Eleven patients with severe Parkinson's disease and on-off-phenomena were included in a controlled double-blind study on the effect of electroconvulsive therapy (ECT). Pharmacological treatment was optimally adjusted before the trial. The severity of extrapyramidal symptoms was measured before, during and after the treatment. The patients were randomly allocated into one group, receiving active ECT and another, receiving sham treatment. The patients given active ECT showed significantly (P less than 0.05) prolonged duration of "on"-phases after ECT, in comparison to the sham-treated group. When collecting data from the controlled part of the study and the subsequent with open administration of ECT, the treatment was in addition found to significantly decrease the time and number of steps required to walk 10 meters. Moreover it reduced the severity of parkinsonian symptoms according to the Webster scale. The improvement induced by ECT was generally short-lasting. Lumbar punctures were performed before and after ECT. The concentrations of monoamine metabolites in cerebrospinal fluid were not affected by the treatment. The results indicate that ECT has an antiparkinsonian effect which probably is mediated via changed responsiveness of dopamine receptors and that further improvement is possible in patients, therapy resistant to the presently available medication.