CCN Colloquium: "Visuomotor perturbations in a redundant task evoke diverse learning dynamics and context-dependent learning"

The motor system displays a wide range of learning dynamics, from one-shot learning to months-long skill mastery. Computational models suggest that this diverse functionality stems from combining multiple computations, such as determining the most relevant previously learned motor skill for a particular context and then adapting the skill as needed. Yet most motor learning tasks are designed to study a single learning computation. We developed a novel sensorimotor interface task that produces a rich range of motor learning dynamics. Our task introduces inherent redundancy between the space of possible movements and the task - a feature of most motor skills. In this talk, I'll present data from monkey and human experiments showing that in this redundant task, some perturbations are learned through rapid adaptation while other perturbations evoke slower "de novo" learning. We also found evidence that prior experience influenced learning in future perturbations. I'll discuss how these results generate new hypotheses about the computations that distinguish different forms of motor learning. I'll also discuss how these findings can be used to inform motor therapies like brain-computer interfaces.