Liber Genesis

Anselm of Laon and his school, PL 113 cols 0067B–0182D. 47 chunks, ~46,600 Latin words → ~70,800 English (1.52×), translated in nine Opus stints, 2026-09-05.

Apparatus as merged: 96 [var:] · 2 [sic:] · 8 [ed:] · 1 [cj:] · 1 [nt:] · 0 [d:]. verify-english OK on all 47 chunks. plate-gate GREEN — 50 recorded plate reads covering 96 of the work's 116 columns (83%), and every column carrying a marker was read.

§0 · WORK-WIDE CONVENTIONS, AND WHAT THE MERGE CHANGED

⛔ This section is the reconciliation, and it is the only place in this file where a stint's text was overridden. Everything below §0 is the nine stint files verbatim (headings demoted one level), including the places where an early stint says something a later one corrected — those corrections are recorded in the later stint's own section and are not back-edited, so the reasoning stays legible. Read §0 before trusting a convention statement further down.

What was already unanimous — measured, not assumed

Nine stints, briefed from the same series file, converged without needing to be reconciled. Each of these was checked mechanically across all 47 chunks at merge time:

⭐ The usque ad result is the one worth naming, because it is the convention most likely to drift and the hardest to check: two ranges appeared to disagree and neither did. Chunk 0016 carries Migne's own abbreviated etc., usque, (no ad), correctly read as the formula; chunk 0023 carries uque ad, also correctly read as the formula. Both stints had documented their reasoning, so the merge only had to confirm it. ⛔ Correction, 2026-09-05, from the Gallica check: the merge first recorded that second site as our twin's typo. It is Migne's — Gallica prints uque ad at 0131C exactly as our twin has it. The rendering is unaffected and still right (it is the splice formula either way), but the attribution was backwards: there is no patch here, and a [sic:] is a live candidate instead.

⚑ THE ONE REAL DRIFT, AND WHY IT MATTERED — pactum

Two renderings were in the work: pactum → compact in the 0019–0023 range (6 sites), and pactum → covenant in three sites elsewhere (0013, 0014, 0025).

⛔ Collapsing them would have destroyed an argument on the page. At 0122D Migne glosses the Hebrew: In Hebraeo berith quod non testamentum, sed foedus, sive pactum significat — the whole point is that one Hebrew word answers to three Latin ones. Render pactum as "covenant" and that sentence says not testament, but covenant, or covenant, which argues nothing. This is the 11551 epistylium/"epistyle" failure exactly: two stints, neither able to see the other, diverging on the vocabulary a third passage's argument rests on.

Resolved work-wide in favour of the range that had the argument in front of it:

testamentum → testament · foedus → covenant · pactum → compact

Three sites were changed by the merge — 0013 receive a compact, 0014 establish my compact, 0025 made a compact with Abimelech. ⚑ foedus appears as a noun only in the 0019–0023 range, so "covenant" is unambiguous everywhere else in the work. ⚠ Not a fourth site: 0011 renders coarctatis "well compacted" of the ark's timber — an ordinary participle from a different root, unrelated to the triad. Leave it; a later sweep should not "harmonise" it.

⚑ ONE AXIS CROSSOVER REPAIRED — sacramentum rendered as its own partner

0032 read nec indignatur cognito sacramento as "nor is he indignant, the mystery being known" — in a chunk that renders sacramento "sacrament" three other times, twelve lines away. Changed to the sacrament being known. The pair is now exactly 9 and 24, 1:1 across the work. ⭐ Both repairs in this section were found by counting Latin against English work-wide, not by reading: a crossover inside one chunk is invisible to the stint that made it and to verify-english.

Conventions settled by a stint and adopted by the rest

Stated once here; the reasoning is in the stint section named.

1. No [var:] on a quotation Migne labels himself as a rival version — Alia translatio, quidam codices, secundum Hebraeum, Hebraica veritas, an announced LXX/Aquila/Symmachus/ Theodotion reading. (Stint 3.) Roughly 40 divergences across the work decline by this rule alone; without it the informative markers are buried.
2. Words follow the plate; punctuation follows the twin. (Stint 3.) Following the plate on a mark manufactures a mismatch in the very instrument that finds real ones. One deliberate exception, at 0178D (stint 9), where the twin's added period cuts the sentence into a verbless fragment — argued in place.
3. An italic Latin word: MEANING is Englished, a word MENTIONED AS A WORD is not. (Stint 4, applied by stints 6, 7, 8, 9.) domus Dei → the house of God; but where the argument is about the letter-sequence itself, the Latin stands.
4. usque ad is decided per occurrence by the etc. splice test, never by italic membership. Work-wide: 245 formulas, the remainder ordinary.
5. Heads mirror the plate's own form — CAPUT PRIMUM → CHAPTER ONE (a word, not a numeral), and Migne's single abbreviated CAP. XLIV. → CHAP. XLIV. (cont.) rides as the chunker set it.
6. The abridgment's tails are parsed as strictly as prose, and no verb or object is ever supplied to complete an usque ad splice. (Stint 2.)
7. Unpaired guillemets are mirrored, never tidied, and no policy is adopted about one without looking at the column — our twin has been caught both supplying a mark Migne does not print (0113C, 0125C) and dropping one he does (0106A, 0115D).

⛔ WHAT A LATER READER MUST NOT TAKE FROM THIS FILE

• Single-witness letter-level plate calls are not confirmed. The archive.org copy is over-inked in places and manufactured 12 divergences at pp. 85–88 and 13 of 15 at pp. 92–96, all refuted by Gallica, with our TEI twin right every time. Several stint sections below were written before that was known and state such calls as findings. See data/briefs/8950-PATCH-GATE-READ-FIRST.md before applying any patch. The rule: a [sic:] needs two witnesses; a dropped word needs one — ink can turn one letter into another, but it cannot put on the page a word that is not there.
• The two [sic:] that stand met that bar: pinguedidine @0152A (on the plate and verbatim in our twin) and trangressionis @0169B (read on both copies).
• ⛔ THREE NON-WORDS WERE FILED AS OURS AND ARE MIGNE'S (Gallica, 2026-09-05): dilicta @0099C, uque ad @0131C, virginet @0130D. The stints that met them declined a [sic:] on the ground that our twin had corrupted a clean plate — the plate carries all three. No patch; three live [sic:] candidates, which is the exact reverse of how they were filed. Anything in the stint sections below that calls these "ours" is superseded here.
• Our Vulgate comparand has typos that manufacture divergences no plate read can catch — Suscipe for Suspice at Gn 15:5, trienem for triennem at Gn 15:9, both inside one column of this work. See the ERRATA in sources/vulgate/README.md.

Open at merge time

• ✅ CLOSED at the merge — the six open-ended lemmata (data/briefs/8950-OPEN-ENDED-FOR-MERGE.md), which no instrument could see until the harvester was fixed. All six collated under 7a″: zero new markers. Four agree with their verse verbatim as far as Migne's etc.; one (0069A, Sir 18:1 vivit/vivet) is the site the 0000–0003 stint had already declined by name; and 0100C only looked like a divergence — terra Naid is absent from the Clementine, but Migne labels the reading LXX: himself and the properly-closed lemma four words earlier agrees with Gn 4:16 exactly. ⚑ On a guillemet book an unclosed « often opens a quotation of ANOTHER VERSION rather than a lemma, and what settles it is the closed lemma standing beside it.
• The TEI patch pass is VERIFIED against a second witness and its result is mostly negative. 125 single-witness claims were checked at Gallica (bpt6k5505319w, f = (col+1)/2, corners read first on every leaf): ~100 REFUTED, and not one letter-level claim survived in any file. What stands is a narrow set — lost Hebrew (whole non-Latin words), Migne's line-head « repeats, a few defects in our own transcription against a clean plate, and two spacing/mark forms. ⛔ The [sic:] backlog those files predicted has evaporated: no [sic:] is owed anywhere in cols 0095–0160, because the defects they accused Migne of are not on his page.
• ⛔⛔ THE DROPPED non AT 0141B DOES NOT EXIST, and it was reported as this work's most important finding. Gallica f71 prints quia in eo quod credidit, vacua non remansit: no non after quia, the phrase closed up at normal spacing, and the sentence's only non standing in the next clause. The reader attributed a word that is genuinely on the page to the wrong slot, in a sentence whose argument wanted a negative there — 7a″ conformation operating inside a plate read, where nothing downstream checks for it. The English has been reverted.
• Three non-words are MIGNE'S, filed as ours — dilicta @0099C, uque ad @0131C, virginet @0130D — so three [sic:] are available where three patches were expected. Not fired; a later session may.
• The blind polarity read has not run. ⚑ Its premise is now cleaner than when this file was written: our Latin is no longer believed to be missing a negation, so the pass compares an English against a Latin twin that two witnesses agree with.

STINT 1 — chunks 0000–0003 (cols 0067B–0082A)

Filed as cruces-0000.md; merged verbatim, headings demoted one level. Nothing in it was rewritten — where a later stint corrected it, the correction stands in that stint's section.

Stint 1 of 8950. Four chunks, 6,252 Latin body words → 9,183 English (1.47×). verify-english.mjs reports no error on 0000–0003: column anchors, [n:] notes, guillemets, italics and question marks all 1:1 with the Latin twins.

Apparatus fired: 5 [var:] · 0 [sic:] · 0 [cj:] · 0 [d:] · 0 [ed:] · 0 [nt:]. Every [var:] stands on a column recorded as read at the plate (0067, 0069, 0070, 0081) and is corroborated by the archive.org scan's own OCR (§2). Nothing was marked on an unread column.

Counts, chunk by chunk (Latin = English in every case): guillemets «/» — 0000: 42/40 · 0001: 36/36 · 0002: 23/26 · 0003: 37/35 (frontmatter included). [n:] notes — 19 / 26 / 23 / 29. Question marks — 3 / 6 / 0 / 1.

§0 · SPAN COUNT AGAINST THE BRIEF — ONE SPAN IS MISSING FROM 8950-lemmata-0000.txt

My Latin carries 168 marked spans; the brief prints 167. Counted mechanically over the four chunk bodies with «[^«»]»|\[^\n]+\, then diffed span-for-span against the brief's inventory lines. The lists agree everywhere except:

• ⛔ « Mente servio legi Dei, » — 0075B, chunk 0002 — is NOT in the brief. Collated here independently: Rom 7:25 Clementine reads Igitur ego ipse mente servio legi Dei : carne autem, legi peccati. Migne's span is an exact whole-word fragment of the Clementine. No divergence. (Its twin at 0082C, chunk 0004, sets the same words with the [n:] inside the guillemets — a different stint's range.)
• One presentational difference, not a count difference: at 0079C the brief prints the span as « Dixit quoque Deus: Producat « terra, » (outer « paired with the closing ») where my regex isolates the inner « terra, ». Same single site, counted once either way.

So: 167 + 1 = 168, and the brief's omission is at the head of chunk 0002, not at a range boundary — i.e. it is not the 9001 mis-split shape. It looks like a single dropped line rather than a bad cut. A later stint should still run its own count.

§1 · CONVENTIONS — inherited unchanged, plus four this stint had to settle

Ruth (8968 §1) and Tobit (8970 §1) govern and were not re-opened: VERS. n.-- passes verbatim; the lemma is Englished; a fragmentary lemma is rendered as the phrase it stands for; sigla ride as [n: …] with Migne's own internal spacing ((BEDA, in Hexaem. ), (ISID., in Gen.? ), (BED., ibid.)); usque ad → as far as with no verb ever supplied — all 31 occurrences in these chunks follow an etc. and are the abridgment formula.

⚠ The usque ad LOOK-ALIKES in this range, all ordinary Latin, all rendered by sense (brief addendum D; none italicised in the Latin, and none follows an etc.): 0000 @0070D ab Adam usque ad Noe → "from Adam to Noah" · 0001 @0073D omnes partes terrae usque ad coelum texerant → "covered all parts of the earth up to the heaven" · 0002 @0075B divina mandata praeteribant usque ad Saulis malitiam → "until the wickedness of Saul" (inside an abridgment tail — the formula and a temporal use in one sentence) · 0002 @0077D quasi carcer sit usque ad tempus judicii → "until the time of judgment". Four of thirty-five.

(a) NEW — a citation siglum that arrives as an ITALIC span, not as an [n:] note

0002 @0077B–C sets Migne's source-siglum in italics rather than as a note: (AUG., lib. II de Gen. ad litt., ex c. 1, 2, 3, 4, [0077C] 5 et 6, tom. III.) — and the column anchor falls inside it. Every one of its twenty-two siblings in the same chunk is an [n:].

Ruling taken: treat it exactly as an [n:] would be treated — Latin verbatim, unanglicized, italics preserved 1:1, anchor kept in place. Convention 4 of the series brief says sigla need no new convention because the locked register already governs them ("markers pass through verbatim, Latin abbreviations unanglicized"); the register does not depend on which wrapper Corpus Corporum gave it. Englishing it because it happens to be italic would make one siglum in this book read differently from the other ninety-seven in the same four chunks. ⚑ This is very likely a TEI defect (a <note> transcribed as <hi rend="italic">), not Migne printing this one differently. Not marked — a [sic:]/[ed:] would assert something about the plate that nothing here supports. Filed for the plate in §4.

(b) NEW — a gloss that DISTINGUISHES LATIN LEXEMES, with one English word for all of them

0002 @0075A–B: Sementis, frugum et arborum; semen, hominum, et animalium. Seminium vel sementum, cujusque rei exordium. English has "seed" for all four. Italics cannot be used — every italic span must be 1:1 with the Latin twin, and the twin has none here.

Ruling: keep the four Latin words in roman inside our own straight quotes — "Sementis" is of crops and of trees; "semen," of men and of animals. "Seminium" or "sementum," the beginning of anything whatever. Pattern 6 makes " " ours and free (guillemet parity is untouched), and the words being mentioned rather than used is the same call the PG register already makes for a mentioned particle. Rendering them all "seed" would have turned a lexical note into a tautology.

(c) The three supplied colons are Pattern 8b, and they are the ONLY punctuation added

A band-by-band comparison (Test 3 — colons, semicolons, ?, «, », !, per column anchor, markers stripped) returned exactly three mismatched bands in the four chunks, all the same class, and zero elsewhere:

Migne's ? is kept in all three; English cannot end an indirect construction with it (8b). ⚑ The scan pass that found these also found thirteen unlicensed marks that had crept in — four supplied colons of apposition at 0069A (Coelum, creatura spiritualis etc.), one each at 0068C, 0069B, 0070C, 0075A, 0076D, four supplied semicolons (0077B, 0078D, 0081A, 0081C) and one DESTROYED colon at 0078A (per haec autem erudimus: Cum igitur) and one at 0075C (infimae parti mundi:). All repaired. The totals had very nearly cancelled — this is Test 3 earning its place a third time.

⛔ Not one comma→semicolon sweep was made. Migne prints semicolons freely in this book (47 in these four chunks), so Tobit's narrow exception does not apply here and was not invoked.

(d) CAPUT PRIMUM. → CHAPTER ONE., and (cont.) rides as the chunker set it

## CAPUT PRIMUM. (cont.) → ## CHAPTER ONE. (cont.). Spelled, per the corpus rule (61 works CHAPTER ONE., zero spelled forms above ONE).

§2 · ⭐ THE ARCHIVE SCAN'S OWN OCR IS ON DISK, AND IT IS A SECOND WITNESS — USE IT

raw/scans/pl113/patrologiaecurs04migngoog_djvu.txt is Google's OCR of the plate scan, not of Corpus Corporum. It is independent of our TEI, it is free, and it settled or strengthened nine calls in this stint. It is noisy (coelnm, Qai, seternum) and it is not a plate read — it cannot license a marker on its own — but where our twin prints a non-word and the OCR prints a word, the OCR is telling you the non-word is ours.

⚠ Two mechanics, or it returns nothing: the file is hard-wrapped and double-spaced, so join lines and squeeze spaces first (tr '\n' ' ' | tr -s ' '); and the two columns of a page interleave in the OCR stream, so a passage can be cut in half by a line from the facing column (this happened at 0069B, §3.2).

Every check run against it, with the result — a list of findings alone cannot be told from a report by an agent that never looked:

§3 · THE DECLINED LIST — suspect readings that took NO marker

The rule this stint worked to: a [sic:] accuses Migne's type and a [var:] asserts Migne's text; our Latin is not the printed page. Where the file could not settle whose the defect was, or where the column has not been read, the crux is the answer and the marker is not.

3.1 ⛔ 0069B se recerunt deformes — a NON-WORD, and it is OURS, so no [sic:]

« Et terram, » id est eos qui postea superbiendo, terram, id est terreni hominis imaginem, portantes, se recerunt deformes. recerunt is no Latin word. The plate's OCR reads fecerunt deformes. « Terra autem erat inanis, » quia — the identical sentence, with the following lemma attached, so it is certainly this site and not a homonym elsewhere in the volume. An f→r substitution is a classic digitization artifact and its direction here is decided by the witness, not by the sense.

Rendered "made themselves deformed", with NO marker. Both alternatives were live and both are refused for the same reason Tobit refused them at Nephtlali (8970 §2): a [sic: recerunt] would publicly accuse Migne of a defect the scan of his own page does not carry, and a [var:] is about scripture, not type. The remaining option — carrying recerunt untranslated under Pattern 7 without a marker — is forbidden by Pattern 12 and would print a hole for a defect that is ours.

⚑ Still owed: an eyes-on read of PDF p. 40 (cols 0069/0070) confirming fecerunt, and then a data/tei-patches/8950.json entry, so the Latin column converges on the English. Until that patch lands the two columns disagree at this word, and §4 records it as a pending patch.

3.2 0069B Eccli 18:1 Qui vivit in aeternum — Clementine has vivet. DECLINED.

Sir/Eccli 18:1 reads Qui vivet in aeternum creavit omnia simul. Migne prints vivit (present). One letter, and sources/vulgate/README.md warns that this copy carries no later upstream corrections, so a call turning on a single letter should say so rather than trust it. A second, unrelated occurrence of the verse elsewhere in this volume also OCRs as vivit, which makes it Migne's usage rather than a slip — but that is an argument about frequency, and frequency proves an error exists, never whose it is. Logged, not marked. Rendered "Who lives for ever".

⚠ The quotation is UNCLOSED in our twin and stays unclosed in the English (Pattern 5 corollary): « Qui vivit in [0069B] « aeternum, creavit omnia simul [n: (Eccli. XVIII)], secundum originem tamen…. Two opens, no close. The second « is Migne's line-continuation mark, which he uses throughout this book; the OCR shows him closing the verse with simul. » at his other printing of it, so our twin has probably lost the », not Migne. Not repairable from here — guillemets must be 1:1 with the twin. Filed in §4.

3.3 0067B ad se imperfectionem creaturae revocare — OCR says perfectionem. DECLINED.

The OCR of p. 39 reads ad se perfectionem creaturce re- vocare. Our twin has imperfectionem. The sense supports imperfectionem (God recalls the creature's imperfection to himself, the whole clause being about imperfectio not imitating the Word's form), Augustine's De Gen. ad litt. I.4 has imperfectionem, and a two-letter loss at a line start is exactly what this scan does elsewhere (CLAUDE.md records this copy clipping the left margin on some pages). So the OCR is the likelier error here — but that is a judgment, not a reading. Rendered imperfectionem ("the imperfection of the creature"); no marker either way; PDF p. 39, left column, is the leaf that settles it.

3.4 ⭐ 0071C primo invisibilibus credit — CHECKED, and it is MIGNE'S. A negative result.

Migne prints the same Isidore excerpt twice, four columns apart, and the two copies differ:

• 0071C (mine): primo die, lucem fidei, quoniam primo invisibilibus credit propter quam fidem Dominus apparere dignatus est.
• 0082C (chunk 0004, another stint): primo die lucem fidei, quando primum visibilibus credit, propter quam fidem Dominus visibiliter apparere dignatus est.

The 0082C form is the coherent one — faith in visible things, for which faith the Lord deigned to appear visibly — and it is very tempting to call 0071C our corruption. The OCR of p. 41 reads quoniam primo invisi- bilibus credit propter qoam fidem Dominus apparere dignatus est, and the OCR of p. 46 reads quando primum visibilibos credit, pro- pter quam fidem Dominus visibiiiter. Both variants are on the plate. Migne set the excerpt twice from different copy. Rendered as printed on each side; no marker; logged as a checked negative, which is worth having precisely because §3.1 came back the other way.

3.5 Real words that read oddly, rendered as printed, no marker (brief addendum A)

• 0073A propter binarium principem alternitatis — "the binary number, the chief of alternation". alternitas is attested medieval Latin; Jerome's Adv. Iovin. I.16 argues from the binary as the number that first departs from unity, which is what stands here. Rendered.
• 0070A quibus alia duo inserta sunt, aqua, scilicet, et terra, quibus insunt ignis et aer — the apposition aqua … et terra sits after inserta sunt but must belong to the FIRST pair, since quibus insunt ignis et aer makes water and earth the containers. Rendered in Migne's order, ambiguity intact ("into which two others were inserted, water, namely, and earth, in which are fire and air"). Reordering to disambiguate would be a silent repair of his syntax.
• 0075B the doubled ablative absolute — separata labe, ac fluctibus carnalium tentationum, tanquam aridam habet terram tentationibus separatis — says the separating twice, in different words. Not a [d:] (the runs are not the same run), and the 0082C twin phrases it once (tanquam aridam habet mentem a perturbationibus maris). Both rendered.
• 0080D quia spiritualis quisque effectus, et Deo similis, omnia judicat — effectus as a participle ("made spiritual"), echoing 1 Cor 2:15. Ordinary. Rendered.

3.6 Unexpressed subjects supplied, with what decided each

• 0072A quia si culpam rectitudo eliminat, cogitationis nostrae foribus assidens … pulsat. Supplied "guilt" as the subject of assidens … pulsat. Decided by the sentence's own argument, not by proximity: Gregory's point is that sin is never so driven out that one rests unshaken — what returns to knock is culpa, the thing rectitudo has just expelled. Rectitudo is the nearer nominative and would reverse the moral.
• 0076C hinc esse aiunt quod multo longinquior a terris atque altior quam luna incedit. Supplied "it" = the sun. Decided by quam luna: the term compared to the moon cannot be the moon.
• 0079C ut inter similia … distinguerentur a caeteris — left with "they", the kinds, as the Latin leaves it; both readings (kinds / individuals) yield the same claim here, so nothing is at stake and nothing was chosen.

§4 · PENDING TEI PATCHES / PLATE DEBTS — the Latin column disagrees with the plate here

None of these is repairable from the English side; each needs an eyes-on read and a data/tei-patches/8950.json entry. The English mirrors the TWIN in every case except §3.1, which is called out there.

1. 0069B recerunt → fecerunt (§3.1). English already renders fecerunt; the Latin does not yet. This is the one live divergence between our two columns in this stint.
2. 0078D possunt,ita → possunt, ita — a lost space; the OCR has the space. Invisible in the English, but it is a defect in the Latin column.
3. 0067C — Migne prints a REPEATED « our TEI drops: « In principio fecit Deus « coelum et terram, ». The OCR's stray c ccelum is that guillemet.
4. 0069C — same class, second instance: « Ubi eras cum me laudarent astra matutina, et « jubilarent omnes filii Dei? ». ⚑ This is a CLASS, not two accidents. Migne repeats « at the head of a continued line throughout this book; our TEI keeps it at 0069B and 0079A and drops it at 0067C and 0069C. Since guillemet parity is checked against the twin and not against the plate, every dropped one is invisible to verify-english — a later stint should expect more, and the count of « against » per chunk is the cheap detector (0000 and 0003 run two opens long, 0002 two closes long).
5. 0069B — the Eccli 18:1 quotation is unclosed in our twin (§3.2); the OCR closes it at Migne's other printing. Possibly the same class as 3–4, in the other direction.
6. 0077B–C — the AUG. siglum is an italic span, not an [n:] note (§1a). Very likely a TEI markup defect. Note parity is unaffected (both sides have 23 [n:]), so nothing downstream can see it, and this book's fontes index will silently lose one citation.

Plate coverage for this range: columns 0067–0070 and 0081–0082 are on the read list (PDF pp. 39, 40, 46; page = (column + 11) / 2). Columns 0071–0080 — pp. 41–45 — are UNREAD, and that is why §5's divergences at 0072D, 0073B, 0074D, 0075A, 0076A, 0078C and 0080B carry no marker.

§5 · 7a″ — THE VULGATE COLLATION, ALL 168 SPANS

Collated against sources/vulgate/clementine-flat.txt, cited by book chapter:verse. The spans that AGREE are listed too — a list of divergences alone cannot be told from a report by an agent who only chased the ⚠ marks.

5a. MARKED — five [var:], every one on a read column and OCR-corroborated

Notes on two of them. 0067C: Migne prints creavit three times in the same paragraph and fecit once, in the sentence whose whole point is that principium = the Son — so the divergence sits inside the argument, not beside it; marked once, at the divergent occurrence only. 0070A: Domini is printed four times in this range (0070A, 0070B ×2, 0070C), and Migne's own Trinitarian gloss two lines later parses the phrase as Spiritus Dei, Spiritus sanctus — i.e. the exposition uses the Vulgate's word and the lemma does not. Marked once, at the first full occurrence, per Pattern 14's practice of noting rather than repeating.

5b. DIVERGENT, and DECLINED FOR THE SHIP TEST — columns 0071–0080 are unread

Each of these is a real difference from the Clementine and would carry a [var:] on a read column. All are rendered as Migne prints them. The leaf that settles each is PDF p. (column + 11)/2.

5c. DIVERGENT and DECLINED ON THE MERITS — the marker was weighed and refused

• 0069C Job 38:7 — Migne: « Ubi eras cum me laudarent astra matutina, et jubilarent omnes filii Dei? »; Clementine 38:7 cum me laudarent simul astra matutina…, and Ubi eras is 38:4. Read column, OCR-confirmed as Migne's, so a marker was available. Refused: the omission of simul changes nothing the gloss argues (it argues that the astra matutina are the angels), and the join of 38:4 to 38:7 stays inside the chapter Migne actually cites, Job XXXVIII — a citation practice, not a text divergence. Firing here would spend the marker on noise. ⚠ The same lemma is clipped to its first half at 0068C, which agrees.
• 0079D Gn 1:22 — Migne: « Benedixitque eis Deus dicens: Crescite et multiplicamini, »; Clementine has no Deus. One supplied noun, no argument on it, unread column. Declined.
• 0081A — « fecit Deus hominem ad imaginem et similitudinem suam, » conflates Gn 1:26 (ad imaginem et similitudinem nostram) with 1:27 (creavit … ad imaginem suam). Read column, so a marker was available, and refused: this is the gloss's own back-reference ("Post haec…"), a recall rather than a proof-text, and no step of the argument rests on the wording. Contrast 0081A's Paul quotation eight words later, which IS adduced as proof and IS marked.
• 0077D — « Secus firmamentum coeli, » for Gn 1:20 sub firmamento caeli. No marker by construction: Migne labels it Alia translatio, so the divergence is already stated on the page in his own voice, and ours would tell the reader twice. (Same instinct as Pattern 12's rule about Migne's own ( sic ).)

5d. COLLATED AND AGREEING — checked, exact, no divergence

Gn 1:1 « In principio creavit, » (0067B, 0068B, 0068C) · « In principio creavit Deus coelum et terram. » (0073D) · « In principio creavit Deus coelum. » + « Et terram, » (0069B, the Mystice split) · Gn 1:2 « Terra autem erat inanis, » · « Et vacua, » · « Inanis et vacua. » · « Et tenebrae erant, » / « Tenebrae erant, » · « super faciem abyssi, » · Gn 1:3 « Dixitque Deus: Fiat lux. » · « Fiat lux, » · « Et facta est lux, » / « Facta est lux; » · Gn 1:4 « Et vidit Deus lucem, » · Gn 1:5 « Factumque est vespere. » and its repeats · Gn 1:6 « Fiat firmamentum, » · « Fiat firmamentum in medio. » · Gn 1:7 « Et fecit Deus firmamentum, » · « Divisitque aquas quae erant, » · « et factum est ita » in all six of its printings · Gn 1:8 « Et factum est vespere, » · Gn 1:9 « Dixit vero Deus, » · « Congregentur aquae, » and « Congregentur aquae? » · Gn 1:10 « Et vocavit Deus aridam terram. » · « Congregationesque aquarum appellavit maria. » · Gn 1:11 « Et ait: Germinet terra, » · « Germinet terra herbam, » · « Facientem semen. » · « lignum pomiferum. » · Gn 1:12 « Et protulit terra herbam, » · « Lignumque. » · Sementem. » · Gn 1:13 « Et factum est vespere et mane. » · Gn 1:14 « Dixit autem Deus: Fiant luminaria, » · Et dividant diem. » · « Et sint in signa, » / « in signa, » · Gn 1:15 « Ut luceant, » · Gn 1:16 « Luminare majus, » · « Et luminare minus, » · Gn 1:17 « Ut lucerent, » · Gn 1:18 « ut praeessent diei, » · « et nocti, » · « dividerent lucem ac tenebras, » · Gn 1:20 « Dixit etiam Deus, » · « Dixit etiam Deus: Producant aquae reptile, » · « Animae viventis. » · « Sub firmamento » , · Gn 1:21 « Creavitque Deus cete grandia, » · « Et omnem animam, » · Gn 1:22 « Crescite. » · « Crescite et multiplicamini, » · Gn 1:24 « Dixit quoque Deus: Producat « terra, » · « Producat terra, » · « In genere suo. » · Gn 1:25 « Et fecit Deus bestias terrae, » · Gn 1:26 « Faciamus hominem. » · « ad imaginem nostram, » · « Et praesit, » · « Et praesit piscibus maris et volatilibus coeli. » · Gn 1:27 « Et creavit Deus hominem ad imaginem, » · « Et creavit Deus, » · « ad imaginem suam. » · « ad imaginem Dei, » · « Masculum et feminam creavit eos, » · Gn 1:28 « Et dominamini piscibus, » · « Et dominamini piscibus maris. » · Gn 1:29 « Dixitque Deus: Ecce dedi, » · Gn 1:30 « Et factum est ita. » · Gn 1:31 « Viditque Deus, » · « Et factum est vespere et mane, » · Ps 101:26 « Initio tu, Domine, terram fundasti, » (exact, including the vocative commas) · Dn 3:72 « Benedicite, lux et tenebrae, Domino, » (exact) · Rom 7:25 « Mente servio legi Dei, » (exact — the span the brief omits, §0) · 1 Cor 4:16 « Imitatores mei estote, » (exact).

