There is an increasing trend to rely on duplex ultrasound rather than angiography to measure an internal carotid artery stenosis. The aim of this study was to determine the validity of ultrasound assessment of carotid stenosis performed in community based vascular laboratories. We compared ultrasound with angiography in 225 patients referred to us for carotid intervention. Mild lesions were diagnosed by ultrasound with a sensitivity of 54%, specificity of 89%, and a positive predictive value of 89% compared with angiography. Severe lesions had a sensitivity of 93%, a specificity of 67%, and a positive predictive value of 45%. Receiver operator characteristic curves demonstrated the optimal ultrasound cut-off value of 66% stenosis as a predictor of >60% stenosis measured angiographically, is associated with a false positive rate of 38%, and a false negative rate of 9%. Similarly, if a cut-off of 76% on ultrasound is used to predict >70% stenosis measured angiographically, it would be associated with a 29% false positive rate and a false negative rate of 11%. Despite the value of non-invasive testing for carotid disease, duplex ultrasonography performed in non-accredited and some accredited laboratories may produce highly variable results. Using ultrasound as the sole diagnostic test to determine the severity of a carotid stenosis may result in a high number of inappropriate operations and a large proportion of patients who may not be offered treatment due to false negative diagnoses.