Signal properties that reduce masking by simultaneous, random-frequency maskers

Large amounts of simultaneous masking can be produced by changing the frequency content of multicomponent maskers with each presentation. Much of this masking appears to be informational, that is, produced by stimulus uncertainty. This study examined whether relatively simple changes in the properties or presentation mode of the signal could increase the saliency of the signal and reduce this masking. The number of masker components varied from 2 to 100 across conditions. The reference signal was a 200-ms, 1000-Hz sinusoid, presented monaurally with the masker. Across experiments, changes in masking relative to the reference condition were examined for different signal types (amplitude-modulated, quasifrequency-modulated, or narrow-band noise), durations (100 or 10 ms), and presentation modes (diotic, dichotic, or cross ear). The use of AM and NBN signals improved performance more consistently than the QFM signal, which degraded performance for some listeners. Lower masking in the reference condition for these listeners may have limited the effects of signal type. Dichotic (versus monaural) presentation produced larger reductions in masking for more listeners and conditions. Comparisons to results with broadband maskers and other patterns in the data, however, suggest the dichotic advantage may not clearly reflect a reduction of masking due to uncertainty. Separating masker and signal onset/offset times by shortening signal duration produced the largest and most consistent reductions in masking produced by masker-frequency uncertainty.