Task errors drive memories that improve sensorimotor adaptation

Traditional views on how humans adapt movements to perturbations of sensory feedback emphasize a fundamental role for automatic, implicit correction of sensory prediction errors. However, it is now clear that adaptive behaviour also involves deliberate, strategic movement corrections. Such strategic processes have recently been argued to underlie the latent retention of sensorimotor adaptation, evident in improved adaptation to previously encountered perturbations; a phenomenon termed “savings”. It remains unclear, however, whether savings results from prior experience of sensory prediction errors, task errors, or both. Here, we used perturbations of target locations and hand position feedback during reaching to dissociate the contributions of task and sensory prediction errors to latent sensorimotor memory. We show that prior learning to correct for task errors is required to improve adaptation to rotated hand position feedback, whereas a history of sensory prediction errors is neither sufficient nor obligatory for savings. A history of correcting for task errors, induced by experimentally perturbing the target location instead of perturbing sensory feedback of movement, improved adaptation to visuomotor perturbations that were never before encountered. Limiting movement preparation time further showed that this learning consists of two distinct components: 1) a strategic component that is flexible enough to facilitate corrective responses in the opposite direction, but that requires substantial preparation time, and 2) a set of inflexible, cached, stimulus-response associations between targets and reach directions, that can be expressed under time-pressure when similar task conditions are experienced. The results emphasise that adaptive responses to sensorimotor perturbations take multiple forms.