⚠ coelum for caelum is orthography, not divergence, and is not listed as one. Migne sets coelum throughout; the Clementine sets caelum. Same for Terram/Maria capitalized as proper names in the Clementine and lowercase in Migne's lemmata at 0073D–0074A.

§6 · SECTIONS READ CLEAN

Read paragraph by paragraph against the Latin, asking only whether the English asserts what Migne asserts (Test 4), after the counts were known clean:

• 0067B–0068C — the In principio / Dixit Deus, Fiat argument, the densest and most negation- heavy passage in the stint (Non enim habet vitam informem Verbum Dei, cui est idem esse quod vivere against Creatura vero spiritualis habet informem vitam, cui non est hoc esse quod vivere). The two limbs are deliberately mirrored and the negatives are the only difference between them; both checked twice.
• 0072A–0073B — Gregory on evening, Augustine on the firmament, Jerome on the binary. Clean.
• 0075C–0077B — Isidore's day-by-day allegory and Bede on the luminaries. Clean.
• 0078A–0078B — Theodoret on the beasts, the longest continuous narrative here. Clean.
• 0079C–0081D — Augustine on secundum genus suum, the image of God, and the sixth day. Clean.

Candidates rejected, with reasons, so a blind reader knows they were looked at: §3.1–3.5 above, plus — nautarum at 0071B (a real word, and divers are what is described; not urinatorum, and no marker on a corpus argument, addendum A) · Septem sunt coeli at 0069B listing seven names for what are then re-ordered by Jerome to three (Migne's source, not a defect) · VERS. 17, 18-- at 0076C printing no period before the dashes where every other address in the range prints one (reproduced verbatim; a spacing artifact of the kind §4.2 records, not worth a patch entry on its own) · the five . . . . ellipsis runs after VERS. n.-- (Migne's, reproduced with his own spacing).

STINT 2 — chunks 0004–0007 (cols 0082A–0095A)

Filed as cruces-0004.md; merged verbatim, headings demoted one level. Nothing in it was rewritten — where a later stint corrected it, the correction stands in that stint's section.

Translated 2026-09-05, one stint, 4 chunks (5,737 Latin words), on Opus. Bands: 0082A–0095A (CAPUT II from Gn 2:1, CAPUT II cont., CAPUT III to Gn 3:14). ⚠ The launch message named this range "0071B–0079C"; the chunks' own frontmatter and the lemma brief both give 0082A–0095A, and the word count (5,737) matches exactly. The band figure in the launch message is wrong; the chunk numbers are right. Nothing was translated outside 0004–0007.

Marked-span count: 215 in my four chunks, against the 215 printed in data/briefs/8950-lemmata-0004.txt. Exact. Method: guillemet spans matched first («.?», non-greedy, DOTALL), then italic spans in what remains — italics nested inside a guillemet span are not counted separately, which is the brief's own convention. Per chunk: 0004 27 « » + 27 …*; 0005 35 + 25; 0006 20 + 12; 0007 43 + 26.

Apparatus fired: 3 [var:] · 1 [nt:] (Migne's own foot note) · 0 [sic:] · 0 [cj:] · 0 [d:] · 0 [ed:]. Parity with the Latin twins, per chunk, all exact: column anchors (14/14/10/14, verbatim and in order), [n:] notes (39/33/15/41), guillemets, question marks (2/4/5/5), italic spans. Punctuation compared band by band (colons, semicolons, guillemet-opens, question marks): zero mismatched bands except the single licensed Pattern 8b colon at 0088D (§5). EN/LA word ratio 1.49 / 1.48 / 1.51 / 1.53.

§1 · CONVENTIONS — inherited, and what this stint had to decide

Ruth (8968) §1 and Tobit (8970) §1 govern unchanged and were not re-opened: VERS. n.-- passes through verbatim; the lemma is Englished; a fragmentary lemma is rendered as the phrase it stands for; the attribution sigla are ordinary [n: …] notes with Migne's own internal spacing, contents unanglicized; the guillemets are the plate's, 1:1, and their contents are Englished.

#### (a) Sigla spacing reproduced character for character (BEDA, in Hexaem. ) ×2 · (BED., ubi supra.) · (PROCOP. in Gen. ) · (ISID. ex Aug. ubi supra. ) · (AUG. ubi supra, c. 6) (no terminal period — his) · (AUG., l. VI de Gen. ad litt., c. I, tom. III.) (Roman I for the chapter number — his). All carried as printed, uncorrected.

#### (b) usque ad → as far as, 77 times, no verb and no object ever supplied Every italic usque ad in these four chunks follows an etc. and names the point at which the abridgment resumes — the etc. splice test passes on all of them. The resumption tails land mid-clause constantly and are left mid-clause:

• 0085D as far as he created that in which he should live (no antecedent for qua).
• 0089D as far as and to subjugate to the dominion of reason, grace cooperating — the object of subjugare is inside the cut. Nothing supplied.
• 0084A as far as and the rest (et reliqua).
• 0084B as far as Therefore on the seventh day no evening came on — the tail opens with a capital Ideo, mid-sentence in ours; kept.
• 0084B as far as and more truly morning? — the tail carries Migne's ? and keeps it.

One variant form of the formula: usque ad haec at 0091D — Oblivio tamen, etc., usque ad haec c. 10: Providet bonos…. The italic span includes haec, so the English italic span is as far as these and the chapter reference stands outside it, as Migne sets it. This is the formula, not ordinary Latin (it follows an etc. and names the resumption).

Three ORDINARY, non-formula usque ad in these chunks — none italic, none after an etc., and each rendered by sense, per the brief's §D:

1. 0084C Numerati sunt sex dies quibus universitas saeculi a capite usque ad finem → "from beginning to end". The head of a sentence whose own splice-formula stands four words later; the two sit in one clause, exactly the 8967 col 0953C shape.
2. 0086C pertingentem usque ad lunarem circulum (Strabo on paradise's height) → "reaching as far as the lunar circle" — the English coincides with the formula's word, but by sense, not by formula.
3. 0089C nomina quae usque ad divisionem linguarum in Genesi legimus (Jerome on Hebrew as the first language) → "until the division of tongues" — temporal.

#### (c) The vocabulary axes

• gentes — two occurrences, rendered two ways, both by the axis: 0089C Significans gentes quae salvae fierent in Ecclesia per Christum → Gentiles (the Church-of-the- Gentiles typology, the lemma being Adam naming the animals and Isa 65:15 nomine alio); 0087B Haec flumina gentibus, etc. → nations (plainly generic, Augustine on the four rivers' fame). No natio, no gentilis, no gentilitas, no ethnicus in range.
• sacramentum → sacrament (0086D, in altero sacramentum vitae aeternae). No mysterium in range, so the 1:1 pair is not exercised.
• diabolus → devil throughout (7 occurrences, all chunk 0007). No daemonium, no satanas.
• ⚑ turb- fires three times and every one is the verb turbare — 0091A turbavit, 0093A turbata est (Ps 41), 0093C turbatur. Zero turba in range. Instrument checked before reporting, per the brief.
• The commentator's own voice is not archaised (dicit/ait → "says"); -est/-eth forms appear only inside quoted scripture and in second-person singular address.
• Pattern 17. Second person singular → thou: 0089A Si comederis → "If thou eat"; 0084B Ex 20:8 sanctifices → "that thou keep holy"; 0088B quod tenes → "what thou holdest"; 0093B Jo 1:48 Cum esses… vidi te → "When thou wast… I saw thee"; 0093D Adam, ubi es? → "Adam, where art thou?"; 0095A repes → "thou shalt creep". Plural → you: 0092A eritis sicut dii and Cur praecepit vobis Deus? → "you shall be as gods", "Why hath God commanded you?" — the serpent addresses the pair, and the number changes inside a single Gregory paragraph, which is the distinction Pattern 17 exists to keep.

#### (d) Heads CAPUT II. → CHAPTER II.; CAPUT II. (cont.) → CHAPTER II. (cont.); CAPUT III. → CHAPTER III. Numerals mirror the plate — Migne sets these in numerals, so the English does, per the corpus measurement (26 works print CHAPTER XIII., none spells above ONE).

#### (e) Frontmatter copied VERBATIM from each Latin chunk, field order included Chunk 0007's Latin frontmatter orders words → noteCount → incipit, unlike its three siblings; the English copies that order rather than normalizing it.

§2 · ⭐ MIGNE'S OWN FOOT-OF-PAGE NOTE AT 0094D — what I did, exactly

The Latin (injected at chunking from the plate) reads:

« Ventrem » LXX addiderunt, sed in Hebraeo habetur « pectus ª [cn: a | In Hebræo est gahon, quod ventrem et pectus significat.] » tantum, …

What I did:

1. The printed body reading stands untouched and is what the English renders. The note emends nothing: pectus is still what the Hebrew is said to have, and the English says so.
2. An [nt: …] at the same position, carrying the note in English, per the 2026-08-24 ruling and Pattern 15's prose-vs-locator test — this note is a sentence that says something, so a Latin-only rendering would strand an English reader: « breast [nt: In the Hebrew it is gahon, which signifies belly and breast.] » only,
3. gahon kept as it stands, in its italics, untransliterated and unnormalized.
4. ventrem / pectus inside the note are TRANSLATED (belly / breast), keeping their italics — they are the two Latin renderings the note is adjudicating between, and the English body two words away already renders the same two words « Belly » and « breast » from the guillemet lemmata. Leaving them Latin inside a translated note would make the note unreadable to the reader it exists for.
5. Migne's raised key ª is NOT carried into the English. Corpus precedent: 9005/0000 (Baruch, the founding asterisk case) and 8996/0001 both place a bare [nt: …] where the Latin carries [cn: | …], with no key character on the English side. The key belongs to the Latin apparatus, which keeps it; on our side the marker is* the key.
6. Guillemet parity is unaffected — the [nt:] sits inside the same « … » the [cn:] sits inside, so the counts are 1:1.

⛔ TWO VERIFIER ERRORS FIRE ON THIS, AND BOTH ARE INFRASTRUCTURE, NOT TRANSLATION

node scripts/verify-english.mjs 8950 reports exactly two things in my range:

`` 0007.md: note markers — Latin 41, English 42 0007.md: [cn: a | …] is malformed — the form is [cn: <Migne's note number> | <his note>] ``

Both have one cause: *this note's key is a LETTER (a), and verify-english.mjs knows only digits and .** Two places in it:

• line ~184, the form check: /^(?:[0-9]+(?:-[0-9]+)?\?|\) \| \S/ — a bare a fails it.
• line ~94, inSequence: n.kind !== 'cn' || /^\\s\|/.test(n.text) — only the asterisk-keyed [cn:] joins the note sequence, so this lettered one does not, and the [nt:] I was told to write has nothing on the Latin side to pair with. Hence 41 vs 42.

⚑ I did not touch the verifier, and I did not touch the Latin chunk: peers are working the same work concurrently and this is a corpus-wide script. The English is as Wilson's ruling requires; the fix is two regexes (admit a single lower-case letter key alongside in both places), and it belongs to whoever merges the stints. ⚠ Note that the ruling in translation-style.md §19 and CLAUDE.md is written for the asterisk* layer specifically; a lettered key is a third shape neither the prose nor the code has met. It behaves exactly like the asterisk layer — editorial prose, not a Forte conjecture — so it takes the English twin, but somebody should say so in §19 rather than leaving it to inference.

§3 · ⚠⚠ 7a″ — THE VULGATE DIVERGENCE LIST. This is the deliverable.

Every lemma in these four chunks was collated against sources/vulgate/clementine-flat.txt by grepping the verse, never from memory. Both the agreements and the divergences are recorded, because a list of findings alone cannot be told apart from a report by an agent that never looked.

3.0 ⛔ WHY ONLY THREE [var:] FIRED — the plate gate, and it is the binding constraint

data/plate-reads.json has no entry for 8950 at all. The launch message states that PDF pages 39, 40, 46, 52, 53, 60, 67, 74, 82, 89 and 96 of raw/scans/pl113/patrologiaecurs04migngoog.pdf have been read; at that volume's map (page = (column + 11)/2, verified in CLAUDE.md) the pages overlapping my range are 46 (cols 0081/0082), 52 (cols 0093/0094) and 53 (cols 0095/0096). Everything from 0083A to 0092D is unread, which is 10 of my 14 columns.

So: markers only on 0093/0094 (three fired), and the declined list for the rest. A [var:] is a public claim that MIGNE'S TEXT diverges from a witness, our Latin is not the plate, and 9003 showed the digitization manufacturing a Vulgate divergence that a collation then dutifully found — the one class no amount of collation can catch. The list below names every site and says which leaf settles it, so the markers can be fired later at the cost of one page render each.

⚠ AND EVEN THE THREE THAT FIRED WILL FAIL plate-gate.mjs UNTIL 8950 IS ADDED TO data/plate-reads.json. They stand on read pages; the read is simply not recorded. Whoever merges must either record pages 52–53 for 8950 or withdraw the three markers. Naming this is not a request to skip the gate.

3.1 The three [var:] fired (all chunk 0007, all on plate-read page 52)

Rejected on those same read columns, and named here because a blind reader will look:

• 0093A « Ad meipsum turbata est anima mea. » [n: (Psal. XLI.)] against Ps 41:7 Ad meipsum anima mea conturbata est. Word order plus simplex for compound. Real, and no marker: the sense is identical and Pattern 14 fires on divergence "in a way that changes the sense." Logged, not asserted.
• 0093D « Vocavitque Deus Adam, » against Gn 3:9 Vocavitque Dominus Deus Adam — Dominus dropped. Not sense-changing; and the same chunk prints the full Vocavitque Dominus Deus Adam correctly at 0093C, twenty lines above, which is an internal control that the shorter form is a real feature of the second citation and not a systematic habit. Logged, no marker.
• 0093B sub ficus for sub ficu. ficus is a well-formed Latin word (nom. sing., gen. sing., nom./acc. pl.), so Pattern 7's non-word clause does not reach it and [sic:] has nothing to wrap; the Addendum-A rule is explicit that a real word that reads oddly is Migne's until a plate says otherwise. Rendered as an ordinary "under the fig tree". A 400 dpi look at p. 52 would settle whether the plate sets ficu.
• 0094B « Ego dixi: Domine, miserere mei, sana animam meam, quia peccavi tibi » — checked verbatim against Ps 40:5. Agrees exactly, including the asyndeton. Recorded because a checked negative is worth having.
• 0094A/0094B/0094C/0094D — « Mulier quam dedisti, » (×3, Gn 3:12), « Et dixit Dominus Deus ad mulierem, » (Gn 3:13), « Serpens decepit, » / « Serpens decepit me, » (Gn 3:13), « Vidit igitur, » (Gn 3:6), « Et ait Dominus Deus, » / « …ad serpentem, » (Gn 3:14), « Super pectus tuum, » (Gn 3:14): every one agrees with the Clementine verbatim as far as it runs. Nine clean lemmata on a read page.
• 0095A « Pectore et ventre repes. » is Gregory's own paraphrase, not a lemma — Gn 3:14 reads super pectus tuum gradieris. No marker: it is the moralist's wording, and the gloss that follows (Ventre repit… Serpit pectore…) is built on it. Logged.
• 0082A–0082D (the other read column) — « Igitur perfecti sunt coeli et terra, » and « …et omnis ornatus eorum. » agree with Gn 2:1 (Migne's coeli for caeli is orthography, and an added comma after coeli is his punctuation); « Complevitque Deus, » ×2 agrees with Gn 2:2; « Mente servio legi Dei » agrees with Rom 7:25 verbatim as far as it runs. Nothing fired on the volume's other read column, and that is a result, not an omission.

3.2 The declined divergences — 0083A–0092D, unread columns. Each names its leaf.

PDF page for a column = (column + 11)/2 on patrologiaecurs04migngoog.pdf; the Gallica second witness (ark:/12148/bpt6k5505319w) is f = (column + 1)/2 and costs one IIIF request.

3.3 Divergences that are NOT divergences — labelled by Migne himself

Each of these is introduced by the plate as an alternative version, so it is data about Migne's practice and takes no [var:] by construction:

• 0082C Alia translatio: « Consummavit Deus die sexto opera quae fecit, » (LXX; Gn 2:2).
• 0082D In Hebraeo habetur « die septima. » (Jerome, on Gn 2:2 — and the Arctabimus ergo Judaeos argument that follows requires exactly this reading, so the plate is coherent).
• 0084D Alia translatio habet: « Hic est liber creaturae coeli et terrae, » (Gn 2:4 LXX).
• 0086C Quidam codices habent: « Eden ad Ortum. » (Strabo).
• 0087B Alia translatio: « Fons exiit de Eden, qui irrigat paradisum, » (Gn 2:10 LXX).
• 0087D Antiqua translatio habet « carbunculum et prasium. » (Gn 2:12 LXX; Clementine bdellium, et lapis onychinus) — and the gloss then etymologizes prasius from πράσον, so the reading is the point.
• 0091C Alii, « prudentissimus, » / Quidam codices habent « sapientissimus, » (Gn 3:1 callidior).
• 0092A « Gustate, et eritis sicut dii, » — Gregory's composition, the devil's speech built out of Gn 3:5, not a citation. Migne's own frame (quodammodo dicit) says so.
• 0087C Evilath for the Clementine's Hevilath (Gn 2:11): H-dropping, orthography, and the name is being etymologized (parturiens), which fixes the form.

#### 3.4 Lemmata collated and found to AGREE with the Clementine — recorded, per 7a″ Gn 2:1 (×2) · 2:2 (×2) · 2:4 (×3: Istae sunt generationes, Istae sunt generationes coeli, In die quo fecit) · 2:5 (×3: Et omne virgultum, Non enim pluerat Dominus, homo non erat qui operaretur terram) · 2:6 (×2) · 2:7 (×2 + the unguilleted Et inspiravit… in animam viventem, verbatim) · 2:8 (×2: Plantaverat autem, In quo posuit) · 2:9 (×2: Lignum etiam vitae, Lignumque scientiae boni et mali) · 2:10 · 2:12 (Bdellium) · 2:15 (×4) · 2:17 (×2 + In quocunque — quocunque/quocumque is orthography) · 2:18 · 2:19 · 2:20 (×4) · 2:21 · 2:25 (×2) · 3:1 (×5) · 3:5 · 3:6 (×4) · 3:6–7 across the verse join (Deditque viro suo, qui comedit. Et aperti sunt oculi amborum — verbatim, which is why a one-verse-per-line phrase search missed it) · 3:7 (Consuerunt folia ficus) · 3:8 (×3) · 3:9 · 3:10 · 3:12 (×3) · 3:13 (×3) · 3:14 (×3) · Ex 20:8 · Rom 7:24 (Migne's ! for the Clementine's comma is his punctuation, kept) · Rom 7:25 · 1Cor 14:35 · Ps 40:5 · Phlp 3:21 (unmarked, in Rupert).

§4 · THE DECLINED LIST — suspect readings, what would settle each, and the leaf

Zero [sic:] fired in 5,737 words. Nine of my fourteen columns have never been read at the plate, our Latin is not the plate, and on 9004 translating agents' "probably our-file" calls were right 5 times and wrong 9. Every candidate below is logged rather than asserted. None is a non-word: there is not a single non-word in these four chunks — no split run, no fused run, no doubled syllable — which is itself worth recording, since that is the class that has been right every time.

4.1 ⭐⭐ 0091C — Quod nudi erant et erubescebant : A MISSING NEGATIVE, AND I DID NOT REPAIR IT

Migne's last sentence of chunk 0006:

« Erat autem uterque nudus, » etc. [n: (ISID., in Gen., ex Aug. de Gen. contra Manich., lib. II, c. 15.)] Quod [0091C] nudi erant et erubescebant, etc., usque ad quia in consensu rea tenetur conscientia.

Rendered literally: "That they were naked and were ashamed". Gn 2:25 reads et non erubescebant, and the printed lemma four lines above in the same paragraph is « Erat autem uterque nudus. » — the verse whose second half is the negative.

The internal corroboration is overwhelming and that is exactly why no marker fired:

• 0090C, same chunk: nuda erant, nec erubescebant;
• 0090C Rupert: quod nuditatem suam non erubescebant;
• 0091A Rupert: nudus erat uterque et non erubescebant, quia quod erubescerent non habebant.

Four independent negatives against one affirmative, inside two columns. Frequency and internal corroboration prove an error EXISTS, never whose it is — this is reference_plate-read-triage's rule and CLAUDE.md 8a's measured version of it, and a [sic:] here would publicly accuse Migne's type of a defect that could just as easily be Corpus Corporum's dropped non. Isidore's source (Aug. de Gen. c. Manich. II.15) has the negative.

⚑ What settles it: PDF p. 51 of patrologiaecurs04migngoog (cols 0091/0092), or Gallica f46. One page render. If Migne prints non, this is a tei-patch, the English gains the negative, and no marker is ever needed; if Migne omits it, it is a [sic:] and one of the more interesting defects in the book. It must not ship unread in either direction without a reader knowing it is here, which is why it is the first entry in this section.

⛔ And note what I did not do: I did not render it "and were not ashamed" on the strength of the argument above. That is 7a's silent repair, and it would have been invisible to every check this edition runs.

4.2 0082B — a » that closes a quotation nothing opened, and a second one

…et intelligat quid sit, » et requievit Deus die septimo ab operibus suis: » quia et ipse in nobis haec bona operatur…

Two closing guillemets, no opening one. Under Pattern 5's corollary the plate is mirrored and nothing is supplied: the English carries both » in the same places. Almost certainly the first should be « (opening the Gn 2:2 quotation) — but supplying it would be exactly the punctuation repair Pattern 8 forbids, and 0082 is a read column (p. 46), so this is cheap to settle from a leaf already on someone's screen. Chunk 0004 therefore runs 28 « against 29 », deliberately.

4.3 0088A and 0088B — three unclosed OPENING guillemets in chunk 0005

Chunk 0005 runs 38 « against 36 », and the English mirrors it exactly. The sites:

1. 0088A Symmachus vero, qui « paradisum florentem ante transtulerat, hic « amoenitatem, » vel « delicias » posuit. — an opening mark before paradisum that is never closed, and it nests inside the following pair.
2. 0088B Ex eo quod additum est, « Et custodiret, innuitur qualis operatio esset: — the mark should close after custodiret,.
3. 0088C « Praecepitque ei dicens: Ex omni, etc. — opens the lemma and never closes. ⚠ Compare 0082D's « Complevitque Deus, » etc. and 0088D's « De ligno autem, » etc., both properly closed: this book closes its lemmata, so this is an anomaly, not a house style. Plus the mirror-image at 0085D, Et inspiravit in faciem ejus… in animam viventem. », which closes a quotation nothing opened. All four render as printed. p. 49 (cols 0087/0088) and p. 48 (0085/0086) settle them.

#### 4.4 Readings suspected and NOT marked — real words, rendered as printed

• 0088B Praecipitque for the expected perfect — §3.2 above; a real word, an internal control four lines away, no marker. Rendered "And he enjoineth him."
• 0093A hypocricis for hypocrisis. An attested medieval spelling of a Greek loan, not a non-word; Pattern 9 governs (attested medieval spellings are correct Latin, translated normally, no italics). Rendered "hypocrisy." Recorded because it looks like a defect.
• 0090B Quamobrem set as one word where the Clementine has Quam ob rem (Gn 2:24): orthography, universal in medieval MSS. No marker.
• 0090B VERS. 22, 24. skips verse 23, and the lemma splices two verses with Migne's own ellipsis . . . . — the spacing of the four stops is reproduced exactly, as is VERS. 5.--. . . . . at 0092C. Migne's ordering and his ellipses are not ours to tidy.
• ⚑ §4 of the brief, the "suspiciously tidy" class: I have nothing to raise. These four chunks are uniformly clean prose — the only rough spots are the guillemets above, and a normalization would not show there. Recorded as a checked negative, since that is the one form of evidence this class can produce from the Latin alone.

#### 4.5 Unexpressed subjects supplied — named, with what decided each

• 0082B et post ab operibus requiescere — a second infinitive still governed by dicitur eleven words back, not a new finite clause. Rendered "and afterwards to rest from his works", keeping it under "is rightly said."
• 0083D quod in tempore gratiae sublatum est, in qua perpetuum observat sabbatum — the antecedent of qua is gratia (feminine), not tempore (neuter) and not quod. Rendered "in which" with no gloss; the Latin's own morphology decides it.
• 0085A praecedens formationem non tempore sed origine — the participle can agree with the subject (God) or with formabilitatem. Both construe; I left the participle unattached in English too ("…preceding the forming not in time but in origin"), which preserves the ambiguity rather than choosing silently.
• 0086A Sed forte animae sexta die factae… — dative animae, so "to the soul made on the sixth day", not a genitive "of the soul". The head is cut by the abridgment; the case decides it.
• 0090C quia nullam legem in membris senserant legi mentis repugnantem, quae inobedientiam secuta est — quae is the law in the members, not the law of the mind: it is concupiscence that followed on disobedience. Rendered "which followed upon disobedience" in that position.
• 0094D sed de illo diabolus, et per illum, et in illo — the three illes are the serpent, the subject is diabolus. Rendered "but the devil out of him, and through him, and in him."

#### 4.6 A negation whose SCOPE the abridgment has cut away — 0094C, and I declined it > [n: (AUG., ibid., c. 28.)] Notandum quod non diabolus est permissus feminam tentare, > etc., usque ad sed ex hac homini invidisse ratio ostendit.

This fires test 2a's trigger exactly: a negative on a verb with a sed correlative nearby. Two readings construe — "the devil was not permitted to tempt the woman" (negative on permissus) and "it was not the devil that was permitted to tempt the woman" (negative on diabolus, with the sed naming the real agent or motive). The half of the correlative that would decide it is inside the usque ad cut, so 2a's own instruction — prove the host from the sentence's own structure, never from what reads well — cannot be carried out. I took the first, because Latin habitually sets non before the verb phrase and constituent negation needs positive structural proof I do not have; the rival is recorded here so that a reader meets it. This is the one place in the range where I chose between two live readings.

⚑ Contrast 0094D, where the same shape was resolvable and I did act on it: quia non ipse propria voluntate et natura fecit, sed de illo diabolus — here the sed clause survives and names diabolus against ipse, so the negative belongs on ipse, not on fecit. Rendered "because it was not he that did it by his own will and nature, but the devil out of him" — an earlier draft read "he himself did not do it", which floats the negative onto the verb and is the 8948/8990 defect exactly.

#### 4.7 Other places test 2a fires, checked and found correctly hosted

• 0084A requies Dei non ipsi Deo, sed rerum perfectioni inchoatur → "begun not for God himself, but for the perfection of things." Host = ipsi Deo, proved by sed.
• 0092D nullo enim regente per suam potestatem solus Deus beatus est → "for, with none ruling him, God alone is blessed through his own power." The negative is inside the ablative absolute; solus is affirmed** of God. This is the 8990 @0614B shape with the opposite answer, and the ablative absolute is the structural proof.
• 0092B quando nulli bene est ab ipso recedere → "when it is well for no one to withdraw from him." Host = the dative nulli, not est.
• 0088B non sic eo fruerentur, ut integritatem naturae suae usurpando vetitum violarent → "should not so enjoy it as to* violate…". The sic…ut* correlative fixes the host: what is forbidden is not the enjoying but the manner.
• 0090C Non erat hoc ignominiae, sed honoris… non erat insipientiae, sed securitatis → both negatives on the predicate genitives, both proved by sed.
• 0090C Aut nunquid confusionem hujusmodi natura, et non potius conscientia facit? → "Or is it nature, and not rather conscience, that makes confusion of this kind?" The non potius attaches to the alternative subject, and the English keeps it there.
• 0090D origo ejus pullulavit non ex conditione primaeva, sed ex juniore culpa → host = the prepositional phrase, proved by sed.

§5 · PUNCTUATION AND THE MARKS

• Question marks are at exact parity in all four chunks (2 / 4 / 5 / 5), and every one is Migne's. Three sit on resumption tails of the abridgment formula (0084B et verius mane?, 0092B …praescientiam de seipsis?, 0092B …incerta sua beatitudo?) — a tail is parsed as strictly as ordinary prose and keeps its mark.
• ⭐ ONE Pattern 8b recasting, and it is the only punctuation this stint added anywhere: 0088D Quaeritur quomodo loqui potuerunt, vel loquentem intelligere…? → "It is asked: how could they speak, or understand one speaking…?" Migne closes an indirect question with ? and English cannot; the colon carries the quaeritur frame so the mark survives. This is the single colon by which band 0088D's English exceeds its Latin, and it is the only band-level punctuation mismatch in 5,737 words. Every other Quaeritur in the range is truncated by etc. before the mark (0082C, 0084A, 0090A, 0092A) and takes no colon.
• Nothing else was supplied, softened or deleted. Band-by-band counts of colons, semicolons, guillemet-opens and question marks are identical in every one of the 56 bands except that one. Three drafting slips were caught by that test and repaired against the Latin (a colon at 0089B, semicolons at 0092A and 0093D) — recording it because the test is what found them, not a re-read.
• 0093A …erubescenda videbatur. etc., usque ad… — Migne sets a full stop before etc. The English does the same. Not a defect (cf. Tobit §1(a) « Et factum est. » etc.).
• Em-dashes: seven characters across five sentences (0085A ×2 in one apposition, 0085C, 0089B, 0091B ×2 around Rupert's quod impossibile est, 0094B), each where English needs the break. Not a defect class, not marked, not reported.

§6 · SECTIONS READ CLEAN

Read against the Latin sentence by sentence with the counts already known clean, asking only "does the English assert what he asserts": 0083A–0084B (Augustine on the seventh day, sixteen consecutive abridged notes, the densest usque ad stretch in the range), 0086C–0087D (Strabo and the etymologies — Evilath, bdellium, onyx, prase, with the two Greek words ὄνυξ and πράσον carried in Greek script as printed), 0090C–0091B (Rupert on nakedness, the longest continuous prose in the range and the only paragraph with no lemma at all), and 0094A–0095A. No findings in any of them beyond what is logged above.

