Construing Experience Through Meaning
Construing Experience Through Meaning
Construing Experience Through Meaning
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
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
9.2.3
Lexicogrammatical generation
9.3.2
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
398
398
9.3.3
9.3.4
403
411
309
312
314
10.3.3
11.3.4
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
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.
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. Theoretical Preliminaries
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.).
* 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. Theoretical Preliminaries
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.
saw
textual
Theme
Rheme
interpersonal
Adjunct
Subject
Resi-
Mood
Time
Senser
Process
Phenomenon
nom. gp.
nom. gp.
verbal gp.
clause: nonfmite
ideational
-due
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.
11
textual
interpersonal
am
real
Given
New
Given
Theme
Rheme
Subject
Finite/ Predicator
Mood
ideational
Complement
Residue
Camer
Process
Attribute
nom. gp.
verbal gp.
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.
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 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.
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.
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
17
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:
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
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 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
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.
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).
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).
29
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.
31
different aspects
32
1. Theoretical Preliminaries
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
33
34
1. Theoretical Preliminaries
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).
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:
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."
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
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).
39
of
+ 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
41
more concepts specify another concept, they may constitute a disjoint covering, which
means that their disjunction is exclusive.
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.
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.
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
45
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
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.
47
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
49
50
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.
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,
52
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.
53
2.3 Figures
54
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
56
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
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
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
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,
61
general
person, man, woman,
boy, girl, baby
horse, stallion, mare,
foal; dog, bitch, puppy
ant, butterfly, slug
meteorological
scattered clouds
history, mathematics
government, school
culinary
cook ['you']
[only as ingredients]
[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
implements: knife, saucepan. These and other examples are tabulated in Table 2(2)
above.
63
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
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
66
67
68
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.
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
71
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
dimension in space
intersecting dimensions
nw)ti4tmfitionality
region (along dimension in space)
chreodization of dimension
curves in space
72
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
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:
75
76
Fig. 2-11: Construing the abstraction 'balance' within the ideation base
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.
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
78
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
Why did the cat go out? Mummy, why did the cat go out?
It gets fed up, having its tail squashed.
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
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
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).
80
Nigel
cat
creature:
'mouse*
food:
vegetable &
fruit; milk
Actor :
Nigel
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.
81
82
(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
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
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.)
85
life form
(unique
beginner)
(kind)
few in number;
polytypic
generic (basic)
specific
varietal
large in number;
core in folk
taxonomy
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).
86
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
87
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
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.
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
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.
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
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)
93
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
In discourse, the whole may serve as the anaphoric domain of a pan ('bridging1, cf. Clark,
1975):
95
and 2-6. We can trace one further example from sequence via figure to element: see Figure
2-17.
96
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
98
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).
99
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
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)
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
(i) ideational
process
participant
circumstance
TRANSITIVITY: role in
figures
Process
class of unit
verbal
nominal
adverbial;
prep, phrase
DEIXIS
TENSE
(temporal)
MODIFICATION
grammatical items
INDIVIDUATK^
DETERMINATION (reference
spatial)
lexical & grammatical items
Proper names
as well as class names
(ii) interpersonaJ
MOOD:
WH-SELECTION
(excqrtifcoostnjedasRange:
fawkat&c)
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]
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)
103
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.
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).
105
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
107
between the two orders of metaphenomena and the presence of consciousness in the act of
projection.
108
3. Sequences
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
110
3. Sequences
3.3 Projection
111
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
113
3.3 Projection
information proposition
statement
I've
question
Have you done the laundry?
command
Do the laundry!
The term proposition is not used here in its sense in logic, but in contrast with proposal (see
immediately below).
3. Sequences
114
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! "
proposal
reporting
bound: finite
quoting
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
to do the laundry
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.
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:
1 figure:
xp
+2
+2
a
+2
a
xp
1
=2
fleures In sequence
[1] if you go into a bank
2
1
"2
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?
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.
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.
N.B. Highton & R.B. Highton. 1964. The Home Book of Vegetarian Cookery. London
Faber and Faber.
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
cook
for
only 10 minutes.
Drain
from OPEN.
it m
corresponding to the
122
3.Sequences
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.
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
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:
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
the accounts
(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
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.
129
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).
