Maximum-likelihood yes-no procedure for gap detection: effect of track length.

A maximum-likelihood yes-no procedure was used to measure minimum detectable gaps (MDGs) at 1 and 4 kHz using two track lengths (15 and 30 trials). Results from 11 normal listeners show no difference between MDGs for the two track lengths, and variability of the MDGs did not differ significantly for the track lengths. Results from simulations indicate that the variability of MDGs from real listeners is considerably larger than that imposed by the psychophysical procedure. Additional simulations show that random variability of listeners' response criterion is a compelling explanation for the excess variability. These findings indicate that changes in a listener's threshold over time impose a lower bound on the variability obtainable with a yes-no procedure. They imply that increasing the number of trials in a track, beyond the minimum number required to obtain a stable threshold estimate, offers little or no advantage for the test-retest reliability of a clinical threshold measurement.