IF, LIKE the senior authors of the article by Regier et al,1 you have the responsibility of providing accurate data about the prevalence of mental disorders in the community as a basis for justifying the allocation of scarce treatment resources, you are haunted by 2 findings: major discrepancies in the prevalence of mental disorders in 2 large community studies and lifetime and 1-year prevalence rates for major disorders that are higher than, to many, seem reasonable. Unfortunately, there is no sharp boundary between mental disorder and psychological health or between the various mental disorders. Readers, therefore, should not be surprised that prevalence rates can vary markedly with even minor changes in diagnostic criteria and thresholds for defining mental disorders and changes in assessment questions used in community surveys. As the authors note, however, the risk factors for the DSM-III and DSM-III-R disorders2,3 remained relatively constant even when the prevalence rates varied greatly. This suggests that, despite these often puzzling differences in prevalence, both surveys measured something valid.