The established test for disease in the internal carotid artery using continuous wave Doppler is to listen for flow velocity changes over the supraorbital artery with ipsilateral temporal (or facial) artery compression. This is only reliable when there is a reduction in mean pressure (and flow) distal to disease in the internal carotid artery, ie reduction of lumen diameter by more than 85%. In this study, 101 vessel segments (48 with disease at the carotid junction, 53 normal) were compared with the results of angiography. Seven gave a positive temporal artery occlusion test, all of which showed severe disease. However, spectral analysis of the Doppler signals from supraorbital and common carotid arteries showed sonagram changes both with ageing and with disease. In particular, the ratio of primary peak (A) to secondary peak (B) in systole falls, the A/B ratio being lower in disease than in health. At A/B ratios less than 1.05 there was an 88% probability of disease at the carotid junction. 36/48 (75%) diseased junctions were detected, including almost all major lesions. The method did not so reliably detect small lesions (less than 2 mm plaques, less than 60% lumen diameter stenosis, and 'minimal atheroma'). In 5/53 normal junctions the A/B ratio was in the disease range. Scanning the carotid junction for turbulence yielded additional information in some cases.