1. We have stimulated over the cerebellum of intact human subjects by applying single electrical stimuli through electrodes placed on the back of the head, approximately at the level of the inion. The intensity of stimulation used was below that required to produce direct EMG responses in pre-activated muscles of the hand. 2. In ten subjects the effect of the stimulus over the cerebellum was to reduce the size of the EMG response in first dorsal interosseous muscle evoked by a magnetic stimulus to the cerebral cortex. In all subjects the onset of the period of suppression occurred when the test magnetic cortical shock followed the conditioning cerebellar shock by 5 ms. The duration of the suppression lasted from 3 to 7 ms. 3. The amount of suppression was related to the intensity of stimulation over the cerebellum. At 15% below the threshold for direct motor activation there was no effect; increasing suppression was evident at 10, 5 and 0% below motor threshold. 4. With a conditioning-test interval of 5-6 ms the suppression was the same whether the target muscle was relaxed or active. With longer conditioning-test intervals (12 and 15 ms) the amount of suppression was greater in active than relaxed muscles. 5. The short-latency suppression was greatest when the stimulating anode was ipsilateral to the target muscle and contralateral to the stimulated sensorimotor cortex. The later period of suppression was insensitive to the polarity of stimulation. When the stimulating electrodes were moved 2 cm caudally or cranially the short latency suppression disappeared whereas the longer latency suppression was still observed with the electrodes in the lower position. 6. Different results were obtained when the test EMG response was produced by an electrical (rather than magnetic) stimulus over the sensorimotor cortex. The short latency effect was no longer visible whereas the longer latency effect was the same as when testing with a magnetic cortical stimulus. 7. We suggest that a single electrical stimulus across the base of the skull (particularly with the anode over one cerebellar hemisphere) produces a short latency (5-6 ms) disfacilitation of the contralateral motor cortex through activation of cerebellar structures. A later (12 and 15 ms), less specific suppression which is present when testing in active muscles is thought to be mediated by a different mechanism and probably produces its effect at the level of the spinal cord.