Understanding the relations between mind, brain and behavior is an enduring challenge for science and philosophy. The present article focuses on the concept of internal states, of which emotions are the most studied subtype, in the recent neuroscientific literature. Internal states are conceived by neuroscientists as functional states implemented in neural circuits that drive behavior. To begin, we discuss current definitions of internal states, that emphasize both their intrinsic and relational properties. We stress the difference between preliminary characterizations of these states (that allow researchers to track them) and findings related to their intrinsic nature. We analyze three experimental studies interpreted within this framework, and outline the role of innovative methods in the process of discovery. We interpret the families of states under investigation as homeostatic property clusters, and suggest that the work of Anderson and Adolphs offers a solution to the problem of what constitutes a natural kind in neuroscience. Concerning the relationship between physical states of the brain and mental states, we make explicit the discrepancy between an eliminativist and a reductive project within the literature, and underline the importance of a choice between them. Finally, we suggest that studies of internal states have three interrelated objectives: a better grasp of brain-behavior relationships, a more principled attribution of mental states to nonhuman animals, and better animal models of our own internal states in clinical contexts. With this plurality of objectives comes a plurality of possible outcomes of ongoing research.