The nature of acid sites on amorphous silica-alumina (ASA) is strongly debated, as well as their infrared signature. We report a combined experimental and computational study to unravel this challenging question at the atomic scale, focusing on proton transfer from ASA to lutidine (2,6-dimethylpyridine), an experimentally widely used molecule for probing Bronsted acid sites. The ASA surface model obtained by density functional theory (OFT) calculations is validated by the comparison of infrared frequencies of OH-groups with experimental spectra. The bands observed are assigned to the various OH-groups present, as a function of their hydrogen-bond donor character and of the proximity of silanols toward aluminum atoms. The affinity of lutidine (2,6-dimethylpyridine) for each site of the ASA surface is then evaluated by sampling the DFT model and varying the experimental pretreatment conditions. A general rule is established for Bronsted acidity of ASA, by comparison with calculations on reference silica, alumina, and mordenite models: the driving force for the proton transfer from OH-groups to lutidine is the stabilization of the conjugated base (after deprotonation) of the hydroxyls, more than the intrinsic acidity of the OH-group. Pseudo-bridging silanols (PBS) are thus found to be capable of proton transfer, thanks to the stabilization of silanolate species by the formation of additional O-Al and O-Si bonds. A prominent role of water molecules adsorbed on Al atoms is also shown: they act as a proton reservoir to express intrinsic acidity and to promote the acidity of neighboring silanols. Finally, we suggest that the (nu) over bar 8a, and the (nu) over bar (8b) modes of lutidinium species are inverted with regards to lutidine, contrary to what was previously thought on the basis of empirical data. (C) 2011 Elsevier Inc. All rights reserved.