Muscarinic cholinergic receptors can appear to be more numerous when labeled by [(3)H]quinuclidinylbenzilate (QNB) than by N-[(3)H]methylscopolamine (NMS). The nature of the implied heterogeneity has been studied with M(2) receptors in detergent-solubilized extracts of porcine atria. The relative capacity for [(3)H]NMS and [(3)H]QNB was about 1 in digitonin-cholate, 0.56 in cholate-NaCl, and 0.44 in Lubrol-PX. Adding digitonin to extracts in cholate-NaCl increased the absolute capacity for both radioligands, and the relative capacity increased to near 1. The latency cannot be attributed to a chemically impure radioligand, instability of the receptor, an irreversible effect of NMS, or a failure to reach equilibrium. Binding at near-saturating concentrations of [(3)H]QNB in cholate-NaCl or Lubrol-PX was blocked fully by unlabeled NMS, which therefore appeared to inhibit noncompetitively at sites inaccessible to radiolabeled NMS. Such an effect is inconsistent with the notion of functionally distinct, noninterconverting, and mutually independent sites. Both the noncompetitive effect of NMS on [(3)H]QNB and the shortfall in capacity for [(3)H]NMS can be described quantitatively in terms of cooperative interactions within a receptor that is at least tetravalent; no comparable agreement is possible with a receptor that is only di- or trivalent. The M(2) muscarinic receptor therefore appears to comprise at least four interacting sites, presumably within a tetramer or larger array, and ligands appear to bind in a cooperative manner under at least some conditions.