The interpretation of the gas-phase and matrix-isolation vibrational spectra of the water dimer and monomer had led to contradictory conclusions as to the effect of dimerization on the OH-stretching fundamentals. Ab initio MPn/6-311+G(2d, 2p) and MPn/+VPS(2d)(S) calculations (with n=2-4) have been performed on the water dimer to determine accurately that dimerization shift. In order to assess the quality of the computational results, frequency shifts were calculated at the same level also for several other binary complexes involving hydrogen bonds, and for which there are not large discrepancies between the gas-phase and solid-matrix data. Full geometry optimization of the complexes was performed. Analytically (SCF, MP2) or numerically (MP3, MP4) determined frequencies at the optimized geometries were employed to obtain the frequency shifts in question. In all cases a satisfactory agreement between the theoretical and solid-matrix data was found. SCF calculations underestimate and MP2 calculations overestimate the shifts in all cases. However, in only one of the complexes studied, the N-2.HF dimer, were the MP2 results far off the experimental ones. Both in this case and for the water dimer it was necessary to use a higher order of perturbation theory (MP3, MP4) to calculate the correlation correction, in order to obtain satisfactory convergence of the frequency shifts. In conclusion, our theoretical calculations indicate that the gas-phase experimental data on the dimerization shift of the OH fundamentals of the water dimer need to be reinterpreted. As a side result, we suggest a reassignment of the v(3) vibrational transition of the water monomer, performed in the recent experiments of water trapped in a Ne solid matrix.