Prophetia Jeremiae

Columns 0009A–0018C: front matter (Jerome's prologue, PRAEFATIO, ARGUMENTUM) + CAPUT PRIMUM–V. 3,538 Latin words → 5,613 English (≈1.59×). One translator, 2026-08-29.

Parity, verified mechanically per chunk and again band by band: column anchors 1:1 in order and in place (each checked against the Latin word it stands beside, not merely present) · [n:] notes 1:1 verbatim · italic spans 1:1 outside markers · question marks 1:1 by band · paragraph/line structure 1:1 · frontmatter copied verbatim · zero guillemets in either column · zero ! in either column.

Apparatus fired: 10 [var:] · 1 [cj:] · 0 [sic:] · 0 [d:] · 0 [ed:] · 0 [nt:].

⚠ Superseded in part by §9 (repair pass, 2026-08-29), which also extends this file's coverage to chunks 0005–0007. After that pass the apparatus in chunks 0000–0007 is 14 [var:] · 2 [sic:] · 1 [cj:]; three conformed lemmata and three parses were repaired. Read §9 before relying on anything above it.

⛔ Corrected 2026-08-29 (reconciliation pass): the [var:] figure above was written as 18 and is 14. 18 is 11 (chunks 0000–0004, after §9d added the 0013B marker to the ten of §1a) + 7, but the 7 is cruces-0005.md's whole batch, chunks 0005–0009, and this file reaches only to chunk 0007. The three [var:] in chunks 0005–0007 are 0019D, 0022B, 0024D. Counted against the files: chunks 0000–0007 carry 14 [var:] · 2 [sic:] · 1 [cj:]. The old figure is corrected rather than deleted because the arithmetic, not the count of markers, was the error, and the next reader should see which two bands were being added. This file's own five chunks (0000–0004) stand at 11 [var:] · 2 [sic:] · 1 [cj:], against the 10 · 1 · 0 printed two lines above.

§0 · ⛔ PLATE-GATE DEBT — read these ten columns before this work is marked ours

data/plate-reads.json has no entry for 9006 and data/plate-notes/coverage.json records no spot check. Under CLAUDE.md hard rule 8a every [var:] below stands on an unread column and scripts/plate-gate.mjs 9006 will name all ten. They are listed here so the read is a half-hour's work and not an archaeology:

0010A · 0010C · 0011D (×3) · 0013C · 0014B · 0015C · 0017A · 0017D.

✅ DISCHARGED 2026-08-29, and this paragraph's premise is now false. The full plate gate was run before the work shipped — 24 pages of patrologiaecurs06saingoog at 300 dpi, three readers, recorded in data/plate-reads.json (entry 9006) and data/tei-patches/9006.json. node scripts/plate-gate.mjs 9006 now returns gate OK. The debt is kept on the page because the markers below were fired under it, and the reasoning in the next two paragraphs is why they were safe to fire before it was paid. ⚠ Two of the ten columns named above are named from an older anchor placement: the gate resolves them as 0014C and 0017B.

⚑ The line I drew, and why the markers were fired at all rather than left as cruces. The digitization's known ways of manufacturing a Vulgate divergence are single-letter alteration (concilio→consilio, judicabit→judicavit), single-word loss, and normalization. Eight of the ten [var:] rest on a divergence of several words or a whole different clause, which no OCR path produces. Every candidate that turned on one letter alone or on one omitted word alone was declined and logged in §2 instead.

⚠ The two exceptions, named so nobody has to find them: 0013C Adversatrix is a single letter and was fired only because Migne repeats it consistently four lines later (adversatricem), which a stray slip does not do; 0011D Gn 9:25 bundles a name-letter with an omitted word, neither of which would have earned a marker alone. If only two columns can be read, read those two.

§1 · THE DIVERGENCE LIST — every lemma and inline citation in bands 0009A–0018C

This is the 7a″ deliverable. Agreements are recorded as well as divergences, because a list of findings alone cannot be told apart from a report by an agent that never looked. Every entry was collated against sources/vulgate/clementine-flat.txt at the verse actually being glossed, not merely phrase-searched.

1a · Fired [var:] — ten

1b · Fired [cj:] — one

0012A · de cujus sarculo plantavit Israel. sarculum is a real word — a hoe — so Pattern 12 has nothing to wrap and Pattern 7 renders it. But the faithful English ("from whose hoe he planted Israel") asserts something the author did not: the sentence is about soreth, a kind of vine, and one plants from a shoot, surculo. Jerome's own commentary on Jr 2:21 has de cujus surculo. Additive marker per Pattern 18, Migne's word first and kept in the English: from whose hoe [cj: sarculo; read surculo, "shoot"]. The rival I weighed and rejected: reading sarculo straight as an instrument of planting ("with whose hoe"). It lost because cujus is the vine, not the planter, and a vine has no hoe.

1c · Checked and found to agree with the Clementine EXACTLY

Every lemma below was read word by word against its own verse and diverges in nothing:

Jr 1 — 1:1 Verba Jeremiae · 1:2 In diebus · 1:5 Priusquam te (fragment; ⛔ the expansion "Before I formed thee" was WITHDRAWN in the repair pass — see §9a — and the lemma now reads "Before thee") · 1:6 A, a, a, Domine Deus · 1:9 Et misit · 1:13 Et factum est, Ollam succensam · 1:14 Habitatores terrae · 1:17 Tu ergo accinge · 1:18 Et in columnam, Regibus Juda · 1Tim 3:15 Columna et firmamentum veritatis.

Jr 2 — 2:2 Adolescentiam · 2:3 Omnes qui devorant · 2:4 Israel · 2:5 Quid invenerunt patres · 2:6 Neque habitavit homo · 2:7 Et induxi vos · 2:10 Cethim, Et considerate · 2:11 Populus meus · 2:12 Obstupescite, coeli · 2:13 Me dereliquerunt, Fontem aquae vivae · 2:14 Nunquid servus · 2:15 Leones · 2:16 Usque ad verticem · 2:19 Arguet te malitia tua · 2:20 A saeculo confregisti · 2:21 Vineam electam, Quomodo ergo conversa es? · 2:22 Borith · 2:23 Vide vias tuas · 2:25 Prohibe pedem tuum · 2:28 Secundum numerum · 2:29 Quid vultis? · 2:30 Filios vestros · 2:32 Nunquid obliviscetur? · 2:34 In omnibus · 2:35 Et dixisti, Ecce ego judicio · 2:36 Et ab Aegypto · 2:37 Super caput tuum · Gn 1:26 Faciamus hominem ad imaginem et similitudinem (truncated, not altered) · Is 1:6 A planta pedis usque ad verticem (verbatim) · Ex 20:2 Ego sum Dominus Deus tuus qui eduxi te de terra Aegypti (truncated before de domo servitutis, not altered).

Jr 3 — 3:1 Vulgo dicitur: si dimiserit, Si dimiserit vir, Tu autem fornicata es, Tamen revertere · 3:2 Leva oculos tuos, Quasi latro · 3:3 Frons mulieris · 3:4–5 Dux virginitatis meae tu es. Nunquid irasceris in perpetuum · 3:6 Nunquid vidisti quae fecerit? (both occurrences) · 3:7 Et vidit · 3:11 Justificavit animam · 3:12 Et non avertam faciem · 3:13 Verumtamen scito · 3:14 Convertimini · 3:16 Cumque multiplicati fueritis · 3:17 Vocabunt · 3:18 In diebus illis · 3:19 In filios · 3:20 Mulier amatorem · 3:21 Vox in viis audita est · 3:22 Aversiones vestras · 3:25 Dormiemus · 2:35 non peccavi (quoted back from the previous chapter, correctly) · Lc 18:14 Descendit hic justificatus in domum suam ab illo (verbatim) · Ps 50:11 Averte faciem tuam (verbatim) · Ps 79:2 Qui sedes super Cherubim, manifestare (verbatim).

Jr 4 — 4:1 Si reverteris, Si abstuleris · 4:2 Et jurabis, vivit, Novate · 4:4 Circumcidimini, Ne forte · 4:7 Ascendit leo, ascendit leo de cubili suo, Et praedo · 4:8 Super hoc · 4:9 Principum, Sacerdotes, Prophetae consternabuntur, peribit cor regis · 4:11 Ventus urens · 4:12 Spiritus plenus, Et nunc ego · 4:13 Ecce quasi, Vae nobis quoniam vastati sumus (all three Vae nobis) · 4:14 Lava a malitia, Lava · 4:15 Vox enim annuntiantis, De monte · 4:16 Custodes · 4:18 Viae tuae · 4:19 Ventrem meum · 4:21 Usquequo videbo · 4:22 Quia stultus populus, Sapientes sunt · 4:23 Aspexi terram · 4:27 Haec enim dicit · 4:28 Cogitavi et non poenituit me · 4:29 A voce equitis · 4:30 Tu autem vastata, Cum vestieris · 4:31 Vocem enim quasi · 3:17 in tempore illo vocabunt Jerusalem solium Domini (quoted back, verbatim) · Ps 72:2 Mei autem pene moti sunt pedes, pene effusi sunt gressus mei (verbatim) · Ps 9:26 Omnium inimicorum suorum dominabitur (verbatim — and Migne's (Psal. IX) is right) · Jn 14:30 Venit enim princeps mundi hujus, et in me non habet quidquam (verbatim) · Is 1:16 Lavamini, mundi estote (verbatim).

Jr 5 — 5:1 Circuite vias · 5:3 Domine, oculi tui, Percussisti eos, Et non doluerunt · 5:4 Forsitan pauperes sunt · 5:5 Ecce magis, Vincula · 5:6 Pardus vigilans · 5:7 Saturavi eos · 5:8 Equi amatores · 5:10 Et dissipate · 5:12 Negaverunt Dominum · 5:14 Ecce ego do · 5:15 Ignorabis linguam · 5:17 Et comedet segetes · 5:18 Verumtamen · 5:19 Quod si dixeritis · 5:21 Audi, popule stulte, Qui non habes cor, Qui habentes · 5:25 Iniquitates vestrae · 5:27 Sicut decipula · 5:28 Incrassati sunt, Et impinguati · Ps 2:3 Dirumpamus vincula eorum et projiciamus a nobis jugum ipsorum (verbatim; Migne's (Psal. I) is his, untouched) · Ps 113:13 oculos habent, et non videbunt and Ps 113:16 Similes illis fiant qui faciunt ea, et omnes qui confidunt in eis (both verbatim) · Dt 32:15 Incrassatus, impinguatus, dilatatus, dereliquit Deum factorem suum and Incrassatus est dilectus et recalcitravit (both verbatim, and note Migne quotes the verse's two halves in reverse order — that is his arrangement, reproduced, not a divergence).

§2 · DIVERGENCES FOUND AND DECLINED — the marker was withheld, the finding is here

2a · Declined because the divergence is ONE LETTER

• 0012C · Jr 2:31 Videre verbum Domini — the Clementine has Videte ("See ye"), an imperative; Migne's is an infinitive standing as a lemma title, which is odd but construes ("To see the word of the Lord"), and the gloss below (Moyses videbat vocem Domini) works with either. t→r is exactly the single-letter alteration class the digitization is documented to produce. Rendered as printed, no marker. ⭐ Top plate candidate of this batch after the two in §2c: one look at col. 0012C settles it, and if the plate reads Videte the fix is a data/tei-patches/9006.json entry, not a [var:].
• 0011D · Gn 9:25 Chanaam for Chanaan. Carried inside the Gn 9:25 marker rather than standing on its own, and flagged there. A proper-name letter is the weakest possible independent claim.
• 0015A · Job 2:2 Circumivi for Circuivi — both are legitimate forms of circumeo. Pattern 9: no defect, no marker, no crux beyond this line.
• 0011A · Jr 2:12 coeli for caeli — the flat Clementine expands ligatures to ae; Migne's oe is PL house orthography throughout. Not a divergence in any sense. Likewise Nunquid/Numquid at Jr 2:14, 2:32, 3:5, 3:6 — orthographic, four occurrences, all silent.

2b · Declined because the divergence is a single OMITTED word (a live digitization-loss risk)

• 0010A · 1Ptr 1:1 — Migne: Electis advenis Ponti…; Clementine: electis advenis dispersionis Ponti…. One word absent. It does not change what the gloss argues (the scattering of the habitatores terrae), so the marker fails Pattern 14's "changes the sense" test on its own terms as well.
• 0011D · Gn 9:25 — servus for servus servorum. This is the one I came closest to firing alone: the omission flattens the famous phrase. It is inside the marker's text because the marker was fired for the whole reading, but the omission by itself would not have earned one.
• 0011C · Rom 6:19 — sicut exhibuistis for sicut enim exhibuistis. Connective.
• 0015C · Jn 1:16 — De plenitudine ejus omnes accepimus for Et de plenitudine ejus nos omnes accepimus.
• 0014B · Jn 1:12 — Quotquot enim for Quotquot autem. A connective substitution, not an omission, so my own rule would license a marker; declined because nothing in the sense moves, and because 8960's Jdt 7:7 autem/tamen lean is the standing warning against exactly this reflex.

2c · Declined because the form is a NON-WORD — raise at the plate, never accuse Migne

Addendum A is unambiguous: every suspicion aimed at a non-word has been confirmed as our corruption, and seven false [sic:] were withdrawn in one batch for publicly blaming Migne for the digitization's error. So both of these are carried into the English exactly as printed, with no [sic:], and named here as plate candidates.

• ⭐ 0011B–C · Velut anthropophormitae — no Latin word; a metathesis of anthropomorphitae (the Anthropomorphite heretics), morph→phorm, the same shape as 8970's Nephthali→Nephtlali. Hapax in src/latin/. Carried as plain roman anthropophormitae — not italicised, because italic Latin in our English already means "our editorial choice of a technical term", and because adding an italic span would have broken the 1:1 italic structure. The relative clause after it tells a reader it is a sect name. If the plate reads anthropomorphitae, this is a tei-patch; if it reads as ours, then and only then a [sic:].
• ⭐ 0014B · Jr 3:19 gloss opens i In numero filiorum — a stranded lower-case i with no grammatical slot. Carried verbatim as i, unmarked. Two conjectures, and I adjudicate rather than merely naming them: i = an abbreviated id est (Migne uses id est constantly in this work and the phrase that follows is exactly an id est apposition to the lemma In filios) beats a stray broken sort, because a broken sort would not land in the one slot where a discourse marker belongs. Neither conjecture entered the text; a plate read decides which repair applies. 7a⁗ check: the clause loses nothing — In numero filiorum qui de gentibus crediderunt is a verbless apposition either way, and no negation, conjunction, interrogative or object sits inside the carried token.

2d · Declined because Migne NAMES the other witness himself

Pattern 14 marks a Vulgate divergence. Where Migne says which text he is quoting, there is no claim to make:

• 0009A · LXX: Verbum Dei (Jr 1:1) and 0011C · LXX: Obstupuit coelum (Jr 2:12) — labelled LXX.
• 0014D · Septuaginta: Si conversus fuerit Israel, ait Dominus, ad me convertatur (Jr 4:1) — labelled.
• 0014D · Alia littera, converteris — Migne's own "another reading". ⚠ The word is not italic in the Latin, so it is carried untranslated in plain roman rather than being given an italic span our source does not have. Adding one would have broken italic parity.
• 0015B · Aquila, amentes erunt (Jr 4:9) — Aquila, named.
• 0018A · In Hebraeo: Non intelliges quid loquatur (Jr 5:15, where the Clementine has nec intelliges) — labelled In Hebraeo.
• 0012A · In Hebraeo, soreth — the Hebrew word, carried as printed. (The Clementine's Latin at Is 5:2 has sorec; irrelevant, since Migne is transliterating, not quoting.)

2e · Scripture quoted OUTSIDE any italic span — recorded, not marked

• 0015D · Cum conversus ingemueris, salvus eris — the Old-Latin/LXX shape of Is 30:15, where the Clementine reads Si revertamini et quiescatis, salvi eritis. A substantial divergence, and I would have fired on it — except that Migne sets it as running prose with no italics and no reference, so the page does not present it as a citation and a [var:] would be asserting a citation he did not make. Rendered literally; recorded here.
• 0013D · Justificata est Sodoma soror tua ex te (after Ez 16:52) — same treatment, same reason. Not italic, no reference.
• 0012C · Joannes supra pectus Domini in coena recubuit — italic, but it is the glossator's own sentence about Jn 13, not a quotation of it (the nearest Vulgate wording is Jn 21:20 qui et recubuit in coena super pectus ejus). Pattern 14 does not reach a paraphrase.
• 0015B · Stultam enim fecit Deus sapientiam hujus mundi (1Cor 1:20) — the words are verbatim; Migne has turned Paul's Nonne … ? into a statement with enim. A change of sentence type, not of wording. Recorded; no marker; and no ? supplied (Pattern 8).
• 0014A · Rom 9:29 — sicut Sodoma essemus, et similes Gomorrhae fuissemus for sicut Sodoma facti essemus, et sicut Gomorrha similes fuissemus. Word order and one dropped participle; the sense is untouched. Declined.
• 0013D · Ps 13:4 — Qui devorant populum meum for plebem meam. Declined for a reason worth stating: English has no exponent for it. Both render "my people", so a marker would annotate a distinction our page cannot show, which is Pattern 9's logic rather than Pattern 14's.
• 0018C · Lc 12:19 — Anima, multa bona habes posita… for Anima, habes multa bona posita…, and truncated before epulare. Word order + truncation. Declined.
• 0018C · Omnis dives, aut iniquus, aut haeres iniqui — Migne calls it illam philosophorum sententiam; it is not scripture and takes no Vulgate collation.

§3 · MIGNE'S OWN PRINTING — reproduced, not tidied

• 0014A · VERS. 17. Vocabunt. — the verse address prints without the -- that every other address in the work carries. Reproduced exactly. Not a defect of ours and not corrected.
• 0014B · VERS 22.-- — and here without the period after VERS. Reproduced exactly. Two different malformations of the same address, four columns apart, which is itself mild evidence that they are the compositor's and not systematic.
• 0009A–B · PRAEFATIO prints with no closing period where ARGUMENTUM. and every CAPUT head has one. The English head PREFACE matches it, period for period.
• 0015B · cum principes et reges et sacerdotes quondam suos sub leone conspexerit. — a plain rhetorical question (Quis enim non insaniat…) that Migne closes with a full stop. Preserved. No ? supplied; Pattern 8, and it is the reason the band check is run rather than the total.
• 0015B · Et dixi: Heu, heu, heu, Domine Deus. — the Clementine prints Heu ! heu ! heu ! Migne prints commas. Punctuation, not wording; his marks ride as printed and the batch carries zero ! on either side.
• 0015C · Virgil, Quos ego; sed motos praestat componere fluctus — Aeneid 1.135, the locus classicus of aposiopesis, and modern texts print a dash. Migne prints a semicolon. Preserved as a semicolon. (No em-dash was introduced anywhere in the batch.)
• 0015A · [n: (Job. XIV.)] keyed to a quotation of John 14:30. Migne's reference, and it passes through verbatim and untranslated like every other. Not corrected — and note a following [n: (Ibid.)] chain exists in this work (runbook false positive 2).
• 0009C, 0014B · bare italic (ID) (no period, no [n:] wrapper) where the same siglum is elsewhere [n: (ID.)]. Reproduced as an italic span, untranslated, exactly as printed.
• 0009A · [n: (Vide Operum ejus tomo IX, col. 847.)] kept as [n:], not converted to [nt:]. Judged by content per Pattern 15: it is a locator with an imperative, not a statement. Corpus control: Vide … notes are shipped untranslated as [n:] in the existing English (Vide inter Opera Hieronymi, tom. III., Vide scholia ad calcem opusculi., Vide tomo VIII operum.). No new convention invented.

§4 · READINGS THAT CONSTRUE ODDLY AND ARE MIGNE'S — rendered as printed

Addendum A: a real, well-formed word that reads oddly is Migne's until a plate says otherwise. None of these took a marker.

• 0009B · defugatis (postulatione sua defugatis ab eodem loco serpentibus) — hapax in src/latin/, and fugatis is likewise a hapax, so frequency says nothing. But defugare is attested medieval Latin (Du Cange) and the form is well-made. Pattern 9: rendered "driven away", no italics, no marker.
• 0009A · Non errore neque reprehensione superiori, sed certo consilio — reprehensione superiori is the hardest phrase in the batch. Rendered literally, "nor by any higher censure". Two readings construe and I could not choose: superiori as "from above / on high", or superiori as "the foregoing" (the censure mentioned earlier — except that nothing earlier censures anything, which is what makes it hard). Neither reading was smuggled into the English; the rendering is neutral between them. Negation hosts proved from the sed certo consilio correlative: non is on errore, neque on reprehensione superiori, and neither is on intelliguntur.
• 0017D · Audiat hoc qui acceptis divitiis luxuriae serviunt — a number disagreement: singular Audiat … qui against plural serviunt. Rendered so that the disagreement survives into English — "Let him hear this who, riches having been received, serve luxury" — rather than being repaired by a plural "them", which is the silent-repair move Pattern 12's second test names (rendering that hides the defect). English does have an exponent here (him/serve), so Pattern 7 governs and no marker is needed.
• 0013B · Tamen revertere. Si revertebaris ad me, dicit Dominus. — Jr 3:1 reads tamen revertere ad me, dicit Dominus, an imperative; the gloss opens with an imperfect indicative, Si revertebaris, which yields "If thou wert returning to me, saith the Lord" and hangs without an apodosis. Two conjectures weighed: (a) Sive: revertere ad me, on the model of the Sive: that opens the immediately preceding gloss in the same column, and (b) Jerome's alternative rendering si reverteris ad me*. Neither was taken into the text and the printed reading stands, because (a) requires deleting a syllable Migne prints and (b) requires changing tense — and the two conjectures pull in different directions, which is itself the reason to render and log rather than choose. Plate candidate, low priority.
• 0013C · quae ad nullum declinare erubescit — carried with its negation intact ("who is ashamed to turn aside to no man"), which is the double-negative Latin idiom for shamelessness and reads as a contradiction in English. That is the required treatment: 7a forbids translating a word as its opposite to rescue a sentence.
• 0012B · convallem filiorum Ennon (for vallis filii Ennom, Jos 15:8) · 0009B · Thamnas (the usual form is Taphnas) · 0016A · Paneas · 0016A · circumitu for circuitu. Proper names and an attested orthographic variant, all in gloss prose rather than in a lemma. Rendered as printed; Pattern 9; no markers.

§5 · THE usque ad LOOK-ALIKES — decided per occurrence, as addendum D requires

usque ad occurs 44 times in these five chunks (43 lower-case plus the capitalised lemma at 0011D). 41 are the abridgment formula, rendered as far as in their own italic span per the settled convention (8978 → 8970; 1,361 shipped instances corpus-wide). No verb was supplied to complete any of them, and where the resumption lands mid-clause the English lands mid-clause.

Three are NOT the formula and were caught by reading what stands on either side:

1. 0011D · Jr 2:16 lemma Usque ad verticem. — the verse's own words. Rendered "Unto the crown of the head."
2. 0011D · Is 1:6 A planta pedis usque ad verticem — the verse's own words, inside a quoted citation, and the very next usque ad in the same sentence IS the formula. The two stand nine words apart. This is the trap addendum D was written about.
3. 0015C · Jr 4:10 pervenerit gladius usque ad animam — the verse's own words, again followed by the formula in the same sentence.

§6 · NEGATION AND THE UNEXPRESSED SUBJECT — the sites I proved from structure

Every non · nec · neque · nisi · nunquam · nullus · nihil · ne · nequaquam · nolle in the five chunks is in the English, and none is in the English that is not in the Latin. Test 2a's trigger population — a negative resting near a verb with a quantifier, correlative or sed pivot — fired at these sites, and in each the host was proved from the sentence's structure, not from what reads well:

Unexpressed subjects supplied — the antecedent named, per site. In a gloss the antecedent is usually in the lemma or the verse above, not in the sentence:

• 0009C nihilominus populum ad poenitentiam vocat → "he calls" = God, from Mirabilis Dei clementia two clauses earlier.
• 0009C Non quod ante conceptionem fuerit → "he existed" = Jeremiah, from the lemma Priusquam te (Jr 1:5, of Jeremiah).
• 0010A sed ille … corripitur, quasi puero huic venia datur → "the one" = Moses, named in the preceding clause; "this one" = Jeremiah, the puer of Jr 1:6.
• 0011D pro quo secutus est daemones → "he followed" = the people, from Populus meus / Me dereliquerunt above; the speaker in the same sentence (Qui praeceptum dedi) is God, and the subject changes at secutus est.
• 0015B turbatur propheta, et in se putat Deum mentitum → the prophet, printed.
• 0016B Loquitur hoc Dominus → printed.
• ⚠ 0018C venientium ad nos, non manus contemplentur, sed ora — two readings construe and I did not choose silently. contemplentur is third plural, so the subject is either indefinite ("they/men") or, taken from the preceding et non a nobis, "we" of the Church, in which case the third-plural is a loose generalising shift. The English uses "they", which is the reading nearer the printed morphology; the rival is recorded here.

