Flow-sensitive experiments (FAIR) have been performed on a tube-flow phantom in order to validate quantitative perfusion measurements on humans. A straight-forward correspondence between perfusion and bulk-flow is found. It is shown that the flow phantom model only holds when the slice profiles of the involved RF pulses are taken into account. A small flow-independent off-set may be present in the data. The off-set is explained by the model. Based on the correspondence between the phantom and the in vivo models, it is shown that the lowest flow values that could be measured in the phantom correspond to perfusion values lower than the cortical perfusion in the brain. Thus, the experimental accuracy and the computational methods for quantitative perfusion measurements in vivo can be validated by a tube-flow phantom