Principal component analysis of M1 neurophysiology data suggests a motor-control system-architecture template.

Stereotyped reaching tasks are used to study how primate subjects learn and recall motor skills required to compensate for different external forces during arm movements. To unveil mechanisms accounting for skilled performance under a wide range of rapidly switching task dynamics conditions, we recorded neural data from the primary motor-cortex (M1). Here we present a systematic analysis of changes in the M1 activity of a monkey with extensive practice compensating for five different dynamic fields in an elbow flexion/extension task. We show how they reflect differences in task kinematics and dynamics. Making extensive use of principal component analysis (PCA) and in preparation for computational modeling (see the companion paper) we demonstrate how M1 activity can be related functionally to the dynamics of feed-forward (FF), fast- and slow- feedback (FB) loops of the adaptive controller implemented by the brain to guide skilled motor behavior.