1 Introduction: linguistic forms and functions
The functions of language
The analysis of discourse is, necessarily, the analysis of language in use.
As such, it can not be restricted to the description of linguistic forms independent of the purposes or functions which those forms are designed to serve in human affairs.
While some linguists may concentrate on determining the formal properties of a language, the discourse analyst is committed to an investigation of what that language is used for.
While the formal approach has a long tradition, manifested in innumerable volumes of grammar, the functional approach is less well documented.
Attempts to provide even a general set of labels for the principal functions of language have resulted in vague, and often confusing, terminology.
We will adopt only two terms to describe the major functions of language and emphasise that this division is an analytic convenience.
It would be unlikely that, on any occasion, a natural language utterance would be used to fulfil only one function, to the total exclusion of the other.
That function which language serves in the expression of ' content ' we will describe as transactional, and that function involved in expressing social relations and personal attitudes we will describe as interactional.
Our distinction, ' transactional / interactional ', stands in general correspondence to the functional dichotomies  ' representative / expressive ', found in Bhler (1934), ' referential / emotive ' (Jakobson, 1960), ' ideational / interpersonal ' (Halliday, 1970b) and ' descriptive / social-expressive ' (Lyons, 1977),
The transactional view
Linguists and linguistic philosophers tend to adopt a limited approach to the functions of language in society.
While they frequently acknowledge that language may be used to perform many communicative functions, they nonetheless make the general assumption that the most important function is the communication of information.
Thus Lyons (1977: 32) observes that the notion of communication is readily used ' of feelings, moods and attitudes' but suggests that he will be primarily interested in ' the intentional transmission of factual, or propositional, information '.
Similarly Bennett (1976: 5) remarks' it seems likely that communication is primarily a matter of a speaker's seeking either to inform a hearer of something or to enjoin some action upon him '.
The value of the use of language to transmit information is well embedded in our cultural mythology.
We all believe that it is the faculty of language which has enabled the human race to develop diverse cultures, each with its distinctive social customs, religious observances, laws, oral traditions, patterns of trading, and so on.
We all believe, moreover, that it is the acquisition of written language which has permitted the development within some of these cultures of philosophy, science and literature (see Goody, 1977).
We all believe that this development is made possible by the ability to transfer information through the use of language, which enables man to utilise the knowledge of his forebears, and the knowledge of other men in other cultures.
We shall call the language which is used to convey ' factual or propositional information ' primarily transactional language.
In primarily transactional language we assume that what the speaker (or writer) has primarily in mind is the efficient transference of information.
Language used in such a situation is primarily ' message oriented '.
It is important that the recipient gets the informative detail correct.
Thus if a policeman gives directions to a traveller, a doctor tells a nurse how to administer medicine to a patient, a householder puts in an insurance claim, a shop assistant explains the relative merits of two types of knitting wool, or a scientist describes an experiment, in each case it matters that the speaker should make what he says (or writes) clear.
There will be unfortunate (even disastrous) consequences in the real world if the message is not properly understood by the recipient.
The interactional view
Whereas linguists, philosophers of language and psycho-linguists have, in general, paid particular attention to the use of language for the transmission of ' factual or propositional information ', sociologists and sociolinguists have been particularly concerned with the use of language to establish and maintain social relationships.
In sociological and anthropological literature the phatic use of language has been frequently commented on  particularly the conventional use of language to open talk-exchanges and to close them.
Conversational analysts have been particularly concerned with the use of language to negotiate role-relationships, peer-solidarity, the exchange of turns in a conversation, the saving of face of both speaker and hearer (cf.
Labov, 1972a; Brown and Levinson, 1978; Sacks, Schegloff &amp; Jefferson, 1974; Lakoff, 1973).
It is clearly the case that a great deal of everyday human interaction is characterised by the primarily interpersonal rather than the primarily transactional use of language.
When two strangers are standing shivering at a bus-stop in an icy wind and one turns to the other and says' My goodness, it's cold ', it is difficult to suppose that the primary intention of the speaker is to convey information.
It seems much more reasonable to suggest that the speaker is indicating a readiness to be friendly and to talk.
Indeed a great deal of ordinary everyday conversation appears to consist of one individual commenting on something which is present to both him and his listener.
The weather is of course the most quoted example of this in British English.
However a great deal of casual conversation contains phrases and echoes of phrases which appear more to be intended as contributions to a conversation than to be taken as instances of information-giving.
Thus a woman on a bus describing the way a mutual friend has been behaving, getting out of bed too soon after an operation, concludes her turn in the conversation by saying
Aye, she's an awfy woman.
(awfy = Sc awful)
This might be taken as an informative summary.
Her neighbour then says reflectively (having been supportively uttering aye, aye throughout the first speaker's turn):
Aye, she's an awfy woman.
Pirsig (1976: 313) remarks of such a conversation: ' the conversation's pace intrigues me.
It isn't intended to go anywhere, just fill the time of day... on and on and on with no point or purpose other than to fill the time, like the rocking of a chair. '
What seems to be primarily at issue here is the sharing of a common point of view.
Brown &amp; Levinson point out the importance for social relationships of establishing common ground and agreeing on points of view, and illustrate the lengths to which speakers in different cultures will go to maintain an appearance of agreement, and they remark ' agreement may also be stressed by repeating part or all of what the preceding speaker has said ' (1978:117)
Whereas, as we shall note, written language is, in general, used for primarily transactional purposes, it is possible to find written genres whose purpose is not primarily to inform but to maintain social relationships-'thank you' letters, love letters, games of consequences, etc.
Spoken and written language
Manner of production
From the point of view of production, it is clear that spoken and written language make somewhat different demands on language-producers.
The speaker has available to him the full range of ' voice quality ' effects (as well as facial expression, postural and gestural systems).
Armed with these he can always override the effect of the words he speaks.
Thus the speaker who says' I'd really like to ', leaning forward, smiling, with a ' warm, breathy ' voice quality, is much more likely to be interpreted as meaning what he says, than another speaker uttering the same words, leaning away, brow puckered, with a ' sneering, nasal ' voice quality.
These paralinguistic cues are denied to the writer.
We shall generally ignore paralinguistic features in spoken language in this book since the data we shall quote from is spoken by co-operative adults who are not exploiting paralinguistic resources against the verbal meanings of their utterances but are, rather, using them to reinforce the meaning.
Not only is the speaker controlling the production of communicative systems which are different from those controlled by the writer, he is also processing that production under circumstances which are considerably more demanding.
The speaker must monitor what it is that he has just said, and determine whether it matches his intentions, while he is uttering his current phrase and monitoring that, and simultaneously planning his next utterance and fitting that into the overall pattern of what he wants to say and monitoring, moreover, not only his own performance but its reception by his hearer.
He has no permanent record of what he has said earlier, and only under unusual circumstances does he have notes which remind him what he wants to say next.
The writer, on the contrary, may look over what he has already written, pause between each word with no fear of his interlocutor interrupting him, take his time in choosing a particular word, even looking it up in the dictionary if necessary, check his progress with his notes, reorder what he has written, and even change his mind about what he wants to say.
Whereas the speaker is under considerable pressure to keep on talking during the period allotted to him, the writer is characteristically under no such pressure.
Whereas the speaker knows that any words which pass his lips will be heard by his interlocutor and, if they are not what he intends, he will have to undertake active, public ' repair ', the writer can cross out and rewrite in the privacy of his study.
There are, of course, advantages for the speaker.
He can observe his interlocutor and, if he wishes to, modify what he is saying to make it more accessible or acceptable to his hearer.
The writer has no access to immediate feedback and simply has to imagine the reader's reaction.
It is interesting to observe the behaviour of individuals when given a choice of conducting a piece of business in person or in writing.
Under some circumstances a face-to-face interaction is preferred but, in others, for a variety of different reasons, the individual may prefer to conduct his transaction in writing.
Whereas in a spoken interaction the speaker has the advantage of being able to monitor his listener's minute-by-minute reaction to what he says, he also suffers from the disadvantage of exposing his own feelings ('leaking'; Ekman &amp; Friesen, 1969) and of having to speak clearly and concisely and make immediate response to whichever way his interlocutor reacts.
The representation of discourse: texts
So far we have considered in very general terms some of the differences in the manner of production of writing and speech.
Before we go on to discuss some of the ways in which the forms of speech and writing differ, we shall consider, in the next two sections, some of the problems of representing written and spoken language.
We shall place this within a general discussion of what it means to represent ' a text '.
We shall use text as a technical term, to refer to the verbal record of a communicative act.
(For another approach to text cf. discussion in Chapter 6.)
Written texts
The notion of ' text ' as a printed record is familiar in the study of literature.
A ' text ' may be differently presented in different editions, with different type-face, on different sizes of paper, in one or two columns, and we still assume, from one edition to the next, that the different presentations all represent the same ' text '.
It is important to consider just what it is that is' the same '.
Minimally, the words should be the same words, presented in the same order.
Where there are disputed readings of texts, editors usually feel obliged to comment on the crux; so of Hamlet's
O, that this too too sullied flesh would melt (1.
ii.129)
Dover Wilson makes it clear that this is an interpretation, since the second Quarto gives' too too sallied ' and the first Folio ' too too solid ' (Dover Wilson, 1934).
Even where there is no doubt about the identity of words and their correct sequence, replicating these alone does not guarantee an adequate representation of a text.
Consider the following extract of dialogue from Pride and Prejudice:
' Mr. Bennet, how can you abuse your own children in such a way?
You take delight in vexing me.
You have no compassion on my poor nerves. '
' You mistake me, my dear.
I have a high respect for your nerves.
They are my old friends.
I have heard you mention them with consideration these twenty years at least. '
It is clear that more than simply reproducing the words in their correct order is required.
It is necessary to replicate punctuation conventions, as well as the lineation which indicates the change of speaker.
The extract reads as gobbledygook if it is read as a speech by one individual.
An adequate representation of a text must assign speeches to the correct characters, sentences to the correct paragraphs, and paragraphs to the correct chapters.
The author's organisation and staging of his work must be preserved.
In a piece of expository prose, the author's indication of the development of the argument contributes to the reader's experience of the text.
Thus titles, chapter headings, sub-divisions and sub-headings all indicate to the reader how the author intends his argument to be chunked.
The detail of lineation rarely matters in expository or descriptive prose.
However it clearly becomes crucial in the reproduction of poetry.
The work of those seventeenth-century poets who created poems in the shape of diamonds or butterflies would be largely incomprehensible if the form were not preserved.
The notion of ' text ' reaches beyond the reproduction of printed material in some further printed form.
A letter, handwritten in purple ink with many curlicues, may have its text reproduced in printed form.
Similarly, neutral printed versions may be produced of handwritten shopping lists, slogans spray-painted on to hoardings, and public notices embossed on metal plates.
In each case the ' text ' will be held to have been reproduced if the words, the punctuation and, where relevant, the lineation are reproduced accurately.
Where the original text exploits typographical variety, a text reproduced in one type-face may lack some of the quality of the original.
An obvious example is a newspaper item which may exploit several different type-faces, different sizes of type and a particular shape of lay-out.
It is interesting to observe that publishers regularly reproduce conscious manipulation of the written medium on the part of the writer.
Thus Jane Austen's expression of contrast is reproduced by publishers in italics:
' Nay, ' said Elizabeth, ' this is not fair.
You wish to think all the world respectable, and are hurt if I speak ill of any body.
I only want to think you perfect... '
Similarly Queen Victoria's use of underlining in her handwritten journal is represented by her publishers in the printed version with an italic type-face to represent the emphasis she wishes to indicate when writing of Lord Melbourne:
he gave me such a kind, and I may say, fatherly look (Thursday, 28 June 1838)
Where the writer is deliberately exploiting the resources of the written medium, it seems reasonable to suggest that that manipulation constitutes part of the text.
A further illustration of this is to be found in the conventions governing spelling.
In general we assume that words have a standardised spelling in British English.
The fact of the standardisation enables authors to manipulate idiosyncratic spelling to achieve special effects.
Thus in Winnie-the-Pooh the publishers reproduce the notice outside Owl's house in one inset line, using capitals, and with the author's own spelling:
PLEZ CNOKE IF AN RNSR IS NOT REQID
The point that the author makes with this particular spelling would be lost if the words were reproduced in their standard form.
It might then be claimed that such a form of the text was incomplete or inadequate, because the point which the author wishes to make is no longer accessible from the written text.
Indeed the importance of the correct citing of an author's spelling is regularly marked by the insertion of sic into a citation by a second author who wishes to disclaim responsibility for an aberrant spelling.
We have so far been making the simplifying assumption that it is clear, in all cases, what the original text consists of.
Where handwritten texts are at issue, it is often the case that the individual reproducing the text in a printed version has to make a considerable effort of interpretation to assign a value to some of the less legible words.
In literature, as we have remarked already, uncertainty may give rise to cruces, to disputed texts.
In letters, prescriptions, shopping lists, school essays, the reader normally pushes through a once-for-all interpretation of a text which may never be read again.
It must be clear however, that a printed version of a handwritten text is, in an important sense, an interpretation.
This is particularly clear in the handwritten attempts of very young children where the adult is obliged to assign each large painstakingly formed letter token to a particular type of letter, which he may then re-interpret in the light of the larger message.
Thus we have before us a page with a drawing of a large animal (reported to be a lion) and a table with a goldfish bowl on it.
The five-year-old writes below what might be transliterated as:
1
the lion wos the fish to ti it
2
the cat wants to get dwon the steis
3
with qwt to dsthhb thelion
A possible interpretation of the text thus represented might be:
The lion wants the fish, to eat it.
The cat wants to get down the stairs without to disturb the lion.
The transliteration of the original with qwt, in line 3, reasonably accurately represents the first letter (which might also be represented as a figure nine if nine has a straight back stroke).
A more charitable and interpretive transliteration would render it as a (i.e. ' unhatted ' a with a long backstroke (a.
). We shall return to the problem of the interpretive work of the reader / listener in identifying the words which constitute the text, in the next section.
Spoken texts
The problems encountered with the notion of ' text ' as the verbal record of a communicative act become a good deal more complex when we consider what is meant by spoken ' text '.
The simplest view to assume is that a tape-recording of a communicative act will preserve the ' text '.
The tape-recording may also preserve a good deal that may be extraneous to the text  coughing, chairs creaking, buses going past, the scratch of a match lighting a cigarette.
We shall insist that these events do not constitute part of the text (though they may form part of the relevant context, cf.
Chapter 2).
In general the discourse analyst works with a tape-recording of an event, from which he then makes a written transcription, annotated according to his interests on a particular occasion  transcriptions of the sort which will appear in this book.
He has to determine what constitutes the verbal event, and what form he will transcribe it in.
Unless the analyst produces a fine-grained phonetic transcription (which very few people would be able to read fluently) details of accent and pronunciation are lost.
In general, analysts represent speech using normal orthographic conventions.
The analyst may hear an utterance which might be transcribed phonemically as / greipbritn /.
Is he to render this orthographically as grape britain?
Hardly.
He will interpret what he hears and normalise to the conventional orthographic form Great Britain inserting conventional word boundaries in the orthographic version which do not, of course, exist in the acoustic signal.
If he hears a form / gene /, is he to render this in the orthography as gon na (which for some readers may have a peculiarly American association) or gointuh or going to?
The problem is a very real one, because most speakers constantly simplify words phonetically in the stream of speech (see Brown, 1977: ch. 4).
If the analyst normalises to the conventional written form, the words take on a formality and specificity which necessarily misrepresent the spoken form.
Problems with representing the segmental record of the words spoken pale into insignificance compared with the problems of representing the suprasegmental record (details of intonation and rhythm).
We have no standard conventions for representing the paralinguistic features of the utterance which are summarised as' voice quality ', yet the effect of an utterance being said kindly and sympathetically is clearly very different from the effect if it is said brutally and harshly.
Similarly it is usually possible to determine from a speaker's voice his or her sex, approximate age and educational status, as well as some aspects of state of health and personality (see Abercrombie, I968; Laver, 1980).
It is not customary to find any detail relating to these indexical features of the speaker in transcriptions by discourse analysts.
In general, too, rhythmic and temporal features of speech are ignored in transcriptions; the rhythmic structure which appears to bind some groups of words more closely together than others, and the speeding up and slowing down of the overall pace of speech relative to the speaker's normal pace in a given speech situation, are such complex variables that we have very little idea how they are exploited in speech and to what effect (but, cf.
Butterworth, 1980).
It seems reasonable to suggest, though, that these variables, together with pause and intonation, perform the functions in speech that punctuation, capitalization, italicization, paragraphing etc. perform in written language.
If they constitute part of the textual record in written language, they should be included as part of the textual record in spoken language.
If it is relevant to indicate Queen Victoria's underlining, then it is surely also relevant to indicate, for example, a speaker's use of high pitch and loudness to indicate emphasis.
The response of most analysts to this complex problem is to present their transcriptions of the spoken text using the conventions of the written language.
Thus Cicourel (1973) reproduces three utterances recorded in a classroom in the following way:
1
Ci: Like this?
2
T: Okay, yeah, all right, now...
3
Ri: Now what are we going to do?
In I and 3 we have to assume that the? indicates that the utterance functions as a question  whether it is formally marked by, for instance, rising intonation in the case of I, we are not told.
Similarly the status of commas in the speech of the T(eacher) is not made explicit  presumably they are to indicate pauses in the stream of speech, but it may be that they simply indicate a complex of rhythmic and intonational cues which the analyst is responding to.
What must be clear in a transcript of this kind is that a great deal of interpretation by the analyst has gone on before the reader encounters this' data '.
If the analyst chooses to italicise a word in his transcription to indicate, for example, the speaker's high pitch and increased loudness, he has performed an interpretation on the acoustic signal, an interpretation which, he has decided, is in effect equivalent to a writer's underlining of a word to indicate emphasis.
There is a sense, then, in which the analyst is creating the text which others will read.
In this creation of the written version of the spoken text he makes appeal to conventional modes of interpretation which, he believes, are shared by other speakers of the language.
It must be further emphasised that, however objective the notion of ' text ' may appear as we have defined it ('the verbal record of a communicative act'), the perception and interpretation of each text is essentially subjective.
Different individuals pay attention to different aspects of texts.
The content of the text appeals to them or fits into their experience differently.
In discussing texts we idealise away from this variability of the experiencing of the text and assume what Schutz has called ' the reciprocity of perspective ', whereby we take it for granted that readers of a text or listeners to a text share the same experience (Schutz, 1953).
Clearly for a great deal of ordinary everyday language this assumption of an amount of overlap of point of view sufficient to allow mutual comprehension is necessary.
From time to time however we are brought to a halt by different interpretations of ' the same text '.
This is particularly the case when critical attention is being focused on details of spoken language which were only ever intended by the speaker as ephemeral parts, relatively unimportant, of the working-out of what he wanted to say.
It seems fair to suggest that discourse analysis of spoken language is particularly prone to over-analysis.
A text frequently has a much wider variety of interpretations imposed upon it by analysts studying it at their leisure, than would ever have been possible for the participants in the communicative interaction which gives rise to the ' text '.
Once the analyst has' created ' a written transcription from a recorded spoken version, the written text is available to him in just the way a literary text is available to the literary critic.
It is important to remember, when we discuss spoken ' texts', the transitoriness of the original.
It must be clear that our simple definition of ' text ' as' the verbal record of a communicative act ' requires at least two hedges:
i
the representation of a text which is presented for discussion may in part, particularly where the written representation of a spoken text is involved, consist of a prior analysis (hence interpretation) of a fragment of discourse by the discourse analyst presenting the text for consideration
ii
features of the original production of the language, for example shaky handwriting or quavering speech, are somewhat arbitrarily considered as features of the text rather than features of the context in which the language is produced.
The relationship between speech and writing
The view that written language and spoken language serve, in general, quite different functions in society has been forcefully propounded, hardly surprisingly, by scholars whose main interest lies in anthropology and sociology.
Thus Goody &amp; Watt (1963) and Goody (1977) suggest that analytic thinking followed the acquisition of written language ' since it was the setting down of speech that enabled man clearly to separate words, to manipulate their order and to develop syllogistic forms of reasoning ' (Goody, 1977: 11).
Goody goes on to make even larger claims about the ways in which the acquisition of writing, which permits man to reflect upon what he has thought, has permitted the development of cognitive structures which are not available to the non-literate (cf. also the views of Vygotsky, 1962).
He examines the use of ' figures of the written word ' in various cultures, particularly the ' non-speech uses of language ' which develop systems of classification like lists, formulae, tables and ' recipes for the organisation and development of human knowledge ' (1977: 17).
Goody suggests that written language has two main functions: the first is the storage function which permits communication over time and space, and the second is that which ' shifts language from the oral to the visual domain' and permits words and sentences to be examined out of their original contexts, ' where they appear in a very different and highly ' abstract ' context' (1977: 78).
It seems reasonable to suggest that, whereas in daily life in a literate culture, we use speech largely for the establishment and maintenance of human relationships (primarily interactional use), we use written language largely for the working out of and transference of information (primarily transactional use).
However, there are occasions when speech is used for the detailed transmission of factual information.
It is noteworthy, then, that the recipient often writes down the details that he is told.
So a doctor writes down his patient's symptoms, an architect writes down his client's requirements, Hansard records the proceedings of the British Parliament, we write down friends' addresses, telephone numbers, recipes, knitting patterns, and so on.
When the recipient is not expected to write down the details, it is often the case that the speaker repeats them sometimes several times over.
Consider the typical structure of a news broadcast which opens with the ' headlines'  a set of summary statements  which are followed by a news item that consists of an expansion and repetition of the first headline, in which is embedded a comment from ' our man on the spot ' that recapitulates the main points again, then, at the end of the broadcast, there is a repetition of the set of headlines.
