Task relevance selectively modulates sensorimotor adaptation in the presence of multiple prediction errors.

Adaptation to consistently occurring sensorimotor errors is considered obligatory in nature. We probed the robustness of this finding by asking if humans can selectively attenuate adaptation based on the task-relevance of error signals. Subjects made planar reaches to three different targets: an arc (experiment 1), a bar (experiment 2), and a point (experiment 3). During the reach, perturbations in extent (visuomotor gain), direction (visuomotor rotation) or both simultaneously were employed. In experiment 1, subjects showed robust adaptation to the rotation when reaching to the arc even though the presence of this perturbation was irrelevant for achievement of the task goal. Interestingly however, rotation adaptation was strongly attenuated when it was presented simultaneously with a task-relevant gain perturbation. In experiment 2, which involved reaches to the bar, again, subjects successfully adapted to the task-irrelevant gain perturbation when it occurred in isolation. However, adaptation was attenuated when the gain co-occurred with a task-relevant rotation. Experiment 3 revealed that the attenuation observed in the first two experiments was not due to an inability to adapt to co-occurring rotation and gain perturbations. Collectively, our results suggest that the sensorimotor system selectively tunes learning in the presence of multiple error signals, a finding that can potentially be explained by a biased competition mechanism. That is, given limited processing capacity, a salient attribute - the relevance of the error to the task goal in this case - is prioritized for processing and drives subsequent adaptive changes in motor output.