Having addressed myself to Antaki et al.'s six key points, in part by extending them, I move briefly now to add three to them. A seventh would be under-analysis through uncontested readings. Rather than formulating a monovocal account, good discursive analyses acknowledge the multiple and contested character of the interplay of discourses by showing how different discursive representations are built to interact with and ward off others. They have their own integrity of categories (in the sense that subject positions are elaborated in complementary ways) but these are frequently contested (if only to refute and reassert them) within any text. Highlighting such contests provides some internal reliability regarding the function and effects of any analysis, and the better discourse analyses that I have more recently encountered have formulated, or else demonstrated consideration of, several alternative ways of conceptualising the discursive domain under analysis.
An eighth proposition would be under-analysis through decontextualisation, whether this decontextualisation is on the part of the analyst in their account, or of their representation of the text. This point connects with my comments about lack of engagement, by which I mean the need to situate the text and one's analysis of it socially, historically, culturally and politically. While it is clearly logically impossible (as well as probably undesirable) to claim to identify the 'whole context' of the text, the issue is to provide both a rationale for it as a meaningful text to analyse (how it has come about, why it is important, who has which kinds of stake in it, and why and how), and to indicate the stance from which the analysis is conducted. Doing this helps to ward off the incipient objectivism that dogs so much psychological research, and (as I discuss in Burman, 1992, Burman, 1998), it also wards off the corresponding alternative (but equally mistaken) position of subjectivism. To use Antaki et al.'s phrase 'insightful and technically sophisticated work' (p. 16) can also, and I would argue inevitably, includes being politically committed and institutionally located. The alternative to 'anything goes' is surely not 'nothing goes'.
My final - but perhaps most important - addition is perhaps one that was assumed or taken for granted by Antaki et al., but I think is still worth making explicit: under-analysis through not having a question. The most uninteresting and weak examples of discourse work that I have encountered principally founder through the failure to specify why this analysis is being done, and is worth doing. Just as under-analysis, as Antaki et al. so well highlight, arises through presumption and premature selection of quotations and correspondingly unwarranted interpretation, it can also arise through an insufficiency of critical selectivity on the part of the analyst. In this respect the production of the data is actually necessary to the analysis, and we might expect innovative forms of data collection to always operate already analytically.