analysing talk
and text
a course for the Universitat Autònoma
de Barcelona
Charles Antaki
Loughborough University
lectures   1     2     3     4     5     6     7     8     9     10     11     12  

Lecture/Seminar 11: Varieties of Discourse Analysis II


Iñiguez, L (2003) Análisis del Discurso (chapter 3). Editorial UOC.

Wetherell, M (2001) Themes in Discourse Research: the case of Diana. In Wetherell et al (eds) Discourse Theory and Practice London: Sage

Antaki, C, Billig, M, Edwards, D and Potter, J (2002) Discourse Analysis means doing analysis. Discourse Analysis On-Line, vol 1. In Spanish in Athenea, vol 3.

van Dijk, T (2001) Principles of Critical Discourse Analysis. Chapter 22 in Wetherell et al Discourse Theory and Practice, London: Sage

Last week I was talking about various methods people used in Discourse Analysis. I mentioned the more linguistic approach of working with grammatical and stylistic features, then the more general approach of looking for 'repertoires'. I described how Gilbert and Mulkay identified two linguistic repertoires in scientists' talk and showed how they functioned together to perpetuate a discourse of reliable Science. Then I talked about Potter and Wetherell's identification of apparently contradictory 'attitudes' in discourses of immigration and national identity.

We shall stay with Potter and Wetherell's classic work, because they give a useful account of what the discourse analyst actually does when he or she sits down to do their work.

Potter and Wetherell's 8 steps (as described in their 1987 Discourse and Social Psychology).

Protesting (rightly) that mechanical methodology is wrong for DA, Potter and Wetherell nevertheless lay out 10 steps of what to do. We must remember not to take these as being hard and fast 'rules'.

1. Decide your research question

2. Select your sample of data. Not a trivial decision: it flows from what your research object is and how you think it is manifested in the world of language. If you're investigating (say) 'heterosexuality', do you seek out the views of heterosexuals? Or would you get a better fix on it by talking to non-heterosexuals? What's the difference? Or, another example: if you're interested in 'conspiracy theories' you could get one classic account (e.g. the notorious 'Protocols of the Elders of Zion' fake) and subject it to close analysis, or alternatively you might collect modern examples (eg about the 11th September 2001 attacks on the World Trade Center in New York and other US targets) and see how they are similar among themselves.

3. Collect records. You'll need fairly accurate records; your memory of what people say won't do. So you're committed to transcripts, cuttings, letters, diaries, videotapes, and so on.

4. Interviews. Potter and Wetherell like these because they force the issue; they apply some of the rigour of traditional positivist psychology (the same questions can be asked to many people, allowing comparison). But, of course, this introduces the Researcher as a potentially disturbing or intrusive presence into the data - a significant problem, as psychologists well know. Potter and Wetherell are alive to this, but it is not always the case that they remember that their respondents were answering interviewers' questions. It is worth saying that Potter has become much more cautious about interviews, and now takes a position which is much closer to CA: that it is better to analyse more 'naturally-occuring' talk.

5. Transcribe. Then you can go over it.

6. Coding. Do a first pass through the data and keep your eye out for anything that looks interesting (like Gilbert and Mulkay did and noticed 'error' talk). But these 'codings' are only provisional, and always qualitative. No discourse analyst would code their material for the sake of counting them up in a quantitative analysis.

7a) Analyse. This is the crux of it all, of course, and it is here that Potter and Wetherell admit that there is no standard method - no 'recipe'. Rather, there are a few rules of thumb: "Why am I reading the passage this way? What features produce this reading?". Ignore nothing ; anything may turn out to be significant. Specifically, look for variability and consistency in what is said (and perhaps what is not said).

7b) go through all the data again, checking the appearance of any 'repertoires' or 'discourses' that you have started to find. "It should be clear, then, that there is no analytic method, at least as understood in social psychology. Rather, there is a broad theoretical framework, which focuses attention on the constructive and functional dimensions of discourse, coupled with the reader's skill in identifying patterns of consistency and variation" (Potter and Wetherell, 1987, p 169).

7. Validate. Look for
a) how the 'discourses' you have identified helps understand the coherence of the data
b) how the participants themselves orient to the discourses (that will remind you of 'participant orientation' in Conversation Analysis)
c) new problems - what is still unexplained?.
d) Fruitfulness - what it tells you - the 'aha' experience

8. Write up. Be frank about what your argument is and how the data are to be interpreted.

Although those steps from Potter and Wetherell's 1987 book, which is now sixteen years old, they are still useful as a way of describing the general 'method'. Where things have changed is in step 6 and 7. There is now a great variety of what counts as coding and analysis. This has led to a recent debate about methods.

Some common problems in doing Discourse Analysis

No one in DA believes that the steps above are a 'recipe' that can be followed automatically. Doing DA needs a strong element of scholarly insight that simply can't be put down as cook-book instructions.

Nevertheless some DA work does appear that seems to lack much insight. Of course, this can happen in any discipline. But as things are still developing in DA, some authors are concerned to correct this error (as they see it). A group of researchers at Loughborough University have written an account of what they believe are 'six mistakes' in doing DA (see the link in the Readings above).

Critical Discourse Analysis

So far I have been talking about a kind of DA which does not explicitly start from a certain political position. If there is political implication in what it finds in the material, then it will notice it. But it does not specifically set out to find it.

There is a different view, expressed by those who promote Critical Discourse Analysis. They differ among themselves, but most would agree with van Dijk when he says that they focus on the role of discourse in the (re)production and challenge of dominance. (van Dijk, 2001, p 300).

In other words, for Critical Discourse Analysis (CDA), you first have a theory of dominance (or power), then you find materials in which you can see how discourse promotes one group's power over another. Or, still better, you might have a theory or hypothesis about how such power can be challenged, and again you look into discourse to see how that can be done (this, however, is still rather rare in DA).

These theories need not be political, or left wing, or feminist, although in fact CDA tends to be some combination of these. Norman Fairclough, for example, or Erica Burman and Ian Parker, three discourse analysts closely associated with the critical movement, are clear about their political perspectives. See, for example, Fairclough, N (2001) Language and Power, London: Longmans, and Burman, E and Parker, I (eds) (1993) Discourse Analytic Research : Repertoires and Readings of Texts in Action. London: Routledge)

CDA Method

There is no special method, as such, which distinguishes CDA from other forms of DA. The difference is in the degree to which the analysis depends on the analyst's prior theory about the data. Compared to Conversation Analysis, for example, CDA is very dependent on its prior categorisation of its participants (for example, it may view a speaker as male or female, or as Black or White, even if the participants in the scene do not give off any orientation to gender or ethnicity).

This dependence on prior theory is essential to CDA, yet brings it into conflict with those discourse analysts who want to stay closer to the materials in their data. The clearest conflict is with Conversation Analysis, and recently there has been an illuminating dispute between Emmanuel Schegloff, Margaret Wetherell and Michael Billig in the pages of Discourse and Society. If there is time on the course, I hope we will be able to have a look at it.

This week's seminar

Discussion of the varieties of discourse analysis.

For next week's seminar

Next week will be a review week, so bring along any questions and queries you may have about any part of the course.