Discourse analysis ought ideally to be an interdisciplinary under-taking. This follows from the conception of discourse I have been
The Corpus
The perspective of specialist disciplines, and those being re-searched, is also important in the selection of data, the construc-tion of a corpus of discourse samples, and in deciding what supplementary data to collect and use. The nature of the data will ,vary to the project and the research ques-nons, but there are certam general principles to bear in mind.
One can only make a sensible decision about the content and structure of a corpus in the light of adequate information on the
Doing Discourse Analysis 227
Enhancing the Corpus
There are various ways in which a corpus can be enhanced with supplementary data. One can, for instance, judgements about aspects of discourse samples in corpus from .of people who are in some significant relation to the practlce in focus. If classroom practices and classroom discourse were being studied, the panels
members of different majority and mmonty commumnes, and educational administrators (on the use of panels, see Gumperz
1982). A widely used way of enhancing a corpus i.s.
interviews. One can interview those involved as participants ID corpus samples, not only to elicit their interpretations of those samples, but also as an opportunity for the researcher to into issues which go beyond the sample as such, to try to diSCOV-er, for example, whether a is conscious the investment of a particular discursive convennon ID 'archive'. (This term is used in a way which extends it beyond its historical usage, to refer to the totality of discursive practice, either recorded past practice or ongoing practice, that falls within the of the research project.) This is partly a practical matter of knowing what is available, and how to get access to it, but it is partly a matter .of having a .model of order of discourse of the institution or domain one tS researching, and the processes of change it is undergoing, as a preliminary to deciding upon where to collect samples a.corpus. Of course, work the corpus may change the prehmmary mental map. The dis-course analyst should depend upon people.in
and people working within the research .slte, for
which samples are typical or representative of a certam pracnce;
whether the corpus adequately the of. practice and changes of practice across different types of Situation, and both normative and innovative practice; and whether the corpus includes cruces and moments of crisis (these notions are explained below). There are particular problems in collecting a corpus of material that gives access to processes of one obviously needs to try to incorporate reasonable nrne spansID the data.
Doing Discourse Analysis 226
advocating, which involves an interest in properties of texts, the production, distribution, and consumption of texts,
tive processes of producing and interpreting texts, social practice ,in various institutions, the relationship of social practice to
relations, and hegemonic projects at the societal level.
facets of discourse impinge upon the concerns of various sciences and humanities, including linguistics, psychology social psychology, sociology, history, and political science.
What is specific about a particular discursive practice delPettds upon the social practice of which it is a facet. Research projects discourse analysis are, therefore, most sensibly defined first terms of questions about particular forms of social practice, their relations to social structure; given the focus I have adopting, in terms of particular aspects of social or cultural change. This means that it is the disciplines which deal with these questions - sociology, political science, history - that we should look to in the first instance in defining research projects. Dis-course analysis should best be regarded as a method for conduct-ing research into questions which are defined outside it. I shall give an example shortly. This is however, an inordinately 'top-down' way of seeing it: in many cases, interdisciplinary teams of researchers may be able to work with people in, for example, education or health, in investigations of issues and problems which they experience during the course of change. Indeed, it may be possible to engage in 'eo-research', the individuals or groups who are the subjects of research being involved in the design, implementation, writing up, and use of the research (see Ivanic and Simpson forthcoming).
An Example
Doing Discourse Analysis 229
period, as of a wider study of the development of power relations in quality circles. This corpus might be enhanced with recordings of the training of managers as quality circle leaders or 'facilitators', presentations of proposals by quality circles to man-agement panels, and communication between quality circles and members the workforce who are not involved in them. It could also be enhanced through interviews about quality circles with quality circle members, senior management, trade union
sentatives, and other shop-floor workers. The analyst might involve members of quality circles in arriving at definitions of research questions analytical focus. I shall develop this example a little in discussing 'Results'.
