Reliability of magnetic resonance (MR) velocity mapping to assess severity of stenosis was assessed in vitro. Steady flow at different flow rates through five stenoses with a central orifice area ranging from 17 to 176 mm2 was measured with velocity mapping performed perpendicular to the stenotic jet. Besides determination of the stenotic cross‐sectional area and flow rate, the pressure gradient was calculated with the modified Bernoulli equation and compared with manometer measurements. Cross‐sectional areas were measured with an accuracy of ⩾76%, a precision of ⩾91%, and an error of ⩽19 mm2. Flow rates had an accuracy of ⩾72%, a precision of ⩽94%, and an error of ⩽1.4 L/min. The modification of the Bernoulli equation limited its reliability to stenoses with areas of 35‐113 mm2. Pressure gradients were calculated with an accuracy of ⩾80%, a precision of ⩾88%, and an error of ⩽15 mm Hg. The method was applied in a single patient with aortic stenosis and gave estimates that agreed with those obtained by heart catheterization.