Event Distribution in the Biblical Narrative (−4025 to 100 CE): Quantifying Temporal Clustering with Preregistered Methods

Engineering notes

Engineering notes will be added by the aipentium editorial team.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为大语言模型、生成式AI、ChatGPT相关技术、计算机视觉、深度学习等高价值论文补充中文说明。

Original abstract

January 8, 2026 - Update: Added "repobundle_A1_Sensor.zip" The original confirmatory analysis (preregistration-Locked) has been packaged (while maintaining all previous versions) for ease of reproducing the confirmatory analysis. Also, the event data set evolved from a V1 (legacy) 109 events, to V2 with 5100 dated events; and this analysis on the V2 dataset is being named "sensor_A1" as part of a larger research program. Version 2 Update - December 26, 2025 Status: Extraction + preregistered confirmatory analysis complete (manuscript in progress). Uploaded code and analysis results from V2 Project updates (OSF): https://osf.io/as8mq What Changed from Version 1 Version 1 reported extreme early-first-century clustering in a 109-event list (p < 3e-7). Version 2 tests the same core hypothesis on a dramatically larger, machine-generated, machine-reproducible event series produced by an explicit ruleset and frozen pipeline. The preregistered confirmatory results are even stronger, and the global peak remains centered in the early first century CE. Dataset Expansion Biblical events analyzed: 5,100 (with 303 exclusions documented)Total dated dataset rows (including excluded): 5,403Coverage: Full canonical narrative (Creation -> Revelation)Extraction method: Rule-based, machine-reproducible extraction using extractor_v1.2.1.py on the American Standard Version (1901)Methodology: Preregistered rules + extraction system applied before confirmatory analysis Major Confirmatory ResultsFamily A - Primary gatekeeper hypotheses H1 (Location): Peak bin midpoint in 0-75 CE?[OK] SUPPORTED - Peak at 37.5 CE (25-50 CE bin) H2 (Dominance): Peak-to-background ratio (R)?[OK] STRONGLY SUPPORTED - peak_count = 698, mean(others) = 26.84, R = 26.0Interpretation: the peak bin contains ~26x more events than the average non-peak bin. H3 (Extremity): Monte Carlo test vs uniform-timing null[OK] EXTRAORDINARILY SIGNIFICANT - 10,000,000 simulations, 0 exceedancesp ≤ 1.0e-7 (raw); p ≤ 3e-7 (familywise correction for 3 tests)Interpretation: in ten million random timelines, none produced a peak as extreme as observed. Family B - Secondary evidence-weighted tests H4 (Overdispersion / concentration): Negative binomial vs Poisson (LRT)[OK] SIGNIFICANT - theta_hat = 0.104, 95% CI [0.077, 0.140], LRT = 16336.4, p < 0.0001 (after FDR correction) H5 (Temporal scan across the timeline):[OK] SIGNIFICANT - Maximum observed span = 3,795 events, range -1650 to -425 BCE, p < 0.0001 (after FDR correction) H6 (Multiple change-points; PELT + permutation test):Candidate change-points were suggested by the deterministic algorithm, but the preregistered permutation test was not significant (10,000 permutations, seed = 42; p = 1.0). Therefore, H6 is not supported. Robustness Checks H7 (Bin alignment shift): 12.5-year shift of 25-year bins[OK] H1 HOLDS - Peak remains within the H1 window H9 (Interval-weighted analysis):[OK] H1 HOLDS - Peak at 37.5 CE under interval weighting Files Available events_merged_with_years_combo.csv - Full dated dataset (including excluded rows)exclusions_confirmatory.csv - Documented exclusions (missing/ambiguous dating)bin_counts_25y.csv - 25-year bin counts (165 bins across the analysis window)ColabCode.py - End-to-end analysis implementationfamily_a_summary.csv - Primary (Family A) resultsfamily_b_summary.csv - Secondary (Family B) resultsanalysis_report.txt - Detailed result dump Structural Notes Sample: ~5,100 events analyzedExcluded rows: 303 events with missing/ambiguous datingTime window: -4025 to 100 CE, discretized into 165 x 25-year bins AI-Assisted Development Disclosure The technical infrastructure for this study was developed with extensive AI-assisted support (ChatGPT, Gemini, Grok), used for: (i) iterative generation/refinement of Python extraction code, (ii) technical refinement of rule logic, and (iii) assembly of data-processing utilities and checksum protocols. The authors maintained full executive control over the research design. All logic patterns, event-type hierarchies, exclusion filters, and chronological anchor points were defined by the authors. No AI system was used for hypothesis formulation, final clustering analysis, or the selection of reported results. All code was reviewed, tested, and executed by the authors; all numerical outputs are author-produced and reproducible from the archived scripts and frozen inputs. The authors accept full responsibility for the manuscript's content and the integrity of the automated pipeline. Next Steps Manuscript preparation for preprint release and journal submission. How to Cite This Work McCain, W. (2025). Event Distribution in the Biblical Narrative (−4025 to 100 CE): Quantifying Temporal Clustering with Preregistered Methods (Version v2) [Computational notebook]. Zenodo. https://doi.org/10.5281/zenodo.18065787 BibTeX @misc{mccain_2025_biblical_clustering, author = {McCain, Wiley}, title = {{Event Distribution in the Biblical Narrative (−4025 to 100 CE): Quantifying Temporal Clustering with Preregistered Methods}}, month = dec, year = 2025, publisher = {Zenodo}, version = {v2}, doi = {10.5281/zenodo.18065787}, url = {https://doi.org/10.5281/zenodo.18065787}, note = {Computational notebook}} Preregistration: For the study protocol and preregistered hypotheses, see OSF Registries at https://osf.io/as8mq

Links and sources

Comments