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
Sayer
being &
having
she
says
Carrier
she
Token
is
she
is
Receiver
to her brother
Receiver
to her brother
Attribute
a lawyer
Value
his lawyer
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
131
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
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)
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).
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
135
David thought
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
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.
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
4. Figures
138
I think
They want
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
Birthday Party)
emotive &
perceptive
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
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).
140
4. Figures
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 ...'.
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)
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.
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
membership ascriptive:
Carrier + Process + Attribute
frogs are amphibians
(amphibious)
their heads;
verbs are
defined
on the
basis
aspect,
of
potential
especially
ua
4. Figures
146
occurrence after
ke
membership ascriptive:
Carrier + Process + Attribute
The sports centre has
(includes) a health clinic
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.
147
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
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
icicles formed
(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
We have seen that figures of being, other than the existential, may be elaborative
(intensive), extending (possessive) or enhancing (circumstantial). This enables us to
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
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'
'kite + fly'
doing &
happening
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. Figures
150
generalizing over the whole. These constitute distinct but complementary perspectives:
see Figure 4-7.
151
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).
153
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.
155
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
We can also see this in relation to more delicately specified subtypes such as:
type of figure
shine
ache
decant
neigh
dress
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*
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
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
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
159
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
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
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
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.
ext.
enb.
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
[b]
elab.
doing
being
doing to/with
assigned
lava
bard
non-assigned
ext.
enh.
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.
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
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
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.
12
Cf. Hopper & Thompson (1980) on degrees of transitivity and Taylor (1989) on
categorization in general.
167
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
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
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]
[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.
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 tennis
make a mistake
do the regular cleaning
(2) saying
ask
ask politely
ask a 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
(1) doing
cross the street
climb the mountain
enter 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
(1) saying
discuss the issue
talk politics
describe your holiday
(2) sensing
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).
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.
(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),
two volumes,
173
entity as circumstance:
(1) doing
shoot the pianist :
grab somebody :
paint the wall :
buy mother a present :
(2) sensing
guess the answer :
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
175
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
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
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.
verbal group
nominal group
adverbial group; prepositional phrase
conjunction group
participants:
nominal group
a shallow tin
invaluable advice
a tasteless vegetable
powdered white sugar
thin slices of lemon
ballpoint pen remover
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).
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
prototypes, see Hopper & Thompson, 1985.)This complementarity is reflected both in the
similarities and in the differences between the two.
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
181
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.
183
5.3 Participants
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
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.
185
5.3 Participants
where passengers
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
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'
187
5.3 Participants
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
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
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)
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
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
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
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
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;
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)
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
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)
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 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
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.
5. Elements
200
sex
formality
part
supporting
: firm
supporting
: soft
nonsupporting
[soft]
"
inner &
"
"
"
body
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
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
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.
203
5.3 Participants
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
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
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
5. Elements
204
extension
subtype
symbol
kind
instance
part
amount
collection
extension
enhancement
time
place
cause
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
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:
doing:
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.
quality as Manner:
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
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.
211
5.3 Participants
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
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
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
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).
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
(Range)
(Actor)
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
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
Cause: condition
223
Fractal types:
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
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
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.
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
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
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.
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.
(nominal group)
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
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.
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
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!
235
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.
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
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
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
The circumstance could be either some quality of the process or some participant that was
indirectly involved:
circumstance (quality)
circumstance (indirect
participant)
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
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
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. :
These Anglo-Saxon resources were reinforced and largely overtaken as productive devices
by those borrowed from Greek and Latin, e.g.
analyse verb:
nation noun:
behave verb:
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
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
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;
244
6. Grammatical metaphor
245
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
[aA P:'purpose']They broke the door down so that ('in order that') they
[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
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
with
phase;
contingency
circumstance:
minor process
accompaniment; to
destination
ii
prepositional
phrase
Location, Extent
&c => Classifier
conjunction
noun
verb
adjective
Conjunctive
Thing
minor process +
thing
dust
4
5
relator
tfsf> condition
process:
quality
247
[poverty J increases
event
s> increasing
[poverty]
Auxiliary
Epithet/ Classifier
ii
was/ used to
tense; modality
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
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
Conjunctive ->
Location, Extent
&c
so 3*> as a result,
circumstance:
minor process
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
(phase &c)
13
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
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
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.
