Ars Inquirendi

← All conjectures · Texts, scribes & transmission

The One-Way Mirror

Status: Supported

The verdict’s fine print — quoted from the resolution record: “SUPPORTED on the registered primary clause, with the anti-connection wager holding at a 3.2x margin: under the most INCLUSIVE pinned reading, 1 in ~650 Byzantine book epigrams carries any explicit translation-from-a-named-language marking.” Read the full caveats ↓

Status is derived only from the shepherd-authored triage/prediction data above -- community submissions and claims are a separate overlay and can never change it (see the participation panel below).

This is a conjecture imagined by a language model — drawn from its trained weights and held to falsifiability, novelty, and plausibility, not to any one method: it may join two or more fields, or none. It is not an article and not evidence: it sits below the evidence/publication boundary. A quantitative prediction and a named kill-dataset are attached (when registered) so the claim stays falsifiable rather than merely evocative.

Claim (verbatim)

Byzantine book culture, 800-1453, told itself it imported nothing. Whatever actually crossed - Arabic astrology and medicine, Latin theology after 1204, Syriac ascetica - the verse paratexts that Greek scribes attached to their books celebrate authors, patrons, and the labor of copying, and will be found almost never to mark a book as translated from a foreign tongue. The wager is AGAINST visible connection: in the thousands of book epigrams gathered by DBBE, explicit translation-marking should be vanishingly rare, an order of magnitude below what any diffusionist reading of middle and late Byzantine literature would predict - the empire's self-image as source, never receiver, enforced at the level of the page.

Prediction clause (verbatim)

Prediction: fewer than 0.5% of DBBE epigrams with full recorded text contain any explicit marker of translation into Greek from a named foreign language (Syriac, Arabic, Latin, Georgian, Slavonic, or other), and epigrams naming Greek fathers or classical authors outnumber epigrams naming any foreign-language author by at least 50 to 1; coverage guard: at least 5,000 epigram occurrences with full text searched, else inconclusive.

Kill-dataset (verbatim)

Kill: the Database of Byzantine Book Epigrams (DBBE, Ghent) - full-text search across occurrences for translation verbs and foreign-language glottonyms, counting matches against the total of full-text occurrences.

On Inferpedia

This conjecture is linked to the following pages on Inferpedia, an encyclopedia of the missing — working atlas pages, some still early scaffolding.

Provenance

Run: Fresh agent generation · model: claude-fable-5

Fresh blind generation by claude-fable-5, 2026-07-20, for the connected-Old-World (movement-signatures) wave. The sixteen items lean on digit-string error phylogeny, first-translation-lag profiles (backlog cliffs and two-pulse corridors), bridge-language fingerprints (Sogdian, Aramaic-chancery, Judeo-Arabic), attribution drift, and pulse/directional asymmetries, with five anti-connection wagers (ordinals 2, 9, 13, 14, 15) where independence or non-transmission is the predicted outcome.

Novelty / leakage triage

anticipated in the literature — this exact test has never been run

Both halves of the mechanism are established in located print: Byzantium's self-presentation as source-culture (Mango's 'distorting mirror' is this item's thesis avant la lettre), and the reality that translations into Greek existed but were few and lightly advertised — Arabic dream-books and astrology, Syriac ascetica, later Planudes' Latin versions. The pinned census, however — a full-text search of DBBE occurrences for translation verbs and foreign glottonyms with a 0.5-percent ceiling and a 50-to-1 author-name ratio — has not been run in located print; the Ghent circle has studied scribal formulae and epigram conventions but I locate no count of translation-marking. The outcome is plausible but not guaranteed: verse paratexts attached to translated traditions could produce nonzero counts near the ceiling, so the wager is falsifiable as posed. Adjacent: established mechanism, genuinely unrun census.

