Kinetics of L-3,4-dihydroxy-6-[18F]fluorophenylalanine (FDOPA) in striatum and cerebellum were measured in 10 normal human subjects with positron emission tomography (PET) from 0 to 120 min after an intravenous bolus injection of the tracer. The time course of the arterial plasma concentrations of the tracer and its metabolites was also assayed biochemically. FDOPA compartmental models that are based on biochemical information were investigated for their consistency with the measured striatal and cerebellar tissue kinetics. A modeling approach was also developed for separating plasma FDOPA and metabolite time-activity curves from the measured total 18F time-activity curve in plasma. Results showed that a model consisting of three separate compartments for tissue FDOPA, tissue 6-[18F]fluorodopamine (FDA) and its metabolites, and tissue L-3,4-dihydroxy-6-[18F]fluoro-3-O-methylphenylalanine (3-OMFD) could describe adequately the striatal kinetics in humans. Based on this model, the FDOPA transport constant across the blood-brain barrier (BBB) (K1), the FDOPA decarboxylation rate constant (k3), and the turn-over rate constant of FDA and its metabolites (k4) could be estimated by model fitting to the tissue kinetics and were found for the normal subjects to be 0.031 +/- 0.006 ml/min/g (mean +/- SD), 0.041 +/- 0.015/min, and 0.004 +/- 0.002/min, respectively. About 50% of the FDOPA that crossed the BBB from plasma to striatum was decarboxylated. The decarboxylation constant with respect to plasma FDOPA (K3) was 0.015 +/- 0.003 ml/min/g. The BBB transport corresponded to a permeability-surface area product of 0.032 ml/min/g for FDOPA. For 3-OMFD, the BBB transport was 1.7 times faster. The effects of tissue heterogeneity on the FDOPA kinetics and on the estimated model parameters were also investigated. The usefulness and implications of these findings for interpretation of PET FDOPA studies are discussed.