Chapter One
Introduction: Bringing Things into Focus
1.1 The field of scrutiny
In its comparatively short lifespan, generative linguistics has conventionally concentrated on a range of data whose boundaries are restricted to the single sentence. Thus, faced with such sentences as those in (1), generative linguists have had no hesitation in recognizing each and every one as perfectly acceptable1:
1.(a) Most people enjoy brandy after dinner
(b) Most people ENJOY [brandy] after [dinner]
(c) [Brandy] is [enjoyed] by most people after dinner
(d) It's BRANDY that [most] people [enjoy] after [dinner]
(e) What MOST people [enjoy] after [dinner] is BRANDY
(f) [Brandy], most people enjoy after dinner
On the other hand, there are sequences that would often be regarded either as unacceptable outputs of sentence-grammar2 (henceforth, S-grammar), or else as possible outputs, but only at the expense of a considerable machinery of special rules and constraints:
From the viewpoint of the autonomous S, i.e. the single-sentence output of an S-grammar, the examples in (2) show two kinds of difficulty:
a. where there is a syntactic infelicity of some kind, that is, a sequence which cannot be generated in a standard grammar; examples include unorthodox word-orders such as (2a), and deviant complementisation as in (2c,d)
b. where there is an information-gap of some kind: particularly the zero-anaphora type of example, as in (2b), but also to a lesser extent, the indeterminate pronoun, as in (2c,d). The latter type presents problems of interpretation rather than derivation.
However, in certain circumstances, the judgments indicated by (*) can be reversed. These circumstances always involve specific CONTEXTS. Note that in all cases, it is the second S with which we shall be concerned.
3. (a) The Russians invaded Afghanistan - @Most people enjoy brandy after dinner
(b) George enjoys malt whisky after dinner -@[Most] people ENJOY [brandy] after [dinner]
(c) What do people like to drink after dinner? - @[Brandy] is [enjoyed] by most people after dinner
(d) Most people enjoy brandy with their breakfast- @It's BRANDY that [most] people [enjoy] after [dinner]
(e) Most people enjoy brandy with their breakfast - @What MOST people [enjoy] after [dinner] is BRANDY
(f) Tell me about people's drinking habits here — @(Brandy], most people enjoy after dinner
4. (a) You know Mrs. Goldstein said she was going to have a brandy? Well, [have] a [brandy] she DID
(b) George, you bring the soup. JOHN, the FISH
or: George ordered the steak; JOHN, the FISH
(c) Don't try looking for the artist's signature on the front of the painting. You'll [find] it BEHIND
(d) Do you want me to throw the parcel on to the table or put it there? - PUT it
Perhaps we should complete this comparison by suggesting appropriate contexts for the responses in (1) and the cues in (3):
5. (a) Tell me about people's drinking habits here - Well, most people enjoy brandy after dinner
(b) You don't have to FORCE that Remy Martin down you. After all, most people ENJOY [brandy] after [dinner]
(c) The brandy-producers have recently sponsored a market survey of their product. Apparently, [brandy] is [enjoyed] by most people after dinner
(d) Really? I heard there'd been a change to VODKA - No, no: it's BRANDY that [most] people [enjoy] after [dinner]
(e) And I've also read somewhere that after dinner, cocoa is gaining in popularity. - Nevertheless, what MOST people [enjoy] after [dinner] is still BRANDY
(f) Do Belgians drink a lot of brandy? —[Brandy], most people here [enjoy] after dinner
6. (a) The Russians invaded Afghanistan. They claimed to have been invited in by the regine
(b) George enjoys malt whiskey after dinner. Most people [enjoy] BRANDY after [dinner]
(c) Most people enjoy brandy after dinner However, rising prices are preventing them from INDULGING in this luxury
(d) Most people enjoy brandy with their breakfast- That's not true: what MOST people [have] with their [breakfast] is COFFEE
(e) Most people enjoy brandy with their breakfast - No, it's DINNER that [most] people [enjoy] [brandy] with
(f) Tell me about people's drinking habits here Most people enjoy brandy after dinner
Given the appropriate or inappropriate context, therefore, perfectly grammatical (that is, S-grammatical) sentences may be rendered unacceptable (that is, D-ungrammatical). On the other hand, S-ungrammatical, or S-semi-grammatical, sequences may turn out to be perfectly D-grammatical. One of the aims of this study, then, is to provide an explanation for such facts as these, and indeed to attempt what amounts to a definition of the term "appropriateness".
What we are investigating here is essentially the semantics of discourse. This crucially concerns what we conceive of as the "flow" of information through a linguistic sequence, and how it is deployed and manipulated by contextual factors. To put it simply: why do particular types of context demand particular forms of successive sentence structure? Why are particular linguistic forms appropriate to particular contexts, and inappropriate to others?
The answers to such questions are not simple, and moreover, they are extraordinarily far-reaching in terms of their consequences for the theory and practice of grammar. We shall maintain in the course of this book that the restrictions involved are essentially semantic-pragmatic in nature, though actually manifesting themselves at all linguistic levels:
- phonetic/phonological (e.g. stress, intonation)
- syntactic (structural variation)
- semantic (information-structure, reference, deixis)
- pragmatic (interaction, implicature).
