Modelling English diphthongs with dynamic articulatory targets

The nature of English diphthongs has been much disputed. By now, the most influential account argues that diphthongs are phoneme entities rather than vowel combinations. However, mixed results have been reported regarding whether the rate of formant transition is the most reliable attribute in the perception and production of diphthongs. Here, we used computational modelling to explore the underlying forms of diphthongs. We tested the assumption that diphthongs have dynamic articulatory targets by training an articulatory synthesiser with a three-dimensional (3D) vocal tract model to learn English words. An automatic phoneme recogniser was constructed to guide the learning of the diphthongs. Listening experiments by native listeners indicated that the model succeeded in learning highly intelligible diphthongs, providing support for the dynamic target assumption. The modelling approach paves a new way for validating hypotheses of speech perception and production