpersonal
metafunctions, although they are not in any direct correspondence with them; thus the
"personal" signs expressing curiosity, or pleasure/ displeasure, constitute the beginning
of the semiotic exploration of experience and open the way to naming and classifying
phenomena, while the interactional signs are the ones whereby a child enacts social
relationships with caregivers and others who are close (Halliday, 1975; 1984b). Here we
see the earliest context for the later emergence of types of process within the grammar
(Halliday, 1991). But the immediate significance of the protolanguage is that by acting
semiotically in these particular contexts children construe the fundamental distinction
between "self* and "other", and the further distinction of "other" into persons and objects
(cf. the discussion and figure in Halliday, 1978b). The consciousness of the self arises at

15.2 Construing experience: ontogenetic perspective

613

the intersection of the various semiotic roles defined by each of these systems2 as well
as, of course, from awareness of being one interactant in the general dialogic process
(Halliday, 1991).
The protolanguage is typically associated with the stage of crawling, when children
are mobile, but not yet walking and running: typically about 0;8 - 1;4, but with wide
variation around these times. The elements of the protolanguage are "signs" (that is,
content/ expression pairs); they are thus formally identical with the semiotic resources of
higher mammals (primates and cetaceans) but with one important difference: the signs
of other species become codified as the form of communication among adults, whereas
those of human children are transitional to a system of a different kind, and hence do
not stabilize into a settled pattern but are constantly shifting on both semiotic planes.3
How, and why, do children discard their functioning protolanguage and move on to
"language" in its adult form? To take up the "why?" first: because the protolanguage sets
limits on both dimensions of meaning. You can converse in it, but you cannot build up a
dialogue: that is, it allows exchange of meaning, but it precludes any form of an
interpersonal dynamic, in which meanings expand on the basis of what went before. You
can point with it, but you cannot refer that is, it allows focus on an object, but it
precludes any form of ideational systematic, in which phenomena are construed as
configurations and in taxonomies. For these to be possible you need a semiotic of a
different kind, one that allows for a purely abstract level of representation "in between"
the two faces of the sign, To put this another way (as we did at the beginning of the
book), the sign has to be deconstructed so that, instead of content interfacing directly with

There is a sense in which these roles anticipate the functions in the transitivity structure of
the clause: proto-Beneficiary (instrumental), proto-Agent (regulatory), proto-Carrier
(interactional), and proto-Senser (personal).

thashewas physican).Whenhe as

It is clear that animals such as chimpanzees and gorillas, whales and dolphins, communicate
with signs in this denned sense, and it appears that these are in some way organized into
sign systems. It is possible that some of these species have already moved towards a
human-like, stratified form of language; but this has not yet been demonstrated, as far as we
know, by any of the available evidence.
An interesting case is that of domesticated cats and dogs. They communicate with signs to
their human companions, but apparently not, or only very rarely, to each other. lhe
affinity often felt between such pets and small children is not merely one of a shared
material plane (they are more like each other in size) but also one of a shared semiotic
plane: they share a common form of language.

614

15. Language and the making of meaning

expression, the relationship is mediated by a systematic organization of form (a


lexicogrammar). In other words, the semiotic has to become stratified.
Children make the transition from protolanguage to language typically in the second
half of the second year of life. The transition has been described in detail elsewhere, based
on intensive observations of individual children (see especially Painter, 1984, 1989;
Oldenburg, 1987); we may assume that in general terms they are recapitulating the
phylogenetic evolution of language, although of course we can only speculate about the
way that evolution took place (it is important to say explicitly that all human languages
known today are equally far removed from that phase in our semiotic history). During
that stage they learn to construe elements and figures, and in this way "semanticize" both
the construction of experience and the enactment of interpersonal relations. In terms of
the grammar, they learn to form groups and clauses, and to select systemic options
simultaneously in transitivity and in mood.
Since our concern in this book is with the ideation base, we have not been
considering interpersonal aspects of meaning, and we have not put major emphasis on
dialogic patterns in discourse. So in this final glance at ontogenesis we should foreground
very clearly the fact that meaning is an interactive process and that children learning to
mean construe their semiotic resources through dialogue. This is not simply an optional
extra, something that makes the learning processes easier; it is an inherent property of
semiosis itself. Semiotic systems are social systems, and meaning arises in shared social
consciousness; this is evident already in the protolanguage, when infants depend on being
treated as communicating beings, and those within their "meaning group" are tracking
them unconsciously creating the language along with them (see Halliday, 1979b). We
find this manifested also in the forms of discourse, in the way children participate in
constructing narratives of shared experience (see Halliday, 1975: 112; Painter, 1989: 55).
When we talk of "construing experience" as the metafunctional realm of the ideation base,
we are referring to the shared experience of the group, the culture and the species; it is by
means of dialogue that children gain access to this shared experience and are enabled to
construe their own experience with reference to it. And the dialogic nature of discourse
serves the child also as a metaphor, as the semiotic manifestation of the social conditions
of human existence.

15.2 Construing experience: ontogenetic perspective

615

15.2.2 Generalization, abstraction and metaphor

A prerequisite for the semiotic construal of experience is generalization: the move


from "proper" to "common" as the basic principle of referring. The protolanguage, as
already remarked, is non-referring; children move into reference by gradually deconstruing
the proto-linguistic sign in a sequence of steps such as 'I want Mummy to ...', 'I want
Mummy!', 'Where is Mummy?1, Mummy' (see Hailiday, 1992). The sign has now
become a word, functioning as a proper name. Typically one or two other signs have
been deconstrued at the same time in similar fashion, e.g. 'I want my (toy) bird!', 'Where
is my bird?', 'My bird'; and by a further step these then become common names l)ird(s)1.
The child has now learnt to name a class of things; this then opens the way (i) to
constructing hierarchies of classes a 'pigeon' is a kind of "bird1, and so on, and (ii) to
naming other kinds of element, processes and qualities, which can be construed only as
"common" terms. Since these other elements have distinct and complementary functions
it becomes possible to combine them into organic structures, as complex elements or as
figures, such as 'blue bird', *birds flying', 'tiny bird flew away1. The resources are now in
place for construing experience in lexicogrammatical terms.
This principle of generalization that is, naming general classes rather than specific
individuals is what makes it possible to construct an ideation base. When they have
reached this stage, children can make the transition from protolanguage to mother tongue,
building up figures and sequences of figures, and simultaneously structuring these as
moves in dialogic exchanges (question, statement, etc. the interaction base), and as
messages or quanta of information (the text base). In other words, they learn "how to
mean" according to the metafunctional principle of adult human semiosis. But there are
two further developments to come before the ideation base can take the form it has to take
if it is to produce discourse of the kind we have been assuming throughout the book (and
illustrating with weather reports and recipes); and there is some elapse of time before
children take these further steps. The first of these developments is abstractness; the
second is (grammatical) metaphor.
General terms are not necessarily abstract; a bird is no more abstract than a pigeon.
But some words have referents that are purely abstract words like cost and clue and
habit and tend and strange; they are construing some aspect of our experience, but there
is no concrete thing or process with which they can be identified. Small children simply

616

15. Language and the making of meaning

ignore them, but by the age of about four or five they begin to cope with abstract
meanings; in literate societies, this is the time we consider that children are "ready to start
school", no doubt because you have to cope with abstractness in meaning in order to be
able to learn to read and write (cf. wordings like spell stand for, beginning of a sentence).
But it is not only the written medium; rather it is the whole world of educational
knowledge that demands such abstractness in meaning. Consider examples taken from
primary text books such as Some animals rely on their great speed to escape from danger,
or The time taken by the earth to rotate once on its own axis is a day.
What happens here is that experience is being reconstruid in order to build up a
form of knowledge that is systematically organized and explicit. Children already know
that animals run away because they're frightened, and that the sun goes round the earth
once in a day; but they have to learn these things over again in a new, more abstract
semiotic frame. When Nigel was 4; 11 he had the following conversation with his father:
Nigel: Why does as plasticine gets longer it gets thinner? (sic)
Father: That's a very good question. Why does it?
Nigel Because more of it is getting used up.
Father (doubtfully): Well . . .
Nigel: Because more of it

is getting used up to make it

longer,

that's why; and so it goes thinner.

Here Nigel displayed a clear understanding of the principle of conservation. Some years
later in school he was studying the following passage:
Put a label on each [container] to show two things:
(a) The quantity it holds.
(b) What fraction of a litre in it. (sic)
Put all

that measure 1 litre together. Some will be tall,

some

short, some rectangular, some cylindrical for milk or drinks, some


wine bottles or carafes. But they all

contain 1 LITRE. A litre is

a litre, whether long/ round or square. . . . So all kinds of shapes


can be made to have the same capacity.
K. Penet & G. Fiddes, New Primary Maths, Book 5. Sydney: School Projects, 19687
1977. (p. 71)

15.2 Construing experience: ontogenetic perspective

617

Note the difference in the way the grammar construes the same domain of experience, first
in its everyday commonsense form, as spoken by a child, and then in its reconstruction as
educational knowledge.
But there is a further transformation still to come, when experience is once again
reconstrued, this time as technical knowledge. This reconstmal too is institutionalized,
in the transition from primary to secondary education: when children move into secondary
school, as adolescents, they learn to organize their experience according to the disciplines
mathematics, science (chemistry, physics, biology), geography, history, and so on.
Semiotically, the critical factor is that of metaphor; the semiotic bonds that had enabled
the child to learn the mother tongue in the first place, bonds between figures and their
elements on the one hand and clauses and their transitivity functions on the other, ate
systematically (and more or less ceremonially!) untied. The categories of experience aie
deconstrued, to be recategorized over the remaining years of schooling in the
"objectifying" framework of grammatical metaphor. We have described these effects in
Part II above (Chapter 6) and will not repeat the exposition here (cf. Derewianka, 1995,
for a longitudinal study of one child). By the time children reach the llth and 12th year of
education their experience is being construed in terms such as these:
Every similarity transformation, if not a translation, reflection,
rotation, or enlargement, is

the product of two or irore such

transformations.
A. McMullen & J.L. Williams, On Course Mathematics 4, Advanced Level. Melbourne:
Macmillan, 19657 1975. (p. 153)
What would be the order of magnitude of the moment of inertia of
the Earth about its axis of rotation?
E.D. Gardiner & B.L. McKittrick, Problems in Physics. Sydney: McGraw-Hill, 19697
1985. (p. 58)
The elements are processes and qualities that have been metaphorically reconstrued to
become participants: rotation, magnitude, enlargement, and so on; together with the
relation of identity construed as a process by the verb be. When our adolescents' ideation

618

15. Language and the making of meaning

base comes to accommodate a meaning potential of this technicalized kind, we consider


that they have reached semiotic maturity.
The developmental dynamic of "generalization abstractness metaphor" provides
the semiotic energy for the grammar, enabling it to serve as the powerhouse for
construing experience in the form of scientific knowledge. Presented in this very sketchy
fashion the movement may seem catastrophic and discontinuous; but this is misleading.
Rather, it is a steady progression, marked by three periods of more rapid development at
the transitions: from protolanguage to language (generalization, associated with bipedal
motion), from commonsense (spoken) language to written language (abstractness: the
move into primary school), and from non-specialized written language to technical
language (metaphor the move into secondary school). There is a clear grammatical and
semantic continuity between the various versions of experience, which we can bnng out
by analysing the grammar of particular instances (such as those cited above). At the same
time, the ontogenetic perspective shows that in fact our experience is being ongoingly
reconstruedandrecategorizedas we grow from infancy to maturity. This is the outcome
of processes taking place in human history evolutionary events that are at once both
material and semiotic, and that cannot be reduced to either purely physical processes
driven by technology or purely discursive processes driven by ideology. There is no point
in asking whether the ideation base of our technologized natural languages necessarily had
to evolve the way it did. But it is extremely pertinent to ask, given the enormous
demands now being made on both the material and the semiotic resources of the human
species, what the options are for the way it may evolve in future.

References
Allen, J. 1987. Natural language understanding. Menlo Park, CA: The Benjamins/
Cummings Publishing Company.
Amsler, R. 1981. A taxonomy for English nouns and verbs. Proceedings of the 19th
Annual Meeting of the Association for Computational Linguistics.
Anderson, J. 1983. The architecture of cognition. Cambridge, Mass.: Harvard University
Press.
Argyle, M., A. Furnham & G.A. Graham. 1981. Social situations. Cambridge:
Cambridge University Press.
Bach, E. 1968. Nouns and noun phrases. E. Bach & R. Harms (eds.).
Bach, E. 1989. Informal lectures on formal semantics. Albany, NY: State University of
New York Press. [323J
Bach, E. & R. Harms (eds.). 1968. Universals in linguistic theory. New York: Holt,
Rinehart and Winston.
Ballim. A. & Y. Wilks. 1991. Artificial believers: the ascription of belief. Hillsdale, NJ:
Lawrence Erlbaum.
Ban*, A. & E. Feigenbaum. 1981. The handbook of artificial intelligence. Menlo Park,
CA: Addison-Wesley. Barwise, J. 1988. On the circumstantial relation between
meaning and content Eco et al.
Barwise, J. & J. Perry. 1983. Situations and attitudes. Cambridge, Mass.: The MIT
Press. Bateman, J.A. 1985. Utterances in context: towards a systemic theory of the
intersubjective achievement of discourse. Ph.D. Thesis, Department of Artificial
Intelligence / School of Epistemics, University of Edinburgh.
Bateman, J.A. 1989, Dynamic systemic-functional grammar: a new frontier. Word 40.12 (Systems, structures and discourse: selected papers from the fifteenth international
systemic congress).
Bateman, J.A. 1996. KPML development environment multilingual linguistic
resource development and sentence generation. Manual for release 0.9, March 1996.
IPSI7 GMD, Darmstadt, Germany.
Bateman, J.A., Kasper, R., Moore, J. and Whitney, R. 1990. A general organization of
knowledge for natural language processing: the Penman Upper Model. Research
report, Information Sciences Institute, University of Southern California.
Bateman, J.A. & C.M.I.M. Matthiessen. 1993. The text base in generation. K. Hao, H.
Bluhme & R. Li (eds.), Proceedings of the international conference on texts and
language research, Xi'an, 29-31 March 1989. Xi'an: Xi'an Jiaotong University Press.
Bateman, J.A. & S. Momma. 1991. The nondirectional representation of Systemic
Functional grammars and semantics as Typed Feature Structures. Technical report,
GMD/ Institut fr Integrierte Publikatioas- und Infonnatioassysteme, Darmstadt and
Institut fr Maschinelle Sprachverarbeitung, Universitat Stuttgart
Bateman, J.A., C.M.I.M. Matthiessen, K. Nanri and L. Zeng. 1991. The rapid
prototyping of natural language generation components: an application of functional

