I present and evaluate three interpretations of methodological naturalism (MN), the principle that scientific explanations may only appeal to natural phenomena: as an essential feature of science, as a provisional guideline grounded in the historical failure of supernatural hypotheses, and as a synthesis of these two approaches. In doing so, I provide both a synoptic overview of current scholarship on MN, as well a contribution to that discussion by arguing in favor of a restricted version of MN, placing it on (...) a firmer theoretical foundation than that supplied by previous studies, and replying to recent objections. (shrink)