§7 · SECTIONS READ CLEAN

Named so a blind reader can tell a checked passage from an unvisited one:

• The PRAEFATIO (0009A) in full — the LXX-vs-Hebrew argument, which is the passage most exposed to 7a″ conformation because it is about textual divergence. Nothing in it is a scripture quotation; the only crux is reprehensione superiori (§4). utraque … firmata est is singular by construction, not a disagreement.
• The ARGUMENTUM (0009B) in full — Jeremiah's death and Egyptian cult. One flagged word (defugatis), no quotations, no anchors inside it.
• CAPUT PRIMUM entire (0009B–0010C) — 12 glosses, 6 anchors, 8 notes, zero question marks in either column.
• The whole of chunk 0004 (CAPUT V, 0017B–0018C) carries zero question marks on both sides — checked band by band, not by total, and worth stating because a supplied ? in the rhetorical Quis enim non insaniat family is the exact defect this book invites (see §3, and the parallel site at 0015B in chunk 0003).
• All 44 usque ad occurrences individually classified (§5) rather than swept — the count was taken mechanically per chunk against the English as far as count, which is what exposed the capitalised lemma the lower-case grep had missed.
• Fourth test (does the English ASSERT what he asserts) run paragraph by paragraph on all five chunks with the counts already known clean. It produced two changes and no third: the sarculo [cj:] (§1b), where the containment checks all passed and the sentence was simply false about vines; and the negation host at 0017A, where "great is the love of justice, that God does not free the city for ten just men" reads perfectly and is not what the sed correlative permits.

§8 · CONVENTIONS APPLIED WITHOUT RE-OPENING THEM

VERS. n.-- passes through verbatim · the lemma is Englished · a fragmentary lemma is rendered as the phrase it stands for — ⛔ but see §9: the Priusquam te. expansion was CONFORMATION, not convention 3, and has been withdrawn (0010A Regibus Juda. → "To the kings of Juda", dative, from Jr 1:18, stands: nothing is imported there, the dative is in the printed word) · sigla ride as [n:] verbatim, unanglicized, and passing one through is not endorsing it · heads translated 1:1, matching Ruth's shipped forms (CAPUT PRIMUM. → CHAPTER ONE., CAPUT II. → CHAPTER II.) and 8977's for the prologue (PROLOGUS HIERONYMI. → THE PROLOGUE OF JEROME., so PROLOGUS HIERONYMI IN JEREMIAM. → THE PROLOGUE OF JEROME ON JEREMIAH.; ARGUMENTUM. → THE ARGUMENT.) · Pattern 17 observed throughout, and the batch has a worked contrast at 0012C, where disceretis … diceretis (plural, the sons' generation) sits in the same sentence as noluisti (singular, addressed to God): "that you might learn … lest you should say: Thou wouldst not rebuke us."

⛔ Nothing was copied from Ruth (8968) except the head forms and the settled conventions; no rendering was carried across from another book.

§9 · REPAIR PASS, 2026-08-29 — the conformation class, the two plate patches, and four new cruces

Chunks 0000–0007 (cols 0009A–0025D) re-read against the re-chunked Latin (14 TEI patches of 2026-08-29; two of them fall in this range). Sections 0–8 above were written before those patches and before the blind polarity read; nothing in them is deleted, but where §8 recorded a rendering that has now been withdrawn it is corrected in place and the reason is here.

⚠ Scope note. This pass owns chunks 0000–0007, so it reaches two columns documented in cruces-0005.md (0019A in chunk 0005 and the chunk-0006 candidates). Those entries are recorded here, not there, because cruces-0005.md is another agent's file.

9a · ⛔ THE LEMMA-EXPANSION FINDING — DECLINED AND REVERTED; only the conformation site stands

⛔ Correction, same day, after the ruling was applied. Two of the three repairs below were WRONG and have been undone. The entry is kept, not deleted, so that the next reader sees the finding, sees why it fails, and does not raise it a fourth time.

What was originally written here (2026-08-29, first repair pass, now superseded): three blind polarity readers reported a class they called "the lemma silently completed from the Vulgate" — Migne prints a short lemma, our English carries the whole phrase — and this section ordered the lemmata contracted back to word-for-word stumps. That repair pass was stopped mid-run.

The governing ruling (GLOSSA-NT-AGENT-BRIEF.md, final section, Before you call something inconsistent, NAME THE CONTROL'S CLASS):

A clipped lemma whose governed word is missing and a complete one-word lemma are different classes. The first is a fragment and convention 3 renders it as the phrase it stands for; the second has nothing elided and is rendered as it stands. Contrasting the two manufactures an inconsistency that is not there — three readers did exactly that in one batch, and all three findings were declined. … the defect is never expansion or its absence, it is CONFORMATION — following the familiar English Bible where Migne's own plate diverges from it. Expanding a fragment per convention 3 is not conformation. Where a work really does treat the same bare lemma two ways, resolve it by EXPANDING the stump, not by contracting its twin.

Convention 3's own model (brief, The Glossa conventions): Ruth's VERS. 3.-- Non te. — Ruth 3:3, non te videat homo — is rendered "Let no man see thee", not "Not thee."

⚠ Where the reasoning above went wrong. It read Pattern 8's "never supply" as outranking convention 3. It does not. Pattern 8 governs content the plate does not stand for; a clipped lemma does stand for its verse-phrase, and supplying that phrase is what convention 3 is for. The list of correctly-short fragments cited as proof (Verba Jeremiae, In diebus, Et factum est, Ab extremitatibus, Lignum, Borith, Grex) is a different class — complete one-word or self-standing lemmata with nothing elided — and so cannot control clipped ones. That is exactly the class error the ruling names.

⛔ This is now the third time this non-defect has been raised by fenced readers. It is settled. Do not re-raise it, and do not accept a finding that contrasts a clipped lemma with a complete one.

⭐ 0016C is the only genuine defect of the three, and it is genuine because it is CONFORMATION, not because it was long. Sapientes malitiosi is not a clipped lemma at all: it is the head of an abridged gloss, and Migne prints an adjective. "Wise to do mischief" is Douay's rendering of Jer 4:22's sapientes sunt ut faciant mala — a phrase Migne's plate does not carry — so the English was following the familiar English Bible where the plate diverges from it. The proof is the gloss's own resumption tail, non sapientia, sed versutia et calliditas vocatur, which glosses the adjective and was left explaining a word the English no longer had.

⚑ The test to apply, in order. (1) Name the class: is the lemma clipped (a governed word missing) or complete? (2) If clipped, convention 3 applies and the phrase it stands for is the right rendering — say so and stop. (3) A defect exists only if the English follows an English Bible against what Migne prints. (4) If one work renders the same bare lemma two ways, expand the stump; never contract its twin.

⚠ Sites restored in the corrective pass (2026-08-29, second pass), for the record — 25 in all across chunks 0000–0023; the two in this file's range are 0009C and 0019A above. The other 23 lie in chunks 0008–0022, and the rulings that license them were already written, correctly, by the batch agents before the contraction pass ran: see cruces-0005.md §2d(ii) (Sicut frigidam · Si mutare potest · Nunquid avis discolor), cruces-0010.md §1b (the eleven bare fragments of cols 0031B–0044D, including the three Si non reliquiae rendered alike), cruces-0015.md §1 (fragments rendered as the phrase they stand for) and cruces-0020.md §3 (Ad filios · Confusa est · Dissoluta · Subito cecidit · Stultus factus est · Confusus est). Those sections need no correction — the contraction pass contradicted them, and the restoration puts the English back in agreement with them. The 0055D ejus repair (chunk 0019, "their" → "his") is conformation and stands, like 0016C.

9b · PARSE AND SUBJECT REPAIRS — three sites where the English did not assert what Migne asserts

Fourth-test class (brief §5): everything was present, in order, at parity, and the page said something else.

• 0013A (chunk 0002) — Habuisti pastores multos in offensionem tui, was "unto thine own hurt". The gloss that follows exists solely to fix the genitive as subjective — ut Deum scilicet offenderes — so as written the second half explained something the first half had not said. Now "unto thy giving of offence, that is, that thou mightest offend God."
• 0012D (chunk 0001) — Audiat haeresis nova ex veteri, iram Dei esse vel maximam, nolle peccatum humiliter confiteri, sed impudenter jactare justitiam. The English had made three coordinate objects of audiat; the Latin predicates iram Dei esse vel maximam of the infinitive phrase that follows. Now "…hear that it is the very greatest wrath of God, to be unwilling to confess sin humbly, but shamelessly to boast one's righteousness." The sentence's actual claim is no longer floating as an apposition.
• 0017D (chunk 0004) — Audiat hoc Ecclesia negligens: et providentiam Dei refutans, quod et gladium et famem sustineat… The English had followed Migne's stray colon into inventing a second hearer ("the negligent Church, and that which rejects the providence of God"). refutans is a nominative participle agreeing with Ecclesia: one hearer. Now "Let the Church, negligent and rejecting the providence of God, hear this: that she shall suffer…" ⚠ Migne's colon is kept exactly as printed — the English was fixed, not his punctuation; the 0017D band still carries one colon on each side.

9c · TWO NEW [sic:] — both restored by TEI patch, both Migne's own defect

Corpus Corporum had silently supplied the correct form in each case, so our Latin read correctly and no check we own could see it (brief §4, the normalization direction). Both are plate-verified in data/tei-patches/9006.json; these are the only two [sic:] in chunks 0000–0007 and the first in this work.

• 0011D (chunk 0001) — the Latin now prints fratibus (no r) in Maledictus Chanaam, servus erit fratibus suis; plate p.11 at 9×, clean type. English: "a servant shall he be to his brethren [sic: fratibus] [n: (Gen. IX)]".
• 0015C (chunk 0003) — the Latin now prints Quos egos in the Virgil tag (Aen. 1.135, correctly Quos ego—); plate p.13 at 5×, the final s a full sort distinct from the semicolon. English: "Whom I [sic: egos]; but it is better to settle the troubled waves."

⛔⛔ BOOKKEEPING GAP — RAISED HERE, AND SINCE CLOSED. As written this section reported that the 2026-08-29 plate gate lived only in data/tei-patches/9006.json and that data/plate-reads.json had no 9006 entry at all, so plate-gate.mjs saw 0 recorded reads and refused all 44 markers then standing. ✅ The entry was written. data/plate-reads.json now carries works.9006 — 24 reads, pp. 10–36, each with the page, the source, the corner-number calibration and what was found — and node scripts/plate-gate.mjs 9006 returns gate OK on all 50 plate-claiming markers (37 [var:] + 13 [sic:]; [cj:], [ed:] and [nt:] make no claim about the plate and the gate does not read them). The warning is corrected in place, not deleted: it is the record of a real gap between the evidence and the file the gate reads, and that gap is exactly the failure mode hard rule 8a exists to catch.

⛔ Neither marker swallows the word it marks. Strip every marker from both sentences and what is left reads complete — "a servant shall he be to his brethren", "Whom I; but it is better…" — no stranded article, no stranded preposition, no doubled mark. And the fourth question 7a⁗ does not ask: the clause still has the thing it acts on — fratibus suis is the dative serviet governs and it is present as "to his brethren"; the Virgil tag is an aposiopesis by Migne's own word for it, so "Whom I" is complete precisely by being incomplete, and its object quos is rendered. [sic:] content is verbatim in the Latin twin in both cases (fratibus, egos).

9d · ONE NEW [var:] — fired for parity, verified against the file before writing

• 0013B (chunk 0002) — Migne prints Si revertebaris ad me, dicit Dominus where the Clementine (Jr 3:1, checked in sources/vulgate/clementine-flat.txt, not from memory) reads …tamen revertere ad me, dicit Dominus, et ego suscipiam te. An imperative in the received text against Migne's indicative behind si, which is a change of what the sentence does, not of its wording only — and the gloss that follows (Arguit impudentiam meretricis) reads on either. The English was already literal and correct; it fired no [var:] where the same translator fired ten elsewhere in these chunks, so this is a parity repair, not a new reading. The lemma above it, Tamen revertere., agrees with the Clementine exactly and is left unmarked.

9e · ONE NEW CRUX, NO MARKER — In divinis confisi (0018C, chunk 0004)

Migne prints Et impinguati, etc. In divinis confisi: quasi dicat: Anima, multa bona habes posita in annos plurimos… The gloss attaches the Lucan rich fool, so the presumption at issue is on wealth, and divinis is very likely the plate's slip for divitiis. The literal rendering "Trusting in divine things" is correct and stands (Pattern 7).

⛔ No [sic:], deliberately. The column was read at the plate in the 2026-08-29 gate and Migne does print divinis: it is a real, well-formed word, so the defect is his sense, not his type as we can prove it, and addendum A is explicit that a real word reading oddly is Migne's until a plate says otherwise. A [cj:] was weighed and rejected as well — Pattern 18 requires the conjecture to be one the printed word's own context forces, and here the rich fool is adduced for impinguati as much as for confisi, so divinis is odd rather than false. A crux asserts nothing and is always safe; this is where the finding lives.

9f · OPEN QUESTION LEFT FLAGGED, ENGLISH UNCHANGED — 0021B (chunk 0005)

Ne dubitetis indurare cervices, etc., usque ad nunc saltem loquere eis verbis meis, glossing Jr 7:27 (Et loqueris ad eos omnia verba haec, et non audient te). The English reads "Hesitate not to harden your necks" and is left exactly as it stands. Two readings, both construing:

1. Ours, as printed: dubitetis second plural + prolative infinitive — "hesitate not to harden your necks", an ironic imperative addressed to the people.
2. The rival: accusative-and-infinitive with an unexpressed subject — "do not doubt that they harden their necks" — addressed to the prophet, which is what Jr 7:27 is about (et non audient te) and which fits nunc saltem loquere eis verbis meis, an imperative plainly addressed to Jeremiah.

⛔ Neither was chosen silently, and neither could be settled from this repo. The reading turns on what Jerome writes at the corresponding place in In Hieremiam (PL 24), which the abridgment is cutting into and which is not on disk here. The rival is the stronger on sense; the printed words are the stronger on syntax, since dubito + bare infinitive is Jerome's ordinary idiom and the acc.+inf. would need a subject the plate does not print. Pattern 7 breaks the tie for the running text. Whoever gets PL 24 should settle this and it is the one open reading in these eight chunks.

9g · NEW NON-WORD CANDIDATES FOUND IN CHUNK 0006 — plate read requested, NO markers fired

Chunk 0006 (cols 0021D–0024A) was not in the 2026-08-29 plate gate — that read covered the 37 fired markers, and chunk 0006 fires only one [var:], at 0022B. Four non-words stand in its Latin, none of them named in the findings I was given, all four read through in the English:

⛔ No [sic:] on any of them, and that is the ruling, not an omission. Addendum A: every suspicion aimed at a non-word has been confirmed as our corruption, and seven false [sic:] were withdrawn in one batch for publicly blaming Migne for the digitization's error. All four have the exact shape of Corpus Corporum damage (Qui dam, quorum dam, quisbusdam), and Jesusalem in particular is a dittographic OCR slip no compositor makes. Ship test B: an unread candidate is safe iff it fires no marker. If the plate reads as ours in any of the four, that one is owed a tei-patch and then a [sic:]; if it reads correctly, our Latin needs the repair and the English is already right. The same holds for the two candidates §2c already named at 0011B–C (anthropophormitae) and 0014B (i).

⚑ These four are worth the reader's half-hour precisely because chunk 0006 has no plate cover at all — the gate's coverage follows the markers, so a chunk that fires nothing is a chunk nobody has looked at.

9h · PARITY RE-CHECKED AGAINST THE FRESH LATIN TWINS

The Latin word count changed by design (patches restore Migne's own defective forms), so Latin↔English ratios shift slightly; anchors and notes did not move. Re-measured after every edit above:

Column anchors verbatim, in order, and checked in place against the Latin word they stand beside (test 2b), including the two the repairs moved text around — 0013B and 0017D both still open on the word they opened on. scripts/verify-english.mjs 9006 passes on all 24 chunks. Italic spans counted with markers stripped, so the italics inside the two new [sic:] do not enter the count and structure stays 1:1. Apparatus in chunks 0000–0007 after this pass: 18 [var:] (17 standing + 1 new at 0013B) · 2 [sic:] · 1 [cj:] · 0 [d:] · 0 [ed:] · 0 [nt:] · 0 « ».

9i · READ CLEAN IN THIS PASS, AND WHAT I LOOKED FOR

• Every lemma in chunks 0000–0007 re-read word by word against the fresh Latin for the 9a class specifically. Beyond the three repaired, no fourth conformation — checked hardest at the fragments most likely to invite one: A, a, a, Domine Deus (0010A), Tu ergo accinge (0010B), Quid invenerunt patres (0011A), Vulgo dicitur: si dimiserit (0013A), Lava a malitia (0015D), Circuite vias (0017A), Sicut decipula (0018C), State super vias (0019D), Quis dabit (0022D), Nolite ergo timere (0024B), Ab extremitatibus (0024D). All stop where Migne stops.
• Chunks 0006 and 0007 carry no repair of any kind — they were re-read in full against the fresh Latin and the only findings are the four non-words of §9g, which are candidates against our own text, not defects in the English.
• Negation host named for every printed negative in the two repaired sentences (9b): the non of non quod ante conceptionem fuerit (0009C) rests on the clause, not on fuerit alone, and survives the Priusquam te repair untouched; 0017D's nisi stays on crediderit and did not move when the subject was unified.
• Migne's punctuation untouched everywhere, including the 0017D colon that caused the defect.

Glossa ordinaria, Anselmus Laudunensis et schola. Cols 0018D–0031A (Jeremiah 6–14), 5 chunks, 4,919 Latin words, 168 [n:] notes. One translator.

Apparatus fired in this batch: 7 [var:] · 0 [sic:] · 0 [cj:] · 0 [d:] · 0 [ed:] · 0 [nt:] · 0 [cn:] · 0 « ».

⛔ CORRECTED 2026-08-29, after the plate gate: the [sic:] count is 2, not 0 — 0026C multc (chunk 0008) and 0028C ,umbare (chunk 0009). The figure above was true when it was written and is kept because the reasoning it belongs to — §3, where six non-words were raised and none fired — is still the reasoning that produced these two markers; it produced them once the plate was read. Neither is a reversal of §3's policy. Both are of the opposite class from anything §3 anticipated: Corpus Corporum silently repaired a defect in Migne's own type, so our Latin read CORRECTLY and no check we own could ever have seen it. The Latin has since been patched back to what the plate prints (data/tei-patches/9006.json) and the [sic:] shows the reader that the defect is Migne's. See §6. Final for chunks 0005–0009: 7 [var:] · 2 [sic:] · 0 [cj:].

Column anchors 12 · 10 · 7 · 10 · 11 = 50, all verbatim, in order, and checked against the word they stand beside in the Latin (brief test 2b) — the adjacency table was dumped for all fifty and every one matches, including the four that split a construction (usque [0020A] ad, etc., [0026B] usque ad, as far as [0021C], as far as [0027B]).

Line structure 1:1 with the Latin twin in all five chunks (101 · 93 · 70 · 83 · 85), including the double blank line the Latin prints after the Virgil citation at 0024B.

Punctuation checked BAND BY BAND, never by total (brief test 3): ? : ; ! « » compared inside every anchor-delimited band of every chunk. Nine mismatched bands were found and repaired before filing — all of them English colons/semicolons where Migne prints a comma or the reverse, i.e. exactly the class that has twice cancelled at the level of totals. Final state: zero mismatched bands in all five chunks. Question marks: 2 · 2 · 0 · 4 · 1, matching the Latin band for band; every one is a direct question in the Latin and stays direct in the English, so Pattern 8b never fires here. Italic-span parity checked band by band: zero mismatches.

§1 · CONVENTIONS — INHERITED, NOT RE-DERIVED

Ruth (8968) cruces.md §1 governs and was not reopened: VERS. n.-- passes through verbatim; the lemma is Englished; a fragmentary lemma is rendered as the phrase it stands for; the attribution sigla arrive as ordinary [n: …] notes with contents unanglicized. CAPUT VI. → CHAPTER VI., the Roman numeral kept.

usque ad → as far as, in its own italic span, per 8978 and Tobit (8970) §1(b), and no verb is ever supplied to complete the abridgment. Sites where the resumption tail lands mid-clause and reads broken are listed in §4 and are the formula working, not defects.

⚠ usque ad LOOK-ALIKES — decided per occurrence (brief addendum D). Six occurrences in this batch are not the abridgment formula and are rendered as ordinary words:

The first and the fourth are the dangerous shape (a formula-looking pair standing inside a lemma or a proof-text); the Sap 8:1 trap named in addendum D is exactly this class.

§2 · THE 7a″ DIVERGENCE LIST — every lemma in bands 0018D–0031A

Collated against sources/vulgate/clementine-flat.txt verse by verse (Jr 6–14 pulled entire, plus every non-Jeremiah proof-text). ⛔ A [var:] is a public claim about Migne's printed page and our Latin is not that page (CLAUDE.md 8a; data/plate-reads.json records no read on any column of 9006). The policy adopted, and applied uniformly:

Fire [var:] only where the divergence is (a) multi-word or a whole distinct word — so that no one-letter digitization slip could have manufactured it — AND (b) changes what the sentence asserts. Everything else is logged here and fires nothing.

That is the ship test applied to Pattern 14 rather than to [sic:], and it is why six markers stand where twenty-odd divergences are recorded.

2a · THE SEVEN [var:] FIRED

1. 0019D · Isa 40:9 — Migne prints In montem excelsum ascende tu qui evangelizas Sion, exalta sicut tuba vocem tuam, qui annuntias Jerusalem. The Clementine has Super montem excelsum ascende, tu qui evangelizas Sion ; exalta in fortitudine vocem tuam, qui evangelizas Jerusalem. Three separate substitutions, and the trumpet is drawn in from Isa 58:1 (quasi tuba exalta vocem tuam). Material to the gloss, which is expounding Vocem tubae — the proof-text was chosen for a word the Clementine does not have there.
2. 0022B · Jr 6:13 / Jr 8:10 — Migne quotes a minimo usque ad maximum omnes avaritiae student, citing Jer. VI. Jr 6:13 reads a minore quippe usque ad majorem omnes avaritiae student; Jr 8:10 — the verse actually under gloss — reads a minimo usque ad maximum omnes avaritiam sequuntur. The quotation is a cross-contamination of the two near-identical Jeremiah verses, taking the opening of one and the predicate of the other. Both halves multi-word.
3. 0024D · Isa 1:2 — Migne prints Audi, coelum, et auribus percipe, terra, quia Dominus locutus est. The Clementine has Audite, caeli, … quoniam Dominus locutus est. Singular addressee for plural, two words changed. Fired because the addressee's number is the thing the sentence does.
4. 0026A · Gn 49:8 — Migne prints manus tuae super dorsum inimicorum tuorum; the Clementine has manus tua in cervicibus inimicorum tuorum. A different image (hands on the back vs. hand on the necks), three words.
5. 0027A · Mt 23:8–9 — Migne prints one sentence, Nolite vocari magistri super terram, unus quippe est magister vester qui in coelis est, which exists in neither verse: 23:8 gives nolite vocari Rabbi : unus est enim magister vester, and super terram + qui in caelis est belong to 23:9, which is about fathers, not masters. A conflation that makes a sentence of its own.
6. 0029B · Dn 12:2 — Migne prints Isti resurgent in vitam aeternam, et illi in opprobrium et confusionem sempiternam; the Clementine has alii in vitam aeternam, et alii in opprobrium ut videant semper. The closing clause is different matter, not a different word.
7. 0030D · Ps 88:45 — Migne prints Thronum ejus in terra allisisti; the Clementine has sedem ejus in terram collisisti. Two content words, and Thronum is the word the lemma Solii gloriae is being matched to.

Distribution, verified against the files: 0005 ×1 (0019D) · 0006 ×1 (0022B) · 0007 ×1 (0024D) · 0008 ×2 (0026A, 0027A) · 0009 ×2 (0029B, 0030D) — seven markers, one per site above.