§7 · ADDENDUM (added after the coordinator's harvester fix) — two recovered spans, collated

The lemma harvester paired guillemets positionally and lost spans on any line whose first delimiter was a » closing a quotation opened in an earlier paragraph. Regenerated, my brief prints 216, and two spans in my range were absent from the file I worked against:

[0085D] « flavit, vel sufflavit, » [0088A] « delicias »

Both were translated in the delivered English — they were never missing from the page, only from the inventory — but neither had been collated under 7a″. That is done below. No English text changed as a result.

7.1 ⭐ THE RECOUNT — 216/216, and my own 215 was one short by the SAME class of bug

Re-run with the span scan bounded to the paragraph, my four chunks give 216: 0004 27 « » + 27 …; 0005 35 + 26; 0006 20 + 12; 0007 43 + 26.

⛔ My original 215/215 was a false reconciliation, and the fault was mine as well as the file's. My scan ran an open cursor across the whole chunk rather than stopping at the paragraph, so an unclosed « swallowed the next paragraph's span. The exact site:

0088A … Ex eo quod additum est, « Et custodiret, innuitur qualis operatio [0088B] esset: …vetitum violarent. 0088B VERS. 16.-- « Praecipitque ei, » etc.

The « before Et custodiret never closes in its own paragraph (§4.3, site 2), so my cursor ran on, ignored the « of the next paragraph's lemma, and closed on its » — fusing two paragraphs into one span and deleting « Praecipitque ei, » from my tally. The same thing happened one paragraph later at 0088C's unclosed « Praecepitque ei dicens: Ex omni, etc., which swallowed 0088D's « Priusquam sciat puer… » — there the two errors cancelled, which is why the total came out one short rather than two.

⚑ The lesson, and it is the general one: an unclosed guillemet is a hazard for any span counter, not only for a positional pairer, and the two bugs are the same bug wearing different clothes — mine let an unclosed open run forward, the harvester's let an unmatched close shift the pairing. This work carries five unmatched delimiters in my range alone (§4.2, §4.3), all in chunks 0004–0005, which is why the class bit here rather than elsewhere. Any future count on a guillemet book should be paragraph-bounded and should report its unmatched-delimiter count alongside the total — had I printed "5 unmatched" beside "215", the disagreement would have been visible immediately instead of looking like agreement.

⚠ Note also that an unclosed « produces NO span at all under the paragraph-bounded rule, so 0088A's « Et custodiret and 0088C's « Praecepitque ei dicens are in nobody's inventory — mine or the brief's. They are logged at §4.3 and are not lemmata anyone can collate; recording that here so a later reader does not go looking for them in the brief.

7.2 [0085D] « flavit, vel sufflavit, » — a MENTION, not a quotation. No marker.

Et inspiravit in faciem ejus spiraculum vitae, et factus est homo in animam viventem. » [n: (AUG., ibid., l. VII, c. 3, 5, 7.)] Sine dubitatione, « flavit, vel sufflavit, » dicendum est, etc., usque ad poenam si contemneret?

Collation. Neither word stands in Gn 2:7, which reads et inspiravit in faciem ejus spiraculum vitae — and the lemma printed immediately above carries inspiravit, verbatim and correct (already recorded clean at §3.4). So the two are not competing readings of the same line: the lemma quotes the verse, and this span quotes two Latin verbs Augustine says ought to be used instead. De Gen. ad litt. VII.3 is arguing that the Greek ἐνεφύσησεν is better Latinized sufflavit than inspiravit, because inspiravit imports something the text does not say.

Therefore no [var:], and the reason is categorical, not evidential: Pattern 14 fires where Migne's printed scripture diverges from the received text, and this is not printed scripture at all — it is a discussion of how to render it. It is the mention-vs-use distinction the PG register section already draws (never silently English a mentioned particle into its meaning), applied to a Latin verb. Firing a marker here would tell a reader that Migne's Genesis reads flavit, which it does not, four words away.

Corroboration that the type is sound — both forms are ordinary Latin and the Clementine itself carries them: sufflavit at Is 40:7, insufflavit at Jo 20:22 and Sap 15:11, bare flavit at Is 40:24 and Sir 43:22. No non-word, nothing to [sic:], no plate read owed.

⭐ What the English had to protect, and did. The whole point of the sentence is the preposition. Our rendering keeps inspiravit as "he breathed into" ("And he breathed into his face the breath of life") against sufflavit as "he breathed upon" — so Augustine's distinction survives in English. Collapsing both into a single "breathed" would have been 7a″ conformation in its purest form: the exposition would have rested on a contrast the page no longer made. Recorded because it was got right by accident of literalism rather than by the brief, which did not list the span.

7.3 [0088A] « delicias » — Jerome reporting Symmachus. No marker.

Symmachus vero, qui « paradisum florentem ante transtulerat, hic « amoenitatem, » vel « delicias » posuit.

Collation. Gn 2:8 reads paradisum voluptatis a principio; delicias is not the Vulgate's word there, nor anywhere in Genesis 2–3 (the Clementine's only Genesis instance of the noun is Gn 49:20, praebebit delicias regibus). But the span is explicitly framed as a report — Symmachus vero … posuit — so it belongs with §3.3's labelled alternatives and takes no [var:] by construction, exactly like « Eden ad Ortum » and « carbunculum et prasium ». No plate read owed.

Two things worth recording:

• ⭐ The Vulgate does use this very word of paradise elsewhere: Ez 28:13, In deliciis paradisi Dei fuisti. So Jerome's report is not an exotic reading but the Latin Bible's own alternative vocabulary for Eden, which is precisely why he offers it.
• ⭐ Internal consistency, and the English keeps it. Two columns earlier, at 0086B, the gloss itself renders in Eden as « …in Eden. » Id est in deliciis — the same word, unmarked, in the commentator's own voice. Our English gives "in delights" there and "« delights »" here, so a reader can see that Jerome's report of Symmachus and the gloss's own paraphrase land on the same Latin word. Had the two been varied ("pleasures" / "delights") the connection would have been silently destroyed — the within-work reconciliation rule.

7.4 Consequential amendments to what I filed above

• §1's span figure and §3's opening claim now read against 216, not 215. The two spans join §3.3 (Migne's own labelled alternatives) — « delicias » — and a new category the file did not previously need: a mentioned word that is neither lemma nor citation — « flavit, vel sufflavit, ». Nothing in §3.1 or §3.2 changes; no [var:] is added or withdrawn, and the apparatus total for my range stands at 3 [var:] · 1 [nt:] · 0 [sic:] · 0 [cj:].
• §3.0's warning about data/plate-reads.json is now spent: the entry exists, pp. 39, 40, 46 and 52 are read, and all three [var:] at 0093A/0093B/0093D are confirmed as Migne's own type. The paragraph stays as filed as a record of the state at delivery.
• §2's verifier finding is likewise spent: the letter-keyed [cn:] is now accepted in both places, and Pattern 19 documents the letter as a third key shape. Chunk 0007 verifies clean.
• ⛔ §4.1 — 0091C Quod nudi erant et erubescebant — IS NOT SETTLED AND STAYS DECLINED. Its column is on p. 51, which nobody has read. One page render (or Gallica f46) decides whether this is a tei-patch restoring a dropped non or a [sic:] on Migne's own type. It remains the most important unresolved item in this range, and it is deliberately unmarked.
• The band range in §Preamble (launch message said 0071B–0079C; the chunks are 0082A–0095A) is confirmed as a launch-message error, not a translation one. Left as filed.

STINT 3 — chunks 0008–0011 (cols 0095A–0106D)

Filed as cruces-0008.md; merged verbatim, headings demoted one level. Nothing in it was rewritten — where a later stint corrected it, the correction stands in that stint's section.

Translated 2026-09-05, one translator, 4 chunks (5,041 Latin words). verify-english.mjs: zero errors on 0008–0011 (column anchors, [n:] notes, guillemets and question marks all 1:1 with the Latin twins; 0007's two errors are a neighbouring stint's and are untouched here).

Apparatus fired: 5 [var:] · 0 [sic:] · 0 [cj:] · 0 [d:] · 0 [ed:] · 0 [nt:]. Per-chunk parity (English / Latin): anchors 12/12 · 14/14 · 11/11 · 10/10 · notes 23/23 · 23/23 · 13/13 · 14/14 · « 29/29 · 46/46 · 29/29 · 25/25 · » 28/28 · 44/44 · 29/29 · 25/25 · ? 3/3 · 1/1 · 0/0 · 0/0.

SPAN COUNT — the brief's 183, confirmed. Counted in my own four Latin chunks: 123 closed « … » spans + 60 … spans = 183, matching the brief's header exactly. ⚠ The count reconciles only because the harvester pairs guillemets: the raw marks are 125 « and 124 », and the three unpaired ones (§3 below) appear in no inventory line. They are Migne's, not ours — the plate was read for all three.

⭐ THE WHOLE RANGE WAS READ AT THE PLATE, before the English was written, not after: archive.org patrologiaecurs04migngoog, PDF page = (column + 11) / 2, corner numbers read first on each page — pp. 53 (095/096), 54 (097/098), 55 (099/100), 56 (101/102), 57 (103/104), 58 (105/106). Nineteen twin/plate divergences were confirmed at 600–1200 dpi and are filed in data/briefs/8950-PENDING-TEI-PATCHES-0008.md; thirteen more were seen only at 300 dpi and are listed there as not acted on. Foot-of-page apparatus: checked zero on all six pages, both layers.

§1 · CONVENTIONS — decisions this stint made, stated for the merge

Ruth (8968 §1) and Tobit (8970 §1) govern and were not re-opened: VERS. n.-- passes through verbatim; the lemma is Englished; a fragmentary lemma is rendered as the phrase it stands for; sigla ride as ordinary [n:] notes with Migne's own internal spacing; usque ad → as far as, with no verb ever supplied. What follows is what this range had to settle.

#### (a) The Historice / Allegorice / Mystice tags are TRANSLATED, in their italic span Mystice. → Mystically. (0096A), Allegorice. → Allegorically. (0098B, 0101D). Corpus precedent, not invention: the shipped English carries 14 Mystically. and 2 Historically. and no untranslated tag. A note inside the tag's italic span stays inside it — 0096A prints Mystice. [n: (AUG., …)] and the English does the same.

#### (b) ⭐ NO [var:] on a quotation Migne HIMSELF labels as a rival version This range is dense with them — LXX (×7), Aquila, Symmachus, Theodotion, in Hebraeo, alia translatio. Where Migne announces the version, its divergence from the Clementine is the sentence's own subject, and a [var:] restates what the text already says. Marking them would have put ~12 markers in four chunks and buried the four that carry information. The rule taken: [var:] fires where the words are offered AS the text and diverge. Worked pair at 0095B, which is exactly why the rule was needed: Jerome's first quotation is given bare, under a naked (HIERON.), so it took the marker; his second is introduced Melius in Hebraeo habetur, so it did not.

#### (c) ⛔ WORDS follow the plate; PUNCTUATION follows the TWIN Nineteen twin/plate divergences were confirmed in this range (patch file, Part A), and they split into two classes that must be handled differently:

• A divergent WORD — the English renders what Migne printed, per the locked "plate where read" rule. Five sites: Deos (0097A), potest (0098A), interpretatur (0100A), intellectu (0102C), plus the four non-words whose SENSE the English renders.
• A divergent MARK — ⛔ the English follows the TWIN, not the plate. Three sites were first set from the plate and then reverted: « septempliciter: » at 0100A (twin ;), the semicolon after positio and the absent one after dedicatio at 0103C, and the dropped opening « at 0106A. The reason is mechanical and not a softening of Pattern 8: verify-english compares guillemets against the twin as a hard check, and the band-by-band colon/semicolon comparison (the test that caught 8989 and Tobit) partitions on the twin's anchors — so an English that follows the plate on a mark manufactures a mismatch in exactly the instrument that exists to find real ones, for a difference no reader can act on. The plate readings are in the patch file; when the patch lands, both columns move together. ⚑ Band comparison run and clean: colons, semicolons, question marks and guillemets match the Latin twin in every band of all four chunks, zero mismatches.

#### (d) The gentes axis in this range gentes of the peoples subdued by Rome, and of "all nations" out of which the Church gathers (0100B, 0106A) → nations, generic. sub gentibus sive sub Christianis regibus (0100B) → Gentiles, the Judaei/gentes opposition, with gens Judaea → the Jewish nation in the same sentence. Dispersus in gentibus of the scattered Jews (0099C) → nations. diversitas gentium (0106B) → nations. daemonibus (0104C) → demons, per the axis. ⚠ Instrument check run: natio appears nowhere in this range (every hit is damnatio-/condemnatio-/generatio-/interpretatio-); turba/multitudo — one multitudine (0099C) and one multitudinem (0106A), both multitude, no turba at all.

#### (e) usque ad — 20 formulas, 6 ordinary. Decided per occurrence, per the etc. splice test Formula (italic, after an etc., 20×): 0095B, 0095C, 0095D, 0096A, 0096B, 0096C ×2, 0098A, 0098B, 0098D, 0100B ×2, 0100C, 0101B, 0103D, 0104B, 0104D, 0105D ×2, 0106B. Ordinary Latin, unitalicised (6×), and each takes its own English: 0097A usque ad ultimum judicium → "until the last judgment"; 0099D and 0100A usque ad septimam generationem / usque ad tempus illud → "until"; 0102A ab Adam usque ad Christum → "from Adam to Christ"; 0103B per denarium usque ad Noe → "up to Noah"; 0102D usque ad diluvium → "up to the flood". Not one of the six sits inside an italic span, and not one follows an etc.

#### (f) The abridgment's tails are parsed as strictly as prose, and none was completed Three resumption tails land mid-clause and stay there: 0095C sed prius reluctatum est cum dolore consuetudini malae; 0096C quae falsum verum esse putavit; 0098A, where the plate reads sed potest in exsilio generata and the complementary infinitive has been cut out by the splice. Only there is a word supplied, in brackets — "but it can [be] begotten in exile" — because potest demands a complement that the abridgment removed. This is not the formula being completed (Tobit §1(b)); it is a bracketed English exponent inside a resumption. Recorded because it is the one place in the stint where anything was supplied.

§2 · THE FLAGGED SITE — 0104D, and it is a REAL divergence. [var:] FIRED.

VERS. 5.-- « Videns autem Dominus, » etc.

Verdict: Migne prints Dominus; the Clementine's Gn 6:5 reads Deus. The ⚑ MATCHES ONLY ELSEWHERE mark was right, and the marker stands.

Evidence, in order:

1. The file. sources/vulgate/clementine-flat.txt, Gn 6:5: Videns autem Deus quod multa malitia hominum esset in terra… The gloss is on Genesis 6 — its head is CAPUT VI., its neighbours are VERS. 3, 4. (Gn 6:3, Non permanebit spiritus meus) and VERS. 6-8. (Gn 6:6, Poenituit eum), and the lemma is addressed VERS. 5. So the verse is fixed independently of the words, and it is Gn 6:5.
2. The brief's other address, checked and rejected as the source. Gn 29:31 does read Videns autem Dominus quod despiceret Liam — but that is Jacob and Lia, four books' worth of narrative away, in a chapter this commentary has not reached and does not cite. It explains why a mechanical phrase search matched; it is not a candidate reading here.
3. ⭐ The plate, read at 1200 dpi (p. 57, col 104, band D, corner numbers checked first): VERS. 5. — « Videns autem Dominus, » etc. Non in Deum poenitentia cadit… Migne's type reads Dominus. So this is not the digitization manufacturing a divergence — the class plate-reads.json exists to catch, and the class that would have made a collation "find" a defect nobody printed. The claim is against the plate and the plate supports it.
4. Not the gloss's own word. The comment immediately below reads Non in Deum poenitentia cadit and ira Dei, i.e. the expositor says Deus while the lemma says Dominus — so the divergence is in the quotation alone, which is exactly Pattern 14's case.

Marker as fired: [var: Vulg. Gn 6:5 reads Deus, "God"].

⚑ And the same divergence recurs 25 columns earlier, unflagged by the brief because the phrase matches Gn 4:25 well enough for the pre-scan: 0102B « Posuit mihi Dominus semen aliud, » against Clementine Gn 4:25 Posuit mihi Deus semen aliud pro Abel. Plate read at p. 56 (corners 101/102) — Migne prints Dominus. [var:] fired there too. The check that fired once found a pair; a ⚑ mark is a place to start, not the extent of the class.

§3 · THE THREE UNPAIRED GUILLEMETS — all three are MIGNE'S, plate-read

• 0100C, unclosed « — LXX: « Habitavit in terra Ναΐδ Naid, … dum huc atque illuc profugus aberrat. Plate (p. 55): the quotation opens and is never closed. Mirrored unclosed, per Pattern 5's corollary.
• 0101C, unclosed « — VERS. 22.-- . . . « Tubalcain. [n: (RAB.)] Qui secundum Josephum… Plate (p. 56): likewise open. Mirrored.
• 0106A, orphan » — Ostium autem arcae, » etc. ⛔ This one is NOT Migne's. The plate (p. 58) prints « Ostium autem arcae, » etc., balanced. Corpus Corporum has dropped the opening mark. The English nevertheless mirrors the twin and prints the orphan », because guillemet parity against the twin is a hard verify-english check and the English cannot carry a mark the Latin column does not show. Patch filed; when it lands, the English takes the « too. Recorded prominently because a blind reader will meet this and read it as our defect — it is our SOURCE's, and the plate is clean.
• ⚠ The corollary that matters for the merge: two of three unpaired marks were Migne's and one was the digitization's, and no test but the plate separates them. Do not adopt a policy of mirroring unpaired guillemets as "his" without reading the column.

§4 · THE VULGATE DIVERGENCE LIST — every lemma collated, agreements included

Collated against sources/vulgate/clementine-flat.txt, never from memory. Agreements are listed because a list of findings alone cannot be told from a report by an agent that never looked.

Fired — 5 [var:]

Checked and DECLINED, with the reason — these are not defects

• 0095B « Ipsa conteret caput tuum, et tu conteres calcaneum ejus, » — diverges from Gn 3:15 in the second half, and is introduced Melius in Hebraeo habetur. §1(b): announced.
• 0097B, three spans (« Et ejecit, et habitare fecit contra paradisum… », « et ejecit Dominus, », « et habitare fecit ante paradisum… ») — all against Gn 3:24, all explicitly LXX or in Hebraeo. Announced.
• 0098C « Et inflammavit Dominus super Abel… non inflammavit? » — announced as Theodotion's. Not in the Clementine at all, and could not be.
• 0099C « gemens et tremens. », 0100A « Omnis qui occiderit Cain septem vindictas exsolvet, » / « Septem vindictas exsolvet. », « septempliciter » / « septulum » / « per hebdomadem » — announced LXX, Aquila, Symmachus, Theodotion. Announced.
• 0104B « filios deorum » / « Videntes filii potentum filias hominum. », 0104C « Non judicabit spiritus meus homines in sempiternum, quoniam caro sunt. » — announced Aquila, Symmachus, and the Hebrew against Gn 6:3's Non permanebit spiritus meus in homine in aeternum, quia caro est. Announced.
• 0105C, 0106A « Bicamerata et tricamerata… » — announced LXX against Gn 6:16.
• 0096A « Et vocavit Adam nomen uxoris suae Eva, » — Gn 3:20 spells Heva. Spelling only; no sense turns on it. Same for 0100D « Porro Enoch, » against Gn 4:18 Porro Henoch, and note that our own gloss spells it Henoch nine lines later at 0100C.
• 0101D « Non dico tibi septies, sed usque septuagies septies, » — Mt 18:22 reads Non dico tibi usque septies. A dropped usque in the first member; the sense and the arithmetic the gloss builds on are untouched. Declined.
• 0102A « Non enim judicabit Deus bis in idipsum, » — not a Vulgate verse. The nearest is Nah 1:9 non consurget duplex tribulatio, which shares no wording. Jerome's own maxim, not a quotation. No received reading exists to name, so [var:] has nothing to compare against. Declined and recorded, because its absence from the list would otherwise look like an oversight.
• 0100B « Disperge illos in virtute tua [n: (Psal. LIII)]. » — the text agrees exactly with Ps 58:12; it is Migne's citation that is wrong by five psalms. The note rides verbatim (refDisplay keeps what he prints); no [var:], since Pattern 14 marks the text.
• 0106A « Qui non colligit mecum, dispergit. » cited Matth. XI — the wording is Lc 11:23 exactly; Mt 12:30 reads qui non congregat mihi, spargit. So Migne quotes Luke under Matthew's name. Text agrees with its real source, so no [var:]; citation logged here. ⚠ The previous Matth. XI in the same chunk (0105A, Discite a me quia mitis sum et humilis corde) is correct — Mt 11:29. The two are not one error.
• 0104A « Et fuerunt omnes dies Mathusalae quos vixit anni nongenti sexaginta novem… » — Gn 5:27 reads Et facti sunt omnes dies Mathusala nongenti sexaginta novem anni. Jerome is quoting his own version inside a discussion of the codices' arithmetic, and Migne prints the Vulgate form as the lemma four words earlier (« Et facti sunt omnes dies, »), so the contrast is deliberate. Sense identical. Declined.
• 0100C « Habitavit in terra ad orientalem plagam Eden. » — Gn 4:16 reads habitavitque profugus in terra ad orientalem plagam Eden. A clipped lemma, not a divergence.

Checked and found to AGREE with the Clementine exactly

Gn 3:15 (Inimicitias ponam… Ipsa conteret caput tuum, with Migne's own ellipsis) · 3:16 (Mulieri quoque, Multiplicabo aerumnas, Multiplicabo aerumnas tuas) · 3:17 (Maledicta terra, Maledicta terra in opere tuo ×2, In laboribus) · 3:19 (Pulvis es) · 3:21 (Fecit quoque) · 3:22 (Et ait: Ecce Adam quasi unus ex nobis, and all five re-quotations of quasi unus ex nobis) · 3:23 (Et emisit eum) · 3:24 (Ejecitque Adam, Et collocavit ×2, Collocavit, flammeum gladium) · 4:1 (Adam vero, Concepit et peperit) · 4:2 (Fuit autem Abel, agricola, pastor ovium) · 4:4 (Abel quoque obtulit, Et respexit Dominus ad Abel) · 4:7 (In foribus) · 4:8 (Dixitque Cain) · 4:9 (Et ait Dominus ad Cain) · 4:10 (Vox sanguinis) · 4:11 (Nunc igitur maledictus eris ×2, Maledictus eris super terram) · 4:12 (Cum operatus, Vagus et profugus) · 4:13 (Major est iniquitas mea ×2) · 4:14 (Omnis igitur qui invenerit me) · 4:15 (Nequaquam ita fiet, Omnis qui occiderit Cain, Qui occiderit Cain, Sed omnis qui occiderit Cain, septuplum punietur, Posuitque Dominus Cain signum) · 4:17 (Cognovit autem Cain uxorem suam, Et aedificavit civitatem) · 4:19 (Qui accepit duas uxores) · 4:20 (Genuitque Ada Jabel) · 4:21 (Et nomen fratris ejus Jubal) · 4:23 (Dixitque Lamech, Occidi virum in vulnus meum, Occidi virum, in vulnus meum, et adolescentulum in livorem meum) · 4:24 (Septuplum ultio dabitur) · 4:25 (Cognovit quoque adhuc) · 4:26 (Quem vocavit Enos ×2, coepit invocare nomen Domini ×2) · 5:1 (Hic est liber generationis Adam, In die qua creavit Deus hominem, Ad similitudinem Dei) · 5:3 (Vixit autem Adam, Et genuit) · 5:24 (Ambulavitque cum Deo) · 5:25 (Vixit quoque Mathusala) · 5:32 (Noe vero cum quingentorum esset annorum) · 6:1 (Cumque coepissent homines multiplicari super terram) · 6:2 (Videntes filii Dei) · 6:3 (Dixitque Deus: Non permanebit Spiritus meus, Eruntque dies illius) · 6:9 (Noe vir justus, In generationibus suis) · 6:13 (Disperdam eos cum) · 6:14 (Fac tibi arcam de lignis, Mansiunculas) · 6:15 (Trecentorum cubitorum) · 6:16 (In cubito consummabis summitatem ejus, Coenacula et tristega) · 6:19 (Et ex cunctis) · 6:21 (Tolles igitur) · Ps 62:6 (Sicut adipe et pinguedine repleatur anima mea) · Mt 11:29 (Discite a me quia mitis sum et humilis corde).

⚠ One address anomaly, Migne's, reproduced and not reordered: the lemma at VERS. 31. (« Et facti sunt omnes dies, », 0104A) glosses Mathusala, i.e. Gn 5:27, and the lemma under the same address at 0104B (« Noe vero cum quingentorum esset annorum, ») is Gn 5:32. The recension's verse numbering is its own; Ruth's §1 rule (never reorder, log it) applies.

§5 · THE DECLINED LIST — suspect words, what would settle them, and why NO [sic:] fired

Zero [sic:] and zero [cj:] in 5,041 words, and this is a finding, not a gap. Every plate defect this range actually contains is one our twin has already tidied away, so a [sic:] on it would fail the verbatim-substring guard — the guard is right and the marker is owed, not withheld. All nineteen confirmed divergences are in data/briefs/8950-PENDING-TEI-PATCHES-0008.md; the ones that will take a marker after the patch are named there. Below is what a reader should know from the English side.

[sic:] OWED, after the patch — plate defect, twin corrected (all read at 600–1200 dpi):

1. 0097A Deos enim et aeternus est (twin: Deus). A real word with no grammatical slot — accusative plural as subject of a singular est — which is Pattern 12's own test (1). English renders the sense, "For God is both eternal and happy"; nothing in the English discloses it until the marker can be fired.
2. 0098C » et. Quia justus (twin: etc.). Migne's broken abbreviation.
3. 0100A « septulum; » (twin: septuplum). Non-word. ⚠ And a second thing is lost here even when it is patched: septempliciter (Aquila, adverbial) and septuplum (Symmachus, accusative) both render "sevenfold" in English, which flattens a distinction Migne is drawing between two Greek translators. There is no English exponent for it; Pattern 9's reasoning applies, and this note is the only record.
4. 0101C id est vulnera quod infixi (twin: vulnere). Neuter plural against a singular relative — no slot. English renders "by the wound which I inflicted".
5. 0104D contesimo anno (twin: centesimo). Non-word. English: "the hundredth year".
6. 0105D Potet in trecentis cubitis (twin: Potest). Non-word — and trecentis prints correctly in the very same clause, so it is a broken sort, not an orthography.
7. 0105D Arca quae treeentis cubitis (twin: trecentis). Non-word, two lines below (6), and the two together are the compositor having trouble with one word on one page.

Suspicions raised and REFUTED at the plate — recorded because a checked negative is worth having (brief §4 asks for exactly these):

• 0099C dilicta — our twin's reading, a non-word, and the natural suspicion is Migne's type. The plate reads delicta. This one is ours, and a [sic:] here would have been a public accusation against a clean page. It is also the shape brief §A calls decisive: the suspicion aimed at a non-word was right that an error existed and wrong about whose it was — which is the whole of plate-read-triage's rule with an instance attached.
• 0097D Remotus est — read oddly against gladius; the plate has Remotum est, an impersonal. English has no exponent ("it was removed" serves both), so nothing changes in the text and no marker is possible either way.
• 0102C intellecta diligeret — reads smoothly as intelligendo → intellecta → dilecta, a three-term chain, and that smoothness is exactly what brief §4 warns about. The plate reads intellectu. The English follows the plate — "might love with understanding" — and the rival (intellecta, "might love the things it understood") loses because the plate decides it, not because it reads worse; it reads better, which is the point.
• 0098A sed post in exsilio generata — construes cleanly ("but afterwards begotten in exile"). The plate reads sed potest, read at 1200 dpi. The tidy twin reading was the suspect and the rough plate reading won. See §1(f) for the bracketed supply.
• 0100A interpretantur — plural, distributing over Aquila, Symmachus and Theodotion. The plate reads interpretatur, singular, so the verb belongs to Theodotion alone and the first two clauses are elliptical. English follows the plate.

Raised, unresolved, and deliberately NOT acted on — thirteen punctuation and orthography divergences seen only in the 300-dpi full-page render (Part B of the patch file). A mark is precisely the class a 300-dpi glance gets wrong, so the rule this stint applied is: act on the plate only where the line was re-cropped at 600 dpi or better; otherwise log and leave the twin alone. Two are worth naming here because they touch structure rather than a comma: VERS. 23.-- at 0097B may be set VERS. — 23. on the plate (the dash before the number, against every other address in the range), and CAPUT IV. may lack its period.

§6 · UNEXPRESSED SUBJECTS — every supplied pronoun, named

• 0095B, et Dominus conteret Satan sub pedibus ejus → "under her feet". ejus is genitive singular and common in gender, so the Latin does not decide it. The antecedent taken is the ipsa/mulier of the lemma directly above, which the whole paragraph is expounding (Ipsa conteret caput tuum) — the head-crushing member of the verse, of which this clause is the gloss. ⚠ The rival is real and is named here rather than buried: Rom 16:20 has sub pedibus vestris, and a translator reaching for the remembered verse gets "your feet" or "his feet". Neither is what Migne prints, and ejus cannot be plural.
• 0095A, qui cum ab initio deprehenditur, percutere calcaneo molitur → "when he is caught": the serpent, from serpentis two clauses earlier; the same subject carries through non percutit and decipere tendit.
• 0096D–0097B, the whole Rupert paragraph → "he" is Adam throughout (tam miser factus Adam), and "let us" is the divine plural of quasi unus ex nobis which the passage is glossing. The one place the referent shifts is ob hoc veracem se esse rideat oculus nequam diaboli — subject oculus, not Adam.
• 0100D, Figurate progenies impiorum … civitatem exstruxit quam vocavit Henoch → "he called", singular, agreeing with progenies; the plurals that follow (dedicant, plantant, floreant, arescant) are Migne's own shift to the collective and are carried as plurals. Both numbers are in the printed Latin; neither was harmonised.
• 0103A, Quasi et iste portavit imaginem terreni qui pro Abel justo repositus est → "he who was put in the place of the just Abel" = Seth, named nowhere in the sentence but fixed by pro Abel and by ante Seth nullum legitur genuisse four lines below.
• 0106D, qui etiam ut sine cibo viverent divina facultate donaret → "they" are the animals of the preceding sentence, not Noe and his household; viverent is plural and the paragraph's subject throughout is animalia.

