Methods

Corpus scope (time windows)

Each tradition on the timeline uses a published overlap window — not “every manuscript ever.” Details and witness counts live in Coverage and DATA.md.

Greek New Testament

Witnesses whose paleographic or catalog range overlaps 1–400 CE (frozen window). Papyri from INTF Liste cache; majuscules from scripts/uncial-seed.json. Not the full ~5,700+ witness tradition.

Qurʾān (Hijazi catalog)

1–100 AH (~622–719 CE overlap) for dated or overlapping leaves in scripts/quran-seed.json (Corpus Coranicum–aligned metadata). Later display mushafs are generally excluded unless their published range genuinely overlaps.

Hebrew Bible & Septuagint

250 BCE – 400 CE for this corpus only (BCE allowed on the Hebrew/LXX timeline). Hand-curated DSS and LXX papyri in scripts/hebrew-lxx-seed.json — not BHQ or Rahlfs apparatus dumps.

Nag Hammadi

Codex-level witnesses positioned by paleography (~300–400 CE overlap) from the Claremont NHA catalog seed — catalog cards, not a Coptic variant census.

Greek NT variation units (CNTR vs SR GNT)

Our Greek NT explorer and downloadable census compare committed diplomatic transcriptions from CNTR to the open Society of Biblical Literature Greek New Testament (SR GNT), also sourced from CNTR. This is a mechanical, word-aligned disagreement count — not the Nestle-Aland 28th edition apparatus.

Unit definition

One word-aligned variation unit is a single aligned word position where a witness's surviving letters disagree with SR GNT (spelling, omission, addition, substitution, etc.). Units are classified by kind via scripts/lib/variant-classify.mjs.

Fragment-safe counting

We count only extant text on each witness — lacunae and broken edges do not generate false disagreements. Compare logic respects per-verse availability in src/data/witness-texts.json.

Exports

Flat browse index: src/data/variant-index.json. Researcher-facing rows: public/variant-census.json / public/variant-census.csv (see Cite → For researchers).

Curated vs computed layers

Computed at build time

Greek disagreement totals, per-book summaries, variant index, SR witness ESN export, and coverage aggregates in src/data/coverage.json.

Famous passages (hand-curated)

Block-level cards for debaters and classrooms in scripts/famous-passages.json → src/data/famous-passages.json. Verified by npm run famous-passages; not an exhaustive apparatus.

Use / claims hub (hand-curated)

Claim discipline and audience routes in scripts/claims-evidence.json (synced to src/data/ by the famous-passages script). Verdicts describe this site's scope, not every field debate.

Uthmanic regional rasm (hand-curated)

Illustrative table of published regional muṣḥaf reports (~35–40 in Cook / van Putten summaries) in scripts/uthmani-regional-variants.json. We cite published totals and secondary collations — we have not re-census every rasm report against every leaf in our Qurʾān seed. See Uthmanic regional rasm.

Intentional-response tags (provisional)

Optional LLM/heuristic labels on a bounded Greek NT sample. Treat as hypotheses unless tagged with a model-judged pipeline; see Coverage for tagging limits.

Image licensing policy

Leaf photos are shown only when we can host or link responsibly.

• Hosted plates: Wikimedia Commons and other openly licensed facsimile scans in public/witnesses/, each with a .attribution.json sidecar (author, license, source URL).
• Link-only: Holding-institution viewers (Vatican, BL, Leon Levy DSS, etc.) when reuse is restricted — metadata still cites INTF, Corpus Coranicum, or the holding library as documented on witness cards.
• Not rehosted: Copyrighted reading-room photography, NA28/NIV/ESV text, or proprietary critical editions.

Reproducibility (npm scripts)

After changing seeds or CNTR-derived JSON, regenerate and commit outputs before publishing.

• npm run data — refresh Liste cache + witness generation spine (when NTVMR reachable).
• npm run texts — rebuild src/data/witness-texts.json from CNTR exports.
• npm run reclassify — re-run variant classifier on committed transcriptions.
• npm run coverage — write src/data/coverage.json aggregates.
• npm run variant-index — write src/data/variant-index.json.
• npm run export-census — write public/variant-census.json / .csv (also runs in prebuild).
• npm run famous-passages — verify famous cards + sync claims JSON.
• npm run uthmani-regional — validate + copy Uthmanic rasm table to src/data/.
• npm run quran / npm run hebrew-lxx / npm run nag-hammadi — tradition-specific JSON builds.

Known limitations

• Not a full INTF Liste mirror, NA28/ECM apparatus, or tradition-scale Greek NT variant total (cf. Gurry/Ehrman scope notes on Coverage).
• Qurʾān, Hebrew/LXX, and Nag Hammadi timelines are catalog + excerpt layers — no mechanical word census comparable to Greek NT yet.
• Uthmanic rasm rows follow published regional reports (Cook 2004; al-Dānī summaries; Alrahawan & Said 2021 collation, etc.) — not a complete digital re-count of all ~40 loci.
• Intentional tags remain provisional; do not cite them as established text-critical judgments.
• English glosses on cards (WEB, Pickthall, etc.) are labeled PD reference translations — not diplomatic text.