Effects of stimulus uncertainty, consonantal context and training on formant frequency discrimination

Abstract

Words in natural speech vary in their predictability, e.g., stimulus uncertainty ranges from low to very high across sentences. The ability to discriminate formants is strongly affected by longer phonetic context [Kewley-Port and Zheng (in press)]. This study investigates the effects of stimulus uncertainty from minimal to high uncertainty and the phonetic contexts /V/ or /bVd/. The primary formants were from four female vowels /ɪ, ε, æ, ə/. Using adaptive tracking, DLs for discriminating a small change in a formant was calculated in Delta Barks. In experiment 1, performance for five listeners, optimized by extensive training, began with minimal uncertainty, subsequently increasing uncertainty from 8- to 16- to 22 formants per block. Effects of higher uncertainty were less than expected, only decreasing performance by about 33%, although DLs for CVCs were 25% poorer than for isolated vowels. In experiment 2, performance in the 22-formant condition was tracked over 1 h for 37 listeners without formal laboratory training. DLs for untrained listeners were about 300% worse than for trained listeners, and comparable to untrained listeners in Kewley-Port and Zheng (in press). Results indicate that the effects of longer phonetic context degrade formant frequency discrimination more than higher stimulus uncertainty. [Work supported by NIHDCD-02229.]