We present experimental evidence in favor of Harley & Haugen's (2007) proposal that the acceptability of denominals is determined by the degree of similarity of their prepositional objects to their nominal roots in various respects. In contrast to Kiparsky (1997), we argue that all denominals are derived from nominal roots regardless of their alleged (in)ability to combine with PP-objects different from the root (hammer with a shoe versus ??tape with pushpins). On the basis of an experiment eliciting English similarity judgments, we show that nominal roots are not just defined by one predominant semantic aspect (as in Dowd 2010), but by a variety of aspects (shape, function, material, a.o.), along the lines of the Qualia Structure theory of concepts proposed by Pustejovsky (1995). Shape, function and material all have a large effect on denominal sentence acceptability, and we conclude that acceptability is best accounted for by all of these aspects, and possibly others.