On the robustness of high variability phonetic training effects: A study on the perception of non-native Dutch contrasts by French-speaking learners

There is growing evidence in the literature for the positive effect of high variability phonetic training (HVPT) on the perception of non-native contrasts. In the present study, we aim to examine the robustness of perceptual training effects. We define robustness along three dimensions: (1) the generalizability of the training to novel tokens and talkers, (2) the long-term retention effects of the training, and (3) the effect of training in non-optimal listening conditions, i.e., with noise added to the signal. The participants are 48 adult L1 French learners of Dutch in Belgium, 27 of whom are enrolled in secondary education, while the others are university students (N=21). Participants are assigned to an experimental (N=27) or a control group (N=21). Both groups take a pre-test, post-test and delayed post-test, which consists of a lexical identification task with and without noise. The experimental group is trained on five Dutch sound contrasts in five multimodal HVPT sessions, consisting of perceptual identification tasks with feedback and metalinguistic information. The results show a nuanced picture: overall, where training effects are found, learners are able to generalize these to novel tokens and talkers, thus confirming the effectiveness of HVPT for pronunciation training. However, the results also reveal considerable variability in the effectiveness of HVPT along most robustness variables, which can to a large extent be attributed to the moderating variables we examined, being the type of learners (secondary education vs. university) and the type of sound contrast