Abstract:
Compared to thresholds obtained with psychophysical procedures [Kewley-Port, J. Acoust. Soc. Am. 97, 3139--3146 (1995)], formant-frequency discrimination tasks using less well-trained listeners and relatively higher stimulus uncertainty yield severely degraded performance. The present study is the first step to examine formant discrimination under more natural listening conditions simulated by increasing the trial-to-trial stimulus uncertainty, the phonetic context of the target vowels, and the level of cognitive load. Difference limens (DLs) were obtained for F1 and F2 from four vowels in either CVC syllables, a three-word phrase or a nine-word sentence under medium stimulus uncertainty. Cognitive load was presumably increased in the sentence task with the addition of an identification response. Analyses of effects of the phonetic context demonstrated that DLs were significantly degraded between CVCs, phrases, and sentences in the F2 region, but in the F1 region were relative similar. For the sentence task, the effect of training was shown to improve discrimination significantly by about a factor of 2. However, after training, little difference was observed between the two tasks, discrimination only and discrimination and identification combined. These results suggest that effects of more central levels of auditory processing on discrimination need to be further investigated. [Supported by NIH-NIDCD-02229.]