Patterns of individual differences in second language vowel perception

Abstract

A growing body of research has focused on individual differences in speech perception (Idemaru et al., 2012). Individual differences are especially evident in second language (L2) speech perception in which learners show different cue weighting strategies in phonetic categorization (Schertz et al., 2015). However, little is known about whether the individual differences in cue weighting strategies are related across perceptual tasks and contrasts. The present study aims to further our understanding of individual differences in L2 speech perception by examining whether learners’ perceptual cue weighting strategies are related to their discrimination abilities and how individual cue weights are related across contrasts for individual learners of English. Twenty-four native Korean learners of English participated in cue weighting experiments and an AX discrimination test after one year of residence in Canada. For perceptual cue weighting, stimuli were resynthesized speech continua based on natural recordings of the endpoints of both /bɪt/-/bit/ and /bɛt/-/bæt/ contrasts produced by a male native speaker of Canadian English. Five steps varying in formant frequency were crossed with 5 duration steps and presented in a two-alternative forced-choice identification task. To test learners’ ability to discriminate L2 vowels, a two-talker AX discrimination task was employed, where participants heard natural recordings by two female native speakers and judged whether they heard “same” or “different” stimuli. Linear mixed-effects models were used to examine the potential relationship between individual learners’ cue weighting strategies (β) and their discrimination abilities in the discrimination task (dʹ). Correlation analyses were carried out to examine the relation between individual learners’ cue weights across vowel contrasts. Results showed that there was a significant effect of Contrast (β = –0.45, p = 0.01), suggesting that the learners’ discrimination abilities were overall better for the /ɪ/-/i/ contrast than the /ɛ/-/æ/ contrast. A significant effect of Spectrum was found (β = 1.05, p < 0.001); that is, learners who attended more to spectral information in categorizing vowels were more sensitive to the phonological contrast. However, individual learners’ vowel duration weight was not correlated with discrimination performance (β = 0.10, p = 0.68), suggesting that attending to vowel duration (which is a secondary cue for native listeners) did not affect discrimination performance. A marginally significant interaction between Contrast and Spectrum was found (β = 0.63, p = 0.07), suggesting that the effect of Spectrum on discrimination abilities differed across contrasts. That is, the learners’ spectral weight had a greater influence on their discrimination performance for the /ɪ/-/i/ contrast than the /ɛ/-/æ/ contrast. Correlation analyses between cue weights across vowel contrasts for each of the cues showed that the learners’ cue weights were correlated across contrasts for both cues (Spectrum: r = 0.60, p = 0.001, Duration: r = 0.55, p = 0.004). That is, learners who showed greater spectral reliance on the /ɪ/-/i/ contrast also relied more on the same cue for the /ɛ/-/æ/ contrast. The same pattern was also observed for the duration cue. The findings of this study suggest that the perceptual cue weighting strategies in individual L2 learners for a given phonological contrast are also manifested in these same individuals’ discrimination performance for that contrast. In other words, more native-like performance in a cue weighting task is related to better recognition of naturally produced vowels. We also found that learners whose perceptual weighting strategies are stable across contrasts show better discrimination abilities. Thus, the present results confirm earlier reports that individual L2 learners vary greatly in the extent to which they rely on particular phonetic cues. However, we further demonstrated that these individual differences are not random, but are systematically associated with how well L2 learners discriminate contrasts and that successful learners use a stable cue weighting strategy across contrasts.

Date
Location
Carleton University, Ottawa, Canada