It has been suggested that mean velocity of flow could be estimated by the time-averaged maximum frequency over an integral number of cardiac cycles (Gill 1985). The present study verified this theory experimentally with a computer-controlled flow phantom. The effects of some parameters on the relationship between mean velocity and time-averaged maximum frequency were also studied. Parameters investigated included beam-vessel angle, diameter of tubing, pulsatility, flow rate and stenosis. The velocities measured by the Doppler system were compared with the actual velocities. A simple theoretical model was also developed to compare with the experimental results. The results showed that, in a long straight tube, the mean velocity can be estimated to within about 5% from the time-averaged maximum Doppler shift at various flow rates and pulsatilities. The error due to geometrical spectral broadening, especially for large beam-vessel angles, can be estimated to within 3% and therefore corrected.