§7 · NEGATION — the hosts, named (test 2 / 2a)

Fired on the trigger population (a negative on a verb with a quantifier or correlative near):

• 0095D Terrae maledixit, non aquis — non is on aquis, not on the verb; the non … quia … non de aquis bibit chain confirms it. English keeps it on "the waters".
• 0096A sed non ut laborem homini inferrent — on the purpose clause, not on erant: the thorns did exist. English: "Thorns were there before, but not so as to bring labor". ⚠ This is the 8987 shape exactly (a sed non ut beside a verb that would take the negative plausibly), and the plate was checked: Migne prints erant sed non ut.
• 0099B qui noluit esse sub gratia, sed sub lege — the sed correlative fixes the negative on sub gratia, not on esse; rendered "who would not be under grace, but under the law".
• 0099D tanquam aut omnia non possit, aut invideat saluti — non on possit inside the first arm of aut … aut; the second arm is affirmative and must stay so.
• 0103B justorum enim est omnia viriliter agere, nihil femineum, nihil fragile usurpare — two nihil, both on usurpare, both carried.
• 0104C Non est incredibile ab hominibus, non ab angelis — the second non is on ab angelis, and the whole point is that the giants were begotten by men. Rendering it onto the verb would reverse the claim. English: "by men, not by angels".
• 0105A nihil fecerit quod poenitendo resipiscat — nihil is the object of fecerit; 7a⁗'s fourth question (does the clause still have the thing it acts on) answered explicitly.
• 0106B nec impulsu ventorum, nec impetu fluctuum, nec inquietudine animalium … inclinari posset aut mergi — three nec, all on the ablatives, all carried; the aut is inside the scope of the negation and English keeps it there ("neither tilted nor sunk").
• 0106C nihil est quod Deus suave et salubre facere non possit — a double negative Migne prints and the English prints: "there is nothing which God cannot make sweet and wholesome".

§8 · READ CLEAN — sections examined and found without defect

0095A–0095D (Gregory on crushing the serpent's head; the Augustine splices on Gn 3:16–18) · 0096B–0096C (the perizomata / tunics gloss; the two Augustine splices on quasi unus ex nobis) · 0096D–0097B (Rupert entire — the longest continuous passage in the range, and the one place where the printed Latin's antitheses could most easily have been smoothed; the aeterna felicitas, felix aeternitas chiasmus is Migne's and is carried in his order both times) · 0098B–0098D (Isidore on the two peoples; Theodotion on the sacrifice) · 0099A–0099B · 0100A–0100B (the seven vengeances, all four versions) · 0101A–0101B · 0102A–0102B · 0102C–0103A · 0103C–0103D (the name-interpretation catalogue) · 0104A–0104B · 0105A–0105C (the ark's chambers, Strabo and Augustine; the two accounts genuinely disagree in Migne and are left disagreeing) · 0106A–0106D.

⚠ One passage that reads oddly and was checked rather than repaired: 0106C prints habens in se longitudinem, id est, Trinitatis fidem ac longitudinem vitae et immortalitatis; latitudinem in charitate … altitudinem in spe. The word repeats where the triad length/breadth/height would lead a reader to expect something else, and the shape invites a [d:]. It is not a dittography: the two occurrences govern different phrases (Trinitatis fidem and vitae et immortalitatis) and neither is an eye-skip with no grammatical slot, which is Pattern 11a's test. Plate read at p. 58: both are printed. Rendered as printed.

STINT 4 — chunks 0012–0018 (cols 0106D–0120D)

Filed as cruces-0012.md; merged verbatim, headings demoted one level. Nothing in it was rewritten — where a later stint corrected it, the correction stands in that stint's section.

Stint 4 of 8950. Seven chunks, 5,898 Latin body words → 8,477 English (1.437×). Bands 0106D–0120D: CAPUT VII (the flood) through CAPUT XIV (Melchisedech). verify-english.mjs reports no error on 0012–0018 — column anchors, [n:] notes, guillemets, italic spans and question marks all 1:1 with the Latin twins. (The run exits non-zero only on the chunks other stints have not yet filed.)

⭐⭐ THE WHOLE RANGE WAS READ AT THE PLATE — all eight leaves, cols 0106D–0120D. archive.org patrologiaecurs04migngoog, PDF page = (column + 11) / 2, corner numbers read first on every page: pp. 58 (105/106), 59 (107/108), 60 (109/110), 61 (111/112), 62 (113/114), 63 (115/116), 64 (117/118), 65 (119/120). Reads are recorded in data/briefs/8950-PLATE-READS-0012.json for the coordinator to merge — data/plate-reads.json was NOT written, and was verified byte-identical afterwards. plate-gate.mjs 8950 returns gate OK against a simulated merge.

Apparatus fired: 16 [var:] · 3 [ed:] · 0 [sic:] · 0 [cj:] · 0 [d:] · 0 [nt:]. ⛔ Zero [sic:], and that is a finding, not caution. Every defect this range contains is Corpus Corporum's, not Migne's — nine sites where the digitization silently corrected the plate — so a [sic:] would fail verify-english's verbatim-substring guard and would publicly accuse Migne of the transcription's error. Five [sic:] are OWED after the patch pass and are named in data/briefs/8950-PENDING-TEI-PATCHES-0012.md. This is the same result the 0008–0011 stint reached over its own range, independently.

Per-chunk parity (English = Latin in every column):

Em-dashes: ZERO in all seven chunks (four written in draft, all four removed for a comma).

⚑ Ten places where the English now renders MIGNE'S TYPE rather than our twin's, under translation-style.md § "What the English translates — THE PLATE WHERE READ": 0108B VERS. 20-22.--, 0110B non dicit and the dropped et, 0113A Japhet ×2, 0115D the comma inside as far as, 0117B the head CHAPTER XIII., 0118B a full stop for a comma, 0119B a comma for a full stop and a full stop for a comma four lines apart, 0119B the two tag periods, 0120D a semicolon inside a quotation. Every one is filed as a pending patch, so the Latin converges on the English rather than the two drifting.

§0 · SPAN COUNT AGAINST THE BRIEF — 245 against 245, EXACT

Counted mechanically over the seven chunk bodies (frontmatter excluded): guillemet spans matched first with a non-greedy «[\s\S]?», then \[^\n]+\ over what remained, so an italic span nested inside a guillemet span is not double-counted — the brief's own convention.

245 = the 245 printed in data/briefs/8950-lemmata-0012.txt. Every inventory line was matched to a span in my own Latin and every span to a line; there is no residue in either direction, and the seven spans the rebuilt harvester recovered in this range (0108C ×2, 0108D ×2, 0115D, 0116C ×2) are all present and all collated below. The mis-split shape reported on 9001 is not present here.

⚠ The raw marks are unbalanced — 159 « against 161 » — and that is Migne, not a count error (§3). One extra open at 0013 and three extra closes at 0015/0016; the pairing regex absorbs them, which is why the span total still reconciles.

§1 · CONVENTIONS — inherited unchanged, and the three this stint had to settle

Ruth (8968 §1) and Tobit (8970 §1) govern and were not re-opened, nor were the three earlier stints of this same work (cruces-0000.md, cruces-0004.md, cruces-0008.md), which I read before drafting: VERS. n.-- passes through verbatim; the lemma is Englished; a fragmentary lemma is rendered as the phrase it stands for; sigla ride as ordinary [n: …] notes with Migne's own internal spacing, contents unanglicized; the guillemets are the plate's, 1:1, and their contents are Englished; usque ad → as far as, with no verb and no object ever supplied; the Historice / Allegorice / Mystice tags are translated inside their italic span (stint 3 §1a) — Mystice. → Mystically., Allegorice. → Allegorically., [n: (STRAB.)] Historice. → [n: (STRAB.)] Historically. with the note kept inside the span, exactly where Migne sets it (0110B).

(a) ⭐ NEW — an italic Latin word: MEANING is Englished, a word MENTIONED AS A WORD is not

This range is the etymological heart of the book (Genesis 10–14 is a name-list), and it carries 53 italic spans that are Latin rather than Hebrew. They split into two classes that must be treated oppositely, and getting it wrong destroys an argument either way:

1. The Latin gloss that gives the MEANING of a Hebrew name is Englished inside its span. Bethel … domus Dei, → the house of God, · Heber, transitor, … transitores. → a passer-over, … passers-over (and the next sentence, qui … transeunt, turns on it) · declinans → declining · descensum, → a going down, · rex justitiae, … rex pacis, → king of justice, … king of peace, · and the whole Genesis 14 chain at 0118C–0119C (excussio dentium, fetor eorum, Decorus manipulus, Sciens jugum, in angustia constitutum,, cinguli maeroris,, praecipitatio, absorbitio: …). Leaving these Latin would leave an English reader the name and not the interpretation, which is the only thing the paragraph says.
2. ⛔ The Latin word MENTIONED as a word — where the argument is about the form itself — stays Latin. Two sites, both cross-linguistic: - 0115B in Latino sidera non ferrea significat, sicut in Graeco, sed astra. The whole point is that one letter-sequence means different things in the two languages; Englishing sidera would make the sentence say that "stars" does not signify "iron" but "stars". The transliteration in raudo sidera. is likewise carried untouched, as is the Greek ἐν ραβδῳ σιδηρᾶ. - 0118B quod in quibusdam codicibus, mare pro occidente ponitur. A statement about what stands in certain codices, not about what a name means. Hebrew and Greek forms are never touched at all: heth, χ chi, Cham, Ham,, samech,, schin, scheba, seba., tau ×2, Salissam, Zoara, in Ur Chesdim.

⚑ The test, stated so a later stint can apply it without re-deriving it: is the italic word the TRANSLATION of something, or the SUBJECT of the sentence? A translation is Englished; a subject is quoted. This is the same line Pattern 6's corollary draws for quotation marks and the same line stint 1 §1(b) drew for sementis/semen/seminium — that stint had no italics in its twin and used our own straight quotes; here the italics exist, so the span carries the distinction instead.

⚠ Two spans resisted the rule and are recorded as such, both because the Latin gloss contains an untranslatable Hebrew element:

• 0118C civitas, arbe quatuor. — CHARIAT enim Hebraice civitas, arbe quatuor. Rendered a city, arbe four.: civitas and quatuor are the Latin meanings, arbe is the Hebrew word being glossed, and all three sit in ONE italic span that cannot be split without breaking span parity. The English therefore mixes the two classes inside one span. Its twin four words earlier, civitas quatuor; → the city of four;, is a pure meaning-gloss and took the ordinary rule.
• 0119D transfluviali; → one from beyond the river; — the Seventy's rendering of Hebraeus (περάτης). Englished as a meaning, because the next clause (transito fluvio Euphrate) is the argument for it.

(b) usque ad — 16 formulas, 7 ORDINARY. Decided per occurrence by the etc. splice test

Formula (italic, immediately after an etc.), 16×, all → as far as, no verb, no object and no antecedent supplied anywhere: 0107A · 0107B · 0108C · 0110B · 0111B · 0112C · 0113A · 0114D · 0115C · 0115D · 0116A ×2 · 0116C · 0116D · 0119D · 0120D. Fifteen are usque ad; the sixteenth is 0115D's usque,, discussed below. ⚠ Two of the fifteen land mid-clause and are left mid-clause: 0112C aversi operuerunt, etc., usque ad « tanquam in pacem anguli concurrentes. » — the resumption is itself inside guillemets and has no finite verb; and 0113A etc., usque ad id est opus eorum — a bare apposition.

⛔ Seven ORDINARY, non-formula usque ad, none italic, none after an etc., each rendered by sense (brief addendum D):

⭐ 0115D usque, is a VARIANT OF THE FORMULA and is listed above only because it lacks ad. « Tulit itaque Thare Abram, » etc., usque, facti sunt dies Thare ducentorum quinque annorum, et mortuus est in Haran. » — it follows an etc., it names the point at which the abridgment resumes, and the resumption is Gn 11:32 verbatim. Rendered as far as,, with the comma inside the span as Migne sets it. Precedent: stint 2's usque ad haec at 0091D, ruled the formula on the same grounds. Said here explicitly, per the series brief's requirement that an exception name the Latin that forced it. ⚠ 0118C is the 8967 col 0953C shape and is the one that must not be swept: Ps 71:8 prints usque ad as part of the verse, four lines below a formula at 0116D. Rendered "from sea to sea".

(c) The vocabulary axes — every instance, with the instrument checked before reporting

• gentes / gentilis. 25 occurrences (instrument checked: gentesimo, indigentia and ducentorum are excluded, and every remaining hit is listed). Gentiles where the Judaei/gentes opposition is live: 0111B in vocatione gentium ad fidem (Peter's dish); 0112A « Judaeis scandalum, gentibus stultitiam … Judaeis et gentibus, » glossed quasi Sem et Japheth; 0112D–0113A in populo gentium orbem terrae occupavit Ecclesia; 0113A de Japheth gentiles (the gentilis → Gentile rule, added on 9000); 0116C non tantum gentis Judaicae, sed omnium gentium fidem suam sequentium; 0118D stultitiam gentilium; 0120C Ecclesiam gentium. nations where the sense is plainly generic: 0113B nomina locis et gentibus relinquentes; 0114A caeterae sex gentes nobis ignotae and deletae sunt gentes bello Aethiopico; 0114B hae gentes a fluvio Cono; 0115B diversis gentibus divisit; 0117D inter gentes Chananaei et Pheresaei; 0118C haereditas omnium gentium and Ps 2:8's dabo tibi gentes; 0118D rex gentium (Gn 14:1's own words); 0119D nomina locorum et gentium; 0118A gente, sicut omnes Judaei → by nation, the axis's singular-gens-of-the-Jewish-people rule; 0112D Deus omnium gentium and in ipsis gentibus.
• ⚠ The one place the axis had to be applied twice in one sentence, and it is deliberate: 0116C sets the exposition non tantum gentis Judaicae, sed omnium gentium against its own proof-text « Et benedicentur in te omnes gentes terrae. » The exposition is the Judaei/gentes opposition → Gentiles; the verse is the axis's own generic exemplar shape (omnes gentes, Mt 28:19) and is universal, covering the Jews it has just named → nations. Singular gens Judaica → the Jewish nation, per the axis. Recorded because it reads as an inconsistency and is not one.
• natio — ZERO in this range. ⚑ Instrument checked before reporting, per the series brief: a raw natio grep returns five hits and every one is damnatione (0112D), dominatio (0115B), cognationem (0116B), cognatione (0116B) or cognatione (0118A). There is no natio in the range at all.
• turba — ZERO. Every turb- is absent entirely; there is no verb turbare here either. multitudo ×2, both "multitude" (0113A lata est multitudo credentium, 0120B non in multitudine exercitus).
• daemonium — ZERO. diabolus ×2 → devil (0113C, 0119C). No satanas.
• sacramentum ×7 → sacrament (0107D, 0108D, 0109B, 0110A, 0110C, 0110D, 0120A); mysterium ×1 → mystery (0110A nullis egeamus mysteriis). The 1:1 pair is exercised in adjacent columns and was kept 1:1.
• The commentator's own voice is not archaised: dicit/ait/aiunt → "says / say", never saith. -eth/-est forms appear only inside quoted scripture (Gal 5:17 concupiscit → "lusteth", 1 Cor 3:7 plantat/rigat → "planteth/watereth", Job 5:2 interficit/occidit → "killeth/slayeth") and in second-person singular address.
• Pattern 17. Singular → thou: 0110C's whole Mystice paragraph (addas · adjicias · pervenisti · egeas · invenies → "thou add … thou hast come … thou needest … thou shalt find"); 0115D « Ego sum qui eduxi te de Ur Chaldaeorum. »; 0116D « Dic ergo, obsecro, »; 0117B quomodo potuit dives esse valde is third person and untouched; 0118A si attenuatus fuerit frater tuus et vendiderit se tibi; 0118A–B « Leva oculos, » and quantum vides; 0118C « Tibi dabo. » and Ps 2:8 « Postula a me, et dabo tibi … ». Plural → you: 0111B « Et terror vester, »; 0119D Jos 24 « Quid vultis servire diis patrum vestrorum? ».

(d) Heads — numerals mirror the plate, INCLUDING a missing period

CAPUT VII. → CHAPTER VII. · CAPUT VIII. → CHAPTER VIII. · CAPUT IX. → CHAPTER IX. · CAPUT X. → CHAPTER X. · CAPUT XI. → CHAPTER XI. · CAPUT XII. → CHAPTER XII. · CAPUT XIV. → CHAPTER XIV. — numerals throughout, per the corpus measurement. ⚠ Chunk 0017's head is ## CAPUT XIII with NO terminal period, and its heads: frontmatter field carries the same defect ("CAPUT XIII"). The English head is ## CHAPTER XIII, also with no period, and the frontmatter is copied verbatim. This is the opposite call from Tobit's CAPUT XIV case (8970 §1e), where the plate was READ and shown to carry the period, so the English head restored it while the frontmatter kept the TEI's reading. Here the plate is unread, so supplying a period would be a silent repair of a head I have no witness for. Filed in §5 as a plate debt: PDF p. 64 settles it.

(e) Frontmatter copied VERBATIM from each Latin chunk, field order included

Every field, in the Latin twin's own order, including words: figures that count the Latin. Chunk 0012's colContext: "0106D" against colFirst: "0107A" is the chunker's, and is copied.

§2 · THE THREE [ed:] — Corpus Corporum has dropped MIGNE'S HEBREW at four sites

Migne sets Hebrew square script in this book wherever Jerome is quoted on a name, and our transform has lost every occurrence in my range. All four were read at 1200 dpi.

Three take an [ed:]; the fourth does not, and the line is Pattern 13a's own. Clause 2 gives the marker to a loss that would otherwise be invisible as an absence; clause 1 leaves an ordinary mid-sentence loss to a silent patch and a crux. So:

• 0113C, 0115C and 0118C are MARKED — each leaves a visible scar in our own text that misleads a reader who has only the two columns: a correlative that is not one, an empty bracket, and a fused italic span. ⭐ 0118C's scar is the one I had already flagged as a convention problem in §1(a) before I read the plate — "two spans that resisted the rule" — and the plate shows the fusion is the digitization's, not Migne's oddity. That is the crux file catching something the Latin alone could not.
• 0117A is NOT marked — the loss leaves the sentence sound and says nothing false. Crux and patch only.

⛔ None of the four is a [sic:] or a [var:]. They are Pattern 13 source loss: the plate is perfectly legible and it is our transcription that has the hole. Pattern 13's rule that an [ed:] never supplies the lost text is kept — the recovered Hebrew is recorded here and in the patch file, not in the running English.

⚑ Stint 3 recorded the identical class at 0104C ("two Hebrew words dropped by CC"), so this is a known, recurring behaviour in this book and not a local accident. Eight Hebrew words are now known lost across cols 0104–0118.

§3 · THE GUILLEMETS — the plate settles all four unpaired marks, and it splits them 2–2

Before the plate pass I reported four unpaired marks and called them all Migne's. That was half right, and the half that was wrong is the more interesting half.

⭐ So the class is not "Corpus Corporum drops Migne's repeated «." It is "Corpus Corporum normalizes the quotation marks in whichever direction makes the pair balance" — dropping an opener at 0106D, 0107C, 0115C and 0115D, and supplying a closer at 0113C. That is a stronger and more worrying statement than the one the earlier stints could make, and it means a balanced pair in our Latin is not evidence that Migne's pair was balanced.

⛔ Nothing was repaired. Guillemet parity is checked against the twin, so the English mirrors the twin at all four sites and the corrections belong in the patch file, where all six dropped line-head repeats in this range are listed (0106D ×2, 0107C, 0113C, 0115C, 0115D).

§4 · ⚠⚠ 7a″ — THE VULGATE COLLATION. This is the deliverable.

All 158 guillemet spans and the scriptural italic spans were collated against sources/vulgate/clementine-flat.txt, line by line, cited by book chapter:verse — never from memory, and never from the brief's pre-scan, whose ✓ addresses were read and in four places found to point at the wrong book (Ps 119:7 filed under 1Par; Gn 11:8 under 1Par 24:3; Ps 71:8 and Ps 2:8 correctly). The spans that AGREE are listed too (§4d): a list of divergences alone cannot be told apart from a report by an agent that only chased the ⚠ marks.

4a. ⭐ THE PLATE WAS READ, SO THE MARKERS COULD BE FIRED — 16 of the 26

All eight leaves were read before any marker was placed, and every one of the 26 divergences was confirmed as MIGNE'S TYPE at 200–1200 dpi. Not one turned out to be the digitization manufacturing a Vulgate divergence — the class plate-gate.mjs exists to catch, and the class 9003's collation showed a Vulgate collation can never catch by itself.

Sixteen took a [var:]; ten are recorded and declined. The line I drew, stated so it can be checked rather than admired — fire where the divergent word is (a) a different lexeme, (b) a different referent or grammatical person, or (c) a word the gloss itself argues from; decline where the difference is orthography, a clipped fragment boundary, a synonymous connective, or word order alone. Pattern 14's bar is "in a way that changes the sense", and a marker on word order is noise that buries the fourteen that carry information.

⚑ The ten declines are in §4b's own table, flagged DECLINED, with the reason on each row — not hidden in a separate list, because a reader comparing the two needs them side by side.

4b. DIVERGENT — 26 sites, ALL plate-confirmed as Migne's. 16 FIRED · 10 DECLINED.

4c. DIVERGENT AND DECLINED ON THE MERITS — Migne labels the version himself

Stint 3's ruling (§1b of cruces-0008.md) governs: [var:] fires where the words are offered AS the text and diverge; where Migne announces the version, its divergence IS the sentence's subject and a marker would tell the reader twice. Five sites, and this range would have carried five extra markers without the rule:

• 0108B « flatum vitae. » — introduced sed quidam codices habent melius. Migne's own variant-report.
• 0113D « quia ipse coepit esse gigas super terram. » against Gn 10:8 potens in terra — introduced vel « gigas, » secundum alios, and the next sentence argues with it.
• 0117A–B « Abiit unde venerat in desertum usque Bethel, » — introduced ut in LXX legitur, and expressly contrasted with the next.
• 0117B « Abiit in itinere suo per austrum usque Bethel. » — introduced secundum Hebraeum. Gn 13:3 reads Reversusque est per iter quo venerat a meridie in Bethel; the whole paragraph is Jerome adjudicating between the two, so the divergence is the text.
• 0117A gravis vehementer: for Gn 13:2 dives valde — introduced sed in Hebraica veritate habetur, and the gloss then argues from it (Aegypti enim pondere gravabatur).
• 0108D [Al., unum] — Migne's own bracketed variant, reproduced verbatim and untranslated per the locked register. It offers Gn 1:9's unum against the text's suum, and it is his conjecture layer, not ours (Pattern 19's instinct).

4d. COLLATED AND AGREEING — checked, exact or a clean whole-word fragment

Recorded so this can be told from a report by an agent who only read the ⚠ lines.

Gn 7 — « Dixitque Dominus, » (7:1) · « Ex omnibus animantibus, » (7:2) · « Ego pluam, » (7:4) · « Fecit ergo Noe omnia, » (7:5) · « Eratque sexcentorum annorum, » (7:6) · « Ingressus est Noe et filii ejus, uxor, » and « Ingressus est Noe et filii ejus, » and « uxor ejus et uxores filiorum ejus. » (7:7 — the lemma opens mid-verse, dropping only the Clementine's initial Et) · « Cumque transissent, » (7:10) · « Anno sexcentesimo, » (7:11) · « Cataractae. » (7:11) · « Bina et bina, » and « In qua erat spiritus vitae, » (7:15) · « Quadraginta diebus, » (7:17) · « Quindecim cubitis altior, » and « Quindecim. » (7:20) · « Deleta sunt de terra. » (7:23). Gn 8 — « Adduxit spiritum super terram, » (8:1) · « centum quinquaginta dies » (8:3) · « Requievitque arca mense septimo, » (8:4) · « Decimo enim mense. » (8:5) · « Qui egrediebatur, » (8:7) · « Emisit quoque. » (8:8) · « Reversa est. » (8:9) · « Exspectatis, » (8:10 — Expectatis is orthography, not divergence) · « Portans ramum. » (8:11) · « Exspectavitque nihilominus septem alios dies, » (8:12, same orthography) · « Sexcentesimo primo anno, » (8:13) · « Mense secundo. » and « Mense secundo, » (8:14) · « Egredere de arca, » (8:16) · « Et tollens, » (8:20) · « Sensus enim, » (8:21). Gn 9 — « Crescite et multiplicamini. » (9:1) · « Et terror vester, » (9:2) · « Carnem cum sanguine. » (9:4) · « Ecce ego statuam pactum, » (9:9) · « Arcum meum ponam, » (9:13) · « Bibensque vinum, » and « In tabernaculo suo, » (9:21) · « Quod cum vidisset, » (9:22) · « At vero Sem, » (9:23) · « Benedictus Dominus Deus Sem. » (9:26) · « Dilatet Deus Japheth. » ×3 and « Et habitet in tabernaculis Sem, » (9:27) · « Chanaan. ». Gn 10 — « Hae sunt generationes, » (10:1) · « Gomer. » (10:2) · « Thubal. » (10:2) · « Chanaan. » (10:6) · « Saba. » (10:7) · « Nemrod. » and « Ipse coepit esse potens, » and « gigas, » (10:8) · « Robustus venator. » (10:9) · « Babylon. » (10:10) · « Assur. » ×2 (10:11, 10:22) · « At vero Mesraim, » (10:13) · « De Sem. » (10:21) · « Heber. » (10:24) · « Natique sunt Heber filii, » (10:25). Gn 11–14 — « Erat autem terra labii unius, » (11:1) · « Cumque proficiscerentur. » (11:2) · « Venite. » (11:3) · « Venite igitur, » (11:7) · « et divisit eos, » and « Divisit eos. » (11:8) · « quia tulit eum Deus. » (5:24, exact) · « In Ur Chaldaeorum. » (11:28) · « Tulit itaque Thare Abram, » (11:31) · facti sunt dies Thare ducentorum quinque annorum, et mortuus est in Haran. » (11:32, exact, including the terminal clause) · « Egredere de terra, » and « de domo patris, » and « Et veni in terram quam monstrabo tibi. » (12:1) · Faciamque te in gentem magnam, » (12:2) · « Septuaginta quinque annorum. » (12:4) · « Bethel. » · « Hai. » · « Dic ergo, obsecro, » (12:13, dropping only te) · « Sublata est mulier. » (12:15) · « Ascendit ergo Abram, » (13:1) · « In Bethel, » (13:3) · « Erat quippe substantia eorum multa, » (13:6) · « Ne quaeso sit, » and « Fratres enim sumus. » (13:8) · « Sicut paradisus, » (13:10) · « Coram Domino. » (13:13) · « Leva oculos, » (13:14) · « Omnem terram, » and « Tibi dabo. » (13:15) · « Surge et perambula terram, » (13:17) · « In Hebron. » (13:18) · « Amraphel. » ×2 · « Arioch. » · « Chodorlahomor. » · « Thadal. » (14:1) · « Balae. » (14:2) · « Duodecim, » (14:4) · « Percusseruntque. » (14:5) · « Hebraeo. » (14:13) · « Quod cum audisset Abram, » (14:14) · « At vero Melchisedech, » (14:18) · « Dedit ei decimas. » (14:20) · « Qui venerunt, » (14:24). Non-Genesis proof-texts that agree exactly — Gn 1:2 « Spiritus Dei ferebatur super [0108D] aquas: » (the anchor falls inside the quotation and is reproduced inside it) · Gn 1:26 « Faciamus hominem ad imaginem et similitudinem nostram. » · Ps 106:25 « Et stetit spiritus procellae, » · Ps 118:104 « A mandatis tuis intellexi. » · Lc 10:21 « Abscondisti haec a sapientibus et prudentibus, et revelasti ea parvulis. » (and Migne's Luc X is right) · Ps 119:7 « Cum his qui oderunt pacem eram pacificus: [0109D] cum loquebar illis, » · 1Cor 3:7 « Neque qui plantat est aliquid, neque qui rigat, sed qui incrementum dat Deus » · Ps 2:9 « In virga ferrea, » · Ps 2:8 « Postula a me, et dabo tibi gentes haereditatem tuam » · Ps 71:8 « Et dominabitur a mari usque ad mare. » · Ps 109:4 / Hbr 7:17 « Tu es sacerdos in aeternum secundum ordinem Melchisedech, » · Hbr 7:2 rex justitiae, and rex pacis, · Job 5:2 « Stultum interficit iracundia, parvulum occidit invidia. » (dropping only Vere and et, and Migne's Job V is right).

⚠ coelum/coeli for caelum, Sennaar for Senaar, Hierusalem for Jerusalem and quaeso/caeteri spellings are Migne's orthography, not divergences, and are not listed as such.

4e. TWO CITATION SIGLA THAT POINT ELSEWHERE — reproduced verbatim, uncorrected

• 0109C [n: (Psal. XIX)] on « Cum his qui oderunt pacem eram pacificus … ». The verse is Ps 119:7, and the Clementine agrees with Migne's words exactly. XIX for CXIX is a dropped C. Notes pass through verbatim (series brief convention 4) and this is not the English's to fix; the correction belongs in data/citation-corrections.json at index time.
• 0119C [n: (I Cor. XV)] on « Comedamus et bibamus, cras enim moriemur ». 1Cor 15:32 reads Manducemus et bibamus; Is 22:13 reads comedamus et bibamus, cras enim moriemur — Migne's words exactly. So the words are Isaiah's under a Corinthians siglum (Paul is himself quoting Isaiah there). ⚑ This is the reason a bare ✓ is not clearance: the brief's own pre-scan matched it to Is 22:13, which is the mark telling you the siglum names a different verse. Reproduced as printed; no [var:], because against the address Migne's own words are the ones he printed and there is no divergence to report** — the anomaly is in the siglum, not the text.

§5 · THE DECLINED LIST — suspect readings, what would settle each, and the leaf

PDF page = (column + 11) / 2 on raw/scans/pl113/patrologiaecurs04migngoog.pdf. Cols 0106–0120 are pp. 58–65 and every one is unread. Where the archive OCR was consulted the result is given, and it is not a plate read: it licenses no marker on its own.

5.1 ⭐⭐ 0110B — SETTLED AT THE PLATE. All three words go against our twin.

