After short preincubations with N-[(3)H]methylscopolamine ([(3)H]NMS) or R(-)-[(3)H]quinuclidinyl benzilate ([(3)H]QNB), radioligand dissociation from muscarinic M(1) receptors in Chinese hamster ovary cell membranes was fast, monoexponential, and independent of the concentration of unlabeled NMS or QNB added to reveal dissociation. After long preincubations, the dissociation was slow, not monoexponential, and inversely related to the concentration of the unlabeled ligand. Apparently, the unlabeled ligand becomes able to associate with the receptor simultaneously with the already bound radioligand if the preincubation lasts for a long period, and to hinder radioligand dissociation. When the membranes were preincubated with [(3)H]NMS and then exposed to benzilylcholine mustard (covalently binding specific ligand), [(3)H]NMS dissociation was blocked in wild-type receptors, but not in mutated (D99N) M(1) receptors. Covalently binding [(3)H]propylbenzilylcholine mustard detected substantially more binding sites than [(3)H]NMS. The observations support a model in which the receptor binding domain has two tandemly arranged subsites for classical ligands, a peripheral one and a central one. Ligands bind to the peripheral subsite first (binding with lower affinity) and translocate to the central subsite (binding with higher affinity). The peripheral subsite of M(1) receptors may include Asp-99. Experimental data on [(3)H]NMS and [(3)H]QNB association and dissociation perfectly agree with the predictions of the tandem two-site model.