Robert D. Irwin
Daniel L. Weber
Dept. of Psych., Wright State Univ., Dayton, OH 45435
The distribution discrimination procedure can be used to evaluate a listener's ability to discriminate along a single dimension of an auditory signal. This can be accomplished by sampling from two different distributions of the parameter and asking the listener to identify which sample was most likely to have come from the distribution with the higher mean. Information integration can be measured as the improvement in performance when multiple samples are presented. Whether information integration appears to be different for frequency discrimination and duration discrimination was examined. Subjects discriminated duration or frequency in a 2IFC task when different numbers of samples were presented. In the discrimination of the duration of a 1-kHz sinusoid, samples were drawn from normal distributions with either a mean of 100 or 130 ms; both had a variance of 900 ms. In the discrimination of the frequency of a 100-ms sinusoid, samples were drawn from normal distributions with means of 565 and 570 Hz and a variance of 25 Hz and from distributions with means of 565 and 585 Hz and a variance of 400 Hz. These conditions should be alike to an ideal observer in that all involve a standard deviation equal to the difference between their means. The majority of the listeners show better information integration for signal duration than signal frequency. [Work supported by AFOSF through WPAFB AL/CFBA.]