Lesions of the cerebellum severely impair the classically conditioned nictitating membrane response (NMR) in rabbits. Thus, the cerebellum is essential for the production of conditioned responses (CRs), either because it is actively involved in NMR conditioning or because damage to it causes motor or other general deficits. To distinguish between these alternatives, the cerebellum may be inactivated during training. Inactivation of the cerebellum during acquisition training might result in the absence of CRs on initial trials of subsequent training without the neuronal blockade. The blockade may have prevented learning but it may have produced other deficits that require time or further training to overcome. This problem can be addressed by inactivating the cerebellum during extinction training. If inactivation during extinction training results in the immediate production of CRs when training is resumed without the blockade, then it may be concluded that extinction learning was prevented by the blockade-the presence of CRs argues against any deficits not associated with learning. We used muscimol to inactivate the cerebellum and test its involvement in acquisition and extinction of NMR conditioning in the same subjects. We injected muscimol close to the interpositus nucleus of the cerebellum 1 h before each of four daily training sessions of delay conditioning. Almost no CRs were produced in these training sessions-there was little or no acquisition of NMR conditioning during cerebellar inactivation. The subjects were then trained for four daily sessions without injections of muscimol. There were no CRs on initial trials of the first session of retraining, but all subjects produced CRs by the end of this session. The subjects then received four daily sessions of extinction training with muscimol inactivation of the nuclei-no CRs were produced. Extinction training then continued for four daily sessions without muscimol inactivation. On the first of these sessions, all subjects immediately produced high levels of CRs. These responses then extinguished within and between sessions with characteristic beginning-of-session spontaneous recovery. There was little or no extinction of NMR conditioning during cerebellar inactivation. After inactivation, the muscimol- inactivated subjects went on to acquire and extinguish NM responses at rates similar to those of appropriate controls. We conclude that cerebellar circuitry is essential for, and actively engaged in, both acquisition and extinction of this simple form of motor learning.