Migne's Historice on Gn 8:14 reads, in our twin:

Id est, Maio, Noe annum fecit in arca, et ipsa die qua ingressus fuerat, egressus est. Ingressus es enim vicesima septima die mensis secundi; num dicit quia egressus est septima et vicesima die ejusdem mensis.

The plate (p. 60, col 110, band D, read at 1200 dpi, corner numbers checked first) reads:

Id est, Maio, Noe annum fecit in arca, ipsa die qua ingressus fuerat, egressus est. Ingressus est enim vicesima septima die mensis secundi; non dicit quia egressus est septima et vicesima die ejusdem mensis.

1. ⛔ Ingressus es is OURS. The t is unmistakable. A second-person singular stranded in a narrative about Noah is Pattern 12's test-1 shape and would have earned a [sic:] on the Latin alone — exactly the false public claim against Migne the ship test exists to stop. The English already rendered est; no marker, patch owed.
2. ⛔ num is OURS too, and the plate reads non. Three clear letters, a round o, no m and no c — it is neither our twin's num nor the nunc the sense had tempted me toward. ⭐ This is the entry I am most glad I declined. Before the read I had rendered num as an interrogative particle without supplying a question mark, producing a sentence that read as broken and that I flagged as the stint's headline crux. Had I "repaired" it to the nunc the argument seemed to want, I would have been wrong, and invisibly so. The English now reads non dicit — "it does not say that he went out on the seventh and twentieth day of the same month" — which is what Migne printed, and the lunar computation that follows is what supplies the exit date he says Scripture does not give.
3. The twin's et before ipsa die is OURS — the plate has none. English corrected.
4. ⚠ vicesima septima as the date of ENTRY is MIGNE'S. It contradicts Gn 7:11 (septimodecimo die mensis) and it contradicts this paragraph's own next sentence (decima septima … dies mensis secundi fuit). Rendered as printed, and it is his contradiction, not ours. It takes no [cj:]: a numeral slip is neither a negation nor a correlative, which is Pattern 18's declining class, and the sentence asserts nothing false about the argument once the lunar reckoning is read.

5.2 Real words that read oddly — checked at the plate. Five stand as Migne's; two did not.

Per addendum A: a real, well-formed word that reads oddly is Migne's until a plate says otherwise, and the corpus argument loses even when it is good.

• 0110B Vitae Noe, id est, peractis sexcentis annis — a genitive with no governing noun, the abridgment having cut the head of Bede's sentence. Rendered "Of Noah's life, that is, six hundred years being completed"; this is the formula working, not a defect.
• 0110D Juncti exeunt, qui disjuncti intraverunt — the OCR reads intraverant (pluperfect). Both are real words and the sense is unaffected. Rendered intraverunt ("who entered"), no marker. Recorded because it is a genuine twin/plate disagreement candidate that changes nothing; p. 60.
• ⭐⭐ 0112B — the plate found a defect TWO WORDS to the left of the one I raised. I had flagged debebat for having no infinitive and rendered it as printed. The plate reads ridiculum ei visum ET FACIMUS cui minime debebat, where our twin has visum est facinus — Corpus Corporum has repaired two words at once, and its repair is good Latin that reads so naturally nothing in the Latin column could ever have raised it. Migne's line does not construe: facimus is a first-person verb among third-person narrative about Cham, and visum et leaves the clause with no copula. [sic: et facimus] is OWED and cannot be fired until the patch lands, because the marker's content must be verbatim in the twin. debebat itself is Migne's and stands. ⚑ This is addendum C exactly — "read the whole line, not the target word" — and it is the strongest argument in this stint for reading plates at all: the site I flagged was sound, and the site four characters away was not. p. 61.
• 0113A Quia igitur lata est multitudo credentium, a latitudine, quae Japheth dicitur, latitudo nominatur — the etymology names latitudo twice and says nothing new the second time. Real words; rendered literally. p. 62.
• 0117D Pastores ejus, haereticos et schismaticos, qui … ascendunt ad Dominum, pacem et unitatem continentes; — an accusative in apposition with no verb to govern it (significant has to be carried over from the previous sentence), and continentes is a nominative plural that can attach neither to Dominum (accusative singular) nor coherently to the heretics, of whom the clause has just denied peace and unity. Rendered in Migne's order with the participle left dangling exactly where he leaves it; reordering it to disambiguate would be a silent repair of his syntax. p. 64.
• ⛔ 0119B — REFUTED BY THE PLATE, and it is the one entry on this list that the read reversed. Our twin prints in mare Mortuum versa est. quod a Graecis stagnum Asphalti … dicitur, and I logged the full stop before a lower-case relative as Migne's under addendum A's rule that a real mark reading oddly is his until a plate says otherwise. The plate says otherwise: Migne prints a COMMA. Our twin introduced the sentence break. English corrected to a comma. ⭐ And the same class runs the other way four lines below: the plate prints superant primum Bara. qui interpretatur with a FULL STOP where our twin has a comma. Corpus Corporum swapped both marks, in opposite directions, within four lines — which no totals-based punctuation check could ever see, and which is why Test 3 compares band by band. Both now follow the plate. p. 65.
• 0118C Ibi enim situs est Adam maximus et Abram — Jos 14:15's Adam maximus ibi inter Enacim situs est, a proper name in Migne's copy where modern texts read a common noun. Rendered "Adam the greatest", as printed. p. 64.

5.3 The head at 0117B, and the address anomalies — all reproduced, none tidied

• ✅ ## CAPUT XIII — SETTLED. The plate prints CAPUT XIII. WITH the terminal period (p. 64, read at 1200 dpi), so the omission is our twin's and its heads: frontmatter carries the same defect. The English head is now ## CHAPTER XIII., and the frontmatter is left exactly as copied, because "copy the Latin chunk's frontmatter VERBATIM" is unconditional and the patch will fix both sides at once. This is Tobit 8970 §1e applied, not re-derived — and it is the call I declined to make before the read, correctly, because an unread plate could not license it.
• 0108B VERS. 20-22-- (chunk 0012) — no terminal period and the double hyphen unspaced from the numeral, against VERS. n.-- in the other forty-one addresses of this stint. Reproduced verbatim.
• 0116B VERS. 2-3.--Faciamque — no space after .--, and the lemma's opening « missing (§3.4). Reproduced verbatim.
• 0113C and 0113D both print VERS. 8. — the first on « Regma et Saba, », which is Gn 10:7 (Filii Regma : Saba et Dadan), the second on « Nemrod. », which is Gn 10:8. Migne's own ordering and his own address; reproduced, never renumbered (series brief: a verse number out of sequence is his). And 0113D prints VERS. 10, 12. before VERS. 11. — the same class, four lines later. p. 62.

5.4 Unexpressed subjects supplied — named, with what decided each

• 0108D tunc ferebatur, ut, congregatis in suum locum, terra appareat — the ablative absolute has no noun. Supplied "the waters". Decided by the verse being expounded, not by proximity: Gn 1:9 Congregentur aquae … in locum unum, et appareat arida, which is also what Migne's own bracketed [Al., unum] is pointing at. The nearer candidate, spiritus, cannot be gathered into a place and is the subject of ferebatur in the same clause.
• 0108D nunc adducitur, ut … faciem terrae revelaret — supplied "it", the spirit, from adducitur in the same clause; the alternative (God) would require a change of subject the Latin does not mark.
• 0111C ne perdat eam diluvio — "it" = omnem animam vivam, the nearest and only feminine singular accusative in the sentence.
• 0112B quia forte nesciebat vino posse inebriari — "he", Noah, named in the same sentence.
• 0113B Ad quas misit Dominus septuaginta duos discipulos — "to which", the seventy-two generations, not the three sons: the number is the whole point of the sentence.
• 0114B quia ejus generationem, et generationis lineam prosequitur — "he", Moses the narrator, not Sem: Sem is ejus.
• 0116A Sed per recapitulationem ostenditur vivo patre Dominum praecepisse — the subject is printed; recorded only because the English word order has to move it.
• 0119A–C the whole Allegorice passage: superant · Vincunt · obruunt · hortantur · decipiunt · mittunt all have "they" = the four kings / the contrary powers, fixed by Quatuor reges quinque superant two sentences above and by Tales reges superant primum Bara. No other plural agent is in view.
• 0120C Ex quo vult colligere — "he", the Apostle, named as the subject of refert in the same paragraph; Melchisedech is its object.
• ⚠ One place where both readings construe and nothing was chosen: 0111C Illi enim non pereunt diluvio separati ab Ecclesia, qui … agnoscunt gloriam Christi. separati can be a predicate of illi ("they do not perish in the flood as men separated from the Church do") or a concession ("they, though separated from the Church, do not perish"), and the second reverses this book's own doctrine four columns earlier (0107C: extra Ecclesiam aqua baptismi … non valet ad salutem). Rendered in Migne's order with the participle left where he puts it — "For they perish not in the flood, separated from the Church, who … acknowledge the glory of Christ" — so the ambiguity survives into the English instead of being decided silently. p. 60.

5.5 Test 2a fired and checked — negatives on a verb with a quantifier or correlative nearby

Every site in the trigger population, with the host proved from the sentence's structure:

• 0108B non solum de hominibus, sed et de pecoribus dictum — non solum … sed et; the host is de hominibus, not dictum. English: "is said not of men only, but of cattle also."
• 0106D–0107A Non geminatum septenarium … sed tantum septem and non quatuor, sed tantum duo — the sed tantum correlative fixes the host on the numeral both times.
• 0111B sciant se irrationabilibus, sed non rationabilibus dominari — sed non puts the negative on rationabilibus, not on dominari: they DO have dominion, but not over the rational. Rendered "they have dominion over the irrational, but not over the rational."
• 0110D « Non percutiam omnem animam viventem sicut feci » — a negative on a verb with omnem beside it, the exact 8990 @0611D shape. Left scope-ambiguous in English as it is in Latin ("I will not strike every living soul as I have done"), because the Latin does not disambiguate and the gloss's next clause (quae largitate bonitas donat indignis) does not turn on the scope. See §4b for the separate ultra divergence.
• 0118B Non hoc solum promissum est — host is hoc solum. "Not this only was promised."
• 0116C non tantum gentis Judaicae, sed omnium gentium — non tantum … sed, host on the genitive. ✓
• 0116D non solum in hoc quod non inquinatur corpus nec anima nisi mentis consensu, sed etiam in hoc — four negatives in one clause. Hosts: solum → "not only"; inquinatur + nec → "neither body nor soul is defiled"; nisi → "save by the consent of the mind". All four present, each on its own word.
• 0112C–D Non praetermittit Deus inulta peccata — a negative verb with a privative adjective; both carried ("does not pass over sins unavenged"), which is the reading the following sed ita judicium … incipit requires.
• 0111D ne homines formidolosi timerent altero diluvio deleri — host on timerent, and the purpose clause is the point of the rainbow. ✓
• 0115B ut non intelligeretur jubens homini, qui noluit intelligere ut obediret jubenti Deo — two negatives, and the second is the reason for the first. Host of the first is intelligeretur (the punishment IS unintelligibility), proved by the preceding dominatio imperantis in lingua est; host of the second is voluit. Rendered "so that he who commanded should not be understood by man, the man who would not understand so as to obey God who commanded."

No site in this range required a negative to be relocated, and none was.

5.6 Sections read clean (Test 4 — does the English ASSERT what he asserts?)

Run last, with all counts already known clean, paragraph against paragraph:

• 0106D–0108C — the seven-and-seven arithmetic, Origen's geometrical cubit, and the spiritus/flatus vitae variant. The densest numeral passage in the stint; every figure checked against the Latin digit by digit.
• 0108C–0110A — Augustine on recapitulation, the raven and the dove. The aut … aut and an … an disjunctions were checked for completeness; both limbs survive in each.
• 0110B–0111A — the lunar computation (§5.1), the Mystice on seventeen and twenty-seven, and Gregory on the Moralia. The arithmetic (si decem et septem decem addas, fiunt viginti septem) is internally consistent in the English.
• 0111A–0113A — the rainbow, Cham's sin, and the blessing of Japheth. Procopius's catalogue of aggravations at 0112B is the one place where colons carry the argument and they were matched band-for-band.
• 0113B–0114B — the table of nations. Almost pure onomastics; each name checked against Gn 10 for spelling as Migne prints it.
• 0114C–0117A — Babel, the sidera/astra argument, Thare's years, and Abram in Egypt. The hardest single sentence is 0115B's cross-linguistic pun (§1a).
• 0117B–0118C — Jerome on gravis vehementer, the shepherds' quarrel, and the promise of the land. Read twice for §5.2's dangling continentes.
• 0118C–0120D — the four kings, the etymological chain, and Melchisedech. Every one of the twenty-three etymological glosses was checked against the name it interprets.

Candidates raised and rejected, so a blind reader knows they were looked at: everything in §5.2; plus abusive at 0107D (Cataractae … abusive pro omnibus fenestris ponuntur — a real adverb, "by misuse", and the standard grammarians' term; rendered, no crux); superfluus at 0119C (Duodenarius superfluus est numerus — the medieval arithmetical term for an abundant number, and twelve is the first one, so the sentence is technically exact and not a slip); Helia at 0117B (Aelia Capitolina, Migne's spelling, attested); Phutaei / Laabin / Nephetuim at 0114A (Migne's onomastic spellings, which vary from the Clementine's throughout this chapter and are its own class); and ternarius quadratus at 0107B (= nine, and the abridgment's tail says so).

§6 · PENDING TEI PATCHES — 23 confirmed sites, and the class is NORMALIZATION

Full table with plate readings: data/briefs/8950-PENDING-TEI-PATCHES-0012.md. The summary that matters here:

⭐⭐ Nine of the twenty-three are Corpus Corporum silently CORRECTING Migne — brief §4's class, and not one is detectable from the Latin alone, because every one reads perfectly well: uspue→usque (0108C), et facimus→est facinus (0112B, two words at once), es→est and num→non (0110B), debito→debita and utilitatum→utilitatem (0120D), Japhet→Japheth (0113A ×2), and two punctuation marks swapped in opposite directions four lines apart (0119B versa est,→versa est. and Bara.→Bara,).

⭐ The most consequential is 0112B. Migne prints ridiculum ei visum et facimus cui minime debebat, which does not construe — facimus is a first-person verb among third-person narrative about Cham, and visum et leaves the clause with no copula. Our twin prints visum est facinus, which is good Latin and reads so naturally that nothing in the Latin column could ever have raised it. I had logged debebat at this exact site as an addendum-A candidate ("a real word that reads oddly, rendered as printed"); the plate shows the real defect was two words to its left. That is addendum C's rule earning its place — read the whole line, not the target word.

⛔ Five [sic:] are OWED and none could be fired, because verify-english requires the marker's content to be a verbatim substring of the Latin twin and in every case the twin carries the repaired form: [sic: uspue] · [sic: et facimus] · [sic: debito] · [sic: utilitatum] — and the fifth is not a [sic:] at all but the es/num pair, which needs no marker once patched because est and non are what the English already prints. This is precisely the coordinator's prediction and the 0008–0011 stint's result, reproduced independently: where the plate confirms our twin has tidied Migne's defect away, what is owed is a patch, not a marker.

⚑ Also patched-not-marked: four Hebrew losses (§2), six dropped line-head guillemets and one supplied closer (§3), a dropped head period at 0117B, and five siglum-punctuation differences (0107B Ibid., 0110B (Ibid ), 0115C lib. iii., 0112A capital Χ, 0113B CAPUT ·X.).

Foot-of-page apparatus: CHECKED ZERO on all eight pages, both layers — no numbered notes and no asterisk layer anywhere in cols 0106D–0120D. Neither foot rule nor note appears; p. 58 carries only the signature 4 and p. 65 only a closing rule. ⚑ This is a real result and it is recorded as coverage, not as a skip. With stint 3's six pages it takes PL 113's checked-zero sample to fourteen pages across this book, and with Ruth's five and Tobit's four to twenty-three pages across four books — but per CLAUDE.md it licenses nothing about the columns nobody has opened, and it does not license the claim that this work has no notes.

§7 · PUNCTUATION — compared BAND BY BAND, and the only added marks are four Pattern 8b colons

Test 3 run mechanically over all seven chunks: markers stripped, both twins partitioned by their verified-1:1 column anchors, and colons, semicolons, question marks, «, » and ! counted per band. Exactly four bands differ, all in the same direction, all the same class, and zero elsewhere:

Migne's ? is kept in all four; English cannot end an indirect construction with it (Pattern 8b), and the colon carries the quaeritur frame. Question-mark parity is exact in every chunk (3 / 2 / 1 / 1 / 1 / 1 / 1).

⚠ The scan that produced this table also found 23 unlicensed marks in the draft, and the totals had very nearly cancelled band for band. All were repaired against the Latin:

• 7 supplied colons removed (Migne prints a comma): 0110A quando requies sanctorum erit, · 0110B Id est, Maio, · 0111C ne perdat eam diluvio, and per lapidem, · 0114C quod humanum genus perdit, · 0116B Duo promittuntur Abrae, · 0120C ministerium nostrum veri ordinis signatur,
• 5 destroyed colons restored: 0108D terra appareat: · 0109B immunditia cupiditatis teterrimos: · 0112B facinus detexisset omnibus: · 0117B euntibus Neapolim: · 0120B usque ad Isaac vixisse:
• 7 supplied semicolons removed: 0107C egressi sunt, · 0107D requiem significat, · 0112B Despectui habet parentem, and jacuerit pater, · 0115A Creatorem conspiciendo, · 0117A ascendisse legitur, · 0117C altare Domino aedificavit,
• 4 destroyed semicolons restored: 0107B quod pertinet ad dies; · 0114A Phutaei vocabantur; · 0117B separata sacerdotibus; · 0120C dicta est Hierusalem;

This is Test 3 earning its place again: the raw totals for the stint were within one of each other before the repair, and only the per-band comparison found the sites.

⚑ ONE band now differs from the twin on purpose, and it is a plate reading: 0120D shows semi 1/2, because Migne closes the Tit 1:12 quotation « Malae bestiae, pigri ventres ; » with a semicolon inside the guillemets and our twin has a comma (p. 65, 1200 dpi). The English follows the plate and the divergence is filed as a pending patch. Nine further English/twin punctuation divergences of the same kind are listed in §6; each is a mark Migne prints that our twin has altered, and each is filed. A blind reader comparing the columns will meet them and should not read them as supplied marks.

⛔ No comma→semicolon sweep was made. Migne prints semicolons freely in this range (31 in the seven chunks), so Tobit's narrow exception does not apply and was not invoked.

⚑ Three places where an English question was NOT given a mark, because Migne prints none: 0114D–0115C Magna quaestio nobis nascitur … quomodo Sem biennio post diluvium centum esse annorum dicatur (rendered as an indirect "how", no mark); 0116D Quaeritur cur patriarcha voluit mentiri, etc. (no mark, no colon); 0112C Quid est ergo quod filii verenda patris … operuerunt, nisi quia … occultent: — an interrogative sentence that Migne closes with a colon, reproduced with the colon and no question mark.

STINT 5 — chunks 0019–0023 (cols 0120D–0133C)

Filed as cruces-0019.md; merged verbatim, headings demoted one level. Nothing in it was rewritten — where a later stint corrected it, the correction stands in that stint's section.

Translated 2026-09-05, one stint, 5 chunks (5,245 Latin words by the frontmatter; 5,013 body words with anchors and notes stripped → 7,501 English, 1.50x). Bands: 0120D–0133C — CAPUT XV, XVI, XVII, XVIII, XVIII (cont.), XIX. verify-english.mjs 8950 reports no error on 0019–0023 (the only failures in the work are MISSING English chunk for ranges other stints hold).

SPAN COUNT — 189, of which 3 OPEN-ENDED. Counted paragraph-bounded in my own five Latin chunks: 115 closed « … » spans + 71 … spans + 3 unclosed « = 189, matching the rebuilt brief's header exactly. Per chunk: 0019 23+11+0 · 0020 22+38+0 · 0021 27+4+2 · 0022 5+1+1 · 0023 38+17+0. ⚠ The raw marks are 120 « / 115 », and the five-mark gap is real and accounted for: three unclosed quotations (0127B, 0127D, 0129D — all three plate-confirmed as Migne's, §3) and two nested line-head « inside closed spans (0121A, 0127D), each of which pairs into its host and is counted once. A bare «/» total is not a checkable number on this book. ⚑ The counting method matters: a DOTALL non-greedy «.*?» over the whole chunk body runs an unclosed quote forward across a paragraph break and swallows the next span. Counting paragraph-bounded is what keeps the two errors from cancelling.

Apparatus fired: 11 [var:] · 1 [cj:] · 0 [sic:] · 0 [d:] · 0 [ed:] · 0 [nt:]. Every one stands on a column read at the plate this session.

Per-chunk parity (English / Latin), all exact: column anchors 8/8 · 11/11 · 14/14 · 3/3 · 15/15 · [n:] notes 14/14 · 12/12 · 18/18 · 3/3 · 20/20 · « 24/22/30/6/38 · » 23/22/27/5/38 · question marks 0/2/4/2/2. EN/LA word ratio (apparatus stripped) 1.48 / 1.46 / 1.52 / 1.45 / 1.51. Zero em-dashes supplied in 5,245 words.

⭐ THE WHOLE RANGE WAS READ AT THE PLATE before the English was written: archive.org patrologiaecurs04migngoog, PDF page = (column + 11) / 2, corner numbers read first on every page — pp. 66 (121/122), 67 (123/124), 68 (125/126), 69 (127/128), 70 (129/130), 71 (131/132), 72 (133/134). Foot-of-page apparatus: CHECKED ZERO on all seven pages, both layers. Reads are recorded in data/briefs/8950-PLATE-READS-0019.json for the orchestrator to merge; twin/plate divergences are in data/briefs/8950-PENDING-TEI-PATCHES-0019.md. ⚠ Band 0120D (the two lemmata before the first anchor of chunk 0019) sits on p. 65 and was not read. Nothing fired there; both lemmata agree with the Clementine exactly.

§1 · CONVENTIONS — inherited, and what this stint had to settle

Ruth (8968 §1) and Tobit (8970 §1) govern and were not re-opened: VERS. n.-- passes through verbatim; the lemma is Englished; a fragmentary lemma is rendered as the phrase it stands for; sigla ride as ordinary [n: …] notes with Migne's own internal spacing ((ISID. in Gen. ), ( Glossa Graeca. ), (AUG. Quaest in Gen. ), (AUG., Q. in Gen.)); the guillemets are the plate's, 1:1, and their contents are Englished; usque ad → as far as, with no verb ever supplied.

The three conventions the earlier 8950 stints settled were adopted, not re-derived: (i) no [var:] on a quotation Migne himself labels a rival version — one worked instance here, §4.3; (ii) words follow the plate, punctuation follows the twin — eight instances, patch file Part A3; (iii) unpaired guillemets are mirrored, never tidied, and are usually Migne's — three here, all his, §3.

#### (a) The sense tags are translated in their italic span, and one is ABBREVIATED Tropologice. → Tropologically. (0121B) · Allegorice. → Allegorically. (0123C, 0131B) · Historice. → Historically. (0124C) · Moraliter. → Morally. (0126A, 0129A, 0131A) · Mystice. → Mystically. (0129A, 0131A, 0132A). A note inside the tag's italic span stays inside it ([n: (GREG., lib. XVI Moral., c. 20.)] Tropologice.). ⚑ 0131C prints the tag ABBREVIATED — Moral. — where the same tag is spelled out three times elsewhere in the same range. Rendered Morally.: English has no exponent for a Latin abbreviation, and the facing Latin column preserves it, which is exactly Pattern 9's argument for normalizing a section label. Flagged rather than decided silently, because it is the first abbreviated tag in this book.

#### (b) A gloss word that must NOT be Englished, and three that must

• « Tria sata. » → « Three sata. » (0126B, 0126C ×2), and Satum genus est mensurae → "A satum is a kind of measure." ⛔ Rendering sata as "measures" would have made the gloss read "A measure is a kind of measure" — the tautology trap the 0000 stint met at sementis/semen. The word is carried in roman inside the guillemets, because italicizing it would add an italic span the Latin twin does not have and break span parity.
• aedificatio → a building up (0122A). Jerome's point is that the Hebrew expresses bearing children by the verb "to be built up"; the word is Latin, not Hebrew, so it is translated.
• onager. → a wild ass. and rusticus, → a rustic, (0122B) — Latin gloss-words. The Hebrew beside them (phara) is kept.
• caecitas, → blindness, (0131B) and motus eorum: → their motion: (0123A) — Latin etymologies of Hebrew place-names, translated in their spans.
• ⚠ avidentia, (0130D) is CARRIED, untranslated, in its italics. It is Augustine's coinage for Greek ἀορασία and Migne flags it as one — si possit dici, "if it may be so called" — so the register's italic-Latin rule for a technical word with no clean English equivalent applies, and the flag would be unintelligible beside an ordinary English word. No [sic:]: the type is sound and the coinage is the author's.
• Hebrew letter-names stay Hebrew: scin · res · iod · he · zadich · ain · tau · schin (0124B), and the single letters h / a / r stay letters. centum → a hundred (0124A, the numeral value of Greek ρ, a Latin word being mentioned). princeps mea, → my princess, · princeps, / principem, → a princess, · lepram → leprosy, · sarai and Sara kept.
• ex patre, and de patre both → from the father (0133B, 0133C). English has one preposition for Migne's two; nothing in Jerome's etymology of Moab turns on the difference. Recorded rather than hidden.

#### (c) The foedus / pactum / testamentum triad is kept 1:1, because the gloss distinguishes them 0122D prints the distinction as its whole point — In Hebraeo berith quod non testamentum, sed foedus, sive pactum significat. So, fixed for the range: testamentum → testament · foedus → covenant · pactum → compact. Hence « I will establish my compact » (0124C, Gn 17:19) beside « And an everlasting covenant » in the next line — Migne's two words, not one English word twice. Same reason as the mysterium/sacramentum pair below.

#### (d) The vocabulary axes, with the instrument checked first

• gentes — 8 occurrences, split by the axis, and the split is recorded because it is visible on one page. Gentiles where the Judaei/gentes typology is explicit: 0124A vocatione gentium (Isaac the type of the Church against Ismael the type of the Jews) and 0128A omnes gentes per fidem, opposed in the same clause to gens Judaeorum secundum carnem. Nations where the sense is plainly generic, which is the axis's own named exception: 0122B omnes gentes quibus desertum … jungitur; 0123A pater multarum gentium; 0123C ne misceretur semen ejus gentibus; 0123D pater multarum gentium; 0124B omnium … gentium princeps futura. ⚠ The one place the split shows: 0128A. The lemma « in eo benedicentur omnes gentes terrae » is the generic verse formula and reads "all the nations of the earth"; the exposition four words later, omnes gentes per fidem against gens Judaeorum, is the typology and reads "all the Gentiles through faith". Both are the axis as written; the lemma and its re-quotation therefore differ, which is the one thing 7a″ normally forbids. Flagged here rather than decided silently — if the merge prefers one word for both, it is "nations" in both places, and the typology then lives only in gens Judaeorum → "the Jewish nation".
• Singular gens of a people → nation: 0121A gentem Christianam → "the Christian nation", gens Judaeorum → "the nation of the Jews"; 0128A gens Judaeorum → "the Jewish nation"; 0128A in gentem magnam futurus → "about to be a great nation".
• ⚑ Instrument checked before reporting. natio matches once in 5,245 words (0123B nationibus → "nations"); the other three hits are cognatione / incarnationem / peregrinationem. turb- fires nine times and every one is perturbare / perturbatio — there is no turba at all in this range, so the turba/multitudo pair is not exercised on the turba side. multitudo fires three times (0121A multitudine, 0122A prae multitudine, 0124D multitudo) and is multitude in all three.
• mysterium → mystery, sacramentum → sacrament, and the 1:1 pair IS exercised: 0126A mysterium futuri sacramenti → "the mystery of the sacrament to come".
• Zero daemonium, diabolus, satanas, gentilis, gentilitas, ethnicus in range.
• The commentator's own voice is not archaised (dicit/ait → "says"); -est/-eth forms appear only inside quoted scripture and in second-person singular address.
• Pattern 17. Singular → thou: 0121A cujus tu pater futurus es → "whose father thou art to be"; 0121C Scito praenoscens → "Know thou beforehand"; 0122C non apparebit tibi → "shall not appear to thee"; 0122D Nec ultra vocabitur nomen tuum → "Neither shall thy name any more be called"; 0131A Si vis perfectus esse … quae habes → "If thou wilt be perfect … that thou hast". Plural → you: 0123B ex vobis → "among you"; 0126D 1 Cor 3:2 Lac potum dedi vobis → "Milk I gave you to drink"; 0129D Videte si inveniatis → "See if you find". ⚑ Two plural imperatives are rendered as bare English imperatives, which mark no number: 0130A « Lavate. » → « Wash. » and 0132B « Mementote uxoris Lot, » → « Remember the wife of Lot, ». Wash ye / Be ye mindful were drafted and withdrawn — Pattern 17 gives the plural you, not ye, and an English imperative supplies no pronoun at all.

#### (e) usque ad — 10 formulas, 5 ordinary, decided per occurrence by the etc. splice test Formula (italic, following an etc., 10×): 0121A, 0121B, 0121C, 0121D ×2, 0122A, 0123C, 0127D, 0131C, 0132C. Every one is preceded by etc. and names the point at which the abridgment resumes; no verb and no object was supplied to any of them. Three tails land mid-clause and stay there: 0121C as far as "a disturbance not to be named" (the governing verb is inside the cut — non dicendam perturbationem rendered as the bare accusative phrase it is); 0121D as far as "and they arrive at Solomon"; 0131C as far as "whence it is said to Lot:", whose colon is Migne's and closes on nothing. ⚠ 0131C's formula prints as the non-word uque ad in our twin. The plate reads usque ad (600 dpi, p. 71), so the defect is OURS; the English renders the formula and the patch file carries the repair. No [sic:] — it would accuse a clean page. Ordinary Latin (5×), none italic, none after an etc., each rendered by sense: 0121B ab Adam usque ad Noe, et inde usque ad Abram, deinde usque ad David → "from Adam to Noah, and thence to Abram, and then to David" (three in one clause); 0128D a quo usque ad decem descendit → "from which he descends down to ten"; 0128D Usque ad decem justos non perire Sodomam dixit → "up to ten just men". ⚑ The two at 0128D were deliberately not given the formula's "as far as", to keep the abridgment mark distinguishable on a page where it occurs nowhere else. ⚠ Also present and NOT this word: 0122C Usque hodie → "To this day", and 0123B quantum pertinet → "so far as pertains".

