Constraints on water storage capacity and actual content in the mantle must be derived not only from experimental studies, but also from investigation of natural samples. Olivine is one of the best-studied, OH-bearing "nominally anhydrous" minerals, yet there remain multiple hypotheses for the incorporation mechanism of hydrogen in this phase. Moreover, there is still debate as to whether the mechanism is the same in natural samples vs. experimental studies, where concentrations can reach very high values (up to ~0.6 wt% H2O) at high pressures and temperatures. We present new observations and review IR and TEM data from the literature that bear on this question. Hydrogen incorporation in natural olivine clearly occurs by multiple mechanisms, but in contrast to some previous assertions we find that there are strong similarities between the IR signatures of experimentally annealed olivines and most natural samples. At low pressures (lower than ~2 GPa) in both experiments and natural olivines, hydrogen incorporation might be dominated by a humite-type defect, but the nature of the defect may vary even within a single sample; possibilities include point defects, planar defects and optically detectable inclusions. IR bands between 3300 and 3400 cm-1, ascribed previously to the influence of silica activity, are apparently related instead to increased oxygen fugacity. At higher pressures in experiments, the IR band structure changes and hydrogen is probably associated with disordered point defects. Similar IR spectra are seen in olivines from xenoliths derived from deeper parts of the mantle (below South Africa and the Colorado Plateau) as well as in olivines from the ultra-high pressure metamorphic province of the Western Gneiss Region in Norway.