In a recent article in this journal, Daniel Stephens argues against Chad Hansen’s and Chris Fraser’s interpretations of the later Mohists as realists about the ontology of kinds, contending that the Mohist stance is better explained as conventionalist. This essay defends a realist interpretation of later Mohism that I call “similarity realism,” the view that human-independent reality fixes the similarities that constitute kinds and thus determines what kinds exist and what their members are. I support this interpretation with a new, (...) detailed account of the Mohist conception of a kind on which kind relations lie in inherent similarities between the intrinsic features of objects. This account distinguishes kind relations from “uniting together” and part-whole relations, both of which, unlike kind relations, may be determined by convention. I argue that Stephens’s critique of realist interpretations fails because it confuses the ontological issue of what determines the existence of kinds with the semantic issue of what fixes the names for kinds. (shrink)