A procedure is presented which enables the physician to forecast the course a disease will take by looking through a list of items with attached numbers, i.e. weights, and simply adding the weights of those items which hold true for a specific patient. The application of this procedure to three different diseases, namely schizophrenia, facial paresis, and aphasia is presented and proves its accuracy. Furthermore, a Monte Carlo Analysis of the procedure underlying this method--multiple linear regression of dichotomous items--leads to recommendations concerning the optimum definitions of items and the structuring of data. The procedure is compared with other methods used in computer diagnosis and its main advantage, apart from its high reliability, i.e. independence of computers in its application, is stressed. Its application to other diseases is encouraged and the presentation of data for processing by this Department is invited.