The fact that glioblastoma multiforme possesses antigens differing from those of normal brain has been stressed in early papers from Scheimberg, Mahaley, Eggers, and Brooks. In our work the presence of specific cytoplasmic and nuclear antigens in neoplastic cells has been demonstrated. These specific antigens are present not only in experimental tumours from the rat, but also in human glioblastoma, and are easily demonstrated by immunodiffusion and immunofluorescence techniques. From our work differences between intracellular and membrane antigens are clear, as the latter do not react with IgG immunoglobulin. On the other hand, tumoural antigens in glioblastoma have similar antigenic qualities to those of histocompatible antigens in normal brain. Experimental and human glioblastomas have weak antigens, as demonstrated by frequent tumour recurrence following amputation and the positive cross-reaction of antibody with normal brain in experimental models. Glioblastoma multiforme may have a common antigen as its antibodies easily cross-react positively with different human tumours with similar, histological features. As tumoural membranes did not react as cytoplasm and nuclei, we cannot say that membrane antigens resemble those of intracellular contents. The fact that viral-induced tumours may have common antigens should point to aetiological possibilities in this group of tumours. Delayed cellular response is very useful during the follow-up of these patients. Positive DNCB and intradermal reactions could be elicited in those patients in whom the antigenic overload has been reduced as a consequence of a surgical procedure. On the other hand, patients with extensive and infiltrating tumoural masses exhibited weak or negative delayed cellular responses. Humoral responses from the patient's sera may not have the prognostic value of cellular responses.