Michael Devitt tells us that metaphysical realism has a kind of immunity from considerations concerning the nature of truth. Part of this immunity comes from Devitt’s insistence that realism is a metaphysical issue, not a semantic one. Most of Devitt’s critics have focused on this point, arguing that a proper understanding of the realism question necessarily involves semantic considerations :65–74, 1991; Miller in Synthese 136:191–217, 2003; Putnam in Comments on Michael Devitt’s ‘Hilary and Me’, in: Baghramian Reading Putnam. Taylor and (...) Francis, Hoboken, 2012; Taylor in Models, truth, and realism. Oxford University Press, New York, 2006). But Devitt also argues at length that semantic considerations should not lead us to give up on realism regardless of whether we admit that the correct characterization of realism is metaphysical in character or not. Devitt’s view hinges on the viability of a strategy which he calls Putting Metaphysics First, which essentially involves giving “a certain temporal and explanatory priority to metaphysical concerns”. Fully explicated, Devitt’s strategy is seen to be based on an epistemological naturalism in the vein of Quine. The main idea is that if we assume a naturalism according to which the only way to know anything is empirically through experience, then we will be able to establish a version of metaphysical realism early on in our investigations. According to Devitt, the empirical evidence for this realism is much stronger than the empirical evidence for the numerous semantical and epistemological views which are put forward in favor of antirealism. So, Devitt argues, we should hold on to the realism that we establish first and build our other theories on top of it. After detailing Devitt’s approach, I offer two objections. The first problem for Devitt is that in the process of putting metaphysics first, he puts epistemological naturalism first. This fact is in tension with his claim that epistemology is a weak starting place and a poor basis of theoretical revision. The second problem is that while our experiences of objects give us evidence that those objects exist, our experiences do not tell us that the objects exist mind-independently in all the ways required for metaphysical realism, or so I contend. Thus, I argue, even if we do put naturalized metaphysics first as Devitt recommends, realism is not secured as a starting point. (shrink)