Recently, our laboratory proposed that a single model could account for a large body of behavioral data in numerous motor adaptation paradigms. The idea was that motor memory is supported by at least two interacting processes: a fast process that learns quickly from motor error but rapidly forgets, and a slow process that only gradually learns from error but has long-term retention. The main purpose of this research is to uncover the time and error-dependent properties of these putative processes and to ask whether there is a link between these processes and the functions of the cerebellum and primary motor cortex.
How does passage of time affect retention of motor memories? The typical approach is to learn a task, and then look at retention as a function of time. However, if motor memories are supported by multiple processes, then a better way to reveal their timescales is to adapt, reverse adapt, and then quantify retention. The prediction of our theory is that there should be spontaneous recovery of the initial adaptation with passage of time. Because theory predicts that reverse adaptation will primarily engage the fast system, this experimental design allowed us to quantify how retention changed as the fast and slow processes decayed with passage of time. The different rates of decay in these putative processes resulted in a time-dependent pattern of spontaneous recovery, as well as a time dependent stabilization of the fast memory process.
Previous work by Huang and Shadmehr demonstrated that the statistics of the environment during adaptation altered the time constant of the putative processes that support memory (Huang and Shadmehr 2009). When the perturbation was presented abruptly, the memory decayed quickly, suggesting engagement of the fast process. When the perturbation was presented gradually, the memory was decayed slowly, suggesting engagement of the slow process. We hypothesized that the role of the cerebellum is to respond to large errors, thus supporting the fast process of motor memory. To test for this, we trained cerebellar degeneration patients in both the abrupt (large errors) and gradual (small errors) conditions. Severely affected patients showed improved adaptation in the gradual condition and upon sudden removal of the perturbation the motor memory that was acquired showed a strong resistance to change, exhibiting after-effects that persisted much longer than in healthy controls. Therefore, cerebellar degeneration impairs the ability to learn from large magnitude errors, but has a lesser impact on learning from small errors.
Finally, we hypothesized that the role of the primary motor cortex is to support the late phase of adaptation, during which error is small, motor output has reached a plateau, and the slow process dominates net adaptation. We employed three behavioral conditions, abrupt, gradual and uber to vary the size of the error, the number of trials where the perturbation was at steady state, and the phase of learning. A single pulse TMS paradigm was applied to the primary motor cortex. Disruption of the primary motor cortex caused an impairment in performance when the errors were small and the environment at steady state, independent of the phase of the learning. We conclude that the primary motor cortex contributes to motor learning when the training environment and motor output have reached steady state for an extended number of trials.