Scriptome methods
The Scriptome does not turn catalogue counts directly into production. It keeps each method visible, records the assumptions, and lets disagreement widen the envelope instead of hiding it.
Every cell (a region, a period, a tradition) is, in principle, one of four things: extant and known (a census counted it), extant but unknown (survives, uncatalogued -- capture–recapture estimates this), destroyed but attested (a lost work named in a catalogue we no longer hold -- title-loss estimates this), or destroyed and unattested (inferred only from a survival or production model). The public map shows the first and last of these directly (land and water); the middle two exist today only as tradition-level method runs, not yet computed at the exact region/period/tradition cell grain the map itself uses -- named honestly on the front door rather than force-fit into a per-cell number the project does not actually have.
From counts of surviving manuscripts, corrected for what catalogues miss, divided by an estimated survival rate, to a production estimate. Evidence-calibrated on the Latin West (Ker/MLGB); other traditions currently borrow that calibration, which is provisional.
Dated scribal colophons as an independent production signal that does not depend on survival modelling. The fraction of production that carried a dated colophon is an uncertain multiplier, recorded as such.
Where a manuscript's place of origin is unrecorded - most of the time - its language gives a probabilistic guess at region: sharp for Urdu or Ottoman Turkish, very wide for Arabic, which was written everywhere. Region inferred from language, not recorded origin.
Where two or more independent methods agree, confidence rises; where they diverge, the uncertainty widens honestly. No disagreement is averaged away into false precision.
Survival calibration is not one method borrowed everywhere. This section names, per tradition, what is genuinely identified, what is only provisional, and what has been proved structurally unfittable rather than merely under-searched.
Counted object: codex_volume. Ker / MLGB extant counts; MLGB3 survival fields are extant-count substrate only pending lost-entry coverage work.
The hazard/survival chain self-derives Buringh's own reciprocal-survival factors and production series from first principles; the test suite gates every candidate model on reproducing it within tolerance.
Limitations
See the fit note above for this tradition's own named limitation.
Counted object: codex_volume. Pinakes extant counts; base survival borrowed pending recalibration.
Pinakes' own dated census peaks at c16 then falls at c17/c18 -- a genuine mid-window peak that no before-only-shock, monotonic-decay survival model can ever fit for any parameter choice (a provable structural property of the model class, Wave 5). Byzantine_greek's own 1204/1453 shocks ARE genuinely evidenced for this tradition and are deliberately left untouched rather than dropped to relocate a grid optimum. The unlocking evidence named: an independent time-varying production-rate term for the Greek print-transition centuries. Separately: this tradition's 41,346-record census (39,842 from Pinakes, the rest vHMML/Bodleian) is real but carries zero mapped production region for any of it -- a deliberate Wave-1 abstention, not a gap in this fit. Pinakes and its sibling catalogues record holding repository, not production origin, so no region is invented to place these records on the map (see any Byzantine-zone region page for the honest census-exists-unmapped disclosure).
Limitations
See the fit note above for this tradition's own named limitation.
Counted object: codex_volume. Ktiv / NLI extant counts; base survival borrowed pending recalibration.
This tradition has not had a dedicated survival-model fit attempted; it borrows the Latin-West calibration provisionally where an estimate exists at all.
Limitations
See the fit note above for this tradition's own named limitation.
Counted object: codex_volume. Provisional scaffold borrowing Byzantine & Greek until catalogue data and survival calibration land.
Reproduced bit-for-bit stable across Waves 4a, 4b, and 5 (shape_k=1.40, scale_lambda=2.5) -- the one non-Latin tradition whose parametric-survival fit is not boundary-adjacent.
Limitations
Scaffold status: provisional, borrowing another tradition's calibration until this tradition's own catalogue data lands.
Counted object: codex_volume. Provisional scaffold borrowing Byzantine & Greek until catalogue data and survival calibration land.
This tradition has not had a dedicated survival-model fit attempted; it borrows the Latin-West calibration provisionally where an estimate exists at all.
Limitations
Scaffold status: provisional, borrowing another tradition's calibration until this tradition's own catalogue data lands.
