Understanding the neurocomputational mechanisms underlying speech sensorimotor adaptation

University Of Wisconsin-Madison, Madison WI

Investigators

Abstract

Project Summary Sensorimotor adaptation, learning from the mismatch between the predicted sensory feedback of our intended motor action and the actual sensory feedback we observe, is a fundamental skill that supports accurate movement control, such as speech. One common paradigm used to study the mechanisms of speech adaptation is examining how speakers adjust their speech production in response to an external perturbation applied to the auditory feedback of their speech acoustics in real time. While this paradigm has been proposed as a tool to assess the mechanisms of speech deficits in various neurogenic disorders, reduced speech adaptation is commonly found across various clinical populations, despite impairments in very different neural structures. This pattern is not well-predicted by the current models of speech motor control, potentially because speech adaptation may not involve a single learning process (as in current models) but, as suggested by extensive work in other (nonspeech) motor domains, two distinct processes – a fast process that learns quickly from error but retains less, and a slow process that learns slowly but retains more. Here, we test the hypothesis that adaptation in speech, like in other motor domains, involves separable fast and slow processes by translating a behavioral paradigm known to uncover multiple learning processes across motor domains to speech for the first time (Aim 1). Separately, we will test the neurocomputational basis of adaptation by combining this novel behavioral paradigm with continuous theta-burst transcranial magnetic stimulation to temporarily inhibit regions known to affect adaptation (Aim 2). By assessing behavioral correlates of both fast and slow processes after stimulation, we can separately assess which regions are involved in each process. We test this both in speech and, as a control, in reaching, which is known involved both the fast and slow processes. As the neural networks underlying these separate processes in reaching have not been well established this work additionally provides the first assessment of the neurocomputational basis of adaptation across multiple motor domains, allowing us to test the domain generality of the two-process model of adaptation. Even in the case that we only find a single adaptation process in speech, comparison of the neural findings in speech and those in reaching allows us to test whether the neural substrate underlying speech adaptation maps onto the analogous regions involved for the fast or slow process (or both) in reaching. This research will improve our understanding of the computational processes in speech adaptation and refine the neural mechanisms of sensorimotor adaptation across all motor domains. This work will provide a first step towards developing more sensitive measures that can differentiate control deficits through speech adaptation which may help the development of speech therapy that precisely targets the mechanisms of speech deficits across various clinical populations.