Hydroxyl groups on metal oxide in water are the sites for ion exchange, and the surface hydroxyl site density on oxides is a measure of the ion-exchange capacity. Here, the Grignard reagent method was applied to determine the surface hydroxyl site density of oxide samples. The results were similar to those reported for different oxides with other methods (dehydration by heating, tritium exchange, crystallographic calculations, etc.), and they are comparable with those calculated from the closest packing of hydroxide ions. A mechanism of hydroxylation is proposed: lattice oxide ions (extremely strong bases) are exposed to aqueous solutions and are neutralized by water to become hydroxide ions. Also, the saturated deprotonation method was applied to hematite, and it was found that all the acid hydroxyl groups on hematite were deprotonated in very high concentrations of alkali solutions ( approximately 5 mol dm-3 NaOH), and from the saturated amount of OH- consumed by deprotonation, the same result as that by the Grignard method was obtained. It is shown that all hydroxyl groups take part in ion exchange and that the unusually small values reported elsewhere with the saturated (de)protonation method may contain errors. Hetero- or homogeneity of hydroxyl groups in contact with water as ion-exchange sites is also discussed. It is suggested that intensely hydrated layers at the oxide/water interface may result in homogeneous discrete sites. The development of microstructures in the oxides was suggested from the measured values of specific surface areas, and the effect of the microstructure environments on the reactivity of internal surface hydroxyl sites is discussed. Copyright 1999 Academic Press.