The discrepancy between experimental and harmonically predicted shifts of the OH stretching fundamental of methanol upon hydrogen bonding to a second methanol unit is too large to be blamed mostly on diagonal and off-diagonal anharmonicity corrections. It is shown that a decisive contribution comes from post-MP2 electron correlation effects, which appear not to be captured by any of the popular density functionals. We also identify that the major deficiency is in the description of the donor OH bond. Together with estimates for the electronic and harmonically zero-point corrected dimer binding energies, this work provides essential constraints for a quantitative description of this simple hydrogen bond. The spectroscopic dissociation energy is predicted to be larger than 18 kJ/mol and the harmonic OH-stretching fundamental shifts by about -121 cm(-1) upon dimerization, somewhat more than in the anharmonic experiment (-111 cm(-1)).