Tracking changes in resident physician training since the landmark Libby Zion case in 1984, Drs. Halpern and Detsky review both the intended and unintended effects of ACGME work hour and supervision regulations. They describe the incremental loss of the traditional model of graded autonomy for residents and point out the lack of evidence in support of this new approach. For instance, a recent study showed that having in-house critical care attendings overnight did not improve outcomes compared with having in-house residents with as-needed telephone access to their supervisors. The authors call for the ACGME and other training program regulators to promote evaluations of various models of graded autonomy, rather than set "one rigid standard on the basis of conjecture alone." Studies should examine outcomes of future patients cared for by physicians that were exposed to different training environments, as well as shorter-term evaluations of residents' current clinical competence. A prior AHRQ WebM&M interview with Dr. Thomas Nasca, head of the ACGME, discussed duty hours and the balance of autonomy with oversight.