Acoustic realization and category split in L1 English speakers’ L2 Korean stop production*

Abstract

This study investigated L2 Korean lenis and aspirated stops produced by L1 English adult learners in the early stages of L2 learning, examining how they utilize voice onset time (VOT) and F0 for category distinction over the course of L2 learning. Narrative speech samples (568 word-initial stops) were collected from 34 university students enrolled in first-year (25 instruction hours) and second-year (125 instruction hours) Korean courses in Canada. Bayesian mixed-effects logistic regression models were constructed to quantify the acoustic structures of the VOT and F0 for lenis-aspirated types across learner groups. The results revealed that the predictive roles of VOT and F0 were patterned differently by proficiency: first-year speakers solely utilized VOT, whereas second-year speakers exhibited increased use of F0 and reduced use of VOT. In addition, higher prediction accuracies based on the acoustic models were observed for second-year students than for first-year students, and for lenis stops than for aspirated stops in both groups. This prediction accuracy distribution suggests that the initial acoustic structure of the undifferentiated category is closer to the lenis target in its production representation. The current findings are significant in depicting the category split of L2 sounds assimilated into a single L1 category in terms of English speakers' acoustic realization of Korean stops.