The purpose of this study is to ascertain possible efficacy and to understand possible mechanisms of action of intramuscular or intravenous immunoglobulin (IVIg) in the treatment of intractable epilepsy, through a review of all identifiable articles on this topic. In 24 studies, none with a placebo controlled design, 368 patients with epilepsy receiving IVIg were identified. Patients' ages ranged from < 1 to 35 years, mean 7.3 years. Female/male ratio was 0.6. All patients were reported to suffer from intractable epilepsy. The average percentage of patients with an IgG2 deficiency was 25%. The total dose of IVIg varied between 0.3 and 6.8 g/kg for a period of 0.15 to 12 months. Whenever reported, adverse effects of IVIg were minimal. None of the studies reported the need of cessation of IVIg administration due to adverse effects. On the average, the mean clinical seizure reduction and the mean EEG improvement were 52% and 45%, respectively. On the average the percentage of patients with complete seizure remission and the percentage of patients with behavioral improvement were 23% and 63%, respectively. Cumulative meta-analysis of the identified articles is not possible due to the lack of controlled studies, the heterogeneity of the available studies, and the possible publication bias of unpublished negative data. Given these pitfalls, this literature study nevertheless allows some conclusions: (i) There is no formal proof of efficacy of IVIg treatment in epilepsy, and the present review underscores the need of controlled clinical trials before firm conclusions concerning efficacy can be drawn. The uncontrolled clinical observations discussed in this 'state-of-the-art' review generate suggestive evidence at best. They suggest that IVIg might be effective in some patients with intractable epilepsy, and may be considered as a safe add-on medication in various types of idiopathic and symptomatic intractable epilepsy. (ii) Review of the literature did not help in explaining intractable epilepsy or the mechanism of action of IVIg, but did permit some inferences that could serve to design future clinical and experimental approaches to IVIg administration in epilepsy.