IN RECENT YEARS there has been increased interest in using computers to arrive at a clinical diagnosis. Part of the present well-documented unreliability of psychiatric diagnoses1,2lies in the variability in the operations by which clinicians use the raw data of observation to make a diagnosis. This source of unreliability is completely eliminated by the use of a computer program which will always arrive at the same diagnosis when given the raw data describing a subject. The availability of a computer program for psychiatric diagnosis with demonstrated validity would make possible meaningful comparisons of the diagnostic composition of various populations. Such comparisons are now difficult to interpret because of the use of different diagnostic criteria by clinicians. Although several attempts at developing computer programs for classifying patients according to the standard psychiatric nomenclature have been made, they have all relied on