Continuous wave Doppler echocardiography was performed before cardiac catheterisation in 69 consecutive patients with suspected aortic stenosis. Agreement between the maximum and the mean Doppler gradients and catheterisation gradients was good. Doppler echocardiography, however, systematically underestimated the maximum and mean gradients, particularly in the high range. Stepwise regression analysis of the small pressure difference between the two methods showed that it could not be explained by age, sex, stroke volume, differences in heart rate, ejection fraction, the presence of coronary artery disease, or severity of aortic regurgitation. There was a negative curvilinear correlation between the maximum and mean Doppler gradients and the aortic valve areas that were measured at catheterisation in patients with pure aortic stenosis. The degree of correlation decreased when patients with concomitant aortic regurgitation were included. The scatter of gradients above and below the correlation line was large and this was caused by low and high transvalvar flow. These results show that the usefulness of Doppler gradients for judging the severity of aortic stenosis, both in relation to immediate diagnosis and follow up, is severely limited if transvalvar flow is not taken into account.