This paper considers how the idea of ‘discipline’ can best be conceptualised, both in general and particular terms. Much previous research has employed a strong essentialist approach, a model of disciplines which exaggerates the homogeneity of specific disciplinary features and accords disciplines generative powers which they rarely possess. That approach is disabling because it closes down the appreciation of the heterogeneity within disciplines, as well as occluding the reasons for that heterogeneity. However, for researchers this oversimplified model offers the attractions of simple research questions and research designs. The consequence of using such a model is, though, that its distortions threaten the robustness of higher education research. The paper argues for a more sophisticated conceptualisation of disciplines, one which deploys a moderate form of essentialism. It applies Wittgenstein’s notion of family resemblances to the task of depicting disciplines and explores the implications for research of this more nuanced model.