#### (f) Heads and frontmatter CAPUT XV. → CHAPTER XV., CAPUT XVI. → CHAPTER XVI., CAPUT XVII. → CHAPTER XVII., CAPUT XVIII. → CHAPTER XVIII., CAPUT XVIII. (cont.) → CHAPTER XVIII. (cont.), CAPUT XIX. → CHAPTER XIX. Numerals mirror the plate. Frontmatter copied VERBATIM from each Latin chunk, field order included.

#### (g) Three addresses Migne prints irregularly, all reproduced as printed

• VERS. 27-19.-- at 0127D — the plate's own reading, and our TEI had normalized it. See §2.
• « VERS. 10.--Et habebit filium, » at 0127A — Migne sets the verse address inside the guillemets. Reproduced there, untranslated, exactly as the twin has it.
• VERS. 37, 38. « at 0133B — no -- after the period, where every other address in the range prints one. Reproduced bare.

§2 · ⭐⭐ 0127D — VERS. 27-19. IS WHAT THE PLATE PRINTS, and our twin had tidied it

twin: VERS. 17-19.-- « Num celare potero Abrahae quae « gessurus sum. » plate: VERS. 27-19. — « Num celare potero Abrahæ quæ « gessurus sum. »

Read at 600 dpi on p. 69 (corner numbers 127/128 checked first). The first digit is a 2, not a 1, and the control is on the same line: the 1 of the adjacent 19 is a plain vertical stroke with a foot-serif, while the first glyph carries the flat base and curled head of a 2. A second control sits on p. 71, where Migne sets VERS. 17-19. correctly for Gn 19:17–19 — the same two digits, in the same fount, unmistakably different.

The English renders the plate: VERS. 27-19.--. Three reasons, and they agree:

1. The locked rule — where a translator has opened the column at the plate and Migne's type differs from our TEI twin, the English renders what Migne printed, with the pending patch filed.
2. It is not a mark, so it is not covered by the punctuation-follows-the-twin carve-out. It is content: a citation address, and this edition's addresses are its citation-resolving apparatus.
3. Migne's own printing is not ours to tidy — the series brief's rule for a verse number out of sequence (Jude prints VERS. 23 before VERS. 21) is this case exactly, one step further along.

⛔ No [sic:], and it could not have taken one: the marker's content must be verbatim in the Latin twin, and the twin prints 17-19. The patch is what makes the two columns agree again. ⚑ This is the sharpest instance in the range of brief §4 — the digitization "corrects" as well as corrupts — and nobody could have flagged it from the Latin alone, because VERS. 17-19. is exactly the address the lemma requires. Only the plate shows it.

§3 · THE THREE UNPAIRED GUILLEMETS — all three MIGNE'S, all plate-read

The rebuilt brief emits these as ⚠ UNCLOSED « spans running to the end of their paragraph, and warns the printed span may be short of what Migne quotes. Each was collated at the line, not from the brief.

All three are mirrored unclosed (Pattern 5's corollary). In each the etc. or the quasi: that follows shows the quotation is complete for Migne's purpose; the missing » is his. ⚠ The fourth apparent anomaly is not one: at 0121A and 0127D an inner « stands at a line head inside a closed span (« Dixitque Abram. [n: …] « Ego vado sine liberis, »; « Num celare potero Abrahae quae « gessurus sum. »). That is Migne's line-continuation mark, the class the 0000 stint identified, and it is reproduced in place. At 0121A the plate actually sets a CLOSING » there and our twin sets « — a mark, so the English follows the twin and the patch file carries the plate reading.

§4 · ⚠⚠ 7a″ — THE VULGATE COLLATION. This is the deliverable.

Every lemma in these five chunks was collated against sources/vulgate/clementine-flat.txt by grepping the verse, cited book chapter:verse — never from memory. Agreements are listed with divergences, because a list of findings alone cannot be told apart from a report by an agent that never looked.

4.0 ⭐⭐ THE FLAGGED SITE, 0121A — « Suspice coelum, » — NOT a divergence, and the reason is in OUR comparand

[0121A] VERS. 5-8.-- « Suspice coelum, » ⚑ MATCHES ONLY ELSEWHERE — Job 35:5 (lemma is Gn 15)

Verdict: Migne prints Gn 15:5's own words. No marker. The flag fired on a defect in the Clementine copy on disk, compounded by orthography. Walked at the line and settled with the file:

1. The verse is fixed independently of the words. The address is VERS. 5-8. under ## CAPUT XV.; its neighbours are VERS. 2. (Gn 15:2 Dixitque Abram) and VERS. 9-10. (Gn 15:9 Sume, inquit, mihi); and the paragraph's own next lemma is « Sic erit semen tuum, » — the closing words of Gn 15:5, which our file carries verbatim. So the lemma is Gn 15:5 and cannot be Job.
2. ⛔ sources/vulgate/clementine-flat.txt line 15072 and clementine/Gn.lat line 365 both read Suscipe cælum at Gn 15:5. The received Sixto-Clementine reads Suspice cælum, as does Weber. The defect is in our copy, not in Migne — and sources/vulgate/README.md says in terms that later upstream corrections are not applied and that a call turning on a single letter must say so rather than trust this copy to that precision. This is that case, on the comparand side.
3. coelum for caelum is Migne's orthography throughout this book and is not a divergence (cruces-0000 settled it). With Suscipe in the flat file, a normalized phrase search could match neither Gn 15:5 nor anything else in Genesis — so it fell through to Job 35:5 (Suspice caelum, et intuere), a different book that the commentary has not reached and does not cite. The ⚑ is telling you where the string matched, which is exactly what the brief says it does.
4. The plate, read at 600 dpi (p. 66, col 121, band A, corner numbers checked first): VERS. 5-8. — « Suspice cœlum, » etc. Superius dixit : — Migne's type reads Suspice coelum. So there is nothing here to mark against him.

⭐ AND THE RECURRENCE, which is why the brief tells you to look for one. The same class fires a second time nine lines below, at 0121B « Vaccam triennem, et capram trimam, » — marked ⚠ NOT in Clementine verbatim on the same grounds. Our copy reads vaccam trienem at Gn 15:9 (one n); the received Clementine reads triennem, exactly as Migne prints it. No divergence there either.

⛔ The finding worth carrying past this stint: the comparand can manufacture a divergence. CLAUDE.md 8a records the digitization manufacturing one on the Latin side (9003's judicabit→judicavit); this is the same failure on the Vulgate side, twice in one column, and no plate read can catch it — only reading the received text can. Both would have produced a false public claim against Migne had the ⚠/⚑ been trusted as findings. Every single-letter divergence in §4.2 below is therefore reported as a single-letter divergence and declined.

4.1 The eleven [var:] fired — all on columns read at the plate this session

4.2 DIVERGENT and DECLINED ON THE MERITS — every one on a read column, so the marker was available

These are the ones a blind reader will find, so each says why it was refused.

• 0121A « Erit semen tuum sicut arena maris, » — introduced Superius dixit. Not a Vulgate verse. Gn 13:16 has sicut pulverem terrae, Gn 22:17 velut arenam quae est in littore maris, Gn 32:12 sicut arenam maris of Jacob. It is the gloss's own summary back-reference. [var:] has no received reading to name; declined and recorded, as the 0008 stint recorded 0102A.
• 0121A « Sic faciam semen tuum, » — likewise no Vulgate verse: Rabanus's coined antithesis to Gn 15:5's Sic erit semen tuum, which stands three lines below with the sand instead of the stars. Declined.
• 0122B « Tu laborem et dolorem consideras, et tibi derelictus est pauper, » — Ps 9:35 Vides, quoniam tu laborem et dolorem consideras, ut tradas eos in manus tuas. Tibi derelictus est pauper. Migne drops the opening Vides quoniam and the ut tradas clause and joins the two halves with et. One verse abridged, and the omitted clause is not what the gloss adduces it for (gratuito consulit miseris). Declined; the joining et is recorded.
• 0122B « Vocavit autem Agar. » — Gn 16:13 Vocavit autem nomen Domini qui loquebatur ad eam. Migne's lemma names Agar, whom the Vulgate leaves as the implicit subject, and cuts the object. Declined as a clipped lemma with the subject supplied for the address, convention 3's shape: the gloss immediately restores what is cut (Recte nomen ejus ita vocavit), so nothing rests on the difference. Recorded because the substitution of a proper name into a lemma is unusual.
• 0123A « Daboque tibi et semini tuo. . . .et omnem terram Chanaan. » — Gn 17:8 Daboque tibi et semini tuo terram peregrinationis tuae, omnem terram Chanaan. Migne's own four-point ellipsis marks the cut, and the et before omnem is his join. A signalled abridgment, not a variant. Declined.
• 0123B « Circumcidetur ex vobis masculinum, » — Gn 17:10 circumcidetur ex vobis omne masculinum. One word dropped; nothing in Strabo's three causes turns on it. Rendered without "every", literally. Declined.
• 0124B « Cecidit Abraham in faciem suam, et risit in corde suo. » — Gn 17:17 Cecidit Abraham in faciem suam, et risit, dicens in corde suo. Dropping dicens moves in corde suo onto risit, which makes the laughter interior — and interiority is close to Jerome's point (Isaac is named from Abraham's laughter, not Sara's). ⚠ The nearest of the declines to the line. Refused because a single dropped word yields a perfectly natural reading either way, and because Jerome's argument is about whose laughter, not about where it was.
• 0124C « Et vocabis nomen ejus Isaac: » — Gn 17:19 vocabisque nomen ejus Isaac. A connective. Declined.
• 0124C « Et foedus sempiternum, » — Gn 17:19 constituam pactum meum illi in foedus sempiternum. A clipped lemma with in → Et. Declined.
• 0125A « Nos sumus enim circumcisio, qui spiritu Deo servimus. » cited Phil. III — Phlp 3:3 Nos enim sumus circumcisio, qui spiritu servimus Deo. Word order twice, sense identical, citation correct. Declined.
• 0126D « Lac potum dedi vobis, non escam, » cited I Cor. III — 1 Cor 3:2 lac vobis potum dedi, non escam. Word order only; citation correct. Declined.
• 0127B « Laetare sterilis, quae non parturis, » cited Isa. LIV et Gal. IV — Gal 4:27 Laetare sterilis, quae non paris; erumpe et clama, quae non parturis; Is 54:1 Lauda, sterilis, quae non paris. Migne joins Galatians' opening to Galatians' second relative clause, skipping the middle. ⚑ Declined because Migne cites BOTH sources himself, so the conflated form is a citation-form question, not a text divergence — the 0004 stint's 0082B precedent. And nothing in the gloss rests on paris against parturis. Rendered parturis as "travailest", keeping the two verbs distinguishable.
• 0127D « Num celare potero Abrahae quae gessurus sum. » — Gn 18:17 Num celare potero Abraham quae gesturus sum. Two divergences, both plate-confirmed at 600 dpi: Abrahae (dative, the late/medieval construction of celare) for the accusative, and gessurus (a perfect-stem analogical future participle) for gesturus. Declined: neither has an English exponent — both readings give "hide from Abraham" and "what I am about to do", so Pattern 14's "in a way that changes the sense" is not met. ⛔ And no [sic:] on gessurus: it is a well-formed Latin word-shape, not broken type, and Addendum A's rule is that a real word that reads oddly is Migne's until a plate says otherwise — here the plate says it IS Migne's.
• 0128A « in gentem magnam futurus, » — Gn 18:18 cum futurus sit in gentem magnam. The verse's own words, re-jointed into Migne's sentence as a participial phrase. Declined.
• 0128A « in quo sunt omnes thesauri sapientiae et scientiae absconditi » cited II Cor. II — the text agrees with Col 2:3 exactly; it is Migne's citation that is wrong, and the plate prints it barer still ((II Cor), no chapter). No [var:], since Pattern 14 marks the text; the note rides verbatim per refDisplay. Same shape as the 0008 stint's Psal. LIII at 0100B.
• 0129A « Memento, Domine, quoniam pulvis sum » cited Psal. X — no psalm carries these words. The nearest is Ps 102:14 recordatus est quoniam pulvis sumus, third person and plural; Ps 10 carries neither the words nor the sense. Gregory's adaptation, cited to a psalm that is not its source. [var:] has no received reading to name; declined and recorded, with the citation logged.
• 0130C « Percusseruntque caecitate. » — Gn 19:11 percusserunt caecitate. An enclitic -que. Declined.
• 0132B « Nemo ponens manum in aratro, et respiciens retro, aptus est regno Dei. » cited Luc. IX — Lc 9:62 Nemo mittens manum suam ad aratrum. Two words, and the citation is correct. Declined: no sense change — "putting his hand on the plough" and "putting his hand to the plough" assert the same thing, and the gloss is entirely about the looking back. Rendered in aratro as "on the plough", not conformed to the familiar "to the plough".
• 0132B « Mementote uxoris Lot, » cited Luc. XVII — Lc 17:32 Memores estote uxoris Lot. A synthetic imperative for the periphrasis; identical sense, correct citation. Declined.
• 0132C « Recordatus est Abrahae. » — Gn 19:29 recordatus Abrahae (participle). Migne makes it finite. Declined.
• 0132D « et nescivit quod dormissent cum eo; » — no Vulgate equivalent: Gn 19:33 reads at ille non sensit, nec quando accubuit filia, nec quando surrexit and 19:35 et ne tunc quidem sensit. Migne's clause is the LXX/Old-Latin form. ⚑ Declined because the sentence around it is ITSELF a note on the reading's textual status — appungunt enim desuper Hebraei obelo, quasi superfluum — so the divergence is the passage's own subject and a [var:] would tell the reader twice. Same instinct as §4.3.
• 0133B/0133C the Ammon etymologies — filius generis mei, benammi, filius populi mei, populus meus, de patre — are Jerome's Hebrew glosses, not quotations. Gn 19:38 does print id est, Filius populi mei, which agrees. No marker.

4.3 A divergence Migne LABELS as a rival version — no [var:] by construction

• 0131A « Egressus est Lot, et locutus est ad sponsos qui accepturi erant filias ejus: », introduced Sed Hebraica veritas habet. Gn 19:14 reads Egressus itaque Lot, locutus est ad generos suos qui accepturi erant filias ejus. The divergence is the sentence's whole argument — sponsi against generos, which is what decides whether the daughters were married — and Migne announces the witness in his own voice. Convention (b), settled by the 0008 stint: the marker would restate what the text already says. Declined by construction, and it is the only instance of the class in this range — which is worth recording, since the 0095–0106 range carried twelve.

4.4 Lemmata collated and found to AGREE with the Clementine — recorded, per 7a″

Gn 15:1 (« His itaque transactis, » · « Noli timere. ») · 15:5 (« Suspice coelum, » — §4.0 · « Sic erit semen tuum, ») · 15:9 (« Sume, inquit, mihi, » · « Vaccam triennem, et capram trimam, » — §4.0) · 15:11 (« Descenderuntque volucres, ») · 15:13 (« Scito praenoscens, » ×2) · 15:16 (« Generatione autem, ») · 16:1 (« Igitur Sarai uxor Abram, ») · 16:2 (« Ingredere ad ancillam, ») · 16:12 (« Hic erit ferus homo. ») · 16:13 (« Profecto hic vidi posteriora videntis me. ») · 16:14 (« Propterea appellavit puteum illum, ») · 17:2 (« Ponamque foedus, ») · 17:5 (« Nec ultra vocabitur nomen tuum Abram. ») · 17:8 (« In possessionem aeternam. ») · 17:11 (« Carnem praeputii. ») · 17:14 (« Masculus cujus praeputii caro, ») · 17:15 (« Sarai uxorem tuam, ») · 17:19 (« Nomen ejus Isaac. » · « Constituam pactum meum, ») · 17:20 (« Super Ismael, ») · 17:21 (« In anno altero. ») · 17:23 (« Tulit autem Abraham, ») · 18:1 (« Apparuit autem ei Dominus in convalle Mambre. ») · 18:2 (« Tres viri. » · « Quos cum vidisset. ») · 18:4 (« Sed afferam. ») · 18:6 (« Tria sata. » ×3 · « Subcinericios. ») · 18:7 (« Vitulum tenerrimum. ») · 18:8 (« Butyrum et lac ») · 18:9 (« Ecce in tabernaculo est, ») · 18:10 (« VERS. 10.--Et habebit filium, » · « Quo audito, ») · 18:11 (« Erant autem, ») · 18:12 (« Voluptati operam dabo? » — including Migne's question mark, which is the Vulgate's own) · 18:15 (« Non risi, ») · 18:20 (« Clamor Sodomorum et Gomorrhae, ») · 18:21 (« Descendam et videbo, ») · 18:24 (« Si fuerint quinquaginta justi, ») · 18:27 (« Loquar ad Dominum, » clipped · « Loquar ad Dominum meum, cum sim pulvis, » clipped · « Loquar ad Dominum meum cum sim pulvis et cinis, » verbatim) · 18:28 (« Quinque fuerint. ») · 18:32 (« Quid si inventi, » · « Decem. » · « Dixit. ») · 19:1 (« Veneruntque duo angeli, » · « vespere, » · « Qui cum vidisset, ») · 19:2 (« Lavate. ») · 19:3 (« Compulit illos. ») · 19:8 (« Habeo duas filias. ») · 19:10 (« Et introduxerunt ad se Lot. ») · 19:14 (« Egressus, ») · 19:15 (« Cumque esset mane, ») · 19:17 (« In monte. ») · 19:19 (« Nec possum in monte: ») · 19:20 (« Est civitas, » · « Est civitas haec juxta, » · « Juxta » · « Ad quam possum, ») · 19:21 (« Ecce etiam in hoc suscepi preces tuas, ») · 19:24 (« Igitur Dominus pluit super Sodomam, ») · 19:25 (« Universos habitatores. ») · 19:26 (« Respiciensque uxor ejus. ») · 19:29 (« Cum enim subverteret, ») · 19:30 (« Ascenditque, » · « Ascenditque Lot de Segor, ») · 19:31 (« Dixitque major ad minorem, ») · 19:32 (« Veni, inebriemus, ») · 19:33 (« Dederunt, ») · 19:36 (« Conceperunt ergo, ») · 19:37 (« Vocavit nomen ejus Moab, ») · Rom 2:28 (« Non enim qui in manifesto Judaeus est… » — verbatim, including the comma placement) · Lc 8:8 (« Qui habet aures audiendi, audiat; ») · Mt 5:8 (« Beati mundo corde ») · Act 15:9 (« Fide purificans corda eorum. ») · 1 Cor 7:5 (« Nolite fraudare invicem, »).

⚠ One near-agreement recorded as such: 1 Cor 7:2 « Propter fornicationem unusquisque uxorem habeat. » (0131B) drops autem and suam. Neither is sense-bearing and the citation is correct; listed here rather than in §4.2 because the omission is of two function words. ⚠ coelum for caelum, Chanaan/Sodomam/Gomorrhae capitalization, and Samsom for Samson (0124C) are orthography, not divergence. The name is rendered by its established English form; the plate's spelling stands in the facing Latin.

§5 · THE DECLINED LIST — suspect readings, what settles each, and why NO [sic:] fired

Zero [sic:] in 5,245 words, and on this book that is now four stints and 22,000+ words. The reason is stated once and holds for every entry: a [sic:] accuses Migne's type, its content must be verbatim in our twin, and every plate defect this range actually contains is one our twin has already tidied away — so the marker is owed after the patch, not withheld now. The mirror case is equally firm: every defect our twin carries is OURS, and a [sic:] on one would blame a clean page. On 9004, translating agents' "probably our-file" calls were right 5 times and wrong 9; here the plate decided all nine, and it went both ways.

5.1 Non-words in the twin — ALL THREE ARE OURS, settled at the plate

⚑ The split run is the dangerous shape (7a⁗-b): quarantining agno scebat under Pattern 10 would have taken the clause's only finite verb out of quem in angelis agnoscebat. It was not quarantined, because it is not Migne's. A fourth, supplicio-ignis at 0131A, is a spurious hyphen in the twin (plate: a space) and is likewise ours.

5.2 Non-words on the PLATE — three, all normalized away by our twin, none markable

eic. for etc. (0129B) · Sodomorm. for Sodomorum (0129C) · susperstites for superstites (0133A). All three read at 600 dpi. These are Migne's, and they are exactly the class brief §4 warns cannot be seen from the Latin alone — our file prints three perfectly good words. Each becomes a live [sic:] after the patch lands; none is one now, because the guard would reject it and rightly.

5.3 ⭐ A SUSPICION WITHDRAWN AT THE PLATE — 0123B (RAB. STRAR.)

Raised at 150 dpi as an obvious STRAB. mis-transcribed. Read at 600 dpi, the plate sets STRAR. — Migne's own broken small-cap sort, and our twin is faithful. No patch, no marker. ⚠ Migne's ( sic ) rule applies in spirit: this is inside an [n:], whose contents ride verbatim and untranslated, so nothing of ours goes near it. Recorded because it is the exact shape of the 8955 @0629D ulit-for-Tulit finding and went the other way, and because a checked negative is worth having beside §5.2's three positives.

5.4 Real words that read oddly — rendered as printed, no marker (Addendum A)

• 0123D pueri circumcisi — this one took a [cj:], see §6. The plate reads circumcisi, so it is Migne's and there is nothing to [sic:].
• 0124D Annus alter … tempora sunt regenerationis — a singular subject with a plural predicate nominative. Rendered as printed, "The other year … is the times of regeneration". English carries the disagreement, so Pattern 12's second test (a defect with no English exponent) is not triggered and no marker is needed. Logged.
• 0132A ne posset adverti impiissimum facinus Sodomitarum — Alcuin on why the infants were burned. adverti will not construe comfortably in any of its senses ("be turned aside", "be attended to"); the sense the argument wants is that the crime should be left without trace. Rendered literally — "lest the most impious deed of the Sodomites could be turned aside" — and the conjecture is here and not in the text. Candidates weighed: animadverti ("be punished", the likeliest, and a haplography of anim- is a plausible compositor's slip) and averti. ⛔ No [cj:]: the printed word does not make the sentence assert something Alcuin is arguing against — it makes it assert nothing very clearly, and Pattern 18's own boundary is that a sentence too broken to assert is not the marker's business. Plate read (p. 71): Migne's type is adverti.
• 0127B mirum est: nisi homines prius arbitratus est, sed in quibus Deum loqui intelligeret — the roughest sentence in the stint; the mirum est has no complement and the nisi clause hangs. Rendered clause by clause as printed. Plate (p. 69) agrees with the twin word for word. No marker.
• 0130A ut sciamus Deo benefacere, jucundum et placitum Deo proprie esse ulcisci vero Deo ingratum esse — the Glossa Graeca's Latin, with no punctuation between esse and ulcisci. Rendered on the reading that proprie esse governs benefacere ("that to do good is proper to God") and that ulcisci … ingratum esse is the answering member. Plate agrees with the twin. No marker; the alternative construal (Deo benefacere = "to do good to God") is recorded here.
• 0122A et omnibus bonis contrarius — bonis is masculine or neuter and the plate does not disambiguate. Rendered "contrary to all good things"; "to all good men" construes equally and is what the sentence's own subject (the persecuting people) suggests. Nothing in the passage decides it. Logged rather than chosen silently.

5.5 Unexpressed subjects supplied, with what decided each

• 0122B Recte nomen ejus ita vocavit — supplied "she" = Agar, from the lemma directly above (« Vocavit autem Agar. »), and ejus = the Lord's, from Gn 16:13's nomen Domini qui loquebatur ad eam. The alternative (an impersonal "one called") would detach the naming from the woman the whole gloss is about.
• 0122D Propterea appellavit puteum illum — the Vulgate leaves the subject unexpressed at Gn 16:14 too. Rendered "she called", continuous with the lemma two paragraphs above, and noted.
• 0129A Quosdam per tumorem sensus non elevat (chunk 0022's opening, mid-sentence from the previous chunk) — the Latin prints no subject at all, and the chunk boundary cuts off what would have supplied it. Rendered with a bare "it", which is as open as the Latin: candidates are abundantia/wealth (from the terrenis abundare two clauses on) and God. ⛔ Not chosen — this is the 11613 class, and naming the subject here would be a guess wearing Gregory's voice.
• 0127B illius risus admirationis fuit, hujus dubitationis — ⚠ the one place two readings both construe and the choice matters. The argument requires Abraham's laughter to be of wonder and Sara's of doubt: Sara alone is reproved, she alone denies it, and Abraham vero Deum intelligebat follows. But the classical pointer runs the other way — the nouns stand in the order Sara … Abraham, so illius (the former) is Sara and hujus (the latter) is Abraham. Rendered so as to keep the ambiguity the Latin has: "because the laughter of the one was of wonder, that of the other of doubt." Nothing is decided in the text; the argument for Abraham=wonder is recorded here. Augustine's own Quaest. in Hept. I.36 has admirantis / dubitantis in that order.
• 0128A duo illi promittuntur — read as dative, "two things are promised him". duo illi as "those two" also construes; the colon that follows lists the two, so the numeral is doing the work either way. Logged.

5.6 What was checked and found NOT to be a defect

• 0125C the missing colon after mundabantur. The plate prints mundabantur : unde :; our twin prints mundabantur unde:. The English follows the twin and therefore reads "as before circumcision holy men were cleansed whence: « … »" — abrupt, and deliberately so. A mark, so convention (ii) governs. Recorded prominently because a blind reader will meet it and read it as ours.
• 0125C the opening guillemet before « Fide purificans corda eorum. » is not on the plate — our twin supplied it. The English carries it, because guillemet parity against the twin is a hard verify-english check. ⭐ This is the inverse of the 0008 stint's 0106A, where the twin dropped an opening mark. Two directions, one class, and only the plate separates them: the corollary the 0008 stint drew — never adopt a policy about unpaired marks without reading the column — is confirmed from the other side.
• 0128A « in quo sunt omnes thesauri sapientiae et scientiae absconditi » cited (II Cor. II) in the twin and (II Cor) on the plate. The note rides verbatim either way; the patch file carries it.
• 0126C [ Al., rancore] — Migne's own bracketed variant, reproduced verbatim and untranslated per the register. The word it varies, angore, is rendered ("anguish").
• 0129C the paragraph that ends without a period (ad exaggerationem iniquitatis Sodomorum) and 0123A (in anteriora extenditur) and 0131A (supplicio-ignis aeterni) — three paragraphs the twin leaves unstopped. Mirrored unstopped (Pattern 8, never supply). The plate does print a period at 0129C; a mark, so the twin governs.
• 0127D « Num celare potero … gessurus sum. » closes with a PERIOD, not a question mark. Rendered as an interrogative sentence closing on Migne's period: « Can I hide from Abraham what « I am about to do. » ⛔ The mark is not supplied and not deleted. Pattern 8b does not apply — this is a direct question that English can carry; only the terminal mark is odd, and recording his oddity is the point (the 11537 1063D–1064A precedent, in the other direction).

5.7 The three licensed Pattern 8b colons, and the band scan that found them

A band-by-band comparison (colons, semicolons, question marks, guillemets, per column anchor, apparatus stripped) returned exactly three mismatched bands in five chunks, all : +1, all the same class, and zero elsewhere:

Migne's ? is kept in all three; English cannot end an indirect construction with it. ⚑ The scan also found four unlicensed marks that had crept into the draft and were repaired: a supplied semicolon at 0125B (ne mala faciant, pedes — his comma), one at 0126A (solis necessariis intenta, nesciat — his comma), one at 0131D (Sulphur, fetor carnis, ignis — his comma), and a semicolon standing where Migne prints a colon at 0130B (ne fiant graviora:). Totals had not cancelled here, but the sites were invisible until the bands were compared. ⛔ No comma→semicolon sweep was made. Migne prints 48 semicolons in these five chunks; Tobit's narrow exception rests on his printing zero and does not apply.

§6 · THE ONE [cj:] — 0123D pueri circumcisi

[n: (ISID.)] Quaeritur quomodo pueri circumcisi (qui pactum Dei irritum non fecerunt, sed qui eos circumcidere neglexerunt) perire debeant?

Fired: circumcised [cj: circumcisi; read incircumcisi, "uncircumcised"] boys.

The printed word is a real Latin form, so Pattern 7 renders it and Pattern 12 has nothing to wrap — and the English then asserts what Isidore is arguing against. That is Pattern 18's founding shape.

Why the conjecture, from the sentence's own structure and not from what reads well:

1. The parenthesis assigns the guilt away from the boys — qui pactum Dei irritum non fecerunt, sed qui eos circumcidere neglexerunt. The boys did not void the covenant; their parents did, by neglecting to circumcise them. A circumcised boy has no parents who neglected anything, and the parenthesis has nothing to distinguish.
2. The lemma is Gn 17:14 — Masculus cujus praeputii caro circumcisa non fuerit, delebitur anima illa — whose whole subject is the uncircumcised male. The question is asked of the verse's own class.
3. Isidore's answer is about those not freed: Ideo pereunt omnes qui per regenerationem non liberantur. The perishing class is defined by the absence of the sign, not its presence.

Why the marker and not a silent repair or a bare crux. Pattern 7 forbids the repair; and a crux alone puts the reading where the reader does not meet it, which is 11081 @0030B's exact failure. The gloss is additive — Migne's word keeps its English and the conjecture stands beside it — and the verbatim-in-the-Latin-twin guard is satisfied by circumcisi, so nothing is smuggled in. The class is 18a's admitted one: a lost negation (here the privative in-), which is the defect class the whole 7a⁗ apparatus exists for and which has nowhere else to live when no type is broken. The 7561 @0648 precedent is exact — the gloss opens on the word Migne did print and supplies what fell out.

⭐ Plate-read (p. 67, 600 dpi, corner numbers checked first): Migne's type reads circumcisi, with no prefix. So the reading is his and not the digitization's — which is what makes this a [cj:] and not a patch. Had the plate read incircumcisi, the entry would have moved to the patch file and no marker would stand.

§7 · SECTIONS READ CLEAN (Test 4 — does the English ASSERT what he asserts?)

Run last, with counts, bands and anchors already known clean, paragraph by paragraph against the Latin.

