Observability gradients and institutional capture in Silk Road cultural transmission: a pre-registered exploratory replication

Pre-registered exploratory replication of the observability-fidelity framework (Allan, 2026) on n = 26 documented Silk Road cultural transmission cases (13 Tier A1 material technologies, 5 Tier A2 disease protocols and explanations, 8 Tier A3 documented transmission failures). The pre-registered inter-rater reliability ICC failed across two scoring iterations; an amendment to the OSF pre-registration H3D7F filed and publicly timestamped on 2026-05-12 invoked a three-tier resumption rule whose exploratory-only outcome downgrades the study to descriptive/exploratory framing. Within that framing, two substantive findings emerge alongside the methodological case-study contribution: (i) Firth-corrected logistic regression of transmission outcome on observability, institutional flag, and time depth yields a separation-robust institutional flag coefficient (β = −3.05, Firth-corrected Wald p = 0.036, profile-likelihood 95% CI [−8.38, −0.18]), with both pre-coded institutional-flag = 1 items (silk sericulture protection, Byzantine Greek fire) failing transmission despite high observability — the finding rests on two pre-specified cases; (ii) the partial Spearman correlation between observability and fidelity controlling for first-attestation year on transmitted items is ρ = 0.746 (bootstrap percentile 95% CI [0.50, 0.90], p = 0.0006); the scorer-coupling threat is directionally tested but not statistically resolved by three post-v1 sensitivity analyses with prospectively-declared decision rules. A Tier A2 paired comparison (disease protocol > explanation; Cohen's d_paired = 1.10 at n = 5) is reported as a transparency observation, not as a substantive finding. The composite fidelity rubric does not decompose into mechanically derived transmission-faithfulness and cross-cultural-stability constructs. EXPLORATORY OBSERVATIONAL STUDY per OSF amendment §f exploratory-only outcome; pre-registered confirmatory replication claims on H1–H5 are precluded. All scripts, data, and intermediate JSON outputs are included; the OSF pre-registration H3D7F and 2026-05-12 amendment are publicly viewable.