Effects of perceptual training on the perception and production of Korean boundary tones by Chinese learners

In many languages, the pitch contour at the end of a sentence signals whether a speaker is asserting, asking, or conveying an attitude such as regret or admiration. Korean uses such boundary tones extensively, whereas Mandarin uses pitch mainly to distinguish words. This study asked whether perceptual training with multi-talker input helps Mandarin-speaking learners acquire three Korean boundary-tone contrasts: statement L% vs question H%, regret L% vs supposition H%, and boast H% vs admiration LHL%. In Experiment I, mixed-effects models showed reliable training gains in all three identification tasks, largest for regret-supposition (+43.3% points) and smallest for boast-admiration (+17.6% points), with generalization to untrained talkers and items. Continuum-based analyses showed more limited change: the statement-question boundary moved into the native-like range, whereas the regret-supposition boundary did not. In Experiment II, native listeners' identification of learner productions showed no reliable training effect, and acoustic analyses revealed persistent difficulty producing the intermediate pitch peak that LHL% requires. Perceptual training thus produced robust gains in interpreting Korean boundary tones but no reliable immediate transfer to production, indicating that L1 pitch experience, form-meaning mapping, and contour complexity jointly constrain learning. Training for L2 intonation should combine perceptual practice with explicit production instruction.