Ultrasound production training of American English retroflex/ɹ/: A case of Korean L2 learners

This study investigates the efficacy of ultrasound tongue imaging as a tool for training Korean learners of English in producing the American English retroflex/ɹ/. Due to the absence of a direct equivalent of/ɹ/in Korean and the complex phonetic relationship between Korean and English liquid sounds, Korean learners often struggle with accurate production. In the experiment, nine Korean participants of intermediate-level proficiency were trained to produce American English retroflex/ɹ/. Ultrasound imaging was employed to provide real-time visual feedback on tongue movements, aiming to enhance the articulatory accuracy of both onset and coda variants. Participants underwent six training sessions over three weeks, and results revealed significant improvements in articulatory accuracy after the training, particularly in tongue tip elevation and dorsum lowering. Also, these improvements were sustained a month later, indicating effective generalization. This suggests that ultrasound feedback can successfully facilitate the learning of complex L2 sounds like the English retroflex/ɹ/. In addition, there was a notable asymmetry in the learning process between onset and coda retroflex allophones, with the coda variant ([ɹ]) being acquired more readily than the onset variant ([ɹ^w]). This discrepancy is attributed to the greater phonetic dissimilarity between the Korean L1 [l] and the English L2 [ɹ] in coda positions, compared to the dissimilarity between [ɾ] and [ɹ^w] in onset positions. This finding suggests that Korean learners are likely to acquire the coda variant earlier than the onset variant in their learning development of the English L2 retroflex.