Cerebral angiography is still suggested as a first line investigation for patients with putative transient ischaemic attacks (TIA) and it is considered by most surgeons as a necessary prelude to carotid endarterectomy. That conventional cerebral angiography involves risk is well known, although the magnitude of this risk is not. Prospective studies of cerebral complications from this technique, published over the last decade, have been analysed to show that the major stroke rate after conventional cerebral angiography for patients with TIAs is likely to be about 2.4 per cent. This must be taken into account if the true morbidity of carotid endarterectomy is to be appreciated (assuming conventional angiograms have been used). Furthermore, since many sufferers are assessed by angiography but not submitted to surgery, a policy of conventional angiography for patients with TIAs puts a far greater number at risk than that actually having carotid endarterectomy.