620

References

typology. Proceedings of the 12th international conference on artificial intelligence,


Sydney, 24-30 August 1991. San Mateo, CA: Morgan Kaufman.
Bateman, J.A., CM.LM. Matthiesssen & L. Zeng. forthc. A general architecture of
multilingual resources for natural language processing. MS, Macquarie University &
GMD/ IPSI.
Becker, J.D. 1975. The phrasal lexicon. R. Schank & B. Nash-Webber (d.), Theoretical
issues in natural language processing, Association for Computational Linguistics.
Benson, JX>. & W.S. Greaves (eds.). 1985. Systemic perspectives on discourse: selected
papers from the ninth international systemic workshop, Volume L Norwood, NJ:
Ablex.
Benson, JJD. & W.S. Greaves (eds.). 1988. Systemic functional approaches to discourse:
aelected papers from the twelfth international systemic workshop. Norwood, NJ:
Ablex.
Benson, J.D., M.J. Cummings, & W.S. Greaves (eds.). 1988. Linguistics in a systemic
perspective. Amsterdam: Benjamins.
Berger, P.L. & T. Luckmann. 1966. The social construction of reality. Garden City, NY:
Doubleday.
Berlin, B. 1972. Speculations on the growth of ethnobotanical nomenclature. Language
in Society 1.
Berlin, B., D. Breedlove, & P. Raven. 1973. General principles of classification and
nomenclature in folk biology. American Anthropologist, 75.
Berlin, B. & P, Kay. 1969. Basic color terms: their universality and evolution. Berkeley,
CA: University of California Press.
Berry, M. 1981. Systemic linguistics and discourse analysis: a multi-layered approach to
exchange structure. M. Coulthard & M. Montgomery (eds.), Studies in discourse
analysis. London: Routledge & Kegan Paul. Berry, M., C.S. Butler, R.P. Fawcett &
G. Huang (eds.). 1996. Meaning and form: systemic functional interpretations.
Meaning and choice in language: studies for Michael Halliday. (Volume LVn in the
series Advances in Discourse Processes.) Norwood, NJ: Ablex.
Biagi, M.L.A. 1995. Diacronia dei linguaggi scientifici. R. R. Favretti (d.), Proceedings
of the international conference "Languages of Science", Bologna, 25-27 October
1995. In press.
Bickhart, M.H. & RJL. Campbell. 1992. Some fundamental questions concerning
language studies: with a focus on categora! grammars and model-theoretic possible
world semantics. Journal of Pragmatics 17.
Bohm, D. 1979. Wholeness and the implicate order. London: Routledge & Kegan Paul.
Boole, G. 1854. An investigation of the laws of thought: on which are founded the
mathematical theories of logic and probability. London: Macmillan. [New York:
Dover. 1958.]
Brachman, RJ. 1978. A structural paradigm for representing knowledge. BBN Report
No. 3605, Bolt Beranek and Newman, Inc. Cambridge, MA.

References

621

Brachman, RJ. 1979. On the epistemological status of semantic networks. N.V. Findler
(d.), Associative networks: representation and use of knowledge by computers. New
York: Academic Press. Reprinted in Brachman & Levesque (1985).
Brachman, R.J. & HJ. Levesque. (eds.) 1985. Readings in knowledge representation.
Los Altos, CA: Morgan Kaufman.
Brachman, RJ., R. Ftkes & HJ. Levesque. 1983. KRYPTON: A functional approach to
knowledge representation. IEEE Computer, 16.10. Reprinted in Brachman &
Levesque (1985).
Brachman, RJ. & J. Schmolze. 1985. An overview of the KL-ONE knowledge
representation system. Cognitive Science, 9.2.
Brew, C. 1991. Systemic classification and its efficiency. Computational Linguistics
17.4.
Brill, D. (1991). LOOM: reference manual Technical report, Information Sciences
Institute, University of Southern California.
Bundy, A. (ed), 1986. Catalogue of artificial intelligence tools. Second, revised edition.
Berlin & New York: Springer-Verlag.
Bursill-Hall, G.L. 1971. Speculative grammars of the Middle Ages: the doctrine of partes
or aliones of the Modistae. The Hague & Paris: Mouton.
Butt, DJ. 1983. Semantic "drift" in verbal art. Australian Review of Linguistics 6.1.
Butt, DJ. 1987. Randomness, order and the latent patterning of text. D.I. Birch &
M.L.M. OTooie (eds.), Functions of style. London: Pinter.
Butt, DJ. 1991. Some basic tools in a linguistic approach to personality: a Firthian
concept of social process. F. Christie (d.), Literacy in social processes: papers from
the Inaugural Australian Systemic Functional Linguistics Conference, Deakin
University, January 1990. Darwin: Centre for Studies of Language in Education,
Northern Territory University.
Caffarei, A. 1990. Mediating between grammar and context: a bi-stratal exploration of
the semantics of French tense. B.A. Honours thesis, Department of Linguistics,
University of Sydney.
Caffarei, A. 1992. Context projected onto semantics and the consequences for
grammatical selection. Language Sciences, 14.4.
Cann, R. 1993, Formal semantics: an introduction. Cambridge: Cambridge University
Press. (Cambridge Textbooks in Linguistics.)
Chafe, W.L. 1970. Meaning and the structure of language. Chicago: Chicago University
Press.
Chafe, W.L. 1979. The flow of thought and the flow of language. T. Givn (d.), Syntax
and Semantics 12: Discourse and Syntax. New York: Academic Press.
Chafe, W.L. 1987. Cognitive constraints on information flow. R. Tomlin (d.),
Coherence and grounding in discourse. Amsterdam: Benjamins.
Clark, H.H. 1975. Bridging. TlNLAP-1. Reprinted in P.N. Johnson-Laird & P. Wason
(eds.). 1977. Thinking: readings in cognitive science. Cambridge: Cambridge
University Press.

622

References

Conklin, H. 1962. Lexicographical treatment of folk taxonomies. F. Householder & S.


Saporta (eds.), Problems in lexicography. Bloomington, Ind.: Indiana University
Research Center in Anthropology, Folklore, and Linguistics. (Publication 21.)
Cook, W.A, 1977. Case grammar: development of the matrix model (1970 - 1978).
Washington, DC: Georgetown University Press.

Cross, M. 1991. Choice in text: a systemic approach to computer modelling of variant


text production. Ph.D. thesis, Macquarie University.
Cross, M. 1992. Choice in lexis: computer generation of lexis as most delicate grammar.
Language Sciences, 14.4.
Cross, M. 1993. Collocation in computer modelling of lexis as most delicate grammar.
M. Ghadessy (ed), Register analysis: theory and practice. London: Pinter.
Cruse, D.A. 1986. Lexical semantics. Cambridge: Cambridge University Press.
(Cambridge Textbooks in Linguistics.)
D'Andrade, R. 1987. A folk model of the mind. Holland & Quinn (eds).

Dahlgren, K. 1988. Naive semantics for natural language understanding. Dordrecht:


Kluwer.
Davey, A. 1978. Discourse production: a computer model of some aspects of a speaker.
Edinburgh: Edinburgh University Press.
Davidse, K. 1991. Categories of experiential grammar. Ph.D. thesis, Catholic University
of Leu ven
Davidse, K. 1992a, Existential constructions: a systemic perspective. Leuvense Bijdragen
81.
Davidse, K. 1992b. A semiouc approach to relational clauses. Occasional Papers in
Systemic Linguistics 6.
Davidse, K. 1992c. Transitive/ ergative: the Janus-headed grammar of actions and events.
M. Davies & L. Ravelli (eds.).
Davidse, K. 1996a, Turning grammar on itself: identifying clauses in linguistic
discourse. Berry et al. (eds.).
Davidse, K. 1996b. Ditransitivity and possession. R. Hasan, C. Cloran & D. Butt (eds.).
Davidson, D. 1967. Truth and meaning. Synthse 17.
Davidson, D. 1970. Semantics for natural languages. Reprinted in D. Davidson & G.
Hannan (eds.), The logic of grammar. Encino, CA: Dickenson, 1975. Davies, M. &
L. Ravelli (eds). 1992. Advances in systemic linguistics. London: Pinter.

de Beaugrande, R. 1980. Text, discourse, and process: toward a multidisciplinary science


of texts. Norwood, NJ: Ablex.
de Beaugrande, R. 1994. Function and form in language theory and research: the tide is
turning. Functions of Language 1.2.

de Beaugrande, R. in press. New foundations for a science of text and discourse.


Norwood NJ: Ablex.

References

623

Dennett, D. 1981. Three kinds of intentional psychology. Healy (ed), Reduction, time
and reality. Cambridge: Cambridge University Press. Reprinted in J.L. Gaifield (ed),
Foundations of cognitive science: the essential readings. New York: Paragon House,
1990.
Derewianka, B. 1995. Language development in the transition from childhood to
adolescence: the role of grammatical metaphor. Ph X). thesis, Macquarie University.
Dik, S. 1978. Functional grammar. Amsterdam: North-Holland.
Dik, S. 1986. Linguistically motivated knowledge reprsentation. Working Papers on
Functional Grammar 9. Amsterdam.
Dik, S. 1987. Generating answers from a linguistically coded knowledge base. Kempen
(d.).
Downing, P. 1980. Factors influencing lexical choice in narratives. W. Chafe (d.),
Cognitive, cultural and linguistic aspects of narrative production. Norwood, N.J.:
Ablex.
Dowty, D., R. Wall & S. Peters. 1981. Introduction to Montague semantics. Dordrecht:
Reidel.
Dowty, D. 1979. Word meaning and Montague grammar. Dordrecht: Reidel.
Dunbar, R. 1992. Quoted in New Scientist 1850, 5 Dec, 1992. (Supplement Number 4:
Secret Life of the Brain.)
Eco, U., M. Santambrogio & P. Violi. (eds.) 1988. Meaning and mental representations.
Bloomington, Ind.: Indiana University Press.
EdelmaD, G. 1992. Bright air, brilliant fire: on the matter of the mind. New York: Basic
Books.
Eggins, S. 1990. Conversational structure: a systemic-functional analysis of
interpersonal and logical meaning in multiparty sustained talk. PhD thesis,
Department of Linguistics, University of Sydney.
Eggins, S. 1994. An introduction to systemic functional linguistics. London: Pinter.
Eggins, S., P. Wigneil & J.R. Martin. 1993. The discourse of history: distancing the
recoverable past M. Ghadessy (ed), Register analysis. Theory and practice. London:
Pinter.
Ellis, J.M. 1993. Language, thought and logic. Evanston, 111.: Northwestern University
Press.
Enrich, V. 1987. The generation of tense. G. Kempen (d.).
Fang, Y., E. McDonald & M. Cheng. 1995. On Tbeme in Chinese: from clause to
discourse. R. Hasan & P.H. Ries (eds.), On subject and theme: a discourse functional
perspective. Amsterdam: Benjamins.
Fawcett, R.P. 1980. Cognitive linguistics and social interaction: towards an integrated
model of a systemic functional grammar and the other components of an interacting
mind. Heidelberg: Julius Gross; Exeter: Exeter University Press.

624

References

Fawcett, R.P. 1981. Generating a sentence in systemic-functional grammar. M.A.K.


Halliday & J.R. Martin (eds.). Fawcett, R.P. 1984. System networks, codes and
knowledge of the universe. R.P. Fawcett et al (eds).
Fawcett, R.P. 1987. The semantics of clause and verb for relational processes in English.
M.A.K. Halliday & R.P. Fawcett (eds.).
Fawcett, R.P. 1988a. The English personal pronouns: an exercise in linguistic theory.
Benson, Cummings & Greaves (eds.).
Fawcett, R.P. 19885. Language generation as choice in social interaction. M. Zock & G.
Sabah (eds.), Advances in natural language generation. London: Pinter.
Fawcett, R.P. 1994. A generationist approach to grammar reversibility in natural
language processing. T. Strzalkowski (d.), Reversible grammar in natural language
processing. Dordrecht: Kluwer.
Fawcett, R.P., G.H. Tucker & Y.Q. Lin. 1992. The COMMUNAL Project: how to get
from semantics to syntax. Proceedings of the 14th International Conference on
Computational Linguistics.
Fawcett, R.P., M.A.K, Halliday, S.M. Lamb & A. Makkai. (eds). 1984. The semiotics
of culture and language. 2 Volumes. London: Frances Pinter.
Fawcett, R.P. & D.J. Young. 1988. New developments in systemic linguistics. Volume
2: theory and application. London: Pinter.
Fillmore, C. 1968. The case for case. Bach & Harms (eds.).
Firth, J.R. 1950. Personality and language in society. Sociological Review 42.2.
Reprinted in Firth (1957).
Firth, J.R. 1956. Linguistic analysis and translation. For Roman Jakobson: essays on
the occasion of his sixtieth birthday. The Hague: Mouton. Reprinted in F.R. Palmer
(d.), Selected papers J.R. Firth 1952 - 7959. London: Longman, 1968.
Firth, J.R. 1957. Papers in linguistics 1934-1951. London: Oxford University Press.
Fischer-J0rgensen, E. 1975. Trends in phonological theory: a historical introduction.
Copenhagen: Akademisk Forlag.
Fodor, J. 1975. The language of thought. Cambridge, Mass: Harvard University Press.
Foley, W.A. 1986. The Papuan languages of New Guinea. Cambridge: Cambridge
University Press (Cambridge Language Surveys).
Foley, W.A. & R. Van Valin. 1984. Functional syntax, and universal grammar.
Cambridge: Cambridge University Press.
Fox, B. 1987. Discourse structure and anaphora: written and conversational English.
Cambridge: Cambridge University Press.
Frake, C. 1962. The ethnographic study of cognitive systems. T. Gladwin & W.
Sturtevant (eds), Anthropology and human behavior. Washington: Anthropological
Society of Washington. Frege, G. 1879. Begriffsschrift> eine der arithmetischen
nachgebildete Formelsprache des reinen Denkens. Halle: Nebert

625

References

Frege, G. 1892. Uber Sinn und Bedeutung. Zeitschrift


philosophische Kritik 100.

fur Philosophie und

Fries, P.M. 1981. On the status of theme in English; arguments from discourse. Forum
Linguisticum 6.1. Reprinted in J. Petfi & E. Sozer (eds.), 1983, Micro and macro
connexity of texts. Hamburg: Helmut Buske (Papers in Textlinguistics 45).
Fries, P.M. 1992. Sttucture of information in written English text. Language Sciences
14.4.
Fries, P.H. 1995. Themes, methods of development, and texts. R. Hasan & P.H. Fries
(eds.), On subject and theme: a discourse functional perspective. Amsterdam:
Benjamins.
Fries, P.H. in press. Towards a discussion of the flow of information in a written
English text. MJ. Gregory, MJ. Gumming s & J. Copeland (eds.), Relations and
functions in and around language.
Givn, T. 1979. On understanding grammar. New Yoric: Academic Press.
Givn, T. 1980. The binding hierarchy and the typology of complements. Studies in
Language 4.3.
Goldman, N. 1974. Computer generation of natural language from a deep conceptual
base. PhD. Thesis, Stanford University.
Goodenough, W. 1956. Componential analysis and the study of meaning. Language 32.
Grosz, B. 1978. Discourse knowledge. D. Walker (d.), Understanding spoken language.
Amsterdam: North Holland.
Gruber, J. 1965. Studies in lexical relations. Ph.D. thesis, Massachusets Institute of
Technology.
Gruber, J. 1976. Lexical structures in syntax and semantics. Amsterdam: North-Holland.
Haack, S. 1978. Philosophy of logics. Cambridge: Cambridge University Press.
Haiman, J. (ed.) 1985. Iconicity in syntax. Amsterdam: Benjamins.
Halliday, M.A.K. 1956. Grammatical categories in Modern Chinese. Transactions of the
Philological Society. Abridged version reprinted in Halliday (1976).
Halliday, M.A.K. 1959. The Language of the Chinese "Secret History of the Mongols".
Oxford: Blackwell (Publications of the Philological Society 17.)
Halliday, M.A.K. 1961. Categories of the theory of grammar. Word 17.3. Abridged
version reprinted in HaUiday (1976).
Halliday, M.A.K. 1963a, The tones of English. Archivum Linguisticum 15.1.
Halliday, M.A.K. 1963b. Intonation in English grammar. Transactions of the
Philological Society.
Halliday, M.A.K. 1966. Some notes on "deep" grammar. Journal of Linguistics 2.1.
Abridged version reprinted in Halliday (1976).

