The validity of the classical Mulliken model of electron donor-acceptor complexes is examined for the benzene-tetracyanoethylene (TCNE) complex. Some theoretical predictions based on this model are verified by means of the CNDO/S CI-l method. The observed discrepancies are analysed with the aid of configuration analysis. The Mulliken model is shown to be too crude for a quantitative description of the benzene-TCNE supermolecule. A comparison between the experimental and theoretical data is made.