Somatosensory information drives modification of a motor memory

Abstract The role of somatosensory information in motor memory modification remains controversial. One view holds that somatosensory information plays only a supportive role, such as feedback or sensory calibration, and is not directly involved in memory modification itself. An alternative view is that afferent somatosensory information, on its own, can modify motor memory. Here we resolve this controversy by isolating somatosensory information from voluntary motor execution using a hand-exoskeleton robot. Participants first learned a sequential finger movement task through voluntary practice. Their learned finger movements were recorded using a data glove. On subsequent days, they underwent a brief, 30-sec reinstatement session: either exoskeleton-based reinstatement, which reproduced their learned movements without voluntary execution, or voluntary reinstatement, in which they executed the trained task. Exoskeleton-based reinstatement led to subsequent performance gains comparable to those induced by voluntary reinstatement. This gain was not explained by replay of visual information, mental rehearsal, enhanced arousal, or warm-up effects induced by exoskeleton-driven movements. Rather, the results indicate that brief reinstatement of somatosensory information alone is sufficient to drive performance gain. We further found that exoskeleton-based exposure to a sequence that only partially matched the trained sequence yielded performance gains, whereas voluntary execution of the same partially matched sequence did not. This dissociation suggests that somatosensory information broadly activates a relevant motor skill memory, whereas voluntary execution activates memory in a sequence-specific manner. Together, these findings resolve a long-standing controversy in motor learning research by showing that afferent somatosensory information alone enables sequential motor memory modification through a route distinct from voluntary execution.