To determine how the response to levodopa is altered by long-term therapy, we examined the dose response to 2-hour infusions of levodopa in three groups of parkinsonian patients: those who were previously untreated, those who exhibited stable responses, and those who exhibited fluctuating responses to levodopa therapy, using tapping speed as an index of bradykinesia. The baseline tapping speed was greater in the patients with stable responses than in the untreated patients, probably representing a "long-duration response" to levodopa therapy. A "short-duration response," indicated by an increase in tapping speed lasting hours, was observed in most patients in all groups. The onset of the short-duration effect was more rapid and the incremental increase in tapping speed was twice as large in the patients with fluctuating responses compared with the untreated patients and patients with stable responses. The duration of the short-duration effect was greatest in the untreated group but did not differ between the groups with stable and fluctuating responses. Dyskinesia was not observed in any of the de novo patients but was observed in three of 12 patients with stable responses and eight of nine patients with fluctuating responses to levodopa therapy. Dyskinesia appeared before or with the antiparkinsonian effects in patients with stable responses, giving no indication of a higher threshold for dyskinesia in these patients compared with those with fluctuating responses. The plasma half-life clearance, volume of distribution, and maximum plasma concentrations of levodopa did not differ among groups.(ABSTRACT TRUNCATED AT 250 WORDS)