One hundred cases of isolated angiopathy of the central nervous system (CNS) have been described, the majority of whom (71%) were diagnosed histologically. The pathological findings were those of angiitis, most frequently granulomatous angiitis. These patients should be classified as isolated angiitis of the CNS. The minority of patients (29%) were diagnosed according to clinical and angiographic findings which lack sensitivity and specificity. They represent a heterogeneous group, comprising several cases of probable isolated angiitis of the CNS, some clinical and angiographic syndromes associated with acute hypertension, subarachnoid haemorrhage or migraine and other more obscure ''entities'' such as post-partum cerebral angiopathy and isolated benign cerebral angiopathy. In the absence of a histological diagnosis, it is preferable to classify them as isolated angiopathy of the CNS. Although isolated angiitis of the CNS can be considered a discrete, or at least definable, subset of isolated angiopathy of the CNS, it is unclear whether it is a specific disorder or part of a spectrum of disease or even one of a number of reactions to various antigens. The finding of granulomatous angiitis of the CNS (GACNS) in several patients with various systemic disorders suggests that GACNS may be a non-specific pathological reaction. Until the last decade, the prognosis of isolated angiitis of the CNS was extremely poor. Most patients now survive and return to active lives. The major influence on outcome appears to have been the use of combination corticosteroid and cyclophosphamide therapy, although no randomized controlled trials have been undertaken. Before prolonged treatment with such potent medication is prescribed, it recommended that other causes of CNS vasculitis be carefully excluded and the diagnosis of isolated angiitis of the CNS be established histologically by combined leptomeningeal and parenchymal biopsy. The risks of biopsy would not appear to exceed those of untreated isolated angiitis of the CNS. Multicentre collaboration of interested groups should be encouraged if controlled therapeutic trials are to be achieved.