Cerebral oximetry (cerebral oxygen saturation; ScO2) is used to noninvasively monitor cerebral oxygenation. ScO2 readings are based on the fraction of reduced and oxidized hemoglobin as an indirect estimate of brain tissue oxygenation and assume a static ratio of arterial to venous intracranial blood. Conditions that alter cerebral blood flow, such as acute changes in PaCO2, may decrease accuracy. We assessed the performance of two commercial cerebral oximeters across a range of oxygen concentrations during normocapnia and hypocapnia. Casmed FORE-SIGHT Elite (CAS Medical Systems, Inc., USA) and Covidien INVOS 5100C (Covidien, USA) oximeter sensors were placed on 12 healthy volunteers. The fractional inspired oxygen tension was varied to achieve seven steady-state levels including hypoxic and hyperoxic PaO2 values. ScO2 and simultaneous arterial and jugular venous blood gas measurements were obtained with both normocapnia and hypocapnia. Oximeter bias was calculated as the difference between the ScO2 and reference saturation using manufacturer-specified weighting ratios from the arterial and venous samples. FORE-SIGHT Elite bias was greater during hypocapnia as compared with normocapnia (4 ± 9% vs. 0 ± 6%; P < 0.001). The INVOS 5100C bias was also lower during normocapnia (5 ± 15% vs. 3 ± 12%; P = 0.01). Hypocapnia resulted in a significant decrease in mixed venous oxygen saturation and mixed venous oxygen tension, as well as increased oxygen extraction across fractional inspired oxygen tension levels (P < 0.0001). Bias increased significantly with increasing oxygen extraction (P < 0.0001). Changes in PaCO2 affect cerebral oximeter accuracy, and increased bias occurs with hypocapnia. Decreased accuracy may represent an incorrect assumption of a static arterial-venous blood fraction. Understanding cerebral oximetry limitations is especially important in patients at risk for hypoxia-induced brain injury, where PaCO2 may be purposefully altered.