The short-duration response, long-duration response, and dyskinetic response to levodopa change during long-term levodopa therapy. How these responses evolve, and which changes contribute to the emergence of motor fluctuations, remain unclear. We studied 18 subjects with Parkinson's disease before they began levodopa therapy and after 6, 12, 24, and 48 months of long-term levodopa therapy. The responses to 2-hour levodopa infusions after overnight and after 3 days of levodopa withdrawal were studied from 6 months onward. The mean magnitude of the short-duration response and the long-duration response measured after overnight without antiparkinsonian medications did not change during the 4 years. However, after 3 days without levodopa, it was apparent that the short-duration-response magnitude was progressively increasing (p < 0.0001) and that the long-duration response was decaying more rapidly (p = 0.0004). The short-duration-response magnitude at 4 years was inversely related to the long-duration-response magnitude (p = 0.022), suggesting that the long-duration response was one determinant of the short-duration-response magnitude. Dyskinesia increased progressively in severity during the study (p = 0.013). The duration of the short-duration response and dyskinesia did not change during the 4 years. Subject reports of motor fluctuations tended to be associated with a large short-duration response (p = 0.054). We suggest that a larger long-duration response, rather than a shortened one, is more important to the development of fluctuations. Improving the baseline or practical-off motor function to reduce the magnitude of the short-duration response may be a strategy to treat fluctuations.