UCSF

Motor cortex is known to be required for both motor learning and the execution of certain dexterous motor skills; however, much remains unknown about the communication between primary motor and premotor areas, and how it may or may not change with learning. We addressed this question by making simultaneous electrophysiological recordings of neural populations in rat primary motor (M1) and premotor (M2) cortex, first in the context of natural motor learning, and then in the context of M1 brain-computer interface (BCI) learning. In both cases, we found that activity was coordinated between the two regions. In the case of natural motor learning, we found that learning did not change the strength of M2-M1 interactions, but did increase the amount of task-related information available in the cross-area dynamics, and that these cross-area dynamics were necessary for learned behavioral improvements. In the case of BCI learning, we found that M2-M1 interactions occurred on a broader timescale than did M1-internal interactions, and that M2 population activity evolved at a slower pace. These results are consistent with a model in which M2 provides top-down contextual information to M1, which more precisely controls the output.