Although the literature on Wittgenstein's notion of criteria is extensive, it seems unsatisfactory. Most interpretations of criteria not only misrepresent Wittgenstein; more importantly, they misconstrue the relation between a mental state and the behavior characteristic of that state. If by “criteria” Wittgenstein meant what he has been taken to mean, it is unlikely that any mental states have criteria. In this paper I shall argue that a proper interpretation of Wittgenstein's notion provides an account of the relation between some mental states and the behavior characteristic of them which is at least approximately correct.According to current interpretations of criteria the relation between a form of behavior C and the mental state of which it is a criterion, S, is such that the proposition “A person displaying C is (or: very likely is) in S” is either a necessary truth, known a priori to be true, or “true in virtue of a definition, convention, or rule of language.”