OBJECTIVE The accuracy of colour-coded duplex sonography (CCDS) for differentiating preocclusive stenoses from occlusions of the internal carotid artery (ICA) is a crucial point in non-invasive quantification of atherosclerotic lesions prior to carotid endarterectomy. METHODS A total of 401 consecutive patients with CCDS followed by ICA arteriographies as gold standard was available for comparison. The entire number was divided into groups of <90%, 90-94%, preocclusive (95-99%) stenoses and occlusions. Sensitivity, specificity, and predictive value for distinguishing these groups were calculated using a contingency table. RESULTS With CCDS we found a sensitivity of 88% and a specificity of 99% in 43 preocclusive ?95% stenoses. Similar findings were seen in 31 occlusions of the ICA (SE 87%, SP 99%). CCDS accurately differentiates the subgroups of severe carotid obstructions (90-94%, ?95% and occluded) shown by a predictive accuracy of 97, 96 and 93%. Carotid endarterectomies were performed in two of three angiographically occluded but sonographically preocclusive arteries. Intraoperatively preocclusive ICAs were seen in both cases. CONCLUSIONS CCDS showed a high accuracy for differentiating preocclusive stenoses and occlusion of the ICA. Intraoperative findings indicated that angiography is not the absolute gold standard for preocclusive carotid disease in every case. Irregularities of the stenosis channel make it impossible to estimate the true area reduction in stenoses ?90%. The hemodynamic estimation of degree of stenosis by Doppler ultrasound may be closer to reality than angiographic measurement.