626

References

Halliday, M.A.K. 1967. Intonation and grammar in British English. The Hague: Mouton
(Series Practica XLVni).
Halliday, M.A.K. 1967/8. Notes on transitivity and theme in English. Journal of
Linguistics 3.1, 3.2, & 4.2.
Halliday, M.A.K. 1971. Linguistic function and literary style: an enquiry into the
language of William Golding's The inheritors. In S. Chatman (ed), Literary style: a
symposium. New York: Oxford University Press. Reprinted in Halliday (1973).
Halliday, M.A.K. 1973. Explorations in the Junctions of language. London: Edward
Arnold.
Halliday, M.A.K. 1975. Learning how to mean. London: Edward Arnold.
Halliday, M.A.K. 1976. System and function in language: selected papers, d. G.R.
Kress. London: Oxford University Press.
Halliday, M.A.K. 1977. Aims and perspectives in linguistics. Applied Linguistics
Association of Australia (Occasional Papers Number 1).
Halliday, M.A.K. 1978a. Language as social semiotic: the social interpretation of
language and meaning. London: Edward Arnold
Halliday, M.A.K. 1978b. Meaning and the construction of reality in early childhood. In
HJ-. Pick Jr & E. Saltzman (eds), Modes of perceiving and processing information.
Hillsdale, NJ: Erlbaum.
Halliday, M.A.K. 1979a. Modes of meaning and modes of expression. D.J. Allerton, E.
Carney, and D. Holdcroft (eds), Function and context in linguistic analysis: essays
offered to William Haas. Cambridge: Cambridge University Press.
Halliday, M.A.K. 1979b. One child's protolanguage. M. Bullowa (d.), Before speech:
the beginnings of interpersonal communication. Cambridge: Cambridge University
Press.
Halliday, M.A.K. 1984a. On the ineffability of grammatical categories. A. Manning, P.
Martin & K. McCalla (eds.), The Tenth LACUS Forum 1983. Columbia: Hornbeam
Press. Reprinted in Benson, Cummings & Greaves (eds.).
Halliday, M.A.K. 1984b. Language as code and language as behaviour: a systemicfunctional interpretation of the nature and ontogenesis of dialogue. Fawcett et al (eds).
Halliday, M.A.K. 1984c. Listening to Nigel Department of Linguistics, University of
Sydney.
Halliday, M.A.K. 1985. An introduction to functional grammar. London: Edward Arnold.
Second, revised edition 1994.
Halliday, M.A.K. 1985/9. Spoken and minen language. Geelong, Vic: Deakin
University Press. Reprinted London: Oxford University Press.
Halliday, M.A.K. 1987. Language and the order of nature. D. Attridge, A. Durant, N.
Fabb & C. MacCabe (eds), The Linguistics of Writing. Manchester Manchester
University Press. Reprinted in Halliday & Martin (1993).

627

References

Halliday, M.A.K. 1988. On the language of physical science. M. Ghadessy (d.),


Registers of written English: situational factors and linguistic features. London:
Pinter. Reprinted in Halliday & Martin (1993).
Halliday, M.A.K. 1990. New ways of meaning: a challenge to applied linguistics.
Journal of Applied Linguistics (Greek Applied Linguistics Association) 6. Reprinted
in Piitz (d.),
Halliday, M.A.K. 1991. The place of dialogue in children's construction of meaning. S,
Stati, E. Weigand & F. Hundsnurscher (eds.), Dialoganalyse HI: Referate der 3.
Arbeitstagung, Bologna 1990. Tubingen: Niemeyer (Beitrage zur Dialogforschung,
Bd. D.Teill.
Halliday, M.A.K. 1992. How do you mean? Davies & Ravelli (eds.).
Halliday, M.A.K. 1993a. Towards a language-based theory of learning. Linguistics and
Education 5. Halliday, M.A.K. 1993b. Systemic theory. R.E. Asher (d.), The
encyclopedia of language and linguistics. Oxford: Pergamon Press. Volume 8.
Halliday, M.A.K.
consciousness.
Symposium 92
1994. London.

1995a, On language in relation to the evolution of human


S. Allen (d.), Of thoughts and words: proceedings of Nobel
"The relation between language and mind", Stockholm 8-12 August
Imperial College Press; Singapore: World Scientific Publishing Co.

Halliday, M.A.K. 1995b. Computing meaning: some reflections on past experience and
present prospects. Paper presented to PACLING 95, Brisbane, April 1995.
Halliday, M.A.K. 1996. On grammar and grammatics. Hasan et al (eds.).
Halliday, M.A.K. in press a. Things and relations: regrammaticizing experience as
technical knowledge. J.R. Martin & R. Veel (eds.).
Halliday, M.A.K. in press b. The grammatical construction of scientific knowledge: a the
framing of the English clause. In R.R. Favretti (d.), Proceedings of the international
conference "Languages of Science", Bologna, 25-27 October 1995.
Halliday, M.A.K. in press c. Grammar and daily life: construing pain. Bessy Dendnnos
(d.), Proceedings of the fourth international symposium in Critical Discourse
Analysis, Athens, December 1995.
Halliday, M.A.K. & R.P. Fawcett (eds). 1987. New developments in systemic
linguistics. Volume 1: theory and description. London: Pinter.
Halliday, M.A.K. & R. Hasan. 1976. Cohesion in English. London: Longman.
Halliday, M.A.K. & R. Hasan. 1985/89. Language, context, and text: aspects of
language in a social-semiotic perspective. Geelong, Vic.: Deakin University Press.
Reprinted London: Oxford University Press, 1989.
Halliday, M.A.K & Z. James. 1993. A quantitative study of polarity and primary tense
in the English finite clause. J. M. Sinclair, G. Fox & M. Hoey (eds.), Techniques of
description: spoken and written discourse. London: Routledge.
Halliday, M.A.K. & J.R. Martin, (eds.) 1981. Readings in systemic linguistics. London:
Batsford.
Halliday, M.A.K. & J.R. Martin. 1993. Writing science: literacy and discursive power.
London: Palmer Press.

628

References

Harvey, A. 1997. Equivalence and depersonalisation in definitions: an exploration of


lexicogrammatical and rhetorical patterns in English technical discourse. Ph.D. thesis,
University of Sydney.
Hasan, R. 1978. Text in the systemic-functional model. W. Dressier (d.), Current trends
in textlinguistics. Berlin: de Gniyter.
Hasan, R. 1984a. What kind of resource is language? Australian Review of Applied
Linguistics 7.1. Reprinted in Hasan (1996).
Hasan, R. 1984b. The nursery tale as a genre. Nottingham Linguistic Circular, 13.
(Special Issue on Systemic Linguistics.) Reprinted in Hasan (1996).
Hasan, R. 1985a. Lending and borrowing: from grammar to lexis. J.E. Clark (d.), The
cultivated Australian. Hamburg: Helmut Buske.
Hasan, R. 1985b. Meaning, context and text fifty years after Malinowski. Benson &
Greaves (eds.).
Hasan, R. 1985/9. Linguistics, language and verbal art. Geelong, Vic: Deakin University
Press. Reprinted London: Oxford University Press.
Hasan, R. 1987. The grammarian's dream: lexis as most delicate grammar. M.A.K.
Halliday & R.P. Fawcett (eds.). Reprinted in Hasan (1996).
Hasan, R. 1989. Semantic variation and sociolinguistics. Australian Journal of
Linguistics, 9.
Hasan, R. 1992. Rationality in everyday talk: from process to system. J. Svartvik (d.),
Directions in corpus linguistics: proceedings of Nobel Symposium 82, Stockholm,
4-8 August 1991. Berlin & New York: Mouton de Gniyter (Trends in Linguistics
Studies & Monographs 65).
Hasan, R. 1995. The conception of context in text P.H. Fries & M. Gregory (eds.),
Discourse in society: systemic functional perspectives. Norwood, NJ: Ablex.
(Meaning and Choice in Language: studies for Michael Halliday [Advances in
Discourse Processes L])
Hasan, R. 1996. Ways of saying: ways of meaning, ed. C. Cloran, D.G. Butt & G.
Williams. London: Cassell.
Hasan, R., C. Cloran & D.G. Butt (eds.). 1996. Functional descriptions: theory in
practice. Amsterdam: Benjamins.
Havrnek, B. 1932. The functional differentiation of the standard language. P. Garvn
(d.), 1964. A Prague School reader on esthetics, literary structure, and style.
Originally in B. Havrnek & M. Weingart (eds.), Spisovn cestina a jazykov
kultura
Hendrix, G. 1978. Semantic knowledge. D. Walker (ed), Understanding natural language.
New Yoik: North Holland.
Hendrix, G. 1979. Encoding knowledge in partitioned networks. N.V. Finder (ed),
Associative networks. New York: Academic Press.
Henrici, A. 1966. Some notes on the systemic generation of a paradigm of the English
clause. Working paper, O.S.T.I. Programme in the Linguistic Properties of Scientific

References

629

English, Communication Research Centre, University College London. Reprinted in


Halliday & Martin (eds., 1981).
Henschel, R. 1994. Declarative representation and processing of systemic grammars. C.
Martin-Vide (d.), Current issues in mathematical linguistics. Amsterdam: Elsevier.
Hjelmslev, L. 1943. Omkring sprogteoriens grundlaeggelse. K0benhavn: Akademisk
Forlag.
Hobbs, J.R. 1984. Building a large knowledge base for a natural language system.
COUNG 84.
Hobbs, J.R. 1987. World knowledge and world meaning. TINLAP 3.
Holland, C. & N. Quinn (eds.). 1987. Cultural models in language and thought.
Cambridge: Cambridge University Press.
Hopper, P. & S. Thompson. 1980. Transitivity in grammar and discourse. Language 56.
Hopper, P. & S. Thompson. 1985. The iconicity of the universal categories 'noun' and
Verb1. Haiman (d.).
Hopper, P. & E. Traugott. 1993. Grammaticalization. Cambridge: Cambridge University
Press. (Cambridge Textbooks in Linguistics.)
Hovy, E. 1988a. Planning coherent multisentential text. Proceedings of the 26th Annual
Meeting of the Association for Computational Linguistics.
Hovy, E. 1988b. Generating natural language under pragmatic constraints. Hillsdale, NJ:
Lawrence Erlbaum.
Hovy, E. 1991. Approaches to the planning of coherent text. C. Paris, W.R. Swartout &
W.C. Mann (eds.).
Hovy, E. &. K. McCoy. 1989. Focussing your RST: a step towards generating coherent
multisentential text. Proceedings of the llth Annual Conference of the Cognitive
Science Society, Ann Arbor, Agust 1989.
Huddleston, R.D. 1969. Some observations on tense and deixis in English. Language 45.
Jackendoff, R. 1972. Semantic interpretation in generative grammar. Cambridge, Mass.:
MIT Press.
Jackendoff, R. 1976. Toward an explanatory semantic representation. Linguistic Inquiry
Jackendoff, R. 1983. Semantics and cognition. Cambridge, Mass.: MIT Press.
Jackendoff, R. 1988. Conceptual semantics. In Eco, Santambrogio & Violi (eds.).
Jackendoff, R. 199L Parts and boundaries. Levin & Pinker (eds.).
Jacobs, P.S. 1985. PHRED: a generator for natural language interfaces. American
Journal of Computational Linguistics, 11.
Jacobs, P.S. 1987. KING: a knowledge-intensive natural language generator. In Kempen
(d.).

630

References

Jakobson, R. 1949. On the identification of phonemic entities. Recherches Structurales,


Travaux du Cercle Linguistique de Prague V.
Jerison, H. 1973. Evolution of the brain and intelligence. New York: Academic Press.
Jespersen, 0.1924. The philosophy of grammar. London: Allen & Un win.

Johnson-Laird, P.N.. 1983. Mental models: towards a cognitive science of language,


inference, and consciousness. Cambridge: Cambridge University Press.
Johnston, T. 1989. AUSLAN dictionary: a dictionary of the Sign Language of the
Australian Deaf Community. Petersham, N.S.W.: Deafness Resources Australia
Limited.
Johnston, T. 1992. The realization of the linguistic metafunctions in a sign language.
Language Sciences 14.4.
Kasper, R. 1989. A flexible interface for linking applications to Penman's sentence
generator. Proceedings of the DARPA Workshop on Speech and Natural
Language. Information Sciences Institute, University of Southern California.

Kasper, R. 1987. Feature structures: a logical theory with application to language


analysis. Ph,D. Thesis, University of Michigan.
Kasper, R. 1988. Systemic grammar and functional unification grammar. Benson &
Greaves (eds.).
Katz, J.J. & J. A. Fodor. 1963. The structure of a semantic theory. Language 39.
Katz, J. & P. Postal. 1964. An integrated theory of linguistic description. Cambridge,
Mass.: M.I.T. Press.
Kay, M. 1979. Functional grammar. Proceedings of the fifth annual meeting of the
Berkeley Linguistic Society.
Kay, M. 1985. Parsing with functional unification grammar. K. SparckJones & B.
Webber (eds.), Readings in natural language processing. Los Altos, CA: Morgan
Kaufman.
Keenan, E. & A. Faltz. 1985. Boolean semantics for natural language. Dordrecht: Reidel.

