Compensation for complete assimilation in speech perception: The case of Korean labial-to-velar assimilation

In connected speech, phonological assimilation to neighboring words can lead to pronunciation variants (e.g., ‘garden bench’ → “gardem bench”). A large body of literature suggests that listeners use the phonetic context to reconstruct the intended word for assimilation types that often lead to incomplete assimilations (e.g., a pronunciation of “garden” that carries cues for both a labial [m] and an alveolar [n]). In the current paper, we show that a similar context effect is observed for an assimilation that is often complete, Korean labial-to-velar place assimilation. In contrast to the context effects for partial assimilations, however, the context effects seem to rely completely on listeners’ experience with the assimilation pattern in their native language.

► Listeners use phonetic contexts to reconstruct intended words from assimilated forms.
► Such compensation is evident often with phonetic residuals in the assimilated form.
► Koreans showed compensation even when there was no bottom-up phonetic support.
► Cross-linguistic data with Korean, Dutch and English listeners were compared.
► It is implied that language-specific phonological knowledge plays a role.