L1 phoneme distributional features influence L2 perception: a comparison between audiovisual and auditory input using HVPT

We investigated second language (L2) phoneme acquisition among adult learners using High Variability Phonetic Training (HVPT) with audio, audiovisual, and viseme + audio cues. Theoretical insights from the PAM-L2 and the Analysis-by-Synthesis model guided the experimental predictions and informed the interpretation of the results. Two experiments tested how the presence of visual articulatory cues during training influenced the auditory identification of Southern Standard British English phonemes /d, ɛ, æ, ʌ, ɒ, θ/ by native Salento Italian speakers. Audio, audiovisual, and viseme + audio training were equally effective in enhancing L2 phoneme identification, confirming HVPT’s robustness. Nevertheless, visually prominent articulatory information did not boost auditory identification, despite observed benefits of audiovisual integration during the training sessions. Our results highlight that L2 speakers rely primarily on auditory input to construct new phonetic-phonological categories. Our findings are compatible with both PAM-L2 predictions and models based on the distinctive features specifying L1 and L2 phonemes. The /ʌ–ɒ/ contrast, an Uncategorized-Categorized Partially Overlapping assimilation for which moderate to poor learning is predicted, was the most resistant to training: here the distinctive feature [±round] is redundant (unspecified) in the learners’ L1. The /ɛ–æ/ and /d–θ/ contrasts, respectively Two-Category and Uncategorized-Categorized Non-Overlapping assimilations for which good learning is predicted, improved more readily, as L1 and L2 share specification for [±low] and [±continuant, ±voice, ±coronal, ±sonorant]. For /d–θ/, however, differences in manner, place, and voicing may also play a crucial role. We discuss the theoretical, experimental, and pedagogical implications of these findings for L2 research and teaching practices.