Presented in this work are the results of a quantum chemical study of oxygen adsorption on small Au-n and Au-n(-) (n=2,3) clusters. Density functional theory (DFT), second order perturbation theory (MP2), and singles and doubles coupled cluster theory with perturbative triples [CCSD(T)] methods have been used to determine the geometry and the binding energy of oxygen to Au-n. The multireference character of the wave functions has been studied using the complete active space self-consistent field method. There is considerable disagreement between the oxygen binding energies provided by CCSD(T) calculations and those obtained with DFT. The disagreement is often qualitative, with DFT predicting strong bonds where CCSD(T) predicts no bonds or structures that are bonded but have energies that exceed those of the separated components. The CCSD(T) results are consistent with experimental measurements, while DFT calculations show, at best, a qualitative agreement. Finally, the lack of a regular pattern in the size and the sign of the errors [as compared to CCSD(T)] is a disappointing feature of the DFT results for the present system: it is not possible to give a simple rule for correcting the DFT predictions (e.g., a useful rule would be that DFT predicts stronger binding of O-2 by about 0.3 eV). It is likely that the errors in DFT appear not because of gold, but because oxygen binding to a metal cluster is a particularly difficult problem. (C) 2003 American Institute of Physics.