Armstrong (1997: 65–8, 1999) argues that all natural kinds supervene upon more simple monadic and relational universals. Moreover, since kinds supervene, then they are not an ontological addition to their base. In contrast, Ellis (1999; 2001; 2002) argues that natural kinds find a sui generis place in our ontology. Natural kinds require a distinct ontological category and are irreducible. Both Armstrong and Ellis agree that our scientific worldview should inform our general ontology. One should not proceed a priori to posit an ontological theory and then seek to reconcile it with the scientific worldview. Thus, the correct ontological account of natural kinds should be scientifically informed. In this paper, I argue that science suggests a more complex view of natural kinds than has been permitted by either of the above ontologies.