There is a general expectation that people will not remember detailed facts correctly if they are only exposed to them in the spoken mode, especially if they are required to remember them over an extended period of time.
This aspect of communication is obviously what written language is supremely good at, whether for the benefit of the individual in remembering the private paraphernalia of daily life, or for the benefit of nations in establishing constitutions, laws and treaties with other nations.
The major differences between speech and writing derive from the fact that one is essentially transitory and the other is designed to be permanent.
It is exactly this point which D. J. Enright makes in the observation that ' Plato may once have thought more highly of speech than of writing, but I doubt he does now! '
(Review in The Sunday Times, 24 January 1982).
Differences in form between written and spoken language
It is not our intention here to discuss the many different forms of spoken language which can be identified even within one geographical area like Britain.
Clearly there are dialectal differences, accent differences, as well as' register ' differences depending on variables like the topic of discussion and the roles of the participants (see e.g. Trudgill, 1974 and Hudson, 1980 for discussion of these sorts of differences).
There is however, one further distinction which is rarely noted, but which it is important to draw attention to here.
That is the distinction between the speech of those whose language is highly influenced by long and constant Immersion in written language forms, and the speech of those whose language is relatively uninfluenced by written forms of language.
It is of course the case that it is the speech of the first set whose language tends to be described in descriptions of the language (grammars), since descriptions are typically written by middle-aged people who have spent long years reading written language.
In particular situations the speech of, say, an academic, particularly if he is saying something he has said or thought about before, may have a great deal in common with written language forms.
For the majority of the population, even of a ' literate ' country, spoken language will have very much less in common with the written language.
This, again, is a point appreciated by Goody: ' Some individuals spend more time with the written language than they do with the spoken.
Apart from the effects on their own personalities... what are the effects on language?
How do written languages differ from spoken ones? ' (1977: 124).
In the discussion which follows we shall draw a simplistic distinction between spoken and written language which takes highly literate written language as the norm of written language, and the speech of those who have not spent many years exposed to written language (a set which will include most young undergraduate students) as the norm for spoken language.
In 1.2.1 we discussed some of the differences in the manner of production of speech and writing, differences which often contribute significantly to characteristic forms in written language as against characteristic forms in speech.
The overall effect is to produce speech which is less richly organised than written language, containing less densely packed information, but containing more interactive markers and planning ' fillers'.
The standard descriptive grammars of English (e.g. Quirk, Greenbaum, Leech &amp; Svartvik, 1972) typically describe features of the written language, or that form of the spoken language which is highly influenced by written language.
From the descriptive work of a number of scholars studying spoken language (e.g. Labov, 1972a; Sinclair &amp; Coulthard, 1975; Chafe, 1979; Ochs, 1979; Cicourel, 1981; Goffman, 1981) we can extract some (by no means all) features which characterise spoken language:
a
the syntax of spoken language is typically much less structured than that of written language
i
spoken language contains many incomplete sentences, often simply sequences
ii
spoken language typically contains rather little subordination
iii
in conversational speech, where sentential syntax can be observed, active declarative forms are normally found.
In over 50 hours of recorded conversational speech, Brown, Currie and Kenworthy (1980) found very few examples of passives, it-clefts or wh-clefts.
Crystal(1980) also presents some of the problems encountered in attempting to analyse spontaneous speech in terms of categories like sentence and clause
As a brief example, notice how this speaker pauses and begins each new ' sentence ' before formally completing the previous one:
it's quite nice the Grassmarket since + it's always had the antique shops but they're looking + they're sort of + em + become a bit nicer +
b
in written language an extensive set of metalingual markers exists to mark relationships between clauses (that complementisers, when I while temporal markers, so-called ' logical connectors' like besides, moreover, however, in spite of, etc.), in spoken language the largely paratactically organised chunks are related by and, but, then and, more rarely, if.
The speaker is typically less explicit than the writer: I 'm so tired (because) I had to walk all the way home.
In written language rhetorical organisers of larger stretches of discourse appear, like firstly, more important than and in conclusion.
These are rare in spoken language.
c
In written language, rather heavily premodified noun phrases (like that one) are quite common  it is rare in spoken language to find more than two premodifying adjectives and there is a strong tendency to structure the short chunks of speech so that only one predicate is attached to a given referent at a time (simple case-frame or one-place predicate) as in: it's a biggish cat + tabby + with torn ears, or in: old man McArthur + he was a wee chap + oh very small + and eh a beard + and he was pretty stooped.
The packaging of information related to a particular referent can, in the written language, be very concentrated, as in the following news item: A man who turned into a human torch ten days ago after snoozing in his locked car while smoking his pipe has died in hospital.
(Evening News (Edinburgh), 22 April 1982)
d
Whereas written language sentences are generally structured in subject-predicate form, in spoken language it is quite common to find what Givo/1n (1979b) calls topic-comment structure, as in the cats + did you let them out.
e
in informal speech, the occurrence of passive constructions is relatively infrequent.
That use of the passive in written language which allows non-attribution of agency is typically absent from conversational speech.
Instead, active constructions with indeterminate group agents are noticeable, as in: Oh everything they do in Edinburgh + they do it far too slowly
f
in chat about the immediate environment, the speaker may rely on (e.g.) gaze direction to supply a referent: (looking at the rain) frightful isn't it.
g
the speaker may replace or refine expressions as he goes along: this man + this chap she was going out with
h
the speaker typically uses a good deal of rather generalised vocabulary: a lot of, got, do, thing, nice, stuff, place and things like that.
i
the speaker frequently repeats the same syntactic form several times over, as this fairground inspector does: I look at fire extinguishers + I look at fire exits + I look at what gangways are available + I look at electric cables what + are they properly earthed + are they properly covered
j
the speaker may produce a large number of prefabricated ' fillers': well, erm, I think, you know, if you see what I mean, of course, and so on.
Some of the typical distinctions between discourse which has been written and that which has been spoken can be seen in the following two descriptions of a rainbow.
(No direct comparison is intended, since the two pieces of discourse were produced in strictly non-comparable circumstances for very different purposes.)
And then, in the blowing clouds, she saw a band of faint iridescence colouring in faint shadows a portion of the hill.
And forgetting, startled, she looked for the hovering colour and saw a rainbow forming itself.
In one place it gleamed fiercely, and, her heart anguished with hope, she sought the shadow of iris where the bow should be.
Steadily the colour gathered, mysteriously, from nowhere, it took presence upon itself, there was a faint, vast rainbow.
(D. H. Lawrence, The Rainbow, chapter 16)
In the first extract (I), the rich lexis and well-organised structure are indications that the writer has taken time in the construction, and possibly reconstruction after several rewritings, of the final product.
There are complete sentences, containing subordinations, frequent modifications via adjectives and adverbs, and more than single predicates per referential expression.
In extract (2), there are frequent pauses, often interrupting major syntactic units, repetitions, incomplete sentences, generalised vocabulary, fillers and one example of a tongue-slip.
normally after + very heavy rain + or something like that + and + you're driving along the road + and + far away + you see + well + er + a series + of + stripes + + formed like a bow + an arch + + very very far away + ah + seven colours but + + I guess you hardly ever see seven it's just a + a series of + colours which + they seem to be separate but if you try to look for the separate (kAz)  colours they always seem + very hard + to separate + if you see what I mean ++ (Postgraduate student speaking informally)
The speaker planning in the here-and-now, possibly threatened with his interlocutor wanting to take a turn, typically repeats himself a good deal, using the same syntactic structure, the same lexical items, using the first word that comes to mind rather than hunting for the mot juste, filling in pauses with ' fillers'.
The overall effect is of information produced in a much less dense manner than is characteristic of written language.
We must assume that the density of information packing in spoken language is appropriate for the listener to process comfortably.
Most people have experienced expository prose read aloud which they have found difficult to follow in the spoken mode.
Few people can extract a great deal from a lecture which is read aloud with no visual support.
Goody points out that the written form of language releases us from the linear experiential mode: ' the fact that it takes a visual form means that one can escape from the problem of the succession of events in time, by backtracking, skipping, looking to see who-done-it before we know what it is they did.
Who, except the most obsessive academic, reads a book as he hears speech?
Who, except the most avant-garde of modern dramatists, attempts to write as they speak? ' (1977:124).
Sentence and utterance
It might seem reasonable to propose that the features of spoken language outlined in the preceding section should be considered as features of utterances, and those features typical of written language as characteristic of sentences.
In this convenient distinction, we can say, in a fairly non-technical way, that utterances are spoken and sentences are written and that we will apply these terms to what Lyons describes as' the products of ordinary language-behaviour '.
In the case of the term sentence, it is important to be clear about the type of object one is referring to.
Lyons makes a distinction between ' text-sentences' and ' system-sentences'.
He describes the latter in the following way:
system-sentences never occur as the products of ordinary language-behaviour.
Representations of system-sentences may of course be used in metalinguistic discussion of the structure and functions of language: and it is such representations that are customarily cited in grammatical descriptions of particular languages.
(Lyons, 1977:31)
Since the linguistic exemplification presented in support of our discussion throughout this book is overwhelmingly drawn from ' ordinary language behaviour ', we shall generally employ the term ' sentence ' in the ' text-sentence ', and not the ' system-sentence ' sense.
Although the linguist who undertakes the analysis of discourse has ultimately the same aims as a linguist who uses' system-sentences' in his grammatical description of a language, there are important methodological differences involved in the two approaches.
Both linguists wish to produce accurate descriptions of the particular language studied.
In pursuit of this goal, the grammarian will concentrate on a particular body of data and attempt to produce an exhaustive but economical set of rules which will account for all and only the acceptable sentences in his data.
He will not normally seek to account for the mental processes involved in any language-user's production of those sentences, nor to describe the physical or social contexts in which those sentences occur.
On each of these issues, concerning ' data ', ' rules', ' processes' and ' contexts', the discourse analyst will take a different view.
On ' data'
The grammarian's ' data ' is inevitably the single sentence, or a set of single sentences illustrating a particular feature of the language being studied.
It is also typically the case that the grammarian will have constructed the sentence or sentences he uses as examples.
This procedure is not often made explicit, but an overt commitment to the constructed-data approach has recently been expressed in the following terms:
I shall assume... that invented strings and certain intuitive judgements about them constitute legitimate data for linguistic research.
(Gazdar, 1979: 11)
In contrast, the analysis of discourse, as undertaken and exemplified in this book, is typically based on the linguistic output of someone other than the analyst.
On the few occasions where constructed data is used as illustration (of a paradigm, for example, in Chapter 4), it is inevitably directed towards accounting for the range of formal options available to a speaker or writer.
More typically, the discourse analyst's ' data ' is taken from written texts or tape-recordings.
It is rarely in the form of a single sentence.
This type of linguistic material is sometimes described as' performance-data ' and may contain features such as hesitations, slips, and non-standard forms which a linguist like Chomsky (1965) believed should not have to be accounted for in the grammar of a language.
Although these two views of ' data ' differ substantially, they are not incompatible, unless they are taken in an extreme form.
A discourse analyst may regularly work with extended extracts of conversational speech, for example, but he does not consider his data in isolation from the descriptions and insights provided by sentence-grammarians.
It should be the case that a linguist who is primarily interested in the analysis of discourse is, in some sense, also a sentence-grammarian.
Similarly, the sentence-grammarian can not remain immured from the discourse he encounters in his daily life.
The sentence he constructs to illustrate a particular linguistic feature must, in some sense, derive from the ' ordinary language ' of his daily life and also be acceptable in it.
A dangerously extreme view of ' relevant data ' for a discourse analyst would involve denying the admissibility of a constructed sentence as linguistic data.
Another would be an analytic approach to data which did not require that there should be linguistic evidence in the data to support analytic claims.
We shall return to the issue of ' relevant data ' for discourse analysis in Chapter 2.
An over-extreme view of what counts as data for the sentence-grammarian was, according to Sampson (1980), noticeable in some of the early work of generative grammarians.
Chomsky gave an indication of the narrowness of view which could be taken, when, immediately before his conclusion that ' grammar is autonomous', he stated:
Despite the undeniable interest and importance of semantic and statistical studies of language, they appear to have no direct relevance to the problem of determining or characterising the set of grammatical utterances.
(Chomsky, 1957: 17)
The essential problem in an extreme version of the constructed-sentence approach occurs when the resulting sentences are tested only against the linguist's introspection.
This can (and occasionally did) lead to a situation in which a linguist claims that the ' data ' he is using illustrates acceptable linguistic strings because he says it does, as a result of personal introspection, and regardless of how many voices arise in disagreement.
The source of this problem, as Sampson (1980: 153) points out, is that the narrow restriction of ' data ' to constructed sentences and personal introspection leads to a ' non-testability ', in principle, of any claims made.
One outcome of this narrow view of data is that there is a concentration on ' artificially contrived sentences isolated from their communicative context ' (see Preface to Givn (ed.), 1979).
Although we shall appeal frequently, in the course of this book, to the insights of sentence grammarians, including those working within a generative framework, we shall avoid as far as possible the methodology which depends on what Lyons (I968) described as regularised, standardised and decontextualised data.
Rules versus regularities
A corollary to the restricted data approach found in much of Chomskyan linguistics is the importance placed on writing rules of grammar which are fixed and true 100% of the time.
Just as the grammarian's ' data ' can not contain any variable phenomena, so the grammar must have categorial rules, and not ' rules' which are true only some of the time.
It is typical of arguments concerning the ' correct rules' of the language in the Chomskyan approach, and that of most other sentence-grammarians, that they are based on the presentation of ' example ' and ' counterexample '.
After all, a single (accepted) sentence, which is presented as a counterexample, can be enough to invalidate a rule of the categorial type.
In this sense, the ' rules' of grammar appear to be treated in the same way as' laws' in the physical sciences.
This restricts the applicability of such rules since it renders them unavailable to any linguist interested in diachronic change or synchronic variation in a language.
It should be emphasised that this is an extreme version of the sentence-grammarian's view and one which is found less frequently, in contemporary linguistics, than it was fifteen years ago.
The discourse analyst, with his' ordinary language ' data, is committed to quite a different view of the rule-governed aspects of a language.
Indeed, he may wish to discuss, not ' rules' but regularities, simply because his data constantly exemplifies non-categorial phenomena.
The regularities which the analyst describes are based on the frequency with which a particular linguistic feature occurs under certain conditions in his discourse data.
If the frequency of occurrence is very high, then the phenomenon described may appear to be categorial.
As Givn says:
what is the communicative difference between a rule of 90% fidelity and one of 100% fidelity?
In psychological terms, next to nothing.
In communication, a system with go% categorial fidelity is a highly efficient system.
(Givn, 1979a: 28)
Yet the frequency of occurrence need not be as high as 90% to qualify as a regularity.
The discourse analyst, like the experimental psychologist, is mainly interested in the level of frequency which reaches significance in perceptual terms.
Thus, a regularity in discourse is a linguistic feature which occurs in a definable environment with a significant frequency.
In trying to determine such regularities, the discourse analyst will typically adopt the traditional methodology of descriptive linguistics.
He will attempt to describe the linguistic forms which occur in his data, relative to the environments in which they occur.
In this sense, discourse analysis is, like descriptive linguistics, a way of studying language.
It may be regarded as a set of techniques, rather than a theoretically predetermined system for the writing of linguistic ' rules'.
The discourse analyst attempts to discover regularities in his data and to describe them.
Product versus process
The regularities which the discourse analyst describes will normally be expressed in dynamic, not static, terms.
Since the data investigated is the result of ' ordinary language behaviour ', it is likely to contain evidence of the ' behaviour ' element.
That is, unless we believe that language-users present each other with prefabricated chunks of linguistic strings (sentences), after the fashion of Swift's professors at the grand academy of Lagado (Gulliver's Travels, part 3, chapter 5), then we must assume that the data we investigate is the result of active processes.
The sentence-grammarian does not in general take account of this, since his data is not connected to behaviour.
His data consists of a set of objects called ' the well-formed sentences of a language ', which can exist independently of any individual speaker of that language.
We shall characterise such a view as the sentence-as-object view, and note that such sentence-objects have no producers and no receivers.
Moreover, they need not be considered in terms of function, as evidenced in this statement by Chomsky (I968: 62):
If we hope to understand human language and the psychological capacities on which it rests, we must first ask what it is, not how or for what purposes it is used.
A less extreme, but certainly related, view of natural language sentences can also be found elsewhere in the literature which relates to discourse analysis.
In this view, there are producers and receivers of sentences, or extended texts, but the analysis concentrates solely on the product, that is, the words-on-the-page.
Much of the analytic work undertaken in ' Textlinguistics' is of this type.
Typical of such an approach is the ' cohesion ' view of the relationships between sentences in a printed text (e.g. the approach in Halliday &amp; Hasan, 1976).
In this view, cohesive ties exist between elements in connected sentences of a text in such a way that one word or phrase is linked to other words or phrases.
Thus, an anaphoric element such as a pronoun is treated as a word which substitutes for, or refers back to, another word or words.
Although there are claims that cohesive links in texts are used by text-producers to facilitate reading or comprehension by text-receivers (cf.
Rochester &amp; Martin 1977, 1979; K?llgren, 1979), the analysis of the ' product ', i.e. the printed text itself, does not involve any consideration of how the product is produced or how it is received.
We shall describe such an approach as deriving from a text-as-product view.
This view does not take account of those principles which constrain the production and those which constrain the interpretation of texts.
In contrast to these two broadly defined approaches, the view taken in this book is best characterised as a discourse-as-process view.
The distinction between treating discourse as' product ' or ' process' has already been made by Widdowson (1979b: 71).
We shall consider words, phrases and sentences which appear in the textual record of a discourse to be evidence of an attempt by a producer (speaker / writer) to communicate his message to a recipient (hearer / reader).
We shall be particularly interested in discussing how a recipient might come to comprehend the producer's intended message on a particular occasion, and how the requirements of the particular recipient(s), in definable circumstances, influence the organisation of the producer's discourse.
This is clearly an approach which takes the communicative function of language as its primary area of investigation and consequently seeks to describe linguistic form, not as a static object, but as a dynamic means of expressing intended meaning.
There are several arguments against the static concept of language to be found in both the ' sentence-as-object ' and ' text-as-product ' approaches.
For example, Wittgenstein (1953: 132) warns that ' the confusions that occupy us arise when language is like an engine idling, not when it is doing work '.
In the course of describing how a sentence-as-object approach, based exclusively on syntactic descriptions, fails to account for a variety of sentential structures, Kuno (1976) concludes that ' it is time to re-examine every major syntactic constraint from a functional point of view '.
Similar conclusions are expressed by Creider (1979), Givon (1976, 1979b), Rommetveit (1974) and Tyler (1978).
In criticising the text-as-product view of cohesion in text, Morgan (1979) argues that we see a link between a particular pronoun and a full noun phrase in a text because we assume the text is coherent and not because the pronoun ' refers back ' to the noun phrase.
We seek to identify the writer's intended referent for a pronoun, since a pronoun can, in effect, be used to refer to almost anything.
That is, what the textual record means is determined by our interpretation of what the producer intended it to mean.
The discourse analyst, then, is interested in the function or purpose of a piece of linguistic data and also in how that data is processed, both by the producer and by the receiver.
It is a natural consequence that the discourse analyst will be interested in the results of psycholinguistic processing experiments in a way which is not typical of the sentence-grammarian.
It also follows that the work of those sociolinguists and ethnographers who attempt to discuss language in terms of user's purposes will also be of interest.
In the course of this book, we shall appeal to evidence in the psycholinguistic and sociolinguistic literature which offers insights into the way in which discourse, produced in describable contexts for recognisable purposes, is processed and comprehended.
On ' context'
We have constantly referred to the ' environment ', ' circumstances' or context in which language is used.
In Chapter 2 we shall explore the problem of specifying the relevant context.
Here we simply remark that in recent years the idea that a linguistic string (a sentence) can be fully analysed without taking ' context ' into account has been seriously questioned.
If the sentence-grammarian wishes to make claims about the ' acceptability ' of a sentence in determining whether the strings produced by his grammar are correct sentences of the language, he is implicitly appealing to contextual considerations.
After all, what do we do when we are asked whether a particular string is' acceptable '?
Do we not immediately, and quite naturally, set about constructing some circumstances (i.e. a ' context ') in which the sentence could be acceptably used?
Any analytic approach in linguistics which involves contextual considerations, necessarily belongs to that area of language study called pragmatics.
' Doing discourse analysis' certainly involves' doing syntax and semantics', but it primarily consists of ' doing pragmatics'.
When the principles which we have expounded in 1.3 are placed alongside Morris's definition of pragmatics as' the relations of signs to interpreters' (1938: 6), the connection becomes quite clear.
In discourse analysis, as in pragmatics, we are concerned with what people using language are doing, and accounting for the linguistic features in the discourse as the means employed in what they are doing.
In summary, the discourse analyst treats his data as the record (text) of a dynamic process in which language was used as an instrument of communication in a context by a speaker / writer to express meanings and achieve intentions (discourse).
Working from this data, the analyst seeks to describe regularities in the linguistic realisations used by people to communicate those meanings and intentions.
2
The role of context in interpretation
Pragmatics and discourse context
In Chapter I, we emphasised that the discourse analyst necessarily takes a pragmatic approach to the study of language in use.
Such an approach brings into consideration a number of issues which do not generally receive much attention in the formal linguist's description of sentential syntax and semantics.
We noted, for example, that the discourse analyst has to take account of the context in which a piece of discourse occurs.
Some of the most obvious linguistic elements which require contextual information for their interpretation are the deictic forms such as here, now, 1, you, this and that.
In order to interpret these elements in a piece of discourse, it is necessary to know (at least) who the speaker and hearer are, and the time and place of the production of the discourse.
In this chapter we shall discuss these and other aspects of contextual description which are required in the analysis of discourse.
