The merits of the suggestion that CuA in cytochrome oxidase is a mixed-valence binuclear site is reviewed on the basis of recent analytical and spectroscopic studies. First an alternative mononuclear model is presented. Metal analyses indicate that homogeneous oxidase preparations with high activity contain 3Cu/2Fe. Multifrequency EPR measurements demonstrate a close similarity with a copper site in nitrous oxide reductase, and this is also supported by optical and MCD spectra. Strong evidence for a binuclear site is provided by a 7-line hyperfine structure in the EPR spectra of both enzymes. A binuclear model consistent with amino acid sequence data can be formulated.