In order that cellular immune responses afford protection without risk to sensitive normal tissue, they must be adapted to individual tissues of the body. Nowhere is this more critical than for the brain, where various passive and active mechanisms maintain a state of immune privilege that can limit high magnitude immune responses. Nevertheless, it is now clear that immune responses are induced to antigens in the brain, including those expressed by cerebral malignancies. We discuss hypotheses of how this can occur, although details such as which antigen presenting cells are involved remain to be clarified. Antitumor responses induced spontaneously are insufficient to eradicate malignant astrocytomas; many studies suggest that this can be explained by a combination of low level immune response induction and tumor mediated immunosuppression. A clinical objective currently pursued is to use immunotherapy to ameliorate antitumour immunity. This will necessitate a high level immune response to ensure sufficient effector cells reach the tumor bed, focused cytotoxicity to eradicate malignant cells with little collateral damage to critical normal cells, and minimal inflammation. To achieve these aims, priority should be given to identifying more target antigens in astrocytoma and defining those cells present in the brain parenchyma that are essential to maintain antitumour effector function without exacerbating inflammation. If we are armed with better understanding of immune interactions with brain tumor cells, we can realistically envisage that immunotherapy will one day offer hope to patients with currently untreatable neoplastic diseases of the CNS.