The aim of this study was to obtain an MRI severity-related classification of diffuse astrocytic tumours able to integrate the histological data in the grading of such tumours. We studied presurgical MR images of 91 patients with a histological diagnosis of astrocytoma, anaplastic astrocytoma and glioblastoma. A score ranging from 1 to 3 was assigned by two independent readers to each of the following MR features: oedema, mass effect, contrast enhancement, borders, signal homogeneity, necrosis, haemorrhage and flow void. Statistical analysis showed significant differences in the mean MRI scores between the three histological grades. Contrast enhancement was found to be the best predictor of the histological grade followed by necrosis, signal homogeneity and border scores. This classification represents a simple and reproducible means of carefully evaluating some macroscopic characteristics of these tumours. It could be used to integrate histological data especially in cases in which tissue sampling defects may affect the validity of this examination.