2b · DIVERGENCES LOGGED, NO MARKER FIRED — with the reason each lost

• 0018D · Jr 6:1 Confortamini, — Clementine Confortamini, filii Benjamin. Fragment, not divergence.
• 0019A · Jr 6:4/6:5 — the gloss quotes Surgite et ascendamus for what Jr 6:4 prints as consurgite, et ascendamus in meridie, and at VERS. 5 prints, in roman, consurgite et ascendite. The verse being glossed at that point is 6:5 (Surgite, et ascendamus in nocte), which Migne then quotes correctly in italic. The roman ascendite is Migne's own loose back-reference, not a lemma, and no marker fires on a re-quotation the plate does not set as scripture.
• 0019A · Isa 5:25 Et adhuc manus extenta sive excelsa — the Clementine has sed adhuc manus ejus extenta. A clipped citation offering two renderings; nothing asserted differently.
• 0019A · Mt 24:12 Refrigescet charitas multorum — agrees (Clem. caritas, an orthographic variant only). Migne's [n:] cites Matth. XIV; the verse is Matt 24:12. The note rides verbatim per the locked register — Migne's mis-citation is his refDisplay, not our text.
• 0019A · Ps 27:1, Mt 11:15, Jn 14:6 — all agree (Jn 14:6 lacks the Clementine's two et, which is not material).
• 0020A · Pr 13:8 — Migne Redemptio animae viri propriae divitiae ejus; Clementine divitiae suae. One word expanded to two, sense unchanged. Lost on materiality.
• 0020D · Lk 18:8 — Migne Cum venerit Filius hominis putas inveniet fidem super terram; Clementine Verumtamen Filius hominis veniens, putas, inveniet fidem in terra. Inside an abridgment tail, and super terram/in terra asserts nothing new. Lost on materiality.
• 0020B · Jr 7:8 Ecce vos confiditis in vobis — Clementine confiditis vobis in sermonibus mendacii. A preposition added. Lost on materiality.
• 0021A · Jr 7:31 Excelsa Tophet for Clementine excelsa Topheth. Spelling of a Hebrew place-name; Pattern 9. Not a divergence.
• 0021B · Jr 7:29 Tonde capillum. — the Clementine has Tonde capillum tuum. ⭐ The English deliberately reads "Cut off the hair," NOT "thy hair": supplying the possessive would import a word Migne does not print, which is 7a″'s exact failure mode.
• 0021D · Jr 8:1 In tempore illo for Clementine In illo tempore. Word order only.
• 0021D · Gn 8:21 Omne g enus hominum pronum est ad malum — a free paraphrase of sensus enim et cogitatio humani cordis in malum prona sunt. Paraphrase, not variant. (For g enus see §3.)
• 0022D · Mt 17:16 Usquequo sustinebo vos? — Clementine usquequo patiar vos ? A different verb; declined because Migne's own [n:] names the place and the clause is quoted from memory in a chain of memory-quotations.
• 0023A · Oui intelligit sedebit, quoniam tempus pessimum est — ⭐ Migne's own note adjudicates this. He cites (Thren. III; Mich. II), and the two halves are exactly Lam 3:28 (Sedebit solitarius, et tacebit) and Mch 2:3 (quoniam tempus pessimum est). A composite the plate itself labels is not a variant; no marker, on the same principle as the Nicetas note-settles-the-[lat:] calibration.
• 0023A · Lk 12:52–53 — Migne Et duo in tres, et tres in duo dividentur, Clementine tres in duos, et duo in tres dividentur; and Et tradet pater filium, et filius patrem is Mt 10:21 / Mk 13:12, not Luke. Order and conflation. Declined: the clause pair says the same thing either way round.
• 0023B · Sir 27:6, Ps 11:7 · 0022A · Ps 13:3, Ps 11:2 — all agree verbatim.
• 0023C · Lv 26:12 Inhabitabo et inambulabo in eis — the wording is 2 Cor 6:16's citation of Lv 26:12 (Ambulabo inter vos, et ero Deus vester). Migne cites Lev. XXVI. Declined: the words are a real Vulgate sentence, and identifying which of two places he meant is a research question Pattern 14 forbids the marker to answer.
• 0023C · 2 Cor 10:17 / 1 Cor 1:31 Qui gloriatur, in Domino glorietur — agrees verbatim with 1 Cor 1:31; Migne cites II Cor. X, where the Clementine reads Qui autem gloriatur. Note rides verbatim.
• 0024C · 1 Cor 1:19 — agrees verbatim.
• 0024D · Exspectet terra, ut pluviam verba mea, cited (Exod. IX) — nearest received text is Dt 32:2 Exspectetur sicut pluvia eloquium meum. ⛔ Declined, and the reason matters: Pattern 14 requires the marker to name the witness, and naming Dt 32:2 would be my identification of a verse Migne assigns elsewhere. A [var:] resting on my own re-attribution is not a claim about a text, it is a conjecture wearing a marker.
• 0024D · Jr 10:13 Fulgura in pluvias for Clementine fulgura in pluviam facit. Number of a noun in a lemma. Lost on materiality.
• 0024B · Jr 10:12 Dominus qui facit terram in fortitudine — Clementine Qui facit terram in fortitudine sua, with no Dominus. Origen's citation form; the added subject and the dropped possessive change nothing. Logged.
• 0025D · Ps 36:23 — Migne A Domino gressus hominis dirigetur; Clementine Apud Dominum gressus hominis dirigentur. ⚠ Two divergences of different weight: A Domino for Apud Dominum is a real preposition-and-case change, but dirigetur/dirigentur is one letter and is precisely the class the 9003 collation showed the digitization manufacturing. A marker resting on a pair one of whose members may be ours is not safe.** Crux, no marker. Plate read requested (§5).
• 0025D · Jr 10:24 In judicio, et non in furore — agrees; Clementine adds tuo.
• 0026B · Isa 43:3 — Ego sum Dominus Deus tuus sanctus Israel, cited (Ezech. XIV). The words are Isa 43:3 minus enim and salvator tuus. Same class as Dt 32:2 above; declined for the same reason.
• 0026B · Ero Deus eorum (Exod. III.) — the Vulgate formula is ero vobis in Deum / ero Deus vester. A formulaic third-person recast to fit the sentence. Logged.
• 0026B · Ex 3:6 and Lk 20:38 — both agree verbatim (Migne's (Luc. XXIX) is a chapter that does not exist; the note rides as printed).
• 0026B · Jr 11:5 lemma Lacte et melle, gloss Melle et lacte — ⭐ Migne prints the pair in one order in the lemma and the reverse in his own gloss, on the same line. Both are rendered as printed and neither is harmonized to the other. The Clementine has lacte et melle; the gloss's inversion is Migne's, not ours. (This is the 7914 @1079B cinis et pulvis class, and the whole point is that the reader should see it.)
• 0026C · Jr 11:8 Et non fecerunt, ab ipsis — a lemma cut out of the middle of the verse (quod praecepi ut facerent, et non fecerunt) with ab ipsis attached. Rendered as printed.
• 0026C · Jr 11:9 Inventa est conjuratio / Inventa est colligatio — Migne gives the second under [n: (ORIG.)] as the Septuagint-side reading. A variant the plate itself labels takes no [var:]; both are rendered.
• 0026D · Ps 142:8 and Ps 93:10 — Vel notum fac mihi, Domine, et cognoscam against Clementine Notam fac mihi viam in qua ambulem; Dominus docet hominem scientiam against qui docet hominem scientiam (an interrogative relative recast as a statement with a supplied subject). Both logged; both lost on materiality, the second being an ordinary proof-texting recast.
• 0026D · Jn 14:30 — agrees (Clementine venit enim princeps).
• 0026D · Ps 72:1 Quam bonus Israel Deus his qui recti sunt — Clementine his qui recto sunt corde. Migne's clipped citation ends where he ends it, with recti agreeing with qui. Fragment plus one adjusted ending; lost on materiality, and the etc. shows he knows it is clipped.
• 0027C · Jr 12:5 Si cum pedibus — Clementine Si cum peditibus currens laborasti. ⚠ The most tempting decline in the batch. pedibus is a real, well-formed word; the gloss contrasts these nations with the Chaldeans and Persians who equis gaudent, so "footmen" is plainly the sense wanted. Rendered as printed — "If with feet" — and no marker. Two candidates were weighed: a [cj: pedibus; read peditibus, "footmen"], and silence. The [cj:] lost on 18a's bar, which admits a false assertion, a negation, or a stranded correlative, and this is none of the three (a bare lemma stump asserts nothing); and it lost again on the ship test, because the difference is two letters inside a word and our Latin is not the plate. Plate read requested (§5).
• 0027D · Jr 12:9 Nunquid avis discolor? — the Clementine's Numquid against Migne's Nunquid throughout this work is orthography (Pattern 9), noted once here and nowhere else.
• 0028A · Gal 6:8 Seminamus non in carne, sed in spiritu… — cited (I Cor. XVII). The Clementine's Gal 6:8 is third-person conditional (qui seminat in carne sua…); Migne's is a first-person plural exhortation. Declined: this is Origen applying the verse in his own voice inside a resumption tail, not quoting it.
• 0028A · Lk 12:48 Cui plus committitur, plus ab eo exigitur cited (Sap. VI) — a proverbializing compression of cui commendaverunt multum, plus petent ab eo. Declined, same class.
• 0028A · 2 Cor 12:9 in infirmitatibus suis ut inhabitet in eo virtus Christi — Clementine in infirmitatibus meis … in me. The person shift is forced by the sentence, which hangs all three quotations off Qui autem gloriatur. Declined.
• 0028A · 1 Cor 1:31 and Gal 6:14 — both agree verbatim.
• 0028C · Rom 7:24 Miser ego homo for Clementine Infelix ego homo. A whole word, but synonymous; lost on materiality (criterion (b) of the policy above).
• 0028D · Soph 1:15 Dies tenebrarum et turbinis — the Clementine has dies tenebrarum et caliginis, dies nebulae et turbinis: Migne takes the head of one clause and the tail of the next. Compression of a doublet, nothing asserted differently. Logged.
• 0028D · Jr 13:17 Quod si hoc non audieris — Clementine audieritis, plural. Two letters. Same reasoning as Ps 36:23: declined**, plate read requested.
• 0029B · Jr 9:23 self-quotation — Non igitur glorietur sapiens in sapientia sua, vel fortis in fortitudine sua against Clementine Non glorietur sapiens… et non glorietur fortis… Migne adds igitur and collapses the second non glorietur into vel. ⚠ Negation-scope note: the English reads "Let not the wise man therefore glory in his wisdom, or the strong man in his strength." "or", not "nor", is deliberate — the Latin prints ONE Non, and "nor" would be a second English negative word standing for a Latin negative that is not on the page. English "Let not X … or Y" carries the scope correctly and adds nothing.
• 0029B · Ps 50:6 Tibi peccavimus, et malum coram te fecimus — Clementine Tibi soli peccavi, et malum coram te feci. Pluralized, and soli dropped. Declined: Migne has just written Dicamus et nos, so the plural is his own adaptation announced one clause earlier — the same self-labelling that saves the Lam/Mch composite above.
• 0029B · 1 Cor 3:9 Dei agricultura estis agrees; Dei cooperatores sumus against Clementine Dei enim sumus adjutores. A whole distinct word, but exactly synonymous in the argument (both render συνεργοί), so criterion (b) is not met. Logged.
• 0029D · Gal 6:2 Invicem onera vestra portate against Clementine Alter alterius onera portate. Multi-word, but the two say the same thing; lost on materiality.
• 0030C · Jr 14:18 abierunt in terram quam ignorabant — agrees verbatim.

2c · LEMMATA CHECKED AND AGREEING WITH THE CLEMENTINE — named, not summarized

Naming these is the point of the exercise: a divergence list without its negatives cannot be told from a list nobody made.

Jr 6 (chunk 0005): Speciosae. (6:2) · Pastores. (6:3) · Vae nobis, (6:4) · Surgite et ascendamus in nocte, (6:5) · Sicut frigidam. (6:7) · Erudire. (6:8) · Usque ad racemum. (6:9) · Incircumcisae aures, (6:10) · Idcirco furore. (6:11) · Extendam. (6:12) · Et curabant contritionem. (6:14) · Confusi sunt. · Nescierunt. (6:15) · State super vias. (6:16) · Vocem tubae. (6:17) · Ut quid mihi thus. (6:20) · Patres et filii. (6:21) · Ecce populus, (6:22) · Vox ejus quasi, (6:23) · Nolite exire, (6:25) · Et probabis. (6:27) · Omnes isti. (6:28).

Jr 7 (chunk 0005): Sta in porta domus Domini, (7:2) · Templum Domini. (7:4) · Spelunca latronum, (7:11) · Ite ad locum meum in Silo, (7:12) · Ad vos mane consurgens, (7:13) · Sicut feci Silo. (7:14) · Tu ergo noli, (7:16) · Nonne vides. (7:17) · Conflatur. (7:20) · Quia non sum locutus. (7:22) · Et loqueris ad eos. (7:27) · Haec est gens. · Periit fides et ablata est. (7:28) · Ecce dies venient. (7:32) · Et quiescere faciam. (7:34).

Jr 8–9 (chunk 0006): Et dices. · Aversus est, / avertit. (8:4) · Quare ergo, (8:5) · Nemo quod bonum, (8:6) · Milvus in coelo. (8:7) · Propterea dabo. (8:10) · Et sanabant contritionem. (8:11) · Idcirco cadent. (8:12) · Non est uva, (8:13) · Quare sedemus. · Convenite et ingrediamur civitatem munitam. (8:14) · A Dan auditus est. (8:16) · Mittam vobis serpentes. (8:17) · Dolor meus. (8:18) · Et ecce vox. (8:19) · Transiit messis. (8:20) · Super contritione. (8:21) · Quis dabit. (9:1) · Quis dabit me, (9:2) · Quia de malo, (9:3) · Unusquisque, · A proximo suo. (9:4) · Ut inique agerent laboraverunt. (9:5) · Et probabo eos. (9:7) · Nunquid super his (9:9) · Super montes. (9:10) · Et dabo Jesusalem. (9:11) · Cordis sui. (9:14) · Idcirco haec dicit Dominus. (9:15) · Gladium. (9:16) · Et vocate. / vocate lamentatrices, (9:17) · Oculi nostri. (9:18) · Quomodo vastati. (9:19) · Audite ergo. (9:20) · Quia ascendit. (9:21) · Loquere: Haec dicit, (9:22) · Qui facio misericordiam. (9:24) · Ecce dies venient. (9:25).

Jr 10 (chunk 0007): Audite verbum, (10:1) · Lignum. (10:3) · Argento et auro. (10:4) · Nolite ergo timere. (10:5) · Non est similis tui. (10:6) · Tuum est enim. (10:7) · Involutum. (10:9) · Coelos et terram. (10:11) · Qui facit. (10:12) · Nebulas. / Ab extremitatibus. / Et educit ventum de thesauris. (10:13) · Stultus. / Falsum est. / Non est spiritus. (10:14) · Vana sunt. / In tempore visitationis. (10:15) · Congrega. (10:17) · Vae mihi, (10:19) · Tabernaculum. (10:20) · Grex. (10:21) · Vox auditionis. / stulte egerunt, / Draconum. (10:22) · Scio, (10:23) · In judicio, et non in furore. (10:24).

Jr 11–12 (chunk 0008): Verbum quod factum est. (11:1) · Audite, / Viros Juda, (11:2) · Maledictus. (11:3) · Eduxi eos. / Et eritis. (11:4) · Lacte et melle. (11:5) · Quia contestans, / Audite verba pacti hujus, (11:6–7) · Inventa est conjuratio. (11:9) · Clamabunt. (11:11) · Secundum numerum. (11:13) · Tu ergo. (11:14) · Quid est, / Nunquid carnes sanctae. (11:15) · Ad vocem loquelae. (11:16) · Et Dominus. (11:17) · Tu autem. (11:18) · Tibi enim. (11:20) · Propterea haec dicit Dominus. (11:22) · Justus quidem, (12:1) · Et tu, Domine. (12:3) · Quomodo. (12:5) · Nam et fratres, (12:6) · Reliqui domum. (12:7) · Haereditas mea, (12:8) · Pastores multi, (12:10) · Desolatione. (12:11) · Non est pax, (12:12) · Seminaverunt. / A fructibus. (12:13) · Vicinos. (12:14).

Jr 13–14 (chunk 0009): Haec dicit Dominus. (13:1) · Et abscondi, (13:5) · Omnis laguncula. (13:12) · Et dispergam. (13:14) · Date Domino, (13:16) · Captus est. (13:17) · Dic regi et dominatrici. (13:18) · Tu enim docuisti eos, (13:21) · Propter multitudinem, (13:22) · Si mutare potest, (13:23) · Nudavi. (13:26) · Super colles. (13:27) · Et portae ejus, (14:1) · Agricolae. (14:4) · Nam et cerva, / in agro peperit (14:5) · Traxerunt ventum. / Defecerunt oculi. (14:6) · Exspectatio Israel. / Quasi colonus. / Et quasi viator. (14:8) · Movere pedes suos, (14:10) · Noli orare. (14:11) · Prophetae dicunt eis, (14:13) · Haec dicit Dominus. (14:15) · Deducant oculi mei. (14:17) · Si egressus, (14:18) · Nunquid projiciens, (14:19) · Cognovimus, Domine, (14:20) · Solii gloriae. (14:21) · Nunquid sunt in sculptilibus, (14:22).

2d · TWO FRAGMENT-RENDERING RULINGS

(i) The Nunquid fragments are NOT expanded — negation fidelity outranks convention 3. Migne prints VERS. 9.-- Nunquid super his (Jr 9:9) and Nunquid carnes sanctae. (Jr 11:15). Convention 3 would render each as the phrase it opens — but Jr 9:9 continues non visitabo, so "Shall I not visit for these things" would put on the page a non Migne does not print. The rule that every negative in the English answers to one on the page is not suspended by a lemma convention. Both are therefore rendered from the printed words alone and identically as a class: "Shall it be for these things" and "Shall the holy flesh."** The stumpiness is the plate's.

(ii) Where convention 3 DID expand, and why each is safe. Sicut frigidam. → "As a cistern maketh her water cold" (Jr 6:7, and the gloss is entirely about coldness) · Ut quid mihi thus. → "To what purpose do ye bring me frankincense" · Nemo quod bonum, → "No man speaketh that which is good" · Quia ascendit. → "For death is come up" (Jr 9:21) · Si mutare potest, → "If the Ethiopian can change" · Nunquid projiciens, → "Hast thou utterly cast off" · Nunquid avis discolor? → "Is my inheritance a speckled bird?", where the Septuagint reading Migne prints beside it carries the same full clause and the contrast is unreadable without it. None of these imports a negative, and none imports a word the verse's own Clementine text does not have. Complete one-word lemmata (Speciosae. · Pastores. · Nescierunt. · Gladium. · Lignum. · Grex. · Tabernaculum. · Vicinos. · Nudavi. · Agricolae.) are rendered as they stand — a fragment and a complete one-word lemma are different classes and are not to be contrasted.

⚠ Tabernaculum. (Jr 10:20 Tabernaculum meum) is rendered "The tabernacle," not "My tabernacle": same reasoning as Tonde capillum above.

§3 · NON-WORDS — SIX PLATE CANDIDATES, AND WHY NONE FIRED A [sic:] (at the time)

Brief addendum A: a suspicion aimed at a non-word has been confirmed as OUR corruption ten times out of ten. CLAUDE.md 8a: a [sic:] accuses Migne's type, and no column of 9006 has been read at the plate. The two together mean: raise it here, render the evident sense, fire nothing. All six are therefore Englished normally and the defect is recorded only here.

⛔ The last row was RIGHT ABOUT THE WORD AND WRONG ABOUT WHOSE IT WAS, and it is corrected in §6a rather than deleted. tuh is Corpus Corporum's corruption of Migne's tuli — so "And I took the girdle" was a silent emendation, the repair Pattern 7a forbids, wearing no marker. And on the same line CC had silently repaired a real defect of Migne's. The Latin now reads Et tuli ,umbare and the English carries [sic: ,umbare].

⚑ Note that the last three sit inside italic lemmata, which is where a swallowed run does the most damage (7a⁗-b): quarantining Oui would have taken the relative pronoun out of a clause that has no other subject, and quarantining tuh would have left "And ____ the girdle" with no verb at all. Rendering the sense and logging the type is the only treatment that is both honest and complete here.

⭐ Requested at the plate (§5). If Migne prints any of these six as we have them, the entry becomes a [sic:]; if he prints them correctly, the fix is data/tei-patches/9006.json and nothing reaches the reader — which is the whole reason the marker was withheld.

✅ Read 2026-08-29. The disposition of all six, and of one word this table never suspected, is §6. The prediction held exactly: one of the six (0028C) went to data/tei-patches/9006.json and nothing about it reached the reader as a [sic:] — while the [sic:] that did fire on that line is for a word standing beside it that no one had suspected.

§4 · THIS BATCH'S CRUCES — sense, subject, and negation

4a · ⛔ THE ONE SITE WHERE THE PRINTED NEGATIVE MAKES THE SENTENCE CONTRADICT ITSELF

0028C · Jr 13:5, resumption tail: …usque ad si non adhaerens Domino, uno spiritu fuerit in Christo.

Englished as printed: "if, not cleaving to the Lord, he shall have been one spirit in Christ." The allusion is 1 Cor 6:17, qui adhaeret Domino unus spiritus est — so the clause as printed says the opposite of the doctrine it is built on. The non attaches to adhaerens and to nothing else (it stands immediately before the participle, and fuerit has its own predicate in uno spiritu). It is carried. Per Pattern 7a and the standing rule, a word is never translated as its opposite to rescue a sentence, and the tail is a resumption whose head the abridgment removed — the missing head may well be a negative of its own (nemo …, nisi …), in which case the sentence is sound and only looks broken. No conjecture is offered and no marker fires, because both candidates (a lost governing negative; a plate slip for adhaerens) require the head of Jerome's sentence, which is exactly what usque ad deleted.

4b · UNEXPRESSED SUBJECTS — every "he/she/it/they" I supplied, and what fixed it

• 0021C · offerre debuisset → "ought he to have offered it": the subject is the master, lifted from primum occurrisset domino in the same clause. The rival (the animal) is impossible — an animal does not offer.
• 0021C · Quando recessit… duxit Sedeciam → "When he departed from fealty to his lord": Sedecias, named four clauses later as the object of duxit. The dominus he departs from is Nabuchodonosor, which is why Nabuchodonosor illum fugavit follows.
• 0022A · Unde ipse: before Ps 11:2 → "Whence he himself." ⚠ Two readings both construe and I have not chosen silently: (a) the Psalmist, since the preceding quotation is Ps 13:3 and this is another psalm; (b) the Saviour, since the sentence opens In tempore Salvatoris. The English "he himself" is deliberately as underdetermined as the Latin.
• 0024C · dum coeli sibi colorem ostendunt → "while they display upon themselves the colour of heaven": the heretics, from falsitas involvitur and the haeretici of VERS. 7 two paragraphs above. The abridgment removed the nearer antecedent.
• 0025A · Quod ordine suo fingit → "That which he fashions in his own order": the artifex of Jr 10:14 (confusus est artifex omnis in sculptili), the verse being glossed.
• 0026C · in Septuaginta non habentur → "they are not found in the Septuagint": the words of the verse from Quia contestans down to ut facerent haec, which is what the sentence has just delimited.
• 0027B · sermo Salvatoris in eam descendit → "descends into it": feminine, and the antecedent (the bitter water of Mara, Ex 15:25, with the lignum thrown into it) lies in the part of the sentence usque ad removed. Rendered neutrally rather than named.
• 0030D · Tunc alliditur… quia culpa perditur… sustentatur → "Then is it dashed down… because it is lost by fault, it is upheld": the subject throughout is the solium / templum of the lemma, which is what in quo… Deus offenditur requires.
• 0026D · Dicamus hoc principibus… putant se Dei mereri clementiam → "who… think that they merit": the princes and the rich, from the dative just before; putant is indicative against the subjunctives of the cum clause and so belongs to the relative, not to dicamus.

4c · NEGATION HOSTS — the sites in brief-test 2a's trigger population

Every printed negative in the batch was checked, and its host named from structure rather than from what reads well. The ones with a quantifier, correlative, or pivot beside a verb:

• 0019D · non solum non cavere, sed nec intelligere velle peccata, et nullam habere distantiam — three negatives, three hosts: non on solum; non on cavere; nec on velle; nullam on distantiam. The English carries each on its own word, and the object peccata is carried through to both infinitives ("not to beware of sins… not even to be willing to understand them") — the fourth question the brief names, which 7a⁗'s three do not ask.
• 0019A · non temporis esse victoriam, sed virtutum — host is temporis, proved by the sed virtutum correlative, NOT the infinitive esse.
• 0025B · Non enim Deus, sicut prius, longe post futura minatus, sed statim futura — ⭐ the sharpest case in the batch. The Non stands before the subject and the participle, and the sed names the contrast: long-after things versus at-once things. So the host is the time-phrase, not minatus, and the English reads "God has threatened not, as before, things to come long after, but things to come at once" — never "God has not threatened." Migne is not denying the threat; he is dating it.
• 0026B · Non est Deus omnium, sed eorum tantum… — host is omnium, proved by the sed … tantum pivot: "not the God of all, but only of those…"
• 0027A · non humiliter creatorem… intellexisti, sed contra Dominum granditer locutus — the humiliter/granditer pair across the sed fixes the host on the adverb, not on intellexisti. English: "not humbly hast thou understood."
• 0029C · non humiliatur sub potenti manu Dei, sed confidit in sceleribus suis — here the sed pivots on the verbs (not humbled / but trusting), so the host really is humiliatur. Recorded because it is the same shape as the previous entry and gets the opposite answer; the pivot, not the shape, decides.
• 0030A · ut Dominus auditus non est, non possessor agri, sed colonus — two negatives, two hosts: the first on auditus est; the second on possessor agri, proved by sed colonus.
• 0029C/0030A pair, and 0030B nullis precibus eruitur — nullis on precibus.
• 0026A · ne maledicti simus · 0029A · ne et ipse conteratur · 0021A · ne videatur… non impetrare — purpose-clause negatives, each on its own verb.

