Professional Documents
Culture Documents
M. A. K. Halliday
Continuum
On Grammar
Collected Works of M. A. K. Halliday
Volume 1: On Grammar
Volume 2: Linguistic Studies of Text and Discourse
Volume 3: On Language and Linguistics
Volume 4: The Language of Early Childhood
Volume 5: The Language of Science
Volume 6: Computational and Quantitative Studies
Volume 7: Studies in English Language
Volume 8: Studies in Chinese Language
Volume 9: Language and Education
Volume 10: Language and Society
Volume 1 in the Collected Works of M. A. K. Halliday
On Grammar
M. A. K. Halliday
Edited by Jonathan Webster
Continuum
The Tower Building, 11 York Road, London, SE1 7NX
370 Lexington Avenue, New York, NY 10017-6503
쑔 M. A. K. Halliday 2002
Preface vii
Acknowledgements ix
Introduction: a personal perspective 1
Professor M. A. K. Halliday
Bibliography 419
Index 433
vi
PREFACE
1 On Grammar
2 Linguistic Studies of Text and Discourse
3 On Language and Linguistics
4 The Language of Early Childhood
5 The Language of Science
6 Computational and Quantitative Studies
7 Studies in English Language
vii
preface
viii
ACKNOWLEDGEMENTS
ix
acknowledgements
x
INTRODUCTION:
A PERSONAL PERSPECTIVE
1
on grammar
what I did stress was how much linguistics had in common with other
scholarly pursuits and, when it came to asking questions about
language, I always found myself lining up with the outsiders. It seemed
fine for us, as linguists, to determine the content or domain of our
own discipline – sociologists studied society, psychologists studied . . .
whatever psychologists do study, linguists studied language.1 That’s
what we were there for. But it did not seem fine, to me at least, for us
to determine what questions should be asked about that domain. I was
interested in what other people wanted to know about language,
whether scholars in other fields or those with practical problems to be
faced and solved – including that undervalued and under-rewarded
group who have to be both scholars and practical problem-solvers,
namely teachers.
There was nothing surprising about this last perspective: not only
had my parents both been teachers, but I myself had taught languages
for thirteen years before transferring myself into a teacher of linguistics.
But even before that, while still at school, I had been trying to find out
about language – because I was keen on literature and wanted to
understand why its language was so effective, what was special about
it. There is no separating one’s personal history from the academic
paths one pursues, nor any way of detaching cause from effect in
explaining one’s chosen approach to a field of study. One way or
another, I have always found myself asking the kinds of questions about
language that arose, as it were, from outside language itself.
Of course linguists always have been located, and located themselves,
within some broader context; there is nothing unusual about that. But
at any given “moment” in space-time, there are likely to be only a few
predominant motifs by which the context of linguistic scholarship is
defined. This may even be legislated from on high, as when Stalin
writing in 1950 (or Chikobava, writing on his behalf) instructed Soviet
linguists to get on with the job of demonstrating the linguistic unity of
the Slav nations – reasonably enough, since the Soviet Union had just
taken them all over. Usually it is determined by less overtly political
factors: by particular social movements and demands, or notable
advances in knowledge in some other field.
The present era provides a noteworthy example of the latter. Since
about 1985 there has been spectacular progress in the field of neuro-
science; the combination of new technology – positron-emission
tomography, magnetic resonance imaging and its derivatives – with
new insights in evolutionary theory and its contributing disciplines has
transformed the way we understand the human brain, how it has
2
a personal perspective
evolved in the species and how it develops in the individual from birth
(and before) to maturity. And this new understanding has radically
redefined the place of language. It is now clear that language and the
brain evolve together, and that these develop together in infancy and
childhood. The development of the brain is the development of the
ability to mean; as in every aspect of human history, so in the ontogeny
of the individual human being the material and the semiotic inter-
penetrate, as complementary aspects of the characterology of the species
(McCrone; Edelman; Deacon; Dawkins; Jones).
To say this is not to proclaim that the human species is unique in
this respect or that no other species has evolved, or could evolve, a
similar type of higher-order semiotic. On the contrary. The work of
Duane Rumbaugh and Susan Savage-Rumbaugh has brought out the
point that the bonobo chimpanzees can operate with sets of arbitrary
symbols in a way that is analogous to our own system of wording
(lexicogrammar). They lack an analogous vocal apparatus, but that is
beside the point. It is tempting to assume that they have been following
the same evolutionary path and are simply less far advanced along the
way – this is the assumption that prompts questions like ‘what age have
they reached, in terms of a human child?’ But this assumption is
probably wrong, or at least misleading, if it is used to describe an adult
chimpanzee in terms of an immature human; the adult bonobo’s brain
is fully wired up in terms of the construing of experience, and enacting
of social relations, that constitute bonobo culture. The question can
fairly be asked about bonobos brought up from birth in a human-like
semiotic environment, like Kanzi, and it is too early to say yet whether
Kanzi and the other youngsters’ development of the power of meaning
tracked that of human children and stopped at a certain level or
whether it was proceeding along a somewhat different route.
It might be argued that such new knowledge about how the brain
functions, and how it evolves and develops, has no significance for the
way linguists describe and explain language, especially at the ‘inner’
strata of lexicogrammar and phonology (wording and sounding). Pos-
sibly; although even here it seems to me to set certain constraints and
more importantly perhaps to favour certain explanations over others. It
suggests “systems thinking” rather than compositional thinking (Mat-
thiessen), grammatical logic rather than formal logic (Sugeno 1995),
fuzzy and probabilistic categories rather than clearly bounded and
deterministic ones. Since the brain is more like a jungle than like a
computer (Edelman 1992), it disfavours representations of grammar and
phonology that are influenced, however indirectly or subconsciously,
3
on grammar
4
a personal perspective
5
on grammar
6
a personal perspective
7
on grammar
the century were technological ones: the invention of the tape recorder
and the evolution of the computer. The tape recorder made it possible
to record natural speech. The computer made it possible to process
large quantities of data. The two together have given us the modern
computerized corpus, with natural speech as a significant component,
on which we can undertake quantitative analyses on a statistically
significant scale. As a bonus, the computer enables us to test our
descriptive generalizations, through text generation and analysis (“pars-
ing”), and to observe and represent sound waves in a wealth of
complementary perspectives.
These resources have transformed (or at least are in process of
transforming) the way language looks from the inside. Patterns are
being revealed that we have known must be there, because there was a
gap where the approaches from the lexical and the grammatical poles
of the lexicogrammar converged, but which we could not see: the
nature of grammatical logic is beginning to be understood; the semo-
genic (meaning-creating) power of discourse is coming into view, both
in monologic and in dialogic mode; quantitative mechanisms of lin-
guistic change are beginning to appear on the agenda. From all this it
should be possible in the next decade or two to crack the semiotic
code, in the sense of coming fully to understand the relationship
between observed instances of language behaviour and the underlying
system of language – something that has eluded us up till now, so that
we have even turned the two into different disciplines, calling only one
of them “linguistics” and labelling the other “pragmatics”.
Some people will feel threatened by this new understanding. We
know this because there are those who already do. To bring to light
the systems and processes of society is already threatening enough, as
witness the panic reactions to Bernstein thirty years ago when he
demonstrated how social class structures are transmitted, but semiotic
systems and processes are even nearer the bone. As long as linguists
confined their attention to dead languages, codified texts or sanitized
examples like John kissed Mary and It’s cold in here, no one would feel
really at risk. But when grammar extends to the study of the meaning-
creating power of everyday real life talk, it starts to become dangerous.
Some people feel worried that the grammarian is someone who knows
what they are going to say next and even if they can be reassured that
that is not what theory is about, it is scarcely less threatening (appar-
ently) to be told what proportion of positive to negative clauses they
are going to use in their speech. And for others, just to be faced with a
record of real life conversation can be unnerving; they feel embarrassed
8
a personal perspective
9
on grammar
10
a personal perspective
11
on grammar
12
a personal perspective
Notes
1. Psychologists, in fact, study psychology – the domain is defined by the
discipline, rather than the other way round. Hence the rather odd locutions
like “criminal psychology”, meaning the mind-set, or psyche, of criminals,
rather than psychological theories that criminals have devised. I was once
put down rather scathingly by a psychologist for suggesting that their
domain of study might be the human psyche.
2. See Volume 4 in this series.
3. Language is a system of meaning (a “semiotic” system); and semiotic
systems are of the fourth order of complexity, being also physical and
biological and social. This means that one and the same linguistic phenom-
enon (whether “a language” or a single utterance by one speaker) will
appear in all these various guises.
4. I hope it will be clear that I am not seeking either to justify this approach
or to apologize for it. These bits of personal history are brought in simply
to provide a context, to explain the way the papers in these volumes
13
on grammar
14
SECTION ONE
EARLY PAPERS ON BASIC CONCEPTS
This page intentionally left blank
EDITOR’S INTRODUCTION
17
early papers on basic concepts
18
editor’s introduction
ordered repetition of like events that make up the patterns’. There are
both primary and secondary structures, distinguished in terms of delicacy, or
depth of detail. Whereas class involves ‘the grouping of like events by
their occurrence in patterns’, system deals with ‘the occurrence of one
rather than another among a number of like events’. To help the reader
better understand the application of the categories of grammar, Halliday
presents a framework of categories for the description of another very
familiar kind of patterned activity, namely, eating a meal. Looking back
on this chapter after forty years, Professor Halliday provides some
background from his personal history to help readers better understand
his very careful concern for assigning things to categories:
19
early papers on basic concepts
20
Chapter One
First published in Studies in Linguistic Analysis (Special Volume of the Philological Society),
1957, Oxford: Blackwell, pp. 54–67.
21
early papers on basic concepts
22
systematic description and comparison
23
early papers on basic concepts
Firth’s expression, ‘on the agenda’, since what is formal when particular
or comparative tends to become imaginative when universal. One
might then summarize diagrammatically:
Descriptive Historical Evolutionary
—
Particular — —
—
—
— —
Comparative — — — —
— —
— —
——
Universal —— ——
——
——
The “structural” linguist, while attempting to develop descriptive
methods that are general, in the sense that they are scientific methods
universally valid in linguistic description and forming part of general
linguistic theory, will be unwilling to claim universality for any formally
established category; since, while, for example, it may be convenient
in the description of all languages so far studied to give the name ‘verb’
to one class of one unit, this is not a universal statement: the ‘verb’ is
redefined in the description of each language. Even if any one category
can be identified (presumably contextually) across all languages studied
so far – and this is still a long way off: possibly, if a unit ‘word’ can be
satisfactorily universalized, the word-class ‘personal pronoun’, as rela-
tively easy to identify contextually, may be the first – such a category
is universal only in the limited range of what is and has been, not what
might be. At present any universal system of categories must rest on
other than formal linguistic criteria: if such can be provided, for
example, by mathematics, so much the better – the “structural” linguist
will not “reject” it, but he cannot be expected to provide it within his
own terms of reference.
What “structural” linguistics has done has been to concentrate, partly
in redress of the balance in linguistics, on the first of the methods of
description outlined in the first paragraph above. It is in fact by the
study of systems and structures within the framework of particular
description that this body of theory, and the methods derived from it,
24
systematic description and comparison
25
early papers on basic concepts
the text other criteria, phonetic or graphic, may contribute and may be
taken, where to do so is compatible with the general aims of simplicity
and comprehensiveness, as the primary or even sole criteria (for
example, punctuation or spacing in a written text, features of intonation
of pause in a spoken text, all of which then enter into the grammatical
description). It is probable that in the description of any language at
least two units will be required: these would be such as could be named
the sentence and the word.3
A descending order of procedure seems preferable not only for the
presentation (where indeed it may be varied for a particular purpose),
but also for the analysis of the grammar, where in such a hierarchic
progression the classification made at the level of each unit will itself
determine the classes that are to be set up for the lower units. One may
begin by establishing, and delimiting the exponents of, that unit (which
we may call the “sentence”) which, while within the scope of gram-
matical statement – not so extensive as to be incapable of systematic
analysis – is yet enabled to operate as the linguistic action of participants
in a situation: which is, in fact, “living language” and constitutes the
unit of analysis at the contextual level. In the subsequent establishment
and classification of the lower units, the statements made about each
unit will be related to values set up in the structure of the higher unit.4
For each unit there will then be set up systems of classes, formally
established in the grammar and exhaustive, such that statements may be
made which are valid for all exponents of a given unit. These classes
are set up independently of structure: that is to say, a unit (for example
clause) having been established, it is then classified by reference to
various sets of formal criteria (for example presence of, or ordering of,
certain formally defined elements); each set of criteria permits the
establishment of one system of classes (“clause–classes”). Such mutually
independent systems of classes of any unit are referred to as ‘dimen-
sions’: thus one dimension of clause–classes might be the aspect
dimension, with a system, say, of two terms, perfective and imperfective.
There may be any number of dimensions of classes for each unit, and
the system of any dimension may admit of a neutral or ‘unmarked’
term, but each dimension will by itself form an exhaustive system of
classification such that every exponent of a given unit may be placed
within it. Thus it might be that all words are either red or blue or
yellow and all words are either square or round or neutral in shape.
Two dimensions of classes are implicit in the taxonomic hierarchy of
the unit system: as characteristic of each unit except the lowest there
appears the dimension compound / simple, while for each unit except
26
systematic description and comparison
27
early papers on basic concepts
28
systematic description and comparison
29
early papers on basic concepts
30
systematic description and comparison
3 tā { he
she
12(2), 13(3), 12(2)3(3)
12(2), 12(2)3(3)
wǒmen
zámen
} we
31
early papers on basic concepts
4 Descriptive comparison
In a descriptive comparison there is no implication of genetic relation-
ship; but beyond the application to particular description permitted by
comparison of this type (since any l.u.d. may be compared with any
l.o.d.) systematic comparison itself connotes a wider purpose, seen
initially in the establishment of degrees of relationship. Here the gram-
matical identification of terms across languages is an essential basis of
comparison; and while much can be done on contextual criteria (though
not as much as in the establishment of lexical systems for comparative
purposes) some formal procedure must be found in order that statement
of degrees of relationship may assume a general significance.
Possibly, however, we should not impose a strict demarcation
between identification of grammatical terms on grammatical and on
non-grammatical (e.g. contextual) criteria. Allen, rightly rejecting both
terminological identification and identification by translation, considers
that identification by formal grammatical criteria (as opposed to the
situational–contextual criteria which are available for the semantic
identification of lexical items) seems not to be possible (Allen 1953: 99,
100). Some attempt at identification may be made with the Chinese
dialects, where speakers of more than one dialect constantly make such
identifications in practice, with or without phonological resemblances:
for example, from the personal pronouns, a Cantonese speaker equates
ne˛i with Pekingese nı̌ and, as readily, ke˛udei with Pekingese tāmen. This
identification in practice demonstrates the¯ contextual basis of the iden-
tification; its validity on grammatical criteria may be tested by reference
to the place of the class “pronoun” in the system of word–classes.
When we find that it is possible to describe both Modern Pekingese
and Modern Cantonese in terms of the same units of sentence, clause,
word and character; that at the word level we can set up for both a
three-term system of classes, verbal, nominal and adverbial, and that
one term in the system of nominal word–classes, the auxiliary noun,
enters in both into the noun group with identical position and value in
the structure, we can regard the class of auxiliary noun in the two
dialects as comparable, and are then justified in seeking to identify
32
systematic description and comparison
33
early papers on basic concepts
there may be only one nasal final. If we compare these systems in certain
dialects of the Mandarin, Wu and Yüeh dialect groups, we find that it
is prosodically linked, and identical in number of terms, with the system
of plosive finals or, if the latter is absent, with that of vocalic finals.
Cantonese shows the final plosive system p/t/k with prosodically
identical nasals m/n/ŋ; in Shanghai, where there is a single final plosive,
the glottal closure, there is one nasal final, with varying point of contact;
while in Pekingese, where there are no final plosives, the two terms of
the final nasal system are prosodically identical with the oral finals, so
that n:ŋ::i:u.11 The calculation of the degree of relationship will depend
on the number of final systems set up for each dialect and the number
of these in which all terms can be identified; but the comparison
immediately suggests the historical interconnection between the disap-
pearance of the final plosive system and the elimination of one term
from the final nasal system in the dialects where these have occurred.
With regard to the study of languages in geographical proximity in
an area where no, or only partial, genetic groupings have been estab-
lished, systematic comparison can determine whether the question of
grammatical affinity is to be posed at all, and if so in what form. The
superficially apparent affinity among the languages of certain areas has
long been the basis of traditional typology. An instance of how it may
be demonstrated or disproved, initially as a function of systems, might
be found in the grammar of East Asian languages. It seems possible to
set up in, for example, Pekingese, Cantonese, Vietnamese, Siamese and
Malay systems of nominal word–classes which would permit the iden-
tification as comparable of certain terms, including the category discussed
with reference to Pekingese and Cantonese above and which may be
called the auxiliary noun. If the place of the auxiliary noun in the system
of word–classes, and the terms in the various systems of the auxiliary
noun, can be compared as between these languages on the basis of
adequate criteria, we may determine whether or not a formal description
of these languages would reveal for these and other systems anything
that could be regarded as grammatical affinity – such as, that is, to
exclude from such affinity any systems set up for other languages in
other areas. Only if such affinity could be established among a significant
number of systems in the grammar of these languages would it be
possible to raise further questions of general grammatical affinity and –
as yet only an interesting speculation – of geographical gradation.
Such a systemic comparison may help to resolve the difficulty that,
on the one hand, there appears an obvious but unformulated grammat-
ical similarity among the East Asian languages as contrasted, for example,
34
systematic description and comparison
with the Indo-European languages, while on the other hand there exists
the quite justifiable scepticism among linguists either as to how this is
to be explained, in comparative historical terms, where there are no
correspondences, or as to whether there is anything to be explained (or,
in systematic terms, to be stated comparatively) at all.
Notes
1. The study of the “form of the content” by plerematics, as envisaged by
Hjelmslev, would set up particular or comparative (i.e. non-universal)
categories; but the criteria for such categories, while as yet inadequately
defined, would also be such as to be regarded as formal-linguistic. A
method of classification of words on the basis of universal categories of
relation and description is to be found in the work of Brøndal (1948).
2. Seventh International Congress of Linguists, Preliminary Reports, London,
1952, p. 53: ‘Can a purely formal grammatical analysis be carried out on
languages such as Chinese, in which all or nearly all the words are
invariable, and if so, on what principle?’
3. Provided the principle of particular formal description is adhered to, the
choice of current terms seems preferable to the creation of new ones.
There is, however, no reason why, especially in the initial stage of the
process of analysis, completely non-committal terms should not be
employed. The practice of some linguists of talking (at least to themselves)
about e.g. red words and blue words can equally well be extended, for
example, to “strings” and “bits”.
4. Compare the descending order within the levels of linguistic analysis in
which meaning may be stated, as suggested by Firth (1951: 121). For an
example of the employment of this method of descending analysis in
grammatical statement, cf. Robins (1953) and Halliday (1956).
5. As pointed out by Robins (1953: 109), Firth has indicated how system
and structure require to be kept separate in General Linguistic theory. I
have attempted to follow Firth’s view of the two as distinct but related
concepts.
6. The linguistic unit of which the written character is the graphic symbol
corresponding to the syllable at the phonological level. The use of the
term ‘character’ follows the Chinese practice of calling both the linguistic
and the graphic units by the single name tzǔ (zı̀).
7. The concept of ‘context of situation’ here intended is as developed by
Firth in linguistic analysis, where it ‘is best used as a suitable schematic
construct to apply to language events’ and should be regarded as ‘a group
of related categories at a different level from grammatical analysis but
rather of the same abstract nature’ (Firth 1952: esp. 7).
8. Professor Allen has kindly drawn my attention to the paper entitled
35
early papers on basic concepts
36
Chapter Two
37
early papers on basic concepts
derives most of all from the work of J. R. Firth.4 At the same time I
do not of course imply that I think Professor Firth would necessarily
have found himself in accord with all the views expressed, which in
some places depart from his own; nor do I underestimate the debt to
my present colleagues and the many others whose work I have,
obviously, drawn on.5
No excuse is needed, I think, for a discussion of General Linguistic
theory. While what has made linguistics fashionable has been, as with
other subjects, the discovery that it has applications, these applications
rest on many years of work by people who were simply seekers after
knowledge. It would not help the subject if the success of these
applications led us into thinking that the theoretical problems were
solved and the basic issues closed.
1 Starting-point
It will perhaps be helpful if the point of departure is first made clear.
The following is a summary of what is taken as “given” for the
purposes of this paper.
38
categories of the theory of grammar
1.5 The primary levels are form, substance and context. The substance
is the material of language: phonic (audible noises) or graphic (visible
marks). The form is the organization of the substance into meaningful
events: meaning is a concept, and a technical term, of the theory (see
below, 1.8). The context is the relation of the form to non-linguistic
features of the situations in which language operates, and to linguistic
features other than those of the item under attention: these being
together “extratextual” features.
39
early papers on basic concepts
1.8 Language has formal meaning and contextual meaning. Formal mean-
ing is the “information” of information theory, though (i) it can be
stated without being quantified and was in fact formulated in linguistics
independently of the development of information theory as a means of
quantifying it,11 and (ii) formal meaning in lexis cannot be quantified
until a method is found for measuring the information of non-finite
(“open”) sets (see below, 2.1 and 8.2). The formal meaning of an item
is its operation in the network of formal relations.
1.10 It follows from 1.8 and 1.9 that, in description, formal criteria
are crucial, taking precedence over contextual criteria; and that the
statement of formal meaning logically precedes the statement of contex-
tual meaning.14
2 Grammar
2.1 Grammar is that level of linguistic form at which operate closed
systems.16,17 Since a system is by definition closed, the use of the term
“closed” here is a mnemonic device; but since “system” alone will be
used as the name of one of the four fundamental grammatical categories
(see below, 6) it is useful to retain “closed system” when referring to
the system as the crucial criterion for distinguishing grammar from
lexis.
A closed system is a set of terms with these characteristics:
40
categories of the theory of grammar
2.2 The fundamental categories for the theory of grammar are four:
unit, structure, class and system. These are categories of the highest
order of abstraction: they are established, and interrelated, in the theory.
If one asks: “why these four, and not three, or five, or another four?”,
the answer must be: because language is like that – because these four,
and no others, are needed to account for the data: that is, to account
for all grammatical patterns that emerge by generalization from the
data. As the primary categories of the theory, they make possible a
coherent account of what grammar is and of its place in language, and
a comprehensive description of the grammars of languages, neither of
which is possible without them.
Each of the four is specifically related to, and logically derivable
from, each of the others. There is no relation of precedence or logical
priority among them. They are all mutually defining: as with theoretical
categories in general, “definition” in the lexicographical sense is impos-
sible, since no one category is defined until all the others are, in the
totality of the theory.19 The order chosen here for exposition is
therefore simply that which seemed the easiest: namely the order in
which they are listed above.
The relation of these categories to each other and to the data involve
three distinct scales of abstraction, those of rank, exponence and delicacy;
these are considered separately (see below, 7) but have also to be
41
early papers on basic concepts
3 Unit
3.1 Language is patterned activity. At the formal level, the patterns
are patterns of meaningful organization: certain regularities are
42
categories of the theory of grammar
3.2 The category set up to account for the stretches that carry
grammatical patterns is the unit. The units of grammar form a hierarchy
that is a taxonomy. To talk about any hierarchy, we need a conver-
sational scale; the most appropriate here might seem that of size, going
from “largest” to “smallest”; on the other hand size is difficult to
represent in tables and diagrams, and may also trap one into thinking
in substantial terms, and a vertical scale, from “highest” to “lowest”,
has advantages here. For the moment we may use both, eventually
preferring the latter. The relation among the units, then, is that, going
from top (largest) to bottom (smallest), each consists of one, or of
more than one, of the unit next below (next smaller). The scale on
which the units are in fact ranged in the theory needs a name, and may
be called rank.
“Consists of”, like “unit” and “rank”, also belongs to the theory: its
43
early papers on basic concepts
44
categories of the theory of grammar
when they turn up. For English, for the two units between sentence
and word the terms clause and phrase are generally used. It is at the
rank of the phrase that there is most confusion – because there are here
the greatest difficulties – in the description of English; one reason is
that in English this unit carries a fundamental class division (see below,
5), so fundamental that it is useful to have two names for this unit in
order to be able to talk about it: I propose to call it the group, but to
make a class distinction within it between group and phrase. Below the
word, English has one unit, called by the general name for the unit of
lowest rank, the morpheme.29
So in the description of English the sentence30 consists of one or
more complete clauses, the clause of one or more complete groups, the
group of one or more complete words and the word of one or more
complete morphemes. The descriptive meaning of “consists of”, and
the possibilities of rankshift (including recursive rankshift), are stated as
and where applicable. One distinction that is often useful is between a
member of a unit that consists of only one member of the unit next
below and one that consists of more than one; the former may be
called simple and the latter compound, but if this is done the terms must
be kept rigorously to this, and no other, use.31
3.4 The theory requires that each unit should be fully identifiable in
description. This means that, if the description is textual, every item of
the text is accounted for at all ranks, through the various links of the
exponence chain which involve, of course, the remaining theoretical
categories. If the description is exemplificatory, exactly the same is
implied, except that the description proceeds from category to expo-
nent instead of from exponent to category.
It will be clear from the discussion in the next sections that there
can be no question of independent identification of the exponents of
the different units, since criteria of any given unit always involve
reference to others, and therefore indirectly to all the others. A clause
can only be identified as a clause if a sentence can be identified as a
sentence and a group as a group, and so on up and down the line. For
this reason description is not and can never be unidirectional: it is
essential to “shunt”, and “shunting” is a descriptive method that is
imposed on description by theory.
45
early papers on basic concepts
4 Structure
4.1 The unit being the category of pattern-carrier, what is the nature
of the patterns it carries? In terms once again of language as activity,
and therefore in linear progression, the patterns take the form of the
repetition of like events. Likeness, at whatever degree of abstraction, is
of course a cline, ranging from “having everything in common” to
“having nothing in common”. The commonplace that no two events
are ever identical, that the same thing can never happen twice, is of no
relevance whatever to linguistics; as soon as description starts, however
little the generalization involved, absolute identity is a necessary
hypothesis, which is then built into the theory, as one endpoint of the
likeness cline. Likeness, including absolute identity, is of course re-
defined for each level and each category.
In grammar the category set up to account for likeness between
events in successivity is the structure.32 If the relation between events
in successivity is syntagmatic, the structure is the highest abstraction of
patterns of syntagmatic relations. The scale used for talking about it,
and for its graphic display, will most naturally be the orthographic
scale: to those of us brought up on the roman alphabet this happens to
run horizontally from left to right, which is enough reason for adopting
this version of the scale. But, as in the case of the unit, it must be
stressed that linear progression itself is a feature of substance. A structure
is made up of elements which are graphically represented as being in
linear progression; but the theoretical relation among them is one of
order. Order may, but does not necessarily, have as its realization
sequence, the formal relation carried by linear progression; sequence is
at a lower degree of abstraction than order and is one possible formal
exponent of it.33
46
categories of the theory of grammar
the structure of a given unit is defined with reference to the unit next
below. Each place is the place of operation of one member of the unit
next below, considered as one occurrence. Each element represents the
potentiality of operation of a member of one grouping of members of
the unit next below, considered as one item–grouping.36 It follows
from this that the lowest unit has no structure;37 if it carried structure,
there would be another unit below it.
47
early papers on basic concepts
48
categories of the theory of grammar
5 Class
5.1 The structure is set up to account for likeness between events of
the same rank, and it does so by referring them to the rank next below.
To one place in structure corresponds one occurrence of the unit next
below, and at each element operates one grouping of members of the
unit next below. This means that there will be certain groupings of
members of each unit identified by restriction on their operation in
structure. The fact that it is not true that anything can go anywhere in
the structure of the unit above itself is another aspect of linguistic
patterning, and the category set up to account for it is the class.
The class is that grouping of members of a given unit which is
defined by operation in the structure of the unit next above. It accounts
for a paradigmatic relation, being a grouping of items “at risk” under
certain conditions. It is related primarily to elements of structure: the
first degree of classification yields classes which stand in one / one
relation to elements of primary structures, and these we may call
primary classes.
49
early papers on basic concepts
XYZ, YZ and XYZY, the primary classes of the unit next below are
“class operating at X”, “class operating at Y” and “class operating at
Z”. If, however, there is a further restriction such that in XYZY,
which will now be (secondarily) rewritten XYaZYb, only a section of
the members of the class at Y can operate at Ya and only another
section (not necessarily mutually exclusive) at Yb, this yields as second-
ary classes “class operating at Ya” and “class operating at Yb”. Secondly,
with increased delicacy the elements of primary structure will be
differentiated into secondary elements. A primary structure generalized
as X . . . nYZ, of which XXXXYZ is an instance, shows a generalized
relation of X to (say) Y; but there may be internal relations within X
. . . n such that XXXX is rewritten pqrs. These will yield secondary
classes “class operating at p”, “class operating at q”, etc.
In the second place, more delicate classes appear whenever a
restriction is found which differentiates among the members of a
primary class. There may be a relation of mutual determination, or
“concord”, between two classes; each divides into two sections such
that a member of one section of one class is always accompanied by a
member of one section of the other class. Thus if the primary class at
X is 1 and that at Y is 2, a structure XY must have as its exponent
either 1.1+2.1 or 1.2+2.2. Secondary classes arrived at in this way in
description may be referred to distinctively as sub-classes, to indicate
that they are derived by differentiation from primary classes without
reference to secondary structures; but it is important to state that there
is no theoretical difference here. The relation between structure and
class is a two-way relation, and there is no question of “discovering”
one “before” the other. In any given instance there may be descriptive
reasons for stating the one without the other; but all structures
presuppose classes and all classes presuppose structures.
50
categories of the theory of grammar
51
early papers on basic concepts
6 System
6.1 Up to this point the theory has accounted for three aspects of
formal patterning: the varying stretches that carry patterns, the ordered
repetition of like events that makes up the patterns and the grouping
of like events by their occurrence in patterns. What remains to be
accounted for is the occurrence of one rather than another from among
a number of like events.
The category set up for this purpose is the system.52 This falls under
the general definition of system given above (2.1). But this does not
yet state its place in grammatical theory, its relation to the other
fundamental categories.
The class is a grouping of items identified by operation in structure:
that is, what enters into grammatical relations of structure is not the
item itself considered as a formal realization53 but the class, which is
not a list of formal items but an abstraction from them. By increase
in delicacy, the primary class is broken down into secondary classes of
the same rank. This set of secondary classes now stands in the relation
of exponent to an element of primary structure of the unit next above.
This gives a system of classes. If class 1 is the primary class (say of
the group) operating at X in (clause) structure, and this has secondary
classes 1.1, 1.2 and 1.3, then 1.1, 1.2 and 1.3 form a system of classes
operating at X. X is now shown to presuppose a choice – a choice that
is implied by the nature of the class (as a grouping of items) but that is
displayed first still in abstraction, by reference to the category of class
itself.54
52
categories of the theory of grammar
6.3 There comes a point, however, when one is forced out to the
exponents; and this happens in one of two ways. In the first case the
53
early papers on basic concepts
54
categories of the theory of grammar
55
early papers on basic concepts
7.2 The scale of rank has been discussed (above, 3) with reference to
the unit, the basic category which operates on this scale. The syntactic
(‘downward’) determination of classes is a feature of the theory, so that
with respect to the category of class the rank scale appears as one
of logical precedence running from highest to lowest unit. But this
precedence applies to class criteria only, so that even in the theory it is
not a one-way relation: the theory itself embodies “shunting” (moving
up and down the rank scale) as crucial to the interrelation of the
categories. In description, all statements presuppose shunting; the
description of the sentence cannot be complete until the description of
the morpheme is complete, and vice versa. In presentation, of course,
procedure varies according to purpose and scope: downward presenta-
tion seems easier to make clear, but this may well be overridden – for
example if grammar is an adjunct to lexical description, as sometimes
in the statistical study of lexis.
Rank is distinct both from exponence and from delicacy. A shift in
56
categories of the theory of grammar
one never by itself entails a shift in either of the others. The reason
why rank is often confused with other scales is that there are cases
where a shift in rank does accompany a shift in something else; but this
is always by virtue of the logical relations among the categories
involved. The fact that by moving from structure to class, which is (or
can be) a move on the exponence scale, one also moves one step down
the rank scale, is due to the specific relation between the categories of
class and structure, and not to any inherent interdetermination between
exponence and rank. The descriptive relevance of keeping the scales
distinct is that it is important to be able to display what happens if one
shifts on one scale, keeping the other two constant.66
7.3 Exponence is the scale which relates the categories of the theory,
which are categories of the highest degree of abstraction, to the data.67
Since categories stand in different relations to the data, it might seem
necessary to recognize four different scales of exponence, one leading
from each category. In fact, however, exponence can be regarded as a
single scale.
In the first place, each category can be linked directly by exponence
to the formal item: it is in fact a requirement of the theory that any
descriptive category should be able to be so linked. This may be stated
by way of exemplification, as when we say “the old man is (an example
of ) an exponent of S in clause structure”. This is, however, not a
description of the element S, since by relating it to its exponent at a
stage when it was not necessary to do so we should have lost generality
(cf. above, 6.3). So instead of throwing up the grammatical sponge and
moving out to lexis while this is still avoidable, the description takes
successive steps down the exponence scale, changing rank where
necessary, until (at the degree of delicacy chosen) it is brought
unavoidably face to face with the formal item.
In the second place, therefore, the step by step move from any one
category to the data can proceed via any or all of the other categories.
This is then a move down the exponence scale, and at each step, given
that delicacy is constant, one category is replaced by another (either
with or without change of rank, according to which category is
replaced by which). While therefore the categories are distinct, they
are interrelated in such a way that the relation of exponence has the
status of a single scale.
The ultimate exponent in form is the formal item. This has then to
be related, in turn, to the substance. But this relationship, though it
may also be called “exponence”, entails a new scale, in which the
57
early papers on basic concepts
nature of the abstraction is different and the formal item is now at the
other end – it is itself the abstraction, and is not in any way delimited
or categorized by grammar.68
When grammar reaches the formal item, either it has said all there is
formally to be said about it or it hands over to lexis. The formal item
is the boundary of grammar on the exponence scale. It is not of course
the boundary on the rank scale: whenever the formal item is anything
other than a single morpheme (whether in closed system like “seeing
(that)”, or in open set like “pickup”) the grammar can be taken further
down in rank, since it can state the structures in terms of elements
whose exponents are words and morphemes. But seeing (that) enters
into a system at group rank, while pickup emerges from the grammar as
a word, though being a lexical item it would not necessarily be an
exponent of any whole grammatical unit. Once it has been taken over
by lexis, the grammatical categories, and the grammatical exponence
scale, no longer impinge on it (see below, 8.2).
