The role of L1 feature specification in naive L2 vowel perception: Evidence from feature fission and fusion

This study investigates how feature specification in a learner’s native language (L1) – specifically, whether a feature is contrastively specified or overspecified – influences discrimination in naive perception. We focus on the perception of four high vowels ([i], [y], [ʉ], [u]) by Mandarin and Italian speakers. We propose that Italian specifies /u/ as [labial], while Mandarin overspecifies this vowel as both [labial] and [dorsal]. The vowel /y/ is native in Mandarin but non-native in Italian, while the vowel /ʉ/ is not native in either language. We test two possible feature redeployment mechanisms, fusion and fission, using an ABX task to examine four contrasts, /i/–/y/, /u/–/y/, /ʉ/–/u/, and /ʉ/–/y/. Feature fusion predicts that the non-native /y/ can be built by Italian naive learners by combining [coronal] and [labial] from /i/ and /u/, such that Italian participants should show similar discrimination accuracy for both /i/–/y/ and /u/–/y/ contrasts as Mandarin participants. In contrast, feature fission predicts that [labial] can detach from /u/ in Mandarin to form /ʉ/ without causing a merger. However, this process should not be possible in Italian, where fission would cause /ʉ/ to merge with /u/, since /u/ is exclusively represented with [labial]. Thus, Mandarin participants should show better discrimination accuracy for the /ʉ/–/u/ and /ʉ/–/y/ contrasts than Italian participants. Results from the discrimination task align with these predictions, supporting the proposed feature specifications and demonstrating the effects of L1 feature specification and feature redeployment (fusion/fission) on naive second language (L2) vowel perception. Importantly, while phonetic cues form the perceptual basis for sound discrimination, our findings indicate that language-specific feature representations guide how these cues are phonologically interpreted.