4d · READINGS DECLINED, WITH THE REJECTED CANDIDATE NAMED

• 0021B · Ne dubitetis indurare cervices (Jr 7:27 tail). Two construals both parse. (a) dubito + infinitive = "hesitate": "Hesitate not to harden your necks" — which is what the argument does not want. (b) dubito + accusative-and-infinitive = "doubt that", giving "Do not doubt that they harden their necks", which fits Jr 7:27 exactly (and they will not hear thee). (a) was taken and (b) rejected, on one ground: (b) requires an accusative subject that the plate does not print, so choosing it would be supplying a word to make the sentence behave. The English is therefore harsh, and the harshness is the evidence.
• 0026C · invenimus Judam et Jeremiam multo pejora — ⚠ the Latin of this very clause has since changed: multo is multc on the plate, and now carries the batch's first [sic:] (§6a). The ruling on Jeremiam below is unaffected and stands. The reader who found multc was reading this clause word by word and left Jeremiam unpatched, which is corroboration but not a recorded verdict: data/tei-patches/9006.json says nothing about Jeremiam either way. — Jerusalem is the obvious conjecture (the pairing Judam et Jerusalem occurs four times in these five chunks and Judam et Jeremiam nowhere else in the work). ⛔ Declined: Jeremiam is a real, well-formed, contextually possible word, and brief addendum A is emphatic that a corpus argument against a real word has been refuted every time it was tested. Rendered as printed. Plate read requested (§5).
• 0030B · Caveat pseudoprophetae — a singular verb against a plural nominative (or a nominative plural against a singular verb; the form is ambiguous between them). Englished "Let the false prophets beware," which is number-neutral in English and therefore shows the reader nothing — Pattern 12's second test would ordinarily make that a marker case. It does not fire here, for the ship-test reason: the marker asserts a defect in Migne's type, and a one-letter ending is exactly what the digitization has been shown to alter. Logged; plate read requested.
• 0019A · Ecclesia, spiritu ferventes * — a plural participle against singular Ecclesia*. Pattern 9: a concord mismatch with no English exponent. No marker, no repair; the English reads "the Church, fervent in spirit," and the Latin column shows the disagreement.
• 0025D · In judicio, et non in furore. (HIER) — ⚠ this siglum is NOT a [n:] note. Migne sets it inside the italic lemma as running text, without the parenthesis-note typography the other 168 sigla have, and the chunk's note count (22) confirms it is not one. It is therefore reproduced inside the italic span, untranslated, exactly as printed: In judgment, and not in fury. (HIER). Do not "fix" it into a note; that would change the marker count against the Latin twin.
• 0026A · Alleg. — Migne abbreviates the sense-label here and prints Allegorice. in full at 0020A, 0028A and 0029B. Rendered "Allegorically." in all four places, on Pattern 9's section-label ruling: there is no English exponent for a Latin abbreviation, and the Latin column preserves the difference in the facing text. Logged so that a later reader knows the uniformity is ours and the variation is his.
• 0029D · Nam et cerva (vel cervae) in agro peperit (vel pepererunt). Migne's own parenthetical variant readings. Reproduced verbatim, untranslated, in place, on the bracketed-variant rule of the locked register ([ al., …] family) and the L8 precedent — they are variant Latin, not prose. The lemma around them is Englished as usual.
• 0019B · lacum — Migne discusses the word as a word (Semel sufficiat dixisse lacum, non stagnum sonare). It is not italicized in the Latin, so it could not be italicized in the English without breaking span parity; it is carried in our own double quotes (Pattern 6: " " is ours) — that "lacum" sounds not a pool. The quotation is one we supply, so guillemet parity is unaffected.
• 0021C · et ibi orbatus. — a participle with no copula, closing the sentence. Rendered "and there deprived," with nothing supplied. The historical referent is the blinding of Sedecias (4 Rg 25:7), and est is what the sentence wants; it is not printed, so it is not written. No marker: nothing is broken type, and Pattern 18's bar is not met by an ellipsis.

4e · SECTIONS READ CLEAN

Read for sense against the Latin, paragraph by paragraph, and found to assert what Migne asserts, with nothing to log: 0019D–0020B (Jr 6:16–6:28, the pearl-merchant paragraph through Omnes isti), 0022B–0022D (Jr 8:13–8:21), 0023B–0023D (Jr 9:5–9:16), 0024B–0024C (Jr 10:1–10:8), 0027D–0028B (Jr 12:7–12:14), 0029C–0030A (Jr 14:1–14:8). The batch's whole Quasi dicat: formula set (17 occurrences) is rendered "As if he should say:" without variation, and the four bare Quasi / Quasi, / Ac si aperte diceretur variants are rendered distinctly, because Migne distinguishes them.

§5 · WHAT IS OWED AT THE PLATE BEFORE THIS WORK SHIPS

✅ PAID, 2026-08-29 — every page asked for below was read, and §6 records what came back. The paragraph that follows is left standing because it is the request the read answered; its opening statement of fact is no longer true.

data/plate-reads.json records no read on any column of 9006, and data/plate-notes/coverage.json has no entry for this work — so the spot check that chunk-work.mjs prints as ⬜ PLATE SPOT CHECK OWED is still owed, and PL 114's sampled zero elsewhere must not be inherited for this book.

PL 114 map, verified in CLAUDE.md against the printed corner numbers: patrologiaecurs06saingoog, PDF page = (column + 11) / 2. Bands 0018D–0031A are therefore pp. 15–21. Seven pages settles everything below.

1. The six non-words of §3 — quod tibet (p. 16) · uque ad and g enus (p. 16) · Oui (p. 17) · Jesusalem (p. 17) · tuh (p. 20). Each becomes a [sic:] if Migne prints it, or a data/tei-patches/9006.json entry if he does not.
2. Si cum pedibus (Jr 12:5, p. 19) — the strongest conjecture in the batch and the one most worth settling.
3. invenimus Judam et Jeremiam (p. 18).
4. Caveat pseudoprophetae (p. 21) and Quod si hoc non audieris (p. 20) — both one-ending questions.
5. A Domino gressus hominis dirigetur (p. 18) — read for dirigetur/dirigentur.
6. ⭐ Read the whole line, not the target word (addendum C): all six of §3's sites and the four of §4d sit in paragraphs that are otherwise smooth, which is precisely where the normalization class hides. The two normalizations found in the last batch were both spotted while reading the plate for a different word four words away.
7. All seven [var:] columns — 0019D, 0022B, 0024D, 0026A, 0027A, 0029B, 0030D — must pass node scripts/plate-gate.mjs 9006 before the work is marked ours. They are fired on multi-word divergences precisely so that a single manufactured letter cannot be what they rest on, but the gate is the gate.

§6 · THE PLATE GATE, 2026-08-29 — what came back, and the two [sic:] it fired

24 pages of patrologiaecurs06saingoog at 300 dpi, three readers, corner numbers read first on every page; recorded in data/plate-reads.json (works.9006) and patched in data/tei-patches/9006.json. Bands 0018D–0031A fall on pp. 15–21, all read.

⛔ The finding this batch did not anticipate, and could not have. §3 was built for one direction — our Latin corrupt, Migne's plate clean — and named six candidates on that model. The plate returned the opposite direction twice inside these five chunks: Corpus Corporum had silently repaired a defect in Migne's own type, so our Latin read correctly, every mechanical check we own passed, and nothing whatever pointed at the site. Frequency analysis cannot see this class: the token it would have to be suspicious of is a perfectly ordinary word. The Latin is now patched back to what the plate prints, and the [sic:] is what tells the reader the defect is Migne's and not ours.

6a · THE TWO [sic:]

1. 0026C (chunk 0008) · multc for multo.

⚠ The reader flagged this as the LEAST CERTAIN of the seven finds on its slice, and the caveat is recorded here on purpose, in its own words: a worn o and a c can look alike, and only the unthinned terminals at 12× decided it. A later reader should weigh this one below the unambiguous six — it is the single find in the whole 9006 gate carrying a stated doubt. The unthinned terminal is real evidence and the marker stands; the point is that it stands on a narrower base than ,umbare or pμtantes, and nobody should discover that by re-deriving it.

⚠ Column label: data/tei-patches/9006.json files this patch under 0026D; by anchor position in the chunk the word stands under 0026C, which is where plate-gate.mjs resolves it and where §4d above independently placed the same clause. The word sits at the very foot of the band. Nothing was moved to settle this — the marker is where the translator put it; the discrepancy is noted so a reader who follows the patch file to col. 0026D and finds nothing knows why.

2. 0028C (chunk 0009) · ,umbare for lumbare — and tuli corrupted on the same line.

⭐ This is the 8991 pattern on a single line, and it is the most instructive site in the work. Both directions of departure at once:

• CC → us, corrupting. tuh is a non-word. §3 above spotted it and got the word right (tuli) — but then Englished the sense, "And I took the girdle," with no marker, which is a silent emendation of a text we had not verified: exactly the repair Pattern 7a forbids, reaching a reader as though it were Migne's own smooth Latin.
• Migne → CC, repairing. ,umbare is a genuine defect of Migne's type, and CC's tidy lumbare concealed it completely. Our Latin was clean, so there was nothing to detect.

Neither direction is visible to frequency analysis, and the two cancel to a clean-looking line. The Latin now reads Et tuli ,umbare; the English translates tuli on the plate's authority and carries [sic: ,umbare] for the defect. ⚑ Strip the marker and the sentence is complete — "And I took the girdle." — so 7a⁗ is satisfied: the marker does not carry the force.

6b · THE OTHER FIVE CANDIDATES OF §3, AND THE §4d CONJECTURES

The read covered every page these sit on and returned no patch for any of them, which under the policy of §3 is the outcome that fires nothing and changes nothing: quod tibet (0021C) · uque ad (0021D) · g enus (0021D) · Oui (0023A) · Jesusalem (0023C). Likewise Si cum pedibus (0027C), invenimus Judam et Jeremiam (0026C), Caveat pseudoprophetae (0030B), Quod si hoc non audieris (0028D) and A Domino … dirigetur (0025D) — every decline recorded in §2b and §4d stands, and none of them acquired a marker.

⚠ What that silence is worth, stated exactly. data/plate-reads.json records per page what was found, not a per-token verdict, so for these ten sites the evidence is that a reader had the page on screen at 300 dpi, was reading it word by word (§3's "read the whole line" rule is what turned up egos and multc), and patched nothing. That is good corroboration and it is not the same thing as a recorded confirmation. No marker rests on it, which is why it costs nothing.

6c · THE SEVEN [var:] OF §2a

All seven — 0019D · 0022B · 0024D · 0026A · 0027A · 0029B · 0030D — were read and stand. node scripts/plate-gate.mjs 9006 returns gate OK.

One translator, five chunks, 5,417 Latin words (Jer. 15–30, CAPUT XV–CAPUT XXX). verify-english.mjs clean. EN/LA ratio 1.48–1.55 (markers stripped), against the Tier-2 expectation of ~1.5×.

Apparatus fired in this batch: 4 [var:] · 1 [ed:] · 0 [sic:] · 0 [cj:] · 0 [d:] · 0 [nt:].

⛔ CORRECTED 2026-08-29, after the plate gate: the [sic:] count is 2, not 0 — 0033D pμtantes (chunk 0010) and 0037A Nuntlis (chunk 0012). Final for chunks 0010–0014: 4 [var:] · 2 [sic:] · 1 [ed:] · 0 [cj:]. See §7. The zero above and the ⛔ paragraph below it are kept, not deleted, because the policy they state was never overturned — neither new marker is a letter-level accusation against Migne of the kind that policy withheld. Both are the opposite direction: Corpus Corporum silently repaired a defect in Migne's own type, our Latin therefore read correctly, and no check we own could see the site at all. Only a plate could.

Notes carried verbatim: 35 · 33 · 26 · 10 · 17 = 121, exact parity with the Latin twins. Column anchors 11 · 12 · 12 · 12 · 8 = 55, verbatim, in order, each checked against the word it stands beside in the Latin (§6). Guillemets: zero in either language.

⛔ THE SHIP TEST GOVERNED EVERY MARKER DECISION IN THIS BATCH, AND IT IS WHY THE [sic:] COUNT IS ZERO. No column in 0031B–0044D has been read at Migne's plate (data/plate-reads.json carries no 9006 entry). A [sic:] accuses Migne's type and a [var:] asserts his text diverges from a witness; our Latin is Corpus Corporum's transcription, not the page. So the rule applied throughout was:

• A non-word, or a real word with no grammatical slot, is CARRIED in italics with a crux and named here as a plate candidate — and takes no marker. Three sites (§3).
• A [var:] fires only where the divergence is of a shape the digitization cannot manufacture — an added or dropped phrase, a different lexeme, a changed word order that alters the syntax. A one- or two-letter difference gets a crux and no marker, because that is exactly what OCR produces (§4b). Four [var:] fired; eleven candidate divergences were declined on this ground and are listed with reasons.

§1 · THE 7a″ DIVERGENCE LIST — every lemma in cols 0031B–0044D, checked against the Clementine

Method: each italic span in data/briefs/9006-lemmata.txt for these bands was read against sources/vulgate/clementine-flat.txt at the verse it glosses, not merely searched for somewhere in the canon (the pre-scan's ✓ is not clearance). Book/chapter/verse cited.

1a. Lemmata that agree with the Clementine EXACTLY (allowing for clipping at the lemma's end)

These were checked and found clean. They are recorded because a findings-only list cannot be told apart from a report by an agent that never looked.

Jer. 15 — 15:1 Et dixit Dominus · 15:3 Et visitabo super eos · 15:4 Et dabo eos in fervorem · 15:5 Quis enim miserebitur tui? · 15:6 Retrorsum abiisti · 15:8 Multiplicatae sunt mihi · 15:9 Septem · 15:10 Vae mihi, mater mea · 15:11 Si non reliquiae / Si non reliquiae tuae in bonum / occurri tibi · 15:13 Et thesauros / In omnibus peccatis tuis · 15:14 Ardebit · 15:15 Noli in patientia tua suscipere me · 15:17 Non sedi / A facie manus tuae / Solus sedebam · 15:19 Si separaveris / Quasi os meum. Jer. 16 — 16:1 Et factum est verbum · 16:4 Mortibus aegrotationum · 16:5 Ne ingrediaris domum · 16:6 Neque plangentur · 16:9 Ecce ego auferam · 16:12 Sed et vos · 16:14 Ecce dies veniunt · 16:16 Ecce ego mittam piscatores / Cavernis petrarum · 16:18 Duplices iniquitates · 16:19 Ad te gentes venient / Ab extremis · 16:20 Deos. Jer. 17 — 17:1 Peccatum Juda scriptum est (twice) · 17:3 Excelsa · 17:5 Maledictus homo · 17:7 Benedictus vir · 17:8 Quod ad humorem · 17:11 In dimidio dierum · 17:14 Sana me, Domine · 17:15 Ecce ipsi dicunt ad me: Ubi est verbum · 17:21 Custodite animas · 17:24 Ut non inferatis onera per portas civitatis hujus in die sabbati (verbatim, whole) · 17:27 Si autem non audieritis me, ut sanctificetis diem sabbati, et ne portetis onus, et ne inferatis per portas Jerusalem (verbatim, whole) · Si autem. Jer. 18 — 18:1 Verbum quod factum est ad Jeremiam a Domino dicens · 18:4 Et dissipatum est · 18:7 Repente loquar · 18:11 Ecce ego · 18:12 Post cogitationes · 18:16 Ut fieret terra · 18:17 Sicut ventus · 18:19 Attende, Domine, ad me, et audi vocem adversariorum meorum (verbatim, whole) · 18:23 Ne propitieris iniquitati eorum. Jer. 19 — 19:1 Haec dicit Dominus: Vade et accipe lagunculam · 19:4 Eo quod · 19:5 Quae non praecepi · 19:6 Topheth, et vallis filii Ennom · 19:7 Et dabo cadavera · 19:11 Sic conteram populum istum / Et in Topheth sepelientur · 19:13 Et erunt domus Jerusalem et domus regum. Jer. 20 — 20:1 Et audivit Phassur · 20:3 Cumque illuxisset / Non Phassur · 20:7 Seduxisti me, Domine / Factus sum in derisum · 20:12 Et tu, Domine exercituum / Videam / Tibi enim · 20:13 Quia liberavit · 20:14 Maledicta / Maledicta dies in qua natus sum · 20:15 Maledictus vir. Jer. 21 — 21:1 Verbum quod / Quando misit / Phassur filium · 21:4 Ecce ego · 21:7 Dabo Sedeciam / In manu · 21:8 Et ad populum / Ecce ego do coram vobis · 21:12 Judicate mane · 21:13 Ecce ego ad te habitatricem · 21:14 In saltu. Jer. 22 — 22:1 Haec dicit · 22:3 Facite judicium · 22:6 Si non posuero · 22:10 Nolite flere mortuum · 22:11 Ad Sellum · 22:13 Vae qui aedificat · 22:18 Non plangent eum · 22:24 Vivo ego dicit / Quia si fuerit · 22:28 Vir iste · 22:30 Scribe / Sterilem. Jer. 23 — 23:1 Vae pastoribus · 23:5 Ecce dies veniunt, dicit · 23:7 Propter hoc · 23:9 Quasi homo madidus a vino / A facie Domini · 23:10 Luxit terra · 23:11 Propheta namque · 23:12 Afferam super eos mala (Vulg. afferam enim super eos mala — the lemma simply begins one word later) · 23:14 Et in prophetis · 23:15 A prophetis · 23:18 Quis enim? · 23:19 Ecce turbo · 23:25 Audivi quae dixerunt / somniavi · 23:26 usquequo istud · 23:27 Somnia eorum quae · 23:28 Quid paleis? · 23:29 Quasi ignis · 23:30 Qui furantur · 23:39 Et derelinquam. Jer. 24 — 24:1 Ostendit / Duo calathi / Positi ante templum · 24:2 Et calathus unus · 24:4 Et factum est · 24:6 Et ponam oculos meos. Jer. 25 — 25:1 Verbum / Anno quarto · 25:4 Et misit Dominus ad vos omnes servos / Neque inclinastis · 25:6 Post deos · 25:9 Et assumam / Nabuchodonosor servum meum · 25:10 Perdamque · 25:11 Et servient · 25:15 Sume calicem vini furoris · 25:18 Jerusalem et civitatibus Juda · 25:19 Pharaoni regi Aegypti · 25:22 Tyri et universis regibus · 25:25 Regibus Elam et cunctis · 25:26 Et rex Sesach · 25:31 Judicium Domino · 25:34 Ululate, pastores. Jer. 26 — 26:1 In principio regni · 26:2 In atrio domus · 26:3 Si forte audiant · 26:4 Si non audieritis · 26:6 Dabo domum istam sicut Silo · 26:10 Portae domus · 26:11 Et locuti sunt · 26:12 Et ait / In veritate enim misit me Dominus ad vos (verbatim, whole) · 26:13 Nunc ergo bonas · 26:16 Et dixerunt · 26:20 Fuit quoque · 26:22 Et misit rex · 26:24 Igitur manus Ahicam. Jer. 27 — 27:1 In principio regni · 27:2 Haec dicit Dominus · 27:5 Terram et homines / In fortitudine / Et dedi · 27:6 In manu Nabuchodonosor / Insuper et · 27:7 Donec veniat · 27:12 Et ad Sedeciam · 27:14 Nolite audire / Non servietis · 27:16 Et ad sacerdotes · 27:17 Ut vivatis · 27:18 Occurrant Domino · 27:19 Ad columnas. Jer. 28 — 28:1 Et factum est / In domo Domini · 28:3 Omnia vasa · 28:6 Amen, sic · 28:7 Verumtamen · 28:10 Et tulit · 28:13 Ananias · 28:14 Et facies / Catenas · 28:15 Et dixit · 28:16 Hoc anno · 28:17 Mense septimo. Jer. 29 — 29:1 Et haec / Verba libri / Ad reliquias seniorum · 29:6 Accipite uxores · 29:8 Non vos seducant · 29:10 Cum coeperint / Septuaginta anni · 29:14 Et reverti · 29:17 Quasi ficus · 29:19 De nocte · 29:21 Ad Achab · 29:24 Ad Semeiam · 29:27 Quare non increpasti · 29:31 Haec dicit / Ecce ego visitabo. Jer. 30 — 30:1 Hoc verbum · 30:7 Vae quia magna · 30:9 Et David · 30:10 Serve meus · 30:12 Plaga · 30:16 Propterea · 30:18 Ecce ego convertam · 30:19 Ludentium, et multiplicabo eos.

Inline NT and other scripture that agrees exactly: Matt. 21:43 Auferetur a vobis regnum Dei, et dabitur genti facienti fructus ejus (0032A) · Rom. 3:2 credita sunt illis eloquia Dei (0032A) · 1 Cor. 11:30 Ideo inter vos multi infirmi et imbecilles (0032C) · Phil. 4:4 Gaudete in Domino semper, iterum dico gaudete (0033A; and again 0040A, where Migne sets a comma after dico) · Matt. 27:25 sanguis ejus super nos et super filios nostros (0035B) · Ps. 72:23 Ut jumentum factus sum apud te (0038B) · Jer. 23:4 [Et] suscitabo super eos pastores (0038B — see §4a.1) · Ex. 32:6 Sedit populus manducare et bibere, et surrexerunt ludere (0044B) · John 3:18 qui non credit jam judicatus est (0041A, Vulg. qui autem non credit) · Jer. 16:13 Et ejiciam vos (0042C).

1b. Where Migne's lemma is a FRAGMENT — how it was rendered, and the rule used

Convention 3 (Ruth cruces.md §1.2) renders a fragment as the phrase it stands for. This work prints fragments in two shapes, and the shape decided the treatment:

• A fragment Migne himself marks as truncated with a comma + etc. is rendered from the printed words only, with the comma and etc. carried through — Migne has already told the reader that words are elided, so supplying them would add text he chose not to print. So Ecce ego, etc. (0036D, Jer. 21:4) is "Behold I, etc.", not an expansion.
• A bare fragment that ends in a full stop, with nothing marking the elision, is expanded, because there the stump is simply not English. Expanded here: Si non reliquiae. → "If thy remnants be not unto good" (three occurrences, all rendered alike — the brief's rule that a stump is resolved by EXPANDING, never by contracting its twin) · Quod ad humorem. → "Which sendeth forth its roots toward moisture" · Ecce ego. (0035A, Jer. 18:11) → "Behold I frame evil against you" · Ut fieret terra. → "That the land might become a desolation" · Nunquid de petra: → "Shall the snow of Libanus fail from the rock:" (Migne's colon kept) · Non Phassur. → "Not Phassur hath the Lord called thy name" · Maledicta. → "Cursed be the day wherein I was born" · Nunquid? → "Is this man Jechonias a broken earthen vessel?" · Quis enim? → "For who hath stood in the counsel of the Lord?" · Quid paleis? → "What hath the chaff to do with the wheat?" · Nunc ergo bonas. → "Now therefore make your ways good."
• ⚠ TWO BARE FRAGMENTS WERE DELIBERATELY NOT EXPANDED, and the reason is not inconsistency but DOUBLING. At 0038A Migne glosses Scribe. and then, four lines later, Sterilem, — both out of Jer. 22:30 Scribe virum istum sterilem. Expanding the first would print "barren" twice and make the second lemma look like a repetition of the first. Same at 0038D–0039B, where Nunquid non verba mea? is followed by Quasi ignis. out of Jer. 23:29 Numquid non verba mea sunt quasi ignis. Both were therefore rendered from the printed words: "Write." and "Are not my words?" ⭐ The general form, offered for the rest of the book: convention 3 expands a fragment to the phrase it stands for, EXCEPT where the next lemma in the same gloss is the continuation of that same phrase.
• Complete one-word lemmata are rendered as they stand and were not expanded, per the brief's "name the control's class": Septem. · Ardebit. · Deos. · Excelsa. · Verumtamen. · Ananias. · Catenas. · Plaga. · Propterea. · Ostendit. · Verbum. · Videam. · Perdamque. · Venit. (on which see §3.3).

