The prognostic assessment of a patient with intra-cerebral hemorrhage (IH) requires simultaneous appraisal of several parameters. We have attempted this with a multivariate method: discriminant analysis. We studied retrospectively 142 patients with non-operated IH, not due to vascular malformation, distributed two months after the initial event in two groups: 92 living patients and 50 dead. Discriminant analysis of 21 parameters from the initial examination and CT scan, selected five factors which best separate the two groups, since 89% of the patients were well classified. These five parameters (age, consciousness impairment, temperature, volume of the hematoma and ventricular hemorrhage) combined, give a prognostic score which gives for each patient his probability of survival or death. The validity of the proposed model was controlled on a test-sample of 66 patients from another department. The possibility of giving a trustworthy spontaneous prognosis on the first day can enable the evaluation of the possible benefit from surgery, which we illustrated with a group of 23 operated patients.