Transcription
Spoken discourse, such as quality circle meetings, needs to be transcribed. Transcription is a difficult and time-consuming pro-cess. Depending upon the system of transcription used,. it can take anything from six to twenty hours or m.ore to
hour of recorded speech. There are a vanety of transcnpnon
systems available which features of
with different degrees of detail - mtonanon, stress, pausing, changes in loudness and tempo, and so forth (Atkinson and Heritage 1984: ix-xvi; Tannen 202-4). No system .could conceivably show everything, and It IS always a matter of JUdge-ment, given the nature of the project and the research questions, what son of features to show, and in how much detail. A fairly minimal type of transcription, which is adequate for many pur-poses, shows overlaps between speakers, pauses, and silences (the transcriptions on pp. 139'and 145 are examples). .
What is, perhaps, less obvious is that transcription necessanly imposes an interpretation on speech; in the words of one.paper on this issue, transcription is theory (Ochs 1979). Consider a situation where three people are talking, with one of them accounting for 80 per cent of the talk. The way this is set out on paper might either represent it as a 'conversation', in which all three take turns at but one happens to take longer turns and more turns than the others; or it might be represented as a monologue with various interruptions or supportive additions Doing Discourse Analysis
228
Anexample of a possible research project will give these instances. I shall use the example referred to in the Introduction (p. 7 above) of changes in discursive practice associated with the shift from Fordist to post-Fordist produc-tion, focusing upon a projected study of 'quality circles' (this example is based upon planning for a joint sociological and lin-guistic project at Lancaster University). A characteristic of post-Fordist industry is the growing importance. of communication on the shop-floor. New forms of interaction between shop-floor workers, supervisers, and managers are emerging, such as 'quality circles', groups of five to ten employees who usually work together and meet regularly to discuss ways of improving quality, productivity, and other work-related issues. One unanswered question about quality circles, which may well bear upon trade union suspicions of them and their high failure rate, is whether they really break down old divisions between employees, and give shop-floor workers more power, or whether they are a management tool for using the valuable experience of shop-floor workers, .and integrating them into management priorities. Re-search is needed into how quality circles actually function - how they select topics, how they debate issues, how they generate proposals and present them to management - and whether con-trol of these activities is shared and negotiated, or exercized more or less ovenly by management. These questions can be investi-gated using discourse analysis as a method. The corpus for such a study might consist of video recordings of quality circles over a one-year period from their inception. One focus might be on how conventions for conducting meetings emerge over the some situationsthan in others. Or, in eo-research, one
closer and less formal access to the perspective of those researched. The point to emphasize is that interviews, panels, so forth, are further discourse samples, and that one way in they enhance the corpus is simply by being added to it.
corpus should be seen not as constituted once and for all bet,ore one starts the analysis, but as open to ongoing enhancement response to questions which arise in analysis.
Coding and Selecting Samples within the Corpus
from the other speakers, perhaps by setting out the more loqua-cious speaker's talk in a column down the middle of the page with the other contributions set out at the margins (see Edelsky 1981 for examples of this sort). Similarly,ifthere is silence on the tape, the transcriber has to decide whether to attribute it to one or other of the participants; if there is overlap, whether to repre-sent it as one speaker interrupting another.
Researchers may well wish to code a whole corpus or large parts of it in broad terms, perhaps summarizing the discourse, or code it in terms of topics, Or they may scan the whole corpus for particular sorts of feature - certain types of question, or formula-tions (see p. 157 above). However, the conception of discourse I have presented, and the view of analysis I shall summarize below, are especially relevant to detailed analysis of a small number of discourse samples. This raises the problem of how to select sam-ples for detailed analysis. The answer is broadly that samsam-ples should be carefully selected on the basis of a preliminary survey of corpus, taking advice where one can get it from those being researched, or from colleagues in relevant social science disci-plines, so that they yield as much insight as possible into the contribution of discourse to the social practice under scrutiny.
