Stability in perceptual assimilation: Talker and vowel context effects

Abstract

A study was conducted to examine methodological and phonetic context variables in the discrimination and identification of non-native speech sounds and the stability of these patterns across experimental conditions. Native speakers of American English were presented with two voicing contrasts produced by three Hindi speakers, [k]–[g] and [t■]−[d■]. They appeared word-initially and in three vowel contexts [a, i, u] in a forced choice identification test and in a categorical AXB discrimination test. In the discrimination test, listeners were assigned to two conditions: (1) trials in which stimuli were produced by a single talker and (2) trials in which each stimulus was produced by a different talker. Single vs multiple talkers within trials were used to determine if a stronger correlation would be observed between identification and discrimination if listeners were forced to rely on more abstract representations in long-term memory when coping with talker variability. Multiple vowel contexts were employed to examine the stability of perceptual assimilation patterns. The results of pilot studies have demonstrated a significant vowel context effect on both identification and discrimination effect as well as a significant talker effect. The implications of the results for models of cross-language speech perception will be discussed.