Nonnative phonetic category training in varying acoustic environments

Abstract

Past research has shown that reverberation has a pronounced detrimental effect on speech intelligibility, but no previous studies investigated how it affects the acquisition of novel phonetic categories. Here, we present a follow-up to a previous study that examined adult nonnative phonetic learning using speech stimuli presented in reverberant environments [Vlahou et al. (2014), ARO Abstract #806]. Listeners were trained to discriminate a novel dental-retroflex contrast in Hindi. Using virtual acoustics, stimuli were presented in anechoic space, in a single room, or in multiple rooms. For some subjects, training was supervised, consisting of a 2AFC categorization task with trial-by-trial feedback. For other subjects, training was unsupervised, in the form of a videogame which promoted stimulus-reward contingencies. Performance was evaluated on trained and untrained sounds presented in familiar and unfamiliar rooms. Overall, trained listeners outperformed untrained listeners. Supervised training induced more robust learning of the trained material and generalization of learning to untrained voice. Introducing multiple reverberant environments was beneficial for unsupervised, but not for supervised, training. These results suggest that phonetic adaptation to reverberation is robust and that experiencing novel phonemes in various acoustic environments can enhance unsupervised phonetic category learning. [Work supported by APVV-0452-12, TECHNICOM ITMS: 26220220182, and NSF-BCS-1057625.]