58
categories of the theory of grammar
8 Lexis
8.1 This section is intended merely to bring lexis into relation with
grammar, not to discuss the theory of lexis as such. As has been pointed
out (above, 3.3. and 6.3), there is no one / one correspondence in
exponence between the item which enters into lexical relations and
any one of the grammatical units. It is for this reason that the term
lexical item is used in preference to word, “word” being reserved as the
name for a grammatical unit, that unit whose exponents, more than
those of any other unit, are lexical items.
Not only may the lexical item be coextensive with more than one
different grammatical unit; it may not be coextensive with any gram-
matical unit at all, and may indeed cut right across the rank hierarchy.
Moreover, since the abstraction involved is quite different, what is for
lexis “the same” lexical item (that is, different occurrences of the same
formal item) may be a number of different grammatical items, so it is
59
early papers on basic concepts
not true that one lexical item always has the same relation to the rank
hierarchy. So that, in English, (i) a lexical item may be a morpheme,
word or group (at least); (ii) a lexical item may be assigned to no rank,
being for example more than a word but less than a whole group, or
even both more and less than a word – part of one word plus the
whole of another, sometimes discontinuously; and (iii) one and the
same lexical item may in different occurrences cover any range of the
possibilities under (i) and (ii).72
This does not mean that lexical items cannot be identified in
grammar; it means that they are not identified by rank. They are
identified, as has been suggested (above, 6.3), by their being unac-
counted for in systems. But it is an additional, descriptive reason
(additional, that is, to the theoretical one that lexical items lend
themselves to different relations of abstraction) for keeping grammar
and lexis apart. When the two have been described separately, the next
stage is to relate them; and it is here that the complex relation between
lexical item and grammatical unit must be accounted for. This is exactly
parallel to what was said above (7.1) about grammar and phonology;
and, of course, it applies equally to phonology and lexis, where, after
separate description, is displayed the relationship between the lexical
item and the categories of phonology.
8.2 The task of lexis can be summed up, by illustration, as that it has
to account for the likelihood of wingless green insects and for the, by
contrast, unlikelihood of colourless green ideas.73 As in grammar, we shall
expect language to work by contrasting “more likely” with “less likely”
rather than “possible” with “impossible”; but, as has often been pointed
out,74 this particular type of likelihood is not accounted for by
grammar, at least not by grammar of the delicacy it has yet attained. It
is, however, too often assumed that what cannot be stated grammat-
ically cannot be stated formally: that what is not grammar is seman-
tics, and here, some would add, linguistics gives up.75 But the view
that the only formal linguistics is grammar might be described as a
colourless green idea that sleeps furiously between the sheets of
linguistic theory, preventing the bed from being made. What are
needed are theoretical categories for the formal description of lexis.
It seems that two fundamental categories are needed, which we may
call collocation and set.76 The first basic distinction between these and
the categories of grammar is that in lexis there are no scales of rank and
exponence. There is no hierarchy of units; therefore no rank scale.
There is only one degree of abstraction – a set is a set of formal items
60
categories of the theory of grammar
61
early papers on basic concepts
9 An analogy
9.1 Eating, like talking, is patterned activity, and the daily menu may
be made to yield an analogy with linguistic form. Being an analogy, it
is limited in relevance; its purpose is to throw light on, and suggest
problems of, the categories of grammar by relating these to an activity
which is familiar and for much of which a terminology is ready to
hand.
The presentation of a framework of categories for the description of
eating might proceed as follows:
Units:
Daily menu
Meal
Course
Helping
Mouthful
62
categories of the theory of grammar
63
early papers on basic concepts
At the rank of the “course”, the primary class “entrée” has secondary
classes “meat dish” and “poultry dish”. Each of these two secondary
classes carries a grammatical system whose terms are formal items. But
this system accounts only for simple structures of the class “entrée”,
those made up of only one member of the unit “helping”. The class
“entrée” also displays compound structures, whose additional elements
have as exponents the (various secondary classes of the) classes “cereal”
and “vegetable”. We will glance briefly at these:
64
categories of the theory of grammar
65
early papers on basic concepts
tion involved in relating one unit to another at the same level is quite
different from that involved in relating one level to another. The sense
in which a sentence “consists of ” morphemes is stated in description
with reference to its definition in the theory, and is totally different
from the sense in which a morpheme “consists of ” phonemes – indeed
it is doubtful if there is any meaningful sense in which a morpheme
consists of phonemes. (The contrastive use of “morph” and “mor-
pheme” is designed to build in an extra stage to account for the two
kinds of abstraction. But it does not get over the first (descriptive)
difficulty; nor in fact does it solve the theoretical one, since morph and
phoneme differ in the extent and kind of formal determination under-
lying their phonological abstraction.83)
10.2 The relation among the units also tends to be confused with
the relation between a category and its exponent(s). It is assumed that
in moving up the rank scale, from morpheme to sentence, one is also
moving up the exponence scale. It is true (as said above, 6.2 and 7.3)
that in comprehensive description, in order to display the full gram-
maticalness of language, one takes the final step on the exponence
scale at the lowest rank possible (though this, as already shown, is
by no means always the morpheme),84 and this is probably the reason
for the confusion of rank with exponence. But the scales of rank
and exponence are again different dimensions of abstraction, and
one can link any unit directly to its formal exponent (and through
this to its exponent in substance): the relation of an exponent of the
unit “sentence” to the category of sentence is exactly the same as
that of an exponent of the unit “morpheme” to the category of
morpheme.
66
categories of the theory of grammar
carries it.88 For example: the units may not be coextensive, and a
variable relation of phonological unit to grammatical unit may be the
very thing that carries a formal system (cf. above, 10.1 n. 82); or a
single system in grammar may be carried by different phonological
distinctions, say two of its terms by tone and a third by addition of a
segment; or a phonological system, such as tone associated with a given
unit (recognized as phonological because it carries some formal distinc-
tion), may carry different formal distinctions, part grammatical and part
lexical, or terms in different grammatical systems.89
The categories required by the grammar, and the criteria for these,
should come from within grammar. They are set up to provide a
description that is comprehensive, consistent and maximally powerful.
In the definition of, for example, the unit “clause”, the requirement is
that it should yield classes and structures which make possible the
description of the sentence, the group and so on in terms of their
structures and classes: hence the mutual definition of all units and of all
grammatical categories (and, procedurally, until a description is “com-
prehensive” (primary delicacy at all ranks) all parts of it remain subject
to revision). A grammatical category is not required to be identifiable
by reference to a particular feature of substance stated phonologically:
it merely carries the potentiality of being stated in phonological terms
through a long chain of exponence.90
(The starting point here of course is the theory of grammar, so that
what is being considered, and objected to, is the identification of
grammatical categories on phonological criteria. For linguistic
theory as a whole, the question must also be formulated the other way
round: do we derive phonological categories from formal ones? For
example, in the last case mentioned in the first paragraph of this section,
would we state one phonological system of tone or more than one?
The nature of phonology as an “interlevel” suggests that its categories
should be derived from those of form; and this is usually done in
prosodic phonology, as developed by Firth and others, though it is not
an inherent requirement of prosodic method.91 Inter-level dependence
in this direction is theoretically justifiable, since the role of phonology
is to account for the formally meaningful organization of phonic
substance. But it has descriptive dangers: first, that a system carried
only by variable relation between grammatical and phonological unit
may be missed, and second, that unless a comprehensive formal
description has first been made, formal features may be distorted into a
phonological mould – the phonologist may take the “word” class
“verb” as a phonological unit, but he (or someone) must have described
67
early papers on basic concepts
the grammar first, or his “word” and “verb” may not turn out to be
the grammarian’s “word” and “verb”.)
To be precise, then, what is being criticized here has again both a
theoretical and a descriptive aspect. In the theory, it is the use of
phonologically stated features as crucial criteria for grammatical
categories, as when supra-segmental phonemes are used as criteria for
the category of “phrase”.92 A phrase is a phrase because it operates in
the structure of the unit above it and has its own structures in terms of
the unit below it. It has then to be related to the phonological
categories which are arrived at by a different process of abstraction on
the basis of the minimal requirement that some formal distinction is
always involved. If at any point this yields a one / one relation of
categories, so much the better: if the phrase turns out to be exponen-
tially coextensive with the tone group, the latter can be used as a
recognition signal for the former. But it remains a signal, not a criterion.
In description, the trouble arises when phonological features are used
as grammatical criteria even when they clash. If, for example, the
segment which carries tone contrasts, or is bounded by juncture
features, will not work as a grammatical unit, then tone and juncture
are no use even as recognition signals; so to “define”, say, the clause
by reference to tone or juncture one has to set up a phonological
system in which any feature in substance can be an allophone of any
term in the system. There will be clash; and if it is recognized in the
first place that there is nothing at all surprising when, say, units carrying
formal patterns do not coincide with those carrying patterns of the
organization of substance, then the search for one / one phonological
identification signals of grammatical categories, such as a phonological
statement of clause boundary, can be abandoned as being without
profit.93
68
categories of the theory of grammar
10.5 With the smallest unit as fundamental, the description starts off
in an upward direction. It proceeds, unidirectionally, from the mor-
pheme, through the word to somewhere around the group.96 Not
surprisingly, since there is no shunting, it proves extremely difficult to
take it further along the same route. Shunting, or moving up and down
the rank scale, is a part of descriptive method imposed by the theory to
show the relation among the different units: to permit a unified
description with links, through all categories, all the way from mor-
pheme to sentence. In the absence of shunting, the description has to
jump to the top end of the rank scale and proceed downwards from
the sentence by “immediate constituent” analysis:97 still unidirectional,
though with the direction reversed.
The hope is that, by digging the tunnel from both ends, the two
will meet in the middle. With the aid of a good homogeneous
mathematics they might; but two totally different bodies of method are
involved, morphemics and IC analysis, which are difficult to integrate.
Moreover, the middle ranks of the grammar are often the most
complex, presumably since they face both ways; so that a grammar
which starts unidirectionally from the two ends will find it difficult to
avoid leaving the middle ragged. Presentation can of course proceed in
any direction that is desired; but it needs to be based on a description
that permits – indeed presupposes – constant shunting.
10.6 It has been suggested at various places above that the theory
cannot validly regard constrasts and relations that are clearcut, and
statable in “yes / no” terms, as the norm in language. Even at the
primary degree of delicacy the description will encounter features
where “shading” is necessary: where a feature is better stated in terms
69
early papers on basic concepts
70
categories of the theory of grammar
71
early papers on basic concepts
Notes
1. It is in no way to deny the fundamental importance of Chomsky’s work
(1957) and elsewhere, if we suggest that the readiness of linguists who
had previously worked in the “Bloomfieldian” tradition to abandon
these methods in favour of Chomsky’s is in part due to their lack of
72
categories of the theory of grammar
73
early papers on basic concepts
74
categories of the theory of grammar
. . . , and those posterior to PP” (p. 17); “The languages of the dominant
world cultures use vector formulas, and the discussion which follows is
therefore concerned only with the structure of a hypothetical tense
system based on the vector principle” (p. 20); “The system, of course,
would break down if a plus form were to be used to describe a minus
event or if a form indicating anticipation were used for recollection”
(p. 24). This does not invalidate the approach; it does suggest that it will
have to be part of a study of context which starts from form as well as
from “objective reality”, as phonology works both from form and from
substance; context, like phonology, is in a real sense an interlevel.
15. Theoretical validity implies making maximum use of the theory (see
below, 2.3 and 6.2). It is not necessary to add a separate criterion of
“simplicity”, since this is no use unless defined; and it would then turn
out to be a property of a maximally grammatical description, since
complication equals a weakening of the power of the theory and hence
less grammaticalness. It should perhaps also be mentioned here that the
distinction between methods of description and discovery procedures is
here taken for granted throughout (cf. below, 2.3). We are not
concerned with how the linguist “finds out” how an event is to be
described. This is no more capable of scientific exposition than are the
steps by which the theory was arrived at in the first place – in fact less,
since the latter can at least be formulated, while the former can only be
summed up in the words of the song: “I did what I could”.
16. Cf. Firth (1957a: 22), and above, 1.8 n. 11; Garvin (1957); see below,
6.3 n. 60; Robins (1959).
17. “Grammar” is also the name for the study of grammar; as with “level”
(above, 1.4 n. 8), it is unnecessary to distinguish between “the grammar”
of a language and “grammar” in theory and description – though a
distinction is often made between “lexis” and “lexicology”, the latter
being the study of lexis. Again, not a set of discovery procedures, but a
set of properties of what the linguist accounts for grammatically. The
grammar of a language can only be “defined” as that part of the language
that is accounted for by grammatical description.
18. The reference is, of course, in formal meaning: it is form that is under
discussion. It may always happen that the addition of a new term
changes the contextual meaning of at least one of the others, since terms
that are formally mutually exclusive are likely to carry contextual
distinctions; but this is not a property of a system. The “addition” of a
new term is not of course considered as a process (though historical
change is one type of instance of it): it may be displayed in any
comparison of two related systems. For example, two possible systems
of first and second person pronouns used by different speakers of Italian
(quoted in oblique disjunct form; I = “interior to social group”, E =
“exterior . . .”).
75
early papers on basic concepts
1 me me
1+ noi noi
2I te te
2I+ voi
2E
2E+
lei
loro
} voi
(The distinctions made in written Italian are ignored, since they would
not affect the point.) The difference in format meaning is a function of
the different number of terms: in system one me excludes five others, in
system two only three. In contextual meaning only terms of the second
group are affected.
19. Cf. Firth (1957a): “Moreover, these and other technical words are given
their “meaning” by the restricted language of the theory, and by
application of the theory in quoted works” (p. 2). This is true of
descriptive categories too: “noun” can no more be defined in a glossary
than “structure”.
20. I should therefore agree with Palmer (1958) that linguistic levels do not
form a hierarchy. His view is “that there are levels, but only in the
widest sense, and that these are related in specific, but different, ways.
The set of relationships cannot be regarded as a hierarchy, except in the
loosest sense of the word”. Palmer, however, appears to reintroduce
procedural hierarchy when he says, “The procedure is not from phonet-
ics via phonology to grammar, but from grammar via phonology to
phonetics, though with the reminder that the phonetic statement is the
basis, i.e. the ultimate justification for the analysis” (p. 240). I would
rather say that there is order among the levels, determined by their
interrelations, but (a) no hierarchy, in the defined sense of the word,
and (b) no procedural direction. Unfortunately Palmer excludes this use
of “order”. “There is a statable order of levels . . . and, therefore, a
hierarchy” (pp. 231–2, in reference to Hockett).
21. Immediate Constituent analysis, for example, yields a hierarchy that is
not a taxonomy: it does not fulfil criterion (ii). (It may not always fulfil
(i): cf. Hockett (1957): “There must be also at least a few utterances in
which the hierarchical structure is ambiguous, since otherwise the
hierarchical structure would in every case be determined by form, and
order, and hence not a “primitive”” (p. 391).)
22. The theory thus leads to “polysystemic”-ness in description – both
syntagmatically and paradigmatically. Syntagmatic polysystemic state-
ment follows from the linking of classes and systems to places in structure
(see below, 4–6), so that the question “how can we prove that the b of
beak and the b of cab are occurrences of one and the same phoneme?”
(Ebeling: 1960: 17) is regarded as an unreal one; cf. Henderson (1951:
132); Carnochan (1952: 78); Robins (1953: 96); Firth (1955: 93; 1957b:
76
categories of the theory of grammar
77
early papers on basic concepts
the term is here used in grammar, since this use presupposes a rank scale
(as well as the other terms structure, class and system in a system of
related categories), which is absent from lexis. It is probably better to
restrict the term “unit” to grammar and phonology: cf. Bazell (1953:
11) – though Bazell does not here consider lexical form.
29. So, for the description of English:
units
앖 sentence
rank
앗
(
clause
group (phrase)
word
morpheme
30. Statistical work on grammar may yield a further unit, above the sentence:
it will then be possible to set up sentence classes, and account for
sequences of them, by reference to this higher unit. Similarly in
phonology we need a unit in English above the tone group to account
for sequences of different tones. The grammatical and phonological
“paragraph” (and perhaps ‘paraphone’?) is probably within reach of a
team of linguist, statistician, programmer and computer; cf. Firth
(1957a): “Attention must first be paid to the longer elements of text –
such as the paragraph . . .” (p. 18); Harris (1952); for Hill (1958: 406),
and others, this is “stylistics”, but in the present theory it would come
within exactly the same general framework of categories.
31. The “simple / compound” opposition is thus one of structure. It may,
of course, happen that a given realization yields simple membership all
the way up and down the rank scale. Yes may be (i.e. may be an
exponent of) one sentence which is one clause which is one group
which is one word which is one morpheme.
32. Cf. Robins (1953: 109); Firth (1957a, esp. 17, 30; and 1955, esp. 89,
91); Halliday (1959: 49).
33. Cf. Firth (1957a): “Elements of structure . . . share a mutual expectancy
in an order which is not merely a sequence” (p. 17). Since sequence is a
variable, and may or may not be an exponent of structure, we find
difference in sequence without difference in structure (cf. below, 4.3
n. 42), or difference in structure without difference in sequence. I am
indebted to J. M. Sinclair for a recent conversational example of the
latter: orthographically, The man came(,) from the Gas Board. Phonolo-
gically (relevant units: tone group, bounded by //, and foot, by / – these
are unit boundaries and have nothing to do with juncture): what was
said was (tonic syllable underlined):
// 1 the / man_ / came // 1 from the / Gas / Board //
Grammatically, one clause, structure SP; exponent of P came, of S the
78
categories of the theory of grammar
man . . . from the Gas Board, being a nominal group, structure MH +Q.
What might have been said was
// 1 the / man / came from the // Gas / Board //
Grammatically, one clause, structure SPA; exponents, S the man, P came,
A from the Gas Board. The two are different in grammatical structure,
and this difference has its exponent in phonic substance which can be
stated phonologically. (That the phonological patterns, and the distinc-
tion between them, abstracted from the substance along one dimension
correspond regularly (though not one / one) with the grammatical
patterns, and the distinction between them, abstracted along another
dimension from the same substance can be shown by the construction
of other partially like clauses.) But though the difference in structure has
its manifestation in substance (there can of course be ambiguity in
substance, as in Hockett’s old men and women (1957: 390n.), in form the
difference is not realized in sequence. In sequence, from the Gas Board
occupies the same place in both instances; in order, S and A stand in
different relations to P, and from the Gas Board is exponent of (part of) S
in the one case and of (the whole of) A in the other.
Sequence is presumably always manifested in phonic substance as
linear progression; the distinction is then one of exponence, “sequence”
being the name for that formal relation between formal items of which
linear progression is the manifestation in phonic substance.
34. It is useful to make a distinction in the use of symbols between an
inventory of elements of a structure and a structure, by the use of
commas in the former. Thus, X, Y, Z is an inventory of elements, XYZ
a structure composed of these elements.
35. Since a unit that carried only one-place structures would be unnecessary:
if, for example, all words consist of one morpheme (i.e. the unit “word”
has no structure containing more than one place), “word” and “mor-
pheme” would be one and the same unit.
36. For the name and nature of this grouping, see below, 5.
37. Since the morpheme (i) is a grammatical unit and (ii) carries no
grammatical structure, it has no structure. Cf. Palmer (1958: 229–30)
(quoting Hockett 1955: 15): “ ‘Morphemes are not composed of
phonemes at all. Morphemes are indivisible units. A given morpheme is
represented by a certain more or less compact arrangement of phono-
logic material . . . If we call any such representation a morph, then it
becomes correct to say that a morph has a phonological structure – that
it consists of an arrangement of phonemes.’ [Hockett] recognizes that
the units established at each level differ in kind, and not merely in size,
from those established at other levels.” The “morph” does indeed
accommodate the theoretical point (but cf. below, 10. 1 n. 83), that the
units differ in kind; but in accepting Hockett’s view Palmer has not
79
early papers on basic concepts
80
categories of the theory of grammar
43. For example the following two exponents of the (class) nominal (of the
unit) group: all the ten houses on the riverside and the finest old houses on the
riverside have the same primary structure M . . . HQ (or MMMHQ).
But a more delicate statement of M, still at group rank, shows distinct
secondary structures, the first example having D2DbO, the second
DbOE.
44. When Hockett writes (1955: 17) “In general, then, if we find continu-
ous-scale contrasts in the vicinity of what we are sure is language, we
exclude them from language (though not from culture)”, this applies (i)
only to grammar and phonology, not to lexis or context (cf. Bazell
1953: 11), and (ii) only to one type of contrast, that between terms in
systems. It is, indeed, a defining characteristic of a system that it cannot
be a cline. But units and classes are not crucially discrete: in exponence,
units display syntagmatic non-discreteness (syncretism); classes, paradig-
matic non-discreteness (statable in various ways, such as multidimen-
sional classification, assignment of an item to different classes with
variable probability, etc.).
45. I would class it with other dichotomies that Firth rejects: cf. Firth
(1957b), “My own approach to meaning in linguistics has always been
independent of such dualisms as mind and body, language and thought,
word and idea, signifiant et signifié, expression and content. These
dichotomies are a quite unnecessary nuisance, and in my opinion should
be dropped” (p. 227). Cf. Firth (1957c: 2, 3 – though here, I must
admit, Firth also rejects “form and substance”, which I find crucial (as
levels) to an understanding of how language works.
46. Cf. Firth (1957a): “It follows that the distinction between morphology
and syntax is perhaps no longer useful or convenient in descriptive
linguistics” (p. 14).
47. i.e. languages in which inflexional systems are a regular feature of word
structure. Free and bound are generalized class categories, linked to the
generalized structure categories of simple and compound (above, 2.3):
“free” is “able to stand as exponent of one-element structure of the unit
next above”, “bound” is “unable to stand, etc”. A member of a “free”
class can thus be exponent of a “simple” structure, while a member of a
bound class can operate only in “compound” structures.
48. Other terms are of course available, like Haas’ (1954: 68ff.) “synthetic
classification” and “analytic classification”. The terminological objection
to the use of “class” in both (as in “form class” and “function class”) is,
however, theoretically founded: if we say, with Haas (p. 68), that “we
distinguish two ways of classifying” linguistic units, we imply two things:
(i) a choice of (ii) procedural direction. But this is not a procedural
matter, and there is no choice. All forms are to be accounted for, and
this means stating both their class (linking them to the unit next above)
and their own structure (linking them to the unit next below). Whether
81
early papers on basic concepts
Highest
abstraction 䉱 Primary Secondary
structure 䉴 structure
Scale of
Exponence 䉲 䉲
Primary Secondary
䉲 class 䉴 class
Data
51. Or nearly coextensive: the criteria for the setting up of one primary class
or two are descriptive. For example, in English clause structure S and C
are different elements standing in different relation to P. There is a high
degree of overlap between their exponents: one primary class (class
“nominal” of unit “group”) can be set up as exponent of both S and C.
The lack of exact coextensiveness will be stated by secondary elements
and classes, to account for (for example) the occurrence of the old hall,
the old town, the old town hall, this hall / town / town hall is old, this is a hall /
town / town hall, and the non-occurrence of this old is a hall, this is an old
or this hall is town.
82
categories of the theory of grammar
52. Cf. Firth (1957b): “Various systems are to be found in speech activity
and when stated must adequately account for such activity. Science
should not impose systems on languages, it should look for systems in
speech activity, and, having found them, state the facts in a suitable
language” (p. 144). Cf. also references given in 4.1 n. 32, above.
53. Again, abstraction on the exponence scale. The formal item the old
man is exponent of (is a member of) a class (“nominal”, of the unit
“group”). The class “nominal group” is exponent of (operates at the
place of) an element of structure (S or C, of the unit “clause”). The
formal item itself, of course, has its own (and ultimate) exponents in
phonic or graphic substance.
54. Diagrammatically (axes as in 5.4 n. 50, above):
1 (at X) 䉳
2.1 (at Ya)
2 (at Y) 2.2 (at Yb)
3 (at Z) 2.3 (at Yc)
83
early papers on basic concepts
}
sentence structures . . .
clause (system of ) classes . . .
exponents
⬙ structures
group (system of) classes . . .
etc. etc.
59. Without in any way affecting the syntactic nature of the “class”.
60. Cf. Garvin (1957): “Morphemes of limited membership class should be
listed in the grammar and morphemes which belong to classes of
unlimited membership should be exemplified in the grammar and listed
in the dictionary” (p. 55).
61. Except in the sense that the description will always try a move down
the rank scale as a possible way of extending its power (“remaining in
grammar”). But wherever the lexical item is greater than a morpheme,
its further analysis by grammar into morphemes will leave its lexical
relations unaccounted for. For example, in the train left ten minutes late,
but made it up, made up is a discontinuous verbal group analysed as two
words, one (made) of two morphemes, the other simple; but it enters
into an open set qua lexical item make up, which itself is here assigned
to no grammatical unit.
62. See below, 8.
63. Regressive structures can of course be regarded as forming a scale; but
their description does not require the introduction of a separate scale
into the theory. Cf. above, 3.2 n. 26.
64. Cf. Allen (1956), from which the following is taken: “It frequently
occurs that an appropriate ‘bit’ of the corresponding phonological
statement (or of the orthography) is used as a label for the grammatical
unit in question . . . The price of using such labels is constant vigilance
. . . Where the phonological analysis permits of alternatives, that alterna-
tive is to be chosen which is most congruent with the grammatical
84
categories of the theory of grammar
85
early papers on basic concepts
86
categories of the theory of grammar
87
early papers on basic concepts
76. Cf. Firth (1957a: 11–13, 26–7, 31); Mitchell (1958: 108 ff.); Halliday
(1959: 156–75).
77. Analogous to the morphological grouping (the “paradigm”) of grammar
is the lexical “ordered series of words”: cf. Firth (1957b: 228) – though
Firth’s “ordered series of words” includes what I should consider a
“lexical set”, namely his “lexical groups by association”, these being (by
analogy) “syntactical”.
78. The traditional vehicle of lexical statement, the dictionary, states formal
meaning by citation and contextual meaning by definition. The theoret-
ical status of lexicographical definition (Firth 1957a: 11 “shifted terms”)
needs to be carefully examined.
79. Cf. above, 7.4 n. 71. The point, however, is: what is to be regarded as
“one” lexical item? Dictionaries, in general, mix grammatical and lexical
criteria: in the Shorter Oxford Dictionary, for example, cut v. (defined
as “to penetrate so as to sever the continuity of with an edged
instrument”) and cut sb. (“the act of cutting”) are shown as separate
items having the same relation to each other as bear v. (“to support and
remove; to carry”) and bear sb. (“a heavily-built, thick-furred plantigrade
quadruped . . .”). But formally (quite apart from the fact that one
contextual statement will cover both cut v. and cut sb. but not both bear
v. and bear sb.) the pairs are quite distinct: there is a high degree of
overlap in the range of collocation of cut v. and cut sb., which is not the
case with bear v. and bear sb. Collocation provides a formal criterion for
the identification of the lexical item.
80. An analogy which Bloomfield did use was that of the signal: “Accord-
ingly, the signals (linguistic forms, with morphemes as the smallest
signals) consist of different combinations of the signalling-units (pho-
nemes), and each such combination is arbitrarily assigned to some feature
of the practical world (“sememe”)” (1933: 162). This runs the risk of
suggesting the analogy of a code – or even that language “is” a code. If
language is a code, where is the pre-coded message? Cf. my review of
Herdan’s Language as Choice and Chance (1959).
81. Detailed references are not given in this section. It is recognized, as
already remarked, that what is here called “Bloomfieldian” method is an
abstraction from a large body of descriptive work by different linguists,
within which there is considerable variety and disagreement even on
basic issues. Roughly it covers the work based on what Hockett called
the “item and arrangement” model. It is not of course suggested that all
the points made in this section are applicable to all such studies, nor all
of them to any one study, within this type of linguistics. Since I have
been concerned to apply the present theory in the description of English,
many of the points made here were in fact first formulated with reference
to Hill (1958), which incorporates what is probably the best comprehen-
sive account of English grammar yet published and is an example of the
88
categories of the theory of grammar
method here referred to. Hill’s book has recently been the subject of a
review article by Haas (1960); my present aim differs from Haas’ in that
I want to consider certain features of an approach in descriptive
linguistics, exemplified by Hill’s work but also by many other studies, in
the light of the theory here outlined.
82. Especially since a language may make systemic use of this variable (cf.
4.2 n. 37, above, where it is suggested that it is desirable to recognize in
English a grammatical system the exponent of which is precisely the
contrast between coextensiveness of the grammatical unit with one and
with (a sequence of) two exponents of a phonological unit).
83. Cf. above, 4.2 n. 37. Hill’s proportion, however, is
morph : morpheme : allomorph ::
sound : phoneme : allophone.
“Every morpheme must contain one phoneme and may contain several”
(p. 89) and, for English, “the occurrence of any juncture always marks
the boundary of an entity larger than a phoneme. The entity thus
bounded may be word, phrase or sentence, but must always be at least a
morph” (pp. 93–4). The phoneme thus enters into the statement both
of the unit “morpheme” and of its structure.
84. That is, the final step of the formal statement, the move to the formal
item. Strictly, since formal statement includes the placing of all forms at
all units, it would be more accurate to say “though it is by no means
always at morpheme rank that systems of formal items are to be found”.
For example in English the items when, because, if, in case, provided that,
etc., though they can of course be analysed into words and morphemes,
operate as items at the rank of the group and, as such, are members of a
particular class of the group.
85. So if we find in three languages items in substance statable phonetically
as [pata] and [pate], these may yield:
Substance Form
L1 [a] / [c] - (free variant)
L2 ’ singular / plural
(grammar)
L3 ’ ‘cat’ / ‘dog’ (lexis)
(Phonologically restricted variants may be of either type: the language
might have [pata] and [pete], never [pate] or [peta], without prejudice
to whether the contrast between [pata] and [pete] is formally meaningful
(L2,3) or not (L1).)
86. It does not matter, of course, what type of distinction is made in
substance: the pair above could equally be replaced by for example [pàt]
and [pát], or [’pata] and [pa’ta], etc.
87. Again it does not matter where: the following grammatical contrasts in
89
early papers on basic concepts
90
categories of the theory of grammar
structure, and by its own classes and their structures. The difference
between two instances, such as an affirmative clause he saw them and an
interrogative clause did he see them?, is of course ultimately statable in
terms of substance; but grammar is not grammar if it tries to define the
class system in this way – or even to state the difference phonologically
at all. No linguist would ever try to state the grammar of clause–classes
by reference to phonology; yet the attempt to define the unit “clause”
by reference to phonological features such as juncture is no less objec-
tionable – and leads, not surprisingly, to a phonology in which any
substantial feature is a possible exponent of any term in the phonological
system!
91. A phonological description will, in this view, be prosodic if (i) it
incorporates a rank scale, with a hierarchy of units to which contrasts
are assigned, and (ii) it is polysystemic, so that, for example, the /t/ in
10.3 n. 87, above is not “the phoneme /t/” of “the English language”
(no such entity will be postulated) but a phoneme identified by reference
to its place in the structure of the unit concerned; this would still be
true whether the phonological units concerned are (partly) taken over
from grammar or are set up independently in the phonology. Firth
stresses the very different nature of the “phoneme” in a description of
this type, and prefers to use the distinct term “phonematic unit”
(cf. 1957b, Chapter 9, passim).
92. Cf. Haas’ reference to “a structure of a number of pyramids, all inverted”
(1960: 267).
93. I personally feel that English requires a totally different set of phonolog-
ical units not derived from the grammatical units. Intonation in English
needs a carrier unit “tone group” to display the (phonological) system of
intonation; this system, and the terms in it, can then be related to the
grammar. The attempt to describe intonation in a framework of “the
intonation of the clause”, “of the group”, etc. is complicated and may
lead to a misunderstanding of the operation of the intonation system.
But the attempt is not theoretically sinful, as would be the attempt to
describe the “grammar of the tone group”.
94. As in Hill’s description of the English personal pronouns (1958:
pp. 145–8). This “playing games”, or “party linguistics”, is again linked
to the confusion of levels. Cf. Haas (1960: 273).
95. I should thus agree with Robins (1959): “The morpheme must be
recognized as the minimal element of grammatical structure; but this
does not imply that it is the most suitable element to bear the assignment
of all the grammatical functions fulfilled by the word into whose
composition it enters” (pp. 127–8). I would not follow Robins, how-
ever, when he says that “In many ways . . . the word is a unique entity
in grammar, and not just a stage in the progression ‘from morpheme to
utterance’ ” (p. 137). Robins rejects morpheme-based grammar but
91
early papers on basic concepts
92
categories of the theory of grammar
93
early papers on basic concepts
94
Chapter Three
95
early papers on basic concepts
96
class in relation to chain and choice
97
early papers on basic concepts
98
class in relation to chain and choice
99
early papers on basic concepts
other same certain, etc.), together with other items forming an open set
(i.e. that cannot be so reduced: John’s, etc., including compound ones
as in the railway company’s property).
This class of deictic may be variously subdivided along both axes. On
the one hand, there are certain sets whose members can occur in
combination, as in all my other friends; there are in fact three such
secondary groupings, the members occurring respectively in first, second
and third place in a maximum sequence. This gives three secondary
chain classes which may be called “predeictic” (for example all), “deic-
tic” (for example my) and “postdeictic” (for example other). Within each
of these three classes, choices are made. There are many ways of
describing these, according to what are taken to be the principal
dimensions. The deictic, for example, may be “specific” / “non-
specific” (my / every); “selective” / “non-selective” (my / the); and, as a
further subdivision of the class formed by the intersection of “specific”
and “selective”, it may be “possessive” / “demonstrative” (my / this).
These and various other systems eventually yield, by their subdivisions
and intersections, one-member classes: thus my can be uniquely classified
as “deictic: specific, selective, possessive: personal: first person”.
Secondary classes regularly cut across each other. The systems of
“specification” and “selection”, for example, form a matrix as follows:
Specific Non-specific
this / these which both all
that / those what every each
no neither
Selective my your whose
our
their
his / her
its
John’s (etc.)
the a
Non-selective some any
either another
100
class in relation to chain and choice
deictics. But the patterns they display are typical in their complexity: a
given class breaks down by simple subdivision into a system of more
delicate classes, but the same original class will also subdivide in a
number of different ways, so that many dimensions of classification
intersect with one another. Any given item, to be fully identified, may
require to be simultaneously classified on all such dimensions. In this
way it can be assigned to a “microclass”, this representing its value in
respect of all the properties which have been found relevant to the way
it patterns in the language. There will be, of course, a very large
number of such microclasses: for example, in a computational study of
English “phrasal verbs” (items like take up, put on) which is being
carried out at the moment, 557 such items were found to yield 125
microclasses.
Up to this point I have been concerned only with “place-ordered”
structures. These are sometimes thought of as being the normal type of
linguistic structure. By a place-ordered structure I mean one composed
of a limited number of different elements occurring nonrecursively.
