We studied 20 patients with ataxia caused by various disorders using magnetic stimulation over the cerebellum. Results were compared with normal values found for 12 normal volunteers. In normal subjects, a magnetic stimulus over the cerebellum reduced the size of responses evoked by magnetic cortical stimulation when it preceded cortical stimulus by 5, 6 and 7 ms. The grand average of the ratios of the areas of conditioned responses at intervals of 5, 6 and 7 ms to those of control responses was designated the average area ratio (5-7 ms). Suppression of motor cortical excitability was reduced or absent in patients with a lesion in the cerebellum or cerebellothalamocortical pathway, but was normal in patients with a lesion in the afferent pathway to the cerebellum. Normal suppression was observed in Fisher's syndrome. The average area ratio (5-7 ms) correlated well with the severity of ataxia in patients with degenerative late-onset ataxia. These results are consistent with those for electrical stimulation of the cerebellum reported previously. We conclude that magnetic stimulation over the cerebellum produces the same effect as electrical stimulation even in ataxic patients. This less painful method can be used clinically to clarify the pathomechanisms for ataxia. Two other clinical uses of this technique were that it revealed clinically undetectable cerebellar dysfunction in patients whose extrapyramidal signs masked cerebellar signs, and that the slow progression of ataxia could be followed quantitatively in patients with degenerative late-onset ataxia.