Recent measurements of the 4He(y, p)3H and 4He(y, n)3He cross sections suggest that their ratio in the giant dipole resonance region is approximately I . 6. This is in marked disagreement with detailed model calculations, which have given a ratio close to unity. We here explain why the calculations yielded this result, by using a simple argument based on independent experimental data and some apparently reasonable assumptions. In order to remove the discrepancy with experiment, it appears to be necessary to relax one of these assumptions and to include contributions from states of higher configurations. A satisfactory fit to the cross sections is then possible, but the required isospinmixing matrix element is much larger than that derived from the Coulomb interaction plus a phenomenological charge-symmetry breaking interaction.