Reporting heterogeneity in Bell–CHSH experiments: a meta-scientific audit of uncertainty and visibility metadata

Bell–CHSH experiments across photonic, atomic, NV-center, and superconducting platforms over four decades constitute the empirical evidence base for quantum nonlocality. The substantial heterogeneity across these reports — different platforms, entangled states, detection efficiencies, loophole-closing strategies, and noise models — is a feature of the experimental program's diversity rather than a limitation of any individual experiment. We report a meta-scientific audit of the Bell–CHSH literature 1972–2025 (N = 121 strict-scope papers, n = 30 reporting both S and σ; PRISMA-style methodology), focused on the metadata accompanying the reported quantities rather than on the experiments themselves. Two reporting-practice patterns are documented. (i) The type of within-study uncertainty σ_S is unspecified in 40% of σ-reporting papers — statistical-only, systematic-inclusive, or 95% confidence interval form is not stated — and cross-study variance of S exceeds within-study σ by approximately two orders of magnitude (I² = 99.7%; τ ≈ 0.28 on the S scale). The audit cannot disambiguate among the physical sources of this heterogeneity (genuine platform-specific variability), the metadata sources (σ-composition incommensurability across reports), or their interaction, precisely because the metadata required for disambiguation is not consistently reported. (ii) Among n = 8 σ-reporting papers also reporting a visibility V, the textbook scope-conditional relation S = 2√2 V is consistent with the reports in 4 cases where V is operationally identifiable as polarization-correlation interference fringe visibility (including one case requiring human reclassification of an automated regex extraction failure), and deviates by 0.23–0.50 in 4 cases where V designates a different physical observable (Hong–Ou–Mandel dip visibility, classical threshold value, or unspecified semantics). The symbol V in the Bell–CHSH literature therefore designates at least three operationally distinct physical quantities. We propose a minimum metadata-reporting guideline addressing SPAM-uncertainty decomposition between statistical fluctuation and systematic offset, V operational definition, measurement settings disclosure, state-geometry context, and raw count data availability. The guideline is adoptable without additional experimental work and would render the cross-platform Bell–CHSH literature amenable to standardized reproducibility-infrastructure audit. This is a documentation-practice contribution to the reproducibility ecosystem of quantum information experimental physics; it does not challenge quantum mechanics, any individual experimental result, or any specific group's reporting choices.