A technique has been developed to record 18O2 dilution curves of an organ in vivo by use of 51Cr-labeled erythrocytes as a reference tracer. The technique employs anaerobic sampling of venous outflow following an intraarterial injection of tracer-laden blood and off-line determination of [18O2] and [51Cr] profiles in the venous outflow. O2 and reference indicator-dilution curves of cerebral circulation were recorded in eight experiments with six halothane-anesthetized dogs. Autologous blood labeled with the tracers was injected into a carotid artery, and brain venous outflow was sampled from the sagittal sinus. The total net extraction of O2 tracer was equal to the extraction of elemental O2. Instantaneous extraction of 18O2 along the outflow curve fell linearly with time, from an initial value of 0.6-0.7 to very small or even negative values toward the end of a pulse. This indicates that O2 undergoes a flow-limited distribution. In all experiments, the mean transit time of unmetabolized 18O2 was longer than the mean transit time of the Cr tracer. An index of the tissue O2 dilution space, hence the mean tissue PO2, is calculated from this data with the use of a modified central volume principle. This estimate of mean tissue PO2 increases as a linear function of sagittal sinus PO2 with a slope of 0.97. The method may provide an index of the critical PO2 of venous blood, the PO2 below which O2 diffusion from blood to tissue may limit its rate of metabolic uptake.