"Their
13.i(b)
249
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
250
6. Grammatical metaphor
2i/l + 13i(a)
(3)
13i(b) + 2i/l
indecision (indecisiveness)
(4)
6ii + 2i/l
(5)
13i(b) + 2i + 13i(a)
(6)
13ii + 2i + 12i
(7)
13ii + 2i + 13i(a)
accused
(9)
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
(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
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
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
elements:
participant
participant
the accused
thing
quality
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
(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
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
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
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
12iii (phase
process)
start
participant
(they)
elements:
participant
thing
quality
process
circumstance
6.5.2
Paradigmatic
255
participant
they
circumstance
(in the early
afternoon)
2i (process
thing)
discussion
12i (+ process)
took place
13i(b) (thing
expansion of
thing: possessive)
their
and steps
in
"unpacking"
256
6. Grammatical metaphor
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
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!
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
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)
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
(he
is associated with
causes/ is caused by
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).
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
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:
261
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
262
6. Grammatical metaphor
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
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.
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.
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
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
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:
relator:
(equal) so
(unequal) because
267
a happened; so x happened
* happened, because a happened
quality: causal
thing: cause
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
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
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
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
271
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
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.
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)
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,
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
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
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
for
the
mlugglmhu+mm
real GNP
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
financing done
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
278
6. Grammatical metaphor
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
and
then
and
then
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.
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
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.
Deictic Epithet
Token:
prep, phrase complex
prep, phrase
[3] an
prep, phrase
political
Classifier
resurrection
Thing
6. Grammatical metaphor
282
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.
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
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
reduced
or an engine
failed
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
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
6. Grammatical metaphor
288
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
289
metaphorical token-value relation; but tbe expansion is still within the semantic system
itself.
6. Grammatical metaphor
290
::
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
293
294
6. 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
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
projection
expansion
elaboration
extension
enhancement
sequence
He added and
smiled
figure
He ordered the
dismissal of the
chief.
The conclusion of
the treaty was
important
He added witha
smile
298
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
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.
completive:
zh
break
zhdun
break in two
jin
jindun
cut in two
qi
qidun
cut in two
kn
kndun
chop down
shui
shuidun
300
directional:
chuqu
'go out')
kan
look
Z.U
walk
zuch
walkout
fen
separate
fench
distinguish
gn
chase, drive
gnch
drive out
kanchli
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
301
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
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
302
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/
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
(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
(i)
(ii-a)
(ii.b)
(ui)
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'
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
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;
7.5 Elements
But it is in the construction of elements that Chinese and English differ most although
even here the similarities
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
sh
po
'ftir
(b)
(c)
(4)
yl su fangzi
pidgin piece, piecee) which in certain contexts replaces the specific individuating term.
Hence the 'entity1 is anything that can be measured as
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
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.
W5
gngcai
gi
just now
to
ta
da
dinhu,
dbtng
him
call
phone
call-not-through
308
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
mi jin
culminative
neutral
neutral
completive
mijindun
culminative
completive
This contrasts with the English construction of polarity in association with primary and
secondary tense.
huyuan
at
garden
but other, more abstract locations (including temporals) are construed in the same way:
zi
bozhi
shng
7.5 Elements
at newspaper
zi zh zhng
309
top
qngxng zhxi
310
qi
eut
duo
jin
&
cM
xio
xi
li
P*
kan
chai
gu
ji
na
bo
hui
png
in cupped hands
tio
dan
tai
by handle
kng
gong
duan
ji
underarm
han
inside mouth
dido
ding
on head
bi
on back
di
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.
311
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
qch
&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
ynshuj
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
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
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
lu
rate
piril ('repeat')
frequency
slii ('fast')
speed
ylnl ('sound')
musical pitch
hxl (breathe)
respiration rate
turnover
zhshu ('refract')
index of refraction
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
**
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
de
by
two
fragment
straight
between
faults
whose
two faults,
planes
are
construing the
means either as a full process in a separate figure or as a minor one in the form of a
315
316
Process
317
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
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
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
something else through labour and effort by one's strong right hand"); moextending far class prefix, resulting in mo-modyi
("something that goes afar when using one's strong right hand for paddling"); ovamangrove class, resulting in oya-modyi
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.
