The analysis of dynamic games hinges on assumptions about players’ actions and beliefs at information sets that are not actually reached during game play, and that players themselves do not expect to reach. However, it is not obvious how to elicit intended actions and conditional beliefs at such information sets. Hence, key concepts such as sequential rationality, backward induction, and forward induction do not readily translate to testable behavioral assumptions. This paper addresses this concern by introducing a novel optimality criterion, structural rationality. In any dynamic game, structural rationality implies sequential rationality. In addition, if players are structurally rational, their intended actions and conditional beliefs can be elicited via the strategy method (Selten, 1967). Finally, structural rationality is consistent with experimental evidence indicating that subjects behave differently in the strategic and extensive form, but take the extensive form into account even if they are asked to commit to strategies ahead of time.