There are, however, other ways in which the discourse analyst's approach to linguistic data differs from that of the formal linguist and leads to a specialised use of certain terms.
Because the analyst is investigating the use of language in context by a speaker / writer, he is more concerned with the relationship between the speaker and the utterance, on the particular occasion of use, than with the potential relationship of one sentence to another, regardless of their use.
That is, in using terms such as reference, presupposition, implicature and inference, the discourse analyst is describing what speakers and hearers are doing, and not the relationship which exists between one sentence or proposition and another.
Reference
In presenting the traditional semantic view of reference, Lyons (I968: 404) says that ' the relationship which holds between words and things is the relationship of reference: words refer to things'.
This traditional view continues to be expressed in those linguistic studies (e.g. lexical semantics) which describe the relationship between a language and the world, in the absence of language-users.
Yet, Lyons, in a more recent statement on the nature of reference, makes the following point: ' it is the speaker who refers (by using some appropriate expression): he invests the expression with reference by the act of referring ' (1977: 177).
It is exactly this latter view of the nature of reference which the discourse analyst has to appeal to.
There is support for such a pragmatic concept of reference in Strawson's (1950) claim that ' 'referring ' is not something an expression does; it is something that someone can use an expression to do'; and in Searle's view that ' in the sense in which speakers refer, expressions do not refer any more than they make promises or give orders' (1979: 155).
Thus, in discourse analysis, reference is treated as an action on the part of the speaker / writer.
In the following conversational fragment, we shall say, for example, that speaker A uses the expressions my uncle and he to refer to one individual and my mother's sister and she to refer to another.
We will not, for example, say that he ' refers to ' my uncle.
(I)
A: my uncle's coming home from Canada on Sunday + he's due in +
B: how long has he been away for or has he just been away?
A: Oh no they lived in Canada eh he was married to my mother's sister + + well she's been dead for a number of years now +
The complex nature of discourse reference will be investigated in greater detail in Chapters 5 and 6.
Presupposition
In the preceding conversational fragment (I), we shall also say that speaker A treats the information that she has an uncle as presupposed and speaker B, in her question, indicates that she has accepted this presupposition.
We shall take the view that the notion of presupposition required in discourse analysis is pragmatic presupposition, that is, ' defined in terms of assumptions the speaker makes about what the hearer is likely to accept without challenge ' (Givn, 1979a: 50).
The notion of assumed ' common ground ' is also involved in such a characterisation of presupposition and can be found in this definition by Stalnaker (1978: 321):
presuppositions are what is taken by the speaker to be the common ground of the participants in the conversation.
Notice that, in both these quotations, the indicated source of presuppositions is the speaker.
Consequently, we shall, as with reference, avoid attributing presuppositions to sentences or propositions.
Thus, we can see little practical use, in the analysis of discourse, for the notion of logical presupposition which Keenan (1971: 45) describes in the following way:
A sentence S logically presupposes a sentence S' just in case S logically implies S' and the negation of S, C S, also logically implies S'.
If we take the first sentence of extract (I) as S, and present it below as (2a), we can also present the negation of S, as (2b), and the logical presupposition, S', as (2c).
(2)
a.
My uncle is coming home from Canada.
b.
My uncle isn't coming home from Canada.
c.
I have an uncle.
Following Keenan's definition, we can say that (2a) logically presupposes (2c) because of constancy under negation.
However, it seems rather unnecessary to introduce the negative sentence (2b) into a consideration of the relationship between (2a) and (2c) which arises in the conversation presented earlier in (I).
Though it may not be common knowledge that the speaker has an uncle, it is what Grice (1981: 190) terms' noncontroversial ' information.
Moreover, since the speaker chose to say my uncle rather than I have an uncle and he..., we must assume she didn't feel the need to assert the information.
What she appears to be asserting is that this person is coming home from Canada.
Given this assertion, the idea that we should consider the denial of this assertion in order to find out whether there is a presupposition in what the speaker has not asserted seems particularly counterintuitive.
The introduction of the negative sentence (2b) into a consideration of (2a) creates an additional problem.
For example, it has been suggested (cf.
Kempson, 1975) that a sentence such as (2d) is a perfectly reasonable sentence of English and undermines the argument for logical presupposition, as it is defined above
(2d) My uncle isn't coming home from Canada because I don't have an uncle.
Sentences like (2d) always seem typical of utterances made by a speaker to deny another speaker's presupposition in a rather aggressive way.
Yet the circumstances in which (2d) might be uttered are likely to be quite different from those in which the first sentence of extract (I) was uttered.
The speakers, we may suggest, would have different presuppositions, in the two situations.
If we rely on a notion of speaker, or pragmatic, presupposition, we can simply treat (2c) as a presupposition of the speaker in uttering (2a).
Sentences (2b) and (2d) do not come into consideration at all.
In support of a view that hearers behave as if speakers' presuppositions are to be accepted, there is the rather disturbing evidence from Loftus' study (1975) of answers to leading questions.
After watching a film of a car accident some subjects were asked the two questions in (3).
(3)
a.
How fast was car A going when it turned right?
b.
Did you see a stop sign?
We can note that one of the speaker-presuppositions in asking (3a) is that car A turned right.
A number (35%) answered yes to question (3b).
Another group of subjects were asked the questions in (4).
(4)
a.
How fast was car A going when it ran the stop sign?
b.
Did you see a stop sign?
One of the speaker-presuppositions in asking (4a) is that car A ran the stop sign.
In this situation, a significantly larger group (53%) answered yes to question (4b).
It is worth noting that a number of subjects did not answer the b question in terms of truth or falsehood of fact, but according to what the speaker, in asking the preceding question, had appeared to presuppose.
(For a more detailed discussion of this issue, see Loftus, 1975 and Loftus &amp; Zanni, 1975.)
We shall reconsider the notion of presupposition in section 3.3.2, but generally avoid the complex arguments which revolve around the presuppositions of sentences and propositions.
(See the contributions and bibliography in Oh &amp; Dineen (eds.) 1979.)
Implicatures
The term ' implicature ' is used by Grice (1975) to account for what a speaker can imply, suggest, or mean, as distinct from what the speaker literally says.
There are conventional implicatures which are, according to Grice, determined by ' the conventional meaning of the words used ' (1975: 44).
In the following example (5), the speaker does not directly assert that one property (being brave) follows from another property (being an Englishman), but the form of expression used conventionally implicates that such a relation does hold.
(5) He is an Englishman, he is, therefore, brave.
If it should turn out that the individual in question is an Englishman, and not brave, then the implicature is mistaken, but the utterance, Grice suggests, need not be false.
For a fuller discussion of conventional implicature, see Karttunen &amp; Peters (1979).
Of much greater interest to the discourse analyst is the notion of conversational implicature which is derived from a general principle of conversation plus a number of maxims which speakers will normally obey.
The general principle is called the Cooperative Principle which Grice (1975: 45) presents in the following terms:
Make your conversational contribution such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged.
The conversational conventions, or maxims, which support this principle are as follows:
Quantity: Make your contribution as informative as is required (for the current purposes of the exchange).
Do not make your contribution more informative than is required.
Quality: Do not say what you believe to be false.
Do not say that for which you lack adequate evidence.
Relation: Be relevant.
Manner: Be perspicuous.
Avoid obscurity of expression.
Avoid ambiguity.
Be brief (avoid unnecessary prolixity).
Be orderly.
Grice does not suggest that this is an exhaustive list  he notes that a maxim such as Be polite is also normally observed  nor that equal weight should be attached to each of the stated maxims.
(The maxim of manner, for example, does not obviously apply to primarily interactional conversation.)
We might observe that the instruction Be relevant seems to cover all the other instructions.
However, by providing a description of the norms speakers operate with in conversation, Grice makes it possible to describe what types of meaning a speaker can convey by ' flouting ' one of these maxims.
This flouting of a maxim results in the speaker conveying, in addition to the literal meaning of his utterance, an additional meaning, which is a conversational implicature.
As a brief example, we can consider the following exchange:
(6)
A: I am out of petrol.
B: There is a garage round the corner.
In this exchange, Grice (1975: 51) suggests that B would be infringing the instruction Be relevant if he was gratuitously stating a fact about the world via the literal meaning of his utterance.
The implicature, derived from the assumption that speaker B is adhering to the Cooperative Principle, is that the garage is not only round the corner, but also will be open and selling petrol.
We might also note that, in order to arrive at the implicature, we have to know certain facts about the world, that garages sell petrol, and that round the corner is not a great distance away.
We also have to interpret A's remark not only as a description of a particular state of affairs, but as a request for help, for instance.
Once the analysis of intended meaning goes beyond the literal meaning of the ' sentences-on-the-page ', a vast number of related issues have to be considered.
We shall investigate some of these issues in the course of this book, particularly in Chapters 6 and 7.
As a brief account of how the term ' implicature ' is used in discourse analysis, we have summarised the important points in Grice's proposal.
We would like to emphasise the fact that implicatures are pragmatic aspects of meaning and have certain identifiable characteristics.
They are partially derived from the conventional or literal meaning of an utterance, produced in a specific context which is shared by the speaker and the hearer, and depend on a recognition by the speaker and the hearer of the Cooperative Principle and its maxims.
For the analyst, as well as the hearer, conversational implicatures must be treated as inherently indeterminate since they derive from a supposition that the speaker has the intention of conveying meaning and of obeying the Cooperative Principle.
Since the analyst has only limited access to what a speaker intended, or how sincerely he was behaving, in the production of a discourse fragment, any claims regarding the implicatures identified will have the status of interpretations.
In this respect, the discourse analyst is not in the apparently secure position of the formal linguist who has' rules' of the language which are or are not satisfied, but rather, is in the position of the hearer who has interpretations of the discourse which do, or do not, make sense.
(For a more detailed treatment of conversational implicature, see Levinson, forthcoming.)
Inference
Since the discourse analyst, like the hearer, has no direct access to a speaker's intended meaning in producing an utterance, he often has to rely on a process of inference to arrive at an interpretation for utterances or for the connections between utterances.
Such inferences appear to be of different kinds.
It may be the case that we are capable of deriving a specific conclusion (7c) from specific premises (7a) and (7b), via deductive inference, but we are rarely asked to do so in the everyday discourse we encounter.
(7)
a.
If it's sunny, it's warm.
b.
It's sunny.
c.
So, it's warm.
We are more likely to operate with a rather loose form of inferencing which leads us to believe that the hats and coats mentioned in (8) belong to visitors to the house which has the dresser in its kitchen.
(8) in the kitchen there was a huge dresser and when anyone went in you see + the hats and coats were all dumped on this dresser
It may be, of course, that such an inference is wrong, but, as discourse processors, we seem to prefer to make inferences which have some likelihood of being justified and, if some subsequent information does not fit in with this inference, we abandon it and form another.
As an illustration of this, consider the following example (9), taken from Sanford &amp; Garrod (1981: 10):
(9) John was on his way to school.
If we were to take a formal view of the entailments of such a declarative sentence (like that, for example, expressed in Smith &amp; Wilson, 1979: 150f.), we would be obliged to accept as entailments a set of sentences which would include the following:
(10)
a.
Someone was on his way to school.
b.
John was on his way to somewhere.
c.
Someone was on his way to somewhere.
This view of what we infer from reading (9) will only provide us with a limited insight into how readers interpret what they read.
Most readers report that they infer from (9) that John is a schoolboy, among other things.
When sentence (9) is followed later in the same text by sentence (II), readers readily abandon their original inference and form another, for example that John is a schoolteacher.
(11) Last week he had been unable to control the class.
In order to capture this type of inference, which is extremely common in our interpretation of discourse, we need a relatively loose notion of inference based on socio-cultural knowledge.
Gumperz (1977) presents an extended discussion of the types of factors involved in this type of pragmatic, as opposed to logical, inference.
We shall discuss the influence of inference in more detail in Chapter 7.
For the moment, we simply present a view which claims that the terms reference, presupposition, implicature and inference must be treated as pragmatic concepts in the analysis of discourse.
These terms will be used to indicate relationships between discourse participants and elements in the discourse.
Since the pragmatic use of these terms is closely tied to the context in which a discourse occurs, we shall now investigate what aspects of context have to be considered in undertaking the analysis of discourse.
The context of situation
Since the beginning of the 1970s, linguists have become increasingly aware of the importance of context in the interpretation of sentences.
The implications of taking context into account are well expressed by Sadock (1978: 281):
There is, then, a serious methodological problem that confronts the advocate of linguistic pragmatics.
Given some aspects of what a sentence conveys in a particular context, is that aspect part of what the sentence conveys in virtue of its meaning... or should it be ' worked out ' on the basis of Gricean principles from the rest of the meaning of the sentence and relevant facts of the context of utterance?
If we are to begin to consider the second part of this question seriously we need to be able to specify what are the ' relevant facts of the context of utterance '.
The same problem is raised by Fillmore (1977: 119) when he advocates a methodology to which a discourse analyst may often wish to appeal:
The task is to determine what we can know about the meaning and context of an utterance given only the knowledge that the utterance has occurred...
I find that whenever I notice some sentence in context, I immediately find myself asking what the effect would have been if the context had been slightly different.
In order to make appeal to this methodology, which is very commonly used in linguistic and philosophical discussion, we need to know what it would mean for the context to be ' slightly different '.
Features of context
Consider two invented scenarios in which an identical utterance is produced by two distinct speakers.
a
speaker: a young mother, hearer: her mother-in-law, place: park, by a duckpond, time: sunny afternoon in September 1962.
They are watching the young mother's two-year-old son chasing ducks and the mother-in-law has just remarked that her son, the child's father, was rather backward at this age.
The young mother says:
I do think Adam's quick
b
speaker: a student, hearers: a set of students, place: sitting round a coffee table in the refectory, time: evening in March 1980.
John, one of the group, has just told a joke.
Everyone laughs except Adam.
Then Adam laughs.
One of the students says:
I do think Adam's quick
(In each case phonological prominence is placed on Adam.)
Clearly we can do a formal analysis on these tokens and, in both cases, the speaker says of Adam that he is quick.
It is clear, however, that the utterances in the contexts of situation in which they are cited, would be taken to convey very different messages.
In (a) we shall simplistically assume that the referents of I and Adam are fixed by spatio-temporal co-ordinates.
This' Adam ' is being compared (or contrasted), favourably, with his father.
Quick, may be interpreted, in the context of backward, as meaning something like ' quick in developing '.
In (b) different referents for I and Adam are fixed spatiotemporally.
This' Adam ' is being compared (or contrasted) not with his father and favourably, but with the set of other students unfavourably.
In this case quick must be interpreted as meaning something like ' quick to understand / react / see the joke '.
Moreover, since it is said in a context where Adam has just manifestly failed to react to the punch-line as quickly as the set of other students, the speaker (given this type of speaker to this type of hearer in this type of surroundings) will be assumed not to be intending to tell an untruth, but to be implicating the opposite of what she has said.
Is it possible to determine in any principled way what aspects of context of situation are relevant to these different interpretations of the same ' utterance ' on two occasions?
J. R. Firth (regarded by many as the founder of modern British linguistics) remarked:
Logicians are apt to think of words and propositions as having ' meaning ' somehow in themselves, apart from participants in contexts of situation.
Speakers and listeners do not seem to be necessary.
I suggest that voices should not be entirely dissociated from the social context in which they function and that therefore all texts in modern spoken languages should be regarded as having ' the implication of utterance ', and be referred to typical participants in some generalised context of situation.
(1957: 226)
Firth, then, was concerned to embed the utterance in the ' social context ' and to generalise across meanings in specified social contexts.
He proposed an approach to the principled description of such contexts which bears a close resemblance to more recent descriptions which we shall go on to examine:
My view was, and still is, that ' context of situation ' is best used as a suitable schematic construct to apply to language events...
A context of situation for linguistic work brings into relation the following categories:
A
The relevant features of participants: persons, personalities.
i
The verbal action of the participants.
ii
The non-verbal action of the participants.
B
The relevant objects.
C
The effect of the verbal action.

A very rough parallel to this sort of context can be found in language manuals providing the learner with a picture of the railway station and the operative words for travelling by train.
It is very rough.
But it is parallel with the grammatical rules, and is based on the repetitive routines of initiated persons in the society under description.
(1957: 182; for a practical application of Firth's approach, see Mitchell, 1957.)
An approach similarly emphasising the importance of an ethnographic view of communicative events within communities has been developed by Hymes in a series of articles.
Hymes views the role of context in interpretation as, on the one hand, limiting the range of possible interpretations and, on the other, as supporting the intended interpretation:
The use of a linguistic form identifies a range of meanings.
A context can support a range of meanings.
When a form is used in a context it eliminates the meanings possible to that context other than those the form can signal: the context eliminates from consideration the meanings possible to the form other than those the context can support.
(Hymes, 1962, quoted in Wootton, 1975: 44)
Hymes (1964) sets about specifying the features of context which may be relevant to the identification of a type of speech event in a way reminiscent of Firth 's.
Like Firth, he seizes first on the ' persons' participating in the speech event.
Generalising over speech events, he abstracts the roles addressor and addressee.
The addressor is the speaker or writer who produces the utterance.
The addressee is the hearer or reader who is the recipient of the utterance.
(Later Hymes also distinguishes audience, since the presence of overhearers may contribute to the specification of the speech event.)
Knowledge of the addressor in a given communicative event makes it possible for the analyst to imagine what that particular person is likely to say.
Knowledge of his addressee constrains the analyst's expectations even further.
Thus, if you know the speaker is the prime minister or the departmental secretary or your family doctor or your mother, and you know that the speaker is speaking to a colleague or his bank manager or a small child, you will have different expectations of the sort of language which will be produced, both with respect to form and to content.
If you know, further, what is being talked about, Hymes' category of topic, your expectations will be further constrained.
If then you have information about the setting, both in terms of where the event is situated in place and time, and in terms of the physical relations of the interactants with respect to posture and gesture and facial expression, your expectations will be still further limited.
The remaining features of context which Hymes discusses (in 1964) include large-scale features like channel (how is contact between the participants in the event being maintained  by speech, writing, signing, smoke signals), code (what language, or dialect, or style of language is being used), message-form (what form is intended  chat, debate, sermon, fairy-tale, sonnet, love-letter, etc.) and event (the nature of the communicative event within which a genre may be embedded  thus a sermon or prayer may be part of the larger event, a church service).
In later recensions Hymes adds other features, for example key (which involves evaluation  was it a good sermon, a pathetic explanation, etc.), and purpose (what did the participants intend should come about as a result of the communicative event).
Hymes intends that these contextual features should be regarded rather as general phonetic features are regarded.
Just as a phonetician may select, from the general phonetic features available, the features voiced, bilabial and stop, but not lateral, to characterise a [ b ], so, he suggests, the analyst may choose from the contextual features, those necessary to characterise a particular communicative event.
Just as the phonetician may wish to make a more detailed, more specific description of the [ b ] under consideration, for example mentioning delayed onset of voicing and some protrusion of the lips during the period of closure, so may the ethnographer wish to specify some of the contextual features in great detail.
We shall return to this point.
Hymes' features constitute essentially a checklist which would enable a visiting ethnographer to arrive by helicopter in a location where a communicative event is in process and to check off the detail of the nature of the communicative event.
Let us consider such an ethnographer as an invisible witness to a particular speech event.
He would begin, presumably, by noting the larger-scale features of context: what channel is being used (we shall say speech), what language code is being used (we shall specify it is English), what message-form is being performed (we shall specify it is conversation), what event is it embedded in (we shall specify it is part of an interview).
He can identify the participants: the addressor is a young scientist who is being interviewed by the addressee who is doing research on language.
The setting is physically located in the addressee's territory in Edinburgh University and a prominent physical feature is a tape-recorder which is switched on.
The time is during the later 1970s (so it is reasonable to expect that they will speak modern English, with Scottish accents).
It has just been agreed that they will talk about the young scientist's work, the tape-recorder is switched on and he says:
(12) I must admit I 'm very nervous.
His topic at this point, we shall simplistically assume (see further discussion in Chapter 3), is his nervousness.
Given the knowledge of context the analyst has, he should find this a fairly unsurprising utterance.
It is very rarely the case in real life that we can predict in detail the form and content of the language which we will encounter, but, given all of the ethnographic information we have specified, the actual occurring utterance is much more likely (hence, we assume, much more readily processed by the addressee) than any of the following ' utterances' which did not occur:
(13)
a.
Please pass the marmalade.
b.
My cat has just been sick again.
c.
Get into the box.
d.
I am about to make the first incision.
The more the analyst knows about the features of context, the more likely he is to be able to predict what is likely to be said (see 2.4).
It is further the case that the ethnographic features will give us a value for the deictic forms occurring in the utterance which was actually produced.
Thus 1, must, and am must be interpreted with respect to the speaker, the young scientist, at the time of making the utterance.
(The context here makes the other possible reading, that the speaker is characteristically nervous all of the time, so unlikely as not to be considered apparently by the addressee, or indeed by the analyst until the process of analysis was brought to conscious attention.)
In 2.
I we pointed out that deictic elements of the utterances can only be interpreted with respect to the context in which they are uttered.
Hymes' checklist of ethnographic features offers one characterisation of context to which we can relate such deictic elements.
A more elaborate checklist is provided by the philosopher Lewis (1972), specifically to provide an index of those co-ordinates which a hearer would need to have specified in order that he could determine the truth of a sentence.
Like most formal linguists, Lewis assumes that the channel is speech, the code, English, the message-form conversation and the event one where one individual is informing another.
His interests lie, not with these general features of the communicative event, but with those particular co-ordinates which constitute ' a package of relevant factors, an index ' (1972: 173) and which characterise the context against which the truth of a sentence is to be judged.
