This paper sets out an ethnomethodological approach to the study of gender and interaction, and demonstrates how the topic of ‘gender’ can be studied empirically via its categorial reference in talk-in-interaction. I begin by charting the history of ethnomethodological accounts and studies of gender, starting with Garfinkel's groundbreaking work and the subsequent ‘doing gender’ project, alongside a more general discussion of feminism's relationship to ethnomethodology. I then consider two related trajectories of research, one in conversation analysis and the other in membership categorization analysis, both of which deal with the explication of gender's relevance to interaction, but in somewhat different ways that raise different problems. Finally, drawing on data from different institutional settings, I show how ‘categorial’ phenomena such as ‘gender’ can be studied as phenomena of sequential organization using the machinery of membership categorization alongside conversation analysis. I suggest that the analysis of members' categories in their sequential environment allows language and gender researchers to see how everyday notions of gender are taken up, reformulated, or resisted, in turns of talk that accomplish conversational action; that is, how categories ‘might be relevant for the doing of some activity’ (Sacks, 1992 vol. 1: 597).