Investigating how the healthy human motor system acquires and retains motor memories can yield insight into how we might improve rehabilitation for patients recovering from a wide array of movement disabilities. The research projects presented in this document are directed towards the goal of understanding motor skill learning and retention. I use reaching arm movements in healthy human volunteers as a model system, and in particular, I study how the human motor system adapts to external force perturbations during these voluntary movements. Careful analysis of the adaptive responses to these perturbations reveals several key insights into the adaptive mechanisms employed by the motor system. One feature is the presence of multiple timescales in the learning process. Recent work has shown that one adaptive process learns quickly but forgets quickly, whereas another learns slowly but has excellent retention. I explore how interactions between these two processes can explain two key learning phenomena—anterograde interference and the spacing effect. Anterograde interference is the ability of a previously-learned motor task to reduce the rate of subsequently learning a different motor task, and the spacing effect is the observation that increased spacing between learning trials leads to increased long-term retention. Understanding the mechanisms behind these phenomena, which have been described in numerous neural learning systems, may allow us to develop improved training procedures that mitigate unwanted interference and maximize long-term retention.

Another concept that I explore in this document is the characterization of the neural learning elements underlying the adaptation process. By studying the adaptive responses to different types of learning stimuli, we can gain insight into these so-called "motor primitives." I find that these primitives are motion-dependent, and that the nature of this motion-dependence predicts a host of experimental observations about motor adaptation, including why early learning tends to be rapid and stereotyped but late learning is slower and more targeted, and why adaptation to non-motion-dependent force perturbations are surprisingly motion-dependent. Leveraging this knowledge, I design dynamic environments that are intrinsically the easiest and most difficult to learn, suggesting a theoretical basis for the rational design of improved procedures for rehabilitation.