The operative experience in Zürich of forty-one cases of spinal AVM with major intramedullary components showed that it was possible, with the aid of precise microsurgical techniques, to remove completely 60% of these lesions with improvement, or, at least, without deterioration in neurological condition. A further 12% could be apparently effectively palliated by subtotal removal. Radical surgery may be justified in patients with irreversible neurological deficits to treat pain and to prevent fatal SAH. The best results have generally been obtained in patients with less severe neurological deficits and with lesions in the cervical region rather than the thoracolumbar region. The natural history of intramedullary spinal AVMs--that of deterioration after recurrent haemorrhage--is analogous to that of intracranial aneurysms--and the need for earlier diagnosis and for early preventive surgery is the same for both. It would, perhaps, be preferable to treat all cases of spinal AVM by transvascular occlusion to obviate the risk of open surgery and of spinal deformity, but some AVMs will remain impossible to treat by this means and the long term results of embolization still require full analysis before it can be accepted as definitive treatment. Comprehensive and exact superselective spinal angiography is a mandatory prerequisite to surgery and preoperative partial embolization may facilitate operation considerably in the future. However, even the most careful angiographic studies do not always totally define the lesion and the surgeon must be prepared to find unexpected vascular relationships at operation. A simple classification of intramedullary and mixed extra/intramedullary lesions is described. The experiences with dural arteriovenous malformations in Queen Square again show that the best results are obtained in patients who have mild or moderate neurological deficit preoperatively. There is no doubt that progressive neurological deficits finally become irreversible and it is therefore clear that once the diagnosis is suspected, it should be definitively established and operation should follow immediately. The prime, indeed the only, necessary investigation is selective spinal angiography, which demands a high degree of radiological skill and experience, but given these prerequisites, may be performed with little hazard. While embolization of these lesions is possible, the simple surgical disconnection of the nidus of the shunt from the coronal venous plexus is effective in most cases, apparently permanently, and is substantially without risk.