Throughout the 1980s, psychiatry has promoted diagnosis--with its language of categories--as the preeminent measure of psychological problems. In clinical psychiatry, the decade opened with the publication of the Diagnostic and Statistical Manual of Mental Disorders-III (DSM-III). In psychiatric epidemiology, the decade saw the development of the Diagnostic Interview Schedule (DIS) and its use in the large-scale Epidemiologic Catchment Area (ECA) surveys. Proponents of the diagnostic approach herald the DIS as a breakthrough. We argue, on the contrary, that diagnosis as a form of measurement hinders understanding. If the ECA studies provide new insight into the patterns and causes of psychological problems, they will do so despite the use of diagnostic measurement. We present here a critical analysis of the inherent weaknesses of diagnosis as a form of measurement, particularly as a means of representing psychological problems. First, we describe the weaknesses: diagnosis treats attributes as entities; it reduces the signal but not the noise; and it collapses the structural relationships. Second, we offer an interpretation of why psychiatry promotes a form of measurement poorly suited to its subject: the linguistic legacy of nineteenth-century biology and epidemiology; the social construction of the need for mental health services; and the enclosure of a scientific and professional domain. Third, we conclude that diagnostic measurement impedes understanding. We recommend eliminating diagnosis from research on the nature, causes, and consequences of mental, emotional, and behavioral problems.