Encoding of acoustic and visual features in L1 and L2 high vowels: Bangla-English bilinguals

This study investigates how Bangla-English late bilinguals acquire genuinely novel second language (L2) phonological contrasts with a focus on the encoding of acoustic and visual cues in English high vowels. Bangla and English share the front-back dimension of vowels; however, the English tense-lax distinction constitutes a double “NEW scenario” for Bangla learners, requiring them to split native categories. Using a multimodal design, acoustic formants and visible lip geometry (horizontal and vertical aperture) were extracted and analyzed from 10 bilingual speakers(L1 Bangla, L2 English). Results from Bayesian mixed-effects models demonstrate that the shared front-back contrast is robustly encoded in both languages via F2 and horizontal lip aperture. However, the novel tense-lax contrast exhibits an asymmetric, channel-specific acquisition. For acoustics, learners were successful in forming distinct L2 categories for tense and lax vowels through F2. However, the visual articulatory encoding was only partially realized; while speakers showed a global gestural re-tuning of lip postures for L2 speech, the tense-lax contrast was not systematically differentiated visually. Cross-linguistic mapping revealed that speakers handled acoustic features independently of visual features of L2 vowel production. These findings support a stage-wise model of L2 phonological learning where acoustic category formation precedes visible articulatory differentiation.