Spectral integration of synchronous and asynchronous cues to consonant identification

The goal of the present research was to determine how well observers utilize cues to consonant identification from different spectral regions that occur asynchronously as opposed to synchronously across frequency; such an ability would be useful for processing speech in the context of a spectro-temporally complex masker (e.g., competing speech). This was assessed by obtaining masked identification thresholds for VCV speech material, of the form \a\ C \a\, under various conditions of 10-Hz or 20-Hz square-wave amplitude modulation (AM). The speech tokens were filtered into 2, 4, 8, or 16 contiguous log-spaced frequency bands spanning 0.1 to 10 kHz. Bands were then modulated, with the pattern of that AM being either coherent across bands or 180degrees out of phase for adjacent bands. In the out-of-phase conditions the odd-numbered bands had coherent phase AM and the even bands had coherent phase AM, but the AM pattern across these two subsets of bands were out of phase. Results from these two conditions, along with further conditions employing only modulated even- or odd-numbered bands, allowed performance to be compared between stimuli characterized by synchronous and asynchronous cues. Results indicate that observers are able to utilize asynchronously presented cues to consonant identification efficiently across a range of conditions. (C) 2004 Acoustical Society of America.