Abstract:
In everyday listening the acoustic attributes of a sound are often used to recover specific information about the sound-producing object or event. In such cases, correlational analysis may be used to study the rules of representation underlying the discrimination of natural sound events [R. A. Lutfi and E. Oh, J. Acoust. Soc. Am. 95, 2967(A) (1994)]. Applications involving two types of natural discriminations, the discrimination of differences in the material composition of two objects, and the discrimination of changes in the motion of a sound-emitting source, are described. Sounds are synthesized according to the principles of theoretical acoustics such that the nominal values of acoustic parameters uniquely identify the material or the feature of motion to be discriminated. Small perturbations about these values are introduced from trial to trial, and partial correlations with the listener's response are taken as estimates of the degree of reliance placed on each acoustic parameter. The correlations reveal significant differences in individual listening strategies for the same task, but a general tendency to overly rely on one acoustic parameter. The resulting information loss amounts to an 80% reduction in performance efficiency in some cases. [Research supported by NIDCD R01 DC01262-04.]