| Sumario: | Two experiments were conducted to evaluate the efficacy of ultrasound-based training in improving the pronunciation of English /l/ and /ɹ/ sounds in six Japanese speakers in Japan. Participants were trained using ultrasound tongue imaging and shadowing exercises, with half utilizing ultrasound feedback on their own. The analysis involved (i) SSANOVAs of tongue shapes of those sounds with root mean square (RMS) and (ii) pre- and post-training sound discrimination tasks with 795 native English listeners on Amazon Mechanical Turk (MTurk). The results showed significant enhancement in the differentiation of /l/ and /ɹ/ tongue curves post-training, particularly in word-initial positions. Intelligibility notably improved among participants who showed greater increase of lingual contrast at posttest, indicating the training is especially beneficial for those with neutralized /l/ and /ɹ/ pre-training. However, perceived intelligibility increased more for /l/ than for /ɹ/, particularly in participants who utilized individually customized ultrasound feedback. A participant exhibited a delayed tongue tip raising for coda /l/, suggesting temporal coordination of multiple gestures can be learned via visuo-proprioceptive feedback with ultrasound on themselves. These promising findings necessitate further large-scale investigation.
|
|---|