Such a structure may be fully class-defining, in the sense that to each
element corresponds a distinct class of lower rank: for example the
clause structure “Subject + Predicator”, with classes respectively “nom-
inal group” and “verbal group”, as in “my friends have arrived”; or it
may be only partially class-defining, where two or more elements are
expounded by the same class but differentiated in sequence. In this
type of structure, there is no constant relation between successive (or
otherwise paired) elements: for example, in the structure ‘Subject +
Predicator + Complement’ (for example John saw Mary, my friends have
invited me) it is not true that Subject is to Predicator as Predicator is to
Complement.
This is not the only type of structure found in language, and there
seems no particular reason for assuming it to be the norm, especially in
its pure form. Language also exhibits a different kind of structure, the
“recursive” or “depth-ordered” structure. Here, as the name implies,
an element of structure, or a combination of elements, is repeated “in
depth”, a series of such elements (or combinations) thus forming a
progression. It is doubtful whether one should set a theoretical limit to
the degree of depth in recursion; rather there appears to be some
logarithmic scale of diminishing frequency, so that the number of
observations one would expect to have to make before recording a
depth of, say, ten would be extremely high. Spoken English seems to
tolerate greater depth in recursion, or at least to tolerate it more readily,
than written English; and this may be true of language generally. The
101
early papers on basic concepts
102
class in relation to chain and choice
103
early papers on basic concepts
in the garden; but this class can also operate as Adjunct in the clause,
being thus not rankshifted, as in he sat in the garden. Where a class may
occur either rankshifted or not, often ambiguity may arise, as in take
that chair in the garden: is the chair already in the garden (= ‘take that
chair which is in the garden’; rankshift) or not (= ‘take that chair and
put it in the garden’; no rankshift)?
Some of the most complex problems in the description of a
language arise where the same structure combines place-ordered with
recursive elements. Perhaps the most striking example of this in Eng-
lish is the nominal group, which is a troublesome mixture of the two
types; the earlier elements are largely place-ordered, recursive elements
being increasingly tolerated as one approaches the Head (the head
being, e.g., houses in the two old stone houses by the river) and continuing,
by rankshift, thereafter. The element immediately following the
Determiner, which may be called Ordinative and defines the classes
of cardinal and ordinal numeral, together with superlatives, is already
marginally recursive (e.g. the first three second best hotels); with later
modifying elements preceding the Head this potentiality is greatly
increased, giving items such as example 4 above and familiar also in
the language of headlines: holiday coach death crash inquiry verdict. More-
over, the linear succession of the items does not act as a constant in
showing the depth relation: compare 5-millimetre perspex boxes (gamma
beta alpha) with 6-inch perspex boxes (beta beta alpha). The description
of the word–classes entering into the structure of the English nominal
group is extremely complicated if one treats it as a simple place-
ordered structure, with classes defined for each possible position, as
the various attempts to do so have already shown. On the other hand
it is unsatisfactory to treat the whole thing as recursive in structure
and to recognize no secondary classes beyond the primary word–class
“noun”. The facts of the language here lie in between the extremes
of these two types of structure, and the best description seems to be
one which takes this into account.
These seem to me to be some of the problems that arise in intra-
linguistic classification at the level of grammar. Such problems are
probably most acute at this level, but similar ones arise also at other
levels, notably phonology, which has other additional classification
problems of its own. Lexical classification is rather a different matter,
and there are reasons for preferring not to use the term “class” in
talking about lexis; but this subject would require a separate paper. The
field of classification as a whole is one where linguists can learn much
from disciplines faced with similar or related problems. At the same
104
class in relation to chain and choice
Notes
1. I am assuming here the more “abstract” view of grammatical categories
such as morpheme; cf. Palmer (1964b: 232–7).
2. For interdisciplinary purposes I have used here the terms “chain axis” and
“choice axis” in place of their less self-explanatory technical equivalents
“syntagmatic axis” and “paradigmatic axis”.
3. Strictly speaking the ultimate exponent is a token (“occurrence”) of the
type (“formal item”) fetch. If, however, we confine ourselves to the level
of grammar we can regard the formal item as the ultimate exponent.
4. Another way of drawing the same distinction between grammar and lexis
is to say that grammar is “deterministic” by contrast with lexis which is
“probabilistic”; in the sense that in grammar one can distinguish what is
possible from what is impossible (before assigning probabilities, if one
wishes, to what is possible), whereas in lexis one can only distinguish
between what is more and what is less probable.
5. In fact the verse does contain rankshifted nominal groups, as “prepositional
complement” in adverbial groups (for example that lay in the house), but
these have been ignored in the illustration for the sake of simplicity.
6. This restriction to “the rank immediately above” implies a particular
model of grammar more specifically than do most of the other points made
here. A more generally valid formulation would be “enter into a structure
of higher rank”.
105
Chapter Four
First presented at a meeting of the Linguistics Association, Newcastle, March 1965. First
published in Journal of Linguistics, 1966, 2(1), pp. 57–67.
106
some notes on ‘deep’ grammar
107
early papers on basic concepts
108
some notes on ‘deep’ grammar
109
early papers on basic concepts
110
some notes on ‘deep’ grammar
111
early papers on basic concepts
112
some notes on ‘deep’ grammar
This may be illustrated from the example it’s John who has seen the play.
Leaving aside variation that is immaterial to the discussion, there would
seem to be three possible representations of its structure:
(1) it Subject, is Predicator, John who has seen the play Complement
(2) it . . . who has seen the play Subject, is Predicator, John
Complement
(3) it’s John who Subject, has seen Predicator, the play Complement
(1) would presumably be an attempt merely to state the simplest
sequence of classes in the syntagm, although it could be shown to be
unsatisfactory even on class-distributional grounds. (2) is distributionally
acceptable and would account adequately for the syntagmatic relations;
but it fails to account for the paradigmatic relations in that it does not
show the “relatedness” of this clause to John has seen the play, etc. If the
structural description is required to show the paradigmatic as well as
the syntagmatic relations of the grammar we need some representation
such as (3) in which John is the Subject. This leads to complexity in
the realization, since a nominalization of the form it’s John who seems
to add no new insight elsewhere in the grammar. A more serious
difficulty arises in relation to the element “Subject” in English, which
is a complex element within which it is possible to distinguish three
components, or features; each of these may contrast independently of
the other two, although there is a general, and generalizable, tendency
to co-variation among them.
The three contrasts can be seen independently in (i) John has seen the
play, with tonic on play, versus, respectively, (ii) the play has been seen by
John (Subject as actor versus Subject as goal); (iii) the play John has seen (=
“the play, John has seen, but . . .”, Subject as theme versus Subject
nonthematic); (iv) John has seen the play (with John tonic; Subject as
“given” versus Subject as “new”). Each of these three is related paradig-
matically to the original item, and each of them contrasts with it in
respect of one feature only. By a further contrast, that of “unpredicated
theme” versus “predicated theme”, (iv) is related to (v) it’s John who has
seen the play with John tonic.6 Thus, despite the difference in constituent
structure, (v) differs from (iv) in respect of only one feature. Such
patterns, where different complexes of (paradigmatic) features may be
combined in what is syntagmatically one and the same element of
structure, here the Subject, involve some complexity for a structural
description; if they were handled in systemic terms, the structure need
represent only their realization in syntagmatic relations. We could then
adopt a form of structural representation such as (2) above.
113
early papers on basic concepts
114
some notes on ‘deep’ grammar
115
early papers on basic concepts
116
some notes on ‘deep’ grammar
Notes
1. This paper was first presented at a meeting of the Linguistics Association,
Newcastle, March 1965. I am grateful to R. M. W. Dixon and R. D.
Huddleston for their subsequent valuable comments and suggestions.
2. Hass (1966: 131) points out that Firth’s position here has been misinter-
preted; this may be partly due to his use of the term “element of structure”
as a functional term.
3. Firth, perhaps somewhat confusingly, reserved the term “order” precisely
for this non-linear relation among the components (“elements”) of a
structure, contrasting it with “sequence”.
4. I use Lamb’s term “realization” instead of the earlier “exponence”. Lamb’s
term is more widely known; it also corresponds closely to my own use,
whereas as Palmer (1964b) pointed out my use of “exponence” differed
materially from that of Firth.
5. This may be interpreted either as “if there is at least one set of conditions
under which both could occur” or as “if both could occur under the given
set of conditions”. It is the latter interpretation which I take to be the basis
of (one aspect of) Firth’s ‘polysystemic’ approach. Firth himself was
inconsistent in referring to a ‘system’ of word–classes noun, verb, etc.
6. Since the subject normally has the feature “given”, that of “new” being
realized in other elements, the realization of the feature “new” in the
subject is often accompanied by its predication as in it’s John who has seen
the play. This explains why it’s John who has seen the play usually, though
not obligatorily, has John and not play tonic, while the opposite is true of
John has seen the play.
117
Chapter Five
118
the concept of rank: a reply
that is my fault rather than his. I plead guilty unreservedly to the charge
of inadequate documentation of my own work, although the relevant
bibliography of others’ writings is rather longer than Matthews seems
to suggest.
Some of the points raised in my paper, ‘Some notes on “deep”
grammar’ (see above, Chapter 4), are I think relevant to Matthews’
comments; more important in this connection is Huddleston (1965),
written without knowledge of Matthews’ article. Here I shall try to
take up with reasonable brevity some of the issues as Matthews himself
sees them.
Two assertions seem crucial to Matthews’ argument. (1) A rank-free
constituency grammar is to be preferred to a rank constituency gram-
mar. (2) All formulations of rank grammars are either theoretically
insignificant or empirically unsound: unsound if they differ materially
from rank-free grammars and insignificant if they do not. (1) follows
from (2), except that, if they do not differ materially, this by itself gives
us no reason for preferring either. Let us examine each of these in turn.
By a rank grammar I mean one which specifies and labels a fixed
number of layers in the hierarchy of constituents, such that any
constituent, and any constitute, can be assigned to one or other of the
specified layers, or ranks. The European linguistic tradition, by its use
of terms such as “sentence” and “clause”, has always implied the
possibility of such a grammar, although, as is well known, the absence
of criteria regulating the necessary modifications of the simple con-
stituency relation led to various difficulties. I return below to the
concept of rank as suggested in such formulations as “clause used as a
word” and the like.
A rank grammar is, as Matthews observes, a hypothesis about the
nature of language. This leads us to ask, first, whether it can be falsified,
and second, whether it is worth making in the first place. Like many
other hypotheses, both in linguistics and in other disciplines, its
empirical falsification, given that it cannot be logically disproved, is
unlikely to take the form of the discovery of a clear counter-example
– in this case, of a language which it is impossible to describe in rank
constituent terms. It would be likely to take the form rather of
demonstrating that the hypothesis prompts no interesting new questions
and leads to unnecessarily complex or otherwise unsatisfactory accounts
of the facts. Neither this limitation on the conditions of its falsification
nor the fact that disagreements may arise (as in all subjects) over
whether a given hypothesis has in fact been shown to be inadequate of
themselves deprive it of interest. It seems to me that this is a hypothesis
119
early papers on basic concepts
120
the concept of rank: a reply
121
early papers on basic concepts
122
the concept of rank: a reply
123
early papers on basic concepts
124
the concept of rank: a reply
125
early papers on basic concepts
It is perfectly true that the fact that something is “in some sense” a
word does not make it an immediate constituent of anything. But it is
surprising to be told that ‘there is no argument for treating the king of
England’s . . . as (a word) . . . which does not simply take as a premise
the requirement that “each unit should be fully identifiable in descrip-
tion” ’. The suggestion that the king of England’s should be treated as a
word, which is not a new one, has not so far as I know usually rested
on such grounds; in any case we may reasonably ask for more than the
mere reiteration that it “is” a phrase. Many questions need to be raised:
what, for example, are the implications for constituent structure of the
many cases where ’s and of are not parallel, for example that hat is the
king of England’s but not this hat is of the king of England? How does the
grammar take account of the fact that the placement of the ’s is
paralleled by the placement of the tonic (in the king of England) which
likewise depends on location in the syntagm and not on the constituent
structure? and so on. If Matthews is suggesting that the problem is
created, or even that it is complicated, by the introduction of rank into
the grammar, it seems to me that he has given no evidence of this.
Here also the issue seems to be partly one of labelling. This does not
make it any the less appropriate for discussion; but whatever the
usefulness, or otherwise, of the concept of rank in this and other
contexts, it is surely something that can be discussed in its own
linguistic terms.
Notes
1. In this connection I find the concept of ‘a “neo-Firthian” ’, especially one
“committed” to certain “statements”, rather extraordinary. Must we all be
labelled in this way? There is an important principle at stake here: that a
scholar is responsible for what he says and writes, not for what others say
and write. If I express agreement with something another linguist has put
forward this neither makes that linguist responsible for my views nor
commits me to acceptance of the whole of his. I may be wrong, but I
feel that there are undesirable limitations on this principle inherent in
Matthews’ first two paragraphs.
2. I am sorry if my own formulation was unclear here. But in that case the
whole of this part of Matthews’ argument rests on one purely termino-
logical point – since even if I was appearing to suggest that such an item
should be labelled a word, no further use was being made of such a
suggestion.
126
APPENDIX TO SECTION ONE
TABLE 1
Tonality: Distribution of utterance into tone groups (location of tone
group boundaries)
Tonicity: Distribution of tone group into tonic and pretonic (location
of tonic foot)
Tone (primary; pitch movement on tonic)
Tone (secondary)
This paper was written between May and August 1964 and formed the substance of a
course on the description of English at the University of Indiana. First published in System
and Function in Language, 1976, edited by G. R. Kress, London: Oxford University Press,
pp. 101–35.
127
early papers on basic concepts
Symbols
// tone group boundary (always — tonic syllable
also foot boundary) . silent ictus
/ foot boundary . . . pause
Tone, primary and secondary, is shown by Arabic figures, alone or
with diacritics, placed immediately after the tone group boundary
marker.
128
appendix
TABLE 2
Tonicity: location of ‘information focus’
tonic = final lexical item (neutral)
tonic = pre-final item or final grammatical item (contrastive)
Tone (assuming tonality neutral):
Place of clause in sentence structure: final main 1, non-final
coordinate 3, non-final subordinate 4
Declarative clauses
reservation: 1 unreserved, 4 reserved
involvement: 1 neutral, 3 uninvolved, 5 involved
agreement: 1 neutral, 3 confirmatory, 2 contradictory
information: 1 one information point, 13 two information points
‘key’: 1 neutral, 1 + strong, 1 – mild
Interrogative clauses, WH-type
‘key’: 1 neutral, 2 (with final tonic) mild
relation to previous utterance: 1 unrelated, 2 (with WH-tonic) echo
Interrogative clauses, yes/no type
‘key’: 2 neutral, 1 strong
involvement: 2 neutral, 3 uninvolved, 5 involved
place in alternative question: 2 first alternative, 1 second alternative
specification of point of query: 2 unspecified, 2 specified
Imperative clauses
‘key’ (positive): 1 neutral, 3 moderate, 13 mild
‘key’ (negative): 3 neutral, 1 strong, 13 mild
force: 1 neutral, 4 compromising, 5 insistent
function: 1 etc. command, 2 question
‘Moodless’ clauses (also as declarative)
function: 1 answer etc., 2 question, 3 warning, 5 exclamation
In the section under ‘tone’, the headings (e.g. ‘reservation’) indicate
the nature of the choice, the entries under each heading representing
the terms in the choice with their appropriate tone. Thus ‘1 unreserved,
4 reserved’ means ‘in this choice tone 4 indicates reservation, by
contrast with one 1 which indicates no reservation’. Secondary tones
are indicated where relevant.
129
1 Systems of tone
appendix
131
early papers on basic concepts
132
appendix
2.3 Interrogative 1
133
early papers on basic concepts
2.4 Interrogative 2
134
appendix
2.5 Imperative
135
early papers on basic concepts
2.6 Transitivity
136
appendix
2.7 Theme
137
early papers on basic concepts
Examples
1 2 3 4 5 6 7 8 9 10
//1 who said / that // 0 0 1 - 0 - 1 - 0 -
//2 who said / that // 0 0 1 - 1 0 1 - 0 -
//1 who was it / said that // 0 0 1 - 0 - 1 - 1 0
//2 who was it / said that // 0 0 1 - 1 0 1 - 1 0
//2 it was / who said / that // 0 0 1 - 1 0 1 - 1 1
//1 who did you / see // 0 0 1 - 0 - 0 0 0 -
//2 who did you / see // 0 0 1 - 1 1 0 0 0 -
138
appendix
Key to examples
Examples
1 2 3 4 5 6
//2 . did / who do it // 0 1 0 0 - 0
//1 . did / who do it // 0 1 1 0 - -
//2 did he do / what // 0 0 0 0 - 1
//1 did he go / where // 0 0 1 0 - -
//1 who said / what // 1 - - 0 0 -
//1 who was it / said / what // 1 - - 1 0 -
//1 what did / who say // 1 - - 0 0 -
//2 where did / who go // 1 - - 0 1 1
//2 who went / where // 1 - - 0 1 0
//2 where was it / who went // 1 - - 1 1 0
139
early papers on basic concepts
Examples
1 2 3 4 5 6 7 8 9
//1 ask / John // 0 - 0 0 0 0 - 0 0
//1 you ask / John // 0 - 0 0 0 1 - 0 0
//1 you ask / John // 1 0 0 0 0 1 - 0 0
//1 let’s ask / John // 0 - 0 0 1 - - 0 0
//1 let’s ask / John // 1 0 0 0 1 - - 0 0
//1 do ask / John // 0 - 0 1 0 0 - 0 0
//13 do ask / John // 1 0 0 1 0 0 - 0 0
//1 you / do ask / John // 1 0 0 1 0 1 0 0 0
//1 do let’s ask / John // 0 - 0 1 1 - - 0 0
//13 do let’s ask / John // 1 0 0 1 1 - 0 0 0
//3 don’t ask / John // 0 - 1 - 0 0 - 0 0
//13 don’t ask / John // 1 0 1 - 0 0 - 0 0
//4 don’t / you ask / John // 1 0 1 - 0 1 1 1 0
//1 don’t / let’s ask / John // 1 0 1 - 1 - 1 0 0
// 1 let’s / not ask / John // 1 0 1 - 1 - 0 0 0
//13 don’t let’s ask / John // 1 0 1 - 1 - 0 0 0
//13 let’s not ask / John // 1 0 1 - 1 - 1 0 0
//1 you / ask / John // 1 1 0 0 0 1 - 0 0
//5 don’t let’s / ask John // 1 1 1 - 1 - - 1 0
//53 . let’s / ask / John // 0 - 0 0 1 - - - 1
//1 you ask //4 John - - 0 0 0 1 - - 1
140
appendix
Examples
1 2 3 4 5 6 7 8 9
//1 John / saw the / play / yesterday // 0 - 0 0 - 0 0 0 00
//13 John / saw the / play / yesterday // 0 - 0 0 - 0 0 0 01
//4 John //1 saw the / play / yesterday // 0 - 0 0 - 0 0 0 10
//1 John / saw the / play / yesterday // 0 - 0 0 - 0 0 1 00
//1 John / saw the / play / yesterday // 0 - 0 0 - 0 0 1 00
//13 John / saw it / yesterday the / play // 0 - 0 0 - 0 1 0 01
//13 . he / saw the / play / yesterday / John // 0 - 0 0 - 1 0 0 01
//4 . it / wasn’t / John that / saw the / play /
yesterday // 0 - 0 1 1 0 0 1 00
//13 . it was / John that / saw the / play /
yesterday // 0 - 0 1 0 0 0 1 01
//4 . it / wasn’t the / play John / saw /
yesterday // 1 1 0 1 1 0 0 1 00
//13 . it was the / play John / saw / yesterday // 1 1 0 1 0 0 0 1 01
//4 John he //1 saw the / play / yesterday // 0 - 1 0 - 0 0 0 10
//1 yesterday / John / saw the / play // 1 0 0 0 - 0 0 0 00
//4 yesterday //1 John / saw the / play // 1 0 0 0 - 0 0 0 10
//1 . the / play / John saw / yesterday // 1 1 0 0 - 0 0 0 00
//4 . the / play //1 John saw / yesterday // 1 1 0 0 - 0 0 0 10
//4 . it / wasn’t / yesterday John / saw the /
play // 1 0 0 1 1 0 0 1 00
Examples of theme and information systems (from spontaneous
conversation)
in the first month one was too ill to move
adjudicator I thought they were called
the sound that went floating out on the air I didn’t know I had it in me
aged legal gentlemen all like pipes
the metal container somehow it turns your coffee rather sour
141
early papers on basic concepts
142
appendix
143
early papers on basic concepts
7 Verbal group
144
appendix
present 14
future 15
present in past in past 16
present 17
future 18
present in future in past 19
present 20
future 21
future in past in past 22
present 23
future 24
past in future in past in past 25
present 26
future 27
present in past in future in past 28
present 29
future 30
present in future in past in past 31
present 32
future 33
present in past in future in past in past 34
present 35
future 36
145
early papers on basic concepts
146
appendix
147
early papers on basic concepts
8 Nominal group 1
8.1 Principal systems
148
appendix
8.2 Determiners
149
early papers on basic concepts
8.3 Determiners
150
appendix
8.4 Quantifiers
151
This page intentionally left blank
SECTION TWO
WORD–CLAUSE–TEXT
This page intentionally left blank
EDITOR’S INTRODUCTION
155
word–clause–text
156
editor’s introducton
157
Chapter Six
158
lexis as a linguistic level
159
word–clause–text
160
lexis as a linguistic level
and third, strong (but not powerful) is a member of a class entering into
this relation with a class of which tea is a member. It would be hoped
that such classes would reappear elsewhere in the grammar defined on
other criteria. Argument, car and tea will, for example, already have been
distinguished on other grounds on the lines of “abstract”, “concrete
inanimate” and “mass”; but these groupings are not applicable here,
since we can have a strong table and powerful whisky, while a strong device
is at least questionable.
The same patterns do reappear: he argued strongly, I don’t deny the
strength of his argument, his argument was strengthened by other factors.
Strongly and strength are paralleled by powerfully and power, strengthened
by made more powerful. The same restrictions have to be stated, to
account for the power (but not the strength) of his car and the strength (but
not the power) of her tea. But these involve different structures; elsewhere
in the grammar strong, strongly, strength and strengthened have been
recognized as different items and assigned to different classes, so that
the strong of his argument has been excluded on equal terms with the
strong of his car. Strong and powerful, on the other hand, have been
assigned to the same class, so that we should expect to find a powerful
car paralleled by a strong car. The classes set up to account for the
patterns under discussion either will cut across the primary dimension
of grammatical classification or will need to be restated for each primary
class.
But the added complexity involved in either of these solutions does
not seem to be matched by a gain in descriptive power, since for the
patterns in question the differences of (primary) class and of structure
are irrelevant. Strong, strongly, strength and strengthened can all be
regarded for this present purpose as the same item; and a strong argument,
he argued strongly, the strength of his argument and his argument was
strengthened all as instances of one and the same syntagmatic relation.
What is abstracted is an item strong, having the scatter strong, strongly,
strength, strengthened, which collocates with items argue (argument) and
tea; and an item power (powerful, powerfully) which collocates with argue
and car. It can be predicted that, if a high-powered car is acceptable, this
will be matched by a high-powered argument but not by high-powered tea.
It might also be predicted, though with less assurance, that a weak
argument and weak tea are acceptable, but that a weak car is not.
As far as the collocational relation of strong and argue is concerned, it
is not merely the particular grammatical relation into which these items
enter that is irrelevant; it may also be irrelevant whether they enter
into any grammatical relation with each other or not. They may be in
161
word–clause–text
162
lexis as a linguistic level
163
word–clause–text
164
lexis as a linguistic level
kind of pattern and attempt to subsume all formal relations within it:
some grammatical models, as has been noted, envisage that it is the
grammar’s task to distinguish strong from powerful as well as to dis-
tinguish a from the and “past” from “present”; while a lexicographical
model in which a and the, as well as strong and powerful, are entered in
the dictionary and described by means of citations could be regarded as
in a similar way attempting to subsume grammar under lexis. Even
where the model recognizes two distinct kinds of pattern, these still
represent different properties of the total phenomenon of language, not
properties of different parts of the phenomenon; all formal items enter
into patterns of both kinds. They are grammatical items when described
grammatically, as entering (via classes) into closed systems and ordered
structures, and lexical items when described lexically, as entering into
open sets and linear collocations. So in a strong cup of tea the grammar
recognizes (leaving aside its higher rank status, for example as a single
formal item expounding the unit “group”) five items of rank “word”
assignable to classes, which in turn expound elements in structures and
terms in systems; and the lexis recognizes potentially five lexical items
assignable to sets.
But, to take a further step, the formal items themselves vary in
respect of which of the two kinds of pattern, the grammatical or the
lexical, is more significant for the explanation of restrictions on their
occurrence qua items. The items a and of are structurally restricted, and
are uniquely specified by the grammar in a very few steps in delicacy;
collocationally on the other hand they are largely unrestricted. For the
item strong, however, the grammar can specify uniquely a class (sub-
class of the “adjective”) of which it is a member, but not the item itself
within this class; it has no structural restrictions to distinguish it from
other members of the class (and if the members of its ‘scatter’ strong,
strength, etc., turn out to operate collocationally as a single item then
this conflated item is not even specifiable qua class member); colloca-
tionally, however, it is restricted, and it is this which allows its
specification as a unique item. There might then appear to be a scale
on which items could be ranged from “most grammatical” to “most
lexical”, the position of an item on the scale correlating with its overall
frequency ranking. But these are three distinct variables, and there is
no reason to assume a correlation of “most grammatical” with either
“least lexical” or “most frequent”. The “most grammatical” item is one
which is optimally specifiable grammatically: this can be thought of as
“reducible to a one-member class by the minimum number of steps in
delicacy”. Such an item may or may not be “least lexical” in the sense
165
word–clause–text
166
lexis as a linguistic level
167
word–clause–text
168
lexis as a linguistic level
169
word–clause–text
170
lexis as a linguistic level
the results to be interesting; and if they are, this is yet another indication
of the great insight into the nature of language that is so characteristic
of J. R. Firth’s contribution to linguistic studies.
Notes
1. See especially Firth (1935) (reprinted in Firth 1957b).
2. Firth (1957b: 195–6): ‘It must be pointed out that meaning by collocation
is not at all the same thing as contextual meaning, which is the functional
relation of the sentence to the processes of a context of situation in the
context of culture . . . Meaning by collocation is an abstraction at the
syntagmatic level and is not directly concerned with the conceptual or
idea approach to the meaning of words.’ Compare also Firth (1935) (in
“lexical items”): “This (sc. the lexical) function should not be misnamed
semantic”.
3. See Firth (1957a: 12). In the present paper “lexical level” has been used
in preference to “collocational level” in order to suggest greater generality
and parallelism with the grammatical level.
4. It is also stated explicitly by Firth (1957a: 12): “Collocations of a given
word are statements of the habitual or customary places of that word in
collocational order but not in any other contextual order and emphatically
not in any grammatical order”. Note that here “order” refers to the
“mutual expectancy” of syntagmatically related categories, such as ele-
ments of structure in grammar or phonology, and not to linear sequence:
cf. ibid., pp. 5, 17 and Halliday (1961: 254–5 (see above, Chapter 2)).
5. That is, that distinctions are made which involve the recognition of more
finely differentiated syntagmatic relations in the grammar, and that these
in turn define further sub-classes on various dimensions within previously
defined classes.
6. The place of collocational restrictions in a transformational grammar is
considered by Matthews (1961).
7. For a discussion of the relation between grammatical and lexical patterns
see McIntosh (1961).
8. Sinclair (1966).
9. The implication is, in effect, that “wellformedness” is best regarded as
“lexicogrammaticalness”, and that a departure from wellformedness may
be ungrammatical, unlexical or unlexicogrammatical. That the last two
are distinct is suggested by such examples as sandy hair, sandy gold and
sandy desk: sandy desk is unlexical, in that this collocation is unlikely to
occur in any grammatical environment, whereas sandy gold is merely
unlexicogrammatical: there is nothing improbable about golden sand. An
analogous distinction is observable in clichés: in shabby treatment the
mutual expectancy is purely lexical, and is paralleled in they treated him
171
word–clause–text
shabbily, a shabby way to treat him and so on, whereas the collocation faint
praise is restricted to this structure, in the sense that it will not occur with
similar probability under other grammatical conditions.
10. Compare the methods used to assess the disponibilité of lexical items in the
development of “Français fondamental” (Gougenheim, Michea, Rivenc
and Sauvageot 1956).
11. Roget (1960: 8).
12. Cf. Firth (1951).
13. The following text examples (drawn from written work by learners of
English) may be cited in this connection: festive animals, circumspect
beasts, attired with culture, funny art, barren meadows, merry admiration,
the situation of my stockings was a nightmare, lying astray, fashionable
airliner, modern cosy flights, economical experience, delightfully stressed,
serious stupid people, shining values, a wobbly burden, light possibility,
luxurious man, whose skin was bleeding, driving a bicycle, old and
disturbed bits of brick wall, a comprehensive traffic jam, her throat
became sad, my head is puzzled, people touched with assurance, thoughts
are under a strain, a sheer new super car.
14. This research is being undertaken by Dr. A. R. Meetham and Dr. P. K.
T. Vaswani at the National Physical Laboratory, Teddington, Middlesex.
15. Sinclair, Jones and Daley (1970).
172
Chapter Seven
First published in New Horizons in Linguistics, 1970, edited by John Lyons. Harmondsworth:
Penguin, pp. 140–65.
173
word–clause–text
174
language structure and language function
speaker’s experience of the real world, including the inner world of his
own consciousness. We may call this the ideational function, though it
may be understood as easily in behavioural as in conceptual terms
(Firth 1968: 91). In serving this function, language also gives structure
to experience, and helps to determine our way of looking at things, so
that it requires some intellectual effort to see them in any other way
than that which our language suggests to us.
2. Language serves to establish and maintain social relations: for the
expression of social roles, which include the communication roles
created by language itself – for example the roles of questioner or
respondent, which we take on by asking or answering a question; and
also for getting things done, by means of the interaction between one
person and another. Through this function, which we may refer to as
interpersonal, social groups are delimited, and the individual is identi-
fied and reinforced, since by enabling him to interact with others
language also serves in the expression and development of his own
personality.
These two basic functions, to each of which corresponds one broad
division in the grammar of a natural language, are also reflected in
Bernstein’s studies of educational failure (e.g. Bernstein 1970). Bern-
stein’s work suggests that in order to succeed in the educational system
a child must know how to use language as a means of learning, and
how to use it in personal interaction; these can be seen as specific
requirements on his control of the ideational and interpersonal func-
tions of language.
3. Finally, language has to provide for making links with itself and
with features of the situation in which it is used. We may call this the
textual function, since this is what enables the speaker or writer to
construct “texts”, or connected passages of discourse that is situationally
relevant; and enables the listener or reader to distinguish a text from a
random set of sentences. One aspect of the textual function is the
establishment of cohesive relations from one sentence to another in a
discourse (Hasan 1968).
All these functions are reflected in the structure of the clause. Here
we attempt to show, by reference to English, what a clause is: how it
serves for the realization of a number of very general meanings, or
semantic options, relating to the interpersonal, ideational and textual
functions of language; and how these are expressed through various
configurations of structural “roles” – functional elements such as
“process” and “actor” that derive from these basic functions. For a
more detailed exemplification we shall consider an aspect of ideational
175
word–clause–text
176
language structure and language function
177
word–clause–text
178
language structure and language function
This also illustrates the conflict of criteria. In (6i), by the fire might be
considered instrument rather than actor, on the grounds that it is
inanimate. Fillmore (1968) distinguishes actor and instrument as,
respectively, the “typically animate perceived instigator of the action”
(his “agentive”; cf. below, Section 8) and the “inanimate force or
object causally involved in the action”; the latter may also be grammat-
ical subject, and if not may also be expressed by with as in (7):
(7i) the key opened the door / John opened the door with the key
(7ii) the door was opened with the key
But with is not normally used where the action is unintentional (the
window was broken with the ball is odd), nor can it be substituted in (6i).
179
word–clause–text
180
language structure and language function
what was he doing on the bridge? and not what was he throwing stones on?)
However, the sense of “inner” and “outer” is contributed to by various
factors not all of which coincide. For example, in (10) the place
element is obligatory in (i) but optional in (ii):
(10i) he put all his jewels in the wash
(10ii) he lost all his jewels in the wash
In (11), there is a difference of clause type; (i) is a relational clause
(see below, Section 7) whereas (ii) is an action clause (Fillmore, from
whom (11) is taken, gives this as an instance of dependency between
functions: the place element is “outer” if an actor is present and “inner”
otherwise):
(11i) John keeps his car in the garage
(11ii) John washes his car in the garage
5 Inherent functions
The distinction between obligatory and optional roles helps us to relate
transitivity functions to a system of clause types. As, however, this
involves recognizing that an “obligatory” element may in fact be
absent, we shall use the term “inherent” rather than “obligatory”. An
inherent function is one that is always associated with a given clause
type even if it is not necessarily expressed in the structure of all clauses
of that type. (We are not here talking about ellipsis, which is a matter
of textual structure.)
Consider a pair of clauses such as (12):
(12i) Roderick pelted the crocodile with stones
(12ii) the crocodile got pelted
The verb pelt, as it happens, is always associated with three participant
roles: a pelter, a pelted and something to pelt with; and this holds for
(ii) as well as for (i) (cf. Svartvik (1966), on “agentless agentives”).
Similarly there are inherently benefactive clauses without a beneficiary,
such as we’re giving a silver coffee-pot. So:
(12iii) Roderick pelted the crocodile
is “(inherently) instrumental”, and although no instrument is men-
tioned the receiver interprets the process as having an instrumental role
associated with it.
The same verb may occur in clauses of more than one type. But
181
word–clause–text
within one type there may be different sets, and different alignments,
of participants; this is the function of the system of voice – of the choice
between active and passive, though the actual patterns are more
elaborate than this. The options in the voice system (simplifying
somewhat) are (a) middle / non-middle (see next paragraph); if non-
middle, then (b) “active” / “passive” (not exactly equivalent to active
and passive in the verb; see Halliday, 1967–68: § I 39ff., where they
are referred to as “operative” and “receptive”); if “active”, then (c)
plus / minus goal; if ‘passive’, then (d) plus / minus actor. The reason
for choosing one rather than another of these options lies in the textual
function of language (see Sections 11 and 12 below); but which options
are available to choose from depends on transitivity.
Voice is concerned with the roles of Actor and Goal (but see below,
Section 8), both as inherent and as actualized roles. A “middle” clause
is one which has only one inherent participant, which for the moment
we will continue to refer to as the “Actor”; examples are Hector sneezed,
the cat washed. A “non-middle” clause is one which has two, an Actor
and a Goal, but one or the other may not be actualized: if “active”,
there may be no Goal, for example Mary is washing (“the clothes”), and
if ‘passive’, no clear Actor, e.g. the clothes have been washed (“by Mary”).
All actions are classified into those involving one participant role and
those involving two; there are then different ways of presenting the
situation in those cases where there are two.
The point was made earlier that the notion of “participant” derives
from the more fundamental concept of syntactic function, or “role”.
