Diagnostic tests in medicine must satisfy certain validity and accuracy criteria to be clinically useful. In psychiatry, the validity of a diagnostic procedure might be tested independently against clinical diagnosis, treatment response, and family history criteria; a strong relationship to any of the three would suggest clinical usefulness. Predictive value theory provides a model for such a test. The dexamethasone suppression test (DST) has a lower predictive value for major depressive disorder than most conventional laboratory tests used for diagnosis in medicine. In spite of promising early reports, the DST is not predictive of treatment response nor does it appear to identify genetic subtypes of depression. Although no diagnostic laboratory test is currently powerful enough for routine clinical use in psychiatry, laboratory tests may prove useful in predicting relapse and in continuing research on the psychobiology of mental disorders.