Effects of a Complex Reflection on Vowel Identification

Part of a sound heard in a typical room has traveled to the listener directly from the source, but this direct sound becomes mixed with later arriving copies that come to the listener after being reflected from the room's surfaces. Such reflections are often complex in such a way that they have a filtering effect, which distorts the spectral envelope of the reflected sound. As a consequence there is a change in the sound's spectral envelope, or "spectral transition", as the reflections start to mix with the direct sound at the listener's ears. The present experiments ask how such spectral transitions affect the fusion and segregation of parts of a vowel sound. The vowels in "itch" or "etch" were played through a filter to simulate the effects of an undistorted direct sound with a later-arriving, complex pattern of reflections. This filter was designed to impair identification of the two test-vowels in a way that is independent of the reflections' delay. Listeners heard one test-word on each trial and identified it as itch or etch using a rating scale. A signal detection analysis was used to find the bias-free discrimination index p(A). This index was obtained when the reflections' delay was varied, as well as when parts of the vowel before or after the spectral transition were replaced with "signal correlated noise". The pattern of results indicates that the reflections fused with the part of the direct sound that follows the spectral transition, while the part before the spectral transition is hardly affected by the subsequent reflections. In other conditions, this fusion of the reflections persisted when their interaural time difference (ITD) was increased beyond the direct sound's ITD. However, such fusion was less evident when the reflections were replaced by a putative masking sound.