The accuracy of two-dimensional time-of-flight magnetic resonance angiography (MRA) in judging diameter narrowing of the extracranial carotid artery was studied using digital subtraction angiography (DSA) as the gold standard. Particular attention was paid to the specificity of the technique for 70–90% diameter stenoses. MRA and DSA examinations of 90 cervical carotid arteries were graded independently by 2 blinded readers. The examinations were compared in the categories 0–29, 30–69, 70–99 and 100% diameter narrowing (NASCET measurements). Overestimation of the degree of stenosis measured angiographically resulted in MRA sensitivity estimates of 77.8% for 0–29% angiographic stenosis and 31% for 30–69% angiographic stenosis. In the 70–99% stenosis group a false-positive rate of 21.9% resulted in 22 of 42 arteries being inappropriately placed in the high-grade stenosis group. In cases of 100% occlusion there was a false-negative rate of 35.7% (5 of 14 cases). MRA is reliable as a screening test for extracranial carotid occlusive disease. Limited ability of the technique to distinguish between ''surgical'' and ''non-surgical'' disease with inappropriate inclusion of some patients as potential surgical candidates limits its application as a single definitive pre-o