Quantitative coronary analysis is widely used in studies of progression/regression and restenosis of coronary lesions. On-line digital systems are used in diagnostic coronary angiography, and as a guide in coronary interventions. The aim of this investigation was to test the reliability of measures obtained with one commercially available on-line equipment. Well-visualized coronary lesions from patient studies were analyzed for variability in single-frame measurement. Procedural factors affecting the consistency of measurements were identified by repeated visualization of the same coronary lesion with hand- and power-injection of contrast in various positions in the field of the image intensifier, and by imaging of steel phantoms in the same positions. Steel phantoms closely resembling coronary lesions as encountered in practice were visualized in the most favourable radiologic setting compatible with clinical situations. Accuracy and precision of measurements were found to be worse than reported in validation studies. This may be due to a host of variables which may need to be tested in each laboratory performing on-line quantitative coronary angiography, when data so obtained are to be used in clinical decision making or in research studies.