One selection strategy which has much to recommend is to focus on what I earlier called 'cruces' and 'moments ·of crisis'.
These are moments in the discourse where there is evidence that things are going wrong: a misunderstanding which requires parti-cipants to 'repair' a communicative problem, for example through asking for or offering repetitions, or through one participant correcting another; exceptional disfluencies (hesitations, repeti-tions) in the production of a text; silences; sudden shifts of style.
In addition to the evidence of the text and of the participants' conduct of the interaction, one might again use panel judgements or participants' retrospective judgements about points of diffi-culty. Such moments of crisis make visible aspects of practices which might normally be naturalized, and therefore difficult to notice; but they also show change in process, the actual ways in which people deal with the problematization of practices.
231
Analysis
Doing Discourse Analysis
This section consists of a summary of the sorts of analysis I have introduced and illustrated in chapters 3-7. I have not adhered exactly to the order in which topics were discussed in those chapters, but I do follow the same overall progression from (i) analysis of (at a 'macro' level: see p. 85 above), focusmg upon the intertextualiry and interdiscursivity of discourse samples; to (ii) analysis of texts (plus 'micro' aspects of discourse practice); to (iii) analysis of the social practice of which the discourse is a part. These three dimensions of analysiswill overlap in analysts, for example, always begin WIth some sense of the SOCIal practice that the discourse is embed-ded within. But this progression is a useful one for ordering the outcome of one's engagement with a particular discourse sample before presenting it in written or spoken form. Notice that it involves a progression from interpretation to description and back to interpretation: from interpretation of the discourse prac-(processes of and consumption), to descrip-non of the text, to mterpretadescrip-non of both of these in the light of the social practice in which the discourse is embedded. It is not necessary to proceed in this order, and analysts can begin from text analysis, or indeed analysis of social practice. The choicewill depend upon the purposes and emphases of the analysis. 'Front-ing' the analysis of discourse processes seems to be particularly appropriate, given my main concern here with process and change.
Each of the main heads in the summary below is followed by a short description of the sort of analysis it involves, and then in most cases a set of questions to act as pointers during analysis of particular discourse samples. Bear in mind that during analysis there. is a constant alternation of focus from the particularity of the discourse sample, to the type(s) of discourse which it draws upon, and the configurations of discourse types to which it is oriented. Analysis should be directed at both: it should show features, patterns and structures which are typical of certain types of discourse, restructuring. tendencies in orders of discourse, and of using these conventional resources which are specific to this sample. Note that in any particular analysis some of the Doing Discourse Analysis
230
Discourse
categories are likely to be more relevant and useful than othe and analysts are likely to want to focus upon a small number rs, them.
Doing Discourse 233
What sorts of transformation does this (type of) discourse sample undergo?
the intertextual chains and uansformations relatively stable, or are they shifting, or contested?
there signs that the text producer anticipates more than one sort of audience?
(see pp. 83-4 above) The aim here is to look the interpretative implications of the intertexrual and inter-discursive properties of the discourse sample. This could involve the analyst in 'reader research', that is, research into how texts are actually interpreted.
MANIFEST INTERTEXTUALITY (see pp.117-23 above) Manifest intertexmality is a grey area between discourse practice and text:
it raises questions about what goes into producing a text, but it is also concerned with features which are 'manifest' on the surface of the text. The objective is to specify what other texts are drawn upon in constitution of the text being analysed, and how.
Genres differ in the modes of manifest intertextuality with which Is the text produced (consumed) individually or collectively?
(Are there distinguishable stages of production? anima-tor, author, and principal the same or different people?) What sort of non-discursive effects does this sample have?
How heterogeneous and how ambivalent is the text for particu-lar interpreters, and consequently how much inferential work is needed? (This leads directly to intertextual dimen-sions of the consuuction of subjects in discourse: see 'Social practice' below.)
Does this sample receive resistant readings? From what sort of reader?