Kempen, G. (d.) 1987. Natural language generation: recent advances in artificial


intelligence, psychology, and linguistics. Dordrecht: Kluwer.
Kiparsky, P. & C. Kiparsky. 1970. Fact. M. Bierwisch and K. Heidolph (eds.), Progress
in linguistics. The Hague: Mouton.

Kittredge, R. & L. Lehrberger (eds.) 1982. Sublanguage: studies of language in restricted


semantic domains. Berlin: de Gruyter.
Kittredge, R. 1987. The significance of sublanguage for automatic translation. S.
Nirenburg (d.), Machine translation: theoretical and methodological issues.
Cambridge: Cambridge University Press.
Klose, GM E. Lang, Th. Pirlein (eds.) 1992. Ontologie und Axiomatik der Wissensbasis
von LILOG. Wissensmodellierung im IBM Deutschland LILOG-Projekt.

References

631

Kneale, W. & M. Kneale. 1962. The development of logic. London: Oxford University
Press. Knowlson, J. 1975. Universal language schemes in England and France 16001800. Toronto: Toronto University Press.
Kobayashi, I. 1995. A social system simulation based on human information processing.
PhD. thesis, Tokyo Institute of Technology.
Kobsa, A. & W. Wahlster (eds.). 1989. User models in dialogue systems. Berlin:
Springer.
Kress, G.R. & TJ. van Leeuwen. 1990. Reading images. Geelong, Vic: Deakin
University Press.
Kress, G.R. & TJ. van Leeuwen. 1996. Reading images: the grammar of visual design.
London & New York Routledge.
Labov, W. 1973. The boundaries of words and their meanings. C.-J. Bailey & R. Shuy
(eds,), New ways of analysing variation in English. Washington: Georgetown
University Press.
Lakoff, G. 1972. Linguistics and natural logic. D. Davidson & G. Harman (eds.),
Semantics for natural language. Dordrecht Reidel.
Lakoff, G. 1987. Women, fire and dangerous things: what categories reveal about the
mind. Chicago: Chicago University Press.
Lakoff, G. 1988. Cognitive semantics. Eco, Santambrogio & Violi (eds.).
Lakoff, G. 1992. Metaphor and war: the metaphor system used to justify war in the Gulf.
M. Putz (d.).
Lakoff, G. & M. Johnson. 1980. Metaphors we live by. Chicago: Chicago University
Press.
Lakoff, G. & Z. Kvecses. 1987. The cognitive model of anger in American Engush.
Holland & Quinn (eds.).
Lamb, S.M. 1964. The sememic approach to structural semantics. American
Anthropologist 66:3.
Lamb, S.M. 1965. Outline of stratificational grammar. Washington DC: Georgetown
University Press.
Lamb, S.M. 1966a. Prolegomena to a theory of phonology. Language 42.
Lamb, S.M. 1966b. Epilegomena to a theory of language. Romance Philology 19.
Lamb, S.M. 1992. Outline of a cognitive theory of language: a work in progress.
Mimeo.
Lang, E. 1992. Linguistische vs. konzeptuelle Aspekte der LILOG-Ontologie. In G.
Klose et al.
Langacker, R. 1978. The form and meaning of the English auxiliary. Language 54.
Langacker, R. 1984. Active Zones. Proceedings of the tenth annual meeting of the
Berkeley Linguistic Society. .

632

References

Langacker. R. 1987. Foundations of cognitive grammar. Stanford, CA: Stanford


University Press.
Large, A. 1985. The artificial language movement. Oxford: Blackwell.
Leech, G. 1974. Semantics. Hannondsworth: Penguin. Second, revised edition 1981.
Leisi, E. 1955. Der Wortinhalt. Heidelberg: Quelle u. Meyer.
Lemke, J.L. 1984. Semiotics and education. Toronto: Toronto Semiotic Circle
(Monographs, Working Papers and Publications 2).

Lemke, J.L. 1990a. Talking science: language, learning and values. Norwood, NJ:
Ablex.
Lemke, JL. 1990b. Technical discourse and technocratic ideology1. M.A.K. Halliday, J.
Gibbons & H. Nicholas (eds.), Learning, keeping and using language: selected papers
from the SthWorld Congress of Applied Linguistics, Volume II. Amsterdam:
Benjamins.

Lemke, J.L. 1995. Textual politics: discourse and social dynamics. London: Taylor &
Francis.
Lemke, J.L. in press. Multiplying meaning: visual and verbal semiotics in scientific
text. J.R. Martin & R. Veel (eds.).

Levin, B, 1993. English verb classes and alternations: a preliminary investigation.


Chicagor & London: The University of Chicago Press.
Levin, B. & S. Pinker (eds.) 1991. Lexical and conceptual semantics. Cambridge, MA &
Oxford: Blackwell.
Linde, C. 1987. Explanatory systems in oral life stories. Holland & Quinn (eds).
Lockwood, DJ. 1972. Introduction to stratificational linguistics. New York: Harcourt
Brace Javanovich.
Longacre, R. 1974. Narrative vs other discourse genres. R. Brend (ed.) Advances in
Tagmemics. Amsterdam: North-Holland.
Longacre, R. 1976. Anatomy of speech notions. Lisse: Peter de Ridder Press.
Longacre, R. 1985. Sentences as combinations of clauses. T. Shopen (d.), Language
typology and syntactic description. Volume II Complex constructions. Cambridge:
Cambridge University Press.
Lounsbury, F. 1956. A semantic analysis of Pawnee kinship usage. Language 32.
Malinowski, B. 1923. The problem of meaning in primitive languages. Supplement I to
C. K. Ogden & I. A. Richards. The meaning of meaning. New York: Harcourt Brace
& World.
Mann, W.C. 1982. An overview of the Penman text generation system. Information
Sciences Institute, University of Southern California: ISI/RR-83-114.
Mann, W.C. 1983a. The anatomy of a systemic choice. Discourse Processes 8.1.

References

Mann, W.C. 1983b. Inquiry semantics: a functional semantics of natural language


grammar. Proceedings of the First Annual Conference of the European Chapter of the
Association for Computational Linguistics.
Mann, W.C. & C.M.I.M. Matthiessen. 1985. Demonstration of the Nigel Text
Generation Computer Program. Benson & Greaves (eds.).
Mann, W.C. & C.M.I.M. Matthiessen. 1991. Functions of language in two frameworks.
Word 42.
Mann, W.C. & S.A. Thompson. 1987. Rhetorical Structure Theory: a framework for the
analysis of texts. Information Sciences Institute, University of Southern California:
ISI/RS-87-185.
Mann, W.C., C.M.I.M. Matthiessen & S.A. Thompson. 1992. Rhetorical Structure
Theory and text analysis. Mann & Thompson (eds), Discourse desription: diverse
linguistic analysis of a fund-raising text. Amsterdam: Benjamins.
Martin, J.R. 1985. Factual writing: exploring and challenging social reality. Geelong,
Vic.: Deakin University Press (Sociocultural Aspects of Language and Education).
Reprinted London: Oxford University Press, 1989,
Martin, J.R. 1992. English text: system and structure. Amsterdam: Benjamins.
Martin, J.R. 1996a. Transitivity in Tagalog: a functional interpretation of case. Berry et
al. (eds).
Martin, J.R. 1996b. Metalinguistic diversity: the case from case. R. Hasan, C. Cloran &
D.G. Butt (eds.).Martin, J.R. in press. Beyond exchange: appraisal systems in
English. S. Hunston & G. Thompson (eds,), Evaluation in text. London: Oxford
University Press.
Martin, J.R. & C.M.I.M. Matthiessen. 1991. Systemic typology and topology. F.
Christie (d.), Literacy in social processes: papers from the Inaugural Australian
Systemic Functional Linguistics Conference, Deakin University, January 1990.
Darwin: Centre for Studies of Language in Education, Northern Territory University.
Martin, J.R., C.M.I.M. Matthiessen & C. Painter. 1997. Working with functional
grammar. London: Edward Arnold.
Martin, J.R. & R, Veel (eds.). forthc. Reading science: research, popular culture, industry
and schooling. London: Routledge.
Matthiessen, C.M.I.M. 1987. Notes on the environment of a text generation grammar.
G. Kempen (d.).
Matthiessen, C.M.I.M. 1988a. Representational issues in systemic functional grammar.
JX>. Benson & W.S. Greaves (eds.).
Matthiessen, C.M.LM. 1988b. Semantics for a systemic grammar: the chooser and
inquiry framework. J. D. Benson, M. J. Cummings & W. S. Greaves (eds.).
Matthiessen, CM.IM. 1990a. Two approaches to semantic interfaces in text generation.
Proceedings ofCOUNG-90, Helsinki, August 1990.
Matthiessen, C.M.LM. 1990b. Metajunctional
Department of Linguistics, Sydney University.

complementarity

and resonance.

634

References

Matthiessen, C.M.I.M. 1991a. Language on language: the grammar of semiosis. Social


Semiotics 1:2.
Matthiessen, C.M.I.M. 1991b. Lexico(grammatical) choice in text generation. C. Paris,
W. Swartout & W.C Mann (eds.).
Matthiessen, C.M.I.M. 1992. Interpreting the textual metafunction. M. Davies & L.
Ravelli (eds.). Matthiessen, C.M.I.M. 1993a. The object of study in cognitive
science in relation to its construal and enactment in language. Language as Cultural
Dynamic (Cultural Dynamics 6.1-2.).
Matthiessen, C.M.I.M. 1993b. Register in the round: diversity in a unified theory of
register analysis. M. Ghadessy (d.), Register analysis. Theory and practice. London:
Pinter.
Matthiessen, C.M.I.M. 1993c. Insiantial systems and logogenesis. Paper presented to the
3rd National Chinese Systemic Symposium, Hangzhou University, Hangzhou, July
1993.
Matthiessen, C.M.I.M. 1995a. Fuzzness construed in language: a linguistic perspective.
Proceedings of FUZZ/IEEE, Yokohama, March 1995.
Matthiessen, C.M.LM. 1995b. Lexicogrammatical cartography: English systems.
Tokyo: International Language Sciences Publishers.
Matthiessen, C.M.I.M. 1995c. Theme as a resource in ideational Toiowledge'
construction. M. Ghadessy (d.), Thematic development in English texts. London:
Pinter.
Matthiessen, CM.I.M. 1996. Systemic perspective on tense in English. M. Berry et al.
(eds).

Matthiessen, C.M.I.M. & J. Bateman. 1991. Text generation and systemic linguistics:
experiences from English and Japanese. London: Pinter.
Matthiessen, C.M.LM. & M.A.K. Halliday. forthc. Outline of systemic functional
linguistics.

Matthiessen, C.M.I.M., I. Kobayashi & L. Zeng. 1995. Generating multimodal


presentations: resources and processes. Preprint of Artificial Intelligence in Defence
Workshop, Eighth Australian Joint Conference of Artificial Intelligence (AI '95)f
Canberra, November 13 -14,1995.
Matthiessen, C.M.I.M., K. Nanri & L. Zeng. 1991. Multilingual resources in text
generation: ideational focus. Proceedings of the 2nd Japan-Australia Symposium on
Natural Language Processing, Kyushu Institute ofTechnology, October 1991.
Matthiessen, C.M.I.M. & C. Nesbitt. 1996. On the idea of theory-neutral descriptions.
R. Hasan, C Cloran & D. Butt (eds.).
McCawley, J. 1973. Syntactic and logical arguments for semantic structures. O.
Fujimura (d.), Three dimensions in linguistic theory. Tokyo: The TEC Corporation.
McCoy, C. & J. Cheng. 1991. Focus of attention: constraining what can be said. W.
Mann, C. Paris & W. Swartout (eds.).
McDonald, E. 1994. Completive verb compounds in Modern Chinese: a new look at an
old problem. Journal of Chinese Linguistics 22.2.

References

635

McKeown, K. 1982. Generating natural language text in response to questions about


database structure. Ph.D. Thesis, University of Pennsylvania.
McKeown, K. 1985. Text generation: using discourse strategies and focus constraints to
generate natural language text. Cambridge: Cambridge University Press. McKeown,
K. & W.R. Swartout. 1987. Language generation and explanation. Annual Review of
Computer Science. 2.
Mel'chuk, I. 1982. Lexical functions in lexicographic description. Proceedings of the
Eighth Anuual Meeting of the Berkeley Linguistic Society.
Meiiish, C. 1988. Implementing systemic classification by unification. Journal of
Computational Linguistics, 14.1.
Miller, G.A & P.N. Johnson-Laird. 1976. Language and perception. Cambridge, MA:
Belknap Press of Harvard University Press.
Miller, G.A. & C. Fellbaum. 1991. Semantic networks of English. B. Levin & S.
Pinker (eds.).
Mitchell, T.F. 1957. The language of buying and selling in Cyrenaica: a situational
statement. Hesperis 26. Reprinted in T.F. Mitchell, Principles of Neo-Firthian
Linguistics. London: Longman, 1975.
Montague, R. 1974. The proper treatment of quantification in ordinary English. R
Thomason (d.), Montague, R.: Formal philosophy. New Haven: Yale University
Press.
Nakamura, J-I. & N. Okada 1991. Modeling, accumulating and evaluating machine
dictionary of Japanese noun concepts. Proceedings of the 2nd Japan-Australia
Symposium on Natural Language Processing, Kyushu Institute of Technology,
October 199L
Nesbitt, C. 1994. Construing linguistic resources: consumer perspectives. Ph.D. Thesis,
University of Sydney.
Nesbitt, C. & G. Plum. 1988. Probabilities in a systemic grammar: the clause complex
in English. R.P. Fawcett & DJ. Young (eds).
Nida, E. 1975. Exploring semantic structures. Munich: Fink.
Nordenfelt, L. 1977. Events, actions, and ordinary language. Lund Doxa
Noreen, A. 1904-12. Vdrtsprdk. Lund.
ODonnell, M. 1990. A dynamic model of exchange. Word 41.3.
OTtonnell, M. 1994. From theory to implementation: analysis and generation with
systemic grammar. Ph.D. thesis, University of Sydney.
OToole, L.M. 1989. Semiotic systems in painting and poetry. M. Falchikov, C. Poke
and R. Russell (edsM Festschrift for Dennis Ward. Nottingham: Astra Press.
OToole, LM. 1992. Institutional sculpture and the social semiotic. Social Semiotics
2.1. OTooie, L.M. 1994. The language of displayed art. London: Leicester
University Press (Pinter).
OToole, L.M. 1995. A systemic-runctional semiotics of art P.H. Fries and M.J.
Gregory (eds), Discourse in society: junctiond perspectives. Norwood, NJ: Ablex.

