Evidence-based medicine (EBM) is an important new paradigm of the medical profession. While the quantitative approach of EBM has its place, clinical medicine must take into account many subtleties that EBM fails to consider. In this article, the authors describe three caveats to this quantitative approach: (1) the detection of "maybe disease" (physiologic, anatomic, or histologic abnormalities that may not ever be overtly expressed in the patient's lifetime) inflates apparent diagnostic test performance; (2) probability revision is valuable primarily as an exercise to gain qualitative insights; and (3) patients are likely to be interested more than just central tendencies in making treatment decisions. They then consider some challenging questions facing clinician-educators: how do they prepare students for situations where there is an absence of rigorous evidence? Should they teach students that the burden of proof lies in demonstrating efficacy or in demonstrating ineffectiveness? And what should they tell students about when to seek evidence to aid diagnostic and treatment decisions?