In a previous paper, the authors showed that mannitol causes cerebral vasoconstriction in response to blood viscosity decreases in cats. The present paper describes the changes in intracranial pressure (ICP) and cerebral blood flow (CBF) after mannitol administration in a group of severely head-injured patients with intact or defective autoregulation. The xenon-133 inhalation method was used to measure CBF. Autoregulation was tested by slowly increasing or decreasing the blood pressure by 30% and measuring CBF again. Mannitol was administered intravenously in a dose of 0.66 gm/kg; 25 minutes later, CBF and ICP were measured once again. In the group with intact autoregulation, mannitol had decreased ICP by 27.2%, but CBF remained unchanged. In the group with defective autoregulation, ICP had decreased by only 4.7%, but CBF increased 17.9%. One of the possible explanations for these findings is based on strong indications that autoregulation is mediated through alterations in the level of adenosine in response to oxygen availability changes in cerebral tissue. The decrease in blood viscosity after mannitol administration leads to an improved oxygen transport to the brain. When autoregulation is intact, more oxygen leads to decreased adenosine levels, resulting in vasoconstriction. The decrease in resistance to flow from the decreased blood viscosity is balanced by increased resistance from vasoconstriction, so that CBF remains the same. This might be called blood viscosity autoregulation of CBF, analogous to pressure autoregulation. Vasoconstriction also reduces cerebral blood volume, which enhances the effect of mannitol on ICP through dehydration of the brain. When autoregulation is not intact there is no vasoconstriction in response to increased oxygen availability; thus, CBF increases with decreased viscosity. With the lack of vasoconstriction, the effect on ICP through dehydration is not enhanced, so that the resulting decrease in ICP is much smaller. Such a mechanism explains why osmotic agents do not change CBF but decrease ICP in normal animals or patients with intact vasoconstriction, but do (temporarily) increase CBF in the absence of major ICP changes after stroke.