Four experiments investigated memory for timbre using the interpolated-tone paradigm [Deutsch, Science168, 1604–1605 (1970)], in which participants discriminate pairs of tones (standard and comparison) separated by intervening (interpolated) tones. Interpolated tones varied from the standard tone in spectral similarity (within-dimensional variation), fundamental frequency (cross-dimensional variation), and repetition frequency. While the latter two variables had negligible effects on timbre memory, interference with timbre memory increased with the spectral similarity of the interpolated tones to the standard tone. The findings closely parallel those found for pitch memory, and suggest that memory interference depends on perceptual similarity in both cases.