To evaluate the pathogenesis of end-of-dose dystonia in levodopa-treated patients with Parkinson's disease, we discontinued a steady-state optimal-dose levodopa infusion either abruptly or slowly. Although dystonic signs appeared sooner after sudden levodopa termination, in both situations dystonia emerged only when circulating drug levels had fallen to the same concentration and parkinsonian scores had declined by the same amount. Dystonia onset thus appears to reflect the degree, rather than the rate, of reduction in dopaminergic stimulation, and may involve the preferential interaction of dopamine with a receptor subpopulation that does not mediate its antiparkinsonian efficacy.