Oxidized cytochrome c oxidase exhibits an unusual EPR signal at g = 12 when measured at 9 GHz and at g = 9.3 when observed at 15 GHz. Two models have been advanced for the paramagnetic site that gives rise to this unusual EPR signal: (i) a single high-spin Fe(IV) ion; and (ii) a strongly antiferromagnetically superexchange coupled high-spin Fe(III)Cu(II) dimer. These two models have been compared by EPR spectral simulations at 9 and 15 GHz. It is shown that both models can reproduce the positions of the EPR signal observed from oxidized cytochrome c oxidase as well as the temperature dependence of the EPR signal intensity. However, the Fe(IV) model does not reproduce the experimental lineshape well. Moreover, for the positions of the EPR signal to be simultaneously fitted at both frequencies, the Fe(IV) model requires that the EPR signal observed at 9 GHz arises from a "ΔM = 2" transition, whereas at 15 GHz the EPR signal arises from a "ΔM = 4" transition. In contrast, the experimental EPR spectra can be well fitted at both frequencies using a single set of spin-Hamiltonian parameters for the Fe(III)Cu(II) model. These considerations lead to the reasonable conclusion that the g = 12 EPR signal probably arises from a "ΔM = 2" transition between the M = ± 1 sublevels of the S = 2 state formed from strong antiferromagnetic superexchange coupling between high-spin ferriheme a3 and CuB(II) at the catalytic site of the enzyme.