Real-time spoken word recognition: A data explanation approach.
Dr. Michael Tanenhaus, University of Rochester
Monday, October 27th at 5:30 p.m., 118 Psychology
About Dr. Tanenhaus
Michael Tanenhaus is a Professor in the Department of Brain and Cognitive Sciences and the Department of Linguistics at the University of Rochester in New York. He is also the director of the Center for Language Sciences.
Abstract
For most of the last two decades two classes of “pattern-recognition” models have dominated the spoken word recognition literature: (1) pattern recognition models with abstract representations, and (2) exemplar-based pattern recognition models. I will argue for another class of model: “data explanation” models that evaluate hypotheses about the state of the world according to how well those hypotheses predict the observed properties of the perceptual input. I will outline an explanation-based framework in which prediction and learning play a central role and present supporting evidence from four lines of work. The first shows that listeners use information from the first few pitch periods in a preceding vowel to generate lexical hypotheses (e.g., the shwa in “the” in “the lamp…”). The second demonstrates that the information structure of an utterance affects whether or not increased syllable duration is interpreted as evidence for a prosodic boundary in an onset embedded word (“hamster”). This result is unexpected given traditional approaches to spoken word recognition. The third line of work shows that listeners maintain gradient representations of VOT and combine that information rationally with right context. The fourth line of work shows that changes in speech rate affect whether people hear the same acoustic input as “…a yellow duck swimming” or “…yellow ducks swimming”, and then uses appearing and disappearing indefinites as a domain for exploring real-time adaptation and cue integration within a belief-updating framework.
Suggested Readings
Farmer, T. A., Brown, M. & Tanenhaus, M. K. (2013). Prediction, explanation, and the role of generative models in language processing. Behavior and Brain Sciences, 36, 211-212.[.pdf]
Meredith Brown, Anne Pier Salverda, Christine Gunlogson & Michael K. Tanenhaus. (2013): Interpreting prosodic cues in discourse context. Language, Cognition and Neuroscience, DOI: 10.1080/01690965.2013.862285[.pdf]
Anne Pier Salverda, Dave Kleinschmidt & Michael K. Tanenhaus. (2013): Immediate effects of anticipatory coarticulation in spoken-word recognition. Journal of Memory and Language, 71, 145-163.[.pdf]
Dr. Tanenhaus will also be holding a brown bag discussion
Title: Dynamic updating of perceptual representations in situated language
The discussion with be at 12:00 in Psychology 230. You can find out more about the event and access supplementary readings here.