Generalization indicates asymmetric and interactive control networks for multi-finger dexterous movements

Finger dexterity is manifested by coordinated patterns of muscle activity and generalization of learning across contexts. Some fingers flex, others extend, and some are immobile. Whether or not the neural control processes of these direction-specific actions are independent remains unclear. We characterized behavioral principles underlying learning and generalization of dexterous flexion and extension movements, within and across hands, using an isometric dexterity task that precisely measured finger individuation, force accuracy, and temporal synchronization. Two cohorts of participants trained for 3 days in either the flexion or extension direction. All dexterity measures in both groups showed post-training improvement, although finger extension exhibited inferior dexterity. Surprisingly, learning of finger extension generalized to the untrained flexion direction, but not vice versa. This flexion bias was also evident in the untrained hand. Our study indicates direction-specific control circuits for learning of finger flexion and extension that interact by partially, but asymmetrically, transferring between directions.