Acquisition of the complex three-way Korean plosive contrast by native English speakers

Learning to perceive foreign language speech sounds is a core challenge in adult second language acquisition. Previous research has considered how listeners learn novel foreign language categories for a known phonetic continuum [e.g., voice onset time (VOT)], or how listeners learn to use a previously unattended phonetic feature (e.g., F3). We investigated perceptual learning of the Korean three-way plosive contrast (lenis, aspirated, and fortis) by native English speakers. Unlike VOT continua in other languages, this contrast is distinguished by complex trading relations between VOT and pitch, with place of articulation differences in VOT adding further complexity. In this study, participants (N = 38) learned a vocabulary of 18 Korean pseudowords comprised of six minimal triplets (e.g., pan, ban, and ppan) by undergoing four days of high-variability (multi-talker) training on a lexical identification task. Mixture model analysis suggested two learner groups: (1) two-thirds of the participants were partially successful at learning words beginning with the fortis stops, but did not differentiate the lenis and aspirated stops; and (2) one-third of the participants successfully learned words beginning with the fortis stops, and exhibited progress distinguishing the lenis and aspirated stops. (Fortis stops most closely resembled listeners' existing English voiced stop categories.) Both groups acquired these contrasts best for bilabial stops and least accurately for alveolar stops.