A retrospective review of 61 cases of angiographically confirmed occlusion of one or both carotid arteries was carried out to look at the evidence that the presence of collateral blood supply influenced the extent and type of cerebral infarction. Forty six patients had bilateral angiography from which it was possible to assess collateral filling of the internal carotid and middle cerebral arteries. As expected, patients with no CT scan evidence of infarction more frequently had transient ischaemic attacks or retinal infarcts than did those with visible cerebral infarcts. Patients with small cortical ischaemic lesions usually had appropriate neuropsychological or neurological deficits. Three patients with cortical watershed infarcts had a fluctuating deficit. Of the patients with bilateral angiograms 67% showed some filling via collaterals of the middle cerebral artery, and 43% of the internal carotid artery in the siphon. Those with collateral filling were mostly found to have normal CT scans or evidence of peripheral cortical lesions. By contrast most watershed and full territory infarcts were found in individuals whose angiograms showed no collateral filling. It is suggested that in many cases infarction occurs despite collateral flow and is due to distal embolism and that this may be relevant to the recently reported failure of the EC/IC bypass operation.