James McQueen March 29th, 2011

Post-talk update: slides available here.

James McQueen will be delivering the first ICESL-sponsored departmental colloquium in Psychology at 4 pm on March 29th in Tobin Hall Room 423.

Speech recognition depends on abstract knowledge about sounds, voices and words
James M. McQueen, Donders Institute for Brain, Cognition & Behaviour, Centre for Cognition, and Behavioural Science Institute, Radboud University Nijmegen, and Max Planck Institute for Psycholinguistics, Nijmegen, The Netherlands

Each spoken utterance is a novel perceptual episode that arises through a unique combination of sounds and words, talker voice characteristics, and multiple contextual influences. In this talk I will use data on perceptual learning to argue that speech comprehension depends on abstraction over these episodes. In the first part the focus will be on experiments which demonstrate that listeners – adults and children – can tune in to unusual speech sounds using their lexical knowledge (their knowledge about how words ought to sound). Critically, lexically-guided retuning is based on phonological abstraction. Because the listener retunes boundaries between abstract speech-sound categories, learning can be generalized across the lexicon, to all words containing those sounds. In the second part of the talk I will present evidence on abstraction in voice learning. Listeners – adults and infants – recognize novel talkers using abstract speech-sound categories and by forming abstract voice categories. In the final part of the talk I will turn to the suprasegmental properties of words. Listeners appear to have abstract knowledge about prosodic structure – syllable duration in Dutch and lexical stress features in Italian – and use this knowledge during the recognition of newly-learnt words. Thus, abstract representations of speech sounds, talker voices and spoken words mediate the way we map episodic speech inputs onto utterance interpretations.

Co-sponsored by Developmental Science and Cognitive Science