§2 · THE FOUR [var:] MARKERS — what each claims, and why the shape licenses the claim

2.1 · 0032D · 1 Cor. 7:29 — Migne's quotation carries a CLAUSE the Vulgate does not

⚠ This heading read 0032C and is corrected to 0032D. Ruled 2026-08-29 by reading the chunk: 0010.md prints the [0032D] anchor in the paragraph before this one and the next anchor is [0033A], three paragraphs later, so the [var:] stands under 0032D by anchor position — which is also where plate-gate.mjs resolves it, where data/tei-patches/9006.json files the plate read, and what migne.app/pl/114/32d will show a reader who follows the citation. ⛔ The marker itself was not moved; only this heading was wrong.

Quia tempus breve est et consummatio imminet, ut qui uxores habent, sint quasi non habentes [n: (I Cor. VII)]

Clementine 1 Cor. 7:29: Hoc itaque dico, fratres: tempus breve est: reliquum est, ut et qui habent uxores, tamquam non habentes sint. The four-word clause et consummatio imminet is in no part of 1 Cor. 7 and no digitization inserts a clause. [var:] fired, on that clause and on the word order.

⛔ The marker's second claim has been WITHDRAWN, and the quotation above is corrected. As written, this entry read ut qui exores habent and said the [var:] "also carries the Vulgate's uxores, which is what lets a reader decode exores (§3.1)." The plate prints uxores, in italic, unmistakably (p. 21; data/tei-patches/9006.json, the work's one corrupted patch). exores was Corpus Corporum's, not Migne's — so there was never anything to decode, and a [var:] clause contrasting our reading with the Vulgate's was contrasting the Vulgate with a digitization error. The Latin is patched to uxores; the marker now reads without et consummatio imminet and in a different word order, and nothing else. The clause about et consummatio imminet and the word order stands unaltered — it was never in question, and the plate confirms it.

2.2 · 0034B · Jer. 17:11 — the divergence is what the gloss ARGUES FROM

Novissimum ejus erit insipiens, cum omnia subjecta fuerint Christo. Novissimus inimicus destruetur mors.

Clementine Jer. 17:11: et in novissimo suo erit insipiens — "in his latter end he shall be a fool," a prepositional phrase. Migne's Novissimum ejus makes "his last" the subject, and the gloss immediately reads it as "his last [enemy]" and yokes it to 1 Cor. 15:26. ⭐ The exposition rests on the diverging words, which is 7a″'s own test, and the divergence is a change of construction, not of a letter. [var:] fired.

2.3 · 0038D · Jer. 23:11 — singular against plural, and English has an exponent

Propheta namque et sacerdotes polluti sunt.

Clementine: Propheta namque et sacerdos polluti sunt. Rendered "the priests", with [var:]. Fired rather than declined because (a) English has an exponent for number, so rendering it silently would conceal the divergence; (b) the divergence is an appended syllable, which is not an OCR shape (the digitization drops and confuses letters, it does not supply a plural ending); (c) it is a re-quotation — the VERS. 11 lemma five words earlier prints Propheta namque, alone — and it is plural precisely because the gloss has just said in principibus ejus. ⚠ The rival reading (our transcription's slip) is recorded here and is the reason this is the batch's least secure marker; a plate read of col. 0038 settles it.

2.4 · 0043C · Ps. 141:8 — a different LEXEME, and the gloss turns on it

Educ de carcere animam meam

Clementine Ps. 141:8: Educ de custodia animam meam. Migne's sentence is de carceribus corporum liberamur … secundum illud quod male interpretantur haeretici: the whole argument is about prisons of the body, so carcere is not an accident of type but the word the heretics' misreading hangs on. A different lexeme is not a transcription shape. [var:] fired.

§3 · PLATE CANDIDATES — carried in italics, NO marker fired, and each needs one page

⚠ Status after the 2026-08-29 plate gate: §3.1 is settled (ours, not Migne's — and the column relabelled 0032D); §3.2 sits on a page that was read with nothing recorded against it; §3.3–3.5 are on pp. 23 and 26, the two pages the gate did not reach, and remain owed. §3.6's [ed:] is unaffected — it makes no claim about type.

⭐ All three are on unread columns and all three are named here so a plate sweep is cheap. Under addendum A the first is a real word in an impossible slot and the second and third are non-words or near-non-words: a non-word suspicion has been right every time.

3.1 · 0032D · exores for uxores — RAISED, AND SETTLED AT THE PLATE: OURS, NOT MIGNE'S

✅ RESOLVED 2026-08-29. Read the ⛔ below before the entry. Plate p. 21 prints italic uxores. exores was Corpus Corporum's corruption; the Latin is patched (data/tei-patches/9006.json, direction: "corrupted" — the only one in the work) and the English now reads plainly, "that they who have wives be as those that have not." No marker is owed and none fired: a CC corruption is not a defect of Migne's type and takes no [sic:].

⛔ What this cost a reader, and it is the reason the entry is kept. For as long as it stood, our English published the non-word untranslated — "they who have exores be as those that have not" — presenting a digitization error to a reader as though it were Migne's Latin. The quarantine reasoning below is sound as reasoning and reached the wrong reader anyway, because the premise it rests on ("a real word reading oddly is Migne's until a plate says otherwise") is a rule about whose the word is, and the answer to that question was one page render away the whole time. The heading's column is also corrected, 0032C → 0032D, per §2.1.

The entry as originally written, kept intact:

ut qui exores habent, sint quasi non habentes. Exores is a well-formed Latin form (2nd sg. pres. subj. of exorare), which is why no [sic:] was fired — addendum A's finding is that a real word reading oddly is Migne's until a plate says otherwise. But it has no grammatical slot: qui … habent wants an accusative object and a finite verb cannot supply one. The Vulgate's uxores is certain, an x/u confusion at the head of the word is the commonest OCR shape in this corpus, and the whole passage is about a prophet forbidden a wife (the gloss's own ne … uxoris quoque et liberorum miseriis torquearis). Rendered by carrying the word untranslated in italics inside the already-italic quotation (nested italics are not expressible — Pattern 10's rule), with the Vulgate's uxores given in the [var:] beside it, so the sense is outside the quarantine (7a⁗). ⬜ Plate read owed, col. 0032, PL 114 p. 21. ← ✅ read; verdict above.

3.2 · 0031B · serta lignea for serra lignea — RAISED AS A PLATE CANDIDATE

Isaiam serta lignea secuit, of Manasses sawing Isaias asunder. Serta is a real noun (sertum, garlands) but only as a neuter plural; the slot here is a feminine ablative singular of instrument, which serta is not, while serra ("saw") is exactly that and is the received form of the legend. Rendered "he cut Isaias with a wooden serta" — the verb and its object stay in plain English outside the carried word, so the clause still has the thing it acts on. No [sic:]: a t/r confusion is the digitization's signature and blaming Migne's type for it is the class from which seven false [sic:] were withdrawn in the 2026-08-23 batch. ⬜ Plate read owed, col. 0031.

⚠ Page read, no verdict recorded. Col. 0031 is on p. 21, which the 2026-08-29 gate read; the read note for that page speaks only of 0031C/0032C ("heavy ink spatter … scan debris over legible type, NOT broken sorts — our Latin there is correct") and data/tei-patches/9006.json carries no patch at 0031B, so our Latin still reads serta lignea. That is corroboration, not a recorded confirmation, and it is worth exactly what it costs: no marker rests on it. Left as written.

3.3 · 0035B · Venit. for Jer. 18:18 Venite — RAISED, AND RENDERED AS PRINTED

VERS. 18's lemma is Venit.; Jer. 18:18 reads Venite, et cogitemus contra Jeremiam cogitationes, and the gloss is about the Jews' plotting contra Jeremiam vel Christum. Rendered "He came." — as printed, per Pattern 7 — and not expanded, because expanding a stump presupposes knowing which word the stump is. A dropped final -e is a transcription shape; no marker. ⬜ Plate read owed, col. 0035 — STILL OWED. Col. 0035 falls on p. 23, which is one of the pages the 2026-08-29 gate did not read (it read pp. 10–22, 24, 27–36). No marker stands here, so nothing is blocked.

3.4 · 0035A · ege for lege — a NON-WORD, carried

quando nullus Judaeorum terram et urbem sanctam ingredi ege permittitur. Ege is no Latin word; lege ("by law") construes and is what Jerome's sentence about Hadrian's edict requires. Carried in italics — "is permitted by ege to enter" — with the negation nullus and the verb outside the carried run. No [sic:], for the reason in §3.2. ⬜ Plate read owed, col. 0035 — STILL OWED, same unread page as §3.3.

3.5 · 0041A · in furore. Domini. — a full stop inside a phrase

conticescent bona pacis et auferentur in furore. Domini. Rendered exactly so: "shall be taken away in fury. Of the Lord." in furore Domini is the obvious phrase and the stop is almost certainly not Migne's, but Pattern 8 forbids deleting a printed mark and the ship test forbids asserting the plate is at fault. Recorded; not repaired, not marked. ⬜ Plate read owed, col. 0041 — STILL OWED; col. 0041 is p. 26, also unread. No marker stands here.

3.6 · 0044A · the empty parenthesis — the ONE [ed:] of the batch

quae Hebraice suarim ( ). An empty parenthesis where Migne sets a Hebrew word, in a sentence that exists to give the Hebrew (Theodotion's suarinas, Symmachus's novissimas, "which in Hebrew are suarim"). This is the 11055 signature — an empty element where the plate has text — and rendered faithfully it reads as our bug to any reader who does not open the notes. [ed:] fired per Pattern 13a clause 3: the hole is marked, nothing is supplied, and the note says recovery needs the plate. It is the only place in five chunks where the edition speaks in its own voice.

§4 · DIVERGENCES FOUND AND DECLINED — the eleven, with reasons

Declining is not overlooking. Each of these is a real difference from the Clementine; none took a marker, and the ground is stated per item.

4a. Declined because the SENSE is unchanged (Pattern 14's own bar)

1. 0038B · Super eos suscitabo pastores [n: (Ezech. XXXIV)]. Not Ezech. 34 at all: the words are Jer. 23:4, Et suscitabo super eos pastores — the very chapter being glossed, with only the word order inverted. So there is no textual divergence; the SIGLUM is wrong. Citation repair belongs to data/citation-corrections.json at index time, never to the running text (the Ruth precedent). Two more of the same class in these chunks: 0032A credita sunt illis eloquia Dei is cited (I Cor. I) and is Rom. 3:2; 0033C Voluntarie nobis peccantibus… is cited (Hebr. VI) and is Heb. 10:26. ⚠ Do NOT correct any of these in the text.
2. 0035B and 0038B–C · Rom. 11:25–26, quoted twice and differently. 0035B ends an abridgment usque ad tunc omnis Israel salvus fiet; 0038C has ut postquam intraverit, plenitudo gentium, tunc omnis Israel salvus fiat. The Clementine reads donec plenitudo gentium intraret, et sic omnis Israel salvus fieret. Both are the glossator weaving Paul into his own syntax (a resumption tail; a ut … tunc clause) and both assert exactly what Paul asserts. Declined — and declined together, deliberately: firing on one and not the other would have been the incoherence the within-work reconciliation rule forbids.
3. 0033A · Jer. 1:6 A a a Domine, nescio loqui for A, a, a, Domine Deus, ecce nescio loqui, quia puer ego sum. Compression of a familiar verse, no word changed.
4. 0033C · Luke 12:47 servus sciens et non faciens voluntatem Domini, vapulabit multis for the Vulgate's much longer Ille autem servus qui cognovit voluntatem domini sui…. A condensation; every element of the sense survives, both negatives included.
5. 0033C · Heb. 10:26, 0032A · Matt. 21:43, 0034B · John 14:6 (Ego sum vita for Ego sum via, et veritas, et vita) — all truncations or compressions that change nothing.
6. 0033C · 1 Cor. 10:4 Petra enim Christus for petra autem erat Christus; 0036C · Eph. 5:13 Omne quod manifestatur lux est for lumen est (and English has no exponent — both are "light"); 0031B · 1 Reg. 12:23 Absit a me hoc peccatum, quia cessem orare pro vobis for Absit autem a me hoc peccatum in Dominum, ut cessem (late-Latin quia introducing a noun clause is not a different assertion).
7. 0033B · Jer. 31:29 / Ezech. 18:2 patrem comedisse uvam acerbam, et dentes filiorum obstupescere. The accusative-and-infinitive is forced by the gloss's own Injusta sententia est; the singular patrem against the Vulgate's Patres is the one real difference, and it is a two-letter one — §4b.
8. 0040C · Jer. 25:25 Elam et reges Medorum for cunctis regibus Elam, et cunctis regibus Medorum: a compression inside the gloss's own sentence, not a quotation offered as text.

4b. Declined on the SHIP TEST — real differences of a shape the digitization manufactures

⛔ These are the eleven-minus-list's sharp end, and every one of them would have been a plausible [var:]. None was fired, because a one- or two-letter difference is exactly what Corpus Corporum produces, and 9003's collation showed the digitization MANUFACTURING a Vulgate divergence that a collation then dutifully "found."

• 0031D · Nunquid for Numquid (Jer. 15:12; also 0035A, 0037D, 0038D) — orthography, Pattern 9, not a defect at all.
• 0040A · A tertio decimo anno for A tertiodecimo anno (Jer. 25:3) — word division of a compound numeral; both forms are ordinary Latin.
• 0040C · Philisthiim et Ascalonis for …et Ascaloni (Jer. 25:20) — genitive for dative, and English has no exponent in a list of place names.
• 0040C · Cum impleti for Cumque impleti (Jer. 25:12) — one enclitic, at a lemma head.
• 0041A · Ecce egredietur for Ecce afflictio egredietur (Jer. 25:32) — a dropped noun. Raised because a dropped word is what §4b exists to catch, and declined because a dropped word is also what OCR does. ⬜ plate read would settle it.
• 0042C · Verba prophetae for verba prophetarum (Jer. 27:14) — singular for plural, with an English exponent, and the gloss's own next words are plural (pseudoprophetas vocari prophetas). Rendered "the words of the prophet". ⬜ plate read owed, col. 0042.
• 0043B · Haec dixit Dominus for Haec dicit Dominus — rendered as the perfect, "Thus hath the Lord said," so the divergence survives into the English rather than being smoothed into the formula every reader expects. That is the 7a″ discipline; the marker was still withheld because dixit/dicit is one letter.
• 0034B · Novissimus inimicus destruetur mors for 1 Cor. 15:26 Novissima autem inimica destruetur mors — the masculine is what ties it to Novissimum ejus in §2.2, but English has no exponent for the gender (Pattern 9), so there is nothing for a marker to show. Recorded, not marked.
• 0031D · Reliquiae enim tuae in bonum against Jer. 15:11 Si non reliquiae tuae in bonum. This one is not a variant at all and the proof is on the same line: Migne prints the verse verbatim, Si non reliquiae tuae in bonum, two sentences later. The enim form is the glossator restating the Hebrew oath-formula as the assertion he says it is (Noli considerare praesentia tantum, sed futura). Declined, and it would have been a false claim.
• 0043C · Mors viro requies est — in no verse of the Clementine. The words correspond to Job 3:23 as the Septuagint has it, against the Vulgate's wholly different viro cujus abscondita est via. No [var:], because identifying the source verse is itself the inference, and Pattern 14 forbids speculating about why two readings differ. Recorded.
• 0033A · Cum hujusmodi nec cibum sumere for 1 Cor. 5:11 cum ejusmodi nec cibum sumere — "with such a one" either way.

§5 · UNEXPRESSED SUBJECTS — every supplied pronoun, named

Per the brief, each "he/she/it/they" the Latin does not print is named here with what fixes it.

• 0031B si liberi nepotesque similia gesserint ad posteros perveniunt — "they come down upon the descendants." Subject = the fathers' sins and their punishment, from the abridged Jerome on Ex. 20:5 that the usque ad has cut out. The alternative (the children themselves reaching their descendants) does not construe with ad posteros.
• 0031B Aegyptias carnes desideravit — "he longed for the flesh of Egypt." Subject = populus, masculine, lifted from the lemma's addressee (Retrorsum abiisti, Jer. 15:6, of Jerusalem/Juda). Not Jeremias: it is the people who went backward.
• 0031C confusa est in aeternum … perdens populum suum — "she is confounded … destroying her own people." Feminine subject = quae peperit septem of Jer. 15:9, the synagogue. Fixed by the verse, not by the sentence.
• 0032D si … voluerit — "if it shall have willed to deliver." The nearest and only nominative is sermo doctoris. A rival reading takes the doctor himself as subject and construes equally well; the nominative decided it, and the difference changes nothing in the argument.
• 0033A et quotidie deceptorum funera prosternunt in Ecclesia? — "they cast down." Third plural with no antecedent in our text: the abridgment removed it, and Jerome's sentence is about the deceivers/heretics. Supplied as a bare "they"; nothing narrower is warranted.
• 0033C sed et Petro donavit — "he gave also to Peter": Christ, named in the same clause (Petra enim Christus).
• 0034B et derelinquimus per diem — "we forsake [them] through the day." The object is Jer. 17:11's eas (the riches), which Migne does not print; supplied in brackets, our convention for a supplied word.
• 0035A Ubi est absque gratia Dei — "Where he is apart from the grace of God." The abridgment has cut the antecedent; a bare third singular is all the words license.
• 0036B Propheta putaverat … et populus aestimabat quod statim non venerat, non esse venturum — both negatives kept and both on their own hosts (venerat; venturum).
• 0042C sacerdotibus loquitur, et populo, qui jam interitum prophetis annuntiaverat — taken as the Lord who had already announced destruction by the prophets (prophetis ablative), because the sentence's next words are his own quoted Et ejiciam vos (Jer. 16:13). ⚠ The rival takes qui as populus and prophetis as dative ("the people, who had already announced destruction to the prophets"); it construes, and it is recorded here rather than chosen silently. The reading taken is the one the quoted verse supports.
• 0043C ideo mortuum mentiuntur — "they falsely say that he died": the haeretici named in the preceding gloss (0043B–C), the only third plural in view.
• 0044A qui idem erat regni Sedeciae — "which was the same [year]": annus is elided; supplied in brackets.

§6 · WHAT WAS CHECKED AND FOUND CLEAN

• Column anchors, all 55, checked against the word each stands beside in the Latin, not merely for presence (brief §2b). Ten stand mid-construction and were deliberately split to match Migne, and in each of these the English word order was bent to keep the anchor beside its Latin neighbour rather than moved to the nearest comfortable place: 0031C (ut | qui), 0032D (errore | quempiam), 0033C (apostolicis | viris), 0034B (mortuus | est), 0035A (arbitrium | significat), 0038A (Jechonia | dicatur), 0038C (plenitudo | gentium), 0040A (tredecimo | anno), 0040B (servum | meum), 0044D (principio | Epistolarum).
• Punctuation compared PARAGRAPH BY PARAGRAPH (finer than band; the twins have identical paragraph structure, 34 · 50 · 54 · 53 · 38, verified). Colons, semicolons and question marks are now zero-mismatch in every paragraph of all five chunks. ⚠ Seven mismatches were found and fixed by that test after the first draft read clean by eye — five supplied marks (0010 ×2, 0012 ×1, 0013 ×2, 0014 ×2) and one deleted semicolon at 0012 para 8, where Migne's quam in Ecclesia illa permanere; in qua fames sit had been smoothed into one clause. The by-total count would have hidden four of the seven, exactly as the brief warns.
• Question marks: 2 · 1 · 6 · 2 · 1, at parity, each on the Latin's own sentence. Two indirect questions (Quaeritur cum Urias eadem, Quaeritur quomodo Joachim rex, 0041D) carry no mark in Migne and were given none — Pattern 8b was not needed anywhere here.
• usque ad — decided per occurrence, addendum D. The five chunks print it 107 times, by far the heaviest run in the Glossa block so far. 106 are the abridgment formula and are rendered as far as in their own italic span, per the settled 8978/8970 convention, with no verb ever supplied to complete a resumption tail. ⭐ One is NOT the formula and was caught: 0032B, A quintodecimo anno Tiberii Caesaris usque ad subversionem templi — an ordinary "from … to", and unitalicized in Migne, which is the tell (every one of the 106 sits inside asterisks). Rendered "From the fifteenth year of Tiberius Caesar to the overthrow of the temple". ⚠ Three further occurrences look like look-alikes and are not: at 0033D, 0035B and 0041C the formula and its resumption sit inside one italic span (usque ad Quiescite ab homine… · usque ad tunc omnis Israel salvus fiet. · usque ad: In veritate enim misit me Dominus ad vos.), so the English keeps them inside one span too.
• Negation fidelity. Every printed non/nec/neque/ne/nisi/nullus/nihil/nunquam/nondum was matched to the single Latin word it attaches to and checked in the English. The 2a trigger population — a negative near a quantifier or correlative — fired at seven sites, and at each the host was proved from structure, not from what read well: 0031A ut Creatore neglecto non universa creatura consurgeret (host universa, proved by the enumeration of exactly four plagues that the clause exists to explain — "not the whole creation") · 0031B qui non in toto orbe terrarum (host the phrase, not fuit) · 0032A non apparuit ultra … sed tantum credentibus · 0032A Non est … causa ardoris in Domino, sed in his · 0033D Non solum de corporeis simulacris · 0035D vocabula non urbem significant, sed locum · 0037B vitio suo accideret eis, non duritia comminantis. Also 0040C non solum contra Babylonem, sed contra omnes gentes and 0043D's four-negative chain (ne accipiat … nec faciat … non habentibus … ne accipiant), all four carried.
• Sections read clean, with nothing to report: CAPUT XXIV entire (0039C–0039D), CAPUT XXVI entire (0041A–0041D), CAPUT XXX (0044C–0044D). Their lemmata agree with the Clementine, their negatives are simple, and no word in them is a candidate.
• Migne's own printing, reproduced and not tidied: ## CAPUT XXV prints without a final stop where every other head has one — carried as ## CHAPTER XXV. VERS. 14. at 0038D prints without the .-- that every other verse address carries — carried verbatim. At 0040C the siglum for Jerome is set as italic body text, (ID), not as a note, and is carried as printed rather than promoted to [n: …]. Chunk 0014 ends with no terminal stop (…ante arcam Domini [n: (I Par. XIII)]) and the English ends with none either.
• Mention vs. use, 0035C. Lagunculam figuli pro laguncula figuli — Jerome is discussing the case of the noun, which English cannot show. Both forms are kept in Latin, the first italic as Migne sets it and the second roman as Migne sets it. The same treatment at et pro Hierihoh Chiricho., and for the versions' readings at 0038A (suarinas, sterilem, non crescentem, vacuum, abominabilem et abdicatum) and at 0038B (excelsus Domini), which are Aquila's, Symmachus's and Theodotion's words, not the Vulgate's, and take no [var:] for that reason.
• Pattern 4 [f:] was NOT used anywhere, per the runbook's known-false-positive 4: this is a lemma-and-gloss commentary, not a florilegium.
• Em-dashes: none in these five chunks.

§7 · THE PLATE GATE, 2026-08-29 — two [sic:], one withdrawn claim, one restored address

Bands 0031B–0044D fall on pp. 21–28 of patrologiaecurs06saingoog. The gate read pp. 21, 22, 24, 27 and 28 of that span and did not read pp. 23, 25, 26 — recorded in data/plate-reads.json (works.9006), patched in data/tei-patches/9006.json. What came back changed four things in this file, in three different directions.

⛔ The direction this batch's whole policy was blind to. §1's ⛔ rule withheld [sic:] wherever a letter-level defect might be the digitization's rather than Migne's — sound, and it is why no false accusation was fired. But it is a rule about one direction of departure. The plate returned the other: Corpus Corporum silently repairing a defect in Migne's own type. In that direction our Latin reads correctly, the token is an ordinary word, every mechanical check passes, and there is nothing at all to be suspicious of. Both [sic:] below are of that class. The Latin is now patched back to what the plate prints, and the [sic:] is what tells the reader the defect is Migne's.

7a · 0033D (chunk 0010) · pμtantes — a Greek mu in a Latin word

A foul case from the Greek sorts, standing in an ordinary Latin word in an ordinary sentence. ⚑ The marker does not carry the force: strip it and "thinking those things to be true which they had feigned" is complete (7a⁗).

7b · 0037A (chunk 0012) · Nuntlis — an i set as an l

⚠ This is precisely the shape §1's rule refused to accuse Migne of — one letter, and i/l is a classic OCR confusion. The rule was right and this marker does not contradict it: the rule forbids firing on suspicion, and this fires on a plate. Note also that the confusion runs the opposite way from the one the rule anticipated — the digitization did not manufacture the defect, it erased one.

7c · 0043A (chunk 0014) · a VERSE ADDRESS restored — and no [sic:] is owed

⚠ Logged here because it appears nowhere else in this file, and it is not a marker.

⛔ CC was right on the merits, and that is exactly what made the departure invisible. Jer 28:3 is the vasa verse — §1a above lists 28:3 Omnia vasa among the lemmata checked and found to agree with the Clementine, and it agreed because it had been silently corrected before it reached us. A checker collating our address against the Clementine confirms a number Migne did not print. No collation we own can catch that class.