636

References

Oldenburg, J. 1987. From child tongue to mother tongue: a case study of language
development in the first two and a half years. PhD. thesis, University of Sydney.
Padley, G.A. 1985. Grammatical theory in Western Europe 1500 - 1700. Trends in
vernacular grammar I. Cambridge: Cambridge University Press.
Painter, C. 1984. Into the mother tongue: a case study in early language development.
London: Pinter.

Painter, C. 1989. Learning language: a functional view of language development. R.


Hasan & J. R. Martin (eds).
Painter, C. 1993. Learning through language: a case study in the development of
language as a resource for learning from 21/2 to 5 years. Ph.D. thesis, University of
Sydney.

Painter, C. 1996. The development of language as a resource for thinking: a linguistic


view of learning. R. Hasan & G. Williams (eds), Literacy in society. London &
New Yoik: Longman.
Paris, C.L., W.R. Swartout & W.C Mann. (eds.). 1991. Natural language generation in
artificial intelligence and computational linguistics. Boston: Kluwer.

Parker-Rhodes, A.F. 1978. Inferential Semantics. New York, NY: The Humanities
Press.
Partee, B. 1975. Montague grammar and transformational grammar. Linguistic Inquiry 6.
Pattabhiraman. T. 1992. Aspects of salience in natural language generation. Ph.D.
thesis, Simon Fraser University.
Patten, T. 1988. Systemic text generation as problem solving. Cambridge: Cambridge
University Press.
Patten, T. & G. Ritchie. 1987. A formal model of systemic grammar. G. Kempen (ed).
Paul, H. 1909. Prinzipien der Sprachgeschichte. Halle.
Pawley, A. 1987. Encoding events in Kalam and English: different logics for reporting
experience. R.S. Tomlin (d.).
Phillips, J. 1985. The development of comparisons and contrasts in young children's
language. M.A. Honours thesis, University of Sydney.

Phillips, J. 1986. The development of modality and hypothetical meaning: Nigel 1; 7


1/2 - 2; 7 1/2. Working Papers in Linguistics 3. University of Sydney, Linguistics
Department
Platzack, C. 1979. The semantic interpretation of aspect and aktionsarten: a study of
internal time reference in Swedish. Dordrecht Foris Publications.

Pollard, C. & I. Sag. 1987. Information-based syntax and semantics. Vol. 1:


Fundamentals. CSLI Lecture Notes No. 13. Stanford University, Center for the Study
of Language and Information.
Pollard, C. & I. Sag. 1993. Head-driven phrase structure grammar. Chicago & London:
The University of Chicago Press.

References

637

Postal, P. 1966. On so-called "pronouns" in English. F. Dineen (ed), Report of the


Seventeenth Annual Round Table Meeting on Linguistics and Language Studies.
Washington, DC: Georgetown University Press. (Monograph on languages and
linguistics 19).
Postal, P. 1970. On (preferential complement subject deletion. Linguistic Inquiry 1.4.
Ptz, M. (d.), 1992, Thirty years of linguistic evolution. Amsterdam: Benjamins.
Quillian, M.R. 1968. Semantic memory. M. Minsky (ed), Semantic information
processing. Cambridge, MA: MIT Press.
Quirk, R., S. Greenbaum, G. Leech, & J. Svartvik. 1985. A comprehensive grammar of
the English language. London: Longman.

Ransom, E. 1986. Complementation: its meaning and forms. Amsterdam: Benjamins.


Ravelli, L. 1985. Metaphor, mode and complexity: an exploration of co-varying patterns.
B A Honours thesis, Department of Linguistics, University of Sydney.
Ravelli, L. 1988. Grammatical metaphor: an initial analysis. E. Steiner & R. Veltman
(eds.).
Reddy, M. 1979. The conduit metaphor a case of frame conflict in our language about
language. A. Ortony (d.), Metaphor and thought. Cambridge: Cambridge University
Press.
Reinhart, T. 1982. Pragmatics and linguistics: an analysis of sentence topics.
Bloomington: Indiana University Linguistics Club.

Renton, N.E. 1990. Metaphors, an annotated dictionary. A concise overview of 3800


picturesque idiomatic expressions normally used subconsciously. Melbourne:
Schwartz & Wilkinson.
Restak, R.M. 1988. ne mind. New York: Bantam.

Roget, M. 1852. A Thesaurus of English Words and Phrases, Classified and Arranged so
as to facilitate the Expression of Ideas and assist in Literary Composition.
Rosner, M. & R. Johnson (eds.). 1992. Computational linguistics and formal semantics.
Cambridge: Cambridge University Press.

R.R. 1641. An English grammar: or, a plain exposition of Lilies Grammar. In English,
with ease and profitable rules for parsing and making Latine: very usefull for young
scholars, and others, that would in a short time learn the Latine tongue. Which may
srveos a comment for them that learn Lillie's Grammar. London, Printed by Felix
Kyngston for Mathew Walbank and Laurence Chapman.
Salmon, V. 1966. Language-planning in seventeenth-century England: its contexts aid
aims. C.E. Bazell, J.C. Catford, M.A.K. Halliday & R.H. Robins (eds.), In memory
ofJ.R. Firth. London: Longman.
Salmon, V. 1979. The study of language in 17th-century England Amsterdam:
Benjamins. (Amsterdam Studies in the Theory and History of Linguistic Science
3.17.)

Samlowski, W. 1976. Case grammar. E. Chamiak & Y. Wilks (eds.), Computational


Semantics. Amsterdam: North-Holland.

638

References

Sampson, G. 1987. Probabilistic models of analysis. R. Garside, G. Leech & G.


Sampson (eds), The computational analysis of English: a corpus-based approach.
London: Longman.
Sampson, G. 1992. Probabilistic parsing. J. Svartvik (d.), Directions in corpus
linguistics: proceedings of Nobel Symposium 82, Stockholm, 4-8 August 1991.
Berlin & New York: Mouton de Gruyter (Trends in Linguistics Studies &
Monographs 65).
Schank, R. 1972. Conceptual dependency: a theory of natural language understanding.
Cognitive Psychology 3.4.

Schank, R. 1982. Dynamic memory: a theory of reminding and learning in computers


and people. New York: Cambridge University Press.
Schank, R. & R. Abelson. 1977. Scripts, plans, goals and understanding: an inquiry into
human knowledge structures. Hillsdale, NJ: Lawrence Erlbaum.
Schank, R. & A. Kass. 1988. Knowledge representation in people and machines. In Eco.
Santambrogio & Violi (eds.).

Schneider, K. 1977. Aktionsart och aspekt i svenskan och danskan jmfrd med tyskan
och nederlandskan. Turku.
Sefton, P. 1995. State-potentials and social subjects in systemic-functional theory:
towards a computational socio-semiotics. Ph.D. thesis, University of Sydney.
Sheldrake, R. 1988. The presence of the past. Glasgow & London: Fontana.
Shieber, S. 1986. An Introduction to Unification-Based Approaches to Grammar.
Stanford University: Center for the Study of Language and Information.

Sigurd, B. 1977. Om textens dynamik. Papers from the Institute of Linguistics,


University of Stockholm 34.
Simon, H.E. & C.E. Kaplan. 1989. Foundations of cognitive science. M.I. Posner (d.),
Foundations of cognitive science. Cambridge, Mass: MIT Press.
Sinclair, J.M. 1992. The automatic analysis of corpora. J. Svartvik (d.), Directions in
corpus linguistics: proceedings of Nobel Symposium 82, Stockholm, 4-8 August
1991. Berlin & New York: Mouton de Gruyter.
Slade, D. 1996. The texture of casual conversation in English. Ph.D. thesis, University
of Sydney.

Slaughter, M. 1986. Universal languages and scientific taxonomy in the seventeenth


century. Cambridge: Cambridge University Press.
Sowa, J. 1983a. Conceptual structures. Menlo Park: Addison-Wesley.
Sowa, J. 1983b. Generating language from conceptual graphs. Computers and
Mathematics with Applications 9.1.

Sowa, J. (ed.) 1991. Principles of semantic networks: explorations in the representation


of knowledge. San Mateo, CA: Morgan Kaufmann.
Starosta, S. 1988. The case for lexicase: an outline of lexicase grammatical theory.
London: Pinter.

References

639

Steiner, E. 1984. Language and music as semiotic systems: the example of a folk ballad
Trier: LAUDT.
Steiner, E. 1988a. Describing language as activity: an application to child language.
R.P. Fawcett & DJ. Young (eds.).
Steiner, E. 1988b. The interaction of language and music as semiotic systems: the
example of a folk ballad. J.D. Benson, M. Crownings & W.S. Greaves (eds.).
Steiner, E. 1991. A functional perspective on language, action and interpretation: en
initial approach with a view to computational modelling. Berlin & New York:
Mouton de Gniyter.
Steiner, E. & R. Veltman. (eds.) 1988. Pragmatics, discourse, and text: some
systemically inspired approaches. London: Pinter.
Stiilings, N.A. et al. 1987. Cognitive science: an introduction. Cambridge, Mass.: MIT
Press.
Sugeno, M. 1993. Toward intelligent computing. Proceedings of the 5th IFSA World
Congress, Seoul 1993.
Svartvik, J. & R. Quirk, (eds.) 1980. A corpus of English conversation. Lund: Gleerup.
(Lund Studies in English 56.)
Talmy, L. 1985. Lexicalisation patterns. T. Sbopen (d.), Language typology and
syntactic description. Grammatical categories and the lexicon. Cambridge: Cambridge
University Press.
Taylor, J. 1989. Linguistic categorization: prototypes in linguistic theory. Oxford
Clarendon Press.
Teich, E. 1995. A proposal for dependency in systemic functional grammar:
metasemiosis in computational systemic functional linguistics. Ph.D. thesis,
Universit des Saarlandes, Saarbrucken.
Teich, E. & J.A. Bateman. 1994. Towards an application of text generation in an
integrated publication system. Proceedings of the 17th International Workshop on
Natural Language Generation, Kennebunkport, Main, USA June 21-24.
Thibault, PJ. 1992. Grammar, ethics, and understanding: functionalist reason and clause
as exchange. Social Semiotics 2.1.
Thibault, PJ. 1993, Using language to think interpersonally: experiential meaning and
the cryptogrammar of subjectivity and agency in English. Language as Cultural
Dynamic (Cultural Dynamics 6.1-2.).
Thompson, S.A. 1988. A discourse approach to the cross-linguistic category 'adjective'.
J. Hawkins (d.), Explaining language universal. Oxford: Blackwell.
Thomson, JJ. 1977. Acts and other events. Ithaca & London: Cornell University Press.
Tomlin, R. 1987. Linguistic reflections of cognitive events. R, Tomlin (d.).
Tornlin, R. (d.). 1987. Coherence and grounding in discourse. Amsterdam: Benjamins.
Trevarthen, C. 1987. Sharing making sense: intersubjectivity and the making of an
infant's meaning. R. Steele & T. Threadgold (eds), Language topics: essays in honour
of Michael Halliday, Volume L Amsterdam: Benjamins.

640

References

Tmbetzkoy, N. 1939. Grundzge der Phonologie. Prague. (Travaux du Cercle


Linguistique de Prague 7.)
Tung, Y.-W., C.M.I.M. Matthiessen & N. Sondbeimer. 1988. On Parallelism and the
Penman Language Generation System. University of Southern California,
Information Sciences Institute: ISI/RR-88-195.
Unsworth, L. 1995. How and why: recontextualizing science explanations in school
science books. PhD. thesis, University of Sydney.
Vachek, J. 1964. A Prague School reader in linguistics. Prague: Academia.
van Leeuwen, TJ. 1988. Music and ideology: towards a socio-semantics of mass media
music. Sydney: Sydney Association for Studies in Society and Culture. (Working
Papers 2.)
van Leeuwen, TJ. 1991. The sociosemiotics of easy listening music. Social Semiotics
1.1.
Vendler, Z. 1967. Linguistics in philosophy. Ithaca: Cornell University Press.
Ventola, E. 1987. The structure of social interaction: a systemic approach to the
semiotics of service encounters. London: Pinter.
Verkuyl, H J. 1972. On the compositional nature of the aspects. Dordrecht: Reidel.
Waddington, C.H. 1977. Tools for thought. Frogmore: Paladin.
Webber, B.L. 1987. Event Reference. TINLAP: theoretical issues in Natural Language
Processing -3. Position Papers. Computing Research Laboratory, New Mexico State
University (Memoranda in Computer and Cognitive Science).
Whorf, BJL 1956. Language thought and reality: selected writing of Benjamin Lee
Whorf, ed. J.B, Carroll. Cambridge, MA: The MIT Press.
Wierzbicka, A. 1975. Why "kill" does not mean "cause to die": the semantics of action
sentences. Foundations of Language 13.
Wierzbicka, A. 1980. Lingua mentalis. New York: Academic Press.
Wierzbicka, A. 1985. Oats and wheat: the fallacy of arbitrariness. J. Haiman (d.),
Iconicity in syntax. Amsterdam: Benjamins.
Wierzbicka, A. 1988. The semantics of grammar. Amsterdam: Benjamins.
Wignell, P., J.R. Martin & S. Eggins. 1990. The discourse of geography: ordering and
explaining the experiential world Linguistics and Education 1.4. Reprinted in
M.AJC, Halliday & J.R. Martin, 1993.
Wilkins, J. 1668. An Essay Towards a Real Character and a Philosophical Language.
London.
Winograd, T. 1968. Linguistics and the computer analysis of tonal harmony. Journal of
Music Theory. Reprinted in M.A.K. Halliday & J.R. Martin (eds.), 1981.
Winograd, T. 1972. Understanding natural language. Edinburgh: Edinburgh University
Press.

References

641

Winograd, T. 1983. Language as a cognitive process, Volume 1: syntax. Reading, Mass:


Addison Wesley.
Witherspoon, G. 1977'. Language and art in the Navajo universe. Ann Arbor: University
of Michigan Press.
Woods, W. 1975. What's in a link: foundations for semantic networks. D.G. Bobrow &
A. Collins (eds), Representation and understanding: studies in cognitive science. New
York: Academic Press.
Wortman, C.B. & E.F. Loftus. 1985. Psychology. New York: Knopf.
Wurm, S.A. 1987. Semantics and world view in languages of the Santa Cruz
Archipelago, Solomon Islands. R. Steele & T. Threadgold (eds.), Language topics,
essays in honour of Michael Halliday> Volume 1. Amsterdam: Benjamins.
Wuthnow, R. et al. 1984, Cultural analysis: the work of Peter L. Berger, Mary Douglas,
Michel Focaultt andJUrgen Habermas. Boston: Routledge & Kegan Paul. Yates, F.
1966. The an of memory. London: Routledge & Kegan Paul.
Zadeh, L.A. 1987. Fuzzy sets and applications: selected papers by LA. Zadeh, d. R.R.
Yager, S. Ovchinnikov, R.M. long, H.T. Nguyen. New York: Wiley.
Zeng, L. 1993. Coordinating ideational and textual resources in the generation of
multisentential texts in Chinese. Proceedings of PACLING 93, Vancouver, April
1993.
Zeng, L. 1996. Planning text in an integrated multilingual meaning-space: a systemiclinguistic perspective. Ph.D thesis, University of Sydney.
Zipf, G.K. 1935. The psycho-biology of language: an introduction to dynamic
philology. New York: Houghton Mifflin Company.

