A model for the mechanism of ATP synthase was proposed previously (Cox, G.B., Jans, D.A., Fimmel, A.L., Gibson, F. and Hatch, L. (1984) Biochim. Biophys. Acta 768, 201-208) in which the b subunit of the Fo of Escherichia coli rotated. The driving force was proposed to be an interaction between two charged residues in the membrane, namely, Lys-23 of the b subunit and Asp-61 of the c subunit. To test this proposal the Lys-23 of the b subunit was replaced by threonine using site-directed mutagenesis. The resulting mutant, although it had an impairment in the assembly of the F1F0-ATPase, was normal with respect to oxidative phosphorylation. The role of the a subunit, which had been previously proposed to be a structural one, was reassessed by examination of the possible secondary and tertiary structure of the analogous proteins from several sources. Not only did these subunits appear to have very similar structures, but in each there was a highly conserved helical arm on one of the transmembrane helices which could form a proton channel if it interacted with the Asp-61 of the c subunit. A revised model is therefore presented in which five transmembrane helices from the a subunit and two from the b subunit are surrounded by a ring of c subunits. The highly conserved nature of the structures of the a, b and c subunits from various organisms suggests that the model may have relevance for ATP synthases from bacterial plasma membranes, mitochondria and chloroplasts.