Three methods for estimating peripheral artery pressure gradients from ultrasound data were assessed by means of an acute canine aortic coarctation with a variable stenosis and retrospective data from 18 patients with iliac stenoses who had duplex scanning studies and pressure measurements at the time of angiography. The measured pressure difference was correlated with end-diastolic velocity, the presence or absence of reverse flow in diastole, and a pressure difference calculated with the modified Bernoulli equation. Although the calculated pressure gradients correlated well with measured values in animal studies (11 animals, r = 0.78, n = 224, SD = 8.1), they did not in the clinical studies (r = 0.54, n = 33, SD = 28). In both cases, pressure gradients were consistently overestimated for mild stenoses. There was a strong correlation between end-diastolic velocity and pressure gradient (r = 0.71, n = 94, SD = 5.2 for animal studies; r = 0.81, n = 36, SD = 23 for clinical studies), but the data were too variable to provide useful pressure estimates. In clinical studies the absence of reverse flow in diastole at the site of the stenosis was the best indicator of a resting pressure gradient of greater than 15 mm Hg. We conclude that the modified Bernoulli equation and end-diastolic velocity correlate highly with the pressure gradient but are not clinically useful because the variability is too great. The absence of reverse flow in diastole is a more reliable indicator of hemodynamically significant stenosis.