Index
A

Abelard 455
Abelson 431, 435
ability 526, 558
abstract,-ness 615-7
- space 278
material semiotic abstraction
191-2
Accompaniment 174
accomplishment 469-70, 472
achievement 469-70, 473
act of meaning 612
action (/event) 148, 432, 479,
485
activity 469-71
Actor 80, 148, 151, 188-90, 207,
478, 481, 491-3
ACTS 432
actual, -ized, -ization 113, 154,
215
acyclic directed graph 41
adjacency pair 531
adjective 185-6, 209-10, 217,
304, 446-8, 454
adverb, -ial 178, 196, 209, 215,
217
-group 100,178,221,518
advertisement 123
agency 468,478, 552, 559
Agent 154, 159,188, 350,476-8,
487, 491
agnate, -tion 7,209-10, 219, 2212, 227-8, 256, 286-7
fractal-224
Akan93, 315
Aktionsart 469
algebra of thought 421
algorithm 120
Allen 111
alternative descriptions 505
ambiguity, -ous 231, 549-50, 558
ambivalent 205
Amsler 431
analogy, -ous 68, 263, 510, 537,
607
anaphora, -ic 530
Anderson 54
Angle 175
Anglo-Saxon 242, 542
animal 191
antonym, -y 531
appearance (phase) 216
apposition 72
arbitrary, -iness 3, 98, 534
architecture 509, 606
argumentation 545
Argyle 610
Aristarchus of Somos 540

Aristotelian categories 46, 72


artefact 88
artificial: intelligence (AI) 2, 419
435
~ language 419-21, 447, 575
ascribing 338
ascription, -tive 73, 143
~ figure 183, 304, 338-42
aspect 64, 215, 304, 469
association [see also possession]
145
ATRANS 432
ATTEND 433
Attribute 74, 143, 149-52, 167,
183-4, 188, 207, 211, 350, 450
460, 464
attribution, -tive 211, 338-42,
514
Attributor 132
AUSLAN 30, 535
Austronesian 530
auto-genetic potential 224
autonomous (syntax) 3, 438
auxiliary 180, 306, 467

BABEL 433
Bach 422, 446
Bacon 445
Ballimlll
Banckes 570
Ban 120
Barwise 416, 422, 429
basic level [see generic level]
Bateman 1n, 2n, 35, 42, 42n,
43n, 122, 239, 323n, 374, 394,
398, 404, 435
Bauhin 571
Becker 6
becoming 160
Bedeutung 421
Begnftsschnft 421
Behaver 136, 456
behaviour, -al 135-6, 509, 514,
559
~ process 135-6, 142-4, 551
being 52, 128, 151, 561
~ & having 72, 98, 144-7, 209,
213, 370, 449, 473-4, 488, 504
benef active 484
Beneficiary 147, 151, 157, 188
Benson 1n
Berger 429
Berlin 72, 83-4
Berry 122
Biagi 240
Bickhart 422
binary (opposition) 206, 211
biological (systems) 507, 607-10
bio-semiotic (system) 606-10
birds of prey 86
blend 549-50, 558
Bohm 576

643

Index

Boole 185,421
Boolean algebra 421
borderline case 135-6, 551
bounded 181, 214,470
Brachman 31, 34n, 35, 36n, 40,
54, 419, 430
brain 509, 607-10
Breedlove 83
Brew 42
bridging (reference) 94,407
Brill 35
Bundy406
Bursill-Hall4l9
Butt 610
C

Caffarel 323n
calling [see naming]
Compte//422
Cam 422
Cantonese 315
cardinal: numeral 209
- vowel 69
Camap42l
Carrier 74, 151, 183-4, 188, 207,
349, 450
case55n
-grammar 441,491,494
~ relation 442-3
cases ('thing' type) 203
casual conversation 566
cats (construal of) 78
cataphora, -ic 530
categorial grammar 423
category, -ization 48, 68, 72-82,
89, 92, 96
Aristotelian ~ 46, 72
Chafe 426, 446, 457, 476, 479-85
chances ('thing' type) 203
change (as process) 213,468,512
chaos (/order) 329,547
characterology 523
charger 353
Chaucer 542
Cheng, J. 410
Cheng, M. 305
child language [see also
ontogenesis] 72
Chinese 206, 215, 221, 297-315,
459, 503
Chomsky 425, 438
chooser 353,394
&-inquiry interface 379
chreodization 71
circle diagram 34
circumstance 52, 63-4, 99,172-7,
217-22, 351, 466, 485-6
- of expansion/projection 63-4,
174-5, 219-20, 224

circumstantial (element) 116,

132, 148, 183, 303, 308, 351, 490, 519


490, 519

circumstantiation 504
Clark 94, 407
class 19, 27, 339n, 492-503, 513,
615
-of role filler 40
-name 73
-shift 259-62
classification [see also
categorization] 14, 612, 615
criteria in- 86
Classifier 197, 206, 211
clause 5, 8-9, 54-5, 178, 213,
227-32, 236-8, 466, 512, 536
- complex 121-5, 227-32, 236-8,
301
- nexus 303 [Chinese]
- nucleus 494
cleft construction 400
client system 372
climate 328
cline 46, 72
- of instantiation 381, 536
closed system 209
clothing 198
COBUILD 24n
codifying 22-5
cognition (type of sensing) 13744
cognitive: science 2, 72, 415,
565-601, 608
- semantics 424-9, 425n
cohesion 528-30
collocation 531
comic strip 106, 584
command 112, 525
comment (speaker's) 218
commodity (in exchange) 112,
524
common noun 73, 513
commonsense (knowledge,
theories) 565
COMMUNAL 374, 504
compacting 23-4
comparative, comparison 212,
531
complementarity 133-4, 137,
141, 154-5, 158, 270, 506, 522,
549, 551-2, 559, 562
- of 'doing' and 'being' 158-65,
338
of ergative and transitive 149-50,
158, 318
- of Medium and Process 158
complementation 505
completive phase 299, 307
complex (phenomena) [see also
sequence, clause complex] 48

644

component (semantic) 38-40, 497


componential analysis 38-40, 42
composition, -al, -ality [see also
structure] 145, 418, 422
- offigures128-65
compound 310
computer memory 590
conation, -live 216, 307
concept frame 54
conceptual: dependency (CD) 4313
- semantics 425-6, 425n
concrete 225
condensing 24, 232, 576
condition 118, 120-1, 303
[Chinese], 364-5, 369-70
conduit metaphor 273
configuration, -al 48, 52-7, 128,
160-1, 165-6
-organization 118
- phenomena 466
conflation 44
conflict (/ consensus) 429
congruent, -ence 227-32, 535-9
conjunction 104, 117, 178, 530-1
Conklin 83
connective 104
connotation 527
connotative semiotic 34, 605
conscious 61, 190
- process 577
conscious (/nonconscious) beings
93-4, 189
consciousness 130, 222, 481,
565-601
collective ~ 609
content of- 579
critical - 568
emergence of ~ 608
higher order - 604-9
primary - 608
consensus 422, 429
constraints (participant roles) 152
constnictivist 17
construe, -al, -ing 1-3, 5, 7-8, 82103, 398, 565-601, 607-10
- cats 73
~ experience 66-98, 424, 609
content 20, 236-8
-plane 4-5, 7, 69, 108, 511,
603
context 121, 375, 388-9, 438
-of culture 565
- of situation 354, 388, 434,
530, 602
contextual configuration 331
control 476-90
conventional [see also arbitrary]
534

Index

Cook 483
cooking 355-71, 606
copula 455
core meaning 69
count (/mass) 181
covert (category, criteria) 23, 26
86, 569, 575
creative (/ dispositive) 148
cieolize 298
Cross 5, 380n, 435
crossover (semiotic modes) 278
Cruse 90
cryptogrammar 569
cryptotype, -typic 4, 26, 134,
306, 443n, 569, 575
culinary (text) 563
cultural model 565
Cummings 1n

D'Andrade 598
Dahlgren 431, 435-7, 469-71
dance 606
Darwin 507
data fusion 444
database system 372
dative (object) 147
Davey 2n
Davidse 29, 151, 158n
Davidson
Davies
deBeaugrande
de Morgan 421
deconstrual 89
deductive reasoning 105
deep: case 441-42
- generation 444
- structure 126, 447
definition 76
Deictic 180, 197
deixis 133, 179, 302
delicacy 14-15, 20, 41, 87, 144,
325-27
- hierarchy 83
- of categorization 92
-of focus 92
demetaphorizing 284
Dennett 587, 598
denominal adjective 198,218
denotation 417
dependent clause 302
depictive 208
Derewianka
derivational: morpheme 242
- morphology 575
- priority 235
desideration, -tive 137
designed (/ evolved) system 544,
557

645

Index

determiner 209, 530


development (child) [see also
ontogenesis] 18
dialogue 525, 531, 613-14
dialogic construal (of experience) 610
dictionary 5n, 568
Dik 36, 477, 480n, 485-90
dimension [see also
multidimensional] 186, 206
diminutive 527
direct (/ mediated) realization 387
directional phase 299
disciplines 617
discontinuity 98
discourse 512
~ semantics 417, 605
discrete semiotic abstraction 193
discrimination (net, network) 37,
433
discursive mind 603
diseases 83
disjunction 104
dispositive (/ creative) 148
distilling 72
distinctive feature 39
doing 52, 128, 147, 151-52, 216,
366, 473, 488, 551
- and being 128, 132-4, 158-65,
340n
~ & happening 98, 147-9, 213,
449, 504
domain 320, 323, 351
-model 15, 323-7, 563-65
domestic pet 195, 613n
Downing 84
downranked sequence 77
Dowty 422-23, 469, 475-77, 495
dress 509
drift (metaphorical) 265
drinks 196
Dunbar 6l0
Durkheim 507, 609
dynamic (/ synoptic) 468, 485,
507

E
Eco 419
eco-functional 83
- selection 94-6
Edelman 593,604-9
educational knowledge 270,616
effector 491
effectuai (/ affectum) 148
Eggins In, 86, 329, 429, 576
Ehrich474
elaborating, -tion (expansion
type) 34, 63, 72, 106, 116, 144,
148, 169, 183, 204, 211, 222,
267, 288-93, 334, 433, 490,
520, 531

elasticity (of semantic space) 314


element 48, 59-65, 95, 99, 177227, 237, 305-9 [Chinese], 34650

ellipsis 530-1
Ellis 37, 68, 72
embodied mind 603
emotion 138
enabling 398, 512, 528
enacting, -ment 511, 524
enhancing, -ement (expansion
type) 63, 106, 116, 146-48, 170,
183, 204, 222, 267, 317, 490,
520, 531
entry condition 42
epistemic logic 111
Epithet 183, 197, 206, 211, 218,
436
equal (status in sequence) 51, 117
ergative 303, 317, 489, 559
ethnoscience 418
Eurasian culture band 297
Event 180, 213, 265, 469
evolution 17, 508, 545
-of the brain 607
evolved (/ designed) system 104
exchange 9
~ relationship 377
Existent 151, 348
existential 132, 336-8, 514
exophoric 73
expansion 11, 106-8, 116-8, 1447, 162, 183, 202, 209, 267,
302, 314, 364, 490, 520
- nexus 303 [Chinese]
-sequence 116-8, 144-7
experience, -tial 1, 3, 29, 511,
604, 612
~ chain 520
- class 212
~ complexity 180, 205
~ meaning 270
~ metafunction 314-5, 512-9
-resource 317
Experienced, -cer 484, 488
experientialist cognition 429, 598
experimental science 544-5, 569
expert: model 568
- system 372
expression 20
extended reference 102
extending, -sion (expansion type)
34, 63, 94, 102, 106, 116, 148,
183, 204, 222, 267, 302, 333,
417, 421, 423, 433, 490, 520,
531
extensional 423
exterior symbolic processing 129
extra-linguistic deception 444
phenomena 97

646

~-stratal438
world 423

Facet 204, 308


fact 102,138,140, 489
~ clause 203,579
- noun 203
factual writing 72
Faltz 37,422
familiarity 527
family resemblance 72
Fang 305
Fawcett In, 41, 374,429, 504,
605
feature:
~ network 198-202
~ structure 35,42
distinctive - 39
systemic ~ 42-6,65-7
Feigenbaum 120
Fellbaum 431
field (contextual) 320-7, 331,
398, 567
specific semantics 322
figurative expansion 226
figure 9,48, 52-9, 92, 95,98,
128-76, 178,193, 213, 222,
227, 236, 241, 303-5 [Chinese],
336-46, 366-70, 393, 432, 448,
466, 485
- of being & having 132-3, 1447,336
~ of doing & happening 132-3,147-9
~ of sensing & saying 128-32,137-44
Fikes 35
Fillmore 443,443n, 494
finger-spelling 534
Finite 9, 180
Finnish 221
first person 525
first-order logic 428
-reality 106
Firth 17, 30, 321, 446, 562, 602,
610
Fischer-Jprgensen 39
flux (of experience) 68
focus space 406
Fodor 39, 418
Foley 316n, 469,476,491-5,
502
folk: model 566, 573, 577-95
~ psychology 587
~ taxonomy 83, 418
force 477
formal:
- grammar 3-4, 6,16
~ semantics 419-24
- theory 4,427

Index

forward chaining 394


Fox 410
fractal 144
~ agnation 224
~ complementarity 562
~ pattern 294
~ type 223
Frakett
frame 40
based inheritance network 35,
40-1, 45, 96n
system 572
-of reference 206
free variant 238
Frege 421
frequency 553
Fries 238, 262, 410
functional:
- grammar 3-7, 15-6
- logic 36
- theory 4, 5-8, 416, 427
- variety 376
Functional Unification Grammar
(FUG) 32, 380n
future (tense, time) 64, 214
fuzzy 547
- logic 35
-set 46

Galileo 507
general noun 189
generalization 615-7
generalized model (of figures)
153-8, 164
generalized phrase structure
grammar (GPSG) 380n, 438
generation (system) 283,372-87
generic level (in taxonomy) 83
generic structure potential (GSP) 388-9
GENESYS 605
genie [see also register, situation
type] 605
Gerard51l
gesture 533
Given 529
~ + New 238
Givon 37, 505
global: norm 557
- probabilities 553
Goal 80, 148,151, 167, 188,
207, 472
Goldman 433
Goodenough 39
goods-&-services 112,524
grammar, grammatical 3-4, 29,
98, 123, 216, 510-1
- and meaning 15-22
~ for science 544

