The science of emotion is in a crisis: despite an enormous literature studying emotion from the perspectives of psychology and neuroscience, there is little agreement on any theory of emotion. The most acute problem is that our commonsense concept of "emotion" essentially involves conscious experience, whereas the scientific concept should not. I suggest that this state of affairs is, however, no different than in any other domain of cognitive psychology or cognitive neuroscience. Numerous examples show the need for a scientific concept of emotion. Both psychological and neurological data require such a concept in order to make sense of the data. I argue that (a) we cannot dispense with a scientific concept of "emotion"; and (b) such a concept needs to be grounded in broad data across a range of species; and (c) we need to begin formulating specific criteria for its application. In this paper I begin by outlining what the problem is, give some examples from my own research, and conclude with a framework for thinking about emotions that gives them scientific purchase.