Rabbit and guinea-pig cerebellum membranes contain a very high (greater than 80%) proportion of mu- and kappa-opioid receptors, respectively. Rabbit (mu) and guinea-pig (kappa) cerebellum membranes were (i) labeled either with the opiate agonist, [3H]etorphine (Kd = 0.1-0.2 nM), or with the opiate antagonist, [3H]diprenorphine (Kd = 0.1 nM), in the absence or presence of Na+ and/or 5'-guanylylimidodiphosphate (GppNHp), (ii) solubilized with digitonin (1%, w:v) and (iii) the radioactivity in the soluble extracts analyzed by ultracentrifugation in sucrose gradients. In the soluble extracts from rabbit cerebellum (mu) membranes, bound [3H]etorphine sedimented faster (S20,w congruent to 12S) than bound [3H]diprenorphine (10S), while in those from guinea-pig cerebellum (kappa) membranes, bound [3H]etorphine and bound [3H]diprenorphine sedimented at the same position (12S). Na+ selectively decreased recovery of the bound tritiated agonist in the two soluble preparations. When they had been generated in the presence of GppNHp but in the absence of Na+, the [3H]etorphine complexes of the mu- and kappa-opioid receptors as well as the [3H]diprenorphine complex of the kappa-opioid receptor were all recovered at position 10S, indicating that GppNHp had induced a decrease of the apparent molecular size of the two types of opioid receptors. These data are interpreted in terms of mu- and kappa-opioid receptors being capable of physically interacting with a G protein (GTP binding regulatory protein) yet, unlike the mu-opioid receptor which does so only in the presence of an agonist, the kappa-opioid receptor appears to be precoupled with a G protein.