A closer look at perceptual epenthesis in cross‐language perception.

Abstract

Previous studies have shown that listeners have difficulty distinguishing non‐native CC sequences from their CVC counterparts (e.g. Japanese [ebzo]‐[ebuzo], English [zgamo]‐[zəgamo]). Some have argued that the phonology “repairs” non‐native consonant clusters with vowel insertion (“perceptual epenthesis”), causing listeners to respond that CC and CVC sequences are the same. Production studies, however, have shown that speakers produce non‐native CC sequences many ways, including epenthesis, changing C1, deleting C1, and prothesis. To test whether these other repairs of non‐native clusters are also difficult to distinguish, English listeners (n=37) participated in an AX discrimination task that paired C1CVCV stimuli with CəCVCV (epenthesis), əCCVCV (prothesis), CVCV (deletion), or C2CVCV (C1 change). Each of these repairs caused some difficulty and there was an interaction between repair type and manner combination (stop‐stop, stop‐nasal, fricative‐stop, fricative‐nasal). Listeners were more accurate when C1 was deleted (81% accurate), followed by epenthetic and C1 changes (both 77%), then by prothesis (59%). Furthermore, compared to previous studies testing only the epenthetic repair, presenting listeners with various repair types dramatically improves discrimination in the epenthesis condition. These results indicate that perceptual epenthesis may be a task effect and that top‐down phonological influences are more complex than previously assumed. [Research supported by NSF.]