Grammatical metaphor is that type of metaphoric process in which the moving parts […] are not lexical items but grammatical categories. It is the process whereby the congruent form […] is transformed into its metaphorical agnate […] . This is perhaps the single most important source of the power that language has of reconstruing the reality that it has previously put in place.
But given that “system” is simply the name given to the potential that is instantiated in text, when we talk of the system changing what we mean is that the text potential has undergone a shift of some kind or other. This may be quantitative change, a shift in the probabilities of the terms in some grammatical system … ; or it may be qualitative change, as happens when a term disappears altogether … . Quantitative changes may stabilise; or they may move on to become qualitative … .
For the discourse analyst, the main problem has been the lack of any coherent account of the semantic system. A text is a construction of meaning; and while the meaning potential of a language is structured and powered by the lexicogrammar, it is the multidimensional organisation of semantic space that defines the overall topology within which speakers and writers are operating.
For the grammarian (and I should say, by the way, that by “the grammarian” here I understand one who works in lexicogrammar and semantics: the systems thinker on the content plane), the problem has been one that is very simple to state: namely lack of data. The grammarian has never had access to the data that are needed for construing an adequate theory.
… whichever of these rôles we are adopting, we need to observe from both ends. The grammarian, however system-oriented he may be, has to monitor instances of discourse; the discourse analyst, however text-oriented, has to keep an eye on the overall potential. The complementarity means that, unless you shift your angle, you will distort the picture: you cannot know all that is going on if you keep to just one observational perspective.
It is particularly in accounting for different kinds of variation in language — dialect, register and genre, Bernstein’s “code” — that the complementarity becomes important for our understanding; and especially in overcoming those unfortunate disjunctions that we have come to live with, that between linguistics and pragmatics at one level, and that between theory of grammar and theory of discourse at another.
The system observer is viewing language from a greater distance in space and time. Any phenomenon in language can be observed from either end, in its environment within the system or within the text; each stance will afford certain insights — but the phenomenon will look different according to which we choose.
In […] grammar and lexis, […] the complementarity is one of focus, based on the scale, or vector, of delicacy. System and text, on the other hand, form a complementarity of angle, based on the vector of instantiation.
In the most general sense, complementarity is a special form of complexity; one can think of it perhaps as the management of contradiction. … I am using it here as a way of bringing together what are in fact three different types of complexity that we find in language, but which are all based on some kind of contradiction.
In other words, system and text form a complementarity, one that is defined by the vector of instantiation. … What I am here calling system and text are two complementary positions of the observer, two observational perspectives on the phenomenon — in this case, on the phenomenon of language. Thus “text” is the instantiation of the system; or, alternatively, “system” is text potential.
The system and the text are not two different orders of reality … . There can only be one set of phenomena here, not two. But since there are two possible angles from which to approach and observe them, the same question of point of view arises [as with climate and weather]. Consider the phenomenon of functional variation in language. If we look at this from the “instance” end, we will find in our corpus certain groups of texts that are alike in certain respects; we will be able to recognise something that we call a “text type”. If we look at these same phenomena from the “system” end, we will see this as a sub-system, a recognisable alignment of the meaning-making resources of the system; this is what we refer to as a “register”. Either we see a recurrent pattern of linguistic weather, or we see a definable motif within the linguistic climate. Both of these are possible points of view; but they engender rather different modes of explanation.
Like all analogies, that between climate and language is oblique and partial. It is an abstract tool for thinking with, not in any sense a strict proportion. Semiotic systems are not like physical systems. For one thing, an instance of a semiotic system carries value. Instances of physical systems do not. You may prefer one kind of weather to another, but that has no relevance whatsoever to the significance of an instance of that kind of weather to another in relation to the theory of climate. Such an instance is simply to be observed, measured and taken account of like any other. But a text has its own differential value. The value is not something fixed and determinate, in most cases; if it is a recorded text, its value may vary at different times, in different places, or under different conditions.
So the analogy whereby language is to text as climate is to weather is useful to think with — up to a point. It reminds us that there are not two different things at issue here, but only one. What we call “climate”, and what we call “weather”, are the same phenomenon seen from different angles, or different moments of time. So it is with “language” and “text”. But in the case of language, so much misunderstanding has been caused by counterposing these two terms, with language and text being treated as if they were different orders of reality, that I shall not make use of these terms in opposition. Instead I shall talk of “text” and “system”, these being two aspects, or guises, of the single phenomenon “language”: thus, language as system, and language as text. Here I am borrowing “system” from its place in the generalised opposition between system and instance.
