The goal of all graduate medical education is to ensure that the graduating physician is competent to practice in his or her chosen field of medicine. The evaluation of a resident's competency to practice, however, has never been clearly defined, nor has the fixed period of time given for residency training in each specialty been shown to be the right amount of time for each individual resident to achieve competency. To better ensure that new physicians have the competencies they need, the author proposes the replacement of the current approach to residents' education, which specifies a fixed number of years in training, with competency-based training, in which each resident remains in training until he or she has been shown to have the required knowledge and skills and can apply them independently. Such programs, in addition to tailoring the training time to each individual, would make it possible to devise and test schemes to evaluate competency more surely than is now possible. The author reviews the basis of traditional residency training and the problems with the current training approach, both its fixed amount of time for training and the uncertainty of the methods of evaluation used. He then explains competency-based residency education, notes that it is possible, indeed probable, that some trainees will become competent considerably sooner than they would in the current required years of training, quotes a study in which this was the case, and explains the implications. He describes the encouraging experience of his neurosurgery department, which has used competency-based training for its residents since 1994. He then discusses issues of demonstrating competency in procedural and nonprocedural fields, as well as the evaluation of competency in traditional and competency-based training, emphasizing that the latter approach offers hope for better ways of assessing competency.