Generalizing timbre space data across stimulus contexts: The meta-analytic approach

Abstract

Studies of the multidimensionality of the complex auditory attribute called timbre often represent the perceptual structure in terms of a spatial model in which distance has a monotonic relation to perceived dissimilarity. Acoustical correlates are sought for the dimensions of these timbre spaces in order to quantify them psychophysically. An important question concerns the extent to which the results obtained with one stimulus set generalize to other stimulus sets. A meta-analytic study of data from several timbre spaces using musical tones was performed with the same multidimensional scaling algorithm. The effect of changing a subset of the sounds on the perceptual relations among common sounds was evaluated in conditions where the degree of variation along perceptual dimensions across sets was similar or different (new salient dimensions added). A change in the sounds present in the stimulus set, keeping salient dimensions constant, had little effect on perceptual structure of the common sounds in the two sets. Changing salient acoustical correlates present, while resulting in new dimensions, did not strongly affect the perceptual structure of the common sounds. Results demonstrate that, while a change in stimulus set may introduce new perceptual dimensions, the perceptual structure among sounds common to two sets remains similar.