The semantics of natural kind terms has recently been seen as a problem of reference. Kripke and Putnam have suggested that their meaning begins with rigid designation, with any further implications emerging after empirical study. I part ways with this approach and instead offer an account that focuses on the contribution that these terms make to the inferential roles of different sorts of sentences. I note that natural kind terms play an odd array of grammatical roles, both as subject and as parts of predicates, and explain why this dual role contributes something pragmatically significant to explanations of natural phenomena.