How muscle synergies fail to solve the muscle redundancy problem during human reaching

The production of movement involves integrating biomechanical, neural, and environmental factors. The biomechanics is complex enough that neural sensorimotor circuits must embed its dynamics for efficient and robust control. However, a problem of redundancy exists, i.e., the problem of choosing among multiple muscles and combinations of joint angles that are possible for a given desired hand position or motion. This problem may be resolved by reducing the dimensionality of the space of motor commands by the central nervous system, i.e., through muscle synergies or motor primitives. Other studies have obtained muscle synergies using decomposition methods. However, we posit that it is not sufficient to show the existence of a low dimensional space, one needs to demonstrate the utility of the obtained synergies in controlling movement. Here we defined a muscle synergy as a single control signal producing specific force direction. We then tested a hypothesis that such muscle synergies exist using dimensionality reduction method. Our approach takes advantage of the close relationship between the temporal profiles of muscle activity observed with electromyography and the joint moments they produce during reaching derived from motion capture. We recorded electromyography of 12 muscles and the kinematics of both arms in 14 right-handed participants performing reaching movements in multiple directions from different starting positions. We used principal component analysis to evaluate the contribution of individual muscles to supporting the arm against gravity and producing propulsive forces. Results show that the joint torques in specific directions (flexion or extension) required to move toward a target were not produced by consistent muscle groups in most conditions as would be expected from the muscle synergy definition outlined above. We further show that both agonistic and antagonistic muscles coactivate in flexible muscle groups but to a different extent between the dominant and non-dominant arms. Our main findings indicate that the nervous system solves the problem of choosing which muscles to activate and when by taking into account limb dynamics rather than reducing the dimensionality through muscle synergies. Furthermore, our data supports the idea of two neural controllers that target different muscle groups in the arm and hand for gross postural and fine goal-directed control of reaching.