Changes in the exchange rate of water across the blood-brain barrier, denoted k(w), may indicate blood-brain barrier dysfunction before the leakage of large-molecule contrast agents is observable. A previously proposed approach for measuring k(w) is to use diffusion-weighted arterial spin labeling to measure the vascular and tissue fractions of labeled water, because the vascular-to-tissue ratio is related to k(w). However, the accuracy of diffusion-weighted arterial spin labeling is affected by arterial blood contributions and the arterial transit time (τ(a)). To address these issues, a two-stage method is proposed that uses combinations of diffusion-weighted gradient strengths and post-labeling delays to measure both τ(a) and k(w). The feasibility of this method was assessed by acquiring diffusion-weighted arterial spin labeling data from seven healthy volunteers. Repeat measurements and Monte Carlo simulations were conducted to determine the precision and accuracy of the k(w) estimates. Average grey and white matter k(w) values were 110 ± 18 and 126 ± 18 min(-1), respectively, which compare favorably to blood-brain barrier permeability measurements obtained with positron emission tomography. The intrasubject coefficient of variation was 26% ± 23% in grey matter and 21% ± 17% in white matter, indicating that reproducible k(w) measurements can be obtained.