Kinetic methods were used to obtain regional estimates of benzodiazepine receptor concentration (Bmax) and equilibrium dissociation constant (Kd) from high and low specific activity (SA) [11C]flumazenil ([11C] Ro 15-1788) positron emission tomography studies of five normal volunteers. The high and low SA data were simultaneously fit to linear and nonlinear three-compartment models, respectively. An additional inhibition study (pretreatment with 0.15 mg/kg of flumazenil) was performed on one of the volunteers, which resulted in an average gray matter K1/k2 estimate of 0.68 +/- 0.08 ml/ml (linear three-compartment model, nine brain regions). The free fraction of flumazenil in plasma (f1) was determined for each study (high SA f1: 0.50 +/- 0.03; low SA f1: 0.48 +/- 0.05). The free fraction in brain (f2) was calculated using the inhibition K1/k2 ratio and each volunteer's mean f1 value (f2 across volunteers = 0.72 +/- 0.03 ml/ml). Three methods (Methods I-III) were examined. Method I determined five kinetic parameters simultaneously [K1, k2, k3 (= konf2Bmax), k4, and konf2/SA] with no priori constraints. An average kon value of 0.030 +/- 0.003 nM-1 min-1 was estimated for receptor-rich regions using Method I. In Methods II and III, the konf2/SA parameter was specifically constrained using the Method I value of kon and the volunteer's values of f2 and low SA (Ci/mumol). Four parameters were determined simultaneously using Method II. In Method III, K1/k2 was fixed to the inhibition value and only three parameters were estimated. Method I provided the most variable results and convergence problems for regions with low receptor binding. Method II provided results that were less variable but very similar to the Method I results, without convergence problems. However, the K1/k2 ratios obtained by Method II ranged from 1.07 in the occipital cortex to 0.61 in the thalamus. Fixing the K1/k2 ratio in Method III provided a method that was physiologically consistent with the fixed value of f2 and resulted in parameters with considerably lower variability. The average Bmax values obtained using Method III were 100 +/- 25 nM in the occipital cortex, 64 +/- 18 nM in the cerebellum, and 38 +/- 5.5 nM in the thalamus; the average Kd was 8.9 +/- 1.0 nM (five brain regions).