Carbon-11-labeled flumazenil combined with positron emission tomography (PET) was used to measure the concentration (Bmax) of the benzodiazepine (Bz) receptor in the brain and its equilibrium dissociation constant (KD) for flumazenil in five normal subjects. The steady-state approach was used injecting the tracer as a bolus of high specific activity. In each subject two studies were carried out. The first study was performed at essentially zero receptor occupancy, the tracer alone study. The second study was performed at a steady-state receptor occupancy of about 50%, achieved by a prolonged constant infusion of nonlabeled ("cold") flumazenil starting 2h before the bolus tracer injection and continuing until the end of scanning period. In this second study the free concentration of unmetabolized flumazenil in plasma water was measured in multiple blood samples. The observed tissue and plasma tracer curves, calibrated in the same units of radioactivity per millimeter, were analyzed in two ways: (a) by the noncompartmental (stochastic) approach making no assumptions regarding number of compartments in the tissue, and (b) by the single-compartment approach assuming rapid exchange (mixing) of tracer between all tissue compartments. The noncompartmental and the compartmental analyses gave essentially the same values for the distribution volume of the tracer, the parameter used for quantitation of the Bz receptor. As the compartmental approach could be applied to a shorter observation period (60 min instead of 120 min) it was preferred. The five subjects had a mean KD value of 12 nM/L of water and Bmax values of the grey matter ranging from 39 +/- 11 in thalamus to 120 +/- 14 nM/L of brain in occipital cortex. Most previous studies have been based on the pseudoequilibrium approach using the brain stem as a receptor-free reference region. This yields practically the same KD but lower Bmax values than the steady-state approach presented here.