Restored, per the standing rule that Migne's own printing is not ours to tidy (GLOSSA-NT-AGENT-BRIEF.md): VERS. n.-- addresses pass through verbatim, and out-of-sequence addresses are reproduced and logged, never reordered. §1c of cruces-0020.md records three more of his versification disagreements on the same principle.

⛔ No [sic:], deliberately. A [sic:] accuses Migne's type, and marks a word of his text. VERS. 5.-- is an address, not a word of the text — it is the apparatus by which he points at a verse, it takes no marker in any of the twenty-four chunks, and marking it would put a defect claim on a piece of furniture. The record of the departure is this entry.

Bands 0045A–0056A, CAPUT XXXI through CAPUT XLVIII (Jer 31–48, including the opening of the oracles against the nations). Five chunks, 4,272 Latin words, one translator.

Conventions taken as settled from GLOSSA-NT-AGENT-BRIEF.md and src/english/8968/cruces.md §1: VERS. n.-- verbatim, lemma Englished, fragments rendered as the phrase they stand for, sigla as ordinary [n: …]. usque ad → as far as (the corpus's settled rendering, 1,344 instances), with no verb ever supplied to complete an abridged clause.

Apparatus fired: 10 [var:] · 1 [sic:] · 1 [nt:] · 0 [cj:] · 0 [d:] · 0 [ed:].

⛔ CORRECTED 2026-08-29. That line and the table below it are BOTH out of date, and in two different ways. Counted against the chunks, this batch now stands at 10 [var:] · 5 [sic:] · 2 [cj:] · 1 [nt:].

• [sic:] 1 → 5. Four were added by the plate gate: 0047A Jean, 0048C uo n, 0048C Qomodo, 0055C erum (§3). All four are the normalization direction — Corpus Corporum silently repaired a defect in Migne's own type, so our Latin read correctly and no check we own could see the site. §2.1's sedet is unaffected and was confirmed at the plate.
• [cj:] 0 → 2. ⛔ This is a correction of a claim, not a stale count. §2.2 and §2.3 below are each headed [cj:] DECLINED, and in both places the marker is in the chunk: 0046D carries [cj: sic; read sicut, "as a garment"] and 0054C carries [cj: gladium; read gladio, "with the sword"]. Each section is corrected in place and the reason the decline was reversed is given there.

Column anchors: verbatim, in order, and each checked against the word it stands beside in the Latin (brief §2b), not merely for presence. Punctuation compared band by band with the Latin twin — colons, semicolons, question marks, exclamations, guillemets — zero mismatched bands after the one repair recorded at 0046A below. verify-english.mjs clean on all five.

§1 · THE 7a″ DIVERGENCE LIST — every lemma in bands 0045A–0056A, checked against the Clementine

Checked against sources/vulgate/clementine-flat.txt. Agreements are listed as well as divergences, because a findings-only list cannot be told from a report by an agent that never looked. Where Migne differs only in punctuation or in an orthographic variant of the corpus's own kind (quaecunque/quaecumque, coelum/caelum, exsecrabilis/execrabilis), that is recorded as agreement, not divergence — Pattern 9.

1a. Lemmata that agree with the Clementine exactly (allowing Migne's own pointing)

Jer 31 (chunk 0015): 31:1 In tempore · 31:3 Longe Dominus apparuit · 31:5 Plantabunt · 31:7 Reliquias Israel · 31:8 Ecce ego adducam · 31:8 Claudus · 31:9 In fletu venient et in (Clem. has a comma after venient; wording identical) · 31:9 Et Ephraim · 31:11 De manu potentioris · 31:14 Animam sacerdotum · 31:14 Pinguedine, et populus meus bonis meis adimplebitur, ait Dominus (verbatim) · 31:15 Vox in · 31:16 Et revertentur · 31:18 Eruditus sum · 31:19 Quoniam sustinui · 31:20 Si filius honorabilis (both occurrences) · 31:20 Idcirco conturbata · 31:21 Statue tibi · 31:21 Dirige cor tuum · 31:21 Revertere, virgo, Israel, revertere ad civitates tuas istas (Migne adds a comma after virgo) · 31:23 Adhuc dicent · 31:23 Benedicat tibi · 31:25 Quia inebriavi · 31:26 Ideo quasi · 31:27 Ecce dies veniunt dicit Dominus · 31:30 Obstupescent dentes · 31:31 Ecce dies · 31:34 Et non docebit · 31:34 Omnes enim cognoscent · 31:35 Qui dat solem · 31:36 Si defecerint · 31:37 Si mensurari · 31:38 Ecce dies.

Jer 32–33 (chunk 0016): 32:1 Verbum quod · 32:2 In atrio carceris · 32:3 Clauserat enim eum · 32:3 Quare vaticinaris · 32:5 Visitem eum · 32:7 Ecce Hanameel, filius (Migne's comma) · 32:7 Eme tibi agrum (Clem. agrum meum; the lemma stops before meum, a clipped fragment, not a divergence) · 32:8 In terra Benjamin · 32:13 Et praecepi Baruch · 32:14 In vase fictili · 32:15 Adhuc possidebuntur domus et agri · 32:18 Qui facis misericordiam · 32:18 Fortissime, magne · 32:19 Magnus consilio · 32:19 Ut reddas unicuique · 32:20 Qui posuisti signa · 32:20 Et portenta · 32:21 Et eduxisti populum tuum · 32:23 Et ingressi sunt · 32:23 Et non obedierunt · 32:24 Ecce munitiones exstructae · 32:24 Ecce munitiones · 32:24 Et quaecunque locutus es (orthographic only) · 32:24 Et tu dicis mihi, Domine Deus, eme agrum · 32:25 Et tu dicis · 32:27 Ecce ego Dominus · 32:28 Ecce ego tradam civitatem · 32:28 Civitatem istam · 32:29 Et venient Chaldaei · 32:29 Et domos in quarum domatibus · 32:30 Jugiter facientes malum · 32:33 Et verterunt ad me terga · 32:33 Cum docerem eos · 32:34 Et posuerunt idola sua · 32:35 Et aedificaverunt · 32:37 Ecce ego congregabo · 32:37 Et reducam eos ad locum istum, et habitare · 32:38 Erunt mihi in populum, et ego · 32:39 Et dabo eis · 32:40 Et timorem meum dabo in corde eorum · 32:41 In toto corde meo, et in tota anima mea · 32:41 Quia haec dicit Dominus · 33:1 Et factum est · 33:3 Clama ad me, et exaudiam te (the gloss's SECOND quotation — verbatim Clementine; see §1b.6) · 33:10 Adhuc audietur in loco isto · 33:12 Accubantium gregum · 33:14 Ad domum Israel: (colon inside the italics is Migne's; carried) · 33:17 Haec dicit Dominus · 34:1 Verbum quod factum · 35:1 Verbum quod factum est.

Jer 36–45 (chunk 0017): 36:3 Si forte audiente domo Juda · 36:4 Vocavit ergo · 36:9 Praedicaverunt jejunium in conspectu Domini · 36:16 Igitur cum audissent · 36:18 Ex ore suo loquebatur · 36:23 Et projecit in ignem · 36:26 Abscondit autem · 36:30 Propterea · 37:1 Et regnavit rex Sedecias · 37:3 Ora pro · 37:4 Igitur exercitus · 38:7 Audivit autem Abdemelech · 38:17 Si profectus exieris · 38:22 Ecce · 39:17 Et liberabo te · 40:1 Sermo qui factus est · 40:2 Tollens ergo · 40:3 Et factus est vobis · 41:8 Decem viri reperti sunt inter eos (Clem. Decem autem viri; the lemma opens after autem) · 44:15 Responderunt autem Jeremiae · 44:26 Vivit Dominus · 45:1 Verbum quod locutus est Jeremias.

Jer 46–47 (chunk 0018): 46:1 Quod factum est verbum · 46:9 Ascendite equos · 46:11 Ascende in Galaad · 46:11 Tolle resinam · 46:12 Audierunt gentes · 46:14 Annuntiate Aegypto · 46:18 Quoniam sicut Thabor · 46:20 Vitula · 46:21 Mercenarii · 46:26 Et post haec · 47:1 Quod factum est verbum · 47:4 Et dissipabitur · 47:4 Reliquias insulae Cappadociae · 47:5 Venit · 47:7 Quomodo quiescet.

Jer 48 (chunk 0019): 48:1 Ad Moab haec dicit Dominus · 48:6 Salvate animas vestras · 48:9 Date florem · 48:10 Maledictus qui facit · 48:10 Maledictus · 48:10 Et maledictus · 48:11 Fertilis · 48:12 Mittam ei · 48:17 Consolamini · 48:25 Abscissum · 48:29 Audivimus · 48:29 Audivimus superbiam Moab · 48:31 Ideo super Moab ejulabo · 48:31 Moab universam clamabo (Clem. ad Moab universam clamabo; clipped) · 48:32 Plorabo tibi, vinea Sabama · 48:32 Vinea Sabama · 48:34 A Segor usque ad Oronaim · 48:34 Aquae quoque Nemrim · 48:35 Et auferam de Moab · 48:37 Omne enim caput calvitium · 48:38 Quoniam contrivi Moab · 48:41 Et erit cor fortium Moab · 48:45 In umbra Hesebon steterunt · 48:46 Vae tibi, Moab · 48:47 Et convertam captivitatem Moab.

Inline scripture that agrees verbatim: Isa 9:2 qui ambulabat in tenebris, vidit lucem magnam · Ps 24:7 Delicta juventutis meae et ignorantias meas ne memineris · Ps 18:3 Dies diei eructat verbum, et nox nocti indicat scientiam · Mt 24:35 Coelum et terra transibunt · Isa 2:2 (unitalicized in Migne, and left unitalicized here) · Isa 19:22 Percutiet Dominus Aegyptum plaga et sanabit eam · Gal 5:24 Qui autem sunt Christi carnem suam crucifixerunt · Deut 32:42 Gladius meus devorabit carnes (see §1b.13) · Acts 20:27 Non enim subterfugi, quominus annuntiarem omne consilium Dei vobis · Isa 15:6 Aquae enim Nemrim desertae erunt · John 16:20 Vos autem contristabimini, sed tristitia vestra vertetur in gaudium.

1b. Divergences — WITH a [var: …] fired

1. 0045B · III Reg. 18:21 · Usquequo claudicatis utroque pede? — Clementine reads Usquequo claudicatis in duas partes? Migne prints utroque pede, "on both feet", and the gloss's whole point is the lame man of Jer 31:8 (Claudus), so the image of the feet is what the exposition rests on. [var: Vulg. III Reg. 18:21 in duas partes, "on two sides"]. Rendered "How long halt ye upon both feet?"
2. 0045D · Jer 31:18 · Erudisti me, Domine, et eruditus sum — Clementine Castigasti me, et eruditus sum. Migne substitutes Erudisti for Castigasti and adds Domine. Material: the whole VERS. 18 gloss above turns on correction (Correctio proficit in salutem), and Migne's own lemma at VERS. 18 is Eruditus sum, which agrees with the Vulgate — so the divergence is confined to this re-quotation. [var: …] fired, and the VERS. 18 lemma left untouched.
3. 0046B · Jer 31:26 · ideo expergefactus sum — ⚠ this heading read 0046A and is corrected to 0046B (2026-08-29): in 0015.md the [0046B] anchor stands four lines above the marker and the next anchor is [0046C], so the [var:] is under 0046B, which is where plate-gate.mjs resolves it and what migne.app/pl/114/46b will show. The marker was not moved. — Clementine Ideo quasi de somno suscitatus sum. Migne's citation drops the quasi, turning a simile into a plain awakening, and his own next lemma (VERS. 26) is Ideo quasi, i.e. he prints both readings four lines apart. [var: Vulg. Jer. 31:26 Ideo quasi de somno suscitatus sum].
4. 0048B · Jer 32:24 · Ne capiatur — Clementine ut capiatur. A negation divergence in a lemma: Migne's "lest it be taken" against the Vulgate's "that it may be taken", in a verse about the bulwarks raised against the city. Rendered as printed, negation carried ("Lest it be taken"), [var: Vulg. Jer. 32:24 ut capiatur, "that it may be taken"]. The gloss beneath it (Non habent quos vincant, sed quos capiant) reads with the Vulgate's sense, which is exactly the kind of internal tension the marker exists to expose; it is not grounds for emending the lemma.
5. 0048D · Zach 7:11 · Vertunt contra me scapulam recedentem — Clementine averterunt scapulam recedentem. Migne's contra me is not in the Vulgate at all, and the gloss argues precisely from turning the back against God (qui vero tergum vertit… se negligere comminantem). [var: …] fired.
6. 0049A · Jer 33:3 · Clama ad me, et ego exaudiam eum — Clementine Clama ad me, et exaudiam te. Migne's lemma reads "and I will hear him"; four lines later he prints the Vulgate reading himself (Nunc autem dicitur: Clama ad me, et exaudiam te), and the gloss distinguishes the two. Both rendered as printed; [var:] on the lemma only. ⚠ Note that this [var:]'s quoted Vulgate reading also occurs in the Latin twin, in the second quotation — the verifier's mirror test passes because the marker's content (witness + reading + gloss) is not a verbatim substring, and the coincidence is itself the finding.
7. 0050D · Ps 24:4 · Confundentur omnes iniqua agentes — Clementine Confundantur omnes iniqua agentes supervacue. Indicative for subjunctive: Migne turns an imprecation into a prediction, and the gloss uses it as a prediction (Ecce principes… captivantur). [var: Vulg. Ps. 24:4 Confundantur omnes iniqua agentes supervacue].
8. 0051B · Jer 41:8 · habemus thesaurum — Clementine habemus thesauros in agro (plural). Migne's singular is what the exposition immediately builds on (Thesaurus enim in agro, Mt 13:44). [var: Vulg. Jer. 41:8 habemus thesauros in agro].
9. 0054C · I Cor 15:50 · Caro enim et sanguis regnum Dei non possidebunt — Clementine caro et sanguis regnum Dei possidere non possunt. Future indicative for the modal: "shall not possess" against "cannot possess". [var:] fired. (Migne's note reads (II Cor. XV); his citation, carried verbatim per the register, and not corrected — runbook item 2.)
10. 0055C · Isa 15:5 · vitulam consternantem — Clementine vitulam conternantem. This is the sharpest of the set: the gloss immediately explains the phrase as id est, perfectae aetatis, "that is, of perfect age", and then computes the third year — which only follows from conternantem (three-year-old), not from consternantem ("dismaying"). Migne's printed word is rendered as printed and the received reading marked: [var: Vulg. Isa. 15:5 vitulam conternantem, "a heifer of three years old"]. ⚠ The Clementine's own conternantem is the odd form here and sources/vulgate/README.md warns that a call turning on a single letter should say so: it does, and this one turns on an s.

1c. Divergences found and DECLINED the marker — with the reason

Pattern 14's bar is a divergence "in a way that changes the sense". These are real, checked, and recorded; none of them meets it, and Addendum B's ship test says a marker is a public claim.

• 0045A · Deut 28:30 (Migne's note) · Plantate plantaria et vindemiate — the Clementine's Deut 28:30 reads Plantes vineam, et non vindemies eam, the exact opposite in polarity, and the citation cannot be made to fit any Deuteronomy verse. Two possibilities — Migne's note is misplaced, or he quotes a non-Vulgate text — and a [var:] must name a book chapter:verse it can stand behind. Declined; the reference is his and passes through verbatim.
• 0045B · Jer 31:7 (Migne's note (III Reg. XVIII)) · Personate, canite — the words are Jer 31:7's own (personate, et canite), minus the et, so the quotation agrees with the verse it glosses; only Migne's citation points elsewhere. Declined (runbook item 2: never correct a reference).
• 0045B · Rachel plorans — Clementine Jer 31:15 fletus Rachel plorantis filios suos. A citation-form nominative for the Vulgate's genitive, with no change of sense. Declined; rendered "Rachel weeping".
• 0046A · Apoc 2:5 · Memento unde excideris — Clementine Memor esto itaque unde excideris. A lexical substitution of exactly synonymous force ("remember"). Declined.
• 0046D · Jer 31:38 · Usque ad anguli portam (lemma at VERS. 39) — Clementine 31:38 ends usque ad portam anguli; Migne inverts the genitive and numbers it 39. Word order only, no sense change; the verse number is his and is reproduced as printed. Declined. ⚠ This is an usque ad LOOK-ALIKE, not the abridgment formula (brief addendum D): the words are the verse's. Rendered "Even to the gate of the corner", not "as far as".
• 0047D · John 17:6 · Pater, manifestavi nomen tuum hominibus — Migne prefixes the vocative Pater (from Jn 17:1/5); the rest is verbatim. Declined.
• 0048A · Exod 24 · Quaecunque praeceperit Dominus faciemus — no citation is printed, and the nearest Clementine readings are Exod 24:3 Omnia verba Domini, quae locutus est, faciemus, 24:7 Omnia quae locutus est Dominus, faciemus and 19:8 Cuncta quae locutus est Dominus faciemus. A [var:] needs a verse to cite; three candidates is none. Declined.
• 0049A · Amos 5:21 · festivitates vestras odi — Clementine Odi, et projeci festivitates vestras. Compression and word order, same words, same assertion. Declined.
• 0049A · Jer 32:35 · In valle filii Ennon — Clementine in valle filii Ennom, final m. ⚠ This is the class the launch brief warned would tempt hardest. Rendered Ennon as Migne prints it, not the familiar Ennom/Hinnom; a one-letter transliteration variant of a place name asserts nothing different, so no marker.
• 0050C · Ps 144:20 · Custodit enim Dominus omnes diligentes se… — Migne inserts a connective enim; otherwise verbatim. Declined.
• 0050C · Prov 28:9 · Qui enim avertit aurem suam ne audiat legem… — Clementine Qui declinat aures suas ne audiat legem. Lexical and number variation of the same act; ne audiat and oratio ejus erit exsecrabilis both stand. Declined (and exsecrabilis for execrabilis is Pattern 9 orthography, not a defect).
• 0050D · Jer 38:5 · Et dicit rex Sedecias — Clementine Et dixit. A tense shift in a narrative formula; rendered as the present Migne prints ("And the king Sedecias saith"). Declined on the sense bar.
• 0050D · Qui sperant in Domino non confundentur [n: (Sap. XIX)] — Wisdom 19 carries nothing of the kind and the words are not in the Clementine anywhere; the nearest is Ps 24:3 universi qui sustinent te, non confundentur. No verse can be cited with confidence. Declined; the negation non is carried into the English.
• 0051B · Jer 41:16 · Johanan filius Careae — Clementine Johanan filius Caree. Spelling of a patronymic. Rendered "the son of Careas" (the nominative Migne's genitive implies) and recorded here; no marker.
• 0052A · Sir 15:9 · Non enim est pulchra laus in ore peccatoris — Clementine Non est speciosa laus in ore peccatoris. pulchra for speciosa, both "comely"; plus a connective enim. Declined.
• 0052D · John 10:12 · Mercenarius autem, et qui non est pastor, cum videt lupum venientem… — Clementine has cujus non sunt oves propriae between, and videt for cum videt. The span sits inside an usque ad resumption tail, where removal of words is the formula's own work; declined. Both printed negations (non est pastor) are carried.
• 0052D · Rom 11:25–26 · Quia ex parte caecitas contigit in Israel, et cum plenitudo gentium subintraverit, tunc omnis Israel salvus fiet — Clementine has quia caecitas ex parte contigit… donec plenitudo gentium intraret / et sic omnis Israel salvus fieret. Migne conflates the two verses with cum… tunc for donec… et sic. The assertion is identical and the span is an usque ad resumption tail; declined, with the conflation recorded here.
• 0052D · Gaza futura est in tumulum sempiternum — not in the Clementine; the nearest is Soph 2:4 Gaza destructa erit. No citation is printed. Declined for want of a verse.
• 0053D–0054C · Heb 9:22 · sine sanguinis effusione non fit remissio peccatorum — Migne adds peccatorum, which the Clementine does not have. An addition that specifies rather than alters; declined. The printed non is carried ("there is no remission").
• 0054C · Acts 20:26 · Contestor hodie quia mundus sum a sanguine omnium — Clementine Quapropter contestor vos hodierna die, quia…. Compression only; v. 27 is verbatim. Declined.
• 0055C · Isa 15:5 · usque Segor — Clementine usque ad Segor. Migne drops ad. ⚠ Another usque ad look-alike neighbourhood: usque Segor here, and in the same paragraph usque ad extremos fines provinciae — both the ordinary preposition, neither the formula, and neither italicized. Declined.
• 0055D · Isa 15:2 · In cunctis capitibus ejus calvitium, omnis barba radetur — Clementine has et omnis barba radetur. A dropped connective. Declined.
• 0055D · Joel 2:32 · quicunque invocaverit nomen Domini, salvus erit — Clementine omnis qui invocaverit. Synonymous. Declined.
• 0052B, 0052D · Manasse, Diocaesaraea, Nephthalim, Chebron, Enachim, Charchamis, Nabo, Chamos, Beelphegor, Oronaim, Sabama, Hesebon, Anathot, Ananehel, Rablatha — every proper name is rendered in the form Migne prints, not the familiar English Bible's. Three worth naming: Anathot (0047A; Clem. Anathoth), Ananehel (0046D; Clem. Hananeel), and Nemrim / Nimrim, which Migne spells both ways in one paragraph at 0055C (Aquae quoque Nemrim … Est enim Nimrim oppidum) — both carried as printed, and the inconsistency is his.

§2 · CRUCES — the printed text, where it is odd

2.1 · 0047A · sedet for sed et — the one [sic:], and why it fired

Migne prints Non solum verba, sedet opera prophetarum, sunt nobis in exempla virtutum. sedet ("he sits") is a real word with no grammatical slot in a clause whose verb is sunt and whose skeleton is the correlative non solum … sed et. That is Pattern 12's first test (11632's four markers on real words were ruled to keep it), and sedet for sed et is one of the fused runs Pattern 10 itself names.

⚑ 7a⁗ applied: the marker does not carry the force. The English reads Not only the words, but also [sic: sedet] the works… — the correlative's second member is in plain English beside the marker, so stripping every marker leaves a readable sentence.

⚠ Provenance is the open question and I am not asserting it. Addendum A's ten-for-ten split concerns non-words; a fused run is a well-formed-looking token that no mechanical test sees (7a⁗-b), and the withdrawn-[sic:] class of Addendum B was digitization damage blamed on Migne. Worth a plate read (PL 114 col. 47, first line of the CAPUT XXXII gloss). If the plate prints sed et, this marker comes out and the fix is a data/tei-patches/9006.json entry.

✅ READ, AND THE MARKER STANDS. Plate p. 29, 2026-08-29: Migne prints sedet fused, and at 6× the flanking word-spaces (verba | sedet | opera) are visibly wider than any inter-letter gap inside the run — which is the measurement that decides a fusion, and the reason the open question above is now closed rather than merely unrefuted. data/plate-reads.json works.9006 records it as the finding of the p. 29 read. ⚠ The same page then produced three further [sic:] that nobody had suspected (§3): reading the whole line, not the token, is what turned them up.

2.2 · 0046D · et sic vestimentum veterascent — will not construe. [cj:] DECLINED, THEN FIRED

Si coeli peribunt, et sic vestimentum veterascent… The subject of veterascent is coeli (plural), so vestimentum (neuter singular) has no slot; the sentence wants sicut vestimentum (cf. Ps 101:27 omnes sicut vestimentum veterascent), which is what the argument requires.

Rendered literally — "and so shall grow old a garment" — and no [cj:]: the two conjectures weighed were sicut (near-certain, from the psalm the clause echoes) and reading vestimentum as a stranded nominative (which fails on number). sicut wins on evidence, but Pattern 18a's bar is that the printed word makes the English assert something false, and an incoherent clause was ruled to pass that bar only for a negation or a correlative whose answering member is printed. Nothing answers sic here. Declined, with the conjecture recorded.

⛔ REVERSED, 2026-08-29 — and this heading said DECLINED while the marker was in the file. 0015.md reads: "If the heavens shall perish, and shall grow old so, a garment [cj: sic; read sicut, "as a garment"], how can the seed of Israel be everlasting?" The [cj:] is fired; the count in the header is corrected with it.

Why the decline was reversed. The reading above declined on a reading of 18a's bar as negation-or-answered-correlative only. That is too narrow for this site. sic here is not an unanswered correlative — it is a word doing a job it cannot do: the English forced to render it literally must write "shall grow old so, a garment," which asserts an apposition Migne's sentence does not make and leaves vestimentum dangling as though it were a second subject of veterascent. That is 18a's actual test — the printed word makes the English assert something false — and it is met. The conjecture is not mine either: sicut vestimentum veterascent is the clause of Ps 101:27 that this sentence is quoting, so the correction is the psalm's own words.

