John F. Culling
Quentin Summerfield
MRC Inst. of Hear. Res., University Park, Nottingham NG7 2RD, UK
Four 150-Hz-wide noise bands, centered on (1) 225, (2) 625, (3) 975, and (4) 1925 Hz, represented the formants of whispered vowels. Presented simultaneously, the four bands formed a competing perceptual organization, interpretable as /i/+/(open aye)/ (bands 1 and 4 vs 2 and 3) or /u/+/(backward eh)/ (bands 1 and 3 vs 2 and 4). Listeners were encouraged to perceive a particular organization by presenting pairs of bands with different interaural delays ((plus or minus)700 (mu)s) or with different interaural level differences (dichotically, one vowel to each ear). In control conditions, the four bands were presented with the same interaural delay (-700 (mu)s) and monotically (left ear). Listeners were instructed to identify the vowel heard on the left. The nominal target vowel was presented at different gain levels relative to the competing vowel. Listeners identified the target more accurately than chance only when the vowels were presented dichotically or with a positive target gain. Thus listeners were unable to group the formants of concurrent, whispered speech sounds by common interaural delay. In contrast, a similar experiment found that identification was facilitated when the noise bands representing the target vowel were interaurally uncorrelated, while the competing vowel was diotic. This result is consistent with a within-channel mechanism for segregation, such as interaural cancellation.