Properties come in clusters. It seems impossible, for instance, that a mass could float free, unattached to any other property. David Armstrong takes this as a reductio of the bundle theory and an argument for substrata, while Peter Simons and Arda Denkel reply by supplementing the bundle theory with accounts of property interdependencies. I argue against both views. Virtually all plausible ontologies turn out to be committed to the existence of free masses. I develop and defend the view that the clustering of properties is a mere contingent truth, on grounds that properties can be subtracted one-by-one. This opens the door not just to the (unsupplemented) bundle theory, but also to any plausible account of the relation between objects and properties.