The co-ordinates of the index are specified as follows:
a
possible-world co-ordinate: this is to account for states of affairs which might be, or could be supposed to be or are
b
time co-ordinate: to account for tensed sentences and adverbials like today or next week
c
place co-ordinate: to account for sentences like here it is
d
speaker co-ordinate: to account for sentences which include first person reference (I, me, we, our, etc.)
e
audience co-ordinate: to account for sentences including you, yours, yourself, etc.
f
indicated object co-ordinate: to account for sentences containing demonstrative phrases like this, those, etc.
g
previous discourse co-ordinate: to account for sentences including phrases like the latter, the aforementioned, etc.
h
assignment co-ordinate: an infinite series of things (sets of things, sequences of things...)
Rather similar lists are proposed by scholars who are concerned with the construction of formal discourse domains (see discussion in Chapter 3).
For our present purposes we should note that Lewis' list, like Hymes', makes reference to the speaker and hearer in order to assign values to the deictic categories of speaker and audience (addressor / addressee) realised in first and second person pronouns.
Hymes' category setting is expanded to take explicit and distinct account of time and place.
Hymes' generalised feature of topic is now distributed between the deictic co-ordinate indicated object, the assignment co-ordinate and the previous discourse co-ordinate.
This last co-ordinate specifically enables the hearer to interpret what is said in the light of what has already been said.
It builds in a cumulative temporal structure to the index, in that the hearer must continually update the information in his previous discourse component, to take account of what has most recently been added.
It is, obviously, not possible for us in a textbook to permit you to have the experience of everyday discourse in what Stenning (1978) calls a ' normal context ', where the hearer is part of the context and then experiences the text.
We have to have recourse to what Stenning calls' abnormal ' contexts, where the analyst reads the text and then has to try to provide the characteristics of the context in which the text might have occurred.
We are going to provide you with three written fragments, abstracted from the contexts in which they appeared.
The first two are printed, the third spraygunned on a wall.
We ask you to consider what, if any, difficulty you have in understanding them, in terms of the co-ordinates of Lewis' index.
(14).
Place two fingers in the two holes directly to the left of the finger stop.
Remove finger nearest stop.
b.
He seemed to resent them on that occasion and will not wear them today.
c.
SQUASHED INSECTS DO NT BITE MAD MENTAL RULE
We have not, as yet, introduced any satisfactory way of handling your experience of previous similar texts (see discussion in 2.4).
For the moment we shall suppose that you probably recognise the type of writer in (a) as some impersonal / institutionalised writer addressing a general reader rather than a particular individual (paying attention to Place and Remove and the ellipsis in the second sentence (the) finger nearest (to the) stop).
If you have difficulty in interpreting this fragment it is probably partly because you are not sure of the referents of the expression the two holes and the finger stop.
You may work out that the two holes have to be of a suitable size for an individual (?) to put two fingers in, possibly near enough to each other to put two fingers of the same hand in, and, having established this scale, it seems likely that the object referred to as the finger stop is only centimetres removed, rather than kilometres removed.
It would certainly help you to have the following information:
The addressor is the Post Office.
The addressee is you as a telephone user.
You can probably work out the rest if you did not know it already.
However we shall spell out some more:
The time of utterance in clock or calendar time does not seem relevant, but what certainly is relevant is that you should know whether this instruction still applies.
(It does.)
The place of the original utterance is hardly relevant but where you would encounter the text is.
(Look in your telephone directory.)
The possible world that is relevant is specified in the previous discourse: ' It is worth remembering how to dial 999 in darkness or smoke. '
(We should point out that you are not here being asked to use the co-ordinates for the purpose Lewis intended them for, to determine the truth of a sentence.
It is a matter of debate whether truth can be assigned to sentences in the imperative form.)
In text b the problem of interpretation arises because of not knowing the referents for the expressions He, them, on that occasion and them and not having a value to fix the time expression today.
You may be able to work out that He refers to an animate masculine entity, the subject of both clauses.
You may wonder why it is reported that He seemed to resent them, which may suggest that he was unable to express his own resentment, which may limit your range of potential interpretations of the expression He.
You note that he resented them, where them is plural, and you may consider what plural entity may be both resented and worn (or not worn).
This example has all the characteristics of a sentence occurring within a larger piece of text, and illustrates quite clearly the need for a ' previous discourse ' co-ordinate, as well as the more obvious' time ' and ' place ' co-ordinates.
This text appeared in The Sporting Chronicle on 4 June 1980.
In the preceding part of the text, the writer has been describing a particular racehorse (He) which had been fitted with blinkers (them) for its previous race (on that occasion).
The third text, c, offers more thorny problems.
Whereas the language of a and b is quite straightforward and all you require to arrive at an interpretation are values for expressions being used to refer, you may feel that the language here is obscure, perhaps not even meaningful.
It is relevant that the time at which this text appeared was in the late 1970s.
Your experience of previous similar texts in the 1970s may have familiarised you with the form X rule OK which may permit you to divide this unpunctuated sequence into two parts:
SQUASHED INSECTS DO NT BITE MAD MENTAL RULE
The place at which it appeared is relevant.
It was spraygunned on a wall in Glasgow.
The form of the text, together with the information about place, may suggest to you, if you have previous experience of such texts, that this text derives from an interaction between street gangs.
Encyclopedic knowledge of the world might inform you that the writer is a member of ' Mad Mental ' (a street gang) and that the intended addressees are members of ' The Insects' (another street gang).
You then need to make appeal to previous discourse in which the Insects had proclaimed INSECTS BITE.
(You might then appeal to your knowledge of what Hymes calls' message-form ' which informs you that street gang interactions on walls consists of taunts and counter-taunts.
Thus you might arrive at an attribution of intention in the warning SQUASHED INSECTS DO NT BITE and the straight assertion MAD MENTAL RULE  without the OK tag, which might be taken to invite assent on the part of the addressee.)
Texts a and b, addressed to the general reader, are relatively accessible fragments of language which require only specification of the intended referents to make them readily interpretable.
Text c is intended for specific addressees, not for the general public, and it is hard for the general public to interpret without access to shared presuppositions and previous experience which can not comfortably be forced into the framework proposed by Lewis.
In order to take account of this, we are going to need some way of making appeals to notions like ' shared presuppositions', 'encyclopedic knowledge', ' intention / purpose in uttering ' and ' experience of previous similar text ' which we have simply appealed to in an ad hoc way in our discussion so far.
We return to these questions in 2.3.
What we have shown in this section is that the contextual features suggested by Hymes, supplemented with the index of co-ordinates proposed by Lewis (put forward, remember, with quite different purposes in mind) do enable us to give a partial account of what the undifferentiated term ' context ' may mean.
From this it follows that we could give some account of what it might mean to ' change the context ' in the sense in which Fillmore (1977: 119) envisages this when he says' I... find myself asking what the effect would have been if the context had been slightly different. '
We could reply that if you alter the condition specified by any of the co-ordinates, you alter the context.
At this point we shall consider only the alteration of one co-ordinate, the speaker co-ordinate.
Obviously, if Jane says I 'm skipping and Mary says I 'm skipping we observe that on one occasion it is Jane who announces that she is skipping and on another it is Mary.
In each case the sentence is true if the person who spoke was skipping at the time of the utterance.
However, if we are further told that speaker Jane is only three years old, we may, in addition to paying attention to the announcement, consider that it is a remarkable feat for a three-year-old.
Whereas if Mary is eight years old and known to be an intrepid skipper, the announcement may be one of a depressingly predictable series.
We pay different amounts of attention to the announcements and react to them differently, because one aspect of the context, the speaker, is significantly different.
Consider the following fragment of conversation:
(15)
A: are you often here
B: quite often + about once a month + actually ++ I come up to see my children
You have to suppose of B that B is of an age to have children.
What we are interested in is the different sorts of inferences which we make as addressees, depending on variables like the age and sex of the speaker, as a result of hearing what B says.
Suppose B is a man of seventy.
We assume that B's children will be grown-up.
Nothing particular follows from the fact that he visits them once a month, except perhaps we infer that he has a close relationship with them.
Suppose the speaker is a young man in his thirties.
We assume that children he has will be young children, children of an age who usually live with their parents.
We may then wonder why B's children are not living with their father, wonder whether the exigences of his professional life, or of his relationship with the children's mother, constrains him to live apart from them.
Suppose the speaker is a young woman in her thirties.
Again we assume that she would have young children, children who would normally be expected to be living with her.
Since, in the case of the parents being separated, young children usually live with their mother in our society, we might infer that the woman's children are in some form of institutional or educational care.
(In the conversation we quote from, the speaker was a man in his early thirties and the children were living with his estranged wife, all inferences which had been drawn by A before B went on to explain that this was the case.)
Observe that the sorts of inferences we have been discussing are not sanctioned by the form of language used.
The different inferences arise because of the alteration of the context, in the simple manipulation of age and sex of the addressor.
It is the interpretation of the utterance in context which permits the hearer to draw such inferences (see Chapter 7 for further discussion of inferences).
Co-text
In our discussion so far we have concentrated particularly on the physical context in which single utterances are embedded and we have paid rather little attention to the previous discourse co-ordinate.
Lewis introduced this co-ordinate to take account of sentences which include specific reference to what has been mentioned before as in phrases like the aforementioned.
It is, however, the case that any sentence other than the first in a fragment of discourse, will have the whole of its interpretation forcibly constrained by the preceding text, not just those phrases which obviously and specifically refer to the preceding text, like the aforementioned.
Just as the interpretation of the token  in the child's representation of ' without to disturb the lion ' and the token [ p ] in [ greipbritn ] are determined by the context in which they appear, so the words which occur in discourse are constrained by what, following Halliday, we shall call their co-text.
Consider the following lexical items in a number of verbal contexts cited almost at random from Darwin's Journal during the Voyage of HMS Beagle round the World:
(16)
a.
The children of the Indians are saved, to be sold or given away as servants, or rather slaves for as long a time as the owners can make them believe themselves slaves.
But I believe in their treatment there is little to complain of (114)
b.
The same evening I went on shore.
The first landing in any new country is very interesting.
(169)
c.
When we came within hail, one of the four natives who were present advanced to receive us and began to shout most vehemently, wishing to direct us where to land.
When we were on shore the party looked rather alarmed.
(206)
d.
After crossing many low hills, we descended into the small land-locked plain of Guitron.
In the basins, such as this one, which are elevated from one thousand to two thousand feet above the sea, two species of acacia... grow in large numbers.
(257) (1892 edition)
The point we wish to make here should be an obvious one and can of course be made with respect to many of the other items which we have not italicised in the cited texts.
However, consider the sort of lexical content you would expect to find associated with the forms treatment, landing, party and basin in a dictionary entry, and note how finding the forms embedded within a co-text constrains their interpretation.
Just as the interpretation of individual lexical items is constrained by co-text, so is the interpretation of utterances within a discourse.
Consider this text of the beginning of a sixteen-year-old Scottish pupil's account of a Semp cartoon:
(17)
a. a man and woman sitting in the living room + the woman sitting reading quite happily  the man's bored goes to the window looks out the window + and gets himself ready and goes out +
The reader must interpret the woman sitting reading quite happily as the ' woman ' already mentioned, hence must construct an interpretation which has her ' sitting reading quite happily in the living room '.
Similarly the window which the man approaches must be interpreted as' the window of the living room '.
The speaker continues with a change of location and we have to assume that what follows is within the newly introduced location:
b. goes to his goes to a club + has a drink talks to the barman + then he starts dancing with a beautiful girl long black hair + has a good time +
We interpret everything that happens here as happening to the man we met in the living room who is now at a club.
So he has a drink, talks to the barman, starts dancing and has a good time all at the ' club '.
The speaker announces another change of location
c. then he goes home and he calls her + and his wife overhears him +
Again we assume that we are still talking about the same man, that he has returned home to the location where the ' living room ' we first met was located.
Now the analyst may be in some doubt how to interpret and he calls her, since the man might reasonably go into the house and call (shout for) his wife.
However this interpretation is ruled out by the following co-text and his wife overhears him.
So we are obliged to interpret calls as meaning ' phones' and her as referring to ' the beautiful girl with long black hair with whom he danced and had a good time '.
Within the co-text, as we have seen in (17) above, a further context may be constructed which has its own index of coordinates.
Indeed within that constructed context, further contexts may be nested.
Consider the following passages:
(18) About four months before the time I am writing of, my Lady had been in London, and had gone over a Reformatory...
The matron, seeing my Lady took an interest in the place, pointed out a girl to her, named Rosanna Spearman, and told her a most miserable story: which I haven't the heart to repeat here; for I don't like to be made wretched without any use, and no more do you.
The upshot of it was, that Rosanna Spearman had been a thief...
(Wilkie Collins, The Moonstone)
The actual place and time of writing of the manuscript by the author, Wilkie Collins, or indeed the identity of the author, is not a necessary piece of information for the reader to interpret the text.
We may assume, however, that he will have a better understanding of the purpose of the author in constructing the text in the way it is constructed if he knows that it is written in the late nineteenth century (which will account for some differences in code, in Hymes' terms) in Victorian England (which will account for the reference to a Reformatory) and that the author is constructing the first English detective story, narrating the events from the point of view of four different participants, whose characters are in part revealed by the narrative style which the author assigns to them.
We have then, an author and an actual time and place of writing the novel (or a series of times and places).
Then to each narrator is assigned a time and place of the writing of his contribution.
It is presumably that time which is relevant to the comment which I haven't the heart to repeat here where I refers to the current narrator.
Immediately preceding this extracted fragment, the narrator has been describing an incident relevant to the main story.
This is referred to in the expression the time I am writing of.
The narrator then proceeds to give some background information, which he situates in a previous time About four months before.
He introduces Rosanna Spearman, who, at the time four months before was a resident of the Reformatory, but at some previous time to that, Rosanna Spearman had been a thief.
Within the time domain of ' four months before ' a new speaker and hearer are introduced:
(19) My Lady... said to the matron upon that, ' Rosanna Spearman shall have her chance, in my service '.
In a week afterwards, Rosanna Spearman entered this establishment as our second housemaid.
At the time of utterance, four months before the time I am writing of, the beneficent lady speaks of the future, shall have her chance.
In the following sentence the narrator comments on what happened a week later than the time of the lady's speech, from the point of view of his context at the time of writing his contribution to the novel, In a week afterwards...
This brief introduction does scant justice to the interest of the temporal structure of this passage.
It does, however, indicate the complexity of nested contexts established by co-text which, as hearers / readers, we are capable of interpreting.
In Chapter 6 we shall discuss the issue of anaphoric reference which is generally held to depend crucially on co-text for interpretation.
For the moment the main point we are concerned to make is to stress the power of co-text in constraining interpretation.
Even in the absence of information about place and time of original utterance, even in the absence of information about the speaker / writer and his intended recipient, it is often possible to reconstruct at least some part of the physical context and to arrive at some interpretation of the text.
The more co-text there is, in general, the more secure the interpretation is.
Text creates its own context.
As Isard (1975: 377) remarks: ' communications do not merely depend on the context for their interpretation, they change that context '.
The expanding context
In our discussion so far, we have been concerned to impose some sort of analytic structure on the lumpen mass of context.
We have abstracted away from particular contexts, across communicative contexts in general, to arrive at a set of features, some of which seem relevant to the identification of a speech event as being of a particular kind, to the ability of the hearer to predict what sort of thing the speaker is likely to say in a given type of context, and to the constraining of interpretation in context.
The observant reader will have noticed that we have helped ourselves to the content of the features proposed by Hymes and the co-ordinates proposed by Lewis in a fairly arbitrary way.
So we have given variable amounts of information about the speaker or the hearer or the time or the place as we have discussed different fragments of discourse.
This behaviour is consistent with Hymes' own expectations about how his framework would be used.
You will remember that he thought that contextual features might be considered in the way that general phonetic features are considered: sometimes, but not always relevant, and specifiable to variable degrees of delicacy for different purposes (2.2.
I).
A problem for the discourse analyst must be, then, to decide when a particular feature is relevant to the specification of a particular context and what degree of specification is required.
Are there general principles which will determine the relevance or nature of the specification, or does the analyst have to make ad hoc judgements on these questions each time he attempts to work on a fragment of discourse?
For the moment, we shall limit our discussion of this question to those features which relate directly to the deictic context, those features which will permit interpretation for deictic expressions like the temporal expression now, the spatial expression here, and the first person expression I. Are there standard procedures for determining what information is relevant to the interpretation of these expressions?
Lyons (1977: 570) suggests that there might, in principle, be such standard procedures:
Every actual utterance is spatiotemporally unique, being spoken or written at a particular place and at a particular time; and provided that there is some standard system for identifying points in space and time, we can, in principle, specify the actual spatiotemporal situation of any utterance act.
There clearly are standard systems for locating points in time and space.
It would be possible to specify the time of an utterance as stretching between say 9.33 a.m. and 9.34 a.m. on 5 June 1961, specifying the utterance in terms of clock and calendar time, good standard systems.
We could, then, presumably, if we had the relevant instrumentation, specify the place of the utterance in terms of a fine interaction of latitude and longitude.
It is not at all clear, however, that these particular standard systems produce the relevant information on all occasions.
Presumably some patrol ship on the high seas might log messages in this way, but it is clear that, as humans, our experience of utterances is not that we have recorded in memory a list of utterances to which are attached standard tags specifying time and place in these terms.
A friend can attempt to recall to your mind some utterance which you both experienced by a variety of place and time tags:
(20)
a.
But you just said he wasn't.
(Place: maintained; time: only minutes ago)
b.
You said in the staff meeting yesterday that he wasn't.
c.
You said last week at the staff meeting that he wasn't.
d.
You said last year when we met in Toronto that he wasn't.
The further away in time the message was situated, the less likely the speaker is to remember precisely the date and time at which it occurred, and the larger the time-span he is likely to make available for it to have occurred in.
It seems unlikely then, that ' standard procedures' of recording space and time are going to be relevant to the unique identification of utterance acts.
Perhaps the standard procedures will enable us to fix the relevant space spans for the interpretation of deictic expressions like here.
Suppose X is talking to Y, standing on the blue border of the carpet in X's office, in a given street, in Manchester, in England, in
Britain, in Western Europe...
Y might produce any of the following utterances:
(21)
a.
There's another worn section which needs repair here.
b.
You've got a very nice room here.
c.
It's a really nasty day here.
d.
You have a comparatively mild climate here.
It must be clear that the spatial location identified by here in each of these expressions could be interpreted as a series of concentric rings spreading out from the speaker and encompassing different amounts of physical space, but the interpretation of the spatial range of the expression here on any particular occasion of use will have to be sought in the context of what the speaker is talking about.
What appears to be stable in interpretations of here (apart from curious usages deriving from long-distance telephonic communication and long-distance travel, discussed in Lyons, 1977) is that the deictic centre is located where the speaker is.
Very similar problems arise with the interpretation of the temporal deictic expression now.
Consider the following possible utterances:
(22)
a.
Clap altogether NOW.
(gym mistress to class)
b.
I think you should begin the next chapter now.
(supervisor to student)
c.
Now I 'm getting older I really do find policemen look younger.
d.
From the iron age till now, man has been making increasingly complex artefacts.
In c and d the utterances appear to be located within different temporal spans, one relating to the speaker's advancing age (involving a span of 20C30 years) as opposed to the advancement of man (involving a span at least of decades and possibly centuries).
Utterances a and b are different in that the action specified is to follow the utterance, immediately in the case of a, but after some expanse of time in b.
Once again we suggest that the deictic centre is located within the context of utterance by the speaker, but that the interpretation of the expression now as relating duratively or subsequently to the utterance, and the time-span involved, must be determined with respect to the content of the utterance.
We should note that this fixing of the deictic centre is particularly appropriate to what Lyons (1977: 637) calls
the canonical situation of utterance: this involves one-one, or one-many, signalling in the phonic medium along the vocal-auditory channel, with all the participants present in the same actual situation able to see one another and to perceive the associated non-vocal paralinguistic features of their utterances, and each assuming the role of sender and receiver in turn.
It is, of course, possible to use the expressions here and now in what might be described as' displaced contexts'.
Consider how you would interpret the utterance We 'll land here said by one astronaut to another, on earth, as they study a map of the moon.
Or, how you interpret the message on each sheet of one brand of government-issue toilet roll, which reads NOW WASH YOUR HANDS, PLEASE.
Speakers, or writers, do have the option of transferring the deictic centre to the hearer 's, or reader 's, spatio-temporal situation in which the text will be encountered.
From our discussion of the spatio-temporal co-ordinates which seem, in principle, peculiarly accessible to standard specification, it must be obvious first, that deictic expressions may retain a standard deictic centre but must be interpreted with respect to the content of the utterance in which they occur and, second, that the relevant standard temporal description of an utterance, for instance 9.22 a.m. on Tuesday 28 June 1873, as opposed to in the late nineteenth century, will vary depending on the knowledge and intention of the analyst (or speaker) in referring to the utterance as located in time.
That is to say, even if there were an agreed, standard system for tagging utterances with spatio-temporal features, there is no guarantee that that tagging system provides the relevant information.
Thus in 2.2.
I. we discussed a fragment of discourse:
He seemed to resent them on that occasion and will not wear them today
where we specified the time of utterance as 4 June 1980.
The newspaper article from which this fragment was extracted did indeed appear on that date.
However, for anyone who knows what the expression the Derby means, it would almost certainly have been more informative to tag the time of utterance as Derby Day, 80.
The space-time co-ordinates can not be regarded as simple unstructured cues to interpretation in context.
Similarly, the other co-ordinates relevant to the deictic context, speaker, hearer and indicated object, can not be regarded as simple unstructured cues which demand standard specification.
What does it mean to specify, for instance, the indicated object co-ordinate?
We could identify a person by name.
We could report Ellen Blair said she'd like to come.
This might be adequate to identify the speaker, indeed the expression Ellen might be sufficient.
If, however, you do not know who this person is, or might be, it would be more helpful if we were to give some indication of why we have introduced her into the conversation.
