A consensus is emerging that psychiatric nosology should become more "scientific." The application of the scientific method to psychiatric nosology has a number of important advantages, including (1) providing a clear criterion by which to evaluate nosologic proposals, (2) preventing rapid changes due to nosologic "fashions" not supported by research findings, (3) increasing the prestige and acceptability of our nosologic system to individuals outside of psychiatry, and (4) optimizing the reliability and validity of our diagnostic constructs. However, many critical issues confronting nosologists are fundamentally nonempirical and cannot be addressed by the scientific method. Examples of such issues include (1) disagreements about the proper construct for a psychiatric disorder; (2) the interpretation of results when different validators provide opposing answers; (3) defining how different syndromes must be to be considered subtypes of the same disorder, distinct disorders within the same overall category, or entirely independent disorders; and (4) the proper balance of reliability and validity. The optimal use of the scientific method in our nosologic process requires a recognition of both its strengths and its limitations.