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,
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
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
324
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.
326
still being specific enough to allow us to construe particular semantic features (types of
figure, elements, etc.).
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
Introduction
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.
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
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)
332
II. Tenor
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
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
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.
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
(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.
337
(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
(1) be in existence
ntinwitioctur
338
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
^groJxnflS
may
set
off
scattered
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)
^ 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.
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
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
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.
severe thunderstorms
are possible
in
Colorado and
342
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.
(1)
Bel = 'exist, come to exist, continue to exist, cease to exist'
343
344
Junctions
Probability Attribute
Probability = Existent
Tune = Carrier
Attribute Existent
process (weaken,
diminish, end,
increase, clear)
participant: quality
participant: thing
participant: thing
participant: thing
345
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;
Sunday,
in the north.
highs 80s.
346
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:
(2) participants:
(3) circumstances:
(Time)
347
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
Be3
become ...
keep ...
remain, stay
become ...
make, keep
Attribute =>
Process
Be4
ascribed)
BeS
Be6
have, get
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.
348
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
349
(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
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
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).
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
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
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
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
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
356
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].
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
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.
358
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.
360
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.
361
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
the
362
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
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.
363
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]
sensemeasure
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
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
Extension:
adversative: They mil become discoloured
Enhancements:
and ['then'] warm in the
temporal: point & unequal: When the pan and fat are hot,
pour in the
until the
brown.
happening limit: (i) 'until1 place in a hot oven 375 F. 425 F.
until
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 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
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
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.
367
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).
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
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
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
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.
371
Table 8(9): Figures of doing, showing types of Agent and Medium found in culinary
domain (italicized)
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.
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.
374
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.
375
376
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]
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
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
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.
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;
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
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
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
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
385
386
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".
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.
(5) continue
388
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:]
stuffing
Use
Variation
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.
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
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
metafunctional regions (ideation base, interaction base, text base), as shown in Figure 9-
9:
392
(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.
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.
394
395
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
[nominal group]
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
397
Subject
Finite/ Predicate*
Complement
Medium/ Carrier
Process
Range/ Attribute
[nominal group]
398
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
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.
400
fun out
of
'sidewalk1,
etcetera .
B: ton.
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.
(cf. Halliday, 1985: 324-5), may serve to create two quanta of information,
402
the
the
Blue Mountains
as
they would
in
the
Catskills:
turn
Here the metaphoncal clause summer finds campers and hikers descending on the
mountains in throngs
waves, was
403
404
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.
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,
surviving subsequent
Alternatively, we can define processes in terms of the semantic network processes for
406
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).
407
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.
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).
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
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
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
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.
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
message:
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
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
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).
416
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
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
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
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
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
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
421
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
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
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
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
combination
of
adverse
currency
effects
and
eluggish
of the
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.
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
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
(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
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)
427
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
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
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
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.
430
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
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
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.
433
434
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
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)},
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
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.
43 7
438
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.
439
440
442
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
443
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?
We should note that FiUmore (1968) was carefully worked out, taking grammatical criteria
into consideration, including Whorfian cryptotypes.
444
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.
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
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,
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)
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
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.
the
[Process]
449
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.
450
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.
451
DOCK. (...) The ducks are a large and well-defined group of birds
of
the
order
Anseres
and
the
family
Anatidae.
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
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
(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
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.
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'
453
454
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.
Position
Transitivity model
Classes
[il
relational (ascriptive)
[ii]
material
[iii]
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.
455
relational
- Predicate
Process
Predicate
Attribute
(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
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
[...]
splendidly,
t...]
The ducks are easy to tame or
The ducks are easily tameable
i...]
457
erosion:
TYPE OF OCCURRENCE
habitual
potential
instantiation
material
relational
Socrates is a good
swimmer
Socrates is a
matutinal swimmer
instance
Socrates swam across
the river; Socrates is
swimmmg
4-58
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.
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.
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
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
[Process] (very)
[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
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
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
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.
463
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
(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
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
He
had
hatred
of dogs
metaphorical
Carrier
Process
Attribute
Thing
Qualifier
Senser
Process
Phenomenon
He
hated
dogs
Deictic
congruent
465
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.
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
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.
468
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
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).
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).
