Synthese. forthcomingThe scientific realism debate has long been characterized as a deadlock, and many have sought to escape its persistent stalemate through a pragmatic turn. Stance vol- untarism, developed most fully by Anjan Chakravartty, exemplifies this turn. Rather than adjudicating between realism and anti-realism, it diagnoses the gridlock as rooted in differing epistemic stances: clusters of attitudes, commitments, and policies that are not themselves truth-apt. The stance voluntarist argues that both the realist and empiricist stances are rationally permissible. This permissibility claim rests on a ra- tionality norm requiring merely that stances be logically and pragmatically coherent. This paper raises the grounding problem for stance voluntarism: what justifies this thin rationality norm, and from what perspective can it be coherently defended? I argue that the rationality norm underwriting equal permissibility cannot itself be justified from within any first-order stance. Crucially, I show that this problem survives Elder’s response to Baumann’s irrationality objection, since Elder’s level-distinction clarifies where the norm is asserted without explaining what justifies it. I then argue that the most viable grounding strategy commits stance voluntarism to epistemic instrumental- ism. Epistemic instrumentalism is a controversial and widely contested thesis that the voluntarist assumes rather than defends. This reveals that stance voluntarism is far less neutral than advertised. I conclude by mapping two tractable paths forward for the realism debate, if the conciliatory promise of stance voluntarism cannot be vindicated.