Stability of arterial whole blood and plasma concentrations is a basic requirement in the application of the oxygen-15 (15O2) steady-state inhalation technique for measuring regional cerebral blood flow and oxygen use. The level of stability obtainable in practice is reported in the form of retrospectively analysed blood data from 626 consecutive studies in patients with a range of clinical conditions. Serial arterial whole blood and plasma concentrations were measured during both C15O2 and 15O2 inhalations, and coefficients of variation were calculated. In addition, these concentrations were compared with the corresponding values recorded at the start of each study and maximum variations were calculated. For all four concentrations, mean and median coefficients of variation were around 5 and 4%, respectively. Mean and median maximum variations were around 9 and 7%, respectively. The effects of these variations on the calculations of regional cerebral blood flow, oxygen extraction, and oxygen use were estimated. Mean expected errors were found to be between 4 and 9%, and median expected errors between 3 and 6%. Inherent blood sampling errors were assessed from blood volume studies using 11CO-labeled red cells. These errors were found to be less than 3%.