Key points are not available for this paper at this time.
Masked threshold for spondees were obtained in (1) steady-state speech spectrum noise, (2) speech spectrum noise modulated by seven talker combinations, and (3) the seven talker combinations. These combinations were 1, 2, 3, 16, 32, 64, and 128 voices speaking continuous discourse. The longterm spectra of all maskers were equalized to a common level by averaging their intensity disparities across the 20 AI bands that contribute to intelligibility. Subsequent excesses in masking by a talker combination over masking by the companion condition of modulated noise are attributable to centrally mediated interference arising because the speech masker and the target signal are more perceptually similar than the target and the modulated noise are. This interference, which we have termed perceptual masking, appeared in the presence of all seven combinations of voices. It averaged 6.2 dB with the one talker and 7.2 dB with the two-voice background. The maximum (9.8 dB) was reached with three talkers. Thereafter it decreased, stabilizing at about 3 dB with 64 talkers. Supported by NINCDS, NS-09442.
Carhart et al. (Sat,) studied this question.