A technique is described for the measurement of specific [3H]diazepam binding to subcellular fractions of rat brain. Binding occurred in fresh or hypotonically shocked, frozen and thawed preparations; was independent of the buffer used; and was concentrated in the synaptosomal fraction. The total binding capacity of crude brain homogenate was entirely recovered in the particulate fractions (P1, P2 and P3). Binding was saturable and reached 34 pmol/g of tissue; half-maximal binding (KD) occurred at 3.2 nM. Hill and Scatchard analysis indicated that the binding was noncooperative and to a single class of sites. Binding was time dependent and reversible; the bimolecular association constant (K1) was 1.13 . 10(6) sec-1 M-1 and the first order dissociation constant (K-1) was 2.69 . 10(-3) sec-1. Binding was highest in cerebral cortex, hippocampus and cerebellum; intermediate in midbrain, hypothalamus, corpus striatum and medulla oblongata/pons; and lowest in spinal cord. Benzodiazepines inhibited binding of [3H]diazepam in a manner correlated with pharmacological activity in vivo, and binding was not inhibited by non-benzodiazepine anxiolytics, muscle relaxants, anticonvulsants or by strychnine and glycine. Distribution of [3H]diazepam binding in several regions of the rat central nervous system correlated with Na+-independent binding of gamma-aminobutyric acid in the same regions. The results may be in accord with the possible involvement of gamma-aminobutyric acid in the mechanism of action of the benzodiazepines but provide no support for a mechanism based on the interaction of benzodiazepines with central glycine receptors.