The basic elements of transitivity structure are the various roles associ-
ated with processes; and two or more such roles may be combined
in one participant, as in a reflexive clause such as John is washing
(“himself ”) where John is both Actor and Goal at the same time. The
elements that operate as Actor, Goal, etc. also play a part, simul-
taneously, in other structures of the clause, expressing aspects of the
interpersonal and textual functions of language. The principle of
combining a number of roles in a single complex element of structure
is fundamental to the total organization of language, since it is this that
makes it possible for the various functions of language to be integrated
in one expression. We return to this below, Section 9.
182
language structure and language function
{
n active
m passive
i
d passive
d
l passive
e
Actor, (Goal) active the Council won’t sell
Goal
Goal, Actor
active the gazebo won’t sell
passive the gazebo has been sold by the Council
Goal, (Actor) passive the gazebo has been sold
183
word–clause–text
voice (15i), and those having the Processer (15ii). In the first types, the
passive form is much more frequent than the passive in action clauses;
in the second type it is much less so:
(15i) the gift pleased her / she was pleased by (with) the gift
(15ii) she liked the gift / the gift was liked by her
This is because the passive is a means of bringing the element governed
by by into prominence as the focus of information (see below, Section
12); in (15ii) the by element, i.e. her, is the Processer, and in English
this tends to be the “given” element in the situation (she must have
been referred to already in the text), and thus does not appropriately
carry such prominence.
Mental process clauses express (a) perception, e.g. see, look; (b)
reaction, e.g. like, please; (c) cognition, e.g. believe, convince; (d) verbali-
zation, e.g. say, speak. They are distinct in that the Phenomenon – that
which is perceived, reacted to, etc. – is not limited, as are the
participants in action clauses, to the class of “thing”, namely persons,
objects, abstractions and the rest of the phenomena on the plane of
experience.
What is perceived or felt or thought of may be a simple phenom-
enon of this kind, but it may also be what we might call a metaphe-
nomenon: a fact or a report – a phenomenon that has already as it were
been filtered through the medium of language. Here words as well as
things may participate in the process.
For example, in (16) all the “processed” entities are simple phenom-
ena, or “things”:
(16i) I noticed Helen over there [person]
(16ii) I noticed a discrepancy [abstraction]
(16iii) I noticed a quarrel (going on) / them quarrelling [event]
(16iv) I noticed what (the thing that) she was wearing [object]
In (17) and (18), however, they are metaphenomena; facts in (17),
reports in (18):
(17i) I noticed what (the fact of what) she was wearing
(17ii) it worries me that you look so tired
(18i) I notice the bank rate’s going up again
(18ii) he says the bank rate’s going up again
We could insert the fact (that) in (17) and the report (that) in (18i); not,
however, in (18ii), which is a clause of verbalization, since such clauses
accept only reports, and “reported speech” is the meaning of clauses of
184
language structure and language function
this type. The difference between fact and report is that a “fact” is a
representation at the semantic level, where the truth lies in the meaning
– (she regretted) that he had gone away; whereas a “report” is a represen-
tation at the lexicogrammatical, or syntactic, level, where the truth lies
in the wording – (she said) that he had gone away.
In relational clauses, the “process” is simply a form of relation
between two roles. One type is the attributive, such as:
(19i) Marguerite is a poet
(19ii) Marguerite looks desperate
where the relation is one of class membership: “Marguerite belongs to
the class of poets,” “. . . the class of people who look desperate”. This
is a relation between entities of the same order of abstraction but
differing in generality.
The other type, exemplified by (20):
(20i) Templecombe is the treasurer
(20ii) the treasurer is Templecombe
has two functions, resembling the two terms of an equation, where the
one serves to identify the other, as in x = 2. Here the two entities are
alike in generality but differ in abstraction: the identifying element may
be of a higher order of abstraction, as in (21i), where the treasurer
expresses Templecombe’s function, or of a lower order, as in (21ii)
where the fat one expresses Templecombe’s form, how he is to be
recognized:
(21i) (which is Templecombe?) Templecombe is the treasurer
(21ii) (which is Templecombe?) Templecombe is the fat one
(21i) could be interpreted in the sense of (21ii) if the committee were
in view on the platform; there is in fact partial ambiguity between
these two sub-types.
These two major types of relational clause, the attributive and the
equative, differ in various respects. The attributive are non-reversible
(e.g. we can say that man is a poet but not a poet is that man), have the
role Attribute which may be an adjective and is usually indefinite,
express class inclusion, are usually questioned by what? or how? and are
expressed by the verbs be, get, turn, keep, remain, seem, sound, look, etc.
The equative are reversible (i.e. have a “voice” system), have the role
Identifier which must be as noun and is usually definite, express class
identity, are usually questioned by who? or which? and are expressed by
the verbs be, equal, represent, resemble, stand for, etc.
185
word–clause–text
8 The ergative
As far as the ideational component of grammar is concerned, the
English clause shows the three principal types – action, mental process
and relation – and associates with each a set of different inherent roles,
or structural functions. The system of clause types is a general frame-
work for the representation of processes in the grammar; possibly all
languages distinguish three such categories. We need to ask, at this
point, whether the structural functions can be generalized across clause
types; whether, for example, an Actor in an action clause can be shown
to be equivalent to a Processer (one who does the thinking, etc.) in a
mental process clause. This may be approached through a reconsidera-
tion of the functions in action clauses, a reconsideration which such
clauses demand anyway.
If we look at examples like (23i and ii):
(23i) the sergeant led the recruits
(23ii) the sergeant marched the recruits
they appear to be clearly distinct, (i) being transitive, with Actor and
Goal, (ii) causative, with Initiator and Actor. However, there is a
problem with (23iii):
(23iii) the sergeant trained the recruits
Is it like (23i) or like (23ii)?
Actually it is like both; (23i) and (23ii) are not really different as far
as transitivity is concerned. In English no very clear distinction is made
between doing something to someone and making someone do some-
thing, so that (23iii) can be interpreted in either way without any sense
of ambiguity. This is why so many verbs are labelled “vb trans. &
intrans.” in the dictionary.
The concepts of actor and goal are not well suited to describing this
situation, since with these we are forced to describe (23i) and (23ii)
differently. The distinction between them is by no means entirely
186
language structure and language function
unreal, since there are verbs like lead which are normally transitive (two
inherent participants) and others like march which are normally intran-
sitive (one inherent participant). But with a large number, especially of
the more frequently used verbs, either form seems equally normal:
there is nothing to choose, as regards the more typical use of the verb
bounce, between he bounced the ball and the ball bounced. In addition there
are a number of verbs which, while themselves clearly transitive or
clearly intransitive, group into pairs differing only in transitivity, so that
Mary put out the fire is to the fire went out as Polly lit the fire is to the fire
lit.
It has been pointed out by various linguists (Halliday 1967–68: § 3;
Anderson 1968; Fillmore 1968) that action clauses in English seem to
be organized on an ergative rather than on a transitive (or “nomina-
tive”) basis. This means that, with any action clause, there is associated
one inherent role which is that of the participant affected by the process
in question. Fillmore describes this as the “semantically most neutral”
function, and labels it the “objective”; I used the term “affected”,
which I will retain here. In (23) the recruits has the role of “affected” in
every case, even through it is Goal (if an Actor–Goal analysis is used)
in (23i) and Actor in (23ii); in general, the affected is the Goal in a
transitive and the Actor in an intransitive clause.
We have now turned what was the borderline case, such as (23iii),
into the most central clause type. This is the type in which both middle
(one-participant) and non-middle (two-participant) forms are equally
normal; it may be considered the “favourite” clause type of Modern
English. The transitive and intransitive types – those with non-middle
as norm and with middle as norm respectively – are the marginal ones,
and they seem to be becoming more marginal as time goes on.
Hence all the examples in (24i) have the same structure, with a
Process and an Affected. Those in (24ii) also have a Causer (Fillmore’s
“agentive”):
(24i) they’re being led
they’re being trained / they’re training
they’re being marched / they’re marching
(24ii) he’s leading them
he’s training them
he’s marching them
These two ways of representing processes, the transitive and the
ergative, are very widely distributed; possibly all languages display one
or the other, or (perhaps always) both, in different mixtures. In English,
187
word–clause–text
the two occur side by side. The transitive system asks “does the action
extend beyond the active participant or not?”, the ergative, “is the
action caused by the affected participant or not?” The ergative com-
ponent is more prominent now than it was in Middle English, and this
appears in various ways, for example, the change from impersonal to
personal forms in mental process clauses (formerly methinks, it likes me).
In the modern form I like, I cannot be explained as an Actor (among
other things we cannot say what he does to jam is like it); but it can be
shown on various grounds to have the function Affected.
As this suggests, the ergative pattern, whereby a Process is accom-
panied by an obligatory Affected participant and an optional Causer, is
more readily generalizable than that of Actor and Goal. It extends
beyond action clauses to those of mental process, and perhaps even to
clauses of relation as well. We want to say that Paul has the same
function in both (25i) and (25ii):
– not that they are identical in meaning, but that the transitivity roles
are the same. This is not possible in Actor–Goal terms. But in an
ergative system there is considerable evidence for regarding Paul as the
‘Affected’ participant in both cases. The ergative, therefore, represents
the more general model of the transitivity patterns of Modern English
– that is, of the options available to the speaker of English for talking
about processes of all kinds.
188
language structure and language function
10 Mood
As we have said, one function of language is to provide for interaction
between people, by allowing the expression of statuses, social and
individual attitudes, assessments, judgements and the like; and this
includes participation in linguistic interaction. Language itself defines
the roles which people may take in situations in which they are
communicating with one another; and every language incorporates
options whereby the speaker can vary his own communication role,
making assertions, asking questions, giving orders, expressing doubts
and so on. The basic “speech functions” of statement, question,
response, command and exclamation fall within this category (though
they do not exhaust it), and these are expressed grammatically by the
system of mood (cf. Sweet 1891: 105), in which the principal options
are declarative, interrogative (yes / no and wh- types) and imperative.
The difference between he can and can he? is a difference in the
communication role adopted by the speaker in his interaction with a
listener.
The notion “grammatical subject” by itself is strange, since it implies
a structural function whose only purpose is to define a structural
189
word–clause–text
11 Theme
The basic unit of language in use is not a word or a sentence but a
“text”; and the “textual” component in language is the set of options
by means of which a speaker or writer is enabled to create texts – to
use language in a way that is relevant to the context. The clause, in
this function, is organized as a message; so in addition to its structure
in transitivity and in mood, it also has structure as a message, what is
known as a “thematic” structure. (It was linguists of the Prague school
who first studied this aspect of language, cf. Mathesius 1928; Firbas
1959; 1964; Svoboda 1968 and references therein.)
The English clause consists of a Theme and a Rheme. The Theme
is another component in the complex notion of subject, namely the
“psychological subject”; it is as it were the peg on which the message
is hung, the theme being the body of the message. The Theme of a
clause is the element which, in English, is put in first position; in (28
190
language structure and language function
i–v) the Theme is the item outside the brackets, what is inside being
the Rheme:
(28i) I (don’t know)
(28ii) yesterday (we discussed the financial arrangements)
(28iii) his spirit (they could not kill)
(28iv) suddenly (the rope gave way)
(28v) people who live in glasshouses (shouldn’t throw stones)
As we have seen, Theme, Actor and modal Subject are identical unless
there is good reason for them not to be (cf. (26) above). Where they
are not, the tendency in Modern English is to associate Theme and
modal Subject; and this is the main reason for using the passive. The
passive has precisely the function of dissociating the Actor from this
complex, so that it can either be put in focal position at the end or,
more frequently, omitted, as in (29):
(29i) this gazebo was built by Sir Christopher Wren
(29ii) this gazebo is being restored
The typical theme of declarative clause is thus the modal Subject (or
“grammatical subject” – this gazebo in both cases); in interrogatives,
however, the picture is different. If we ask a question, it is usually
because we want to know the answer, so that the typical Theme of an
interrogative is a request for information. Hence we put first, in an
interrogative clause, the element that contains this request for infor-
mation: the polarity-carrying element in a yes / no question, and the
questioning element in a “wh-” question, as in (30):
(30i) didn’t (Sir Christopher Wren build this gazebo?)
(30ii) how many gazebos (did Sir Christopher Wren build?)
In English there is a definite awareness of the meaning expressed by
putting something in first position in the clause. The Theme is the
point of departure for the message; a paradigm form of it is the
headword in a definition, for example a gazebo in (31):
(31) a gazebo is a pavilion or summerhouse on an eminence, open
for the view
In addition to the selection of a particular element as the theme, the
speaker has other options in thematic structure open to him (Halliday
1967–68: § 2); for example, any clause can be split into two parts by
the use of nominalization, as in
191
word–clause–text
(32) the one who built this gazebo was Sir Christopher Wren
where the Theme is the whole of whichever part comes first – here the
one who built this gazebo.
12 Information structure
Thematic structure is closely linked to another aspect of the textual
organization of language, which we may call information structure. This
refers to the organization of a text in terms of the functions Given and
New. These are often conflated with Theme and Rheme under the
single heading “topic and comment”; the latter, however, is (like the
traditional notion of “subject”) a complex notion, and the association of
Theme with Given, Rheme with New, is subject to the usual “good
reason” principle already referred to – there is freedom of choice, but
the Theme will be associated with the Given and the Rheme with the
New unless there is good reason for choosing some other alignment.
In English, information structure is expressed by intonation. Con-
nected speech takes the form of an unbroken succession of distinctive
pitch contours, or tone groups; each tone group represents what the
speaker decides to make into one unit of information. This is not
necessarily the same length as a clause, though it often is so. The
information unit consists of an obligatory New element – there must
be something new, otherwise there would be no information – and an
optional Given element; the main stress (tonic nucleus) marks the end
of the New element, and anything that is Given precedes it, unless
with good reason – which means, here, unless it is a response to a
specific question, either asked or implied. The function Given means
‘treated by the speaker as non-recoverable information’: information
that the listener is not being expected to derive for himself from the
text or the situation.
(33) illustrates the interaction of information structure with thematic
structure (information unit (“i.u.”) boundaries are marked by //; main
stress is indicated by bold type; 4 = falling-rising tone; 1 = falling
tone):
(33) //4 this gazebo //1 can’t have been built by Wren//
(clause: Theme . . . Rheme. . . . . . . . . . . . . . . . . . . .)
(i.u.(1): New. . . . . . .; (2): New. Given. . . . . . . . . . . . . . . .)
meaning ‘I am talking (Theme), specifically, (New) about this gazebo:
the fact is (Rheme) that your suggestion (Given) that Wren built it is
192
language structure and language function
193
word–clause–text
13 Conclusion
The subject, in its traditional sense, is thus a complex of four distinct
functions, three in the structure of the clause (cf. Lyons 1968: 343–4):
1. Actor (“logical subject”): ideational
2. modal Subject (“grammatical subject”): interpersonal
3. Theme (“psychological subject1”): textual
together with a fourth function which is in the structure of the
‘information unit’
4. Given (‘psychological subject2’): textual
These coincide unless there is “good reason” for them not to do so;
thus in (35i) the Borough Council is Actor, modal Subject and Theme,
whereas in (35ii) the Borough Council is Actor, this gazebo is modal
Subject and next year is Theme:
(35i) the Borough Council will restore this gazebo next year
(35ii) next year this gazebo will be restored by the Borough Council
No mention has been made of subject and predicate as a logical
relation. We might introduce “predication” as another dimension of
clause structure, with the Borough Council in (35i) being also “subject in
predication” and the rest predicate; but the subject in this sense would
be identical with the modal subject. The subject-predicate structure is
entirely derivable from mood, and has no independent significance (cf.
Fillmore 1968: 17; and Fillmore’s reference, ibid., to Tesniēre 1959:
103–5). As a form of generalization, it may be useful in that it expresses
the fact that Actor, modal Subject and Theme are regularly associated;
but it obscures the equally important fact that they are distinct and
independent structure roles.
The multiple function of language is reflected in linguistic structure;
this is the basis for the recognition of the ideational (including logical),
interpersonal and textual functions as suggested here. It is not necessary
to argue that one function is more abstract, or “deeper”, than another;
all are semantically relevant. The investigation of these functions
enables us to relate the internal patterns of language – its underlying
options, and their realization in structure – to the demands that are
made on language in the actual situations in which it is used. As
performers and receivers, we simultaneously both communicate
through language and interact through language; and, as a necessary
condition for both of these, we create and recognize discourse (the
194
language structure and language function
195
Chapter Eight
1 Preamble
Let us say that a code is a system of signs having a content and an
expression: for example, a traffic control code (Figure 1).
Figure 1
The relation between the content and the expression is one of
realization. Then, a semiotic, or semiotic system, is a code having two
or more realizational cycles in it, so that the expression of content1 (call
it expression1) is itself a content (content2) that in turn has its own
expression (expression2). Hence there will be at least three levels, or
strata, in such a system:
level one: content1
two: (realized as) expression1 = content2
three: (realized as) expression2 . . .
A semiotic, in other words, is a stratified, or stratal, system, in which
the output of one coding process becomes the input to another.
First published in Function and Context in Linguistic Analysis: A Festschrift for William Haas,
1979, edited by D. J. Allerton, Edward Carney and David Holdcroft. Cambridge:
Cambridge University Press, pp. 57–79.
196
modes of meaning and modes of expression
Figure 2
Now it is typical of semiotic systems that the different strata are not
isomorphic; there is no relation of biuniqueness (one–one correspon-
dence) between one level and the next. This is bound to be the case in
a system such as language, where the coding not only converts elements
of one kind into elements of another kind – meanings into wordings
into sounds – but also reduces both the size and the inventory of the
basic components. By any usual definition of linguistic units, units of
speech sound are both smaller than and fewer than units of form; and
units of form are both smaller than and fewer than units of meaning.
197
word–clause–text
EXPERIENTIAL LOGICAL
The first of these is language as representation: the semantic system
as expression of experience, including both experience of what is round
about us in the outside world and experience of the world of conscious-
ness that is inside us. This we are calling the ideational component.
There are two subcategories: an experiential, where we represent
experience “directly” in terms of happenings (actions, events, states,
relations), entities that participate in these happenings (persons, animate
and inanimate objects, institutions, abstractions) and circumstantial
features (extent, location, time and space, cause, manner and so on);
and a logical, where we represent experience “indirectly” in terms of
certain fundamental logical relations in natural language – ‘and’,
‘namely’, ‘says’, ‘is subcategorized as’, etc. – which are not those of
formal logic but rather are the ones from which the operations of
198
modes of meaning and modes of expression
formal logic are ultimately derived. These two, the logical and the
experiential, together make up the ideational component in the seman-
tic system: that of meaning in the reflective mode.
The second main component, the interpersonal, is language as
interaction: it is meaning in the active mode. Here the semantic system
expresses the speaker’s intrusion in the speech event: his attitudes,
evaluations and judgements; his expectations and demands; and the
nature of the exchange as he is setting it up – the role that he is taking
on himself in the communication process, and the role, or rather the
role choice, that he is assigning to the hearer. This component is
therefore both speaker- and hearer-oriented; it is interpersonal – what
Hymes called “socio-expressive” – and represents the speaker’s own
intrusion into the speech situation.
All discourse involves an ongoing simultaneous selection of meanings
from both these components, which are mapped into a single output
in the realization process. But there is also a third component, which
we are calling the textual, whereby the meanings of the other two
kinds take on relevance to some real context. Here the semantic system
enables the speaker to structure meaning as text, organizing each
element as a piece of information and relating it significantly to what
has gone before. If the ideational component is language as reflection
(the speaker as observer of reality), and the interpersonal component is
language as action (the speaker as intruder in reality), the textual
component is language as relevance (the speaker as relating to the
portion of reality that constitutes the speech situation, the context
within which meanings are being exchanged). The textual component
provides what in modern jargon we might refer to as the ecology of
the text. For example, from the Walrus and the Carpenter, in Alice
Through the Looking-Glass, when the Carpenter says to the Walrus
199
word–clause–text
200
modes of meaning and modes of expression
201
word–clause–text
202
modes of meaning and modes of expression
elements. But we can formulate the general principle that this is how
processes are represented in languages. This means that a structure
which represents experiential meanings will tend to have this form: it
will be a configuration, or constellation, of discrete elements, each of
which makes its own distinctive contribution to the whole.
We usually represent this kind of structure linguistically as a func-
tionally labelled constituent structure as shown, for example, in Figure
3.
Figure 3
There is no particular reason at this stage why the representation should
have to be linear; that is no doubt the form of the final output, after
the other structures have been mapped on to it, but experiential
structures are not in fact sequential and we could just as well represent
this as in Figure 4. If ordering is to be introduced into the representa-
tion, there is the possibility of using
Figure 4
a dependency construct having the Process at the centre, as in Figure
5. But a more appropriate ordering would have a nucleus consisting of
Process plus Goal, with the other elements clustering around it, as in
Figure 6. (We will leave aside here the question of the appropriateness
of the functional labels themselves.)
Figure 5 Figure 6
203
word–clause–text
Figure 7 Figure 8
introduced into the structure in this way are the source of the linguist’s
classic mode of representation of constituent structure in the form of
trees, or labelled bracketings. But the appropriate bracketing for func-
tionally labelled structures is minimal, not maximal (“many ICs” rather
than “few ICs”, in the terms of Hudson 1967). Maximal bracketing
imposes too much structure for a functional grammar: for example, in
four young oysters it is reasonable to recognize young oysters as a
constituent provided the labelled elements are classes, since it is a
nominal group, but in no way does this correspond to any meaningful
functional constituent. The nonlinear representation implies more of a
molecular model of structure, with a taxonomy similar to cell: mol-
ecule: atom: subatomic particle.
Experiential meanings are typically realized as elemental structures of
this type. The basic structural mechanism is that of constituency, with
larger units constituted out of layered clusters or bracketed strings of
smaller units, each part having its own specific function with respect to
the whole. We could call this “segmental”, except that it is better
perhaps to reserve the term “segment” for an element in the final
output – the syntagm – that serves as input to the next realizational
cycle. So let us say that experiential meanings are realized through
some kind of constituent structure.
This expresses the particular way in which we order our experience
of reality when we want to turn it into meaning. The bounded entities
that enter into constituent structures with specific functions like Pro-
cess, Actor, Goal, Extent or Manner offer a presentation of reality in
terms of “things” – doings by, and happenings to, persons and objects,
in the environment of other persons and objects, with yet other persons
and objects, and also times and places and so on, as attendant circum-
stances; and including various “metathings” (facts and reports), which
are complex things that have already been encoded in language and so
acquired a status which enables them to participate in certain types of
204
modes of meaning and modes of expression
205
word–clause–text
206
modes of meaning and modes of expression
place where the greatest pitch movement takes place. (This may be a
falling movement or a rising movement or some kind of complex
movement, depending on which kind of melody it is. It corresponds
to what is sometimes called “primary stress”, although it is not, in fact,
a stress feature.) Suppose we represent the intonation unit (the tone
group) as bounded by double slash, and intermediate rhythmic units
(the feet) by single slash, with tonic prominence as bold type; the likely
form of the utterance would be:
207
word–clause–text
Figure 9
we should presumably have the structure shown in Figure 10. But
because a clause is typically one information (focal) unit, and the focus
in the information unit is typically at the end, final position in the
clause carries a potential for prominence which highlights it in the
same way that initial position is highlighted by prominence of a
thematic kind.
Figure 10
The structures that realize options in the textual component are
what we may call “culminative” structures. They are not configura-
tions or clusters of elements such as we find in the ideational compo-
nent; nor are they prosodic chains of the interpersonal kind. What the
textual component does is to express the particular semantic status of
elements in the discourse by assigning them to the boundaries; this
gives special significance to “coming first” and “coming last”, and so
marks off units of the message as extending from one peak of promi-
nence to the next.
The effect of this is to give a periodicity to the discourse. The clause,
in its status as a message, begins with prominence of one kind, thematic
prominence, and ends with prominence of another kind, prominence
due to information focus. The latter is expressed through the assign-
ment of the tonic accent to a particular place in the tone group; so the
prominence is also in part phonological – and can be heard. The
periodicity is further reinforced by the use of conjunctives to link one
sentence with another; these contribute to the texture by relating a
clause cohesively to what went before it, and they also occur at the
boundaries – usually the beginning, but sometimes, especially in casual
speech, at the end.
208
modes of meaning and modes of expression
Figure 11
Given structures of these very general kinds, it is clear that each can
be reduced to some form of constituency; but not all with the same
success. Experiential structures are quite constituent-like; whereas inter-
personal ones are not, and the attempt to represent them in constitu-
ency terms involves idealizing them to an extent that is tantamount to
209
word–clause–text
210
modes of meaning and modes of expression
Figure 13
211
word–clause–text
212
modes of meaning and modes of expression
Figure 14
These are of two kinds: paratactic and hypotactic. The paratactic
involve relations like “and” and “equals”, which are logically transitive
(A ‘&’ B, and B ‘&’ C, implies A ‘&’ C; A ‘=’ B, and B ‘=’ C, implies
A ‘=’ C). The hypotactic are logically non-transitive; these include
relations such as “if” and “says”, where a “if” ß and ß “if” y does not
imply a “if” y, nor does a “says” ß and ß “says” y imply a “says” y.
In paratactic structures, because they are transitive, recursive order
is expressed by structural (linear) sequence: there is no other way in
which it can be expressed – no way, that is, in which a sequence A
and B and C could realize an order “A & C & B”. The only departure
from this strict sequence is nesting, where a sequence A and B and C
and D represents an order, say, “A & (B & C) & D”. These are
typically signalled in the phonological structure, e.g.:
213
word–clause–text
Table 1
214
modes of meaning and modes of expression
the same way, the days of the week do not form a closed system, in
the linguistic sense; they form a lexical set whose members happen to
be limited and fixed. In both instances, the infiniteness of the set of
natural numbers and the finiteness of the set of days of the week are
properties not of the language but of the social system.
Logical structures are different in kind from all the other three. In
the terms of systemic theory, where the other types of structure –
particulate (elemental), prosodic and periodic – generate simplexes
(clauses, groups, words, information units), logical structures generate
complexes (clause complexes, group complexes, etc.) (Huddleston
1965). The apparent exception is the sentence, which is generated by
logical structures; but this is merely a terminological exception – the
sentence is, in fact, simply a name for the clause complex. While the
point of origin of a non-recursive structure is a particular rank – each
one is a structure “of” the clause, or of the group, etc. – recursive
structures are in principle rank-free: coordination, apposition, subcate-
gorization are possible at all ranks. The more restricted ones, like tense
and report, are also the ones that are nearer the borderline; they are
only just logical structures. Tense is particularly interesting because it
has only come into this category within the last two to three centuries,
and English appears to be unique in treating tense in this way.
9 Postscript
As a postscript, it should be noted that what is conceptually one and
the same kind of relation may be coded in the semantic system in more
than one way, i.e. may be realized through more than one of the
functional components. The “and” relation, for example, may be coded
in a logical system, expressed as coordination; or in a textual system,
expressed as conjunction. The same is true of “yet”, “so” and “then”;
these are much more weakly represented in the logical mode, but on
the other hand they can also be coded experientially. Table 2 shows
different codings of the temporal “then” relation.
Although the distinct types of structure discussed in this paper appear
at all ranks throughout the grammar, in English at least it is at the rank
of the clause that they are most clearly in evidence.
As an experiential construct, the clause is the locus of transitivity: it
is the representation of the processes, participants and circumstances
that constitute our experience of the real world. As an interpersonal
construct it is the locus of mood and modality: the speaker’s adoption
and assignment of speech roles and his judgement of the validity of the
215
word–clause–text
Table 2
Functional mode Example
logical : paratactic He sang, then people applauded, then . . .
logical : hypotactic After he had sung, people applauded
Textual (First) he sang. Afterwards, people applauded
Experiential Applause followed his song. / His song was followed
by applause
10 Summary
If we study the semantic system of a language we find that it consists
of three major functional components: an ideational, an interpersonal
and a textual; with the ideational further subdivided into an experiential
and a logical.
This pattern appears clearly at the semantic level itself: within each
component, the networks of systemic options are closely inter-
connected, whereas between one component and another there are
216
modes of meaning and modes of expression
Notes
1. Written language may be (i) an alternative coding of meanings (ideograms);
(ii) an alternative coding of wordings (characters, as in Chinese); or (iii) an
alternative coding of sounds (syllabaries and alphabets).
2. Mathesius’s observation relates to ‘functional styles’ (orientation towards
different functions in the use of language), not to functionally derived
components of the system; but it is pertinent nevertheless: ‘The influence
of functional styles on the lexical and semantic aspects of speech was
stressed especially by Gröber, . . . [who] distinguishes the subjective
expression . . . and the objective expression . . . The subjective expression
differs from the objective both quantitatively (inasmuch as it expresses by a
pause, by tone or by gesture what the latter expresses by words; and
further, as it repeats what could be expressed only once) and qualitatively (by
217
word–clause–text
choosing other words than factual names of the things referred to), and,
finally, locally (by placing sentence elements into positions not pertaining to them
in objective speech). Both ways of expression are often combined in actual
speech.’ (My italics throughout.)
218
Chapter Nine
219
word–clause–text
220
text semantics and clause grammar
221
word–clause–text
222
text semantics and clause grammar
223
word–clause–text
224
text semantics and clause grammar
225
word–clause–text
high school. She has shown how the conjunctive relation of specific
instance to general principle, coded in Martin’s network as:
(internal / comparative) : similarity : nonexhaustive
is the major conjunctive factor giving coherence to these texts. For
example, in one part of a text there occurs the sentence
Odysseus’ friends and anyone else who heard his story respected him.
Elsewhere in the same text we find:
Heroic men are very much respected and idolized.
(Either member of such a pair may come first.) There is lexical cohesion
between individual items – the repetition of respect. There is cohesion
between Odysseus and heroic (men), with ‘Odysseus’ being a hyponym
of ‘hero’. But between the two sentences as wholes there is a conjunct-
ive relation – itself an extension of hyponymy – such that the second
one stands as a general principle of which the first one offers a specific
instance; and this type of conjunction is a distinguishing feature of the
sort of expository discourse she is investigating.
226
text semantics and clause grammar
227
word–clause–text
228
text semantics and clause grammar
229
word–clause–text
230
text semantics and clause grammar
231
word–clause–text
We should not press the analogy too far. But if it seemed useful to
set up simultaneous structures in a text along these lines we might ask
whether there is the same kind of structural variation as we find in the
clause, with the eidological structure being “particulate” (represented
by definable segments of the text) and the ethological being “field-
like” (represented by overlapping prosodies in the text). (The “wave-
like” periodic movement corresponding to the textual dimension of
clause structure has already been referred to under 1.1.4 above; cf.
further below.) (Cf. Halliday 1979.)
To say that a text resembles a clause in having coherence is not to
say very much, since the coherence in a clause is created by its
structure, whereas coherence in a text largely depends on cohesion.
Cohesion is the resource that takes over, as it were, when grammatical
structure no longer holds (i.e. above the clause complex). We could
point out that cohesion also obtains within clauses; we find reference,
substitution, ellipsis, conjunction and lexical cohesion all operating
between elements in the same clause, for example:
M’s evening speech caused more fuss than his morning one had.
C R L S E
232
text semantics and clause grammar
That she died was caused by her not knowing the rules.
Her ignorance of the rules caused her death.
233
word–clause–text
closely analogous to what takes place in a text; not only over the whole
text but also in structurally defined intermediate units within the text.
The classic movement of a paragraph, beginning with a topic sentence
(from theme to elaboration) and culmination – having a high point,
unmarkedly but not obligatorily final – in a climax (from prelude to
main point), is one of the clearest manifestations of the analogy between
clause and text. It is in the clause that this movement is displayed in
the most systematic and clearly motivated form.
Finally a clause can be said to have “character” in both the generic
and the specific sense. If a text is typified by virtue of its being
organized around the expression of processes of a particular type, the
clause is the unit in which these processes are realized and categorized.
The clause is the locus of the transitivity system: the system of material,
mental and relational processes, together with their numerous subcat-
egories. Thus analogous to the major types or genres of text are the
major types or classes of the clause, each being characterized by the
selection of a dominant process type.
But each clause is also a unique combination, or potentially unique
combination, of features deriving from the different semantic functions,
ideational, interpersonal and textual. Moreover any one or other of
these may be foregrounded: the clause may display an orientation
towards any one, or any combination, of the various systems and their
subsystems. The extract from J. B. Priestley’s An Inspector Calls
(Appendix 5, pp. 257–9) gave an instance of a clause that is oriented
towards a certain type of modality: interpersonal meanings are high-
lighted, with the speaker’s skeptical doubting as the predominant
rhetorical function. The passage cited is a unique interplay between the
exploration of probabilities and the assertion of obligations, and so is
the entire text. No one clause can recapitulate the whole; but all
contribute, and some achieve a remarkable likeness – a likeness that is
possible because the systems of the clause not only embody all the
semantic components from which the text is built but do so in a way
that allows an infinitely varied, almost text-like balance among them.
Thus the properties that we recognize in a text are also, in a
transformed way, properties that we recognize in a clause. A clause is a
kind of metaphor for a text – and a text for a clause. That this is
possible is due to two things: one, that a text is not only (typically)
larger than but also more abstract than a clause; two, that on the other
hand there is no line of arbitrariness between clause and text as there is
between clause and syllable. Hence it is not only in a formal sense that
a text is like a clause. It is no accident that it is possible to illustrate so
234
text semantics and clause grammar
235
word–clause–text
236
text semantics and clause grammar
237
word–clause–text
238
text semantics and clause grammar
succeeding clause. Thus a clause is at one and the same time particle,
field and wave, as Pike suggested more than twenty years ago (Pike
1959), although the details of this interpretation are not quite the same
as those worked out by Pike.
Now, the significance of this step in our interpretation lies not only
in establishing that these three distinct patterns of realization go to
make up the English clause, but also in the fact that they appear to be
non-arbitrary; this is clearly important when we come to ask whether
such tendencies are found in every language. The grammar of languages
is a natural grammar; as I expressed it earlier, there is no line of
arbitrariness between semantics and grammar as there is between
grammar and phonology. If the clause is at once particle, field and
wave this is because the meanings it has to express have different
semiotic contours, to which these three realizational forms correspond
in a natural, non-arbitrary way.
The particular nature of ideational structures reflects the relative
discreteness of the phenomena of our experience. Consider cows eat
grass: we know where the cow begins and ends, what eating is and is
not, which part of reality consists of grass and which part consists of
other things. Many of our perceptions are schematized into entities that
are bounded in this way, and the constituent-like form of the wording
reflects this fact: the word cow has an outline because the object cow
has an outline. Of course not all experience is like this; indeed I have
always tended to emphasize the unboundedness of many phenomena,
the indeterminacy and the flux; and I share Whorf’s view that language
itself, once it has been constituted in this way, strongly influences the
forms that our perceptions take. Nevertheless there is a basic fit between
the discreteness of words and the discreteness of things; otherwise we
should not be able to talk about the things at all, or explain contrastively
those instances where the fit of words to things is less than perfect.