Sources cited by the triage
  • Cyril Mango, Byzantine Literature as a Distorting Mirror: An Inaugural Lecture, Clarendon Press, Oxford, 1975.
  • Maria Mavroudi, A Byzantine Book on Dream Interpretation: The Oneirocriticon of Achmet and Its Arabic Sources, Brill, 2002. — A documented Arabic-to-Greek translation, the kind of import the epigrams should ignore.
  • Floris Bernard and Kristoffel Demoen (eds.), Poetry and its Contexts in Eleventh-century Byzantium, Ashgate, 2012. — The book-epigram scholarship around DBBE.

Predictions

Supported registered 2026-07-20 calibration prediction (parent triage: leaked/adjacent)

Resolution: Supported

Caveats: SUPPORTED on the registered primary clause, with the anti-connection wager holding at a 3.2x margin: under the most INCLUSIVE pinned reading, 1 in ~650 Byzantine book epigrams carries any explicit translation-from-a-named-language marking. Robustness runs one way by construction: (i) the registered rule deliberately over-counts - five of the twenty accepted hits are bare-metaphrasis occurrences confirmed at the verdict desk to be the Symeon Metaphrastes (Greek-to-Greek rewriting) tradition, and the glottonym+language-word collocations need not all mark translation - so the prediction's stricter verbatim reading ('translation into Greek from a named foreign language') yields <= 15 hits, ~0.12%, clearing the bar by ~4x. (ii) The one-sided undercount risk was disclosed at registration; the compute documented exactly one probable false negative (occ 20685, a genuine Arabic-to-Greek translation-prologue epigram missed by literal stem morphology) - including it gives 0.163%; breaching 0.5% would require ~45 further missed genuine marker epigrams against a 213-candidate raw sweep, which the registrant judges implausible; that lone caught Arabic-to-Greek prologue is the exception that displays the rarity. (iii) SECONDARY clause (>= 50:1 Greek-vs-foreign author naming) unresolvable per the registered fallback: DBBE's person schema carries only Byzantine-prosopography cross-references, no language/origin field; the tempting translation_language facet records modern scholarly translations (false lead, documented); geographic epithets were tested and rejected as a proxy (Ioannes of Damascus < Syria wrote in Greek). Per the registered precedence the primary verdict is unaffected; the secondary is recorded unresolved, not failed. (iv) The pinned manual-QC step was load-bearing exactly as the dbbe-014 lesson predicted: 93% of Rhos- raw hits were rhosis 'vigour', all Kopt- hits were kopto 'cut', plus syro/Perseus/phragellion collisions - without QC the raw rate would have been ~1.65%, a tenfold error. (v) Scope honesty: this is a record-altitude fact about Byzantine paratextual self-presentation - the page's silence about receiving - not proof that nothing crossed; the wave's other items measure what did. Blind compute claude-sonnet-5; verdict claude-fable-5.

OPEN prediction (anti-connection wager): fewer than 0.5% of DBBE epigram occurrences with full recorded text carry any explicit marker of translation into Greek from a named foreign language, and Greek-author naming outnumbers foreign-language-author naming by at least 50:1 - Byzantine book paratexts enforce the empire's self-image as source, never receiver.