• 0121A–0121D — the Rabanus stars/sand antithesis and the two Augustine quaestiones. The densest negation in the stint sits at 0121D (Non sic accipiendum est, tanquam … quadringentos annos populus Dei fuerit), where the whole point is that the four hundred years are not the years of the servitude. Checked twice.
• 0122C — the Ex 33 face/hinder-parts passage, six negations in five lines (non apparebit · Non enim … et vivet · nemo potest videre et vivere), each traced to its host.
• 0123B–0124B — Strabo's three causes and Jerome's Hebrew letters. The letter-names and the three empty ( ) were checked span by span against the brief's inventory.
• 0125A–0125C — the circumcision-of-the-senses catalogue. ⚑ The site test 2a exists for: non solum se ab illicitis temperat, sed etiam a concessis — the negative is on solum, not on temperat, and the sed etiam correlative proves it. The English puts "not only" on "from things unlawful". Same check run on Qui autem non obscoenis desideriis … moechatur, sed fidei puritatem … custodit.
• 0127C — the physicians' passage, five negations turning on senior/senex/juvencula. ⚑ Emortuum enim corpus non omnino ad generationem erat is the trigger shape (a quantifier beside a verb): the negative belongs on omnino, because the next clause has Abraham begetting of Cethura. English reads "was not altogether dead for generation".
• 0128B–0128D — Augustine on God's anger and Hilary on God's not-knowing, the heaviest double negation in the range (Habemus nescientem Deum, quod tamen non nesciat; non ea tunc scire, quia prius nesciat). ⚑ non est ignorantiae diminutio was drafted as "a diminution of knowledge" and corrected to "a diminution of ignorance" — the genitive is ignorantiae and reversing it would have made Hilary say the opposite about what is being diminished.
• 0130C–0131B — Gregory on the blinded Sodomites and Strabo on Segor, and the avidentia passage, whose four negations (non videri, non omnia, sed quod non est opus … ut nihil viderent) were traced individually.
• 0132A–0133C — Alcuin on the infants, Isidore on Lot's wife and on the law, Jerome's Ammon etymologies. Est etiam qualecunque beneficium, non reum esse, qui gloriosus non est was re-ordered in English so that each negative sits on its own host.

Candidates rejected, with reasons, so a blind reader knows they were looked at: §5.1–5.6 above, plus — Samsom at 0124C (Migne's spelling of Samson, rendered by the established English name, logged, no marker) · frater Abraham at 0131A of Lot, who is Abraham's nephew (the Latin idiom, and the Vulgate uses it; rendered "brother") · Marci XIX at 0131A (a book with sixteen chapters; the citation rides as printed and the [var:] names Mt 19:21) · the four-dot ellipsis runs at 0122D (. . . . . .) and 0123A (. . . .), reproduced with Migne's own spacing · [n: ( Glossa Graeca. )] at 0124D and 0130A, reproduced with his internal spaces · the double blank line the twin prints at 0123B before VERS. 10., mirrored.

STINT 6 — chunks 0024–0030 (cols 0133C–0147C)

Filed as cruces-0024.md; merged verbatim, headings demoted one level. Nothing in it was rewritten — where a later stint corrected it, the correction stands in that stint's section.

Stint 6 of 8950. Seven chunks, 5,994 Latin body words → 8,245 English (1.38×). verify-english.mjs reports no error on 0024–0030: column anchors, [n:] notes, guillemets, italics and question marks all 1:1 with the Latin twins. (Its VERIFY FAILED is the unwritten chunks of peer ranges, nothing in mine.)

Apparatus fired: 16 [var:] · 0 [sic:] · 0 [cj:] · 0 [d:] · 0 [ed:] · 0 [nt:]. Every one of the sixteen stands on a column I read at the plate myself (pp. 72–79, cols 0133–0148, corner numbers read first on all eight). The reads are recorded in data/briefs/8950-PLATE-READS-0024.json for the orchestrator to merge into data/plate-reads.json — until that merge lands plate-gate.mjs reports fifteen of them as standing on unread columns, which they do not (0133D already passes on the 0019–0023 stint's read of p. 72).

Counts, chunk by chunk (Latin = English in every case, body only): guillemets «/» — 0024: 18/18 · 0025: 25/25 · 0026: 34/34 · 0027: 6/6 · 0028: 38/37 · 0029: 2/2 · 0030: 37/37. [n:] notes — 5 / 15 / 17 / 4 / 21 / 1 / 21. Question marks — 0 / 2 / 1 / 0 / 1 / 0 / 2. Column anchors — 4 / 12 / 9 / 2 / 15 / 1 / 13.

§0 · SPAN COUNT AGAINST THE BRIEF — 219 / 1 open-ended, EXACT, and checked span-for-span

My Latin carries 218 paired spans plus 1 open-ended = 219; the brief's header says 219 spans, of which 1 OPEN-ENDED. Counted paragraph-bounded, per the launch instruction, with «[^«»]»|\[^\n]+\ run inside each paragraph and the unmatched-« residue counted separately, so an unmatched delimiter cannot run forward across a paragraph break.

Per chunk (paired + italic / open-ended): 0024 22/0 · 0025 35/0 · 0026 45/0 · 0027 10/0 · 0028 49/1 · 0029 4/0 · 0030 53/0.

⛔ The totals were not trusted on their own — the 0004–0007 stint's false 215/215 is the reason. I diffed my 219 against the brief's 219 span for span, normalized, and the two lists are identical with no residue on either side. The only apparent differences were my own extractor not stripping the VERS. n-m.-- prefixes the brief prints, i.e. presentational.

⭐ The open-ended span is MIGNE'S, confirmed at the plate. 0141D « Si vobis spiritualia seminavimus, magnum est si nos carnalia vestra metamus. — p. 76, 600 dpi: the plate carries no closing » either. It stays unclosed in the English (Pattern 5 corollary). The brief's warning that the shown text may be SHORT of what Migne quotes does not bite here: I read it at the line and the quotation ends exactly where the paragraph does.

⚠ One inventory line describes a span Migne did not set. At 0143C the brief lists a single enormous span, « Inclinata, etc. Quia passionum exercitia … sacrificium vespertinum. ». The plate prints TWO spans there — « Inclinata, » closes, and a fresh »-marked quotation opens before Elevatio manuum — and our twin has lost both marks. The English mirrors the twin (marks follow the twin, §1c), and the plate reading is filed as a pending patch. This does not change the count, because the twin's run is closed; it does mean one of the 219 is a TEI artifact.

§1 · CONVENTIONS — inherited unchanged, and the one this stint had to settle

Ruth (8968 §1) and Tobit (8970 §1) govern and were not re-opened, nor were the four earlier stints of this work (cruces-0000.md, -0004.md, -0008.md, -0012.md), all of which I read before drafting: VERS. n.-- passes through verbatim; the lemma is Englished; a fragmentary lemma is rendered as the phrase it stands for; sigla ride as ordinary [n: …] notes with Migne's own internal spacing, contents unanglicized; the guillemets are the plate's, 1:1, and their contents are Englished; usque ad → as far as, with no verb and no object ever supplied; the Historice / Allegorice / Mystice tags are translated inside their italic span; no [var:] on a quotation Migne labels himself as a rival version; words follow the plate, punctuation follows the twin; a Latin gloss giving the MEANING of a name is Englished inside its span, a Latin word MENTIONED as a word is not; usque ad is decided per occurrence by the etc. splice test.

(a) ⭐ NEW — how a plate/twin WORD divergence is rendered, stated as a rule

Twelve sites in this range have the plate and the twin printing different words, which is far more than any earlier stint of this work met at once. The English was set by one rule, uniformly, and it is set out with every site in data/briefs/8950-PENDING-TEI-PATCHES-0024.md §"The convention this stint applied". In brief:

1. Twin defective, plate clean → render the PLATE (the locked rule's founding case). Five sites: probare (0137C), haesitavit (0138D), praedicator (0141C–D), animale (0145A), quam (0136C, no English exponent).
2. Plate defective and its type CANNOT render in place — a non-word, a form with no grammatical slot, a dropped word → render the sense the type is defective for (Pattern 7 as amended 2026-09-05), with no marker, because the [sic:] guard requires the twin to carry the type verbatim and the twin has been tidied. Eight sites: sonorem (0133C), par quem (0136D), accdperat (0137C), incanationem (0139C), est. for etc. (0140A), the dropped sub (0140C), electorem (0141A), caanem (0141C). Precedent: the 0008–0011 stint's "four non-words whose SENSE the English renders."
3. Plate defective but its type DOES render in place — a real word in a real slot → Pattern 7's main rule is untouched and the English renders the plate's word. One site: « plenius. » at 0145A, where the English reads « more fully. » against an argument that requires the bare plenus. That oddity on the page is the disclosure, and it is deliberate.
4. Negation is the exception to (2). Pattern 7a is unconditional, so at 0141B the plate's non is carried into the English even though our twin lacks it. §5.1.
5. A form CARRIED untranslated follows the TWIN, not the plate (seyr, at 0146C), because the English there is the Latin and a divergence would be indistinguishable from our own typo. A form that is TRANSLATED follows the plate.

⚑ Why no marker fires at any of the twelve, stated once: a [sic:] may only wrap type our Latin twin carries, and in every case here the twin has been tidied; a [cj:] may only open on a word Migne prints AND that our twin carries, and its form (printed; read conjectured) cannot express a deletion. Each of the eight sites in (2) is a [sic:] OWED, markable the moment the TEI patch lands. This is the 0008–0011 stint's finding reproduced exactly: every plate defect in this range is one our twin had already tidied away.

(b) usque ad — 21 formulas, 11 ordinary, 1 the verse's own words

Formula (italic usque ad, immediately after an etc.), 21×, all → as far as, no verb, no object and no antecedent supplied anywhere: 0024 @0133D · 0025 @0136A, 0136D, 0137A · 0026 @0139D ×3 · 0027 @0140A ×4 · 0028 @0140D, 0141A, 0142C ×2, 0143A, 0143C · 0030 @0144C ×2, 0146A.

⛔ Eleven ORDINARY, non-formula usque ad, none italic, none after an etc., each rendered by sense (brief addendum D):

⭐ And one that is the LEMMA'S OWN WORDS, the 8967 col 0953C shape that must not be swept: 0145C « Habitavit autem ab Hevila usque Sur, quae respicit Aegyptum, » — Gn 25:18 verbatim, bare usque without ad, rendered "from Hevila as far as Sur". It sits four lines from an ordinary usque ad and eleven from a formula. Said here explicitly, with the Latin that forced it, per the series brief's requirement.

(c) The vocabulary axes — every instance, with the instrument checked before reporting

• gentes / gentilis / gentilitas — 9 occurrences, and the instrument was checked first. A raw gent- grep in this range also returns gentem magnam (0136C) and nothing spurious; there is no indigentia/ducentorum noise here, but the check was run. Gentiles where the Judaei/gentes opposition is live: 0134C de gentili conversatione ("out of Gentile converse", the gentilis rule added on 9000); 0138C « plenitudo gentium intraret » and « ubi plenitudo gentium intraverit » (Rom 11:25, set against Israel in the same sentence); 0138C sacrificium crucis per gentes fuerit praedicatum; 0147B « dabitur gentibus » (the whole gloss opposes Judaicum populum). the Gentile world: 0144A Ecclesia ex gentilitate veniens. nation(s) where the sense is plainly generic: 0136C futurum in gentem magnam; 0146B « Duae gentes » and __gentem__ quam propagavit Esau; 0146D « Duae gentes et duo populi ». ⚠ 0146B–D reads as an inconsistency and is not one: Gn 25:23's Duae gentes is the axis's generic exemplar shape (two peoples in one womb, expounded of Israel/Edom and then of vices/virtues) → nations; the gentibus of Mt 21:43 at 0147B is the typological opposition → Gentiles.
• natio — ZERO in this range. ⚑ Instrument checked before reporting: a raw natio grep returns six hits and every one is incarnationem (0139C ×2), cognatione (0140D, 0144A) or praedestinatione (0140D, 0144A). There is no natio here at all.
• turba — ZERO. No turb- of any kind, and no verb turbare, in the seven chunks. multitudo ×1 → multitude (0139A in multitudine prophetarum).
• daemonium — ZERO. diabolus / satanas — ZERO. ⚑ Checked with the pattern that does not miss daemones/daemonum/daemone; nothing.
• sacramentum ×1 → sacraments (0138A quae portabat omnia sacramenta — of the ass that bore them and knew them not). mysterium ×3 → mystery (0135C prophetiae mysterio compulsa and Significabatur ergo hoc mysterio; 0143D quanto Christi mysteria penetrat). The two words stand five columns apart and were kept 1:1; neither was ever rendered by the other's English.
• The commentator's own voice is not archaised: dicit / ait / aiunt / inquit → "says / say", never saith. -eth forms appear only inside quoted scripture (Jo 7:37 sitit → "thirsteth"; Eph 5:29's habuit is a perfect and takes "hated"; Rom 12:3's oportet → "behoveth" inside the woven Pauline clause at 0141B) and in second-person singular address.
• Pattern 17. Singular → thou: 0133D Chrysostom to Abimelech (Scias autem quod illius preces tibi vitam praebebunt → "know thou … grant thee life"); 0135A Unde cresce, et manducabis me → "grow, and thou shalt eat me"; 0141A voluntatem tuam persuasioni meae inclinem and Humilia saecularem facundiam → "Humble thy worldly eloquence"; 0143D Quem enim fructum habuisti tunc in illis in quibus nunc erubescis? → "hadst thou … thou art now ashamed"; 0146A Ne putes → "Think not thou" (rendered "Think not"). Plural → you: 0133D Mt 23:9 « unus est enim Pater vester »; 0137D Dt 13:3 « Tentat vos Deus vester »; 0141C Rom 13:14 « Carnis curam ne feceritis »; 0142A 2Cor 6:12 « Non angustiamini in nobis, sed in visceribus vestris »; 0143C Lc 21:19 « In patientia vestra possidebitis animas vestras »; 0147B Mt 21:43 « Auferetur a vobis regnum Dei ».

(d) Heads — numerals mirror the plate

CAPUT XX. → CHAPTER XX. · CAPUT XXI. → CHAPTER XXI. · CAPUT XXII. → CHAPTER XXII. · CAPUT XXIII. → CHAPTER XXIII. · CAPUT XXIV. → CHAPTER XXIV. · CAPUT XXIV. (cont.) → CHAPTER XXIV. (cont.) · CAPUT XXV. → CHAPTER XXV. — numerals throughout, per the corpus measurement (26 works CHAPTER XIII., zero spelled forms above ONE). All seven carry their terminal period on both sides.

(e) Frontmatter copied VERBATIM from each Latin chunk, field order included

Every field in the Latin twin's own order, including the words: figures, which count the Latin. Chunk 0029's colFirst/colLast are both 0144B and chunks 0024/0028/0030's colContext runs one band behind colFirst; both are the chunker's and are copied unaltered.

(f) Two mentioned-word sites, decided by the 0012 test (translation, or subject?)

• 0137C sed usualiter, tentare, pro eo quod est probare, dicimus. Neither word is in an italic span in the twin, so italics were unavailable (span parity), and the 0000 stint's straight-quote device applies: rendered we say "to tempt" for that which is "to prove". These are the SUBJECT of the sentence in one sense — but the sentence is about the two verbs' usage, not about their letters, and leaving them Latin would tell an English reader nothing. Recorded because it is the borderline case of the 0012 rule.
• 0143D « teristrum, » is left untranslated inside its guillemets — the one place a guillemet span is not Englished in this range. The clause exists to explain what a teristrum is (quod etiam nunc genus est Arabici vestimenti), so the word is mentioned, not used; this is 0012 §1a's sidera class carried into a guillemet span. Same footing as Bersabee, sabee, sabaa, Alon, debir, Beka, Schekel, siclus, alma, seyr, Edom, all left as printed.
• ⚑ Their opposites, Englished inside their spans per the meaning rule: fons judicii → the fountain of judgment · angustia → narrowness · illuminans / lucens → illuminating / shining · veprem / spinetum → a briar / a thorn-brake · negotiatores → merchants · copulata → joined · risus / patientia → laughter / patience · medius / fluvius → middle / a river · Rubeum / fulvum → Red / tawny · juramentum → an oath · quercus → an oak (inside ( Alon Hebraice Latine quercus ), rendered ( Alon in Hebrew, in Latin an oak )).

§2 · PLATE READS — EIGHT PAGES, ALL EIGHT NEW, CORNERS FIRST

⭐ My range was entirely unread when the stint opened. data/plate-reads.json covered cols 0067–0070, 0081–0082, 0093–0106 and (from a peer's merge) 0121–0133; cols 0134–0147 had no read at all. All eight pages were rendered and read: PDF pp. 72–79 = cols 0133/0134, 135/136, 137/138, 139/140, 141/142, 143/144, 145/146, 147/148.

⛔ The printed corner numbers were read FIRST on every one of the eight pages, before anything else on the page, and each confirmed the map PDF page = (column + 11) / 2. 200 dpi for the full page, 500–900 dpi crops at 22 contested sites.

The foot apparatus: CHECKED ZERO on all eight pages — a real result

No foot rule and no notes of either layer on any of the eight — not numbered, not asterisk-keyed, and not letter-keyed, which is the layer this book is known to carry (cols 0068 and 0094). p. 74's foot carries only PATROL. CXIII. and the signature 5; the rest simply end. Recorded as a checked zero, not a skip. With the earlier stints' pages this puts PL 113's "no conjecture apparatus in these columns" finding on 33 pages across five books — and it is still a sample, not a licence to claim the volume has none.

What the plates found: 21 twin/plate divergences, filed in full

data/briefs/8950-PENDING-TEI-PATCHES-0024.md carries every one with its column, both readings, the dpi it was settled at, and which marker it will take once the patch lands. The headline numbers: 13 are MIGNE'S defect tidied by Corpus Corporum · 4 are OUR corruption of a clean plate · 4 are mark-, address- or Hebrew-level. Two deserve naming here:

• ⭐⭐⭐ 0141B — our twin has DROPPED a printed non. §5.1 below.
• ⭐ 0145A — the plate prints the ADDRESS VERS 13. (no period after VERS, a clear 3 at 900 dpi) where our twin prints VERS. 12.--, and the next address on the plate is VERS. 13-15. So Migne's page repeats the number and CC silently corrected the first to 12. This is the most substantive normalization I found and it is invisible from the Latin alone — brief §4's class, at the level of a structural address rather than a word. The English carries the twin's VERS. 12.--, because the address is copied verbatim rather than translated and a divergence there would misalign the two columns' addresses; the patch is the right instrument.

§3 · ⚠⚠ 7a″ — THE VULGATE COLLATION. This is the deliverable.

Every one of the 219 marked spans was collated against sources/vulgate/clementine-flat.txt, by book chapter:verse, never from memory. The ✓ addresses in the brief were read, not taken as clearance: three of the fires below are spans the brief ticked against a different verse.

3a. FIRED — 16 [var:], every one on a column I read at the plate

3b. DIVERGENT AND DECLINED ON THE MERITS — Migne labels the version himself

Per the 0008–0011 stint's §1(b), settled and not re-opened: where Migne announces the version, its divergence from the Clementine is the sentence's own subject, and a [var:] restates what the text already says. Eight sites, and this range is dense with them:

• 0134A « soror mea est, filia patris mei, sed non filia matris meae, » — Gn 20:12 reads et non filia. Introduced in Hebraeo habetur vegam … id est, etiam vere. Labelled.
• 0134A « a patre, et non ex matre; » — introduced Alia translatio, apertius. Labelled.
• 0136C « Non videbo mortem pueri mei. » — Gn 21:16 reads Non videbo morientem puerum. Introduced In Hebraeo enim prius hoc scriptum est. Labelled.
• 0143B « Egressus est Isaac, ut loqueretur in agrum, inclinata jam vespera; » — Gn 24:63 reads egressus fuerat ad meditandum in agro, inclinata jam die. Introduced In Hebraeo legitur. Labelled.
• 0143D « teristrum, » — introduced Alia editio. Labelled.
• 0144D « Deficiens mortuus est Abraham, » — introduced male in LXX additum est. Labelled, and Migne is condemning the reading in the same breath.
• 0145A « in senectute bona senex », « plenus dierum; », « plenus. » — Jerome comparing the Greek codices with his own. Labelled.
• 0145B « contra faciem omnium fratrum suorum habitabat, » (juxta LXX legimus) and « coram omnibus fratribus suis occubuit, » (verius est quod posuimus) — Gn 25:18 reads coram cunctis fratribus suis obiit. Both labelled, and the sentence exists to prefer the second.
• 0145D « ludebant » · « recalcitrabant: » (LXX) · « confligebant filii in utero ejus. » (Aquila) · « in similitudine navis in superficie ferebantur. » (Symmachus) — four announced versions in one sentence. Labelled.
• 0146C « pilosum » — Ubi nos « pilosum » posuimus, Jerome's own rendering against the Hebrew seyr; Gn 25:25 reads hispidus. Labelled.

⚠ Worked pair inside this range, which is why the rule earns its keep: at 0136C the same paragraph carries both classes. Quod sequitur: « Et projecit puerum subter unam abietem … » is offered as the text and took the marker; « Non videbo mortem pueri mei. » is introduced In Hebraeo and did not. What Jerome is comparing there is the order of the clauses; the word divergences (abietem, jactus sagittae) are not the sentence's subject and are not disclosed by it.

3c. DIVERGENT AND DECLINED AS IMMATERIAL — the bar is "changes the sense"

Recorded because a findings list alone cannot be told from a report by an agent that never looked.

3d. COLLATED AND FOUND TO AGREE — recorded, per 7a″

Exact, or a clean whole-word fragment of, the Clementine: 0133C Profectus inde (Gn 20:1) · Dixitque de Sara (20:2) · 0133D Orabit pro te (20:7) · Alias autem (20:12) · 0134A Filia patris (20:12) · Postquam autem eduxit me Deus de domo patris mei, dixi ad eam (20:13, exact) · 0134B Tulit igitur (20:14) · Ubicunque tibi (20:15, ubicumque orthographic) · Mille argenteos (20:16) · Mirabilis Deus in sanctis suis (Ps 67:36, exact) · 0134C Orante autem (20:17) · 0135A Circumcidit (21:4) · Crevit igitur puer (21:8) · 0135B Lac vobis potum dedi, non escam (1Cor 3:2, exact) · Cumque vidisset Sara (21:9) · Sedit populus manducare et bibere, et surrexerunt ludere (Ex 32:6, exact) · Ejice ancillam hanc (21:10) · 0135C Dure accepit (21:11) · 0135D Quia semen tuum (21:13) · Tollens panem (21:14) · 0136B Imposuit scapulae ejus (21:14) · Abjecit puerum (21:15) · 0136D Exaudivit autem (21:17) · 0137A Eodem tempore (21:22) · Jura ergo (21:23) · Idcirco vocatus est (21:31) · 0137B Abraham vero (21:33) · 0137C Ne forte tentaverit vos is qui tentat (1Th 3:5, exact) · 0137D Tolle filium (22:2) · Quem diligis (22:2) · 0138A In Isaac vocabitur tibi semen (21:12, exact) · Vade in terram (22:2) · Asinum suum ducens (22:3) · 0138B Die autem tertio (22:4) · Exspectate hic (22:5) · 0138C cum asino · plenitudo gentium intraret (Rom 11:25) · postquam adoraverimus (22:5) · Ego et puer (22:5) · 0138D Erunt duo in carne una (Gn 2:24, exact) · 0139A Cumque alligasset (22:9) · 0139B Cornua in manibus ejus (Hab 3:4, exact) · 0139C Dominus dixit ad me (Ps 2:7, exact) · Dixit Dominus Domino meo (Ps 109:1, exact) · 0139D Arietem inter vepres (22:13) · Appellavitque nomen (22:14) · 0140A Vixit autem Sara (23:1) · Et mortua est in civitate Arbee (23:2) · Surrexit Abraham (23:7) · Quod cum audisset (23:16) · Confirmatusque est ager (23:17) · 0140C Erat autem Abraham senex … Dixitque ad servum seniorem: Pone manum (24:1–2, with Migne's own ellipsis) · Ut non accipias (24:3) · 0140D Posuit ergo (24:9) · Ex omnibus bonis ejus (24:10) · 0141A Juxta puteum (24:11) · Igitur puella (24:14) · Inclina hydriam tuam (24:14) · 0141B Occurritque ei servus, et ait: Pauxillum (24:17) · 0141C Sapientibus et insipientibus debitor sum (Rom 1:14, exact) · Carnis curam ne feceritis in desideriis (Rom 13:14, exact) · Nemo enim unquam carnem suam odio habuit (Eph 5:29, exact; unquam/umquam orthographic) · Protulit vir (24:22) · 0141D Palearum quoque (24:25) · Locus spatiosus (24:25) · 0142A Habebat / Habebat autem (24:29) · 0142B Non comedam (24:33) · Et adjuravit (24:37) · 0142C Veni ergo hodie (24:42) · Prolatisque (24:53) · Dedit ea Rebeccae (24:53) · 0142D Et nutricem illius (24:59) · Igitur Rebecca (24:61) · 0143A Qui festinus (24:61) · Habitabat (24:62) · 0143B Et egressus fuerat (24:63) · Egressus es in salutem populi tui (Hab 3:13, exact) · 0143C In patientia vestra possidebitis animas vestras (Lc 21:19, exact) · Elevatio manuum mearum sacrificium vespertinum (Ps 140:2, exact) · Descendit de camelo (24:64) · 0143D Quis est ille homo (24:65) · At illa (24:65) · 0144A Qui introduxit eam (24:67) · Et intantum dilexit eam (24:67; intantum/in tantum orthographic) · 0144B Spe gaudentes, in tribulatione patientes (Rom 12:12, exact) · Abraham vero aliam duxit uxorem (25:1) · 0144C Una est columba mea (Ct 6:8, exact) · 0144D Deditque Abraham (25:5) · Et deficiens mortuus est in senectute bona, provectaeque aetatis, et plenus dierum, congregatusque est ad populum suum. Et sepelierunt eum (25:8–9, exact but for one comma) · 0145A Hae sunt generationes (25:12, autem dropped) · 0145C Duodecim principes tribuum suarum (25:16) · Habitavit autem ab Hevila usque Sur, quae respicit Aegyptum (25:18, exact) · 0145D De Mesopotamia sororem Laban (25:20) · 0146A Perrexitque (25:22) · Sexagenarius, inquit, erat quando nati sunt parvuli (25:26, Chrysostom's inquit inserted) · 0146B Qui respondens ait: Duae gentes (25:23, exact) · 0146C Jam tempus (25:24) · Protinus alter (25:25) · 0146D et major serviet minori (25:23, exact) · Factus est Esau (25:27) · 0147A Simplex (25:27) · Coxit autem Jacob (25:29) · Cui dixit Jacob (25:31) · 0147B Juravit Esau (25:33).

3e. TWO CITATION SIGLA THAT POINT ELSEWHERE — reproduced verbatim, uncorrected

• 0137C Unde I Thess. I: — the words are 1 Thess 3:5. Migne's chapter number.
• 0139C Quod autem dictum est ibidem, c. 9: introducing « Dixit Dominus Domino meo » — the verse is Ps 109:1, not psalm 9. Migne's own back-reference.
• 0147B dicitur Matth. XII: introducing Auferetur a vobis regnum Dei — the verse is Matt 21:43. Migne's book-and-chapter.
• 0139B [n: (Jer. XXXVIII.)] on a sequence that is in no chapter of Jeremias.

All four ride as printed. Passing a siglum through is not endorsing it (series brief, convention 4), and this recension's citations are known to be unreliable.

§4 · THE DECLINED LIST — suspect readings, what settles each, and the page it was settled on

Zero [sic:] fired, and the reason is structural, not timid: every plate defect in this range is one our twin has already tidied away, so a [sic:] would have failed the verbatim-substring guard and would have accused Migne of the digitization's work. That is the 0008–0011 finding reproduced. Each candidate below names the page that settled it.

⛔ Three suspicions I raised and the plate refuted (4.2, 4.3, 4.4) — all three aimed at non-words, all three came back ours, exactly as addendum A predicts. And two came back the other way (4.1, 4.9): a real, well-formed word that reads perfectly was the defective one, and only the plate could say so. Addendum A's rule holds in both directions here, and it is the reason nothing was marked on argument alone.

§5 · NEGATION — the hosts, named (tests 2 and 2a)

5.1 ⭐⭐⭐ 0141B — THE HEADLINE. Our twin has dropped a printed non, and the English carries it.

plate (p. 76, 600 dpi): Vel aquam praebuit, quia non in eo quod credidit, vacua non remansit. our twin: Vel aquam praebuit, quia in eo quod credidit, vacua non remansit.

Rendered: "Or she supplied water, because it was not in that which she believed that she did not remain empty."

Hosts, named: the first non stands on the prepositional phrase in eo quod credidit (constituent negation, fronted before it); the second on remansit. Both sit on those hosts in the English. The result is a self-contradiction against the next sentence (Mox enim praedicavit quod audivit), and that is the point — Pattern 7a is unconditional and a printed negation is carried "including when carrying it makes the sentence contradict itself."

⛔ This is the one place in the stint where the English asserts something its own Latin column does not, with no marker able to say so. A [cj:] was weighed and rejected: its fixed form (printed; read conjectured) cannot express a deletion, and the verbatim-in-the-twin guard would be satisfied only by the other non in the same sentence, which is gaming it. The TEI patch must land before this work ships, or a reader will meet an English negative with no Latin exponent and read it as ours. Filed in the patch file, Part D.

5.2 Test 2a fired and checked — negatives on a verb with a quantifier, correlative or pivot near

Six sites in the range meet the trigger. Each host was proved from the sentence's own structure, not from what reads well:

• 0140C Verumtamen non super femur, sed sub femore manum ponere jubet. The non stands before the verb-adjacent super femur, and the sed sub femore correlative proves the host is the prepositional phrase, not jubet. Rendered "he bids him put his hand not upon the thigh but under the thigh." (This is the 8960 @0740B model case.)
• 0140D quia Christo nullae reproborum animae conjunguntur. Quantifier host: nullae is on animae, not on the verb. "no souls of the reprobate are joined."
• 0141C Studet enim non plus sapere quam oportet sapere. The comparative quam fixes the host on plus sapere. "studies not to be more wise than it behoveth to be wise."
• 0141C quia verbum vitae non solum prudentibus, sed etiam stultis praedicatur. non solum … sed etiam; host prudentibus. Carried as "not only … but also."
• 0142C quod in aliquibus verbis non omnino conveniunt. ⚠ The 8960 @0734B shape exactly — omnino with a negation, where English word order can reverse the scope. non omnino is "not altogether", i.e. partial disagreement, and English "do not altogether agree" holds the same scope. Checked, correct.
• 0147B __non solum__ primatus sui honorem amisit, sed et coelestis regni praemium … adipisci non meruit. Two negations, two hosts (amisit under non solum; meruit). Both carried.
• 0143A quos non corporis frigus, sed fervor charitatis tenuit. The sed pivot proves the host is corporis frigus. "whom not the cold of the body but the fervour of charity held."

