Making up new sets of diagnostic criteria in American psychiatry has become a cottage industry with little attempt at quality control. Currently there are at least three sets in common use, the Feighner et al criteria (Washington University criteria), the Research Diagnostic Criteria (RDC), and the DSM-III criteria (DSM-III-R is a fourth set on the horizon). In research articles, how often do we meet with such simple statements as "Diagnoses were made using the Feighner criteria with the agreement of two raters," or "All patients met the RDC for schizophrenia," or "The patients were reviewed and all met DSMIII criteria for bipolar illness; the controls met criteria for other psychiatric illnesses"? The answer is often. Presumably, having read these statements, we are supposed to accept the reliability of the diagnosis as well as the validity. The latter is a complex problem that we need not go into here except to