Counted object: codex_volume. Provisional scaffold borrowing Byzantine & Greek until catalogue data and survival calibration land.
The borrowed Byzantine 1204/1453 placeholder shocks pinned the grid; two genuinely-researched, cited replacement shocks (Timurid Mesopotamian campaigns, ~1400) were tried first and also failed to unpin it. Only removing the shock term entirely reaches a genuinely interior fit (Wave 5), at a small honestly-reported likelihood cost. The deeper recency-concentration in the underlying vHMML age distribution remains an open, named data-representativeness question, not resolved by this fix.
Limitations
Scaffold status: provisional, borrowing another tradition's calibration until this tradition's own catalogue data lands.
Counted object: codex_volume. Provisional scaffold borrowing Byzantine & Greek until catalogue data and survival calibration land.
This tradition has not had a dedicated survival-model fit attempted; it borrows the Latin-West calibration provisionally where an estimate exists at all.
Limitations
Scaffold status: provisional, borrowing another tradition's calibration until this tradition's own catalogue data lands.
Counted object: codex_volume. Provisional scaffold borrowing Byzantine & Greek until catalogue data and survival calibration land.
Same diagnosis and fix as syriac: the borrowed Byzantine placeholder shocks were tested against a genuine cited replacement (the Ethiopian-Adal War church destructions, 1529-43) and still pinned the grid; removing the shock entirely reaches an interior fit (Wave 5).
Limitations
Scaffold status: provisional, borrowing another tradition's calibration until this tradition's own catalogue data lands.
Counted object: codex_volume. vHMML Reading Room census (Wave 2): HMML's named Christian-Arabic/Garshuni collection area, 6,370 counted records reclassified onto this slug in Wave 3a. No independent survival model exists yet.
This tradition has not had a dedicated survival-model fit attempted; it borrows the Latin-West calibration provisionally where an estimate exists at all.
Limitations
Scaffold status: provisional, borrowing another tradition's calibration until this tradition's own catalogue data lands.
Counted object: codex_volume. vHMML Reading Room census (Wave 2): Cyrillic/Glagolitic residue, 1,954 counted records reclassified onto this slug in Wave 3a. No independent survival model exists yet.
This tradition has not had a dedicated survival-model fit attempted; it borrows the Latin-West calibration provisionally where an estimate exists at all.
Limitations
Scaffold status: provisional, borrowing another tradition's calibration until this tradition's own catalogue data lands.
Counted object: codex_volume. FIHRIST extant plus colophon-rate triangulation in Phase B.
Estimated via extant-survival + colophon-rate + language-attribution triangulation instead; a dedicated parametric-survival fit has not been attempted.
Limitations
See the fit note above for this tradition's own named limitation.
Counted object: manuscript_and_woodblock_juan_title. IDP/Dunhuang census plus Chinese imperial bibliography title/juan anchors; estimates gated.
East Asian rows are juan/title or explicit manuscript/woodblock metadata, firewalled from the physical-manuscript survival chain by design until a dedicated physical-survival model is authorized.
Limitations
See the fit note above for this tradition's own named limitation.
Counted object: manuscript_and_woodblock_title. IDP/Dunhuang and Silk Road census substrate; estimates gated.
This tradition has not had a dedicated survival-model fit attempted; it borrows the Latin-West calibration provisionally where an estimate exists at all.
Limitations
See the fit note above for this tradition's own named limitation.
Counted object: manuscript_and_woodblock_title. Shosoin sealed deposit and Nihonkoku genzaisho mokuroku anchors; estimates gated.
This tradition has not had a dedicated survival-model fit attempted; it borrows the Latin-West calibration provisionally where an estimate exists at all.
Limitations
See the fit note above for this tradition's own named limitation.
Counted object: manuscript_and_woodblock_title. Tripitaka Koreana woodblock deposit anchor; estimates gated.
This tradition has not had a dedicated survival-model fit attempted; it borrows the Latin-West calibration provisionally where an estimate exists at all.
Limitations
See the fit note above for this tradition's own named limitation.
Counted object: manuscript_title. IDP Silk Road census substrate; estimates gated.
