Range of performance for two tasks with random-frequency context tones

Abstract

Forty listeners were tested on two tasks known to show large individual differences in performance. In one task, termed sample discrimination, pairs of target tones were drawn from each of two overlapping frequency distributions, and listeners indicated which pair came from the higher distribution. Target tones were presented alone and in the presence of two flanking context tones that were either fixed at 500 and 4000 Hz or had Gaussian variation centered at these two frequencies. In the second task, masked thresholds were determined for a 1000-Hz tone in the presence of ten-component, random-frequency, simultaneous maskers. The maskers were drawn on each presentation from a pool of either 200 or 10 waveforms. For sample discrimination, the range of performance was large for both the no-context and fixed-frequency context conditions. For the Gaussian-context condition, few listeners exceeded chance performance. Significant effects of testing order were found, with exposure to Gaussian-context conditions degrading later performance. For masking, performance did not differ for large versus small waveform pools, and the 35-dB range of performance corroborated previous work. For both tasks, random-frequency context degraded performance for the majority of listeners. [Work supported by NIDCD.]