1. Inactivation of the portions of globus pallidus pars interna (GPi) containing the greatest concentration of wrist-related neurons was achieved in two rhesus monkeys with microinjections of muscimol (temporary) and kainic acid (permanent). 2. After muscimol injection, there was onset within 30 s of 1) tonic and phasic coactivation of wrist flexors and extensors; 2) slightly greater activation of the flexors, giving a flexor bias in postural holds and the endpoint of movements; and 3) slowness of all movements with a prolonged movement time. Nevertheless, 4) movements made by lessening prior loaded muscle activity (to move in the direction of the load) were slower than movement made by increasing muscle activity (to move against the direction of the load). Despite marked slowing of all movements, there was 5) a normal reaction time for movement onset. Finally, there was 6) a reduced amplitude of most movements. Open room behavior included 7) spiraling contralateral to the lesion while walking. Effects were reproducible (12 injections), were apparent for 7-8 h and were usually completely gone by the next day's testing. 3. After kainic acid injection, there was a period of mixed effects, followed by a period of permanent defects (observed for up to 24 days) that duplicated the temporary effects of muscimol. 4. By contrast, muscimol inactivation of the cerebellar dentate nucleus resulted in 1) a prolonged reaction time and 2) an increased variability of movement trajectory, but 3) without change in movement time or peak velocity. Open room behavior included overshoot in reaching for fruit with the forelimb ipsilateral to the injection. 5. From the facts that normal pallidal neurons fire constantly, that pallidal neurons inhibit their target neurons, and that the muscimol effect was immediate, we conclude that the release of the target neurons from the tonic inhibition allowed them to fire in patterns that promoted a maintained state of cocontraction of agonist and antagonist muscles. From the fact that movement time was prolonged, we conclude that the maintained state of neural activity that caused the muscle cocontraction interfered with the commands for voluntary movement, which were generated by other mechanisms. From the fact that reaction time for movement onset was normal, we conclude that the pallidal neurons may play little or no role in the voluntary initiation of these movements, which are instead generated by other structures that include the anterior cerebral cortex and the lateral cerebellum.