By contrast, the interpersonal kind of meaning is a motif that runs
throughout the clause; and this is represented by lexicogrammatical or
phonological motifs that are likewise strung unboundedly throughout.
The speaker’s attitudes and assessments; his judgements of validity and
probability; his choice of speech function, the mode of exchange in
dialogue – such things are not discrete elements that belong at some
particular juncture, but semantic features that inform continuous
stretches of discourse. It is natural that they should be realized not
segmentally but prosodically, by structures (if that term is still appropri-
ate) that are not particulate but field-like. The linguist’s tree is an
inappropriate construct for representing structures of this kind.
239
word–clause–text
in answer to the Cheshire Cat’s question What became of the baby?, she
begins with the Theme it (‘I’m going to tell you about something’) and
ends with the New a pig (‘here’s what is news to you’). In this case,
Alice has obligingly chosen as her Theme the thing that the Cat had
asked her about, namely the baby, realized by the anaphoric reference
item it; and she has kept the news, its change of state, till last. Alice is
being helpful, keeping the wave pattern of the dialogue in phase. But
she need not do this; the Theme is the speaker’s choice, and in any
case there is not always a ready-made candidate for thematic status.
Compare the following instance:
240
text semantics and clause grammar
The Footman, just like the rest of us, favours himself as an unmarked
Theme.
So the patterns of wording in the clause, which is the basic unit of
lexicogrammatical organization, display a variation that derives from
the different kinds of meaning they express; and the structural shape is
in each case a natural product of the semantic functions. A functional
grammar is an interpretation of the primary semiotic purposes that
language has evolved to serve, and of the different ways in which
meanings relating to these different purposes tend to be encoded (and
the patterns just described are only tendencies). When we go on to
observe the developmental processes whereby young children construct
their language, we gain a further insight into the steps by which
grammar may have evolved on the way towards its present form.
241
word–clause–text
242
text semantics and clause grammar
243
word–clause–text
244
text semantics and clause grammar
245
word–clause–text
246
text semantics and clause grammar
Appendices
Appendix 1
from Hasan (1980).
Text A
1. once upon a time there was a little girl
2. and she went out for a walk
3. and she saw a lovely little teddybear
4. and so she took it home
5. and when she got home she washed it
6. and when she took it to bed with her she cuddled it
7. and she fell straight to sleep
8. and when she got up and ( ) combed it with a little wirebrush
the teddybear opened his eyes
9. and ( ) started to speak to her
10. and she had the teddybear for many many weeks and years
11. and so when the teddybear got dirty she used to wash it
12. and every time she brushed it it used to say some new words
from a different country
13. and that’s how she used to know how to speak English Scottish
and all the rest
Text B
1. the sailor goes on the ship
2. and he’s coming home with a dog
3. and the dog wants the boy and the girl
4. and -----
they don’t know the bear’s in the chair
5. and the bear’s coming to go to sleep in it
6. and -----
they find the bear in the chair
7. they wake him up
-----
8. and (-----) chuck him out the room
9. and (-----) take it to the zoo
10. the sailor takes his hat off
247
word–clause–text
11. and the dog’s chased the bear out the room
12. and the boy will sit down in their
----- chair what the bear was
sleeping in
Underlined items are those which enter into lexico-referential chains.
Broken underlining indicates that one item incorporates more than one
token, for example -----
they referring to the girl, the boy, the sailor and
the dog. Empty underlining within parenthesis ( ), (-----) indicates a
token or tokens presupposed by ellipsis.
The number of lexico-referential tokens in the two texts is not very
different: 66 in Text A, 56 in Text B. But whereas 43 of those in Text
A (65 per cent) occur in chain interaction, the comparable figure for
Text B is only 20 (36 per cent). Text A thus displays considerably
greater cohesive harmony.
When subjects were asked to judge the coherence of the two texts,
Text A was consistently rated “more coherent” than Text B.
248
Chain Interaction – Text A
girl 왗 왘
went
home 왗 왘
girl got
home girl
girl 왗 왘
took 왗 왘
teddybear
girl had teddybear
girl took-to-bed
girl 왗 왘
fell-to-sleep
girl got-up
lovely 왗 왘
teddybear
dirty teddybear
speak
왗 왘
teddybear say
teddybear words
English
speak 왗 왘
Scottish
All-the-rest
Chain Interaction – Text B
bear chair
bear chair
왗 왘
go-to-sleep 왗 왘
bear chair
sleep bear chair
chuck 왗 왘
bear 왗 왘
room
chase bear room
go 왗 왘
sailor
come sailor
Each rectangle corresponds to one chain; subdivisions in the chain are indicated by boxing. Each box contains those
items which are in a constant functional relation (shown by a double-headed arrow) to items in some box in a
different chain; for example in Text A, between girl (4) and the box containing washed . . . brushed there is an actor–
action relation; between the latter and teddybear (4), a relation of action–goal.
text semantics and clause grammar
Appendix 2
from James R. Martin (forthcoming)
1. B: Lips are a must.
2. They’re in fashion.
3. So . . . what are you using in your skin care?
4. A: Oh I just – I don’t know.
5. Something my mom gets: Ponds or something.
6. B: Yes.
7. A: I don’t know.
8. B: Well really uh that’s not good enough really.
9. You want something that’s going to treat the skin.
10. You need to cleanse your skin well
11. uh to use a good toner
12. A: Hmm.
13. B: and moisturiser is a must
14. and of course then you can go into the make-up.
15. But if you do all these things
16. your skin will start to improve.
17. A: Yeah.
18. B: You’re finding a few little spots under your skin, aren’t you?
Talking Shop: scene 21 Halliday and Poole (1978).
251
word–clause–text
252
text semantics and clause grammar
Figure 2
253
word–clause–text
Appendix 3
from Peter Fries (1981).
1. A. The English Constitution – that indescribable entity – is a
living thing, growing with the growth of men, and assuming
ever-varying forms in accordance with the subtle and com-
plex laws of human character.
2. B. It is the child of wisdom and chance.
3. C. The wise men of 1688 moulded it into the shape we
know.
4. C. but the chance that George I could not speak English
gave it one of its essential peculiarities – the system of a
cabinet independent of the Crown and subordinate to
the Prime Minister.
5. C The wisdom of Lord Grey saved it from petrification and
set it upon the path of democracy.
6. C. Then chance intervened once more.
7. D. A female sovereign happened to marry an able and
pertinacious man,
8. D. and it seemed likely that an element which had been
quiescent within it for years – the element of irre-
sponsible administrative power – was about to
become its predominant characteristic and change
completely the direction of its growth.
9. C. But what chance gave chance took away.
10. D. The Consort perished in his prime,
11. D. and the English Constitution, dropping the dead limb
with hardly a tremor, continued its mysteric life as if
he had never been.
Queen Victoria, Lytton Strachey (p. 192)
254
text semantics and clause grammar
Appendix 4
From Robin Melrose (1979).
The remaining eleven sections deal with every aspect of life, regulating
it at every stage and aspect, ordering everything, forcing everything into
a symmetrical pattern: the cities are uniform, married life is strictly
controlled, education is minutely prescribed. Philosophy is confined
within rigid limits, the fine arts somewhat less so (. . .) This planned
paradise is enforced by drastic penal laws. Machinery of government is
paternalistic and pyramidal. It is based on division into families, tribes,
cities and provinces, and, in the case of the different crafts and profes-
sions, on units of ten. To each unit of work is assigned its “master”
(. . .) Each paterfamilias over fifty is a senator, each family in turn
provides a tribal chief, each town in turn a city chief. Subordinate
senates of cities are controlled by the Supreme Senate. At the head of
the state is the General.
Totalitarianism, Leonard Schapiro (pp. 87–8)
255
word–clause–text
Theme:
The remaining eleven sections ; the cities ; married life ; education ;
Philosophy ; This planned paradise ; Machinery of government ; It ;
To each unit of work ; Each paterfamilias over fifty ; each family ;
Subordinate senates of cities ; At the head of state
Lexical Cohesion:
Group A :
regulating ; ordering ; forcing ; uniform ; controlled ; prescribed ;
confined ; planned ; enforced ; penal laws ; paternalistic and
pyramidal
This is a particularly clear example of an attributive message. It
begins with a summation, “every aspect of life”, with the general noun
“aspect” acting as Head of a nominal group. It is this summation which
determines the Theme of the clauses that follow: thematic prominence
is assigned precisely to aspects of life, so that there is a relationship of
superordinate to hyponym between summation and Theme in the
message, reinforced by Theme in the last five clauses, which is in a
relationship of hyponym to “machinery of government”, itself an aspect
of life.
There is no explicit conjunction of interest : the chief conjunctive
relationship is an implicit one, of the internal additive type. More
worthy of attention is the lexical string of Group A. Just as Theme was
determined by the summation, so the lexical string of Group A is
determined by the non-finite clauses dependent on the clause of which
the summation is an element. Together with the three verbs in these
non-finite clauses, nine lexical items of the message proper constitute a
string of synonyms, near-synonyms, and collocates – and of these nine,
six function as Complement, and so form a kind of pattern in the
Rheme. Thus it may be seen that in this attributive message the
summation and the clause complex of which it is an element are both
closely related to the message proper that follows: the summation is
hyperonymously linked to the Theme, and the clause complex (or,
more precisely, the non-finite clause) is synonymously connected –
with one exception – to the Rheme. Or, to put it another way, it is
most often the case that “aspects of life” are encoded in the Subject,
while “regulation” is realised in the Complement.
256
text semantics and clause grammar
Appendix 5
from M. A. K. Halliday (1982).
Mrs. Birling: I think we’ve just about come to an end of this
wretched business.
Gerald: I don’t think so. Excuse me.
(He goes out. They watch him go in silence. We hear
the front door slam.)
Sheila: (to Inspector) You know, you never showed him that
photograph of her.
Inspector: No. It wasn’t necessary. And I thought it better not
to.
Mrs. Birling: You have a photograph of this girl?
Inspector: Yes. I think you’d better look at it.
Mrs. Birling: I don’t see any particular reason why I should –
Inspector: Probably not. But you’d better look at it.
Mrs. Birling: Very well.
(He produces the photograph and she looks hard at
it.)
Inspector: (taking back the photograph) You recognize her?
257
word–clause–text
258
text semantics and clause grammar
259
indicative imperative
(MODALITY) (MODULATION)
‘it is’ ‘do!’
probability temporality obligation inclination
modal ‘it either is ‘it both is ‘you do!’ ‘me do!’
or isn’t’ and isn’t’
categorical positive is do
high: must ought to need has/had to subjective, congruent: must &c. (modal auxiliary verbs)
median: will would shall should subjective, explicit: I think &c. (mental process clauses)
low: can could may might objective: certain(ly) &c. (modal adverbs, etc.)
Chapter Ten
First published in The Handbook of Discourse Analysis, Vol. 2: Dimensions of Discourse, 1985.
London: Academic Press, pp. 29–56.
261
word–clause–text
262
dimensions of discourse analysis
Figure 1
263
word–clause–text
Clause complex 1 ⳯b ^ a ( 1 ^ +2 )
1.1 ⳯b ||| when you have a small baby in the house ||
1.2 1 || do you call it ||
1.3 +2 || or do you call it she or he ||
Figure 3a
264
dimensions of discourse analysis
Clause complex 2 a ( ⳯b ( a ^ ‘b ) ^ a) ^ ⳯b ( 1 (a ^ ‘b ) ^ +2 ( ⳯b
( 1 ^ +2 ( 1 ^ ⳯ 2 (a ^ ‘b ) ) ) ^ a ( a ^ ⳯b ) ) )
2.1 a ⳯b a ||| well if you don’t know ||
2.2 a b ‘b || what it is ||
2.3 a a || I think you ought to call it it ||
2.4 ⳯b 1 a || because you don’t know ||
2.5 b 1 ‘b || whether you’re calling it a boy or a girl ||
2.6 b +2 ⳯b 1 || and if it gets on ||
2.7 b 2 b +2 1 || and if you start calling it she ||
2.8 b 2 b 2 ⳯2 a || then you find out ||
2.9 b 2 b 2 2 ‘b || that it’s a boy ||
2.10 b 2 a a || you can’t stop yourself ||
2.11 b 2 a ⳯b || ’cause you’ve got so used to calling it she |||
Clause complex 3 a ^ ‘‘b ( ⳯b ( 1 ^ + ( a ^ ‘b ) ) ^ a ( ⳯b ^ a ) )
3.1 a ||| Mrs. Siddons says ||
3.2 ‘‘b ⳯b 1 || that if some neighbour has a new baby next door ||
3.3 b b +2 a || || and you don’t know ||
3.4 b b 2 1b || whether it’s a he or a she ||
3.5 b a ⳯b || if you refer to it as it ||
3.6 b a a || well then the neighbour will be very offended |||
Clause complex 4 1 ( ⳯b ^ a ) ^ ⳯2 ( a ( a 具具 ⳯b ) 典典 ^ 1b )
4.1 1 ⳯b ||| well if it’s in your family ||
4.2 1 a || I think you should call it either he or she ||
4.3 ⳯ 2 a a || or else the poor thing 具具 典典 won’t know ||
4.4 2 ⳯b 具具 when it grows up 典典
4.5 2 a ‘b || what it is |||
Clause complex 5 ‘‘b a 具具 ) a ( 典典 ^ ⳯b (1 ^ +2 ( a ^ 1b ) ) )
5.1 ‘‘b a ||| well what 具具 典典 you should do ||
5.2 a 具具 did Mrs. Siddons suggest 典典
5.3 b ⳯b 1 || if your neighbour has a baby ||
5.4 b b +2 a || and you don’t know ||
5.5 b b 2 ‘b || whether it’s a boy or a girl |||
Clause complex 6
6.1 ||| she didn’t |||
Clause complex 7
7.1 ||| I don’t suppose she knew |||
Clause complex 8
8.1 ||| call it the |||
Clause complex 9
9.1 ||| hello the |||
Clause complex 10
10.1 ||| oh I know |||
Clause complex 11
11.1 ||| call it baby |||
Figure 3b Clauses and Clause Complexes
265
word–clause–text
The major portion of the text consists of five turns, each made up
grammatically of one clause complex. These contain, respectively, 3,
11, 6, 5, and 5 clauses. They show a preference for hypotactic (17)
over paratactic (7) interdependencies; the predominant logical-semantic
relation is that of enhancement, typically hypothetical (11 instances) –
the discussion centers around what to do if and when a certain situation
arises. Of the 15 other instances, 5 are extension, “and / or”; the
remainder (8) are projection, of which 2 are saying (the teacher as
Sayer) and 6 are knowing, mainly negative and with a generalized
Senser “(if) you don’t know” – this being an aspect of the problem
under discussion. (The two instances of I think [new addition] are
metaphorical modalities, not projections, as can be shown by adding a
tag: the tagged form of I think you should call it he or she is shouldn’t you?
not don’t I?)
All these clause complexes are not only complex but also impeccably
well formed, as is typical of casual spontaneous speech (including that
of children!).
So much for the “reasoning” component of the discussion. The
remainder consists of “suggesting,” partly humorous and partly serious,
and here the turns are short, one or two clauses each. The clause
complexes are even shorter, since each consists of just one clause.
There is no parataxis or hypotaxis. (6.1 and 7.1, She didn’t. I don’t
suppose she knew [new addition] could be considered to form a paratactic
elaboration, given the tone concord; but the latter, though necessary,
is probably not a sufficient condition.)
As far as the organization of the information is concerned, a
comparison of the two transcriptions shows that in the majority of
instances one clause is one information unit, this being the unmarked
(default) situation in English. This holds throughout, with the following
systematic exceptions: (1) seven out of the eight projections have both
projecting and projected clause on one tone group, for example, // if
you / don’t know / what it / is // – this is the locally unmarked form;
and (2) one clause consists of two information units, one for the Theme
and one for the Rheme: // if some / neighbour has a // new / baby next
/ door // – this is the predominant pattern when two information units
are mapped on to one clause. For the analysis of theme and information
structure see the next two sections.
266
dimensions of discourse analysis
3 Theme
Each clause, and each clause complex characterized by rising depend-
ency (b ^ a), is analysed for thematic structure:
Clause Theme
Markedness of
Clause Textual Interpersonal Topical topical theme
267
word–clause–text
268
dimensions of discourse analysis
4 Information structure
Each information unit is analysed for information structure (numbers
refer to clause complexes, letters to information units):
Figure 5
269
word–clause–text
In the “New” column, the focal element is outside the brackets. Square
brackets enclose other New material; within this, the items shown
within curved brackets are those previously mentioned.
The New is what the speaker presents as being for the listener to
attend to: “this is what’s news”. It may be previously unknown, or
contrary to expectation, or picked out for special prominence. It is
realized by means of the tonic accent.
Phonologically, spoken discourse in English consists of a linear
succession of tone groups, each characterized by one intonation con-
tour or tone. The tone group, in turn, consists of a tonic segment that
carries the characteristic tone contour: 1, falling; 2, rising; 3, level
(phonetically realized as low rising); 4, falling–rising; 5, rising–falling;
13, falling plus level; 53, rising–falling plus level. The tonic segment
begins with the tonic accent, which embodies the distinctive pitch
movement. Optionally, the tonic segment may be preceded by a
pretonic segment that also forms part of the same tone contour. Both
tonic and pretonic segments display a range of more delicate options
within each tone: wide fall, narrow fall, low pretonic, high pretonic,
and so on.
Grammatically, the tone group realizes a unit of information, which
is one piece of news, so to speak. It consists of a New component
optionally accompanied by a component that is Given. Typically, the
New comes at the end; but unlike thematic structure, information
structure is not realized by the order in which things are arranged, but
by tonic prominence – the New is the element containing the tonic
accent. The particular word on which the tonic accent falls is said to
carry the information focus. Anything after the focal element is thereby
marked as Given, while anything preceding it may be Given or may
be New (there are rather subtle intonational and rhythmic variations
serving as signals).
An information unit is not necessarily coextensive with a clause, but
that is its unmarked status: each ranking clause (i.e. independent or
dependent, but not embedded) is typically one information unit. The
principal systematic variants are (1) two clauses / one information unit:
failing dependencies, that is, a ^ b sequences; (2) one clause / two
information units: thematic focus, that is, // Theme // Rheme //
information patterns.
Analysis of the New elements will reveal the “main point” of the
text. In the example it is to do with babies, what sex they are, and
how they are to be referred to in cases of doubt. Subsidiary points of
270
dimensions of discourse analysis
attention are the baby’s growing up, the children’s understanding and
obligations, and the adult world’s possible displeasure.
The Theme in a clause is what is prominent for the speaker; it is
“what I am on about”. The New in an information unit is what is
made prominent (by the speaker) to the listener; it is “what you are
being invited to attend to”. When clause and information unit are
mapped on to each other, the result is a wavelike movement from
speaker to listener, the diminuendo of the speaker’s part being as it
were picked up by the crescendo of the listener’s part.
The effect of this movement is cumulative over the text as a whole.
The present text is typical in the way that the sequence of Themes
represents the “method of development” of the dialogue, while the
sequence of News represents the “main point” of the discussion, with
each speaker contributing her part to the construction of the overall
pattern – all unconsciously, of course.
This interplay of two distinct waves of prominence is possible
because Theme–Rheme and Given–New are not (as often conceived)
one single structure, but two distinct structures that interpenetrate. As
a result, they can vary independently, allowing for all possible combina-
tions of the two kinds of texture. In unit 2h, for example, Elsie might
have chosen a different distribution by combining thematic and infor-
mation prominence (mapping New onto Theme):
’cause you’ve got //.1. so / used to / calling it / she //
This would have strongly highlighted used to, as a marked focus, and
marked calling it she as Given; the effect of the latter would have been
to bring out the repetitive facet of calling it she, thus reinforcing its
cohesion with 2e, but by the same token to deprive it of its status as a
main point for attention. The interaction of the thematic and informa-
tional systems is the clause grammar’s contribution to the creation of
texture in discourse.
5 Mood
Each finite clause is analysed for mood; its Subject and Finite element
are shown, together with any modality:
271
word–clause–text
Figure 6
272
dimensions of discourse analysis
Figure 7
273
word–clause–text
274
dimensions of discourse analysis
Figure 8
275
word–clause–text
The remaining six clauses include three characterizing the baby (two
material: grow, get on; one circumstantial: be in + family), one character-
izing the neighbour (affective: be offended), and two others, one minor
(greeting: hello) and one a WH- process (do what).
There are three major types of process in English: Type I, doing
(material and behavioural); Type II, sensing / saying (mental and
verbal); Type III, being (relational and existential). They are distin-
guished in the grammar in various ways; the principal distinctions are
as follows:
Figure 9
276
dimensions of discourse analysis
In the example text, the clauses in (i–iii) above are all type III,
relational; the issue is one of being, partly attribution (having a baby,
which in this text means possessing it, not bearing it; being male or
female) and partly identification (being the name of). Of these, (i) and
(ii) are middle – there is a Medium (the neighbour, the baby) but no
Agent; (iii) however is effective – there is an Agent in the identification
process, always represented as you, but moving from personal “you” in
clause complex 1 to impersonal “you” thereafter. The clauses in (iv)
and (v) are type II, mental-verbal; (iv) are cognitive, with the Medium
(Senser) being you, I, Mrs. Siddons or the baby when it grows up; (v)
are verbal, the Medium (Sayer) being Mrs. Siddons. All interactants,
real and hypothetical, are involved in thinking about the problem –
including the baby, at some future date, if a solution is not reached
now; and the teacher has put the problem into words.
A summary of process types and the relevant participant functions
follows:
Figure 10
277
word–clause–text
Figure 11
278
dimensions of discourse analysis
Figure 12
279
word–clause–text
Figure 13
280
dimensions of discourse analysis
Figure 14
281
word–clause–text
9 Grammatical metaphor
Grammatical metaphors are identified, reworded, and reanalysed.
Most examples of adult English contain some instances of grammat-
ical metaphor: clauses in which one type of process is represented in
the grammar of another; for example, the fifth day saw them at the summit
“on the fifth day they arrived at the summit”, or guarantee limited to
refund of purchase price of goods “we guarantee only to refund the price
for which the goods were purchased”.
Children’s speech is largely free of grammatical metaphors of this
kind; this is in fact the main distinction between child and adult
language. There are no examples of it in the present text.
There are however certain grammatical metaphors that have been
built into the language, so that the metaphorical version has become
the norm; for example, she gave a nod “she nodded”, he has a long nose
282
dimensions of discourse analysis
“his nose is long”. One very common type of these is the use of a
mental process (cognitive) clause to express a modality, such as I think,
I don’t believe. It was pointed out above that the tagged form of I think
it’s broken is I think it’s broken, isn’t it?, not I think it’s broken, don’t I?,
showing that this is a metaphor for it’s probably broken. We can use one
of these as an example:
|| I don’t suppose || she came ||| a ˆ ´ b
mental: cognitive material
negative positive
reworded as:
||| she probably didn’t come ||| (single clause)
material
modalized:probability/median
negative
In cases like this it saves time if the analysis moves directly to the
nonmetaphorical version, since the rewording is quite automatic. In
other instances, however, the principle is as follows:
1. Analyse the clause as it stands.
2. Reword it, in nonmetaphorical form.
3. Analyse the reworded version.
Both analyses figure in the interpretation. Sometimes it takes several
steps in rewording to reach a nonmetaphorical version, and there may
be more than one possible route; all are potentially relevant to an
understanding of the text.
10 Context of situation
The context of situation of the text is described in terms of field, tenor,
and mode. The “field” is what is going on: the nature of the social-
semiotic activity. The “tenor” is who are taking part: the statuses and
mutual roles of the interactants. The “mode” is what part the language
is playing: the rhetorical and communicative channels. This description
is then used to interpret the lexicogrammatical features of the text.
Field. A general, imaginary problem of verbal behaviour: how to
refer to a baby whose sex is unknown, without offending
against the parents, the baby (later in life), or the language.
283
word–clause–text
Tenor. Adult and three children: adult (neither teacher nor parent)
interviewer, informal; children self-conscious but relaxed.
Speech roles: adult questioning, children suggesting.
Mode. Informal spontaneous speech. Dialogue: question-and-
answer exchanges. Exploratory; hypothetical and logical in
orientation, moving towards (partly humorous) resolution.
These features determine the choice of register: that is, the kinds of
meanings that are likely to be exchanged. Like the rest of the linguistic
system, the patterns are probabilistic: given these features of the con-
text of situation, we can make semantic (and therefore lexicogram-
matical) predictions with a significant probability of being right – that,
after all, is precisely what the interactants themselves are doing all the
time.
What makes this possible is what makes it possible for a child to
Figure 15
284
dimensions of discourse analysis
learn the language in the first place: the systematic relationship between
these categories of the situation and the metafunctions of the content
system. By and large, characteristics of the field predict experiential
meanings, those of the tenor predict interpersonal meanings, and those
of the mode predict textual meanings.
In analysing a text, we identify those features of the lexicogrammar
which in a text-generation program might reasonably be expected to
be called upon if the situation was represented in this way.
11 Conclusion
Three final points should be made about an outline of this type. One
is that it is just an outline: Obviously, the analysis under every heading
could be developed much further in delicacy, and other headings could
be added. The guiding principle is to select and develop whatever is
needed for the particular purpose in hand. There are many different
purposes for analysing a text, and the scope and direction of the analysis
will vary accordingly. Often we may want to scrutinize only one or
two features, but to follow them through to a considerable depth.
Secondly, a text analysis is a work of interpretation. There are
relatively few absolute and clearcut categories in language; there are
many tendencies, continuities, and overlaps. Many actual instances can
be analysed in two or more different ways, none of which can be ruled
out as impossible; some may be less sensible than others, and so can be
discarded, but we may still be left with valid alternatives. Especially in
a literary text it is to be expected that we will find clauses with multiple
grammars; but this is not a distortion of the system – it is a richer use
of its natural resources. All analyses may need to figure in the
interpretation.
Thirdly, the lexicogrammatical analysis is only a part of the task. It is
an essential part; all text is made of language, and the central processing
unit of the linguistic system is the lexicogrammar. But just as the
grammatical system does not itself create text – text is a semantic
creation, with the grammar functioning largely (though not entirely) as
the automatic realization of the semantic choices – so the analysis of
the grammar does not constitute the interpretation of a text. (There
has been some misunderstanding on this point, for example in the use
of cohesion as a method of text analysis. Cohesion is an essential
property of texts, but it is the way the cohesive resources are deployed
that makes the difference between text and non-text, and between one
text and another.)
285
word–clause–text
286
SECTION THREE
CONSTRUING AND ENACTING
This page intentionally left blank
EDITOR’S INTRODUCTION
289
construing and enacting
one is doing; so when one talks about something, one tends to say that
it happened or was done.” By means of grammatical metaphor, written
language symbolically distances “the act of meaning and its counterpart
in the real world”. Each plays complementary roles when it comes to
using language to acquire knowledge and reflect on one’s experience.
Halliday maintains that this complementarity must figure into any
attempt at developing a linguistic theory of learning.
In ‘How do you mean?’ (Chapter 13), appearing in Advances in
Systemic Linguistics: Recent Theory and Practice (1992), meaning is taken
as a mode of action occurring at the intersection of the conscious and
material modes of experience. In particular, Halliday examines the
evolution of protolanguage into language, or how the two-dimensional
semiotic that defines the mammalian experience evolves into ‘a semiotic
of a new kind: a stratified, tri-stratal system in which meaning is “twice-
cooked”, thus incorporating a stratum of “pure” content form. The
two dimensions of protolanguage, a minimal semiotic system, include
“the ‘inner’ dimension of reflective / active, ‘I think’ as against ‘I want’,
and the ‘outer’ dimension of intersubjective / objective, ‘you and me’
as against ‘he, she, it’ ”. The third dimension results from the introduc-
tion of grammar, or as Halliday describes it: ‘a purely symbolic mode
of being between these two interfaces’. The processes of instantiation
and realization make possible this dynamic open system we call language.
In the two final works in this section, ‘Grammar and daily life’,
which first appeared in Functional Approaches to Language, Culture and
Cognition (2000), a Festschrift for Sydney Lamb, and ‘Grammar and
grammatics’, published as Volume 121 of the series Current Issues in
Linguistic Theory (1996), Halliday describes how grammar enables us,
unconsciously, to construe our reality, and interpret our experience,
while grammatics makes it possible for us to reflect consciously on how
this theory of our human experience works. Halliday introduced the
term “grammatics”, in contradistinction to grammar, to distinguish
between a particular stratum of a natural language and the study of this
stratum. “Thinking grammatically”, or “using grammatics to think
about what grammar thinks about the world”, may help us better
understand this ‘grammatical energy’ or ‘grammatical logic’ that powers
language and also conditions our attitudes to each other and to the
world around us. “To be a linguist”, Halliday writes, “is inevitably to
be concerned with the human condition”, and those who ‘think
grammatically’ will be better prepared not only to address issues of
social injustice and inequality, but also to contribute to the develop-
ment of new applications of linguistics such as intelligent computing.
290
Chapter Eleven
First published in The Tenth LACUS Forum, 1984, edited by Alan Manning, Pierre Martin
and Kim McCalla, pp. 3–18. Amsterdam: John Benjamins.
291
construing and enacting
292
on the ineffability of grammatical categories
293
construing and enacting
294
on the ineffability of grammatical categories
semantic. The same is true of the pair ‘sentence’ and ‘discourse’. But
with the majority of grammatical terms – those for functions, systems
and terms in systems (features) – this does not hold. These too have to
be imported into grammar from outside; but in this case they come
not from folk semantics but from outside language altogether. When
categories of this type come to be named, the terms that are introduced
for the purpose interpret the categories by reference to some aspect of
extralinguistic experience.
A typical example would be a complex of categories such as ‘(system)
number: (features) singular / plural’. Consider a label such as ‘plural’.
This derives from the ideational meaning of the category: it expresses a
relation between that category and the speaker’s experience of the
world. The term ‘plural’ is the name of this relationship. But the same
term is also used as the name of the grammatical category which realizes
this relationship; a noun will be said to be ‘plural (in number)’. And
this can cause problems.
Typically in the history of western linguistics the reasoning has
proceeded as follows. In the morphology we are presented with a
certain category, let us say a two-term system, of the form ‘a : x/y’;
thus, ‘all a are either x or y’. By inspecting, or perhaps introspecting,
typical contexts in which these forms are used we recognize a redun-
dancy, such that x redounds with one of a set of countable things and
y redounds with two or more. The abstract labels ‘a : x/y’ are then
replaced by the substantive labels ‘number : singular / plural’. This is
of course an idealized model of the process; such abstract labels have
never been used, as far as I know, at least until modern times. But it
helps to bring the issue into focus.
What happens next? The categories that have been labelled in this
way then come to be reified and the question is asked what they mean.
The answer given is: ‘singular means one of a thing, plural means more
than one’. In fact, these are definitions of the words singular and plural;
but they are made to serve as definitions of the metawords, the terms
of the metalanguage. Next, the terms are evaluated for their predictive
power: will they correctly predict text from situation, or situation from
text? Given a plural form, will it refer to more than one of a thing?
Given more than one of a thing, will it be referred to by a plural form?
The answer this time is: yes, with a certain degree of probability –
high enough for many purposes, but inadequate for some, and disturb-
ing for those who like their categories pure. This then gives rise to a
theory of ‘core meanings’: a term of the metalanguage is said to
represent the ‘core meaning’ of the category. With this defence, in
295
construing and enacting
spite of fruit and furniture, scissors and oats (or their counterparts in the
language in question), ‘singular’ and ‘plural’ can continue to be used.
This problem – that of interpreting a symptom and then labelling its
interpretation – is common to all sciences. It arises in any realm of
discourse involving explanation and abstraction. Somehow, a metalan-
guage has to be created, and created out of natural language, in order
to assign a Value to a Token. I go to my doctor with a swelling
somewhere on my body. He looks at it, and pronounces ‘You’ve got
an oedema.’ What is oedema? – Greek for ‘swelling’. But what he is
diagnosing is a more abstract swelling – it is that which is manifested
by the swelling on my body. An oedema is the Value of which a
swelling is the Token. The use of a different term in this way allows
for stratal diversification: not all swellings realize oedemata, and not all
oedemata are realized as swellings. The relationship is a probabilistic
one – and hence invites further, more delicate investigation.
Where do such terms come from? Ideally, they come from a parallel
but distinct semiotic. It should be a natural language, in order to
maintain the non-arbitrariness of the relation between the symptom
and the underlying condition – given swelling and oedema we would
predict that, in default of any special circumstances, they will refer to
the same thing. But it should be a different language, or at least a
different sub-language, in order to allow for instances where they do
not. And it should be a higher status code, in order to symbolize a
higher order of abstraction (and also the social value of abstract
knowledge). The ideal source of a metalanguage is thus the ‘high’
variant in a diglossia. A word of a natural language that is at one
remove from primary reality, such as ancient Greek, or classical
Chinese, or Sanskrit, is appropriate for symbolizing a phenomenon that
is at one remove from primary observation.
But when it comes to metalinguistic matters, linguistics presents a
special case. It is not just another science. It is ‘language turned back
on itself’, to use Firth’s (very British) expression; or, in Weinreich’s
(very American) formulation, ‘language as its own metalanguage’. As a
consequence, where other sciences need two terms, we need three:
one for the phenomenon, and two for the metaphenomenon, one
grammatical and the other semantic. To return to the example of
number: we need to be able to say that the grammatical category of
‘plural’ typically expresses (realizes) the semantic category of, say,
‘manifold’, which typically expresses (redounds with) more than one of
a thing.
But notice what has happened. The grammatical category of ‘plural’
296
on the ineffability of grammatical categories
297
construing and enacting
298
on the ineffability of grammatical categories
It is only a very short step from here to the assertion that the Subject
‘has no meaning’. The implication is: whatever it is that is functioning
as Subject in any instance has meaning as actor, or has meaning as
topic; but as Subject it has none – the category of Subject has no
meaning in itself. In this view, Subject is a grammatical function whose
only function is to be a grammatical function.
Such a view is enshrined in the terminology, in the term ‘grammat-
ical Subject’ (used for example in Sweet 1891); this is in contradistinc-
tion to ‘logical Subject’ (i.e. Actor) and ‘psychological Subject’ (i.e.
Theme). Compare the later Prague school interpretation, with ‘syntac-
tic structure’ (Subject–Predicate) contrasted with ‘semantic structure’
299
construing and enacting
300
on the ineffability of grammatical categories
One project in which they play an essential part is the Penman project
at the Information Sciences Institute (Mann and Matthiessen 1983;
Mann 1983). In Penman the motive power is provided by the
grammar, Nigel, which is a systemic grammar consisting of a network
of some hundreds of options. At each choice point a Chooser is
activated; the Chooser consults the environment (the Knowledge Base)
for instructions on which way to go. The Chooser’s questions are
referred to an Inquiry Operator; and they take forms such as the
following:
Is this concept inherently multiple, i.e. a set or collection of things,
or is it unitary?
