Therapeutic activity is a mainstay of clinical neurorehabilitation, but is typically unstructured and directed at compensation rather than restoration of central nervous system function. Newer activity-based therapies (ABTs) are in early stages of development and testing. The ABTs attempt to restore function via standardized therapeutic activity based on principles of experimental psychology, exercise physiology, and neuroscience. Three of the best developed ABTs are constraint-induced therapy, robotic therapy directed at the hemiplegic arm, and treadmill training techniques aimed at improving gait in persons with stroke and spinal cord injury. These treatments appear effective in improving arm function and gait, but they have not yet been clearly demonstrated to be more effective than equal amounts of traditional techniques. Resistance training is clearly demonstrated to improve strength in persons with stroke and brain injury, and most studies show that it does not increase hypertonia. Clinical trials of ABTs face several methodological challenges. These challenges include defining dosage, standardizing treatment parameters across subjects and within treatment sessions, and determining what constitutes clinically significant treatment effects. The long-term goal is to develop prescriptive ABT, where specific activities are proven to treat specific motor system disorders. Activity-based therapies are not a cure, but are likely to play an important role in future treatment cocktails for stroke and spinal cord injury.