Ethnomethodologists (or at least many of them) have been reticent about their theoretical sources and methodological principles. It frequently falls to others to make such matters explicit. In this paper I discuss this silence about theory, but rather than entering the breach by specifying a set of implicit assumptions and principles, I suggest that the reticence is consistent with ethnomethodology's distinctive research 'program'. The main part of the paper describes the pedagogical exercises and forms of apprenticeship through which Garfinkel and Sacks aimed to develop ethnomethodology as a practice. These efforts were not entirely successful, partly because ethnomethodological 'practice' required an engagement with other fully-fledged practices. Aside from the difficulties of mastering such practices, it was unclear what an ethnomethodological study would add to, or take from, them. Whether successful or not, ethnomethodological research points to the specificity of discourse and action in any given practice which a general theory is bound to misconstrue. Current disputes about cultural constructivist versions of natural science illustrate the problems that arise when the terms of a general theory are used to describe and evaluate specific domains of practice. The paper concludes by recommending ethnomethodology as a way to dissolve an unbridgeable gap between cultural theories and socially located practices.