We studied the reproducibility of measurement of carotid stenosis and assessment of plaque surface morphology on 1001 angiograms from a consecutive series of patients entered into the European Carotid Surgery Trial. Inter-observer agreement (Kappa statistic, 95% confidence interval (CI)) for categorization of carotid stenosis, as 0-29%, 30-69% or 70-99% was good (0.68, 0.63-0.73) on 789 conventional or digitally subtracted selective angiograms, and good (0.64, 0.54-0.75) on 174 conventionally and digitally subtracted aortic arch injection angiograms, but was poor (0.29, 0.02-0.80) on 29 intravenous digital subtraction angiograms. Inter-observer agreement did not vary with the method of image acquisition of arterial angiograms, but was dependent on the quality of visualization of the stenosis: kappa = 0.73 (0.67-0.79) for good quality angiograms vs. 0.54 (0.44-0.64) for poor quality angiograms. Inter-observer agreement for assessment of plaque surface morphology was moderate (kappa 0.4-0.6) and did not vary with type of angiography or method of image acquisition. However, ulceration was reported most frequently on selective angiograms and on those angiograms on which the quality of visualization of the stenosis was good. We conclude that the reproducibility of measurement of carotid stenosis and the assessment of plaque surface morphology vary depending on the type of angiography and the quality of visualization of the stenosis. This should be taken into account when validating non-invasive methods of imaging the carotid bifurcation.