The steady-state behaviour of isolated mammalian cytochrome c oxidase was examined by increasing the rate of reduction of cytochrome c. Under these conditions the enzyme's 605 (haem a), 655 (haem a3/CuB) and 830 (CuA) nm spectral features behaved as if they were at near equilibrium with cytochrome c (550 nm). This has implications for non-invasive tissue measurements using visible (550, 605 and 655 nm) and near-IR (830 nm) light. The oxidized species represented by the 655 nm band is bleached by the presence of oxygen intermediates P and F (where P is characterized by an absorbance spectrum at 607 nm relative to the oxidized enzyme and F is characterized by an absorbance spectrum at 580 nm relative to the oxidized enzyme) or by reduction of haem a3 or CuB. However, at these ambient oxygen levels (far above the enzyme Km), the populations of reduced haem a3 and the oxygen intermediates were very low (<10%). We therefore interpret 655 nm changes as reduction of the otherwise spectrally invisible CuB centre. We present a model where small anti-cooperative redox interactions occur between haem a-CuA-CuB (steady-state potential ranges: CuA, 212-258 mV; haem a, 254-281 mV; CuB, 227-272 mV). Contrary to static equilibrium measurements, in the catalytic steady state there are no high potential redox centres (>300 mV). We find that the overall reaction is correctly described by the classical model in which the Michaelis intermediate is a ferrocytochrome c-enzyme complex. However, the oxidation of ferrocytochrome c in this complex is not the sole rate-determining step. Turnover is instead dependent upon electron transfer from haem a to haem a3, but the haem a potential closely matches cytochrome c at all times.