Individual subjects with or without musical training made similarity judgments of pairs of tones on a nine‐point scale. Each subject was run in three or four sessions of 351 trials each. The tones had structures like those of musical instruments, being made of all 27 combinations of three dimensions, each at three levels. In Experiment 1, the dimensions were fundamental frequency F0, envelope, and relative amplitudes of harmonics. In Experiment 2, the dimensions were number of harmonics, envelope, and onset rate of harmonics. Analysis of data by means of multidimensional scaling showed a strong context effect. In Experiment 1, F0 had such high saliency that for most subjects no other dimension was present in perceptual space and thus no differences were found between musical and nonmusical subjects. By holding F0 constant in Experiment 2, subjects were able to use harmonic as well as envelope structure in judgments. Differences between musical and nonmusical subjects appeared, and we discuss the basis for these differences. For both experiments, the curve relating latency of response to similarity was parabolic and, although a given subject’s perceptual space changes little over successive runs, there is some evidence from Experiment 2 that musical subjects have the more stable space of perceptual dimensions.
Subject Classification: 65.52, 65.75; 75.10.