Comodulation masking release for multicomponent signals.

Detection of signals composed of one, two, or three pure-tone components was examined in comodulated and noncomodulated masking noises. The masking noise was either a single 30-Hz-wide narrow band of noise, two narrow bands of noise, or three narrow bands of noise. Comodulation masking release (CMR) was greatest when (1) the signal was a single pure tone; (2) the masker was composed of three noise bands (as opposed to two); and (3) at least one flanking band was lower in frequency than the signal. Substantial CMRs did occur, however, for two- and three-component signals presented in two- and three-component maskers, respectively. The results of the present experiments did not support a "dip listening" hypothesis, nor models based strictly on across-frequency differences in stimulus envelope pattern or correlation. The results were more consistent with a model of CMR in which across-frequency envelope difference is coded by subtracting the envelope at the signal frequency from the envelope at a flanking frequency.