To date, the study of the relationship between drug occupancy and action in the brain has had to rely on the use of either animal models or of indirect kinetic measures in man, e.g. serum concentrations of unbound drug (as a measure of "free" drug in brain). We describe the first set of experiments which directly measure agonist-induced changes in both pharmacodynamic effects and pharmacokinetic parameters simultaneously and which demonstrate the feasibility of these studies in man. Five healthy volunteers each had two PET scans using [11C]flumazenil (a radiolabelled benzodiazepine site antagonist) as part of a study investigating kinetic models and the relationship between occupancy and effect of benzodiazepine site ligands. In both studies the [11C]flumazenil was displaced from the brain by infusion of midazolam administered i.v. 30 min into the scan. In one study a higher dose of midazolam was administered than in the other (range 12.5-50 micrograms/kg). Time-activity curves of the concentration of radioligand were derived in 17 different brain regions using a stereotactic automatic method of region selection. We demonstrated that there are significant differences in an index of occupancy, induced by the two different doses of midazolam, both across brain regions and within subjects. There was a significant correlation between measured occupancy index change and pharmacodynamic effects as measured by the peak change in beta 1 spectral power on EEG. There was no significant correlation between dose administered and EEG changes; plasma concentrations of midazolam were correlated with the occupancy index and with the EEG measures. In addition, we have demonstrated that a non-regional total index of brain occupancy can be obtained by analysing the non-tomographic data obtained with the PET scanner (total radioactivity counts head curve) and that this index shows significant correlations both with the dose administered and with the pharmacodynamic measure. This last finding validates the use of other non-tomographic counting techniques (Malizia et al., 1995a) where an index of displacement can be obtained after the administration of less than 1% of the dose of radiation needed for a PET study. These studies are likely to be useful in human psychopharmacology, in particular in the assessment of tolerance and of putative changes in benzodiazepine sensitivity in anxiety disorders. The same principles can be applied to other ligand studies and will be useful to validate current PK/PD models.