We have examined the steady-state redox behavior of cytochrome c (Fe(c)), Fe(a), and Cu(A) of cytochrome c oxidase during steady-state turnover in intact rat liver mitochondria under coupled and uncoupled conditions. Ascorbate was used as the reductant and TMPD (N,N,N',N'-tetramethyl-1,4-phenylenediamine) as the redox mediator. After elimination of spectroscopic interference from the oxidized form of TMPD, we found that Fe(a) remains significantly more oxidized than previously thought. During coupled turnover, Cu(A) always appears to be close to redox equilibrium with Fe(c). By increasing the amount of TMPD, both centers can be driven to fairly high levels of reduction while Fe(a) remains relatively oxidized. The reduction level at Fe(a) is close to a linear function of the enzyme turnover rate, but the levels at Fe(c) and Cu(A) do not keep pace with enzyme turnover. This behavior can be explained in terms of a redox equilibrium among Fe(c), Cu(A), and Fe(a), where Fe(a) is the electron donor to the oxygen reduction site, but only if Fe(a) has an effective E(m) (redox midpoint potential) of 195 mV. This is too low to be accounted for on the basis of nonturnover measurements and the effects of the membrane potential. However, if there is no equilibrium, the internal Cu(A) --> Fe(a) electron-transfer rate constant must be slow in the time average (about 200 s-1). Other factors which might contribute to such a low E(m) are discussed. In the presence of uncoupler, this situation changes dramatically. Both Fe(c) and Cu(A) are much less reduced; within the resolution of our measurements (about 10%), we were unable to measure any reduction of Cu(A). Fe(a) and Cu(A) remain too oxidized to be in redox equilibrium with Fe(c) during steady-state turnover. Furthermore, our results indicate that, in the uncoupled system, the (time-averaged) internal electron-transfer rate constants in cytochrome oxidase must be of the order of 2500 s-1 or higher. When turnover is slowed by azide, the relative redox levels at Fe(a) and Fe(c) are much closer to those predicted from nonturnover measurements. In presence of uncouplers, Fe(a) is always more reduced than Fe(c), but in the absence of uncouplers, the two centers track together. Unlike the uninhibited, coupled system, the redox behavior here is consistent with the known effect of the electrical membrane potential on electron distribution in the enzyme. Interestingly, in these circumstances (azide and uncoupler present), Fe(a) behaves as if it were no longer the kinetically controlling electron donor to the bimetallic center.