The accuracy of volume flow rate measurements obtained with phase-contrast methods was assessed by means of computer simulation and in vitro experiments. Factors studied include (a) the partial-volume effect due to voxel dimensions relative to vessel dimensions and orientation and (b) intravoxel phase dispersion. It is shown that limited resolution (partial-volume effect) is the major obstacle to accurate flow measurement for both laminar and plug flow. The results show that at least 16 voxels must cover the cross section of the vessel lumen to obtain a measurement accuracy to within 10%. Measurement accuracy also greatly depends on the relative signal intensity of stationary tissue and is better for laminar flow than plug flow.