OBJECTIVE To analyze the value of ultrasound for early diagnosis and follow-up of internal carotid artery dissection. METHODS The carotid arteries were evaluated in 43 consecutive patients using extracranial and transcranial pulsed-wave Doppler and duplex sonography. RESULTS Ultrasound examination was performed, on average, 7.7 days after the first symptoms. The dissections subsequently were verified by MRI (16 patients), angiography (13 patients), or both (14 patients) on average 4.4 days later. The overall sensitivity of the combined examination was 95% (93% for extracranial Doppler, 86% for transcranial Doppler, and 79% for duplex sonography). All three methods detected occlusions or high-grade stenoses in 100% of patients and moderate- or low-grade stenoses in 80% (combined methods), 70% (extracranial Doppler), 40% (transcranial Doppler), and 20% (duplex) of the patients. The findings in 33 patients with an occlusion or high-grade stenosis according to neuroradiology were as follows: absent flow signal in the internal carotid artery (100%) and biphasic (stump) flow in its bulb (86%), high-resistance flow pattern of the ipsilateral common carotid artery (91%), signs of collateral flow across the circle of Willis (97%), and low flow in the middle cerebral artery (79%) on transcranial insonation. In seven patients, a moderate stenosis of the high cervical carotid segment was found because of a retromandibular high-velocity signal. In five of them this was the only abnormal finding. Duplex examination was helpful because it confirmed absent internal carotid artery flow or stump flow in the case of occlusion or high-grade stenosis (100%) and excluded an atherosclerotic origin by demonstrating a patent bulb (100%) and the absence of plaques (95%). Follow-up studies showed recanalization in 63% of patients, occurring at variable intervals. Occlusion persisted in 37%. CONCLUSIONS Ultrasound performed within the first weeks can corroborate a clinically suspected carotid dissection in up to 95% of patients. Repetitive follow-up studies in most cases are sufficient to monitor evolution.