In an empirical study on the classification of the psychoses, 302 patients were rated using the Longitudinal Psychopathology Schedule. The data were condensed by factor analysis, which yielded 10 factors - mania and schizomania, depression and suicidal activity, and 6 factors concerned with psychotic symptoms (verbal hallucinosis/passivity, delusion formation, defect symptoms, social decline, cycloid symptomatology and a factor loading depressive auditory hallucinations and visual hallucinations). Provisional diagnostic groups were obtained using DSM III. Discriminant function analyses showed that the only clearly distinct diagnostic group was bipolar disorder, and this was true for various definitions. Canonical variate analyses were performed using 3- and 4-criterion groups. These showed that a group corresponding approximately to cycloid psychosis also met criteria for being a distinct group. The most detailed examination pf the data, using 4-criterion groups and serial reclassification, suggested that the psychoses might fall into 5 groups - bipolar disorder, cycloid psychosis, depression, defect states and schizoaffective depression.