We reanalyze the issue whether the resonance peak observed in neutron scattering experiments on the cuprates is an exciton, a pi-resonance, or a magnetic plasmon. We consider a toy model on-site Hubbard U and nearest-neighbor interaction in both charge and spin channels. We find that the mixing between pi and spin channels is not negligible, but the resonance remains predominantly an exciton even if the magnetic interaction is absent and the d-wave pairing originates from attractive density-density interaction. Our results indicate that it may be difficult to distinguish between spin- and charge-mediated pairing interactions by just looking at the resonance peak in the spin susceptibility.