In the same — or at least an analogous — sense [as climate being a theory of the weather], language is a theory of the text. Again, it may be a low-level commonsense theory, as when we comment on a particular person’s language, or on the language of advertisers or of politicians. … Or it may be a designed scientific theory, such as we find in learned articles on the grammar or phonology of a given language, or of several languages or even language in general. When this happens we turn the theory itself into a virtual entity and call it “linguistics”, or, in the case of the weather, “climatology”.
I do make a distinction between these two; but it is a difference in point of view, between different angles of vision on the phenomena, not in the phenomena themselves. So we can use either to define the other: “discourse” is text that is being viewed in its sociocultural context, while “text” is discourse that is being viewed as a process of language.
[…] you can, as expected, network through the grammar into the lexis; but what you end up with are not lexical items but lexical features. The lexical item will appear, but it will appear as a conjunct realisation of a number of these terminal features. The features are thus components of the lexical items, but the description differs from a usual componential analysis in two important respects. In the first place, the components are systemic: they are organised in sets of systemic options; and in the second place, more significantly, they are derived by ordered steps in delicacy all the way from the primary grammatical categories.
Each feature (that is, each term in each system) is entered in the network together with its realisation, the contribution it makes to the shape of the final output. In the most general (least delicate) systems, this contribution is typically structural: some change in the functional configuration, like adding new elements or putting them in a certain order. But some items of wording appear as realisations in the network almost from the start: these are the function words of English like the and of and it and and and to. As you move over towards the right, the systems become less general; and more and more of the features get to be lexicalised. First there will be middle order items, those functioning somewhere in between the grammatical and lexical poles: like pronouns, and certain types of conjunction and of adverb; then particular classes of verbs, of adjectives or of nouns.
The vertical axis of the network is unordered; systems are arranged so as to make the two-dimensional representation of the wiring as simple as possible. But the horizontal axis is ordered from left to right on a scale of delicacy, or refinement in detail (referred to in computational grammars as “granularity”). Here the ordering is determined by the interrelations among the various systems, irrespective of how their terms are realised in structure. This is a fundamental consideration: one of the reasons for making the underlying representation paradigmatic was in order to free it from constraints of structure, so that every feature could be located according to its relationship with other features. In other words, describing something, and relating it to everything else (to its agnates), which in a description based on structure are two distinct operations, in a systemic grammar are one and the same operation. Each feature is described by being located in its paradigmatic context.
In a systemic grammar, the most abstract representation takes the form of a system network, modelling the sets of options that the grammar makes available, and the interrelation between one set of options and another: whether they are simultaneous or dependent. The network is a theory of what the speaker can do: what he can mean, at the semantic stratum; what he can say, or “word”, at the stratum of lexicogrammar; what he can say, or “sound”, at the strata of phonology and phonetics. It is a more truly generative grammar, as distinct from the earlier Chomskyan kind of generative grammar which, as Peter Wexler once remarked, should really be called a “contraceptive grammar” since it was all about the enforcement of constraint.
Polarity can be taken as the quintessential example of a grammatical system; it is involved in everything we say — everything that language can turn into meaning. It has a place in all metafunctions — in a sense it is pre-metafunctional; this is why it can be ambivalent, if realised on its own (as yes or no), and can be lexicalised in both ideational and interpersonal combinations (eg allow / forbid; nice / nasty). It provides a model of how meaning gets grammaticalised: no doubt it was central to the evolution of language in the species, though that we can never know; it is the first opposition to appear in the development of language in infancy, in the context of “I want / I don’t want” signalled by sound or gesture.
In principle any meaning can be construed in either perspective. Languages vary, of course, as regards which semantic categories they construe grammatically and which lexically, as well in how the distinction is formally expressed; and any given feature may turn up at different locations along the cline.
Thus in the interpersonal domain, the organisation of meaning into the two regions, the lexical and the grammatical, is less polarised; there is not such a clear demarcation between the general and the particular in the management of human relationships. The two contrasting perspectives are still distinct; but it becomes more apparent that the differences between them is one of depth of focus, not one of discontinuity in the phenomena themselves.
