[11C]Diprenorphine, a weak partial opiate agonist, and positron emission tomography were used to obtain noninvasive regional estimates of opiate receptor concentration (Bmax) and affinity (Kd) in human brain. Different compartmental models and fitting strategies were compared statistically to establish the most reliable method of parameter estimation. Paired studies were performed in six normal subjects using high (769-5,920 Ci/mmol) and low (27-80 Ci/mmol) specific activity (SA) [11C]diprenorphine. Two subjects were studied a third time using high SA [11C]diprenorphine after a pretreatment with 1-1.5 mg/kg of the opiate antagonist naloxone. After the plasma radioactivity was corrected for metabolites, the brain data were analyzed using a three-compartment model and nonlinear least-squares curve fitting. Linear differential equations were used to describe the high SA (low receptor occupancy) kinetics. The k3/k4 ratio varied from 1.0 +/- 0.2 (occipital cortex) to 8.6 +/- 1.6 (thalamus). Nonlinear differential equations were used to describe the low SA (high receptor occupancy) kinetics and the curve fits provided the konf2 product. The measured free fraction of [11C]diprenorphine in plasma (f1) was 0.30 +/- 0.03, the average K1/k2 ratio from the two naloxone studies was 1.1 +/- 0.2, and the calculated free fraction of [11C]diprenorphine in the brain (f2) was 0.3. Using the paired SA studies, the estimated kinetic parameters, and f2, separate estimates of Bmax and Kd were obtained. Bmax varied from 2.3 +/- 0.5 (occipital cortex) to 20.6 +/- 7.3 (cingulate cortex) nM. The average Kd (eight brain regions) was 0.85 +/- 0.17 nM.