An improved method for the design and evaluation of follow-up studies of psychiatric disorders is presented. Available indices of diagnostic consistency are reviewed and their limitations discussed. A general causal model of diagnostic concordance for reliability and follow-up studies is described. This model describes the relationship of diagnosis to true clinical status, interview procedures, the consistency of the historian, typicality of the clinical features, and temporal influences. Using the method of path analysis, estimates of the influences of these variables are derived under a wide range of experimental conditions. The method has implications for classification systems, strategies of research design, and the integration of reliability and follow-up studies. Its advantages include distinguishing the effects of the multiple causes of concordance and discordance and facilitating quantitative comparisons between different groups and different experimental designs.