Quantification of water permeability can improve the accuracy of perfusion measurements obtained with arterial spin labeling (ASL) methods, and may provide clinically relevant information regarding the functional status of the microvasculature. The amount of labeled water in the vascular and tissue compartments in an ASL experiment can be estimated based on their distinct diffusion characteristics, and in turn, water permeability determined from the relative vascular and tissue contributions. In the present study, a hybrid magnetic resonance imaging technique was introduced by marrying a continuous ASL method with a twice-refocused spin-echo diffusion sequence. Series of diffusion-weighted ASL signals were acquired with systematically varied b values. The signals were modeled with fast and slow decaying components that were associated with the vascular and tissue compartments, respectively. The relative amount of labeled water in the tissue compartment increased from 61% to 74% and to 86% when the postlabeling delay time was increased from 0.8 to 1.2 and to 1.5 secs. With a b value of 50 secs/mm2, the capillary contribution (fast component) of the ASL signal could be effectively minimized. Using the single-pass approximation model, the water permeability of gray matter in the human brain was estimated based on the derived relative water fractions in the tissue and microvasculature. The potential for in vivo magnetic resonance mapping of water permeability was showed using two diffusion weighted ASL measurements with b=0 and 50 secs/mm2 in both healthy subjects and a case of brain tumor.