Natural versus concatenative speech in high-variability phonetic training: A pilot

In the quest to develop effective speech perception training, research has increasingly reported positive results with high-variability phonetic training (HVPT). Presenting utterances in different phonetic environments and by multiple speakers, HVPT provides a more representative input to map onto developing L2 sound systems. While there is much success surrounding HVPT, we still do not yet know what components of HVPT help second language learners. One characteristic that plays a major role in HVPT is the speech input, such as using “natural,” human voices versus synthetic voices. This study aimed to explore possible effects on segmental perception through high-variability training with either human voices or concatenative synthetic voices. Six non-native English-speaking participants trained with the /i/ - /ɪ/ contrast in one of the conditions. The study followed a pretest-posttest design and included one training session in between. The learning management system Canvas was used to display all materials. While this study highlighted the need to examine HVPT, it did not find conclusive results. Further studies will be needed to explore HVPT and how to maximize its gains