In the first experiment, a continuum of 10 harmonic musical intervals was constructed from a minor to a major third. Four pairs of stimuli with constant physical distances were presented to seven musicians in a two-interval forced-choice discrimination task. Either silence, an interfering tone, or a noise burst was interposed between the two stimuli in a pair. Unbiased discriminability was found to be consistently higher for pairs straddling the boundary between two categories than for the

more »

... int pairs. The interfering tone lowered overall discrimination but left the shape of the function unchanged, whereas the noise burst had no effect. Experiment 2 used a simitar paradigm, but the continuum consisted of the single tone that had cued the minor-major distinction for intervals. Discrimination of this series did not show consistent changes as a function of continuum position. In Experiment 3, triads that varied in either interval or overall pitch were presented to musicians for sorting according to one dimension or another. The result was that there were much longer latencies to sort according to interval when pitch varied irrelevantly than vice versa. These results demonstrate that there are changes in discriminability associated with learned categories and suggest that there may be two hierarchically organized stages. A dual-processing model is discussed in which the listener has available both auditory and categorical information. In a now classic article, Miller (1956) pointed out that identification of unidimensional stimuli was closely linked to the capacity of shortrterm memory; however, in the realm of speech perception an exception to this formulation was reported. Liberman, Harris, Hoffman, and Griffith (1957) found that listeners were not able to perceive graded differences between stimuli but rather tended to hear only discrete changes between one pho-Some of these experiments formed part of a doctoral dissertation submitted to