ABSTRACT
Threshold for a 1000-Hz sinusoidal signal was measured in the presence of simultaneous maskers composed of 2 to 200 sinusoidal components. Masker frequencies were drawn at random for each presentation from a range of 300–3000 Hz, excluding the signal frequency. To mimic the properties of components drawn from noise, the amplitude and phase of each component was drawn at random from Rayleigh or rectangular distributions, respectively. As in an earlier study, large amounts of masking were observed for maskers with very few components spread across a wide frequency range. In the first experiment, eliminating masker components from a 160-Hz wide critical band around the signal reduced the amount of masking, but considerable masking remained even for maskers with 10 or fewer components. In the second experiment, component frequencies, amplitudes, or both, were either fixed or randomized across the two listening intervals of a forced-choice trial; new frequencies were always presented on each successive trial. Amplitude randomization had no effect regardless of the number of components in the masker. Frequency randomization, however, produced large amounts of masking for maskers with 10 or fewer components. These effects typically show little change with extensive practice, and appear to be produced primarily by nonperipheral processes.
