Background and Purpose: Diffusion-weighted magnetic resonance imaging has been shown to be particularly suited to the study of the acute phase of cerebral ischemia in animal models. The studies reported in this paper were undertaken to determine whether this technique is sensitive to the known ischemic thresholds for cerebral tissue energy failure and disturbance of membrane ion gradients.Methods: Diffusion-weighted images of the gerbil brain were acquired under two sets of experimental conditions: as a function of cerebral blood flow after controlled graded occlusion of the common carotid arteries (partial ischemia), as a function of time following complete bilateral carotid artery occlusion (severe global ischemia), and on deocclusion after 60 minutes of ischemia.Results: During partial cerebral ischemia, the diffusion-weighted images remained unchanged until the cerebral blood flow was reduced to 15-20 ml • 100 g" 1 • min~' and below, when image intensity increased as the cerebral blood flow was lowered further. This is similar to the critical flow threshold for maintenance of tissue high-energy metabolites and ion homeostasis. After the onset of severe global cerebral ischemia, diffusion-weighted image intensity increased gradually after a delay of approximately 2.5 minutes, consistent with complete loss of tissue adenosine triphosphate and with the time course of increase in extracellular potassium. This hyperintensity decreased on deocclusion following 60 minutes of ischemia.Conclusions: The data suggest that diffusion-weighted imaging is sensitive to the disruption of tissue energy metabolism or a consequence of this disruption. This raises the possibility of imaging energy failure noninvasively. In humans, this could have potential in visualizing brain regions where energy metabolism is impaired, particularly during the acute phase following stroke. (Stroke 1992;23:1602-1612 KEY WORDS • cerebral ischemia • energy metabolism • magnetic resonance imaging • gerbils