The effect of mismatched acoustic-articulatory information on listener judgments: The case of Korean cluster /lp/*

This study investigates how listeners perceive the presence or absence of the stop consonant /p/ in the Korean /lp/ cluster, focusing on the relationship between acoustic and articulatory cues. Using data from 19 native speakers in the Seoul National University Multilingual Articulatory Corpus, we measured acoustic stop closure duration and lip aperture, and collected perceptual judgments from five native Korean listeners. Results revealed that acoustic and articulatory cues for /p/ strongly aligned for most speakers, with most tokens judged as having /p/ when both cues were present. Mixed-effects modeling showed that acoustic closure was the primary predictor of categorical /p/-identification, while articulatory lip closure alone had little effect. Crucially, several tokens exhibited mismatches between acoustics and articulation, showing either lip closure without acoustic silence interval or acoustic silence without lip contact. Listener judgments for these mismatched tokens were inconsistent and received lower goodness ratings, indicating perceptual uncertainty. These findings suggest that acoustic evidence alone does not always reflect articulatory events, highlighting the importance of incorporating both acoustic and articulatory information in speech perception research, especially when acoustic cues alone may not capture gestural nuances.