CONDITIONS OF PRACTICE (see pp. 78-80 above) The aim is to specify the social practices of text production and consumption associated with the type of discourse the sample represents (which may be related to its genre: see the first ques-tion under 'Interdiscursivity' above).
Doing Discourse 232
Each of the three dimensions of discourse below. 'Interdiscursivity' and 'Manifest I1
upon text production, 'Intertextual chains' upon text
upon text consumption. (See chapter 4 for tailed discussion), I have added 'Conditions of discourse to bring in the social and institutional aspects briefly alluded to chapter 3 (pp. 78-80 above).
(see pp. 124-30 above) The objective is specify what discourse types are drawn upon in the discourse
under and how. Feel free to use the general term type If It IS not clear whether something is a genre, type,. or discourse (see pp. 125-8 above). The mam of Justifying ,an interpretation is through text analysis, by showing that your interpretation is compatible with the fea-tures of the and more compatible than others. Other sorts of evidence were mentioned above under 'Enhancing the corpus'.
Is there an obvious of characterizing the sample overall (in terms of genre)?(if so, what does it imply in terms of how the sample is produced, distributed, and consumed?) Does the draw upon more than one genre?
What aCtIVIty type(s), style(s), discourse(s) are drawn upon?
(can you specify styles according to tenor, mode, and rhetor-ical mode?)
Is the discourse sample relatively conventional in its interdis-cursive properties, or relatively innovative?
.pp. 130-3 above) The objective here IS s.peclfy dlsulbutlon of a (type of) discourse sample by descnbmg the intertextual chains it into that is the series of text types it is transformed into or out of: '
they are associated see (p. 128 above), and one arm here is to explore such differences.
Doing Discourse Analysis 235
by whom? Does one participant evaluate the utterances of others?
To what extent to participants formulate the interaction? What functions do formulations have, and which participant(s) formulate(s)?
COHESION (see pp. 174-7 above) The objective is to show how clauses and sentences are connected together in the text. This information is relevant to the description of the 'rhetorical mode' of the text (see p. 127 above): its structuring as a mode of argu-mentation, narrative, etc.
What functional relations are there between the clauses and sentences of the text?
there explicit surface cohesive markers of functional rela-tions? Which types of marker (reference, ellipsis, conjunc-tion, lexical) are most used?
Doing Discourse Analysis 234
discourse representation Is it direct or indirect?
What is represented: aspects of context and style, or just ideational meaning?
Is the represented discourse clearly demarcated? Is it translated into the voice of the representing discourse?
How is it contextualized in the representing discourse?
presupposition
How are presuppositions cued in the text?
they links to the prior texts of others, or the prior texts of the text producer?
they sincere or manipulative?
they polemical (such as negative sentences)?
And one further question:
there instances of metadiscourse or irony?
POLITENESS (see pp. 162-6 above) The objective is to deter-mine which politeness strategies are most used in the sample, whether there are differences between participants, and what these features suggest about social relations between participants.
Text
INTERACTIONAL CONTROL (see pp. 152-8 above) The objec-tive here is to describe larger-scale organizational properties of interactions, upon which the orderly functioning and control of interactions depends. An important issue is who controls interac-tions at this level: to what extent is control negotiated as a joint accomplishment of participants, and to what extent is it asymmet-rically exercised by one participant?
What turn-taking rules are in operation? the rights and obligations of participants (with respect to overlap or silence, for example) symmetrical or asymmetrical?
What exchange structure is in operation?
How are"topics introduced, developed, and .established, and is topic control symmetrical or asymmetrical?
How are agendas set and by whom? How are they policed and
Which politeness strategies (negative politeness, positive polite-ness, off record) are used, by whom, and for what purposes?
ETHOS (see pp. 166-7 above) The objective is to pull together the diverse features that go towards constructing 'selves', or social
ETHOS (see pp. 166-7 above) The objective is to pull together the diverse features that go towards constructing 'selves', or social