The extension of the Barker–Henderson (BH) theory to diatomic potentials is examined by comparison of the predicted thermodynamic properties with ‘‘exact’’ MC results. Two choices of the potential separation have been considered; corresponding to the BH scheme and that of Weeks–Chandler–Andersen (WCA). Attention has been paid to the effect of the usual approximations to the exact properties of the hard reference system (equation of state and pair correlation functions) and to the convergence rate of the expansion. Conclusions can also be drawn for the extension of the WCA-blip function theory to these systems. It is shown that the behavior of the perturbation expansions exhibits the same features observed in atomic fluids. Nevertheless, the presence of the molecular anisotropy leads to a rather different assessment of the relative usefulness of each theory. Incidentally, small inaccuracies of the well-known Boublik and Nezbeda equation for the compressibility factor of dumbbells have been discovered at very high densities (outside the range previously reported).