Two experiments demonstrated the way in which musicians and nonmusicians process realistic music encountered for the first time. A set of tunes whose members were related to each other by a number of specific musical relationships was constructed. In Experiment 1, subjects gave similarity judgments of all pairs of tunes, which were analyzed by the ADDTREE clustering program. Musicians and nonmusicians gave essentially equivalent results: Tunes with different rhythms were rated as being very dissimilar, whereas tunes identical except for being in a major versus a minor mode were rated as being highly similar. In Experiment 2, subjects learned to identify the tunes, and their errors formed a confusion matrix. The matrix was submitted to a clustering analysis. Results from the two experiments corresponded better for the nonmusicians than for the musicians. Musicians presumably exceed nonmusicians in the ability to categorize music in multiple ways, but even nonmusicians extract considerable information from newly heard music.