OBJECTIVE. This study compared carotid artery sonography with angiography to determine, in retrospect, which types of sonographic errors arose from incorrect interpretation of sonographic images and which errors could be ascribed to the limitations of sonographic imaging. MATERIALS AND METHODS. A review of all patients who underwent carotid artery sonography and angiography between 1993 and 1997 at our institution revealed 66 patients with complete sets of studies, yielding 132 examinations (right or left). Studies were not reinterpreted and angiography was considered to be the gold standard. Only stenoses of 60% or greater were included in our study. If the degree or location of stenosis differed on the two imaging studies, they were reviewed together to classify the type of sonographic error. RESULTS. We found complete agreement of sonography and angiography in 115 cases (87%) and discrepancies in 17 (13%). Thirteen of 17 sonographic errors were false- positive interpretations and three were false-negative interpretations. One was an error in location. Retrospective review showed seven interpretive errors. In all these cases, the color Doppler image better revealed the degree of stenosis. Other complicating factors included inconsistencies between absolute velocities, velocity ratios, and waveforms obtained while a patient was being treated with an intraaortic balloon pump. In the other 10 discrepancies, the sonographic interpretation was accurate. Seven of these cases were false-positive interpretations in patients with contralateral occlusions or stenoses. The other three cases in this group showed long segments of stenosis, ulcerations, or tortuous vessels on angiography. CONCLUSION. Our study suggests that increased accuracy can be achieved in the interpretation of carotid artery sonography by meticulous attention to the color image. When color Doppler sonography is technically limited by tortuosity or ulceration, or if significant contralateral disease is present, misinterpretation is more likely.