Conventional arteriography (CA) was performed for 78 of 688 patients who underwent digital subtraction angiography (DSA) for suspected extracranial carotid arterial disease. Prospective readings by radiologists and retrospective readings by the authors were used to compare the results of the two studies in 142 internal carotid arteries (ICAs). DSA findings were uninterpretable for 10% (prospective) and 16% (retrospective) of the ICAs. For detecting the presence or absence of any disease evident on CA, the sensitivity and specificity of DSA (prospective) were 86% and 88%, respectively. The accuracy for differentiating diameter stenosis greater than 50% from stenosis of lesser degree was 80% sensitivity and 94% specificity. Negative or uninterpretable DSA results occurred in 22% of ICAs with stenosis visible on CA, in 22% of 46 ICAs subjected to endarterectomy, and in 66% of ICAs with ulcerated plaques. DSA obviated CA for 34% of patients undergoing endarterectomy. Symptoms had little influence on the decision to obtain CA, but the degree of stenosis did--the greater the stenosis visible on DSA, the more likely that CA was performed. DSA is a good method for evaluating disease at the carotid bifurcation, but its limitations must be appreciated if it is to be employed rationally.