Classical rationality as accepted by game theory assumes that a human chooser in a given moment has consistent preferences and beliefs and that actions result consistently from those preferences and beliefs, and moreover that these preferences, beliefs, and actions remain the same across equal choice moments. Since, as is widely found in prior experiments, subjects do not follow the predictions of classical rationality, behavioral game theorists have assumed consistent deviations from classical rationality by assigning to subjects certain dispositions— risk preference, cognitive abilities, social norms, etc. All of these theories are fundamentally cognitive theories, making claims about how individual human minds work when choosing. All of them are fundamentally wrong in assuming one kind of consistency or another. Or at least, all of the proposals for consistency in belief, preference, and action with which we are aware turn out to be wrong when tested experimentally.