Why do constitutional provisions survive repeated reform attempts even when their consequences are observable and widely criticized? Existing accounts invoke veto player models, path dependence, and interest group capture, but none explains a specific and troubling regularity: the same reform content fails predictably in one institutional environment while succeeding in another, and the order in which institutional signals arrive matters more than the substantive content of the proposal itself. This paper proposes that constitutional lock-in is a geometric property of the normative space through which institutional belief dynamics operate. Building on Heteronomous Bayesian Updating (HBU), a previously formalized account of how legal professionals form strong constitutional priors through reaction-based learning rather than outcome observation, I argue that institutional belief updating has non-Abelian structure: the order in which authority signals, reform proposals, and professional reactions are received produces systematically different posteriors, and this non-commutativity is not a cognitive bias of individual agents but a structural feature of the institutional manifold as a whole. I formalize this structure by defining a gauge connection over the normative manifold implicit in the Non-Euclidean Normative Space framework (Lerer, 2025f). Three instruments of the research program are assigned differentiated roles: the CLI indexes architectural rigidity (structural dimension); the IHR approximates the scalar curvature as accumulated deformation from past reform cycles (dynamic dimension); and the LSI functions as a curvature-amplifying mechanism. The central result: constitutional lock-in corresponds formally to non-trivial holonomy. A closed institutional cycle does not return to the same belief configuration. The curvature prevents it, and the IHR is the observable that most directly approximates that curvature. Three testable predictions follow. All three are consistent with the record of 23 Argentine labor reform attempts between 1957 and 2024, and with the comparative record across Brazil, Chile, and Spain. Reform strategy should be reconceived as path engineering on a curved institutional manifold, not preference aggregation in a flat policy space. This record is part of the Law as Extended Phenotype research program.
Ignacio Adrián LERER (Sun,) studied this question.