OBJECTIVE The purpose of this study was to evaluate the use of velocity criteria applied to transcranial Doppler (TCD) signals in the detection of stenosis of the middle cerebral (MCA), distal vertebral, and basilar arteries. METHODS Sixty-five patients who underwent both cerebral angiography and transcranial Doppler examinations in the workup of acute cerebral ischemia were reviewed. Angiography was performed a mean of 7 +/- 5 days (range, 1 to 28 days) after TCD. Interpretation of the angiogram was performed without input regarding the TCD findings. TCD interpretation was performed according to standard criteria. RESULTS When we used a mean velocity (MV) cutoff of > or = 80 cm/s in the MCA as the criterion for stenosis, 10 of 12 stenoses of any degree were detected by TCD, with 11 of 87 false-positives. Nine of 12 MCA stem (M1) stenoses were detected when a cutoff of > or = 90 cm/s was used, with 7 of 87 false-positives. When we used an MV cutoff of > or = 70 cm/s as the criterion for > or = 50% stenosis of the vertebrobasilar system, 5 of 6 stenoses were detected, with 15 of 85 false-positives. The most important confounding factor was the presence of > or = 75% stenosis of the extracranial internal carotid artery, resulting in both false-positive (from collateral flow) and false-negative (decreased volume flow from the proximal stenosis without adequate collateral flow) errors in TCD interpretation. When patients with > or = 75% stenosis of the cervical internal carotid artery were excluded from analysis, a TCD MV cutoff of > or = 80 cm/s identified 9 of 10 M1 lesions with 7 of 61 false-positives, and an MV of > or = 70 cm/s identified 3 of 4 vertebrobasilar lesions causing > or = 50% stenosis with 7 of 56 false-positives. CONCLUSIONS TCD may be an effective screening test for M1 stenosis when velocity criteria alone are used. TCD may less reliably detect intracranial vertebral and basilar artery stenosis.