⚑ What the marker does and does not do. It proposes; it does not repair. The English still renders sic as printed ("so"), the dangling a garment is still visibly dangling, and the [cj:] stands beside it naming the conjecture. Strip the marker and the sentence is exactly what this section says it should be — Pattern 7a's silent repair is not what happened here, because nothing was silently repaired.

2.3 · 0054C · resecandum gladium verbi Dei — a stranded accusative. [cj:] DECLINED, THEN FIRED

fundendum est quidquid in membris nostris carnale et terrenum inolevit et resecandum gladium verbi Dei. Both gerundives agree with quidquid; gladium is accusative where the sense demands the ablative of means gladio ("cut away by the sword of the word of God"), which the preceding sentence supplies in so many words (Maledictus ergo qui prohibet gladium… a sanguine) and Gregory's note repeats.

Rendered as printed and left visibly broken: "must be poured out and cut away the sword of the word of God." Declined for the same reason as 2.2: a case slip is not a negation or a correlative, and 18a's ruling is expressly scoped. Writing "by the sword" would have been the silent repair Pattern 7a exists to forbid.

⛔ REVERSED, 2026-08-29 — and this heading said DECLINED while the marker was in the file. 0019.md reads: "…must be poured out and cut away the sword [cj: gladium; read gladio, "with the sword"] of the word of God." The [cj:] is fired; the count in the header is corrected with it.

Why the decline was reversed, and it is the same ground as §2.2. The decline rested on scoping 18a to negations and answered correlatives. The operative test is whether the printed word makes the English assert something false, and an accusative gladium standing as a second object of resecandum does exactly that: the English is made to say that the sword of the word of God is among the carnal things to be cut away, which inverts the argument — the sword is the instrument that cuts. Migne's own preceding sentence prints the instrument in so many words (Maledictus ergo qui prohibet gladium … a sanguine) and Gregory's note repeats it, so the ablative gladio is the conjecture the printed context forces, which is Pattern 18's requirement.

⚑ The English is unchanged and still reads broken — "cut away the sword of the word of God" — and the marker names the conjecture beside it. Nothing was smoothed; a false assertion is now labelled instead of standing unlabelled.

⚠ Related, and not the same site: 0054C also carries the [var:] on 1 Cor 15:50 (§1b). Two markers, one column, different claims.

2.4 · 0054A · versuram doctrinam philosophorum — real word, odd sense. Rendered as printed

fugiant versuram doctrinam philosophorum et haereticorum. The obvious conjecture is versutam ("crafty"), which fits the sentence perfectly. But versuram construes: versurus is the future participle of verto, so versuram doctrinam is "the doctrine that is about to turn" — and Addendum A is explicit that a real, well-formed word that reads oddly is Migne's until a plate says otherwise, with every such suspicion in the last batch refuted. Rendered "the turning doctrine"; no marker; conjecture versutam recorded and rejected here.

2.5 · 0051B · ut faceritis consilium (Isa 30:1) — non-word, or a medieval variant? NO marker

Clementine Isa 30:1 reads ut faceretis consilium. faceritis is neither faceretis (imperfect subj.) nor feceritis (perfect subj.). Corpus control: 2 occurrences of faceritis in src/latin/ (this chunk and 11057/0010, an unrelated work), against 14 feceritis and 1 faceretis. That an unrelated PL work prints the same form is the reason it is not treated as a plate corruption: an a/e confusion in the perfect stem is an attested medieval orthography, and Pattern 9 rules attested spellings correct Latin for their period. Rendered normally, "that you should take counsel"; no [sic:] — Addendum B's ship test forbids a public claim about the plate I cannot stand behind. Flagged for a plate read if one is being done for §2.1 anyway; the two sit twelve columns apart.

2.6 · 0053A · the Thabor entry and the Carmel entry have run together

…distans a Diocaesaraea decem millibus contra orientalem plagam. Qui confinium quoque inter tribus Issachar et Nephthalim fuit Carmelus, ubi Nabal Carmelus fuit et nunc villa est Carmela nomine…

In Jerome's Liber locorum, from which this note is abridged, the Thabor entry closes at fuit and Carmelus opens a new entry. Migne prints them as one sentence, so Carmelus stands as a second predicate to fuit and the mountain of Thabor is made to be Carmel. Rendered as printed, run-on intact: "Which was also the border between the tribes of Issachar and Nephthalim, Carmelus, where Nabal the Carmelite was…". No marker: nothing is defective type, and the collapse is the abridgment's, which is what this edition exists to show. Also in the same sentence: tribus (acc. pl.) after inter is correct; Diocaesaraea is Migne's spelling and is kept.

2.7 · 0047D · sed proprie nomen Dei. Pater est — Migne's full stop, kept

Haec nomina potentiam Creatoris indicant, sed proprie nomen Dei. Pater est, unde: Pater, manifestavi nomen tuum hominibus. The natural sentence is sed proprie nomen Dei Pater est ("but properly the name of God is Father"), which the proof-text then supports. Migne's period falls after Dei. Pattern 8: the mark rides as printed — rendered "but properly the name of God. He is Father, whence:". Recorded, not repaired.

2.8 · 0050C · Nequaquam pro se orare postulavit — two readings, and neither was chosen silently

Jer 37:3 has Sedecias sending to Jeremias Ora pro nobis, so he plainly did ask. The negation's host is therefore genuinely undecided between: (a) pro se — "he by no means asked [him to pray] for himself", i.e. he asked for the people, not for himself; and (b) postulavit — "he by no means asked to pray for himself", i.e. he sought another's prayer because he would not pray himself, which is what quia divinis mandatis obtemperare noluit and the Prov 28:9 proof-text about a man whose own prayer is abominable both support. Both construe. The English keeps Migne's word order and the fronted Nequaquam so that the scope stays where the Latin leaves it: "By no means did he ask to pray for himself, because he would not obey the divine commandments." Reading (b) is the better-supported one and is the one an English reader will take; (a) is recorded here rather than adjudicated in the text.

2.9 · Unexpressed subjects supplied, each named

Every "he/she/it/they" not printed in the Latin, with what fixes it:

• 0045A non merito salvatus est — Israel, from offenderat Dominum Israel in the same clause.
• 0045B in Samaria primus regnum obtinuit — Ephraim, named in the lemma above.
• 0045B interrogant legem (0045C, VERS. 14) — the subject is not the priests, whose mouth is the source (de quorum ore); it is the people who consult them, cf. Mal 2:7. Rendered impersonally, "of whose mouth they ask the law".
• 0045C flere dicitur — Rachel, named two words earlier.
• 0046A convertit — the Lord, printed in the clause.
• 0047A recludi jusserat — the object supplied is Jeremias, from Jer 32:2 (Jeremias propheta erat clausus in atrio carceris), not the king.
• 0047A Temperavit sententiam — the Lord, from ait Dominus in Jer 32:5.
• 0047D quod proprie praestiterit Israel — God, addressed in the second person through the prayer; transit / percurrit in the same sentence are the prophet. Two subjects in one sentence, and the change of person is Migne's.
• 0048A Quibus afflicta est — Egypt, from the lemma's signa et portenta in terra Aegypti (Jer 32:20); rendered "she". Israel ab omnibus separat in the same sentence is God.
• 0048B Primum vallata… capta est — the city, feminine, from the lemma Civitatem istam.
• 0050C ureretur — the body of Joakim, from Jer 36:30 (cadaver ejus projicietur ad aestum per diem, et ad gelu per noctem), which is the verse the gloss abridges. Rendered "he", the king being the person the vengeance falls on; "it" (the corpse) is the rival and would read the same in the Latin.
• 0051D Confitentur… non acquiescunt — the Jews in Egypt who answered Jeremias, from the lemma.
• 0053D exurit… praecipitabitur — the sword/plague of the lemma Venit (Jer 47:5), which the gloss has just moralized; the masculine is the Latin's, not a person.
• 0055D sperabat — Moab, singular, from filiorum Moab and the whole chapter's subject.

2.10 · Negation sites checked one by one (test 2 / 2a)

Every printed non · nec · neque · nisi · nullus · nihil · nunquam · nequaquam · ne appears in the English and none that is not printed. The sites where the host was a quantifier, correlative or predicate adjacent to a plausible verb — 2a's trigger population — were proved from structure, not from what reads well:

• 0045D non merito salvatus est → on merito, not on the verb ("saved not by merit").
• 0046C non austeritate ciborum, sed vitio dentium → on austeritate; the sed decides.
• 0046C non in natura qua… latuit → on the phrase in natura, not on latuit: putting it on the verb would say he did not lie hidden, which the etiam cum videretur concessive contradicts. Recorded because both parse.
• 0047A Non solum verba, sed et opera → on solum.
• 0047A non in carcere, sed in vestibulo → on in carcere; the sed decides.
• 0048A non enim in responsione, sed in opere praemium est → on in responsione.
• 0048B Non habent quos vincant, sed quos capiant → on habent, with the contrast carried by vincant / capiant; rendered "They have not those whom they may conquer, but whom they may take".
• 0048B Non omnium gentium vel populi Israel tantum → on tantum, not on an implied verb. "Not… only."
• 0048D Non solum tunc Judas posuit… sed usque hodie → on solum.
• 0050B aure corporis, sed non cordis → on cordis.
• 0052D non accipient ab eis sanitatem, sed confusionem → on sanitatem, not on accipient: the sed supplies the answering member. Rendered "shall receive from them not health, but confusion" — the host is the noun, and putting the negative on the verb would have made them receive nothing at all, which the sed confusionem denies.
• 0052D non victoriam veritatis accipiunt, sed… cadunt → on victoriam veritatis.
• 0053B non irreparabiliter… esse destructum → on the adverb irreparabiliter, proved by the proof-text: the usque ad tail resumes at Isa 19:22 Percutiet Dominus Aegyptum plaga et sanabit eam, so the destruction is reparable and the negative cannot sit on destructum.
• 0053B non bene convenit Judaeis → on convenit (modified by bene); the qui toto orbe dispersi that follows is the reason the promise does not fit them.
• 0054B ut non solum intueamur quod agimus, sed etiam qua discretione → on solum.
• 0053C nec munitio urbium, nec virtus idolorum valuit (0056A) → correlative pair, both carried.
• 0055C ut philosophi nec haeretici… aliquid inveniant → a single printed nec joining two subjects under one negative force. English's "neither … nor" is the two-word exponent of that one nec, not a supplied second negation; recorded so a counting check does not read it as an addition.
• Carried without incident: nihil medium (0048A) · Nemo ergo (0047D) · ne posset effugere and quasi non… fuerit (0047A) · ne desperent (0049D) · ne crederent et poeniterent (0050B) · nullam necessitatem (0050B) · noluit and non negavit (0049C) · non poterat (0051A) · non confundentur (0050D) · nec sermo eorum nec oratio (0052A) · non debebant, non vocatur (0052A) · nihil afflictionis (0054D) · si stultitiam non correxerint (0055A) · ut nihil probetur (0055A) · nunquam se cogitasse (0049A) · non possidebunt (0054C) · Non enim subterfugi (0054C) · Non potest gladius quiescere, nisi Domino jubente (0053D) · ne memineris (0045D) · nec… declinare consentit (0046A) · non merito (0045A) · sed non eam possedit, hostes expellere non valuit, sed eam non obtinuit (0053C).

2.11 · usque ad — decided per occurrence (addendum D)

Bands 0045A–0056A print the string usque ad 71 times. 67 are the abridgment formula and are italicized by Migne as usque ad (61 of them bare, plus usque ad, at 0046D with the comma inside the italics, usque ad finem capitis. twice at 0050A, and three tails where the resumption is inside the same italic run: usque ad Mercenarius autem…, usque ad: Percutiet…, usque ad: Quia ex parte caecitas… — the colon in the last two is Migne's and is kept). Four are NOT the formula — one lemma and three ordinary prepositions:

1. 0046D, VERS. 39 lemma Usque ad anguli portam — the verse's own words (Jer 31:38–39).
2. 0048A afflicta est usque ad diem hanc — Jer 32:20's own words, unitalicized.
3. 0052C extenditurque per desertum usque ad trans Jordanem — Jerome's topography.
4. 0055C usque ad extremos fines provinciae.

The near neighbours usque Segor (0055C, in the Isaiah citation two lines later) and usque hodie (0048D and 0053C) are look-alikes of the same family and were read the same way.

Each was read from what stands on either side of it, and none was allowed to swallow a lemma's words.

2.12 · Structure and typography read clean

• [nt: …] in band 0051B, CAPITA XLII, XLIII. Migne prints [n: ( Textus sacer sine expositione. )] as the entire content of a two-chapter section: this is editorial prose (a statement about the chapters), not a locator, so Pattern 15 applies and it is translated and re-marked in place — [nt: ( The sacred text without exposition. )], with his internal spacing kept. Note count and order are unchanged.
• Verse numbering. VERS. 39 at 0046D carries words the Clementine gives at 31:38, and CAPUT XXXIX prints a single gloss at VERS. 17. Both are Migne's, reproduced as printed, nothing reordered.
• VERS 1.-- without the period after VERS at 0054A — the only instance in these five chunks (every other verse mark prints VERS.). Reproduced verbatim; not patched, because unlike Ruth's VERS.9 this is a missing period rather than a missing space and I have not read the plate. Worth one look if a plate reader is at col. 54 anyway.
• Greek. ῥητίνην and ῥέω at 0052C–0052D carried verbatim, unaccented characters untouched.
• Guillemets: zero in the Latin, zero supplied. Quotation is by italics throughout, and no " " of ours was added.
• Em-dashes: zero in all five English chunks.
• Column-band arithmetic: 0051B is followed by 0051D with no 0051C, and 0044D by 0045A. Runbook item 1 — not reported as a gap, and no text breaks mid-word across any anchor.
• Pattern 4 [f:]: not applicable. This is a lemma-and-gloss commentary, runbook item 4.
• Frontmatter noteCount: 19 against 20 body notes in 0015, 14 against 15 in 0016. Runbook item 3 — the body count governs, and both English chunks match their Latin twins exactly.

2.13 · Read clean, and named as such

Read against the Latin sentence by sentence with nothing to report beyond §1: the whole of CAPUT XLVI–XLVII (chunk 0018, which fired no marker at all), CAPUT XL, XLI, XLIV, XLV (0051A–0052B), and CAPUT XXXIII–XXXV (0049C–0050A). The candidates examined there and rejected were: mistim (0052B — an attested variant of mixtim, Pattern 9, rendered "mixedly"); exsperturus (0052A — the -xs- orthography Migne uses in exspecto, readable, Pattern 9); decem tribubus (0049A — a dative of reference where a nominative subject would be expected, which construes, so rendered "for the ten tribes"); Haec est terminus (0053C — a gender mismatch with no English exponent, Pattern 9 and runbook item 6, left alone); and de Sedecias (0050A — a nominative after de, which English cannot expone either way in a proper name). None of these took a marker and none should.

§3 · THE PLATE GATE, 2026-08-29 — four new [sic:], all Migne's, none of them suspected

Bands 0045A–0056A fall on pp. 28–33 of patrologiaecurs06saingoog; all six were read at 300 dpi with the corner numbers checked first. Recorded in data/plate-reads.json (works.9006), patched in data/tei-patches/9006.json.

⛔ Read this before the four entries: not one of these sites was on anybody's list. §2 above raised every candidate it could see and every one of them is a suspicion pointing the same way — our text corrupt, Migne's clean. These four point the other way. In each case Corpus Corporum silently repaired a defect in Migne's own type, so our Latin read correctly, the token was an ordinary word in an ordinary sentence, and no automated check we own could ever have fired on it. There is nothing for frequency analysis to be suspicious of; the defect exists only on the page. All four were found by the standing rule to read the whole line, not the target word — three of them on p. 29, which was rendered to settle §2.1's sedet.

The Latin has been patched back to what the plate prints, and the [sic:] is what tells the reader the defect is Migne's and not ours.

3a · 0047A (chunk 0016) · Jean for Jam

⚑ The marker sits on "now", which is what Jam means and what the sentence needs; strip the marker and "The captivity being now at hand, the city being now consumed with pestilence and famine" is complete (7a⁗).

⚠ Column label. data/tei-patches/9006.json files this patch under 0047B; the word stands immediately before the [0047B] anchor, so by anchor position — and in plate-gate.mjs, and at migne.app/pl/114/47a — it is under 0047A, the chunk's colContext. The word sits at the very foot of the band. Nothing was moved; the discrepancy is recorded so a reader who follows the patch file to 0047B and finds nothing knows why. Note that sedet (§2.1) is on 0047A too: two [sic:] in one band, from the same page read.

3b · 0048C (chunk 0016) · uo n for non — ⛔ a negation particle CC restored for us

⚑ This is the most consequential of the four and deserves its own sentence. The site is a negation, and the blind polarity read traced it to a named host and found it correct. It was correct — but only because the digitization had already repaired it. Our polarity apparatus was reading a word Migne does not print. The finding is not that the polarity read was wrong; it is that a polarity read run over a normalized text can only ever check the normalizer's work, and nothing short of the plate can tell you which you are looking at.

3c · 0048C (chunk 0016) · Qomodo for Quomodo

A single dropped sort, at the head of a sentence, in Migne's own type. ⚠ Exactly the one-letter shape this file's policy refused to accuse Migne of on suspicion — and rightly; the difference is that this rests on a plate, and that the letter was dropped by Migne and supplied by CC, not the reverse.

3d · 0055C (chunk 0019) · erum for erunt

⭐ This one lands on a column that already carried a marker, and the shape is worth naming. 0055C also carries the [var:] on Isa 15:5 (consternantem / conternantem, §1b) — and that [var:] was confirmed by the same read, self-witnessing, since the gloss directly beneath computes the third year, which follows only from the Vulgate's conternantem. So on one column: our marker correctly accuses one word while CC had silently repaired another four words away, and the two are unrelated. A collation checks the word it is pointed at; it cannot check the line it is standing on.

3e · WHAT THE READ DID NOT CHANGE

• §2.1 0047A sedet — confirmed; see §2.1.
• §2.4 0054A versuram — no patch; the decline stands.
• §2.5 0051B faceritis — no patch; the decline stands, and the corpus argument from 11057 is untouched.
• All ten [var:] — read and standing. node scripts/plate-gate.mjs 9006 returns gate OK.

⚠ For §2.4 and §2.5 the evidence is that a reader had the page at 300 dpi and patched nothing, not a per-token verdict recorded in data/plate-reads.json. That is corroboration, not confirmation — and it costs nothing, because neither site fires a marker.

CAPUT XLIX–LII, PL 114 cols 0056A–0062D (the oracles against Ammon, Edom, Damascus, Cedar and Elam; against Babylon; and the closing historical appendix). 2,614 Latin words, one translator.

Conventions: Ruth (8968) cruces.md §1 unchanged — VERS. n.-- verbatim, the lemma Englished, the sigla ordinary [n: …]. Tobit (8970) §1(b) unchanged — usque ad → as far as in its own italic span, no verb ever supplied to complete the abridgment. Heads: CAPUT XLIX. → CHAPTER XLIX., etc., matching Ruth's shipped form for the numbered chapters.

Apparatus fired: 5 [var:] · 0 [sic:] · 0 [cj:] · 0 [d:] · 0 [ed:] · 0 [nt:].

⛔ CORRECTED 2026-08-29. Counted against the chunks, this batch stands at 5 [var:] · 2 [sic:] · 1 [cj:]. The zeroes above were true when written and are kept because §4's closing claim — "No [sic:] was fired anywhere in the batch, and none was withdrawn — none was ever drafted" — is the record of a policy, and the policy was not overturned. What changed is the evidence:

• [sic:] 0 → 2. 0061D Scripturae and 0062A iniquitates, both at the foot of the work (§11). Both are the normalization direction — Corpus Corporum silently repaired a defect in Migne's own type, so our Latin read correctly and no check we own could see the site.
• [cj:] 0 → 1. 0061B [cj: mersa; read non mersa, "was not sunk"] (§11c), fired on Pattern 18a's own case: a printed word that makes the English assert something false. Notes 7 / 12 / 16 / 8, verbatim and in order against the Latin twins. Column anchors 6 / 7 / 8 / 6, verbatim, in order, and each checked against the Latin word it stands beside. Question marks 0 / 1 / 0 / 0, matching. Colons, semicolons, ! and guillemets compared band by band (the anchors partition both twins identically) — zero mismatched bands after eight repairs on our side, listed in §5. Zero guillemets in either language. Paragraph structure 1:1 (36 / 36 / 39 / 36 lines).

§1 · THE LEMMA DIVERGENCE LIST — every italic span in bands 0056A–0062D

Collated against sources/vulgate/clementine-flat.txt, cited book chapter:verse. Agreements are listed as well as divergences, because a list of findings alone cannot be told apart from a report by an agent that never looked. A ✓ below means checked at the verse being glossed, not that the pre-scan said ✓.

1a. Divergences that took a [var:] — 5

⚠ On the Rom. VI locator: Migne prints Rom. VI for what is Rom 9:22. The note contents are his citation and pass through verbatim (refDisplay keeps the printed reference); not corrected and not marked — that is the settled treatment, and it is recorded here so a blind reader does not read the silence as an oversight.

1b. Divergences examined and deliberately not marked — 7, with the reason

These are the candidates a marker was considered for and lost. Under Pattern 14 a [var:] is for a divergence that changes the sense; each of these does not.

1. 0057A · Jr 49:35 — Migne Ecce ego confringam; Clementine confrigam. Here the Clementine carries the aberrant form and Migne the ordinary one. Nothing turns on it. No marker.
2. 0057D · Is 13:17 — Migne …non quaerant, neque aurum velint; Clementine nec aurum velint. Same particle, same scope, same sense.
3. 0059B · Jr 50:44 — Migne Quis enim similis mei?; Clementine at 50:44 quis est enim similis mei. Migne's form is exactly the Clementine's at 49:19, which the same work glosses at 0056D. A copy-text agreement, not a defect.
4. 0060D · Jr 51:34 — Migne Comedit me et devoravit me; Clementine Comedit me, devoravit me. Migne supplies a coordinating et where the Clementine has asyndeton.
5. 0061A · Apc 18:4 — Migne Exite de illa, populus meus, ne participes sitis; Clementine …: ut ne participes sitis. A dropped ut; the negative purpose is unaffected and both negations (ne participes sitis, non accipiatis) are carried in the English.
6. 0061B · Apc 18:21 — Migne drops the opening Et and stops the quotation at …mittetur Babylon, where the Clementine continues civitas illa magna, et ultra jam non invenietur. An ordinary citation trim, of the same family as etc.
7. 0061C · 4 Rg 25:4 — Migne, quoting in libro Regum, prints Fugit Sedecias per viam, etc. where the Clementine has Fugit itaque Sedecias per viam. A trimmed connective inside a quotation he himself closes with etc.

⚠ One further scripture divergence, in unmarked prose. At 0058D the ORIG. abridgment tail runs legi enim Dei non subjicitur, neque enim potest; the Clementine's Rom 8:7 reads legi enim Dei non est subiectus: nec enim potest. It is not italicized, carries no note, and is not offered as a citation — so Pattern 14 does not reach it. Rendered as printed ("for it is not subject to the law of God, neither indeed can it be"), logged here, no marker.

1c. Migne's verse addresses disagree with the Clementine's versification — 3, all his

Reproduced exactly as printed and never reordered (brief: "Migne's own printing is not yours to tidy").

In each case the lemma itself agrees with the Clementine verbatim at the verse named on the right. The divergence is the address, not the words.

1d. Lemmata checked and found to agree with the Clementine exactly — 77

Counted per chapter: 16 · 18 · 24 · 19.

VERS. n.-- addresses are given as Migne prints them; the verse cited is where the words stand.

CAPUT XLIX — Ad filios (49:1) · Filiae Rabbath (49:3) · Et post haec (49:6) · Et captivos (49:6) · Ad Idumaeam (49:7) · Si fures (49:9) · Audivi (49:14) · Ecce enim (49:15) · Parvulum dedi te in gentibus (49:15, verbatim) · Cum exaltaveris (49:16) · Ecce quasi leo (49:19) · A voce ruinae (49:21) · Confusa est (49:23) · Dissoluta (49:24) · Ad Cedar (49:28) · Tabernacula (49:29).

CAPUT L — Verbum (50:1) · Confusa sunt (50:2) · Venient filii Israel (50:4) · Grex perditus (50:6) · Recedite (50:8) · Confusa est mater vestra nimis (50:12) · Ecce novissima erit (50:12) · Sicut fecit, facite ei (50:15) · A facie gladii columbae (50:16) · Leones ejecerunt (50:17) · Ecce ego visitabo (50:18) · Super terram dominantium (50:21) · Quomodo confractus (50:23) · Aperuit Dominus thesaurum (50:25) · Vox fugientium (50:28) · Ecce ego ad te (50:31) · Et erunt quasi mulieres (50:37) · terra sculptilium est, et in portentis … gloriantur (50:38, verbatim).

