Trust the Text

Language, Corpus and Discourse

John Sinclair, Ronald Carter

• Publisher: Routledge
• Year: 2004
• Print ISBN: 9780415317672
• eBook ISBN: 9781134369911
• Topic: Sprachen & Linguistik
• Subtopic: Sprachwissenschaft

John Sinclair is one of the major figures in applied linguistics and his work is essential study for students. This accessible book collects in one volume Sinclair's key papers on written discourse structure, lexis patterns, phraseology, corpus analysis, lexicography and linguistic theory from the 1990s. All the papers have been edited and updated for this book. The clear and accessible introduction helps students to navigate his key themes and arguments, making the volume an ideal companion for those coming to Sinclair's more recent writings for the first time.

Part I

Foundations

1 Trust the text

This chapter was edited from a transcript of a keynote lecture given at the Seventeenth International Systemic Conference in Stirling, Scotland, in July 1990. Although not the earliest paper chronologically in this collection, it brings up to date previous work in discourse analysis and maps out the approach to corpora which is now known as ‘corpus-driven linguistics’. In discourse analysis the chapter sets up a model of structural cohesion that applies equally to spoken and written discourse, and concentrates on written discourse. The background to this project is set out in Chapters 3 and 4, and the written analysis is pursued in Chapter 5. In the second half of the chapter it is argued that the received theories of language are not adequate to account for the mass of new evidence coming from early corpus study. This line of research is followed up in Part III – chapters 8–12 inclusive.

By way of a sub-title to this chapter, I should like to quote a short sentence from an article in The European (1–3 June 1990), by Randolph Quirk.

The implications are daunting.

I shall refer to the discourse function of this sentence from time to time, but at present I would like to draw attention to its ominous tone. The implications of trusting the text are for me extremely daunting, but also very exciting and thought-provoking.

The argument that I would like to put forward is that linguistics has been formed and shaped on inadequate evidence and in a famous phrase ‘degenerate data’. There has been a distinct shortage of information and evidence available to linguists, and this gives rise to a particular balance between speculation and fact in the way in which we talk about our subject. In linguistics up till now we have been relying very heavily on speculation.

This is not a criticism; it is a fact of life. The physical facts of language are notoriously difficult to remember. Some of you will remember the days before tape recorders and will agree that it is extremely difficult to remember details of speech that has just been uttered. Now that there is so much language available on record, particularly written language in electronic form, but also substantial quantities of spoken language, our theory and descriptions should be re-examined to make sure they are appropriate. We have experienced not only a quantitative change in the amount of language data available for study, but also a consequent qualitative change in the relation between data and hypothesis. In the first part of this chapter I hope to raise a point about description based on the appreciation of this fairly fundamental appraisal.

Apart from the strong tradition of instrumental phonetics, we have only recently devised even the most rudimentary techniques for making and managing the recording of language, and even less for the analysis of it. In particular we should be suspicious of projecting techniques that are suitable for some areas of language patterning on to others.

This is my first point. Until recently linguistics has been able to develop fairly steadily. Each new position in the major schools has arisen fairly naturally out of the previous one. However, the change in the availability of information which we now enjoy makes it prudent for us to be less confident about reusing accepted techniques.

My second main point is that we should strive to be open to the patterns observable in language in quantity as we now have it. The growing evidence that we have suggests that there is to be found a wealth of meaningful patterns that, with current perspectives, we are not led to expect. We must gratefully adjust to this new situation and rebuild a picture of language and meaning which is not only consistent with the evidence but also exploits it to the full. This will take some time, and the first stage should be an attempt to inspect the data with as little attention as possible to theory.

It is impossible to study patterned data without some theory, however primitive. The advantage of a robust and popular theory is that it is well tried against previous evidence and offers a quick route to sophisticated observation and insight. The main disadvantage is that, by prioritizing some patterns, it obscures others. I believe that linguists should consciously strive to reduce this effect until the situation stabilizes.

The first of my points takes us into the present state of the analysis of discourse, which is now some 20 years old and worth an overhaul; the second plunges us into corpus linguistics, which, although even more venerable, has been rather furtively studied until becoming suddenly popular quite recently. They might seem to have very little in common, but for me they are the twin pillars of language research.

What unites them is:

1. They both encourage the formulation of radically new hypotheses. Although they can be got to fit existing models, that is only because of our limited vision at present.
2. The dimensions of pattern that they deal with are, on the whole, larger than linguistics is accustomed to. Both to manage the evidence required, and even to find some of it in the first place, there is a need to harness the power of modern computers.

The most important development in linguistic description in my generation has been the attempt from many different quarters to describe structures above the sentence and to incorporate the descriptions in linguistic models. The study of text, of discourse, including speech acts and pragmatics, is now central in linguistics. Since the early 1950s a number of approaches have been devised that attempt to account for larger patterns of language. Although large-scale patterns are clearly affected by, for example, sociological variables, they still lie firmly within the orbit of linguistic behaviour for as long as linguistic techniques can be used as the basis of their description.

No doubt we quite often begin a new study by projecting upwards the proven techniques of well described areas of language. To give an example, consider distributional techniques of description which began in phonology. These led in the early 1950s to attempts by, for example, Zellig Harris, to describe written text using essentially the same methods, by looking for repeated words and phrases which would form a basis for classifying the words and phrases that occur next to them. This is just the way in which phonemes were identified and distinguished from allophones, the basis of the famous ‘complementary distribution’. Now there are only a relatively small number of phonemes in any language, numbered in tens, and there are a relatively large number of words, numbered in tens of thousands. The circumstances are quite different, and in the pre-computer era this kind of research faced very serious problems. The unlikelihood of finding exactly repeated phrases led Harris to the idea that stretches of language which, though physically different, were systematically related, could be regarded as essentially the same. This was articulated as grammatical transformation. It is an object lesson in what can go wrong if you project your techniques upwards into other areas without careful monitoring and adaptation. In the event, transformations provided the key feature with which Chomsky (1957) launched a wave of cognitive, non-textual linguistics.

Discourse study took off when speech acts (Austin 1962) were identified in philosophy. It took a development in a discipline outside linguistics to offer a reconceptualization of the function of the larger units of language. However, much of the description of discourse since then has been the upward projection of models, worked out originally for areas like grammar and phonology. I cheerfully admit mea culpa here, in having projected upwards a scale and category model in an attempt to show the structure of spoken interaction (Sinclair et al. 1972a). It has been a serviceable model, and it is still developing, along lines which are now suitable for capturing the general structure of interactive discourse. Recent work on conversation by Amy Tsui (1986), on topic by Hazadiah Mohd Dahan (1991) and by others incorporating the relations between spoken and written language are continuing within the broad umbrella of that model while making it more convenient as a vehicle for explaining the nature of interaction in language.

Louise Ravelli’s study of dynamic grammar (1991) is an interesting exercise in turning the new insights of a theoretical development back on to familiar ground. It is in effect a projection downwards from the insights of discourse into some aspects of language form.

While using familiar tools is a reasonable tactic for getting started, we should also work towards a model of discourse which is special to discourse and which is not based upon the upward projection of ...

Table of contents

• Cover Page
• Title Page
• Copyright Page
• Introduction
• Part I: Foundations
• Part II: The Organization of Text
• Part III: Lexis and Grammar
• Notes
• References