Project Summary Motor cortex historically has been viewed as a relatively stable region of the brain, with static cortical dynamics mapping to particular movements. Recent work has questioned these assumptions, asserting that the brain’s internal map of the external world drifts over time. This phenomena, known as representational drift, may provide insight into how the brain adapts to learn new behavior. Modeling representational drift and establishing its connection to motor learning are the primary aims of this proposal. Currently, there is no standardized metric to track representational drift over time. Leveraging our repository of motor and premotor neuronal recordings, we look to establish a common framework for calculating the rate at which this drift occurs. To do this, we will apply unbiased, probabilistic methods (e.g., continuous-time hidden Markov models, switching linear dynamical systems) that extract discrete latent states to the neural data. Models will be evaluated based on offline decoding performance, and drift will be calculated by the transition rate across latent states. After establishing a common modeling framework, we will analyze how drift rates change across various motor tasks and levels of exertion. To accomplish this, we will employ our novel, unconstrained recording platform, which allows for the simultaneous capture of markerless kinematics and wirelessly transmitted neural data. By analyzing the neural data collected during diverse movements, we can assess how both overall levels activity and different types of behavior impact rates of representational drift. By determining how physical activity impacts representational drift, we can better understand the relationship between drift and rates of motor learning. To do this we will study to cohorts recovering from motor deficit. One cohort will integrate drift- enhancing activities during recovery and another will not. By monitoring rates of representational drift and motor recovery, we can evaluate the effect that drift has on motor learning. This knowledge can then be leveraged to improve treatments for those recovering from brain-related movement disorders, like stroke and Parkinson’s.