Perception of native language speech sounds does not predict non-native speech sound learning

Individual differences are often observed in laboratory studies of non-native speech sound learning. One possible explanation for this variability is that better detection of fine-grained contrasts within native language categories might facilitate non-native learning. For instance, Diaz et al. (2008) found larger MMN responses to both native and non-native speech contrasts in good compared to poor perceivers of a non-native contrast, suggesting a general speech-related skill. The current study explores whether the ability to discriminate subtle differences in native language speech sounds correlates with non-native speech sound learning. To test this, we trained participants on a non-native, Hindi dental/retroflex contrast and assessed their categorization and discrimination of a native /da/-/ta/ continuum. Additionally, participants completed a visual Flanker task in order to control for general motivation in the experimental setting. Neither native language measures nor the Flanker task predicted non-native speech sound learning abilities. Rather, using k-means cluster analysis, we found two distinct groups of learners and non-learners that did not significantly differ on native language or Flanker measures, suggesting that non-native speech sound learning may be independent of those skills. Instead, non-native learning success was best predicted by an ability to discriminate the non-native contrast at pretest.