Perceptual compensation when isolated test words are heard in room reverberation.

Room reverberation usually degrades speech reception, such as when listeners identify test words from a 'sir'-to-'stir' continuum. Here, substantial reverberation introduces a 'tail' from the [s], which tends to fill the gap that cues the [t], and a degradation effect arises as listeners report correspondingly fewer 'stir' sounds. This effect is particularly clear when test words are preceded by a precursor phrase (e.g. 'next you'll get…') that contains much less reverberation than the test word. When the precursor's reverberation is increased to be the same as in the test word, the degradation diminishes as more 'stir' sounds are heard once again. This last effect has been attributed to a perceptual compensation mechanism that is informed by the precursor's reverberation level. However, a recent claim is that the degradation is caused by 'modulation masking' from precursors with a low level of reverberation. Such masking is likely to diminish when the precursor's reverberation level is raised, because reverberation acts as a low-pass modulation filter. Support for this hypothesis comes from results in conditions where degradation effects seem to be entirely absent, despite substantial reverberation. In these conditions, test words were played in isolation, with no precursor, and reverberation was kept at the same level in the test words of every trial. The experiments reported here have conditions that are similar, except that reverberation in test words is varied unpredictably from trial to trial, so that substantial-level trials are interspersed with trials that have a much lower level of reverberation. The result is that under these conditions, the degradation effect is entirely restored, allowing rejection of the modulation-masking hypothesis. An alternative is that some perceptual compensation comes from reverberation information within test words, and its effects accumulate over sequences of trials as long as the test word's reverberation level stays the same from trial to trial.