One day during an aneurysm surgery under the preoperative administration of mannitol, the feeding artery was accidentally occluded temporarily. Since then, investigation on whether or not mannitol can prevent infarctions in ischemic brain tissues have been carried out utilizing the thalamus infarction model in dogs. Twenty dogs were divided into equal groups, the control and the mannitol group. Temporary clipping for sixty minutes was adopted as a standard for both groups. The only difference between the groups was the administration of 2 g/kg of mannitol before the arterial occlusions. In the control, six of ten dogs showed infarctions verified as histological changes on the seventh post-ischemic day. In the mannitol group only one of ten showed infarction. Each of the difference is statistically significant (P less than 0.03).