The adsorption of methanol on cluster models of Bronsted acid sites of zeolite catalysts has been investigated by ab initio quantum chemical methods at the Hartree-Fock self-consistent, field (SCF) and at the second-order Moller-Plesset perturbation theory (MP2) levels. Among the two possible structures of the adsorption complex, the neutral methanol H-bonded to the zeolite OH group and the methoxonium cation attached to the zeolite surface (ion-pair), only the former is a minimum. The ion-pair structure is a transition structure for the proton transfer from one lattice oxygen to a neighboring one via the adsorbed methanol. However, the energy difference between both structures is only a few kJ/mol. There is a broad and shallow potential well which accommodates two symmetry-equivalent neutral complexes with the Bronsted proton attached to different O-sites of the lattice and the ion-pair structure connecting them. For the complex of methanol with the largest zeolite model optimized at the MP2 level, H-1 NMR chemical shifts of 10.8 and 17.4 ppm are predicted for the neutral and the ion-pair structure, respectively. The former value agrees well with the observed shift and therefore explains the observed signal as caused by fast exchange of the zeolite and methanol hydroxyl protons of the neutral structure. The vibrational frequencies calculated for the ion-pair structure do not permit an interpretation of the observed infrared spectrum. For the neutral structure, we predict frequencies of 1353 and 1015 cm(-1) for the zeolitic in-plane and out-of-plane modes, respectively, while a range of 2300-2600 cm(-1) is estimated for the zeolitic OH stretching mode. These data support a recent interpretation of the IR spectrum which explains the observed triplet of bands as a result of Fermi resonance between the strongly perturbed zeolitic OH stretch and the OH bending overtones. The required large frequency shifts are only predicted when electron correlation is included. For the methanol OH stretching frequency, we predict a range of 3260-3360 -1. This is too low compared to the observed frequency at about 3500 cm(-1) and leaves the question open whether cm the observed band can be explained by a weakly perturbed methanol OH stretch or whether another surface species is responsible for that band. We conclude that a nonconventional treatment of the dynamics may be necessary to understand the observed vibrational transitions.