We undertook a prospective double-blind study of 128 carotid artery bifurcations using colour Doppler and duplex sonography and angiography. Sixty-four patients with cerebrovascular events were admitted for angiography. All underwent sonography within 24 h of angiography. Standard duplex sonography and colour Doppler imaging without spectral analysis were performed, on the same device, by two sonographers, using defined morphological and haemodynamic criteria. Digital radiological data on vessel diameter were interpreted independently by two radiologists. The two sonographic methods gave similar grading of stenosis, compared to angiography, with an accuracy ranging from 96% in severe to 83% in minor stenoses. Colour Doppler studies gave better area measurements than standard duplex sonography, except for major stenoses. Discrepancies between ultrasonography and angiography were due mainly to minor stenoses and large plaques of calcification on the vessel walls, which masked very segmental 70% stenoses in 2 cases. Angiography is limited by its own resolution, does not show uncalcified vessel walls and does not give cross-sectional data. It would therefore be inappropriate for showing small plaques, the full extent of ectasia or for defining the carotid bulb accurately. The advantages of colour methods were in investigating sinuous or deep vessels and hypoechoic plaques. Analysis of the residual lumen of a stenosis and its extent could be determined more rapidly. Haemodynamic quantification of stenoses by standard duplex sonography may be difficult because of limited sample volume and error in estimation of angle, whereas colour Doppler allows semiquantitative estimation of haemodynamics.(ABSTRACT TRUNCATED AT 250 WORDS)