471
472
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
have, possess,
dominate, rule
(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
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
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
475
activities
accomplishments
achievements
states
doing
sensing
saying
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
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.
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.
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
Intentionality
478
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
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.
480
motivated"; agents are "largely animate beings" but also include e.g. heat, wind, ship.3
Chafe characterizes these things as 'potent*.
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.
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
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
Circumstances
cit.)
the
sense of
progress
10
checked
by
depression
or
new
systems
of
transportation
and
corntuni cation
. . . (opcit, 278)
own distinctive
483
outer participant =
state
[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.
484
process
action
action-process
experiential
benefactive
OsProc
broken, dry, dead,
tight
[relational]
Obj Proc
break, die, dry,
tighten
AgProc
dance, laugh, play,
sing
[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
[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
dynamic
static
controlled
uncontrolled
action
process
position
state
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
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 remained
in the hotel
for my sake
John kept
himself in
balance with a
counter -weight
The t*:* *U
dovm for sty
****
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
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
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
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).
489
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
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
process
John fell
down
sensing
being &
having
The dog
frightened
John [effective]
position
state
saying
John stayed
motionless
John saw a
beautiful bird
[middle]
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.
491
492
Relationship between
RRG
SFG
difference of delicacy
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
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
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
specific case roles. Foley and Van Valin give the following examples to illustrate this
point:
Colin killed the taipan
Actor/
Agent
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:]
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
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
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.
496
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
..
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
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
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.
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.)
500
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.)
have
501
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
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
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.
in systemic
504
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
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
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
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
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
511
provide an opportunity to put the ideation base hack into its wider social-semiotic
environment.
512
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
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
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
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.
516
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
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
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
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
he acted brilliantly as
Hamlet
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
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
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
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
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
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
demand + goods-&-services:
command
give + information:
statement
demand + information:
question
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]
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
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.
527
528
13.2.5 Textual
metafunction
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
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
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
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
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
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
535
536
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'
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
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
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.
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
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
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:
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
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.
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
544
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
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
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
547
548
"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;
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
(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
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
(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
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
:
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
^^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
552
'11 dry/wipe
Process
'11 dry/ wipe
they
Actor
they
transitivity
- Goal
the dishes
+ Agent
the sun
Process
will dry/ fade
they
Medium
them
11 dry/fade
Process
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
(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.)
555
556
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.)
557
558
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,
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
region are more strongly aligned to one perspective, other parts to the other, but the total
picture requires the confrontation of the two.
561
preposition, as in they get caught for taking bribes. What happens here is that the fact
that there is
logical relationship is
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
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
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
complementarities: the
complementarity: the
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
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.
566
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).
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
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
571
5 72
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.
573
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
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'
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
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
577
reviewing the general folk model of such processes, and then explore how this folk model
has been reconstrued in cognitive science.
578
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
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.
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
581
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
'P*P"P
[?].
I think
583
He's working
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
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
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.
585
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
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
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
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
and that,
together
with
uncontroversial,
but does
folk
appropriately
related
psychology
also
have
it
that
iitplicit
belief-deire
introspectible,
and if
conplexes
so,
what
authority
Are
do the
belief
believer's
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
believe
ota:
>
(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
perceptions
Value
nom. gp.
Thing:
nominalized
process of
perception
[Process:] are
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
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
that he's a
at
party
given
by a
friend
to
celebrate
the
thinking
rcogniz
is
organized.
Michael's
face,
Why is
rmmbr
it
that
significant
Michelle can
facts
about
the
590
recall
(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
TEXT
(Processes in bold)
Type of figure
doing sensing
saying being
&
having
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.
592
uhow we
593
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
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:
Actor:
9
?
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
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: [...]
595
596
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
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).
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.
598
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
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.
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
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.
600
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.
601
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
603
604
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.
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;
606
in
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
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
608
609
610
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
environments
for the
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.
612
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
615
616
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
longer,
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
some
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
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
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
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
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
625
References
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. 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
References
629
630
References
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
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.).
References
complementarity
and resonance.
634
References
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.
References
635
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.
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.
References
637
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.)
638
References
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.
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
References
641
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
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
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
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
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
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
Index
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
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
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
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
Padley 455
Partee 422
652
Index
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
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
Index
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