Underlying Factors in Non-native Listeners’ Recognition of High-Variability Speech

The present study explores how non-native listeners recognize high-variability L2 speech by the function of task paradigm, masking type, and individual listener differences. Eighteen Korean-speaking learners of English who were different in English fluency, immersion experience, and working memory capacity participated in a series of four sentence recall tasks - one was presented in a single sentence paradigm with single-talker babble and the other three were presented in a sequential sentence paradigm (with single-talker babble vs. multi-talker babble with 8 dialects vs. mixed-gender multi-talker babble with 8 dialects). The arcsine-transformed accuracy of recalled keywords indicates that the sequential sentence paradigm provides a more sensitive measure to the cognitive processing load required for high-variability speech recognition than the traditional recall paradigm. Of the two indexical maskers (talker dialect and gender), only dialect variety significantly increased listeners’ processing load during L2 speech recognition. Advanced non-native listeners were better at recognizing words in both low- and high-variability L2 speech than less advanced non-native listeners, suggesting L2 fluency is a more reliable index of non-native listeners’ ability to recognize variable L2 speech than other listener factors such as immersion experience and working memory capacity.