In this paper I defend the claim that Paul Feyerabend held a robust metaphilosophical position for most of his philosophical career. This position I call Decision-Based Epistemology and reconstruct it in terms of three key components: a form of epistemic voluntarism concerning the justification of philosophical positions and a behaviorist account of philosophical beliefs, which allows him to cast normative arguments concerning philosophical beliefs in scientific methodology, such as realism, in terms of means-ends relations. I then introduce non-naturalist and naturalist (...) variants of his conception of normativity, which I trace back to his mentors Viktor Kraft and Karl Popper, respectively. This distinction, introduced on the metaphilosophical level, can can be put to use to explain key changes in Feyerabend’s philosophical proposals, such as the viability of his methodological argument for realism. I conclude that this Decision-Based Epistemology should be further explored by historically embedding Feyerabend’s metaphilosophy in a voluntarist tradition of scientific philosophy. (shrink)