The basic principles of measuring cerebral blood flow (CBF) using arterial spin labeling (ASL) are reviewed. The measurement is modeled by treating the ASL method as a magnetic resonance imaging (MRI) version of a microsphere study, rather than a diffusible tracer study. This approach, particularly when applied to pulsed ASL (PASL) experiments, clarifies that absolute calibration of CBF primarily depends on global properties of blood, rather than local tissue properties such as the water partition coefficient or relaxation time. However, transit delays from the tagging region to the image voxel are a potential problem in all standard ASL methods. The key to quantitative CBF measurements that compensate for this systematic error is to create a well-defined bolus of tagged blood and to ensure that all of the bolus has been delivered to an imaging voxel at the time of measurement. Two practical technical factors considered here are 1) producing a tagged bolus with a well-defined temporal width and 2) accounting for reduction in magnitude of the tagged magnetization due to relaxation. The ASL approach has the potential to provide a robust estimation of CBF, although the timing of water exchange into tissue and the effects of pulsatile flow require further investigation.