Both panegyric and criticism of evidence-based dentistry tend to be clumsy because the concept is poorly defined. This analysis identifies several contributions to the profession that have been made under the EBD banner. Although the concept of clinicians integrating clinical epidemiology, the wisdom of their practices, and patients' values is powerful, its implementation has been distorted by a too heavy emphasis of computerized searches for research findings that meet the standards of academics. Although EBD advocates enjoy sharing anecdotal accounts of mistakes others have made, faulting others is not proof that one's own position is correct. There is no systematic, high-quality evidence that EBD is effective. The metaphor of a three-legged stool (evidence, experience, values, and integration) is used as an organizing principle. "Best evidence" has become a preoccupation among EBD enthusiasts. That overlong but thinly developed leg of the stool is critiqued from the perspectives of the criteria for evidence, the difference between internal and external validity, the relationship between evidence and decision making, the ambiguous meaning of "best," and the role of reasonable doubt. The strongest leg of the stool is clinical experience. Although bias exists in all observations (including searches for evidence), there are simple procedures that can be employed in practice to increase useful and objective evidence there, and there are dangers in delegating policy regarding allowable treatments to external groups. Patient and practitioner values are the shortest leg of the stool. As they are so little recognized, their integration in EBD is problematic and ethical tensions exist where paternalism privileges science over patient's self-determined best interests. Four potential approaches to integration are suggested, recognizing that there is virtually no literature on how the "seat" of the three-legged stool works or should work. It is likely that most dentists choose to wait for collective professional standards to reveal acceptable practice or follow a strategy of punctuated equilibrium, only switching out established practice habits when very conspicuous advantages are identified. Integration in medicine appears to follow the statistically sophisticated practice of updating estimates of clinical parameters (probabilities) for diagnoses, treatments, prognoses, and side-effects. This approach is likely beyond the skill or interest of clinical dentists and it fails to incorporate values in the integration. The use of decision trees to integrate both research and experiential parameters and values is illustrated and it is shown that such a technique identifies why there are very few cases in dentistry where evidence needs to be consulted and indicates what such cases are.