Six experiments examined the cognitive reality of hierarchic structure in music. The first experiment showed that subjects were only moderately able to match a performed rendition of a hierarchic structure to the piece of music from which it was derived, with performance slightly better than chance. Metric accent emerged as a significant predictor of the tones perceived by subjects as structurally important. The second experiment showed that subjects' correct matchings were unlikely to be based on aesthetic preference, because performed renditions of (rule-governed) structures were not regarded as aesthetically preferable to nonstructures. The third experiment attempted to increase subjects' matching of structures and their original pieces through various task modifications, but these modifications did not increase performance over the success rate achieved in Experiment 1. Three additional experiments determined whether evidence of the distinction between structureand surface could be obtained in various similarity judgment tasks. Pairs of musical fragments were composed so that the members of each pair embodied (1) the same structure and same surface harmony, or (2) the same structure but different surface harmony, or (3) a different structure but same surface harmony. The rating task used in Experiments 4 and 5 showed that the members of Type-2 pairs of fragments were perceived to be just as similar as the members of Type-1 pairs, but Type-3 pairs were rated significantly lower in similarity. Thus, similarity judgments were based on underlying hierarchic structure, even in the face of radical harmony differences on the surface. This effect increased in strength with repeated hearing. The results support the cognitive reality of hierarchic structure, but are influenced by the type of perception used in a particular similarity judgment task and by the experience of repeated hearings.