Variable pathways in L2 sibilant category formation among Cantonese-English bilinguals

This study examines how Hong Kong Cantonese (HKC) speakers with different L1 sibilant systems produce sibilants in their L2 English. Some HKC speakers exhibit discrete s, ts, tsʰ vs tɕ•, tɕ•ʰ allophones conditioned by vowel rounding, i.e., tsi vs tɕ•y. Others use a single place of articulation in all contexts, predicting variable cross-linguistic mappings to English /s, ʃ, tʃ, dʒ/. Fifty-four HKC-English bilinguals (9M/9F × 3 age groups) and seven native English controls completed a Cantonese + English production experiment. Dynamic measurements from midsagittal tongue ultrasound, high-speed lip video, and synchronized audio were analyzed with generalized additive mixed models. Results show that L2 English /ʃ, tʃ, dʒ/ are typically native-like in terms of acoustics. Yet while younger speakers produce English postalveolars with the same tongue posture as their HKC tɕ, tɕ•ʰ allophones, older speakers recruit a novel (non-HKC) tongue position, reflecting the formation of a new L2 category. Speakers further vary in the temporal extent of anticipatory lip rounding on /s/, the use of secondary rounding for English /ʃ/, and the phonological distribution of /s/-/ʃ/ substitutions. These findings shed light on the nature of L2 category formation, which may combine novel L2 acoustic and articulatory targets with phonetic categories selectively transferred from L1.