main content

Issues of method in Critical Discourse Analysis

The methodology of discourse analysis creates some areas of concern. Several critics point to its lack of systematicity, its lack of transparency, and its lack of strict guidelines or governing principles (Cohen, Manion & Morrison, 2008, p.391; Flick, 2002, p.201; Coyle, 1995, p. 247). Denscombe (2007, p.310) sums up the position:

A disadvantage of using discourse analysis . . . is that it does not lend itself to the kind of audit trail that might be needed to satisfy conventional evaluations of the research. It is not easy to verify its methods and findings . . . because the approach places particularly heavy reliance on the insights and intuition of the researcher for interpreting the data.

In response, a number of points can be forwarded in defence of discourse analysis. The first is that many now recognise that discourse analysis is better not understood as a discrete methodology but as a ‘field of research’ (Wetherell et al., 2001, p5), or a ‘scholarly orientation’ (Locke, 2004, p.2). In that sense a range of different methodological approaches exist within a broad developing tradition. Potter (2004) suggests that there are at least four quite distinct domains within discourse analysis, reflecting roots in linguistics, cognitive psychology, sociolinguistics and poststructuralism. Within the educational sphere and in particular within discourse analysis of policy, Taylor (2004) and Fairclough (2003) recognise an important distinction in the approach of ‘textually oriented discourse analysis’ and that of social theory.

Work which belongs to the tradition rooted in continental social philosophy and cultural analysis (Potter & Wetherell, 1994, p.47; MacLure, 2003, p.174) is generally not focused on fine-grained linguistic analysis. The method does involve a careful reading of the textual material in question, however, with interpretation supported by the linguistic evidence, but it recognises that discourse is more than language: ‘it is constitutive of the social world that is a focus of interest or concern’ (Bryman, 2008, p.499). This means that the text and the interpretation are related to broader social issues in Fairclough’s ‘transdisciplinary dialogue’ (2003, p.6), which would not be available if one were engaged merely in textually-oriented analysis.

Relevance of validity and reliability

Discourse analysis as reflexive research recognises that the outcomes are a reading of the text, an interpretation rather than the truth, the whole truth. This epistemological position which sees knowledge as provisional and incomplete must also turn upon itself and recognise the imperfect nature of discourse analysis as a practice. Thus, at least some of the criticisms of discourse analysis can be countered by simply pointing to its postpositivist position. Rapley (2008, p.128) suggests that the conventional research values of reliability, validity, and generalizability or replicability, are inappropriate for CDA, a position also held by Taylor (2001, p.319) because the very epistemological claims upon which CDA rests actually problematize such concepts. Rapley instead suggests that the claims of CDA should be judged in terms of their being ‘credible’ and ‘plausible’, and that this can be addressed by the research being open and transparent both about the textual evidence under review and about the basis of the claims made about it. Potter and Wetherell (1994) similarly suggest two features of critical discourse analysis methodology: that it should impart ‘coherence’ to a text, showing how it fits together in terms of content, functions and effect, and that it should be ‘fruitful’ in that it provides insights that may prove useful.


The ‘robustness’ of CDA research – the extent to which it can withstand intellectual challenge - can best be defended if the claims made are in alignment with the methods employed. If the discourse analyst presents findings in the terms indicated above – as a reading, an interpretation – then that would constitute robustness in that it does not overstate its claims. CDA is not claiming to be a positivist exercise and is not making positivist claims to knowledge. Its robustness instead must lie in the tight logical relationship between methods and claims.

The argument that discourse analysis is tendentious or slanted is an obvious danger of any ‘reading’ of a text but can also be answered by being explicit about the nature of one’s research perspective. For example, van Dijk (2001) sees critical discourse analysis as having an explicitly emancipatory role, and Fairclough (2003) sees it as being ‘socially transformative’. The foregrounding of such a perspective is one means of making the analysis open to question.

Rigour and Transparency

However, on methodological rigour and transparency, there are perhaps more points to answer. Numerous researchers within discourse analysis do outline clear frameworks and methodological tools for their analysis (Fairclough, 1992, 2005b; Wodak, 2001; Jäger,2001; van Dijk, 2001). To some extent these reflect the different traditions within which they are working but also the nature of the texts under question: conversations, policy texts, media reports, speeches. A further variable is that discourse analysis within the social theory tradition is highly dependent upon the broader knowledge and understanding of the particular analyst: the intertextual issues raised, the non-discursive practices addressed, the links and connections with theory and practice beyond the text cannot be set down as a method but emanate from the particular awareness of the individual involved. Thus, in a sense, the analysis is personal. That does not mean that the analysis is subjective and ephemeral and so unfit for academic standards: instead it recognises the role of the reader, the fluidity of interpretation, and the fact that as Derrida (1981) suggests reading and interpretation is a process, a dialogue between reader and text. The reflexive element of this understanding must also be understood: just as the analyst’s reading of the text is an interpretation, so a similar interpretative process will involve the reader of the analysis. The analyst cannot prescribe what the reader reads, understands, and knows. This phenomenon is part of what Giddens (1987, p.20) means by the ‘double hermeneutic’, a useful concept which also embraces what was identified earlier about the way in which the subject can be shaped by the discourse.

Thus, rather than resting on a single methodological toolkit, it is possible for work in this area to be founded on the broader principles of discourse analysis: a careful reading of the text; interpretation and comment clearly related to textual evidence; and an analysis which aims to be credible, plausible, coherent, and fruitful, its robustness founded on the limits of its claims.

Adapted from: Gillies, D. (2009). Critical discourse analysis and current education policy. Unpublished doctoral dissertation, University of Strathclyde, Glasgow, Scotland.