In this paper I present a proposal to reformulate the argument of Alvin Plantinga (2011) against metaphysical naturalism. Contrary to Plantinga’s argument, in this new version I propose to consider the probability of the reliability of cognitive faculties, not with regard to any kind of beliefs, but only with respect to metaphysical beliefs. I claim that those who accept naturalism have a defeater for the belief that their cognitive faculties are reliable with respect to metaphysical beliefs and, thus, they have a defeater for any of their metaphysical beliefs, including the belief in metaphysical naturalism. Therefore, those who accept naturalism have a defeater for naturalism; in other words, metaphysical naturalism is self-defeating.