The general structure of cytochrome oxidase is reviewed and evidence that the enzyme acts as a redox-linked proton pump outlined. The overall H+/e- stoichiometry of the pump is discussed and results [Wikström (1989), Nature 338, 293] which suggest that only the final two electrons which reduce the peroxide adduct to water are coupled to protein translocated are considered in terms of the restrictions they place on pump mechanisms. "Direct" and "indirect" mechanisms for proton translocation are discussed in the context of evidence for redox-linked conformational changes in the enzyme, the role of subunit III, and the nature of the CuA site.