How coordination can be achieved in isolated, one-shot interactions without com- munication and in the absence of focal points is a long-standing question in game theory. We show that a cost-benefit approach to reasoning in strategic settings deliv- ers sharp theoretical predictions that address this central question. In particular, our model predicts that, for a large class of individual reasoning processes, coordination in some canonical games is more likely to arise when players perceive heterogeneity in their cognitive abilities, rather than homogeneity. In addition, and perhaps contrary to common perception, it is not necessarily the case that being of higher cognitive sophistication is beneficial to the agent: in some coordination games, the opposite is true. We show that sub jects’ behavior in a laboratory experiment is consistent with the predictions of our model, and present evidence against alternative coordination mechanisms. Overall, the empirical results strongly support our model.