Studies of central benzodiazepine receptors in the human brain in vivo are now possible using positron emission tomography (PET) and [11C]flumazenil. With the aim of measuring Bmax and Kd in brain regions, we used a two-injection [11C]flumazenil (at high and low specific radioactivity, respectively) pseudo-equilibrium paradigm to evaluate, in seven unmedicated healthy volunteers, the relative merits of three 'reference' structures (pons, hemispheric white matter and corpus callosum) in which the free radioligand concentration in brain tissue was estimated 15-40 min after i.v. injection of the radioligand. By means of high-resolution PET, the Bmax and Kd were calculated for each subject in 18 gray matter structures, based on a two-point Scatchard plot. We found that the use of the corpus callosum as reference often resulted in spurious Bmax and Kd values. The pons was the best reference structure because it provided satisfactory Bmax values (closest to in vitro data) and most consistent Kd values, and was the region easiest to sample on PET images. The pattern of regional Bmax was consistent with that expected from in vitro studies, with values highest in the cerebral cortex, intermediate in the cerebellum, and lowest in the striatum and the thalamus. The Kd values were uniform among regions and were consistent with earlier in vitro and in vivo data. This work documents the feasibility of estimating Bmax and Kd of central benzodiazepine receptors in multiple brain regions for clinical research.