Masking produced by spectral uncertainty with multicomponent maskers

The first experiment examined thresholds for signals at 250, 1000, and 4000 Hz presented simultaneously with maskers that had from 1 to 100 sinusoidal components. The number of masker components remained constant throughout a block of trials, but the component frequencies were drawn at random from a 5000-Hz range for each presentation. Maskers with as few as 10 components produced more masking than maskers with up to 100 components or than broadband noise. The amount of masking decreased dramatically if the same masker occurred in both intervals of a forced-choice trial, even if the masker changed between trials. The second experiment examined whether judgments of the number of components present in a stimulus might aid signal detection. Such a strategy would not be effective, since about 50% more components are needed to discriminate between stimuli, regardless of the initial number of components. These experiments indicate that listeners sometimes are unable to use a single-filter detection strategy, even when it is advantageous to do so.