To use a range of techniques to explore diffusion limitation as a mechanism of cellular hypoxia in the setting of head injury. A prospective interventional study. A specialist neurocritical care unit. Thirteen patients within 7 days of closed head injury underwent imaging studies. Tissue for ultrastructural studies was obtained from a cohort of seven patients who required surgery. Cerebral tissue PO2 (PtO2) was obtained using a multiple-variable sensor, and images of oxygen extraction fraction (OEF), derived from positron emission tomography, were used to calculate cerebral venous PO2 (PvO2). These data were used to derive the PvO2-PtO2 gradient in a region of interest around the sensor, which provided a measure of the efficiency of microvascular oxygen delivery. Measurements were repeated after PaCO2 was reduced from 37 +/- 3 to 29 +/- 3 torr (4.9 +/- 0.4 to 3.9 +/- 0.4 kPa) to assess the ability of the microvasculature to increase oxygen unloading during hypocapnia-induced hypoperfusion. Pericontusional tissue was submitted to electron microscopy to illustrate the structural correlates of physiologic findings. Tissue regions with hypoxic levels of PtO2 (<10 torr) had similar levels of PvO2 compared with nonhypoxic areas and hence displayed larger PvO2-PtO2 gradients (27 +/- 2 vs. 9 +/- 8 torr, p <.001). Despite similar cerebral blood flow reductions with hyperventilation, hypoxic regions achieved significantly smaller OEF increases compared with normoxic regions (7 +/- 5 vs. 16 +/- 6 %, p <.05). Pericontusional tissue showed varying degrees of endothelial swelling, microvascular collapse, and perivascular edema. Increased diffusion barriers may reduce cellular oxygen delivery following head injury and attenuate the ability of the brain to increase oxygen extraction in response to hypoperfusion. Global or regional OEF underestimates tissue hypoxia due to such mechanisms.