This tradition has not had a dedicated survival-model fit attempted; it borrows the Latin-West calibration provisionally where an estimate exists at all.
Limitations
Provisional census substrate; estimates are gated pending fuller coverage.
Counted object: manuscript_title. IDP Silk Road census substrate; estimates gated.
This tradition has not had a dedicated survival-model fit attempted; it borrows the Latin-West calibration provisionally where an estimate exists at all.
Limitations
Provisional census substrate; estimates are gated pending fuller coverage.
Counted object: manuscript_title. IDP Silk Road census substrate; estimates gated.
This tradition has not had a dedicated survival-model fit attempted; it borrows the Latin-West calibration provisionally where an estimate exists at all.
Limitations
Provisional census substrate; estimates are gated pending fuller coverage.
Counted object: manuscript_title. IDP Silk Road census substrate; estimates gated.
This tradition has not had a dedicated survival-model fit attempted; it borrows the Latin-West calibration provisionally where an estimate exists at all.
Limitations
Provisional census substrate; estimates are gated pending fuller coverage.
Counted object: manuscript_and_woodblock_title. IDP Silk Road census substrate; estimates gated.
This tradition has not had a dedicated survival-model fit attempted; it borrows the Latin-West calibration provisionally where an estimate exists at all.
Limitations
Provisional census substrate; estimates are gated pending fuller coverage.
Everything above assumes a manuscript sat in a library and slowly ran the risk of fire, war, and neglect. That story is simply wrong for the ancient material now in our census — the papyri of Greco-Roman Egypt and the cuneiform tablets of Mesopotamia. Nothing there survived on a shelf. A tax receipt from Roman Oxyrhynchus was written, used, and thrown on the town rubbish mound within a few years. Whether we can read it today was decided by four gates it then had to pass, each of which threw most things away:
Two consequences follow, and both are on display in our numbers. First, find-maps are climate maps, not writing maps. The Roman army fort at Vindolanda, in rainy northern Britain, happened to seal its rubbish in oxygen-free waterlogged ground — and out came two thousand ink letters about socks, beer, and birthday parties. Herculaneum's library survived by being carbonised under a volcano. Writing was everywhere; survival was a lottery with wildly different odds per climate. So for regions where the odds were effectively zero — including Alexandria itself, the ancient world's greatest book city, sitting in the damp Nile Delta — we estimate production by analogy with documented practice (the same empire ran the same paperwork everywhere), never from the near-empty find record.
Second, a sealed archive is a photograph, not a film. The Dunhuang cave (50,000 manuscripts sealed around 1000 CE) or the Ebla palace archive (buried in one destruction, c. 2300 BCE) show what was standing in one room on one day — a stock. Turning that into "how much was written per year" requires knowing how long documents were kept before disposal, and we say so explicitly instead of quietly treating one room's shelf as a century's output.
Chaining the four gates backwards from our census counts gives production estimates with honestly enormous ranges — a factor of a hundred or more between the low and high ends. We publish the ranges with their anatomy: for cuneiform, nearly half the uncertainty is the excavation gate (we simply do not know what fraction of tablet-bearing ground has been dug); for papyri the ignorance is spread more evenly, led by deposition and excavation. Strikingly, the publication backlog — the stage everyone suspects first — is the best-measured gate and contributes the least uncertainty. Every gate's assumed range is tied to a named, citable anchor or flagged plainly as an educated guess.
These bands now live on the off-by-default Ancient World lens — not just as recorded method runs. The keep-us-honest loop this paragraph promised has already run: 3 registered predictions from this chain have been scored against reality so far (2 of them LOSE, styled with the same weight as a win) — see every one, win or loss, on the track-record page.
The public region ledgers expose the relevant run rows, input snapshot hashes, spec labels, parameter provenance, method ledgers, and triangulation flags. They are provenance surfaces, not new estimate generation.
Staff-only survival recalibration, work-loss diagnostics, covariate checks, and integrated-posterior ledgers now exist as method infrastructure, including a self-deriving Buringh survival-chain audit. They do not replace the public canonical Scriptome map.