Two potential errors in velocity estimation, Doppler angle misalignment and intrinsic spectral broadening (ISB), were determined and used to correct recorded blood velocities obtained from 20 patients (38 bifurcations). The recorded and corrected velocities were used to grade stenoses of greater than 70% using two duplex classification schemes. The first scheme used a peak systolic velocity (PSV) of > 250 cm/s in the internal carotid artery (ICA), and the second a PSV ratio of > 3.4 (ICA PSV/common carotid artery PSV). The "gold standard" was digital subtraction angiography (DSA). The maximum error in velocity estimation due to Doppler angle misalignment was 33 cm/s, but this did not alter sensitivity of stenosis detection. ISB correction caused a reduction in PSV that decreased the sensitivity of the PSV scheme from 65% to 45%. The PSV ratio classification was not affected by ISB errors. Centres using a PSV criterion for grading stenosis should use a fixed Doppler angle and should establish velocity thresholds in-house.