The evolution of crises depends upon interpreting intentions under uncertainty. We model crises as a game of two-sided incomplete information. Players are uncertain about their own payoffs from war because of differences between observable and actual capabilities. We derive four hypotheses, testing them against crises in Europe between 1815 and 1970. We show a nonmonotonic relationship between ex ante observable capabilities and the likelihood of violence in a crisis, as well as the ex ante likelihood of a negotiated settlement. We answer five questions: (1) How do differences in observable capabilities between rivals influence the likelihood of a crisis and the escalation to violence? (2) How do intangible capabilities alter the effects of observable capabilities on the likelihood of conflict and violence? (3) What do national leaders learn from the responses of their adversaries in crises? (4) Under what conditions can deterrence succeed? (5) Under what conditions are the strong likely to give in to the weak or vice versa in a crisis?