5.3 Four relocations found and undone in my own pass

Recorded because they are the class that passes every count-based check:

• 0133D quae putant homines vel non esse peccata — first drafted "either to be no sins", which moves the negative onto peccata. Restored to "either not to be sins."
• 0140D jam extraneam non habebat — first drafted "already held for no stranger", the same relocation. Restored to "did not already hold as a stranger."
• 0142B si … fructum non inveniunt — first drafted "find no fruit". Restored to "do not find fruit."
• 0142C cum veritas narrationis non exigit, nisi ut … — first drafted "requires nothing but that", which merges the non and the nisi into one word. Restored to "does not require anything except that."

§6 · UNEXPRESSED SUBJECTS SUPPLIED — every one named, with what decided it

The dominant defect class, and in a gloss the antecedent is usually in the lemma above or in the verse the lemma comes from.

• 0133C–D Tacuit uxorem, dixit sororem … certus quod a Deo inviolata servaretur → Abraham throughout, from the lemma « Dixitque de Sara » and Gn 20:2. inviolata is feminine and fixes Sara as the one kept.
• 0134C Orante Filio Patrem, sanatur infecunditas → "When the Son prays the Father"; both are printed, nothing supplied.
• 0135C Quod non facit zelo accensa → Sara, from zelo accensa (feminine) and from the lemma she speaks.
• 0135D per Saram nescientem prophetizatum est cum muliebri animo mota est → Sara, named in the same clause.
• 0136A sed subaudiendum est, dedit puerum suum → Abraham. Decided from Gn 21:14 (tradiditque puerum), which is the verse the whole paragraph is arguing about, and from the fact that Agar is the one being loaded, not the loader. ⚠ Both readings construe in the Latin; named here rather than chosen silently. Rendered "he gave over her boy."
• 0136C __Dixit__ enim: Non videbo mortem parvuli mei, et sedit contra eum → Agar. Decided by the paragraph's own closing sentence — Mater enim non suam mortem, sed filii deplorat — and by Gn 21:16, where the speaker is Agar. Rendered "she said … she sat." Likewise Quod sequitur: « Et projecit puerum … et abiens sedit » → "she cast … going away sat." (abiens is common gender and decides nothing on its own.)
• 0136D __illa__ fontem aspicit → Agar, printed.
• 0139C Erat enim in angelo Deus, et ex ejus persona loquebatur → God, from the preceding clause; the gloss's argument is that the speaker is the Son.
• 0139D __non putavimus__, sed verus est → first person plural, the commentator, printed.
• 0142B sed illi nisi prius obtineant aeterna, recipere nolunt → the teachers (doctores), from doctores ex temporalibus stipendiis retinere parati sunt — ⚠ not the people who are ready to retain them. Decided by pedes non lavant, which is the teachers' act (Gn 24:32).
• 0143D Quae se mox pallio cooperuit → Rebecca, from Gn 24:65 and from Rebecca viso Isaac two clauses above.
• 0146A Quaeritur quo ierit → Rebecca. Decided from the lemma « Perrexitque » (Gn 25:22, perrexitque ut consuleret Dominum, of Rebecca) and from the paragraph's own close, Rebecca tamen secundum Scripturam Dominum interrogavit. Rendered "whither did she go". ⚠ cum orando allegassent … dormiebant in the same paragraph is plural and is rendered "they", which is Migne's own shift and not ours.
• 0146A Ne putes … nam viginti annis mansit orans → Isaac, named four words later (quot annorum fuerit Isaac). ⭐ This is the site where Migne's own lemma (exaudivit eam) points the other way; the gloss decides it, and the divergence took the [var:] at §3a.
• 0146D __Praecedit__ Esau … deinde exit Jacob → both printed.

§7 · PUNCTUATION — compared BAND BY BAND, and the only added marks are three Pattern 8b colons

Test 3 run mechanically over all seven chunks: colons, semicolons, question marks, «, » and ! counted per column-anchored band (markers stripped), Latin against English. Four mismatched bands in 63, all accounted for, and none of them a supplied or destroyed mark of any other class:

⚑ The scan that found these also found six unlicensed marks I had introduced and repaired: a supplied colon at 0135A (Unde cresce — Migne prints a comma), a colon for Migne's full stop at 0136C (et statim jungitur.), a supplied semicolon at 0136D (after fontem aspicit), a supplied colon at 0138B (tres aetates significat, ante legem), a supplied semicolon at 0138C (non errore impietatis,) and a supplied semicolon at 0138D (Deus autem non senescit,). Test 3 earning its place again — none of them showed in any total.

⛔ The second ? of chunk 0030 is Migne's own and is preserved as a direct question: an per aliquem sacerdotem? → "was it through some priest?" It is not recast, because it needs no recasting; only the indirect ones take the 8b colon.

⛔ Not one comma→semicolon sweep was made. Migne prints semicolons freely in this range (55 in the seven chunks), so Tobit's narrow exception does not apply and was not invoked.

Em-dashes: two, both inside [var: …] markers (our own apparatus voice), none in the running prose. The verifier's corpus note reports 0.3/chunk for the work against a median of 5.3.

§8 · TEST 4 — sections read back against the Latin asking only does it assert what he asserts?

Run last, with the counts already clean. Read clean, paragraph by paragraph:

• 0024 entire (0133C–0134C). The Abimelech chapter. The Sara/Church allegory at 0134B and the thousand-pieces-of-silver arithmetic at 0134C both assert what Migne asserts.
• 0025 @0134C–0135A, the circumcision-and-eighth-day paragraph — the longest single sentence in the stint, with four subordinate chains; every clause is rendered and the octavo die / octava aetas figure survives intact.
• 0025 @0135C, the Agar/Sara allegory, where three negations run in series (non manere … nec esse … nec … possidere). Read twice; all three are on their own verbs.
• 0026 @0138B–C, the Isidore passage of nested quotations. ⚠ This is the hardest paragraph in the range: seven guillemet spans alternating scripture and gloss with hoc est between them, and it is easy to render as though the gloss were the quotation. Checked span for span against the twin; the alternation is preserved and no span moved.
• 0026 @0139A, the Abraham/God-the-Father typology. The Abraham senex … Deus autem non senescit antithesis is carried, including the concessive sed promissio … quodammodo senuerat.
• 0027 @0140B, the Spelunca paragraph — one 90-word Gregorian period with a non aliter … nisi frame. Rendered as one sentence with both negatives in place.
• 0028 @0141B–C, the Rebecca-as-Church allegory. ⚠ Contains §5.1's site; read three times.
• 0028 @0142C, Augustine on the evangelists' verbal disagreement — the paragraph whose whole subject is whether words differ, so a smoothed rendering would be self-refuting. The non omnino conveniunt and the non exigit, nisi ut frames are both literal.
• 0030 @0146A–B, Augustine's chain of Forsitan … an … vel … Vel forte conjectures. Every one of the five is a conjecture in the English too; none was hardened into a statement.
• 0030 @0146D, Bede on the two peoples within us. The iste minor, ille major / plures … mali quam boni inversions are carried without smoothing.

⚑ Two paragraphs that read more smoothly in English than the Latin does, flagged as suspects and then cleared: 0136A's Augustinian arithmetic of Ismael's age (the Latin's Sed ut … accipiamus concessive is genuinely elliptical; the English keeps the ellipsis) and 0139D's non putavimus, sed verus est (§4.6 — not cleared, and rendered with a bracketed pronoun rather than made to read well).

§9 · WHAT THIS STINT OWES DOWNSTREAM

1. Merge data/briefs/8950-PLATE-READS-0024.json into data/plate-reads.json (append, do not substitute). Until then plate-gate.mjs fails on fifteen of my sixteen [var:].
2. Apply data/briefs/8950-PENDING-TEI-PATCHES-0024.md, Part D first. 0141B is the one site where the English outruns its Latin.
3. At the merge, reconcile with the 0000–0003 stint on 1Cor 9:11 — it fired at 0081A, I declined at 0141D, and the two Migne printings genuinely differ (§3c). The two calls are consistent; they should not be swept into agreement.
4. [sic:] markers owed once the patches land, at eight columns: 0133C (sonorem), 0136D (par quem), 0137C (accdperat), 0139C (incanationem), 0140A (est.), 0140C (the dropped sub), 0141A (electorem), 0141C (caanem).

STINT 7 — chunks 0031–0035 (cols 0147C–0160A)

Filed as cruces-0031.md; merged verbatim, headings demoted one level. Nothing in it was rewritten — where a later stint corrected it, the correction stands in that stint's section.

Translated 2026-09-05, one stint, 5 chunks (5,320 Latin words by the frontmatter; 5,151 body words with anchors, notes and markers stripped → 7,527 English, 1.46×). Bands: 0147C–0160A — CAPUT XXVI, XXVII, XXVII (cont.), XXVIII, XXIX, XXX, XXXI, XXXII. verify-english.mjs 8950 reports no error on 0031–0035 (the only failures in the work are MISSING English chunk for ranges other stints hold).

SPAN COUNT — 222, of which 1 OPEN-ENDED. Counted paragraph-bounded in my own five Latin chunks: 155 closed « … » spans + 66 … spans + 1 unclosed « = 222, matching the brief's header (222 spans, of which 1 OPEN-ENDED) exactly. Per chunk: 0031 27+13+0 · 0032 37+9+0 · 0033 35+17+1 · 0034 15+10+0 · 0035 41+17+0.

⚠ And the agreement is NOT the false kind the brief warns of — the three raw-mark asymmetries are accounted for individually, not netted. Body marks run 157 « / 156 », and the gap resolves as: (a) an orphan closing » at 0148D, Migne's (§3); (b) a nested line-head « inside a closed span at 0152A, which pairs into its host and is counted once; (c) the one genuinely unclosed « at 0154A, Migne's (§3). Chunk by chunk the raw marks are 27/28 · 38/37 · 36/35 · 15/15 · 41/41 — three chunks unbalanced in two different directions, and each direction has its own cause. ⚑ A DOTALL «.*?» over a chunk body runs the 0154A quotation forward across two paragraph breaks and swallows two spans; counting paragraph-bounded is what keeps that from cancelling against (a).

Apparatus fired: 6 [var:] · 1 [sic:] · 1 [ed:] · 0 [cj:] · 0 [d:] · 0 [nt:]. Columns: [var:] 0147D · 0150C · 0152A · 0152B · 0152D · 0160A · [sic:] 0152A · [ed:] 0148C. ⚑ The Num. 6:27 [var:] resolves to band 0152D in plate-gate.mjs because Migne's [0152D] anchor falls inside the quotation it marks; the lemma opens at 0152C. Both columns are on the same read (p. 81). Every one stands on a column read at the plate this session (data/briefs/8950-PLATE-READS-0031.json).

Per-chunk parity (English / Latin), all exact: column anchors 7/14/10/8/11 · [n:] notes 14/13/10/14/10 · « 27/38/36/15/41 · » 28/37/35/15/41 · question marks 0/2/2/3/1 · italic spans 13/9/17/10/17. Band-by-band punctuation scan (: ; ? « » ! per column anchor, apparatus stripped): three mismatched bands in five chunks, all : +1, all Pattern 8b, and zero elsewhere (§7). Zero em-dashes supplied in 5,151 words — the one — in the files is inside a [var:].

⭐ THE WHOLE RANGE WAS READ AT THE PLATE: archive.org patrologiaecurs04migngoog on disk, PDF page = (column + 11) / 2, corner numbers read first on every page — pp. 79 (147/148), 80 (149/150), 81 (151/152), 82 (153/154), 83 (155/156), 84 (157/158), 85 (159/160). Reads at 200 dpi full-page with 600 dpi crops at twenty sites and one 1200 dpi crop (§6). Foot-of-page apparatus: CHECKED ZERO on all seven pages, all three key shapes — no numbered Forte notes, no asterisk layer, and no letter-keyed (a) notes of the kind found at cols 0068 and

1. There is no foot rule on any of the seven pages; the columns simply end. A [cn:]-anchor grep over my five Latin chunks (raised letters, per Pattern 19's note that this book's TEI keeps the key while dropping the note) returns zero, which agrees.

Twin/plate divergences are in data/briefs/8950-PENDING-TEI-PATCHES-0031.md.

§1 · CONVENTIONS — inherited unchanged, and what this stint had to settle

Ruth (8968 §1) and Tobit (8970 §1) govern and were not re-opened, nor were the earlier stints of this same work (cruces-0000.md, -0004.md, -0008.md, -0012.md, -0019.md), which I read before drafting: VERS. n.-- passes through verbatim; the lemma is Englished; a fragmentary lemma is rendered as the phrase it stands for; sigla ride as ordinary [n: …] notes with Migne's own internal spacing, contents unanglicized; the guillemets are the plate's, 1:1, and their contents are Englished; usque ad → as far as, with no verb and no object ever supplied; the Historice / Allegorice / Mystice tags are translated inside their italic span.

The four conventions the earlier 8950 stints settled were adopted, not re-derived: (i) no [var:] on a quotation Migne himself labels a rival version — nine instances here, §4.3; (ii) words follow the plate, punctuation follows the twin — heavily exercised, patch file Parts A and C; (iii) a Latin gloss giving the meaning of a name is Englished inside its span, a Latin word mentioned as a word stays Latin — six of each, §1(b); (iv) unpaired guillemets are mirrored, never tidied, and run both ways — two here, both Migne's, §3.

(a) The sense tags, and one the twin has DE-ITALICIZED

Moraliter. → Morally. (0148A, 0154B) · Mystice. → Mystically. (0149B, 0150C, 0151A, 0159A, 0159D) · Allegorice. → Allegorically. (0150B, 0155C) · Historice. → Historically. (0150C, 0157A, 0158B). A note inside the tag's italic span stays inside it ([n: (HIPP. mart., …)] Mystice. at 0149B, [n: (GREG. hom. 14, …)] Mystice. at 0159D).

⛔ 0154D is the exception, and it is our source's doing, not Migne's. The twin prints etc. Mystice designat mandatum divinitatis with no italics, where the plate italicizes it like every other sense-tag in the book (read at 200 dpi, p. 82). Italic-span parity against the twin is a hard verify-english check, so the English renders it roman — making 0154D the one place in this work where a sense-tag is not set off. Filed as a patch; flagged here because a blind reader will read it as our inconsistency and it is not.

(b) Which italic Latin is Englished, and which is not — convention (iii) applied to twelve spans

Englished, because the span gives the MEANING of a name or of a foreign word:

• contrarium → contrary (0148C) — the meaning of Satana, and the sentence is unde patet Satanam contrarium interpretari.
• latitudines. → breadths. (0148C) — the meaning of Rehobot; and the lemma « Latitudo. » is Englished « Breadth. » for the same reason, the gloss being about nothing else.
• supplantator → a supplanter (0152D) — the meaning of Jacob, and Supplantavit enim me stands four words above it.
• unctus → anointed (0154A) — χρίστος enim Graece, unctus Latine, and the argument runs quem Jacob unxit, ut Christus signaretur. Leaving it Latin would break the chain unxit → unctus → Christus, which is the whole gloss.
• domum Dei, → the house of God, · nux → a nut · amygdalum. → an almond. · prius. / prius → before. / before · ante vestibulum, → before the porch, · superliminare, → a lintel, · postes, → doorposts, (all 0155B) — Jerome's Latin equivalents of Bethel, Luza and ulam.
• dealbatio. → a whitening. (0158A) — the meaning of Laban, and the next sentence turns on it (Diabolus vero, cum tenebrosus sit, transfigurat se in angelum lucis).
• ovis, → a sheep, (0158B) — the meaning of Rachel.

Kept Latin, because the span is a word MENTIONED as a word:

• valle, torrentem (0148B) — Pro valle, torrentem habent in Hebraeo. The 0012 stint's own worked shape at 0118B (mare pro occidente ponitur), and it is Latin-for-Latin: the sentence is about which word stands where, not about what either means. ⚑ The reader loses nothing, because Migne's very next sentence uses both words unitalicised and the English translates them there — habitare potuit in valle, sed in torrente → "could dwell in a valley, but in a torrent".
• pronubo, (0149A) — Pro « Ochozath, » pronubo, in Hebraeo habetur « collegium amicorum ejus, ». The identical frame ("for X, Y, the Hebrew has Z"), so it takes the identical treatment; it is a rival rendering being named, not a gloss.
• castra (0159B) — Ubi castra posita sunt, in Hebraeo mahanaim. Same frame again. ⚑ Note that Gn 32:2 itself glosses Mahanaim, id est, Castra, so the Latin word here is the object of the remark, not its content.
• Hebrew and Greek stay untouched: Rehobot, rabooth, (0148C) · ulam, / ulam / olam (0155B, four spans) · monim (0157C) · theraphim ×2 (0158A) · mahanaim (0159B) · Phanuel (0160A) · χρίστος (0154A). The guilleted « Satana; » (0148C) is likewise the Hebrew word and is carried in roman inside the guillemets, because italicizing it would add an italic span the twin does not have.

⚠ olam, at 0155B is the PLATE'S reading; our twin prints ulam,. See §2 and the patch file. Jerome's sentence is Unde ridicule quidam verbum Hebraicum olam, nomen urbis esse putant: cum ulam interpretetur prius — two forms of one word, which the next sentence explicitly pairs (verbo ulam vel olam plenae sunt). The plate's reading is the harder one and it construes.

(c) usque ad — 23 formulas, 5 ordinary, decided per occurrence by the etc. splice test

Formula (italic, immediately following an etc.), 23×, all → as far as, with no verb, no object and no antecedent supplied anywhere: 0147C · 0147D · 0148A · 0148D · 0149A ×2 · 0150C · 0151B · 0151C · 0154A · 0156C ×3 · 0156D ×3 · 0157A · 0157B · 0159A · 0159D ×2 · 0160A.

Ordinary Latin (5×), none italic, none after an etc., each rendered by sense:

⚑ 0156B is addendum D's trap and it was not swept: the words are Philippians' own, four lines below a formula at 0156C, in a paragraph otherwise thick with the abridgment. ⚠ Also present and NOT this word: 0151B eum usque in finem bonis operibus placare → "unto the end", and 0148C et quousque verae intelligentiae aqua redeat → "until", which is the plate's reading against the twin's non-word quosque (§2).

⛔ 0156D's formula prints as usque ab ON THE PLATE; our twin reads usque ad. Read at 600 dpi, p. 83. The tail is unmistakably the formula — it follows etc.,, it is italic, and it names the resumption (ad unum ergo tendendum). Rendered as the formula, "as far as", with the plate reading in the patch file. ⛔ No [sic:]: the marker's content must be verbatim in the twin, and the twin prints the good word. This is the mirror of the 0019–0023 stint's uque ad at 0131C, where the twin was defective and the plate clean — the same class, running the other way.

(d) The vocabulary axes — every instance, with the instrument checked before reporting

• gentes family — 20 real occurrences, split by the axis. Gentiles where the Judaei/gentes opposition is live and usually explicit in the same clause: 0149D populus gentilium and Jacob gentilem populum significat (against populum Hebraicum); 0150A electi de gentibus and praesciens populum gentium multo meliorem (against populum Judaeorum); 0150B mater gratia gentiles fidem avide suscipientes (against Judaei) and confessionem peccatorum gentilis populi; 0150C Per unum peccatorem populum gentium, per nonaginta novem justos, Judaeos ; 0151A id est gentiles conversi; 0152A cum intraverit plenitudo gentium, tunc omnis Israel and 0153A the same in the Rom 11 quotation; 0153A relicta Judaea, transirent ad gentes; 0153D relictis parentibus … id est Judaeis … abiit in gentes; 0158B Ecclesiam ex gentibus conjunxit; 0159A tam Judaeos quam gentiles; 0159C paries de gentibus, ei qui ex Judaeis. Nations where the sense is plainly generic — the axis's own named exception: 0148D baptizantes omnes gentes in nomine Patris (Mt 28:19, the axis's own exemplar); 0155A Ps 2:8 dabo tibi gentes haereditatem tuam; 0156B Christus diversarum gentium varietatem sibi congregavit. Singular gens of a people → nation: 0148C apud alienigenam gentem puteos fodit → "among a foreign nation". ⚑ Instrument checked before reporting, per the series brief: a raw gent grep returns 22 hits and two are false — intelligentiae (0148C) and ingenti crimine (0156C).
• natio — ZERO in this range. ⚑ A raw natio grep returns six hits and every one is peregrinatione (0148C), venatione ×2 (0150A, 0150B), alienatio (0152B), machinationem / machinatione (0153B, 0153C). There is no natio here at all.
• turba — ONE, and it is a real one: 0149A amicorum turbam → "the crowd of friends", per the axis. ⚑ Instrument checked: it is the only turb- in 5,151 words, and there is no turbare / perturbatio in the range — the reverse of the 0019 stint's finding, where nine turb- hits were every one the verb. So this range exercises the turba side and the 0019 range exercised neither.
• multitudo ×4 → multitude in all four (0151A ×2, 0151D, 0159A). The turba/multitudo pair is genuinely exercised here, in one work and one book, and was kept 1:1.
• daemonium — zero; daemonum ×1 → demons (0148A). diabolus ×5 → devil (0154C, 0158A ×2, 0158B, 0158B). Satanas → Satan (0148C the Hebrew word and its accusative, 0159B post Satanam). ⚠ 0149B prints zabuli, a late form of diabolus; rendered "of the devil" per the axis, no italics and no crux-worthy defect (Pattern 9 — an attested medieval spelling). ethnicus, gentilitas — zero.
• mysterium ×1 → mystery (0151C prophetiae mysteria → "the mysteries of prophecy"); sacramentum ×9 → sacrament (0147C, 0147D, 0148B, 0151A ×2, 0152A, 0152C, 0155D, 0159D). The 1:1 pair is not exercised in a single phrase here, but both words occur and neither was varied.
• The commentator's own voice is not archaised (dicit / ait / dicat → "says", never saith); -est / -eth forms appear only inside quoted scripture and in second-person singular address.
• Pattern 17. Singular → thou: 0150A « Sume arma [tua], » → "Take thy arms"; 0150D « tam cito invenire potuisti, fili mi? » → "couldst thou find it so quickly, my son?"; 0151D « Det tibi Deus de rore coeli, » → "God give thee"; 0152A erit benedictio tua; 0152B « Vives in gladio », « Et servies fratri tuo minori », « cum excutias et solvas jugum ejus de cervicibus tuis » and the gloss's deposueris · non servies · vocaberis; 0152D « Et fratri tuo servies »; 0153C Si diligis vitam meam, mitte eum; 0155A Ps 2:8 « dabo tibi … possessionem tuam »; 0158B « Obliviscere populum tuum »; 0158D « Timui ne violenter auferres filias tuas … furti me arguis » and non tibi manifestavi, ne … superares; 0159D « appellabitur nomen tuum ». Plural → you: 0150A « Si custodieritis mandata mea, dabo vobis pluvias »; 0150C « Ut quid justificatis vos coram hominibus »; 0154C « In quacunque die comederitis, aperientur oculi vestri »; 0153B poterit esse quod ambo eritis interfecti. ⚑ 0153B is worth naming because it mixes them inside one sentence, as the Latin does: Si invadat te, poterit esse quod ambo eritis interfecti → "If he assail thee, it may be that you shall both be slain." Rebecca turns from Jacob to the two brothers mid-clause, and Pattern 17 is what makes that visible. ⚑ Three plural imperatives are rendered as bare English imperatives, which mark no number, per the 0019 stint's ruling that Pattern 17 gives you and never ye: 0148A « Declinate a me, maligni, » → « Depart from me, O malignant ones, »; 0152C « Ponite nomen meum super filios Israel. » → « Put my name upon the children of Israel. »; 0158C « Surgite postquam sederitis. » → « Rise up after you have sat. » Depart ye / Put ye / Rise ye were drafted and withdrawn.

(e) Heads and frontmatter

CAPUT XXVI. → CHAPTER XXVI. · CAPUT XXVII. → CHAPTER XXVII. · CAPUT XXVII. (cont.) → CHAPTER XXVII. (cont.) · CAPUT XXVIII. → CHAPTER XXVIII. · CAPUT XXIX. → CHAPTER XXIX. · CAPUT XXX. → CHAPTER XXX. · CAPUT XXXI. → CHAPTER XXXI. · CAPUT XXXII. → CHAPTER XXXII. Numerals mirror the plate, per the corpus measurement, and all eight heads carry their terminal period on the plate as in the twin (read at 200 dpi; CAPUT XXXI. re-checked at 600 dpi). Frontmatter copied VERBATIM from each Latin chunk, field order included, words: figures (which count the Latin) unaltered. Chunks 0034 and 0035 open with a bare column anchor on its own line ([0155C], [0157C]) between the head and the first paragraph; mirrored as its own line.

§2 · ⭐⭐ THE HEADLINE FINDING — our twin has silently NORMALIZED FOUR VERSE ADDRESSES

This is the largest thing in the stint, it is a class rather than four accidents, and nobody could have flagged any of it from the Latin alone.

All four read at 600 dpi with the page's corner numbers checked first (pp. 80 and 83). The digits are not marginal: on p. 80 the 5 of 5-13 and the 9 of 19-24 are unambiguous against the adjacent 3-5 and 16-19 set in the same fount on the same page.

⭐ The direction is uniform, and that is what makes the class credible rather than four misreadings. In every one of the four, Migne's range OVERLAPS its neighbour or skips a verse, and the twin's is the tidy, non-overlapping one. On p. 80 the plate's sequence runs 1,2 · 2 · 3-5 · 5-13 · 14 · 16-19 · 19-24 · 25,26; our twin's runs 1,2 · 2 · 3-5 · 6-13 · 14 · 15-19 · 20-24 · 25,26. On p. 83 the plate runs 11-19 · 16-19; the twin runs 11-16 · 16-19. And in three of the four the twin's number is also the righter one against the lemma it heads — which is exactly what an editorial tidy looks like from the inside.

The English renders the plate in all four, on the 0019–0023 stint's ruling at VERS. 27-19 (cruces-0019 §2), and for its three reasons unchanged: the locked "plate where read" rule; a verse address is content, not a mark, so the punctuation-follows-the-twin carve-out does not reach it; and Migne's own printing is not ours to tidy — the series brief's rule for a verse number out of sequence (Jude printing VERS. 23 before VERS. 21) is this case exactly.

⛔ No [sic:] was available on any of them — the marker's content must appear verbatim in the Latin twin, and the twin prints the other numeral. The patch is what makes the two columns agree.

⚑ What a later stint should carry forward: the cheap detector is an OVERLAPPING or verse-skipping range in the plate's own sequence. Our twin never leaves one. Where the twin's addresses partition the chapter neatly and Migne's neighbours in this book do not, suspect a tidy and read the leaf. ⚠ Nine other addresses in my range were checked against the plate and AGREE — 0147C 1-7, 8-11, 12,13; 0148A 14-16; 0148B 17, 18,19; 0148C 20,21, 22-25; 0148D 26-31; 0149A 32-35; 0149B 1,2; 0149D 2; 0150A 3-5; 0150C 14; 0152A 33-35; 0152D 36-39, 40; 0153A 41; 0153B 42; 0153C 1-9; 0153D 10; 0154A 11, 12; 0154D 13-15; 0155A 16,17, 18, 19-21; 0155B 22; 0155C 1-10; 0155D 16-19; 0156A 20-35; 0156D 26, 14-31; 0157B 32-43; 0157C 1-11; 0157D 12, 13-31; 0158A 32-39; 0158D 40-45, 46-52; 0159A 53,54, 55; 0159B 2-6, 7-9; 0159C 10, 11-19, 20-27; 0159D 28,29; 0160A 30-32. Forty-six checked, four divergent — the negatives are what size the finding.

§3 · THE TWO UNPAIRED GUILLEMETS — both MIGNE'S, both plate-read, neither tidied

The brief warns that an open-ended span "may be SHORT of what Migne quotes — read it at the line." Both were collated at the line, not from the brief.

⭐ 0154A is the stint's one open-ended span, and the brief's warning does NOT apply to it: the plate stops exactly where the twin stops. The paragraph ends on the siglum, mid-quotation; the next paragraph opens VERS. 12.--. So the brief's rendering of the span is complete, not short. The [n: (I Cor. XI.)] sits inside the unclosed quotation and is reproduced there.

⚠ Both directions occur in one stint, which is the corollary the 0008 and 0019 stints drew from opposite ends and it is now confirmed a third time: 0148D is a close without an open, 0154A an open without a close, and only the plate separates "Migne's" from "our twin's". Six further sites in this range are the other kind — marks our twin supplies or drops — and every one of those is OURS (patch file, Part C1/C2). ⛔ Nothing was repaired; guillemet parity is checked against the twin, so the English mirrors the twin at all eight sites.

⚑ Migne's repeated line-head « is dense here — thirteen sites across the five chunks, and our twin drops every one (patch file C2). The class the 0000 stint identified; it is not two accidents in this book, it is his normal practice at a turned line.

§4 · ⚠⚠ 7a″ — THE VULGATE COLLATION. This is the deliverable.

Every marked span in these five chunks was collated against sources/vulgate/clementine-flat.txt by grepping the verse, cited book chapter:verse — never from memory and never from the brief's pre-scan. The spans that AGREE are listed in §4.4, because a list of divergences alone cannot be told apart from a report by an agent that only chased the ⚠ marks.

4.0 ⛔ THE COMPARAND WAS CHECKED FIRST — the errata class, and what it did and did not touch here

sources/vulgate/README.md records that our 2005 Quasimodo copy carries typos which manufacture divergences shaped exactly like real ones, and that two were found in a single column of this very work. Every one of my six [var:] was tested against that class before it was fired, by the README's own cheap control — grep the phrase, not the word, and look for a correctly-spelled twin elsewhere in the file: