The rapid intracarotid injection technique was used to determine the unidirectional brain uptake of a number of benzodiazepines in the rat. The drugs varied considerably in their lipophilicity and, within the series oxazepam, lorazepam, chlordiazepoxide, desmethyldiazepam and diazepam, brain extraction of unbound moiety was enhanced as the octanol-water (pH = 7.4) partition coefficient increased. However, with flunitrazepam and midazolam, two fluorine-containing benzodiazepines, extraction was more and less extensive, respectively, than predicted from their lipophilicities. Importantly, the uptake findings were consistent with the characteristic onsets of central effects of the drugs established clinically in humans. The effects of reversible protein binding on uptake also were investigated by the addition of albumin (0-8 g.dl-1) to the injectate. This affected markedly the unbound fraction, determined in vitro by equilibrium dialysis, and also the brain uptake of all drugs. As the unbound fraction was reduced, the unidirectional brain extraction ratio decreased in a curvilinear fashion toward zero. However, attempts to describe the data were unsuccessful using a conventional model based on transcapillary uptake of only unbound drug whose binding kinetics with albumin were assumed to be the same as those indicated by equilibrium dialysis. The observed brain extraction was greater than predicted, and the discrepancy became more apparent as binding and albumin concentration increased. The data for all of the benzodiazepines could be fitted, however, if the equilibrium association constant was assumed to be smaller in vivo than in vitro, so that the effective unbound fraction in the brain capillaries was substantially higher (5- to 25-fold, dependent on the particular drug) than that estimated in vitro.(ABSTRACT TRUNCATED AT 250 WORDS)