Is the process one which conceptually has some sort of entity which
causes the process to occur?
Does this represent a concept which the speaker expects the hearer
to find novel, not previously mentioned or evoked, and thus does
not expect the hearer to identify uniquely by reason of the attention
which it currently holds, its inherent uniqueness in culture, or its
association with an identifiable entity?
All such glosses are attempts to get at the grammar beneath the skin;
and they may be supported by a variety of different beliefs. First, it
may be assumed that all grammatical generalizations have some signific-
ance at a higher stratum; or alternatively, that some are simply house-
keeping devices and have no semantic function. Secondly, those
grammatical categories that are regarded as semantically significant may
be thought of as universal, or as particular to the given language, or as
particular to a given register, a functional variety of a language. (These
would represent fairly well the respective views of Jakobson, Hjelmslev
and Firth.) Thirdly, it may be held that every such category has one
meaning that is common to all its manifestations, and the problem is to
find the right semantic generalization to cover all cases; or alternatively,
that some categories at least are polysemous, so that their meaning
varies in ways that are not predictable from the context (cf. Ikegami
1980: 59). Fourthly, there is a range of beliefs about the place of
grammar, and the need to postulate some higher level semiotic system
(‘semantics’, ‘semology’, ‘the conceptual level’, etc.) to which gram-
matical categories can be related in a systematic and in some sense
‘natural’ way. Positions taken on these issues may complicate the task
of semantic interpretation: for example, if categories are assumed to be
universal, and yet are established at an insufficiently abstract level.
But whatever beliefs are held about them, grammatical categories
301
construing and enacting
will remain ineffable. Some of the more recalcitrant ones are categories
that Whorf originally called ‘covert’: ‘having no mark other than
distinctive reactances with overtly marked categories’. But by no means
all are of this kind – there is nothing covert about definiteness in
English, for example. The phenomenon we are concerned with has
more to do with Whorf’s follow-up notion of a ‘cryptotype’.
Whorf remarks of these that ‘they easily escape notice and may be
hard to define, and yet may have profound influence on linguistic
behavior’ (1956: 92). Among cryptotypes in English Whorf cites
gender, transitivity (of the verb), inherence (of the adjective), and
various more delicate categories, such as that of verbs that may be
phrasalized with up. There is, of course, a connection between the two
senses in which Whorf is using ‘crypto-’: a category may be hard to
define precisely because it is hidden from view. But hidden from whose
view? It is not because they are hidden from the linguist that grammat-
ical categories are hard to define; once the linguist has found them, the
fact that they had escaped his notice ceases to matter. The significance
of this concept of a cryptotype is that it is something that escapes the
notice of the speakers of the language.
Franz Boas long ago drew attention to the unconscious nature of
language, contrasting it in this respect with the other meaning systems
of a culture; and although his observations have often been quoted, it
seems to me that their significance is seldom fully taken into account.
There is a fundamental relationship between the unconsciousness of
language and the nature of its semantic categories. I have often pointed
out, in the many years since I began the study of informal speech, that
it is only in the most spontaneous, un-self-monitored kinds of discourse
that a speaker stretches his semantic resources to the utmost (cf.
Halliday 1966). This does not happen in formal speech; and it certainly
does not happen in writing. It is in unconscious spoken language that
we typically find the truly complex sentences, with their labyrinths of
hypotaxis and all their projections and expansions, from which, while
we blunder through such sequences often losing ourselves completely
when we are engaged in the planned self-monitoring discourse of an
academic lecture, we emerge in good order and with every node
unravelled provided we are completely unaware of what we are saying
and attending only to whatever it is we are involved in at the time.
(That sentence is best taken orally, at high speed.) Our ability to use
language depends critically on our not being conscious of doing so –
which is the truth that every language learner has to discover, and the
contradiction from which every language teacher has to escape.
302
on the ineffability of grammatical categories
303
construing and enacting
child. Semogenesis begins well before the mother tongue, as the infant
creates his own protolanguage or ‘child tongue’: he constructs this
symbolic system, in interaction with those who share in his meanings,
for the twofold purpose of doing and thinking that characterizes all
such systems; and in the same process he also constructs the objects of
his action and his reflection. But the child tongue has its limitations,
for both these purposes; so he moves into the adult mode, and takes
over the mother tongue with its ready-made grammatical categories.
The symbols of the mother tongue, which have been around him from
the start, now become his reality, at once a part of, and a key to, the
complex phenomena of his experience. Language and culture are
construed as one.
Does a child, then, know what a Subject is? We cannot ask him;
nor can we set up a test situation to find out – if only because children,
given an unnatural task, will respond with unnatural behavior. (It is
not intended to suggest that there is any contrast here with adults. The
problem lies again in the unconscious nature of linguistic processes,
which adults cannot reproduce experimentally either.) Nevertheless it
is clear, surely, that a child does know what a Subject is, because he
uses one a hundred times a day. We only have to listen to a five-year-
old in ordinary, real life situations, and we will hear the categories of
the grammar that we find most difficult to explain, deployed in their
appropriate semantic roles.
What we observe, of course, even with a tape recorder on perman-
ent duty, is only a limited set of instances. We have to infer the
system that lies behind them; for language (if I may be allowed to
invert Chomsky’s famous dictum) is an infinite system that generates
only a finite body of text. But what we can observe is already very
convincing.
If I assert that a five-year-old knows what a Subject is, it is because
I have listened to children for many years, and heard them talking in
clauses which have Subjects. In my own detailed record of one
particular child, there are about 2,500 of them; but since child language
studies became fashionable there has been an abundance of such
material available, if one does not feel one can rely just on one’s
listening. Now, any one of these clauses could have had the appropriate
Subject by chance. Moreover, since no linguistic category is chosen in
isolation – in choosing the Subject one is always making other choices
besides (and this will apply whatever category is used as illustration) –
in any one instance we could always claim that the appropriateness of
the Subject was a consequence of some other choice. But if countless
304
on the ineffability of grammatical categories
305
construing and enacting
306
on the ineffability of grammatical categories
307
construing and enacting
The speaker keeps the listener’s attention by varying the Theme and
the Actor:
Theme Actor
1 about me (‘I . . .’) what I did (‘did you?’)
2 about me (‘I . . .’) what happened to me (‘were you?’)
3 about the ball (‘the 3rd’) what I did (‘did you?’)
4 about the ball (‘by the 4th’) what happened to me (‘were you?’)
Clause 1 has the speaker in all three functions of prominence: interper-
sonal (I as Subject), ideational (I as Actor) and textual (I as Theme).
Clause 2 is marked by ideational modesty (this is what happened to
me, not what I did), and clause 3 is marked by textual modesty (now
I’ll tell you about the ball, not about myself). In clause 4, the speaker
gains further merit by ending on a doubly modest note, in which he is
neither the Actor nor the Theme.
But in regard to the speech function, the picture is quite different.
The speaker retains himself in the role of Subject throughout. There is
no sign here of interpersonal modesty; the assertion is made to rest on
I every time, and the listener’s response, correspondingly, must always
have a you in it – Did you?, Were you? In other words, every step in
308
on the ineffability of grammatical categories
309
construing and enacting
features and functions; and since grammar is the central processing unit,
where meanings of different kinds are brought together as wordings,
we expect its categories to be valid for the language as a whole. All
uses of English involve ‘Mood : Subject + Finite’, or ‘tense : past /
present / future’; and these are assumed to be in some sense ‘the same
thing’ in all contexts, since otherwise we would not be looking for
definitions of them.
No such constraint figures in our conception of semantics. The
grammar is the grammar: it has internal organization, of a metafunc-
tional kind; it has some special purpose sub-grammars; and it has
considerable indeterminacy – but there is such a thing as ‘the grammar
of English’. We do not operate with a separate grammar for each
register. No doubt we can also conceive of such a thing as ‘the
semantics of English’; but we also feel that (at least at the present state
of our knowledge) it is not counter-productive to envisage a more
restricted domain for semantic generalizations, and to operate with
semantic sub-systems each relevant to a specific universe of discourse.
In principle, the domain of a semantic description may be anything
from ‘the whole language’ down to a single text. At one end of the
scale, I have found it useful to set up a semantic system relating to just
one dialogue of 35 words long; this was a child–adult dialogue, and the
purpose was to explore what meaning potential the child must have in
order to be able to construe such a discourse (see Appendix, p. 313).
Geoffrey Turner’s (1973) semantic networks define a rather broader
range of texts, such as mother–child control patterns in specific exper-
imental situation types. More general again is Ruqaiya Hasan’s (1983)
‘message function’ network, which describes spontaneous interaction
between children and parents, for the purpose of investigating the
development of children’s learning patterns. At the other end of the
scale, J. R. Martin’s (1983a, 1992) conjunction networks are like
grammatical networks in that they are set up for the language as a
whole.
When we describe semantic systems, we are saying what it is that
‘preselects’ the grammatical categories: what choices in meaning call on
what features in the grammar for their realization. It is by this process
that the grammatical categories are defined; when this is done, there is
no need to gloss them further. Once the semantic system is made
explicit, it can only be misleading to attach separate semantic descrip-
tions as glosses to the categories of the grammar.
At the same time, if the semantic system is set up only for a restricted
domain, some particular register variety, then the meanings of any
310
on the ineffability of grammatical categories
311
construing and enacting
312
on the ineffability of grammatical categories
Appendix
Nigel at 5;4 (from Halliday (1984))
Nigel: Shall I tell you why the North Star stays still?
Father: Yes, do.
Nigel: Because that’s where the magnet is, and it gets attracted by the
earth; but the other stars don’t so they move around.
//2 shall I / tell you / why the / North / Star / stays / still//
//1 yes // .1 do//
//4 . because / that’s //1 where the / magnet / is and it //1 gets at/
tracted //1 by the / earth //4 . but the / other / stars //4 don’t so /
/4 they //1 move a/round //
1. Ideational
1.1 Experiential
Lexicogrammar: clause – transitivity [ Nigel ]
shall I tell you why the North Star stays still
Sayer Process: Receiver Cause Carrier Process: Range
Medium Verbal Beneficiary Medium relational
intensive
313
construing and enacting
it gets attracted by the earth the other stars don’t [get attracted by the earth]
Goal Process: Actor Goal Process: material Actor
Medium material Agent Medium Agent
1.2 Logical
Lexicogrammar: clause complex – interdependency, logical-semantic
relation
314
on the ineffability of grammatical categories
Network: experiential
attributive a
{
왘
material 31 identifying b
{
PROCESS relational 32
왘
clause TYPE verbal 33 intensive x
왘 circ.: place y
VOICE effective 41 possessive z
왘
middle 42
Network: logical
INTERDEPENDENCY paratactic 123...
{
왘
(‘TAXIS’) hypotactic abg...
clause alaborating =
complex expansion 왘 extending +
LOGICAL-SEMANTIC enhancing ⳯
왘
RELATION
projection locution “
왘
idea ‘
315
construing and enacting
2 Ideational semantics
2.1 Experiential
reactants (‘magnet’)
North Star
objects 왘 stars 왘
bodies 왘 other stars
earth
processes 왘 move
motion 왘
Moving (‘do’) 왘 not-move
(‘stay still’)
force
alternatively:
positive
moving
(‘do’) { 왘
왘
negative
uncaused (motion)
caused (force)
2.2 Logical
contrastive (‘but’)
additive 왘
LOGICAL positive (‘and’)
SEMANTIC
왘
RELATIONS effect (‘so’)
causal 왘
unknown (‘why’)
cause 왘
known (‘because’)
316
on the ineffability of grammatical categories
Interpersonal semantics
COMMODITY (proposal) (proposition)
ROLE goods-&-services information
give offer statement
demand command question
give
{
ROLE IN
왘
EXCHANGE
demand
tentative (‘shall I? / will you?’)
VEIN
goods-&-service 왘 neutral (‘let . . . / let me!’)
COMMODITY (proposal)
왘
EXCHANGED definitive (I’ll . . . / you . . .’)
information
(proposition)
2 Interpersonal
Lexicogrammar: clause – mood
shall I tell you
Finite Subject
Mood Residue
do
Finite
mood
317
construing and enacting
it gets attracted
the other stars don’t
they move around
Subject Finite
Mood Residue
Network: interpersonal
declarative 01
indicative 왘 yes/no 02
interrog 왘
WH- 03
{
Mood
왘
1st person 04
clause
2nd person 06
direct 11
MOOD
왘
PROJECTION
indirect 12
major
full 21
ELLIPSIS
왘
elliptical 22
318
on the ineffability of grammatical categories
3 Textual
3.1 Structural (THEME and INFORMATION)
Lexcogrammar: clause – theme
Shall I tell you why the North Star stays still
Inter- topical Inter- topical
personal Rheme personal Rheme
Theme Theme
319
construing and enacting
and it gets attacted by the earth but the other stars don’t so they
move around
Focus
왗 New
320
on the ineffability of grammatical categories
{
mood finite TONALITY unmarked (clause = info.unit)
왘 왘
— WH- marked thematic
INTER- (other)
PERSONAL modality
왘
THEME — fresh
STATUS
왘
vocative contrastive
왘
—
{
continuative
왘
—
TEXTUAL structural
왘
THEME —
conjunctive
왘
—
3.2 Cohesive
stays still 1
왗 왘 the North Star
|R
that (=there) 왗 2 왘 the magnet
|R |C
it 1
왗 왘 gets attracted 3
왗 왘 the earth
S A
S H E E
Key:
SA: synonymy – antonyms 1: Process + Medium
SH: synonymy – cohyponyms 2: Process + Location
C: collocation 3: Process + Agent
E: ellipsis columns: lexical chain
R: reference boxes: referential chain
321
construing and enacting
4. Field
4.1 Context of culture (the system)
Copernican – Galileo – Newtonian universe:
heavenly bodies move:
movement is the natural state of (such) things:
non-movement is exceptional.
Natural phenomena are subject to general ‘laws’:
exceptions need to be explained:
explanation is in terms of cause-&-effect.
Construing this universe (Nigel at 5 years perhaps [escentric: geostatic]).
5. Tenor
5.1 Context of culture (the system)
Family: parent–child as hierarchical relationship (age / generation):
Parent as authority (‘+knowledge’ and ‘+power’).
Class: middle class, intellectual:
Role relationship personal rather than positional, hence
(i) child can impart knowledge (which may be corrected);
(ii) child announces intent (but seeks permission, which may be refused).
- i.e. both forms of authority negotiable.
322
Chapter Twelve
First published in Comprehending Oral and Written Language, 1987. Orlando, FL: Academic
Press, pp. 55–82.
323
construing and enacting
324
spoken and written modes of meaning
325
construing and enacting
going to have been being tested makes five serial tense choices, present in
past in future in past in future, and is also passive. This passed quite
unnoticed by both the speaker and the person it was addressed to; yet
at the time it was being seriously questioned whether a simple verb
form like has been being tested, which one can hear about once a week,
could ever occur in English. Five-term tense forms are, predictably,
very rare – one can in fact make a reasonable guess as to how rare, on
the basis of observed frequencies of two- and three-term tense forms
together with the constraints of the tense system; but they are provided
for within the resources of the spoken language. Another instance I
observed was they said they’d been going to’ve been paying me all this time,
only the funds just kept on not coming through.
Other things I noted regularly included present in present participial
non-finites like being cooking in I never heard you come in – it must have
been with being cooking; marked thematic elements with reprise pronoun,
as in that poor child I couldn’t get him out of my mind; and relatives
reaching into dependent clauses, such as that’s the noise which when you
say it to a horse the horse goes faster. These are all systematic features that
people are unaware that they incorporate in their speech, and often
deny having said even when they are pointed out; or at least reject as
unsystematic – after “I didn’t say it”, the next line of defence is “well
it was a mistake”. But of course it was not a mistake; it was a regular
product of the system of spoken English.
But perhaps the most unexpected feature of those early observations
was the complexity of some of the sentence structures. Here are two
examples from recordings made at the time:
(i) It’s very interesting, because it fairly soon is established when you’re
meeting with somebody what kind of conversation you’re having: for
example, you may know and tune in pretty quickly to the fact that
you’re there as the support, perhaps, in the listening capacity – that
you’re there, in fact, to help the other person sort their ideas; and
therefore your remarks, in that particular type of conversation, are aimed
at drawing out the other person, or in some way assisting them, by
reflecting them, to draw their ideas out, and you may tune in to this, or
you may be given this role and refuse it, refuse to accept it, which may
again alter the nature of your conversation.
(ii) The other man who kicks is the full-back, who usually receives the ball
way behind the rest of his team, either near his line or when somebody’s
done what the stand-off in the first example was doing, kicked over the
defenders; the full-back should be able then to pick it up, and his job is
326
spoken and written modes of meaning
usually to kick for touch – nearly always for touch because he’s miles
behind the rest of his side, and before he can do anything else with the
ball he’s got to run up into them, before he can pass it, because he can’t
pass the ball forward, and if he kicks it forward to another of his side
the other man’s automatically off-side.
And you get a penalty for that, do you, the other side?
Depending on whether it’s kicking or passing forward. Passing forward
– no, it’s a scrum. If you kick it forward and somebody else picks it up
that will be a penalty.
And if not, if the other side picks –
If the other side picks it up that’s all right; but the trouble is this is in
fact tactics again, because you don’t want to put the ball into the hands
of the other side if you can avoid it because it’s the side that has
possession, as in most games of course, is at an advantage.
3 Lexical density
These two examples have been around for a long time; so let me turn
to some recent specimens taken from recordings made by Guenter
Plum to whom I am indebted for drawing them to my attention. In
these spontaneous narratives Plum regularly finds sequences such as the
following:
1A I had to wait, I had to wait till it was born and till it got to about eight
or ten weeks of age, then I bought my first dachshund, a black-and-tan
bitch puppy, as they told me I should have bought a bitch puppy to
start off with, because if she wasn’t a hundred percent good I could
choose a top champion dog to mate her to, and then produce something
that was good, which would be in my own kennel prefix.
327
construing and enacting
This displays the same kind of mobility that the earlier observations
had suggested was typically associated with natural, unselfconscious
speech – which is what it was. I asked myself how I would have
expressed this in writing, and came up with two rewordings; the first
(1B) was fairly informal, as I might have told it in a letter to a friend:
1B I had to wait till it was born and had got to about eight or ten weeks of
age; that was when I bought my first dachshund, a black-and-tan bitch
puppy. By all accounts I should have bought a bitch puppy at the start,
because if she wasn’t a hundred percent good I could mate her with a
top champion dog and produce a good offspring – which would carry my
own kennel prefix.
My second rewording (1C) was a more formal written variant:
1C Some eight or ten weeks after the birth saw my first acquisition of a
dachshund, a black-and-tan bitch puppy. It seems that a bitch puppy
would have been the appropriate initial purchase, because of the
possibility of mating an imperfect specimen with a top champion dog,
the improved offspring then carrying my own kennel prefix.
The aim was to produce a set of related passages of text differing
along one dimension, which could be recognized as going from “most
likely to be spoken” to “most likely to be written”. How such variation
actually correlates with difference in the medium is of course problem-
atic; the relationship is a complicated one, both because written /
spoken is not a simple dichotomy – there are many mixed and
intermediate types – and because the whole space taken up by such
variation is by now highly coded: in any given instance the wording
used is as much the product of stylistic conventions in the language as
of choices made by individual speakers and writers. Here I am simply
moving along a continuum which anyone familiar with English usage
can readily interpret in terms of “spoken” and “written” poles.
The kind of difference that we find among these three variants is
one that is often referred to as a difference of ‘texture’, and this familiar
rhetorical metaphor is a very appropriate one: it is as if they were the
product of a different weave, with fibres of a different yarn. But when
we look behind these traditional metaphors, at the forms of language
they are describing, we find that much of the difference can be
accounted for as the effect of two related lexicosyntactic variables. The
written version has a much higher lexical density; at the same time, it
has a much simpler sentential structure. Let us examine these concepts
in turn.
328
spoken and written modes of meaning
Table 1
Lexical Density of Texts 1A, 1B, and 1C
(1) (2) (1:2) (3) (1:3)
Lexical Running Clauses
items words
1A 23 83 1:3.6 13 1.8:1
1B 26 68 1:2.6 8 3.3:1
1C 25 55 1:2.2 4 6.3:1
329
construing and enacting
grammatical words – goes down; and the number of clauses goes down
even more.
Let us attempt a similar rewording the other way round, this time
beginning with a passage of formal written English taken from Scientific
American:
2A Private civil actions at law have a special significance in that they provide
an outlet for efforts by independent citizens. Such actions offer a means
whereby the multiple initiatives of the private citizens, individually or
in groups, can be brought to bear on technology assessment, the
internalization of costs and environmental protection. They constitute a
channel through which the diverse interests, outlooks and moods of the
general public can be given expression.
The current popular concern over the environment has stimulated
private civil actions of two main types.
2B is my attempt at a somewhat less “written” version; while 2C is in
another step nearer to speech:
2B Private civil actions at law are especially significant because they can be
brought by independent citizens, so enabling them to find an outlet for
their efforts. By bringing these actions, either as individuals or in groups,
private citizens can regularly take the initiative in assessing technology,
internalizing costs and protecting the environment. Through the use of
these actions as a channel, the general public are able to express all their
various interests, their outlooks, and their moods.
Because people are currently concerned about the environment, they
have been bringing numerous private civil actions, which have been
mainly of two types.
2C One thing is especially significant, and that is that people should be
able to bring private civil actions at law, because by doing this
independent citizens can become involved. By bringing these actions,
whether they are acting as individuals or in groups, private citizens can
keep on taking the initiative; they can help to assess technology, they
can help to internalize costs, and they can help to protect the environ-
ment. The general public, who want all kinds of different things, and
who think and feel in all kinds of different ways, can express all these
wants and thoughts and feelings by bringing civil actions at law.
At present, people are concerned about the environment; so they have
been bringing quite a few private civil actions, which have been mainly
of two kinds.
330
spoken and written modes of meaning
4 Grammatical intricacy
We have characterized the difference in general terms by saying that
written language has a higher lexical density than spoken language; this
expresses it as a positive feature of written discourse and suggests that
writing is more complex, since presumably lexical density is a form of
complexity. Could we then turn the formulation around, and express
the difference as a positive characteristic of spoken language? To say
that spoken discourse has more words in it, or even more clauses, does
not seem to convey anything very significant about it. We need to
look at how the words and clauses are organized.
Let us consider a shorter example of a pair of texts related in the
same way, one “more written” (Text 3A), the other “more spoken”
(Text 3B). I have constructed these so that they resemble the originals
of Texts 1 and 2; but they are based on a natural example occurring in
two texts in which a person had described the same experience twice
over, once in speech and once in writing.
More “written”:
3A Every previous visit had left me with a sense of the risk to others in
further attempts at action on my part.
More “spoken”:
3B Whenever I’d visited there before I’d end up feeling that other people
might get hurt if I tried to do anything more.
331
construing and enacting
Table 3
Figure 1
332
spoken and written modes of meaning
Figure 2
Let us return to Text 1, in its original spoken form (Text 1A). This
consisted of 13 clauses. However, these 13 clauses were not strung out
end to end; they were constructed into a small number of clause
complexes of mixed paratactic and hypotactic construction: arguably
just one clause complex throughout. Here is its interpretation as one
clause complex:
Figure 3
333
construing and enacting
Table 4
Other examples from the same source but from different speakers
show similar patterns; there are, obviously, individual differences
(including perhaps in the preference for one or other type of interde-
pendency), but the same free-flowing intricacy is noticeable all the
time, as in Texts 4–6:
4 Roy was always interested in dogs and unfortunately he’d never had the
opportunity to have a dog of his own, just because of circumstances –
where he lived and what not, and so I bought him a Shepherd pup, which
was supposedly, you know, pure-bred Shepherd, but unfortunately people
sold it because it didn’t have papers with it, so it was a ‘pup’.
5 Now how I got a German Shepherd was that I worked with a veterinary
surgeon, as I’ve told you before, and there used to be a lady that brought
her Shepherds along to the clinic and I used to admire them greatly, and
she said, ‘Well,’ she said, ‘if you get married I’ll give you one as a
wedding present,’ so immediately I bustled around looking for someone to
marry so I could get a Shepherd given to me for a wedding present, you
see, so that’s how that worked out well, not quite! However I got my
Shepherd and he was my first dog, mainly because when I was a youngster
I always wanted a dog but I lived with grandparents who wouldn’t have
dogs or cats and I was a very frustrated animal lover at that stage of the
game, so as soon as I got out on my own I sort of went completely berserk!
6 So we rang up the breeder, and she sort of tried to describe the dog to us,
which was very hard to do over the phone, so we went over to have a look
to see what they were like, and we bought Sheba, because at that stage
Bob was away a lot on semitrailers with the army and it used to get quite
334
spoken and written modes of meaning
bad with the exercises – you’d have prowlers and perverts through the
married quarters, so if we, you know, got a dog, which we could do
because it didn’t matter what sort of dog anyone had, it’d bark and they
wouldn’t bother us.
5 Types of complexity
Two distinct points need to be made here, and both of them run
counter to received attitudes towards spoken language. One is that
speech is not, in any general sense, ‘simpler’ than writing; if anything,
it is more complex. There are, of course, many different kinds of
complexity, and we have already noted one measure – lexical density
– whereby speech will appear as the simpler of the two. But the
patterns we have been illustrating, which are the patterns of the
organization of the clause complex, referred to above as grammatical
intricacy, would seem to be at least as central to any conception of
complexity; and in this respect, speech appears as the more complex.
The “syntactic complexity expected in writing”, with which Deborah
Tannen (1982) introduces her discussion of oral and literate strategies,
does not turn out to be a characteristic of written discourse.
Of course, there are many other variables. Some writers achieve
considerable intricacy in the structure of the clause complex; it can be
learnt and consciously developed as a style. Some forms of spoken
discourse, on the other hand, militate against it: rapid-fire dialogue
presents no scope for lengthy interdependencies – complex semantic
patterns can be construed between interactants, but usually without
being realized in syntactic terms. And the categories of “written” and
“spoken” are themselves highly indeterminate – they may refer to the
medium in which a text was originally produced, or the medium for
which it was intended, or in which it is performed in a particular
instance; or not to the medium at all, but to other properties of a text
which are seen as characteristic of the medium. So it is important to
indicate specifically which variable of discourse is being referred to,
when one variety is being said to display some distinctive characteristic.
My point here is to question the assumption that written language is
syntactically more complex than spoken, and to suggest that, as far as
one particular kind of syntactic complexity is concerned – the intricacy
(I do not want to call it “structure” because that assumes a particular
interpretation) of the sentence or “clause complex” – this is more a
characteristic of the most unconscious spontaneous uses of language.
The more natural, un-self-monitored the discourse, the more intricate
335
construing and enacting
336
spoken and written modes of meaning
337
construing and enacting
338
spoken and written modes of meaning
silent thought; this would then tell us what the writer actually wrote.
Figure 4 is a specimen.
339
construing and enacting
for most of a century. But for many purposes the discarded first
attempts are merely trivial; they clutter up the text, making it hard to
read, and impart to it a spurious air of quaintness. What is much more
serious, however, is that transcribing spoken discourse in this way gives
a false account of what it is really like. It may seem a harmless piece of
self-gratification for a few academics to present spoken language as a
pathological phenomenon; one might argue that they deceive nobody
but themselves. But unfortunately this is not the way. Just when we
are seeing real collaboration between linguists and educators, and the
conception of “language in education” is at last gaining ground as a
field of training and research, it seems we are determined to put the
clock back to a time when spoken language was not to be taken
seriously and could have no place in the theory and practice of
education.
Let us recapitulate the argument. Speech and writing as forms of
discourse are typically associated with the two modal points on the
continuum from most spontaneous to most self-monitored language:
spontaneous discourse is usually spoken, self-monitored discourse is
usually written. We can therefore conveniently label these two modal
points “spoken” and “written” language. Spoken and written language
do not differ in their systematicity: each is equally highly organized,
regular, and productive of coherent discourse. (This is clearly implied
once we recognize them both as “language”.)
Discourse in either medium can be characterized by hesitation,
revision, change of direction, and other similar features; these tend to
arise when attention is being paid to the process of text production.
Since highly monitored discourse is typically written, these features are
actually more characteristic of writing than of speech; but because most
written text becomes public only in its final, edited form, the hesitations
and discards are lost and the reader is shielded from seeing the process
at work. Where they are likely to remain in is precisely where they
occur least, in the more spontaneous kinds of writing such as personal
letters. (Not all discourse features that are regarded as pathological, or
assigned negative value, are of this self-monitoring kind. One form of
discourse that has received a lot of critical attention is casual conver-
sation, where the well-recognized characteristics are those of turn-
taking, such as interruptions and overlaps. But the strictly linguistic
“deviations” of casual conversation are mainly systematic features that
would not seem deviant if we had a grammar that took into account
the specifically “spoken” resources of the linguistic system.)
Spoken and written language do differ, however, in their preferred
340
spoken and written modes of meaning
341
construing and enacting
simply with the evangelizing aim [[to which I referred earlier]]. ||| It
is not absolutely incompatible with that aim, however, for the following
reason.||| [[What counts as indoctrination and the like]] depends upon
a number of criteria,|| to do with the degree [[to which a teacher fails to
mention alternative beliefs]] , the tone of voice [[used]], the lack of
sympathy for the criticism [[levelled at Christianity or Humanism]]
and so on. |||A dogmatic teacher or lecturer differs from an open
one.||| The non-dogmatic teacher may be tepid;|| the open one may
be fervent.||| Fervour and indifference are not functions of closedness
and openness.|||
To see how this lexical density is achieved, we can look at the first
clause. After the cohesive thus, it begins with a nominal group the
sympathetic induction of people into a proper and deep understanding of what
Christianity is about. The Head is induction; the Postmodifier consists of
a series of alternating embedded prepositional phrases and nominal
groups, mainly one inside the other, and ending with an embedded
clause:
342
spoken and written modes of meaning
343
construing and enacting
344
spoken and written modes of meaning
8 Grammatical metaphor
Where then in the linguistic system do spoken and written discourse
diverge? A language, if it is not written down, consists of three
interrelated subsystems: a semantic system (meanings), coded into a
lexicogrammatical system (wordings), recoded into a phonological
system (sounds). A language that has a writing system has an alternative
form of expression: visual symbols as well as sounds. In such a language,
a written text could, in principle, be a spoken text that has been
written down (a transcription); here the written version is a transcod-
ing of something that has already been coded in sound. Most writing is
not like this. Secondly, a written text could be an alternative
expression of a given wording: in this case meanings are coded as
words and structures (“wordings”), which are then expressed either in
sound or in writing. If this was the norm, there would be no systematic
difference between spoken and written texts; the medium would not
be a significant register variable. But there are such differences; so, to
some extent at least, spoken and written discourse must represent
alternative wordings. In this third case, meanings are coded either as
“speakable wordings” or as “writeable wordings”, the former appropri-
ate to the dynamic nature of the text process, the latter appropriate to
the synoptic nature of the text product. This is the sort of interpretation
we have been offering.
But is it the whole story? There is still a fourth possibility – that
speech and writing can diverge already at the semantic level, so that
spoken and written discourse embody different meanings. Is there any
sign that this can happen? It would of course be only a very partial
effect; no one has suggested that the two derive from different semantic
systems (or even two different lexicogrammatical systems, for that
matter). But we should consider the possibility that there is some
flowback into the meaning.
Consider the last sentence of Text 2, in its original written form
(2A):
The current popular concern over the environment has stimulated private civil
actions of two main types.
We “translated” it into something more speech-like as:
At present, people are concerned about the environment; so they have been
bringing quite a few private civil actions, which have been mainly of two
kinds.
345
construing and enacting
346
spoken and written modes of meaning
processes, one mental and one material, have been dressed up as one
which is neither. This coding of a semantic relation between two
processes as if it was the single process is very common in writing; the
sentence immediately preceding Text 2A contained another example
of the same thing, here with the verb leads to:
A successful tort action leads to a judgment of damages or an injunction
against the defendant company.
But this is just one type of a more general phenomenon, something
that I call grammatical metaphor (Halliday 1985, Chapter 10). Written
language tends to display a high degree of grammatical metaphor, and
this is perhaps its single most distinctive characteristic.
Here are three further examples of grammatical metaphor taken
from various written sources, together with suggested rewordings
which are less metaphorical:
Issue of the specially-coded credit cards will be subject to normal credit
checking procedures.
“Credit cards have been specially coded and will be issued only when
credit has been checked in the normal way.”
Strong Christmas sales were vital to the health of the retail industry,
particularly in the present depressed climate.
“Unless many goods were sold at Christmas the retail industry would
not be healthy, particularly when the economy is depressed as it is
now.”
He also credits his former big size with much of his career success.
“He also believes that he was successful in his career mainly because he
used to be big.”
In all these examples nominalization plays a significant part, as it
does in many types of grammatical metaphor; so it is perhaps worth
stressing that nominalization is well motivated in English. It is not
simply a ritual feature that has evolved to make written language more
ambiguous or obscure; like the passive, which is another feature whose
functions are widely misunderstood, nominalization is an important
resource for organizing information. Take the example youth protest
mounted, which is not a headline but a complete sentence from a feature
article. We might reword this as more and more young people protested, or
young people protested more and more; but the only way to get the
combination of youth and protest as the Theme of the clause is by means
347
construing and enacting
348
spoken and written modes of meaning
349
construing and enacting
350
spoken and written modes of meaning
351
Chapter Thirteen
I realize that the title might well prompt someone to ask, ‘How do
you mean, “How do you mean?”?’ I could have written, ‘How are
meanings made?’ – although I prefer the more personalized version.
The question is meant theoretically; but, like so many theoretical
questions, it becomes relevant in practice the moment we want to
intervene in the processes we are trying to understand. And some
processes of meaning are involved in more or less everything we do.
I shall need to talk about two fundamental relationships, those of
realization and instantiation; so let me begin by distinguishing these
two. Instantiation I take to be the move between the system and the
instance; it is an intrastratal relationship – that is, it does not involve a
move between strata. The wording fine words butter no parsnips is an
instance, or an instantiation, of a clause. Realization, on the other hand,
is prototypically an interstratal relationship; meanings are realized as
wordings, wordings realized as sound (or soundings). We often use the
term to refer to any move which constitutes a link in the realizational
chain, even one that does not by itself cross a stratal boundary (for
example, features realized as structures); but the phenomenon of realiza-
tion only exists as a property of a stratified system. To anticipate the
discussion a little, I shall assume that realization may be formalized as
metaredundancy, as this is defined by Jay Lemke (1985). Instantiation I
shall define by making reference to the observer; it is variation in the
observer’s time depth. Firth’s concept of exponence is the product of
these two relations: his “exponent” is both instantiation and realization.1
First published in Advances in Systemic Linguistics: Recent Theory and Practice, 1992, edited by
Martin Davies and Louise Ravelli. London: Pinter, pp. 20–35.
352
how do you mean?
form of
consciousness
action reflection
domain
of experience
1st/2nd person regulatory interactional
3rd person instrumental personal
353
construing and enacting
354
how do you mean?
