Motor learning in multijoint virtual arm movements with novel kinematics

Abstract Humans move their hands towards precise positions, a skill supported by the coordination of multiple joint movements, even in the presence of inherent redundancy. However, it remains unclear how the central nervous system learns the relationship between redundant joint movements and hand positions when starting from scratch. To address this question, a virtual-arm reaching task was performed in which participants were required to move a cursor corresponding to the hand of a virtual arm to a target. The joint angles of the virtual arm were determined by the heights of the participants’ fingers. The results demonstrated that the participants moved the cursor to the target straighter and faster in the late phase than they did in the initial phase of learning. This improvement was accompanied by a reduction in the amount of angular changes in the virtual limb joint, predominantly characterized by an increased reliance on the shoulder joint as opposed to the wrist joint. Moreover, increased shoulder joint use relative to that of the other joints was positively correlated with the number of successful target acquisitions. These findings suggest that the central nervous system selects a combination of multijoint movements that minimize motor effort while learning novel upper-limb kinematics.