Resolution criteria — the registered fine print

Resolution criteria: VERBATIM PREDICTION: 'fewer than 0.5% of DBBE epigrams with full recorded text contain any explicit marker of translation into Greek from a named foreign language (Syriac, Arabic, Latin, Georgian, Slavonic, or other), and epigrams naming Greek fathers or classical authors outnumber epigrams naming any foreign-language author by at least 50 to 1; coverage guard: at least 5,000 epigram occurrences with full text searched, else inconclusive.' OPERATIONALIZATION (pinned ex ante): Instrument = DBBE (dbbe.ugent.be), live at compute date; the compute records query date and the total count of occurrences with non-empty full text (snapshot pin). Unit = occurrence with non-empty full text; denominator = all such occurrences. A translation-marker hit = an occurrence whose text contains (a) a term of the pinned translation lexicon - stems metaphra-, metaphras-, metaglott-, or hermeneu-/hermenei- and metabal-/metabol- ONLY when the same occurrence also contains a pinned glottonym - or (b) a pinned language-source term - Syr-/Syriak-, Arabik-/Agaren-/Saraken-, Latinik-/Ital-/Phrang-, Iber-/Georgian-, Sklab-/Boulgar-/Rhos-, Hebraik-, Aigypti-/Kopt-, Pers-, Armeni- (Greek script; stem-match) - in a translation context, i.e. collocated in the same occurrence with glossa/phone/dialektos or a lexicon-(a) term. Rhomai- does NOT count as Latin (Byzantine self-designation). Every raw hit passes a manual QC read per the dbbe-014 collision lesson; each accept/reject is logged and may cite only the written rule, never the running tally. CLAUSE PRECEDENCE (ex ante): PRIMARY, verdict-bearing = the <0.5% marker rate. SECONDARY, qualifying only = the 50:1 author-naming ratio, computed from DBBE's person/author metadata associations (Greek patristic + classical persons vs foreign-language authors), method documented before counting; if the metadata cannot support it, the secondary is reported unresolvable with reasons and does not affect the primary verdict. VERDICTS: guard unmet -> INCONCLUSIVE-BY-COVERAGE; guard met and rate >= 0.5% -> KILLED; guard met and rate < 0.5% -> SUPPORTED (qualified by the secondary outcome).

Known-priors disclosure — what the registrant already knew

Known priors disclosure: Triaged adjacent 2026-07-20: that translation into Greek was rare is qualitative common ground in Byzantine studies; the pinned epigram-corpus census is unlocated. The registrant knows the qualitative direction and has computed nothing; no marker counts, rates, or DBBE totals are known. Neighbors disclosed: the byzantine-dbbe-pinakes wave's epigram items measure single-witness structure, renewal notices, and the post-1453 tail - different quantities; dbbe-014's lexical-collision lesson binds the QC step here. Risk disclosed: hermeneu- is polysemous (interpret/translate); the glottonym-collocation requirement is the ex-ante control, and its strictness may undercount - direction disclosed (undercounting favors the wager; the QC log makes this auditable).

Method and dataset — how it was measured

Register-before-compute held (registration 15:05:52Z, packet 9c516ba; compute launched after). The blind compute applied the registered marker rule - unconditional metaphra-/metaphras-/metaglott- stems; hermeneu-/metabal- stems conditional on a pinned glottonym in the same occurrence; glottonym+glossa/phone/dialektos collocations; Rhomai- never counts - with accent-normalized stem matching and a mandatory per-hit QC log. Raw sweep: 236 stem rows = 213 unique occurrences; accepted after QC: 20 unique occurrences. PRIMARY QUANTITY: 20/12,911 = 0.155% vs the registered < 0.5% bar; coverage guard (>= 5,000 full-text occurrences) met at 2.6x. Orchestrator desk-check: all 22 accept rows read individually; one accepted occurrence live-sampled (occ 17146: 'Metaphraste, syggrapheis ego lego') confirming the five bare-metaphrasis accepts are the Symeon Metaphrastes tradition.

Dataset: DBBE (dbbe.ugent.be), queried 2026-07-20 via its public search_api plus individual occurrence pages (bulk CSV export capped at ~1000 rows, documented and not used; no robots.txt exists - polite throttled access with identified UA). Population: 12,911 occurrences with non-empty full text, of 13,036 total - cross-validated two independent ways (empty-field subtraction; union of 24 letter-prefix queries).

computed 2026-07-20

Weigh in

No community feedback yet.

New here? Create an account first

Create an account or sign in and your feedback is tied to you — you can track it, get replies, and claim this conjecture so others know you’re working on it. Prefer not to? Just leave your take below as a guest — only the name you type is shown.

Add your take

Posted immediately (spam is removed). Community feedback is never an adjudicated verdict and never changes this conjecture's triage label or status above.