The usefulness of the flush-ended electrode forin vivo measurements of oxygen tension in tissue has been restricted by the lack of a satisfactory theoretical or empirical relationship between the polarographic current and the oxygen tension in the tissue. It is shown here that, if the electrode is small enough (so that its oxygen consumption is negligible compared with that of the nearby tissue), the current is proportional to the mean value over the electrode surface of the values that the oxygen tension would have taken if the potential had not been applied. The proportionality factor depends on the capillary-tissue structure in the neighbourhood of the probe. However, a membrane covering the surface of the electrode can be constructed so as to eliminate this effect and so enable comparisons between the levels of oxygen tension to be made at different points in the tissue. At the same time, the loss of sensitivity to variations of oxygen tension in time can be reduced to negligible proportions. The current, in this case, is proportional to the mean undisturbed value of the oxygen tension over the surface of the membrane.