Why do masker fluctuations as in interfering speech lower the speech-reception threshold?

Abstract

Two experiments will be presented to explain the difference in speech-reception threshold (SRT) between conditions with a steady-state noise masker and an interfering voice. In experiment I, the possible role of comodulation masking release is investigated by manipulating the comodulation in the interfering voice by introduction of temporal shift among filter bands of various width. The spectral spread of masking from the manipulated interfering voice was controlled by interleaving mutually shifted speech bands with 1/3-octave bands of noise. Although comodulation in interfering speech appears to be very important for the low SRT, the contribution of across-frequencyprocessing of masker fluctuations—commonly considered as the origin of CMR—is only 1.3 dB. In experiment II, the level dependence of masking release with an interfering voice is investigated. The data confirm the hypothesis by Festen and Plomp (1990) that the release from masking with an interfering voice is limited by forward masking. It appears that up to about 55 dBA, the release from masking increases with level up to about 7 dB. Above 55 dBA, the difference in SRT obtained with a noise masker or an interfering voice is constant due to the limited modulation depth of speech.