A precondition for quantitative angiography is strict standardization of all geometric parameters, i.e., position of the patient and radiation-source-object-image geometry. Correction for respiratory and cardiac cycles is equally important. When all these variables are controlled, the absolute size of a vessel can be a major source of error, since border recognition becomes progressively difficult with decreasing vascular diameter. Finally, contrast density, independent of vascular geometry, will induce errors by virtual magnification of minimal diameters. The results in this paper show the influences of image intensifier position, angulation of the imaging device, and contrast concentration on the geometry of model coronary stenoses when evaluated with computer-assisted quantitative methods. Increase of object-image intensifier distance led to an underestimation of size which increased with decreasing vascular diameter. Decrease in contrast concentration led to a significant overestimation of actual size, up to 20% (p less than 0.05). This effect could be confirmed in human coronary arteriograms (n = 11, p less than 0.05). In conclusion, realistic contrast-perfused calibration devices will have to be developed to replace wire or plaster models in order to control systemic errors that may impede the measurement of absolute vascular size.