Multiplicative effects on sentence comprehension for combined acoustic distortions

This study was designed to explore the effect on speech comprehension of combining two types of signal distortion. A tape of clearly-enunciated sentences in quiet was distorted in each of four ways: low-pass (LP) filtering, time compression, interruption, and noise masking. Data are reported on a population of normal-hearing young men for multiple-choice answer tests of colloquial sentences of either LP filtered at 1, 2, 3, or 4 kHz, time compressed by computer at 250 words/min, interrupted (50 msec on--50 msec off), masked by speech-spectrum noise at +2 dB S/N, or given each of the 12 possible combinations of LP filtering plus the other three distortions. Individual distortion conditions were adjusted to reduce speech comprehension performance to about 90% accuracy, Low-pass filtering above 1 kHz reduced comprehension by no more than five to 10 percentage points, but when LP filtering was added to the other distortions in turn, latent effects were uncovered. The reduction in comprehension with the combined distortions was much greater than the simple additive effects of the distortion and LP filtering by themselves. For example, LP filtering above 2 kHz produced no measurable effect on sentence comprehension but this same distortion in combination with noise masking reduced performance from 89.4 to 59.7% correct (where 25% was chance). This study further validates the multiply-compounded nature of simultaneous types of distortion. The use of LP filtering extends the multiplicative principle to the simulated case of high-frequency hearing losses.