The model used for calculating perfusion by MRI techniques that use endogenous water as a tracer assumes that arterial water is a freely diffusible tracer. Evidence shows that this assumption is not valid in the brain at high blood flow rates, at which movement of water into and out of the microvasculature becomes limited by diffusion across the blood-brain barrier. In this work, the arterial spin-labeling technique is used to show that fraction of arterial water that is dependent on blood flow rate remains in the vasculature and does not exchange with brain tissue water. By using perfusion measurements without and with magnetization transfer (MT) effects, one can distinguish arterial label that exchanges into tissue because blood has much smaller MT than brain tissue. Using this technique, the extraction fraction for water is measured in the rat brain at various cerebral blood flow rates. At high flow rates (approximately 5 ml/g/min), the extraction fraction for water is found to be about 45% in rat brain. Disruption of the blood-brain barrier with D-mannitol caused an increase in the extraction fraction for water. It was possible to form an image related to the extraction fraction for water. The ability to estimate the amount of vascular water exchanging with tissue water by MRI may represent a noninvasive approach to detect the integrity of the blood-brain barrier.