647

Index

- metaphor 7, 46, 52, 124-5,


184, 227-96, 312 [Chinese],
331, 399, 401, 532-47, 574
- reactance 89
~ schematization 87
- structure 24, 380, 395
~ system 41-4
-zone 87
grammaticalization 23, 434
grammatics 6, 510, 556
graphic (representation) 34
Greaves In
Greek, Greece 242, 539-41
Grosi 406, 407n, 409

gruber 494

Haack 421n
Hamlet factor 555
happening [see also doing &
happening] 128, 369
Harvey 72, 17
has-a 40
Hasan 1n, 5, 34, 42, 75, 97, 100,
102, 155, 263, 331, 378, 388,
417, 424, 429, 435, 605
having [see being & having]
Havranek 321
Head 183, 312, 540
head-driven phrase structure grammar
(HPSG) 380n, 438
hearer modelling 111
Hendrix 406
Henrici 4l
Henschel 42
herbs 569
here & now 179
higher: animal, primate 88, 610
higher-order, consciousness 3,
604, 608-9
history, -ical 576
- change 555
Hjelmslev 4, 17, 20, 34, 510,
605
Hobbs 431
Holland 565
Hopper 23, 166n
Hovy 6, 374, 410
Huddlesion 446
human.*
~ body (as orientational
framework) 225
- collective 193
~ consciousness 106
-experience 511
~ language 299
hybrid system 35, 37
hyponym, -ic, -y 40, 83, 145,
197, 310, 531

- taxonomy 145
hypotaxis, -tactic 117, 302, 521

iconicity 533
idea (/locution) 108, 129, 138,
578
ideation, -al 8, 99-100, 417, 51123
~ metafunction 7-10,16n, 99101,511-23
~ model 565-601
- semantics 2,441
ideation base 11,13-5,68, 222-6,
314, 428, 434-7, 565, 602-4
building - 320-71
using - 372-414
Identified, -ier 152
identifying (figure) 514
identity (process type) 145,212
ideology 34,605
immanent (theory of meaning)
416, 603
immediate (/ remote) 558
impacting 153
imperfective 304
implementation (computational)
31
in rebus/ in verbis 262-3
inceptive 307
inclination 478, 526, 558
indelicate (categories) 433
indeterminate, -acy 35,70,98,
118,503-5,547-62
axial - 303
individual, -ized 18,513,610
- constant 36
- mind [see also mind] 565, 599
inferencing 35, 111, 122, 387
information 112, 398, 512, 524
~ theory 557
INGEST 433
inherent (quality) 211
inheritance 327
insertion 43
instance, -tial 12, 321-3, 555,
566
- system 385
discoursal- 211-12
phenomenological ~ 66
instantiation 14-5, 40, 45, 323,
327, 382-6, 514
institutions 192, 567
instrumental (microfunction) 612
intelligent computing 372n, 448
intension, -al 421
- logic 105n, 111
intensive 73, 132, 148, 183, 211,
490
- ascription, -tive 15,74

648

intentionality 477-9
interne tant 111,600
interaction base 11, 46, 111, 122,
179, 376, 602
interactional (microfunction) 612
interaxial 95
interior symbolic processing 129
interlingua 446
interpersonal 8, 99-100, 215,
218, 227, 417, 478, 511-3, 5238, 558, 569, 600, 612
-distance 111
~ meaning 241
~ metafunction 9-12, 99-101,
523-7
- metaphor 583
~ semantics 112
~ sequence 122
~ system 198
interpretation (in formal
semantics) 424
interpretive semantics 438-9
interstratal 25, 438, 443
~ preselection 375
- realization 288, 375
~ relation 382
intersubjective 601
intonation 72
- contour 527
intra: -linguistic 444
semantic 443
~-stratal288,437
intransitive 148,168
involvement (degree of) 172-6
is-a40

Jackendoff36,181,424-6,425n,
432, 439, 494
Jacobs 289, 379n
Jakobson 39
James 553,557
Japanese 297-8
JerisondQl
Jespersen 185,455
Johnson 233n, 273,426, 430
Johnson-Laird 416,423
Johnston 30, 534-5
judgement (speaker's) 215
junction, -al 271, 283, 344
~ categories 278
-element 243
semantic ~ 344

K
Kalam316-8
Kamp 417n
kana298
kanji 298

Index

Kaplan 380nT 597


Kasper 42, 393n
Kass 431,435
ATafz39,481
Kay, M. 42, 380n
ATry, P. 72
toermzn 37,422
kind type 437
kingdom 83
Kiparsky, P. a C 505
Kittredge 321
KL-ONE 35, 40, 54
Klose 431
A>ura/e, W a M. 419, 455
knowledge 1, 3, 603
- base 2, 441-3, 604
~ representation 29, 54,419
Knowlson 419
Kobayashi 373,606
fotefllll
Adverses 143, 585
Kress 34, 427, 606
KRYPTON 35

L
Lflfov 72
Lakoffll, 72, 143, 233n, 273,
424-5, 425n, 429, 585, 598
Lamarck 507
Lamb 4n, 416, 430, 510
Lang 443-4
Langacter 92,426-7,455
language: and other semiotic
systems 602-11
as action 528
~ as information 528
~ as prototype 606
~ as reflection 528
~ based (case relations) 442
- free 431, 434
~ of settlement 539
~ typology 315
- universals 299
langue (/ parole) 328
Large 419
Latin 242, 541, 575
laws of reasoning, of thought 421
learned (language, vocabulary)
298
Leech 40, 83
Lehrberger32l
Leibniz 421
Leisi 505
Lemkein, 330,429, 606, 611
level (of taxonomy) 83
levels (of language) [see also
strata] 4
Levesque 35, 40, 419, 430
Levin 431,505

Index

lexeme 568
lexical: class 24
- cohesion 530-1
- compacting 23
- function 505
- item 198, 568
- metaphor 221, 232, 585
- verb 306, 308
- zone 89
lexicalization, -ized 23,426
- token 76
lexicogrammar, -grammatical 4-5,
108, 151, 299, 380, 438, 441,
511, 604
~ generation 387, 394
- metaphor 277
lexicon 5n, 199
lexis 5, 98, 197-202, 216, 380n,
568
~ as delicate grammar 198
life form 83
LILOG 431, 443
Lily 206
Lin In, 374
Lind 568
linear (/ serial) time 300
Linnaeus 572
local plan 379, 392
location, -uve 152, 484,491
Lockwood 4n
locution (/idea) 108, 129
Lofts 597
logic, -al 29, 35-7, 104-6, 511,
519
- form 455
- structure (of predicate) 497
-(philosophical) tradition (in
linguistics) 415-9
logical: metafunction 314-5, 51923
~ mode of constatai 317
- relation 36,301
--semantic relation 262
logogenesis, -etic 18, 235, 537
~ time frame 72,384
Longacre315,356
LOOM 35, 40
loss of information 230, 235
Lounsbury 39
Luckmann 429
M
machine translation 431,446
macro: -being 355
circumstance 63,172,218
figure 122
operation 356
participant 60,182
phenomenon 102
-role 491-4, 502

649

Malinowski 97,328
Mandarin (Chinese) 315
Mann 122n, 374, 379
Manner 208, 217,478
marking 22
Martin In, 29, 34, 46, 72, 86,
122, 137, 202, 227, 239, 263,
271, 312, 329, 356, 407, 417,
435, 443, 503-4, 566n, 572, 605
Marx 507
mass (/count) 181
material 99, 134, 190, 448-60,
504, 514, 559
~ clause, process 99,134-7,44855, 504, 514-6
- implication 104
- realm 190, 194
material (quality type) 206
mathematics 508
Mafhesius 529
Matter 587
McCawley 422
McCoy 410
McDonald 299, 305
McKeown 388,445
meaning 1-4, 108, 415-40, 509,
603
- as expansion 222
~ as projection 222
-base 1-3, 372,441,608
-group 98, 614
~ in NLP 430-7
making activity, behaviour
603-6
~ representation 29
act of ~ 612
[see also semogenesis]
Means 487
measure-word 306
Medium 154, 160, 188, 207,
450, 493-4
Mel'chuh 6, 505
Mellish 42
membership (process type) 145
memory organization package
(MOP) 435
mental 99, 130, 134, 464, 504,
514, 521, 559
~ clause, process 130,134-7,142,175,514
7,578,583
-model 424
meronym, -ic, -y 40, 83, 89-94,
197, 531
~ extension 89
~ taxonomy 145
message 9, 241, 378, 393
metafunction, -al 7-13, 15, 27,
507-46
- complementarity 562
metalanguage, -linguistic 30-5

650

- typology 505
metaphenomenon 106,140
metaphor 197,425-6,489, 500,
536, 606, 615-7
grammatical - 7,46,52,124-5,184, 22796,312-4 [Chinese], 331,399,401,532-47,
574
lexical ~ 221, 232-5, 273, 585,590
metaphoric, -al: complementarity
562
-compound 311
~ distance 598
-drift 265
- instability (of relators) 265
- reconstrual (of sensing) 596
- relationship 235
- shift 238-43
metasemiotic (space) 603
metathing 210
meteorological: process 563
- text 65, 328-54
metonymy 606
micror-category 194,204
function 612
middle (voice) 448
M*7/er416,431
mind 440, 577-601,603
- as container 585
oriented (tradition) 416,429
Miniprocess 305
Minirange 188, 218
minor process 266, 308
Mitchell 435
modal (form, verb) 64, 306
modality 478, 525, 558, 583
mode (contextual) 321, 332, 398,
567
mode: of dressing 606
-of expression 535
-of meaning 46
model 213, 417
- theory, -etic (semantics) 421-3
modification 183,197, 312
modulation 478
Momma 35, 42
Montague 422,41$
- semantics 422
mood 9,113, 396,455,485, 525
Mood 9
morphology, -ical 16, 23,242-3,
305 [Chinese], 527
mother tongue 73, 537
motion (process type) 88,471
move (in dialogue) 9, 241, 393
MTRANS432
MULTEX 374,386
multidimensional: organization
39
- semantic space 69, 522

Index

multilingual text generation 313


music 509, 606
musical instrument 195

N
N-Rheme262
naive semantics 431, 435-7
Nakamura43l
naming 73-4
Nanri 374
natural (/ arbitrary, conventional)
3, 15
natural: force 193,476
- language 421-2,445-7,572
processing (NLP) 2, 34n,
372-414, 430-6
- logic 104-6, 421, 445
Navajo 58
near (/rote) 179,209
needs ('thing* type) 203
negation 308,447
negative (/ positive) 525, 557-8
Nesbitt5n, 111,556,602
nesting 117,120
network:
- of related categories 82-95
discrimation - 37-9,433
feature ~ 198-202
neural - 607
semantic - 6,15, 34n, 40
system - 41-5,198,433
neural network 607
neutral (aspect) 304
neutral term (in system) 300
neutralization, -ized 549, 551,
560
New 262,529
Newton 240, 507, 542, 575
nexus [see also clause complex]
116, 520
clause - 303 [Chinese]
Afafo505
Nigel grammar 556n
NIKL35,40
NLP 2, 34n, 372-414,430-6
nomenclature 570
nominal:
- deixis 179
- group 100, 178, 220, 230,
265, 305-6 [Chinese], 312
[Chinese], 518, 536, 540
nominalization, -izing 269
- processes 243
~ suffix 541, 576
- tendency 301
non-actual 113
non-conscious 61
non-discreteness 46,135
non-spatial (realm) 225

651

Index

non-state 181, 468, 479


Nordenfel 468,469,477-8
Noreen 505
Norman French 542
noun 186, 189, 209, 237, 304,
446-8, 539
~ adjective 206
- base 318
-classifier 306
- substantive 206
- substitute 207
common-73,513
count-181
mass ~181
proper ~ 73
now 214
nucleus:
~ of figure 154,158,171,174-5,218-9
clause - 156,213,494
nuisances 196
numeral 209
Numerative 197, 212

O'Donnelt 35, 374, 376n, 409n


O'Toole 427, 606
object (material/ semiotic) 191-2
objective 558
objectivist (cognition) 422-24,
429
obligation 526, 558
oblique case 218
observer (/ phenomenon) 510
offer 112, 525
Okada43l
Oldenburg 614
ontogenesis, -etic 18, 72-82, 235,
537, 610-20
- time frame 72
ontology 36,426,441
open set 209
order (/chaos) 329, 547
order of complexity 95
-of reality 106
ordinal (numeral) 209
organism 610
otber-agentive 154,163
outcome (of change) 148,159-60
overlap 549, 551, 559
overt (category, criteria) 26, 86,
569
P

Padley 455

Painter In, 72-3,83, 580, 614


painting 509,606
paradigm, -atic 13,24, 249,2559,509
~ axis 24,83

- organization 83, 418


- orientation 28
parataxis, -tactic 117, 302, 521
Parker-Rhodes 406n
parole (/ langue) 328
pars pro toto 94
part-whole 93n, 197

Partee 422

partial association 556


participant 52, 60-1, 99, 167,
177-212, 218-20, 237, 265, 3034 [Chinese], 347,436, 463-5,
466-8, 476, 512-8
- perspective 81
- role 57, 79, 128, 190
participation (models of) 149-77
participle, -ial 210
particularistic model (of figures)
150-3, 164
partition 405
past (tense, time) 64, 214
patient 479,491
Pattabhiraman 409n
Patten 2n, 42, 322n, 379
Paul 185
Pawley3l5
Penman 353, 374, 393n, 394,
431, 504, 556n
-Upper Model 35,431
perception (type of sensing) 88,
138, 424, 500, 606
perceptual space 225
perfect 215
perfective (aspect) 304
peripheral meaning 69
Perry 422
person (grammatical) 525
person (social) 610
personal (microrunction) 612
personal pronoun 517
perspective: ergative ~ 552
participant -81
topological - 68-72
transitive - 551
typological - 69-72
phase 64, 91, 159,216,299
[Chinese], 469
completive ~ 299
directional ~ 300
phase: of change 160
- of generation process 387
phatic communion 328
phenomenon (/ observer) 510
Phenomenon 9, 48, 80, 151, 188,
210, 465
Phillips 123
phoneme 39
phonology, -ical:
- representation 30
- system 39

652

pbora (anaphora, cataphora, etc.)


