The experimental manipulation method is in general not feasible in studies of the etiology of severe mental illness, whereas this method can be used in studies pertaining to primary prevention, because possible preventive methods would most likely be of a protective or beneficial nature. Preventive experimentation requires early detection of those individuals who will eventually become seriously mentally deviant. A study of children at a high genetic risk for schizophrenia is described. This study suggested certain autonomic nervous system variables as predictors of later serious mental breakdown. In a WHO-sponsored project in Mauritius the authors used these results for the selection of a group of children at risk for mental illness. 1,800 3-year-old Mauritian children were studied socially, psychologically, pediatrically and psychophysiologically between 1972 and 1973. 100 of these children were selected for experimental preventive intervention during their daily presence in two modern kindergartens. Members of this group were characterized by either abnormally fast autonomic recovery or autonomic nonresponding, or normal autonomic responding. A matched group of another 100 children served as a community control. The paper presents preliminary results from the screening of the 1,800 children.