This study critiques and extends the DSM-III use of fixed and explicit criteria by applying principles and statistics common within actuarial decision theory (eg, conditional probabilities). The value and limitations of sensitivity and specificity rates are discussed and compared with an interesting but rarely used statistic, positive predictive power. The statistics and analyses provided herein also provide an empirical method for developing diagnostic criteria and determining when and how the DSM-III cutoff points might be adjusted, recognize the importance of base rates and utilities to efficient diagnosis, and provide an explicit, quantitative means by which to make optimal differential diagnoses and to make use of the overlap among psychiatric diagnoses. The issues are illustrated in the differential diagnosis of the Borderline Personality Disorder, but they have a relevance to the development and application of the other DSM-III diagnoses.