355
construing and enacting
356
how do you mean?
357
construing and enacting
Figure 2 Metaredundancy
358
how do you mean?
interpretation with a dynamic one. This leads us into the other critical
concept, that of instantiation.
Consider the notion of climate. A climate is a reasonably stable
system; there are kinds of climate, such as tropical and polar, and these
persist, and they differ in systematic ways. Yet we are all very concerned
about changes in the climate, and the consequences of global warming.
What does it mean to say the climate is changing? Climate is instantiated
in the form of weather: today’s temperature, humidity, direction and
speed of wind, etc., in central Scotland are instances of climatic
phenomena. As such they may be more, or less, typical: today’s
maximum is so many degrees higher, or lower, than average – meaning
the average at this place, at this time of year and at this time of day.
The average is a statement of the probabilities: there is a 70 per cent
chance, let us say, that the temperature will fall within such a range.
The probability is a feature of the system (the climate); but it is no
more, and no less, than the pattern set up by the instances (the weather),
and each instance, no matter how minutely, perturbs these probabilities
and so changes the system (or else keeps it as it is, which is just the
limiting case of changing it).
The climate and the weather are not two different phenomena. They
are the same phenomenon seen by two different observers, standing at
different distances – different time depths. To the climate observer, the
weather looks like random unpredictable ripples; to the weather
observer, the climate is a vague and unreal outline. So it is also with
language;8 language as system, and language as instance. They are not
two different phenomena; they are the same phenomenon as seen by
different observers. The system is the pattern formed by the instances;
and each instance represents an exchange with the environment – an
incursion into the system in which every level of language is involved.
The system is permeable because each instance redounds with the
context of situation, and so perturbs the system in interaction with
the environment. Thus both realization and instantiation are involved
in the evolution of language as a dynamic open system.
359
construing and enacting
360
how do you mean?
361
construing and enacting
362
how do you mean?
363
construing and enacting
Here for the first time Nigel is selecting in two systems of meaning at
once, and by this token the initial move into grammar has been made.
Through the second year of life this new stratified system will gradually
replace the protolinguistic one, and all meanings (except for a few
protolanguage remnants that persist into adult life like hi! and ah! and
yum! and ouch!) will come to be stratally and metafunctionally complex.
So in Nigel’s first exemplar, just cited, we have (i) proto-metafunctions
(proto-ideational – different persons; proto-interpersonal – seeking /
finding), and (ii) proto-strata, with the meaning ‘first’ construed as
wording (the ideational as contrasting names; the interpersonal as
contrasting mood) and ‘then’ (re)construed as sounding (names as
articulation, mood as intonation). At this second interface the child can
now combine the segmental and the prosodic choices, in this way both
realizing and also iconically symbolizing the two different modes of
meaning that are combined at the first interface. The resources for
making meaning are now in place.
It is probably not a coincidence that, as the ideational grammar
evolved, so in the system of transitivity the field of processes was
construed into different process types along precisely the lines that (if
my understanding is right) went into the making of meaning in the first
place. If meaning arises out of the impact of the conscious and the
material, as mutually contradictory forms of experience, then it is not
surprising that when experience is construed semantically, these two
types of process, the material and the conscious, should come to be
systematically distinguished. But there is a further twist. The semogenic
process, as we saw, involves setting up a relationship between systems
such that one is the realization of the other – that is, they stand to each
other in a relation of Token and Value. This Token–Value relationship
is set up at both interfaces, and it is also what makes it possible to prise
the two apart and wedge in a grammar in between. Here then we find
the third of the kinds of process construed by the grammar: the relational
process, based on identifying a Token with a Value. The grammar of
natural language, in its ideational metafunction, is a theory of human
experience; thus it may reasonably be expected to take as its point of
departure the very set of contrasts from which its own potential is
ultimately derived.
Let me return once again, finally, to the suggestion that meaning is a
mode of action engendered at the intersection of the material (or
phenomenal) and the conscious, as complementary modes of experience.
Now, the effect of this impact is to construe order. By the act of
meaning, consciousness imposes order on the phenomena of experience.
364
how do you mean?
365
construing and enacting
366
how do you mean?
Notes
1. For Firth’s concept of ‘exponence’ see especially his ‘Synopsis of Linguistic
Theory’ in Firth (1957a).
2. I am speaking here of phylogenesis; but the process is recapitulated in the
growth of the individual, where it can be observed in the form of
behaviour. A child experiences certain phenomena as ‘out there’ – as lying
beyond the boundary between ‘me’ and ‘non-me’: some perturbation seen
or heard, like a flock of birds taking off, or a bus going past, or a coloured
light flashing. At the same time, he also experiences a phenomenon that
is ‘in here’: his own consciousness of being curious, or pleased, or
frightened. At first these two experiences remain detached; but then
(perhaps as a result of his success in grasping an object that is in his line of
sight – in Trevarthen’s terms, when “pre-reaching” becomes reaching,
typically at about four months) a spark flies between them by which the
material is projected on to the conscious as ‘I’m curious about that’, ‘I like
that’ and so on. Now, more or less from birth the child has been able to
address others and to recognize that he is being addressed (Catherine
Bateson’s “proto-conversation”). The projection of the material on to the
conscious mode of experience maps readily on to this ability to address an
other; and the result is an act of meaning – such as Nigel’s very high-
pitched squeak, which he first produced at five months, shortly after he
had learnt to reach and grasp.
3. Other microfunctions were added as the protolanguage evolved by degrees
into the mother tongue; but these were the original four. See Halliday
(1975, 1978).
4. At first labelled, somewhat misleadingly, the level of “context”. See the
discussion of levels in Halliday, McIntosh and Strevens (1964). See also
Ellis (1966).
5. Based on giving and demanding – that is, on exchange. Initially this meant
the exchange of goods-and-services; but eventually, by a remarkable
dialectic in which the medium of exchange became itself the commodity
exchanged, it extended to giving and demanding information. By this
step, meaning evolved from being an ancillary of other activities to being
a form of activity in its own right.
6. See the chapters entitled ‘Towards a model of the instructional process’,
‘The formal analysis of instruction’ and ‘Action, context and meaning’ in
Lemke (1984).
7. It is impossible to have metaphor in a protolanguage at all, unless one
chooses to call “metaphor” (or perhaps “proto-metaphor”) what is taking
place when, for example, Nigel transfers a particular sign [gωg gωg gωg]
from ‘I’m sleepy’ to ‘let’s pretend I’m going to sleep’. See Halliday (1975:
Chapter 2).
8. The analogy should not, of course, be pressed too far. Specifically, while
367
construing and enacting
368
Chapter Fourteen
Let me first say what I mean by “grammar” in the title of the paper. I
mean the lexicogrammatical stratum of a natural language as tradition-
ally understood, comprising its syntax and vocabulary, together with
any morphology the language may display: Lamb’s “lexical system”, in
his current (1992: Chapter 5) ‘three-level architecture’ – in common-
sense terms, the resources of wording in which the meanings of a
language are construed. And here I have in mind particularly the
evolved, spontaneous grammar that construes the discourse of daily life.
This is not to exclude from the picture the elaborated grammars of
scientific and other metalanguages; but these can only be understood as
what they are: an outgrowth, supported by design, of the original
grammar that is learnt at mother’s knee and on father’s shoulders.
Now English is not very efficient at creating technical nomenclature,
since it tends to confuse the study of a phenomenon with the
phenomenon itself. So while the term “grammar” is commonly used
in the way in which I have defined it, to mean the wording system,
the central processing unit of a natural language, it is also used
indiscriminately to mean the study of that system: grammar2 meaning
‘the study of grammar1’. Since the study of language is called “linguis-
tics”, I have been calling the study of grammar “grammatics” in order
to make the distinction clearer. A grammatics is thus a theory for
explaining grammar.
But is not a grammar itself also a theory? Clearly it is. A grammar is
First published in Functional Approaches to Language, Culture and Cognition, 2000, edited by
Teun A. van Dijk. Amsterdam: John Benjamins, pp. 221–37.
369
construing and enacting
370
grammar and daily life
in the way that it has been done countless other times by countless
other people, so it is reasonable to talk about the condition being
construed ‘by the grammar’.
In I have a headache the grammar construes a kind of thing, called an
ache; it then uses a part of the body to classify this thing, setting up a
taxonomy of aches including stomachache, backache and various others.
(Not all the parts of the body are allowed to ache, however; you cannot
have a footache or a thighache.) The grammar then sets up a configuration
of possession between the ache and some conscious being, in this case
the speaker I. The speaker becomes the owner of one specimen of that
complex class of things. It is not a prototypical form of possession; the
possessor does not want the thing possessed but cannot get rid of it –
cannot give it away, or put it back where it came from. Why then
does the grammar not favour my head aches; or my head’s aching? – in
which the aching is a process, a state of being, rather than a thing, and
the entity involved in that state of being is my head rather than me.
The grammar has no trouble in constructing the clause my head aches;
yet it is not the most usual way in which the experience is worded.
Why is I have a headache preferred instead?
In English, as in many other languages (though not all), there is a
particular meaning associated with being the first element in the clause.
What is put first is being instated by the speaker as the theme of the
coming message; it is the setting for the information that follows (Fries
1995). This pattern of the clause, a structure of “Theme + Rheme”,
was apparently identified by the earliest rhetorical grammarians of
ancient Greece, the sophists, who seem to have recognized in the
thematic organization of the clause a potent resource for constructing
legal and political discourse. In modern times it was first investigated in
detail by Mathesius, the founder of the Prague school; it is a particularly
prominent feature of English, appearing not only in the clause but also
as a “fractal” pattern in both smaller and larger structures – inside word
groups, both nominal and verbal, on the one hand and extending over
a nexus of clauses on the other. The following example, taken from
natural conversation, shows thematic predication of a whole clause
complex (from Svartvik and Quirk 1980: 304):
371
construing and enacting
I’m dazzled, you know . . . It’s being able to see your working life
will span a period in which so-&-so is the topmost industry which I
find so dazzling.
Now if I say my head aches, the first element in that clause is my head:
I have constructed a message in which my head is enunciated as Theme.
My head is instated as what I want to elaborate on. But it isn’t; I’m the
one that’s suffering, so the Theme of the clause should more appropri-
ately be ‘me’. How does the grammar accommodate this alternative?
Most naturally, by making ‘me’ the Subject, since there is a strong
association of these two functions in English. The ‘ache’ becomes a
thing separated from myself, something that I possess, with my head
identified as its location: I have an ache in my head. Better still, if my
head is used as a classifier, the ache and its location become a single
complex thing; and this now occupies the culminative position in the
clause: I have a headache. The flow of information here is very different
from that of my head’s aching.
If this was just a feature of the grammar of localized aches and pains,
it might remain a curiosity, a special effect rather than a principle. But
this pattern has evolved in English as the prototypical form for
construing bodily qualities and states; rather than her hair is long, his
throat is sore, we tend to say she has long hair, he has a sore throat, putting
the person rather than the body part into the thematic role.1 And in
certain other languages where initial position is thematic we also
regularly find the person, rather than the body part, lodged at the
beginning of the clause. The overall patterns are of course different: in
particular, there may be no strong bond between Theme and Subject,
and this makes it clear that the relevant function is that of Theme. We
can give examples from Chinese, Russian and French. In Chinese it is
possible to say wǒdi tóu tèng ‘my head aches’, where as in the English
wǒdi tóu ‘my head’ is a single element in the clause and so functions as
the Theme. The preferred form, however, is wǒ tóu tèng ‘me the head
aches’, where the ‘head’ is detached from the personal pronoun; wǒ
‘me’ and tóu ‘head’ are now independent elements in the clause and
only the first one, wǒ, is thematic. Again, this is the typical pattern for
all such expressions in Chinese: tā tóufǎ cháng ‘her the hair (is) long’, tā
hóulóng tòng ‘him the throat (is) sore’ and so on. In Russian, likewise,
one can say moja golova bolit ‘my head aches’; but this also is not the
preferred form. Russian however displays a different pattern: u menja
golova bolit ‘at me the head aches’, where again it is the ‘me’ that has
thematic status. In French instead of ma tête me fait mal ‘my head is
372
grammar and daily life
hurting me’ one can use possession as in English: j’ai mal à la tête ‘I
have an ache at the head’. French also has a further device, of detaching
the Theme altogether from the structure of the clause, and announcing
it as a key signature at the beginning: moi j’ai mal à la tête ‘me I’ve got
an ache at the head’. Neither Chinese wǒ nor Russian u menja nor
French moi is Subject; what they have in common is the status of
Theme.
At this point we might think once more of the sufferer and say to
him or her: pity you’ve got a headache. But try de-construing this, in
the grammar, and then re-construing it – rewording it – as my head
aches; or better still my head’s aching, which makes it an external rather
than an internal phenomenon. This is rather less self-centered: it is no
longer a fact about me, and my inner self, but an external fact about
my head. This won’t make the headache go; but it does put it in its
place. It has now become a problem of my head, which is just one part
of my physical make-up. One might offer this as a form of logotherapy,
a kind of grammatical acupuncture. But here I just want it to serve as
an instance of “thinking grammatically”.
Thus the grammar enables us, unconsciously, to interpret experience;
and the metagrammar, or grammatics, enables us to reflect consciously
on how it does so. The grammatics, of course, is part of a more general
theory of meaning: of language as a semiotic system, and of other
semiotic systems brought into relation with language. Without such a
general theory, the excursion into other languages is no more than a
piece of tourism; it assumes significance only when we can show how
this small corner of experience is construed in relation to the meaning
potential of each language as a whole.
But this requires much more than a purely local explanation. Taking
a fragment of the grammar of daily life, and exploring it cross-
linguistically in this way, still leaves it as an isolated fragment, detached
from its environment in the overall system of the language. Yet this is
the critical environment to take into account. The grammar construes
a unitary semantic space, elastic and many-dimensioned; and whatever
aspect of the grammar we are considering (such as the selection of
person as Theme, in the examples above), there will usually be various
other grammatical features, many of them not obviously related in any
formal sense, which are associated topologically within this semantic
space (cf. Martin and Matthiessen 1992). Such features may cluster into
a recognizable syndrome, needing to be interpreted not piecemeal but
as a whole: this is the principle of “frames of consistency” as formulated
by Whorf. Illustrations of this phenomenon may be found in Hasan’s
373
construing and enacting
(1984b) ‘Ways of saying, ways of meaning’, where she shows how the
grammar of Urdu construes experience as collectively shared; and in
Martin’s (1988) account of “grammatical conspiracies” in Tagalog. If
we are comparing the different “realities” of one language with
another, it is the syndrome rather than the single feature that is likely
to be significant.
Side by side with such frames of consistency, however, there are also
frames of inconsistency: regions where the grammar construes a pattern
out of tensions and contradictions – where the different “voices” of
experience conflict. To put this another way, the grammar’s theory of
experience embodies complementarity as well as concurrence. Meta-
phorically the grammar is representing the fact that human experience
is too complex, and has too many parameters, to be construed from
any one angle alone. It is the combination of these two perspectives –
concurrence and complementarity – that is the salient characteristic of
the grammar of daily life.
Let me first try to illustrate the complementarity, and then use this
as a point of departure for exploring concurrence, looking at a more
general syndrome of features within which the earlier, more particular
example might be located. Many grammars (perhaps all) make a rather
clear distinction between the two fundamental modes of human
experience referred to above: between what we experience as taking
place in the world outside of ourselves and what we experience as
processes of our own consciousness – seeing and hearing, liking,
disliking, fearing, hoping, thinking, knowing, understanding and the
like. In English, the conscious or mental processes differ from the
other, material kind in various respects: (1) they have a less exact
present time; (2) they presume a conscious being taking part; (3) they
do not fall within the scope of ‘doing’, and (4) they can project – that
is, they can construe any meaning as taking place in someone’s
consciousness (as “direct or indirect thought”). In addition, these inner
processes display another feature not found with the grammar of
processes of the external, material type: they are bi-directional. Pro-
cesses of consciousness can be construed with the conscious participant,
the Senser, either as object (active Complement), as in it frightens me,
or as active Subject, as in I fear it; likewise it pleases/convinces/strikes me,
I like/believe/notice it, and so on. These are two different and in fact
contradictory constructions of the same class of phenomena. Inner
experience is complex and difficult to interpret; the grammar offers
two complementary models, one with the Senser in the more active
role (by analogy with material processes), one with the Senser appearing
374
grammar and daily life
to be acted upon. Each of these brings out different agnate forms; the
grammar of daily life, in English, accommodates both.
In late Middle to early Modern English the common verbs of
consciousness such as like and think changed their allegiance from the
one pattern to the other: from ‘it likes/thinks (to) me’ to I like it, I
think so. This happened at about the same time as the emergence of the
pattern discussed earlier: I have a headache, etc. For very general
processes of consciousness the grammar came to favour the type of
construction in which the Senser, the participant credited with con-
sciousness, was the Theme. What was explained above as a preference
for a person rather than a part of the body as the starting point for
bodily states and conditions is part of a broader picture whereby the
grammar of all inner processes and physiological states tended to orient
the message towards the human, or humanlike, participant – perhaps
with ‘I’, the individual self, as the prototypical member of this class.
This in turn leads us to another feature. At the same period of
history another shift took place affecting processes of the external kind,
those experienced as happening ‘out there’. In earlier English the
grammatical Subject in such processes had been overwhelmingly the
active participant, whether human or not (in fact the distinction
between human and non-human, or conscious and non-conscious,
plays no part in the construction of these processes of the external
world). Thus in an arrow pierced his eye the arrow was the natural Subject,
and remained Subject even if the narrative required the thing acted on
to function as Theme. To use a constructed example, the pattern was
that of:
The king fell to the ground; his eye an arrow had pierced.
with the Actor remaining as Subject even when displaced from initial
position in the clause. Subsequently, as already noted, this bond
between Subject and Actor was deconstructed and replaced by a
different bond, that of Subject with Theme; this gave the modern
pattern:
The king fell to the ground; his eye had been pierced by an arrow.
This change led to an increase in the frequency of passive verbs, which
was followed by a change in the tense system as passive tenses caught
up with the active ones; and various other changes took place besides.
What this new alignment of grammatical forces amounted to was that
relatively less prominence was being given to the structure of the
experience – which partner is the doer and which the done-to, so to
375
construing and enacting
376
grammar and daily life
me, you say I gave a lot of money to my brother. The preposition to makes
explicit your brother’s role as a participant in the process, and is added
just in those positions which are prominent in the information flow
(likewise if the brother appears as a marked Theme: to my brother I
gave a lot of money). It is precisely this same principle which adds by to
the Actor when the clause is passive: his eye had been pierced by an arrow/
by an arrow his eye had been pierced.
All the features I have sketched in here are features of the grammar
of daily life: some more global, some more local, but all of them
characteristic of unconscious, spontaneous, everyday linguistic encoun-
ters. These, and others that could be added, form a syndrome, a
concurrence of related developments, that has helped to shape the
meaning potential of Modern English, giving the language its charac-
teristic flavour – that “certain cut”, in Sapir’s terms, which makes each
language unique. What all these have in common is that they tend
towards giving greater prominence to the organization of discourse as
a flow of information, making more explicit how each element is to
be construed as part of a message. As a corollary to this, less prominence
is given to the experiential patterning, much of which is in fact left
implicit once the concern with the message begins to take over. Most
of these effects are fairly recent in history; they reflect the changing
social conditions of the language over the past five hundred years. Or
rather: they do not reflect them – they help to bring them about.
These features in the grammar construe the kind of discourse that can
be addressed to a stranger, who does not necessarily share the same
expectations and norms of interaction. They can be written down in a
book that is going to be printed in thousands of copies and read by
people who have never met the author and do not even know who he
is. In other words, they are features of a standard language: a form of
discourse in which the flow of information will typically be rendered
explicit rather than being taken for granted. (Interestingly, many of
these changes appear not to have taken place in the surviving British
rural dialects.)
Effects like these are not the result of sudden catastrophic changes.
They are trends and tendencies in a long process of evolution; and at
any given time they are quantitative – changes in the relative frequency
with which this or that pattern is selected from within the system. The
grammatics is thus a theory of probabilities, in which possible/imposs-
ible is only a special case of more and less probable – and a rather
uninteresting case, because meaning is a product of choice and when
something becomes impossible there is no more choice. So, for
377
construing and enacting
378
grammar and daily life
379
construing and enacting
Mum is the doer and the jacket is the done-to. This is a configurational
model; there is no re-entry to the choice of agency. Then there’s a
snap, and the grammar takes up the story again:
and so on. This second theory says that there is a Medium, an entity
through which the process is actualized (here the thread), plus, option-
ally, something else as Causer that brings it about. This is an iterative
model; here the agency relation is construed in such a way that it can
recur.
Thus there are two ways of looking at a process: one according to
which participant a acts, and the action may (or may not) extend to
another participant x (a is the constant, x the variable); the other
according to which participant x “eventuates” (that is, permits the
process to eventuate), and the event may (or may not) be brought
about by another participant a (x is the constant, a the variable). The
first of these (let us call it type A) is the transitive theory of processes,
the second (type X) is the ergative; and probably all languages embody
some tension between the two. Transitive and ergative are two points
of view on the same phenomenon, that of the nature of material
processes and the relationship of the participants to the process and to
each other; but they are also two distinct phenomena – some processes
pattern ergatively and others transitively (cf. Halliday 1967–68; Davidse
1992). This constitutes another strand in the pattern of changes that
have been taking place in English: type X has tended increasingly to
prevail over type A.
Let us follow this up in a related corner of the grammar. When I
last worked in the United States I was living in Orange County; I
frequently travelled on the local bus services, and there was a notice on
the buses which read:
380
grammar and daily life
If you are standing, on the bus, you are a standee. Why not a stander?
You are a passenger, not a passengee (and if you cannot get on the bus
you may be a bystander); but once you are a standing passenger you
become a standee, and you have your standee line, and must keep
behind it. What kind of participant is construed in the grammar as an -
ee?
There are familiar ones like nominee, trainee, appointee, and more
recent instances of this type like superannuee and oustee, all of which are
modelled on the pattern of employee ‘person employed’. This forms one
term in the transitive opposition employer/employee; the latter form was
derived from the French passive participle and matched up with the
English active termination -er, giving ‘the one who is acting’/’the one
who is acted upon’; cf. trainer/trainee. Here the -ee is functioning as
participant x in type A.
Then there are some instances where rather indirect relationships are
involved: biographee ‘person whose biography is being compiled’, ampu-
tee ‘person who has had a limb amputated’ (note that it does not refer
to the limb; the -ee’s are all human), transplantee (I have a letter
beginning “I am a heart transplantee”), ticketee (in airline parlance); and
various banking terms like advisee, favouree, assignee and so on. These
are modelled on words like referee ‘person to whom a dispute is referred
for decision’, refugee ‘person to whom a place of refuge is offered’.
Then, with escapee ‘person who escapes’ as an early model, we now
have conferee and attendee ‘person attending a conference or lecture’,
retiree ‘person having retired’, and returnee ‘person trying to get back to
original country’. All these are like standee. When we examine them,
we find that they pattern ergatively: the -ee corresponds to the function
of the Medium in the process, to participant x in a process construed
as type X. There is no implication that these are functioning as the
Goal: a standee is not someone who has been or is being stood. If these
were following type A we would have stander, returner, retirer, attender
and so on. The pattern is given in Figure 1:
381
construing and enacting
382
grammar and daily life
Notes
1. Notice on the other hand that in the interrogative this pressure is much
less strong: we readily say does your head ache? is your throat sore? as well as
have you got a headache/a sore throat? This is because in the interrogative the
grammar preempts the thematic slot to signal that the clause is, in fact, a
question, by putting at the beginning the part of the verb that selects for
‘yes or no’, the Finite operator, does/is: does your head ache? signals ‘my
message is concerned with whether it does or not’. As a result there is
relatively little thematic weight left over; the difference in information
flow between is your throat sore? and have you got a sore throat? is very much
less noticeable than that between the agnate declarative pair my throat’s sore
and I’ve got a sore throat, where the full thematic weight is felt on either my
throat or I.
2. Thus the grammar signals metaphorically that meaning is a social process.
We might put this together with the recent neurobiological finding by
Robin Dunbar (1992), that species living in large social groups have
proportionally larger cortices. “Dunbar’s explanation is that large group
sizes demand greater social cohesion and hence more advanced skills for
communicating . . .”.
383
Chapter Fifteen
1 The problem
Most of us are familiar with the feeling that there must be something
odd about linguistics. We recognize this as a problem in the interper-
sonal sphere because as linguists, probably more than other profession-
als, we are always being required to explain and justify our existence.
This suggests, however, that others see it as a problem in the ideational
sphere.
The problem seems to arise from something like the following. All
systematic knowledge takes the form of ‘language about’ some
phenomenon; but whereas the natural sciences are language about
nature, and the social sciences are language about society, linguistics is
language about language – “language turned back on itself ”, in Firth’s
often quoted formulation. So, leaving aside the moral indignation some
people seem to feel, as if linguistics was a form of intellectual incest,
there is a real problem involved in drawing the boundary: where does
language end and linguistics begin? How does one keep apart the
object language from the metalanguage – the phenomenon itself from
the theoretical study of that phenomenon?
The discursive evidence rather suggests that we don’t, at least not
very consistently. For example, the adjective linguistic means both ‘of
language’, as in linguistic variation, and ‘of linguistics’ as in linguistic
association (we never know, in fact, whether to call our professional
bodies linguistic associations or linguistics associations). But a situation
First published in Functional Descriptions: Theory in Practice, 1996, edited by Ruqaiya Hasan,
Carmel Cloran and David G. Butt. Amsterdam: John Benjamins, pp. 1–38.
384
on grammar and grammatics
385
construing and enacting
386
on grammar and grammatics
3 Defining grammar
In the simplest definition grammar is part of language. If we pick up a
book purporting to describe a language, or to help us to learn it, we
expect to find some portion or portions of the book – but not the
whole of the book – devoted to grammar. In my own work, I have
operated with the concept of “lexicogrammar” (that is, grammar and
vocabulary as a single unity), while usually referring to it simply as
grammar for short; this is a stratal concept, with grammar as one among
an ordered series comprising (at least) semantics / lexicogrammar /
phonology. But whatever part–whole model is adopted, language
remains the more inclusive term.
But there is a further step, by which grammar is not just one among
various parts of language; it is a privileged part. The exact nature of
this privilege will be interpreted differently by different linguists, and
some might deny it altogether; but most would probably accept it in
one form or another. I would be inclined to characterize grammar in
the first instance as the part of language where the work is done.
Language is powered by grammatical energy, so to speak.
Let me approach the definition of grammar, however, from a
somewhat different angle. I shall assume here, as a general theoretical
foundation, the account of language given by Lemke (1993). Lemke
characterizes human communities as eco-social systems which persist in
time through ongoing exchange with their environment; and the same
holds true of each of their many sub-systems. The social practices by
which such systems are constituted are at once both material and
semiotic, with a constant dynamic interplay between the two. Note
that by semiotic I mean ‘having to do with meaning’, not ‘having to do
with signs’; thus, practices of doing and practices of meaning. The
important feature of the material–semiotic interplay is that, as Lemke
points out, the two sets of practices are strongly coupled: there is a
high degree of redundancy between them. We may recall here Firth’s
concept of ‘mutual expectancy’ between text and situation.
Underlying the semiotic practices are semiotic systems of various
kinds. In fact, we usually use the term “system” to cover both system
and process: both the potential and the instances that occur; thus a
semiotic system is a meaning potential together with its instantiation in
acts of meaning. Now, one special kind of semiotic system is one that
has a grammar in it: such a system “means” in two phases, having a
distinct phase of wording serving as the base for the construction of
meaning. In other words, its “content plane” contains a grammar as
387
construing and enacting
S S S S S
1 2 3 4.1 4.2
Figure 1 Evolutionary typology of systems
388
on grammar and grammatics
389
construing and enacting
390
on grammar and grammatics
these two modes of meaning together into a single current, such that
everything we say (or write, or listen to, or read) “means” in both
these functions at once. Thus every instance of semiotic practice –
every act of meaning – involves both talking about the world and
acting on those who are in it. Either of these sets of phenomena may
of course be purely imaginary; that in itself is a demonstration of the
constitutive power of a grammar.
6 Grammar as theory
So far I have been talking about various properties of grammar. But in
“talking about” grammar, I have been “doing” grammatics – it is my
discourse that has been construing grammar in this way. Naturally, I
have also been ‘doing’ grammar: the properties have been being
construed in lexicogrammatical terms. In other words I have been
using grammar to construct a theory about itself.
Every scientific theory – in fact every theory of any kind, whether
‘scientific’ or otherwise – is constructed in similar fashion, by means of
the resources of grammar. A theory is a semiotic construct (see Lemke
(1990) for a powerful presentation of this point). That we are able to
use a grammar as a resource for constructing theories is because a
grammar is itself a theory. As I suggested in the previous section, the
grammar functions simultaneously as a mode of knowing and a mode
of doing; the former mode – the construction of knowledge – is the
transformation of experience into meaning. A grammar is a theory of
human experience.
Construing experience is a highly theoretical process, involving
setting up categories and relating each category to the rest. As Ellis
(1993) points out, there are no natural classes: the categories of
experience have to be created by the grammar itself. Or, we might say,
there are indefinitely many natural classes: indefinitely many ways in
which the phenomena of our experience may be perceived as being
alike. In whichever of these terms we conceive the matter, the grammar
has to sort things out, assigning functional value selectively to the
various possible dimensions of perceptual order. The grammar’s model
of experience is constantly being challenged and reinforced in daily life;
thus it tends to change when there are major changes in the conditions
of human existence – not as a consequence, but as a necessary and
integral element, of these changes.
The difference between a grammar, as a “commonsense” theory of
experience, and a scientific theory (such as grammatics) is that grammars
391
construing and enacting
392
on grammar and grammatics
393
construing and enacting
394
on grammar and grammatics
81 systems in it, Mann was probably not far wrong when he estimated
off the cuff that it would generate somewhere between 108 and 109
clause types.
Of course there are lots of mistakes in these complex networks, and
the only way to test them is by programming them and setting them
to generate at random. It is not difficult to generate the paradigm of
selection expressions from a reasonably small network (already in 1966
Henrici developed a program for this purpose; cf. Halliday and Martin
1981), where you can inspect the output and see where it has gone
wrong. But even if the program could list half a billion expressions it
would take a little while to check them over. As far as their overall
capacity is concerned, however, they are probably not orders of
magnitude out.
It has been objected that the human brain could not possibly process
a grammar that size, or run through all the alternative options whenever
its owner said or listened to a clause. I am not sure this is so impossible.
But in any case it is irrelevant. For one thing, this is a purely abstract
model; for another thing, the number of choice points encountered in
generating or parsing a clause is actually rather small – in the network
of the verbal group it took only 28 systems to produce some 70,000
selection expressions, and in any one pass the maximum number of
systems encountered would be even less – probably under half the
total, in a representative network. In other words, in selecting one out
of half a billion clause types the speaker/listener would be traversing at
the most about forty choice points. So although the system network is
not a model of neural processes, there is nothing impossible about a
grammar of this complexity – that is, where the complexity is such that
it can be modelled in this way, as the product of the intersection of a
not very large number of choices each of which by itself is extremely
simple.
395
construing and enacting
like those associated with the recent move into nanotechnology (engin-
eering the very small); but the expansion may take place anywhere in
the lexicogrammar, as new wording, in any form. The grammar is not
simply tagging along behind; technological developments, like other
historical processes, are simultaneously both material and semiotic – the
two modes are interdependent. Early on in his researches into science
and technology in China, Needham noted how in the medieval period,
when there was no adequate institutional mechanism for keeping new
meanings alive, the same material advances were sometimes made two
or three times over, without anyone realizing that the same technology
had been developed before (Needham 1958).
On the other hand, grammars grow by increasing the delicacy in
their construction of existing domains. (This has been referred to by
various metaphors: refining the grid or mesh, sharpening the focus,
increasing the granularity and so on. I shall retain the term “delicacy”,
first suggested by Angus McIntosh in 1959.) This is a complex notion;
it is not equivalent to subcategorizing, which is simply the limiting case
– although also the one that is likely to be the most easily recognized.
The grammar does construct strict taxonomies: fruit is a kind of food, a
berry is a kind of fruit, a raspberry is a kind of berry, a wild raspberry is a
kind of raspberry; these are typically hyponymic and can always be
extended further, with new words or new compositions of words in a
grammatical structure, like the nominal group in English and many
other languages. But greater delicacy is often achieved by intersecting
semantic features in new combinations; and this is less open to casual
inspection, except in isolated instances which happen to be in some
way striking (like certain “politically correct” expressions in present-
day English). The massive semantic innovations brought about by
computing, word processing, networking, multimedia, the information
superhighway and the like, although in part construing these activities
as new technological domains, more typically constitute them as new
conjunctions of existing meanings, as a glance at any one of thousands
of current periodicals will reveal. On a somewhat less dramatic scale,
we are all aware of the much more elaborate variations in the discourse
of environmental pollution and destruction than were available a
generation ago. Even a seemingly transparent piece of wording such as
smoke-free construes a new confluence of meanings; indeed the whole
semogenic potential of -free as a derivational morpheme has recently
been transformed. (Similar expansions have happened with -wise and
-hood.)
There is a special case of this second heading – perhaps even a third
396
on grammar and grammatics
397
construing and enacting
398
on grammar and grammatics
categories must make sense as a whole. And this means that it needs to
be founded on compromise. The grammar of every natural language is
a massive exercise in compromise, accommodating multiple perspect-
ives that are different and often contradictory.
Such compromise demands a considerable degree of indeterminacy
in the system.
10 Indeterminacy in grammar
It seems obvious that grammars are indeterminate (or “fuzzy”, to
borrow the term from its origins in Zadeh’s “fuzzy logic”), if only
because of the effort that goes into tidying them up. Formal logic and
even mathematics can be seen as the result of tidying up the indeter-
minacies of natural language grammars.
The typology of indeterminacy is itself somewhat indeterminate. For
the present discussion I will identify three types: (a) clines, (b) blends,
and (c) complementarities, with (d) probability as a fourth, though
rather different case.
Clines are distinctions in meaning which take the form of continuous
variables instead of discrete terms. The prototype examples in grammar
are those distinctions which are construed prosodically, typically by
intonation (tone contour): for example, in English, “force”, from strong
to mild, realized as a continuum from wide to narrow pitch movement
– if the tone is falling, then from wide fall (high to low) to narrow fall
(midlow to low). But one can include in this category those distinctions
where, although the realizations are discrete (i.e. different wordings are
involved), the categories themselves are shaded, like a colour spectrum:
for example, colours themselves; types of motorized vehicles (car, bus,
van, lorry, truck, limousine . . . etc.); types of process (as illustrated on
the cover of the revised edition of my Introduction to Functional Grammar
1994). In this sense, since in the grammar’s categorization of experience
fuzziness is the norm, almost any scalar set will form a cline: cf. humps,
mounds, hillocks, hills and mountains; or must, ought, should, will, would,
can, could, may, might.
