Current folding models for the nicotinic acetylcholine receptor (AChR) predict either four or five transmembrane segments per subunit. The N-terminus of each subunit is almost certainly extracellular. We have tested folding models by determining biochemically the cellular location of an intermolecular disulfide bridge thought to lie at the delta subunit C-terminus. Dimers of AChR linked through the delta-delta bridge were prepared from Torpedo marmorata and T.californica electric organ. The disulfide's accessibility to hydrophilic reductants was tested in a reconstituted vesicle system. In right-side-out vesicles (greater than 95% ACh binding sites outwards), the bridge was equally accessible whether or not vesicles had been disrupted by freeze--thawing or by detergents. Control experiments based on the rate of reduction of entrapped diphtheria toxin and measurements of radioactive reductant efflux demonstrated that the vesicles provide an adequate permeability barrier. In reconstituted vesicles containing AChR dimers in scrambled orientations, right-side-out dimers were reduced to monomers three times more rapidly than inside-out dimers, consistent with the measured rate of reductant permeation. These observations indicate that in reconstituted vesicles the delta-delta disulfide bridge lies in the same aqueous space as the ACh binding sites. They are most easily reconciled with folding models that propose an even number of transmembrane crossing per subunit.