Babble and random-noise masking of speech in high and low context cue conditions.

"Perceptual" masking of speech by multitalker speech (babble) has been widely reported but poorly quantified. Furthermore, the validity of the construct of perceptual masking is questionable. This report describes an experiment using a newly standardized test of speech perception in noise (SPIN) with both babble and spectrally matched random-noise maskers. Classical psychophysical ogive curves were used to model speech recognition as a function of signal-to-noise ratio (S/N). The two maskers yielded speech recognition functions of the same steepness but different locations on the S/N axis. The high-context items of SPIN yielded speech recognition curves with steeper slope and different locations on the S/N axis than the low-context items. These data are used to argue that perceptual masking was not documented (under certain assumptions) and that the superior masking of babble may be explained in purely acoustical terms. Speculations are offered about the possible acoustical differences that could be responsible for the differences in masking effect.