The computed tomography (CT) scans of 110 consecutive patients who presented, over a 4.5-year period, following spontaneous subarachnoid haemorrhage (SAH), were reviewed. All 110 patients also had one or more 4-vessel digital subtraction cerebral angiograms. The CT scans were reviewed in each case without knowledge of the angiographic result. In nine patients (8%), SAH was confined to the perimesencephalic area, interpeduncular cistern and/or prepontine region at CT. All nine patients had at least two, and some as many as four, negative cerebral angiograms. Eighteen of the 110 patients (16%) ultimately had negative angiography. Hence, the patients with isolated perimesencephalic haemorrhage (PMH) accounted for 50% of the negative angiograms. There was a significant association between isolated PMH and negative angiography (chi 2 = 50.1, P < 0.005). The specificity of PMH for negative angiography was 100% (95% confidence interval (CI) = 97-100%) and the sensitivity of PMH for a negative study was 50% (95% CI = 16-84%). Six of the 110 patients had basilar artery aneurysms demonstrated angiographically as the cause of their SAH but none of these six had isolated PMH at CT. All patients with isolated PMH were alive and well at follow up and none had suffered repeat SAH or vasospasm-related ischaemic cerebral injury. Perimesencephalic haemorrhage should be distinguished from SAH in general, because of the good prognosis associated with it and the doubtful need for repeat cerebral angiography after an initial negative study.