1.2 A programme
1.21 What do we know about texts?
"We" in this section heading refers not to "we native speakers of English", or some other language - who in a sense "know" everything there is to know about the grammar of their discourse. Nor does it refer to "we the author", but to "we the community of linguists". Surprisingly, perhaps, the short answer to the question is "very little"; we are hardly in a position to do much more than chart our ignorance. However, this rather glib response is also true of almost every other area of language, a fact which is presumably due to the complexity of the phenomena themselves as well as the inaccessibility of the biological and psychological systems underlying them. Yet it is also true of trans-sentential studies (as of semantics, for example, too) that, historically speaking, they have been even further hindered by the structuralist diktat, due to Bloomfield originally, that only the structural elements of sentences form the proper concern of linguistics.
Likewise, the generativist legacy has been equally stifling, with its insistence that sentence-sequences fall under the heading of performance. Obviously, it is true that sentence-sequences (or rather, utterance-sequences) are performed; but this is no less true of individual sentences (or rather, utterances). Just as a sentence, with its grammatical description etc., is the more abstract representation of an utterance, so is a text (a sequence of sentences) the more abstract representation of a discourse (a sequence of utterances). Thus, utterances and discourses are performed, in real time and in actual situations, but sentences and texts are their abstract symbolic representations. The generativist position, then, in regarding sentential facts as representational, and part of competence, but suprasentential facts as phenomenal, hence part of performance, is thereby confounding two levels of abstraction.
If we can therefore set aside these traditional objections to our field of study, perhaps we can attempt to formulate a consensus view of the commonly accepted assumptions underlying the study of text and discourse. Let us consider the following:
- - Sequences of utterances (and the sentence-sequences underlying them) are not simply random collections.
- - They display connections which are both syntactic and semantic-pragmatic in nature.
- - They occur in relation to practical situations.
(A) is of course the sine qua non of text-study, and in consequence it might seem trivial to spell it out. However, centuries of traditional language study based on sentences, together with the anti-text strictures of structuralism and generativism, have combined so powerfully as to foster the popular presupposition that grammar IS sentence-grammar. Sentence-sequencing is then dealt with, if at all, under some other branch of study, be it rhetoric, macrolinguistics, or performance. Often enough, indeed, its study is postponed indefinitely. (A) says, then, that an S-grammar alone is insufficient to account for the non-random character of sentence-sequences.
(B) introduces an assumption about the substance of this non-random character. As it stands, (B) is neutral on the question of the relative importance of syntax vis-a-vis semantics, all gradations of which have been argued for3. Our own contribution to this debate will argue that many apparent syntactic characteristics are profoundly semantic-pragmatic in nature.
(C) is probably at one and the same time the most obvious and the most problematical of these assumptions. Clearly, an essential component of the production and comprehension of discourses and the investigation of texts is the wider context in which they occur and/or relate to. Yet at the same time, our ability as linguists to analyse, categorise and account for situations in general terms is extremely limited.
Nevertheless, this shortcoming is in itself no excuse for rejecting the study of discourse/text in principle (as do, for example, Katz and Postal 1964). Recent work on this area of experience has, moreover, produced some very encouraging lines of enquiry. We may cite the work of Minsky 1975 and Charniak 1975 on "frames", of Schank and Abelson 1977 on "plans" and "scripts", of Lakoff 1977 on "linguistic gestalts", and of Fillmore 1982 on "frame semantics" and "idealised cognitive models". In the spirit of this work, let us now, for the time being, take over the term "frame", defining it as follows: frames are to actual situations as texts are to discourses and sentences are to utterances. In other words, texts occur in frames, while discourses occur in actual situations: see Fig. 1.1.
1.22 What do we need to find out?
Let us assume the positions A, B, and C outlined in the previous section: we may now call them sequentiality, connectivity, and contextuality, respectively. We return to their more detailed description in Chs. 2 and 3, but for now we shall confine ourselves to outlining the further issues they raise.
Since sequentiality (or syntagmaticity) is also a basic property of syntax, one question we shall ask is, how compatible is the study of texts with extant S-based theories of grammar? We will argue that the interconnections between sentence, text and context should be used to test the explanatory power of this or that model of S-grammar.
Connectivity (or COHERENCE) is, we shall claim, the single most important principle of textuality. However, we need to establish exactly what the ground-rules of connectivity are; we also need to know what the machinery is by which it takes place.
Finally, we want to investigate how texts plug in to frames. What, in other words, is the machinery of contextuality?
To summarise, therefore:
- - For sequentiality: how compatible is text -linguistics with sentence-linguistics?
- - For connectivity: what are the principles of sentence-connectedness; and what is the machinery of this relationship?
- - For contextuality: what are the principles and mechanisms of the text/frame interface?
1.3 Defining terms
The term FOCUS is used in this book mainly as a non-technical term r...