A retrospective review of 63 patients undergoing duplex scanning and angiography for suspected carotid artery disease was performed to evaluate the need for routine angiography before carotid endarterectomy. A consultant surgeon (M.H.) made a simulated management decision on the basis of a clinical summary and a duplex scan report. Twenty-four patients were selected for surgery without angiography; duplex scanning had a sensitivity of 100 per cent and a specificity of 90 per cent in the detection of internal carotid artery stenosis. In two cases duplex scanning misdiagnosed a total occlusion as a critical stenosis. Eighteen patients failed to meet the criteria for surgery and were referred for angiography. Twenty-one patients were selected for conservative treatment on the basis of the duplex scan report. Combining the surgical and conservative groups (45 patients), duplex scanning had a sensitivity of 96 per cent and specificity of 95 per cent for the detection of stenosis greater than 50 per cent. In the identification of a total occlusion, duplex scanning had a poor sensitivity of 50 per cent. These results suggest that routine angiography before carotid endarterectomy is unnecessary in selected patients but that a suspected occlusion should be confirmed by angiography.