Consistent EMG encoding during reach and grasp by neurons in motor cortex

Abstract

Identifying what the activity of neurons in the primary motor cortex (M1) represents is essential for understanding motor computation. Yet the relation of M1 to muscles and movement remains unresolved, in part because the kinematic variables M1 appears to encode differ across the limb: hand velocity during reaching, but joint position during grasping. This discrepancy has been taken to imply fundamentally distinct control strategies for proximal and distal segments. Yet, here we show that a single, muscle-based control principle accounts well for both observations. We recorded neural activity, electromyographic (EMG) signals, and movement kinematics from macaque monkeys performing planar reaching, wrist movement, or free-form grasping tasks. Across behaviors, EMG-based encoding models explained M1 firing rates as accurately as the best-performing kinematic model: velocity for reaching, position for grasping, and either variable for wrist movements. Impulse response analysis revealed that these task-dependent kinematic relationships arise from the differences in mechanical impedance of each limb segment: the inertial and intersegmental coupling which dominate the dynamics of the proximal arm are far less important in the hand. These findings indicate that M1 output has a relatively simple, consistent relation to muscle activation and that the apparent divergence in kinematic encoding is a consequence of limb mechanics rather than distinct cortical strategies.