When it was published in 1980, the Diagnostic and Statistical Manual o
f Mental Disorders, third edition - universally known as DSM-III - emb
odied a new method for identifying psychiatric illness. The manual's a
uthors and their supporters claimed that DSM-III's development was gui
ded by scientific principles and evidence and that its innovative appr
oach to diagnosis greatly ameliorated the problem of the unreliability
of psychiatric diagnoses. In this paper we challenge the conventional
wisdom about the research data used to support this claim. Specifical
ly,-we argue that the rhetoric of science, more than the scientific da
ta, was used convincingly by the developers of DSM-III to promote thei
r new manual. We offer a re-analysis of the data gathered in the origi
nal DSM-III field trials in light of the interpretations that had been
offered earlier for the reliability studies of others. We demonstrate
how the standards for interpreting reliability were dramatically shif
ted over time in a direction that made it easier to claim success with
DSM-III when, in fact, the data were equivocal.