Seminar Team Language 26/06/2025

Abbie Bradshaw - Leverhulme Early Career Fellow - Medical Research Council Cognition and Brain Sciences Unit

The role of prediction of sensory feedback in speech motor control

Models of speech motor control place high importance on the prediction of sensory feedback, with sensory prediction errors being used to inform and modify movements. This mirrors claims in ‘active inference’ accounts; domain-general theories of brain functioning which reconceptualize the nature of the perception-action interface in terms of a common process of minimization of prediction errors. Such accounts have been extensively applied to the control of manual action; however, they have received relatively little attention in speech motor control. In this talk, I will provide a brief introduction to active inference and how it differs from existing models of speech motor control. I will then present some experimental work testing different claims of active inference with regards to the use of predictions during speech. Firstly, I will present some work testing for the use of a shared set of predictions across speech perception and production. In a series of experiments using synchronous speech, I found evidence that interaction with another voice results in an updating of the same predictions that are used for sensorimotor control of the self-voice. Secondly, I will present some work investigating the mechanism by which predictions are combined with auditory feedback during speech motor control; specifically, whether predictions suppress (as in prediction error accounts) or enhance (as in sharpening accounts) the representation of speech auditory feedback. Overall, these findings suggest the utility of an active inference framework for driving novel investigations into the role of prediction in speech perception and production.