Arterial spin labeling (ASL) uses proximally inverted blood water spins as an endogenous contrast agent to measure blood perfusion in tissue. In this work pulsed ASL was used to measure the kinetic curves of cerebral blood flow (CBF) across eight anatomically based regions of interest (ROIs) in five normal healthy subjects. CBF estimates based on these data were compared when obtained using a single inversion time versus fitting a kinetic model to all 10 measured inversion times. CBF estimates were also compared when fitting to the 10 inversion time data using a standard two-parameter approach (CBF and bolus arrival time) or a more recently proposed three-parameter model (CBF, bolus arrival time, and arrival time dispersion). Variations in the shape of the kinetic curve were found across the brain that were consistent across subjects. The arrival time in the occipital ROI was found to be long enough to lead to underestimation of the CBF when using a single inversion time of 1.5 s. In four out of the eight ROIs there was significant underestimation of CBF using the standard two-parameter model compared to the three-parameter approach. These results have important implications for the development of a robust, quantitative ASL protocol. © 2009 Wiley-Liss, Inc.