1. The potency of a series of selective muscarinic antagonists has been measured on two functional isolated tissue preparations (rat ileum and atria) and these compared with their potency on a range of binding preparations in order to determine whether the subtypes of M2 receptor measured functionally are the same as those measured in binding studies. 2. On the functional preparations pirenzepine, hexahydrosiladiphenidol (HSD) and 4-diphenylacetoxy-N-methylpiperidine (4-DAMP) were more potent on the ileum than on the atrium (3 fold, 29 fold and 5 fold respectively), whereas himbacine, AF-DX 116 and methoctramine showed the opposite selectivity (5 fold, 3 fold and 56 fold respectively). Atropine had a similar potency on the ileum and atrium. 3. [3H]-N-methyl scopolamine was used to study M2 binding sites on membranes from rat heart and rat submandibular gland. Each preparation appeared to contain a homogeneous binding site population. The potencies of the five M2 selective antagonists (and pirenzepine) in binding studies to heart membranes were very similar to those observed in functional studies of rat atria (correlation coefficient = 0.98). Similarly the binding to submandibular gland membranes was very similar to that observed in functional studies on rat ileum (correlation coefficient = 0.97). 4. [3H]-pirenzepine was used to examine the binding of these antagonists to M1 binding sites on membranes from rat cerebral cortex. The affinities of 4-DAMP, HSD, AF-DX116 and himbacine at M1 sites were similar to their affinities on the gland. Only pirenzepine and methoctramine had higher affinity on M1 sites than on the gland. 5. Himbacine had a 20 fold lower affinity at M1 binding sites than at heart sites, and it should therefore be an important tool in identifying M1 sites. 6. Inhibition of [3H]-N-methyl scopolamine binding to rat ileum and rat brainstem by M2-selective antagonists was best described by a two-site model. In both cases the major population of sites (70-90%) appeared to be similar to sites found on the heart (correlation coefficients = 0.95 and 0.97). The other site appeared to be similar to that on the submandibular gland (correlation coefficients = 0.96 and 1.00). 7. The correlations observed in these studies in which a range of selective muscarinic antagonists was used lend weight to previous studies indicating the presence of three functionally important muscarinic receptor subtypes, typified by the binding sites studied in the cerebral cortex, submandibular gland and heart. 8. We propose that the sub-classification of the M2 muscarinic receptor into M2 and M3 subtypes on the basis of ligand binding studies should be extended to cover functionally-defined receptors as well.