Oxygen self‐diffusion experiments on single crystals of San Carlos olivine (∼Fo92) at 1200° ≤ T ≤ 1400°C, oxygen fugacities (ƒO2) along the Ni‐NiO and Fe‐FeO buffers, and silica activity at the olivine‐orthopyroxene buffer yielded results that follow the relationship D = 2.6 × 10−10 ƒO20.21±0.03 exp [−266±11 (kJ mol−1/RT)], where D is the diffusion coefficient in m2 s−1 and ƒO2 is given in pascals. The activation energy compares reasonably well with results for pure forsterite. The positive dependence of ƒO2 implies that the oxygen defect responsible for diffusion is an interstitial rather than a more sterically reasonable oxygen vacancy. Diffusion of oxygen in other close‐packed oxides has also shown a positive dependence on ƒO2. The rate of creep of single‐crystal olivine at fixed orthopyroxene activity also shows a positive ƒO2 dependence. If oxygen interstitials should be shown to be unimportant in oxygen diffusion in oxides, then coupled mechanisms such as countervacancy diffusion must be appealed to in order to explain the positive ƒO2 dependence. Such processes are rate‐limited by the diffusion of metal vacancies which also display a positive ƒO2 dependence in olivine. Compared with data for silicon diffusion in forsterite, our data indicate that oxygen is not the slowest diffusing species in olivine. The activation energy for oxygen diffusion is also low compared to that obtained in a majority of measurements of creep in single‐crystal olivine. Hence if oxygen diffusion contributes to the control of creep in olivine, it must be coupled to another process having a nonzero activation energy. A subinvestigation of the rate of sublimation from polished olivine surfaces, using a high‐resolution thermal balance, showed surface erosion rates at 1300°C of 0.13±0.10 nm/h. Such values are far too slow to noticeably affect our diffusion measurements at temperatures ≤1400°C.