Understanding the Role of Internal Predictions in Sensorimotor Adaptation

Abstract

Our ability to perform intricate tasks with remarkable precision is made possible through a highly complex neuromotor system. In order to maintain this fine control of movement, the motor system needs to continually adapt to internal physiological changes such as bodily growth, injury or fatigue as well as novel environmental dynamics and tools such as swimming or riding a bicycle. Such adaptation is postulated to be grossly driven by errors arising during motor actions, however, the specific mechanisms by which the teaching signal for adaptation is constructed remain unclear. A widely-hypothesized idea is that the brain predicts the sensory consequences of an action and uses those predictions to affect subsequent actions. In this dissertation, we focus on determining the role of internal predictions in constructing the teaching signal for motor adaptation. First, we examine the ability of the brain to predict the effects of motor output noise. This motor output noise is what prevents us from repeating actions with exact precision. Since errors arising due to motor output noise are random on each trial, any adaptation to them would be futile. However, a major challenge in compensating specifically for these errors is that the motor error reflects contributions from internal as well as external sources that are loaded as a single, combined input in the sensory stream. Remarkably we show that errors arising due to motor output noise are predictively canceled from the teaching signal for motor learning. Furthermore, we demonstrate that this teaching signal is constructed by subtracting these predictions from motor performance errors rather than sensory feedback. These results indicate that motor learning is driven by the mismatches between task performance and an internal prediction of it, and is uncontaminated by the effects of motor output noise Given that the internal predictions of actions made by the motor system include the effects of motor output noise, we next examined the plasticity of these predictions. If the noise in actions were to change systematically, the internal predictions of that noise should change accordingly for perfect decontamination of motor learning from output noise. Here, we show that this is indeed the case, and internal predictions adapt completely to systematic changes in output noise. Moreover, we find that this adaptation is more than halfway complete within a single exposure to the change. Together, our results demonstrate that the internal predictions made by the motor system include effects of motor output noise, are highly plastic and get subtracted from motor performance errors to create the teaching signal for motor learning.