The nearly perfect {100} surfaces of MgO smoke cubes formed in air do not show significant v(OH) absorption in infrared spectra from thin (10 mg cm–2), coherent films exposed to H2O vapour for several hours. It is shown that perfect, five-fold-coordinated sites are not protonated and that the proportion of protonated, low-coordination (i.e. less than five-fold) sites is < 5%. These results are in accord with theoretical predictions for H2 adsorption. In contrast, v(OH)[and v(CO3)] absorptions are observed in identical preparations subjected to prior abrasion. The increase in protonated sites is more than ten-fold. Electron microscopy shows that only minor initial alteration of surface structure at edges and corners is caused by abrasion but a major increase in the rate of surface roughening is observed when new, nucleating sites appear and develop in regions across the surface. This process results in a major, time-dependent increase in hydroxylation but it is substantially complete before infrared spectra can normally be obtained. B.E.T. studies, using N2 adsorption, do not measure the change in surface area produced by this surface roughening. Multilayer water adsorption followed by desorption more than doubles the B.E.T. surface area because of the formation of platelets decomposed from the Mg(OH)2 brucite surface layer. The particle-size distribution is altered and the B.E.T. method correctly measures this change.