Structure/agency theories presuppose that there is a unity to structure that distinguishes it from the (potential) diversity of agents' responses. In doing so they formally divide the robust social processes shaping the social world (structure) from contingent agential variation (agency). In this article we question this division by critically evaluating its application to the concept of role in critical realism and structural functionalism. We argue that Archer, Elder-Vass and Parsons all mistakenly understand a role to have a singular structural definition which agents may then diverge from. Drawing on the work of Gross, Mason and McEachern we argue instead that if agents diverge in their conceptions of what role incumbents should do, there is no single role definition, but rather a range of diverse role-expectations. Acknowledging this can help us to understand variation in role behaviour, with different incumbents potentially being more exposed to some expectations than others. We argue that considering roles in this way can extend the ability of social scientists to identify robust social processes shaping role behaviour and decrease the extent to which they need to call on contingent factors in such explanations.