The duration of the antiparkinsonian action of levodopa was studied in 48 patients with various response patterns to the oral administration of the dopamine precursor. Deterioration in motor scores after abrupt cessation of a steady-state intravenous levodopa infusion occurred at two successive rates: an initial rapid phase followed by a terminal slower phase. Efficacy half-time decreased and initial efficacy decay slope increased with progression of levodopa response groups from never treated to stable responders, and then to fluctuating responders of the wearing-off type and finally of the on-off type. Efficacy half-time exceeded plasma levodopa half-life in the 2 nonfluctuating groups, approximated it in those patients with wearing-off responses, and was significantly shorter in patients with fluctuations of the on-off type. The half-times for the decline in antiparkinsonian efficacy and dyskinesia severity differed significantly, suggesting different pharmacological mechanisms. Motor fluctuation severity correlated best with initial efficacy decay slope, and both were best predicted by parkinsonian symptom severity. The dyskinesia decay rate correlated most closely with levodopa dose. These results support the view that progressive dopamine neuron degeneration reduces the brain's ability to buffer shifts in levodopa availability attending its periodic oral administration; the clinical result is wearing-off phenomenon. The on-off phenomenon as well as dyskinesia apparently reflects additional secondary changes related to levodopa therapy and occurring postsynaptically.