Magnetic resonance phase difference techniques are commonly used to study flow velocities in the human body. Acceleration is often present, either in the form of pulsatile flow, or in the form of convective acceleration. Questions have arisen about the exact time point at which the velocity is encoded, and also about the sensitivity to (convective) acceleration and higher order motion derivatives. It has become common practice to interpret the net phase shifts measured with a phase difference velocity technique as being the velocity at a certain (Taylor) expansion time point, chosen somewhere between the RF excitation and the echo readout. However, phase shifts are developed over the duration of the encoding magnetic field gradient wave form, and should therefore be interpreted as a more or less time‐averaged velocity. It will be shown that the phase shift as measured with a phase difference velocity technique represents the velocity at the “gravity” center of the encoding bipolar gradient (difference) function, without acceleration contribution. Any attempt to interpret the measured phase shift in terms of velocity on any other time point than the gradient gravity point will automatically introduce acceleration sensitivity.