Perturbation variance suppresses error sensitivity in the implicit learning system

Abstract

When variability is added to a sensorimotor perturbation, total adaptation is impaired. In Albert et al.1 we explored this phenomenon, and observed that it is the brains implicit, i.e., subconscious learning system that is most affected by perturbation variance. We observed that perturbation variability impaired implicit learning by downregulating its sensitivity to error. Recently, Wang et al.2 present an alternate viewpoint: implicit error sensitivity does not change with experience, only the errors observed by the implicit system change. Here we evaluated this alternate view by empirically measuring error sensitivity as a function of error size. We found that perturbation variability strongly downregulates implicit error sensitivity when controlling for error size, consistent with our original results, counter to the inflexible model argued by Wang et al. With that said, a pre-existing relationship between error sensitivity and error magnitude noted by Wang et al. can contribute at least in part to implicit behavior. State-space models that start with this pre-existing error sensitivity curve and then update it with training according to a memory of errors most accurately tracked measured behavior.