Underestimation of coronary arterial stenosis at clinical arteriography has been explained by inadequate radiological views of crescentic or slit-shaped lumens. Postmortem perfusion-fixation of the coronary arteries at physiological pressures shows that most stenotic lesions result in circular, elliptical, or D-shaped lumens. Crescentic lumens in fully distended vessels are associated only with acute mural thrombus projecting into the lumen or with massive intra-intimal thrombus and plaque fissuring. Elliptical and D-shaped lumens, however, adequately explain the clinically observed phenomenon that some stenotic lesions can be seen only in one x-ray plane and not in another. It is no longer correct to postulate a crescentic or slit-like lumen. Furthermore, not only can elliptical and D-shaped lumens result in underestimation of stenosis, they may also result in serious overestimation. Lumens need only to deviate slightly from being circular for error to be introduced into the estimation of stenosis by currently used methods. The calculation of percentage stenosis from densitometric measurements or from computerised reconstruction of the arterial lumen is, however, more accurate.