So we might say my friend Ellen Blair, or the former chairman Ellen Blair, or a nurse in the ward called Ellen Blair, giving, in some sense, ' credentials' for her existence and for her relationship to the speaker who is responsible for introducing her into the conversation.
Morgan (1975: 442) asks' What can we infer about the speaker's intentions from the fact that he has chosen this particular description, rather than any of the others which would call to mind the same referent? '
For any individual there will be an immense number of possible descriptions which will be more or less appropriate in different contexts.
We may identify the person from external physical cues: the woman in the corner, the man with a beard, the student who has had his hair dyed, the child in the pink dress or, more or less flatteringly, the tall distinguished looking man I the man with a big nose and stringy hair.
We may identify people from a description of what they are doing: the woman who is chatting up the Admiral, the man who's fixing the car, etc.
The variable which interests us most is that which is concerned with the various roles played by the individual.
Lyons (1977: 574ff.) distinguishes between the deictic role of an individual (which assigns, for instance, first, second and third person pronouns) and his social role or ' status'.
Lyons points out that, for example, the terms of address used by a social inferior to a social superior may be different from those used between peers, as in vocative terms like ' Sir ' or ' Doctor ' or ' My Lord ' (in the courtroom).
In different social contexts, then, different terms of address will be found.
(Consider for instance, the distribution of the tu / vous pronouns in French.)
In general we may assume that, in a particular social context, only one role is taken by an individual at one particular time.
A glance at any newspaper will yield a rich crop of identifications of individuals in terms of the social role relevant to the news item.
Here are just a few:
(23)
a.
Daily Telegraph cartoonist Nicholas Carland showing how he sees the Prime Minister.
(Stop Press, 27 February 1982)
b.
Frank Silbey, chief investigator for the Senate Labor and Human Resources Committee, picked up his telephone.
(Time, 31 May 1982)
c.
Sophia Loren, the film actress, awoke in a prison cell in Caserta, near Naples, today.
(The Times, 21 May 1982d.
d.
Mr. Robert Mugabe, the Prime Minister of Zimbabwe yesterday sought to reassure prospective investors in his country.
(The Times, 21 May 1982)
e.
Senor Jorge Blanco of the ruling Revolutionary Party was officially declared winner.
(The Times, 21 May 1982)
In each case the individual is identified either by the role which is relevant to the content of the article, or by the role by which he is known to the public.
Each individual may play many other roles  parent, child, niece, brother, chess player, gardener, diarist, but these roles are not relevant in this context, so not mentioned on this occasion.
It is possible for more than one social role to be relevant at one time.
Rommetveit (1974: 45) discusses a sentence introduced in Chomsky (1972: 67):
I am not against MY FATHER, only against THE LABOR MINISTER
Rommetveit argues that the sentence is not necessarily self-contradictory even if the individual referred to by the two nominal expressions is the same individual.
It merely expresses the ambivalence which is a common human experience where some aspect of an entity pleases you and some other aspect fails to please.
Rommetveit argues against ' the notion of identifying reference as an unequivocally defined point in a monistic and epistemological transparent space, constructed on axiomatic prerequisites for specific operations within formal logic '... where ' the severe laws of truth values prescribe that the speaker must know him (the indicated entity) fully or not at all ' (1974: 48).
It is possible for speakers, hearers or indicated entities to be regarded from the perspective of more than one role.
Consider:
(24)
a.
As his neighbour I see quite a lot of him, as his colleague I hardly ever see him.
b.
As a colleague you're deficient but as a neighbour you're marvellous.
c.
I quite like her as a colleague and she's very pleasant as a casual friend but she's impossible to live with.
It is clear that we can hold partially or severely differing opinions about the same individual in different roles.
In the following extract from a report in The Times (15 May 1982) the same individual is referred to by a number of different expressions which relate to the multiple roles that the reporter considers relevant to the incident:
(25) Priest is charged with Pope attack (Lisbon, May 14) A dissident Spanish priest was charged here today with attempting to murder the Pope.
Juan Fernandez Krohn, aged 32, was arrested after a man armed with a bayonet approached the Pope while he was saying prayers at Fatima on Wednesday night.
According to the police, Fernandez told the investigating magistrates today he had trained for the past six months for the assault.
He was alleged to have claimed the Pope ' looked furious' on hearing the priest's criticism of his handling of the church's affairs.
If found guilty, the Spaniard faces a prison sentence of 15C20 years.
We have italicised the expressions relating to the man identified in the headline as Priest.
The relevance of his role as priest (referred to by the expressions Priest, a dissident... priest, the priest's) is presumably as a priest of the Roman Catholic Church of which the Pope is Head.
Since the incident reported takes place in Portugal (Lisbon) and any subsequent prison sentence will be served in Portugal, it is relevant that the priest is not Portuguese (a...
Spanish priest, the Spaniard).
A potentially confusing indefinite referring expression, a man armed with a bayonet, apparently relates back to the period before he was identified as' a dissident Spanish priest '.
He is identified by his name, as an individual, in the set constituted by the intersection of the various relevant roles (Juan Fernandez Krohn, Fernandez).
As Levy (1979: 193) remarks, ' the speaker by making reference may not simply identify but may construct the object by selecting from a field of relations those properties that are relevant at the moment of utterance '.
Consider the response of a five-and-a-half-year-old girl in a Yorkshire infant school where she is asked to say how two pictures are different from each other.
She replies:
(26)
a.
That one's over there in that but it in't there.
The teacher then holds the little girl's hands, so she can't point, shuts her own eyes and says to the child:
b.
Now I can't see the picture.
Tell me the difference again.
This time the child says:
c.
In this picture the teddy's on the chair but there ain't no teddy in that one.
The pictures are identical except in three respects: the presence or absence of a teddy bear sitting on the chair, a difference in the pattern on the counterpane, a difference in the position of a mirror.
For the child the teddy bear is clearly the salient object.
She relies in her first response on the teacher's access to the shared visual context to interpret what she says.
She points to the teddy bear (that one) in the first picture and then points to the empty chair in the second picture (there) and assumes that the teacher is paying attention to what she is pointing to in their shared context of situation.
When the teacher inhibits the child from pointing and pretends not to be able to see the picture, the child understands that the communicative situation has changed, that she can no longer rely on the shared visual context and she makes her reference explicit (the teddy), locates him verbally rather than by pointing to him (on the chair) and makes explicit how the second picture differs from the first (there ain't no teddy).
A salient aspect of the addressee, her ability to see what the child can see, has been changed by the utterance of b and the acts accompanying the utterance.
Speakers, hearers and indicated objects are not featureless, colourless spheres.
Nor do they come simply tagged with proper names appropriate to all occasions together with one identifying description appropriate to all occasions.
They are, characteristically, endowed with immense numbers of physical and social properties, any one of which may be the property which is relevant to a particular communicative act.
The philosopher's crisp index, which permits the identification of speaker and hearer as X and Y, is only relevant in a restricted model world.
The discourse analyst working in the real world has to be able to extract, see as relevant, just those properties of the features of context which are relevant to the particular communicative act which he is describing, and which contribute to the interpretation (or intended meaning) of the utterance.
As Enkvist (1980: 79) remarks, ' The context analyst's first embarrassment is richness. '
How is he to determine which properties of which features of context are relevant on a particular occasion?
Are there general principles to appeal to?
Is it reasonable to assume, as we tend to do, that those features of context which are salient to the speaker are equally salient to the hearer?
Ought we not rather to think in terms of partially intersecting views of context?
Bar-Hillel (1970: 79) states that ' the depth of the pragmatic context which is necessary for the full understanding of various sentence-tokens, is different, of course, from case to case '.
As yet we have only a very limited understanding of how we might set about determining ' the depth of the pragmatic context which is necessary ' for interpretation.
We outline a possible approach to the problem in the next section and in Chapter 3.
The principles of ' local interpretation ' and of' analogy'
In 2.3 we have discussed the problems for the discourse analyst in specifying what aspects of the apparently illimitable features of context are to be taken into account in the interpretation of discourse.
How is he to determine the relevant span of time in the interpretation of a particular utterance of ' now ' or the relevant aspects of a character referred to by the expression ' John '?
We must assume that the problem for the discourse analyst is, in this case, identical to the problem for the hearer.
There must be principles of interpretation available to the hearer which enable him to determine, for instance, a relevant and reasonable interpretation of an expression ' John ' on a particular occasion of utterance.
One principle which we can identify we shall call the principle of local interpretation.
This principle instructs the hearer not to construct a context any larger than he needs to arrive at an interpretation.
Thus if he hears someone say ' Shut the door ' he will look towards the nearest door available for being shut.
(If that door is shut, he may well say ' It's shut ', rather than consider what other doors are potentially available for being shut.)
Similarly if his host says' Come early ', having just invited him for eight o'clock, he will interpret ' early ' with respect to the last-mentioned time, rather than to some previously mentioned time.
Consider again extract (17) presented here as (27).
(27) a man
and woman sitting in the living room... the man's bored goes to the window looks out the window... and goes out + goes to his goes to a club + has a drink talks to the barman
In our discussion in 2.2.2, we pointed out the effect of ' co-text ' in limiting the interpretation of what follows.
The initial setting of the co-text determines the extent of the context within which the hearer will understand what is said next.
He assumes that entities referred to will remain constant, that the temporal setting will remain constant, that the locational setting will remain constant, unless the speaker indicates some change in any of these, in which case the hearer will minimally expand the context.
Not only does the hearer assume it is the same ' man ' who is being talked about throughout, he also assumes that the man will stay in the same place unless the speaker announces that he moves.
When the hearer hears goes to the window, he assumes it is' the window ' in that same ' living room ' which has already been mentioned, and he assumes that the man ' goes to the window ' on the same occasion, within minutes of the original setting ' sitting in the living room '.
When the man goes to a club, the hearer assumes that the ' club ' is in the same town, that the man has not caught an aeroplane and flown to Las Vegas.
Again the minimal expansion of the spatio-temporal setting will suggest that the man has a drink and talks to the barman within that same club and on that same occasion, within a restricted time-span, say an hour rather than a year.
It is this principle, which instructs the hearer not to construct a context any larger than necessary to secure an interpretation, which accounts for how we understand Sacks' (1972) much-quoted sequence:
(28) The baby cried.
The mommy picked it up.
It is possible, of course, to imagine that the first of these sentences describes one event and the second describes another, quite unrelated, event (so the person identified as' a mother ' may be picking up a chair in the course of cleaning a room).
The principle of local interpretation however, will guide us to construct a limited context in which ' the mother ' is the mentioned baby's mother and the expression it is used to refer to the previously mentioned baby.
Moreover the sequence of events will be understood as happening adjacently in time and situated adjacently in place.
It does not even occur to the reader that the baby might have cried one year in Singapore and be picked up by its mother a year later in Aden.
It would, of course, be possible to establish a setting in which such a sequence of events would be plausible, but, if no such setting is established, the reader will assume a local interpretation in respect of time, place and participants.
It must be obvious that ' local interpretation ' may only be vaguely conceptualised.
It seems unlikely that in interpreting (28) the reader postulates any exact physical distance between the mother and the baby at the point before the mother picks the child up, or that he bothers to wonder whether the mother picks the child up after it has finished crying (and if so how long after, in terms of minutes or seconds) or whether the child was still crying when the mother picked it up.
Similarly it seems unlikely that the reader will bother to construct a three-dimensional, photographic representation of ' the baby ' which cries in the first sentence and which is picked up in the second sentence.
' Local interpretation ' probably relates to another strategy which instructs the hearer / reader to do as little processing as possible, only to construct a representation which is sufficiently specific to permit an interpretation which is adequate for what the hearer judges the purpose of the utterance to be.
Everything that we have said so far in this section leans heavily on the hearer's / reader's ability to utilise his knowledge of the world and his past experience of similar events in interpreting the language which he encounters.
It is the experience of similar events which enables him to judge what the purpose of an utterance might be.
It is his knowledge of the world which constrains his local interpretation.
Consider again (27) presented here as (29).
(29) a man and woman sitting in the living room... the man's bored goes to the window... goes out... goes to a club
We suggested that goes to the window will be interpreted as meaning that ' he goes to the window in the living room ', whereas goes to a club will be interpreted as meaning ' goes to a club in the same town ', i.e. not ' in the living room ', nor even ' in the same house '.
Knowledge of the world tells us that houses which contain living rooms do not usually contain bars.
Goes out can not be simply interpreted as meaning ' goes out of the room ', it has to be interpreted as meaning ' goes out of the house '.
(In Chapter 7 we return to a discussion of ' knowledge of the world '.)
We must suppose that an individual's experience of past events of a similar kind will equip him with expectations, hypotheses, about what are likely to be relevant aspects of context.
Bartlett, one of the founders of modern psychology, comments on the importance of relating a particular experience to other similar experiences:
it is legitimate to say that all the cognitive processes which have been considered, from perceiving to thinking, are ways in which some fundamental ' effort after meaning ' seeks expression.
Speaking very broadly, such effort is simply the attempt to connect something that is given with something other than itself.
(1932: 227, our emphasis)
The individual, he suggests, generalises over particular experiences and extracts from these a number of types of experience.
This notion is, of course, implicit in the construction of the sets of features of context which we have been considering in this chapter.
In order to construct a notion of ' speaker in a context ' it is necessary to generalise over contexts and to determine what characteristics speakers in different contexts share.
Similarly, in order to construct a notion of ' genre ', it is necessary to generalise across experience and determine what it is that is common to fairy stories, chats, news broadcasts, epic poems, debates or salesmen's routines which enables us to recognise one as being a token of the generalised type.
On the basis of experience then, we recognise types of communicative events which take place against the background of a mass of below-conscious expectations also based on past experience which we might summarise, following van Dijk (1977: 99), as' the ASSUMED NORMALITY of the world.
We assume that our muscles will continue to move normally, that doors which normally open will continue to open, that hair grows on heads, that dogs bark, that towns retain their geographical locations, that the sun will shine, and so on.
It is interesting to observe the powerful constraints on creators of surrealist or science fiction in this respect.
Alice may enter a looking-glass world where unexpected things happen, but she is still constituted like a human being: walking may take her in an unexpected direction, but the nature of the physical act of walking is taken for granted.
If too many expectations are flouted, the writer may be suspected of being mentally unbalanced, of being incapable of seeing the world in a normal way.
Thus, on the one hand, expectations make interpretation possible and, on the other, they constitute an extension or further affirmation of their own validity.
Popper makes the point cogently: ' we are born with expectations: with ' knowledge ' which, although not valid a priori, is psychologically orgenetically a priori, i.e. prior to all observational experience.
One of the most important of these expectations is the expectation of finding a regularity.
It is connected with an inborn propensity to look out for regularities, or with a need to find regularities' (1963: 47, original emphasis).
Furthermore, as Lewis (1969: 38) points out, ' fortunately we have learned that all of us will mostly notice the same analogies'.
Not only are we all primed to look for regularities, we tend to perceive the same regularities.
Clearly the smaller the community, the more notions of regularity will be shared, since the contexts which the members of the community share will be very similar.
Once the individual begins to establish regularities, to generalise over experience, it becomes possible for him not only to recognise a particular experience as being one of a type, say a scolding or an interview, it also becomes possible to predict what is likely to happen, what are likely to be the relevant features of context, within a particular type of communicative event.
It follows that the hearer in a speech situation is not in the position of trying to pay attention to every feature of the context (in principle an impossible task).
He only pays attention to those features which have been necessary and relevant in similar situations in the past.
Bartlett suggests that the individual has' an overmastering tendency simply to get a general impression of the whole; and on the basis of this he constructs the probable detail ' (1932: 206).
We pay attention to those salient features which are constitutive of the type of genre, and expect that the peripheral features will be as they have been in the past.
Obviously there will be types of occasions which have not occurred within our past experience.
We have cultural stereotypes which suggest that such occasions are difficult for us, potentially embarrassing, because we do not know the appropriate responses.
Thus, if it is the first time someone tells you a particular genre of joke, you may not know the appropriate type of response.
The second time around, however, you feel more confident of what to expect.
(Tolstoy, in War and Peace, gives a brilliant account of the insecurity engendered by the first occasion of a new type of experience in his description of Pierre's induction into membership of a masonic brotherhood.)
Our experience of particular communicative situations teaches us what to expect of that situation, both in a general predictive sense (e.g. the sort of attitudes which are likely to be expressed, the sort of topics which are likely to be raised) which gives rise to notions of ' appropriacy ', and in a limited predictive sense which enables us to interpret linguistic tokens (e.g. deictic forms like here and now) in the way we have interpreted them before in similar contexts.
We must assume that the young child's acquisition of language comes about in the context of expanding experience, of expanding possible interpretations of forms like here and now in different contexts of situation, contexts which come to be recognised, and stored as types.
Against the background of this mass of expectations which derives from and constitutes our experience, it must become possible to identify the relevant properties of features of the context of situation in terms of norms of expectation within a particular genre.
The more highly constrained and ritualised the genre, the more likely we are to be able to identify norms.
Thus it seems likely that examination questions in chemical engineering at degree level will bear certain similarities of form and content, and share certain presuppositions, in institutions throughout the world.
The less constrained the genre, primarily interactional ' chat ', for example, the less likely it is that we can confidently state norms of expectation which will generalise even over the experience of the English-speaking population.
For the individual participant in a chatting relationship, this does not constitute a difficulty, because he has plenty of previous personal and local experience to call upon.
For the discourse analyst, on the other hand, the more personal and particular the occasion for the participants, the more limited and circumspect he must be in his interpretation.
Confronted with data of the following sort, an extract from a private diary only intended to remind the elderly writer of how she passed a day in January 1982, the discourse analyst may not be able to proceed very far in his analysis.
(30) Did more to Ivy's letter.
A.A. rang me at 4 o/c she returned on 2nd and had had grand time with Gwenda and families.
As was nice p.m.
I went to Evensong (rev.
Carlil) and walked back with Mrs. Nicholls (85!!) and daughter.
Cos' Doris rang 8.15 and will come tomorrow!
Bed.
11.15.
Of course, if the discourse analyst experiences a great deal of data like this, he will feel more confident in his description and interpretation.
He, too, is constrained in his interpretation by past similar experience, by interpreting in the light of what we might call the principle of analogy.
The principle of analogy will provide a reasonably secure framework for interpretation for the hearer and for the analyst most of the time.
Most of the time, things will indeed conform to our expectations.
However, conventions can be flouted and expectations upset, either deliberately for a stylistic effect, or by accident or oversight.
Note that where the speaker / writer is deliberately flouting a convention, upsetting an expectation for a stylistic effect, he can only bring off that effect because the convention / expectation exists.
The ' non-limerick ' which follows only makes an effect in the light of the conventional structure for limericks which have a characteristic rhythm and an aabba rhyme scheme:
(31) There was a young girl of St Bees, Who was stung on the nose by a wasp When asked ' Does it hurt? '
She replied ' Yes it does, But I 'm glad it wasn't a hornet. '
The principle of analogy is one of the fundamental heuristics which hearers and analysts adopt in determining interpretations in context.
They assume that everything will remain as it was before unless they are given specific notice that some aspect has changed.
Dahl (1976: 46) formulates a principle for speakers: ' Indicate only things which have changed and omit those which are as they were before. '
To repeat what is known to be shared knowledge, ' things as they were before ', flouts Grice's maxim of quantity.
(Speakers do, of course, remind each other of knowledge which they share, in order to make that knowledge part of the activated context of discourse, as McCawley (1979) points out.)
Discourse is interpreted in the light of past experience of similar discourse, by analogy with previous similar texts (remember the relevance of experience of previous similar texts in the interpretation of (14C) in Chapter 2, SQUASHED INSECTS DO NT BITE MAD MENTAL RULE).
Relevant previous experience, together with the principle of local interpretation, will impel hearers / readers to try to interpret sequential utterances as relating to the same topic.
When two sentences are placed together in sequence by a writer who does not want us to consider them as a continuous text, their separateness or disconnectedness must be positively indicated.
In a linguistics textbook, the following two sentences were presented as separate citation examples to illustrate structural ambiguity.
(32)
1.
The bride and groom left early last night.
2.
He greeted the girl with a smile.
(Brown &amp; Miller, 1980: 84)
In the context of a linguistics textbook, especially one on syntax, we would not expect to have to interpret two continuous cited sentences as describing an event sequence.
In most contexts, however, the natural ' effort after meaning ' will impel the hearer / reader to try to co-interpret chunks of language which he finds close to each other on a page, or a stone or a wall and, where possible, to interpret the language as relevant to the physical context.
This last point leads us to an important, but frequently misunderstood, concept in the analysis of discourse.
The imperative 'need to find regularities' which Popper speaks of, coupled with Bartlett's ' effort after meaning ', constitute a powerful expectation in human beings that what is said or written will make sense in the context in which it appears.
Even in the most unpropitious circumstances, the natural reaction of man appears to be to make sense of any sign resembling language, resembling an effort to communicate.
The reaction of the man who finds what are apparently signs etched in a stone in the middle of a desert is to try to decipher their meaning.
The reaction of parents to infants, and of friends to the speech of those who are gravely ill, is to attribute meaning to any murmur which can be interpreted as relevant to the context of situation, and, if at all possible, to interpret what appears to be being said as constituting a coherent message, permitting the hearer to construct a coherent interpretation.
The natural effort of hearers and readers alike is to attribute relevance and coherence to the text they encounter until they are forced not to.
The normal expectation in the construction and interpretation of discourse is, as Grice suggests, that relevance holds, that the speaker is still speaking of the same place and time, participants and topic, unless he marks a change and shows explicitly whether the changed context is, or is not, relevant to what he has been saying previously.
Similarly the normal expectation is that the discourse will be coherent.
