ALTHOUGH STUDIES SHOW THAT PSYCHIATRISTS DO NOT OFTEN MAKE VALID PREDICTIONS OF VIOLENT CRIMINAL BEHAVIOR, THEY CONTINUE TO BE CONSULTED EITHER FOR LACK OF BETTER PREDICTORS OR FOR IGNORANCE OF THEIR SHORTCOMINGS. PSYCHIATRISTS' PROBLEM AREAS INCLUDE CLINICAL DIAGNOSIS AND TRANSLATING DIAGNOSTIC LABELS INTO LIKELY BEHAVIOR. DIAGNOSIS INVOLVES PLACING AN INDIVIDUAL AT A PARTICULAR TIME INTO CLINICAL CATEGORY FOR SOME PURPOSE. ALTHOUGH THIS IMPLIES AN INDIVIDUAL'S HAVING A SET OF 'CORE' PERSONALITY TRAITS INDEPENDENT OF ENVIRONMENT, DANGEROUS BEHAVIOR IS POOR FUNCTIONING IN A STRESSFUL ENVIRONMENT. THE PSYCHIATRIST'S JOB IS MADE MORE DIFFICULT SINCE PEOPLE PERCEIVE STRESS DIFFERENTLY; IN ADDITION, A PERSON WITH AN OBSERVED SET OF TRAITS FUNCTIONS DIFFERENTLY AT DIFFERENT TIMES. DIAGNOSTIC CATEGORIES ALSO ARE POORLY DEFINED AT PRESENT, YET PUTTING PEOPLE INTO CATEGORIES WOULD REMAIN A PROBLEM EVEN IF THE IDEAL DIAGNOSTIC CLASSIFICATION SYSTEM EXISTED. FINALLY, DIAGNOSIS IS DESIGNED FOR THERAPY AND COMMUNICATION RATHER THAN PREDICTION OF BEHAVIOR. ACCURATELY PREDICTING VIOLENT BEHAVIOR WOULD REQUIRE, IN ADDITION TO A SOUND DIAGNOSIS CLASSIFICATION SYSTEM, CATEGORIES TO DEFINE DEGREE OF DANGEROUSNESS, AND USEFUL RULES TO CONNECT THESE WITH PEOPLE AND EACH OTHER. IN ADDITION, VARIABLES MAKING IT POSSIBLE TO ASSIGN PERSONS WITH THE SAME DIAGNOSIS TO DIFFERENT CLASSES OF DANGEROUSNESS WOULD BE USEFUL. A CITED STUDY BLAMES LACK OF PREDICTIVE SUCCESS ON PSYCHIATRISTS' LACK OF KNOWLEDGE OF THE INDIVIDUALS THEY ARE ASKED TO ASSESS, AND SUGGESTS BETTER IDENTIFICATION OF INTERVENING VARIABLES, OR BEHAVIORS, HISTORIES, ETC. WHICH CONNECT PERSONS' BEHAVIOR WITH THE DIAGNOSTIC AND PREDICTION SYSTEMS. IT IS RECOMMENDED THAT A LARGE-SCALE STUDY BE DONE OF CRIMINALS BASED ON THESE IDEAS. AN EDITOR'S NOTE AND REFERENCES ARE INCLUDED. (PAP)