Both conversation analysis (inspired by ethnomethodology) and discourse analysis (of the kind proposed and practised by Potter and Wetherell) are usually treated as self-sufficient approaches to studying the social world, rather than as mere methods that can be combined with others. And there are two areas where their conflict with other approaches is clearest. First, they reject the attribution of substantive and distinctive psychosocial features to particular categories of actor as a means of explaining human behaviour. Second, they reject use of what the people they study say about the world as a source of information that can ever be relied on for analytic purposes. These two negative commitments mark conversation analysis and discourse analysis off from almost all other kinds of social scientific research. In this article, I consider how sound the justifications are for these commitments. I conclude that they are not convincing and that neither approach should be treated as a self-sufficient paradigm.