The reaction of some scholars to the question of ' coherence ' is to search for cues to coherence within the text and this may indeed yield a descriptive account of the characteristics of some types of text.
It ignores, however, the fact that human beings do not require formal textual markers before they are prepared to interpret a text.
They naturally assume coherence, and interpret the text in the light of that assumption.
They assume, that is, that the principles of analogy and local interpretation constrain their experience.
There are as many linguistic ' cues to coherence ' (a concept to be discussed in detail in Chapter 6) holding between the pairs of sentences:
(33)
1.
The bride and groom left early last night.
2.
He greeted the girl with a smile.
as there are between:
(34) The baby cried.
The mommy picked it up.
It is not the sequence of sentences which represents' coherent discourse '.
Rather it is the reader, driven by the principles of analogy and local interpretation, who assumes that the second sequence describes a series of connected events and interprets linguistic cues (like baby  it) under that assumption.
Encountering the first pair of sentences in the context in which they occur, the reader does not assume that they describe a connected sequence of events and consequently does not interpret the potential linguistic cues (like groom  he) as referring to the same entity.
The principles of analogy (things will tend to be as they were before) and local interpretation (if there is a change, assume it is minimal) form the basis of the assumption of coherence in our experience of life in general, hence in our experience of discourse as well.
3
Topic and the representation of discourse content
In the course of this chapter, we shall examine some of the uses of the term topic in the study of discourse.
In the process, we shall explore some recent attempts to construct a theoretical notion of ' topic ', a notion which seems to be essential to concepts such as' relevance ' and ' coherence ', but which itself is very difficult to pin down.
We shall suggest that formal attempts to identify topics are doomed to failure, but that the discourse analyst may usefully make appeal to notions like ' speaking topically ' and ' the speaker's topic ' within a ' topic framework '.
We shall also consider briefly how markers of ' topic-shift ' may be identified in written and spoken discourse.
In particular, we shall insist on the principle that it is speakers and writers who have topics, not texts.
We shall then go on to consider how the notion of ' topic ' relates to representations of discourse content.
Since many of the representations proposed are based on a hierarchical organisation of discourse content, we shall consider critically the possibility of characterising ' topic ' in terms of the top-most elements in the hierarchical representation.
Discourse fragments and the notion ' topic'
We have already argued that the data used in discourse analysis will inevitably reflect the analyst's particular interests.
Moreover, the piece of data chosen for study can only be partially analysed.
If the investigation is undertaken by someone primarily interested in intonation, for example, the data selected has to meet certain requirements.
It must be spoken, audible, and, depending on the level of investigation involved, clear enough to allow instrumental analysis, and accompanied by additional information on the age, sex and linguistic background of the speaker.
In practice, any single investigation will have much stricter data requirements than this rather general list.
Having selected the data, the investigators will study features such as the pitch, rhythm and loudness of syllables in the data, and spend relatively little or no time studying the lexis or the morphology.
In its most extreme form, this narrowing of the investigation in terms of the data selected and the analysis undertaken can lead to a constructed text being carefully read aloud in a phonetics laboratory by a speaker of standard Southern British English.
The results of the investigation may then be used to make ' empirical ' claims about the intonation of English.
Although this is an extreme example, it serves to illustrate the selectiveness which characterises linguistic investigation generally, and which is also present to a certain degree in most analysis of discourse.
The data studied in discourse analysis is always a fragment of discourse and the discourse analyst always has to decide where the fragment begins and ends.
How does the analyst decide what constitutes a satisfactory unit for analysis?
There do exist ways of identifying the boundaries of stretches of discourse which set one chunk of discourse off from the rest.
Formulaic expressions such as' Once upon a time... and they lived happily ever after ' can be used explicitly to mark the boundaries of a fragment.
Other familiar markers are ' Have you heard the one about...? ',
' Did I tell you what happened to me last week.
? ' and various other forms which can be used to mark the beginning of a joke or anecdote.
These markers can help the analyst decide where the beginning of a coherent fragment of discourse occurs.
However, speakers often do not provide such explicit guidelines to help the analyst select chunks of discourse for study.
In order to divide up a lengthy recording of conversational data into chunks which can be investigated in detail, the analyst is often forced to depend on intuitive notions about where one part of a conversation ends and another begins.
There are, of course, points where one speaker stops and another starts speaking, but every speaker-change does not necessarily terminate a particular coherent fragment of conversation.
Which point of speaker-change, among the many, could be treated as the end of one chunk of the conversation?
This type of decision is typically made by appealing to an intuitive notion of topic.
The conversationalists stop talking about ' money ' and move on to ' sex '.
A chunk of conversational discourse, then, can be treated as a unit of some kind because it is on a particular ' topic '.
The notion of ' topic ' is clearly an intuitively satisfactory way of describing the unifying principle which makes one stretch of discourse ' about ' something and the next stretch ' about ' something else, for it is appealed to very frequently in the discourse analysis literature.
Yet the basis for the identification of ' topic ' is rarely made explicit.
In fact, ' topic ' could be described as the most frequently used, unexplained, term in the analysis of discourse.
Sentential topic
One use of the term ' topic ' is associated with descriptions of sentence structure.
According to Hockett, a distinction can be made between the topic and the comment in a sentence, in that ' the speaker announces a topic and then says something about it...
In English and the familiar languages of Europe, topics are usually also subjects and comments are predicates' (1958: 201).
It is clear from Hockett's examples, reproduced here as (1) and (2), that this' sentential topic ' may coincide with the grammatical subject, as in (1), but need not, as in (2).
(1) John / ran away
(2) That new book by Thomas Guernsey / I haven't read yet
The treatment of ' topic ' as a grammatical term, identifying a constituent in the structure of a sentence (or the deep structure analysis, at least) is also noticeable in the work of grammarians such as Dahl (1969) and Sgall et al.
(1973).
Transformational generative grammars would also account for the structure of example (2) in terms of a movement transformation called ' topicalisation '.
The term ' topic ', then, as found in descriptions of sentence structure, is essentially a term which identifies a particular sentential constituent.
As such, it has been used in the study of discourse, by Grimes (1975: 337) for example, to describe the different methods used in various languages to mark the ' topic constituent ' of sentences.
It has also been used by Givn (1979a) in his argument that, in the development of a language, sentential subjects are derived from ' grammaticalised topics'.
However, we are not, for the moment, concerned with the structure of linguistic units comparable to the simple sentence (see Chapter 5).
Nor are we considering ' topic ' as a grammatical constituent of any kind.
We are primarily interested in the general pretheoretical notion of ' topic ' as' what is being talked about ' in a conversation.
This type of ' topic ' is unlikely to be identifiable as one part of a sentence.
Accordingly, we agree with Morgan that ' it is not sentences that have topics, but speakers' (Morgan, 1975: 434).
Discourse topic
In an attempt to distinguish their notion of topic from the grammarians' sentential topic, Keenan &amp; Schieffelin (1976) used the term discourse topic.
They were particularly anxious to avoid having ' topic ', in discourse study, treated as if it were somehow expressible by a simple noun phrase, as often happens in the treatment of sentential topics.
(Some ontological reasons for this type of treatment are suggested by Lyons, 1977: 502.)
What Keenan &amp; Schieffelin (1976: 380) emphasise is that ' discourse topic is not a simple NP, but a proposition (about which some claim is made or elicited) '.
It may be because their investigation is primarily concerned with children's speech, but, in describing the discourse topic as the ' question of immediate concern ', Keenan &amp; Schieffelin appear to replace the idea of a single correct noun phrase as expressing the topic with the idea of a single correct phrase or sentence.
The implication in their study is that there must be, for any fragment of conversational discourse, a single proposition (expressed as a phrase or sentence) which represents the discourse topic of the whole of the fragment.
Such a view is certainly too simplistic, as we hope to show by considering some experimental work in which ' the topic ' was treated as the equivalent of a title.
(We shall consider the possibility of representing ' the discourse topic ' as a proposition when we investigate the proposition-based analysis of discourse in section 3.7.)
In a series of experiments reported by Bransford &amp; Johnson (1973) subjects were presented with constructed texts to read, comprehend, and, later, recall.
The aim of the experiments was to demonstrate that the comprehension of English texts depends not only on knowledge of the language, but also on extra-linguistic knowledge, particularly related to the contexts in which the texts occur.
There are examples of texts which appear to depend on accompanying visual material for comprehension and others, such as example (3) reproduced below, for which ' the topic ' must be provided.
(3) The procedure is actually quite simple.
First you arrange things into different groups.
Of course, one pile may be sufficient depending on how much there is to do.
If you have to go somewhere else due to lack of facilities that is the next step, otherwise you are pretty well set.
It is important not to overdo things.
That is, it is better to do too few things at once than too many.
In the short run this may not seem important but complications can easily arise.
A mistake can be expensive as well.
At first the whole procedure will seem complicated.
Soon, however, it will become just another facet of life.
It is difficult to foresee any end to the necessity for this task in the immediate future, but then one never can tell.
After the procedure is completed one arranges the materials into different groups again.
Then they can be put into their appropriate places.
Eventually they will be used once more and the whole cycle will then have to be repeated.
However, that is part of (from Bransford &amp; Johnson, 1973: 400)
Because it was constructed for a specific purpose, this text is fairly unusual in that there are few lexical clues to what the text might be ' about '.
Predictably, the experiments showed that comprehension and recall of this passage were significantly better when subjects were provided, before reading, with what Bransford &amp; Johnson called ' the topic of the passage '.
The topic of this passage was' Washing clothes'.
The reader can judge for himself whether his comprehension would have been fuller if he had known this topic.
The use of the word ' topic ' in this type of experiment suggests that the topic of a text is equivalent to the title and that, for any text, there is a single correct expression which is' the topic '.
This would be the case if texts could only be understood completely as long as they were accompanied by the single, correct title.
However, it should not be too difficult to imagine several different titles for passage (3), each of which could equally facilitate comprehension.
One could indicate that the text contains a set of instructions by producing a title such as' How to Do the Laundry ' or ' A Guide to Getting your Clothes Cleaner '.
One could incorporate the text's philosophical final statement in a title such as' Doing the Laundry as a Philosophy of Life ' or ' An Orderly Life through Good Laundry Procedure '.
These latter titles contain as much information for the reader as the title ' Washing Clothes', which Bransford &amp; Johnson describe as' the topic '.
The implication, surely, is that, for any text, there are a number of possible titles.
Correspondingly, we will suggest, there is, for any text, a number of different ways of expressing ' the topic '.
Each different way of expressing ' the topic ' will effectively represent a different judgement of what is being written (or talked) about in a text.
As an illustration of this point, consider the text in (3) as a dusty fragment, recovered during an archaeological dig in the ruins of Minneapolis in the year 2500 A.D. When asked what the text is' about ', the discourse analyst in the expedition might report that it is about ' procedures used in mid-twentieth-century American middle-class culture for maintaining cleanliness in their garments'.
(Note the temporal and locational elements included here-elements which we shall consider more fully later.)
Another discourse analyst, providing a second opinion, might report that it is about something else entirely, and a debate would ensue in the discourse analysis literature.
The same ' text ' is considered by both analysts.
Their disagreement would be over ways of expressing ' the topic '.
(Literary critics are still exercised about the topic of Hamlet.)
The difficulty of determining a single phrase or sentence as' the topic ' of a piece of printed text is increased when fragments of conversational discourse are considered.
In any conversation, ' what is being talked about ' will be judged differently at different points and the participants themselves may not have identical views of what each is talking about.
People do, however, regularly report on what a conversation was' about '.
There are informal ways of expressing the topic, even in conversational discourse.
Topic framework
The discourse analyst, then, is faced with several problems when he wishes to use the very attractive pretheoretical notion of ' topic ' as' what is being talked/written about '.
The notion is attractive because it seems to be the central organising principle for a lot of discourse.
It may enable the analyst to explain why several sentences or utterances should be considered together as a set of some kind, separate from another set.
It might also provide a means of distinguishing fragments of discourse which are felt to be good, coherent, examples of English from those that are, intuitively, incoherent concatenations of sentences.
Consider, for example, the following discourse fragment, taken from Rochester &amp; Martin (1979: 95).
(4)
Interviewer: A stitch in time saves nine.
What does that mean?
Thought-disordered Speaker: Oh! that's because all women have a little bit of magic to them  I found that out  and it's called  it's sort of good magic  and nine is sort of a magic number + like I've got nine colors here you will notice  I've got yellow, green, blue, grey, orange, blue, and navy  and I've got black  and I've got a sort of clear white  the nine colors to me they are the whole universe  and they symbolize every man, woman and child in the world +
Rochester &amp; Martin attempt to describe the connections existing between sentences in discourse of this type, produced by thought-disordered and schizophrenic speakers, in terms of conceptual associations and lexical ties.
They point out, however, that such connections are ' unrelated to the conversational topic '.
The notion of ' topic ', though undefined, seems to provide Rochester &amp; Martin with a natural criterion for distinguishing between the connected, yet incoherent, discourse of thought-disordered speakers and the coherent discourse of normal speakers.
If there are, as we have already argued, a potentially large number of different ways of expressing ' the topic ' of even a short written text, how does the analyst determine which is the one correct expression of the topic for the text?
One answer, of course, is to say that, for any practical purposes, there is no such thing as the one correct expression of the topic for any fragment of discourse.
There will always be a set of possible expressions of the topic.
In the terms used by Tyler (1978: 452), the ' topic ' can only be ' one possible paraphrase ' of a sequence of utterances.
What is required is a characterisation of ' topic ' which would allow each of the possible expressions, including titles, to be considered (partially) correct, thus incorporating all reasonable judgements of ' what is being talked about '.
We suggest that such a characterisation can be developed in terms of a topic framework.
In Chapter 2, we discussed the problem for the discourse analyst of deciding just what features of context were relevant in the interpretation of a particular fragment of discourse.
We suggested there that the strategy available to him would be, on the one hand, to work predictively in terms of his previous experience (similar speakers, similar genres, etc.) and on the other hand to examine the content of the text.
From the content of the text the analyst can, in principle, determine what aspects of the context are explicitly reflected in the text as the formal record of the utterance.
Those aspects of the context which are directly reflected in the text, and which need to be called upon to interpret the text, we shall refer to as activated features of context and suggest that they constitute the contextual framework within which the topic is constituted, that is, the topic framework.
As a way of characterising the type of feature which will be required in a topic framework, we shall examine a fragment of conversational discourse and try to determine what is' being talked about '.
The fragment, presented as (5), is not a constructed piece of text, it is taken from a recorded conversation.
As an example of discourse analysis data, it has been selected for a particular purpose.
It is not a difficult fragment to work with, it has a definable beginning and end, and, for most of the fragment, there is one participant talking, in response to another's request for information.
This request for information provides a direction for the conversational fragment, so that we are considering speech with some purpose and not just social chat used to pass the time.
One might also say that the content of the request for information could provide some basis for the content of the response, especially when the request is for the meaning of an expression to be given.
That is, it would seem, at first glance, to be a simple matter to produce ' the topic ' for this discourse fragment, for it is contained in the question asked.
Immediately prior to the following extract, the speaker has been asked the meaning of the expression, ' smoke the houses'.
(5)
R: in those days + when we were young + there was no local fire engine here + it was just a two-wheeled trolley which was kept in the borough + in the borough eh store down on James Street + and whenever a fire broke out + it was just a question of whoever saw the fire first yelling ' Fire ' + and the nearest people ran for the trolley and how they got on with it goodness knows + nobody was trained in its use + anyway everybody knew to go for the trolley + well + when we were children + we used to use this taw [ tC: ] + it smouldered furiously + black thick smoke came from it and we used to get it burning + and then go to a letter box and just keep blowing + open the letter box + and just keep blowing the smoke in + you see + till you'd fill up the lower part of the house with nothing but smoke + there was no fire + but just fill it up with smoke + just to put the breeze up + just as a joke + and then of course + when somebody would open a window or a door the smoke would come pouring out + and then + everybody was away then for the trolley + we just stood and watched all of them + +
S: so that's what ' smoke the houses' is?
R: probably + probably + we called it ' the taw ' +
If we were to say that the topic of this discourse fragment is' the meaning of the expression ' smoke the houses' ', we could not claim to have said very much of analytic interest.
It may be that, for participant S, the above expression represents the best way of summarising what speaker R was talking about, as evidenced by her response.
However, even if we take that summarising phrase as one possible expression of the topic of speaker R's lengthy contribution, we have surely not adequately characterised what this speaker was talking about.
We might suggest that the speaker is talking about a joke or a prank.
In doing so, he talks about an object called ' the taw ' which produces a lot of smoke.
He talks about the process of putting the smoke into houses through the letter box and how smoke would come out of the window or door.
He also talks about an object known as the trolley, a type of fire engine, and the events associated with its use.
He talks about people going for the trolley when the smoke comes out of a house.
Thus one account of what this speaker is talking about would contain the following elements: a joke  the taw  smoke  into houses  out of houses  people get trolley  the use of the trolley.
This set of objects and events could be taken as a set of elements which would have to be included in a representation of this speaker's topic, i.e. what he was talking about.
It is not a complete set.
In this fragment, the speaker is also talking ' about ' a particular time and place, and ' about ' a specific person.
He is talking about his own childhood (when we were children) in Stornoway (here).
This last element presents a problem, because there is nothing in the text of the conversational fragment to indicate this location.
Yet it is a piece of knowledge relevant to what the speaker is talking about and, importantly, knowledge which the speaker assumes is available, to his hearer.
Presumably, the speaker can also assume that, because his hearer knows, approximately, the speaker's age, the hearer can judge the time (i.e. forty years before and not ten years before) of the events described.
Aspects of the speaker's assumptions about his hearer's knowledge must also be considered in relation to the elements which the speaker does make explicit in his contribution.
Do the first lines of this fragment contribute to answering the question asked?
Strictly speaking they do not.
Yet one would hesitate to describe these lines as irrelevant.
They are relevant to what the speaker wishes to provide as an answer to the question, given the particular hearer he has.
This young American hearer, visiting Stornoway, may have a quite inappropriate idea of the type of object, and the associated behaviour, involved in dealing with a fire in Stornoway forty years before.
Without knowing about the trolley, the hearer may not (in the speaker's assessment perhaps) appreciate the full flavour of the joke or prank being described.
It may be argued that this last point has more to do with why the speaker talked about something than with what he talked about.
Any consideration of topic involves asking why the speaker said what he said in a particular discourse situation.
As Coulthard (1977: 76), following Sacks (1971), points out, there is a constant analysis in conversation of what is said in terms of ' why that now and to me '.
In the present discussion we have already partially answered the reader's primary ' why ' question about the discourse fragment being studied by providing the previous speaker's question.
That is, attempting to provide an account of what a person is talking about is always built on an assumption that we know why that person says what he says.
The point may be clearer if we consider a possible reaction to the expression, ' Roses are red, violets are blue ' being included in (5) after the speaker has said nobody was trained in its use.
Would the expression simply be included in the list of what was talked about, or would it prompt the question ' Why does he say that here? '
The acceptance of extract (5) as a reasonable piece of English conversational discourse involves implicitly assessing each expression in terms of the ' why? ' question above and finding a suitable answer.
Part of the process of analysing discourse in terms of ' topic ' is an attempt to make explicit the basis for our intuitive ability to recognise why what is said is appropriate in a particular discourse fragment.
Certain elements which constrain the topic can be determined before this discourse begins.
These elements are part of what, in the previous chapter, were described as the context of a speech event.
In relating contextual features to a particular speech event, however, we are particularly interested in only those activated features of context pertaining to the fragment of discourse being studied.
For example, aspects of the time and place of the discourse in (5) are important because they have a bearing on what the speaker says in the fragment (forty years after the described event took place, but still in Stornoway).
Similarly, certain facts about the speaker and hearer, as we pointed out earlier, have to be included.
As a first approximation, then, we could produce a partial representation of a ' framework ' for extract (5) in terms of the following set of activated contextual features.
Conversation between Participant R (50+ years, Scottish, male,...) and Participant S (20+ years, American, female,...) in location p (Stornoway,...) at time t (late 1970s,...)
This simple set of features which we have claimed are necessary for a discussion of topic are required, quite independently of topic considerations, in any form of discourse analysis.
For ethnographers and sociolinguists considering linguistic interaction, these elements and others have to be made explicit in the analysis of features such as code-switching and role-relationships.
For the formal semanticist, these elements are required in the assignment of values to indexicals such as I, you, here and now.
That is, in building a framework for the analysis of topic, we are not adding any machinery to the apparatus of the discourse analyst which he does not have to employ already.
Those contextual features we have described above are, of course, derived from the physical context.
They are external to the text.
There is, for most conversational fragments, a set of discourse-internal elements which are derived from the conversation prior to the particular fragment being studied.
These elements are introduced in the preceding co-text and form part of what has been described as' the domain of discourse ' (cf.
Karttunen, 1974).
Within the domain of a particular discourse fragment are the people, places, entities, events, facts, etc. already activated for both participants because they have been mentioned in the preceding conversation.
If the fragment of discourse one wished to study was only the part of (5) beginning, when we were children we used to use this taw, then accounting for the speaker's mention of the trolley near the end of this fragment would have to be done in terms of the preceding discourse (i.e. all the first section before the taw is mentioned) in which the trolley is introduced and characterised.
We have introduced some basic components which would be required in a characterisation of the topic framework for any discourse fragment.
The topic framework consists of elements derivable from the physical context and from the discourse domain of any discourse fragment.
Notice that we have concentrated on only those elements which are activated, that is, relevant to the interpretation of what is said.
If we say that characterising the topic framework is a means of making explicit some of the assumptions a speaker can make about his hearer's knowledge, we are not talking about the total knowledge which the speaker believes he shares with his hearer.
We are describing only that activated part which is required in the analysis of the discourse fragment under consideration.
This approach is crucially different from some other proposals we shall examine.
