In this paper I criticize the interpretations of John Locke on natural kinds offered by Matthew Stuart and Pauline Phemister who argue that Locke’s Essay Concerning Human Understanding allows for natural kinds based on similar real essences. By contrast, I argue for a conventionalist reading of Locke by reinterpreting his account of the status of real essences within the Essay and arguing that Locke denies that the new science of mechanism can justify the claim that similarities in corpuscular structure imply (...) similarities in sensible qualities. I argue further that Locke rejects as meaningless any talk of kinds that appeals to similarities among real essences. On my reading of Locke, similarities in real essences are not only irrelevant to species, but natural kind theories based on themare unintelligible. (shrink)