Unlike detective novels, where the last chapter provides unambiguous evidence about the truth of all hypotheses, scientific mysteries sometimes remain un solved. The problem often lies not so much in figuring out whether the butler did it as it does in reaching agreement about what a butler is. That is to say, many seemingly empirical issues turn out to be nonempirical matters of definition. The problem cannot be solved simply by insisting on defining one’s terms, because explicit definitions presuppose a developed theory. Nevertheless, some sort of implicit definition presumably underlies any claim about a phenomenon. In the study of emotion too, explicit definitions are more likely to be an outcome than a starting point. Our focus, then, is not on defining terms explicitly but on explor ing the issues raised by a comparison of the definitions of anger implicit in the model proposed by Berkowitz (this volume) and in an alternative cognitive model we propose.