Contemporary issues in masking hailing from Zwicker’s work.

Abstract

Recent work on masking has focused on cases that appear inconsistent with the power-spectrum model of masking. This model holds that simultaneously masked thresholds are determined by the signal-to-masker ratios within frequency-selective channels. More than a decade ago, Zwicker noted that an amplitude-modulated broadband masker is much less effective than a steady-state masker with the same energy. This effect, known as comodulation masking release (CMR), might be mediated by detection of envelope correlation or by listening during brief periods of low masker intensity (in the valleys). Recent experiments show that CMR is obtained for brief tones presented in the valleys of the masker, but not at the peaks. This finding favors the ‘‘listening in the valleys’’ explanation. Zwicker also was among the pioneers in the exploration of ‘‘overshoot’’ of masking. Recent experiments show that the overshoot depends on the onset of energy outside the critical band centered on the signal and not on energy within it. Finally, Zwicker had a long-standing interest in how detection thresholds depend on signal bandwidth. Recent experiments show that the energy at detection threshold is nearly independent of bandwidth for brief signals, but increases with bandwidth for long-duration signals. This finding indicates that the decision rule used for detection depends on duration. All three phenomena can be modeled if the weights applied to the frequency-selective channels of the power-spectrum model are assumed to vary over time because they are driven by the stimulus. [Work supported by NIDCD.]