Daniel L. Weber
Stephen P. Sanger
Dept. of Psychol., Wright State Univ., Dayton, OH 45435
This experiment examined a listener's ability to judge signal duration in the context of a 2IFC, distribution discrimination procedure. A ``standard'' distribution for the duration of a 1-kHz sinusoid had a mean of 10, 30, 100, 200, or 500 ms. Comparison distributions had means 1.05, 1.1, 1.2, and 1.4 times that of the standard. Both were normal distributions with standard deviations equal to the difference in the means. (All conditions thus yield a d' of 1.0 for an ideal observer making a single observation. Obtained d's vary as the separation of the means changes with respect to the duration jnd.) In each observation interval, listeners were presented with 1, 2, 3, 4, 5, 6, 8, 12, or 16 different samples (a fixed number within a block of trials) from one of the two distributions and had to identify which interval was drawn from the comparison (longer mean) distribution. The observed increase in d' with up to six presentations is well described by the observation-integration model [D. M. Green, J. Acoust. Soc. Am. 30, 904--911 (1958)] in all conditions. Our listeners generally fail to show improvement beyond 8 presentations, so the partitioned-variance model [D. E. Robinson and B. G. Berg (1986)] described by Lutfi [R. A. Lutfi, J. Acoust. Soc. Am. 86, 934--944 (1989)] provides a better fit to the entire set of data. [This research was supported by a grant from AFOSR through WPAFB AL/CFBA.]