In the present study, the occupancy of flumazenil (Ro 15-1788; 1-30mg/kg p.o.) at the benzodiazepine site of rat brain GABA(A) receptors was compared using in vivo and ex vivo binding methodologies with [(3)H]flumazenil as the radioligand. Animals either received tracer quantities of [(3)H]flumazenil 3min before being killed for the in vivo binding, or were killed and brain homogenates incubated with 1.8nM [(3)H]flumazenil. The flumazenil dose required to inhibit in vivo binding of [(3)H]flumazenil by 50% (ID(50)) was 2.0mg/kg, which represents the most accurate measure of benzodiazepine site occupancy by flumazenil in vivo. Occupancy measured in crude brain homogenates using the ex vivo method was time dependent with a 3mg/kg dose giving occupancies of 77% and 12% using 0.5 or 60min ex vivo incubations times, respectively, presumably due to dissociation from the binding site during the ex vivo incubation. When incubation time was minimised (0.5min), and despite being under non-equilibrium conditions, the ex vivo method gave an ID(50) of 1.5mg/kg which was not too dissimilar from that observed using in vivo binding (2.0mg/kg). As expected, ex vivo binding can give an underestimation of receptor occupancy but this can be minimised by careful attention to the kinetics of unlabelled drug and radioligand.