Asymptotic limits of sensorimotor adaptation

In many forms of motor adaptation, performance approaches a limit at which point learning stops, despite the fact that errors remain. What causes this adaptation limit? Here we found that while reach adaptation exhibited an asymptotic limit, this limit was not fixed: when the variance of the perturbation decreased, the adaptation limit increased, and performance improved. Moreover, the limit could be altered in real-time by changing perturbation variance. The same was true at low reaction times, indicating that implicit processes influenced the adaptation limit. Using a mathematical model that describes adaptation as a balance between learning and forgetting, we found that participants altered their adaptation limit by changing their sensitivity to reaching errors. Changes in error sensitivity were linked to the consistency of past errors. These observations suggest that during adaptation, time-varying error sensitivities compete with constant forgetting, setting the boundary conditions that produce an apparent limit in total adaptation.