This paper proposes an alternative reading of what conventions are and how they might be used by social scientists in theoretical and empirical work. In the first section of the paper, I trace the modern conception of convention to two characterisations offered by David Hume. I claim that Hume’s two notions of convention provide the basic intuition behind the majority of modern approaches. The second section highlights an important and often implicit characteristic that most theories of convention share: the desire to explain the normativity of conventional practices has led commentators to characterise convention as a sub-category of social rules. I go on to argue that the Wittgensteinian literature on rule-following undermines this strategy and that rules cannot provide the normative guidance required of them by social theorists. The third section describes a promising alternative. I argue that the notion of exemplar, first proposed by Thomas Kuhn in the history and philosophy of science, can be used to clarify and advance the study of convention. The paper concludes with a illustration of how this alternative framework can be used by social scientists.