Presupposition pools
What we have described as a topic framework has much in common with Venneman's proposal that, for a discourse, there is a presupposition pool which contains information ' constituted from general knowledge, from the situative context of the discourse, and from the completed part of the discourse itself ' (Venneman, 1975: 314).
In this approach, each participant in a discourse has a presupposition pool and his pool is added to as the discourse proceeds.
Each participant also behaves as if there exists only one presupposition pool shared by all participants in the discourse.
Venneman emphasises that this is true in ' a normal, honest discourse '.
Within the presupposition pool for any discourse, there is a set of discourse subjects and each discourse is, in a sense, about its discourse subjects.
Because it is part of the shared assumptions of the discourse participants that these discourse subjects exist, they do not need to have their existence asserted in the discourse.
Examples of expressions used for discourse subjects might be the Queen, John, John's wife (in the presupposition pool by virtue of general knowledge), your hat, today (from the situative context) and a concert of the Berlin Philharmonic's last year, several essays (from the preceding part of the text of the discourse).
The number of discourse subjects in a presupposition pool shared by participants in a discourse, particularly participants who know each other quite well, is potentially very large.
How does the discourse analyst decide which discourse subjects to include in the presupposition pool for a particular piece of conversational discourse?
Remembering that any discourse data to which the analyst has access will only be a fragment, it would be extremely difficult for the analyst to predetermine the complete set of discourse subjects which participants share prior to a particular discourse fragment.
The most he could hope to provide would be a partial set.
The problem to be faced is that of limiting the choice of the contents of even a partial set, in some non-arbitrary way.
The most important principle involved in this selection of Venneman's discourse subjects must have to do with their relevance to the particular discourse fragment under consideration.
If, in a stretch of conversational discourse, the participants involved can be independently known to have potential discourse subjects such as' the Queen ', ' the Pope ', or even ' the King of Siam ', within their shared presupposition pool, but do not mention the individuals, so identified, in their conversation, it is surely unnecessary to refer to those individuals in the analysis of that particular discourse fragment They are, in our terms, not ' activated '.
This would lead to the conclusion that the relevant ' discourse subjects' for a particular discourse fragment must be those to which reference is made in the text of the discourse.
If ' mentioned-in-the-text ' is taken as the basis for selection of discourse subjects, it should be noted that the analyst is, in fact, attempting to reconstitute the presupposition pool which the participants must have had prior to the discourse fragment being analysed.
Such a process may be comparable to the experience one has when switching on the radio in the middle of a discussion programme and trying to understand the discussion through a partial reconstruction of what must have been said already, who the participants must be, and so on.
It does suggest that the only information the discourse analyst has access to is that contained in the text of a discourse fragment.
Sentential topic and the presupposition pool
Of course, the data for discourse analysis is not limited to anonymous, decontextualised texts, as we have argued already in Chapter 2.
Concentrating solely on the text, however, remains a common approach in many accounts of discourse.
It is also characteristic of this approach that the text to be analysed is constructed by the analyst to illustrate the points he wishes to make.
This, unfortunately, is the method used by Venneman who, despite the promising breadth of analysis suggested by the concept of a presupposition pool shared by participants, restricts his investigation to describing the relationship between pairs of sentences.
The notion of ' topic ' considered by Venneman reflects the limitations of his investigation.
He considers
the expression ' topic ' or ' topic of a discourse ' as referring to a discourse subject on which the attention of the participants of the discourse is concentrated.
Such concentration of attention is usually, though not always, brought about by an immediately preceding textual mentioning of the discourse subject.
(Venneman, 1975 317)
This definition of topic has a certain intuitive appeal, in the sense that what two participants are concentrating on, in their conversational talk for example, is a reasonable candidate for ' the topic '.
There are, however, two basic problems here.
First, this definition of topic seems to be based on the same ' topic = single term title ' notion which we challenged earlier.
As we pointed out then, although a stretch of discourse can appear to be largely concerned with a single individual, or one discourse subject, so that the discourse may be loosely reported as being ' about ' that individual, this should not lead us to claim that all discourses are about single individuals or can be given convenient one-word titles.
A second objection is that it is far from clear how we would decide, in any principled way, what the participants in a discourse fragment are, in fact, ' concentrating ' on.
An attempt is made by Venneman to provide a formal means of identifying the topic in a discourse fragment.
He suggests that like ' all phenomena whose unique existence is presupposed, topics can be referred to by means of individual names, deictic expressions, and definite descriptions' (Venneman, 1975: 317).
Using this guide, the analyst must find that the following two discourse fragments, one each from stretches of spoken and written discourse, have several such ' topics'.
(6) what was interesting was that little Richard came home from his Toronto school with his Newfie jokes the content of which the substantive content was identical to Irish jokes which my son comes home with from Edinburgh schools
(7) so can he, but the main point about this system is the strain it puts on the other players
What is' the topic ' of (6)  little Richard or his Toronto school or his Newfie jokes, etc.; and is he, this system or the other players the topic of (7)?
It is possible to make a guess at what the speaker of (6) and the writer of (7) were concentrating on, but the guess is probably based on an elaborate reconstruction of what the most probable context was, both verbal and non-verbal, for these two discourse fragments.
That is, the reader will be forced to use these ' texts' to reconstruct, not just some relevant discourse subjects in the presupposition pool, following Venneman, but rather some of the elements of the topic framework existing when these discourse fragments were produced.
It is also likely that the reader, if asked to give the topic for each fragment, would not simply produce a single-term ' title '.
If the same reader were faced with the type of ' discourse ' fragment created by Venneman, reproduced as (8) below, he might quite readily provide support for Venneman's analysis by saying that ' the topic ' is Mary.
(8) Mary is singing strangely.
The reader presumably can just as easily reconstruct an alternative context (e.g. a description of the effects of marijuana on a Nativity play performance) in which Mary would not be proposed as' the topic of the discourse '.
Thus, while there may be preferences discernible in the choice of elements most-likely-to-be-concentrated-on within a sentence if that sentence is presented in isolation, such preferences may reflect the rather trivial fact that names are more salient than anything else, in isolation.
That these preferences do have significance for an analysis of the syntactic structure of sentences has been argued by Kuno &amp; Kaburaki (1977).
However, it is, in principle, impossible for a discourse to consist of a single decontextualised sentence and, in practice, rare for discourse participants to have to work out ' the topic of discourse ' one sentence at a time.
The most a discourse analyst could say about a discourse fragment such as the sentence in (8) above is that Mary is potentially part of the topic of the discourse in which (8) occurred, but more information is required, as indeed is also the case for both extracts (6) and (7).
It should be apparent that the use of single constructed sentences as the basis for making claims about notions such as' the topic of a discourse ' is extremely misleading.
Relevance and speaking topically
The topic framework, as we have described it, represents the area of overlap in the knowledge which has been activated and is shared by the participants at a particular point in a discourse.
Once the elements in the topic framework and the interrelationships between them have been identified, the analyst has some basis for making judgements of relevance with regard to conversational contributions.
The technical use of the term ' relevance ' in the analysis of conversation is derived from the conversational maxims proposed by Grice (1975).
If, as Grice suggests, there is a general agreement of co-operation between participants in conversation, then each participant can expect the other to conform to certain conventions in speaking.
These conventions or maxims have to do with the quantity (or informativeness), the quality (truthfulness), the manner (clearness) and relevance of conversational contributions.
Although he discusses and exemplifies the other maxims, Grice does not elaborate on the simple instruction ' Be relevant. '
The discourse analyst wishing to make use of this notion is immediately confronted with the problem of deciding ' relevant to what? '
One way of solving this problem is to translate the maxim ' Be relevant ' into a more practically useful form as' Make your contribution relevant in terms of the existing topic framework. '
What we have characterised as a convention of conversational discourse  ' making your contribution relevant in terms of the existing topic framework '  could be captured more succinctly in the expression speaking topically.
We could say that a discourse participant is' speaking topically ' when he makes his contribution fit closely to the most recent elements incorporated in the topic framework.
This is most noticeable in conversations where each participant ' picks up ' elements from the contribution of the preceding speaker and incorporates them in his contribution, as in the following fragment:
(9)
E: I went to Yosemite National Park
F: did you
E: yeah  it's beautiful there right throughout the year +
F: I have relations in California and that's their favourite Park because they + enjoy camping a lot
E: oh yeah
F: they go round camping +
E: I must admit I hate camping +
This type of ' speaking topically ' is an obvious feature of casual conversation in which each participant contributes equally and there is no fixed direction for the conversation to go.
In contrast, there is the type of conversational situation in which the participants are concentrating their talk on one particular entity, individual or issue.
In such a situation, the participants may, in fact, ' speak topically ', but they might also be said to be speaking on a topic.
An extreme example of ' speaking on a topic ' would be in a debate where one participant ignored the previous speaker's contribution on ' capital punishment ', for example, and presented his talk quite independently of any connection with what went before.
In practice, we should find that any conversational fragment will exhibit patterns of talk in which both ' speaking topically ' and ' speaking on a topic ' are present.
Both forms are based on the existing topic framework, but the distinction derives from what each individual speaker treats as the salient elements in the existing topic framework.
It is quite often the case that a speaker will treat what he was talking about in his last contribution as the most salient elements and what the other speaker talked about, though more recent, as less salient.
This facet of conversational discourse quite naturally leads to a consideration of the individual speaker's topics within what we have been discussing as the conversational topic.
Before we explore the influence of ' speaker's topic ', we shall try to illustrate in some detail the way in which conversational participants' speak topically ', by making their contributions relevant to the existing topic framework.
In the representation of the topic framework, we shall present the elements involved as a list.
It is difficult to imagine an appropriate ' diagram ' which could incorporate both the sequential pattern of elements introduced and the interrelatedness of those elements with each other and with the contextual features.
For the moment, we shall identify some of the elements and links which are pertinent to an analysis of one fragment.
(10)
Partial topic framework existing in a conversation between K (20+, female, Edinburgh-resident, university student,...) and J (60+, male, Edinburgh-resident, retired,...) in P Working Men's Club, Edinburgh,...) at T (early evening, spring, 1976,...) mentioning (J's three children  J ' s brothers  the schools they attended  the schools J attended  that J did badly at school  J left school at fourteen) when K asks J what he did after he left school J: oh I done odd jobs like + paper boy + chemist's shop worked in a chemist shop + and done two or three others+ and I finally started in the bricklaying + so I served my time as a bricklayer + K: that's good money J: nowadays it is but in that + when my time was out it wasn't+ it was only three pounds nine a week + so + + K: my father was a stonemason and he started at home + and they were paid a halfpenny an hour extra for being left-handed + +
Given a fragment of conversation and a topic framework as in (10), it is possible for the analyst to point out some ways in which each participant ' speaks topically '.
Such an undertaking can appear to be a matter of stating the obvious  that speaker J, in his first contribution, for example, is answering the ' what ' question in terms of an understood-to-be-known location and a time which is known from an interaction between knowledge of J's age (context) and knowledge that J was at least fourteen (domain).
We might highlight the ' topicality ' or ' relevance ' of J's first contribution by asking how K might have reacted if J had talked about one of his brothers, or about the type of work to be had in Australia, or training to be a brain surgeon.
Given this topic framework, J is constrained from talking about these things unless he introduces into the topic framework some additional information which he could then treat as shared by his hearer-that one of his brothers had gone to Australia to train as a brain surgeon and he considered doing the same, but settled for bricklaying instead.
Thus, J's first contribution here can be judged to be relevant in terms of the existing topic framework and also to add some information to the topic framework.
In this first contribution, he doesn't talk about ' being fourteen or older ' or ' Edinburgh ', but he does talk about ' starting work as a bricklayer ' (when I was fourteen or older, in Edinburgh) and, as a co-operative conversationalist, he would have to state explicitly if the information ' being fourteen or older, in Edinburgh ' was not applicable.
More interesting is speaker K's first contribution in (10).
First, its connection to the preceding discourse depends on a general inference that if one works (e.g. as a bricklayer) one receives money.
(We shall discuss the role of inference in discourse in Chapter 7.)
Second, this contribution has the potential to produce some conflict within the conversation, since ' what is being talked about ' up to this point is not present time.
The speaker appears to be generalising to a time which includes her own experience.
Within the existing topic framework, speaker K's saying that's good money is an example of speaking topically, for her, but, for speaker J, the time co-ordinate within the topic framework has been narrowed down by his preceding remarks.
There is, then, a discrepancy between what each participant is talking about, within the topic framework.
We shall examine this effect of individual speaker's topics in the next section.
Speaker J relates his subsequent remarks to the two salient time co-ordinates within the topic framework and adds some specific information on the ' money ' element introduced by speaker K.
Speaker K's next contribution exhibits a series of complex ties with the existing topic framework.
Speaker J, in his preceding contribution, has talked about the money received for his work at a particular point in the past.
Speaker K's contribution ' picks up ' the past time element, moving closer to speaker J's time while maintaining the personal reference in my father, who also did work (stonemason) comparable to J's (bricklayer) and received money for this work.
Putting her contribution even closer to J's preceding remarks, K makes her comments about her father relate to his' starting ' work and so comparable to J's started and when my time was out.
With these complex connections made, speaker K adds some new elements to the conversation (extra pay for being left-handed).
We have tried to list the connections existing across contributions in this discourse fragment to emphasise the ways speakers make what they're talking about fit into a framework which represents what we (as discourse participants) are talking about in conversational discourse.
For the discourse analyst, as an overhearer, those connections can signal the coherence relations which make each contribution relevant to the discourse as a whole.
Identifying the elements in the topic framework at any point in the discourse allows the analyst to make claims about what is involved in ' speaking topically '.
It also enables him to produce a version of ' what is being talked about ', i.e. the topic of conversation, which is much more comprehensive, and certainly of greater analytic interest, than the single word-or-phrase-type title which is often used in a fairly trivial way to characterise ' topic ' in the study of conversation.
Speaker's topic
So far we have considered the notion of ' topic ' in discourse in terms of what the participants share.
The ' topic framework ', as an analytic device, is essentially a means of characterising the area of overlap in contributions to a discourse.
By concentrating on the way conversational contributions overlap, however, we may neglect aspects of conversational discourse associated with different speakers having different personal ' topics'.
So far, we have been concentrating on describing the ' conversational topic ', but neglecting the notion of speaker's topic.
As we have already pointed out, the analyst typically treats conversational data as something complete, as a static product of some recorded interaction.
In doing so, he may lose sight of the fact that conversational discourse is dynamic, and that his data represents a process.
If we can treat any piece of conversational data as a process in which two or more participants speak within the topic framework, we should also find in their contributions elements which characterise their own personal ' speaker's topics'.
We shall look at a fragment of spoken discourse, not in terms of how we would characterise the participants' shared information, but in terms of a process in which each participant expresses a personal topic within the general topic framework of the conversation as a whole.
Prior to extract (11), the participants, L (female, 20+, unmarried, Edinburgh-resident, and M (female, 30+, married with young children, Edinburgh-resident), have been talking about recent improvements to old buildings in different areas in Edinburgh.
(11)
L: I quite like the way they've done the Mile though + I think it's quite 
M: yes[AhA] yes
L: the bottom of it anyway
M: it is  it is quite good they've certainly kept within the + em + + preserved it reasonably well or conserved it but we were up in Aberdeen this year for a holiday and we were staying right within the University complex there in Old Aberdeen+ and + oh some of the buildings there are beautiful really they really are nice + but er I was quite impressed with it  it's the first holiday we've had up there +
L: I was noticing  I was down by Queen Street or+the bottom of Hanover Street or somewhere + and they've just cleaned up some of the buildings down there + and what a difference it makes +
M: yes I know because there are some beautiful buildings
L: oh it was really nice
Extract (11) is representative of a common conversational situation in which each of the participants give examples from their personal experience to illustrate some general point.
The general point in this case is something like ' the effect of restoring old buildings' which is already part of the topic framework established by the preceding discourse.
Notice that speaker M's second contribution in this extract is not just 'about ' that general point.
She is also talking about her recent holiday in Aberdeen, for example.
We could describe this' holiday in Aberdeen ' element as, at this point, a part of speaker M's personal topic which could become, in the developing conversation, a shared topic area for both speakers.
Speaker L could have followed on, with a question, for example, about the holiday, Aberdeen, or even with some personal observations on the buildings in Old Aberdeen or the University.
Speaker L, however, does not ' pick up ' any elements from speaker M's personal topic, but continues on her own personal topic area (i.e. Edinburgh's old buildings after restoration).
When participant M speaks again near the end, she does not return to her ' holiday ' or ' Old Aberdeen ', but makes her contribution relate closely to L's immediately preceding remarks.
There are two points worth noting about this fragment of conversational discourse.
First, it is a feature of a lot of conversation that ' topics' are not fixed beforehand, but are negotiated in the process of conversing.
Throughout a conversation, the next ' topic ' of conversation is developing.
Each speaker contributes to the conversation in terms of both the existing topic framework and his or her personal topic.
It is clear from extract (11) that some elements in a speaker's personal topic do not become salient elements in the conversation if neither the other participant nor the speaker herself mention them again.
To use the ' negotiation ' metaphor, we can say that speaker M offers elements in her personal topic (in her second contribution) as possible elements to be included in the conversational business, but speaker L does not take up the offer.
A second point to be noted in this, and in a large number of other conversational fragments, is that personal topics are frequently introduced through first person reference in one form or another.
Although the points made in extract (11) could have been expressed objectively as statements that certain buildings in certain locations are more beautiful since restoration, both speakers relate such statements to personal experience.
It is as if speakers feel obliged to offer some personal warrant for the statements they will make about the world.
A statement that the buildings in Old Aberdeen are beautiful is embedded within an assertion that the speaker was recently in Old Aberdeen, and stayed there for a period, and so she has a warrant for making the statement.
If we reconsider the earlier extract (5) as one participant wanting to know the meaning of an expression and the other offering a possible explanation, we can see that the explanation is offered in personal terms (when we were young and we called it ' the taw ') based on the speaker's personal experience.
It may be that this explanation is not an acceptable answer to the question, but it is presented by the speaker in a form which conveys' what I think we're talking about ' in this part of the conversation.
Characterising the individual speaker's topic as' what I think we're talking about ' incorporates both that element which the conversational analyst tends to abstract as the ' topic of conversation ' for the participants ('What we're talking about') and the individual speaker's version ('I think'), as he/she makes a conversational contribution.
That speakers do introduce what they want to say via some form of personal reference has a noticeable effect on the structure of contributions in conversational discourse.
We shall return to this point in the discussion of further details of discourse structure in Chapter 4.
From what we have proposed as speakers' topics in conversational discourse, it must occasionally happen that there are at least two versions of ' What I think we're talking about ' which are potentially incompatible.
It is a noticeable feature of co-operative conversational discourse, however, that this potential incompatibility rarely leads to conflict over the topic of conversation.
What typically happens is that, in the negotiation process, one speaker realises that his version is incompatible with what the other appears to be talking about and makes his contributions compatible with ' what I think you (not we) are talking about '.
We can illustrate this process in two conversational fragments and note two different strategies used to avoid conflict in the ' negotiations'.
In the first extract, (12), one piece of continuous conversational discourse has been divided up into chunks.
Immediately before this extract, speaker B (female, 50+, aunt of speaker A) has been describing to speaker A (female, 20+) the first type of radio she had, forty years before.
(12)
A: but you'd have telephones around +
B: mm oh yes oh aye oh aye I've had the telephone since nineteen thirty eight +
A: hmm
B: oh they were on a long while I think before that +
Speaker B had been talking about the radio she had in the 1930s and speaker A's first line here seems to continue within the temporal, locational and personal indices of the existing topic framework while introducing telephones.
Speaker B treats this contribution as requiring an answer, following a pattern described by Labov in the rule: ' If (speaker) A makes a statement about a (speaker) B-event, it is heard as a request for confirmation ' (1972b: 254).
Speaker B expands on her answer, in personal terms, regarding the telephone.
Speaker A offers no contribution and speaker B adds some additional information about telephones.
We might characterise speaker B's view of ' what I think we're talking about now ' as something involving herself, the 1930s, and the existence of telephones (as well as radios) at that time.
The conversation continues:
A: 'cause there was a man in  my father's in the Scouts +
B: oh yes he was  is he still
A: he's a county commissioner now
B: oh is he +ah ha+
Speaker A appears to be offering some new elements as part of the conversational topic, again deriving from some personal reference (as in my father) which speaker B appears to accept.
That is, speaker B does not insist on mentioning telephones, but moves on to this new area.
Speaker B's view of ' what I think we're talking about now ' must now involve speaker A, A's father, the Scouts and a man (who may have something to do with telephones).
We might expect speaker B to be a little confused about how these elements relate to the preceding conversation.
Speaker A continues, as follows:
A: and eh one of his oldest+ scoutmasters wa-ha-was reaching his hundredth birthday +
B: is that so +
We suspect that, by this point, although speaker B can identify ' what's being talked about ', she can play no part in negotiating the topic, because she may not be able to see why this individual entity is being talked about.
The contributions of speaker B cease to be attempts to add anything to the conversational topic.
Speaker B's view of the conversation has consequently become one in which she is no longer expressing a personal topic, but is waiting to discover ' what I think you (not we) are talking about '.
Throughout the rest of this fragment, speaker B simply makes' interested ' noises as speaker A gradually gets to the point.
A: so father was making up a big + sort of remembrance book 
B: aha
A: to give him and he was writing just at the beginning he was  writing the whole  for each year of his life he wrote something in that had  had been invented or +
B: oh yes
A: ah a book that had been written or a piece of music that had been written or a painting or a 
B: very interesting yes
A: or whatever you know and + within his lifetime the telephone had been invented +
B: had it + really + fancy +
In this extract as a whole, we can trace speaker B's attempt to contribute to what she thinks they're about, by first offering some remarks on telephones and then on the father, but gradually reducing her comments to the type of contentless noises described by Duncan (1973) as back channels.
