It is often thought that the boundaries and properties of art-kinds are determined by the things we say and think about them. More recently, this tendency has manifested itself as concept-descriptivism, the view that the reference of art-kind terms is fixed by the ontological properties explicitly or implicitly ascribed to art and art-kinds by competent users of those terms. Competent users are therefore immune from radical error in their ascriptions; the result is that the ontology of art must begin and (...) end with conceptual analysis. Against this tendency towards concept-driven ontology, I offer a trio of objections derived from: the cultural and temporal variability of concepts of art, the systematic tendency, on the part of would-be ontological assessors, to err on the side of familiar categories or, conversely, to exaggerate minor differences between familiar and unfamiliar practices, and the influence artworld precedents exert over expert and folk concepts alike. These considerations, I argue, mandate an epistemic humility that is simply unavailable to the concept-descriptivist. (shrink)