Paper I (Referential Displacement and Ritual Capture in Organizational KPIs: A Five-Stage Unified Framework for Measuring Epistemic Transfer, KPI Validity Degradation, and Organizational Reform Capacity) established a five-stage epistemic transfer mechanism and derived two formal instruments for its detection: V(m, O, t), a structured scoring model grounded in information theory and cybernetics, and RD(m, t), a linguisticallycomputable proxy for ritual capture derived from the ratio of metric-referential to system-referential language in organizational documents. This paper reports the empirical validation of both instruments against a corpus of twenty-nine publicly available engineering documents spanning three register categories. All primary hypotheses are confirmed: Pearson r(RV , RRD ) = 0.973 (H1, conver-gent validity), category ordering B < C < A with a 351-times discrimination ratio (H2, register discrimination), Mann-Whitney U = 0 (H3, large effect), and three of four metrics showing the predicted C-first, R-lagging degradation sequence (H5, trajectory direction). A large language model validation round achieves Pearson r = 0.958 against the lexicographic scores, establishing triple convergence across three methodologically independent instruments. Analysis of systematic divergences between the two primary instruments reveals two interpretable failure modes — DIAGNOSTIC (lexicographic over-fires on analytical documents) and REFORMER (LLM over-fires on critical/meta-aware documents) — and motivates a combined signal model that classifies documents into four interpretive quadrants. The REFORMER quadrant is the empirical discovery that motivated the extension of Paper I’s four-stage mechanism to five stages: practitioners who have already detected ritual capture are identifiable from documents before Stage 5 suppression completes. All scripts, corpora, lexicons, and scoring protocols are available in the companion repository.
Luciano Federico Pereira (Tue,) studied this question.