530
phrasal verb 265
phylogenesis, -etic 17, 235, 537
~ time frame 17
physical systems 507
physiological systems 607
pitch: prominence 529
plants 83, 569
Platzack 469-70
Plum 111,556
pointing (/referring) 613
polar, -ity 37, 64, 525
Pollard 380n
polysystemic 85, 506, 562-4
Popper
~'s World Three 577
position 485
positive (/ negative) 525,557-8
possession, -ive 132, 145, 148,
462, 490, 504
Possession 348
Post-Deictic 212
post(positive)-verb 299
Postal 446
potent, -cy 476,479-85
potential 15, 321, 555, 566
auto-genetic ~ 224
meaning ~ 7,10,14-5,18,20,32-33,74
power 527
~ to determine 480n, 485-90
pragmatics 12,417
Prague School 321,415, 529
predicate 36,455,457,467
-logic 36,207,441,447
pie-language 612
preposition, -al 308
~ phrase 100,178, 218-20, 3089 [Chinese], 518
preselection 44,95, 353
presence 43
present (tense, time) 64, 214
pre(positive)-verb 308
primary (clause in nexus) 302
primary consciousness 608
primitive (act) 431-3
probabilistic (system) 72
probability 329, 364, 526, 552-8
-profile 557
procedure, -al 120
- text 356
process 9, 52, 64-5, 91, 99, 159,
168, 177, 206, 213-8, 237, 265,
303, 309-12 [Chinese], 347,
463, 466, 512, 517
~ configuration 512
participant complementarity
133
-role 57

Index

- type 134-65, 303,460-5,4689


- typology 466-406
process (/ resource) 374-86
- of instantiation 382
Process 8-9, 10-1, 13, 24, 100,
151-65, 168-72, 228, 265, 464,
467
programming language' 31
projected world 429
projection, -ing 11, 28, 108, 137
170, 202, 209, 302, 314, 364,
487, 490, 520-1, 578, 600
~ nexus 303 [Chinese]
~ sequence 108-15,128-32
[see also idea, locution]
pronoun 189, 530
proofs ('thing' type) 203
proper noun 73
proposal (/ proposition) 113, 140
203, 241, 525
proposition (/proposal) 36,112,
140, 203, 241, 417, 525
prepositional: calculus 417
- logic 36,104-6
~ semantics 417
prosody, -ic (feature, system) 527
protolanguage 73, 537,611-5
prototype theory 72
pro-verb 99
psychoanalysis 598
psycholinguistics 416
psychology: of cognition 590
academic ~ 598
folk ~ 587
purr (/ snarl) word 527

Qualifier 81, 203, 465, 541


qualifying [see enhancing]
quality 62-3, 182, 205-12, 265,
304 [Chinese], 448-65
- in a clause 304
~ in a nominal group 304
- of expansion 62,209-12
-of projection62, 209
quantifier 36
quantities (type of quality) 211
quantum: of change 116,132,
213
~ of experience 314
~ of infonnation 239, 530
question 112, 525
Quillian 6,40
Quinn 565
Quirk 181,468
quoted,-ing 110,579

653

Index

randomness 547
Range 100, 157-9, 168-72, 188,
218, 274, 289-90, 314, 460,
464,472
ranging 460
rank 4, 230-2,568
- scale 10,230-2
-shift 259-63
Ransom 505
Ravelli 227
Raven 83
reactance 27,99,134, 500
readiness 526, 558
reality construction 608
realization, -al 4, 19, 25-6, 98,
134-7, 144, 327, 437, 447, 510,
514
- potential 294
- statement 43
-variant 283
direct - 387
mediated ~ 387
Receiver 80, 130, 151, 315, 580
receptor systems 607
recipe 355-71
Recipient 80
reconstrual (of experience) 241,
546
Ready 212, 585
reference (textual) 100-2, 530-1
~ opacity 111
extended ~102
reference -rial (denotation) 416-7
referring (/ pointing) 613
register 302, 351, 448, 557
- variation 563
regressive (modification) 312
regulatory (microfunction) 612
reification, reined 143
- of experience 575
- sensing 594
Reinhart401
relational 99,134, 448-50, 461,
464, 504, 514
- clause, process 72, 134-7,1447, 448-55
-organization 118
~ type 437, 504
relator 58,116,177-8, 245, 266,
303
remote (/immediate, near) 179,
558
Renaissance 542
Renton 533n
reported,-ing 110,579
representation, -al 29-46,417
~ meaning 423

~ of grammatical metaphor 27896


resetting (of probabilities) 557
resource (/ process) 374-86
resource (/ rule) 415
Restak 589, 598
restricted: language 321
~ register 105
restriction (realization operator)
44
resultative 208
- Attribute 217, 450
Rheme 9, 529
rhetorical (-ethnographic) tradition
(in linguistics) 415-9
Rhetorical Structure Theory
(RST) 122n, 410
Ritchie 42
Roget 38
~'s Thesaurus 195, 544, 568
role 54-60, 125, 149-52, 167
- infigure178,613n
- restriction 58
Role and Reference Grammar
(RRG)491
Rosch 72
roses 86
Rosner 430
rule (/resource) 415
Russe II 421

Sag 380n
Salmon 419
Samlowski 36
Sampson 556n
Santa Cruz language family 318
Saussure 509-10
SayerSO, 106, 130, 151, 153,
188, 315, 599
saying [see also sensing &
saying] 52, 98, 128-32, 151-3,
213, 222, 474, 488, 504, 577-80
scalability, scalar 143, 206, 210
Schank 431, 435
Schmolze 35, 54
Schneider412
scientific: discourse 298
[Chinese], 539-47, 569-77
- model 567, 572-7, 596
- plane 507
~ taxonomy 86, 570
- theory 565
Scope 151
script 431
sculpture 606
second-order 398
- level 108
~ reality 106
second person 525

654

Sefton 610
selection (in system) 95
self-agentive 154,163
self-constnial 73
semantic, -s 4, 108, 299, 376,
511,604
-component 418
- decomposition 418,497
-environment 223
-field69, 325
- generation 387-9
-junction 260
- net 6, 40
- network 15, 34n, 605
- profile 323
-role 491-503
- space 15,46,222,466
- structure 417
- style 298
- system [see also meaning
potential] 1, 61, 68, 71-76, 823, 96-8, 293, 320, 323, 364,
418, 428
- theory 437
- type 66, 78n, 83
-verb 457
-world order 80
cognitive - 425-30
conceptual - 426-9
formal - 419-24
functional ~ 3-7,418
grammatical - 466,604n
lexical - 604n
situational - 422
semanticized, -ation 20, 512
semiotic, -s 190
- abstraction 203
-design 420
- plane 520
- reality 129
- realm 190, 194
- system 2, 507-9
socio 605,609
semogenesis, -etic 17
- process 17,259
semohistory 18, 235
Senser 9, 80,106,130, 151,18990, 210, 517, 579
sensing 52, 94, 98, 106, 128-32,
137-44, 151-2, 209-13, 222,
370, 473, 504, 551, 577-86
- & saying 128
extensions of - 601
sequence 50,69,92, 95, 104-27,
222, 227, 267, 301-3 IChinese],
317,333-6,364-6,560
expanding-51,116-8
projecting-51,108-16
serial (/ linear) time 215

Index

Sheldrake lOn
Shieber 380n
sign (deaf-) 533-5
signed English 534
signing space 533
sign (symbol) 19,237,613
signified, -fier 533
signification 97
Simon 597
simple: circumstance 63, 172,
217
- participant 60
- quality 62
- thing 61, 182, 370
Sinclair 6
Sinn 421
situation, -al:
- semantics 422
- specification 386
-type 567
context of - 354,388,530
skewing (of probabilities) 557
SLANG 379
Slaughter 83, 86, 419, 570
slot 54
small human collective 195
snarl (/purr) word527
social: mind 603
-relationships 511
- system 507, 602
socio-semiotic (system) 607-10
Sondheimer394
sorted logic 36
sounding (/ meaning, wording) 4,
29
Sowa40, 406n,419, 431
space, spatial 212
- metaphor 68,276
speaker- listener complementarity
530
specialization (of classes) 14
specific level (in taxonomy) 83
speech (/writing) 536
- synthesizer 374
speech: act (theory) 417
- function 524
spoken language (/ sign) 533-5
stack 409
Standard Average European (S AE)
272
standard of comparison 212
StarostaMl
state, stative, -ity 181, 432, 4689, 476-90
-predicate 207
-verb 207
state of affairs 485
statement 112, 525
Steiner In, 54,469n, 606

655

Index

Stillings 594
strata, -14
- ascent 386
- organization 299, 375
~ perspectives (on meaning) 437440
stratification, stratified (system)
4, 96-8, 375-82, 505, 614
strict taxonomy 38, 433
structural role 89
structuralism 508
structure [see also composition,
configuration] 8-11, 24, 380,
388, 395, 417, 520, 527-8, 541n
deep-126,445,447
generic - 388-9
surface - 445
Subject 9, 100, 171,207,455
subjective558, 601
sublanguage 321
subpersonal (level of
consciousness) 593
substance 191
substantive 185
substitute (element) 531
subsumption 14, 40,491-2
subtheory 563
Sugeno 35, 372n, 373, 444, 448
suggestion 113
superordinate310
surface generation 445
Swartoul 445
symbol, -ic 509
-exchange 112
- logic 421
- source 130
Symbolizer 129
syndrome 233, 244, 268, 274
- of elemental metaphors 250-5
-of features239, 466
synecdoche 94
synonym, -ic, -y 531
synoptic (/dynamic) 507
syntagm, -atic 9, 13, 24, 83, 24955
-organization 417
system 12, 323, 553, 612
-network 41-5,433
- of systems 562
structure cycle 430
---process 507-11
biological - 507,607-10
closed-209
grammatical - 41-4
physical - 507
physiological - 607
semantic - 1, 61,68,71-76,82-3,96-8,
293,320,323,364,418,428
semiotic - 2,507-9

social - 507,602
systemic: complementarity 562
-feature 497
- parameters (of grammar) 87
- representation 41-6
systemic (functional): grammar
(SFG) 3, 28, 503-5
~ theory 12, 72

Tacitus 431
Tagalog 491, 503-4
Talmy 426
Target 151,188
7ar5&421,421n
taxonomy, -ic 36-7, 83-9, 184,
197, 206, 211-12, 309
[Chinese], 419, 518, 570
- of systems 507
- of things 265
-order 576
~ relation 73-4
- stage (of science) 86
folk - 83,481
scientific - 86,570
strict - 38,433
technical - 312
Taylor 12, 84, 166n
technical, -ize 539
-knowledge 617
- taxonomy 312
- term 286
Teich 2n, 42, 374
temporal 212
- complementarity 178
- features (of process) 214
- profile 214, 468
- stability 205
- unfolding 133
tenor (contextual) 320, 332, 377,
398, 567
tense 28, 181, 215, 300, 302,
304, 553
- sequence 302
term (in system) 42
terminology 568
text 18, 121, 553, 566
-base 11, 46, 378, 398-414
- generation system 283, 372-87
- processing 372-414
- structure 388
textual 8, 99-100, 417, 511, 528,
569
- meaning 238, 270
- metafunction 9-12, 99-101,
528-32
-prominence 409
thematic relation 494
thematic system 330

656

theme identification (system) 399


theme predication (system) 400
Theme 9, 100t 262, 398, 491,
529
- + Rheme 238
theory, theoretical 1, 31-2, 374,
565, 595, 605
-model 32,415-8,565-77
thesaurus [see also Roget] 5n, 38
Thibault 429, 610
thing 60-2, 81, 180, 183, 184205, 309-12 [Chinese], 370-1
Thing 180,183-4, 203-5
thinginess 264
third person 525
Thomason 422
Thompson 122n, 166n, 209n
Thomson 488
time 213-7
linear - 300
serial-215
Token 152,188, 288
- + Value 561
token-specific 502
-role 494-5
- semantic relation 491
token-value: chain 293
- relation 76, 294
Tomlin 426
topological 46, 98
- perspective 68-72,162n
traditional grammar 466
transcategorization 242-4,259-63,
344
transcendent (theory of meaning)
416, 423
transcendental particles 575
transformation, -al 88
- grammar 126
transforming experience into
meaning 214
transition 409
- network 388
protolanguage-language ~ 614
transitive 317, 559
- perspective (/ergative) 551
transitivity 11, 28, 58, 99, 283,
395, 448-65, 559
translation 298
transphenomenal 144
- category 126, 223
-type 209
Traugott 23
traversal algorithm 394
Trevarthen 611
Frier 69,325
binocular perspective 504
Trique 315
Trubetzkoy 39

Index

truth: condition, -al 421


- function, -al 36, 104
Tucker In, 374
Tung 394
two-pronged implement 196
typed feature structure 35,42
typology, -ical:
~ perspective 68-72
-variation 315
semantic - 68

unbounded 181,214, 470


undergoer 491-2
unequal (status in sequence) 51,
117
unfolding: of text 18
- (/ culminating) 304
uninstantiated 113
unique beginner (in taxonomy) 83
unit [see also rank, rankscale j 810, 16, 44, 45, 100, 118, 128,
377, 417, 423
suniversalist 447
unmarked 22
unpacking (of metaphor) 255-9,
284, 545
Unswonh 576
Upper Model (Penman) 35,431
usuality 526, 558
V
Vachek 321
value 69, 96, 152, 188, 288, 509
- restriction 57
Value 561
van Leeuwen 427, 606
Van Valin 469, 476, 491-5, 502
variable 36
variation, variety 321, 324, 503-4
varietal level 83
Vender 437,469, 469n, 495
Ventola 378
verb 237, 304,446,448, 494-502
-phrase 467
- stem 316
verbal 99,130, 134, 504, 514,
521, 559
- clause, process 130,134-6,141,514
- deixis 179
- group 100, 178, 213, 265,
299-300,306-8 [Chinese], 466
Verbiage 151,188
Verkuyl 469
VIEW relation 289
virtual object 275
visual: arts, semiotic 606
-experience 73
vocabulary 199, 568

Index

learned ~ 298
[see also lexis, lexicon]
vocal (/ gestural) 533
voice quality 527
vowel space 69
W
Waddington 71
WAG 374
Wahlsterlll
weather 328-54
~ forecast 552
Webber 102
Weber 507
western science 539
WH-element 100
Whorftn, 26, 233, 272, 424,
445, 568-9
Wienbicka 40, 89,416
Wignell 86, 329, 576
Wilkins 421, 455, 575
Wilks 111
Winograd2n, 380n, 606
Witherspoon 58
Wittgenstein 72, 421
Woods 15, 34n
word 16, 568
based (system) 26, 568
~ class, classification 16,448,
515-7
wording [see also lexicogrammar;
strata] 4, 108, 380
WordNet431
world-oriented (tradition) 416,429
Wortman 597
Wright 411
writing (/ speech) 536
written: language 567
-medium
Wurm 318
Wuthnow 429

X
X-bar425
Y
Yates 596n
Young In

Zadeh35
Zeng 2n, 12,42, 42n, 374, 379,
386, 404, 435, 606
Zipf 553

657

You might also like