Phase effects in masking: Within- versus across-channel processes

The effects of bandwidth and component phase on masking were investigated using 200-ms narrowband $\text{(1-}{\mathrm{ERB}}_{\mathrm{N}})$ and broadband $\text{(5-}{\mathrm{ERB}}_{\mathrm{N}})$ cosine-phase (CP) and random-phase (RP) harmonic complex maskers, centered at 1 or 6 kHz. A continuous notched-noise was used to restrict off-frequency listening. The masker fundamental frequency (F0) was 25 Hz. In experiment 1, thresholds were measured for sinusoidal signals at 1 and 6 kHz, gated with the maskers. Thresholds were lower in the CP than in the RP masker, for both bandwidths, but the effect was markedly greater for the wider bandwidth. For the CP maskers, thresholds were markedly lower for the $\text{5-}{\mathrm{ERB}}_{\mathrm{N}}$ than for the $\text{1-}{\mathrm{ERB}}_{\mathrm{N}}$ bandwidth; for the RP maskers, there was a small effect in the opposite direction. Experiment 2 used 1- and 6-kHz CP maskers. The masker components in the ${\mathrm{ERB}}_{\mathrm{N}}$ around the signal frequency were presented to one ear, and the remaining components were presented contralaterally. Thresholds were much higher than when all components were presented to the same ear, and were higher than for the $\text{1-}{\mathrm{ERB}}_{\mathrm{N}}$ masker alone, suggesting that the low thresholds for broadband monaural presentation do not depend on “high level” across-channel comparisons. Simultaneous masked thresholds could be predicted well using a model based on a simulated auditory filter, a level-dependent compressive nonlinearity, and a sliding temporal integrator; it was not necessary to assume the involvement of across-channel processes or of selective listening in the masker dips.