Back channel behaviour, which can also include nods and sentence completions is used when a participant wants to indicate to the person speaking that he should continue.
Speaker B stops trying to take turns in the negotiation of topic and waits for speaker A to make it clear how what she is saying has some connection to the existing topic framework.
Eventually, as we can see in A's final remarks, a connection is made.
There is evidence in speaker A's contributions that what she is trying to say is not very well organised before she starts to speak.
There are false starts, hesitations and repetitions.
Everyday conversational discourse is, not infrequently, characterised by this lack of preplanning.
The resulting structure of speaker A's contributions is, in fact, quite common in discourse and will be discussed in some detail later in terms of ' staging ' (see Chapter 4).
Speaker B's strategy, then, in a situation where she finds that she is unsure about what she thinks they're talking about, is to stop talking.
In the following extract (13), there is another example of a mismatch between speakers' topics, brought about by a misunderstanding of the intended meaning of a particular word.
In the immediately preceding conversation, speaker C (female, 20+, American, visiting Edinburgh) has been finding out from speaker D (male, 40+, Edinburgh-resident) where there are good places to go for bicycle rides in and around Edinburgh.
(13)
C: what about going down by the  the Firth of Forth
D: that should be fun shouldn't it yes you could
C: is it
D: yes you can cycle all  you can ride right along the edge you know + without falling in you can ride right along the edge eh without em + going  keeping on the main road + that should be great actually + you could do that +
C: is it very rough down there though
D: well there are no cobbles as far as I remember  have you tried riding on the cobbles
C: yes yes
D: you must have done
C: I went down to Muirhouse
D: which is almost all cobbles isn't it
C: it was rather rough
D: hmm
C: no but I was  I was thinking rather more rough in terms of the em + people +
D: oh I see + you well I don't think so + I don't know + I  I  eh  parts of it are quite poor + particularly the Pilton area +
Looking back to speaker C's third question, we can propose two versions of ' what I think we're talking about '.
For speaker C, it involves' are the people rough? ' and, for speaker D, ' are the roads rough? '
Unlike the hearer (B) in extract (12), however, speaker C appears to be able to recognise speaker D's alternative topic and accepts what she thinks speaker D is talking about as' what we're talking about ', for a few turns.
When speaker D stops talking about cobbles (i.e. rough for cycling on), speaker C can attempt to return to her topic (rough in terms of the em + people).
Speaker D's response at the end of this fragment is, in effect, an answer to the question which speaker C originally intended him to answer.
We might think that by the end of this fragment there is once again a single version for both speakers of ' what I think we're talking about '.
Indeed, most conversational analysis is undertaken with this single ' topic ' concept as a working assumption.
Yet, in extract (13) we can only reconstruct the intended meaning of C's third question because she actually explains her intended meaning later.
If speaker D had gone on at some length about ' cobbles' or rough roads in general, or if the analysis only had part of this fragment, up to C's it was rather rough, then we might have had no evidence of a divergence in speakers' topics within the conversation.
Our argument for the importance of considering individual speaker's topics in conversational discourse would consequently be weaker.
We do not suggest that discourse analysts should spend their time looking for potential alternative meanings in what speakers say in a conversation, but we do suggest that the analyst should not simply assume that there is a single, static ' topic of conversation ' in any conversational fragment.
If there is an entity identifiable as' the topic of conversation ', the analyst should consider what evidence from each individual speaker's contributions he is using to make that identification.
He should also remain aware of the fact that conversation is a process and that each contribution should be treated as part of the negotiation of ' what is being talked about '.
Above all, he should remember that it is speakers, and not conversations or discourses, that have ' topics'.
Topic boundary markers
In our discussion of ' topic ', we have concentrated mainly on considerations of ' content ' and neglected the influence of ' form '.
Yet our interpretation of what a speaker is talking about is inevitably based on how he structures what he is saying.
We shall now investigate some formal aspects of topic-structure in discourse.
In this section we shall look at the formal devices used to mark the boundaries of chunks of both written and spoken discourse which form large units of some kind, such as paragraphs.
Aspects of the internal structuring of these chunks will be discussed in Chapter 4.
It has been suggested (e.g. by Schank, 1977: 424; Maynard, 1980) that instead of undertaking the difficult task of attempting to define ' what a topic is', we should concentrate on describing what we recognise as topic-shift.
That is, between two contiguous pieces of discourse which are intuitively considered to have two different ' topics', there should be a point at which the shift from one topic to the next is marked.
If we can characterise this marking of topic-shift, then we shall have found a structural basis for dividing up stretches of discourse into a series of smaller units, each on a separate topic.
This type of approach to the analysis of discourse is based on the principle that, if we can identify the boundaries of units  where one unit ends and another begins then we need not have a priori specifications for the content of such units.
The burden of analysis is consequently transferred to identifying the formal markers of topic-shift in discourse.
Paragraphs
It might seem that identifying the formal demarcation of chunks of written or printed discourse is a relatively simple task.
After all, written discourse is divided into paragraphs whose boundaries are marked by indentations.
Topic-shifts in written discourse then could be identified with the beginning of each new paragraph.
Unfortunately, it doesn't seem to be as simple as that.
Those who use the term ' paragraph ' to describe a unit in the structural analysis of written discourse go to some trouble to point out that they are not describing the orthographic paragraph.
According to Longacre (1979: 116), the orthographic paragraph can result from a writer's stylistic concerns, ' partially dictated by eye appeal ', or from printing conventions such as an indentation for each change of speaker.
Hinds (1977: 83) also notes that the journalistic paragraph is often determined on the basis of appearance.
He has a worked example in which a single structural paragraph derives from a newspaper article containing five orthographic paragraphs.
Thus, it may be that the beginning of an orthographic paragraph indicates a point of topic-shift, but it need not do so.
Both Longacre (1979) and Hinds (1977) appeal to languages other than English for evidence that there are formal linguistic markers of the beginning and end of paragraphs.
What is immediately noticeable in the discussion of these markers is that they are genre-specific.
There are ways of indicating the beginning of a new paragraph in a piece of narrative, for example, which are not used in explanatory discourse.
This general point is also made by
Grimes (1975: 109), who describes the marking of paragraph boundaries as one form of ' partitioning ' in discourse.
The principles on which partitioning depends are related to change of ' setting ' (time or place) and ' theme ' (the person or thing talked about), in narrative discourse, at least.
Interesting though it may be to learn that there is a narrative-discourse-paragraph-introductory-particle in Huichol or Shipibo, it becomes decidedly less interesting when one discovers that the identification of the significance of these particles depends on a prior identification of the paragraph as a unit in which ' the speaker continues talking about the same thing ' (Grimes, 1975: 103).
Hinds (1977) bases his paragraph divisions on a similar principle, quoting Grimes as support, and emphasising the significance of ' participant orientation '  that is, the unity of a paragraph derives from its being mainly about a single participant.
Longacre (1979) claims that ' in narrative discourse, a narrative paragraph is built around a thematic participant, occasionally a small set of thematic participants' (Longacre, 1979: 118).
In other words, only the paragraph structure of stretches of discourse about individual, primarily human, characters is being discussed.
In effect, this limits the discussion to narrative discourse, or, as in Hinds (1977), a description or an obituary of a particular individual.
It should be obvious why a single structural or ' semantic ' paragraph in Hinds' (1977) analysis can extend over five orthographic paragraphs in a newspaper.
Each of these orthographic paragraphs is' about ' the same individual.
Yet, some obituaries extend to twenty or more orthographic paragraphs' about ' the one person, and whole chapters of novels, containing over a hundred lengthy orthographic paragraphs, may be ' about ' the same individual.
Surely such extended stretches of written discourse are not single ' paragraphs'?
We shall consider a stretch of written discourse, not from a source such as a Paez (Colombia) folk tale or a specially constructed text, but from a recent English novel.
In the extract reproduced below (14), the orthographic paragraph boundaries as they appeared on the printed page have been ignored.
The whole extract has two principal participants, but is quite clearly ' about ' only one of them.
If there are points of ' topic-shift ' in English written discourse which lead writers, or their editors, to begin new orthographic paragraphs, then we should be able to identify likely points where the writer or the editor marked the division of this' text ' into separate chunks.
(14)
1After the first few days, when I come into the room, Birdie is down on the floor of the cage, running back and forth, looking out over the barrier that holds in the gravel.
2 I think she's glad to see me, not just because I give her treat food, but because she's lonely.
3 I 'm her one friend now, the only living being she gets to see.
4 By the end of the week, I rubberband the treat food dish onto the end of an extra perch and put it into the cage through the door.
5 I lock the door open with a paper clip.
6 At first, Birdie's shy, but then she jumps onto the perch I 'm holding and sidehops over to the treat dish.
7 It's terrific to see her without the bars between us. 8 She sits eating the treat food at the opening to the door and looking at me. 9 How does she know to look into my eyes and not at the huge finger next to her.
10 After she's finished eating, she retreats to the middle of the perch.
11 I lift it gently to give her a ride and a feeling the perch is part of me and not the cage.
12 She shifts her body and flips her wings to keep balance, then looks at me and makes a new sound, like peeEP; very sharp.
13 She jumps off the perch to the bottom of the cage.
14 take out the perch and try to talk to her but she ignores me. 15 She drinks water.
16 She doesn't look at me again till she's wiped off her beak and stretched both wings, one at a time.
17 She uses her feet to help stretch the wings.
18 Then, she gives a small qUeeEEP?. 19 Generally, Birdie looks at me more with her right eye than her left.
20 It doesn't matter which side of the cage I stand.
21 She turns so she can see me with her right eye.
22 Also, when she reaches with her foot to hold the treat dish, or even her regular food dish, she does it with her right foot.
23 She'd be right-handed if she had hands; she's right-footed or right-sided. 24 She approaches and does most things from the right side.
(William Wharton, Birdy, Jonathan Cape, 1979, p. 47)
If there are orthographic paragraph divisions in the original version of this text which were made for the sake of appearance on the page, then we have little hope of identifying such divisions in any formal way.
What kind of formal marks, if any, would we expect to find at the beginning of a new paragraph?
The markers Longacre (1979) identifies in narrative discourse are inevitably adverbial expressions indicating temporal sequence.
It may be that the general class of adverbials which can appear initially in a sentence could be taken as possible markers of ' topic-shift '.
Quirk et al.
(1972: ch. 8) provide lists of such adverbials in terms of adjuncts, conjuncts and disjuncts.
In fact, extract (14) begins with an adverbial clause in initial position.
There are two other points in this extract, sentences 4 and 10, where adverbial clauses occur in sentence-initial position.
There are four other points where adverbial expressions occur sentence-initially, sentences 6 (At first), 18 (Then), I9 (Generally), and 22 (Also).
This would give us six possible breaks, formally marked, in the structure of the piece of text.
The next question is  do all these adverbial expressions function in the same way?
After all, we would like to distinguish between adverbials which indicate a connection between one sentence and the next and those adverbials used to link a set of sentences to another set.
The use of then in 18 seems to introduce a final action in a temporal sequence of actions.
We can conceive of this one sentence being separated from the previous set as a form of distinct climax.
We might expect, however, that it would more typically occur as the final sentence of a paragraph, not as a climax, but as describing an action which culminates a series of actions.
It is followed by a sentence which does not continue the series of actions and which begins with what Quirk et al.
(1972: 509) would characterise as a ' style disjunct '.
This use of generally, in I9, effectively separates the previous set of sentences from the next set describing a particular habit of the individual involved.
Within this latter set, one sentence begins with the additive adjunct, also, in 22, which could be indicating that there are two parts to this set.
It is more likely that the sentence beginning with also is adding more detail to support the general conclusion that the individual concerned is right-sided and is part of the internal structure of a paragraph beginning with Generally.
The other adverbial, at first, in 6, seems to be part of a sentence-internal construction, especially when we see the then which follows.
The events described in this sentence fall within the set of events described as happening by the end of the week (in 4).
Thus, we have reduced the number of possible breaks in this text to three, so that we can suggest that there are four paragraphs, beginning at sentences 1, 4, 10 and 19.
The reader may suggest other possible breaks, as, for example, in 9, where there is a sentence structure (an interrogative) quite different from the structure of the rest of the text sentences.
An argument for a break here would seem quite reasonable since this sentence is structurally marked as separate.
No doubt the reader could also think of an argument, mainly in stylistic terms, for treating this sentence as part of the preceding set.
It may be the case that, taking stylistic considerations more generally, the reader would wish to divide this text into separate paragraphs at points where there are no formal markers at all.
We would assume that the discussion, in such a case, would cease to be a discussion which appealed to primarily linguistic evidence in this piece of discourse.
On the basis of some formal linguistic markers, we have suggested that there are four paragraphs in extract (14).
We may have been led to finding those four paragraphs because they are, in fact, the divisions which actually appear in the original and we merely sought additional evidence to support the way the author had divided up his discourse.
Yet this point highlights the fact that the exercise we have performed on extract (14) was an extremely artificial treatment of written discourse.
We began by removing one of the primary indicators of ' topic-shift ' available to a writer, that of indenting a line in his text.
Rather than treat the indenting of the first line of a paragraph as simply some cosmetic device, as Longacre (1979) does, we might look upon it as an indication by a writer of what he intends us to treat as the beginning of a new part of his text.
If the writer also uses adverbial expressions initially in the first sentence of this new part of his text, then we might say we have overwhelming evidence that the writer is marking a ' topic-shift ' in his discourse.
We are, after all, performing a descriptive and not a prescriptive exercise when we undertake discourse analysis.
We do not wish to say how a writer should organise his written discourse into paragraphs before we have managed to characterise, in any comprehensive way, how writers typically do so.
The investigation of what writers typically do when marking the structure of their texts would seem to be a more appropriate goal of discourse analysis.
For example, rather than dismiss the orthographic paragraph format to be found in newspaper articles as, in some way, a deviation from the ' true ' paragraph structure of what is being written, it would be more appropriate for discourse analysts to describe the journalistic format as one form of written discourse organisation.
The paragraph structure of different genres, such as scientific textbook writing, repair manuals, nineteenth-century novels, etc. could then be characterised, and statements could be made about, for example, the ' norms' or regular features of topic-shift in such genres.
On the basis of such genre-specific descriptions of ' topic-shift ' markers, it should be possible to make linguistic, as opposed to literary, statements about the structure of English written discourse which reflect the writer's purpose.
Thus, in producing a narrative, the writer must provide some indications of change of time and place, as Grimes (1975: 102) has pointed out.
In presenting a philosophical argument, however, the writer can range over different times and places within a single paragraph, but must mark out changes in the direction of his argument.
Taking a random page from the writings of Karl Popper, one can see the structure of the discourse in skeleton form by taking the first phrase or sentence of each paragraph.
(15)
para 1: Other questions have sometimes been asked...
para 2: Another question sometimes asked is this...
para 3: The only correct answer is the straightforward one...
para 4: It has also been said that the problem of induction is...
(Popper, 1963: 56)
Eventually, it should also be possible to specify those markers of ' topic-shift ' which occur in all forms of written discourse.
We might find that it is indeed the case that the use of ' But ' at the beginning of a paragraph as described by van Dijk (1977: 139), is a very general marker of topic change.
Other examples of what van Dijk (1977: 150) terms macro-structure connectives are ' furthermore ', ' however ', and ' so '.
We shall discuss the concept of macro-structures in discourse in section 3.
7 on the proposition-based analysis of discourse.
Paratones
So far we have concentrated on structural markers in written discourse.
In spoken discourse, there is not the visual prompt of paragraph-initial line indentation to indicate a division in the discourse structure.
How do speakers mark ' topic-shifts'?
One suggestion is that there are, in fact, structural units of spoken discourse which take the form of ' speech paragraphs' and have been called paratones (see Brown, 1977: 86).
Some support for the notion that there are ways of marking the boundaries of ' speech paragraphs' can be found in a common practice of people who are asked to read pieces of written text aloud.
They use intonational cues to signal the start of a new paragraph.
The ' speech paragraph ', or paratone, like the orthographic paragraph, is identified by its boundary markers.
The marking of the start of a paratone, then, is clearly one device which speakers can use to indicate a topic-shift.
Since the paratone is a much less familiar concept than the orthographic paragraph, it may be useful to have its identifying features described.
At the beginning of a paratone, the speaker typically uses an introductory expression to announce what he specifically intends to talk about.
This introductory expression is made phonologically prominent and the whole of the first clause or sentence in a paratone may be uttered with raised pitch.
The end of a paratone is marked in a way similar to the ' turn signal ' discussed by those who investigate conversational discourse as a process of social interaction (cf.
Duncan, 1974; Sacks et al., 1974).
It can be marked by very low pitch, even on lexical items, loss of amplitude and a lengthy pause.
Alternatively, the speaker can use a summarising phrase, often repeating the introductory expression, not necessarily low in pitch, but also followed by a lengthy pause.
The most consistent paratone-final marker is the long pause, normally exceeding one second.
We shall examine an extract from conversational discourse containing a longish paratone which illustrates the features just described.
It is relevant that the topic framework for this extract (16) should contain information about the speaker (female, 20+, Edinburgh-resident) and the preceding discourse (the types of drinks the participants had encountered in different types of bars during their respective recent holidays in the United States).
It is also worth noting that in Edinburgh Scottish English, phonologically prominent syllables are typically uttered with raised or high pitch and need not have the type of pitch movement associated with phonological prominence in descriptions of standard southern English (cf.
Brown et al., 1980).
(For an explanation of the stave representation of intonation used, see the ' transcription conventions' on p. xii.)
(16)
I found my drink was a great problem with them because () at that time I drank whisky and lemonade + and I would () go and ask for whisky and lemonade and I would get () whisky and lemon + because you have to ask for whisky () or scotch and seven up + you know + I eventually () cottoned on to it + but + and they couldn't get over () the fact that I didn't like ice in whisky and of course () they either gave me ice whether I wanted it or not or () they stacked the glass up + right up to the level that () you would normally have if you had ice in your drink () anyway + and consequently I got ploughed + frequently + () and that's that I + I tended to stick to my drink ++ ()
This paratone begins with an introductory expression my drink, uttered very high in the pitch range, and closes with the same expression, low in the pitch range, as part of the speaker's summing-up.
The internal pauses are brief, none exceeding 0.5 seconds, but the final pause marking the end of the paratone is long (1.6 seconds).
Those are the formal markers of the boundaries of this paratone.
Of course, there are internal aspects, such as the semantic cohesion within the lexical field established by my drink, which could also be appealed to in claiming that this chunk of discourse is a unit of some kind.
However, this type of internal cohesion is not a necessary feature of the structural unit we have described as the paratone.
It might be argued that there are two paratones, and not one, in this extract.
There appears to be a break where+ but + is used.
Indeed, just prior to but, there is what has been described as a ' possible completion point '.
The speaker has come to the end of a sentence and pauses.
It is a point at which those who analyse conversation in terms of ' turn-taking ' (Sacks et al., 1974) would suggest that another speaker could take over the turn.
However, the speaker in this extract immediately produces an ' utterance incompletor '  in this case but, though any clause connector would do making, as Coulthard (1977: 56) points out, a potentially complete utterance into an incomplete one.
After another brief pause, the speaker continues, using and to indicate that what she is going to say is connected to what she has just said.
We would not want to describe this possible completion point (or any other which occurs in this extract) as a paratone-boundary.
The formal markers, low pitch close plus lengthy pause plus raised pitch introductory expression, are not present.
In intuitive terms, we might also say that what follows + but + is not on a separate speaker's topic, but continues the talk ' about ' 'my drink'.
At the end of this extract, there is an obvious' completion point '.
In ' turn-taking ' terms, it is a point at which another speaker is free to take over.
However, in this part of the conversation, one speaker clearly ' has the floor ' and she is allowed to continue, as shown in extract (I 7).
(17) oh apart from once when we went we found em + an Irish bar () in San Francisco that was famous for its Irish coffees + ()
In beginning a new paratone, the speaker marks as intonationally prominent two expressions  an Irish bar and Irish coffees.
In the course of the paratone, she talks about both the bar and the Irish coffee made there.
It seems quite reasonable to assume that, when a speaker is organising a ' speech paragraph ' which has two connected elements as its foci, both elements can be made phonologically prominent in the introduction.
When the speaker closes this paratone, she repeats one of her introductory expressions  it was very good Irish coffee too  not particularly low in the pitch range, but followed by a lengthy pause.
Some of the features we have described as marking paratone boundaries in spoken discourse can, of course, have other functions.
Although the lengthy pause is also identified by Chafe (1979: 176) as an indication of segmentation in his spoken discourse data comparable to paragraphing in written discourse, the intonational features we appealed to can have other, quite different, functions.
Some of these we will discuss in detail in Chapter 5.
What we have described is the use of the combination of these formal markers by speakers to indicate a shift in what they're talking about.
There may be other, more subtle, indicators of topic-shift used by conversationalists which we have ignored.
The significance of ' speaker gaze ', as described by Kendon (1967) and specific ' body movements' (de Long, 1974) in signalling speaker change in conversation may also be relevant in topic change.
The occurrence of different types of ' fillers' such as' well ', ' mmm ', ' you know ', ' er ', and others may also regularly coincide with topic-shifts.
We have concentrated, however, on some of the primary, easily identifiable formal markers used by writers and speakers to indicate structural divisions in the discourse they produce.
We emphasise once again that, although we can regularly identify such structural markers, their appearance in discourse should not be treated in any way as' rule-governed '.
They represent optional cues which writers and speakers may use in organising what they want to communicate.
Failure to mark out explicitly the structural organisation of what a speaker wishes to communicate may make the addressee's task of interpretation more difficult, perhaps, but, by itself, would not necessarily constitute a failure to communicate.
