Current surgical treatment of malignant gliomas largely depends on mechanistic reasoning and data collected in non-randomised studies. Technological advance has enabled more accurate resection of tumours and preservation of eloquent brain areas but ethical considerations have restricted randomised trials on the efficacy of surgery to one small trial that found a 3 month survival advantage for patients over age 65 years who received surgery and interim analysis of a larger trial. There is an argument for surgery as a palliative measure in patients with symptoms caused by mechanisms that are surgically remediable. Whether there is any survival advantage from surgery in patients other than those with immediately life-threatening, surgically remediable complications, such as raised intracranial pressure, is unclear. The available data show that if such an advantage does exist, it is modest at best. Adjuvant treatments given surgically are being studied. Chemotherapy wafers are the most prominent of the adjuvant treatments but the evidence available is insufficient to recommend their use in routine practice. In this review we examine the prevailing mechanistic model and observational data; we assess how these are applied and the priorities they indicate for future research.