Previous studies show that attention plays an important role in second language (L2) phonetic attainment. This study compares the effect of two high variability phonetic training methods (identification (ID) and categorical AX discrimination (DIS)) on specifically targeted sounds and on implicitly exposed but untargeted sounds. Four groups of Spanish/Catalan bilingual learners of English were trained on either a subset of English vowels (/iː ɪ æ ʌ ɜː/) or word-initial and word-final stops. The study also examined if the potential effect of training generalized to new stimuli and persisted two months after training. Results revealed that trainees significantly outperformed the controls in their identification of targeted sounds, and improvement generalized to new stimuli and was maintained after training, showing the efficacy of both training methodologies. However, while all trainees performed similarly with initial stops, ID trainees outperformed DIS trainees in vowel perception. Interestingly, only DIS trainees showed a significant improvement in the perception of untargeted sounds, indicating that this training method (possibly due to the absence of labels and the exposure to two physically present stimuli in each trial) may be more suited to enhance learners’ perception of both targeted and untargeted sounds.

High-Variability Phonetic Training (HVPT) has been shown to be effective in improving the perception of the hardest non-native sounds. However, it remains unclear whether such training can enhance phonological processing at the lexical level. The present study tested whether HVPT also improves word recognition. Late French learners of English completed eight online sessions of HVPT on the perception of English word-initial /h/. This sound does not exist in French and has been shown to cause difficulty both at the prelexical (Mah, Goad & Steinhauer, 2016) and the lexical level of processing (Melnik & Peperkamp, 2019). In pretest and posttest participants were administered a prelexical identification task and a lexical decision task. Results demonstrate that after training the learners’ accuracy improved in both tasks. Moreover, these improvements were retained four months after posttest. This is the first evidence that short training can enhance not only prelexical perception, but also word recognition.