Muscarinic and other G protein-coupled receptors exhibit an agonist-specific heterogeneity that tracks efficacy and commonly is attributed to an effect of the G protein on an otherwise homogeneous population of sites. To examine this notion, M2 muscarinic receptors were purified from Sf9 cells as monomers devoid of G protein and reconstituted as tetramers in phospholipid vesicles. In assays with N-[(3)H]methylscopolamine, seven agonists revealed a dispersion of affinities indicative of two or more classes of sites. Unlabeled N-methylscopolamine and the antagonist quinuclidinylbenzilate recognized one class of sites; atropine recognized two classes with a preference that was the opposite of that of agonists, as indicated by the effects of N-ethylmaleimide. The data were inconsistent with an explicit model of constitutive asymmetry within a tetramer, and the fit improved markedly upon the introduction of cooperative interactions (P < 0.00001). Purified monomers appeared to be homogeneous or nearly so to all ligands except the partial agonists pilocarpine and McN-A-343, where heterogeneity emerged from intramolecular cooperativity between the orthosteric site and an allosteric site. The breadth of each dispersion was quantified empirically as the area between the fitted curve for two classes of sites and the theoretical curve for a single class of lower affinity, which approximates the expected effect of GTP if a G protein were present. The areas measured for 10 ligands at reconstituted tetramers correlated with similar measures of heterogeneity and with intrinsic activities reported previously for binding and response in natural membranes (P ≤ 0.00002). The data suggest that the GTP-sensitive heterogeneity typically revealed by agonists at M2 receptors is intrinsic to the receptor in its tetrameric state. It exists independently of the G protein, and it appears to arise at least in part from cooperativity between linked orthosteric sites.