Predicting vowel discriminantion accuracy through cross-linguistic acoustic analyses

This study compares the acoustic properties of Australian English and Brazilian Portuguese vowels as a means of predicting L2 discrimination difficulty. Euclidean Distances between the vowels of the two languages were computed to quantify acoustic similarity and to predict discrimination difficulty for Australian English learners of Brazilian Portuguese. Results show that Euclidean Distances successfully predict classification patterns in statistical models. We further compared the models' results to those of a previous study reporting Australian English listeners' discrimination of Brazilian Portuguese vowels, showing that real listeners' discrimination difficulty is indeed predicted by the current acoustic analyses.