Over the past decade, many cognitive control researchers have studied to what extent adaptations to conflict are domain-general or rather specific, mostly by testing whether or not the congruency sequence effect (CSE) transfers across different conditions (e.g., conflict type, task sets, contexts, et cetera). The CSE refers to the observation that congruency effects in conflict tasks tend to be reduced following incongruent relative to following congruent trials, and is considered a prime measure of cognitive control. By investigating the transfer of this CSE across different conflict types, tasks, or contexts, researchers made several inferences about the scope of cognitive control. This method gained popularity during the last few years, spawning an interesting, yet seemingly inconsistent set of results. Consequently, these observations gave rise to a number of equally divergent theories about the determinants and scope of conflict adaptation. In this review, we offer a systematic overview of these past studies, as well as an evaluation of the theories that have been put forward to account for the results. Finally, we propose an integration of these various theoretical views in a unifying framework that centers on the role of context (dis)similarity. This framework allows us to generate new predictions about the relation between task or context similarity and the scope of cognitive control. Specifically, while most theories imply that increasing contextual differences will result in reduced transfer of the CSE, we propose that context similarity and across-context control follow a U-shaped function instead.