An expert system is presented which allows appraisal of the prognosis and a computer-supported decision for the therapy of patients with acute spontaneous subarachnoid haemorrhage (SAH). The knowledge of physicians as a synthesis of their own and other clinicians' experience is simulated with methods of artificial intelligence by setting up two data banks. In one data bank, selected information on the correlation between initial clinical parameters, on the one hand, and mortality, outcome and complications, on the other, from about 250 neurological publications is stored, taking into consideration the therapeutic regimens applied. A second data bank receives clinical and laboratory data profiles of a patient population which has already been treated. The expert system is able to compare the individual initial findings with the corresponding parameter combination stored in the data banks regarding the decision for therapy of a patient to be treated. This enables both calculation of the probable complications and prediction of the expected outcome in relation to various possible forms of therapy. The expert system can thus indicate the kind of therapy where the lowest number of complications and the best outcome can be expected, thus supporting the decision of the physician. As each new patient is treated, the volume of stored information increases, so the system possesses self-learning characteristics. To check the validity of the prognoses, the outcome estimated by the expert system for 51 patients with spontaneous SAH was compared with the actual outcome, and a high level of agreement was attained.