There are varying degrees of spontaneous improvement in arm paresis over the first 6 months after stroke. The degree of improvement at 6 months is best predicted by the motor deficit at 1 month despite standard rehabilitative interventions in the ensuing 5 months. Animal studies indicate that the loss of fine motor control, especially individuation of the digits, is due to interruption of monosynaptic corticomotoneuronal connections. Spasticity occurs because of loss of cortical modulatory control on descending brain stem pathways and spinal segmental circuits but is not a major cause of motor dysfunction. Quantitative studies of reaching movements in patients suggest that arm paresis consists of higher-order motor planning and sensorimotor integration deficits that cannot be attributed to weakness or presence of synergies. Cortical stimulation experiments in animals and functional imaging studies in humans indicate that motor learning and recovery after stroke share common brain reorganization mechanisms. Rehabilitation techniques enhance learning-related changes after stroke and contribute to recovery. Future research will benefit from using quantitative methods to characterize the motor impairment after stroke and by applying concepts in motor learning to devise more physiologically based rehabilitation techniques.