CAPUT LI — Quoniam non fuit (51:5) · Vae Assur, virga furoris mei (Is 10:5) · Fugite de medio (51:6) · Calix aureus ×2 (51:7) · Inebrians (51:7) · De vino (51:7) · Subito cecidit (51:8) · Ululate (51:8) · Curavimus (51:9) · Qui levat nubes (51:16) · Fulgura (51:16) · Sagittas suas ardentibus effecit (Ps 7:14, verbatim) · Volavit super pennas ventorum (Ps 17:11, verbatim) · Stultus factus est (51:17) · Confusus est (51:17) · Ecce ego (51:25) · Et dabo te (51:25) · Cessaverunt (51:30) · Filia Babylonis (51:33) · Et desertum faciam mare (51:36) · Quomodo capta est Sesach (51:41) · Egredimini de medio ejus (51:45) · Propterea (51:52) · Sic submergetur Babylon (51:64).

CAPUT LII — Filius viginti (52:1, and see §2) · Factum est autem (52:4) · Per viam portae (52:7) · Persecutus est (52:8) · Et adduxit eum rex (52:11) · In mense autem quinto (52:12) · Nolite confidere (Jr 7:4) · Venit Nabuzardan (52:12) · De pauperibus (52:16) · Columnas quoque (52:17) · De columnis (52:21) · Et tulit magister (52:24) · De civitate tulit (52:25) · Septem viros (52:25) · Et scribam (52:25) · Et sexaginta viros (52:25) · Et percussit eos rex Babylonis (52:27) · Iste est populus (52:28) · Et factum est (52:31).

Not scripture, so no comparand exists — the onomastic glosses poculum sanguinis · sanguinem bibens · sanguis cilicii (Damascus, 0057A) · tenebrae (Cedar, 0057A) · despicientes · tentantes (Elamitae, 0057A) · byssus sacci · manus sacci (Sesach, 0061A) · dissipans · dividens (Sophar, 0062C); the sigla-spans (HIER) and (RAB) at 0056B/0057A, which are attributions and pass through unanglicized like any [n: …]; and the poetic tag Lateque vagantes Agi (0057A, §4).

§2 · THE NUMBERS AND MEASURES OF CAPUT LII — audited one by one, none conformed

The launch flagged this as the class that gets silently squared with a familiar English Bible. Every number that reaches our English, and the one place a number does not:

• 0061B VERS. 1.-- Filius viginti. — the Clementine's Jr 52:1 reads Filius viginti et unius anni erat Sedecias. Migne's lemma is clipped in the middle of the number, and I have rendered it clipped: "The son of twenty." ⛔ This is a deliberate exception to convention 3's expansion of fragments, and the boundary is worth stating because I applied expansion four times elsewhere in the batch (§3): a fragment is expanded from the verse it opens, except where the elided material is a NUMBER or a MEASURE. Supplying "and one years" would put a figure on our page that Migne's plate does not carry, in exactly the class the launch named. The full reading is here; the Latin stands opposite.
• 0061B VERS. 4 — Novenarius minor denario — nine and ten, the gloss's own arithmetic on Jr 52:4's anno nono … mense decimo. Rendered as printed.
• 0061C VERS. 12 — Migne's prose reads decima die mensis; Jr 52:12 has decima mensis. This is the glossator's own sentence, not a lemma; rendered as printed, no marker.
• 0062A VERS. 21 — Migne's abridgment (Quod facturam columnarum describit … usque ad) cuts before the measurements of Jr 52:21–23 (eighteen cubits, a line of twelve cubits, four fingers' thickness, five-cubit capitals). No measurement reaches our English at all — there was nothing here to conform, and nothing was supplied to fill the gap.
• 0062A VERS. 26 Septem viros — Jeremiah has septem; the parallel at 4 Rg 25:19 has quinque. Migne follows Jeremiah, and the gloss is built on the seven (Septiformi Spiritu … septem spiritibus nequioribus). Checked and left; the Kings variant is not a divergence in this book.
• 0062C Et sexaginta viros — sixty in both Jr 52:25 and 4 Rg 25:19.
• 0062D VERS. 28 Tertio vastavit Nabuchodonosor Judaeam — a third time; agrees with the three deportations of Jr 52:28–30.

§3 · FRAGMENTARY LEMMATA — the expansions taken, and the one refused

Convention 3 (Ruth §1.2) renders a fragment as the phrase it stands for. Applied five times, refused once, and both decisions are recorded because a later reader will otherwise see an inconsistency that is not there (brief: name the control's class).

Complete one-word lemmata were left alone, being a different class: Audivi · Verbum · Tabernacula · Fulgura · Ululate · Curavimus · Inebrians. A lemma printed with a trailing comma or etc. (Cum exaltaveris, etc. · Recedite, etc. · Cessaverunt, etc. · Quoniam non fuit, etc.) already announces its own truncation and was carried with the comma and the etc. as Migne prints them, unexpanded.

⚠ VERS. 9. at 0056B prints without the --, alone among the 45 verse addresses in these four chunks. Carried exactly as printed.

§4 · PLATE-READ CANDIDATES — 2, both raised without a marker

Under the ship test neither fires anything: a crux asserts nothing, and neither of these is a non-word I can stand behind as broken type.

(a) 0056A · ἄρκος, where the sense requires Ἄρης

Metropolis Moabitarum, quae nunc Ariopolis dicitur, composito nomine ex Hebraeo et Graeco, quod ἄρκος, id est, Martis civitas vocetur.

The etymology is Hebrew Ar + Greek πόλις, and the gloss glosses the Greek element as Martis civitas — the city of Mars, i.e. Ἄρης. ἄρκος means "bear" and yields nothing of the kind. Rendered as printed. ⛔ No [cj:]. The form is a real Greek word (addendum A: a real word that reads oddly is Migne's until a plate says otherwise), it is the only Greek in the entire work — grep returns one hit in all of src/latin/ — so the corpus offers no control whatever, and a [cj:] here would be a public claim about a Greek accidental in a Latin plate on the strength of an argument from sense alone. Requested: a plate read of col. 0056A. If the plate reads Ἄρης, the fix is a data/tei-patches/9006.json entry, not an emendation of the English.

(b) 0057A · Agi

…et habitant per solitudinem totam, de quibus poeta dicit: Lateque vagantes Agi.

Agi is unique in the corpus — grep -rno "vagantes Ag[a-z]" src/ returns this one line — and the tag is not identified. The obvious candidate is a truncated Agareni, the very word the same sentence uses two clauses earlier (Hi Agareni sunt); Arabes is the other. Neither was adopted, and the reason is the ship test: Agi* is a capitalized, name-shaped token, not a malformed one, so it fails the non-word trigger that has been right every time, and a [sic:] here would blame Migne for what may be his own abbreviation of a proper name. Rendered as a name, "the Agi". Requested: a plate read of col. 0057A, and worth reading the whole line (addendum C), not the token.

⚠ Status after the 2026-08-29 plate gate. Both candidates' pages were read — 0056A on p. 33, 0057A on p. 34 — and data/tei-patches/9006.json carries no patch for either, so ἄρκος and Agi both stand as our Latin has them and neither acquires a marker. ⛔ That is corroboration, not a recorded verdict: data/plate-reads.json records per page what was found, and neither read note mentions these two tokens. Both requests below therefore stay on the page as requests, and cost nothing, since neither fires anything.

⛔ The closing claim of this section is now out of date and is corrected rather than deleted. Two [sic:] were fired in this batch after all (§11) — and the sentence below is the reason they are worth reading twice, because the sweep it reports was looking for the wrong direction of departure. "No other form in these four chunks is split, fused, doubled, or absent from Latin lexica" is true, and it is exactly why neither of the two was found: both are ordinary, well-formed, correctly-spelled words in our Latin. CC had already repaired them.

⭐ Checked negatives, recorded because the positives are only credible beside them: no other form in these four chunks is split, fused, doubled, or absent from Latin lexica. Ariopolis (0056A) is an attested variant of Areopolis and takes nothing under Pattern 9. Teglathphalasar, Sennacherib, Nabuzardan, Sesach, Reblatha, Aretha, Taphnis, Bubastis, Sienni (0058A, for Syene) are all ordinary Migne orthography for the names. No [sic:] was fired anywhere in the batch, and none was withdrawn — none was ever drafted.

§5 · PUNCTUATION — band-by-band, and eight repairs on our side

Compared per column band (the anchors partition both twins identically), never by total. The first pass found eight mismatched bands, every one ours, and all eight were repaired before filing:

⚑ Note the shape: four supplied and two destroyed marks in chunk 0020 and 0021 alone, and the chunk totals would have cancelled in 0020 (two : and one ; against two : and one ;, in different bands). This is the third recorded instance of that cancellation and it is why the band test exists. Zero mismatched bands now. No comma→semicolon sweep was attempted — Migne prints plenty of both in this book.

Question marks: the batch's only one is Quis enim similis mei? at 0059B, and it is a direct question, so Pattern 8b's colon recast did not arise anywhere.

§6 · READINGS, PARSES AND UNEXPRESSED SUBJECTS

6a. Two readings that both construe — named rather than chosen silently

0059C · Non est semper calix aureus Babylonis vel in manu Domini. Two parses: (i) semper … in manu Domini is one predicate and vel is "even" — "The golden cup of Babylon is not always even in the hand of the Lord"; (ii) two predicates — "It is not always a golden cup of Babylon, or in the hand of the Lord." (i) was taken, and the reason is the sentence's own structure rather than what reads well: the two clauses that follow are both, and only, about the Lord's hand — cum … in manu Domini posita fuerit and Nec jugiter in manu Domini continetur; sed tunc tantum. (ii) loses because nothing in the paragraph ever questions that Babylon is a golden cup. The sed tunc tantum correlative also fixes the host of both negatives on the temporal adverbs (semper, jugiter), not on est/continetur — the 2a trigger fired here and the structure, not the reading, settled it.

0056B · quae consurgit adversus spiritum, ut non faciamus ea quae sunt spiritus, et animam, etc. et animam can hang either on adversus ("rises up against the spirit … and the soul") or inside the ut clause. The intervening purpose clause makes both possible and the abridgment's etc. cuts before it is resolved. Rendered in the printed order, preserving the ambiguity — "which rises up against the spirit, that we may not do the things which are of the spirit, and the soul, etc." Neither reading was adopted in the English, because the Latin does not adopt one either.

6b. A construction that will not construe, and was not repaired

0056C — Sed Dominus e contra omnia secreta eorum, et patriarcharum Esau (qui primas scilicet haereses invenerunt) viros ecclesiasticos et doctores profert in medium. Two accusative groups stand asyndetically before one verb. Rendered with the asyndeton intact ("…brings forth into the midst all their secret things, and those of the patriarchs of Esau (…), churchmen and teachers"), which is what the printed words allow. The obvious conjecture is per viros ecclesiasticos et doctores — the Lord publishing the heretics' secrets through churchmen and teachers — and it is not adopted**: nothing is broken here, there is no [cj:] case because no single printed word is misleading, and supplying a preposition would be exactly the silent smoothing Pattern 7 forbids.

6c. A printed negative kept although the argument seems to want the opposite

0061C · Malignis spiritibus populum circumdantibus, rector non fugit, etc., usque ad… The surrounding gloss censures shepherds who run (0061C's own non bellum gerere, sed effugere quaerunt; 0061B's videntes lupum venientem oves dimittunt), so the sentence reads against its neighbours. The non is carried — "the ruler does not flee" — with no repair and no marker. Either Migne's sentence states the good ruler's contrasting case and the abridgment has cut the contrast away, or the plate has lost a word; neither can be settled from this file, and translating fugit as its opposite to rescue the paragraph is the precise failure 7a exists to forbid.

6d. Every unexpressed subject supplied, and what fixed it

• 0058A …quia venturum erat ut a Medis et Persis vastaretur → "that she should be laid waste" = Babylon, from captivandi erant in Babylonem in the same sentence.
• 0058B sicut illa … gentes circumquaque vastavit → "as she … laid waste" = Babylon, the feminine of the lemma Confusa est mater vestra.
• 0058B Etiam Nabuchodonosor columbam vocat → "he calls" = the prophet, the subject running through Isaias dicit, notantur, monet since 0057C. Not the Lord: the sentence is about the naming, and Jeremiah is the one naming.
• 0059C convertit (0059C), reprehendit (0060B), redit propheta (0060C) — the prophet, named outright in the third, which governs the first two.
• 0059D sed quia noluit curari → "because she would not be healed" = Babylon, from the lemma the sentence is running toward, Jr 51:9 Curavimus Babylonem, et non est sanata.
• 0060A In cujus enim corde non cecidit → "in whose heart she has not fallen" = Babylon, from videat quisque Babylonem in pectore suo corruisse immediately before.
• 0058D legi enim Dei non subjicitur, neque enim potest → "it is not subject" — ⚠ the one place I could not name the subject. Rom 8:7's own subject is prudentia carnis; the run resumes at et inventa est Babylon, which would make it Babylon; the abridgment has removed whatever stood between. A neutral "it" is carried and the ambiguity is recorded here rather than resolved in the text.
• 0062D correctum misericorditer respexit → "him corrected he mercifully regarded" = Joachin regarded by God, from Jr 52:31, the verse the lemma opens.

6e. Ellipsis where English cannot leave the verb out

0062C — Quem liber Regum Sophar, qui interpretatur dissipans vel dividens, et significat… The naming verb is elided in the Latin (a gapping English has no way to reproduce), and et significat takes Sophar as its subject. Rendered "Whom the book of Kings calls Sophar." Ordinary gapping, not an emendation — but recorded, because a reader comparing the columns will find no Latin word answering to "calls".

⚠ In the same sentence Migne prints Sophar where the Clementine's 4 Rg 25:19 has Sopher. This is a name in the glossator's own prose, not a quoted lemma, and the two are the same Hebrew name; no [var:], logged here.

6f. Passages read closely and left exactly as printed

• 0059C efficietur terra quae facta est in Job — obscure, and no candidate in Job is close enough to name. Rendered literally ("there shall be brought about the earth which was made in Job"). Real words throughout; no marker.
• 0060B Praedicatores prophetarum fulgura sunt — the genitive is odd where the parallel wants praedicatores and prophetae in apposition (as at 0060B's own praedicatores et prophetas four words earlier). A real, well-formed phrase; rendered as printed, no marker, addendum A applied.
• 0056A Metropolis Moabitarum — Rabbath is the metropolis of the Ammonites (Jr 49:2, Rabbath filiorum Ammon); Ar/Areopolis is Moab's. The confusion is very likely the glossator's own and is content, not type. Rendered as printed.
• 0062D per bonos et malos aeque sua, etc. — sua is left without its noun by the abridgment, which cuts mid-phrase. No noun supplied (Tobit §1(b): nothing is ever supplied to complete the formula). Rendered "through good and evil alike, his own, etc."

§7 · usque ad — 35 formula instances and 1 look-alike, decided per occurrence

Addendum D. The abridgment formula occurs 9 · 9 · 9 · 8 times across the four chunks, each in its own italic span, each rendered as far as, and no verb was supplied to any of them. Where the resumption lands mid-clause the English lands mid-clause.

⛔ The look-alike is at 0061C, in chunk 0023, and it is not italicized: …ducitur in domum carceris usque ad diem mortis, et in angustia mentis includitur. These are Jeremiah's own words — Jr 52:11, posuit eum in domo carceris usque ad diem mortis ejus — the verse the lemma Et adduxit eum rex comes from. Rendered "unto the day of death". Reading it as the formula would have produced an abridgment where the gloss is quoting its own verse.

⭐ Recorded as a checked negative too: the other 35 are all genuinely the formula, all preceded by etc., and all set in italics by Migne, which is a reliable local tell in this work — not offered as a general rule.

§8 · WHAT WAS READ CLEAN

Named so a blind reader can tell a checked section from a skipped one.

• All 64 VERS. n.-- addresses in the four chunks, character by character against the Latin twins. One irregularity (VERS. 9., §3); three versification disagreements (§1c); no verse number out of sequence anywhere in CAPUT XLIX–LII.
• All 43 [n: …] notes — contents verbatim, unanglicized, same count, same order (7 · 12 · 16 · 8, diffed mechanically against the Latin). None is editorial prose, so no [nt:] arises: every one is either an attribution siglum (RAB. · ID. · ORIG. · HIER. · GREG.) or a scripture locator.
• All 27 column anchors — verbatim, in order, and each checked against the word it stands beside, not merely for presence (2b). Six stand mid-construction and were kept there: Babylonios [0057D] et Chaldaeos · Teglathphalasar, [0058C] qui · terra sculptilium [0059B] est · non miscuit [0059D] in vase · Nam secundum [0061B] historiam · Septiformi Spiritu, [0062C] scilicet. The English splits at the same joints.
• Every printed negative, host named: 5 in 0020, 11 in 0021, 24 in 0022, 6 in 0023. The four that fired test 2a's trigger (a negative on a verb with a correlative or quantifier adjacent) were proved from the structure, not from what reads well:
• 0059B quem non corripiendo castigavit, sed saeviendo contrivit — the sed pivots on the two ablative gerunds, so the negative is on corripiendo, not on castigavit. "whom he chastised not by correcting, but crushed by raging." The false reading ("did not chastise him") is the one that reads better and it is wrong: he did chastise, savagely.
• 0059B Non est semper — on the adverb; see §6a.
• 0059B quae non per partes sed repente fiet — on the prepositional phrase.
• 0061B seek not to wage war, but to escape (non bellum gerere, sed effugere quaerunt) — on the first infinitive phrase, not on quaerunt: they do seek, and what they seek is flight.
• and at 0059B non Cyri potentia, sed Domini judicio — on the ablative, not on factum esse.
• No negative was added anywhere. The English carries no/not/nor/neither/lest/unless only where Migne prints non · nec · neque · nisi · ne · nulla · nondum · nolite · noluit.
• The fourth test run last, paragraph by paragraph, with the counts already clean: does the English assert what he asserts? The three sites it stopped at are §6a, §6b and §6c above — none of them a containment failure, all three a question about what the sentence says. No smooth English was left standing over rough Latin: 0056C, 0059C, 0061C and 0062D all read as awkwardly in English as they do in Latin, deliberately.

§9 · STILL OPEN

1. Plate read, col. 0056A — ἄρκος / Ἄρης (§4a). The only Greek in the work. ⚠ Page read (p. 33), nothing recorded either way; still open as a question, closed as a blocker.
2. Plate read, col. 0057A — Lateque vagantes Agi (§4b), reading the whole line. ⚠ Page read (p. 34), nothing recorded either way; same status.
3. Neither blocks the chunks: both are rendered as printed and neither fires a marker.

§10 · PLATE-GATE DEBT (CLAUDE.md hard rule 8a)

✅ PAID IN FULL, 2026-08-29 — every statement of fact in this section is now out of date, and the argument in it is why the read happened. The full gate ran before the work shipped: 24 pages of patrologiaecurs06saingoog at 300 dpi, three readers, corner numbers checked first on every page, recorded in data/plate-reads.json (works.9006). Of this batch's four pages, all four — 33, 34, 35, 36 — were read. All five [var:] stand; two [sic:] and one address restoration came back from pp. 33–36 that nobody had asked for (§11). node scripts/plate-gate.mjs 9006 returns gate OK on all 50 plate-claiming markers in the work.

⭐ And the paragraph below turned out to be right about the class as well as the rule. It warned that a Clementine collation cannot catch the digitization manufacturing a divergence. The 9006 read found the mirror image of that, twelve times: the digitization erasing a defect of Migne's, which a Clementine collation cannot catch either, and for the same reason — the collation is checking our text against a witness, and our text is not the plate.

The debt as it stood, kept for the record:

node scripts/plate-gate.mjs 9006 reports my five [var:] as standing on unread columns: 0056B · 0058B · 0058D · 0060B · 0062B. data/plate-reads.json has no 9006 entry at all, so this is the whole work's debt, not mine alone. Each of the five is a public claim that Migne's plate prints what our Latin twin prints, and 9003's collation measured Corpus Corporum departing from Migne at roughly three sites per page — including one case where the digitization manufactured a Vulgate divergence that a collation then dutifully "found". ⛔ A Clementine collation cannot catch that class, which is exactly why these five must be read at the plate before the work is marked ours, or withdrawn.

PL 114 is patrologiaecurs06saingoog, PDF page = (column + 11) / 2 — verified against the printed corner numbers per CLAUDE.md, so the five columns are pp. 33 · 34 · 34 · 35 · 36. Two further pages are wanted for §4's plate-read candidates (0056A, 0057A → pp. 33, 34), which overlap the same sheets. Four distinct pages — 33, 34, 35, 36 — settle all seven.

§11 · THE PLATE GATE, 2026-08-29 — two [sic:] and one [cj:] at the foot of the work

Bands 0056A–0062D fall on pp. 33–36, all four read at 300 dpi with the corner numbers checked first. Recorded in data/plate-reads.json (works.9006), patched in data/tei-patches/9006.json.

11a · 0061D (chunk 0023) · Scripturae for Scriptura

⛔ The direction matters and it is the opposite of everything §4 was looking for. Scriptura is not a word anyone would suspect: it is the right word, in the right sentence, correctly spelled. CC had made it right. Migne's own type prints an ungrammatical Scripturae against narrat, and until the plate was read there was nothing whatever to detect — no non-word, no fused run, no frequency signal. §4's sweep could not have found this and did not fail by missing it.

⚑ The English still renders the sense Migne's sentence requires ("the Scripture relates"), which is what narrat forces; the [sic:] is what stops that rendering from being a silent emendation. Strip the marker and the clause is complete (7a⁗).

⚠ Column label. data/tei-patches/9006.json files this patch under 0062A; the word stands immediately before the [0062A] anchor, so by anchor position — and in plate-gate.mjs, and at migne.app/pl/114/61d — it is under 0061D. Nothing was moved; the discrepancy is recorded so a reader following the patch file to 0062A and finding nothing knows why. (The same one-band offset occurs twice more in this work, at multc and Jean.)

11b · 0062A (chunk 0023) · iniquitates for iniquitatis

Same class, same page, same sentence: two of Migne's own case errors, both silently normalized by CC, in one gloss. The English renders the construction multitudine requires and the marker carries the defect.

11c · 0061B (chunk 0022) · [cj: mersa; read non mersa] — Pattern 18a's own case

Sic submergetur Babylon. Nam secundum [0061B] historiam Babylon in Euphraten mersa est, quamvis in solitudinem redacta.

English: "For according to the history Babylon was sunk [cj: mersa; read non mersa, "was not sunk"] in the Euphrates, although reduced to a wilderness."

⛔ This marker is fired and had no entry anywhere in this file until now; the header's 0 [cj:] is corrected with it.

Why it fires where §6b and §6c decline. Pattern 18a's bar is that the printed word makes the English assert something false, and here it does, twice over:

1. Migne's own sentence refutes it. quamvis in solitudinem redacta — "although reduced to a wilderness" — is a concessive, and a concessive needs something to concede against. If Babylon was drowned in the Euphrates, being reduced to a wilderness concedes nothing; the sentence only works if the literal city was not sunk and was merely laid waste.
2. The argument the gloss is making requires it. The whole point is the contrast with the next sentence, Sed spiritualis Babylonis aeternum interitum praedicit, quae in profundum inferni … demergetur — the spiritual Babylon is the one that is sunk. The Sed is the pivot, and it pivots on a negative that the printed clause has lost.

⚠ Distinguish this from §6c, which declines on the same page-worth of text. At 0061C a printed non reads against its neighbours and is carried with no marker, because there the abridgment has plainly cut the contrast away and the missing material is exactly what would settle it. Here nothing is abridged: the concessive and the answering Sed are both printed, in full, in the same two sentences, so the answering member Pattern 18a asks for is on the page. That is the difference, and it is why one site takes a marker and the other does not.

⚑ The [cj:] proposes; it does not repair. The English still reads "Babylon was sunk", as printed. Nothing was silently emended, and the plate gate does not read [cj:] — a conjecture is a claim about sense, not about type, and makes no assertion about what the plate prints.

11d · WHAT THE READ DID NOT CHANGE

• All five [var:] — 0056B · 0058B · 0058D · 0060B · 0062B — read and standing.
• §4a ἄρκος and §4b Agi — no patch; both stand as printed, no marker. See §4.
• §1c's three verse-address disagreements — untouched; Migne's addresses pass through verbatim. ⭐ Related, and recorded in cruces-0010.md §7c: at 0043A CC had silently corrected one of Migne's addresses (VERS. 5. → VERS. 3.) and correctly on the merits, which is exactly what made it invisible. It has been restored. No [sic:] is owed on an address — it is not a word of his text.
• ⚠ For §4a and §4b the evidence is that a reader had the page at 300 dpi and patched nothing, not a per-token verdict. Corroboration, not confirmation — and it costs nothing, because neither fires a marker.