Blends are forms of wording which ought to be ambiguous but are
not. Ambiguity in the strict sense, as in lexical or structural puns, is not
a form of indeterminacy as considered here, because it does not involve
indeterminacy of categorization. Blends also construe two (or more)
different meanings; but the meanings are fused – it is not a matter of
selecting one or the other. A favourite area for blends, apparently in
many languages, is modality; in English, oblique modal finites like
399
construing and enacting
should provide typical examples, for example the brake should be on,
meaning both ‘ought to be’ and ‘probably is’. There is then the further
indeterminacy between an ambiguity and a blend, because a wording
which is clearly ambiguous in one context may be blended when it
occurs in another. A metaphor is the limiting case of a blend.
Complementarities are found in those regions of (typically experi-
ential) semantic space where some domain of experience is construed
in two mutually contradictory ways. An obvious example in English is
in the grammar of mental processes, where there is a regular comple-
mentarity between the “like” type (I like it; cf. notice, enjoy, believe, fear,
admire, forget, resent . . . ) and the “please” type (it pleases me; cf. strike,
delight, convince, frighten, impress, escape, annoy . . .). The feature of
complementarities is that two conflicting proportionalities are set up,
the implication being that this is a complex domain of experience
which can be construed in different ways: here, in a process of
consciousness the conscious being is on the one hand ‘doing’, with
some phenomenon defining the scope of the deed, and on the other
hand ‘being done to’ with the phenomenon functioning as the doer.
All languages (presumably) embody complementarities; but not always
in the same regions of semantic space (note for example the striking
complementarity of tense and aspect in Russian). One favourite domain
is causation and agency, often manifested in the complementarity of
transitive and ergative construals.
Strictly speaking probability is not a “fuzzy” concept; but probability
in grammar adds indeterminacy to the definition of a category. Con-
sider the network of the English verbal group in Figure 2 above. As an
exercise in grammatics this network is incomplete, in that there are
distinctions made by the grammar that the network fails to show: in
that sense, as already suggested, no network ever can be complete. But
it is incomplete also in another sense: it does not show probabilities. If
you are generating from that network, you are as likely to come up
with won’t be taken as with took; whereas in real life positive is
significantly more likely than negative, active than passive, and past
than future. Similarly a typical dictionary does not tell you that go is
more likely than walk and walk is more likely than stroll, though you
might guess it from the relative length of the entries. A grammar is an
inherently probabilistic system, in which an important part of the
meaning of any feature is its probability relative to other features with
which it is mutually defining. Furthermore the critical factor in register
variation is probabilistic: the extent to which local probabilities depart
from the global patterns of the language as a whole; for example a
400
on grammar and grammatics
401
construing and enacting
above – the size (and growth) of the grammar, its trinocular perspective,
and its fuzz – how does the grammatics handle these various parame-
ters? To put this in very general terms: how do we construe the
grammatics so as to be able to manage the complexity of language?
It seems to me that there are certain matching properties. The
grammatics copes with the immense size of the grammar, and its
propensity for growing bigger, by orienting itself along the paradig-
matic axis, and by building into this orientation a variable delicacy; this
ensures that the grammar will be viewed comprehensively, and that
however closely we focus on any one typological or topological domain
this will always be contextualized in terms of the meaning potential of
the grammar as a whole. It copes with the trinocular vision of the
grammar by also adopting a trinocular perspective, based on the stratal
organization of the grammar itself. And it copes with the indeterminacy
of the grammar by also being indeterminate, so that the categories of
the theory of grammar are like the categories that the grammar itself
construes.
Theories in other fields, concerned with non-semiotic systems, begin
by generalizing and abstracting; but they then take off, as it were, to
become semiotic constructs in their own right, related only very
indirectly and obliquely to observations from experience. The proto-
type of such a theory is a mathematical model; and one can theorize
grammatics in this way, construing it as a formal system. But a
grammatics does not need to be self-contained in this same manner. It
is, as theory, a semiotic construct; but this does not create any
disjunction between it and what it is theorizing – it remains permeable
at all points on its surface. The grammatics thus retains a mimetic
character: it explains the grammar by mimicking its crucial properties.
One could say that it is based on grammatical logic rather than on
mathematical logic. In some respects this will appear as a weakness: it
will lack the rigour of a mathematical theory. But in other respects it
can be a source of strength. It is likely to be more relevant to
understanding other semiotic systems: not only verbal art, but also
other, non-verbal art forms, as demonstrated by O’Toole’s masterly
interpretation of painting, architecture and sculpture in terms of sys-
temic grammatics, referred to already (O’Toole 1994). And the new
field of “intelligent computing”, associated with the work of Sugeno,
and explicitly defined by him as “computing with (natural) language”,
requires a theory that celebrates indeterminacy (it is a development of
fuzzy computing) and that allows full play to the interface between
wording and meaning (see section 20 below).
402
on grammar and grammatics
In the next few sections I will make a few observations about these
matching properties of the grammatics, as they seem to me to emerge
in a systemic perspective.
403
construing and enacting
404
on grammar and grammatics
13 A note on delicacy
Inherent in the paradigmatic orientation is the concept of variable
delicacy, in which again the grammatics mimics the grammar: delicacy
in the construal of grammar (by the grammatics) is analogous to
delicacy in the construal of experiential phenomena (by the grammar).
Since for the most part the “lexicalized” mode of realization is
associated with fairly delicate categories in the grammar, we can talk of
“lexis as delicate grammar” (this refers to lexical items in the sense of
“content words”; grammatical items, or “function words”, like the, of,
it, not, as, turn up in the realization of very general systemic features).
But this is not the same thing as saying that when one reaches the stage
of lexical realization one has arrived at the endpoint in delicacy.
What is the endpoint, on the delicacy scale? How far can the
grammatics go in refining the categories of the grammar? In one sense
there can be no endpoint, because every instance is categorially
different from every other instance, since it has a unique instantial
context of situation. We tend to become aware of this when an
instance is codified in the work of a major writer and hence becomes
immortalized as a “quotation”. It seems trivial; but it may not be trivial
in the context of intelligent computing, where the program might need
to recognize that, say, turn left!, as instruction to the car, has a different
meaning – and therefore a different description – at every instance of
its use. This is the sense in which a grammar can be said to be an
“infinite” (i.e. indefinitely large) system. But if we are literate, then in
our commonsense engagements with language, in daily life, we behave
as if there is an endpoint in delicacy: namely, that which is defined by
the orthography. We assume, in other words, that if two instances look
different (i.e. are represented as different forms in writing) they should
be described as different types; whereas if two instances are written
alike they should be described as tokens of the same type – however
delicate the description, it will not tease them apart. The orthography
is taken as the arbiter of paradigmatic boundaries: the way things are
written determines their identity.
There is sense in this: writing represents the unconscious collective
wisdom of generations of speakers/listeners. And we do allow excep-
tions. (a) We recognize homonymy and, more significantly, polysemy,
where the delicacy of categorization does not stop at the barrier created
by the writing system. (b) We accept that there are systematic distinc-
tions which orthography simply ignores: for example, in English, all
those realized by intonation and rhythm. (c) And, as already noted, it
405
construing and enacting
never was assumed, except perhaps among a very few linguists, that a
“function word” like of has only one location in the terrain described
by the grammatics. These exceptional cases challenge the implicit
generalization that the orthographic form always defines a “type”
within the wording.
A more explicit principle could be formulated: that, as far as the
grammatics is concerned, the endpoint in delicacy is defined by what
is systemic: the point where proportionalities no longer continue to
hold. As long as we can predict that a : a⬘ :: b : b⬘ :: . . . , we are still
dealing with types, construed as distinct categories for purposes of
grammatical description.
In practice, of course, we are nowhere near this endpoint in writing
our systemic “grammars”. (I find it disturbing when the very sketchy
description of English grammar contained in Halliday (1994) is taken
as some kind of endpoint. Every paragraph in it needs to be expanded
into a book, or perhaps some more appropriate form of hypertext; then
we will be starting to see inside the grammar – and be able to rewrite
the introductory sketch!) We are only now beginning to get access to
a reasonable quantity of data. This has been the major problem for
linguistics: probably no other defined sphere of intellectual activity has
ever been so top-heavy, so much theory built overhead with so little
data to support it. The trouble was that until there were first of all tape
recorders and then computers, it was impossible to assemble the data a
grammarian needs. Since grammars are very big, and very complex, an
effective grammatics depends on having accessible a very large corpus
of diverse texts, with a solid foundation in spontaneous spoken
language; together with the sophisticated software that turns it into an
effective source of information.
406
on grammar and grammatics
407
construing and enacting
15 Trinocular vision
The “trinocular” principle in the grammatics can be simply stated. In
categorizing the grammar, the grammarian works “from above”, “from
roundabout” and “from below”; and these three perspectives are
defined in terms of strata. Since the stratum under attention is the
lexicogrammar, “from roundabout” means ‘from the standpoint of the
lexicogrammar itself ’. “From above” means ‘from the standpoint of
the semantics: how the given category relates to the meaning (what it
“ ‘realizes” ’)’. “From below” means ‘from the standpoint of morphol-
ogy and phonology, how the given category relates to the expression
(what it “is realized by”)’. What are being taken into account are the
regularities (proportionalities) at each of the three strata.
Since the patterns seen from these three angles tend to conflict, the
resulting description of the grammar, like the grammar’s own descrip-
tion of experience, must be founded on compromise. This is easy to
say; it is not so easy to achieve. Often one finds oneself ‘hooked’ on
one oculation – obsessed, say, with giving the most elegant account of
how some pattern is realized, and so according excessive priority to the
view from below; then, on looking down on it from above, one finds
one has committed oneself to a “system” that is semantically vacuous.
If the view from below is consistently given priority, the resulting
description will be a collapsed grammar, so “flat” that only an impov-
erished semantics can be raised up on it. On the other hand, if one is
biased towards the view from above, the grammar will be so inflated
that it is impossible to generate any output. And if one looks from both
vertical angles but forgets the view from roundabout (surprisingly,
perhaps, the commonest form of trap) the result will be a collection of
isolated systems, having no internal impact upon each other. In this
case the grammar is not so much inflated or collapsed; it is simply
curdled.
Thus the categories of the grammatics, like those of the grammar,
rest on considerations of underlying function, internal organization
(with mutual definition) and outward appearance and recognition. But
there is more than a simple analogy embodied here. I referred above to
the notion of semiotic transformation: that the grammar transforms
408
on grammar and grammatics
16 Indeterminacy in grammatics
That the grammatics should accommodate indeterminacy does not
need explaining: indeterminacy is an inherent and necessary feature of
a grammar, and hence something to be accounted for and indeed
celebrated in the grammatics, not idealized out of the picture – just as
the grammar’s construal of experience recognizes indeterminacy as an
inherent and necessary feature of the human condition.
But construing indeterminacy is not just a matter of leaving things
as they are. Construing after all is a form of complexity management;
and just as, in a material practice such as looking after a wilderness,
once you have perturbed the complex equilibrium of its ecosystem you
have to intervene and actively manage it, so in semiotic practice, when
you transform something into meaning (i.e. perturb it semiotically) you
also have to manage the complexity. We can note how the grammar
manages the complexity of human experience. In the first instance, it
imposes artificial determinacy, in the form of discontinuities: thus, a
growing plant has to be construed either as tree or as bush or as shrub
(or . . .); the line of arbitrariness precludes us from creating intermediate
categories like shrush. Likewise, one thing must be in or on another;
you are either walking or running, and so on. At the same time, however,
each of these categories construes a fuzzy set, whose boundaries are
indeterminate: on and run and tree are all fuzzy sets in this sense.
Furthermore, the grammar explicitly construes indeterminacy as a
semantic domain, with expressions like half in and half on, in between a
bush and a tree, almost running and the like. The specific types of
indeterminacy discussed in Section 10 above, involving complex
relationships between categories, are thus only special cases, fore-
grounding something which is a property of the grammar as a whole.
Now consider the grammatics from this same point of view. The
409
construing and enacting
categories used for construing the grammar – things like noun and
subject and aspect and hypotaxis and phrase – are also like everyday
terms: they impose discontinuity. Either something is a noun or it is a
verb (or . . .); we cannot decide to construe it as a nerb. But, in turn,
each one of these itself denotes a fuzzy set. And, thirdly, the same
resources exist, if in a somewhat fancier form, for making the indeter-
minacy explicit: verbal noun, pseudo-passive, underlying subject, and so
on.
What then about the specific construction of indeterminacy in the
overall edifice constructed by such categories? Here we see rather
clearly the grammatics as complexity management. On the one hand,
it has specific strategies for defuzzifying – for imposing discontinuity
on the relations between one category and another; for example, for
digitalizing the grammar’s clines (to return to the example of “force”,
cited in section 10, it can establish criteria for recognizing a small,
discrete set of contrasting degrees of force). A system network is a case
in point: qualitative relationships both within and between systems may
be ironed out, so that (i) the system is construed simply as a or b
(or . . .), without probabilities, and (ii) one system is either dependent
on or independent of another, with no degrees of partial association.
But, at the same time, the grammatics exploits the various types of
indeterminacy as resources for managing the complexity. I have already
suggested that the concept of lexicogrammmar (itself a cline from
“most grammatical” to “most lexical”) embodies a complementarity in
which lexis and grammar compete as theoretical models of the whole.
There are many blends of different types of structure, for example the
English nominal group construed both as multivariate (configurational)
and as univariate (iterative) but without ambiguity between them. And
the two most fundamental relationships in the grammatics, realization
and instantiation, are both examples of indeterminacy.
I have said that a grammar is a theory of human experience. But that
does not mean, on the other hand, that it is not also part of that
experience; it is. We will not be surprised, therefore, if we find that its
own complexity comes to be managed in ways that are analogous to
the ways in which it itself manages the complexity of the rest. In the
last resort, we are only seeing how the grammar construes itself.
410
on grammar and grammatics
the relationship between the strata; the verb realize faces “upwards”,
such that the “lower” stratum realizes the “higher” one. (Realization is
also extended to refer to the intrastratal relation between a systemic
feature and its structural (or other) manifestation.) Instantiation is the
relationship between the system and the instance; the instance is said to
instantiate the system.
It can be said that, in the elements of a primary semiotic (signs), the
signifier “realizes” the signified; but this relationship is unproblematic:
although the sign may undergo complex transformations of one kind
or another, there is no intermediate structure between the two (no
distinct stratum of grammar). With a higher order semiotic, where a
grammar intervenes, this opens up the possibility of many different
types of realization. It is not necessary to spell these out here; they are
enumerated and discussed in many places (for example Berry 1977;
Fawcett 1980; Martin 1984; Hasan 1987; Matthiessen 1988; Eggins
1994).
But there is another opening-up effect which is relevant to the
present topic: this concerns the nature and location of the stratal
boundary between the grammar and the semantics. This is, of course,
a construct of the grammatics; many fundamental aspects of language
can be explained if one models them in stratal terms, such as metaphor
(and indeed rhetorical resources in general), the epigenetic nature of
children’s language development, and metafunctional unity and diver-
sity, among others. But this does not force us to locate the boundary at
any particular place. One can, in fact, map it on to the boundary
between system and structure, as Fawcett does (system as semantics,
structure as lexicogrammar); whereas I have found it more valuable to
set up two distinct strata of paradigmatic (systemic) organization. But
the point is that the boundary is indeterminate – it can be shifted; and
this indeterminacy enables us to extend the stratal model outside
language proper so as to model the relationship of a language to its
cultural and situational environments.
Instantiation is the relationship which defines what is usually thought
of as a “fact” – in the sense of a physical fact, a social fact and so on.
Facts are not given; they are constructed by the theorist, out of the
dialectic between observation and theory. This has always been a
problem area for linguistics: whereas the concept of a physical principle
became clear once the experimental method had been established – a
“law of nature” was a theoretical abstraction constructed mathemati-
cally by the experimenter – the concept of a linguistic principle has
proved much more difficult to elucidate.
411
construing and enacting
412
on grammar and grammatics
413
construing and enacting
(In that respect the original term “ascriptive”, which I had used earlier
to name this type of process, might better have been retained, rather
than being replaced by “attributive”.) Here too, then, the grammar is
construing a significant aspect of human experience – the perception
of a phenomenon as an instance of a general class – in terms of a
property of language itself, where each act of meaning is an instance of
the systemic meaning potential.
Of course, the boot is really on the other foot: the grammatics is
parasitic on the grammar, not the other way around. It is because of
the existence of clause types such as those exemplified above that we
are able to model the linguistic system in the way we do. The
grammatics evolves (or rather one should say the grammatics “is
evolved”, to suggest that it is a partially designed system) as a meta-
phoric transformation of the grammar itself. This is a further aspect of
the special character of grammatics: while all theories are made of
grammar (to the extent that they can be construed in natural language),
one which is a grammar about a grammar has the distinctive metaphoric
property of being a theory about itself.
19 Centricity
Since the grammatics is a theory about a “logo” system, it is “logo-
centric”, or rather perhaps “semocentric”: its task is to put semiotic
systems in the centre of attention. In the same way, biological sciences
are “bio-centric”: biased towards living things; and so on. I think it is
also a valid goal to explore the relevance of grammatics to semiotic
systems other than language, and even to systems of other types. The
grammatics is also “totalizing”, because that is the job of a theory. Of
course, it focuses on the micro as well as on the macro – the semiotic
weather as well as the semiotic climate; but that again is a feature of
any theoretical activity.
It has always been a problem for linguists to discover what are the
properties of human language as such, and what are features specific to
a given language. The problem is compounded by the fact that there is
more than one way of incorporating the distinction (wherever it is
drawn) into one’s descriptive practice. Firth articulated the difference
between two approaches: “what is being sketched here is a general
linguistic theory applicable to particular linguistic descriptions, not a theory of
universals for general linguistic description” (Firth 1957: 21; Firth’s em-
phasis). I have preferred to avoid talking about “universals” because it
seems to me that this term usually refers to descriptive categories being
414
on grammar and grammatics
treated as if they were theoretical ones. As I see it, the theory models
what is being treated as “universal” to human language; the description
models each language sui generis, because that is the way to avoid
misrepresenting it.
Thus while the theory as a whole is logocentric, the description of
each language is what we might call “glottocentric”: it privileges the
language concerned. The description of English is anglocentric, that
of Chinese sinocentric, that of French gallocentric and so on. (Note
that the theory is not anglocentric; the description of English is.) This
is not an easy aim to achieve, since it involves asking oneself the
question: “how would I describe this language as if English (or other
languages that might get used as a descriptive model) did not exist?”
But it is important if we are to avoid the anglocentric descriptions that
have dominated much of linguistics during the second half of the
century.
In practice, of course, English does exist, and it has been extensively
described; so inevitably people tend to think in terms of categories set
up for English – or for other relatively well-described languages. I have
suggested elsewhere some considerations which seem to me relevant to
descriptive practice (Halliday 1992). As far as my own personal history
is concerned, I worked first of all for many years on the grammar of
Chinese; I mention this here because when I started working on
English people told me I was making English look like Chinese! (It
seems ironic that, now that systemic theory is being widely applied to
Chinese studies, the work of mine most often cited as point of reference
is the descriptive grammar of English.)
In my view an important corollary of the characterological approach
(that is, each language being described in its own terms) is that each
language is described in its own tongue. The protocol version of the
grammar of English is that written in English; the protocol version of
the grammar of Chinese is that written in Chinese; and so on. The
principle of “each language its own metalanguage” is important,
because all descriptive terminology carries with it a load of semantic
baggage from its use in the daily language, or in other technical and
scientific discourses; and this semantic baggage has some metalinguistic
value. This applies particularly, perhaps, to the use of theoretical terms
as metacategories in the description; words such as (the equivalents of)
option, selection, rank, delicacy are likely to have quite significant (but
variable) loadings.
But the principle also helps to guard against transferring categories
inappropriately. Even if descriptive terms have been translated from
415
construing and enacting
English (or Russian, or other source) in the first place, once they are
translated they get relocated in the semantic terrain of the new
language, and it becomes easier to avoid carrying over the connotations
that went with the original. So if, say, the term subject or theme appears
in a description of Chinese written in English, its status is as a translation
equivalent of the definitive term in Chinese. Perhaps one should point
out, in this connection, that there can be no general answer to the
question how much alike two things have to be for them to be called
by the same name!
416
on grammar and grammatics
inferencing with fuzzy sets and fuzzy matching processes. But to use
natural language requires a grammatics: that is, a way of modelling
natural language that makes sense in this particular context. Systemic
theory has been used extensively in computational linguistics; and the
Penman nigel grammar, and Fawcett’s communal grammar, are
among the most comprehensive grammars yet to appear in computa-
tional form (Matthiessen 1991a; Matthiessen and Bateman 1992; Faw-
cett and Tucker 1990; Fawcett, Tucker and Lin 1993). But, more
importantly perhaps, systemic grammatics is not uncomfortable with
fuzziness. That is, no doubt, one of the main criticisms that has been
made of it; but it is an essential property that a grammatics must have
if it is to have any value for intelligent computing. This is an exciting
new field of application; if it prospers, then any grammarian privileged
to interact with Sugeno’s enterprise will learn a lot about human
language, as we always do from applications to real-life challenging
tasks.
Note
1. This is not to question the semiotic achievements of the bonobo chimpan-
zees (cf. Introduction, p. 3). The issue is whether their construal of human
language is an equivalent stratified system with a lexicogrammar at the
language is an equivalent stratified system with a lexicogrammar at the
core.
417
This page intentionally left blank
BIBLIOGRAPHY
419
bibliography
420
bibliography
Colby, B. N. and Colby, L. M. (1980) The Daykeeper: The Life and Discourse
of an Ixil Diviner. Cambridge, MA.: Harvard University Press.
Daneš, F. (1966) ‘A three-level approach to syntax’, Travaux linguistiques de
Prague 1.
Davidse, K. (1992a) ‘A semiotic approach to relational clauses’, Occasional
Papers in Systemic Linguistics 6.
Davidse, K. (1992b) ‘Transitivity/ergativity: the Janus-headed grammar of
actions and events’, in M. Davies and L. Ravelli (eds), Advances in Systemic
Linguistics: Recent Theory and Practice. London: Pinter.
Derewianka, B. (1995) ‘Language development in the transition from childhood to
adolescence: the role of grammatical metaphor.’ Unpublished PhD dissertation,
Macquarie University.
Doughty, P. S., Pearce, J. J. and Thornton, G. M. (1971) Language in Use.
London: Edward Arnold (for the Schools Council).
Dunbar, R. (1992) Summary in ‘Secret life of the brain’, New Scientist 1850
(Supplement 4).
Ebeling, C. L. (1960) Linguistic Units. The Hague: Mouton ( Janua Linguarum
12).
Eco, U. (1983) The Name of the Rose, trans. by W. Weaver. London: Secker
& Warburg. (Italian original Il nome della rosa. Milano: Bompiani.)
Edelman, G. M. (1992) Bright Air, Brilliant Fire: On the Matter of the Mind.
New York: Basic Books.
Eggins, S. (1994) An Introduction to Systemic Functional Linguistics. London:
Pinter.
Eggins, S. and Martin, J. R. (1997) ‘Genres and registers of discourse’, in T.
A. van Dijk (ed.), Discourse: A Multidisciplinary Introduction. London: Sage.
Eiler, M. A. (1979) ‘Meaning and choice in writing about literature: a study of
cohesion in the expository texts of ninth graders.’ Unpublished PhD dissertation,
Illinois Institute of Technology.
Ellis, J. M. (1993) Language, Thought and Logic. Evanston Il.: Northwestern
University Press.
Ellis, J. O. (1966a) Towards a General Comparative Linguistics. The Hague:
Mouton (Janua Linguarum).
Ellis, J. O. (1966b) ‘On contextual meaning’, in C. E. Bazell et al. (eds) In
Memory of J. R. Firth. London: Longman.
Fawcett, R. P. (1980) Cognitive Linguistics and Social Interaction. Heidelberg and
University of Exeter: Julius Groos (Exeter Linguistic Studies 3).
Fawcett, R. P. (1988) ‘What makes a “good” system network good?’, in J. D.
Benson and W. S. Greaves (eds), Systemic Functional Approaches to Discourse.
Norwood, N.J.: Ablex.
Fawcett, R. P. and Tucker, G. H. (1990) ‘Demonstration of GENESYS: a
very large semantically based systemic functional grammar’, Proceedings of
421
bibliography
422
bibliography
423
bibliography
424
bibliography
425
bibliography
426
bibliography
427
bibliography
428
bibliography
429
bibliography
430
bibliography
431
bibliography
432
INDEX
abstraction 38, 41, 52–69, 185, 296, 351, arbitrary 3, 47, 114, 294, 300, 378–9,
411 397
degree of 46–7, 53, 114, 158 art 5, 357, 389, 402
higher order of 185, 296 article 98, 299
level of 220–1 articulation 354, 357, 364, 366
accusative 293 articulatory 7, 324
active 28–9, 400 attributive 185, 229, 243, 276, 336,
and passive 115, 182–3 413
Malinowski’s 227, 236 autostable 358, 362
actor 113, 176, 179–94, 204, 212, 224, auxiliary 28, 268
299 noun 32–4
additive 223 axis 23, 95, 97–8, 100, 109–10, 120,
adjective 55, 108, 187, 302, 336, 342, 163–4, 221, 402–3
384
adjectival 343 benefactive 181
adjunct 47, 56, 96, 98–9, 104, 225, Beneficiary 178–81, 193, 238
238 Bernstein, B. 7–8, 175, 403, 412
adverb 28–9, 95, 180 Bloomfield, L. 65, 219
adverbial 28, 32, 343 Bloomfieldian 4, 65
group 96, 103, 180, 263 Boas, F. 246, 262, 302, 311, 325
adversative 225 bonobo 3
Affected 187–8 branching
Agent 238, 244, 277, 298, 300, 381 singulary 122–5
allophone 68 Bühler, K. 173–4, 226–7, 234
alternative 223
ambiguity 104, 159, 163–4, 185–6, 343, Cantonese 32–4, 202
385, 399–400, 410, 416 Carrier 244, 413
ambiguous 298, 346–8, 385, 399–400 category 12–13, 21–34, 37–72, 95,
apposition 215 97–9, 106, 160–4, 167, 170,
appositive 225 177–9, 186–7, 200–2, 209, 215,
Arabic 223, 292 223–33, 242–3, 284–5, 291–313,
433
index
434
index
435
index
436
index
grammatics 11, 365–6, 369–78, 384–417 indeteminacy 179, 239, 310, 335, 338,
trinocular 402–9 365–6, 392, 399–402, 409–11
graphology 39 indicative 111
Greek 294–6 Indo-European 35, 236
ancient 292, 365 infant 3–4, 304, 323, 388–9
ancient Greece 371 inferential relations 11
alphabet 104 information
flow 184
focus 207–8, 233, 270
Hasan, R. 11, 175, 222–5, 229, 231, 242,
prominence 271
244, 261, 285, 351, 404, 406,
retrieval 170
411–13
structure 192, 216, 233, 266, 269–70,
hearer 199, 205, 207, 301
305, 312
hearer-oriented 199, 207, 240
theory 40, 42, 70, 72
hesitation 205, 337–8, 340
unit 192, 207, 215, 262, 266, 270–2
hierarchy 25–6, 42–4, 56, 59–60,
inherence 302
110–11, 115, 119–24, 166, 213,
instantial 10, 279, 405
228, 242
instantiation 12, 262, 352, 359, 362, 387,
Hjemslev, L. 4, 5, 12, 106, 109, 110,
410, 412–13
112, 219, 236, 262, 301, 312, 354,
interaction 175, 189, 199, 201, 210,
401
216, 227, 230, 310, 377, 382, 390,
Hockett, C. 106, 112, 219, 221
392
Huddleston, R. D. 120, 125, 215, 349
interactional 353, 363
Hudson, R. 204, 349
interlevel 39, 56, 67
hyponomy 12, 226
interpersonal 175, 182, 189, 199–217,
hyponym 226, 282
227–45, 268, 273, 284, 307–8, 311,
hyponymic 396
337, 356, 364, 382, 384, 390, 392,
hypotaxis 107, 266, 302, 327, 332–3,
401, 413
333, 343–4, 362, 410
interrogative 111–13, 189, 191, 200, 233,
hypotactic 213–17, 242, 266, 266,
268, 274
282, 333
intersubjective 245, 354–5
see also parataxis
intonation 55, 114, 123–5, 192–3
contour(s) 205, 217, 238, 270, 399
iconicity 312 pattern 114, 123, 193
ideational 210–11, 216–17, 227–44, 268, of pause 26
298, 308, 311, 348–9, 356, 364, and rhythm 114, 262–3, 405
382, 384, 392, 401 intransitive 28, 164, 187
component 186, 198–200, 208, 237, intricacy 331–7, 341, 343
242
features 243 Jacobson, R. 189, 226, 301
function 175, 177, 189, 193 Jesperson, O. 298–9
meaning 177, 188, 193, 199, 201, 217, juncture 55, 68
229, 295 see also tone
semantics 13
structure 231, 241–2, 244 Labov, W. 222, 350
voice 230 Lamb, S. 5–6, 108, 112, 198, 355, 378,
indefinite 185, 299 383, 389
437
index
438
index
439
index
person 70, 230, 236–7, 268, 337, 353, mental 183–8, 282, 398
378 relational 229, 234, 364, 413
phatic 226 semogenic 227, 355, 364
phoneme 65–6, 68, 71, 96, 219–20, 222 social 201, 227, 386, 407
supra-segmental 68 type 177, 235, 243, 263, 274, 277,
phonetic 26, 71 282, 364, 399
phonetics 39, 324 pronominal 29, 31, 305
phonetic/kinetic 354 pronoun 24, 29, 31, 326, 372
phonic 25, 39, 67 prosodic 67, 71, 115, 205–17, 238,
phonology 3, 6, 33, 37–9, 56, 60, 65–71, 364
104, 120, 206, 219–20, 239, 262, prosodies 206, 232
297, 324, 357, 387, 389, 408 protolanguage 4, 304, 349–51, 353–65,
Chinese 292 378, 389
diachronic 324 protolinguistic 354–64
phonological 32–3, 39, 55, 66–71,
114–15, 208, 213, 231, 239, 293, question 174–6, 189, 191–3, 233, 268
324, 338, 345, 357 question-and-answer 283
phonological-lexical 22 Quirk, P. 9, 371, 407
phonological-morphological 22
prosodic 67, 71 Range 238, 300, 348
see also grammar rank 41–69, 95–103, 115–16, 118–26,
phrasal verb 101, 376 159–66, 212–15, 221–2, 242, 268,
phrase 45, 68, 121, 125–6, 180, 206, 292, 343–4, 415
263, 278, 298, 336, 342, 410 scale 51–69, 123
phylogenetic 355, 360 see also exponence
Pike, K. 106, 174, 211, 239, 242 rankshift(ed) 103–4, 121–4
Pinyin 33 upward rankshift 122–4
plosive 34 realization 44, 46, 52, 108, 112–16, 124,
Plum, G. 327, 362 175, 194, 196–212, 217, 220–1,
plural 295, 307 231, 236–46, 273, 285, 307, 310,
polarity 12, 61, 230, 279, 363, 398 352, 357–9, 362–5, 404–13
polarity carrying element 189, 266 realizational chain 292, 352
polysemy 294, 297, 386, 405 realizational cycles 196–7, 204
possession 327, 371, 373 Recipient 305
possessive 100, 275 recursion 101, 213
postmodifier 342 recursive 45, 102–4, 121, 212–17, 242,
pragmatic 10, 11, 412 279, 343
Prague School 174, 190, 262, 292, 299, Reddy, M. 293–4, 297
311, 371 redundancy 65, 70–2, 295, 356–7, 387
predicate 102, 188, 194, 298–9 reflection 197, 227, 236, 353, 355–6,
predicator 47, 99, 101–2, 113, 238 362, 382, 389
preposition 193, 376–7 reflective 197, 237, 329, 356
prepositional 178, 180, 193, 263, 298, 342 relational 106–7, 227, 241, 277–8
present 70, 102, 110, 274–5, 326, 379 clause(s) 181, 185–6
Priestley, J. B. 234, 245 see also process
process 187–8, 203–5, 212, 224, 238, representational, conative and 226–7,
244, 274, 305, 346 236
440
index
441
index
tense 25, 112, 215, 273–4, 279–80, 300, universal 21–31, 95, 120, 301
310, 326, 365, 375, 379, 393–4, universals 12, 209, 414
400 unmarked 33, 164, 167, 207, 233,
text 7–12, 22–30, 38, 45, 49, 109, 166, 240–1, 264–9, 279, 305
169, 175, 184, 192, 199–201, Urdu 374
207–11, 217–47, 261–87, 294–5, Ure, J. 227, 229, 327
300–11, 324, 328–35, 338–47, 350,
360–3, 386–7, 394, 412 verb 25, 28–30, 51, 95, 167, 178,
text-forming 206–7 180–3, 186–7, 190, 236, 297–302,
text-generation 284 312, 326, 346–8, 376, 394, 410–11
text-like 234 postpositive 28
text-linguistic 222 prepositive 28
texture 207–8, 211, 224–5, 233, 271, pro-verb 28
281 verbal group 47, 51, 70, 102, 243, 263,
Theme 111, 113, 190–4, 206–9, 216, 280, 325, 393–5, 400
228, 230, 233–5, 238–46, 266–8, Vietnamese 34
270–3, 297–300, 303–9, 347, 357, voice 28–9, 182–4
371–8, 416 active 29
Theme-Rheme 209, 228, 233, 262, ergative 28–9
271
(un)predicated 111 WH- 189, 206, 268, 275
Token 292–3, 296–8, 364, 366, 413 Whorf, B. 219, 262, 293, 302–3, 306,
tone 67–8, 164, 192, 205, 270 311, 312, 373, 397
concord 266 word 26–34, 44–5, 51, 55, 58–60, 69,
contour 269, 399 95–104, 106, 116, 121–6, 164, 180,
group(s) 55, 192, 207–8, 266, 269 190, 205–6, 217, 221–2, 239, 243,
key 203 280, 296, 303, 371
tonic 113–14, 126, 192, 270 class(es) 28, 31, 34, 95–6, 104
accent 206, 208, 270 wording(s) 3, 222, 345
prominence 206–7, 270 meaning 197, 231, 293–4, 310,
segment 270 356–7, 402
topic 266 patterns of 219–20, 241
and comment 299 see also lexicogrammar
sentence 211, 228, 234, 243 writing 227–9, 302, 405
Topic 300 expository 225
transcription 262–3, 345 scientific 349
transitivity 176–93, 200, 215, 224, 229, speech vs. 323–51
233–7, 243, 246, 263, 268, 274, theory 227
276, 302, 305, 312, 364, 382 see also language, written
typological 13, 120, 397, 402
typology 9, 33–4, 167, 388, 399 Zadeh, L. A. 399, 416
442