Martin E. Rickert
Donald E. Robinson
Dept. of Psychol., Indiana Univ., Bloomington, IN 47405
Consider a two-interval same-different task in which listeners are asked to discriminate between trials on which a sample of noise is presented twice and trials on which two different samples are presented. Previous research has used this method to determine the effects of a wide range of stimulus conditions including temporal location of an uncorrelated segment, duration, bandwidth, correlation, and overall level [T. E. Hanna, Percept. Psychophys. 36, 409--416 (1984); S. F. Fallon and D. E. Robinson, J. Acoust. Soc. Am. 92, 2630--2635 (1992)]. The data indicate that discriminability is best when changes occur at the end of the sample and that the ratio of the duration of the target segment to the total duration is independent of overall duration. In this paper, a stimulus-oriented model for discriminating samples of Gaussian noise is developed. The model is a derivative of the equalization--cancellation theory of binaural masking level differences [N. I. Durlach, J. Acoust. Soc. Am. 35, 1206--1218 (1963)] and of Jeffress' models of monaural masking [L. A. Jeffress, J. Acoust. Soc. Am. 44, 187--203 (1968)]. This approach begins with a cancellation stage in which the waveforms from the two intervals are jittered in amplitude of differenced. Then, the power of the difference waveform is approximated in a manner similar to Jeffress' leaky integration. Predictions of the model will be compared to the data of Fallon et al. [Work supported by AFOSR.]