Interpersonally, the same two perspectives come into play. Some interpersonal meanings are highly generalised, like the enactment of dialogic rôles (speech function); Hasan (1992) and Hasan & Cloran (1990) present semantic networks for interpersonal systems of speech function (questions and commands) showing their realisations in the grammar.
With options in the way something is evaluated (“I approve / I disapprove”) or contended (“I agree / I disagree”), the borderline between grammar and lexis is shaded over; systems of appraisal, as described by Martin & White (2005), represent more delicate (more highly differentiated options within the general region of evaluation.
There are then mixed systems where the two perspectives intersect; for example systems of modality, in which the various degrees of probability and usuality have multiple realisations, including those where items which elsewhere function lexically are organised into systemic sets — wordings like certainly, perhaps, I think, I’m convinced and many others.
So the lexicogrammar adopts two contrasting perspectives for construing all this complexity. The one is specific and open-ended; hence flexible, but low in information: this the lexical perspective, good for seeing phenomena as particular. The other is general and systemic: hence high in information, but creating closure: this is the grammatical perspective, good for seeing phenomena as generality. The two perspectives are complementary; any phenomenon can be looked at in terms of either, but they will present two different images of the whole.
Ideationally, the lexicogrammar sorts out the complex world of our surroundings, including the part that consists of our own bodies, our own “selves”. But the things that surround us, and impinge on us, are immense varied and complex. There are particular things; these have to be sorted into classes, by selecting, out of the innumerable ways in which one thing can be like one another, those analogies that seem to be significant. The classes, in turn, are sorted into classes of classes, or taxonomies. But the critical factor in the way things impinge upon us is that of change. As well as things, there are happenings, which give the things their value; the particular happenings also have to be sorted into their classes, but in addition to this they have to be located in the matrix of space and time. At the same time, however, both things and happenings display certain very general features, properties which can be abstracted out of their particulars and construed as features of whole classes of phenomena, or even of all. All happenings have several “moments” in time; all things have certain qualities or quantities; and so on.
Survival depends on both aspects of our existence. Language has to be both a way of thinking, of construing the world, and a way of doing, of acting on and interacting with the people in it. This is in fact another complementarity; language acts out this complementarity, at a more abstract level, by transforming both thinking and doing into meaning. It is this that has shaped the evolution of lexicogrammar.
Every clause in every kind of discourse combines an ideational and an interpersonal strand of meaning. But weaving these two together is an extremely complicated task; and in managing all this complexity, language evolved a third metafunctional component, by which it is enabled to organise itself in the form of discourse. Discourse is what linguists perceive as text; so we call this the textual metafunction. This is what enables a language to function meaningfully in all its multifarious contexts.
But the principle is usually formulated in terms of just one aspect of this relationship, the ideational: the way our language functions to construe our own experience of the world. Yet the human condition is a social one, and our environment is not just ecological — it is eco-social; so our language functions also interpersonally. And here it is performing a more active rôle: it is not just construing but enacting our relationships with our fellow creatures. In terms of the classical model deriving from Plato, there is the “first–&–second person” component of language, that which is organised around the interaction of “you” and “me”, as well as the “third person” component of language, that which is organised around the construal of others, of “him, her, it, them”. These two metafunctional components are at the foundation of every human language, and both evolved under the same conditions. Our social interactions increased in complexity just as did our relations with all aspects of the environment. Neither component can be instantiated without the other.
I have always held the view that language is a probabilistic system. Word frequency is the manifestation of probability in the lexicogrammar — or rather, of one particular aspect of it. Another, more powerful aspect of lexicogrammatical probability is the relative probability of the terms in a grammatical system. A grammatical system, such as polarity, or mood, or tense, is a small set of contrasting options with a defined condition of entry — but it is more than that: it is a set of options each having a certain measure of probability.
A grammar is an evolved system, and evolved systems readily admit complementarity; a grammatics is a designed system, and designed systems tend to be hostile to it (it’s too messy). Twentieth century linguistics has been a monument to either/or-ism; which is one reason perhaps why we still do not understand too well how natural semiotic systems work.
Time is both linear and aspectual (tense and aspect)Process is both extension and causation (transitive and ergative)Matter is both discrete and concrete (count and mass)Causality is both a relation between processes and a single complex process (clause nexus and clause)