Parameters of spectral/temporal fusion in speech perception.

Abstract

When the distinctive formant transition of a synthetic syllable is presented to one ear while the remainder (the “base”) is presented to the opposite ear, listeners report hearing the original syllable in the ear receiving the base—a phenomenon called “spectral/temporal fusion” by Cutting (1976). We have found that the mere onset (i.e., the first pitch pulse, 10 msec in duration) of an isolated, contralateral third-formant (F3) transition can be sufficient to cue the /da/-/ga/ distinction in this way. We also varied the relative onset times of isolated F3 and base and compared three types of F3 segments (50-msec time-varying, 50-msec constant, 10-msec onset) under both dichotic and diotic presentation. Time-varying F3 segments were superior to constant ones, especially when they lagged behind the base. Diotic performance exceeded dichotic performance, but only when F3 preceded the base, suggesting that upward spread of masking occurred in diotic presentation when F3 coincided with energy in the lower formants. Perhaps most interestingly, subjects’ tolerance of temporal asynchrony (roughly ±50 msec) was about the same in dichotic and diotic conditions, suggesting that the temporal integration mechanism that combines phonetic information from the isolated F3 segment and the base operates similarly in both conditions.